From baldrick at free.fr  Mon Mar  8 03:16:43 2010
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 08 Mar 2010 09:16:43 -0000
Subject: [llvm-commits] [dragonegg] r97940 - /dragonegg/trunk/www/index.html
Message-ID: <20100308091643.1C6AD2A6C12C@llvm.org>

Author: baldrick
Date: Mon Mar  8 03:16:42 2010
New Revision: 97940

URL: http://llvm.org/viewvc/llvm-project?rev=97940&view=rev
Log:
Dragonegg can now compile clang and boost.

Modified:
    dragonegg/trunk/www/index.html

Modified: dragonegg/trunk/www/index.html
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/www/index.html?rev=97940&r1=97939&r2=97940&view=diff
==============================================================================
--- dragonegg/trunk/www/index.html (original)
+++ dragonegg/trunk/www/index.html Mon Mar  8 03:16:42 2010
@@ -37,9 +37,9 @@
 
   <ul>
   <li>C works well, for example you can build a working gcc using it</li>
-  <li>C++ works fairly well, for example you can build a working LLVM using it.
-  On the other hand, <a href="http://dragonegg.llvm.org/">DragonEgg</a> crashes
-  while building clang and boost</li>
+  <li>C++ works fairly well, for example you can build LLVM, clang and boost
+  with it (the resulting LLVM and clang work correctly; I did not test whether
+  boost actually works)</li>
   <li>A lot of Fortran compiles, but the compiled code has not been tested for
   correctness</li>
   <li>It can compile quite a lot of Ada, but the compiled code may not work
@@ -153,6 +153,7 @@
      <pre>	svn co http://gcc.gnu.org/svn/gcc/trunk gcc</pre>
   <p>Apply all of the patches in <tt>dragonegg/gcc-patches/</tt>, if any, to
      <a href="http://gcc.gnu.org">gcc</a>.
+     You need to pass the <tt>-p1</tt> option to <tt>patch</tt>.
      Build and install <a href="http://gcc.gnu.org">gcc</a> in the usual way.</p>
   <p>Doing</p>
      <pre>	GCC=path_to_just_installed_gcc make</pre>




From xuzhongxing at gmail.com  Mon Mar  8 06:45:34 2010
From: xuzhongxing at gmail.com (Zhongxing Xu)
Date: Mon, 8 Mar 2010 20:45:34 +0800
Subject: [llvm-commits] [PATCH] remove redundant 'static'
Message-ID: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>

'static' is redundant in this context:

Index: utils/TableGen/RegisterInfoEmitter.cpp
===================================================================
--- utils/TableGen/RegisterInfoEmitter.cpp	(?? 97892)
+++ utils/TableGen/RegisterInfoEmitter.cpp	(????)
@@ -201,7 +201,7 @@

     // Emit the register list now.
     OS << "  // " << Name << " Register Class...\n"
-       << "  static const unsigned " << Name
+       << "  const unsigned " << Name
        << "[] = {\n    ";
     for (unsigned i = 0, e = RC.Elements.size(); i != e; ++i) {
       Record *Reg = RC.Elements[i];
@@ -223,7 +223,7 @@
     // Emit the register list now.
     OS << "  // " << Name
        << " Register Class Value Types...\n"
-       << "  static const EVT " << Name
+       << "  const EVT " << Name
        << "[] = {\n    ";
     for (unsigned i = 0, e = RC.VTs.size(); i != e; ++i)
       OS << getEnumName(RC.VTs[i]) << ", ";
@@ -252,7 +252,7 @@

       OS << "  // " << Name
          << " Sub-register Classes...\n"
-         << "  static const TargetRegisterClass* const "
+         << "  const TargetRegisterClass* const "
          << Name << "SubRegClasses[] = {\n    ";

       bool Empty = true;
@@ -298,7 +298,7 @@

       OS << "  // " << Name
          << " Super-register Classes...\n"
-         << "  static const TargetRegisterClass* const "
+         << "  const TargetRegisterClass* const "
          << Name << "SuperRegClasses[] = {\n    ";

       bool Empty = true;
@@ -334,7 +334,7 @@

       OS << "  // " << Name
          << " Register Class sub-classes...\n"
-         << "  static const TargetRegisterClass* const "
+         << "  const TargetRegisterClass* const "
          << Name << "Subclasses[] = {\n    ";

       bool Empty = true;
@@ -382,7 +382,7 @@

       OS << "  // " << Name
          << " Register Class super-classes...\n"
-         << "  static const TargetRegisterClass* const "
+         << "  const TargetRegisterClass* const "
          << Name << "Superclasses[] = {\n    ";

       bool Empty = true;



From clattner at apple.com  Mon Mar  8 10:40:33 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 8 Mar 2010 08:40:33 -0800
Subject: [llvm-commits] [PATCH] remove redundant 'static'
In-Reply-To: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>
References: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>
Message-ID: <F18CF852-A0B8-45FC-BD20-B4B312A1EC86@apple.com>


On Mar 8, 2010, at 4:45 AM, Zhongxing Xu wrote:

> 'static' is redundant in this context:

Um, no it's not.  It prevents exporting the symbol from the .o file.  What are you trying to fix?

-Chris

> 
> Index: utils/TableGen/RegisterInfoEmitter.cpp
> ===================================================================
> --- utils/TableGen/RegisterInfoEmitter.cpp	(?? 97892)
> +++ utils/TableGen/RegisterInfoEmitter.cpp	(????)
> @@ -201,7 +201,7 @@
> 
>     // Emit the register list now.
>     OS << "  // " << Name << " Register Class...\n"
> -       << "  static const unsigned " << Name
> +       << "  const unsigned " << Name
>        << "[] = {\n    ";
>     for (unsigned i = 0, e = RC.Elements.size(); i != e; ++i) {
>       Record *Reg = RC.Elements[i];
> @@ -223,7 +223,7 @@
>     // Emit the register list now.
>     OS << "  // " << Name
>        << " Register Class Value Types...\n"
> -       << "  static const EVT " << Name
> +       << "  const EVT " << Name
>        << "[] = {\n    ";
>     for (unsigned i = 0, e = RC.VTs.size(); i != e; ++i)
>       OS << getEnumName(RC.VTs[i]) << ", ";
> @@ -252,7 +252,7 @@
> 
>       OS << "  // " << Name
>          << " Sub-register Classes...\n"
> -         << "  static const TargetRegisterClass* const "
> +         << "  const TargetRegisterClass* const "
>          << Name << "SubRegClasses[] = {\n    ";
> 
>       bool Empty = true;
> @@ -298,7 +298,7 @@
> 
>       OS << "  // " << Name
>          << " Super-register Classes...\n"
> -         << "  static const TargetRegisterClass* const "
> +         << "  const TargetRegisterClass* const "
>          << Name << "SuperRegClasses[] = {\n    ";
> 
>       bool Empty = true;
> @@ -334,7 +334,7 @@
> 
>       OS << "  // " << Name
>          << " Register Class sub-classes...\n"
> -         << "  static const TargetRegisterClass* const "
> +         << "  const TargetRegisterClass* const "
>          << Name << "Subclasses[] = {\n    ";
> 
>       bool Empty = true;
> @@ -382,7 +382,7 @@
> 
>       OS << "  // " << Name
>          << " Register Class super-classes...\n"
> -         << "  static const TargetRegisterClass* const "
> +         << "  const TargetRegisterClass* const "
>          << Name << "Superclasses[] = {\n    ";
> 
>       bool Empty = true;
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From daniel at zuster.org  Mon Mar  8 10:42:42 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 16:42:42 -0000
Subject: [llvm-commits] [zorg] r97944 - in
 /zorg/trunk/buildbot/osuosl/master/config: builders.py slaves.py
Message-ID: <20100308164242.6AE172A6C12C@llvm.org>

Author: ddunbar
Date: Mon Mar  8 10:42:42 2010
New Revision: 97944

URL: http://llvm.org/viewvc/llvm-project?rev=97944&view=rev
Log:
Tweak slave names for baldrick.


Modified:
    zorg/trunk/buildbot/osuosl/master/config/builders.py
    zorg/trunk/buildbot/osuosl/master/config/slaves.py

Modified: zorg/trunk/buildbot/osuosl/master/config/builders.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/osuosl/master/config/builders.py?rev=97944&r1=97943&r2=97944&view=diff
==============================================================================
--- zorg/trunk/buildbot/osuosl/master/config/builders.py (original)
+++ zorg/trunk/buildbot/osuosl/master/config/builders.py Mon Mar  8 10:42:42 2010
@@ -127,7 +127,7 @@
          'category' : 'clang.exp' },
 
         {'name' : 'dragonegg-x86_64-linux',
-         'slavenames' : ['baldrick2'],
+         'slavenames' : ['baldrick16'],
          'builddir' : 'dragonegg-x86_64-linux',
          'factory' : DragonEggBuilder.getBuildFactory(triple='x86_64-pc-linux-gnu'),
          'category' : 'dragonegg.exp' },

Modified: zorg/trunk/buildbot/osuosl/master/config/slaves.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/osuosl/master/config/slaves.py?rev=97944&r1=97943&r2=97944&view=diff
==============================================================================
--- zorg/trunk/buildbot/osuosl/master/config/slaves.py (original)
+++ zorg/trunk/buildbot/osuosl/master/config/slaves.py Mon Mar  8 10:42:42 2010
@@ -47,12 +47,12 @@
         create_slave("adobe1", properties={'jobs' : 2}, max_builds=1),
 
         # GCC Farm Slaves, for DragonEgg
-        create_slave("baldrick1", properties={'jobs' : 4}, max_builds=1),
-        create_slave("baldrick2", properties={'jobs' : 8}, max_builds=1),
+        create_slave("baldrick11", properties={'jobs' : 4}, max_builds=1),
         create_slave("baldrick12", properties={'jobs' : 4}, max_builds=1),
         create_slave("baldrick13", properties={'jobs' : 4}, max_builds=1),
         create_slave("baldrick14", properties={'jobs' : 8}, max_builds=1),
         create_slave("baldrick15", properties={'jobs' : 2}, max_builds=1),
+        create_slave("baldrick16", properties={'jobs' : 8}, max_builds=1),
         create_slave("baldrick17", properties={'jobs' : 8}, max_builds=1),
 
         # Defunct.




From daniel at zuster.org  Mon Mar  8 10:50:05 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 16:50:05 -0000
Subject: [llvm-commits] [zorg] r97945 -
	/zorg/trunk/buildbot/osuosl/master/config/builders.py
Message-ID: <20100308165005.6742F2A6C12C@llvm.org>

Author: ddunbar
Date: Mon Mar  8 10:50:05 2010
New Revision: 97945

URL: http://llvm.org/viewvc/llvm-project?rev=97945&view=rev
Log:
Remove offline auroraux builders.


Modified:
    zorg/trunk/buildbot/osuosl/master/config/builders.py

Modified: zorg/trunk/buildbot/osuosl/master/config/builders.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/osuosl/master/config/builders.py?rev=97945&r1=97944&r2=97945&view=diff
==============================================================================
--- zorg/trunk/buildbot/osuosl/master/config/builders.py (original)
+++ zorg/trunk/buildbot/osuosl/master/config/builders.py Mon Mar  8 10:50:05 2010
@@ -24,10 +24,6 @@
          'builddir':"llvm-arm-linux",
          'factory': LLVMBuilder.getLLVMBuildFactory("arm-pc-linux-gnu", jobs=1, clean=False,
                                                     timeout=40)},
-        {'name': "llvm-i386-auroraux",
-         'slavenames':["evocallaghan"],
-         'builddir':"llvm-i386-auroraux",
-         'factory': LLVMBuilder.getLLVMBuildFactory("i386-pc-auroraux", jobs="%(jobs)s", make='gmake')},
         ]
 
 # Offline.
@@ -35,6 +31,10 @@
  'slavenames':["andrew1"],
  'builddir':"llvm-alpha",
  'factory': LLVMBuilder.getLLVMBuildFactory("alpha-linux-gnu", jobs=2)}
+{'name': "llvm-i386-auroraux",
+ 'slavenames':["evocallaghan"],
+ 'builddir':"llvm-i386-auroraux",
+ 'factory': LLVMBuilder.getLLVMBuildFactory("i386-pc-auroraux", jobs="%(jobs)s", make='gmake')},
 
 # llvm-gcc self hosting builders.
 def _get_llvmgcc_builders():
@@ -85,11 +85,6 @@
          'slavenames' :['dunbar-win32-2'],
          'builddir' :"clang-i686-xp-msvc9",
          'factory' : ClangBuilder.getClangMSVCBuildFactory(jobs=2)},
-        {'name': "clang-i386-auroraux",
-         'slavenames':["evocallaghan"],
-         'builddir':"clang-i386-auroraux",
-         'factory': ClangBuilder.getClangBuildFactory("i386-pc-auroraux",
-                                                      jobs="%(jobs)s", make='gmake')},
         {'name' : "clang-x86_64-darwin10-selfhost",
          'slavenames' : ["dunbar-darwin10"],
          'builddir' : "clang-x86_64-darwin10-selfhost",
@@ -99,6 +94,13 @@
                                                        stage2_config='Debug')},
         ]
 
+# Offline.
+{'name': "clang-i386-auroraux",
+ 'slavenames':["evocallaghan"],
+ 'builddir':"clang-i386-auroraux",
+ 'factory': ClangBuilder.getClangBuildFactory("i386-pc-auroraux",
+                                              jobs="%(jobs)s", make='gmake')},
+
 def _get_experimental_builders():
     return [
         {'name' : "llvm-gcc-x86_64-darwin10-cross-mingw32",




From dpatel at apple.com  Mon Mar  8 12:13:18 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 8 Mar 2010 10:13:18 -0800
Subject: [llvm-commits] [llvm] r86748 - in /llvm/trunk:
	include/llvm/Analysis/ include/llvm/CodeGen/
	include/llvm/Transforms/Utils/ lib/Analysis/
	lib/CodeGen/AsmPrinter/ lib/CodeGen/SelectionDAG/
	lib/Transforms/Utils/
In-Reply-To: <bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>
	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
Message-ID: <7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>

Hi Jeffrey,

I'll update the comment. AbstractScopes map is used to keep track of abstract scopes in entire module. AbstractScopesList is used to construct abstract scopes seen while processing a function. This use of list avoids randomness in DIE ordering. 

-
Devang
On Mar 6, 2010, at 2:40 PM, Jeffrey Yasskin wrote:

> On Tue, Nov 10, 2009 at 3:06 PM, Devang Patel <dpatel at apple.com> wrote:
>> Author: dpatel
>> Date: Tue Nov 10 17:06:00 2009
>> New Revision: 86748
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=86748&view=rev
>> Log:
>> Implement support to debug inlined functions.
>> 
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Nov 10 17:06:00 2009
>> +  AbstractScopes[N] = AScope;
>> +  if (DIDescriptor(N).isSubprogram())
>> +    AbstractScopesList.push_back(AScope);
>> +  return AScope;
>> +}
> 
>>   // Clear debug info
>> -  if (FunctionDbgScope) {
>> -    delete FunctionDbgScope;
>> +  if (CurrentFnDbgScope) {
>> +    CurrentFnDbgScope = NULL;
>>     DbgScopeMap.clear();
>>     DbgScopeBeginMap.clear();
>>     DbgScopeEndMap.clear();
>> +    ConcreteScopes.clear();
>> +    AbstractScopesList.clear();
> 
> Why do you clear AbstractScopesList here but not AbstractScopes? Since
> AbstractScopesList only gets a value added to it when AbstractScopes
> is missing that value, it seems like this will produce a wrong
> AbstractScopesList in subsequent functions.
> 
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
>> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Tue Nov 10 17:06:00 2009
>> @@ -134,52 +134,52 @@
>> +  /// AbstractScopes - Tracks the abstract scopes a module. These scopes are
>> +  /// not included DbgScopeMap.
>> +  ValueMap<MDNode *, DbgScope *> AbstractScopes;
>> +  SmallVector<DbgScope *, 4>AbstractScopesList;
> 
> Reading this comment and the variable names, I'd think that every
> DbgScope in AbstractScopes should appear in the AbstractScopesList.
> However, only the subprograms do. Could you add a comment and/or
> rename the variable?
> 
> Thanks,
> Jeffrey




From dpatel at apple.com  Mon Mar  8 12:21:24 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 8 Mar 2010 10:21:24 -0800
Subject: [llvm-commits] [PATCH] Delete the CompileUnits that DwarfDebug
	allocates
In-Reply-To: <bb55930e1003060151r37644c6at96a3ae4a87db99a2@mail.gmail.com>
References: <bb55930e1003060151r37644c6at96a3ae4a87db99a2@mail.gmail.com>
Message-ID: <C6203620-6118-4974-BC22-A66A7A355B43@apple.com>

Hi Jeffrey,

In fact, I have a patch in my queue (which was waiting for 2.7 branch) that deletes CompileUnits itself. I should be able to apply it to mainline in a day or so.
-
Devang
On Mar 6, 2010, at 1:51 AM, Jeffrey Yasskin wrote:

> DwarfDebug new's CompileUnits in beginModule() but never deleted them.
> This patch deletes them in endModule(), but I don't know enough about
> how DwarfDebug is used to know for sure that's right. Please
> double-check it.
> 
> 
> 
> For those curious, the valgrind report _after_ this patch follows.
> DbgVarScope.ll was produced by `Debug/bin/clang -S -O0 -g -emit-llvm
> ../src/test/FrontendC/2010-02-16-DbgVarScope.c -o DbgVarScope.ll`
> 
> 
> 
> $ valgrind --dsymutil=yes --leak-check=full Debug/bin/llc
> DbgVarScope.ll  -o /dev/null
> ==14590== Memcheck, a memory error detector
> ==14590== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.
> ==14590== Using Valgrind-3.5.0 and LibVEX; rerun with -h for copyright info
> ==14590== Command: Debug/bin/llc
> /Users/jyasskin/src/llvm/clang/obj/DbgVarScope.ll -o /dev/null
> ==14590==
> ==14590== Warning: ignored attempt to set SIGUSR2 handler in sigaction();
> ==14590==          the SIGUSR2 signal is used internally by Valgrind
> ==14590==
> ==14590== HEAP SUMMARY:
> ==14590==     in use at exit: 20,550 bytes in 265 blocks
> ==14590==   total heap usage: 4,197 allocs, 3,932 frees, 830,632 bytes allocated
> ==14590==
> ==14590== 16 bytes in 1 blocks are definitely lost in loss record 15 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x22E89B: createX86MCInstPrinter(llvm::Target const&,
> unsigned int, llvm::MCAsmInfo const&, llvm::raw_ostream&)
> (X86AsmPrinter.cpp:640)
> ==14590==    by 0x45DDEF: llvm::Target::createMCInstPrinter(unsigned
> int, llvm::MCAsmInfo const&, llvm::raw_ostream&) const
> (TargetRegistry.h:257)
> ==14590==    by 0x45C66F:
> llvm::LLVMTargetMachine::addPassesToEmitFile(llvm::PassManagerBase&,
> llvm::formatted_raw_ostream&, llvm::TargetMachine::CodeGenFileType,
> llvm::CodeGenOpt::Level, bool) (LLVMTargetMachine.cpp:136)
> ==14590==    by 0x3478: main (llc.cpp:359)
> ==14590==
> ==14590== 22 bytes in 1 blocks are possibly lost in loss record 47 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x1B64A69: std::string::_Rep::_S_create(unsigned long,
> unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65753: char* std::string::_S_construct<char
> const*>(char const*, char const*, std::allocator<char> const&,
> std::forward_iterator_tag) (in /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65985: std::basic_string<char,
> std::char_traits<char>, std::allocator<char> >::basic_string(char
> const*, unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0xE6F21: llvm::StringRef::str() const (ADT/StringRef.h:148)
> ==14590==    by 0xE28F2: llvm::StringRef::operator std::string() const
> (ADT/StringRef.h:164)
> ==14590==    by 0x7ACEA7: llvm::sys::Path::Path(llvm::StringRef)
> (Unix/Path.inc:82)
> ==14590==    by 0x25F8: GetOutputStream(char const*, char const*) (llc.cpp:137)
> ==14590==    by 0x3018: main (llc.cpp:293)
> ==14590==
> ==14590== 32 bytes in 2 blocks are definitely lost in loss record 131 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x1B64A69: std::string::_Rep::_S_create(unsigned long,
> unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65753: char* std::string::_S_construct<char
> const*>(char const*, char const*, std::allocator<char> const&,
> std::forward_iterator_tag) (in /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65985: std::basic_string<char,
> std::char_traits<char>, std::allocator<char> >::basic_string(char
> const*, unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0xE6F21: llvm::StringRef::str() const (ADT/StringRef.h:148)
> ==14590==    by 0xE28F2: llvm::StringRef::operator std::string() const
> (ADT/StringRef.h:164)
> ==14590==    by 0x678AD1: llvm::MCSymbol::MCSymbol(llvm::StringRef,
> bool) (MCSymbol.h:60)
> ==14590==    by 0x678888:
> llvm::MCContext::GetOrCreateSymbol(llvm::StringRef) (MCContext.cpp:39)
> ==14590==    by 0x22C07D:
> llvm::X86AsmPrinter::GetGlobalValueSymbol(llvm::GlobalValue const*)
> const (X86AsmPrinter.cpp:59)
> ==14590==    by 0x40E298:
> llvm::DwarfDebug::constructGlobalVariableDIE(llvm::MDNode*)
> (DwarfDebug.cpp:1742)
> ==14590==    by 0x40E746: llvm::DwarfDebug::beginModule(llvm::Module*,
> llvm::MachineModuleInfo*) (DwarfDebug.cpp:1825)
> ==14590==    by 0x42D780:
> llvm::DwarfWriter::BeginModule(llvm::Module*,
> llvm::MachineModuleInfo*, llvm::raw_ostream&, llvm::AsmPrinter*,
> llvm::MCAsmInfo const*) (DwarfWriter.cpp:46)
> ==14590==
> ==14590== 38 bytes in 2 blocks are definitely lost in loss record 136 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x1B64A69: std::string::_Rep::_S_create(unsigned long,
> unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65753: char* std::string::_S_construct<char
> const*>(char const*, char const*, std::allocator<char> const&,
> std::forward_iterator_tag) (in /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65985: std::basic_string<char,
> std::char_traits<char>, std::allocator<char> >::basic_string(char
> const*, unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0xE6F21: llvm::StringRef::str() const (ADT/StringRef.h:148)
> ==14590==    by 0xE28F2: llvm::StringRef::operator std::string() const
> (ADT/StringRef.h:164)
> ==14590==    by 0x678AD1: llvm::MCSymbol::MCSymbol(llvm::StringRef,
> bool) (MCSymbol.h:60)
> ==14590==    by 0x678888:
> llvm::MCContext::GetOrCreateSymbol(llvm::StringRef) (MCContext.cpp:39)
> ==14590==    by 0x47CFC9:
> llvm::MachineBasicBlock::getSymbol(llvm::MCContext&) const
> (MachineBasicBlock.cpp:50)
> ==14590==    by 0x232BA2:
> llvm::X86MCInstLower::Lower(llvm::MachineInstr const*, llvm::MCInst&)
> const (X86MCInstLower.cpp:236)
> ==14590==    by 0x233B92:
> llvm::X86AsmPrinter::EmitInstruction(llvm::MachineInstr const*)
> (X86MCInstLower.cpp:423)
> ==14590==    by 0x3FCC58: llvm::AsmPrinter::EmitFunctionBody()
> (AsmPrinter.cpp:427)
> ==14590==
> ==14590== 64 bytes in 4 blocks are definitely lost in loss record 143 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x4050AC: llvm::DIE::addSiblingOffset() (DIE.cpp:118)
> ==14590==    by 0x410497:
> llvm::DwarfDebug::computeSizeAndOffset(llvm::DIE*, unsigned int, bool)
> (DwarfDebug.cpp:2309)
> ==14590==    by 0x410642:
> llvm::DwarfDebug::computeSizeAndOffset(llvm::DIE*, unsigned int, bool)
> (DwarfDebug.cpp:2338)
> ==14590==    by 0x4106C7: llvm::DwarfDebug::computeSizeAndOffsets()
> (DwarfDebug.cpp:2358)
> ==14590==    by 0x40ED26: llvm::DwarfDebug::endModule() (DwarfDebug.cpp:1912)
> ==14590==    by 0x42D7A9: llvm::DwarfWriter::EndModule() (DwarfWriter.cpp:53)
> ==14590==    by 0x3FCEFA:
> llvm::AsmPrinter::doFinalization(llvm::Module&) (AsmPrinter.cpp:467)
> ==14590==    by 0x72CA76:
> llvm::FPPassManager::doFinalization(llvm::Module&)
> (PassManager.cpp:1389)
> ==14590==    by 0x72C360:
> llvm::FunctionPassManagerImpl::doFinalization(llvm::Module&)
> (PassManager.cpp:1266)
> ==14590==    by 0x72C28D: llvm::FunctionPassManager::doFinalization()
> (PassManager.cpp:1244)
> ==14590==    by 0x3663: main (llc.cpp:382)
> ==14590==
> ==14590== 72 bytes in 4 blocks are definitely lost in loss record 144 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x1B64A69: std::string::_Rep::_S_create(unsigned long,
> unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65753: char* std::string::_S_construct<char
> const*>(char const*, char const*, std::allocator<char> const&,
> std::forward_iterator_tag) (in /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0x1B65985: std::basic_string<char,
> std::char_traits<char>, std::allocator<char> >::basic_string(char
> const*, unsigned long, std::allocator<char> const&) (in
> /opt/local/lib/gcc44/libstdc++.6.dylib)
> ==14590==    by 0xE6F21: llvm::StringRef::str() const (ADT/StringRef.h:148)
> ==14590==    by 0xE28F2: llvm::StringRef::operator std::string() const
> (ADT/StringRef.h:164)
> ==14590==    by 0x678AD1: llvm::MCSymbol::MCSymbol(llvm::StringRef,
> bool) (MCSymbol.h:60)
> ==14590==    by 0x678888:
> llvm::MCContext::GetOrCreateSymbol(llvm::StringRef) (MCContext.cpp:39)
> ==14590==    by 0x22C07D:
> llvm::X86AsmPrinter::GetGlobalValueSymbol(llvm::GlobalValue const*)
> const (X86AsmPrinter.cpp:59)
> ==14590==    by 0x3FD530:
> llvm::AsmPrinter::SetupMachineFunction(llvm::MachineFunction&)
> (AsmPrinter.cpp:542)
> ==14590==    by 0x22C175:
> llvm::X86AsmPrinter::runOnMachineFunction(llvm::MachineFunction&)
> (X86AsmPrinter.cpp:79)
> ==14590==    by 0x48EC1A:
> llvm::MachineFunctionPass::runOnFunction(llvm::Function&)
> (MachineFunctionPass.cpp:27)
> ==14590==
> ==14590== 96 bytes in 2 blocks are definitely lost in loss record 146 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0x17177: llvm::LLParser::ParseValID(llvm::ValID&,
> llvm::LLParser::PerFunctionState*) (LLParser.cpp:1985)
> ==14590==    by 0x1B139: llvm::LLParser::ParseValue(llvm::Type const*,
> llvm::Value*&, llvm::LLParser::PerFunctionState&) (LLParser.cpp:2581)
> ==14590==    by 0x13A9A:
> llvm::LLParser::ParseParameterList(llvm::SmallVectorImpl<llvm::LLParser::ParamInfo>&,
> llvm::LLParser::PerFunctionState&) (LLParser.cpp:1434)
> ==14590==    by 0x20B15:
> llvm::LLParser::ParseCall(llvm::Instruction*&,
> llvm::LLParser::PerFunctionState&, bool) (LLParser.cpp:3624)
> ==14590==    by 0x1D8A5:
> llvm::LLParser::ParseInstruction(llvm::Instruction*&,
> llvm::BasicBlock*, llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:3027)
> ==14590==    by 0x1CE12:
> llvm::LLParser::ParseBasicBlock(llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:2892)
> ==14590==    by 0x1CB58:
> llvm::LLParser::ParseFunctionBody(llvm::Function&) (LLParser.cpp:2843)
> ==14590==    by 0xF2F0: llvm::LLParser::ParseDefine() (LLParser.cpp:392)
> ==14590==    by 0xE0F0: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:161)
> ==14590==
> ==14590== 96 bytes in 2 blocks are definitely lost in loss record 147 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xF9D9: llvm::LLParser::ParseMDNodeID(llvm::MDNode*&)
> (LLParser.cpp:491)
> ==14590==    by 0x171CD: llvm::LLParser::ParseValID(llvm::ValID&,
> llvm::LLParser::PerFunctionState*) (LLParser.cpp:1993)
> ==14590==    by 0x1B139: llvm::LLParser::ParseValue(llvm::Type const*,
> llvm::Value*&, llvm::LLParser::PerFunctionState&) (LLParser.cpp:2581)
> ==14590==    by 0x13A9A:
> llvm::LLParser::ParseParameterList(llvm::SmallVectorImpl<llvm::LLParser::ParamInfo>&,
> llvm::LLParser::PerFunctionState&) (LLParser.cpp:1434)
> ==14590==    by 0x20B15:
> llvm::LLParser::ParseCall(llvm::Instruction*&,
> llvm::LLParser::PerFunctionState&, bool) (LLParser.cpp:3624)
> ==14590==    by 0x1D8A5:
> llvm::LLParser::ParseInstruction(llvm::Instruction*&,
> llvm::BasicBlock*, llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:3027)
> ==14590==    by 0x1CE12:
> llvm::LLParser::ParseBasicBlock(llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:2892)
> ==14590==    by 0x1CB58:
> llvm::LLParser::ParseFunctionBody(llvm::Function&) (LLParser.cpp:2843)
> ==14590==    by 0xF2F0: llvm::LLParser::ParseDefine() (LLParser.cpp:392)
> ==14590==
> ==14590== 96 bytes in 2 blocks are definitely lost in loss record 148 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xF9D9: llvm::LLParser::ParseMDNodeID(llvm::MDNode*&)
> (LLParser.cpp:491)
> ==14590==    by 0xFC6F: llvm::LLParser::ParseNamedMetadata() (LLParser.cpp:525)
> ==14590==    by 0xE21C: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:172)
> ==14590==    by 0xD146: llvm::LLParser::Run() (LLParser.cpp:36)
> ==14590==    by 0x2EDC3: llvm::ParseAssembly(llvm::MemoryBuffer*,
> llvm::Module*, llvm::SMDiagnostic&, llvm::LLVMContext&)
> (Parser.cpp:37)
> ==14590==    by 0x4512: llvm::ParseIR(llvm::MemoryBuffer*,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:88)
> ==14590==    by 0x468B: llvm::ParseIRFile(std::string const&,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:105)
> ==14590==    by 0x2AC0: main (llc.cpp:228)
> ==14590==
> ==14590== 288 (144 direct, 144 indirect) bytes in 1 blocks are
> definitely lost in loss record 153 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x40C39D:
> llvm::DwarfDebug::getOrCreateAbstractScope(llvm::MDNode*)
> (DwarfDebug.cpp:1326)
> ==14590==    by 0x40C371:
> llvm::DwarfDebug::getOrCreateAbstractScope(llvm::MDNode*)
> (DwarfDebug.cpp:1323)
> ==14590==    by 0x40C268:
> llvm::DwarfDebug::getUpdatedDbgScope(llvm::MDNode*, llvm::MachineInstr
> const*, llvm::MDNode*) (DwarfDebug.cpp:1303)
> ==14590==    by 0x40F828: llvm::DwarfDebug::extractScopeInformation()
> (DwarfDebug.cpp:2102)
> ==14590==    by 0x40FC8F:
> llvm::DwarfDebug::beginFunction(llvm::MachineFunction const*)
> (DwarfDebug.cpp:2160)
> ==14590==    by 0x42D833:
> llvm::DwarfWriter::BeginFunction(llvm::MachineFunction const*)
> (DwarfWriter.cpp:62)
> ==14590==    by 0x3FC5FD: llvm::AsmPrinter::EmitFunctionHeader()
> (AsmPrinter.cpp:317)
> ==14590==    by 0x22C23B:
> llvm::X86AsmPrinter::runOnMachineFunction(llvm::MachineFunction&)
> (X86AsmPrinter.cpp:90)
> ==14590==    by 0x48EC1A:
> llvm::MachineFunctionPass::runOnFunction(llvm::Function&)
> (MachineFunctionPass.cpp:27)
> ==14590==    by 0x72C7F2:
> llvm::FPPassManager::runOnFunction(llvm::Function&)
> (PassManager.cpp:1350)
> ==14590==    by 0x72C555:
> llvm::FunctionPassManagerImpl::run(llvm::Function&)
> (PassManager.cpp:1301)
> ==14590==
> ==14590== 432 bytes in 9 blocks are definitely lost in loss record 155 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xF9D9: llvm::LLParser::ParseMDNodeID(llvm::MDNode*&)
> (LLParser.cpp:491)
> ==14590==    by 0x1247A:
> llvm::LLParser::ParseInstructionMetadata(llvm::SmallVectorImpl<std::pair<unsigned
> int, llvm::MDNode*> >&) (LLParser.cpp:1092)
> ==14590==    by 0x1CE9D:
> llvm::LLParser::ParseBasicBlock(llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:2905)
> ==14590==    by 0x1CB58:
> llvm::LLParser::ParseFunctionBody(llvm::Function&) (LLParser.cpp:2843)
> ==14590==    by 0xF2F0: llvm::LLParser::ParseDefine() (LLParser.cpp:392)
> ==14590==    by 0xE0F0: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:161)
> ==14590==    by 0xD146: llvm::LLParser::Run() (LLParser.cpp:36)
> ==14590==    by 0x2EDC3: llvm::ParseAssembly(llvm::MemoryBuffer*,
> llvm::Module*, llvm::SMDiagnostic&, llvm::LLVMContext&)
> (Parser.cpp:37)
> ==14590==    by 0x4512: llvm::ParseIR(llvm::MemoryBuffer*,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:88)
> ==14590==
> ==14590== 432 bytes in 9 blocks are definitely lost in loss record 156 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xF9D9: llvm::LLParser::ParseMDNodeID(llvm::MDNode*&)
> (LLParser.cpp:491)
> ==14590==    by 0x1247A:
> llvm::LLParser::ParseInstructionMetadata(llvm::SmallVectorImpl<std::pair<unsigned
> int, llvm::MDNode*> >&) (LLParser.cpp:1092)
> ==14590==    by 0x1CE7D:
> llvm::LLParser::ParseBasicBlock(llvm::LLParser::PerFunctionState&)
> (LLParser.cpp:2899)
> ==14590==    by 0x1CB58:
> llvm::LLParser::ParseFunctionBody(llvm::Function&) (LLParser.cpp:2843)
> ==14590==    by 0xF2F0: llvm::LLParser::ParseDefine() (LLParser.cpp:392)
> ==14590==    by 0xE0F0: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:161)
> ==14590==    by 0xD146: llvm::LLParser::Run() (LLParser.cpp:36)
> ==14590==    by 0x2EDC3: llvm::ParseAssembly(llvm::MemoryBuffer*,
> llvm::Module*, llvm::SMDiagnostic&, llvm::LLVMContext&)
> (Parser.cpp:37)
> ==14590==    by 0x4512: llvm::ParseIR(llvm::MemoryBuffer*,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:88)
> ==14590==
> ==14590== 816 bytes in 17 blocks are definitely lost in loss record 169 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xF9D9: llvm::LLParser::ParseMDNodeID(llvm::MDNode*&)
> (LLParser.cpp:491)
> ==14590==    by 0x171CD: llvm::LLParser::ParseValID(llvm::ValID&,
> llvm::LLParser::PerFunctionState*) (LLParser.cpp:1993)
> ==14590==    by 0x22765:
> llvm::LLParser::ParseMDNodeVector(llvm::SmallVectorImpl<llvm::Value*>&,
> llvm::LLParser::PerFunctionState*) (LLParser.cpp:3928)
> ==14590==    by 0xFEB9: llvm::LLParser::ParseStandaloneMetadata()
> (LLParser.cpp:546)
> ==14590==    by 0xE1FE: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:171)
> ==14590==    by 0xD146: llvm::LLParser::Run() (LLParser.cpp:36)
> ==14590==    by 0x2EDC3: llvm::ParseAssembly(llvm::MemoryBuffer*,
> llvm::Module*, llvm::SMDiagnostic&, llvm::LLVMContext&)
> (Parser.cpp:37)
> ==14590==    by 0x4512: llvm::ParseIR(llvm::MemoryBuffer*,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:88)
> ==14590==    by 0x468B: llvm::ParseIRFile(std::string const&,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:105)
> ==14590==
> ==14590== 864 (432 direct, 432 indirect) bytes in 3 blocks are
> definitely lost in loss record 170 of 173
> ==14590==    at 0x190FAD3: operator new(unsigned long) (vg_replace_malloc.c:220)
> ==14590==    by 0x40F2CA:
> llvm::DwarfDebug::createDbgScope(llvm::MDNode*, llvm::MDNode*)
> (DwarfDebug.cpp:2038)
> ==14590==    by 0x40F669: llvm::DwarfDebug::extractScopeInformation()
> (DwarfDebug.cpp:2081)
> ==14590==    by 0x40FC8F:
> llvm::DwarfDebug::beginFunction(llvm::MachineFunction const*)
> (DwarfDebug.cpp:2160)
> ==14590==    by 0x42D833:
> llvm::DwarfWriter::BeginFunction(llvm::MachineFunction const*)
> (DwarfWriter.cpp:62)
> ==14590==    by 0x3FC5FD: llvm::AsmPrinter::EmitFunctionHeader()
> (AsmPrinter.cpp:317)
> ==14590==    by 0x22C23B:
> llvm::X86AsmPrinter::runOnMachineFunction(llvm::MachineFunction&)
> (X86AsmPrinter.cpp:90)
> ==14590==    by 0x48EC1A:
> llvm::MachineFunctionPass::runOnFunction(llvm::Function&)
> (MachineFunctionPass.cpp:27)
> ==14590==    by 0x72C7F2:
> llvm::FPPassManager::runOnFunction(llvm::Function&)
> (PassManager.cpp:1350)
> ==14590==    by 0x72C555:
> llvm::FunctionPassManagerImpl::run(llvm::Function&)
> (PassManager.cpp:1301)
> ==14590==    by 0x72C249:
> llvm::FunctionPassManager::run(llvm::Function&) (PassManager.cpp:1231)
> ==14590==    by 0x3627: main (llc.cpp:379)
> ==14590==
> ==14590== 6,764 (700 direct, 6,064 indirect) bytes in 5 blocks are
> definitely lost in loss record 173 of 173
> ==14590==    at 0x190F416: malloc (vg_replace_malloc.c:195)
> ==14590==    by 0x7198DB: llvm::MDNode::getMDNode(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int, llvm::MDNode::FunctionLocalness,
> bool) (Metadata.cpp:227)
> ==14590==    by 0x7199A0: llvm::MDNode::get(llvm::LLVMContext&,
> llvm::Value* const*, unsigned int) (Metadata.cpp:237)
> ==14590==    by 0xFF2B: llvm::LLParser::ParseStandaloneMetadata()
> (LLParser.cpp:555)
> ==14590==    by 0xE1FE: llvm::LLParser::ParseTopLevelEntities()
> (LLParser.cpp:171)
> ==14590==    by 0xD146: llvm::LLParser::Run() (LLParser.cpp:36)
> ==14590==    by 0x2EDC3: llvm::ParseAssembly(llvm::MemoryBuffer*,
> llvm::Module*, llvm::SMDiagnostic&, llvm::LLVMContext&)
> (Parser.cpp:37)
> ==14590==    by 0x4512: llvm::ParseIR(llvm::MemoryBuffer*,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:88)
> ==14590==    by 0x468B: llvm::ParseIRFile(std::string const&,
> llvm::SMDiagnostic&, llvm::LLVMContext&) (IRReader.h:105)
> ==14590==    by 0x2AC0: main (llc.cpp:228)
> ==14590==
> ==14590== LEAK SUMMARY:
> ==14590==    definitely lost: 3,466 bytes in 63 blocks
> ==14590==    indirectly lost: 6,640 bytes in 42 blocks
> ==14590==      possibly lost: 22 bytes in 1 blocks
> ==14590==    still reachable: 10,042 bytes in 150 blocks
> ==14590==         suppressed: 380 bytes in 9 blocks
> ==14590== Reachable blocks (those to which a pointer was found) are not shown.
> ==14590== To see them, rerun with: --leak-check=full --show-reachable=yes
> ==14590==
> ==14590== For counts of detected and suppressed errors, rerun with: -v
> ==14590== ERROR SUMMARY: 15 errors from 15 contexts (suppressed: 0 from 0)
> <delete_compile_units.patch>




From dpatel at apple.com  Mon Mar  8 12:25:49 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 18:25:49 -0000
Subject: [llvm-commits] [llvm] r97947 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/AsmPrinter/DwarfDebug.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.h lib/CodeGen/MachineInstr.cpp
 lib/Target/PIC16/PIC16DebugInfo.cpp lib/Transforms/Utils/CloneFunction.cpp
Message-ID: <20100308182549.3AFF82A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 12:25:48 2010
New Revision: 97947

URL: http://llvm.org/viewvc/llvm-project?rev=97947&view=rev
Log:
Avoid using DIDescriptor.isNull().
This is a first step towards eliminating unncessary constructor checks in light weight DIDescriptor wrappers.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/MachineInstr.cpp
    llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
    llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 12:25:48 2010
@@ -67,7 +67,7 @@
     explicit DIDescriptor() : DbgNode(0) {}
     explicit DIDescriptor(MDNode *N) : DbgNode(N) {}
 
-    bool isNull() const { return DbgNode == 0; }
+    bool Verify() const { return DbgNode != 0; }
 
     MDNode *getNode() const { return DbgNode; }
 
@@ -246,7 +246,9 @@
     bool isArtificial() const {
       return (getFlags() & FlagArtificial) != 0;
     }
-
+    bool isValid() const {
+      return DbgNode && (isBasicType() || isDerivedType() || isCompositeType());
+    }
     /// dump - print type.
     void dump() const;
   };
@@ -360,7 +362,7 @@
     /// DIType or as DICompositeType.
     StringRef getReturnTypeName() const {
       DICompositeType DCT(getFieldAs<DICompositeType>(8));
-      if (!DCT.isNull()) {
+      if (DCT.Verify()) {
         DIArray A = DCT.getTypeArray();
         DIType T(A.getElement(0).getNode());
         return T.getName();
@@ -494,6 +496,7 @@
     DILocation getOrigLocation() const { return getFieldAs<DILocation>(3); }
     StringRef getFilename() const    { return getScope().getFilename(); }
     StringRef getDirectory() const   { return getScope().getDirectory(); }
+    bool Verify() const;
   };
 
   /// DIFactory - This object assists with the construction of the various

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 12:25:48 2010
@@ -132,13 +132,12 @@
 /// isBasicType - Return true if the specified tag is legal for
 /// DIBasicType.
 bool DIDescriptor::isBasicType() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_base_type;
+  return DbgNode && getTag() == dwarf::DW_TAG_base_type;
 }
 
 /// isDerivedType - Return true if the specified tag is legal for DIDerivedType.
 bool DIDescriptor::isDerivedType() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_typedef:
   case dwarf::DW_TAG_pointer_type:
@@ -158,7 +157,7 @@
 /// isCompositeType - Return true if the specified tag is legal for
 /// DICompositeType.
 bool DIDescriptor::isCompositeType() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_array_type:
   case dwarf::DW_TAG_structure_type:
@@ -175,7 +174,7 @@
 
 /// isVariable - Return true if the specified tag is legal for DIVariable.
 bool DIDescriptor::isVariable() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_auto_variable:
   case dwarf::DW_TAG_arg_variable:
@@ -194,15 +193,13 @@
 /// isSubprogram - Return true if the specified tag is legal for
 /// DISubprogram.
 bool DIDescriptor::isSubprogram() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_subprogram;
+  return DbgNode && getTag() == dwarf::DW_TAG_subprogram;
 }
 
 /// isGlobalVariable - Return true if the specified tag is legal for
 /// DIGlobalVariable.
 bool DIDescriptor::isGlobalVariable() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_variable;
+  return DbgNode && getTag() == dwarf::DW_TAG_variable;
 }
 
 /// isGlobal - Return true if the specified tag is legal for DIGlobal.
@@ -213,7 +210,7 @@
 /// isScope - Return true if the specified tag is one of the scope
 /// related tag.
 bool DIDescriptor::isScope() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_compile_unit:
   case dwarf::DW_TAG_lexical_block:
@@ -228,32 +225,27 @@
 
 /// isCompileUnit - Return true if the specified tag is DW_TAG_compile_unit.
 bool DIDescriptor::isCompileUnit() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_compile_unit;
+  return DbgNode && getTag() == dwarf::DW_TAG_compile_unit;
 }
 
 /// isNameSpace - Return true if the specified tag is DW_TAG_namespace.
 bool DIDescriptor::isNameSpace() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_namespace;
+  return DbgNode && getTag() == dwarf::DW_TAG_namespace;
 }
 
 /// isLexicalBlock - Return true if the specified tag is DW_TAG_lexical_block.
 bool DIDescriptor::isLexicalBlock() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_lexical_block;
+  return DbgNode && getTag() == dwarf::DW_TAG_lexical_block;
 }
 
 /// isSubrange - Return true if the specified tag is DW_TAG_subrange_type.
 bool DIDescriptor::isSubrange() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_subrange_type;
+  return DbgNode && getTag() == dwarf::DW_TAG_subrange_type;
 }
 
 /// isEnumerator - Return true if the specified tag is DW_TAG_enumerator.
 bool DIDescriptor::isEnumerator() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_enumerator;
+  return DbgNode && getTag() == dwarf::DW_TAG_enumerator;
 }
 
 //===----------------------------------------------------------------------===//
@@ -268,7 +260,8 @@
 }
 
 unsigned DIArray::getNumElements() const {
-  assert(DbgNode && "Invalid DIArray");
+  if (!DbgNode)
+    return 0;
   return DbgNode->getNumOperands();
 }
 
@@ -276,11 +269,9 @@
 /// this descriptor. After this completes, the current debug info value
 /// is erased.
 void DIDerivedType::replaceAllUsesWith(DIDescriptor &D) {
-  if (isNull())
+  if (!DbgNode)
     return;
 
-  assert(!D.isNull() && "Can not replace with null");
-
   // Since we use a TrackingVH for the node, its easy for clients to manufacture
   // legitimate situations where they want to replaceAllUsesWith() on something
   // which, due to uniquing, has merged with the source. We shield clients from
@@ -295,7 +286,7 @@
 
 /// Verify - Verify that a compile unit is well formed.
 bool DICompileUnit::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
   StringRef N = getFilename();
   if (N.empty())
@@ -306,36 +297,36 @@
 
 /// Verify - Verify that a type descriptor is well formed.
 bool DIType::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a composite type descriptor is well formed.
 bool DICompositeType::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a subprogram descriptor is well formed.
 bool DISubprogram::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
@@ -343,24 +334,24 @@
     return false;
 
   DICompositeType Ty = getType();
-  if (!Ty.isNull() && !Ty.Verify())
+  if (!Ty.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a global variable descriptor is well formed.
 bool DIGlobalVariable::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
   if (getDisplayName().empty())
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
 
   DIType Ty = getType();
@@ -375,10 +366,10 @@
 
 /// Verify - Verify that a variable descriptor is well formed.
 bool DIVariable::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DIType Ty = getType();
@@ -388,6 +379,14 @@
   return true;
 }
 
+/// Verify - Verify that a location descriptor is well formed.
+bool DILocation::Verify() const {
+  if (!DbgNode)
+    return false;
+  
+  return DbgNode->getNumOperands() == 4;
+}
+
 /// getOriginalTypeSize - If this type is derived from a base type then
 /// return base type size.
 uint64_t DIDerivedType::getOriginalTypeSize() const {
@@ -398,7 +397,7 @@
     DIType BaseType = getTypeDerivedFrom();
     // If this type is not derived from any type then take conservative 
     // approach.
-    if (BaseType.isNull())
+    if (!BaseType.isValid())
       return getSizeInBits();
     if (BaseType.isDerivedType())
       return DIDerivedType(BaseType.getNode()).getOriginalTypeSize();
@@ -468,7 +467,7 @@
 
 /// dump - Print type.
 void DIType::dump() const {
-  if (isNull()) return;
+  if (!DbgNode) return;
 
   StringRef Res = getName();
   if (!Res.empty())
@@ -521,8 +520,6 @@
 /// dump - Print composite type.
 void DICompositeType::dump() const {
   DIArray A = getTypeArray();
-  if (A.isNull())
-    return;
   dbgs() << " [" << A.getNumElements() << " elements]";
 }
 
@@ -1155,9 +1152,8 @@
 
 /// processLocation - Process DILocation.
 void DebugInfoFinder::processLocation(DILocation Loc) {
-  if (Loc.isNull()) return;
-  DIScope S(Loc.getScope().getNode());
-  if (S.isNull()) return;
+  if (!Loc.Verify()) return;
+  DIDescriptor S(Loc.getScope().getNode());
   if (S.isCompileUnit())
     addCompileUnit(DICompileUnit(S.getNode()));
   else if (S.isSubprogram())
@@ -1177,26 +1173,21 @@
     DICompositeType DCT(DT.getNode());
     processType(DCT.getTypeDerivedFrom());
     DIArray DA = DCT.getTypeArray();
-    if (!DA.isNull())
-      for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
-        DIDescriptor D = DA.getElement(i);
-        DIType TyE = DIType(D.getNode());
-        if (!TyE.isNull())
-          processType(TyE);
-        else
-          processSubprogram(DISubprogram(D.getNode()));
-      }
+    for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
+      DIDescriptor D = DA.getElement(i);
+      if (D.isType())
+        processType(DIType(D.getNode()));
+      else if (D.isSubprogram())
+        processSubprogram(DISubprogram(D.getNode()));
+    }
   } else if (DT.isDerivedType()) {
     DIDerivedType DDT(DT.getNode());
-    if (!DDT.isNull())
-      processType(DDT.getTypeDerivedFrom());
+    processType(DDT.getTypeDerivedFrom());
   }
 }
 
 /// processLexicalBlock
 void DebugInfoFinder::processLexicalBlock(DILexicalBlock LB) {
-  if (LB.isNull())
-    return;
   DIScope Context = LB.getContext();
   if (Context.isLexicalBlock())
     return processLexicalBlock(DILexicalBlock(Context.getNode()));
@@ -1206,8 +1197,6 @@
 
 /// processSubprogram - Process DISubprogram.
 void DebugInfoFinder::processSubprogram(DISubprogram SP) {
-  if (SP.isNull())
-    return;
   if (!addSubprogram(SP))
     return;
   addCompileUnit(SP.getCompileUnit());
@@ -1216,20 +1205,23 @@
 
 /// processDeclare - Process DbgDeclareInst.
 void DebugInfoFinder::processDeclare(DbgDeclareInst *DDI) {
-  DIVariable DV(cast<MDNode>(DDI->getVariable()));
-  if (DV.isNull())
+  MDNode *N = dyn_cast<MDNode>(DDI->getVariable());
+  if (!N) return;
+
+  DIDescriptor DV(N);
+  if (!DV.isVariable())
     return;
 
   if (!NodesSeen.insert(DV.getNode()))
     return;
 
-  addCompileUnit(DV.getCompileUnit());
-  processType(DV.getType());
+  addCompileUnit(DIVariable(N).getCompileUnit());
+  processType(DIVariable(N).getType());
 }
 
 /// addType - Add type into Tys.
 bool DebugInfoFinder::addType(DIType DT) {
-  if (DT.isNull())
+  if (!DT.isValid())
     return false;
 
   if (!NodesSeen.insert(DT.getNode()))
@@ -1241,7 +1233,7 @@
 
 /// addCompileUnit - Add compile unit into CUs.
 bool DebugInfoFinder::addCompileUnit(DICompileUnit CU) {
-  if (CU.isNull())
+  if (!CU.Verify())
     return false;
 
   if (!NodesSeen.insert(CU.getNode()))
@@ -1253,7 +1245,7 @@
 
 /// addGlobalVariable - Add global variable into GVs.
 bool DebugInfoFinder::addGlobalVariable(DIGlobalVariable DIG) {
-  if (DIG.isNull())
+  if (!DIDescriptor(DIG.getNode()).isGlobalVariable())
     return false;
 
   if (!NodesSeen.insert(DIG.getNode()))
@@ -1265,7 +1257,7 @@
 
 // addSubprogram - Add subprgoram into SPs.
 bool DebugInfoFinder::addSubprogram(DISubprogram SP) {
-  if (SP.isNull())
+  if (!DIDescriptor(SP.getNode()).isSubprogram())
     return false;
 
   if (!NodesSeen.insert(SP.getNode()))
@@ -1283,10 +1275,10 @@
     return 0;
 
   for (unsigned i = 0, e = NMD->getNumOperands(); i != e; ++i) {
-    DIGlobalVariable DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
-    if (DIG.isNull())
+    DIDescriptor DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
+    if (!DIG.isGlobalVariable())
       continue;
-    if (DIG.getGlobal() == V)
+    if (DIGlobalVariable(DIG.getNode()).getGlobal() == V)
       return DIG.getNode();
   }
   return 0;
@@ -1378,12 +1370,6 @@
 /// getDISubprogram - Find subprogram that is enclosing this scope.
 DISubprogram llvm::getDISubprogram(MDNode *Scope) {
   DIDescriptor D(Scope);
-  if (D.isNull())
-    return DISubprogram();
-  
-  if (D.isCompileUnit())
-    return DISubprogram();
-  
   if (D.isSubprogram())
     return DISubprogram(Scope);
   
@@ -1395,9 +1381,6 @@
 
 /// getDICompositeType - Find underlying composite type.
 DICompositeType llvm::getDICompositeType(DIType T) {
-  if (T.isNull())
-    return DICompositeType();
-  
   if (T.isCompositeType())
     return DICompositeType(T.getNode());
   

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 12:25:48 2010
@@ -335,7 +335,7 @@
     // Print source line info.
     DIScope Scope = DLT.getScope();
     // Omit the directory, because it's likely to be long and uninteresting.
-    if (!Scope.isNull())
+    if (Scope.Verify())
       CommentOS << Scope.getFilename();
     else
       CommentOS << "<unknown>";
@@ -1287,7 +1287,7 @@
   if (DL.isUnknown())
     return;
   DILocation CurDLT = MF->getDILocation(DL);
-  if (CurDLT.getScope().isNull())
+  if (!CurDLT.getScope().Verify())
     return;
 
   if (!BeforePrintingInsn) {

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 12:25:48 2010
@@ -413,7 +413,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIVariable *V) {
   // If there is no compile unit specified, don't add a line #.
-  if (V->getCompileUnit().isNull())
+  if (!V->getCompileUnit().Verify())
     return;
 
   unsigned Line = V->getLineNumber();
@@ -427,7 +427,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIGlobal *G) {
   // If there is no compile unit specified, don't add a line #.
-  if (G->getCompileUnit().isNull())
+  if (!G->getCompileUnit().Verify())
     return;
 
   unsigned Line = G->getLineNumber();
@@ -441,7 +441,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DISubprogram *SP) {
   // If there is no compile unit specified, don't add a line #.
-  if (SP->getCompileUnit().isNull())
+  if (!SP->getCompileUnit().Verify())
     return;
   // If the line number is 0, don't add it.
   if (SP->getLineNumber() == 0)
@@ -460,7 +460,7 @@
 void DwarfDebug::addSourceLine(DIE *Die, const DIType *Ty) {
   // If there is no compile unit specified, don't add a line #.
   DICompileUnit CU = Ty->getCompileUnit();
-  if (CU.isNull())
+  if (!CU.Verify())
     return;
 
   unsigned Line = Ty->getLineNumber();
@@ -474,7 +474,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DINameSpace *NS) {
   // If there is no compile unit specified, don't add a line #.
-  if (NS->getCompileUnit().isNull())
+  if (!NS->getCompileUnit().Verify())
     return;
 
   unsigned Line = NS->getLineNumber();
@@ -526,12 +526,8 @@
   }
 
   DICompositeType blockStruct = DICompositeType(subType.getNode());
-
   DIArray Elements = blockStruct.getTypeArray();
 
-  if (Elements.isNull())
-    return Ty;
-
   for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Elements.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -677,7 +673,6 @@
   DIDescriptor varField = DIDescriptor();
   DIDescriptor forwardingField = DIDescriptor();
 
-
   for (unsigned i = 0, N = Fields.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Fields.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -688,10 +683,6 @@
       varField = Element;
   }
 
-  assert(!varField.isNull() && "Can't find byref variable in Block struct");
-  assert(!forwardingField.isNull()
-         && "Can't find forwarding field in Block struct");
-
   // Get the offsets for the forwarding field and the variable field.
   unsigned int forwardingFieldOffset =
     DIDerivedType(forwardingField.getNode()).getOffsetInBits() >> 3;
@@ -781,9 +772,7 @@
 
 /// addToContextOwner - Add Die into the list of its context owner's children.
 void DwarfDebug::addToContextOwner(DIE *Die, DIDescriptor Context) {
-  if (Context.isNull())
-    ModuleCU->addDie(Die);
-  else if (Context.isType()) {
+  if (Context.isType()) {
     DIE *ContextDIE = getOrCreateTypeDIE(DIType(Context.getNode()));
     ContextDIE->addChild(Die);
   } else if (Context.isNameSpace()) {
@@ -820,7 +809,7 @@
 
 /// addType - Add a new type attribute to the specified entity.
 void DwarfDebug::addType(DIE *Entity, DIType Ty) {
-  if (Ty.isNull())
+  if (!Ty.isValid())
     return;
 
   // Check for pre-existence.
@@ -906,9 +895,9 @@
     // Add enumerators to enumeration type.
     for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
       DIE *ElemDie = NULL;
-      DIEnumerator Enum(Elements.getElement(i).getNode());
-      if (!Enum.isNull()) {
-        ElemDie = constructEnumTypeDIE(&Enum);
+      DIDescriptor Enum(Elements.getElement(i).getNode());
+      if (Enum.isEnumerator()) {
+        ElemDie = constructEnumTypeDIE(DIEnumerator(Enum.getNode()));
         Buffer.addChild(ElemDie);
       }
     }
@@ -939,18 +928,17 @@
     DIArray Elements = CTy.getTypeArray();
 
     // A forward struct declared type may not have elements available.
-    if (Elements.isNull())
+    unsigned N = Elements.getNumElements();
+    if (N == 0)
       break;
 
     // Add elements to structure type.
-    for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
+    for (unsigned i = 0; i < N; ++i) {
       DIDescriptor Element = Elements.getElement(i);
-      if (Element.isNull())
-        continue;
       DIE *ElemDie = NULL;
-      if (Element.getTag() == dwarf::DW_TAG_subprogram)
+      if (Element.isSubprogram())
         ElemDie = createSubprogramDIE(DISubprogram(Element.getNode()));
-      else if (Element.getTag() == dwarf::DW_TAG_auto_variable) {
+      else if (Element.isVariable()) {
         DIVariable DV(Element.getNode());
         ElemDie = new DIE(dwarf::DW_TAG_variable);
         addString(ElemDie, dwarf::DW_AT_name, dwarf::DW_FORM_string,
@@ -959,8 +947,10 @@
         addUInt(ElemDie, dwarf::DW_AT_declaration, dwarf::DW_FORM_flag, 1);
         addUInt(ElemDie, dwarf::DW_AT_external, dwarf::DW_FORM_flag, 1);
         addSourceLine(ElemDie, &DV);
-      } else
+      } else if (Element.isDerivedType())
         ElemDie = createMemberDIE(DIDerivedType(Element.getNode()));
+      else
+        continue;
       Buffer.addChild(ElemDie);
     }
 
@@ -973,7 +963,7 @@
               dwarf::DW_FORM_data1, RLang);
 
     DICompositeType ContainingType = CTy.getContainingType();
-    if (!ContainingType.isNull())
+    if (DIDescriptor(ContainingType.getNode()).isCompositeType())
       addDIEEntry(&Buffer, dwarf::DW_AT_containing_type, dwarf::DW_FORM_ref4, 
                   getOrCreateTypeDIE(DIType(ContainingType.getNode())));
     break;
@@ -1051,11 +1041,11 @@
 }
 
 /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator *ETy) {
+DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator ETy) {
   DIE *Enumerator = new DIE(dwarf::DW_TAG_enumerator);
-  StringRef Name = ETy->getName();
+  StringRef Name = ETy.getName();
   addString(Enumerator, dwarf::DW_AT_name, dwarf::DW_FORM_string, Name);
-  int64_t Value = ETy->getEnumValue();
+  int64_t Value = ETy.getEnumValue();
   addSInt(Enumerator, dwarf::DW_AT_const_value, dwarf::DW_FORM_sdata, Value);
   return Enumerator;
 }
@@ -1199,7 +1189,7 @@
   DIArray Args = SPTy.getTypeArray();
   unsigned SPTag = SPTy.getTag();
 
-  if (Args.isNull() || SPTag != dwarf::DW_TAG_subroutine_type)
+  if (Args.getNumElements() == 0 || SPTag != dwarf::DW_TAG_subroutine_type)
     addType(SPDie, SPTy);
   else
     addType(SPDie, DIType(Args.getElement(0).getNode()));
@@ -1282,11 +1272,9 @@
     Parent->addScope(NScope);
   } else if (DIDescriptor(N).isLexicalBlock()) {
     DILexicalBlock DB(N);
-    if (!DB.getContext().isNull()) {
-      Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
-      NScope->setParent(Parent);
-      Parent->addScope(NScope);
-    }
+    Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
+    NScope->setParent(Parent);
+    Parent->addScope(NScope);
   }
 
   NScope->setFirstInsn(MI);
@@ -1318,8 +1306,7 @@
   if (Scope.isLexicalBlock()) {
     DILexicalBlock DB(N);
     DIDescriptor ParentDesc = DB.getContext();
-    if (!ParentDesc.isNull())
-      Parent = getOrCreateAbstractScope(ParentDesc.getNode());
+    Parent = getOrCreateAbstractScope(ParentDesc.getNode());
   }
 
   AScope = new DbgScope(Parent, DIDescriptor(N), NULL);
@@ -1422,10 +1409,9 @@
   // Ignore empty scopes.
   if (StartID == EndID && StartID != 0)
     return NULL;
-
-  DIScope DS(Scope->getScopeNode());
-  if (DS.isNull())
+  if (!Scope->getScopeNode())
     return NULL;
+  DIScope DS(Scope->getScopeNode());
   DIE *ScopeDIE = new DIE(dwarf::DW_TAG_inlined_subroutine);
 
   DISubprogram InlinedSP = getDISubprogram(DS.getNode());
@@ -1547,15 +1533,12 @@
     return;
 
   DIArray Args = SPTy.getTypeArray();
-  if (Args.isNull()) 
-    return;
-
   for (unsigned i = 0, e = Args.getNumElements(); i != e; ++i) {
     DIType ATy(Args.getElement(i).getNode());
-    if (ATy.isNull())
+    if (!ATy.isValid())
       continue;
     DICompositeType CATy = getDICompositeType(ATy);
-    if (!CATy.isNull() && !CATy.getName().empty()) {
+    if (DIDescriptor(CATy.getNode()).Verify() && !CATy.getName().empty()) {
       if (DIEEntry *Entry = ModuleCU->getDIEEntry(CATy.getNode()))
         ModuleCU->addGlobalType(CATy.getName(), Entry->getEntry());
     }
@@ -1564,26 +1547,24 @@
 
 /// constructScopeDIE - Construct a DIE for this scope.
 DIE *DwarfDebug::constructScopeDIE(DbgScope *Scope) {
- if (!Scope)
-  return NULL;
- DIScope DS(Scope->getScopeNode());
- if (DS.isNull())
-   return NULL;
-
- DIE *ScopeDIE = NULL;
- if (Scope->getInlinedAt())
-   ScopeDIE = constructInlinedScopeDIE(Scope);
- else if (DS.isSubprogram()) {
-   if (Scope->isAbstractScope())
-     ScopeDIE = ModuleCU->getDIE(DS.getNode());
-   else
-     ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
- }
- else {
-   ScopeDIE = constructLexicalScopeDIE(Scope);
-   if (!ScopeDIE) return NULL;
- }
-
+  if (!Scope || !Scope->getScopeNode())
+    return NULL;
+ 
+  DIScope DS(Scope->getScopeNode());
+  DIE *ScopeDIE = NULL;
+  if (Scope->getInlinedAt())
+    ScopeDIE = constructInlinedScopeDIE(Scope);
+  else if (DS.isSubprogram()) {
+    if (Scope->isAbstractScope())
+      ScopeDIE = ModuleCU->getDIE(DS.getNode());
+    else
+      ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
+  }
+  else {
+    ScopeDIE = constructLexicalScopeDIE(Scope);
+    if (!ScopeDIE) return NULL;
+  }
+  
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
@@ -2069,7 +2050,6 @@
       if (DL.isUnknown()) continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
-      if (DLTScope.isNull()) continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2089,7 +2069,6 @@
       if (DL.isUnknown())  continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
-      if (DLTScope.isNull())  continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2168,7 +2147,7 @@
     DILocation DLT = MF->getDILocation(FDL);
     unsigned LabelID = 0;
     DISubprogram SP = getDISubprogram(DLT.getScope().getNode());
-    if (!SP.isNull())
+    if (SP.Verify())
       LabelID = recordSourceLine(SP.getLineNumber(), 0, 
                                  DLT.getScope().getNode());
     else

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 12:25:48 2010
@@ -346,7 +346,7 @@
                              DICompositeType *CTy);
 
   /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-  DIE *constructEnumTypeDIE(DIEnumerator *ETy);
+  DIE *constructEnumTypeDIE(DIEnumerator ETy);
 
   /// createGlobalVariableDIE - Create new DIE using GV.
   DIE *createGlobalVariableDIE(const DIGlobalVariable &GV);

Modified: llvm/trunk/lib/CodeGen/MachineInstr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineInstr.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineInstr.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineInstr.cpp Mon Mar  8 12:25:48 2010
@@ -1219,7 +1219,7 @@
     DIScope Scope = DLT.getScope();
     OS << " dbg:";
     // Omit the directory, since it's usually long and uninteresting.
-    if (!Scope.isNull())
+    if (Scope.Verify())
       OS << Scope.getFilename();
     else
       OS << "<unknown>";

Modified: llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp Mon Mar  8 12:25:48 2010
@@ -333,7 +333,7 @@
   for (DebugInfoFinder::iterator I = DbgFinder.type_begin(),
          E = DbgFinder.type_end(); I != E; ++I) {
     DICompositeType CTy(*I);
-    if (CTy.isNull())
+    if (!CTy.Verify())
       continue;
     if (CTy.getTag() == dwarf::DW_TAG_union_type ||
         CTy.getTag() == dwarf::DW_TAG_structure_type ) {

Modified: llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp?rev=97947&r1=97946&r2=97947&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp Mon Mar  8 12:25:48 2010
@@ -336,14 +336,14 @@
 
 static MDNode *UpdateInlinedAtInfo(MDNode *InsnMD, MDNode *TheCallMD) {
   DILocation ILoc(InsnMD);
-  if (ILoc.isNull()) return InsnMD;
+  if (!ILoc.Verify()) return InsnMD;
 
   DILocation CallLoc(TheCallMD);
-  if (CallLoc.isNull()) return InsnMD;
+  if (!CallLoc.Verify()) return InsnMD;
 
   DILocation OrigLocation = ILoc.getOrigLocation();
   MDNode *NewLoc = TheCallMD;
-  if (!OrigLocation.isNull())
+  if (OrigLocation.Verify())
     NewLoc = UpdateInlinedAtInfo(OrigLocation.getNode(), TheCallMD);
 
   Value *MDVs[] = {




From dpatel at apple.com  Mon Mar  8 12:26:24 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 18:26:24 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r97948 -
	/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
Message-ID: <20100308182625.00FEA2A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 12:26:24 2010
New Revision: 97948

URL: http://llvm.org/viewvc/llvm-project?rev=97948&view=rev
Log:
Avoid DIDescriptor.isNull() checks.

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=97948&r1=97947&r2=97948&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Mon Mar  8 12:26:24 2010
@@ -1147,7 +1147,7 @@
     MainTy = getOrCreateType(TYPE_MAIN_VARIANT(type));
 
   DIType Ty = createVariantType(type, MainTy);
-  if (!Ty.isNull())
+  if (!Ty.isValid())
     return Ty;
 
   // Work out details of type.




From sabre at nondot.org  Mon Mar  8 12:29:38 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:29:38 -0000
Subject: [llvm-commits] [llvm] r97950 -
	/llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td
Message-ID: <20100308182938.2C6632A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:29:38 2010
New Revision: 97950

URL: http://llvm.org/viewvc/llvm-project?rev=97950&view=rev
Log:
tidy up

Modified:
    llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td

Modified: llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td?rev=97950&r1=97949&r2=97950&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td (original)
+++ llvm/trunk/lib/Target/Alpha/AlphaInstrInfo.td Mon Mar  8 12:29:38 2010
@@ -892,7 +892,7 @@
       (COND_BRANCH_I (immBRCond 2),  GPRC:$RA, bb:$DISP)>;
 def : Pat<(brcond (setgt GPRC:$RA, 0), bb:$DISP), 
       (COND_BRANCH_I (immBRCond 3),  GPRC:$RA, bb:$DISP)>;
-def : Pat<(brcond (and   GPRC:$RA, 1), bb:$DISP), 
+def : Pat<(brcond (and GPRC:$RA, 1), bb:$DISP), 
       (COND_BRANCH_I (immBRCond 6),  GPRC:$RA, bb:$DISP)>;
 def : Pat<(brcond (setle GPRC:$RA, 0), bb:$DISP), 
       (COND_BRANCH_I (immBRCond 4),  GPRC:$RA, bb:$DISP)>;




From sabre at nondot.org  Mon Mar  8 12:36:19 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:36:19 -0000
Subject: [llvm-commits] [llvm] r97952 -
	/llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp
Message-ID: <20100308183619.CC9372A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:36:19 2010
New Revision: 97952

URL: http://llvm.org/viewvc/llvm-project?rev=97952&view=rev
Log:
Node arguments to type casts can have names too.  This code
needs to be majorly refactored, but this spot bugfix allows
things like:

 def vmrghw_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
                             (vector_shuffle (v4i32 node:$lhs), node:$rhs), [{
...


Modified:
    llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp

Modified: llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp?rev=97952&r1=97951&r2=97952&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp (original)
+++ llvm/trunk/utils/TableGen/CodeGenDAGPatterns.cpp Mon Mar  8 12:36:19 2010
@@ -1238,6 +1238,14 @@
                                 std::vector<std::pair<Init*, std::string> >()));
         return ParseTreePattern(Dag);
       }
+      
+      // Input argument?
+      if (R->getName() == "node") {
+        if (Dag->getArgName(0).empty())
+          error("'node' argument requires a name to match with operand list");
+        Args.push_back(Dag->getArgName(0));
+      }
+      
       New = new TreePatternNode(DI);
     } else if (DagInit *DI = dynamic_cast<DagInit*>(Arg)) {
       New = ParseTreePattern(DI);




From sabre at nondot.org  Mon Mar  8 12:44:04 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:44:04 -0000
Subject: [llvm-commits] [llvm] r97953 -
	/llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td
Message-ID: <20100308184405.0164D2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:44:04 2010
New Revision: 97953

URL: http://llvm.org/viewvc/llvm-project?rev=97953&view=rev
Log:
Fix a bunch of ambiguous patterns which tblgen happens to infer types
for, due to a bug.

Modified:
    llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td

Modified: llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td?rev=97953&r1=97952&r2=97953&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCInstrAltivec.td Mon Mar  8 12:44:04 2010
@@ -35,33 +35,33 @@
 
 
 def vmrglb_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGLShuffleMask(cast<ShuffleVectorSDNode>(N), 1, false);
 }]>;
 def vmrglh_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGLShuffleMask(cast<ShuffleVectorSDNode>(N), 2, false);
 }]>;
 def vmrglw_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGLShuffleMask(cast<ShuffleVectorSDNode>(N), 4, false);
 }]>;
 def vmrghb_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGHShuffleMask(cast<ShuffleVectorSDNode>(N), 1, false);
 }]>;
 def vmrghh_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGHShuffleMask(cast<ShuffleVectorSDNode>(N), 2, false);
 }]>;
 def vmrghw_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                             (vector_shuffle node:$lhs, node:$rhs), [{
+                             (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGHShuffleMask(cast<ShuffleVectorSDNode>(N), 4, false);
 }]>;
 
 
 def vmrglb_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),
-                                   (vector_shuffle node:$lhs, node:$rhs), [{
+                               (vector_shuffle (v16i8 node:$lhs), node:$rhs), [{
   return PPC::isVMRGLShuffleMask(cast<ShuffleVectorSDNode>(N), 1, true);
 }]>;
 def vmrglh_unary_shuffle : PatFrag<(ops node:$lhs, node:$rhs),




From sabre at nondot.org  Mon Mar  8 12:51:21 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:51:21 -0000
Subject: [llvm-commits] [llvm] r97955 - in /llvm/trunk/lib/Target/ARM:
 ARMInstrInfo.td ARMInstrNEON.td ARMInstrVFP.td
Message-ID: <20100308185121.A56712A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:51:21 2010
New Revision: 97955

URL: http://llvm.org/viewvc/llvm-project?rev=97955&view=rev
Log:
fix a bunch of partially ambiguous patterns on ARM.  As an
example, this:

(set DPR:$dst, (fsub (fneg (fmul DPR:$a, DPR:$b)), DPR:$dstin))

is ambiguous because DPR contains both f64 and v2f32.  tblgen
currently accidentally picks f64 because it's first in the 
regclass.


Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
    llvm/trunk/lib/Target/ARM/ARMInstrVFP.td

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=97955&r1=97954&r2=97955&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Mon Mar  8 12:51:21 2010
@@ -21,7 +21,7 @@
 
 def SDT_ARMSaveCallPC : SDTypeProfile<0, 1, []>;
 
-def SDT_ARMcall    : SDTypeProfile<0, -1, [SDTCisInt<0>]>;
+def SDT_ARMcall    : SDTypeProfile<0, -1, [SDTCisPtrTy<0>]>;
 
 def SDT_ARMCMov    : SDTypeProfile<1, 3,
                                    [SDTCisSameAs<0, 1>, SDTCisSameAs<0, 2>,

Modified: llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrNEON.td?rev=97955&r1=97954&r2=97955&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrNEON.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrNEON.td Mon Mar  8 12:51:21 2010
@@ -2707,21 +2707,21 @@
 }
 def : Pat<(vector_insert (v16i8 QPR:$src1), GPR:$src2, imm:$lane),
           (v16i8 (INSERT_SUBREG QPR:$src1, 
-                  (VSETLNi8 (v8i8 (EXTRACT_SUBREG QPR:$src1,
+                  (v8i8 (VSETLNi8 (v8i8 (EXTRACT_SUBREG QPR:$src1,
                                    (DSubReg_i8_reg imm:$lane))),
-                            GPR:$src2, (SubReg_i8_lane imm:$lane)),
+                            GPR:$src2, (SubReg_i8_lane imm:$lane))),
                   (DSubReg_i8_reg imm:$lane)))>;
 def : Pat<(vector_insert (v8i16 QPR:$src1), GPR:$src2, imm:$lane),
           (v8i16 (INSERT_SUBREG QPR:$src1, 
-                  (VSETLNi16 (v4i16 (EXTRACT_SUBREG QPR:$src1,
+                  (v4i16 (VSETLNi16 (v4i16 (EXTRACT_SUBREG QPR:$src1,
                                      (DSubReg_i16_reg imm:$lane))),
-                             GPR:$src2, (SubReg_i16_lane imm:$lane)),
+                             GPR:$src2, (SubReg_i16_lane imm:$lane))),
                   (DSubReg_i16_reg imm:$lane)))>;
 def : Pat<(insertelt (v4i32 QPR:$src1), GPR:$src2, imm:$lane),
           (v4i32 (INSERT_SUBREG QPR:$src1, 
-                  (VSETLNi32 (v2i32 (EXTRACT_SUBREG QPR:$src1,
+                  (v2i32 (VSETLNi32 (v2i32 (EXTRACT_SUBREG QPR:$src1,
                                      (DSubReg_i32_reg imm:$lane))),
-                             GPR:$src2, (SubReg_i32_lane imm:$lane)),
+                             GPR:$src2, (SubReg_i32_lane imm:$lane))),
                   (DSubReg_i32_reg imm:$lane)))>;
 
 def : Pat<(v2f32 (insertelt DPR:$src1, SPR:$src2, imm:$src3)),
@@ -3093,16 +3093,17 @@
 
 class N2VSPat<SDNode OpNode, ValueType ResTy, ValueType OpTy, NeonI Inst>
   : NEONFPPat<(ResTy (OpNode SPR:$a)),
-              (EXTRACT_SUBREG (Inst (INSERT_SUBREG (OpTy (IMPLICIT_DEF)),
-                                                   SPR:$a, arm_ssubreg_0)),
+              (EXTRACT_SUBREG (OpTy (Inst (INSERT_SUBREG (OpTy (IMPLICIT_DEF)),
+                                                       SPR:$a, arm_ssubreg_0))),
                               arm_ssubreg_0)>;
 
 class N3VSPat<SDNode OpNode, NeonI Inst>
   : NEONFPPat<(f32 (OpNode SPR:$a, SPR:$b)),
-              (EXTRACT_SUBREG (Inst (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
-                                                   SPR:$a, arm_ssubreg_0),
-                                    (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
-                                                   SPR:$b, arm_ssubreg_0)),
+              (EXTRACT_SUBREG (v2f32
+                                 (Inst (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
+                                                      SPR:$a, arm_ssubreg_0),
+                                       (INSERT_SUBREG (v2f32 (IMPLICIT_DEF)),
+                                                      SPR:$b, arm_ssubreg_0))),
                               arm_ssubreg_0)>;
 
 class N3VSMulOpPat<SDNode MulNode, SDNode OpNode, NeonI Inst>

Modified: llvm/trunk/lib/Target/ARM/ARMInstrVFP.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrVFP.td?rev=97955&r1=97954&r2=97955&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrVFP.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrVFP.td Mon Mar  8 12:51:21 2010
@@ -57,7 +57,7 @@
 let canFoldAsLoad = 1, isReMaterializable = 1 in {
 def VLDRD : ADI5<0b1101, 0b01, (outs DPR:$dst), (ins addrmode5:$addr),
                  IIC_fpLoad64, "vldr", ".64\t$dst, $addr",
-                 [(set DPR:$dst, (load addrmode5:$addr))]>;
+                 [(set DPR:$dst, (f64 (load addrmode5:$addr)))]>;
 
 def VLDRS : ASI5<0b1101, 0b01, (outs SPR:$dst), (ins addrmode5:$addr),
                  IIC_fpLoad32, "vldr", ".32\t$dst, $addr",
@@ -66,7 +66,7 @@
 
 def VSTRD  : ADI5<0b1101, 0b00, (outs), (ins DPR:$src, addrmode5:$addr),
                  IIC_fpStore64, "vstr", ".64\t$src, $addr",
-                 [(store DPR:$src, addrmode5:$addr)]>;
+                 [(store (f64 DPR:$src), addrmode5:$addr)]>;
 
 def VSTRS  : ASI5<0b1101, 0b00, (outs), (ins SPR:$src, addrmode5:$addr),
                  IIC_fpStore32, "vstr", ".32\t$src, $addr",
@@ -116,7 +116,7 @@
 
 def VADDD  : ADbI<0b11100, 0b11, 0, 0, (outs DPR:$dst), (ins DPR:$a, DPR:$b),
                  IIC_fpALU64, "vadd", ".f64\t$dst, $a, $b",
-                 [(set DPR:$dst, (fadd DPR:$a, DPR:$b))]>;
+                 [(set DPR:$dst, (fadd DPR:$a, (f64 DPR:$b)))]>;
 
 def VADDS  : ASbIn<0b11100, 0b11, 0, 0, (outs SPR:$dst), (ins SPR:$a, SPR:$b),
                   IIC_fpALU32, "vadd", ".f32\t$dst, $a, $b",
@@ -126,7 +126,7 @@
 let Defs = [FPSCR] in {
 def VCMPED : ADuI<0b11101, 0b11, 0b0100, 0b11, 0, (outs), (ins DPR:$a, DPR:$b),
                  IIC_fpCMP64, "vcmpe", ".f64\t$a, $b",
-                 [(arm_cmpfp DPR:$a, DPR:$b)]>;
+                 [(arm_cmpfp DPR:$a, (f64 DPR:$b))]>;
 
 def VCMPD  : ADuI<0b11101, 0b11, 0b0100, 0b01, 0, (outs), (ins DPR:$a, DPR:$b),
                  IIC_fpCMP64, "vcmp", ".f64\t$a, $b",
@@ -143,7 +143,7 @@
 
 def VDIVD  : ADbI<0b11101, 0b00, 0, 0, (outs DPR:$dst), (ins DPR:$a, DPR:$b),
                  IIC_fpDIV64, "vdiv", ".f64\t$dst, $a, $b",
-                 [(set DPR:$dst, (fdiv DPR:$a, DPR:$b))]>;
+                 [(set DPR:$dst, (fdiv DPR:$a, (f64 DPR:$b)))]>;
 
 def VDIVS  : ASbI<0b11101, 0b00, 0, 0, (outs SPR:$dst), (ins SPR:$a, SPR:$b),
                  IIC_fpDIV32, "vdiv", ".f32\t$dst, $a, $b",
@@ -151,7 +151,7 @@
 
 def VMULD  : ADbI<0b11100, 0b10, 0, 0, (outs DPR:$dst), (ins DPR:$a, DPR:$b),
                  IIC_fpMUL64, "vmul", ".f64\t$dst, $a, $b",
-                 [(set DPR:$dst, (fmul DPR:$a, DPR:$b))]>;
+                 [(set DPR:$dst, (fmul DPR:$a, (f64 DPR:$b)))]>;
 
 def VMULS  : ASbIn<0b11100, 0b10, 0, 0, (outs SPR:$dst), (ins SPR:$a, SPR:$b),
                   IIC_fpMUL32, "vmul", ".f32\t$dst, $a, $b",
@@ -159,14 +159,14 @@
 
 def VNMULD  : ADbI<0b11100, 0b10, 1, 0, (outs DPR:$dst), (ins DPR:$a, DPR:$b),
                   IIC_fpMUL64, "vnmul", ".f64\t$dst, $a, $b",
-                  [(set DPR:$dst, (fneg (fmul DPR:$a, DPR:$b)))]>;
+                  [(set DPR:$dst, (fneg (fmul DPR:$a, (f64 DPR:$b))))]>;
 
 def VNMULS  : ASbI<0b11100, 0b10, 1, 0, (outs SPR:$dst), (ins SPR:$a, SPR:$b),
                   IIC_fpMUL32, "vnmul", ".f32\t$dst, $a, $b",
                   [(set SPR:$dst, (fneg (fmul SPR:$a, SPR:$b)))]>;
 
 // Match reassociated forms only if not sign dependent rounding.
-def : Pat<(fmul (fneg DPR:$a), DPR:$b),
+def : Pat<(fmul (fneg DPR:$a), (f64 DPR:$b)),
           (VNMULD DPR:$a, DPR:$b)>, Requires<[NoHonorSignDependentRounding]>;
 def : Pat<(fmul (fneg SPR:$a), SPR:$b),
           (VNMULS SPR:$a, SPR:$b)>, Requires<[NoHonorSignDependentRounding]>;
@@ -174,7 +174,7 @@
 
 def VSUBD  : ADbI<0b11100, 0b11, 1, 0, (outs DPR:$dst), (ins DPR:$a, DPR:$b),
                  IIC_fpALU64, "vsub", ".f64\t$dst, $a, $b",
-                 [(set DPR:$dst, (fsub DPR:$a, DPR:$b))]>;
+                 [(set DPR:$dst, (fsub DPR:$a, (f64 DPR:$b)))]>;
 
 def VSUBS  : ASbIn<0b11100, 0b11, 1, 0, (outs SPR:$dst), (ins SPR:$a, SPR:$b),
                   IIC_fpALU32, "vsub", ".f32\t$dst, $a, $b",
@@ -186,7 +186,7 @@
 
 def VABSD  : ADuI<0b11101, 0b11, 0b0000, 0b11, 0, (outs DPR:$dst), (ins DPR:$a),
                  IIC_fpUNA64, "vabs", ".f64\t$dst, $a",
-                 [(set DPR:$dst, (fabs DPR:$a))]>;
+                 [(set DPR:$dst, (fabs (f64 DPR:$a)))]>;
 
 def VABSS  : ASuIn<0b11101, 0b11, 0b0000, 0b11, 0,(outs SPR:$dst), (ins SPR:$a),
                   IIC_fpUNA32, "vabs", ".f32\t$dst, $a",
@@ -195,7 +195,7 @@
 let Defs = [FPSCR] in {
 def VCMPEZD : ADuI<0b11101, 0b11, 0b0101, 0b11, 0, (outs), (ins DPR:$a),
                   IIC_fpCMP64, "vcmpe", ".f64\t$a, #0",
-                  [(arm_cmpfp0 DPR:$a)]>;
+                  [(arm_cmpfp0 (f64 DPR:$a))]>;
 
 def VCMPZD  : ADuI<0b11101, 0b11, 0b0101, 0b01, 0, (outs), (ins DPR:$a),
                   IIC_fpCMP64, "vcmp", ".f64\t$a, #0",
@@ -253,7 +253,7 @@
 
 def VNEGD  : ADuI<0b11101, 0b11, 0b0001, 0b01, 0, (outs DPR:$dst), (ins DPR:$a),
                  IIC_fpUNA64, "vneg", ".f64\t$dst, $a",
-                 [(set DPR:$dst, (fneg DPR:$a))]>;
+                 [(set DPR:$dst, (fneg (f64 DPR:$a)))]>;
 
 def VNEGS  : ASuIn<0b11101, 0b11, 0b0001, 0b01, 0,(outs SPR:$dst), (ins SPR:$a),
                   IIC_fpUNA32, "vneg", ".f32\t$dst, $a",
@@ -261,7 +261,7 @@
 
 def VSQRTD : ADuI<0b11101, 0b11, 0b0001, 0b11, 0, (outs DPR:$dst), (ins DPR:$a),
                  IIC_fpSQRT64, "vsqrt", ".f64\t$dst, $a",
-                 [(set DPR:$dst, (fsqrt DPR:$a))]>;
+                 [(set DPR:$dst, (fsqrt (f64 DPR:$a)))]>;
 
 def VSQRTS : ASuI<0b11101, 0b11, 0b0001, 0b11, 0, (outs SPR:$dst), (ins SPR:$a),
                  IIC_fpSQRT32, "vsqrt", ".f32\t$dst, $a",
@@ -325,7 +325,7 @@
 def VSITOD : AVConv1I<0b11101, 0b11, 0b1000, 0b1011,
                  (outs DPR:$dst), (ins SPR:$a),
                  IIC_fpCVTID, "vcvt", ".f64.s32\t$dst, $a",
-                 [(set DPR:$dst, (arm_sitof SPR:$a))]> {
+                 [(set DPR:$dst, (f64 (arm_sitof SPR:$a)))]> {
   let Inst{7} = 1; // s32
 }
 
@@ -339,7 +339,7 @@
 def VUITOD : AVConv1I<0b11101, 0b11, 0b1000, 0b1011,
                  (outs DPR:$dst), (ins SPR:$a),
                  IIC_fpCVTID, "vcvt", ".f64.u32\t$dst, $a",
-                 [(set DPR:$dst, (arm_uitof SPR:$a))]> {
+                 [(set DPR:$dst, (f64 (arm_uitof SPR:$a)))]> {
   let Inst{7} = 0; // u32
 }
 
@@ -356,7 +356,7 @@
 def VTOSIZD : AVConv1I<0b11101, 0b11, 0b1101, 0b1011,
                        (outs SPR:$dst), (ins DPR:$a),
                  IIC_fpCVTDI, "vcvt", ".s32.f64\t$dst, $a",
-                 [(set SPR:$dst, (arm_ftosi DPR:$a))]> {
+                 [(set SPR:$dst, (arm_ftosi (f64 DPR:$a)))]> {
   let Inst{7} = 1; // Z bit
 }
 
@@ -370,7 +370,7 @@
 def VTOUIZD : AVConv1I<0b11101, 0b11, 0b1100, 0b1011,
                        (outs SPR:$dst), (ins DPR:$a),
                  IIC_fpCVTDI, "vcvt", ".u32.f64\t$dst, $a",
-                 [(set SPR:$dst, (arm_ftoui DPR:$a))]> {
+                 [(set SPR:$dst, (arm_ftoui (f64 DPR:$a)))]> {
   let Inst{7} = 1; // Z bit
 }
 
@@ -514,7 +514,8 @@
 def VMLAD : ADbI<0b11100, 0b00, 0, 0,
                 (outs DPR:$dst), (ins DPR:$dstin, DPR:$a, DPR:$b),
                 IIC_fpMAC64, "vmla", ".f64\t$dst, $a, $b",
-                [(set DPR:$dst, (fadd (fmul DPR:$a, DPR:$b), DPR:$dstin))]>,
+                [(set DPR:$dst, (fadd (fmul DPR:$a, DPR:$b),
+                                      (f64 DPR:$dstin)))]>,
                 RegConstraint<"$dstin = $dst">;
 
 def VMLAS : ASbIn<0b11100, 0b00, 0, 0,
@@ -526,7 +527,8 @@
 def VNMLSD : ADbI<0b11100, 0b01, 0, 0,
                 (outs DPR:$dst), (ins DPR:$dstin, DPR:$a, DPR:$b),
                 IIC_fpMAC64, "vnmls", ".f64\t$dst, $a, $b",
-                [(set DPR:$dst, (fsub (fmul DPR:$a, DPR:$b), DPR:$dstin))]>,
+                [(set DPR:$dst, (fsub (fmul DPR:$a, DPR:$b),
+                                (f64 DPR:$dstin)))]>,
                 RegConstraint<"$dstin = $dst">;
 
 def VNMLSS : ASbI<0b11100, 0b01, 0, 0,
@@ -538,7 +540,8 @@
 def VMLSD : ADbI<0b11100, 0b00, 1, 0,
                  (outs DPR:$dst), (ins DPR:$dstin, DPR:$a, DPR:$b),
                  IIC_fpMAC64, "vmls", ".f64\t$dst, $a, $b",
-             [(set DPR:$dst, (fadd (fneg (fmul DPR:$a, DPR:$b)), DPR:$dstin))]>,
+             [(set DPR:$dst, (fadd (fneg (fmul DPR:$a, DPR:$b)),
+                             (f64 DPR:$dstin)))]>,
                 RegConstraint<"$dstin = $dst">;
 
 def VMLSS : ASbIn<0b11100, 0b00, 1, 0,
@@ -547,7 +550,7 @@
              [(set SPR:$dst, (fadd (fneg (fmul SPR:$a, SPR:$b)), SPR:$dstin))]>,
                 RegConstraint<"$dstin = $dst">;
 
-def : Pat<(fsub DPR:$dstin, (fmul DPR:$a, DPR:$b)),
+def : Pat<(fsub DPR:$dstin, (fmul DPR:$a, (f64 DPR:$b))),
           (VMLSD DPR:$dstin, DPR:$a, DPR:$b)>, Requires<[DontUseNEONForFP]>;
 def : Pat<(fsub SPR:$dstin, (fmul SPR:$a, SPR:$b)),
           (VMLSS SPR:$dstin, SPR:$a, SPR:$b)>, Requires<[DontUseNEONForFP]>;
@@ -555,7 +558,8 @@
 def VNMLAD : ADbI<0b11100, 0b01, 1, 0,
                  (outs DPR:$dst), (ins DPR:$dstin, DPR:$a, DPR:$b),
                  IIC_fpMAC64, "vnmla", ".f64\t$dst, $a, $b",
-             [(set DPR:$dst, (fsub (fneg (fmul DPR:$a, DPR:$b)), DPR:$dstin))]>,
+             [(set DPR:$dst, (fsub (fneg (fmul DPR:$a, DPR:$b)),
+                             (f64 DPR:$dstin)))]>,
                 RegConstraint<"$dstin = $dst">;
 
 def VNMLAS : ASbI<0b11100, 0b01, 1, 0,




From sabre at nondot.org  Mon Mar  8 12:52:55 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:52:55 -0000
Subject: [llvm-commits] [llvm] r97956 - in /llvm/trunk/lib/Target/SystemZ:
 SystemZInstrInfo.td SystemZOperands.td
Message-ID: <20100308185255.F406A2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:52:55 2010
New Revision: 97956

URL: http://llvm.org/viewvc/llvm-project?rev=97956&view=rev
Log:
fix a type compatibility bug.  imm is i32 in the input 
pattern, not i64.

Modified:
    llvm/trunk/lib/Target/SystemZ/SystemZInstrInfo.td
    llvm/trunk/lib/Target/SystemZ/SystemZOperands.td

Modified: llvm/trunk/lib/Target/SystemZ/SystemZInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZInstrInfo.td?rev=97956&r1=97955&r2=97956&view=diff
==============================================================================
--- llvm/trunk/lib/Target/SystemZ/SystemZInstrInfo.td (original)
+++ llvm/trunk/lib/Target/SystemZ/SystemZInstrInfo.td Mon Mar  8 12:52:55 2010
@@ -1112,7 +1112,8 @@
 
 // Arbitrary immediate support.
 def : Pat<(i32 imm:$src),
-          (EXTRACT_SUBREG (MOV64ri32 (i64 imm:$src)), subreg_32bit)>;
+          (EXTRACT_SUBREG (MOV64ri32 (GetI64FromI32 (i32 imm:$src))),
+             subreg_32bit)>;
 
 // Implement in terms of LLIHF/OILF.
 def : Pat<(i64 imm:$imm),

Modified: llvm/trunk/lib/Target/SystemZ/SystemZOperands.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZOperands.td?rev=97956&r1=97955&r2=97956&view=diff
==============================================================================
--- llvm/trunk/lib/Target/SystemZ/SystemZOperands.td (original)
+++ llvm/trunk/lib/Target/SystemZ/SystemZOperands.td Mon Mar  8 12:52:55 2010
@@ -67,6 +67,10 @@
   return getI32Imm(N->getZExtValue() >> 32);
 }]>;
 
+def GetI64FromI32 : SDNodeXForm<imm, [{
+  return CurDAG->getTargetConstant(N->getSExtValue(), MVT::i64);
+}]>;
+
 def i32ll16 : PatLeaf<(i32 imm), [{
   // i32ll16 predicate - true if the 32-bit immediate has only rightmost 16
   // bits set.




From sabre at nondot.org  Mon Mar  8 12:55:16 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:55:16 -0000
Subject: [llvm-commits] [llvm] r97957 - in /llvm/trunk/lib/Target/X86:
 X86Instr64bit.td X86InstrInfo.td
Message-ID: <20100308185516.1CB6D2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:55:15 2010
New Revision: 97957

URL: http://llvm.org/viewvc/llvm-project?rev=97957&view=rev
Log:
Correct immediate sizes.

Modified:
    llvm/trunk/lib/Target/X86/X86Instr64bit.td
    llvm/trunk/lib/Target/X86/X86InstrInfo.td

Modified: llvm/trunk/lib/Target/X86/X86Instr64bit.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Instr64bit.td?rev=97957&r1=97956&r2=97957&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86Instr64bit.td (original)
+++ llvm/trunk/lib/Target/X86/X86Instr64bit.td Mon Mar  8 12:55:15 2010
@@ -2473,7 +2473,7 @@
 defm PINSRQ      : SS41I_insert64<0x22, "pinsrq">;
 
 // -disable-16bit support.
-def : Pat<(truncstorei16 (i64 imm:$src), addr:$dst),
+def : Pat<(truncstorei16 (i16 imm:$src), addr:$dst),
           (MOV16mi addr:$dst, imm:$src)>;
 def : Pat<(truncstorei16 GR64:$src, addr:$dst),
           (MOV16mr addr:$dst, (EXTRACT_SUBREG GR64:$src, x86_subreg_16bit))>;

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=97957&r1=97956&r2=97957&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Mar  8 12:55:15 2010
@@ -5150,7 +5150,7 @@
           (AND32mi8 addr:$dst, i32immSExt8:$src2)>;
 
 // -disable-16bit support.
-def : Pat<(truncstorei16 (i32 imm:$src), addr:$dst),
+def : Pat<(truncstorei16 (i16 imm:$src), addr:$dst),
           (MOV16mi addr:$dst, imm:$src)>;
 def : Pat<(truncstorei16 GR32:$src, addr:$dst),
           (MOV16mr addr:$dst, (EXTRACT_SUBREG GR32:$src, x86_subreg_16bit))>;




From sabre at nondot.org  Mon Mar  8 12:57:56 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:57:56 -0000
Subject: [llvm-commits] [llvm] r97958 - in /llvm/trunk/lib/Target/X86:
	X86InstrMMX.td X86InstrSSE.td
Message-ID: <20100308185757.079192A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:57:56 2010
New Revision: 97958

URL: http://llvm.org/viewvc/llvm-project?rev=97958&view=rev
Log:
fix some more ambiguous patterns, remove another nontemporalstore
pattern which is broken (source and address swapped).

Modified:
    llvm/trunk/lib/Target/X86/X86InstrMMX.td
    llvm/trunk/lib/Target/X86/X86InstrSSE.td

Modified: llvm/trunk/lib/Target/X86/X86InstrMMX.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrMMX.td?rev=97958&r1=97957&r2=97958&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrMMX.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrMMX.td Mon Mar  8 12:57:56 2010
@@ -160,7 +160,8 @@
                            "movq2dq\t{$src, $dst|$dst, $src}",
           [(set VR128:$dst,
             (movl immAllZerosV,
-                  (v2i64 (scalar_to_vector (i64 (bitconvert VR64:$src))))))]>;
+                  (v2i64 (scalar_to_vector
+                              (i64 (bitconvert (v1i64 VR64:$src)))))))]>;
 
 let neverHasSideEffects = 1 in
 def MMX_MOVQ2FR64rr: SSDIi8<0xD6, MRMSrcReg, (outs FR64:$dst), (ins VR64:$src),
@@ -271,9 +272,9 @@
 
 // Shift up / down and insert zero's.
 def : Pat<(v1i64 (X86vshl     VR64:$src, (i8 imm:$amt))),
-          (v1i64 (MMX_PSLLQri VR64:$src, imm:$amt))>;
+          (MMX_PSLLQri VR64:$src, imm:$amt)>;
 def : Pat<(v1i64 (X86vshr     VR64:$src, (i8 imm:$amt))),
-          (v1i64 (MMX_PSRLQri VR64:$src, imm:$amt))>;
+          (MMX_PSRLQri VR64:$src, imm:$amt)>;
 
 // Comparison Instructions
 defm MMX_PCMPEQB : MMXI_binop_rm_int<0x74, "pcmpeqb", int_x86_mmx_pcmpeq_b>;
@@ -577,7 +578,7 @@
 // Clear top half.
 let AddedComplexity = 15 in {
   def : Pat<(v2i32 (X86vzmovl VR64:$src)),
-            (MMX_PUNPCKLDQrr VR64:$src, (MMX_V_SET0))>;
+            (MMX_PUNPCKLDQrr VR64:$src, (v2i32 (MMX_V_SET0)))>;
 }
 
 // Patterns to perform canonical versions of vector shuffling.

Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=97958&r1=97957&r2=97958&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Mon Mar  8 12:57:56 2010
@@ -385,7 +385,7 @@
 // Extract the low 32-bit value from one vector and insert it into another.
 let AddedComplexity = 15 in
 def : Pat<(v4f32 (movl VR128:$src1, VR128:$src2)),
-          (MOVSSrr VR128:$src1,
+          (MOVSSrr (v4f32 VR128:$src1),
                    (EXTRACT_SUBREG (v4f32 VR128:$src2), x86_subreg_ss))>;
 
 // Implicitly promote a 32-bit scalar to a vector.
@@ -827,7 +827,7 @@
 
 
 def : Pat<(movlhps VR128:$src1, (bc_v4i32 (v2i64 (X86vzload addr:$src2)))),
-          (MOVHPSrm VR128:$src1, addr:$src2)>;
+          (MOVHPSrm (v4i32 VR128:$src1), addr:$src2)>;
 
 def MOVLPSmr : PSI<0x13, MRMDestMem, (outs), (ins f64mem:$dst, VR128:$src),
                    "movlps\t{$src, $dst|$dst, $src}",
@@ -860,9 +860,9 @@
 
 let AddedComplexity = 20 in {
 def : Pat<(v4f32 (movddup VR128:$src, (undef))),
-          (MOVLHPSrr VR128:$src, VR128:$src)>;
+          (MOVLHPSrr (v4f32 VR128:$src), (v4f32 VR128:$src))>;
 def : Pat<(v2i64 (movddup VR128:$src, (undef))),
-          (MOVLHPSrr VR128:$src, VR128:$src)>;
+          (MOVLHPSrr (v2i64 VR128:$src), (v2i64 VR128:$src))>;
 }
 
 
@@ -1011,9 +1011,9 @@
                                             (memop addr:$src), imm:$cc))]>;
 }
 def : Pat<(v4i32 (X86cmpps (v4f32 VR128:$src1), VR128:$src2, imm:$cc)),
-          (CMPPSrri VR128:$src1, VR128:$src2, imm:$cc)>;
+          (CMPPSrri (v4f32 VR128:$src1), (v4f32 VR128:$src2), imm:$cc)>;
 def : Pat<(v4i32 (X86cmpps (v4f32 VR128:$src1), (memop addr:$src2), imm:$cc)),
-          (CMPPSrmi VR128:$src1, addr:$src2, imm:$cc)>;
+          (CMPPSrmi (v4f32 VR128:$src1), addr:$src2, imm:$cc)>;
 
 // Shuffle and unpack instructions
 let Constraints = "$src1 = $dst" in {
@@ -1147,7 +1147,7 @@
 // Extract the low 64-bit value from one vector and insert it into another.
 let AddedComplexity = 15 in
 def : Pat<(v2f64 (movl VR128:$src1, VR128:$src2)),
-          (MOVSDrr VR128:$src1,
+          (MOVSDrr (v2f64 VR128:$src1),
                    (EXTRACT_SUBREG (v2f64 VR128:$src2), x86_subreg_sd))>;
 
 // Implicitly promote a 64-bit scalar to a vector.
@@ -2391,9 +2391,6 @@
 def MOVNTDQmr : PDI<0xE7, MRMDestMem, (outs), (ins f128mem:$dst, VR128:$src),
                     "movntdq\t{$src, $dst|$dst, $src}",
                     [(alignednontemporalstore (v4f32 VR128:$src), addr:$dst)]>;
-
-def : Pat<(alignednontemporalstore (v4i32 VR128:$src), addr:$dst),
-          (MOVNTDQmr VR128:$src, addr:$dst)>;
 }
 
 // Flush cache




From sabre at nondot.org  Mon Mar  8 12:59:49 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 18:59:49 -0000
Subject: [llvm-commits] [llvm] r97959 - in /llvm/trunk/lib/Target/CellSPU:
 CellSDKIntrinsics.td SPUInstrInfo.td SPUMathInstr.td SPUNodes.td
Message-ID: <20100308185950.081DC2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 12:59:49 2010
New Revision: 97959

URL: http://llvm.org/viewvc/llvm-project?rev=97959&view=rev
Log:
disambiguate some types, add a fixme about some 
inconsistent intrinsics.

Modified:
    llvm/trunk/lib/Target/CellSPU/CellSDKIntrinsics.td
    llvm/trunk/lib/Target/CellSPU/SPUInstrInfo.td
    llvm/trunk/lib/Target/CellSPU/SPUMathInstr.td
    llvm/trunk/lib/Target/CellSPU/SPUNodes.td

Modified: llvm/trunk/lib/Target/CellSPU/CellSDKIntrinsics.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/CellSDKIntrinsics.td?rev=97959&r1=97958&r2=97959&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/CellSDKIntrinsics.td (original)
+++ llvm/trunk/lib/Target/CellSPU/CellSDKIntrinsics.td Mon Mar  8 12:59:49 2010
@@ -205,6 +205,7 @@
 // Shift/rotate intrinsics:
 //===----------------------------------------------------------------------===//
 
+/* FIXME: These have (currently unenforced) type conflicts. */
 def CellSDKshli:
   Pat<(int_spu_si_shli (v4i32 VECREG:$rA), uimm7:$val),
       (SHLIv4i32 VECREG:$rA, uimm7:$val)>;

Modified: llvm/trunk/lib/Target/CellSPU/SPUInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPUInstrInfo.td?rev=97959&r1=97958&r2=97959&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/SPUInstrInfo.td (original)
+++ llvm/trunk/lib/Target/CellSPU/SPUInstrInfo.td Mon Mar  8 12:59:49 2010
@@ -2370,7 +2370,7 @@
 class ROTHVecInst<ValueType vectype>:
     ROTHInst<(outs VECREG:$rT), (ins VECREG:$rA, VECREG:$rB),
              [(set (vectype VECREG:$rT),
-                   (SPUvec_rotl VECREG:$rA, VECREG:$rB))]>;
+                   (SPUvec_rotl VECREG:$rA, (v8i16 VECREG:$rB)))]>;
 
 class ROTHRegInst<RegisterClass rclass>:
     ROTHInst<(outs rclass:$rT), (ins rclass:$rA, rclass:$rB),

Modified: llvm/trunk/lib/Target/CellSPU/SPUMathInstr.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPUMathInstr.td?rev=97959&r1=97958&r2=97959&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/SPUMathInstr.td (original)
+++ llvm/trunk/lib/Target/CellSPU/SPUMathInstr.td Mon Mar  8 12:59:49 2010
@@ -45,9 +45,9 @@
 def MPYv4i32:
   Pat<(mul (v4i32 VECREG:$rA), (v4i32 VECREG:$rB)),
       (Av4i32
-        (Av4i32 (MPYHv4i32 VECREG:$rA, VECREG:$rB),
-                (MPYHv4i32 VECREG:$rB, VECREG:$rA)),
-        (MPYUv4i32 VECREG:$rA, VECREG:$rB))>;
+        (v4i32 (Av4i32 (v4i32 (MPYHv4i32 VECREG:$rA, VECREG:$rB)),
+                       (v4i32 (MPYHv4i32 VECREG:$rB, VECREG:$rA)))),
+        (v4i32 (MPYUv4i32 VECREG:$rA, VECREG:$rB)))>;
 
 def MPYi32:
   Pat<(mul R32C:$rA, R32C:$rB),

Modified: llvm/trunk/lib/Target/CellSPU/SPUNodes.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPUNodes.td?rev=97959&r1=97958&r2=97959&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/SPUNodes.td (original)
+++ llvm/trunk/lib/Target/CellSPU/SPUNodes.td Mon Mar  8 12:59:49 2010
@@ -26,7 +26,7 @@
 // Operand constraints:
 //===----------------------------------------------------------------------===//
 
-def SDT_SPUCall   : SDTypeProfile<0, -1, [SDTCisInt<0>]>;
+def SDT_SPUCall   : SDTypeProfile<0, -1, [SDTCisPtrTy<0>]>;
 def SPUcall       : SDNode<"SPUISD::CALL", SDT_SPUCall,
                            [SDNPHasChain, SDNPOptInFlag, SDNPOutFlag]>;
 




From dpatel at apple.com  Mon Mar  8 13:20:19 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 19:20:19 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r97962 -
	/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
Message-ID: <20100308192019.140092A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 13:20:18 2010
New Revision: 97962

URL: http://llvm.org/viewvc/llvm-project?rev=97962&view=rev
Log:
Revert r97948.

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=97962&r1=97961&r2=97962&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Mon Mar  8 13:20:18 2010
@@ -1147,7 +1147,7 @@
     MainTy = getOrCreateType(TYPE_MAIN_VARIANT(type));
 
   DIType Ty = createVariantType(type, MainTy);
-  if (!Ty.isValid())
+  if (!Ty.isNull())
     return Ty;
 
   // Work out details of type.




From dpatel at apple.com  Mon Mar  8 13:20:38 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 19:20:38 -0000
Subject: [llvm-commits] [llvm] r97963 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/AsmPrinter/DwarfDebug.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.h lib/CodeGen/MachineInstr.cpp
 lib/Target/PIC16/PIC16DebugInfo.cpp lib/Transforms/Utils/CloneFunction.cpp
Message-ID: <20100308192038.D65212A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 13:20:38 2010
New Revision: 97963

URL: http://llvm.org/viewvc/llvm-project?rev=97963&view=rev
Log:
Revert r97947.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/MachineInstr.cpp
    llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
    llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 13:20:38 2010
@@ -67,7 +67,7 @@
     explicit DIDescriptor() : DbgNode(0) {}
     explicit DIDescriptor(MDNode *N) : DbgNode(N) {}
 
-    bool Verify() const { return DbgNode != 0; }
+    bool isNull() const { return DbgNode == 0; }
 
     MDNode *getNode() const { return DbgNode; }
 
@@ -246,9 +246,7 @@
     bool isArtificial() const {
       return (getFlags() & FlagArtificial) != 0;
     }
-    bool isValid() const {
-      return DbgNode && (isBasicType() || isDerivedType() || isCompositeType());
-    }
+
     /// dump - print type.
     void dump() const;
   };
@@ -362,7 +360,7 @@
     /// DIType or as DICompositeType.
     StringRef getReturnTypeName() const {
       DICompositeType DCT(getFieldAs<DICompositeType>(8));
-      if (DCT.Verify()) {
+      if (!DCT.isNull()) {
         DIArray A = DCT.getTypeArray();
         DIType T(A.getElement(0).getNode());
         return T.getName();
@@ -496,7 +494,6 @@
     DILocation getOrigLocation() const { return getFieldAs<DILocation>(3); }
     StringRef getFilename() const    { return getScope().getFilename(); }
     StringRef getDirectory() const   { return getScope().getDirectory(); }
-    bool Verify() const;
   };
 
   /// DIFactory - This object assists with the construction of the various

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 13:20:38 2010
@@ -132,12 +132,13 @@
 /// isBasicType - Return true if the specified tag is legal for
 /// DIBasicType.
 bool DIDescriptor::isBasicType() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_base_type;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_base_type;
 }
 
 /// isDerivedType - Return true if the specified tag is legal for DIDerivedType.
 bool DIDescriptor::isDerivedType() const {
-  if (!DbgNode) return false;
+  assert(!isNull() && "Invalid descriptor!");
   switch (getTag()) {
   case dwarf::DW_TAG_typedef:
   case dwarf::DW_TAG_pointer_type:
@@ -157,7 +158,7 @@
 /// isCompositeType - Return true if the specified tag is legal for
 /// DICompositeType.
 bool DIDescriptor::isCompositeType() const {
-  if (!DbgNode) return false;
+  assert(!isNull() && "Invalid descriptor!");
   switch (getTag()) {
   case dwarf::DW_TAG_array_type:
   case dwarf::DW_TAG_structure_type:
@@ -174,7 +175,7 @@
 
 /// isVariable - Return true if the specified tag is legal for DIVariable.
 bool DIDescriptor::isVariable() const {
-  if (!DbgNode) return false;
+  assert(!isNull() && "Invalid descriptor!");
   switch (getTag()) {
   case dwarf::DW_TAG_auto_variable:
   case dwarf::DW_TAG_arg_variable:
@@ -193,13 +194,15 @@
 /// isSubprogram - Return true if the specified tag is legal for
 /// DISubprogram.
 bool DIDescriptor::isSubprogram() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_subprogram;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_subprogram;
 }
 
 /// isGlobalVariable - Return true if the specified tag is legal for
 /// DIGlobalVariable.
 bool DIDescriptor::isGlobalVariable() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_variable;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_variable;
 }
 
 /// isGlobal - Return true if the specified tag is legal for DIGlobal.
@@ -210,7 +213,7 @@
 /// isScope - Return true if the specified tag is one of the scope
 /// related tag.
 bool DIDescriptor::isScope() const {
-  if (!DbgNode) return false;
+  assert(!isNull() && "Invalid descriptor!");
   switch (getTag()) {
   case dwarf::DW_TAG_compile_unit:
   case dwarf::DW_TAG_lexical_block:
@@ -225,27 +228,32 @@
 
 /// isCompileUnit - Return true if the specified tag is DW_TAG_compile_unit.
 bool DIDescriptor::isCompileUnit() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_compile_unit;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_compile_unit;
 }
 
 /// isNameSpace - Return true if the specified tag is DW_TAG_namespace.
 bool DIDescriptor::isNameSpace() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_namespace;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_namespace;
 }
 
 /// isLexicalBlock - Return true if the specified tag is DW_TAG_lexical_block.
 bool DIDescriptor::isLexicalBlock() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_lexical_block;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_lexical_block;
 }
 
 /// isSubrange - Return true if the specified tag is DW_TAG_subrange_type.
 bool DIDescriptor::isSubrange() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_subrange_type;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_subrange_type;
 }
 
 /// isEnumerator - Return true if the specified tag is DW_TAG_enumerator.
 bool DIDescriptor::isEnumerator() const {
-  return DbgNode && getTag() == dwarf::DW_TAG_enumerator;
+  assert(!isNull() && "Invalid descriptor!");
+  return getTag() == dwarf::DW_TAG_enumerator;
 }
 
 //===----------------------------------------------------------------------===//
@@ -260,8 +268,7 @@
 }
 
 unsigned DIArray::getNumElements() const {
-  if (!DbgNode)
-    return 0;
+  assert(DbgNode && "Invalid DIArray");
   return DbgNode->getNumOperands();
 }
 
@@ -269,9 +276,11 @@
 /// this descriptor. After this completes, the current debug info value
 /// is erased.
 void DIDerivedType::replaceAllUsesWith(DIDescriptor &D) {
-  if (!DbgNode)
+  if (isNull())
     return;
 
+  assert(!D.isNull() && "Can not replace with null");
+
   // Since we use a TrackingVH for the node, its easy for clients to manufacture
   // legitimate situations where they want to replaceAllUsesWith() on something
   // which, due to uniquing, has merged with the source. We shield clients from
@@ -286,7 +295,7 @@
 
 /// Verify - Verify that a compile unit is well formed.
 bool DICompileUnit::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
   StringRef N = getFilename();
   if (N.empty())
@@ -297,36 +306,36 @@
 
 /// Verify - Verify that a type descriptor is well formed.
 bool DIType::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
-  if (!getContext().Verify())
+  if (getContext().isNull())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.Verify())
+  if (!CU.isNull() && !CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a composite type descriptor is well formed.
 bool DICompositeType::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
-  if (!getContext().Verify())
+  if (getContext().isNull())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.Verify())
+  if (!CU.isNull() && !CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a subprogram descriptor is well formed.
 bool DISubprogram::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
 
-  if (!getContext().Verify())
+  if (getContext().isNull())
     return false;
 
   DICompileUnit CU = getCompileUnit();
@@ -334,24 +343,24 @@
     return false;
 
   DICompositeType Ty = getType();
-  if (!Ty.Verify())
+  if (!Ty.isNull() && !Ty.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a global variable descriptor is well formed.
 bool DIGlobalVariable::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
 
   if (getDisplayName().empty())
     return false;
 
-  if (!getContext().Verify())
+  if (getContext().isNull())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.Verify())
+  if (!CU.isNull() && !CU.Verify())
     return false;
 
   DIType Ty = getType();
@@ -366,10 +375,10 @@
 
 /// Verify - Verify that a variable descriptor is well formed.
 bool DIVariable::Verify() const {
-  if (!DbgNode)
+  if (isNull())
     return false;
 
-  if (!getContext().Verify())
+  if (getContext().isNull())
     return false;
 
   DIType Ty = getType();
@@ -379,14 +388,6 @@
   return true;
 }
 
-/// Verify - Verify that a location descriptor is well formed.
-bool DILocation::Verify() const {
-  if (!DbgNode)
-    return false;
-  
-  return DbgNode->getNumOperands() == 4;
-}
-
 /// getOriginalTypeSize - If this type is derived from a base type then
 /// return base type size.
 uint64_t DIDerivedType::getOriginalTypeSize() const {
@@ -397,7 +398,7 @@
     DIType BaseType = getTypeDerivedFrom();
     // If this type is not derived from any type then take conservative 
     // approach.
-    if (!BaseType.isValid())
+    if (BaseType.isNull())
       return getSizeInBits();
     if (BaseType.isDerivedType())
       return DIDerivedType(BaseType.getNode()).getOriginalTypeSize();
@@ -467,7 +468,7 @@
 
 /// dump - Print type.
 void DIType::dump() const {
-  if (!DbgNode) return;
+  if (isNull()) return;
 
   StringRef Res = getName();
   if (!Res.empty())
@@ -520,6 +521,8 @@
 /// dump - Print composite type.
 void DICompositeType::dump() const {
   DIArray A = getTypeArray();
+  if (A.isNull())
+    return;
   dbgs() << " [" << A.getNumElements() << " elements]";
 }
 
@@ -1152,8 +1155,9 @@
 
 /// processLocation - Process DILocation.
 void DebugInfoFinder::processLocation(DILocation Loc) {
-  if (!Loc.Verify()) return;
-  DIDescriptor S(Loc.getScope().getNode());
+  if (Loc.isNull()) return;
+  DIScope S(Loc.getScope().getNode());
+  if (S.isNull()) return;
   if (S.isCompileUnit())
     addCompileUnit(DICompileUnit(S.getNode()));
   else if (S.isSubprogram())
@@ -1173,21 +1177,26 @@
     DICompositeType DCT(DT.getNode());
     processType(DCT.getTypeDerivedFrom());
     DIArray DA = DCT.getTypeArray();
-    for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
-      DIDescriptor D = DA.getElement(i);
-      if (D.isType())
-        processType(DIType(D.getNode()));
-      else if (D.isSubprogram())
-        processSubprogram(DISubprogram(D.getNode()));
-    }
+    if (!DA.isNull())
+      for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
+        DIDescriptor D = DA.getElement(i);
+        DIType TyE = DIType(D.getNode());
+        if (!TyE.isNull())
+          processType(TyE);
+        else
+          processSubprogram(DISubprogram(D.getNode()));
+      }
   } else if (DT.isDerivedType()) {
     DIDerivedType DDT(DT.getNode());
-    processType(DDT.getTypeDerivedFrom());
+    if (!DDT.isNull())
+      processType(DDT.getTypeDerivedFrom());
   }
 }
 
 /// processLexicalBlock
 void DebugInfoFinder::processLexicalBlock(DILexicalBlock LB) {
+  if (LB.isNull())
+    return;
   DIScope Context = LB.getContext();
   if (Context.isLexicalBlock())
     return processLexicalBlock(DILexicalBlock(Context.getNode()));
@@ -1197,6 +1206,8 @@
 
 /// processSubprogram - Process DISubprogram.
 void DebugInfoFinder::processSubprogram(DISubprogram SP) {
+  if (SP.isNull())
+    return;
   if (!addSubprogram(SP))
     return;
   addCompileUnit(SP.getCompileUnit());
@@ -1205,23 +1216,20 @@
 
 /// processDeclare - Process DbgDeclareInst.
 void DebugInfoFinder::processDeclare(DbgDeclareInst *DDI) {
-  MDNode *N = dyn_cast<MDNode>(DDI->getVariable());
-  if (!N) return;
-
-  DIDescriptor DV(N);
-  if (!DV.isVariable())
+  DIVariable DV(cast<MDNode>(DDI->getVariable()));
+  if (DV.isNull())
     return;
 
   if (!NodesSeen.insert(DV.getNode()))
     return;
 
-  addCompileUnit(DIVariable(N).getCompileUnit());
-  processType(DIVariable(N).getType());
+  addCompileUnit(DV.getCompileUnit());
+  processType(DV.getType());
 }
 
 /// addType - Add type into Tys.
 bool DebugInfoFinder::addType(DIType DT) {
-  if (!DT.isValid())
+  if (DT.isNull())
     return false;
 
   if (!NodesSeen.insert(DT.getNode()))
@@ -1233,7 +1241,7 @@
 
 /// addCompileUnit - Add compile unit into CUs.
 bool DebugInfoFinder::addCompileUnit(DICompileUnit CU) {
-  if (!CU.Verify())
+  if (CU.isNull())
     return false;
 
   if (!NodesSeen.insert(CU.getNode()))
@@ -1245,7 +1253,7 @@
 
 /// addGlobalVariable - Add global variable into GVs.
 bool DebugInfoFinder::addGlobalVariable(DIGlobalVariable DIG) {
-  if (!DIDescriptor(DIG.getNode()).isGlobalVariable())
+  if (DIG.isNull())
     return false;
 
   if (!NodesSeen.insert(DIG.getNode()))
@@ -1257,7 +1265,7 @@
 
 // addSubprogram - Add subprgoram into SPs.
 bool DebugInfoFinder::addSubprogram(DISubprogram SP) {
-  if (!DIDescriptor(SP.getNode()).isSubprogram())
+  if (SP.isNull())
     return false;
 
   if (!NodesSeen.insert(SP.getNode()))
@@ -1275,10 +1283,10 @@
     return 0;
 
   for (unsigned i = 0, e = NMD->getNumOperands(); i != e; ++i) {
-    DIDescriptor DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
-    if (!DIG.isGlobalVariable())
+    DIGlobalVariable DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
+    if (DIG.isNull())
       continue;
-    if (DIGlobalVariable(DIG.getNode()).getGlobal() == V)
+    if (DIG.getGlobal() == V)
       return DIG.getNode();
   }
   return 0;
@@ -1370,6 +1378,12 @@
 /// getDISubprogram - Find subprogram that is enclosing this scope.
 DISubprogram llvm::getDISubprogram(MDNode *Scope) {
   DIDescriptor D(Scope);
+  if (D.isNull())
+    return DISubprogram();
+  
+  if (D.isCompileUnit())
+    return DISubprogram();
+  
   if (D.isSubprogram())
     return DISubprogram(Scope);
   
@@ -1381,6 +1395,9 @@
 
 /// getDICompositeType - Find underlying composite type.
 DICompositeType llvm::getDICompositeType(DIType T) {
+  if (T.isNull())
+    return DICompositeType();
+  
   if (T.isCompositeType())
     return DICompositeType(T.getNode());
   

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 13:20:38 2010
@@ -335,7 +335,7 @@
     // Print source line info.
     DIScope Scope = DLT.getScope();
     // Omit the directory, because it's likely to be long and uninteresting.
-    if (Scope.Verify())
+    if (!Scope.isNull())
       CommentOS << Scope.getFilename();
     else
       CommentOS << "<unknown>";
@@ -1287,7 +1287,7 @@
   if (DL.isUnknown())
     return;
   DILocation CurDLT = MF->getDILocation(DL);
-  if (!CurDLT.getScope().Verify())
+  if (CurDLT.getScope().isNull())
     return;
 
   if (!BeforePrintingInsn) {

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 13:20:38 2010
@@ -413,7 +413,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIVariable *V) {
   // If there is no compile unit specified, don't add a line #.
-  if (!V->getCompileUnit().Verify())
+  if (V->getCompileUnit().isNull())
     return;
 
   unsigned Line = V->getLineNumber();
@@ -427,7 +427,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIGlobal *G) {
   // If there is no compile unit specified, don't add a line #.
-  if (!G->getCompileUnit().Verify())
+  if (G->getCompileUnit().isNull())
     return;
 
   unsigned Line = G->getLineNumber();
@@ -441,7 +441,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DISubprogram *SP) {
   // If there is no compile unit specified, don't add a line #.
-  if (!SP->getCompileUnit().Verify())
+  if (SP->getCompileUnit().isNull())
     return;
   // If the line number is 0, don't add it.
   if (SP->getLineNumber() == 0)
@@ -460,7 +460,7 @@
 void DwarfDebug::addSourceLine(DIE *Die, const DIType *Ty) {
   // If there is no compile unit specified, don't add a line #.
   DICompileUnit CU = Ty->getCompileUnit();
-  if (!CU.Verify())
+  if (CU.isNull())
     return;
 
   unsigned Line = Ty->getLineNumber();
@@ -474,7 +474,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DINameSpace *NS) {
   // If there is no compile unit specified, don't add a line #.
-  if (!NS->getCompileUnit().Verify())
+  if (NS->getCompileUnit().isNull())
     return;
 
   unsigned Line = NS->getLineNumber();
@@ -526,8 +526,12 @@
   }
 
   DICompositeType blockStruct = DICompositeType(subType.getNode());
+
   DIArray Elements = blockStruct.getTypeArray();
 
+  if (Elements.isNull())
+    return Ty;
+
   for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Elements.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -673,6 +677,7 @@
   DIDescriptor varField = DIDescriptor();
   DIDescriptor forwardingField = DIDescriptor();
 
+
   for (unsigned i = 0, N = Fields.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Fields.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -683,6 +688,10 @@
       varField = Element;
   }
 
+  assert(!varField.isNull() && "Can't find byref variable in Block struct");
+  assert(!forwardingField.isNull()
+         && "Can't find forwarding field in Block struct");
+
   // Get the offsets for the forwarding field and the variable field.
   unsigned int forwardingFieldOffset =
     DIDerivedType(forwardingField.getNode()).getOffsetInBits() >> 3;
@@ -772,7 +781,9 @@
 
 /// addToContextOwner - Add Die into the list of its context owner's children.
 void DwarfDebug::addToContextOwner(DIE *Die, DIDescriptor Context) {
-  if (Context.isType()) {
+  if (Context.isNull())
+    ModuleCU->addDie(Die);
+  else if (Context.isType()) {
     DIE *ContextDIE = getOrCreateTypeDIE(DIType(Context.getNode()));
     ContextDIE->addChild(Die);
   } else if (Context.isNameSpace()) {
@@ -809,7 +820,7 @@
 
 /// addType - Add a new type attribute to the specified entity.
 void DwarfDebug::addType(DIE *Entity, DIType Ty) {
-  if (!Ty.isValid())
+  if (Ty.isNull())
     return;
 
   // Check for pre-existence.
@@ -895,9 +906,9 @@
     // Add enumerators to enumeration type.
     for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
       DIE *ElemDie = NULL;
-      DIDescriptor Enum(Elements.getElement(i).getNode());
-      if (Enum.isEnumerator()) {
-        ElemDie = constructEnumTypeDIE(DIEnumerator(Enum.getNode()));
+      DIEnumerator Enum(Elements.getElement(i).getNode());
+      if (!Enum.isNull()) {
+        ElemDie = constructEnumTypeDIE(&Enum);
         Buffer.addChild(ElemDie);
       }
     }
@@ -928,17 +939,18 @@
     DIArray Elements = CTy.getTypeArray();
 
     // A forward struct declared type may not have elements available.
-    unsigned N = Elements.getNumElements();
-    if (N == 0)
+    if (Elements.isNull())
       break;
 
     // Add elements to structure type.
-    for (unsigned i = 0; i < N; ++i) {
+    for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
       DIDescriptor Element = Elements.getElement(i);
+      if (Element.isNull())
+        continue;
       DIE *ElemDie = NULL;
-      if (Element.isSubprogram())
+      if (Element.getTag() == dwarf::DW_TAG_subprogram)
         ElemDie = createSubprogramDIE(DISubprogram(Element.getNode()));
-      else if (Element.isVariable()) {
+      else if (Element.getTag() == dwarf::DW_TAG_auto_variable) {
         DIVariable DV(Element.getNode());
         ElemDie = new DIE(dwarf::DW_TAG_variable);
         addString(ElemDie, dwarf::DW_AT_name, dwarf::DW_FORM_string,
@@ -947,10 +959,8 @@
         addUInt(ElemDie, dwarf::DW_AT_declaration, dwarf::DW_FORM_flag, 1);
         addUInt(ElemDie, dwarf::DW_AT_external, dwarf::DW_FORM_flag, 1);
         addSourceLine(ElemDie, &DV);
-      } else if (Element.isDerivedType())
+      } else
         ElemDie = createMemberDIE(DIDerivedType(Element.getNode()));
-      else
-        continue;
       Buffer.addChild(ElemDie);
     }
 
@@ -963,7 +973,7 @@
               dwarf::DW_FORM_data1, RLang);
 
     DICompositeType ContainingType = CTy.getContainingType();
-    if (DIDescriptor(ContainingType.getNode()).isCompositeType())
+    if (!ContainingType.isNull())
       addDIEEntry(&Buffer, dwarf::DW_AT_containing_type, dwarf::DW_FORM_ref4, 
                   getOrCreateTypeDIE(DIType(ContainingType.getNode())));
     break;
@@ -1041,11 +1051,11 @@
 }
 
 /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator ETy) {
+DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator *ETy) {
   DIE *Enumerator = new DIE(dwarf::DW_TAG_enumerator);
-  StringRef Name = ETy.getName();
+  StringRef Name = ETy->getName();
   addString(Enumerator, dwarf::DW_AT_name, dwarf::DW_FORM_string, Name);
-  int64_t Value = ETy.getEnumValue();
+  int64_t Value = ETy->getEnumValue();
   addSInt(Enumerator, dwarf::DW_AT_const_value, dwarf::DW_FORM_sdata, Value);
   return Enumerator;
 }
@@ -1189,7 +1199,7 @@
   DIArray Args = SPTy.getTypeArray();
   unsigned SPTag = SPTy.getTag();
 
-  if (Args.getNumElements() == 0 || SPTag != dwarf::DW_TAG_subroutine_type)
+  if (Args.isNull() || SPTag != dwarf::DW_TAG_subroutine_type)
     addType(SPDie, SPTy);
   else
     addType(SPDie, DIType(Args.getElement(0).getNode()));
@@ -1272,9 +1282,11 @@
     Parent->addScope(NScope);
   } else if (DIDescriptor(N).isLexicalBlock()) {
     DILexicalBlock DB(N);
-    Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
-    NScope->setParent(Parent);
-    Parent->addScope(NScope);
+    if (!DB.getContext().isNull()) {
+      Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
+      NScope->setParent(Parent);
+      Parent->addScope(NScope);
+    }
   }
 
   NScope->setFirstInsn(MI);
@@ -1306,7 +1318,8 @@
   if (Scope.isLexicalBlock()) {
     DILexicalBlock DB(N);
     DIDescriptor ParentDesc = DB.getContext();
-    Parent = getOrCreateAbstractScope(ParentDesc.getNode());
+    if (!ParentDesc.isNull())
+      Parent = getOrCreateAbstractScope(ParentDesc.getNode());
   }
 
   AScope = new DbgScope(Parent, DIDescriptor(N), NULL);
@@ -1409,9 +1422,10 @@
   // Ignore empty scopes.
   if (StartID == EndID && StartID != 0)
     return NULL;
-  if (!Scope->getScopeNode())
-    return NULL;
+
   DIScope DS(Scope->getScopeNode());
+  if (DS.isNull())
+    return NULL;
   DIE *ScopeDIE = new DIE(dwarf::DW_TAG_inlined_subroutine);
 
   DISubprogram InlinedSP = getDISubprogram(DS.getNode());
@@ -1533,12 +1547,15 @@
     return;
 
   DIArray Args = SPTy.getTypeArray();
+  if (Args.isNull()) 
+    return;
+
   for (unsigned i = 0, e = Args.getNumElements(); i != e; ++i) {
     DIType ATy(Args.getElement(i).getNode());
-    if (!ATy.isValid())
+    if (ATy.isNull())
       continue;
     DICompositeType CATy = getDICompositeType(ATy);
-    if (DIDescriptor(CATy.getNode()).Verify() && !CATy.getName().empty()) {
+    if (!CATy.isNull() && !CATy.getName().empty()) {
       if (DIEEntry *Entry = ModuleCU->getDIEEntry(CATy.getNode()))
         ModuleCU->addGlobalType(CATy.getName(), Entry->getEntry());
     }
@@ -1547,24 +1564,26 @@
 
 /// constructScopeDIE - Construct a DIE for this scope.
 DIE *DwarfDebug::constructScopeDIE(DbgScope *Scope) {
-  if (!Scope || !Scope->getScopeNode())
-    return NULL;
- 
-  DIScope DS(Scope->getScopeNode());
-  DIE *ScopeDIE = NULL;
-  if (Scope->getInlinedAt())
-    ScopeDIE = constructInlinedScopeDIE(Scope);
-  else if (DS.isSubprogram()) {
-    if (Scope->isAbstractScope())
-      ScopeDIE = ModuleCU->getDIE(DS.getNode());
-    else
-      ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
-  }
-  else {
-    ScopeDIE = constructLexicalScopeDIE(Scope);
-    if (!ScopeDIE) return NULL;
-  }
-  
+ if (!Scope)
+  return NULL;
+ DIScope DS(Scope->getScopeNode());
+ if (DS.isNull())
+   return NULL;
+
+ DIE *ScopeDIE = NULL;
+ if (Scope->getInlinedAt())
+   ScopeDIE = constructInlinedScopeDIE(Scope);
+ else if (DS.isSubprogram()) {
+   if (Scope->isAbstractScope())
+     ScopeDIE = ModuleCU->getDIE(DS.getNode());
+   else
+     ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
+ }
+ else {
+   ScopeDIE = constructLexicalScopeDIE(Scope);
+   if (!ScopeDIE) return NULL;
+ }
+
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
@@ -2050,6 +2069,7 @@
       if (DL.isUnknown()) continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
+      if (DLTScope.isNull()) continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2069,6 +2089,7 @@
       if (DL.isUnknown())  continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
+      if (DLTScope.isNull())  continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2147,7 +2168,7 @@
     DILocation DLT = MF->getDILocation(FDL);
     unsigned LabelID = 0;
     DISubprogram SP = getDISubprogram(DLT.getScope().getNode());
-    if (SP.Verify())
+    if (!SP.isNull())
       LabelID = recordSourceLine(SP.getLineNumber(), 0, 
                                  DLT.getScope().getNode());
     else

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 13:20:38 2010
@@ -346,7 +346,7 @@
                              DICompositeType *CTy);
 
   /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-  DIE *constructEnumTypeDIE(DIEnumerator ETy);
+  DIE *constructEnumTypeDIE(DIEnumerator *ETy);
 
   /// createGlobalVariableDIE - Create new DIE using GV.
   DIE *createGlobalVariableDIE(const DIGlobalVariable &GV);

Modified: llvm/trunk/lib/CodeGen/MachineInstr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineInstr.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineInstr.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineInstr.cpp Mon Mar  8 13:20:38 2010
@@ -1219,7 +1219,7 @@
     DIScope Scope = DLT.getScope();
     OS << " dbg:";
     // Omit the directory, since it's usually long and uninteresting.
-    if (Scope.Verify())
+    if (!Scope.isNull())
       OS << Scope.getFilename();
     else
       OS << "<unknown>";

Modified: llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp Mon Mar  8 13:20:38 2010
@@ -333,7 +333,7 @@
   for (DebugInfoFinder::iterator I = DbgFinder.type_begin(),
          E = DbgFinder.type_end(); I != E; ++I) {
     DICompositeType CTy(*I);
-    if (!CTy.Verify())
+    if (CTy.isNull())
       continue;
     if (CTy.getTag() == dwarf::DW_TAG_union_type ||
         CTy.getTag() == dwarf::DW_TAG_structure_type ) {

Modified: llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp?rev=97963&r1=97962&r2=97963&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp Mon Mar  8 13:20:38 2010
@@ -336,14 +336,14 @@
 
 static MDNode *UpdateInlinedAtInfo(MDNode *InsnMD, MDNode *TheCallMD) {
   DILocation ILoc(InsnMD);
-  if (!ILoc.Verify()) return InsnMD;
+  if (ILoc.isNull()) return InsnMD;
 
   DILocation CallLoc(TheCallMD);
-  if (!CallLoc.Verify()) return InsnMD;
+  if (CallLoc.isNull()) return InsnMD;
 
   DILocation OrigLocation = ILoc.getOrigLocation();
   MDNode *NewLoc = TheCallMD;
-  if (OrigLocation.Verify())
+  if (!OrigLocation.isNull())
     NewLoc = UpdateInlinedAtInfo(OrigLocation.getNode(), TheCallMD);
 
   Value *MDVs[] = {




From idadesub at users.sourceforge.net  Mon Mar  8 13:32:18 2010
From: idadesub at users.sourceforge.net (Erick Tryzelaar)
Date: Mon, 08 Mar 2010 19:32:18 -0000
Subject: [llvm-commits] [llvm] r97965 - in /llvm/trunk/docs/tutorial:
 OCamlLangImpl3.html OCamlLangImpl4.html OCamlLangImpl5.html
 OCamlLangImpl6.html OCamlLangImpl7.html
Message-ID: <20100308193218.924FF2A6C12C@llvm.org>

Author: erickt
Date: Mon Mar  8 13:32:18 2010
New Revision: 97965

URL: http://llvm.org/viewvc/llvm-project?rev=97965&view=rev
Log:
Update the OCaml Kaleidoscope tutorial.

Modified:
    llvm/trunk/docs/tutorial/OCamlLangImpl3.html
    llvm/trunk/docs/tutorial/OCamlLangImpl4.html
    llvm/trunk/docs/tutorial/OCamlLangImpl5.html
    llvm/trunk/docs/tutorial/OCamlLangImpl6.html
    llvm/trunk/docs/tutorial/OCamlLangImpl7.html

Modified: llvm/trunk/docs/tutorial/OCamlLangImpl3.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/OCamlLangImpl3.html?rev=97965&r1=97964&r2=97965&view=diff
==============================================================================
--- llvm/trunk/docs/tutorial/OCamlLangImpl3.html (original)
+++ llvm/trunk/docs/tutorial/OCamlLangImpl3.html Mon Mar  8 13:32:18 2010
@@ -98,6 +98,7 @@
 let the_module = create_module (global_context ()) "my cool jit"
 let builder = builder (global_context ())
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 </pre>
 </div>
 
@@ -389,7 +390,7 @@
 <div class="doc_code">
 <pre>
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try
@@ -903,6 +904,7 @@
 let the_module = create_module context "my cool jit"
 let builder = builder context
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 
 let rec codegen_expr = function
   | Ast.Number n -&gt; const_float double_type n
@@ -974,7 +976,7 @@
       let the_function = codegen_proto proto in
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try

Modified: llvm/trunk/docs/tutorial/OCamlLangImpl4.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/OCamlLangImpl4.html?rev=97965&r1=97964&r2=97965&view=diff
==============================================================================
--- llvm/trunk/docs/tutorial/OCamlLangImpl4.html (original)
+++ llvm/trunk/docs/tutorial/OCamlLangImpl4.html Mon Mar  8 13:32:18 2010
@@ -186,9 +186,8 @@
 <div class="doc_code">
 <pre>
   (* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
@@ -213,18 +212,11 @@
 </pre>
 </div>
 
-<p>This code defines two values, an <tt>Llvm.llmoduleprovider</tt> and a
-<tt>Llvm.PassManager.t</tt>.  The former is basically a wrapper around our
-<tt>Llvm.llmodule</tt> that the <tt>Llvm.PassManager.t</tt> requires.  It
-provides certain flexibility that we're not going to take advantage of here,
-so I won't dive into any details about it.</p>
-
 <p>The meat of the matter here, is the definition of "<tt>the_fpm</tt>".  It
-requires a pointer to the <tt>the_module</tt> (through the
-<tt>the_module_provider</tt>) to construct itself.  Once it is set up, we use a
-series of "add" calls to add a bunch of LLVM passes.  The first pass is
-basically boilerplate, it adds a pass so that later optimizations know how the
-data structures in the program are laid out.  The
+requires a pointer to the <tt>the_module</tt> to construct itself.  Once it is
+set up, we use a series of "add" calls to add a bunch of LLVM passes.  The
+first pass is basically boilerplate, it adds a pass so that later optimizations
+know how the data structures in the program are laid out.  The
 "<tt>the_execution_engine</tt>" variable is related to the JIT, which we will
 get to in the next section.</p>
 
@@ -320,8 +312,7 @@
 let main () =
   ...
   <b>(* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in</b>
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in</b>
   ...
 </pre>
 </div>
@@ -351,7 +342,7 @@
               the_execution_engine in
 
             print_string "Evaluated to ";
-            print_float (GenericValue.as_float double_type result);
+            print_float (GenericValue.as_float Codegen.double_type result);
             print_newline ();
 </pre>
 </div>
@@ -796,6 +787,7 @@
 let the_module = create_module context "my cool jit"
 let builder = builder context
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 
 let rec codegen_expr = function
   | Ast.Number n -&gt; const_float double_type n
@@ -867,7 +859,7 @@
       let the_function = codegen_proto proto in
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try
@@ -932,7 +924,7 @@
               the_execution_engine in
 
             print_string "Evaluated to ";
-            print_float (GenericValue.as_float double_type result);
+            print_float (GenericValue.as_float Codegen.double_type result);
             print_newline ();
         with Stream.Error s | Codegen.Error s -&gt;
           (* Skip token for error recovery. *)
@@ -971,16 +963,15 @@
   let stream = Lexer.lex (Stream.of_channel stdin) in
 
   (* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
   TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
 
   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
-  add_instruction_combining the_fpm;
+  add_instruction_combination the_fpm;
 
   (* reassociate expressions. *)
   add_reassociation the_fpm;

Modified: llvm/trunk/docs/tutorial/OCamlLangImpl5.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/OCamlLangImpl5.html?rev=97965&r1=97964&r2=97965&view=diff
==============================================================================
--- llvm/trunk/docs/tutorial/OCamlLangImpl5.html (original)
+++ llvm/trunk/docs/tutorial/OCamlLangImpl5.html Mon Mar  8 13:32:18 2010
@@ -364,7 +364,7 @@
       let start_bb = insertion_block builder in
       let the_function = block_parent start_bb in
 
-      let then_bb = append_block "then" the_function in
+      let then_bb = append_block context "then" the_function in
       position_at_end then_bb builder;
 </pre>
 </div>
@@ -417,7 +417,7 @@
 <div class="doc_code">
 <pre>
       (* Emit 'else' value. *)
-      let else_bb = append_block "else" the_function in
+      let else_bb = append_block context "else" the_function in
       position_at_end else_bb builder;
       let else_val = codegen_expr else_ in
 
@@ -433,7 +433,7 @@
 <div class="doc_code">
 <pre>
       (* Emit merge block. *)
-      let merge_bb = append_block "ifcont" the_function in
+      let merge_bb = append_block context "ifcont" the_function in
       position_at_end merge_bb builder;
       let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
       let phi = build_phi incoming "iftmp" builder in
@@ -704,7 +704,7 @@
        * block. *)
       let preheader_bb = insertion_block builder in
       let the_function = block_parent preheader_bb in
-      let loop_bb = append_block "loop" the_function in
+      let loop_bb = append_block context "loop" the_function in
 
       (* Insert an explicit fall through from the current block to the
        * loop_bb. *)
@@ -804,7 +804,7 @@
 <pre>
       (* Create the "after loop" block and insert it. *)
       let loop_end_bb = insertion_block builder in
-      let after_bb = append_block "afterloop" the_function in
+      let after_bb = append_block context "afterloop" the_function in
 
       (* Insert the conditional branch into the end of loop_end_bb. *)
       ignore (build_cond_br end_cond loop_bb after_bb builder);
@@ -1204,6 +1204,7 @@
 let the_module = create_module context "my cool jit"
 let builder = builder context
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 
 let rec codegen_expr = function
   | Ast.Number n -&gt; const_float double_type n
@@ -1250,7 +1251,7 @@
       let start_bb = insertion_block builder in
       let the_function = block_parent start_bb in
 
-      let then_bb = append_block "then" the_function in
+      let then_bb = append_block context "then" the_function in
 
       (* Emit 'then' value. *)
       position_at_end then_bb builder;
@@ -1262,7 +1263,7 @@
       let new_then_bb = insertion_block builder in
 
       (* Emit 'else' value. *)
-      let else_bb = append_block "else" the_function in
+      let else_bb = append_block context "else" the_function in
       position_at_end else_bb builder;
       let else_val = codegen_expr else_ in
 
@@ -1271,7 +1272,7 @@
       let new_else_bb = insertion_block builder in
 
       (* Emit merge block. *)
-      let merge_bb = append_block "ifcont" the_function in
+      let merge_bb = append_block context "ifcont" the_function in
       position_at_end merge_bb builder;
       let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
       let phi = build_phi incoming "iftmp" builder in
@@ -1297,7 +1298,7 @@
        * block. *)
       let preheader_bb = insertion_block builder in
       let the_function = block_parent preheader_bb in
-      let loop_bb = append_block "loop" the_function in
+      let loop_bb = append_block context "loop" the_function in
 
       (* Insert an explicit fall through from the current block to the
        * loop_bb. *)
@@ -1341,7 +1342,7 @@
 
       (* Create the "after loop" block and insert it. *)
       let loop_end_bb = insertion_block builder in
-      let after_bb = append_block "afterloop" the_function in
+      let after_bb = append_block context "afterloop" the_function in
 
       (* Insert the conditional branch into the end of loop_end_bb. *)
       ignore (build_cond_br end_cond loop_bb after_bb builder);
@@ -1397,7 +1398,7 @@
       let the_function = codegen_proto proto in
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try
@@ -1462,7 +1463,7 @@
               the_execution_engine in
 
             print_string "Evaluated to ";
-            print_float (GenericValue.as_float double_type result);
+            print_float (GenericValue.as_float Codegen.double_type result);
             print_newline ();
         with Stream.Error s | Codegen.Error s -&gt;
           (* Skip token for error recovery. *)
@@ -1501,16 +1502,15 @@
   let stream = Lexer.lex (Stream.of_channel stdin) in
 
   (* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
   TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
 
   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
-  add_instruction_combining the_fpm;
+  add_instruction_combination the_fpm;
 
   (* reassociate expressions. *)
   add_reassociation the_fpm;

Modified: llvm/trunk/docs/tutorial/OCamlLangImpl6.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/OCamlLangImpl6.html?rev=97965&r1=97964&r2=97965&view=diff
==============================================================================
--- llvm/trunk/docs/tutorial/OCamlLangImpl6.html (original)
+++ llvm/trunk/docs/tutorial/OCamlLangImpl6.html Mon Mar  8 13:32:18 2010
@@ -300,7 +300,7 @@
       end;</b>
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
       ...
 </pre>
@@ -1177,6 +1177,7 @@
 let the_module = create_module context "my cool jit"
 let builder = builder context
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 
 let rec codegen_expr = function
   | Ast.Number n -&gt; const_float double_type n
@@ -1241,7 +1242,7 @@
       let start_bb = insertion_block builder in
       let the_function = block_parent start_bb in
 
-      let then_bb = append_block "then" the_function in
+      let then_bb = append_block context "then" the_function in
 
       (* Emit 'then' value. *)
       position_at_end then_bb builder;
@@ -1253,7 +1254,7 @@
       let new_then_bb = insertion_block builder in
 
       (* Emit 'else' value. *)
-      let else_bb = append_block "else" the_function in
+      let else_bb = append_block context "else" the_function in
       position_at_end else_bb builder;
       let else_val = codegen_expr else_ in
 
@@ -1262,7 +1263,7 @@
       let new_else_bb = insertion_block builder in
 
       (* Emit merge block. *)
-      let merge_bb = append_block "ifcont" the_function in
+      let merge_bb = append_block context "ifcont" the_function in
       position_at_end merge_bb builder;
       let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
       let phi = build_phi incoming "iftmp" builder in
@@ -1288,7 +1289,7 @@
        * block. *)
       let preheader_bb = insertion_block builder in
       let the_function = block_parent preheader_bb in
-      let loop_bb = append_block "loop" the_function in
+      let loop_bb = append_block context "loop" the_function in
 
       (* Insert an explicit fall through from the current block to the
        * loop_bb. *)
@@ -1332,7 +1333,7 @@
 
       (* Create the "after loop" block and insert it. *)
       let loop_end_bb = insertion_block builder in
-      let after_bb = append_block "afterloop" the_function in
+      let after_bb = append_block context "afterloop" the_function in
 
       (* Insert the conditional branch into the end of loop_end_bb. *)
       ignore (build_cond_br end_cond loop_bb after_bb builder);
@@ -1396,7 +1397,7 @@
       end;
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try
@@ -1461,7 +1462,7 @@
               the_execution_engine in
 
             print_string "Evaluated to ";
-            print_float (GenericValue.as_float double_type result);
+            print_float (GenericValue.as_float Codegen.double_type result);
             print_newline ();
         with Stream.Error s | Codegen.Error s -&gt;
           (* Skip token for error recovery. *)
@@ -1500,16 +1501,15 @@
   let stream = Lexer.lex (Stream.of_channel stdin) in
 
   (* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
   TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
 
   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
-  add_instruction_combining the_fpm;
+  add_instruction_combination the_fpm;
 
   (* reassociate expressions. *)
   add_reassociation the_fpm;

Modified: llvm/trunk/docs/tutorial/OCamlLangImpl7.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/tutorial/OCamlLangImpl7.html?rev=97965&r1=97964&r2=97965&view=diff
==============================================================================
--- llvm/trunk/docs/tutorial/OCamlLangImpl7.html (original)
+++ llvm/trunk/docs/tutorial/OCamlLangImpl7.html Mon Mar  8 13:32:18 2010
@@ -543,7 +543,7 @@
 <pre>
 let main () =
   ...
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
@@ -1388,6 +1388,7 @@
 let the_module = create_module context "my cool jit"
 let builder = builder context
 let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
 
 (* Create an alloca instruction in the entry block of the function. This
  * is used for mutable variables etc. *)
@@ -1482,7 +1483,7 @@
       let start_bb = insertion_block builder in
       let the_function = block_parent start_bb in
 
-      let then_bb = append_block "then" the_function in
+      let then_bb = append_block context "then" the_function in
 
       (* Emit 'then' value. *)
       position_at_end then_bb builder;
@@ -1494,7 +1495,7 @@
       let new_then_bb = insertion_block builder in
 
       (* Emit 'else' value. *)
-      let else_bb = append_block "else" the_function in
+      let else_bb = append_block context "else" the_function in
       position_at_end else_bb builder;
       let else_val = codegen_expr else_ in
 
@@ -1503,7 +1504,7 @@
       let new_else_bb = insertion_block builder in
 
       (* Emit merge block. *)
-      let merge_bb = append_block "ifcont" the_function in
+      let merge_bb = append_block context "ifcont" the_function in
       position_at_end merge_bb builder;
       let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
       let phi = build_phi incoming "iftmp" builder in
@@ -1555,7 +1556,7 @@
 
       (* Make the new basic block for the loop header, inserting after current
        * block. *)
-      let loop_bb = append_block "loop" the_function in
+      let loop_bb = append_block context "loop" the_function in
 
       (* Insert an explicit fall through from the current block to the
        * loop_bb. *)
@@ -1599,7 +1600,7 @@
       let end_cond = build_fcmp Fcmp.One end_cond zero "loopcond" builder in
 
       (* Create the "after loop" block and insert it. *)
-      let after_bb = append_block "afterloop" the_function in
+      let after_bb = append_block context "afterloop" the_function in
 
       (* Insert the conditional branch into the end of loop_end_bb. *)
       ignore (build_cond_br end_cond loop_bb after_bb builder);
@@ -1723,7 +1724,7 @@
       end;
 
       (* Create a new basic block to start insertion into. *)
-      let bb = append_block "entry" the_function in
+      let bb = append_block context "entry" the_function in
       position_at_end bb builder;
 
       try
@@ -1791,7 +1792,7 @@
               the_execution_engine in
 
             print_string "Evaluated to ";
-            print_float (GenericValue.as_float double_type result);
+            print_float (GenericValue.as_float Codegen.double_type result);
             print_newline ();
         with Stream.Error s | Codegen.Error s -&gt;
           (* Skip token for error recovery. *)
@@ -1831,9 +1832,8 @@
   let stream = Lexer.lex (Stream.of_channel stdin) in
 
   (* Create the JIT. *)
-  let the_module_provider = ModuleProvider.create Codegen.the_module in
-  let the_execution_engine = ExecutionEngine.create the_module_provider in
-  let the_fpm = PassManager.create_function the_module_provider in
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
 
   (* Set up the optimizer pipeline.  Start with registering info about how the
    * target lays out data structures. *)
@@ -1843,7 +1843,7 @@
   add_memory_to_register_promotion the_fpm;
 
   (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
-  add_instruction_combining the_fpm;
+  add_instruction_combination the_fpm;
 
   (* reassociate expressions. *)
   add_reassociation the_fpm;




From idadesub at users.sourceforge.net  Mon Mar  8 13:32:28 2010
From: idadesub at users.sourceforge.net (Erick Tryzelaar)
Date: Mon, 08 Mar 2010 19:32:28 -0000
Subject: [llvm-commits] [llvm] r97966 - in /llvm/trunk: bindings/ocaml/
 examples/ examples/OCaml-Kaleidoscope/
 examples/OCaml-Kaleidoscope/Chapter2/ examples/OCaml-Kaleidoscope/Chapter3/
 examples/OCaml-Kaleidoscope/Chapter4/ examples/OCaml-Kaleidoscope/Chapter5/
 examples/OCaml-Kaleidoscope/Chapter6/ examples/OCaml-Kaleidoscope/Chapter7/
Message-ID: <20100308193228.9B6A12A6C12C@llvm.org>

Author: erickt
Date: Mon Mar  8 13:32:27 2010
New Revision: 97966

URL: http://llvm.org/viewvc/llvm-project?rev=97966&view=rev
Log:
Add OCaml tutorial to the examples.

Added:
    llvm/trunk/examples/OCaml-Kaleidoscope/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/codegen.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/myocamlbuild.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/bindings.c
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/codegen.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/myocamlbuild.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/bindings.c
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/codegen.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/myocamlbuild.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/bindings.c
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/codegen.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/myocamlbuild.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/Makefile
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/_tags
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/ast.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/bindings.c
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/codegen.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/lexer.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/myocamlbuild.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/parser.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/token.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toplevel.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toy.ml
    llvm/trunk/examples/OCaml-Kaleidoscope/Makefile
Modified:
    llvm/trunk/bindings/ocaml/Makefile.ocaml
    llvm/trunk/examples/Makefile

Modified: llvm/trunk/bindings/ocaml/Makefile.ocaml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/bindings/ocaml/Makefile.ocaml?rev=97966&r1=97965&r2=97966&view=diff
==============================================================================
--- llvm/trunk/bindings/ocaml/Makefile.ocaml (original)
+++ llvm/trunk/bindings/ocaml/Makefile.ocaml Mon Mar  8 13:32:27 2010
@@ -66,35 +66,64 @@
 Compile.CMX  := $(strip $(OCAMLOPT) -c $(OCAMLCFLAGS) $(OCAMLDEBUGFLAG) -o)
 Archive.CMXA := $(strip $(OCAMLOPT) -a $(OCAMLAFLAGS) $(OCAMLDEBUGFLAG) -o)
 
+ifdef OCAMLOPT
+Archive.EXE := $(strip $(OCAMLOPT) -cc $(CXX) $(OCAMLCFLAGS) $(UsedOcamLibs:%=%.cmxa) $(OCAMLDEBUGFLAG) -o)
+else
+Archive.EXE := $(strip $(OCAMLC) -cc $(CXX) $(OCAMLCFLAGS) $(OCAMLDEBUGFLAG:%=%.cma) -o)
+endif
+
 # Source files
 OcamlSources1 := $(sort $(wildcard $(PROJ_SRC_DIR)/*.ml))
-OcamlHeaders1 := $(OcamlSources1:.ml=.mli)
+OcamlHeaders1 := $(sort $(wildcard $(PROJ_SRC_DIR)/*.mli))
 
-OcamlSources := $(OcamlSources1:$(PROJ_SRC_DIR)/%=$(ObjDir)/%)
-OcamlHeaders := $(OcamlHeaders1:$(PROJ_SRC_DIR)/%=$(ObjDir)/%)
+OcamlSources2 := $(filter-out $(ExcludeSources),$(OcamlSources1))
+OcamlHeaders2 := $(filter-out $(ExcludeHeaders),$(OcamlHeaders1))
+
+OcamlSources := $(OcamlSources2:$(PROJ_SRC_DIR)/%=$(ObjDir)/%)
+OcamlHeaders := $(OcamlHeaders2:$(PROJ_SRC_DIR)/%=$(ObjDir)/%)
 
 # Intermediate files
-LibraryCMA   := $(ObjDir)/$(LIBRARYNAME).cma
-LibraryCMXA  := $(ObjDir)/$(LIBRARYNAME).cmxa
 ObjectsCMI   := $(OcamlSources:%.ml=%.cmi)
 ObjectsCMO   := $(OcamlSources:%.ml=%.cmo)
 ObjectsCMX   := $(OcamlSources:%.ml=%.cmx)
 
+ifdef LIBRARYNAME
+LibraryCMA   := $(ObjDir)/$(LIBRARYNAME).cma
+LibraryCMXA  := $(ObjDir)/$(LIBRARYNAME).cmxa
+endif
+
+ifdef TOOLNAME
+ToolEXE      := $(ObjDir)/$(TOOLNAME)$(EXEEXT)
+endif
+
 # Output files
 #   The .cmo files are the only intermediates; all others are to be installed.
-LibraryA   := $(OcamlDir)/lib$(LIBRARYNAME).a
-OutputCMA  := $(LibraryCMA:$(ObjDir)/%.cma=$(OcamlDir)/%.cma)
-OutputCMXA := $(LibraryCMXA:$(ObjDir)/%.cmxa=$(OcamlDir)/%.cmxa)
 OutputsCMI := $(ObjectsCMI:$(ObjDir)/%.cmi=$(OcamlDir)/%.cmi)
 OutputsCMX := $(ObjectsCMX:$(ObjDir)/%.cmx=$(OcamlDir)/%.cmx)
 OutputLibs := $(UsedLibNames:%=$(OcamlDir)/%)
 
+ifdef LIBRARYNAME
+LibraryA   := $(OcamlDir)/lib$(LIBRARYNAME).a
+OutputCMA  := $(LibraryCMA:$(ObjDir)/%.cma=$(OcamlDir)/%.cma)
+OutputCMXA := $(LibraryCMXA:$(ObjDir)/%.cmxa=$(OcamlDir)/%.cmxa)
+endif
+
+ifdef TOOLNAME
+ifdef EXAMPLE_TOOL
+OutputEXE := $(ExmplDir)/$(strip $(TOOLNAME))$(EXEEXT)
+else
+OutputEXE := $(ToolDir)/$(strip $(TOOLNAME))$(EXEEXT)
+endif
+endif
+
 # Installation targets
+DestLibs := $(UsedLibNames:%=$(PROJ_libocamldir)/%)
+
+ifdef LIBRARYNAME
 DestA    := $(PROJ_libocamldir)/lib$(LIBRARYNAME).a
 DestCMA  := $(PROJ_libocamldir)/$(LIBRARYNAME).cma
 DestCMXA := $(PROJ_libocamldir)/$(LIBRARYNAME).cmxa
-DestLibs := $(UsedLibNames:%=$(PROJ_libocamldir)/%)
-
+endif
 
 ##===- Dependencies -------------------------------------------------------===##
 # Copy the sources into the intermediate directory because older ocamlc doesn't
@@ -106,18 +135,27 @@
 $(ObjDir)/%.ml: $(PROJ_SRC_DIR)/%.ml $(ObjDir)/.dir
 	$(Verb) $(CP) -f $< $@
 
+$(ObjectsCMI): $(UsedOcamlInterfaces:%=$(OcamlDir)/%.cmi)
+
+ifdef LIBRARYNAME
 $(ObjDir)/$(LIBRARYNAME).ocamldep: $(OcamlSources) $(OcamlHeaders) \
                                    $(OcamlDir)/.dir $(ObjDir)/.dir
 	$(Verb) $(OCAMLDEP) $(OCAMLCFLAGS) $(OcamlSources) $(OcamlHeaders) > $@
 
-$(ObjectsCMI): $(UsedOcamlInterfaces:%=$(OcamlDir)/%.cmi)
-
 -include $(ObjDir)/$(LIBRARYNAME).ocamldep
+endif
+
+ifdef TOOLNAME
+$(ObjDir)/$(TOOLNAME).ocamldep: $(OcamlSources) $(OcamlHeaders) \
+                                $(OcamlDir)/.dir $(ObjDir)/.dir
+	$(Verb) $(OCAMLDEP) $(OCAMLCFLAGS) $(OcamlSources) $(OcamlHeaders) > $@
 
+-include $(ObjDir)/$(TOOLNAME).ocamldep
+endif
 
 ##===- Build static library from C sources --------------------------------===##
 
-ifneq ($(ObjectsO),)
+ifdef LibraryA
 all-local:: $(LibraryA)
 clean-local:: clean-a
 install-local:: install-a
@@ -160,7 +198,7 @@
 	$(Verb) ln -sf $< $@
 
 clean-deplibs:
-	$(Verb) rm -f $(OutputLibs)
+	$(Verb) $(RM) -f $(OutputLibs)
 
 install-deplibs:
 	$(Verb) $(MKDIR) $(PROJ_libocamldir)
@@ -169,11 +207,12 @@
 	done
 
 uninstall-deplibs:
-	$(Verb) rm -f $(DestLibs)
+	$(Verb) $(RM) -f $(DestLibs)
 
 
 ##===- Build ocaml interfaces (.mli's -> .cmi's) --------------------------===##
 
+ifneq ($(OcamlHeaders),)
 all-local:: build-cmis
 clean-local:: clean-cmis
 install-local:: install-cmis
@@ -212,10 +251,16 @@
 	  $(EchoCmd) "Uninstalling $(PROJ_libocamldir)/$$i"; \
 	  $(RM) -f "$(PROJ_libocamldir)/$$i"; \
 	done
+endif
 
 
 ##===- Build ocaml bytecode archive (.ml's -> .cmo's -> .cma) -------------===##
 
+$(ObjDir)/%.cmo: $(ObjDir)/%.ml
+	$(Echo) "Compiling $(notdir $<) for $(BuildMode) build"
+	$(Verb) $(Compile.CMO) $@ $<
+
+ifdef LIBRARYNAME
 all-local:: $(OutputCMA)
 clean-local:: clean-cma
 install-local:: install-cma
@@ -228,10 +273,6 @@
 	$(Echo) "Archiving $(notdir $@) for $(BuildMode) build"
 	$(Verb) $(Archive.CMA) $@ $(ObjectsCMO)
 
-$(ObjDir)/%.cmo: $(ObjDir)/%.ml
-	$(Echo) "Compiling $(notdir $<) for $(BuildMode) build"
-	$(Verb) $(Compile.CMO) $@ $<
-
 clean-cma::
 	$(Verb) $(RM) -f $(OutputCMA) $(UsedLibNames:%=$(OcamlDir)/%)
 
@@ -243,7 +284,7 @@
 uninstall-cma::
 	$(Echo) "Uninstalling $(DestCMA)"
 	-$(Verb) $(RM) -f $(DestCMA)
-
+endif
 
 ##===- Build optimized ocaml archive (.ml's -> .cmx's -> .cmxa, .a) -------===##
 
@@ -251,6 +292,14 @@
 # If unavailable, 'configure' will not define OCAMLOPT in Makefile.config.
 ifdef OCAMLOPT
 
+$(OcamlDir)/%.cmx: $(ObjDir)/%.cmx
+	$(Verb) $(CP) -f $< $@
+
+$(ObjDir)/%.cmx: $(ObjDir)/%.ml
+	$(Echo) "Compiling optimized $(notdir $<) for $(BuildMode) build"
+	$(Verb) $(Compile.CMX) $@ $<
+
+ifdef LIBRARYNAME
 all-local:: $(OutputCMXA) $(OutputsCMX)
 clean-local:: clean-cmxa
 install-local:: install-cmxa
@@ -260,18 +309,11 @@
 	$(Verb) $(CP) -f $< $@
 	$(Verb) $(CP) -f $(<:.cmxa=.a) $(@:.cmxa=.a)
 
-$(OcamlDir)/%.cmx: $(ObjDir)/%.cmx
-	$(Verb) $(CP) -f $< $@
-
 $(LibraryCMXA): $(ObjectsCMX)
 	$(Echo) "Archiving $(notdir $@) for $(BuildMode) build"
 	$(Verb) $(Archive.CMXA) $@ $(ObjectsCMX)
 	$(Verb) $(RM) -f $(@:.cmxa=.o)
 
-$(ObjDir)/%.cmx: $(ObjDir)/%.ml
-	$(Echo) "Compiling optimized $(notdir $<) for $(BuildMode) build"
-	$(Verb) $(Compile.CMX) $@ $<
-
 clean-cmxa::
 	$(Verb) $(RM) -f $(OutputCMXA) $(OutputCMXA:.cmxa=.a) $(OutputsCMX)
 
@@ -295,7 +337,27 @@
 	  $(EchoCmd) "Uninstalling $(PROJ_libocamldir)/$$i"; \
 	  $(RM) -f $(PROJ_libocamldir)/$$i; \
 	done
+endif
+endif
+
+##===- Build executables --------------------------------------------------===##
+
+ifdef TOOLNAME
+all-local:: $(OutputEXE)
+clean-local:: clean-exe
+
+$(OutputEXE): $(ToolEXE) $(OcamlDir)/.dir
+	$(Verb) $(CP) -f $< $@
 
+ifndef OCAMLOPT
+$(ToolEXE): $(ObjectsCMO) $(OcamlDir)/.dir
+	$(Echo) "Archiving $(notdir $@) for $(BuildMode) build"
+	$(Verb) $(Archive.EXE) $@ $<
+else
+$(ToolEXE): $(ObjectsCMX) $(OcamlDir)/.dir
+	$(Echo) "Archiving $(notdir $@) for $(BuildMode) build"
+	$(Verb) $(Archive.EXE) $@ $<
+endif
 endif
 
 ##===- Generate documentation ---------------------------------------------===##
@@ -325,7 +387,10 @@
 	$(Echo) "LibraryCMA   : " '$(LibraryCMA)'
 	$(Echo) "LibraryCMXA  : " '$(LibraryCMXA)'
 	$(Echo) "OcamlSources1: " '$(OcamlSources1)'
+	$(Echo) "OcamlSources2: " '$(OcamlSources2)'
 	$(Echo) "OcamlSources : " '$(OcamlSources)'
+	$(Echo) "OcamlHeaders1: " '$(OcamlHeaders1)'
+	$(Echo) "OcamlHeaders2: " '$(OcamlHeaders2)'
 	$(Echo) "OcamlHeaders : " '$(OcamlHeaders)'
 	$(Echo) "ObjectsCMI   : " '$(ObjectsCMI)'
 	$(Echo) "ObjectsCMO   : " '$(ObjectsCMO)'
@@ -340,4 +405,6 @@
 .PHONY: printcamlvars   build-cmis \
             clean-a     clean-cmis     clean-cma     clean-cmxa \
           install-a   install-cmis   install-cma   install-cmxa \
-		uninstall-a uninstall-cmis uninstall-cma uninstall-cmxa
+          install-exe \
+		uninstall-a uninstall-cmis uninstall-cma uninstall-cmxa \
+		uninstall-exe

Modified: llvm/trunk/examples/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/Makefile?rev=97966&r1=97965&r2=97966&view=diff
==============================================================================
--- llvm/trunk/examples/Makefile (original)
+++ llvm/trunk/examples/Makefile Mon Mar  8 13:32:27 2010
@@ -10,7 +10,8 @@
 
 include $(LEVEL)/Makefile.config
 
-PARALLEL_DIRS:= BrainF Fibonacci HowToUseJIT Kaleidoscope ModuleMaker
+PARALLEL_DIRS:= BrainF Fibonacci HowToUseJIT Kaleidoscope ModuleMaker \
+	OCaml-Kaleidoscope
 
 ifeq ($(HAVE_PTHREAD),1)
 PARALLEL_DIRS += ParallelJIT 

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,22 @@
+##===- examples/OCaml-Kaleidoscope/Chapter2/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 2.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch2
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm
+
+OCAMLCFLAGS += -pp camlp4of
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto = Prototype of string * string array
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,52 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,122 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the primary expression after the binary operator. *)
+        let rhs = parse_primary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_primary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')' *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,15 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,34 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            ignore(Parser.parse_definition stream);
+            print_endline "parsed a function definition.";
+        | Token.Extern ->
+            ignore(Parser.parse_extern stream);
+            print_endline "parsed an extern.";
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            ignore(Parser.parse_toplevel stream);
+            print_endline "parsed a top-level expr";
+        with Stream.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter2/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,21 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+let main () =
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop stream;
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,24 @@
+##===- examples/OCaml-Kaleidoscope/Chapter3/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 3.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch3
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm llvm_analysis
+
+OCAMLCFLAGS += -pp camlp4of
+
+ExcludeSources = $(PROJ_SRC_DIR)/myocamlbuild.ml
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1,2 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto = Prototype of string * string array
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/codegen.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/codegen.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/codegen.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/codegen.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,100 @@
+(*===----------------------------------------------------------------------===
+ * Code Generation
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+exception Error of string
+
+let context = global_context ()
+let the_module = create_module context "my cool jit"
+let builder = builder context
+let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
+
+let rec codegen_expr = function
+  | Ast.Number n -> const_float double_type n
+  | Ast.Variable name ->
+      (try Hashtbl.find named_values name with
+        | Not_found -> raise (Error "unknown variable name"))
+  | Ast.Binary (op, lhs, rhs) ->
+      let lhs_val = codegen_expr lhs in
+      let rhs_val = codegen_expr rhs in
+      begin
+        match op with
+        | '+' -> build_add lhs_val rhs_val "addtmp" builder
+        | '-' -> build_sub lhs_val rhs_val "subtmp" builder
+        | '*' -> build_mul lhs_val rhs_val "multmp" builder
+        | '<' ->
+            (* Convert bool 0/1 to double 0.0 or 1.0 *)
+            let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
+            build_uitofp i double_type "booltmp" builder
+        | _ -> raise (Error "invalid binary operator")
+      end
+  | Ast.Call (callee, args) ->
+      (* Look up the name in the module table. *)
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown function referenced")
+      in
+      let params = params callee in
+
+      (* If argument mismatch error. *)
+      if Array.length params == Array.length args then () else
+        raise (Error "incorrect # arguments passed");
+      let args = Array.map codegen_expr args in
+      build_call callee args "calltmp" builder
+
+let codegen_proto = function
+  | Ast.Prototype (name, args) ->
+      (* Make the function type: double(double,double) etc. *)
+      let doubles = Array.make (Array.length args) double_type in
+      let ft = function_type double_type doubles in
+      let f =
+        match lookup_function name the_module with
+        | None -> declare_function name ft the_module
+
+        (* If 'f' conflicted, there was already something named 'name'. If it
+         * has a body, don't allow redefinition or reextern. *)
+        | Some f ->
+            (* If 'f' already has a body, reject this. *)
+            if block_begin f <> At_end f then
+              raise (Error "redefinition of function");
+
+            (* If 'f' took a different number of arguments, reject. *)
+            if element_type (type_of f) <> ft then
+              raise (Error "redefinition of function with different # args");
+            f
+      in
+
+      (* Set names for all arguments. *)
+      Array.iteri (fun i a ->
+        let n = args.(i) in
+        set_value_name n a;
+        Hashtbl.add named_values n a;
+      ) (params f);
+      f
+
+let codegen_func = function
+  | Ast.Function (proto, body) ->
+      Hashtbl.clear named_values;
+      let the_function = codegen_proto proto in
+
+      (* Create a new basic block to start insertion into. *)
+      let bb = append_block context "entry" the_function in
+      position_at_end bb builder;
+
+      try
+        let ret_val = codegen_expr body in
+
+        (* Finish off the function. *)
+        let _ = build_ret ret_val builder in
+
+        (* Validate the generated code, checking for consistency. *)
+        Llvm_analysis.assert_valid_function the_function;
+
+        the_function
+      with e ->
+        delete_function the_function;
+        raise e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,52 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/myocamlbuild.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/myocamlbuild.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/myocamlbuild.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/myocamlbuild.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,6 @@
+open Ocamlbuild_plugin;;
+
+ocaml_lib ~extern:true "llvm";;
+ocaml_lib ~extern:true "llvm_analysis";;
+
+flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,122 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the primary expression after the binary operator. *)
+        let rhs = parse_primary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_primary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')' *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,15 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,39 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            let e = Parser.parse_definition stream in
+            print_endline "parsed a function definition.";
+            dump_value (Codegen.codegen_func e);
+        | Token.Extern ->
+            let e = Parser.parse_extern stream in
+            print_endline "parsed an extern.";
+            dump_value (Codegen.codegen_proto e);
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            let e = Parser.parse_toplevel stream in
+            print_endline "parsed a top-level expr";
+            dump_value (Codegen.codegen_func e);
+        with Stream.Error s | Codegen.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter3/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,26 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+let main () =
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop stream;
+
+  (* Print out all the generated code. *)
+  dump_module Codegen.the_module
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+##===- examples/OCaml-Kaleidoscope/Chapter4/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 4.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch4
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm llvm_analysis llvm_executionengine llvm_target \
+	llvm_scalar_opts
+
+OCAMLCFLAGS += -pp camlp4of
+
+ExcludeSources = $(PROJ_SRC_DIR)/myocamlbuild.ml
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1,4 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
+<*.{byte,native}>: use_llvm_executionengine, use_llvm_target
+<*.{byte,native}>: use_llvm_scalar_opts, use_bindings

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto = Prototype of string * string array
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/bindings.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/bindings.c?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/bindings.c (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/bindings.c Mon Mar  8 13:32:27 2010
@@ -0,0 +1,7 @@
+#include <stdio.h>
+
+/* putchard - putchar that takes a double and returns 0. */
+extern double putchard(double X) {
+  putchar((char)X);
+  return 0;
+}

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/codegen.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/codegen.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/codegen.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/codegen.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,103 @@
+(*===----------------------------------------------------------------------===
+ * Code Generation
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+exception Error of string
+
+let context = global_context ()
+let the_module = create_module context "my cool jit"
+let builder = builder context
+let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
+
+let rec codegen_expr = function
+  | Ast.Number n -> const_float double_type n
+  | Ast.Variable name ->
+      (try Hashtbl.find named_values name with
+        | Not_found -> raise (Error "unknown variable name"))
+  | Ast.Binary (op, lhs, rhs) ->
+      let lhs_val = codegen_expr lhs in
+      let rhs_val = codegen_expr rhs in
+      begin
+        match op with
+        | '+' -> build_add lhs_val rhs_val "addtmp" builder
+        | '-' -> build_sub lhs_val rhs_val "subtmp" builder
+        | '*' -> build_mul lhs_val rhs_val "multmp" builder
+        | '<' ->
+            (* Convert bool 0/1 to double 0.0 or 1.0 *)
+            let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
+            build_uitofp i double_type "booltmp" builder
+        | _ -> raise (Error "invalid binary operator")
+      end
+  | Ast.Call (callee, args) ->
+      (* Look up the name in the module table. *)
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown function referenced")
+      in
+      let params = params callee in
+
+      (* If argument mismatch error. *)
+      if Array.length params == Array.length args then () else
+        raise (Error "incorrect # arguments passed");
+      let args = Array.map codegen_expr args in
+      build_call callee args "calltmp" builder
+
+let codegen_proto = function
+  | Ast.Prototype (name, args) ->
+      (* Make the function type: double(double,double) etc. *)
+      let doubles = Array.make (Array.length args) double_type in
+      let ft = function_type double_type doubles in
+      let f =
+        match lookup_function name the_module with
+        | None -> declare_function name ft the_module
+
+        (* If 'f' conflicted, there was already something named 'name'. If it
+         * has a body, don't allow redefinition or reextern. *)
+        | Some f ->
+            (* If 'f' already has a body, reject this. *)
+            if block_begin f <> At_end f then
+              raise (Error "redefinition of function");
+
+            (* If 'f' took a different number of arguments, reject. *)
+            if element_type (type_of f) <> ft then
+              raise (Error "redefinition of function with different # args");
+            f
+      in
+
+      (* Set names for all arguments. *)
+      Array.iteri (fun i a ->
+        let n = args.(i) in
+        set_value_name n a;
+        Hashtbl.add named_values n a;
+      ) (params f);
+      f
+
+let codegen_func the_fpm = function
+  | Ast.Function (proto, body) ->
+      Hashtbl.clear named_values;
+      let the_function = codegen_proto proto in
+
+      (* Create a new basic block to start insertion into. *)
+      let bb = append_block context "entry" the_function in
+      position_at_end bb builder;
+
+      try
+        let ret_val = codegen_expr body in
+
+        (* Finish off the function. *)
+        let _ = build_ret ret_val builder in
+
+        (* Validate the generated code, checking for consistency. *)
+        Llvm_analysis.assert_valid_function the_function;
+
+        (* Optimize the function. *)
+        let _ = PassManager.run_function the_function the_fpm in
+
+        the_function
+      with e ->
+        delete_function the_function;
+        raise e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,52 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/myocamlbuild.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/myocamlbuild.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/myocamlbuild.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/myocamlbuild.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,10 @@
+open Ocamlbuild_plugin;;
+
+ocaml_lib ~extern:true "llvm";;
+ocaml_lib ~extern:true "llvm_analysis";;
+ocaml_lib ~extern:true "llvm_executionengine";;
+ocaml_lib ~extern:true "llvm_target";;
+ocaml_lib ~extern:true "llvm_scalar_opts";;
+
+flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
+dep ["link"; "ocaml"; "use_bindings"] ["bindings.o"];;

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,122 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the primary expression after the binary operator. *)
+        let rhs = parse_primary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_primary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')' *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,15 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,49 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop the_fpm the_execution_engine stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop the_fpm the_execution_engine stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            let e = Parser.parse_definition stream in
+            print_endline "parsed a function definition.";
+            dump_value (Codegen.codegen_func the_fpm e);
+        | Token.Extern ->
+            let e = Parser.parse_extern stream in
+            print_endline "parsed an extern.";
+            dump_value (Codegen.codegen_proto e);
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            let e = Parser.parse_toplevel stream in
+            print_endline "parsed a top-level expr";
+            let the_function = Codegen.codegen_func the_fpm e in
+            dump_value the_function;
+
+            (* JIT the function, returning a function pointer. *)
+            let result = ExecutionEngine.run_function the_function [||]
+              the_execution_engine in
+
+            print_string "Evaluated to ";
+            print_float (GenericValue.as_float Codegen.double_type result);
+            print_newline ();
+        with Stream.Error s | Codegen.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop the_fpm the_execution_engine stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter4/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,53 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+open Llvm_target
+open Llvm_scalar_opts
+
+let main () =
+  ignore (initialize_native_target ());
+
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Create the JIT. *)
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
+
+  (* Set up the optimizer pipeline.  Start with registering info about how the
+   * target lays out data structures. *)
+  TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
+
+  (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
+  add_instruction_combination the_fpm;
+
+  (* reassociate expressions. *)
+  add_reassociation the_fpm;
+
+  (* Eliminate Common SubExpressions. *)
+  add_gvn the_fpm;
+
+  (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
+  add_cfg_simplification the_fpm;
+
+  ignore (PassManager.initialize the_fpm);
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop the_fpm the_execution_engine stream;
+
+  (* Print out all the generated code. *)
+  dump_module Codegen.the_module
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+##===- examples/OCaml-Kaleidoscope/Chapter5/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 5.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch5
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm llvm_analysis llvm_executionengine llvm_target \
+	llvm_scalar_opts
+
+OCAMLCFLAGS += -pp camlp4of
+
+ExcludeSources = $(PROJ_SRC_DIR)/myocamlbuild.ml
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1,4 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
+<*.{byte,native}>: use_llvm_executionengine, use_llvm_target
+<*.{byte,native}>: use_llvm_scalar_opts, use_bindings

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,31 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+  (* variant for if/then/else. *)
+  | If of expr * expr * expr
+
+  (* variant for for/in. *)
+  | For of string * expr * expr * expr option * expr
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto = Prototype of string * string array
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/bindings.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/bindings.c?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/bindings.c (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/bindings.c Mon Mar  8 13:32:27 2010
@@ -0,0 +1,7 @@
+#include <stdio.h>
+
+/* putchard - putchar that takes a double and returns 0. */
+extern double putchard(double X) {
+  putchar((char)X);
+  return 0;
+}

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/codegen.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/codegen.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/codegen.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/codegen.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,225 @@
+(*===----------------------------------------------------------------------===
+ * Code Generation
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+exception Error of string
+
+let context = global_context ()
+let the_module = create_module context "my cool jit"
+let builder = builder context
+let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
+
+let rec codegen_expr = function
+  | Ast.Number n -> const_float double_type n
+  | Ast.Variable name ->
+      (try Hashtbl.find named_values name with
+        | Not_found -> raise (Error "unknown variable name"))
+  | Ast.Binary (op, lhs, rhs) ->
+      let lhs_val = codegen_expr lhs in
+      let rhs_val = codegen_expr rhs in
+      begin
+        match op with
+        | '+' -> build_add lhs_val rhs_val "addtmp" builder
+        | '-' -> build_sub lhs_val rhs_val "subtmp" builder
+        | '*' -> build_mul lhs_val rhs_val "multmp" builder
+        | '<' ->
+            (* Convert bool 0/1 to double 0.0 or 1.0 *)
+            let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
+            build_uitofp i double_type "booltmp" builder
+        | _ -> raise (Error "invalid binary operator")
+      end
+  | Ast.Call (callee, args) ->
+      (* Look up the name in the module table. *)
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown function referenced")
+      in
+      let params = params callee in
+
+      (* If argument mismatch error. *)
+      if Array.length params == Array.length args then () else
+        raise (Error "incorrect # arguments passed");
+      let args = Array.map codegen_expr args in
+      build_call callee args "calltmp" builder
+  | Ast.If (cond, then_, else_) ->
+      let cond = codegen_expr cond in
+
+      (* Convert condition to a bool by comparing equal to 0.0 *)
+      let zero = const_float double_type 0.0 in
+      let cond_val = build_fcmp Fcmp.One cond zero "ifcond" builder in
+
+      (* Grab the first block so that we might later add the conditional branch
+       * to it at the end of the function. *)
+      let start_bb = insertion_block builder in
+      let the_function = block_parent start_bb in
+
+      let then_bb = append_block context "then" the_function in
+
+      (* Emit 'then' value. *)
+      position_at_end then_bb builder;
+      let then_val = codegen_expr then_ in
+
+      (* Codegen of 'then' can change the current block, update then_bb for the
+       * phi. We create a new name because one is used for the phi node, and the
+       * other is used for the conditional branch. *)
+      let new_then_bb = insertion_block builder in
+
+      (* Emit 'else' value. *)
+      let else_bb = append_block context "else" the_function in
+      position_at_end else_bb builder;
+      let else_val = codegen_expr else_ in
+
+      (* Codegen of 'else' can change the current block, update else_bb for the
+       * phi. *)
+      let new_else_bb = insertion_block builder in
+
+      (* Emit merge block. *)
+      let merge_bb = append_block context "ifcont" the_function in
+      position_at_end merge_bb builder;
+      let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
+      let phi = build_phi incoming "iftmp" builder in
+
+      (* Return to the start block to add the conditional branch. *)
+      position_at_end start_bb builder;
+      ignore (build_cond_br cond_val then_bb else_bb builder);
+
+      (* Set a unconditional branch at the end of the 'then' block and the
+       * 'else' block to the 'merge' block. *)
+      position_at_end new_then_bb builder; ignore (build_br merge_bb builder);
+      position_at_end new_else_bb builder; ignore (build_br merge_bb builder);
+
+      (* Finally, set the builder to the end of the merge block. *)
+      position_at_end merge_bb builder;
+
+      phi
+  | Ast.For (var_name, start, end_, step, body) ->
+      (* Emit the start code first, without 'variable' in scope. *)
+      let start_val = codegen_expr start in
+
+      (* Make the new basic block for the loop header, inserting after current
+       * block. *)
+      let preheader_bb = insertion_block builder in
+      let the_function = block_parent preheader_bb in
+      let loop_bb = append_block context "loop" the_function in
+
+      (* Insert an explicit fall through from the current block to the
+       * loop_bb. *)
+      ignore (build_br loop_bb builder);
+
+      (* Start insertion in loop_bb. *)
+      position_at_end loop_bb builder;
+
+      (* Start the PHI node with an entry for start. *)
+      let variable = build_phi [(start_val, preheader_bb)] var_name builder in
+
+      (* Within the loop, the variable is defined equal to the PHI node. If it
+       * shadows an existing variable, we have to restore it, so save it
+       * now. *)
+      let old_val =
+        try Some (Hashtbl.find named_values var_name) with Not_found -> None
+      in
+      Hashtbl.add named_values var_name variable;
+
+      (* Emit the body of the loop.  This, like any other expr, can change the
+       * current BB.  Note that we ignore the value computed by the body, but
+       * don't allow an error *)
+      ignore (codegen_expr body);
+
+      (* Emit the step value. *)
+      let step_val =
+        match step with
+        | Some step -> codegen_expr step
+        (* If not specified, use 1.0. *)
+        | None -> const_float double_type 1.0
+      in
+
+      let next_var = build_add variable step_val "nextvar" builder in
+
+      (* Compute the end condition. *)
+      let end_cond = codegen_expr end_ in
+
+      (* Convert condition to a bool by comparing equal to 0.0. *)
+      let zero = const_float double_type 0.0 in
+      let end_cond = build_fcmp Fcmp.One end_cond zero "loopcond" builder in
+
+      (* Create the "after loop" block and insert it. *)
+      let loop_end_bb = insertion_block builder in
+      let after_bb = append_block context "afterloop" the_function in
+
+      (* Insert the conditional branch into the end of loop_end_bb. *)
+      ignore (build_cond_br end_cond loop_bb after_bb builder);
+
+      (* Any new code will be inserted in after_bb. *)
+      position_at_end after_bb builder;
+
+      (* Add a new entry to the PHI node for the backedge. *)
+      add_incoming (next_var, loop_end_bb) variable;
+
+      (* Restore the unshadowed variable. *)
+      begin match old_val with
+      | Some old_val -> Hashtbl.add named_values var_name old_val
+      | None -> ()
+      end;
+
+      (* for expr always returns 0.0. *)
+      const_null double_type
+
+let codegen_proto = function
+  | Ast.Prototype (name, args) ->
+      (* Make the function type: double(double,double) etc. *)
+      let doubles = Array.make (Array.length args) double_type in
+      let ft = function_type double_type doubles in
+      let f =
+        match lookup_function name the_module with
+        | None -> declare_function name ft the_module
+
+        (* If 'f' conflicted, there was already something named 'name'. If it
+         * has a body, don't allow redefinition or reextern. *)
+        | Some f ->
+            (* If 'f' already has a body, reject this. *)
+            if block_begin f <> At_end f then
+              raise (Error "redefinition of function");
+
+            (* If 'f' took a different number of arguments, reject. *)
+            if element_type (type_of f) <> ft then
+              raise (Error "redefinition of function with different # args");
+            f
+      in
+
+      (* Set names for all arguments. *)
+      Array.iteri (fun i a ->
+        let n = args.(i) in
+        set_value_name n a;
+        Hashtbl.add named_values n a;
+      ) (params f);
+      f
+
+let codegen_func the_fpm = function
+  | Ast.Function (proto, body) ->
+      Hashtbl.clear named_values;
+      let the_function = codegen_proto proto in
+
+      (* Create a new basic block to start insertion into. *)
+      let bb = append_block context "entry" the_function in
+      position_at_end bb builder;
+
+      try
+        let ret_val = codegen_expr body in
+
+        (* Finish off the function. *)
+        let _ = build_ret ret_val builder in
+
+        (* Validate the generated code, checking for consistency. *)
+        Llvm_analysis.assert_valid_function the_function;
+
+        (* Optimize the function. *)
+        let _ = PassManager.run_function the_function the_fpm in
+
+        the_function
+      with e ->
+        delete_function the_function;
+        raise e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,57 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | "if" -> [< 'Token.If; stream >]
+      | "then" -> [< 'Token.Then; stream >]
+      | "else" -> [< 'Token.Else; stream >]
+      | "for" -> [< 'Token.For; stream >]
+      | "in" -> [< 'Token.In; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/myocamlbuild.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/myocamlbuild.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/myocamlbuild.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/myocamlbuild.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,10 @@
+open Ocamlbuild_plugin;;
+
+ocaml_lib ~extern:true "llvm";;
+ocaml_lib ~extern:true "llvm_analysis";;
+ocaml_lib ~extern:true "llvm_executionengine";;
+ocaml_lib ~extern:true "llvm_target";;
+ocaml_lib ~extern:true "llvm_scalar_opts";;
+
+flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
+dep ["link"; "ocaml"; "use_bindings"] ["bindings.o"];;

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,158 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr
+ *   ::= ifexpr
+ *   ::= forexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  (* ifexpr ::= 'if' expr 'then' expr 'else' expr *)
+  | [< 'Token.If; c=parse_expr;
+       'Token.Then ?? "expected 'then'"; t=parse_expr;
+       'Token.Else ?? "expected 'else'"; e=parse_expr >] ->
+      Ast.If (c, t, e)
+
+  (* forexpr
+        ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression *)
+  | [< 'Token.For;
+       'Token.Ident id ?? "expected identifier after for";
+       'Token.Kwd '=' ?? "expected '=' after for";
+       stream >] ->
+      begin parser
+        | [<
+             start=parse_expr;
+             'Token.Kwd ',' ?? "expected ',' after for";
+             end_=parse_expr;
+             stream >] ->
+            let step =
+              begin parser
+              | [< 'Token.Kwd ','; step=parse_expr >] -> Some step
+              | [< >] -> None
+              end stream
+            in
+            begin parser
+            | [< 'Token.In; body=parse_expr >] ->
+                Ast.For (id, start, end_, step, body)
+            | [< >] ->
+                raise (Stream.Error "expected 'in' after for")
+            end stream
+        | [< >] ->
+            raise (Stream.Error "expected '=' after for")
+      end stream
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the primary expression after the binary operator. *)
+        let rhs = parse_primary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_primary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')' *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,19 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char
+
+  (* control *)
+  | If | Then | Else
+  | For | In

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,49 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop the_fpm the_execution_engine stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop the_fpm the_execution_engine stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            let e = Parser.parse_definition stream in
+            print_endline "parsed a function definition.";
+            dump_value (Codegen.codegen_func the_fpm e);
+        | Token.Extern ->
+            let e = Parser.parse_extern stream in
+            print_endline "parsed an extern.";
+            dump_value (Codegen.codegen_proto e);
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            let e = Parser.parse_toplevel stream in
+            print_endline "parsed a top-level expr";
+            let the_function = Codegen.codegen_func the_fpm e in
+            dump_value the_function;
+
+            (* JIT the function, returning a function pointer. *)
+            let result = ExecutionEngine.run_function the_function [||]
+              the_execution_engine in
+
+            print_string "Evaluated to ";
+            print_float (GenericValue.as_float Codegen.double_type result);
+            print_newline ();
+        with Stream.Error s | Codegen.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop the_fpm the_execution_engine stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter5/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,53 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+open Llvm_target
+open Llvm_scalar_opts
+
+let main () =
+  ignore (initialize_native_target ());
+
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Create the JIT. *)
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
+
+  (* Set up the optimizer pipeline.  Start with registering info about how the
+   * target lays out data structures. *)
+  TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
+
+  (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
+  add_instruction_combination the_fpm;
+
+  (* reassociate expressions. *)
+  add_reassociation the_fpm;
+
+  (* Eliminate Common SubExpressions. *)
+  add_gvn the_fpm;
+
+  (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
+  add_cfg_simplification the_fpm;
+
+  ignore (PassManager.initialize the_fpm);
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop the_fpm the_execution_engine stream;
+
+  (* Print out all the generated code. *)
+  dump_module Codegen.the_module
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+##===- examples/OCaml-Kaleidoscope/Chapter6/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 6.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch6
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm llvm_analysis llvm_executionengine llvm_target \
+	llvm_scalar_opts
+
+OCAMLCFLAGS += -pp camlp4of
+
+ExcludeSources = $(PROJ_SRC_DIR)/myocamlbuild.ml
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1,4 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
+<*.{byte,native}>: use_llvm_executionengine, use_llvm_target
+<*.{byte,native}>: use_llvm_scalar_opts, use_bindings

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,36 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a unary operator. *)
+  | Unary of char * expr
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+  (* variant for if/then/else. *)
+  | If of expr * expr * expr
+
+  (* variant for for/in. *)
+  | For of string * expr * expr * expr option * expr
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto =
+  | Prototype of string * string array
+  | BinOpPrototype of string * string array * int
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/bindings.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/bindings.c?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/bindings.c (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/bindings.c Mon Mar  8 13:32:27 2010
@@ -0,0 +1,13 @@
+#include <stdio.h>
+
+/* putchard - putchar that takes a double and returns 0. */
+extern double putchard(double X) {
+  putchar((char)X);
+  return 0;
+}
+
+/* printd - printf that takes a double prints it as "%f\n", returning 0. */
+extern double printd(double X) {
+  printf("%f\n", X);
+  return 0;
+}

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/codegen.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/codegen.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/codegen.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/codegen.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,251 @@
+(*===----------------------------------------------------------------------===
+ * Code Generation
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+exception Error of string
+
+let context = global_context ()
+let the_module = create_module context "my cool jit"
+let builder = builder context
+let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
+
+let rec codegen_expr = function
+  | Ast.Number n -> const_float double_type n
+  | Ast.Variable name ->
+      (try Hashtbl.find named_values name with
+        | Not_found -> raise (Error "unknown variable name"))
+  | Ast.Unary (op, operand) ->
+      let operand = codegen_expr operand in
+      let callee = "unary" ^ (String.make 1 op) in
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown unary operator")
+      in
+      build_call callee [|operand|] "unop" builder
+  | Ast.Binary (op, lhs, rhs) ->
+      let lhs_val = codegen_expr lhs in
+      let rhs_val = codegen_expr rhs in
+      begin
+        match op with
+        | '+' -> build_add lhs_val rhs_val "addtmp" builder
+        | '-' -> build_sub lhs_val rhs_val "subtmp" builder
+        | '*' -> build_mul lhs_val rhs_val "multmp" builder
+        | '<' ->
+            (* Convert bool 0/1 to double 0.0 or 1.0 *)
+            let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
+            build_uitofp i double_type "booltmp" builder
+        | _ ->
+            (* If it wasn't a builtin binary operator, it must be a user defined
+             * one. Emit a call to it. *)
+            let callee = "binary" ^ (String.make 1 op) in
+            let callee =
+              match lookup_function callee the_module with
+              | Some callee -> callee
+              | None -> raise (Error "binary operator not found!")
+            in
+            build_call callee [|lhs_val; rhs_val|] "binop" builder
+      end
+  | Ast.Call (callee, args) ->
+      (* Look up the name in the module table. *)
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown function referenced")
+      in
+      let params = params callee in
+
+      (* If argument mismatch error. *)
+      if Array.length params == Array.length args then () else
+        raise (Error "incorrect # arguments passed");
+      let args = Array.map codegen_expr args in
+      build_call callee args "calltmp" builder
+  | Ast.If (cond, then_, else_) ->
+      let cond = codegen_expr cond in
+
+      (* Convert condition to a bool by comparing equal to 0.0 *)
+      let zero = const_float double_type 0.0 in
+      let cond_val = build_fcmp Fcmp.One cond zero "ifcond" builder in
+
+      (* Grab the first block so that we might later add the conditional branch
+       * to it at the end of the function. *)
+      let start_bb = insertion_block builder in
+      let the_function = block_parent start_bb in
+
+      let then_bb = append_block context "then" the_function in
+
+      (* Emit 'then' value. *)
+      position_at_end then_bb builder;
+      let then_val = codegen_expr then_ in
+
+      (* Codegen of 'then' can change the current block, update then_bb for the
+       * phi. We create a new name because one is used for the phi node, and the
+       * other is used for the conditional branch. *)
+      let new_then_bb = insertion_block builder in
+
+      (* Emit 'else' value. *)
+      let else_bb = append_block context "else" the_function in
+      position_at_end else_bb builder;
+      let else_val = codegen_expr else_ in
+
+      (* Codegen of 'else' can change the current block, update else_bb for the
+       * phi. *)
+      let new_else_bb = insertion_block builder in
+
+      (* Emit merge block. *)
+      let merge_bb = append_block context "ifcont" the_function in
+      position_at_end merge_bb builder;
+      let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
+      let phi = build_phi incoming "iftmp" builder in
+
+      (* Return to the start block to add the conditional branch. *)
+      position_at_end start_bb builder;
+      ignore (build_cond_br cond_val then_bb else_bb builder);
+
+      (* Set a unconditional branch at the end of the 'then' block and the
+       * 'else' block to the 'merge' block. *)
+      position_at_end new_then_bb builder; ignore (build_br merge_bb builder);
+      position_at_end new_else_bb builder; ignore (build_br merge_bb builder);
+
+      (* Finally, set the builder to the end of the merge block. *)
+      position_at_end merge_bb builder;
+
+      phi
+  | Ast.For (var_name, start, end_, step, body) ->
+      (* Emit the start code first, without 'variable' in scope. *)
+      let start_val = codegen_expr start in
+
+      (* Make the new basic block for the loop header, inserting after current
+       * block. *)
+      let preheader_bb = insertion_block builder in
+      let the_function = block_parent preheader_bb in
+      let loop_bb = append_block context "loop" the_function in
+
+      (* Insert an explicit fall through from the current block to the
+       * loop_bb. *)
+      ignore (build_br loop_bb builder);
+
+      (* Start insertion in loop_bb. *)
+      position_at_end loop_bb builder;
+
+      (* Start the PHI node with an entry for start. *)
+      let variable = build_phi [(start_val, preheader_bb)] var_name builder in
+
+      (* Within the loop, the variable is defined equal to the PHI node. If it
+       * shadows an existing variable, we have to restore it, so save it
+       * now. *)
+      let old_val =
+        try Some (Hashtbl.find named_values var_name) with Not_found -> None
+      in
+      Hashtbl.add named_values var_name variable;
+
+      (* Emit the body of the loop.  This, like any other expr, can change the
+       * current BB.  Note that we ignore the value computed by the body, but
+       * don't allow an error *)
+      ignore (codegen_expr body);
+
+      (* Emit the step value. *)
+      let step_val =
+        match step with
+        | Some step -> codegen_expr step
+        (* If not specified, use 1.0. *)
+        | None -> const_float double_type 1.0
+      in
+
+      let next_var = build_add variable step_val "nextvar" builder in
+
+      (* Compute the end condition. *)
+      let end_cond = codegen_expr end_ in
+
+      (* Convert condition to a bool by comparing equal to 0.0. *)
+      let zero = const_float double_type 0.0 in
+      let end_cond = build_fcmp Fcmp.One end_cond zero "loopcond" builder in
+
+      (* Create the "after loop" block and insert it. *)
+      let loop_end_bb = insertion_block builder in
+      let after_bb = append_block context "afterloop" the_function in
+
+      (* Insert the conditional branch into the end of loop_end_bb. *)
+      ignore (build_cond_br end_cond loop_bb after_bb builder);
+
+      (* Any new code will be inserted in after_bb. *)
+      position_at_end after_bb builder;
+
+      (* Add a new entry to the PHI node for the backedge. *)
+      add_incoming (next_var, loop_end_bb) variable;
+
+      (* Restore the unshadowed variable. *)
+      begin match old_val with
+      | Some old_val -> Hashtbl.add named_values var_name old_val
+      | None -> ()
+      end;
+
+      (* for expr always returns 0.0. *)
+      const_null double_type
+
+let codegen_proto = function
+  | Ast.Prototype (name, args) | Ast.BinOpPrototype (name, args, _) ->
+      (* Make the function type: double(double,double) etc. *)
+      let doubles = Array.make (Array.length args) double_type in
+      let ft = function_type double_type doubles in
+      let f =
+        match lookup_function name the_module with
+        | None -> declare_function name ft the_module
+
+        (* If 'f' conflicted, there was already something named 'name'. If it
+         * has a body, don't allow redefinition or reextern. *)
+        | Some f ->
+            (* If 'f' already has a body, reject this. *)
+            if block_begin f <> At_end f then
+              raise (Error "redefinition of function");
+
+            (* If 'f' took a different number of arguments, reject. *)
+            if element_type (type_of f) <> ft then
+              raise (Error "redefinition of function with different # args");
+            f
+      in
+
+      (* Set names for all arguments. *)
+      Array.iteri (fun i a ->
+        let n = args.(i) in
+        set_value_name n a;
+        Hashtbl.add named_values n a;
+      ) (params f);
+      f
+
+let codegen_func the_fpm = function
+  | Ast.Function (proto, body) ->
+      Hashtbl.clear named_values;
+      let the_function = codegen_proto proto in
+
+      (* If this is an operator, install it. *)
+      begin match proto with
+      | Ast.BinOpPrototype (name, args, prec) ->
+          let op = name.[String.length name - 1] in
+          Hashtbl.add Parser.binop_precedence op prec;
+      | _ -> ()
+      end;
+
+      (* Create a new basic block to start insertion into. *)
+      let bb = append_block context "entry" the_function in
+      position_at_end bb builder;
+
+      try
+        let ret_val = codegen_expr body in
+
+        (* Finish off the function. *)
+        let _ = build_ret ret_val builder in
+
+        (* Validate the generated code, checking for consistency. *)
+        Llvm_analysis.assert_valid_function the_function;
+
+        (* Optimize the function. *)
+        let _ = PassManager.run_function the_function the_fpm in
+
+        the_function
+      with e ->
+        delete_function the_function;
+        raise e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,59 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | "if" -> [< 'Token.If; stream >]
+      | "then" -> [< 'Token.Then; stream >]
+      | "else" -> [< 'Token.Else; stream >]
+      | "for" -> [< 'Token.For; stream >]
+      | "in" -> [< 'Token.In; stream >]
+      | "binary" -> [< 'Token.Binary; stream >]
+      | "unary" -> [< 'Token.Unary; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/myocamlbuild.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/myocamlbuild.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/myocamlbuild.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/myocamlbuild.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,10 @@
+open Ocamlbuild_plugin;;
+
+ocaml_lib ~extern:true "llvm";;
+ocaml_lib ~extern:true "llvm_analysis";;
+ocaml_lib ~extern:true "llvm_executionengine";;
+ocaml_lib ~extern:true "llvm_target";;
+ocaml_lib ~extern:true "llvm_scalar_opts";;
+
+flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
+dep ["link"; "ocaml"; "use_bindings"] ["bindings.o"];;

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,195 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr
+ *   ::= ifexpr
+ *   ::= forexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  (* ifexpr ::= 'if' expr 'then' expr 'else' expr *)
+  | [< 'Token.If; c=parse_expr;
+       'Token.Then ?? "expected 'then'"; t=parse_expr;
+       'Token.Else ?? "expected 'else'"; e=parse_expr >] ->
+      Ast.If (c, t, e)
+
+  (* forexpr
+        ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression *)
+  | [< 'Token.For;
+       'Token.Ident id ?? "expected identifier after for";
+       'Token.Kwd '=' ?? "expected '=' after for";
+       stream >] ->
+      begin parser
+        | [<
+             start=parse_expr;
+             'Token.Kwd ',' ?? "expected ',' after for";
+             end_=parse_expr;
+             stream >] ->
+            let step =
+              begin parser
+              | [< 'Token.Kwd ','; step=parse_expr >] -> Some step
+              | [< >] -> None
+              end stream
+            in
+            begin parser
+            | [< 'Token.In; body=parse_expr >] ->
+                Ast.For (id, start, end_, step, body)
+            | [< >] ->
+                raise (Stream.Error "expected 'in' after for")
+            end stream
+        | [< >] ->
+            raise (Stream.Error "expected '=' after for")
+      end stream
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* unary
+ *   ::= primary
+ *   ::= '!' unary *)
+and parse_unary = parser
+  (* If this is a unary operator, read it. *)
+  | [< 'Token.Kwd op when op != '(' && op != ')'; operand=parse_expr >] ->
+      Ast.Unary (op, operand)
+
+  (* If the current token is not an operator, it must be a primary expr. *)
+  | [< stream >] -> parse_primary stream
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the unary expression after the binary operator. *)
+        let rhs = parse_unary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_unary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')'
+ *   ::= binary LETTER number? (id, id)
+ *   ::= unary LETTER number? (id) *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+  let parse_operator = parser
+    | [< 'Token.Unary >] -> "unary", 1
+    | [< 'Token.Binary >] -> "binary", 2
+  in
+  let parse_binary_precedence = parser
+    | [< 'Token.Number n >] -> int_of_float n
+    | [< >] -> 30
+  in
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+  | [< (prefix, kind)=parse_operator;
+       'Token.Kwd op ?? "expected an operator";
+       (* Read the precedence if present. *)
+       binary_precedence=parse_binary_precedence;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+        args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      let name = prefix ^ (String.make 1 op) in
+      let args = Array.of_list (List.rev args) in
+
+      (* Verify right number of arguments for operator. *)
+      if Array.length args != kind
+      then raise (Stream.Error "invalid number of operands for operator")
+      else
+        if kind == 1 then
+          Ast.Prototype (name, args)
+        else
+          Ast.BinOpPrototype (name, args, binary_precedence)
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,22 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char
+
+  (* control *)
+  | If | Then | Else
+  | For | In
+
+  (* operators *)
+  | Binary | Unary

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,49 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop the_fpm the_execution_engine stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop the_fpm the_execution_engine stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            let e = Parser.parse_definition stream in
+            print_endline "parsed a function definition.";
+            dump_value (Codegen.codegen_func the_fpm e);
+        | Token.Extern ->
+            let e = Parser.parse_extern stream in
+            print_endline "parsed an extern.";
+            dump_value (Codegen.codegen_proto e);
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            let e = Parser.parse_toplevel stream in
+            print_endline "parsed a top-level expr";
+            let the_function = Codegen.codegen_func the_fpm e in
+            dump_value the_function;
+
+            (* JIT the function, returning a function pointer. *)
+            let result = ExecutionEngine.run_function the_function [||]
+              the_execution_engine in
+
+            print_string "Evaluated to ";
+            print_float (GenericValue.as_float Codegen.double_type result);
+            print_newline ();
+        with Stream.Error s | Codegen.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop the_fpm the_execution_engine stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter6/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,53 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+open Llvm_target
+open Llvm_scalar_opts
+
+let main () =
+  ignore (initialize_native_target ());
+
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Create the JIT. *)
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
+
+  (* Set up the optimizer pipeline.  Start with registering info about how the
+   * target lays out data structures. *)
+  TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
+
+  (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
+  add_instruction_combination the_fpm;
+
+  (* reassociate expressions. *)
+  add_reassociation the_fpm;
+
+  (* Eliminate Common SubExpressions. *)
+  add_gvn the_fpm;
+
+  (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
+  add_cfg_simplification the_fpm;
+
+  ignore (PassManager.initialize the_fpm);
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop the_fpm the_execution_engine stream;
+
+  (* Print out all the generated code. *)
+  dump_module Codegen.the_module
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+##===- examples/OCaml-Kaleidoscope/Chapter7/Makefile -------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+# 
+# This is the makefile for the Objective Caml kaleidoscope tutorial, chapter 7.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL := ../../..
+TOOLNAME := OCaml-Kaleidoscope-Ch7
+EXAMPLE_TOOL := 1
+UsedComponents := core
+UsedOcamLibs := llvm llvm_analysis llvm_executionengine llvm_target \
+	llvm_scalar_opts
+
+OCAMLCFLAGS += -pp camlp4of
+
+ExcludeSources = $(PROJ_SRC_DIR)/myocamlbuild.ml
+
+include $(LEVEL)/bindings/ocaml/Makefile.ocaml

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/_tags
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/_tags?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/_tags (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/_tags Mon Mar  8 13:32:27 2010
@@ -0,0 +1,4 @@
+<{lexer,parser}.ml>: use_camlp4, pp(camlp4of)
+<*.{byte,native}>: g++, use_llvm, use_llvm_analysis
+<*.{byte,native}>: use_llvm_executionengine, use_llvm_target
+<*.{byte,native}>: use_llvm_scalar_opts, use_bindings

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/ast.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/ast.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/ast.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/ast.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,39 @@
+(*===----------------------------------------------------------------------===
+ * Abstract Syntax Tree (aka Parse Tree)
+ *===----------------------------------------------------------------------===*)
+
+(* expr - Base type for all expression nodes. *)
+type expr =
+  (* variant for numeric literals like "1.0". *)
+  | Number of float
+
+  (* variant for referencing a variable, like "a". *)
+  | Variable of string
+
+  (* variant for a unary operator. *)
+  | Unary of char * expr
+
+  (* variant for a binary operator. *)
+  | Binary of char * expr * expr
+
+  (* variant for function calls. *)
+  | Call of string * expr array
+
+  (* variant for if/then/else. *)
+  | If of expr * expr * expr
+
+  (* variant for for/in. *)
+  | For of string * expr * expr * expr option * expr
+
+  (* variant for var/in. *)
+  | Var of (string * expr option) array * expr
+
+(* proto - This type represents the "prototype" for a function, which captures
+ * its name, and its argument names (thus implicitly the number of arguments the
+ * function takes). *)
+type proto =
+  | Prototype of string * string array
+  | BinOpPrototype of string * string array * int
+
+(* func - This type represents a function definition itself. *)
+type func = Function of proto * expr

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/bindings.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/bindings.c?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/bindings.c (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/bindings.c Mon Mar  8 13:32:27 2010
@@ -0,0 +1,13 @@
+#include <stdio.h>
+
+/* putchard - putchar that takes a double and returns 0. */
+extern double putchard(double X) {
+  putchar((char)X);
+  return 0;
+}
+
+/* printd - printf that takes a double prints it as "%f\n", returning 0. */
+extern double printd(double X) {
+  printf("%f\n", X);
+  return 0;
+}

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/codegen.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/codegen.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/codegen.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/codegen.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,370 @@
+(*===----------------------------------------------------------------------===
+ * Code Generation
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+
+exception Error of string
+
+let context = global_context ()
+let the_module = create_module context "my cool jit"
+let builder = builder context
+let named_values:(string, llvalue) Hashtbl.t = Hashtbl.create 10
+let double_type = double_type context
+
+(* Create an alloca instruction in the entry block of the function. This
+ * is used for mutable variables etc. *)
+let create_entry_block_alloca the_function var_name =
+  let builder = builder_at context (instr_begin (entry_block the_function)) in
+  build_alloca double_type var_name builder
+
+let rec codegen_expr = function
+  | Ast.Number n -> const_float double_type n
+  | Ast.Variable name ->
+      let v = try Hashtbl.find named_values name with
+        | Not_found -> raise (Error "unknown variable name")
+      in
+      (* Load the value. *)
+      build_load v name builder
+  | Ast.Unary (op, operand) ->
+      let operand = codegen_expr operand in
+      let callee = "unary" ^ (String.make 1 op) in
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown unary operator")
+      in
+      build_call callee [|operand|] "unop" builder
+  | Ast.Binary (op, lhs, rhs) ->
+      begin match op with
+      | '=' ->
+          (* Special case '=' because we don't want to emit the LHS as an
+           * expression. *)
+          let name =
+            match lhs with
+            | Ast.Variable name -> name
+            | _ -> raise (Error "destination of '=' must be a variable")
+          in
+
+          (* Codegen the rhs. *)
+          let val_ = codegen_expr rhs in
+
+          (* Lookup the name. *)
+          let variable = try Hashtbl.find named_values name with
+          | Not_found -> raise (Error "unknown variable name")
+          in
+          ignore(build_store val_ variable builder);
+          val_
+      | _ ->
+          let lhs_val = codegen_expr lhs in
+          let rhs_val = codegen_expr rhs in
+          begin
+            match op with
+            | '+' -> build_add lhs_val rhs_val "addtmp" builder
+            | '-' -> build_sub lhs_val rhs_val "subtmp" builder
+            | '*' -> build_mul lhs_val rhs_val "multmp" builder
+            | '<' ->
+                (* Convert bool 0/1 to double 0.0 or 1.0 *)
+                let i = build_fcmp Fcmp.Ult lhs_val rhs_val "cmptmp" builder in
+                build_uitofp i double_type "booltmp" builder
+            | _ ->
+                (* If it wasn't a builtin binary operator, it must be a user defined
+                 * one. Emit a call to it. *)
+                let callee = "binary" ^ (String.make 1 op) in
+                let callee =
+                  match lookup_function callee the_module with
+                  | Some callee -> callee
+                  | None -> raise (Error "binary operator not found!")
+                in
+                build_call callee [|lhs_val; rhs_val|] "binop" builder
+          end
+      end
+  | Ast.Call (callee, args) ->
+      (* Look up the name in the module table. *)
+      let callee =
+        match lookup_function callee the_module with
+        | Some callee -> callee
+        | None -> raise (Error "unknown function referenced")
+      in
+      let params = params callee in
+
+      (* If argument mismatch error. *)
+      if Array.length params == Array.length args then () else
+        raise (Error "incorrect # arguments passed");
+      let args = Array.map codegen_expr args in
+      build_call callee args "calltmp" builder
+  | Ast.If (cond, then_, else_) ->
+      let cond = codegen_expr cond in
+
+      (* Convert condition to a bool by comparing equal to 0.0 *)
+      let zero = const_float double_type 0.0 in
+      let cond_val = build_fcmp Fcmp.One cond zero "ifcond" builder in
+
+      (* Grab the first block so that we might later add the conditional branch
+       * to it at the end of the function. *)
+      let start_bb = insertion_block builder in
+      let the_function = block_parent start_bb in
+
+      let then_bb = append_block context "then" the_function in
+
+      (* Emit 'then' value. *)
+      position_at_end then_bb builder;
+      let then_val = codegen_expr then_ in
+
+      (* Codegen of 'then' can change the current block, update then_bb for the
+       * phi. We create a new name because one is used for the phi node, and the
+       * other is used for the conditional branch. *)
+      let new_then_bb = insertion_block builder in
+
+      (* Emit 'else' value. *)
+      let else_bb = append_block context "else" the_function in
+      position_at_end else_bb builder;
+      let else_val = codegen_expr else_ in
+
+      (* Codegen of 'else' can change the current block, update else_bb for the
+       * phi. *)
+      let new_else_bb = insertion_block builder in
+
+      (* Emit merge block. *)
+      let merge_bb = append_block context "ifcont" the_function in
+      position_at_end merge_bb builder;
+      let incoming = [(then_val, new_then_bb); (else_val, new_else_bb)] in
+      let phi = build_phi incoming "iftmp" builder in
+
+      (* Return to the start block to add the conditional branch. *)
+      position_at_end start_bb builder;
+      ignore (build_cond_br cond_val then_bb else_bb builder);
+
+      (* Set a unconditional branch at the end of the 'then' block and the
+       * 'else' block to the 'merge' block. *)
+      position_at_end new_then_bb builder; ignore (build_br merge_bb builder);
+      position_at_end new_else_bb builder; ignore (build_br merge_bb builder);
+
+      (* Finally, set the builder to the end of the merge block. *)
+      position_at_end merge_bb builder;
+
+      phi
+  | Ast.For (var_name, start, end_, step, body) ->
+      (* Output this as:
+       *   var = alloca double
+       *   ...
+       *   start = startexpr
+       *   store start -> var
+       *   goto loop
+       * loop:
+       *   ...
+       *   bodyexpr
+       *   ...
+       * loopend:
+       *   step = stepexpr
+       *   endcond = endexpr
+       *
+       *   curvar = load var
+       *   nextvar = curvar + step
+       *   store nextvar -> var
+       *   br endcond, loop, endloop
+       * outloop: *)
+
+      let the_function = block_parent (insertion_block builder) in
+
+      (* Create an alloca for the variable in the entry block. *)
+      let alloca = create_entry_block_alloca the_function var_name in
+
+      (* Emit the start code first, without 'variable' in scope. *)
+      let start_val = codegen_expr start in
+
+      (* Store the value into the alloca. *)
+      ignore(build_store start_val alloca builder);
+
+      (* Make the new basic block for the loop header, inserting after current
+       * block. *)
+      let loop_bb = append_block context "loop" the_function in
+
+      (* Insert an explicit fall through from the current block to the
+       * loop_bb. *)
+      ignore (build_br loop_bb builder);
+
+      (* Start insertion in loop_bb. *)
+      position_at_end loop_bb builder;
+
+      (* Within the loop, the variable is defined equal to the PHI node. If it
+       * shadows an existing variable, we have to restore it, so save it
+       * now. *)
+      let old_val =
+        try Some (Hashtbl.find named_values var_name) with Not_found -> None
+      in
+      Hashtbl.add named_values var_name alloca;
+
+      (* Emit the body of the loop.  This, like any other expr, can change the
+       * current BB.  Note that we ignore the value computed by the body, but
+       * don't allow an error *)
+      ignore (codegen_expr body);
+
+      (* Emit the step value. *)
+      let step_val =
+        match step with
+        | Some step -> codegen_expr step
+        (* If not specified, use 1.0. *)
+        | None -> const_float double_type 1.0
+      in
+
+      (* Compute the end condition. *)
+      let end_cond = codegen_expr end_ in
+
+      (* Reload, increment, and restore the alloca. This handles the case where
+       * the body of the loop mutates the variable. *)
+      let cur_var = build_load alloca var_name builder in
+      let next_var = build_add cur_var step_val "nextvar" builder in
+      ignore(build_store next_var alloca builder);
+
+      (* Convert condition to a bool by comparing equal to 0.0. *)
+      let zero = const_float double_type 0.0 in
+      let end_cond = build_fcmp Fcmp.One end_cond zero "loopcond" builder in
+
+      (* Create the "after loop" block and insert it. *)
+      let after_bb = append_block context "afterloop" the_function in
+
+      (* Insert the conditional branch into the end of loop_end_bb. *)
+      ignore (build_cond_br end_cond loop_bb after_bb builder);
+
+      (* Any new code will be inserted in after_bb. *)
+      position_at_end after_bb builder;
+
+      (* Restore the unshadowed variable. *)
+      begin match old_val with
+      | Some old_val -> Hashtbl.add named_values var_name old_val
+      | None -> ()
+      end;
+
+      (* for expr always returns 0.0. *)
+      const_null double_type
+  | Ast.Var (var_names, body) ->
+      let old_bindings = ref [] in
+
+      let the_function = block_parent (insertion_block builder) in
+
+      (* Register all variables and emit their initializer. *)
+      Array.iter (fun (var_name, init) ->
+        (* Emit the initializer before adding the variable to scope, this
+         * prevents the initializer from referencing the variable itself, and
+         * permits stuff like this:
+         *   var a = 1 in
+         *     var a = a in ...   # refers to outer 'a'. *)
+        let init_val =
+          match init with
+          | Some init -> codegen_expr init
+          (* If not specified, use 0.0. *)
+          | None -> const_float double_type 0.0
+        in
+
+        let alloca = create_entry_block_alloca the_function var_name in
+        ignore(build_store init_val alloca builder);
+
+        (* Remember the old variable binding so that we can restore the binding
+         * when we unrecurse. *)
+        begin
+          try
+            let old_value = Hashtbl.find named_values var_name in
+            old_bindings := (var_name, old_value) :: !old_bindings;
+          with Not_found -> ()
+        end;
+
+        (* Remember this binding. *)
+        Hashtbl.add named_values var_name alloca;
+      ) var_names;
+
+      (* Codegen the body, now that all vars are in scope. *)
+      let body_val = codegen_expr body in
+
+      (* Pop all our variables from scope. *)
+      List.iter (fun (var_name, old_value) ->
+        Hashtbl.add named_values var_name old_value
+      ) !old_bindings;
+
+      (* Return the body computation. *)
+      body_val
+
+let codegen_proto = function
+  | Ast.Prototype (name, args) | Ast.BinOpPrototype (name, args, _) ->
+      (* Make the function type: double(double,double) etc. *)
+      let doubles = Array.make (Array.length args) double_type in
+      let ft = function_type double_type doubles in
+      let f =
+        match lookup_function name the_module with
+        | None -> declare_function name ft the_module
+
+        (* If 'f' conflicted, there was already something named 'name'. If it
+         * has a body, don't allow redefinition or reextern. *)
+        | Some f ->
+            (* If 'f' already has a body, reject this. *)
+            if block_begin f <> At_end f then
+              raise (Error "redefinition of function");
+
+            (* If 'f' took a different number of arguments, reject. *)
+            if element_type (type_of f) <> ft then
+              raise (Error "redefinition of function with different # args");
+            f
+      in
+
+      (* Set names for all arguments. *)
+      Array.iteri (fun i a ->
+        let n = args.(i) in
+        set_value_name n a;
+        Hashtbl.add named_values n a;
+      ) (params f);
+      f
+
+(* Create an alloca for each argument and register the argument in the symbol
+ * table so that references to it will succeed. *)
+let create_argument_allocas the_function proto =
+  let args = match proto with
+    | Ast.Prototype (_, args) | Ast.BinOpPrototype (_, args, _) -> args
+  in
+  Array.iteri (fun i ai ->
+    let var_name = args.(i) in
+    (* Create an alloca for this variable. *)
+    let alloca = create_entry_block_alloca the_function var_name in
+
+    (* Store the initial value into the alloca. *)
+    ignore(build_store ai alloca builder);
+
+    (* Add arguments to variable symbol table. *)
+    Hashtbl.add named_values var_name alloca;
+  ) (params the_function)
+
+let codegen_func the_fpm = function
+  | Ast.Function (proto, body) ->
+      Hashtbl.clear named_values;
+      let the_function = codegen_proto proto in
+
+      (* If this is an operator, install it. *)
+      begin match proto with
+      | Ast.BinOpPrototype (name, args, prec) ->
+          let op = name.[String.length name - 1] in
+          Hashtbl.add Parser.binop_precedence op prec;
+      | _ -> ()
+      end;
+
+      (* Create a new basic block to start insertion into. *)
+      let bb = append_block context "entry" the_function in
+      position_at_end bb builder;
+
+      try
+        (* Add all arguments to the symbol table and create their allocas. *)
+        create_argument_allocas the_function proto;
+
+        let ret_val = codegen_expr body in
+
+        (* Finish off the function. *)
+        let _ = build_ret ret_val builder in
+
+        (* Validate the generated code, checking for consistency. *)
+        Llvm_analysis.assert_valid_function the_function;
+
+        (* Optimize the function. *)
+        let _ = PassManager.run_function the_function the_fpm in
+
+        the_function
+      with e ->
+        delete_function the_function;
+        raise e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/lexer.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/lexer.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/lexer.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/lexer.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,60 @@
+(*===----------------------------------------------------------------------===
+ * Lexer
+ *===----------------------------------------------------------------------===*)
+
+let rec lex = parser
+  (* Skip any whitespace. *)
+  | [< ' (' ' | '\n' | '\r' | '\t'); stream >] -> lex stream
+
+  (* identifier: [a-zA-Z][a-zA-Z0-9] *)
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+
+  (* number: [0-9.]+ *)
+  | [< ' ('0' .. '9' as c); stream >] ->
+      let buffer = Buffer.create 1 in
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+
+  (* Comment until end of line. *)
+  | [< ' ('#'); stream >] ->
+      lex_comment stream
+
+  (* Otherwise, just return the character as its ascii value. *)
+  | [< 'c; stream >] ->
+      [< 'Token.Kwd c; lex stream >]
+
+  (* end of stream. *)
+  | [< >] -> [< >]
+
+and lex_number buffer = parser
+  | [< ' ('0' .. '9' | '.' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_number buffer stream
+  | [< stream=lex >] ->
+      [< 'Token.Number (float_of_string (Buffer.contents buffer)); stream >]
+
+and lex_ident buffer = parser
+  | [< ' ('A' .. 'Z' | 'a' .. 'z' | '0' .. '9' as c); stream >] ->
+      Buffer.add_char buffer c;
+      lex_ident buffer stream
+  | [< stream=lex >] ->
+      match Buffer.contents buffer with
+      | "def" -> [< 'Token.Def; stream >]
+      | "extern" -> [< 'Token.Extern; stream >]
+      | "if" -> [< 'Token.If; stream >]
+      | "then" -> [< 'Token.Then; stream >]
+      | "else" -> [< 'Token.Else; stream >]
+      | "for" -> [< 'Token.For; stream >]
+      | "in" -> [< 'Token.In; stream >]
+      | "binary" -> [< 'Token.Binary; stream >]
+      | "unary" -> [< 'Token.Unary; stream >]
+      | "var" -> [< 'Token.Var; stream >]
+      | id -> [< 'Token.Ident id; stream >]
+
+and lex_comment = parser
+  | [< ' ('\n'); stream=lex >] -> stream
+  | [< 'c; e=lex_comment >] -> e
+  | [< >] -> [< >]

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/myocamlbuild.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/myocamlbuild.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/myocamlbuild.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/myocamlbuild.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,10 @@
+open Ocamlbuild_plugin;;
+
+ocaml_lib ~extern:true "llvm";;
+ocaml_lib ~extern:true "llvm_analysis";;
+ocaml_lib ~extern:true "llvm_executionengine";;
+ocaml_lib ~extern:true "llvm_target";;
+ocaml_lib ~extern:true "llvm_scalar_opts";;
+
+flag ["link"; "ocaml"; "g++"] (S[A"-cc"; A"g++"]);;
+dep ["link"; "ocaml"; "use_bindings"] ["bindings.o"];;

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/parser.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/parser.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/parser.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/parser.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,221 @@
+(*===---------------------------------------------------------------------===
+ * Parser
+ *===---------------------------------------------------------------------===*)
+
+(* binop_precedence - This holds the precedence for each binary operator that is
+ * defined *)
+let binop_precedence:(char, int) Hashtbl.t = Hashtbl.create 10
+
+(* precedence - Get the precedence of the pending binary operator token. *)
+let precedence c = try Hashtbl.find binop_precedence c with Not_found -> -1
+
+(* primary
+ *   ::= identifier
+ *   ::= numberexpr
+ *   ::= parenexpr
+ *   ::= ifexpr
+ *   ::= forexpr
+ *   ::= varexpr *)
+let rec parse_primary = parser
+  (* numberexpr ::= number *)
+  | [< 'Token.Number n >] -> Ast.Number n
+
+  (* parenexpr ::= '(' expression ')' *)
+  | [< 'Token.Kwd '('; e=parse_expr; 'Token.Kwd ')' ?? "expected ')'" >] -> e
+
+  (* identifierexpr
+   *   ::= identifier
+   *   ::= identifier '(' argumentexpr ')' *)
+  | [< 'Token.Ident id; stream >] ->
+      let rec parse_args accumulator = parser
+        | [< e=parse_expr; stream >] ->
+            begin parser
+              | [< 'Token.Kwd ','; e=parse_args (e :: accumulator) >] -> e
+              | [< >] -> e :: accumulator
+            end stream
+        | [< >] -> accumulator
+      in
+      let rec parse_ident id = parser
+        (* Call. *)
+        | [< 'Token.Kwd '(';
+             args=parse_args [];
+             'Token.Kwd ')' ?? "expected ')'">] ->
+            Ast.Call (id, Array.of_list (List.rev args))
+
+        (* Simple variable ref. *)
+        | [< >] -> Ast.Variable id
+      in
+      parse_ident id stream
+
+  (* ifexpr ::= 'if' expr 'then' expr 'else' expr *)
+  | [< 'Token.If; c=parse_expr;
+       'Token.Then ?? "expected 'then'"; t=parse_expr;
+       'Token.Else ?? "expected 'else'"; e=parse_expr >] ->
+      Ast.If (c, t, e)
+
+  (* forexpr
+        ::= 'for' identifier '=' expr ',' expr (',' expr)? 'in' expression *)
+  | [< 'Token.For;
+       'Token.Ident id ?? "expected identifier after for";
+       'Token.Kwd '=' ?? "expected '=' after for";
+       stream >] ->
+      begin parser
+        | [<
+             start=parse_expr;
+             'Token.Kwd ',' ?? "expected ',' after for";
+             end_=parse_expr;
+             stream >] ->
+            let step =
+              begin parser
+              | [< 'Token.Kwd ','; step=parse_expr >] -> Some step
+              | [< >] -> None
+              end stream
+            in
+            begin parser
+            | [< 'Token.In; body=parse_expr >] ->
+                Ast.For (id, start, end_, step, body)
+            | [< >] ->
+                raise (Stream.Error "expected 'in' after for")
+            end stream
+        | [< >] ->
+            raise (Stream.Error "expected '=' after for")
+      end stream
+
+  (* varexpr
+   *   ::= 'var' identifier ('=' expression?
+   *             (',' identifier ('=' expression)?)* 'in' expression *)
+  | [< 'Token.Var;
+       (* At least one variable name is required. *)
+       'Token.Ident id ?? "expected identifier after var";
+       init=parse_var_init;
+       var_names=parse_var_names [(id, init)];
+       (* At this point, we have to have 'in'. *)
+       'Token.In ?? "expected 'in' keyword after 'var'";
+       body=parse_expr >] ->
+      Ast.Var (Array.of_list (List.rev var_names), body)
+
+  | [< >] -> raise (Stream.Error "unknown token when expecting an expression.")
+
+(* unary
+ *   ::= primary
+ *   ::= '!' unary *)
+and parse_unary = parser
+  (* If this is a unary operator, read it. *)
+  | [< 'Token.Kwd op when op != '(' && op != ')'; operand=parse_expr >] ->
+      Ast.Unary (op, operand)
+
+  (* If the current token is not an operator, it must be a primary expr. *)
+  | [< stream >] -> parse_primary stream
+
+(* binoprhs
+ *   ::= ('+' primary)* *)
+and parse_bin_rhs expr_prec lhs stream =
+  match Stream.peek stream with
+  (* If this is a binop, find its precedence. *)
+  | Some (Token.Kwd c) when Hashtbl.mem binop_precedence c ->
+      let token_prec = precedence c in
+
+      (* If this is a binop that binds at least as tightly as the current binop,
+       * consume it, otherwise we are done. *)
+      if token_prec < expr_prec then lhs else begin
+        (* Eat the binop. *)
+        Stream.junk stream;
+
+        (* Parse the primary expression after the binary operator. *)
+        let rhs = parse_unary stream in
+
+        (* Okay, we know this is a binop. *)
+        let rhs =
+          match Stream.peek stream with
+          | Some (Token.Kwd c2) ->
+              (* If BinOp binds less tightly with rhs than the operator after
+               * rhs, let the pending operator take rhs as its lhs. *)
+              let next_prec = precedence c2 in
+              if token_prec < next_prec
+              then parse_bin_rhs (token_prec + 1) rhs stream
+              else rhs
+          | _ -> rhs
+        in
+
+        (* Merge lhs/rhs. *)
+        let lhs = Ast.Binary (c, lhs, rhs) in
+        parse_bin_rhs expr_prec lhs stream
+      end
+  | _ -> lhs
+
+and parse_var_init = parser
+  (* read in the optional initializer. *)
+  | [< 'Token.Kwd '='; e=parse_expr >] -> Some e
+  | [< >] -> None
+
+and parse_var_names accumulator = parser
+  | [< 'Token.Kwd ',';
+       'Token.Ident id ?? "expected identifier list after var";
+       init=parse_var_init;
+       e=parse_var_names ((id, init) :: accumulator) >] -> e
+  | [< >] -> accumulator
+
+(* expression
+ *   ::= primary binoprhs *)
+and parse_expr = parser
+  | [< lhs=parse_unary; stream >] -> parse_bin_rhs 0 lhs stream
+
+(* prototype
+ *   ::= id '(' id* ')'
+ *   ::= binary LETTER number? (id, id)
+ *   ::= unary LETTER number? (id) *)
+let parse_prototype =
+  let rec parse_args accumulator = parser
+    | [< 'Token.Ident id; e=parse_args (id::accumulator) >] -> e
+    | [< >] -> accumulator
+  in
+  let parse_operator = parser
+    | [< 'Token.Unary >] -> "unary", 1
+    | [< 'Token.Binary >] -> "binary", 2
+  in
+  let parse_binary_precedence = parser
+    | [< 'Token.Number n >] -> int_of_float n
+    | [< >] -> 30
+  in
+  parser
+  | [< 'Token.Ident id;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+       args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      (* success. *)
+      Ast.Prototype (id, Array.of_list (List.rev args))
+  | [< (prefix, kind)=parse_operator;
+       'Token.Kwd op ?? "expected an operator";
+       (* Read the precedence if present. *)
+       binary_precedence=parse_binary_precedence;
+       'Token.Kwd '(' ?? "expected '(' in prototype";
+        args=parse_args [];
+       'Token.Kwd ')' ?? "expected ')' in prototype" >] ->
+      let name = prefix ^ (String.make 1 op) in
+      let args = Array.of_list (List.rev args) in
+
+      (* Verify right number of arguments for operator. *)
+      if Array.length args != kind
+      then raise (Stream.Error "invalid number of operands for operator")
+      else
+        if kind == 1 then
+          Ast.Prototype (name, args)
+        else
+          Ast.BinOpPrototype (name, args, binary_precedence)
+  | [< >] ->
+      raise (Stream.Error "expected function name in prototype")
+
+(* definition ::= 'def' prototype expression *)
+let parse_definition = parser
+  | [< 'Token.Def; p=parse_prototype; e=parse_expr >] ->
+      Ast.Function (p, e)
+
+(* toplevelexpr ::= expression *)
+let parse_toplevel = parser
+  | [< e=parse_expr >] ->
+      (* Make an anonymous proto. *)
+      Ast.Function (Ast.Prototype ("", [||]), e)
+
+(*  external ::= 'extern' prototype *)
+let parse_extern = parser
+  | [< 'Token.Extern; e=parse_prototype >] -> e

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/token.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/token.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/token.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/token.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,25 @@
+(*===----------------------------------------------------------------------===
+ * Lexer Tokens
+ *===----------------------------------------------------------------------===*)
+
+(* The lexer returns these 'Kwd' if it is an unknown character, otherwise one of
+ * these others for known things. *)
+type token =
+  (* commands *)
+  | Def | Extern
+
+  (* primary *)
+  | Ident of string | Number of float
+
+  (* unknown *)
+  | Kwd of char
+
+  (* control *)
+  | If | Then | Else
+  | For | In
+
+  (* operators *)
+  | Binary | Unary
+
+  (* var definition *)
+  | Var

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toplevel.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toplevel.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toplevel.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toplevel.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,49 @@
+(*===----------------------------------------------------------------------===
+ * Top-Level parsing and JIT Driver
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+
+(* top ::= definition | external | expression | ';' *)
+let rec main_loop the_fpm the_execution_engine stream =
+  match Stream.peek stream with
+  | None -> ()
+
+  (* ignore top-level semicolons. *)
+  | Some (Token.Kwd ';') ->
+      Stream.junk stream;
+      main_loop the_fpm the_execution_engine stream
+
+  | Some token ->
+      begin
+        try match token with
+        | Token.Def ->
+            let e = Parser.parse_definition stream in
+            print_endline "parsed a function definition.";
+            dump_value (Codegen.codegen_func the_fpm e);
+        | Token.Extern ->
+            let e = Parser.parse_extern stream in
+            print_endline "parsed an extern.";
+            dump_value (Codegen.codegen_proto e);
+        | _ ->
+            (* Evaluate a top-level expression into an anonymous function. *)
+            let e = Parser.parse_toplevel stream in
+            print_endline "parsed a top-level expr";
+            let the_function = Codegen.codegen_func the_fpm e in
+            dump_value the_function;
+
+            (* JIT the function, returning a function pointer. *)
+            let result = ExecutionEngine.run_function the_function [||]
+              the_execution_engine in
+
+            print_string "Evaluated to ";
+            print_float (GenericValue.as_float Codegen.double_type result);
+            print_newline ();
+        with Stream.Error s | Codegen.Error s ->
+          (* Skip token for error recovery. *)
+          Stream.junk stream;
+          print_endline s;
+      end;
+      print_string "ready> "; flush stdout;
+      main_loop the_fpm the_execution_engine stream

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toy.ml
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toy.ml?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toy.ml (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Chapter7/toy.ml Mon Mar  8 13:32:27 2010
@@ -0,0 +1,57 @@
+(*===----------------------------------------------------------------------===
+ * Main driver code.
+ *===----------------------------------------------------------------------===*)
+
+open Llvm
+open Llvm_executionengine
+open Llvm_target
+open Llvm_scalar_opts
+
+let main () =
+  ignore (initialize_native_target ());
+
+  (* Install standard binary operators.
+   * 1 is the lowest precedence. *)
+  Hashtbl.add Parser.binop_precedence '=' 2;
+  Hashtbl.add Parser.binop_precedence '<' 10;
+  Hashtbl.add Parser.binop_precedence '+' 20;
+  Hashtbl.add Parser.binop_precedence '-' 20;
+  Hashtbl.add Parser.binop_precedence '*' 40;    (* highest. *)
+
+  (* Prime the first token. *)
+  print_string "ready> "; flush stdout;
+  let stream = Lexer.lex (Stream.of_channel stdin) in
+
+  (* Create the JIT. *)
+  let the_execution_engine = ExecutionEngine.create Codegen.the_module in
+  let the_fpm = PassManager.create_function Codegen.the_module in
+
+  (* Set up the optimizer pipeline.  Start with registering info about how the
+   * target lays out data structures. *)
+  TargetData.add (ExecutionEngine.target_data the_execution_engine) the_fpm;
+
+  (* Promote allocas to registers. *)
+  add_memory_to_register_promotion the_fpm;
+
+  (* Do simple "peephole" optimizations and bit-twiddling optzn. *)
+  add_instruction_combination the_fpm;
+
+  (* reassociate expressions. *)
+  add_reassociation the_fpm;
+
+  (* Eliminate Common SubExpressions. *)
+  add_gvn the_fpm;
+
+  (* Simplify the control flow graph (deleting unreachable blocks, etc). *)
+  add_cfg_simplification the_fpm;
+
+  ignore (PassManager.initialize the_fpm);
+
+  (* Run the main "interpreter loop" now. *)
+  Toplevel.main_loop the_fpm the_execution_engine stream;
+
+  (* Print out all the generated code. *)
+  dump_module Codegen.the_module
+;;
+
+main ()

Added: llvm/trunk/examples/OCaml-Kaleidoscope/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/OCaml-Kaleidoscope/Makefile?rev=97966&view=auto
==============================================================================
--- llvm/trunk/examples/OCaml-Kaleidoscope/Makefile (added)
+++ llvm/trunk/examples/OCaml-Kaleidoscope/Makefile Mon Mar  8 13:32:27 2010
@@ -0,0 +1,15 @@
+##===- examples/OCaml-Kaleidoscope/Makefile ----------------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+LEVEL=../..
+
+include $(LEVEL)/Makefile.config
+
+PARALLEL_DIRS:= Chapter2 Chapter3 Chapter4 Chapter5 Chapter6 Chapter7
+
+include $(LEVEL)/Makefile.common




From rjmccall at apple.com  Mon Mar  8 14:02:05 2010
From: rjmccall at apple.com (John McCall)
Date: Mon, 08 Mar 2010 20:02:05 -0000
Subject: [llvm-commits] [llvm] r97972 - in /llvm/trunk: test/lit.cfg
 utils/lit/lit/TestRunner.py utils/lit/lit/TestingConfig.py
Message-ID: <20100308200205.8302F2A6C12C@llvm.org>

Author: rjmccall
Date: Mon Mar  8 14:02:05 2010
New Revision: 97972

URL: http://llvm.org/viewvc/llvm-project?rev=97972&view=rev
Log:
Revert r97726 and r97728 at ddunbar's request; we want to solve this
some other way when it comes to be necessary.


Modified:
    llvm/trunk/test/lit.cfg
    llvm/trunk/utils/lit/lit/TestRunner.py
    llvm/trunk/utils/lit/lit/TestingConfig.py

Modified: llvm/trunk/test/lit.cfg
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/lit.cfg?rev=97972&r1=97971&r2=97972&view=diff
==============================================================================
--- llvm/trunk/test/lit.cfg (original)
+++ llvm/trunk/test/lit.cfg Mon Mar  8 14:02:05 2010
@@ -144,9 +144,6 @@
 def llvm_supports_binding(name):
     return name in bindings
 
-config.conditions["TARGET"] = llvm_supports_target
-config.conditions["BINDING"] = llvm_supports_binding
-
 # Provide on_clone hook for reading 'dg.exp'.
 import os
 simpleLibData = re.compile(r"""load_lib llvm.exp

Modified: llvm/trunk/utils/lit/lit/TestRunner.py
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/lit/lit/TestRunner.py?rev=97972&r1=97971&r2=97972&view=diff
==============================================================================
--- llvm/trunk/utils/lit/lit/TestRunner.py (original)
+++ llvm/trunk/utils/lit/lit/TestRunner.py Mon Mar  8 14:02:05 2010
@@ -353,8 +353,6 @@
 
     return True
 
-import re
-
 def parseIntegratedTestScript(test):
     """parseIntegratedTestScript - Scan an LLVM/Clang style integrated test
     script and extract the lines to 'RUN' as well as 'XFAIL' and 'XTARGET'
@@ -387,21 +385,7 @@
     script = []
     xfails = []
     xtargets = []
-    ignoredAny = False
     for ln in open(sourcepath):
-        conditional = re.search('IF\((.+?)\((.+?)\)\):', ln)
-        if conditional:
-            ln = ln[conditional.end():]
-            condition = conditional.group(1)
-            value = conditional.group(2)
-
-            # Actually test the condition.
-            if condition not in test.config.conditions:
-                return (Test.UNRESOLVED, "unknown condition '"+condition+"'")
-            if not test.config.conditions[condition](value):
-                ignoredAny = True
-                continue
-
         if 'RUN:' in ln:
             # Isolate the command to run.
             index = ln.index('RUN:')
@@ -438,8 +422,6 @@
 
     # Verify the script contains a run line.
     if not script:
-        if ignoredAny:
-            return (Test.UNSUPPORTED, "Test has only ignored run lines")
         return (Test.UNRESOLVED, "Test has no run line!")
 
     if script[-1][-1] == '\\':

Modified: llvm/trunk/utils/lit/lit/TestingConfig.py
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/lit/lit/TestingConfig.py?rev=97972&r1=97971&r2=97972&view=diff
==============================================================================
--- llvm/trunk/utils/lit/lit/TestingConfig.py (original)
+++ llvm/trunk/utils/lit/lit/TestingConfig.py Mon Mar  8 14:02:05 2010
@@ -28,8 +28,7 @@
                                    on_clone = None,
                                    test_exec_root = None,
                                    test_source_root = None,
-                                   excludes = [],
-                                   conditions = {})
+                                   excludes = [])
 
         if os.path.exists(path):
             # FIXME: Improve detection and error reporting of errors in the
@@ -55,7 +54,7 @@
 
     def __init__(self, parent, name, suffixes, test_format,
                  environment, substitutions, unsupported, on_clone,
-                 test_exec_root, test_source_root, excludes, conditions):
+                 test_exec_root, test_source_root, excludes):
         self.parent = parent
         self.name = str(name)
         self.suffixes = set(suffixes)
@@ -67,7 +66,6 @@
         self.test_exec_root = test_exec_root
         self.test_source_root = test_source_root
         self.excludes = set(excludes)
-        self.conditions = dict(conditions)
 
     def clone(self, path):
         # FIXME: Chain implementations?
@@ -77,7 +75,7 @@
                             self.environment, self.substitutions,
                             self.unsupported, self.on_clone,
                             self.test_exec_root, self.test_source_root,
-                            self.excludes, self.conditions)
+                            self.excludes)
         if cfg.on_clone:
             cfg.on_clone(self, cfg, path)
         return cfg




From idadesub at users.sourceforge.net  Mon Mar  8 14:07:32 2010
From: idadesub at users.sourceforge.net (Erick Tryzelaar)
Date: Mon, 08 Mar 2010 20:07:32 -0000
Subject: [llvm-commits] [llvm] r97973 - /llvm/trunk/examples/Makefile
Message-ID: <20100308200732.7CE582A6C12C@llvm.org>

Author: erickt
Date: Mon Mar  8 14:07:32 2010
New Revision: 97973

URL: http://llvm.org/viewvc/llvm-project?rev=97973&view=rev
Log:
Don't always run the ocaml kaleidoscope tutorials.

Modified:
    llvm/trunk/examples/Makefile

Modified: llvm/trunk/examples/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/examples/Makefile?rev=97973&r1=97972&r2=97973&view=diff
==============================================================================
--- llvm/trunk/examples/Makefile (original)
+++ llvm/trunk/examples/Makefile Mon Mar  8 14:07:32 2010
@@ -10,8 +10,7 @@
 
 include $(LEVEL)/Makefile.config
 
-PARALLEL_DIRS:= BrainF Fibonacci HowToUseJIT Kaleidoscope ModuleMaker \
-	OCaml-Kaleidoscope
+PARALLEL_DIRS:= BrainF Fibonacci HowToUseJIT Kaleidoscope ModuleMaker
 
 ifeq ($(HAVE_PTHREAD),1)
 PARALLEL_DIRS += ParallelJIT 
@@ -26,4 +25,8 @@
     endif
 endif
 
+ifeq ($(filter $(BINDINGS_TO_BUILD),ocaml),ocaml)
+	PARALLEL_DIRS += OCaml-Kaleidoscope
+endif
+
 include $(LEVEL)/Makefile.common




From alenhar2 at llvm.org  Mon Mar  8 14:45:53 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Mon, 08 Mar 2010 20:45:53 -0000
Subject: [llvm-commits] [llvm] r97974 - in /llvm/trunk/include/llvm/ADT:
	DenseMap.h DenseSet.h
Message-ID: <20100308204553.1B5982A6C12C@llvm.org>

Author: alenhar2
Date: Mon Mar  8 14:45:52 2010
New Revision: 97974

URL: http://llvm.org/viewvc/llvm-project?rev=97974&view=rev
Log:
Iterator traits and swap.  closes PR6548 and PR6549

Modified:
    llvm/trunk/include/llvm/ADT/DenseMap.h
    llvm/trunk/include/llvm/ADT/DenseSet.h

Modified: llvm/trunk/include/llvm/ADT/DenseMap.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/DenseMap.h?rev=97974&r1=97973&r2=97974&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/DenseMap.h (original)
+++ llvm/trunk/include/llvm/ADT/DenseMap.h Mon Mar  8 14:45:52 2010
@@ -192,6 +192,13 @@
     return true;
   }
 
+  void swap(DenseMap& RHS) {
+    std::swap(NumBuckets, RHS.NumBuckets);
+    std::swap(Buckets, RHS.Buckets);
+    std::swap(NumEntries, RHS.NumEntries);
+    std::swap(NumTombstones, RHS.NumTombstones);
+  }
+
   value_type& FindAndConstruct(const KeyT &Key) {
     BucketT *TheBucket;
     if (LookupBucketFor(Key, TheBucket))

Modified: llvm/trunk/include/llvm/ADT/DenseSet.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/DenseSet.h?rev=97974&r1=97973&r2=97974&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/DenseSet.h (original)
+++ llvm/trunk/include/llvm/ADT/DenseSet.h Mon Mar  8 14:45:52 2010
@@ -45,6 +45,10 @@
     return TheMap.erase(V);
   }
 
+  void swap(DenseSet& RHS) {
+    TheMap.swap(RHS.TheMap);
+  }
+
   DenseSet &operator=(const DenseSet &RHS) {
     TheMap = RHS.TheMap;
     return *this;
@@ -55,6 +59,12 @@
   class Iterator {
     typename MapTy::iterator I;
   public:
+    typedef typename MapTy::iterator::difference_type difference_type;
+    typedef ValueT value_type;
+    typedef value_type *pointer;
+    typedef value_type &reference;
+    typedef std::forward_iterator_tag iterator_category;
+
     Iterator(const typename MapTy::iterator &i) : I(i) {}
 
     ValueT& operator*() { return I->first; }
@@ -68,6 +78,12 @@
   class ConstIterator {
     typename MapTy::const_iterator I;
   public:
+    typedef typename MapTy::const_iterator::difference_type difference_type;
+    typedef ValueT value_type;
+    typedef value_type *pointer;
+    typedef value_type &reference;
+    typedef std::forward_iterator_tag iterator_category;
+
     ConstIterator(const typename MapTy::const_iterator &i) : I(i) {}
 
     const ValueT& operator*() { return I->first; }




From dpatel at apple.com  Mon Mar  8 14:52:55 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 20:52:55 -0000
Subject: [llvm-commits] [llvm] r97975 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/AsmPrinter/DwarfDebug.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.h lib/CodeGen/MachineInstr.cpp
 lib/Target/PIC16/PIC16DebugInfo.cpp lib/Transforms/Utils/CloneFunction.cpp
Message-ID: <20100308205255.469472A6C12D@llvm.org>

Author: dpatel
Date: Mon Mar  8 14:52:55 2010
New Revision: 97975

URL: http://llvm.org/viewvc/llvm-project?rev=97975&view=rev
Log:
Avoid using DIDescriptor.isNull(). 
This is a first step towards eliminating checks in Descriptor constructors.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/MachineInstr.cpp
    llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
    llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 14:52:55 2010
@@ -68,6 +68,7 @@
     explicit DIDescriptor(MDNode *N) : DbgNode(N) {}
 
     bool isNull() const { return DbgNode == 0; }
+    bool Verify() const { return DbgNode != 0; }
 
     MDNode *getNode() const { return DbgNode; }
 
@@ -246,7 +247,9 @@
     bool isArtificial() const {
       return (getFlags() & FlagArtificial) != 0;
     }
-
+    bool isValid() const {
+      return DbgNode && (isBasicType() || isDerivedType() || isCompositeType());
+    }
     /// dump - print type.
     void dump() const;
   };
@@ -360,7 +363,7 @@
     /// DIType or as DICompositeType.
     StringRef getReturnTypeName() const {
       DICompositeType DCT(getFieldAs<DICompositeType>(8));
-      if (!DCT.isNull()) {
+      if (DCT.Verify()) {
         DIArray A = DCT.getTypeArray();
         DIType T(A.getElement(0).getNode());
         return T.getName();
@@ -494,6 +497,7 @@
     DILocation getOrigLocation() const { return getFieldAs<DILocation>(3); }
     StringRef getFilename() const    { return getScope().getFilename(); }
     StringRef getDirectory() const   { return getScope().getDirectory(); }
+    bool Verify() const;
   };
 
   /// DIFactory - This object assists with the construction of the various

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 14:52:55 2010
@@ -132,13 +132,12 @@
 /// isBasicType - Return true if the specified tag is legal for
 /// DIBasicType.
 bool DIDescriptor::isBasicType() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_base_type;
+  return DbgNode && getTag() == dwarf::DW_TAG_base_type;
 }
 
 /// isDerivedType - Return true if the specified tag is legal for DIDerivedType.
 bool DIDescriptor::isDerivedType() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_typedef:
   case dwarf::DW_TAG_pointer_type:
@@ -158,7 +157,7 @@
 /// isCompositeType - Return true if the specified tag is legal for
 /// DICompositeType.
 bool DIDescriptor::isCompositeType() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_array_type:
   case dwarf::DW_TAG_structure_type:
@@ -175,7 +174,7 @@
 
 /// isVariable - Return true if the specified tag is legal for DIVariable.
 bool DIDescriptor::isVariable() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_auto_variable:
   case dwarf::DW_TAG_arg_variable:
@@ -194,15 +193,13 @@
 /// isSubprogram - Return true if the specified tag is legal for
 /// DISubprogram.
 bool DIDescriptor::isSubprogram() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_subprogram;
+  return DbgNode && getTag() == dwarf::DW_TAG_subprogram;
 }
 
 /// isGlobalVariable - Return true if the specified tag is legal for
 /// DIGlobalVariable.
 bool DIDescriptor::isGlobalVariable() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_variable;
+  return DbgNode && getTag() == dwarf::DW_TAG_variable;
 }
 
 /// isGlobal - Return true if the specified tag is legal for DIGlobal.
@@ -213,7 +210,7 @@
 /// isScope - Return true if the specified tag is one of the scope
 /// related tag.
 bool DIDescriptor::isScope() const {
-  assert(!isNull() && "Invalid descriptor!");
+  if (!DbgNode) return false;
   switch (getTag()) {
   case dwarf::DW_TAG_compile_unit:
   case dwarf::DW_TAG_lexical_block:
@@ -228,32 +225,27 @@
 
 /// isCompileUnit - Return true if the specified tag is DW_TAG_compile_unit.
 bool DIDescriptor::isCompileUnit() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_compile_unit;
+  return DbgNode && getTag() == dwarf::DW_TAG_compile_unit;
 }
 
 /// isNameSpace - Return true if the specified tag is DW_TAG_namespace.
 bool DIDescriptor::isNameSpace() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_namespace;
+  return DbgNode && getTag() == dwarf::DW_TAG_namespace;
 }
 
 /// isLexicalBlock - Return true if the specified tag is DW_TAG_lexical_block.
 bool DIDescriptor::isLexicalBlock() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_lexical_block;
+  return DbgNode && getTag() == dwarf::DW_TAG_lexical_block;
 }
 
 /// isSubrange - Return true if the specified tag is DW_TAG_subrange_type.
 bool DIDescriptor::isSubrange() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_subrange_type;
+  return DbgNode && getTag() == dwarf::DW_TAG_subrange_type;
 }
 
 /// isEnumerator - Return true if the specified tag is DW_TAG_enumerator.
 bool DIDescriptor::isEnumerator() const {
-  assert(!isNull() && "Invalid descriptor!");
-  return getTag() == dwarf::DW_TAG_enumerator;
+  return DbgNode && getTag() == dwarf::DW_TAG_enumerator;
 }
 
 //===----------------------------------------------------------------------===//
@@ -268,7 +260,8 @@
 }
 
 unsigned DIArray::getNumElements() const {
-  assert(DbgNode && "Invalid DIArray");
+  if (!DbgNode)
+    return 0;
   return DbgNode->getNumOperands();
 }
 
@@ -276,11 +269,9 @@
 /// this descriptor. After this completes, the current debug info value
 /// is erased.
 void DIDerivedType::replaceAllUsesWith(DIDescriptor &D) {
-  if (isNull())
+  if (!DbgNode)
     return;
 
-  assert(!D.isNull() && "Can not replace with null");
-
   // Since we use a TrackingVH for the node, its easy for clients to manufacture
   // legitimate situations where they want to replaceAllUsesWith() on something
   // which, due to uniquing, has merged with the source. We shield clients from
@@ -295,7 +286,7 @@
 
 /// Verify - Verify that a compile unit is well formed.
 bool DICompileUnit::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
   StringRef N = getFilename();
   if (N.empty())
@@ -306,36 +297,36 @@
 
 /// Verify - Verify that a type descriptor is well formed.
 bool DIType::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a composite type descriptor is well formed.
 bool DICompositeType::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a subprogram descriptor is well formed.
 bool DISubprogram::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
@@ -343,24 +334,24 @@
     return false;
 
   DICompositeType Ty = getType();
-  if (!Ty.isNull() && !Ty.Verify())
+  if (!Ty.Verify())
     return false;
   return true;
 }
 
 /// Verify - Verify that a global variable descriptor is well formed.
 bool DIGlobalVariable::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
   if (getDisplayName().empty())
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DICompileUnit CU = getCompileUnit();
-  if (!CU.isNull() && !CU.Verify())
+  if (!CU.Verify())
     return false;
 
   DIType Ty = getType();
@@ -375,10 +366,10 @@
 
 /// Verify - Verify that a variable descriptor is well formed.
 bool DIVariable::Verify() const {
-  if (isNull())
+  if (!DbgNode)
     return false;
 
-  if (getContext().isNull())
+  if (!getContext().Verify())
     return false;
 
   DIType Ty = getType();
@@ -388,6 +379,14 @@
   return true;
 }
 
+/// Verify - Verify that a location descriptor is well formed.
+bool DILocation::Verify() const {
+  if (!DbgNode)
+    return false;
+  
+  return DbgNode->getNumOperands() == 4;
+}
+
 /// getOriginalTypeSize - If this type is derived from a base type then
 /// return base type size.
 uint64_t DIDerivedType::getOriginalTypeSize() const {
@@ -398,7 +397,7 @@
     DIType BaseType = getTypeDerivedFrom();
     // If this type is not derived from any type then take conservative 
     // approach.
-    if (BaseType.isNull())
+    if (!BaseType.isValid())
       return getSizeInBits();
     if (BaseType.isDerivedType())
       return DIDerivedType(BaseType.getNode()).getOriginalTypeSize();
@@ -468,7 +467,7 @@
 
 /// dump - Print type.
 void DIType::dump() const {
-  if (isNull()) return;
+  if (!DbgNode) return;
 
   StringRef Res = getName();
   if (!Res.empty())
@@ -521,8 +520,6 @@
 /// dump - Print composite type.
 void DICompositeType::dump() const {
   DIArray A = getTypeArray();
-  if (A.isNull())
-    return;
   dbgs() << " [" << A.getNumElements() << " elements]";
 }
 
@@ -1155,9 +1152,8 @@
 
 /// processLocation - Process DILocation.
 void DebugInfoFinder::processLocation(DILocation Loc) {
-  if (Loc.isNull()) return;
-  DIScope S(Loc.getScope().getNode());
-  if (S.isNull()) return;
+  if (!Loc.Verify()) return;
+  DIDescriptor S(Loc.getScope().getNode());
   if (S.isCompileUnit())
     addCompileUnit(DICompileUnit(S.getNode()));
   else if (S.isSubprogram())
@@ -1177,26 +1173,21 @@
     DICompositeType DCT(DT.getNode());
     processType(DCT.getTypeDerivedFrom());
     DIArray DA = DCT.getTypeArray();
-    if (!DA.isNull())
-      for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
-        DIDescriptor D = DA.getElement(i);
-        DIType TyE = DIType(D.getNode());
-        if (!TyE.isNull())
-          processType(TyE);
-        else
-          processSubprogram(DISubprogram(D.getNode()));
-      }
+    for (unsigned i = 0, e = DA.getNumElements(); i != e; ++i) {
+      DIDescriptor D = DA.getElement(i);
+      if (D.isType())
+        processType(DIType(D.getNode()));
+      else if (D.isSubprogram())
+        processSubprogram(DISubprogram(D.getNode()));
+    }
   } else if (DT.isDerivedType()) {
     DIDerivedType DDT(DT.getNode());
-    if (!DDT.isNull())
-      processType(DDT.getTypeDerivedFrom());
+    processType(DDT.getTypeDerivedFrom());
   }
 }
 
 /// processLexicalBlock
 void DebugInfoFinder::processLexicalBlock(DILexicalBlock LB) {
-  if (LB.isNull())
-    return;
   DIScope Context = LB.getContext();
   if (Context.isLexicalBlock())
     return processLexicalBlock(DILexicalBlock(Context.getNode()));
@@ -1206,8 +1197,6 @@
 
 /// processSubprogram - Process DISubprogram.
 void DebugInfoFinder::processSubprogram(DISubprogram SP) {
-  if (SP.isNull())
-    return;
   if (!addSubprogram(SP))
     return;
   addCompileUnit(SP.getCompileUnit());
@@ -1216,20 +1205,23 @@
 
 /// processDeclare - Process DbgDeclareInst.
 void DebugInfoFinder::processDeclare(DbgDeclareInst *DDI) {
-  DIVariable DV(cast<MDNode>(DDI->getVariable()));
-  if (DV.isNull())
+  MDNode *N = dyn_cast<MDNode>(DDI->getVariable());
+  if (!N) return;
+
+  DIDescriptor DV(N);
+  if (!DV.isVariable())
     return;
 
   if (!NodesSeen.insert(DV.getNode()))
     return;
 
-  addCompileUnit(DV.getCompileUnit());
-  processType(DV.getType());
+  addCompileUnit(DIVariable(N).getCompileUnit());
+  processType(DIVariable(N).getType());
 }
 
 /// addType - Add type into Tys.
 bool DebugInfoFinder::addType(DIType DT) {
-  if (DT.isNull())
+  if (!DT.isValid())
     return false;
 
   if (!NodesSeen.insert(DT.getNode()))
@@ -1241,7 +1233,7 @@
 
 /// addCompileUnit - Add compile unit into CUs.
 bool DebugInfoFinder::addCompileUnit(DICompileUnit CU) {
-  if (CU.isNull())
+  if (!CU.Verify())
     return false;
 
   if (!NodesSeen.insert(CU.getNode()))
@@ -1253,7 +1245,7 @@
 
 /// addGlobalVariable - Add global variable into GVs.
 bool DebugInfoFinder::addGlobalVariable(DIGlobalVariable DIG) {
-  if (DIG.isNull())
+  if (!DIDescriptor(DIG.getNode()).isGlobalVariable())
     return false;
 
   if (!NodesSeen.insert(DIG.getNode()))
@@ -1265,7 +1257,7 @@
 
 // addSubprogram - Add subprgoram into SPs.
 bool DebugInfoFinder::addSubprogram(DISubprogram SP) {
-  if (SP.isNull())
+  if (!DIDescriptor(SP.getNode()).isSubprogram())
     return false;
 
   if (!NodesSeen.insert(SP.getNode()))
@@ -1283,10 +1275,10 @@
     return 0;
 
   for (unsigned i = 0, e = NMD->getNumOperands(); i != e; ++i) {
-    DIGlobalVariable DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
-    if (DIG.isNull())
+    DIDescriptor DIG(cast_or_null<MDNode>(NMD->getOperand(i)));
+    if (!DIG.isGlobalVariable())
       continue;
-    if (DIG.getGlobal() == V)
+    if (DIGlobalVariable(DIG.getNode()).getGlobal() == V)
       return DIG.getNode();
   }
   return 0;
@@ -1378,12 +1370,6 @@
 /// getDISubprogram - Find subprogram that is enclosing this scope.
 DISubprogram llvm::getDISubprogram(MDNode *Scope) {
   DIDescriptor D(Scope);
-  if (D.isNull())
-    return DISubprogram();
-  
-  if (D.isCompileUnit())
-    return DISubprogram();
-  
   if (D.isSubprogram())
     return DISubprogram(Scope);
   
@@ -1395,9 +1381,6 @@
 
 /// getDICompositeType - Find underlying composite type.
 DICompositeType llvm::getDICompositeType(DIType T) {
-  if (T.isNull())
-    return DICompositeType();
-  
   if (T.isCompositeType())
     return DICompositeType(T.getNode());
   

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 14:52:55 2010
@@ -335,7 +335,7 @@
     // Print source line info.
     DIScope Scope = DLT.getScope();
     // Omit the directory, because it's likely to be long and uninteresting.
-    if (!Scope.isNull())
+    if (Scope.Verify())
       CommentOS << Scope.getFilename();
     else
       CommentOS << "<unknown>";
@@ -1287,7 +1287,7 @@
   if (DL.isUnknown())
     return;
   DILocation CurDLT = MF->getDILocation(DL);
-  if (CurDLT.getScope().isNull())
+  if (!CurDLT.getScope().Verify())
     return;
 
   if (!BeforePrintingInsn) {

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 14:52:55 2010
@@ -413,7 +413,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIVariable *V) {
   // If there is no compile unit specified, don't add a line #.
-  if (V->getCompileUnit().isNull())
+  if (!V->getCompileUnit().Verify())
     return;
 
   unsigned Line = V->getLineNumber();
@@ -427,7 +427,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DIGlobal *G) {
   // If there is no compile unit specified, don't add a line #.
-  if (G->getCompileUnit().isNull())
+  if (!G->getCompileUnit().Verify())
     return;
 
   unsigned Line = G->getLineNumber();
@@ -441,7 +441,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DISubprogram *SP) {
   // If there is no compile unit specified, don't add a line #.
-  if (SP->getCompileUnit().isNull())
+  if (!SP->getCompileUnit().Verify())
     return;
   // If the line number is 0, don't add it.
   if (SP->getLineNumber() == 0)
@@ -460,7 +460,7 @@
 void DwarfDebug::addSourceLine(DIE *Die, const DIType *Ty) {
   // If there is no compile unit specified, don't add a line #.
   DICompileUnit CU = Ty->getCompileUnit();
-  if (CU.isNull())
+  if (!CU.Verify())
     return;
 
   unsigned Line = Ty->getLineNumber();
@@ -474,7 +474,7 @@
 /// entry.
 void DwarfDebug::addSourceLine(DIE *Die, const DINameSpace *NS) {
   // If there is no compile unit specified, don't add a line #.
-  if (NS->getCompileUnit().isNull())
+  if (!NS->getCompileUnit().Verify())
     return;
 
   unsigned Line = NS->getLineNumber();
@@ -526,12 +526,8 @@
   }
 
   DICompositeType blockStruct = DICompositeType(subType.getNode());
-
   DIArray Elements = blockStruct.getTypeArray();
 
-  if (Elements.isNull())
-    return Ty;
-
   for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Elements.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -677,7 +673,6 @@
   DIDescriptor varField = DIDescriptor();
   DIDescriptor forwardingField = DIDescriptor();
 
-
   for (unsigned i = 0, N = Fields.getNumElements(); i < N; ++i) {
     DIDescriptor Element = Fields.getElement(i);
     DIDerivedType DT = DIDerivedType(Element.getNode());
@@ -688,10 +683,6 @@
       varField = Element;
   }
 
-  assert(!varField.isNull() && "Can't find byref variable in Block struct");
-  assert(!forwardingField.isNull()
-         && "Can't find forwarding field in Block struct");
-
   // Get the offsets for the forwarding field and the variable field.
   unsigned int forwardingFieldOffset =
     DIDerivedType(forwardingField.getNode()).getOffsetInBits() >> 3;
@@ -781,9 +772,7 @@
 
 /// addToContextOwner - Add Die into the list of its context owner's children.
 void DwarfDebug::addToContextOwner(DIE *Die, DIDescriptor Context) {
-  if (Context.isNull())
-    ModuleCU->addDie(Die);
-  else if (Context.isType()) {
+  if (Context.isType()) {
     DIE *ContextDIE = getOrCreateTypeDIE(DIType(Context.getNode()));
     ContextDIE->addChild(Die);
   } else if (Context.isNameSpace()) {
@@ -820,7 +809,7 @@
 
 /// addType - Add a new type attribute to the specified entity.
 void DwarfDebug::addType(DIE *Entity, DIType Ty) {
-  if (Ty.isNull())
+  if (!Ty.isValid())
     return;
 
   // Check for pre-existence.
@@ -906,9 +895,9 @@
     // Add enumerators to enumeration type.
     for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
       DIE *ElemDie = NULL;
-      DIEnumerator Enum(Elements.getElement(i).getNode());
-      if (!Enum.isNull()) {
-        ElemDie = constructEnumTypeDIE(&Enum);
+      DIDescriptor Enum(Elements.getElement(i).getNode());
+      if (Enum.isEnumerator()) {
+        ElemDie = constructEnumTypeDIE(DIEnumerator(Enum.getNode()));
         Buffer.addChild(ElemDie);
       }
     }
@@ -939,18 +928,17 @@
     DIArray Elements = CTy.getTypeArray();
 
     // A forward struct declared type may not have elements available.
-    if (Elements.isNull())
+    unsigned N = Elements.getNumElements();
+    if (N == 0)
       break;
 
     // Add elements to structure type.
-    for (unsigned i = 0, N = Elements.getNumElements(); i < N; ++i) {
+    for (unsigned i = 0; i < N; ++i) {
       DIDescriptor Element = Elements.getElement(i);
-      if (Element.isNull())
-        continue;
       DIE *ElemDie = NULL;
-      if (Element.getTag() == dwarf::DW_TAG_subprogram)
+      if (Element.isSubprogram())
         ElemDie = createSubprogramDIE(DISubprogram(Element.getNode()));
-      else if (Element.getTag() == dwarf::DW_TAG_auto_variable) {
+      else if (Element.isVariable()) {
         DIVariable DV(Element.getNode());
         ElemDie = new DIE(dwarf::DW_TAG_variable);
         addString(ElemDie, dwarf::DW_AT_name, dwarf::DW_FORM_string,
@@ -959,8 +947,10 @@
         addUInt(ElemDie, dwarf::DW_AT_declaration, dwarf::DW_FORM_flag, 1);
         addUInt(ElemDie, dwarf::DW_AT_external, dwarf::DW_FORM_flag, 1);
         addSourceLine(ElemDie, &DV);
-      } else
+      } else if (Element.isDerivedType())
         ElemDie = createMemberDIE(DIDerivedType(Element.getNode()));
+      else
+        continue;
       Buffer.addChild(ElemDie);
     }
 
@@ -973,7 +963,7 @@
               dwarf::DW_FORM_data1, RLang);
 
     DICompositeType ContainingType = CTy.getContainingType();
-    if (!ContainingType.isNull())
+    if (DIDescriptor(ContainingType.getNode()).isCompositeType())
       addDIEEntry(&Buffer, dwarf::DW_AT_containing_type, dwarf::DW_FORM_ref4, 
                   getOrCreateTypeDIE(DIType(ContainingType.getNode())));
     break;
@@ -1051,11 +1041,11 @@
 }
 
 /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator *ETy) {
+DIE *DwarfDebug::constructEnumTypeDIE(DIEnumerator ETy) {
   DIE *Enumerator = new DIE(dwarf::DW_TAG_enumerator);
-  StringRef Name = ETy->getName();
+  StringRef Name = ETy.getName();
   addString(Enumerator, dwarf::DW_AT_name, dwarf::DW_FORM_string, Name);
-  int64_t Value = ETy->getEnumValue();
+  int64_t Value = ETy.getEnumValue();
   addSInt(Enumerator, dwarf::DW_AT_const_value, dwarf::DW_FORM_sdata, Value);
   return Enumerator;
 }
@@ -1199,7 +1189,7 @@
   DIArray Args = SPTy.getTypeArray();
   unsigned SPTag = SPTy.getTag();
 
-  if (Args.isNull() || SPTag != dwarf::DW_TAG_subroutine_type)
+  if (Args.getNumElements() == 0 || SPTag != dwarf::DW_TAG_subroutine_type)
     addType(SPDie, SPTy);
   else
     addType(SPDie, DIType(Args.getElement(0).getNode()));
@@ -1282,11 +1272,9 @@
     Parent->addScope(NScope);
   } else if (DIDescriptor(N).isLexicalBlock()) {
     DILexicalBlock DB(N);
-    if (!DB.getContext().isNull()) {
-      Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
-      NScope->setParent(Parent);
-      Parent->addScope(NScope);
-    }
+    Parent = getUpdatedDbgScope(DB.getContext().getNode(), MI, InlinedAt);
+    NScope->setParent(Parent);
+    Parent->addScope(NScope);
   }
 
   NScope->setFirstInsn(MI);
@@ -1318,8 +1306,7 @@
   if (Scope.isLexicalBlock()) {
     DILexicalBlock DB(N);
     DIDescriptor ParentDesc = DB.getContext();
-    if (!ParentDesc.isNull())
-      Parent = getOrCreateAbstractScope(ParentDesc.getNode());
+    Parent = getOrCreateAbstractScope(ParentDesc.getNode());
   }
 
   AScope = new DbgScope(Parent, DIDescriptor(N), NULL);
@@ -1422,10 +1409,9 @@
   // Ignore empty scopes.
   if (StartID == EndID && StartID != 0)
     return NULL;
-
-  DIScope DS(Scope->getScopeNode());
-  if (DS.isNull())
+  if (!Scope->getScopeNode())
     return NULL;
+  DIScope DS(Scope->getScopeNode());
   DIE *ScopeDIE = new DIE(dwarf::DW_TAG_inlined_subroutine);
 
   DISubprogram InlinedSP = getDISubprogram(DS.getNode());
@@ -1547,15 +1533,12 @@
     return;
 
   DIArray Args = SPTy.getTypeArray();
-  if (Args.isNull()) 
-    return;
-
   for (unsigned i = 0, e = Args.getNumElements(); i != e; ++i) {
     DIType ATy(Args.getElement(i).getNode());
-    if (ATy.isNull())
+    if (!ATy.isValid())
       continue;
     DICompositeType CATy = getDICompositeType(ATy);
-    if (!CATy.isNull() && !CATy.getName().empty()) {
+    if (DIDescriptor(CATy.getNode()).Verify() && !CATy.getName().empty()) {
       if (DIEEntry *Entry = ModuleCU->getDIEEntry(CATy.getNode()))
         ModuleCU->addGlobalType(CATy.getName(), Entry->getEntry());
     }
@@ -1564,26 +1547,24 @@
 
 /// constructScopeDIE - Construct a DIE for this scope.
 DIE *DwarfDebug::constructScopeDIE(DbgScope *Scope) {
- if (!Scope)
-  return NULL;
- DIScope DS(Scope->getScopeNode());
- if (DS.isNull())
-   return NULL;
-
- DIE *ScopeDIE = NULL;
- if (Scope->getInlinedAt())
-   ScopeDIE = constructInlinedScopeDIE(Scope);
- else if (DS.isSubprogram()) {
-   if (Scope->isAbstractScope())
-     ScopeDIE = ModuleCU->getDIE(DS.getNode());
-   else
-     ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
- }
- else {
-   ScopeDIE = constructLexicalScopeDIE(Scope);
-   if (!ScopeDIE) return NULL;
- }
-
+  if (!Scope || !Scope->getScopeNode())
+    return NULL;
+ 
+  DIScope DS(Scope->getScopeNode());
+  DIE *ScopeDIE = NULL;
+  if (Scope->getInlinedAt())
+    ScopeDIE = constructInlinedScopeDIE(Scope);
+  else if (DS.isSubprogram()) {
+    if (Scope->isAbstractScope())
+      ScopeDIE = ModuleCU->getDIE(DS.getNode());
+    else
+      ScopeDIE = updateSubprogramScopeDIE(DS.getNode());
+  }
+  else {
+    ScopeDIE = constructLexicalScopeDIE(Scope);
+    if (!ScopeDIE) return NULL;
+  }
+  
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
@@ -2069,7 +2050,6 @@
       if (DL.isUnknown()) continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
-      if (DLTScope.isNull()) continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2089,7 +2069,6 @@
       if (DL.isUnknown())  continue;
       DILocation DLT = MF->getDILocation(DL);
       DIScope DLTScope = DLT.getScope();
-      if (DLTScope.isNull())  continue;
       // There is no need to create another DIE for compile unit. For all
       // other scopes, create one DbgScope now. This will be translated
       // into a scope DIE at the end.
@@ -2168,7 +2147,7 @@
     DILocation DLT = MF->getDILocation(FDL);
     unsigned LabelID = 0;
     DISubprogram SP = getDISubprogram(DLT.getScope().getNode());
-    if (!SP.isNull())
+    if (SP.Verify())
       LabelID = recordSourceLine(SP.getLineNumber(), 0, 
                                  DLT.getScope().getNode());
     else

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 14:52:55 2010
@@ -346,7 +346,7 @@
                              DICompositeType *CTy);
 
   /// constructEnumTypeDIE - Construct enum type DIE from DIEnumerator.
-  DIE *constructEnumTypeDIE(DIEnumerator *ETy);
+  DIE *constructEnumTypeDIE(DIEnumerator ETy);
 
   /// createGlobalVariableDIE - Create new DIE using GV.
   DIE *createGlobalVariableDIE(const DIGlobalVariable &GV);

Modified: llvm/trunk/lib/CodeGen/MachineInstr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineInstr.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineInstr.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineInstr.cpp Mon Mar  8 14:52:55 2010
@@ -1219,7 +1219,7 @@
     DIScope Scope = DLT.getScope();
     OS << " dbg:";
     // Omit the directory, since it's usually long and uninteresting.
-    if (!Scope.isNull())
+    if (Scope.Verify())
       OS << Scope.getFilename();
     else
       OS << "<unknown>";

Modified: llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16DebugInfo.cpp Mon Mar  8 14:52:55 2010
@@ -333,7 +333,7 @@
   for (DebugInfoFinder::iterator I = DbgFinder.type_begin(),
          E = DbgFinder.type_end(); I != E; ++I) {
     DICompositeType CTy(*I);
-    if (CTy.isNull())
+    if (!CTy.Verify())
       continue;
     if (CTy.getTag() == dwarf::DW_TAG_union_type ||
         CTy.getTag() == dwarf::DW_TAG_structure_type ) {

Modified: llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp?rev=97975&r1=97974&r2=97975&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/CloneFunction.cpp Mon Mar  8 14:52:55 2010
@@ -336,14 +336,14 @@
 
 static MDNode *UpdateInlinedAtInfo(MDNode *InsnMD, MDNode *TheCallMD) {
   DILocation ILoc(InsnMD);
-  if (ILoc.isNull()) return InsnMD;
+  if (!ILoc.Verify()) return InsnMD;
 
   DILocation CallLoc(TheCallMD);
-  if (CallLoc.isNull()) return InsnMD;
+  if (!CallLoc.Verify()) return InsnMD;
 
   DILocation OrigLocation = ILoc.getOrigLocation();
   MDNode *NewLoc = TheCallMD;
-  if (!OrigLocation.isNull())
+  if (OrigLocation.Verify())
     NewLoc = UpdateInlinedAtInfo(OrigLocation.getNode(), TheCallMD);
 
   Value *MDVs[] = {




From dpatel at apple.com  Mon Mar  8 14:53:59 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 20:53:59 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r97977 -
	/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
Message-ID: <20100308205359.6A3012A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 14:53:59 2010
New Revision: 97977

URL: http://llvm.org/viewvc/llvm-project?rev=97977&view=rev
Log:
Avoid using DIDescriptor.isNull(). 

One character "!" makes all the difference!

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=97977&r1=97976&r2=97977&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Mon Mar  8 14:53:59 2010
@@ -1147,7 +1147,7 @@
     MainTy = getOrCreateType(TYPE_MAIN_VARIANT(type));
 
   DIType Ty = createVariantType(type, MainTy);
-  if (!Ty.isNull())
+  if (Ty.isValid())
     return Ty;
 
   // Work out details of type.




From dpatel at apple.com  Mon Mar  8 15:00:27 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 21:00:27 -0000
Subject: [llvm-commits] [llvm] r97979 -
	/llvm/trunk/include/llvm/Analysis/DebugInfo.h
Message-ID: <20100308210028.0896F2A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 15:00:27 2010
New Revision: 97979

URL: http://llvm.org/viewvc/llvm-project?rev=97979&view=rev
Log:
isNull() is not used any more.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97979&r1=97978&r2=97979&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 15:00:27 2010
@@ -67,7 +67,6 @@
     explicit DIDescriptor() : DbgNode(0) {}
     explicit DIDescriptor(MDNode *N) : DbgNode(N) {}
 
-    bool isNull() const { return DbgNode == 0; }
     bool Verify() const { return DbgNode != 0; }
 
     MDNode *getNode() const { return DbgNode; }




From evan.cheng at apple.com  Mon Mar  8 15:05:02 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 08 Mar 2010 21:05:02 -0000
Subject: [llvm-commits] [llvm] r97980 - in /llvm/trunk:
 docs/CodeGenerator.html docs/LangRef.html test/CodeGen/X86/sibcall.ll
 test/CodeGen/X86/tailcall2.ll
Message-ID: <20100308210503.297632A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 15:05:02 2010
New Revision: 97980

URL: http://llvm.org/viewvc/llvm-project?rev=97980&view=rev
Log:
Add documentation on sibling call optimization. Rename tailcall2.ll test to sibcall.ll.

Added:
    llvm/trunk/test/CodeGen/X86/sibcall.ll
      - copied unchanged from r97973, llvm/trunk/test/CodeGen/X86/tailcall2.ll
Removed:
    llvm/trunk/test/CodeGen/X86/tailcall2.ll
Modified:
    llvm/trunk/docs/CodeGenerator.html
    llvm/trunk/docs/LangRef.html

Modified: llvm/trunk/docs/CodeGenerator.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/CodeGenerator.html?rev=97980&r1=97979&r2=97980&view=diff
==============================================================================
--- llvm/trunk/docs/CodeGenerator.html (original)
+++ llvm/trunk/docs/CodeGenerator.html Mon Mar  8 15:05:02 2010
@@ -86,6 +86,7 @@
   <li><a href="#targetimpls">Target-specific Implementation Notes</a>
     <ul>
     <li><a href="#tailcallopt">Tail call optimization</a></li>
+    <li><a href="#sibcallopt">Sibling call optimization</a></li>
     <li><a href="#x86">The X86 backend</a></li>
     <li><a href="#ppc">The PowerPC backend</a>
       <ul>
@@ -1734,6 +1735,50 @@
 </div>
 <!-- ======================================================================= -->
 <div class="doc_subsection">
+  <a name="sibcallopt">Sibling call optimization</a>
+</div>
+
+<div class="doc_text">
+
+<p>Sibling call optimization is a restricted form of tail call optimization.
+   Unlike tail call optimization described in the previous section, it can be
+   performed automatically on any tail calls when <tt>-tailcallopt</tt> option
+   is not specified.</p>
+
+<p>Sibling call optimization is currently performed on x86/x86-64 when the
+   following constraints are met:</p>
+
+<ul>
+  <li>Caller and callee have the same calling convention. It can be either
+      <tt>c</tt> or <tt>fastcc</tt>.
+
+  <li>The call is a tail call - in tail position (ret immediately follows call
+      and ret uses value of call or is void).</li>
+
+  <li>Caller and callee have matching return type or the callee result is not
+      used.
+
+  <li>If any of the callee arguments are being passed in stack, they must be
+      available in caller's own incoming argument stack and the frame offsets
+      must be the same.
+</ul>
+
+<p>Example:</p>
+<div class="doc_code">
+<pre>
+declare i32 @bar(i32, i32)
+
+define i32 @foo(i32 %a, i32 %b, i32 %c) {
+entry:
+  %0 = tail call i32 @bar(i32 %a, i32 %b)
+  ret i32 %0
+}
+</pre>
+</div>
+
+</div>
+<!-- ======================================================================= -->
+<div class="doc_subsection">
   <a name="x86">The X86 backend</a>
 </div>
 

Modified: llvm/trunk/docs/LangRef.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/LangRef.html?rev=97980&r1=97979&r2=97980&view=diff
==============================================================================
--- llvm/trunk/docs/LangRef.html (original)
+++ llvm/trunk/docs/LangRef.html Mon Mar  8 15:05:02 2010
@@ -5149,8 +5149,11 @@
       a <a href="#i_ret"><tt>ret</tt></a> instruction.  If the "tail" marker is
       present, the function call is eligible for tail call optimization,
       but <a href="CodeGenerator.html#tailcallopt">might not in fact be
-      optimized into a jump</a>.  As of this writing, the extra requirements for
-      a call to actually be optimized are:
+      optimized into a jump</a>.  The code generator may optimize calls marked
+      "tail" with either 1) automatic <a href="CodeGenerator.html#sibcallopt">
+      sibling call optimization</a> when the caller and callee have
+      matching signatures, or 2) forced tail call optimization when the
+      following extra requirements are met:
       <ul>
         <li>Caller and callee both have the calling
             convention <tt>fastcc</tt>.</li>

Removed: llvm/trunk/test/CodeGen/X86/tailcall2.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcall2.ll?rev=97979&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/tailcall2.ll (original)
+++ llvm/trunk/test/CodeGen/X86/tailcall2.ll (removed)
@@ -1,218 +0,0 @@
-; RUN: llc < %s -march=x86    -asm-verbose=false | FileCheck %s -check-prefix=32
-; RUN: llc < %s -march=x86-64 -asm-verbose=false | FileCheck %s -check-prefix=64
-
-define void @t1(i32 %x) nounwind ssp {
-entry:
-; 32: t1:
-; 32: jmp {{_?}}foo
-
-; 64: t1:
-; 64: jmp {{_?}}foo
-  tail call void @foo() nounwind
-  ret void
-}
-
-declare void @foo()
-
-define void @t2() nounwind ssp {
-entry:
-; 32: t2:
-; 32: jmp {{_?}}foo2
-
-; 64: t2:
-; 64: jmp {{_?}}foo2
-  %0 = tail call i32 @foo2() nounwind
-  ret void
-}
-
-declare i32 @foo2()
-
-define void @t3() nounwind ssp {
-entry:
-; 32: t3:
-; 32: jmp {{_?}}foo3
-
-; 64: t3:
-; 64: jmp {{_?}}foo3
-  %0 = tail call i32 @foo3() nounwind
-  ret void
-}
-
-declare i32 @foo3()
-
-define void @t4(void (i32)* nocapture %x) nounwind ssp {
-entry:
-; 32: t4:
-; 32: call *
-; FIXME: gcc can generate a tailcall for this. But it's tricky.
-
-; 64: t4:
-; 64-NOT: call
-; 64: jmpq *
-  tail call void %x(i32 0) nounwind
-  ret void
-}
-
-define void @t5(void ()* nocapture %x) nounwind ssp {
-entry:
-; 32: t5:
-; 32-NOT: call
-; 32: jmpl *
-
-; 64: t5:
-; 64-NOT: call
-; 64: jmpq *
-  tail call void %x() nounwind
-  ret void
-}
-
-define i32 @t6(i32 %x) nounwind ssp {
-entry:
-; 32: t6:
-; 32: call {{_?}}t6
-; 32: jmp {{_?}}bar
-
-; 64: t6:
-; 64: jmp {{_?}}t6
-; 64: jmp {{_?}}bar
-  %0 = icmp slt i32 %x, 10
-  br i1 %0, label %bb, label %bb1
-
-bb:
-  %1 = add nsw i32 %x, -1
-  %2 = tail call i32 @t6(i32 %1) nounwind ssp
-  ret i32 %2
-
-bb1:
-  %3 = tail call i32 @bar(i32 %x) nounwind
-  ret i32 %3
-}
-
-declare i32 @bar(i32)
-
-define i32 @t7(i32 %a, i32 %b, i32 %c) nounwind ssp {
-entry:
-; 32: t7:
-; 32: jmp {{_?}}bar2
-
-; 64: t7:
-; 64: jmp {{_?}}bar2
-  %0 = tail call i32 @bar2(i32 %a, i32 %b, i32 %c) nounwind
-  ret i32 %0
-}
-
-declare i32 @bar2(i32, i32, i32)
-
-define signext i16 @t8() nounwind ssp {
-entry:
-; 32: t8:
-; 32: call {{_?}}bar3
-
-; 64: t8:
-; 64: callq {{_?}}bar3
-  %0 = tail call signext i16 @bar3() nounwind      ; <i16> [#uses=1]
-  ret i16 %0
-}
-
-declare signext i16 @bar3()
-
-define signext i16 @t9(i32 (i32)* nocapture %x) nounwind ssp {
-entry:
-; 32: t9:
-; 32: call *
-
-; 64: t9:
-; 64: callq *
-  %0 = bitcast i32 (i32)* %x to i16 (i32)*
-  %1 = tail call signext i16 %0(i32 0) nounwind
-  ret i16 %1
-}
-
-define void @t10() nounwind ssp {
-entry:
-; 32: t10:
-; 32: call
-
-; 64: t10:
-; 64: callq
-  %0 = tail call i32 @foo4() noreturn nounwind
-  unreachable
-}
-
-declare i32 @foo4()
-
-define i32 @t11(i32 %x, i32 %y, i32 %z.0, i32 %z.1, i32 %z.2) nounwind ssp {
-; In 32-bit mode, it's emitting a bunch of dead loads that are not being
-; eliminated currently.
-
-; 32: t11:
-; 32-NOT: subl ${{[0-9]+}}, %esp
-; 32: jne
-; 32-NOT: movl
-; 32-NOT: addl ${{[0-9]+}}, %esp
-; 32: jmp {{_?}}foo5
-
-; 64: t11:
-; 64-NOT: subq ${{[0-9]+}}, %esp
-; 64-NOT: addq ${{[0-9]+}}, %esp
-; 64: jmp {{_?}}foo5
-entry:
-  %0 = icmp eq i32 %x, 0
-  br i1 %0, label %bb6, label %bb
-
-bb:
-  %1 = tail call i32 @foo5(i32 %x, i32 %y, i32 %z.0, i32 %z.1, i32 %z.2) nounwind
-  ret i32 %1
-
-bb6:
-  ret i32 0
-}
-
-declare i32 @foo5(i32, i32, i32, i32, i32)
-
-%struct.t = type { i32, i32, i32, i32, i32 }
-
-define i32 @t12(i32 %x, i32 %y, %struct.t* byval align 4 %z) nounwind ssp {
-; 32: t12:
-; 32-NOT: subl ${{[0-9]+}}, %esp
-; 32-NOT: addl ${{[0-9]+}}, %esp
-; 32: jmp {{_?}}foo6
-
-; 64: t12:
-; 64-NOT: subq ${{[0-9]+}}, %esp
-; 64-NOT: addq ${{[0-9]+}}, %esp
-; 64: jmp {{_?}}foo6
-entry:
-  %0 = icmp eq i32 %x, 0
-  br i1 %0, label %bb2, label %bb
-
-bb:
-  %1 = tail call i32 @foo6(i32 %x, i32 %y, %struct.t* byval align 4 %z) nounwind
-  ret i32 %1
-
-bb2:
-  ret i32 0
-}
-
-declare i32 @foo6(i32, i32, %struct.t* byval align 4)
-
-; rdar://r7717598
-%struct.ns = type { i32, i32 }
-%struct.cp = type { float, float }
-
-define %struct.ns* @t13(%struct.cp* %yy) nounwind ssp {
-; 32: t13:
-; 32-NOT: jmp
-; 32: call
-; 32: ret
-
-; 64: t13:
-; 64-NOT: jmp
-; 64: call
-; 64: ret
-entry:
-  %0 = tail call fastcc %struct.ns* @foo7(%struct.cp* byval align 4 %yy, i8 signext 0) nounwind
-  ret %struct.ns* %0
-}
-
-declare fastcc %struct.ns* @foo7(%struct.cp* byval align 4, i8 signext) nounwind ssp




From daniel at zuster.org  Mon Mar  8 15:10:36 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 21:10:36 -0000
Subject: [llvm-commits] [llvm] r97982 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td test/MC/AsmParser/X86/x86_32-new-encoder.s
Message-ID: <20100308211036.C2AE32A6C12C@llvm.org>

Author: ddunbar
Date: Mon Mar  8 15:10:36 2010
New Revision: 97982

URL: http://llvm.org/viewvc/llvm-project?rev=97982&view=rev
Log:
X86: Fix encoding for TEST{8,16,32}rr.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=97982&r1=97981&r2=97982&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Mar  8 15:10:36 2010
@@ -3135,16 +3135,16 @@
 //
 let Defs = [EFLAGS] in {
 let isCommutable = 1 in {   // TEST X, Y   --> TEST Y, X
-def TEST8rr  : I<0x84, MRMDestReg, (outs),  (ins GR8:$src1, GR8:$src2),
+def TEST8rr  : I<0x84, MRMSrcReg, (outs),  (ins GR8:$src1, GR8:$src2),
                      "test{b}\t{$src2, $src1|$src1, $src2}",
                      [(X86cmp (and_su GR8:$src1, GR8:$src2), 0),
                       (implicit EFLAGS)]>;
-def TEST16rr : I<0x85, MRMDestReg, (outs),  (ins GR16:$src1, GR16:$src2),
+def TEST16rr : I<0x85, MRMSrcReg, (outs),  (ins GR16:$src1, GR16:$src2),
                      "test{w}\t{$src2, $src1|$src1, $src2}",
                      [(X86cmp (and_su GR16:$src1, GR16:$src2), 0),
                       (implicit EFLAGS)]>,
                  OpSize;
-def TEST32rr : I<0x85, MRMDestReg, (outs),  (ins GR32:$src1, GR32:$src2),
+def TEST32rr : I<0x85, MRMSrcReg, (outs),  (ins GR32:$src1, GR32:$src2),
                      "test{l}\t{$src2, $src1|$src1, $src2}",
                      [(X86cmp (and_su GR32:$src1, GR32:$src2), 0),
                       (implicit EFLAGS)]>;

Modified: llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s?rev=97982&r1=97981&r2=97982&view=diff
==============================================================================
--- llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s (original)
+++ llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s Mon Mar  8 15:10:36 2010
@@ -38,4 +38,6 @@
 	movl	%eax, 16(%ebp)
 // CHECK: movl	%eax, -16(%ebp)          # encoding: [0x89,0x45,0xf0]
 	movl	%eax, -16(%ebp)
-        
+
+// CHECK: testb	%bl, %cl                # encoding: [0x84,0xcb]
+        testb %bl, %cl




From daniel at zuster.org  Mon Mar  8 15:10:40 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 21:10:40 -0000
Subject: [llvm-commits] [llvm] r97983 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100308211040.26A982A6C12D@llvm.org>

Author: ddunbar
Date: Mon Mar  8 15:10:39 2010
New Revision: 97983

URL: http://llvm.org/viewvc/llvm-project?rev=97983&view=rev
Log:
MC/Mach-O: Error out instead of crashing on invalid scattered relocation expressions.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=97983&r1=97982&r2=97983&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 15:10:39 2010
@@ -447,13 +447,22 @@
     // See <reloc.h>.
     const MCSymbol *A = Target.getSymA();
     MCSymbolData *SD = SymbolMap.lookup(A);
+
+    if (!SD->getFragment())
+      llvm_report_error("symbol '" + A->getName() +
+                        "' can not be undefined in a subtraction expression");
+
     uint32_t Value = SD->getFragment()->getAddress() + SD->getOffset();
     uint32_t Value2 = 0;
 
     if (const MCSymbol *B = Target.getSymB()) {
-      Type = RIT_LocalDifference;
-
       MCSymbolData *SD = SymbolMap.lookup(B);
+
+      if (!SD->getFragment())
+        llvm_report_error("symbol '" + B->getName() +
+                          "' can not be undefined in a subtraction expression");
+
+      Type = RIT_LocalDifference;
       Value2 = SD->getFragment()->getAddress() + SD->getOffset();
     }
 




From daniel at zuster.org  Mon Mar  8 15:10:43 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 21:10:43 -0000
Subject: [llvm-commits] [llvm] r97984 - in /llvm/trunk:
 lib/MC/MCAssembler.cpp test/MC/MachO/zerofill-4.s
Message-ID: <20100308211043.108A22A6C12E@llvm.org>

Author: ddunbar
Date: Mon Mar  8 15:10:42 2010
New Revision: 97984

URL: http://llvm.org/viewvc/llvm-project?rev=97984&view=rev
Log:
MC/Mach-O: Fix address compution for zero fill sections.

Added:
    llvm/trunk/test/MC/MachO/zerofill-4.s
Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=97984&r1=97983&r2=97984&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 15:10:42 2010
@@ -1049,8 +1049,8 @@
 
       // Align the fragment offset; it is safe to adjust the offset freely since
       // this is only in virtual sections.
-      uint64_t Aligned = RoundUpToAlignment(Address, ZFF.getAlignment());
-      F.setOffset(Aligned - SD.getAddress());
+      Address = RoundUpToAlignment(Address, ZFF.getAlignment());
+      F.setOffset(Address - SD.getAddress());
 
       // FIXME: This is misnamed.
       F.setFileSize(ZFF.getSize());

Added: llvm/trunk/test/MC/MachO/zerofill-4.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/zerofill-4.s?rev=97984&view=auto
==============================================================================
--- llvm/trunk/test/MC/MachO/zerofill-4.s (added)
+++ llvm/trunk/test/MC/MachO/zerofill-4.s Mon Mar  8 15:10:42 2010
@@ -0,0 +1,35 @@
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump --dump-section-data | FileCheck %s
+
+.zerofill __DATA,__bss,_fill0,1,0
+.zerofill __DATA,__bss,_a,4,2
+.zerofill __DATA,__bss,_fill1,1,0
+.zerofill __DATA,__bss,_b,4,3
+.zerofill __DATA,__bss,_fill2,1,0
+.zerofill __DATA,__bss,_c,4,4
+.zerofill __DATA,__bss,_fill3,1,0
+.zerofill __DATA,__bss,_d,4,5
+
+// CHECK: # Symbol 0
+// CHECK: ('n_value', 0)
+// CHECK: ('_string', '_fill0')
+// CHECK: # Symbol 1
+// CHECK: ('n_value', 4)
+// CHECK: ('_string', '_a')
+// CHECK: # Symbol 2
+// CHECK: ('n_value', 8)
+// CHECK: ('_string', '_fill1')
+// CHECK: # Symbol 3
+// CHECK: ('n_value', 16)
+// CHECK: ('_string', '_b')
+// CHECK: # Symbol 4
+// CHECK: ('n_value', 20)
+// CHECK: ('_string', '_fill2')
+// CHECK: # Symbol 5
+// CHECK: ('n_value', 32)
+// CHECK: ('_string', '_c')
+// CHECK: # Symbol 6
+// CHECK: ('n_value', 36)
+// CHECK: ('_string', '_fill3')
+// CHECK: # Symbol 7
+// CHECK: ('n_value', 64)
+// CHECK: ('_string', '_d')




From clattner at apple.com  Mon Mar  8 15:24:55 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 8 Mar 2010 13:24:55 -0800
Subject: [llvm-commits] [llvm] r97983 -
	/llvm/trunk/lib/MC/MCAssembler.cpp
In-Reply-To: <20100308211040.26A982A6C12D@llvm.org>
References: <20100308211040.26A982A6C12D@llvm.org>
Message-ID: <C01E0093-2BB3-4738-BBD9-1CB88E9B7797@apple.com>


On Mar 8, 2010, at 1:10 PM, Daniel Dunbar wrote:

> Author: ddunbar
> Date: Mon Mar  8 15:10:39 2010
> New Revision: 97983
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=97983&view=rev
> Log:
> MC/Mach-O: Error out instead of crashing on invalid scattered relocation expressions.

Is this something that should be producing a diagnostic?  If not, please just use assert.

-Chris

> 
> Modified:
>    llvm/trunk/lib/MC/MCAssembler.cpp
> 
> Modified: llvm/trunk/lib/MC/MCAssembler.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=97983&r1=97982&r2=97983&view=diff
> ==============================================================================
> --- llvm/trunk/lib/MC/MCAssembler.cpp (original)
> +++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 15:10:39 2010
> @@ -447,13 +447,22 @@
>     // See <reloc.h>.
>     const MCSymbol *A = Target.getSymA();
>     MCSymbolData *SD = SymbolMap.lookup(A);
> +
> +    if (!SD->getFragment())
> +      llvm_report_error("symbol '" + A->getName() +
> +                        "' can not be undefined in a subtraction expression");
> +
>     uint32_t Value = SD->getFragment()->getAddress() + SD->getOffset();
>     uint32_t Value2 = 0;
> 
>     if (const MCSymbol *B = Target.getSymB()) {
> -      Type = RIT_LocalDifference;
> -
>       MCSymbolData *SD = SymbolMap.lookup(B);
> +
> +      if (!SD->getFragment())
> +        llvm_report_error("symbol '" + B->getName() +
> +                          "' can not be undefined in a subtraction expression");
> +
> +      Type = RIT_LocalDifference;
>       Value2 = SD->getFragment()->getAddress() + SD->getOffset();
>     }
> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From dpatel at apple.com  Mon Mar  8 15:32:10 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 21:32:10 -0000
Subject: [llvm-commits] [llvm] r97988 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
Message-ID: <20100308213210.CDA262A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 15:32:10 2010
New Revision: 97988

URL: http://llvm.org/viewvc/llvm-project?rev=97988&view=rev
Log:
Remove DbgNode checks in constructor. Debug descriptors are intended to be light weight wrappers.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97988&r1=97987&r2=97988&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 15:32:10 2010
@@ -44,11 +44,6 @@
   protected:
     MDNode *DbgNode;
 
-    /// DIDescriptor constructor.  If the specified node is non-null, check
-    /// to make sure that the tag in the descriptor matches 'RequiredTag'.  If
-    /// not, the debug info is corrupt and we ignore it.
-    DIDescriptor(MDNode *N, unsigned RequiredTag);
-
     StringRef getStringField(unsigned Elt) const;
     unsigned getUnsignedField(unsigned Elt) const {
       return (unsigned)getUInt64Field(Elt);
@@ -104,8 +99,7 @@
   /// DISubrange - This is used to represent ranges, for array bounds.
   class DISubrange : public DIDescriptor {
   public:
-    explicit DISubrange(MDNode *N = 0)
-      : DIDescriptor(N, dwarf::DW_TAG_subrange_type) {}
+    explicit DISubrange(MDNode *N = 0) : DIDescriptor(N) {}
 
     int64_t getLo() const { return (int64_t)getUInt64Field(1); }
     int64_t getHi() const { return (int64_t)getUInt64Field(2); }
@@ -126,10 +120,7 @@
   /// DIScope - A base class for various scopes.
   class DIScope : public DIDescriptor {
   public:
-    explicit DIScope(MDNode *N = 0) : DIDescriptor (N) {
-      if (DbgNode && !isScope())
-        DbgNode = 0;
-    }
+    explicit DIScope(MDNode *N = 0) : DIDescriptor (N) {}
     virtual ~DIScope() {}
 
     StringRef getFilename() const;
@@ -139,10 +130,7 @@
   /// DICompileUnit - A wrapper for a compile unit.
   class DICompileUnit : public DIScope {
   public:
-    explicit DICompileUnit(MDNode *N = 0) : DIScope(N) {
-      if (DbgNode && !isCompileUnit())
-        DbgNode = 0;
-    }
+    explicit DICompileUnit(MDNode *N = 0) : DIScope(N) {}
 
     unsigned getLanguage() const     { return getUnsignedField(2); }
     StringRef getFilename() const  { return getStringField(3);   }
@@ -175,8 +163,7 @@
   /// type/precision or a file/line pair for location info.
   class DIEnumerator : public DIDescriptor {
   public:
-    explicit DIEnumerator(MDNode *N = 0)
-      : DIDescriptor(N, dwarf::DW_TAG_enumerator) {}
+    explicit DIEnumerator(MDNode *N = 0) : DIDescriptor(N) {}
 
     StringRef getName() const        { return getStringField(1); }
     uint64_t getEnumValue() const      { return getUInt64Field(2); }
@@ -199,8 +186,6 @@
     };
 
   protected:
-    DIType(MDNode *N, unsigned Tag)
-      : DIDescriptor(N, Tag) {}
     // This ctor is used when the Tag has already been validated by a derived
     // ctor.
     DIType(MDNode *N, bool, bool) : DIDescriptor(N) {}
@@ -256,8 +241,7 @@
   /// DIBasicType - A basic type, like 'int' or 'float'.
   class DIBasicType : public DIType {
   public:
-    explicit DIBasicType(MDNode *N = 0)
-      : DIType(N, dwarf::DW_TAG_base_type) {}
+    explicit DIBasicType(MDNode *N = 0) : DIType(N) {}
 
     unsigned getEncoding() const { return getUnsignedField(9); }
 
@@ -273,10 +257,7 @@
       : DIType(N, true, true) {}
   public:
     explicit DIDerivedType(MDNode *N = 0)
-      : DIType(N, true, true) {
-      if (DbgNode && !isDerivedType())
-        DbgNode = 0;
-    }
+      : DIType(N, true, true) {}
 
     DIType getTypeDerivedFrom() const { return getFieldAs<DIType>(9); }
 
@@ -319,8 +300,7 @@
   /// DIGlobal - This is a common class for global variables and subprograms.
   class DIGlobal : public DIDescriptor {
   protected:
-    explicit DIGlobal(MDNode *N, unsigned RequiredTag)
-      : DIDescriptor(N, RequiredTag) {}
+    explicit DIGlobal(MDNode *N) : DIDescriptor(N) {}
 
   public:
     virtual ~DIGlobal() {}
@@ -345,10 +325,7 @@
   /// DISubprogram - This is a wrapper for a subprogram (e.g. a function).
   class DISubprogram : public DIScope {
   public:
-    explicit DISubprogram(MDNode *N = 0) : DIScope(N) {
-      if (DbgNode && !isSubprogram())
-        DbgNode = 0;
-    }
+    explicit DISubprogram(MDNode *N = 0) : DIScope(N) {}
 
     DIDescriptor getContext() const     { return getDescriptorField(2); }
     StringRef getName() const         { return getStringField(3); }
@@ -401,8 +378,7 @@
   /// DIGlobalVariable - This is a wrapper for a global variable.
   class DIGlobalVariable : public DIGlobal {
   public:
-    explicit DIGlobalVariable(MDNode *N = 0)
-      : DIGlobal(N, dwarf::DW_TAG_variable) {}
+    explicit DIGlobalVariable(MDNode *N = 0) : DIGlobal(N) {}
 
     GlobalVariable *getGlobal() const { return getGlobalVariableField(11); }
 
@@ -418,10 +394,7 @@
   class DIVariable : public DIDescriptor {
   public:
     explicit DIVariable(MDNode *N = 0)
-      : DIDescriptor(N) {
-      if (DbgNode && !isVariable())
-        DbgNode = 0;
-    }
+      : DIDescriptor(N) {}
 
     DIDescriptor getContext() const { return getDescriptorField(1); }
     StringRef getName() const     { return getStringField(2);     }
@@ -457,10 +430,7 @@
   /// DILexicalBlock - This is a wrapper for a lexical block.
   class DILexicalBlock : public DIScope {
   public:
-    explicit DILexicalBlock(MDNode *N = 0) : DIScope(N) {
-      if (DbgNode && !isLexicalBlock())
-        DbgNode = 0;
-    }
+    explicit DILexicalBlock(MDNode *N = 0) : DIScope(N) {}
     DIScope getContext() const       { return getFieldAs<DIScope>(1);      }
     StringRef getDirectory() const   { return getContext().getDirectory(); }
     StringRef getFilename() const    { return getContext().getFilename();  }
@@ -471,11 +441,7 @@
   /// DINameSpace - A wrapper for a C++ style name space.
   class DINameSpace : public DIScope { 
   public:
-    explicit DINameSpace(MDNode *N = 0) : DIScope(N) {
-      if (DbgNode && !isNameSpace())
-        DbgNode = 0;
-    }
-
+    explicit DINameSpace(MDNode *N = 0) : DIScope(N) {}
     DIScope getContext() const     { return getFieldAs<DIScope>(1);      }
     StringRef getName() const      { return getStringField(2);           }
     StringRef getDirectory() const { return getContext().getDirectory(); }

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97988&r1=97987&r2=97988&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 15:32:10 2010
@@ -69,15 +69,6 @@
   return true;
 }
 
-DIDescriptor::DIDescriptor(MDNode *N, unsigned RequiredTag) {
-  DbgNode = N;
-
-  // If this is non-null, check to see if the Tag matches. If not, set to null.
-  if (N && getTag() != RequiredTag) {
-    DbgNode = 0;
-  }
-}
-
 StringRef 
 DIDescriptor::getStringField(unsigned Elt) const {
   if (DbgNode == 0)




From gohman at apple.com  Mon Mar  8 15:33:02 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 08 Mar 2010 21:33:02 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r97989 -
	/llvm-gcc-4.2/trunk/gcc/llvm-types.cpp
Message-ID: <20100308213302.618102A6C12C@llvm.org>

Author: djg
Date: Mon Mar  8 15:33:02 2010
New Revision: 97989

URL: http://llvm.org/viewvc/llvm-project?rev=97989&view=rev
Log:
ABIs don't support noalias on sret arguments. See PR6525 for details.

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-types.cpp

Modified: llvm-gcc-4.2/trunk/gcc/llvm-types.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-types.cpp?rev=97989&r1=97988&r2=97989&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-types.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-types.cpp Mon Mar  8 15:33:02 2010
@@ -1098,10 +1098,16 @@
     Attrs.push_back(AttributeWithIndex::get(0, RAttributes));
 
   // If this function returns via a shadow argument, the dest loc is passed
-  // in as a pointer.  Mark that pointer as struct-ret and noalias.
+  // in as a pointer.  Mark that pointer as struct-ret.
+  //
+  // It's tempting to want NoAlias here too, however even though llvm-gcc
+  // itself currently always passes a dedicated alloca as the actual argument,
+  // this isn't mandated by the ABI. There are other compilers which don't
+  // always pass a dedicated alloca. Using NoAlias here would make code which
+  // isn't interoperable with that of other compilers.
   if (ABIConverter.isShadowReturn())
     Attrs.push_back(AttributeWithIndex::get(ArgTys.size(),
-                                    Attribute::StructRet | Attribute::NoAlias));
+                                    Attribute::StructRet));
 
   std::vector<const Type*> ScalarArgs;
   if (static_chain) {




From dpatel at apple.com  Mon Mar  8 16:02:50 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 22:02:50 -0000
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
Message-ID: <20100308220250.5FC1F2A6C12E@llvm.org>

Author: dpatel
Date: Mon Mar  8 16:02:50 2010
New Revision: 97990

URL: http://llvm.org/viewvc/llvm-project?rev=97990&view=rev
Log:
Derive DIType from DIScope. This simplifies getContext() where for members the context is a type. This also eliminates need of CompileUnitMaps maintained by dwarf writer.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97990&r1=97989&r2=97990&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 16:02:50 2010
@@ -172,7 +172,7 @@
   /// DIType - This is a wrapper for a type.
   /// FIXME: Types should be factored much better so that CV qualifiers and
   /// others do not require a huge and empty descriptor full of zeros.
-  class DIType : public DIDescriptor {
+  class DIType : public DIScope {
   public:
     enum {
       FlagPrivate          = 1 << 0,
@@ -188,7 +188,7 @@
   protected:
     // This ctor is used when the Tag has already been validated by a derived
     // ctor.
-    DIType(MDNode *N, bool, bool) : DIDescriptor(N) {}
+    DIType(MDNode *N, bool, bool) : DIScope(N) {}
 
   public:
 
@@ -199,7 +199,7 @@
     explicit DIType() {}
     virtual ~DIType() {}
 
-    DIDescriptor getContext() const     { return getDescriptorField(1); }
+    DIScope getContext() const          { return getFieldAs<DIScope>(1); }
     StringRef getName() const           { return getStringField(2);     }
     DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
     unsigned getLineNumber() const      { return getUnsignedField(4); }
@@ -234,6 +234,8 @@
     bool isValid() const {
       return DbgNode && (isBasicType() || isDerivedType() || isCompositeType());
     }
+    StringRef getFilename() const    { return getCompileUnit().getFilename();}
+    StringRef getDirectory() const   { return getCompileUnit().getDirectory();}
     /// dump - print type.
     void dump() const;
   };
@@ -305,7 +307,7 @@
   public:
     virtual ~DIGlobal() {}
 
-    DIDescriptor getContext() const     { return getDescriptorField(2); }
+    DIScope getContext() const          { return getFieldAs<DIScope>(2); }
     StringRef getName() const         { return getStringField(3); }
     StringRef getDisplayName() const  { return getStringField(4); }
     StringRef getLinkageName() const  { return getStringField(5); }
@@ -327,7 +329,7 @@
   public:
     explicit DISubprogram(MDNode *N = 0) : DIScope(N) {}
 
-    DIDescriptor getContext() const     { return getDescriptorField(2); }
+    DIScope getContext() const          { return getFieldAs<DIScope>(2); }
     StringRef getName() const         { return getStringField(3); }
     StringRef getDisplayName() const  { return getStringField(4); }
     StringRef getLinkageName() const  { return getStringField(5); }
@@ -396,8 +398,8 @@
     explicit DIVariable(MDNode *N = 0)
       : DIDescriptor(N) {}
 
-    DIDescriptor getContext() const { return getDescriptorField(1); }
-    StringRef getName() const     { return getStringField(2);     }
+    DIScope getContext() const          { return getFieldAs<DIScope>(1); }
+    StringRef getName() const           { return getStringField(2);     }
     DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
     unsigned getLineNumber() const      { return getUnsignedField(4); }
     DIType getType() const              { return getFieldAs<DIType>(5); }

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97990&r1=97989&r2=97990&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 16:02:50 2010
@@ -243,7 +243,7 @@
 // Simple Descriptor Constructors and other Methods
 //===----------------------------------------------------------------------===//
 
-DIType::DIType(MDNode *N) : DIDescriptor(N) {
+DIType::DIType(MDNode *N) : DIScope(N) {
   if (!N) return;
   if (!isBasicType() && !isDerivedType() && !isCompositeType()) {
     DbgNode = 0;
@@ -412,6 +412,8 @@
 }
 
 StringRef DIScope::getFilename() const {
+  if (!DbgNode)
+    return StringRef();
   if (isLexicalBlock()) 
     return DILexicalBlock(DbgNode).getFilename();
   if (isSubprogram())
@@ -420,11 +422,15 @@
     return DICompileUnit(DbgNode).getFilename();
   if (isNameSpace())
     return DINameSpace(DbgNode).getFilename();
+  if (isType())
+    return DIType(DbgNode).getFilename();
   assert(0 && "Invalid DIScope!");
   return StringRef();
 }
 
 StringRef DIScope::getDirectory() const {
+  if (!DbgNode)
+    return StringRef();
   if (isLexicalBlock()) 
     return DILexicalBlock(DbgNode).getDirectory();
   if (isSubprogram())
@@ -433,6 +439,8 @@
     return DICompileUnit(DbgNode).getDirectory();
   if (isNameSpace())
     return DINameSpace(DbgNode).getDirectory();
+  if (isType())
+    return DIType(DbgNode).getDirectory();
   assert(0 && "Invalid DIScope!");
   return StringRef();
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97990&r1=97989&r2=97990&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:02:50 2010
@@ -417,7 +417,8 @@
     return;
 
   unsigned Line = V->getLineNumber();
-  unsigned FileID = findCompileUnit(V->getCompileUnit())->getID();
+  unsigned FileID = GetOrCreateSourceID(V->getContext().getDirectory(),
+                                        V->getContext().getFilename());
   assert(FileID && "Invalid file id");
   addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
   addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
@@ -431,7 +432,8 @@
     return;
 
   unsigned Line = G->getLineNumber();
-  unsigned FileID = findCompileUnit(G->getCompileUnit())->getID();
+  unsigned FileID = GetOrCreateSourceID(G->getContext().getDirectory(),
+                                        G->getContext().getFilename());
   assert(FileID && "Invalid file id");
   addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
   addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
@@ -447,9 +449,11 @@
   if (SP->getLineNumber() == 0)
     return;
 
-
   unsigned Line = SP->getLineNumber();
-  unsigned FileID = findCompileUnit(SP->getCompileUnit())->getID();
+  if (!SP->getContext().Verify())
+    return;
+  unsigned FileID = GetOrCreateSourceID(SP->getContext().getDirectory(),
+                                        SP->getContext().getFilename());
   assert(FileID && "Invalid file id");
   addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
   addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
@@ -464,7 +468,10 @@
     return;
 
   unsigned Line = Ty->getLineNumber();
-  unsigned FileID = findCompileUnit(CU)->getID();
+  if (!Ty->getContext().Verify())
+    return;
+  unsigned FileID = GetOrCreateSourceID(Ty->getContext().getDirectory(),
+                                        Ty->getContext().getFilename());
   assert(FileID && "Invalid file id");
   addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
   addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
@@ -1233,16 +1240,6 @@
   return SPDie;
 }
 
-/// findCompileUnit - Get the compile unit for the given descriptor.
-///
-CompileUnit *DwarfDebug::findCompileUnit(DICompileUnit Unit) {
-  DenseMap<Value *, CompileUnit *>::const_iterator I =
-    CompileUnitMap.find(Unit.getNode());
-  if (I == CompileUnitMap.end())
-    return constructCompileUnit(Unit.getNode());
-  return I->second;
-}
-
 /// getUpdatedDbgScope - Find or create DbgScope assicated with the instruction.
 /// Initialize scope and update scope hierarchy.
 DbgScope *DwarfDebug::getUpdatedDbgScope(MDNode *N, const MachineInstr *MI,
@@ -1676,8 +1673,6 @@
     ModuleCU = Unit;
   }
 
-  CompileUnitMap[DIUnit.getNode()] = Unit;
-  CompileUnits.push_back(Unit);
   return Unit;
 }
 
@@ -1783,17 +1778,8 @@
          E = DbgFinder.compile_unit_end(); I != E; ++I)
     constructCompileUnit(*I);
 
-  if (CompileUnits.empty()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
-
-    return;
-  }
-
-  // If main compile unit for this module is not seen than randomly
-  // select first compile unit.
   if (!ModuleCU)
-    ModuleCU = CompileUnits[0];
+    return;
 
   // Create DIEs for each subprogram.
   for (DebugInfoFinder::iterator I = DbgFinder.subprogram_begin(),

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97990&r1=97989&r2=97990&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 16:02:50 2010
@@ -62,14 +62,6 @@
   // Attributes used to construct specific Dwarf sections.
   //
 
-  /// CompileUnitMap - A map of global variables representing compile units to
-  /// compile units.
-  DenseMap<Value *, CompileUnit *> CompileUnitMap;
-
-  /// CompileUnits - All the compile units in this module.
-  ///
-  SmallVector<CompileUnit *, 8> CompileUnits;
-
   /// ModuleCU - All DIEs are inserted in ModuleCU.
   CompileUnit *ModuleCU;
 
@@ -357,10 +349,6 @@
   /// createSubprogramDIE - Create new DIE using SP.
   DIE *createSubprogramDIE(const DISubprogram &SP, bool MakeDecl = false);
 
-  /// findCompileUnit - Get the compile unit for the given descriptor. 
-  ///
-  CompileUnit *findCompileUnit(DICompileUnit Unit);
-
   /// getUpdatedDbgScope - Find or create DbgScope assicated with 
   /// the instruction. Initialize scope and update scope hierarchy.
   DbgScope *getUpdatedDbgScope(MDNode *N, const MachineInstr *MI, MDNode *InlinedAt);




From daniel at zuster.org  Mon Mar  8 16:03:42 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 08 Mar 2010 22:03:42 -0000
Subject: [llvm-commits] [llvm] r97991 - in /llvm/trunk:
 lib/MC/MCAssembler.cpp test/MC/MachO/zerofill-4.s
 test/MC/MachO/zerofill-sect-align.s
Message-ID: <20100308220342.DB6152A6C12D@llvm.org>

Author: ddunbar
Date: Mon Mar  8 16:03:42 2010
New Revision: 97991

URL: http://llvm.org/viewvc/llvm-project?rev=97991&view=rev
Log:
MC/Macho-O: Align the zerofill section itself to the maximum alignment.

Added:
    llvm/trunk/test/MC/MachO/zerofill-sect-align.s
Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp
    llvm/trunk/test/MC/MachO/zerofill-4.s

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=97991&r1=97990&r2=97991&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 16:03:42 2010
@@ -1279,9 +1279,19 @@
     if (!isVirtualSection(SD.getSection()))
       continue;
 
+    // Align this section if necessary by adding padding bytes to the previous
+    // section.
+    if (uint64_t Pad = OffsetToAlignment(Address, it->getAlignment())) {
+      assert(Prev && "Missing prev section!");
+      Prev->setFileSize(Prev->getFileSize() + Pad);
+      Address += Pad;
+    }
+
     SD.setAddress(Address);
     LayoutSection(SD);
     Address += SD.getSize();
+
+    Prev = &SD;
   }
 
   DEBUG_WITH_TYPE("mc-dump", {

Modified: llvm/trunk/test/MC/MachO/zerofill-4.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/zerofill-4.s?rev=97991&r1=97990&r2=97991&view=diff
==============================================================================
--- llvm/trunk/test/MC/MachO/zerofill-4.s (original)
+++ llvm/trunk/test/MC/MachO/zerofill-4.s Mon Mar  8 16:03:42 2010
@@ -1,4 +1,4 @@
-// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump --dump-section-data | FileCheck %s
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump | FileCheck %s
 
 .zerofill __DATA,__bss,_fill0,1,0
 .zerofill __DATA,__bss,_a,4,2

Added: llvm/trunk/test/MC/MachO/zerofill-sect-align.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/zerofill-sect-align.s?rev=97991&view=auto
==============================================================================
--- llvm/trunk/test/MC/MachO/zerofill-sect-align.s (added)
+++ llvm/trunk/test/MC/MachO/zerofill-sect-align.s Mon Mar  8 16:03:42 2010
@@ -0,0 +1,15 @@
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump | FileCheck %s
+//
+// Check that the section itself is aligned.
+
+        .byte 0
+        
+.zerofill __DATA,__bss,_a,1,0
+.zerofill __DATA,__bss,_b,4,4
+
+// CHECK: # Symbol 0
+// CHECK: ('n_value', 16)
+// CHECK: ('_string', '_a')
+// CHECK: # Symbol 1
+// CHECK: ('n_value', 32)
+// CHECK: ('_string', '_b')




From enderby at apple.com  Mon Mar  8 16:17:27 2010
From: enderby at apple.com (Kevin Enderby)
Date: Mon, 08 Mar 2010 22:17:27 -0000
Subject: [llvm-commits] [llvm] r97992 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td test/MC/Disassembler/simple-tests.txt
Message-ID: <20100308221727.361572A6C12D@llvm.org>

Author: enderby
Date: Mon Mar  8 16:17:26 2010
New Revision: 97992

URL: http://llvm.org/viewvc/llvm-project?rev=97992&view=rev
Log:
Fix the vmxon entry in the X86InstrInfo.td so it has the correct prefix bytes
for the encoding and is not the same as vmptrld.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/test/MC/Disassembler/simple-tests.txt

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=97992&r1=97991&r2=97992&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Mar  8 16:17:26 2010
@@ -4252,7 +4252,7 @@
 // 0F 01 C4
 def VMXOFF : I<0x01, MRM_C4, (outs), (ins), "vmxoff", []>, TB;
 def VMXON : I<0xC7, MRM6m, (outs), (ins i64mem:$vmxon),
-  "vmxon\t{$vmxon}", []>, XD;
+  "vmxon\t{$vmxon}", []>, XS;
 
 //===----------------------------------------------------------------------===//
 // Non-Instruction Patterns

Modified: llvm/trunk/test/MC/Disassembler/simple-tests.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/Disassembler/simple-tests.txt?rev=97992&r1=97991&r2=97992&view=diff
==============================================================================
--- llvm/trunk/test/MC/Disassembler/simple-tests.txt (original)
+++ llvm/trunk/test/MC/Disassembler/simple-tests.txt Mon Mar  8 16:17:26 2010
@@ -41,4 +41,13 @@
 0x0f 0x01 0xf8
 
 # CHECK: rdtscp
-0x0f 0x01 0xf9
\ No newline at end of file
+0x0f 0x01 0xf9
+
+# CHECK: vmxon
+0xf3 0x0f 0xc7 0x30
+
+# CHECK: vmptrld
+0x0f 0xc7 0x30
+
+# CHECK: vmptrst
+0x0f 0xc7 0x38




From sabre at nondot.org  Mon Mar  8 16:23:36 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:23:36 -0000
Subject: [llvm-commits] [llvm] r97993 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: CMakeLists.txt DIE.cpp DIE.h
 DwarfDebug.cpp DwarfDebug.h DwarfException.cpp DwarfLabel.cpp DwarfLabel.h
 DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100308222336.E3BD02A6C12D@llvm.org>

Author: lattner
Date: Mon Mar  8 16:23:36 2010
New Revision: 97993

URL: http://llvm.org/viewvc/llvm-project?rev=97993&view=rev
Log:
elimiante the DWLabel class, using MCSymbol instead.  Start 
switching some stuff over to passing around MCSymbol* instead
of stem+ID.


Removed:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.h
Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt Mon Mar  8 16:23:36 2010
@@ -3,7 +3,6 @@
   DIE.cpp
   DwarfDebug.cpp
   DwarfException.cpp
-  DwarfLabel.cpp
   DwarfPrinter.cpp
   DwarfWriter.cpp
   OcamlGCPrinter.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp Mon Mar  8 16:23:36 2010
@@ -273,8 +273,7 @@
 
 #ifndef NDEBUG
 void DIEDwarfLabel::print(raw_ostream &O) {
-  O << "Lbl: ";
-  Label.print(O);
+  O << "Lbl: " << Label->getName();
 }
 #endif
 
@@ -310,9 +309,7 @@
 ///
 void DIESectionOffset::EmitValue(DwarfPrinter *D, unsigned Form) const {
   bool IsSmall = Form == dwarf::DW_FORM_data4;
-  D->EmitSectionOffset(Label.getTag(), Section.getTag(),
-                       Label.getNumber(), Section.getNumber(),
-                       IsSmall, IsEH, UseSet);
+  D->EmitSectionOffset(Label, Section, IsSmall, IsEH, UseSet);
   D->getAsm()->O << '\n'; // FIXME: Necesssary?
 }
 
@@ -325,11 +322,8 @@
 
 #ifndef NDEBUG
 void DIESectionOffset::print(raw_ostream &O) {
-  O << "Off: ";
-  Label.print(O);
-  O << "-";
-  Section.print(O);
-  O << "-" << IsEH << "-" << UseSet;
+  O << "Off: " << Label->getName() << "-" << Section->getName()
+    << "-" << IsEH << "-" << UseSet;
 }
 #endif
 
@@ -353,10 +347,7 @@
 
 #ifndef NDEBUG
 void DIEDelta::print(raw_ostream &O) {
-  O << "Del: ";
-  LabelHi.print(O);
-  O << "-";
-  LabelLo.print(O);
+  O << "Del: " << LabelHi->getName() << "-" << LabelLo->getName();
 }
 #endif
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h Mon Mar  8 16:23:36 2010
@@ -14,7 +14,6 @@
 #ifndef CODEGEN_ASMPRINTER_DIE_H__
 #define CODEGEN_ASMPRINTER_DIE_H__
 
-#include "DwarfLabel.h"
 #include "llvm/ADT/FoldingSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/Compiler.h"
@@ -26,6 +25,7 @@
   class DwarfPrinter;
   class TargetData;
   class MCSymbol;
+  class raw_ostream;
 
   //===--------------------------------------------------------------------===//
   /// DIEAbbrevData - Dwarf abbreviation data, describes the one attribute of a
@@ -308,10 +308,11 @@
   //===--------------------------------------------------------------------===//
   /// DIEDwarfLabel - A Dwarf internal label expression DIE.
   //
+  /// FIXME: Merge into DIEObjectLabel.
   class DIEDwarfLabel : public DIEValue {
-    const DWLabel Label;
+    const MCSymbol *Label;
   public:
-    explicit DIEDwarfLabel(const DWLabel &L) : DIEValue(isLabel), Label(L) {}
+    explicit DIEDwarfLabel(const MCSymbol *L) : DIEValue(isLabel), Label(L) {}
 
     /// EmitValue - Emit label value.
     ///
@@ -362,12 +363,12 @@
   /// DIESectionOffset - A section offset DIE.
   ///
   class DIESectionOffset : public DIEValue {
-    const DWLabel Label;
-    const DWLabel Section;
+    const MCSymbol *Label;
+    const MCSymbol *Section;
     bool IsEH : 1;
     bool UseSet : 1;
   public:
-    DIESectionOffset(const DWLabel &Lab, const DWLabel &Sec,
+    DIESectionOffset(const MCSymbol *Lab, const MCSymbol *Sec,
                      bool isEH = false, bool useSet = true)
       : DIEValue(isSectionOffset), Label(Lab), Section(Sec),
         IsEH(isEH), UseSet(useSet) {}
@@ -395,10 +396,10 @@
   /// DIEDelta - A simple label difference DIE.
   ///
   class DIEDelta : public DIEValue {
-    const DWLabel LabelHi;
-    const DWLabel LabelLo;
+    const MCSymbol *LabelHi;
+    const MCSymbol *LabelLo;
   public:
-    DIEDelta(const DWLabel &Hi, const DWLabel &Lo)
+    DIEDelta(const MCSymbol *Hi, const MCSymbol *Lo)
       : DIEValue(isDelta), LabelHi(Hi), LabelLo(Lo) {}
 
     /// EmitValue - Emit delta value.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:23:36 2010
@@ -366,7 +366,8 @@
 /// addLabel - Add a Dwarf label attribute data and value.
 ///
 void DwarfDebug::addLabel(DIE *Die, unsigned Attribute, unsigned Form,
-                          const DWLabel &Label) {
+                          const MCSymbol *Label) {
+  // FIXME: Merge into DIEObjectLabel?
   DIEValue *Value = new DIEDwarfLabel(Label);
   DIEValues.push_back(Value);
   Die->addValue(Attribute, Form, Value);
@@ -384,7 +385,7 @@
 /// addSectionOffset - Add a section offset label attribute data and value.
 ///
 void DwarfDebug::addSectionOffset(DIE *Die, unsigned Attribute, unsigned Form,
-                                  const DWLabel &Label, const DWLabel &Section,
+                                  const MCSymbol *Label,const MCSymbol *Section,
                                   bool isEH, bool useSet) {
   DIEValue *Value = new DIESectionOffset(Label, Section, isEH, useSet);
   DIEValues.push_back(Value);
@@ -394,7 +395,7 @@
 /// addDelta - Add a label delta attribute data and value.
 ///
 void DwarfDebug::addDelta(DIE *Die, unsigned Attribute, unsigned Form,
-                          const DWLabel &Hi, const DWLabel &Lo) {
+                          const MCSymbol *Hi, const MCSymbol *Lo) {
   DIEValue *Value = new DIEDelta(Hi, Lo);
   DIEValues.push_back(Value);
   Die->addValue(Attribute, Form, Value);
@@ -1355,9 +1356,9 @@
  }
 
  addLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-          DWLabel("func_begin", SubprogramCount));
+          getDWLabel("func_begin", SubprogramCount));
  addLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-          DWLabel("func_end", SubprogramCount));
+          getDWLabel("func_end", SubprogramCount));
  MachineLocation Location(RI->getFrameRegister(*MF));
  addAddress(SPDie, dwarf::DW_AT_frame_base, Location);
 
@@ -1382,15 +1383,11 @@
     return ScopeDIE;
 
   addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           StartID ?
-             DWLabel("label", StartID)
-           : DWLabel("func_begin", SubprogramCount));
+           StartID ? getDWLabel("label", StartID)
+                   : getDWLabel("func_begin", SubprogramCount));
   addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           EndID ?
-             DWLabel("label", EndID)
-           : DWLabel("func_end", SubprogramCount));
-
-
+           EndID ? getDWLabel("label", EndID)
+                 : getDWLabel("func_end", SubprogramCount));
 
   return ScopeDIE;
 }
@@ -1418,9 +1415,9 @@
               dwarf::DW_FORM_ref4, OriginDIE);
 
   addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           DWLabel("label", StartID));
+           getDWLabel("label", StartID));
   addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           DWLabel("label", EndID));
+           getDWLabel("label", EndID));
 
   InlinedSubprogramDIEs.insert(OriginDIE);
 
@@ -1643,8 +1640,9 @@
   unsigned ID = GetOrCreateSourceID(Dir, FN);
 
   DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
+  // FIXME: Why getting the delta between two identical labels??
   addSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
-                   DWLabel("section_line", 0), DWLabel("section_line", 0),
+                   getTempLabel("section_line"), getTempLabel("section_line"),
                    false);
   addString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
             DIUnit.getProducer());
@@ -2445,7 +2443,8 @@
 
   Asm->EmitInt32(ContentSize);  EOL("Length of Compilation Unit Info");
   Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF version number");
-  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
+  EmitSectionOffset(getTempLabel("abbrev_begin"),getTempLabel("section_abbrev"),
+                    true, false);
   EOL("Offset Into Abbrev. Section");
   Asm->EmitInt8(TD->getPointerSize()); EOL("Address Size (in bytes)");
 
@@ -2726,8 +2725,8 @@
 
   EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
 
-  EmitSectionOffset("debug_frame_common", "section_debug_frame",
-                    0, 0, true, false);
+  EmitSectionOffset(getTempLabel("debug_frame_common"),
+                    getTempLabel("section_debug_frame"), true, false);
   EOL("FDE CIE offset");
 
   EmitReference("func_begin", DebugFrameInfo.Number);
@@ -2759,8 +2758,9 @@
 
   Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF Version");
 
-  EmitSectionOffset("info_begin", "section_info",
-                    ModuleCU->getID(), 0, true, false);
+  EmitSectionOffset(getDWLabel("info_begin", ModuleCU->getID()), 
+                    getTempLabel("section_info"),
+                    true, false);
   EOL("Offset of Compilation Unit Info");
 
   EmitDifference("info_end", ModuleCU->getID(), "info_begin", ModuleCU->getID(),
@@ -2798,8 +2798,8 @@
   if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DWARF Version");
   Asm->EmitInt16(dwarf::DWARF_VERSION);
 
-  EmitSectionOffset("info_begin", "section_info",
-                    ModuleCU->getID(), 0, true, false);
+  EmitSectionOffset(getDWLabel("info_begin", ModuleCU->getID()),
+                    getTempLabel("section_info"), true, false);
   EOL("Offset of Compilation ModuleCU Info");
 
   EmitDifference("info_end", ModuleCU->getID(), "info_begin", ModuleCU->getID(),
@@ -2961,12 +2961,13 @@
       Asm->OutStreamer.EmitBytes(Name, 0);
       Asm->OutStreamer.EmitIntValue(0, 1, 0); // nul terminator.
     } else 
-      EmitSectionOffset("string", "section_str",
-                      StringPool.idFor(getRealLinkageName(LName)), false, true);
+      EmitSectionOffset(getDWLabel("string",
+                                   StringPool.idFor(getRealLinkageName(LName))),
+                        getTempLabel("section_str"), true);
 
     EOL("MIPS linkage name");
-    EmitSectionOffset("string", "section_str",
-                      StringPool.idFor(Name), false, true);
+    EmitSectionOffset(getDWLabel("string", StringPool.idFor(Name)),
+                      getTempLabel("section_str"), false, true);
     EOL("Function name");
     EmitULEB128(Labels.size(), "Inline count");
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 16:23:36 2010
@@ -243,7 +243,7 @@
   /// addLabel - Add a Dwarf label attribute data and value.
   ///
   void addLabel(DIE *Die, unsigned Attribute, unsigned Form,
-                const DWLabel &Label);
+                const MCSymbol *Label);
 
   /// addObjectLabel - Add an non-Dwarf label attribute data and value.
   ///
@@ -253,13 +253,13 @@
   /// addSectionOffset - Add a section offset label attribute data and value.
   ///
   void addSectionOffset(DIE *Die, unsigned Attribute, unsigned Form,
-                        const DWLabel &Label, const DWLabel &Section,
+                        const MCSymbol *Label, const MCSymbol *Section,
                         bool isEH = false, bool useSet = true);
 
   /// addDelta - Add a label delta attribute data and value.
   ///
   void addDelta(DIE *Die, unsigned Attribute, unsigned Form,
-                const DWLabel &Hi, const DWLabel &Lo);
+                const MCSymbol *Hi, const MCSymbol *Lo);
 
   /// addDIEEntry - Add a DIE attribute data and value.
   ///

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 16:23:36 2010
@@ -230,8 +230,9 @@
 
     EmitLabel("eh_frame_begin", EHFrameInfo.Number);
 
-    EmitSectionOffset("eh_frame_begin", "eh_frame_common",
-                      EHFrameInfo.Number, EHFrameInfo.PersonalityIndex,
+    EmitSectionOffset(getDWLabel("eh_frame_begin", EHFrameInfo.Number),
+                      getDWLabel("eh_frame_common",
+                                 EHFrameInfo.PersonalityIndex),
                       true, true, false);
 
     EOL("FDE CIE offset");
@@ -819,12 +820,14 @@
       // Offset of the call site relative to the previous call site, counted in
       // number of 16-byte bundles. The first call site is counted relative to
       // the start of the procedure fragment.
-      EmitSectionOffset(BeginTag, "eh_func_begin", BeginNumber, SubprogramCount,
+      EmitSectionOffset(getDWLabel(BeginTag, BeginNumber),
+                        getDWLabel("eh_func_begin", SubprogramCount),
                         true, true);
       EOL("Region start");
 
       if (!S.EndLabel)
-        EmitDifference("eh_func_end", SubprogramCount, BeginTag, BeginNumber,
+        EmitDifference(getDWLabel("eh_func_end", SubprogramCount),
+                       getDWLabel(BeginTag, BeginNumber),
                        true);
       else
         EmitDifference("label", S.EndLabel, BeginTag, BeginNumber, true);
@@ -837,7 +840,8 @@
         Asm->OutStreamer.AddComment("Landing pad");
         Asm->OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
       } else {
-        EmitSectionOffset("label", "eh_func_begin", S.PadLabel, SubprogramCount,
+        EmitSectionOffset(getDWLabel("label", S.PadLabel),
+                          getDWLabel("eh_func_begin", SubprogramCount),
                           true, true);
         EOL("Landing pad");
       }

Removed: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.cpp?rev=97992&view=auto
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.cpp (removed)
@@ -1,32 +0,0 @@
-//===--- lib/CodeGen/DwarfLabel.cpp - Dwarf Label -------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// DWARF Labels
-// 
-//===----------------------------------------------------------------------===//
-
-#include "DwarfLabel.h"
-#include "llvm/ADT/FoldingSet.h"
-#include "llvm/Support/raw_ostream.h"
-
-using namespace llvm;
-
-/// Profile - Used to gather unique data for the folding set.
-///
-void DWLabel::Profile(FoldingSetNodeID &ID) const {
-  ID.AddString(Tag);
-  ID.AddInteger(Number);
-}
-
-#ifndef NDEBUG
-void DWLabel::print(raw_ostream &O) const {
-  O << "." << Tag;
-  if (Number) O << Number;
-}
-#endif

Removed: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.h?rev=97992&view=auto
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfLabel.h (removed)
@@ -1,52 +0,0 @@
-//===--- lib/CodeGen/DwarfLabel.h - Dwarf Label -----------------*- C++ -*-===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// DWARF Labels.
-// 
-//===----------------------------------------------------------------------===//
-
-#ifndef CODEGEN_ASMPRINTER_DWARFLABEL_H__
-#define CODEGEN_ASMPRINTER_DWARFLABEL_H__
-
-namespace llvm {
-  class FoldingSetNodeID;
-  class raw_ostream;
-
-  //===--------------------------------------------------------------------===//
-  /// DWLabel - Labels are used to track locations in the assembler file.
-  /// Labels appear in the form @verbatim <prefix><Tag><Number> @endverbatim,
-  /// where the tag is a category of label (Ex. location) and number is a value
-  /// unique in that category.
-  class DWLabel {
-    /// Tag - Label category tag. Should always be a statically declared C
-    /// string.
-    /// 
-    const char *Tag;
-
-    /// Number - Value to make label unique.
-    /// 
-    unsigned Number;
-  public:
-    DWLabel(const char *T, unsigned N) : Tag(T), Number(N) {}
-
-    // Accessors.
-    const char *getTag() const { return Tag; }
-    unsigned getNumber() const { return Number; }
-
-    /// Profile - Used to gather unique data for the folding set.
-    ///
-    void Profile(FoldingSetNodeID &ID) const;
-
-#ifndef NDEBUG
-    void print(raw_ostream &O) const;
-#endif
-  };
-} // end llvm namespace
-
-#endif

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 16:23:36 2010
@@ -37,6 +37,27 @@
   RI(Asm->TM.getRegisterInfo()), M(NULL), MF(NULL), MMI(NULL),
   SubprogramCount(0), Flavor(flavor), SetCounter(1) {}
 
+
+/// getDWLabel - Return the MCSymbol corresponding to the assembler temporary
+/// label with the specified stem and unique ID.
+MCSymbol *DwarfPrinter::getDWLabel(const char *Name, unsigned ID) const {
+  // FIXME: REMOVE this.  However, there is stuff in EH that passes counters in
+  // here that can be zero.
+  
+  //assert(ID && "Should use getTempLabel if no ID");
+  if (ID == 0) return getTempLabel(Name);
+  return Asm->OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix())
+                                           + Twine(Name) + Twine(ID));
+}
+
+/// getTempLabel - Return the MCSymbol corresponding to the assembler temporary
+/// label with the specified name.
+MCSymbol *DwarfPrinter::getTempLabel(const char *Name) const {
+  return Asm->OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix())
+                                           + Name);
+}
+
+
 /// SizeOfEncodedValue - Return the size of the encoding in bytes.
 unsigned DwarfPrinter::SizeOfEncodedValue(unsigned Encoding) const {
   if (Encoding == dwarf::DW_EH_PE_omit)
@@ -193,12 +214,19 @@
 
 /// PrintLabelName - Print label name in form used by Dwarf writer.
 ///
+void DwarfPrinter::PrintLabelName(const MCSymbol *Label) const {
+  // FIXME: REMOVE.
+  O << Label->getName();
+}
+
 void DwarfPrinter::PrintLabelName(const char *Tag, unsigned Number) const {
+  // FIXME: REMOVE.
   O << MAI->getPrivateGlobalPrefix() << Tag;
   if (Number) O << Number;
 }
 void DwarfPrinter::PrintLabelName(const char *Tag, unsigned Number,
                                   const char *Suffix) const {
+  // FIXME: REMOVE.
   O << MAI->getPrivateGlobalPrefix() << Tag;
   if (Number) O << Number;
   O << Suffix;
@@ -207,6 +235,7 @@
 /// EmitLabel - Emit location label for internal use by Dwarf.
 ///
 void DwarfPrinter::EmitLabel(const char *Tag, unsigned Number) const {
+  // FIXME: REMOVE.
   PrintLabelName(Tag, Number);
   O << ":\n";
 }
@@ -235,12 +264,8 @@
 
 void DwarfPrinter::EmitReference(const char *Tag, unsigned Number,
                                  unsigned Encoding) const {
-  SmallString<64> Name;
-  raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix()
-                            << Tag << Number;
-
-  MCSymbol *Sym = Asm->OutContext.GetOrCreateSymbol(Name.str());
-  EmitReference(Sym, Encoding);
+  // FIXME: REMOVE.
+  EmitReference(getDWLabel(Tag, Number), Encoding);
 }
 
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
@@ -260,6 +285,31 @@
 
 /// EmitDifference - Emit the difference between two labels.  If this assembler
 /// supports .set, we emit a .set of a temporary and then use it in the .word.
+void DwarfPrinter::EmitDifference(const MCSymbol *TagHi, const MCSymbol *TagLo,
+                                  bool IsSmall) {
+  if (MAI->hasSetDirective()) {
+    // FIXME: switch to OutStreamer.EmitAssignment.
+    O << "\t.set\t";
+    PrintLabelName("set", SetCounter, Flavor);
+    O << ",";
+    PrintLabelName(TagHi);
+    O << "-";
+    PrintLabelName(TagLo);
+    O << "\n";
+
+    PrintRelDirective(IsSmall);
+    PrintLabelName("set", SetCounter, Flavor);
+    ++SetCounter;
+  } else {
+    PrintRelDirective(IsSmall);
+    PrintLabelName(TagHi);
+    O << "-";
+    PrintLabelName(TagLo);
+  }
+}
+
+/// EmitDifference - Emit the difference between two labels.  If this assembler
+/// supports .set, we emit a .set of a temporary and then use it in the .word.
 void DwarfPrinter::EmitDifference(const char *TagHi, unsigned NumberHi,
                                   const char *TagLo, unsigned NumberLo,
                                   bool IsSmall) {
@@ -272,7 +322,7 @@
     O << "-";
     PrintLabelName(TagLo, NumberLo);
     O << "\n";
-
+    
     PrintRelDirective(IsSmall);
     PrintLabelName("set", SetCounter, Flavor);
     ++SetCounter;
@@ -284,9 +334,8 @@
   }
 }
 
-void DwarfPrinter::EmitSectionOffset(const char* Label, const char* Section,
-                                     unsigned LabelNumber,
-                                     unsigned SectionNumber,
+void DwarfPrinter::EmitSectionOffset(const MCSymbol *Label,
+                                     const MCSymbol *Section,
                                      bool IsSmall, bool isEH,
                                      bool useSet) {
   bool printAbsolute = false;
@@ -300,11 +349,11 @@
     O << "\t.set\t";
     PrintLabelName("set", SetCounter, Flavor);
     O << ",";
-    PrintLabelName(Label, LabelNumber);
+    PrintLabelName(Label);
 
     if (!printAbsolute) {
       O << "-";
-      PrintLabelName(Section, SectionNumber);
+      PrintLabelName(Section);
     }
 
     O << "\n";
@@ -313,11 +362,11 @@
     ++SetCounter;
   } else {
     PrintRelDirective(IsSmall, true);
-    PrintLabelName(Label, LabelNumber);
+    PrintLabelName(Label);
 
     if (!printAbsolute) {
       O << "-";
-      PrintLabelName(Section, SectionNumber);
+      PrintLabelName(Section);
     }
   }
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97993&r1=97992&r2=97993&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:23:36 2010
@@ -14,7 +14,6 @@
 #ifndef CODEGEN_ASMPRINTER_DWARFPRINTER_H__
 #define CODEGEN_ASMPRINTER_DWARFPRINTER_H__
 
-#include "DwarfLabel.h"
 #include "llvm/CodeGen/MachineLocation.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/FormattedStream.h"
@@ -86,6 +85,14 @@
   const MCAsmInfo *getMCAsmInfo() const { return MAI; }
   const TargetData *getTargetData() const { return TD; }
 
+  /// getDWLabel - Return the MCSymbol corresponding to the assembler temporary
+  /// label with the specified stem and unique ID.
+  MCSymbol *getDWLabel(const char *Name, unsigned ID) const;
+  
+  /// getTempLabel - Return an assembler temporary label with the specified
+  /// name.
+  MCSymbol *getTempLabel(const char *Name) const;
+
   /// SizeOfEncodedValue - Return the size of the encoding in bytes.
   unsigned SizeOfEncodedValue(unsigned Encoding) const;
 
@@ -117,54 +124,47 @@
   
   /// PrintLabelName - Print label name in form used by Dwarf writer.
   ///
-  void PrintLabelName(const DWLabel &Label) const {
-    PrintLabelName(Label.getTag(), Label.getNumber());
-  }
+  void PrintLabelName(const MCSymbol *Label) const;
   void PrintLabelName(const char *Tag, unsigned Number) const;
   void PrintLabelName(const char *Tag, unsigned Number,
                       const char *Suffix) const;
 
   /// EmitLabel - Emit location label for internal use by Dwarf.
   ///
-  void EmitLabel(const DWLabel &Label) const {
-    EmitLabel(Label.getTag(), Label.getNumber());
-  }
+  void EmitLabel(const MCSymbol *Label) const;
   void EmitLabel(const char *Tag, unsigned Number) const;
 
   /// EmitReference - Emit a reference to a label.
   ///
-  void EmitReference(const DWLabel &Label, bool IsPCRelative = false,
-                     bool Force32Bit = false) const {
-    EmitReference(Label.getTag(), Label.getNumber(),
-                  IsPCRelative, Force32Bit);
-  }
+  void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,
+                     bool Force32Bit = false) const;
   void EmitReference(const char *Tag, unsigned Number,
                      bool IsPCRelative = false,
                      bool Force32Bit = false) const;
   void EmitReference(const std::string &Name, bool IsPCRelative = false,
                      bool Force32Bit = false) const;
-  void EmitReference(const MCSymbol *Sym, bool IsPCRelative = false,
-                     bool Force32Bit = false) const;
 
   void EmitReference(const char *Tag, unsigned Number, unsigned Encoding) const;
   void EmitReference(const MCSymbol *Sym, unsigned Encoding) const;
   void EmitReference(const GlobalValue *GV, unsigned Encoding) const;
 
   /// EmitDifference - Emit the difference between two labels.
-  void EmitDifference(const DWLabel &LabelHi, const DWLabel &LabelLo,
-                      bool IsSmall = false) {
-    EmitDifference(LabelHi.getTag(), LabelHi.getNumber(),
-                   LabelLo.getTag(), LabelLo.getNumber(),
-                   IsSmall);
-  }
+  void EmitDifference(const MCSymbol *LabelHi, const MCSymbol *LabelLo,
+                      bool IsSmall = false);
   void EmitDifference(const char *TagHi, unsigned NumberHi,
                       const char *TagLo, unsigned NumberLo,
                       bool IsSmall = false);
 
+  void EmitSectionOffset(const MCSymbol *Label, const MCSymbol *Section,
+                         bool IsSmall = false, bool isEH = false,
+                         bool useSet = true);
+  
+#if 0
   void EmitSectionOffset(const char* Label, const char* Section,
                          unsigned LabelNumber, unsigned SectionNumber,
                          bool IsSmall = false, bool isEH = false,
                          bool useSet = true);
+#endif
 
   /// EmitFrameMoves - Emit frame instructions to describe the layout of the
   /// frame.




From dpatel at apple.com  Mon Mar  8 16:27:22 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 08 Mar 2010 22:27:22 -0000
Subject: [llvm-commits] [llvm] r97994 - in /llvm/trunk:
 include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
Message-ID: <20100308222722.7DC7E2A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 16:27:22 2010
New Revision: 97994

URL: http://llvm.org/viewvc/llvm-project?rev=97994&view=rev
Log:
Introduce DIFile. This will be used to represent header files and source file(s) in debug info.

Modified:
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97994&r1=97993&r2=97994&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 16:27:22 2010
@@ -87,6 +87,7 @@
     bool isSubprogram() const;
     bool isGlobalVariable() const;
     bool isScope() const;
+    bool isFile() const;
     bool isCompileUnit() const;
     bool isNameSpace() const;
     bool isLexicalBlock() const;
@@ -158,6 +159,18 @@
     void dump() const;
   };
 
+  /// DIFile - This is a wrapper for a file.
+  class DIFile : public DIScope {
+  public:
+    explicit DIFile(MDNode *N = 0) : DIScope(N) {
+      if (DbgNode && !isFile())
+        DbgNode = 0;
+    }
+    StringRef getFilename() const  { return getStringField(1);   }
+    StringRef getDirectory() const { return getStringField(2);   }
+    DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
+  };
+
   /// DIEnumerator - A wrapper for an enumerator (e.g. X and Y in 'enum {X,Y}').
   /// FIXME: it seems strange that this doesn't have either a reference to the
   /// type/precision or a file/line pair for location info.
@@ -502,6 +515,9 @@
                                     StringRef Flags = "",
                                     unsigned RunTimeVer = 0);
 
+    /// CreateFile -  Create a new descriptor for the specified file.
+    DIFile CreateFile(StringRef Filename, StringRef Directory, DICompileUnit CU);
+
     /// CreateEnumerator - Create a single enumerator value.
     DIEnumerator CreateEnumerator(StringRef Name, uint64_t Val);
 

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97994&r1=97993&r2=97994&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 16:27:22 2010
@@ -219,6 +219,11 @@
   return DbgNode && getTag() == dwarf::DW_TAG_compile_unit;
 }
 
+/// isFile - Return true if the specified tag is DW_TAG_file_type.
+bool DIDescriptor::isFile() const {
+  return DbgNode && getTag() == dwarf::DW_TAG_file_type;
+}
+
 /// isNameSpace - Return true if the specified tag is DW_TAG_namespace.
 bool DIDescriptor::isNameSpace() const {
   return DbgNode && getTag() == dwarf::DW_TAG_namespace;
@@ -424,6 +429,8 @@
     return DINameSpace(DbgNode).getFilename();
   if (isType())
     return DIType(DbgNode).getFilename();
+  if (isFile())
+    return DIFile(DbgNode).getFilename();
   assert(0 && "Invalid DIScope!");
   return StringRef();
 }
@@ -441,6 +448,8 @@
     return DINameSpace(DbgNode).getDirectory();
   if (isType())
     return DIType(DbgNode).getDirectory();
+  if (isFile())
+    return DIFile(DbgNode).getDirectory();
   assert(0 && "Invalid DIScope!");
   return StringRef();
 }
@@ -661,6 +670,20 @@
   return DICompileUnit(MDNode::get(VMContext, &Elts[0], 10));
 }
 
+/// CreateFile -  Create a new descriptor for the specified file.
+DIFile DIFactory::CreateFile(StringRef Filename,
+                             StringRef Directory,
+                             DICompileUnit CU) {
+  Value *Elts[] = {
+    GetTagConstant(dwarf::DW_TAG_file_type),
+    MDString::get(VMContext, Filename),
+    MDString::get(VMContext, Directory),
+    CU.getNode()
+  };
+
+  return DIFile(MDNode::get(VMContext, &Elts[0], 4));
+}
+
 /// CreateEnumerator - Create a single enumerator value.
 DIEnumerator DIFactory::CreateEnumerator(StringRef Name, uint64_t Val){
   Value *Elts[] = {




From sabre at nondot.org  Mon Mar  8 16:31:46 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:31:46 -0000
Subject: [llvm-commits] [llvm] r97995 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DIE.cpp DIE.h DwarfDebug.cpp
 DwarfDebug.h DwarfPrinter.h
Message-ID: <20100308223146.5D36F2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 16:31:46 2010
New Revision: 97995

URL: http://llvm.org/viewvc/llvm-project?rev=97995&view=rev
Log:
merge DIEObjectLabel and DIEDwarfLabel into DIELabel.

Yes, DIE you fiendish labels, die all of you.


Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp?rev=97995&r1=97994&r2=97995&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp Mon Mar  8 16:31:46 2010
@@ -254,54 +254,30 @@
 #endif
 
 //===----------------------------------------------------------------------===//
-// DIEDwarfLabel Implementation
+// DIELabel Implementation
 //===----------------------------------------------------------------------===//
 
 /// EmitValue - Emit label value.
 ///
-void DIEDwarfLabel::EmitValue(DwarfPrinter *D, unsigned Form) const {
+void DIELabel::EmitValue(DwarfPrinter *D, unsigned Form) const {
   bool IsSmall = Form == dwarf::DW_FORM_data4;
   D->EmitReference(Label, false, IsSmall);
 }
 
 /// SizeOf - Determine size of label value in bytes.
 ///
-unsigned DIEDwarfLabel::SizeOf(const TargetData *TD, unsigned Form) const {
+unsigned DIELabel::SizeOf(const TargetData *TD, unsigned Form) const {
   if (Form == dwarf::DW_FORM_data4) return 4;
   return TD->getPointerSize();
 }
 
 #ifndef NDEBUG
-void DIEDwarfLabel::print(raw_ostream &O) {
+void DIELabel::print(raw_ostream &O) {
   O << "Lbl: " << Label->getName();
 }
 #endif
 
 //===----------------------------------------------------------------------===//
-// DIEObjectLabel Implementation
-//===----------------------------------------------------------------------===//
-
-/// EmitValue - Emit label value.
-///
-void DIEObjectLabel::EmitValue(DwarfPrinter *D, unsigned Form) const {
-  bool IsSmall = Form == dwarf::DW_FORM_data4;
-  D->EmitReference(Sym, false, IsSmall);
-}
-
-/// SizeOf - Determine size of label value in bytes.
-///
-unsigned DIEObjectLabel::SizeOf(const TargetData *TD, unsigned Form) const {
-  if (Form == dwarf::DW_FORM_data4) return 4;
-  return TD->getPointerSize();
-}
-
-#ifndef NDEBUG
-void DIEObjectLabel::print(raw_ostream &O) {
-  O << "Obj: " << Sym->getName();
-}
-#endif
-
-//===----------------------------------------------------------------------===//
 // DIESectionOffset Implementation
 //===----------------------------------------------------------------------===//
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h?rev=97995&r1=97994&r2=97995&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h Mon Mar  8 16:31:46 2010
@@ -203,7 +203,6 @@
       isInteger,
       isString,
       isLabel,
-      isAsIsLabel,
       isSectionOffset,
       isDelta,
       isEntry,
@@ -306,13 +305,12 @@
   };
 
   //===--------------------------------------------------------------------===//
-  /// DIEDwarfLabel - A Dwarf internal label expression DIE.
+  /// DIELabel - A label expression DIE.
   //
-  /// FIXME: Merge into DIEObjectLabel.
-  class DIEDwarfLabel : public DIEValue {
+  class DIELabel : public DIEValue {
     const MCSymbol *Label;
   public:
-    explicit DIEDwarfLabel(const MCSymbol *L) : DIEValue(isLabel), Label(L) {}
+    explicit DIELabel(const MCSymbol *L) : DIEValue(isLabel), Label(L) {}
 
     /// EmitValue - Emit label value.
     ///
@@ -323,7 +321,7 @@
     virtual unsigned SizeOf(const TargetData *TD, unsigned Form) const;
 
     // Implement isa/cast/dyncast.
-    static bool classof(const DIEDwarfLabel *)  { return true; }
+    static bool classof(const DIELabel *)  { return true; }
     static bool classof(const DIEValue *L) { return L->getType() == isLabel; }
 
 #ifndef NDEBUG
@@ -332,34 +330,6 @@
   };
 
   //===--------------------------------------------------------------------===//
-  /// DIEObjectLabel - A label to an object in code or data.
-  //
-  class DIEObjectLabel : public DIEValue {
-    const MCSymbol *Sym;
-  public:
-    explicit DIEObjectLabel(const MCSymbol *S)
-      : DIEValue(isAsIsLabel), Sym(S) {}
-
-    /// EmitValue - Emit label value.
-    ///
-    virtual void EmitValue(DwarfPrinter *D, unsigned Form) const;
-
-    /// SizeOf - Determine size of label value in bytes.
-    ///
-    virtual unsigned SizeOf(const TargetData *TD, unsigned Form) const;
-
-    // Implement isa/cast/dyncast.
-    static bool classof(const DIEObjectLabel *) { return true; }
-    static bool classof(const DIEValue *L) {
-      return L->getType() == isAsIsLabel;
-    }
-
-#ifndef NDEBUG
-    virtual void print(raw_ostream &O);
-#endif
-  };
-
-  //===--------------------------------------------------------------------===//
   /// DIESectionOffset - A section offset DIE.
   ///
   class DIESectionOffset : public DIEValue {

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97995&r1=97994&r2=97995&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:31:46 2010
@@ -367,17 +367,7 @@
 ///
 void DwarfDebug::addLabel(DIE *Die, unsigned Attribute, unsigned Form,
                           const MCSymbol *Label) {
-  // FIXME: Merge into DIEObjectLabel?
-  DIEValue *Value = new DIEDwarfLabel(Label);
-  DIEValues.push_back(Value);
-  Die->addValue(Attribute, Form, Value);
-}
-
-/// addObjectLabel - Add an non-Dwarf label attribute data and value.
-///
-void DwarfDebug::addObjectLabel(DIE *Die, unsigned Attribute, unsigned Form,
-                                const MCSymbol *Sym) {
-  DIEValue *Value = new DIEObjectLabel(Sym);
+  DIEValue *Value = new DIELabel(Label);
   DIEValues.push_back(Value);
   Die->addValue(Attribute, Form, Value);
 }
@@ -1704,16 +1694,16 @@
                 dwarf::DW_FORM_ref4, VariableDie);
     DIEBlock *Block = new DIEBlock();
     addUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
-    addObjectLabel(Block, 0, dwarf::DW_FORM_udata,
-                   Asm->GetGlobalValueSymbol(DI_GV.getGlobal()));
+    addLabel(Block, 0, dwarf::DW_FORM_udata,
+             Asm->GetGlobalValueSymbol(DI_GV.getGlobal()));
     addBlock(VariableSpecDIE, dwarf::DW_AT_location, 0, Block);
     addUInt(VariableDie, dwarf::DW_AT_declaration, dwarf::DW_FORM_flag, 1);
     ModuleCU->addDie(VariableSpecDIE);
   } else {
     DIEBlock *Block = new DIEBlock();
     addUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
-    addObjectLabel(Block, 0, dwarf::DW_FORM_udata,
-                   Asm->GetGlobalValueSymbol(DI_GV.getGlobal()));
+    addLabel(Block, 0, dwarf::DW_FORM_udata,
+             Asm->GetGlobalValueSymbol(DI_GV.getGlobal()));
     addBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
   }
   addToContextOwner(VariableDie, GVContext);

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97995&r1=97994&r2=97995&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 16:31:46 2010
@@ -245,11 +245,6 @@
   void addLabel(DIE *Die, unsigned Attribute, unsigned Form,
                 const MCSymbol *Label);
 
-  /// addObjectLabel - Add an non-Dwarf label attribute data and value.
-  ///
-  void addObjectLabel(DIE *Die, unsigned Attribute, unsigned Form,
-                      const MCSymbol *Sym);
-
   /// addSectionOffset - Add a section offset label attribute data and value.
   ///
   void addSectionOffset(DIE *Die, unsigned Attribute, unsigned Form,

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97995&r1=97994&r2=97995&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:31:46 2010
@@ -159,13 +159,6 @@
                          bool IsSmall = false, bool isEH = false,
                          bool useSet = true);
   
-#if 0
-  void EmitSectionOffset(const char* Label, const char* Section,
-                         unsigned LabelNumber, unsigned SectionNumber,
-                         bool IsSmall = false, bool isEH = false,
-                         bool useSet = true);
-#endif
-
   /// EmitFrameMoves - Emit frame instructions to describe the layout of the
   /// frame.
   void EmitFrameMoves(const char *BaseLabel, unsigned BaseLabelID,




From dpatel at apple.com  Mon Mar  8 16:38:08 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 8 Mar 2010 14:38:08 -0800
Subject: [llvm-commits] [PATCH] Delete the CompileUnits that DwarfDebug
	allocates
In-Reply-To: <bb55930e1003061520h8c2cf01xf45c4707a364a550@mail.gmail.com>
References: <bb55930e1003060151r37644c6at96a3ae4a87db99a2@mail.gmail.com>
	<bb55930e1003061520h8c2cf01xf45c4707a364a550@mail.gmail.com>
Message-ID: <2B67EB78-FCA0-43CF-81BE-E8BE27AF4259@apple.com>


On Mar 6, 2010, at 3:20 PM, Jeffrey Yasskin wrote:

> On Sat, Mar 6, 2010 at 1:51 AM, Jeffrey Yasskin <jyasskin at google.com> wrote:
>> DwarfDebug new's CompileUnits in beginModule() but never deleted them.
>> This patch deletes them in endModule(), but I don't know enough about
>> how DwarfDebug is used to know for sure that's right. Please
>> double-check it.
> 
> Here's a new patch which also destroys the DbgScopes.

The patch looks good. I removed CompileUnitsMap from mainline sources today, so you'll have to update this patch accordingly.

Thanks for working on this!
-
Devang

> I'm still
> working on testing it against the FrontendC* set. It passes the rest
> of check-lit.
> This should enable my MDNode deletion patch to go in, but I haven't
> tested that again either.
> <delete_cu_and_dbgscopes.patch>




From sabre at nondot.org  Mon Mar  8 16:44:40 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:44:40 -0000
Subject: [llvm-commits] [llvm] r97996 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfException.cpp
 DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100308224440.B03802A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 16:44:40 2010
New Revision: 97996

URL: http://llvm.org/viewvc/llvm-project?rev=97996&view=rev
Log:
mc'ize EmitLabel.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97996&r1=97995&r2=97996&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:44:40 2010
@@ -1845,14 +1845,14 @@
 
   // Standard sections final addresses.
   Asm->OutStreamer.SwitchSection(Asm->getObjFileLowering().getTextSection());
-  EmitLabel("text_end", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("text_end"));
   Asm->OutStreamer.SwitchSection(Asm->getObjFileLowering().getDataSection());
-  EmitLabel("data_end", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("data_end"));
 
   // End text sections.
   for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
     Asm->OutStreamer.SwitchSection(SectionMap[i]);
-    EmitLabel("section_end", i);
+    Asm->OutStreamer.EmitLabel(getDWLabel("section_end", i));
   }
 
   // Emit common frame information.
@@ -2112,7 +2112,7 @@
   collectVariableInfo();
 
   // Assumes in correct section after the entry point.
-  EmitLabel("func_begin", ++SubprogramCount);
+  Asm->OutStreamer.EmitLabel(getDWLabel("func_begin", ++SubprogramCount));
 
   // Emit label for the implicitly defined dbg.stoppoint at the start of the
   // function.
@@ -2148,7 +2148,7 @@
 
   if (CurrentFnDbgScope) {
     // Define end label for subprogram.
-    EmitLabel("func_end", SubprogramCount);
+    Asm->OutStreamer.EmitLabel(getDWLabel("func_end", SubprogramCount));
     
     // Get function line info.
     if (!Lines.empty()) {
@@ -2320,38 +2320,38 @@
   // Dwarf sections base addresses.
   if (MAI->doesDwarfRequireFrameSection()) {
     Asm->OutStreamer.SwitchSection(TLOF.getDwarfFrameSection());
-    EmitLabel("section_debug_frame", 0);
+    Asm->OutStreamer.EmitLabel(getTempLabel("section_debug_frame"));
   }
 
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfInfoSection());
-  EmitLabel("section_info", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_info"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfAbbrevSection());
-  EmitLabel("section_abbrev", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_abbrev"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfARangesSection());
-  EmitLabel("section_aranges", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_aranges"));
 
   if (const MCSection *LineInfoDirective = TLOF.getDwarfMacroInfoSection()) {
     Asm->OutStreamer.SwitchSection(LineInfoDirective);
-    EmitLabel("section_macinfo", 0);
+    Asm->OutStreamer.EmitLabel(getTempLabel("section_macinfo"));
   }
 
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfLineSection());
-  EmitLabel("section_line", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_line"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfLocSection());
-  EmitLabel("section_loc", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_loc"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfPubNamesSection());
-  EmitLabel("section_pubnames", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_pubnames"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfPubTypesSection());
-  EmitLabel("section_pubtypes", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_pubtypes"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfStrSection());
-  EmitLabel("section_str", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_str"));
   Asm->OutStreamer.SwitchSection(TLOF.getDwarfRangesSection());
-  EmitLabel("section_ranges", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("section_ranges"));
 
   Asm->OutStreamer.SwitchSection(TLOF.getTextSection());
-  EmitLabel("text_begin", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("text_begin"));
   Asm->OutStreamer.SwitchSection(TLOF.getDataSection());
-  EmitLabel("data_begin", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("data_begin"));
 }
 
 /// emitDIE - Recusively Emits a debug information entry.
@@ -2422,7 +2422,7 @@
   DIE *Die = ModuleCU->getCUDie();
 
   // Emit the compile units header.
-  EmitLabel("info_begin", ModuleCU->getID());
+  Asm->OutStreamer.EmitLabel(getDWLabel("info_begin", ModuleCU->getID()));
 
   // Emit size of content not including length itself
   unsigned ContentSize = Die->getSize() +
@@ -2444,7 +2444,7 @@
   Asm->EmitInt8(0); EOL("Extra Pad For GDB");
   Asm->EmitInt8(0); EOL("Extra Pad For GDB");
   Asm->EmitInt8(0); EOL("Extra Pad For GDB");
-  EmitLabel("info_end", ModuleCU->getID());
+  Asm->OutStreamer.EmitLabel(getDWLabel("info_end", ModuleCU->getID()));
   Asm->O << '\n';
 }
 
@@ -2457,7 +2457,7 @@
     Asm->OutStreamer.SwitchSection(
                             Asm->getObjFileLowering().getDwarfAbbrevSection());
 
-    EmitLabel("abbrev_begin", 0);
+    Asm->OutStreamer.EmitLabel(getTempLabel("abbrev_begin"));
 
     // For each abbrevation.
     for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
@@ -2475,8 +2475,7 @@
     // Mark end of abbreviations.
     EmitULEB128(0, "EOM(3)");
 
-    EmitLabel("abbrev_end", 0);
-    Asm->O << '\n';
+    Asm->OutStreamer.EmitLabel(getTempLabel("abbrev_end"));
   }
 }
 
@@ -2514,15 +2513,16 @@
                             Asm->getObjFileLowering().getDwarfLineSection());
 
   // Construct the section header.
-  EmitDifference("line_end", 0, "line_begin", 0, true);
+  EmitDifference(getTempLabel("line_end"), getTempLabel("line_begin"), true);
   EOL("Length of Source Line Info");
-  EmitLabel("line_begin", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("line_begin"));
 
   Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF version number");
 
-  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
+  EmitDifference(getTempLabel("line_prolog_end"),
+                 getTempLabel("line_prolog_begin"), true);
   EOL("Prolog Length");
-  EmitLabel("line_prolog_begin", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("line_prolog_begin"));
 
   Asm->EmitInt8(1); EOL("Minimum Instruction Length");
   Asm->EmitInt8(1); EOL("Default is_stmt_start flag");
@@ -2565,7 +2565,7 @@
 
   Asm->EmitInt8(0); EOL("End of files");
 
-  EmitLabel("line_prolog_end", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("line_prolog_end"));
 
   // A sequence for each text section.
   unsigned SecSrcLinesSize = SectionSourceLines.size();
@@ -2652,8 +2652,7 @@
     // put into it, emit an empty table.
     emitEndOfLineMatrix(1);
 
-  EmitLabel("line_end", 0);
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getTempLabel("line_end"));
 }
 
 /// emitCommonDebugFrame - Emit common frame info into a debug frame section.
@@ -2671,12 +2670,12 @@
   Asm->OutStreamer.SwitchSection(
                               Asm->getObjFileLowering().getDwarfFrameSection());
 
-  EmitLabel("debug_frame_common", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common"));
   EmitDifference("debug_frame_common_end", 0,
                  "debug_frame_common_begin", 0, true);
   EOL("Length of Common Information Entry");
 
-  EmitLabel("debug_frame_common_begin", 0);
+  Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common_begin"));
   Asm->EmitInt32((int)dwarf::DW_CIE_ID);
   EOL("CIE Identifier Tag");
   Asm->EmitInt8(dwarf::DW_CIE_VERSION);
@@ -2694,8 +2693,7 @@
   EmitFrameMoves(NULL, 0, Moves, false);
 
   Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_common_end", 0);
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common_end"));
 }
 
 /// emitFunctionDebugFrame - Emit per function frame info into a debug frame
@@ -2713,7 +2711,8 @@
                  "debug_frame_begin", DebugFrameInfo.Number, true);
   EOL("Length of Frame Information Entry");
 
-  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
+  Asm->OutStreamer.EmitLabel(getDWLabel("debug_frame_begin",
+                                        DebugFrameInfo.Number));
 
   EmitSectionOffset(getTempLabel("debug_frame_common"),
                     getTempLabel("section_debug_frame"), true, false);
@@ -2729,7 +2728,8 @@
                  false);
 
   Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
+  Asm->OutStreamer.EmitLabel(getDWLabel("debug_frame_end",
+                                        DebugFrameInfo.Number));
   Asm->O << '\n';
 }
 
@@ -2744,7 +2744,7 @@
                  "pubnames_begin", ModuleCU->getID(), true);
   EOL("Length of Public Names Info");
 
-  EmitLabel("pubnames_begin", ModuleCU->getID());
+  Asm->OutStreamer.EmitLabel(getDWLabel("pubnames_begin", ModuleCU->getID()));
 
   Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF Version");
 
@@ -2771,8 +2771,7 @@
   }
 
   Asm->EmitInt32(0); EOL("End Mark");
-  EmitLabel("pubnames_end", ModuleCU->getID());
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getDWLabel("pubnames_end", ModuleCU->getID()));
 }
 
 void DwarfDebug::emitDebugPubTypes() {
@@ -2783,7 +2782,7 @@
                  "pubtypes_begin", ModuleCU->getID(), true);
   EOL("Length of Public Types Info");
 
-  EmitLabel("pubtypes_begin", ModuleCU->getID());
+  Asm->OutStreamer.EmitLabel(getDWLabel("pubtypes_begin", ModuleCU->getID()));
 
   if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DWARF Version");
   Asm->EmitInt16(dwarf::DWARF_VERSION);
@@ -2810,8 +2809,7 @@
   }
 
   Asm->EmitInt32(0); EOL("End Mark");
-  EmitLabel("pubtypes_end", ModuleCU->getID());
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getDWLabel("pubtypes_end", ModuleCU->getID()));
 }
 
 /// emitDebugStr - Emit visible names into a debug str section.
@@ -2827,7 +2825,7 @@
     for (unsigned StringID = 1, N = StringPool.size();
          StringID <= N; ++StringID) {
       // Emit a label for reference from debug information entries.
-      EmitLabel("string", StringID);
+      Asm->OutStreamer.EmitLabel(getDWLabel("string", StringID));
 
       // Emit the string itself.
       const std::string &String = StringPool[StringID];
@@ -2931,7 +2929,7 @@
                  "debug_inlined_begin", 1, true);
   EOL("Length of Debug Inlined Information Entry");
 
-  EmitLabel("debug_inlined_begin", 1);
+  Asm->OutStreamer.EmitLabel(getDWLabel("debug_inlined_begin", 1));
 
   Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("Dwarf Version");
   Asm->EmitInt8(TD->getPointerSize()); EOL("Address Size (in bytes)");
@@ -2975,6 +2973,5 @@
     }
   }
 
-  EmitLabel("debug_inlined_end", 1);
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getDWLabel("debug_inlined_end", 1));
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=97996&r1=97995&r2=97996&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 16:44:40 2010
@@ -88,10 +88,10 @@
     O << MAI->getPrivateGlobalPrefix();
   O << "EH_frame" << Index << ":\n";
   
-  EmitLabel("section_eh_frame", Index);
+  Asm->OutStreamer.EmitLabel(getDWLabel("section_eh_frame", Index));
 
   // Define base labels.
-  EmitLabel("eh_frame_common", Index);
+  Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common", Index));
 
   // Define the eh frame length.
   EmitDifference("eh_frame_common_end", Index,
@@ -99,7 +99,7 @@
   EOL("Length of Common Information Entry");
 
   // EH frame header.
-  EmitLabel("eh_frame_common_begin", Index);
+  Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common_begin", Index));
   if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("CIE Identifier Tag");
   Asm->OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
   if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DW_CIE_VERSION");
@@ -171,8 +171,7 @@
   // be 8-byte on 64-bit targets to match what gcc does.  Otherwise you get
   // holes which confuse readers of eh_frame.
   Asm->EmitAlignment(TD->getPointerSize() == 4 ? 2 : 3, 0, 0, false);
-  EmitLabel("eh_frame_common_end", Index);
-  Asm->O << '\n';
+  Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common_end", Index));
 }
 
 /// EmitFDE - Emit the Frame Description Entry (FDE) for the function.
@@ -228,7 +227,7 @@
                    true);
     EOL("Length of Frame Information Entry");
 
-    EmitLabel("eh_frame_begin", EHFrameInfo.Number);
+    Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_begin",EHFrameInfo.Number));
 
     EmitSectionOffset(getDWLabel("eh_frame_begin", EHFrameInfo.Number),
                       getDWLabel("eh_frame_common",
@@ -269,7 +268,7 @@
     // get holes which confuse readers of eh_frame.
     Asm->EmitAlignment(TD->getPointerSize() == sizeof(int32_t) ? 2 : 3,
                        0, 0, false);
-    EmitLabel("eh_frame_end", EHFrameInfo.Number);
+    Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_end", EHFrameInfo.Number));
 
     // If the function is marked used, this table should be also.  We cannot
     // make the mark unconditional in this case, since retaining the table also
@@ -701,7 +700,7 @@
 
   // Emit the LSDA.
   O << "GCC_except_table" << SubprogramCount << ":\n";
-  EmitLabel("exception", SubprogramCount);
+  Asm->OutStreamer.EmitLabel(getDWLabel("exception", SubprogramCount));
 
   if (IsSJLJ) {
     SmallString<16> LSDAName;
@@ -949,7 +948,7 @@
 
   if (shouldEmitMoves || shouldEmitTable)
     // Assumes in correct section after the entry point.
-    EmitLabel("eh_func_begin", ++SubprogramCount);
+    Asm->OutStreamer.EmitLabel(getDWLabel("eh_func_begin", ++SubprogramCount));
 
   shouldEmitTableModule |= shouldEmitTable;
   shouldEmitMovesModule |= shouldEmitMoves;
@@ -966,7 +965,7 @@
   if (TimePassesIsEnabled)
     ExceptionTimer->startTimer();
 
-  EmitLabel("eh_func_end", SubprogramCount);
+  Asm->OutStreamer.EmitLabel(getDWLabel("eh_func_end", SubprogramCount));
   EmitExceptionTable();
 
   MCSymbol *FunctionEHSym =

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=97996&r1=97995&r2=97996&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 16:44:40 2010
@@ -232,14 +232,6 @@
   O << Suffix;
 }
 
-/// EmitLabel - Emit location label for internal use by Dwarf.
-///
-void DwarfPrinter::EmitLabel(const char *Tag, unsigned Number) const {
-  // FIXME: REMOVE.
-  PrintLabelName(Tag, Number);
-  O << ":\n";
-}
-
 /// EmitReference - Emit a reference to a label.
 ///
 void DwarfPrinter::EmitReference(const char *Tag, unsigned Number,

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97996&r1=97995&r2=97996&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:44:40 2010
@@ -129,11 +129,6 @@
   void PrintLabelName(const char *Tag, unsigned Number,
                       const char *Suffix) const;
 
-  /// EmitLabel - Emit location label for internal use by Dwarf.
-  ///
-  void EmitLabel(const MCSymbol *Label) const;
-  void EmitLabel(const char *Tag, unsigned Number) const;
-
   /// EmitReference - Emit a reference to a label.
   ///
   void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,




From sabre at nondot.org  Mon Mar  8 16:47:57 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:47:57 -0000
Subject: [llvm-commits] [llvm] r97997 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100308224757.A79A42A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 16:47:57 2010
New Revision: 97997

URL: http://llvm.org/viewvc/llvm-project?rev=97997&view=rev
Log:
eliminate the non-MCSymbol versions of EmitReference.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97997&r1=97996&r2=97997&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:47:57 2010
@@ -2487,7 +2487,8 @@
   Asm->EmitInt8(0); EOL("Extended Op");
   Asm->EmitInt8(TD->getPointerSize() + 1); EOL("Op size");
   Asm->EmitInt8(dwarf::DW_LNE_set_address); EOL("DW_LNE_set_address");
-  EmitReference("section_end", SectionEnd); EOL("Section end label");
+  EmitReference(getDWLabel("section_end", SectionEnd));
+  EOL("Section end label");
 
   // Mark end of matrix.
   Asm->EmitInt8(0); EOL("DW_LNE_end_sequence");
@@ -2608,7 +2609,7 @@
       Asm->EmitInt8(0); EOL("Extended Op");
       Asm->EmitInt8(TD->getPointerSize() + 1); EOL("Op size");
       Asm->EmitInt8(dwarf::DW_LNE_set_address); EOL("DW_LNE_set_address");
-      EmitReference("label",  LabelID); EOL("Location label");
+      EmitReference(getDWLabel("label", LabelID)); EOL("Location label");
 
       // If change of source, then switch to the new source.
       if (Source != LineInfo.getSourceID()) {
@@ -2718,10 +2719,10 @@
                     getTempLabel("section_debug_frame"), true, false);
   EOL("FDE CIE offset");
 
-  EmitReference("func_begin", DebugFrameInfo.Number);
+  EmitReference(getDWLabel("func_begin", DebugFrameInfo.Number));
   EOL("FDE initial location");
-  EmitDifference("func_end", DebugFrameInfo.Number,
-                 "func_begin", DebugFrameInfo.Number);
+  EmitDifference(getDWLabel("func_end", DebugFrameInfo.Number),
+                 getDWLabel("func_begin", DebugFrameInfo.Number));
   EOL("FDE address range");
 
   EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
@@ -2730,7 +2731,6 @@
   Asm->EmitAlignment(2, 0, 0, false);
   Asm->OutStreamer.EmitLabel(getDWLabel("debug_frame_end",
                                         DebugFrameInfo.Number));
-  Asm->O << '\n';
 }
 
 /// emitDebugPubNames - Emit visible names into a debug pubnames section.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=97997&r1=97996&r2=97997&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 16:47:57 2010
@@ -234,19 +234,6 @@
 
 /// EmitReference - Emit a reference to a label.
 ///
-void DwarfPrinter::EmitReference(const char *Tag, unsigned Number,
-                                 bool IsPCRelative, bool Force32Bit) const {
-  PrintRelDirective(Force32Bit);
-  PrintLabelName(Tag, Number);
-  if (IsPCRelative) O << "-" << MAI->getPCSymbol();
-}
-void DwarfPrinter::EmitReference(const std::string &Name, bool IsPCRelative,
-                                 bool Force32Bit) const {
-  PrintRelDirective(Force32Bit);
-  O << Name;
-  if (IsPCRelative) O << "-" << MAI->getPCSymbol();
-}
-
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, bool IsPCRelative,
                                  bool Force32Bit) const {
   PrintRelDirective(Force32Bit);

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97997&r1=97996&r2=97997&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:47:57 2010
@@ -133,12 +133,7 @@
   ///
   void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,
                      bool Force32Bit = false) const;
-  void EmitReference(const char *Tag, unsigned Number,
-                     bool IsPCRelative = false,
-                     bool Force32Bit = false) const;
-  void EmitReference(const std::string &Name, bool IsPCRelative = false,
-                     bool Force32Bit = false) const;
-
+  
   void EmitReference(const char *Tag, unsigned Number, unsigned Encoding) const;
   void EmitReference(const MCSymbol *Sym, unsigned Encoding) const;
   void EmitReference(const GlobalValue *GV, unsigned Encoding) const;




From sabre at nondot.org  Mon Mar  8 16:50:36 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:50:36 -0000
Subject: [llvm-commits] [llvm] r97998 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfException.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100308225036.86D0D2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 16:50:36 2010
New Revision: 97998

URL: http://llvm.org/viewvc/llvm-project?rev=97998&view=rev
Log:
remove another form of EmitReference.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=97998&r1=97997&r2=97998&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 16:50:36 2010
@@ -236,7 +236,7 @@
 
     EOL("FDE CIE offset");
 
-    EmitReference("eh_func_begin", EHFrameInfo.Number, FDEEncoding);
+    EmitReference(getDWLabel("eh_func_begin", EHFrameInfo.Number), FDEEncoding);
     EOL("FDE initial location");
     EmitDifference("eh_func_end", EHFrameInfo.Number,
                    "eh_func_begin", EHFrameInfo.Number,
@@ -250,7 +250,7 @@
 
       EmitULEB128(Size, "Augmentation size");
       if (EHFrameInfo.hasLandingPads)
-        EmitReference("exception", EHFrameInfo.Number, LSDAEncoding);
+        EmitReference(getDWLabel("exception", EHFrameInfo.Number),LSDAEncoding);
       else
         Asm->OutStreamer.EmitIntValue(0, Size/*size*/, 0/*addrspace*/);
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=97998&r1=97997&r2=97998&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 16:50:36 2010
@@ -241,12 +241,6 @@
   if (IsPCRelative) O << "-" << MAI->getPCSymbol();
 }
 
-void DwarfPrinter::EmitReference(const char *Tag, unsigned Number,
-                                 unsigned Encoding) const {
-  // FIXME: REMOVE.
-  EmitReference(getDWLabel(Tag, Number), Encoding);
-}
-
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
@@ -259,7 +253,7 @@
 
   PrintRelDirective(Encoding);
   O << *TLOF.getSymbolForDwarfGlobalReference(GV, Asm->Mang,
-                                              Asm->MMI, Encoding);;
+                                              Asm->MMI, Encoding);
 }
 
 /// EmitDifference - Emit the difference between two labels.  If this assembler

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97998&r1=97997&r2=97998&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:50:36 2010
@@ -134,7 +134,6 @@
   void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,
                      bool Force32Bit = false) const;
   
-  void EmitReference(const char *Tag, unsigned Number, unsigned Encoding) const;
   void EmitReference(const MCSymbol *Sym, unsigned Encoding) const;
   void EmitReference(const GlobalValue *GV, unsigned Encoding) const;
 




From sabre at nondot.org  Mon Mar  8 16:52:49 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 22:52:49 -0000
Subject: [llvm-commits] [llvm] r97999 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100308225249.CDB7C2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 16:52:49 2010
New Revision: 97999

URL: http://llvm.org/viewvc/llvm-project?rev=97999&view=rev
Log:
eliminate a form of PrintLabelName.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97999&r1=97998&r2=97999&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 16:52:49 2010
@@ -2969,7 +2969,7 @@
       else
         O << MAI->getData64bitsDirective();
 
-      PrintLabelName("label", LI->first); EOL("low_pc");
+      PrintLabelName(getDWLabel("label", LI->first)); EOL("low_pc");
     }
   }
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=97999&r1=97998&r2=97999&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 16:52:49 2010
@@ -219,11 +219,6 @@
   O << Label->getName();
 }
 
-void DwarfPrinter::PrintLabelName(const char *Tag, unsigned Number) const {
-  // FIXME: REMOVE.
-  O << MAI->getPrivateGlobalPrefix() << Tag;
-  if (Number) O << Number;
-}
 void DwarfPrinter::PrintLabelName(const char *Tag, unsigned Number,
                                   const char *Suffix) const {
   // FIXME: REMOVE.
@@ -291,9 +286,9 @@
     O << "\t.set\t";
     PrintLabelName("set", SetCounter, Flavor);
     O << ",";
-    PrintLabelName(TagHi, NumberHi);
+    PrintLabelName(getDWLabel(TagHi, NumberHi));
     O << "-";
-    PrintLabelName(TagLo, NumberLo);
+    PrintLabelName(getDWLabel(TagLo, NumberLo));
     O << "\n";
     
     PrintRelDirective(IsSmall);
@@ -301,9 +296,9 @@
     ++SetCounter;
   } else {
     PrintRelDirective(IsSmall);
-    PrintLabelName(TagHi, NumberHi);
+    PrintLabelName(getDWLabel(TagHi, NumberHi));
     O << "-";
-    PrintLabelName(TagLo, NumberLo);
+    PrintLabelName(getDWLabel(TagLo, NumberLo));
   }
 }
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=97999&r1=97998&r2=97999&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 16:52:49 2010
@@ -125,7 +125,6 @@
   /// PrintLabelName - Print label name in form used by Dwarf writer.
   ///
   void PrintLabelName(const MCSymbol *Label) const;
-  void PrintLabelName(const char *Tag, unsigned Number) const;
   void PrintLabelName(const char *Tag, unsigned Number,
                       const char *Suffix) const;
 




From evan.cheng at apple.com  Mon Mar  8 16:54:36 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 08 Mar 2010 22:54:36 -0000
Subject: [llvm-commits] [llvm] r98000 - in /llvm/trunk:
 lib/Transforms/InstCombine/InstCombineCalls.cpp
 test/Transforms/InstCombine/objsize.ll
Message-ID: <20100308225436.957F82A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 16:54:36 2010
New Revision: 98000

URL: http://llvm.org/viewvc/llvm-project?rev=98000&view=rev
Log:
Re-commit 97860 with fix. getMallocAllocatedType may return null.

Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/test/Transforms/InstCombine/objsize.ll

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=98000&r1=97999&r2=98000&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Mon Mar  8 16:54:36 2010
@@ -337,6 +337,16 @@
         }
         return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy, AllocaSize));
       }
+    } else if (CallInst *MI = extractMallocCall(Op1)) {
+      const Type* MallocType = getMallocAllocatedType(MI);
+      // Get alloca size.
+      if (MallocType && MallocType->isSized()) {
+        if (Value *NElems = getMallocArraySize(MI, TD, true)) {
+          if (ConstantInt *NElements = dyn_cast<ConstantInt>(NElems))
+        return ReplaceInstUsesWith(CI, ConstantInt::get(ReturnTy,
+               (NElements->getZExtValue() * TD->getTypeAllocSize(MallocType))));
+        }
+      }
     } else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Op1)) {      
       // Only handle constant GEPs here.
       if (CE->getOpcode() != Instruction::GetElementPtr) break;

Modified: llvm/trunk/test/Transforms/InstCombine/objsize.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/objsize.ll?rev=98000&r1=97999&r2=98000&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/objsize.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/objsize.ll Mon Mar  8 16:54:36 2010
@@ -118,8 +118,24 @@
   ret i32 0
 }
 
+ at s = external global i8*
+
+define void @test5(i32 %n) nounwind ssp {
+; CHECK: @test5
+entry:
+  %0 = tail call noalias i8* @malloc(i32 20) nounwind
+  %1 = tail call i32 @llvm.objectsize.i32(i8* %0, i1 false)
+  %2 = load i8** @s, align 8
+; CHECK-NOT: @llvm.objectsize
+; CHECK: @__memcpy_chk(i8* %0, i8* %1, i32 10, i32 20)
+  %3 = tail call i8* @__memcpy_chk(i8* %0, i8* %2, i32 10, i32 %1) nounwind
+  ret void
+}
+
 declare i8* @__memset_chk(i8*, i32, i64, i64) nounwind
 
+declare noalias i8* @malloc(i32) nounwind
+
 declare i32 @llvm.objectsize.i32(i8*, i1) nounwind readonly
 
 declare i64 @llvm.objectsize.i64(i8*, i1) nounwind readonly




From bob.wilson at apple.com  Mon Mar  8 16:56:16 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Mon, 08 Mar 2010 22:56:16 -0000
Subject: [llvm-commits] [llvm] r98001 - in /llvm/trunk:
 lib/Target/ARM/ARMInstrThumb2.td lib/Target/ARM/Thumb2InstrInfo.cpp
 test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll
Message-ID: <20100308225616.1220F2A6C12C@llvm.org>

Author: bwilson
Date: Mon Mar  8 16:56:15 2010
New Revision: 98001

URL: http://llvm.org/viewvc/llvm-project?rev=98001&view=rev
Log:
Fix a crash compiling 254.gap for Thumb2.  The Thumb2 add/sub with 12-bit
immediate instructions cannot set the condition codes, so they do not have
the extra cc_out operand.  We hit an assertion during tail duplication
because the instruction being duplicated had more operands that expected.

Added:
    llvm/trunk/test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
    llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=98001&r1=98000&r2=98001&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Mon Mar  8 16:56:15 2010
@@ -331,9 +331,9 @@
      let Inst{15} = 0;
    }
    // 12-bit imm
-   def ri12 : T2sI<(outs GPR:$dst), (ins GPR:$lhs, imm0_4095:$rhs), IIC_iALUi,
-                   !strconcat(opc, "w"), "\t$dst, $lhs, $rhs",
-                   [(set GPR:$dst, (opnode GPR:$lhs, imm0_4095:$rhs))]> {
+   def ri12 : T2I<(outs GPR:$dst), (ins GPR:$lhs, imm0_4095:$rhs), IIC_iALUi,
+                  !strconcat(opc, "w"), "\t$dst, $lhs, $rhs",
+                  [(set GPR:$dst, (opnode GPR:$lhs, imm0_4095:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 1;
      let Inst{24} = 0;

Modified: llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp?rev=98001&r1=98000&r2=98001&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp Mon Mar  8 16:56:15 2010
@@ -164,6 +164,7 @@
       continue;
     }
 
+    bool HasCCOut = true;
     if (BaseReg == ARM::SP) {
       // sub sp, sp, #imm7
       if (DestReg == ARM::SP && (ThisVal < ((1 << 7)-1) * 4)) {
@@ -195,6 +196,7 @@
         NumBytes = 0;
       } else if (ThisVal < 4096) {
         Opc = isSub ? ARM::t2SUBri12 : ARM::t2ADDri12;
+        HasCCOut = false;
         NumBytes = 0;
       } else {
         // FIXME: Move this to ARMAddressingModes.h?
@@ -207,9 +209,12 @@
     }
 
     // Build the new ADD / SUB.
-    AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(Opc), DestReg)
-                                .addReg(BaseReg, RegState::Kill)
-                                .addImm(ThisVal)));
+    MachineInstrBuilder MIB =
+      AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(Opc), DestReg)
+                     .addReg(BaseReg, RegState::Kill)
+                     .addImm(ThisVal));
+    if (HasCCOut)
+      AddDefaultCC(MIB);
 
     BaseReg = DestReg;
   }
@@ -328,7 +333,6 @@
   if (Opcode == ARM::t2ADDri || Opcode == ARM::t2ADDri12) {
     Offset += MI.getOperand(FrameRegIdx+1).getImm();
 
-    bool isSP = FrameReg == ARM::SP;
     unsigned PredReg;
     if (Offset == 0 && getInstrPredicate(&MI, PredReg) == ARMCC::AL) {
       // Turn it into a move.
@@ -342,6 +346,9 @@
       return true;
     }
 
+    bool isSP = FrameReg == ARM::SP;
+    bool HasCCOut = Opcode != ARM::t2ADDri12;
+
     if (Offset < 0) {
       Offset = -Offset;
       isSub = true;
@@ -354,17 +361,24 @@
     if (ARM_AM::getT2SOImmVal(Offset) != -1) {
       MI.getOperand(FrameRegIdx).ChangeToRegister(FrameReg, false);
       MI.getOperand(FrameRegIdx+1).ChangeToImmediate(Offset);
+      // Add cc_out operand if the original instruction did not have one.
+      if (!HasCCOut)
+        MI.addOperand(MachineOperand::CreateReg(0, false));
       Offset = 0;
       return true;
     }
     // Another common case: imm12.
-    if (Offset < 4096) {
+    if (Offset < 4096 &&
+        (!HasCCOut || MI.getOperand(MI.getNumOperands()-1).getReg() == 0)) {
       unsigned NewOpc = isSP
         ? (isSub ? ARM::t2SUBrSPi12 : ARM::t2ADDrSPi12)
         : (isSub ? ARM::t2SUBri12   : ARM::t2ADDri12);
       MI.setDesc(TII.get(NewOpc));
       MI.getOperand(FrameRegIdx).ChangeToRegister(FrameReg, false);
       MI.getOperand(FrameRegIdx+1).ChangeToImmediate(Offset);
+      // Remove the cc_out operand.
+      if (HasCCOut)
+        MI.RemoveOperand(MI.getNumOperands()-1);
       Offset = 0;
       return true;
     }
@@ -380,6 +394,10 @@
     assert(ARM_AM::getT2SOImmVal(ThisImmVal) != -1 &&
            "Bit extraction didn't work?");
     MI.getOperand(FrameRegIdx+1).ChangeToImmediate(ThisImmVal);
+    // Add cc_out operand if the original instruction did not have one.
+    if (!HasCCOut)
+      MI.addOperand(MachineOperand::CreateReg(0, false));
+
   } else {
 
     // AddrMode4 and AddrMode6 cannot handle any offset.

Added: llvm/trunk/test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll?rev=98001&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll (added)
+++ llvm/trunk/test/CodeGen/Thumb2/2010-03-08-addi12-ccout.ll Mon Mar  8 16:56:15 2010
@@ -0,0 +1,266 @@
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin
+
+ at .str41196 = external constant [2 x i8], align 4  ; <[2 x i8]*> [#uses=1]
+
+declare arm_apcscc void @syStopraw(i32) nounwind
+
+declare arm_apcscc i32 @SyFopen(i8*, i8*) nounwind
+
+declare arm_apcscc i8* @SyFgets(i8*, i32) nounwind
+
+define arm_apcscc void @SyHelp(i8* nocapture %topic, i32 %fin) nounwind {
+entry:
+  %line = alloca [256 x i8], align 4              ; <[256 x i8]*> [#uses=1]
+  %secname = alloca [1024 x i8], align 4          ; <[1024 x i8]*> [#uses=0]
+  %last = alloca [256 x i8], align 4              ; <[256 x i8]*> [#uses=1]
+  %last2 = alloca [256 x i8], align 4             ; <[256 x i8]*> [#uses=1]
+  br i1 undef, label %bb, label %bb2
+
+bb:                                               ; preds = %entry
+  br i1 undef, label %bb2, label %bb3
+
+bb2:                                              ; preds = %bb, %entry
+  br label %bb3
+
+bb3:                                              ; preds = %bb2, %bb
+  %storemerge = phi i32 [ 0, %bb2 ], [ 1, %bb ]   ; <i32> [#uses=1]
+  br i1 undef, label %bb19, label %bb20
+
+bb19:                                             ; preds = %bb3
+  br label %bb20
+
+bb20:                                             ; preds = %bb19, %bb3
+  br i1 undef, label %bb25, label %bb26
+
+bb25:                                             ; preds = %bb20
+  br label %bb26
+
+bb26:                                             ; preds = %bb25, %bb20
+  %offset.2 = phi i32 [ -2, %bb25 ], [ 0, %bb20 ] ; <i32> [#uses=1]
+  br i1 undef, label %bb.nph508, label %bb49
+
+bb.nph508:                                        ; preds = %bb26
+  unreachable
+
+bb49:                                             ; preds = %bb26
+  br i1 undef, label %bb51, label %bb50
+
+bb50:                                             ; preds = %bb49
+  br i1 undef, label %bb51, label %bb104
+
+bb51:                                             ; preds = %bb50, %bb49
+  unreachable
+
+bb104:                                            ; preds = %bb50
+  br i1 undef, label %bb106, label %bb105
+
+bb105:                                            ; preds = %bb104
+  br i1 undef, label %bb106, label %bb161
+
+bb106:                                            ; preds = %bb105, %bb104
+  unreachable
+
+bb161:                                            ; preds = %bb105
+  br i1 false, label %bb163, label %bb162
+
+bb162:                                            ; preds = %bb161
+  br i1 undef, label %bb163, label %bb224
+
+bb163:                                            ; preds = %bb162, %bb161
+  unreachable
+
+bb224:                                            ; preds = %bb162
+  %0 = call arm_apcscc  i32 @SyFopen(i8* undef, i8* getelementptr inbounds ([2 x i8]* @.str41196, i32 0, i32 0)) nounwind ; <i32> [#uses=2]
+  br i1 false, label %bb297, label %bb300
+
+bb297:                                            ; preds = %bb224
+  unreachable
+
+bb300:                                            ; preds = %bb224
+  %1 = icmp eq i32 %offset.2, -1                  ; <i1> [#uses=1]
+  br label %bb440
+
+bb307:                                            ; preds = %isdigit1498.exit67
+  br label %bb308
+
+bb308:                                            ; preds = %bb440, %bb307
+  br i1 undef, label %bb309, label %isdigit1498.exit67
+
+isdigit1498.exit67:                               ; preds = %bb308
+  br i1 undef, label %bb309, label %bb307
+
+bb309:                                            ; preds = %isdigit1498.exit67, %bb308
+  br i1 undef, label %bb310, label %bb313
+
+bb310:                                            ; preds = %bb309
+  br label %bb313
+
+bb313:                                            ; preds = %bb310, %bb309
+  br i1 false, label %bb318, label %bb317
+
+bb317:                                            ; preds = %bb313
+  %2 = icmp sgt i8 undef, -1                      ; <i1> [#uses=1]
+  br i1 %2, label %bb.i.i73, label %bb1.i.i74
+
+bb.i.i73:                                         ; preds = %bb317
+  br i1 false, label %bb318, label %bb329.outer
+
+bb1.i.i74:                                        ; preds = %bb317
+  unreachable
+
+bb318:                                            ; preds = %bb.i.i73, %bb313
+  ret void
+
+bb329.outer:                                      ; preds = %bb.i.i73
+  br i1 undef, label %bb333, label %bb329.us.us
+
+bb329.us.us:                                      ; preds = %bb329.us.us, %bb329.outer
+  br i1 undef, label %bb333, label %bb329.us.us
+
+bb333:                                            ; preds = %bb329.us.us, %bb329.outer
+  %match.0.lcssa = phi i32 [ undef, %bb329.us.us ], [ 2, %bb329.outer ] ; <i32> [#uses=2]
+  br i1 undef, label %bb335, label %bb388
+
+bb335:                                            ; preds = %bb333
+  %3 = and i1 undef, %1                           ; <i1> [#uses=1]
+  br i1 %3, label %bb339, label %bb348
+
+bb339:                                            ; preds = %bb335
+  br i1 false, label %bb340, label %bb345
+
+bb340:                                            ; preds = %bb339
+  br i1 undef, label %return, label %bb341
+
+bb341:                                            ; preds = %bb340
+  ret void
+
+bb345:                                            ; preds = %bb345, %bb339
+  %4 = phi i8 [ %5, %bb345 ], [ undef, %bb339 ]   ; <i8> [#uses=0]
+  %indvar670 = phi i32 [ %tmp673, %bb345 ], [ 0, %bb339 ] ; <i32> [#uses=1]
+  %tmp673 = add i32 %indvar670, 1                 ; <i32> [#uses=2]
+  %scevgep674 = getelementptr [256 x i8]* %last, i32 0, i32 %tmp673 ; <i8*> [#uses=1]
+  %5 = load i8* %scevgep674, align 1              ; <i8> [#uses=1]
+  br i1 undef, label %bb347, label %bb345
+
+bb347:                                            ; preds = %bb345
+  br label %bb348
+
+bb348:                                            ; preds = %bb347, %bb335
+  br i1 false, label %bb352, label %bb356
+
+bb352:                                            ; preds = %bb348
+  unreachable
+
+bb356:                                            ; preds = %bb348
+  br i1 undef, label %bb360, label %bb369
+
+bb360:                                            ; preds = %bb356
+  br i1 false, label %bb361, label %bb366
+
+bb361:                                            ; preds = %bb360
+  br i1 undef, label %return, label %bb362
+
+bb362:                                            ; preds = %bb361
+  ret void
+
+bb366:                                            ; preds = %bb366, %bb360
+  %indvar662 = phi i32 [ %tmp665, %bb366 ], [ 0, %bb360 ] ; <i32> [#uses=1]
+  %tmp665 = add i32 %indvar662, 1                 ; <i32> [#uses=2]
+  %scevgep666 = getelementptr [256 x i8]* %last2, i32 0, i32 %tmp665 ; <i8*> [#uses=1]
+  %6 = load i8* %scevgep666, align 1              ; <i8> [#uses=0]
+  br i1 false, label %bb368, label %bb366
+
+bb368:                                            ; preds = %bb366
+  br label %bb369
+
+bb369:                                            ; preds = %bb368, %bb356
+  br i1 undef, label %bb373, label %bb388
+
+bb373:                                            ; preds = %bb383, %bb369
+  %7 = call arm_apcscc  i8* @SyFgets(i8* undef, i32 %0) nounwind ; <i8*> [#uses=1]
+  %8 = icmp eq i8* %7, null                       ; <i1> [#uses=1]
+  br i1 %8, label %bb375, label %bb383
+
+bb375:                                            ; preds = %bb373
+  %9 = icmp eq i32 %storemerge, 0                 ; <i1> [#uses=1]
+  br i1 %9, label %return, label %bb376
+
+bb376:                                            ; preds = %bb375
+  ret void
+
+bb383:                                            ; preds = %bb373
+  %10 = load i8* undef, align 1                   ; <i8> [#uses=1]
+  %cond1 = icmp eq i8 %10, 46                     ; <i1> [#uses=1]
+  br i1 %cond1, label %bb373, label %bb388
+
+bb388:                                            ; preds = %bb383, %bb369, %bb333
+  %match.1140 = phi i32 [ %match.0.lcssa, %bb369 ], [ 0, %bb333 ], [ %match.0.lcssa, %bb383 ] ; <i32> [#uses=1]
+  br label %bb391
+
+bb390:                                            ; preds = %isdigit1498.exit83, %bb392
+  %indvar.next725 = add i32 %indvar724, 1         ; <i32> [#uses=1]
+  br label %bb391
+
+bb391:                                            ; preds = %bb390, %bb388
+  %indvar724 = phi i32 [ %indvar.next725, %bb390 ], [ 0, %bb388 ] ; <i32> [#uses=2]
+  %11 = load i8* undef, align 1                   ; <i8> [#uses=0]
+  br i1 false, label %bb395, label %bb392
+
+bb392:                                            ; preds = %bb391
+  br i1 undef, label %bb390, label %isdigit1498.exit83
+
+isdigit1498.exit83:                               ; preds = %bb392
+  br i1 undef, label %bb390, label %bb395
+
+bb394:                                            ; preds = %isdigit1498.exit87
+  br label %bb395
+
+bb395:                                            ; preds = %bb394, %isdigit1498.exit83, %bb391
+  %storemerge14.sum = add i32 %indvar724, undef   ; <i32> [#uses=1]
+  %p.26 = getelementptr [256 x i8]* %line, i32 0, i32 %storemerge14.sum ; <i8*> [#uses=1]
+  br i1 undef, label %bb400, label %isdigit1498.exit87
+
+isdigit1498.exit87:                               ; preds = %bb395
+  br i1 false, label %bb400, label %bb394
+
+bb400:                                            ; preds = %isdigit1498.exit87, %bb395
+  br i1 undef, label %bb402, label %bb403
+
+bb402:                                            ; preds = %bb400
+  %12 = getelementptr inbounds i8* %p.26, i32 undef ; <i8*> [#uses=1]
+  br label %bb403
+
+bb403:                                            ; preds = %bb402, %bb400
+  %p.29 = phi i8* [ %12, %bb402 ], [ undef, %bb400 ] ; <i8*> [#uses=0]
+  br i1 undef, label %bb405, label %bb404
+
+bb404:                                            ; preds = %bb403
+  br i1 undef, label %bb405, label %bb407
+
+bb405:                                            ; preds = %bb404, %bb403
+  br i1 undef, label %return, label %bb406
+
+bb406:                                            ; preds = %bb405
+  call arm_apcscc  void @syStopraw(i32 %fin) nounwind
+  ret void
+
+bb407:                                            ; preds = %bb404
+  %cond = icmp eq i32 %match.1140, 2              ; <i1> [#uses=1]
+  br i1 %cond, label %bb408, label %bb428
+
+bb408:                                            ; preds = %bb407
+  unreachable
+
+bb428:                                            ; preds = %bb407
+  br label %bb440
+
+bb440:                                            ; preds = %bb428, %bb300
+  %13 = call arm_apcscc  i8* @SyFgets(i8* undef, i32 %0) nounwind ; <i8*> [#uses=0]
+  br i1 false, label %bb442, label %bb308
+
+bb442:                                            ; preds = %bb440
+  unreachable
+
+return:                                           ; preds = %bb405, %bb375, %bb361, %bb340
+  ret void
+}




From sabre at nondot.org  Mon Mar  8 17:02:59 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 23:02:59 -0000
Subject: [llvm-commits] [llvm] r98002 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfException.cpp
 DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100308230259.73BC82A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 17:02:59 2010
New Revision: 98002

URL: http://llvm.org/viewvc/llvm-project?rev=98002&view=rev
Log:
Remove a version of EmitDifference.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98002&r1=98001&r2=98002&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 17:02:59 2010
@@ -2672,8 +2672,8 @@
                               Asm->getObjFileLowering().getDwarfFrameSection());
 
   Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common"));
-  EmitDifference("debug_frame_common_end", 0,
-                 "debug_frame_common_begin", 0, true);
+  EmitDifference(getTempLabel("debug_frame_common_end"),
+                 getTempLabel("debug_frame_common_begin"), true);
   EOL("Length of Common Information Entry");
 
   Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common_begin"));
@@ -2708,8 +2708,8 @@
   Asm->OutStreamer.SwitchSection(
                               Asm->getObjFileLowering().getDwarfFrameSection());
 
-  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
-                 "debug_frame_begin", DebugFrameInfo.Number, true);
+  EmitDifference(getDWLabel("debug_frame_end", DebugFrameInfo.Number),
+                 getDWLabel("debug_frame_begin", DebugFrameInfo.Number), true);
   EOL("Length of Frame Information Entry");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("debug_frame_begin",
@@ -2740,8 +2740,8 @@
   Asm->OutStreamer.SwitchSection(
                           Asm->getObjFileLowering().getDwarfPubNamesSection());
 
-  EmitDifference("pubnames_end", ModuleCU->getID(),
-                 "pubnames_begin", ModuleCU->getID(), true);
+  EmitDifference(getDWLabel("pubnames_end", ModuleCU->getID()),
+                 getDWLabel("pubnames_begin", ModuleCU->getID()), true);
   EOL("Length of Public Names Info");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("pubnames_begin", ModuleCU->getID()));
@@ -2753,7 +2753,8 @@
                     true, false);
   EOL("Offset of Compilation Unit Info");
 
-  EmitDifference("info_end", ModuleCU->getID(), "info_begin", ModuleCU->getID(),
+  EmitDifference(getDWLabel("info_end", ModuleCU->getID()),
+                 getDWLabel("info_begin", ModuleCU->getID()),
                  true);
   EOL("Compilation Unit Length");
 
@@ -2778,8 +2779,8 @@
   // Start the dwarf pubnames section.
   Asm->OutStreamer.SwitchSection(
                           Asm->getObjFileLowering().getDwarfPubTypesSection());
-  EmitDifference("pubtypes_end", ModuleCU->getID(),
-                 "pubtypes_begin", ModuleCU->getID(), true);
+  EmitDifference(getDWLabel("pubtypes_end", ModuleCU->getID()),
+                 getDWLabel("pubtypes_begin", ModuleCU->getID()), true);
   EOL("Length of Public Types Info");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("pubtypes_begin", ModuleCU->getID()));
@@ -2791,7 +2792,8 @@
                     getTempLabel("section_info"), true, false);
   EOL("Offset of Compilation ModuleCU Info");
 
-  EmitDifference("info_end", ModuleCU->getID(), "info_begin", ModuleCU->getID(),
+  EmitDifference(getDWLabel("info_end", ModuleCU->getID()),
+                 getDWLabel("info_begin", ModuleCU->getID()),
                  true);
   EOL("Compilation ModuleCU Length");
 
@@ -2872,7 +2874,8 @@
 
   // Range 1
   EmitReference("text_begin", 0); EOL("Address");
-  EmitDifference("text_end", 0, "text_begin", 0, true); EOL("Length");
+  EmitDifference(getTempLabel("text_end"), getTempLabel("text_begin"),
+                 true); EOL("Length");
 
   Asm->EmitInt32(0); EOL("EOM (1)");
   Asm->EmitInt32(0); EOL("EOM (2)");
@@ -2925,8 +2928,8 @@
   Asm->OutStreamer.SwitchSection(
                         Asm->getObjFileLowering().getDwarfDebugInlineSection());
 
-  EmitDifference("debug_inlined_end", 1,
-                 "debug_inlined_begin", 1, true);
+  EmitDifference(getDWLabel("debug_inlined_end", 1),
+                 getDWLabel("debug_inlined_begin", 1), true);
   EOL("Length of Debug Inlined Information Entry");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("debug_inlined_begin", 1));

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98002&r1=98001&r2=98002&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 17:02:59 2010
@@ -94,8 +94,8 @@
   Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common", Index));
 
   // Define the eh frame length.
-  EmitDifference("eh_frame_common_end", Index,
-                 "eh_frame_common_begin", Index, true);
+  EmitDifference(getDWLabel("eh_frame_common_end", Index),
+                 getDWLabel("eh_frame_common_begin", Index), true);
   EOL("Length of Common Information Entry");
 
   // EH frame header.
@@ -222,8 +222,8 @@
     O << *EHFrameInfo.FunctionEHSym << ":\n";
 
     // EH frame header.
-    EmitDifference("eh_frame_end", EHFrameInfo.Number,
-                   "eh_frame_begin", EHFrameInfo.Number,
+    EmitDifference(getDWLabel("eh_frame_end", EHFrameInfo.Number),
+                   getDWLabel("eh_frame_begin", EHFrameInfo.Number),
                    true);
     EOL("Length of Frame Information Entry");
 
@@ -238,8 +238,8 @@
 
     EmitReference(getDWLabel("eh_func_begin", EHFrameInfo.Number), FDEEncoding);
     EOL("FDE initial location");
-    EmitDifference("eh_func_end", EHFrameInfo.Number,
-                   "eh_func_begin", EHFrameInfo.Number,
+    EmitDifference(getDWLabel("eh_func_end", EHFrameInfo.Number),
+                   getDWLabel("eh_func_begin", EHFrameInfo.Number),
                    SizeOfEncodedValue(FDEEncoding) == 4);
     EOL("FDE address range");
 
@@ -829,7 +829,8 @@
                        getDWLabel(BeginTag, BeginNumber),
                        true);
       else
-        EmitDifference("label", S.EndLabel, BeginTag, BeginNumber, true);
+        EmitDifference(getDWLabel("label", S.EndLabel), 
+                       getDWLabel(BeginTag, BeginNumber), true);
 
       EOL("Region length");
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98002&r1=98001&r2=98002&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 17:02:59 2010
@@ -276,32 +276,6 @@
   }
 }
 
-/// EmitDifference - Emit the difference between two labels.  If this assembler
-/// supports .set, we emit a .set of a temporary and then use it in the .word.
-void DwarfPrinter::EmitDifference(const char *TagHi, unsigned NumberHi,
-                                  const char *TagLo, unsigned NumberLo,
-                                  bool IsSmall) {
-  if (MAI->hasSetDirective()) {
-    // FIXME: switch to OutStreamer.EmitAssignment.
-    O << "\t.set\t";
-    PrintLabelName("set", SetCounter, Flavor);
-    O << ",";
-    PrintLabelName(getDWLabel(TagHi, NumberHi));
-    O << "-";
-    PrintLabelName(getDWLabel(TagLo, NumberLo));
-    O << "\n";
-    
-    PrintRelDirective(IsSmall);
-    PrintLabelName("set", SetCounter, Flavor);
-    ++SetCounter;
-  } else {
-    PrintRelDirective(IsSmall);
-    PrintLabelName(getDWLabel(TagHi, NumberHi));
-    O << "-";
-    PrintLabelName(getDWLabel(TagLo, NumberLo));
-  }
-}
-
 void DwarfPrinter::EmitSectionOffset(const MCSymbol *Label,
                                      const MCSymbol *Section,
                                      bool IsSmall, bool isEH,
@@ -367,7 +341,8 @@
     // Advance row if new location.
     if (BaseLabel && LabelID && (BaseLabelID != LabelID || !IsLocal)) {
       EmitCFAByte(dwarf::DW_CFA_advance_loc4);
-      EmitDifference("label", LabelID, BaseLabel, BaseLabelID, true);
+      EmitDifference(getDWLabel("label", LabelID),
+                     getDWLabel(BaseLabel, BaseLabelID), true);
       Asm->O << '\n';
 
       BaseLabelID = LabelID;

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98002&r1=98001&r2=98002&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 17:02:59 2010
@@ -139,10 +139,6 @@
   /// EmitDifference - Emit the difference between two labels.
   void EmitDifference(const MCSymbol *LabelHi, const MCSymbol *LabelLo,
                       bool IsSmall = false);
-  void EmitDifference(const char *TagHi, unsigned NumberHi,
-                      const char *TagLo, unsigned NumberLo,
-                      bool IsSmall = false);
-
   void EmitSectionOffset(const MCSymbol *Label, const MCSymbol *Section,
                          bool IsSmall = false, bool isEH = false,
                          bool useSet = true);




From sabre at nondot.org  Mon Mar  8 17:18:21 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 23:18:21 -0000
Subject: [llvm-commits] [llvm] r98004 - in /llvm/trunk/lib/Target:
 ARM/ARMMCAsmInfo.cpp X86/X86MCAsmInfo.cpp
Message-ID: <20100308231821.3C6B02A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 17:18:21 2010
New Revision: 98004

URL: http://llvm.org/viewvc/llvm-project?rev=98004&view=rev
Log:
don't reset defaults.

Modified:
    llvm/trunk/lib/Target/ARM/ARMMCAsmInfo.cpp
    llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMMCAsmInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMMCAsmInfo.cpp?rev=98004&r1=98003&r2=98004&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMMCAsmInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMMCAsmInfo.cpp Mon Mar  8 17:18:21 2010
@@ -48,7 +48,6 @@
 
   // Exceptions handling
   ExceptionsType = ExceptionHandling::SjLj;
-  AbsoluteEHSectionOffsets = false;
 }
 
 ARMELFMCAsmInfo::ARMELFMCAsmInfo() {

Modified: llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp?rev=98004&r1=98003&r2=98004&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp Mon Mar  8 17:18:21 2010
@@ -68,7 +68,6 @@
 
   // Exceptions handling
   ExceptionsType = ExceptionHandling::Dwarf;
-  AbsoluteEHSectionOffsets = false;
 }
 
 X86ELFMCAsmInfo::X86ELFMCAsmInfo(const Triple &Triple) {
@@ -90,7 +89,6 @@
 
   // Exceptions handling
   ExceptionsType = ExceptionHandling::Dwarf;
-  AbsoluteEHSectionOffsets = false;
 }
 
 MCSection *X86ELFMCAsmInfo::getNonexecutableStackSection(MCContext &Ctx) const {




From sabre at nondot.org  Mon Mar  8 17:23:26 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 23:23:26 -0000
Subject: [llvm-commits] [llvm] r98005 - in /llvm/trunk:
 lib/CodeGen/AsmPrinter/DIE.cpp lib/CodeGen/AsmPrinter/DIE.h
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
 lib/CodeGen/AsmPrinter/DwarfException.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.h test/CodeGen/X86/aliases.ll
Message-ID: <20100308232326.311BE2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 17:23:25 2010
New Revision: 98005

URL: http://llvm.org/viewvc/llvm-project?rev=98005&view=rev
Log:
simplify EmitSectionOffset to always use .set if it is
available, the only thing this affects is that we produce
.set in one case we didn't before, which shouldn't harm
anything.  Make EmitSectionOffset call EmitDifference
instead of duplicating it.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
    llvm/trunk/test/CodeGen/X86/aliases.ll

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp Mon Mar  8 17:23:25 2010
@@ -285,7 +285,7 @@
 ///
 void DIESectionOffset::EmitValue(DwarfPrinter *D, unsigned Form) const {
   bool IsSmall = Form == dwarf::DW_FORM_data4;
-  D->EmitSectionOffset(Label, Section, IsSmall, IsEH, UseSet);
+  D->EmitSectionOffset(Label, Section, IsSmall, IsEH);
   D->getAsm()->O << '\n'; // FIXME: Necesssary?
 }
 
@@ -299,7 +299,7 @@
 #ifndef NDEBUG
 void DIESectionOffset::print(raw_ostream &O) {
   O << "Off: " << Label->getName() << "-" << Section->getName()
-    << "-" << IsEH << "-" << UseSet;
+    << "-" << IsEH;
 }
 #endif
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h Mon Mar  8 17:23:25 2010
@@ -336,12 +336,11 @@
     const MCSymbol *Label;
     const MCSymbol *Section;
     bool IsEH : 1;
-    bool UseSet : 1;
   public:
     DIESectionOffset(const MCSymbol *Lab, const MCSymbol *Sec,
-                     bool isEH = false, bool useSet = true)
+                     bool isEH = false)
       : DIEValue(isSectionOffset), Label(Lab), Section(Sec),
-        IsEH(isEH), UseSet(useSet) {}
+        IsEH(isEH) {}
 
     /// EmitValue - Emit section offset.
     ///

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 17:23:25 2010
@@ -376,8 +376,8 @@
 ///
 void DwarfDebug::addSectionOffset(DIE *Die, unsigned Attribute, unsigned Form,
                                   const MCSymbol *Label,const MCSymbol *Section,
-                                  bool isEH, bool useSet) {
-  DIEValue *Value = new DIESectionOffset(Label, Section, isEH, useSet);
+                                  bool isEH) {
+  DIEValue *Value = new DIESectionOffset(Label, Section, isEH);
   DIEValues.push_back(Value);
   Die->addValue(Attribute, Form, Value);
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 17:23:25 2010
@@ -249,7 +249,7 @@
   ///
   void addSectionOffset(DIE *Die, unsigned Attribute, unsigned Form,
                         const MCSymbol *Label, const MCSymbol *Section,
-                        bool isEH = false, bool useSet = true);
+                        bool isEH = false);
 
   /// addDelta - Add a label delta attribute data and value.
   ///

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 17:23:25 2010
@@ -232,7 +232,7 @@
     EmitSectionOffset(getDWLabel("eh_frame_begin", EHFrameInfo.Number),
                       getDWLabel("eh_frame_common",
                                  EHFrameInfo.PersonalityIndex),
-                      true, true, false);
+                      true, true);
 
     EOL("FDE CIE offset");
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 17:23:25 2010
@@ -278,39 +278,18 @@
 
 void DwarfPrinter::EmitSectionOffset(const MCSymbol *Label,
                                      const MCSymbol *Section,
-                                     bool IsSmall, bool isEH,
-                                     bool useSet) {
+                                     bool IsSmall, bool isEH) {
   bool printAbsolute = false;
   if (isEH)
     printAbsolute = MAI->isAbsoluteEHSectionOffsets();
   else
     printAbsolute = MAI->isAbsoluteDebugSectionOffsets();
 
-  if (MAI->hasSetDirective() && useSet) {
-    // FIXME: switch to OutStreamer.EmitAssignment.
-    O << "\t.set\t";
-    PrintLabelName("set", SetCounter, Flavor);
-    O << ",";
-    PrintLabelName(Label);
-
-    if (!printAbsolute) {
-      O << "-";
-      PrintLabelName(Section);
-    }
-
-    O << "\n";
-    PrintRelDirective(IsSmall);
-    PrintLabelName("set", SetCounter, Flavor);
-    ++SetCounter;
-  } else {
-    PrintRelDirective(IsSmall, true);
-    PrintLabelName(Label);
-
-    if (!printAbsolute) {
-      O << "-";
-      PrintLabelName(Section);
-    }
-  }
+  if (!printAbsolute)
+    return EmitDifference(Label, Section, IsSmall);
+  
+  PrintRelDirective(IsSmall, true);
+  PrintLabelName(Label);
 }
 
 /// EmitFrameMoves - Emit frame instructions to describe the layout of the

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 17:23:25 2010
@@ -139,9 +139,9 @@
   /// EmitDifference - Emit the difference between two labels.
   void EmitDifference(const MCSymbol *LabelHi, const MCSymbol *LabelLo,
                       bool IsSmall = false);
+
   void EmitSectionOffset(const MCSymbol *Label, const MCSymbol *Section,
-                         bool IsSmall = false, bool isEH = false,
-                         bool useSet = true);
+                         bool IsSmall = false, bool isEH = false);
   
   /// EmitFrameMoves - Emit frame instructions to describe the layout of the
   /// frame.

Modified: llvm/trunk/test/CodeGen/X86/aliases.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/aliases.ll?rev=98005&r1=98004&r2=98005&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/aliases.ll (original)
+++ llvm/trunk/test/CodeGen/X86/aliases.ll Mon Mar  8 17:23:25 2010
@@ -1,6 +1,6 @@
 ; RUN: llc < %s -mtriple=i686-pc-linux-gnu -asm-verbose=false -o %t
 ; RUN: grep { = } %t   | count 7
-; RUN: grep set %t   | count 16
+; RUN: grep set %t   | count 18
 ; RUN: grep globl %t | count 6
 ; RUN: grep weak %t  | count 1
 ; RUN: grep hidden %t | count 1




From evan.cheng at apple.com  Mon Mar  8 17:28:08 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 08 Mar 2010 23:28:08 -0000
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100308232808.91F8F2A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 17:28:08 2010
New Revision: 98007

URL: http://llvm.org/viewvc/llvm-project?rev=98007&view=rev
Log:
Restrict machine cse to really trivial coalescing. Leave the heavy lifting to a real coalescer.

Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98007&r1=98006&r2=98007&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 17:28:08 2010
@@ -91,7 +91,10 @@
     unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
     if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
         TargetRegisterInfo::isVirtualRegister(SrcReg) &&
+        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
         !SrcSubIdx && !DstSubIdx) {
+      DEBUG(dbgs() << "Coalescing: " << *DefMI);
+      DEBUG(dbgs() << "*** to: " << *MI);
       MO.setReg(SrcReg);
       DefMI->eraseFromParent();
       ++NumCoalesces;




From evan.cheng at apple.com  Mon Mar  8 17:49:12 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 08 Mar 2010 23:49:12 -0000
Subject: [llvm-commits] [llvm] r98009 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100308234912.443DB2A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 17:49:12 2010
New Revision: 98009

URL: http://llvm.org/viewvc/llvm-project?rev=98009&view=rev
Log:
Don't waste time trying to CSE labels, phis, inline asm. Definitely avoid cse implicit-def for obvious performance reason.

Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98009&r1=98008&r2=98009&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 17:49:12 2010
@@ -171,6 +171,10 @@
 }
 
 bool MachineCSE::isCSECandidate(MachineInstr *MI) {
+  if (MI->isLabel() || MI->isPHI() || MI->isImplicitDef() ||
+      MI->isKill() || MI->isInlineAsm())
+    return false;
+
   // Ignore copies or instructions that read / write physical registers
   // (except for dead defs of physical registers).
   unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;




From sabre at nondot.org  Mon Mar  8 17:58:37 2010
From: sabre at nondot.org (Chris Lattner)
Date: Mon, 08 Mar 2010 23:58:37 -0000
Subject: [llvm-commits] [llvm] r98010 - in /llvm/trunk:
 include/llvm/CodeGen/AsmPrinter.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.h test/CodeGen/X86/aliases.ll
Message-ID: <20100308235837.ED6672A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 17:58:37 2010
New Revision: 98010

URL: http://llvm.org/viewvc/llvm-project?rev=98010&view=rev
Log:
move .set generation out of DwarfPrinter into AsmPrinter and 
MCize it.

Modified:
    llvm/trunk/include/llvm/CodeGen/AsmPrinter.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
    llvm/trunk/test/CodeGen/X86/aliases.ll

Modified: llvm/trunk/include/llvm/CodeGen/AsmPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/AsmPrinter.h?rev=98010&r1=98009&r2=98010&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/AsmPrinter.h (original)
+++ llvm/trunk/include/llvm/CodeGen/AsmPrinter.h Mon Mar  8 17:58:37 2010
@@ -136,6 +136,7 @@
     mutable const MachineInstr *LastMI;
     mutable const Function *LastFn;
     mutable unsigned Counter;
+    mutable unsigned SetCounter;
     
     // Private state for processDebugLoc()
     mutable const MDNode *PrevDLT;
@@ -275,6 +276,13 @@
     /// EmitInt64 - Emit a long long directive and value.
     ///
     void EmitInt64(uint64_t Value) const;
+    
+    
+    /// EmitLabelDifference - Emit something like ".long Hi-Lo" where the size
+    /// in bytes of the directive is specified by Size and Hi/Lo specify the
+    /// labels.  This implicitly uses .set if it is available.
+    void EmitLabelDifference(const MCSymbol *Hi, const MCSymbol *Lo,
+                             unsigned Size) const;
 
     //===------------------------------------------------------------------===//
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98010&r1=98009&r2=98010&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 17:58:37 2010
@@ -61,7 +61,7 @@
   : MachineFunctionPass(&ID), O(o),
     TM(tm), MAI(T), TRI(tm.getRegisterInfo()),
     OutContext(Ctx), OutStreamer(Streamer),
-    LastMI(0), LastFn(0), Counter(~0U), PrevDLT(NULL) {
+    LastMI(0), LastFn(0), Counter(~0U), SetCounter(0), PrevDLT(NULL) {
   DW = 0; MMI = 0;
   VerboseAsm = Streamer.isVerboseAsm();
 }
@@ -893,6 +893,33 @@
   OutStreamer.EmitIntValue(Value, 8, 0/*addrspace*/);
 }
 
+/// EmitLabelDifference - Emit something like ".long Hi-Lo" where the size
+/// in bytes of the directive is specified by Size and Hi/Lo specify the
+/// labels.  This implicitly uses .set if it is available.
+void AsmPrinter::EmitLabelDifference(const MCSymbol *Hi, const MCSymbol *Lo,
+                                     unsigned Size) const {
+  // Get the Hi-Lo expression.
+  const MCExpr *Diff = 
+    MCBinaryExpr::CreateSub(MCSymbolRefExpr::Create(Hi, OutContext),
+                            MCSymbolRefExpr::Create(Lo, OutContext),
+                            OutContext);
+  
+  if (!MAI->hasSetDirective()) {
+    OutStreamer.EmitValue(Diff, Size, 0/*AddrSpace*/);
+    return;
+  }
+
+  // Otherwise, emit with .set (aka assignment).
+  MCSymbol *SetLabel =
+    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) + "set" +
+                                 Twine(SetCounter++));
+  OutStreamer.EmitAssignment(SetLabel, Diff);
+  
+  OutStreamer.EmitValue(MCSymbolRefExpr::Create(SetLabel, OutContext),
+                        Size, 0/*AddrSpace*/);
+}
+
+
 //===----------------------------------------------------------------------===//
 
 // EmitAlignment - Emit an alignment directive to the specified power of

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98010&r1=98009&r2=98010&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 17:58:37 2010
@@ -35,7 +35,7 @@
                            const char *flavor)
 : O(OS), Asm(A), MAI(T), TD(Asm->TM.getTargetData()),
   RI(Asm->TM.getRegisterInfo()), M(NULL), MF(NULL), MMI(NULL),
-  SubprogramCount(0), Flavor(flavor), SetCounter(1) {}
+  SubprogramCount(0), Flavor(flavor) {}
 
 
 /// getDWLabel - Return the MCSymbol corresponding to the assembler temporary
@@ -243,7 +243,7 @@
   O << *TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);;
 }
 
-void DwarfPrinter::EmitReference(const GlobalValue *GV, unsigned Encoding)const {
+void DwarfPrinter::EmitReference(const GlobalValue *GV, unsigned Encoding)const{
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
   PrintRelDirective(Encoding);
@@ -255,25 +255,8 @@
 /// supports .set, we emit a .set of a temporary and then use it in the .word.
 void DwarfPrinter::EmitDifference(const MCSymbol *TagHi, const MCSymbol *TagLo,
                                   bool IsSmall) {
-  if (MAI->hasSetDirective()) {
-    // FIXME: switch to OutStreamer.EmitAssignment.
-    O << "\t.set\t";
-    PrintLabelName("set", SetCounter, Flavor);
-    O << ",";
-    PrintLabelName(TagHi);
-    O << "-";
-    PrintLabelName(TagLo);
-    O << "\n";
-
-    PrintRelDirective(IsSmall);
-    PrintLabelName("set", SetCounter, Flavor);
-    ++SetCounter;
-  } else {
-    PrintRelDirective(IsSmall);
-    PrintLabelName(TagHi);
-    O << "-";
-    PrintLabelName(TagLo);
-  }
+  unsigned Size = IsSmall ? 4 : TD->getPointerSize();
+  Asm->EmitLabelDifference(TagHi, TagLo, Size);
 }
 
 void DwarfPrinter::EmitSectionOffset(const MCSymbol *Label,

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98010&r1=98009&r2=98010&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 17:58:37 2010
@@ -70,9 +70,6 @@
   /// unique labels.
   const char * const Flavor;
 
-  /// SetCounter - A unique number for each '.set' directive.
-  unsigned SetCounter;
-
   DwarfPrinter(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T,
                const char *flavor);
 public:

Modified: llvm/trunk/test/CodeGen/X86/aliases.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/aliases.ll?rev=98010&r1=98009&r2=98010&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/aliases.ll (original)
+++ llvm/trunk/test/CodeGen/X86/aliases.ll Mon Mar  8 17:58:37 2010
@@ -1,5 +1,5 @@
 ; RUN: llc < %s -mtriple=i686-pc-linux-gnu -asm-verbose=false -o %t
-; RUN: grep { = } %t   | count 7
+; RUN: grep { = } %t   | count 16
 ; RUN: grep set %t   | count 18
 ; RUN: grep globl %t | count 6
 ; RUN: grep weak %t  | count 1




From sabre at nondot.org  Mon Mar  8 18:00:15 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:00:15 -0000
Subject: [llvm-commits] [llvm] r98011 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfException.cpp
 DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100309000015.79A122A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:00:15 2010
New Revision: 98011

URL: http://llvm.org/viewvc/llvm-project?rev=98011&view=rev
Log:
now that the debug and eh emitters use a common .set counter,
we can eliminate "flavor".

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98011&r1=98010&r2=98011&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 18:00:15 2010
@@ -288,7 +288,7 @@
 } // end llvm namespace
 
 DwarfDebug::DwarfDebug(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T)
-  : DwarfPrinter(OS, A, T, "dbg"), ModuleCU(0),
+  : DwarfPrinter(OS, A, T), ModuleCU(0),
     AbbreviationsSet(InitAbbreviationsSetSize), Abbreviations(),
     DIEValues(), StringPool(),
     SectionSourceLines(), didInitial(false), shouldEmit(false),

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98011&r1=98010&r2=98011&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Mon Mar  8 18:00:15 2010
@@ -39,7 +39,7 @@
 
 DwarfException::DwarfException(raw_ostream &OS, AsmPrinter *A,
                                const MCAsmInfo *T)
-  : DwarfPrinter(OS, A, T, "eh"), shouldEmitTable(false),shouldEmitMoves(false),
+  : DwarfPrinter(OS, A, T), shouldEmitTable(false),shouldEmitMoves(false),
     shouldEmitTableModule(false), shouldEmitMovesModule(false),
     ExceptionTimer(0) {
   if (TimePassesIsEnabled)

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98011&r1=98010&r2=98011&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 18:00:15 2010
@@ -31,11 +31,10 @@
 #include "llvm/ADT/SmallString.h"
 using namespace llvm;
 
-DwarfPrinter::DwarfPrinter(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T,
-                           const char *flavor)
+DwarfPrinter::DwarfPrinter(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T)
 : O(OS), Asm(A), MAI(T), TD(Asm->TM.getTargetData()),
   RI(Asm->TM.getRegisterInfo()), M(NULL), MF(NULL), MMI(NULL),
-  SubprogramCount(0), Flavor(flavor) {}
+  SubprogramCount(0) {}
 
 
 /// getDWLabel - Return the MCSymbol corresponding to the assembler temporary

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98011&r1=98010&r2=98011&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 18:00:15 2010
@@ -66,12 +66,7 @@
   /// SubprogramCount - The running count of functions being compiled.
   unsigned SubprogramCount;
 
-  /// Flavor - A unique string indicating what dwarf producer this is, used to
-  /// unique labels.
-  const char * const Flavor;
-
-  DwarfPrinter(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T,
-               const char *flavor);
+  DwarfPrinter(raw_ostream &OS, AsmPrinter *A, const MCAsmInfo *T);
 public:
   
   //===------------------------------------------------------------------===//




From sabre at nondot.org  Mon Mar  8 18:00:57 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:00:57 -0000
Subject: [llvm-commits] [llvm] r98012 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100309000057.3230F2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:00:57 2010
New Revision: 98012

URL: http://llvm.org/viewvc/llvm-project?rev=98012&view=rev
Log:
remove the suffix form of PrintLabelName, which was only 
used for 'flavor'.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98012&r1=98011&r2=98012&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 18:00:57 2010
@@ -218,14 +218,6 @@
   O << Label->getName();
 }
 
-void DwarfPrinter::PrintLabelName(const char *Tag, unsigned Number,
-                                  const char *Suffix) const {
-  // FIXME: REMOVE.
-  O << MAI->getPrivateGlobalPrefix() << Tag;
-  if (Number) O << Number;
-  O << Suffix;
-}
-
 /// EmitReference - Emit a reference to a label.
 ///
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, bool IsPCRelative,

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98012&r1=98011&r2=98012&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 18:00:57 2010
@@ -117,8 +117,6 @@
   /// PrintLabelName - Print label name in form used by Dwarf writer.
   ///
   void PrintLabelName(const MCSymbol *Label) const;
-  void PrintLabelName(const char *Tag, unsigned Number,
-                      const char *Suffix) const;
 
   /// EmitReference - Emit a reference to a label.
   ///




From gohman at apple.com  Mon Mar  8 18:07:36 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 09 Mar 2010 00:07:36 -0000
Subject: [llvm-commits] [llvm] r98013 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
Message-ID: <20100309000736.537482A6C12C@llvm.org>

Author: djg
Date: Mon Mar  8 18:07:36 2010
New Revision: 98013

URL: http://llvm.org/viewvc/llvm-project?rev=98013&view=rev
Log:
Print the correct index in the "match failed at index" message.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=98013&r1=98012&r2=98013&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Mon Mar  8 18:07:36 2010
@@ -2680,7 +2680,7 @@
       NodeStack.append(LastScope.NodeStack.begin(), LastScope.NodeStack.end());
       N = NodeStack.back();
 
-      DEBUG(errs() << "  Match failed at index " << MatcherIndex
+      DEBUG(errs() << "  Match failed at index " << (MatcherIndex-1)
                    << " continuing at " << LastScope.FailIndex << "\n");
     
       if (LastScope.NumMatchedMemRefs != MatchedMemRefs.size())




From sabre at nondot.org  Mon Mar  8 18:17:59 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:17:59 -0000
Subject: [llvm-commits] [llvm] r98014 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100309001759.1430A2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:17:58 2010
New Revision: 98014

URL: http://llvm.org/viewvc/llvm-project?rev=98014&view=rev
Log:
eliminate an argument from PrintRelDirective, sinking
the one special case into EmitSectionOffset.  MCize
the non-special case in EmitSectionOffset.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98014&r1=98013&r2=98014&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 18:17:58 2010
@@ -77,10 +77,8 @@
   return 0;
 }
 
-void DwarfPrinter::PrintRelDirective(bool Force32Bit, bool isInSection) const {
-  if (isInSection && MAI->getDwarfSectionOffsetDirective())
-    O << MAI->getDwarfSectionOffsetDirective();
-  else if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
+void DwarfPrinter::PrintRelDirective(bool Force32Bit) const {
+  if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
     O << MAI->getData32bitsDirective();
   else
     O << MAI->getData64bitsDirective();
@@ -253,17 +251,23 @@
 void DwarfPrinter::EmitSectionOffset(const MCSymbol *Label,
                                      const MCSymbol *Section,
                                      bool IsSmall, bool isEH) {
-  bool printAbsolute = false;
+  bool isAbsolute;
   if (isEH)
-    printAbsolute = MAI->isAbsoluteEHSectionOffsets();
+    isAbsolute = MAI->isAbsoluteEHSectionOffsets();
   else
-    printAbsolute = MAI->isAbsoluteDebugSectionOffsets();
+    isAbsolute = MAI->isAbsoluteDebugSectionOffsets();
 
-  if (!printAbsolute)
+  if (!isAbsolute)
     return EmitDifference(Label, Section, IsSmall);
   
-  PrintRelDirective(IsSmall, true);
-  PrintLabelName(Label);
+  // On COFF targets, we have to emit the weird .secrel32 directive.
+  if (const char *SecOffDir = MAI->getDwarfSectionOffsetDirective())
+    O << SecOffDir << Label->getName();
+  else {
+    unsigned Size = IsSmall ? 4 : TD->getPointerSize();
+    Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(Label, Asm->OutContext),
+                               Size, 0/*AddrSpace*/);
+  }
 }
 
 /// EmitFrameMoves - Emit frame instructions to describe the layout of the

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98014&r1=98013&r2=98014&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 18:17:58 2010
@@ -89,8 +89,7 @@
   unsigned SizeOfEncodedValue(unsigned Encoding) const;
 
   void PrintRelDirective(unsigned Encoding) const;
-  void PrintRelDirective(bool Force32Bit = false,
-                         bool isInSection = false) const;
+  void PrintRelDirective(bool Force32Bit = false) const;
 
   /// EOL - Print a newline character to asm stream.  If a comment is present
   /// then it will be printed first.  Comments should not contain '\n'.
@@ -130,6 +129,8 @@
   void EmitDifference(const MCSymbol *LabelHi, const MCSymbol *LabelLo,
                       bool IsSmall = false);
 
+  /// EmitSectionOffset - Emit Label-Section or use a special purpose directive
+  /// to emit a section offset if the target has one.
   void EmitSectionOffset(const MCSymbol *Label, const MCSymbol *Section,
                          bool IsSmall = false, bool isEH = false);
   




From sabre at nondot.org  Mon Mar  8 18:26:09 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:26:09 -0000
Subject: [llvm-commits] [llvm] r98015 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100309002609.E8A1E2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:26:09 2010
New Revision: 98015

URL: http://llvm.org/viewvc/llvm-project?rev=98015&view=rev
Log:
mc'ize the last use of PrintLabelName and eliminate PrintLabelName.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98015&r1=98014&r2=98015&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 18:26:09 2010
@@ -15,6 +15,7 @@
 #include "llvm/Module.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
+#include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCAsmInfo.h"
@@ -2967,12 +2968,11 @@
       DIE *SP = LI->second;
       Asm->EmitInt32(SP->getOffset()); EOL("DIE offset");
 
-      if (TD->getPointerSize() == sizeof(int32_t))
-        O << MAI->getData32bitsDirective();
-      else
-        O << MAI->getData64bitsDirective();
-
-      PrintLabelName(getDWLabel("label", LI->first)); EOL("low_pc");
+      // FIXME: "Labels" should hold MCSymbol*'s
+      MCSymbol *L = getDWLabel("label", LI->first);
+      if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("low_pc");
+      Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(L, Asm->OutContext),
+                                 TD->getPointerSize(), 0/*AddrSpace*/);
     }
   }
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98015&r1=98014&r2=98015&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 18:26:09 2010
@@ -209,13 +209,6 @@
 }
 
 
-/// PrintLabelName - Print label name in form used by Dwarf writer.
-///
-void DwarfPrinter::PrintLabelName(const MCSymbol *Label) const {
-  // FIXME: REMOVE.
-  O << Label->getName();
-}
-
 /// EmitReference - Emit a reference to a label.
 ///
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, bool IsPCRelative,
@@ -229,7 +222,7 @@
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
   PrintRelDirective(Encoding);
-  O << *TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);;
+  O << *TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);
 }
 
 void DwarfPrinter::EmitReference(const GlobalValue *GV, unsigned Encoding)const{
@@ -300,8 +293,6 @@
       EmitCFAByte(dwarf::DW_CFA_advance_loc4);
       EmitDifference(getDWLabel("label", LabelID),
                      getDWLabel(BaseLabel, BaseLabelID), true);
-      Asm->O << '\n';
-
       BaseLabelID = LabelID;
       BaseLabel = "label";
       IsLocal = true;

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98015&r1=98014&r2=98015&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Mon Mar  8 18:26:09 2010
@@ -113,10 +113,6 @@
                    unsigned PadTo = 0) const;
 
   
-  /// PrintLabelName - Print label name in form used by Dwarf writer.
-  ///
-  void PrintLabelName(const MCSymbol *Label) const;
-
   /// EmitReference - Emit a reference to a label.
   ///
   void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,




From sabre at nondot.org  Mon Mar  8 18:31:02 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:31:02 -0000
Subject: [llvm-commits] [llvm] r98016 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <20100309003102.B01C62A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:31:02 2010
New Revision: 98016

URL: http://llvm.org/viewvc/llvm-project?rev=98016&view=rev
Log:
make InlineInfoLabels hold MCSymbol*'s, avoiding 
recomputation of the labels.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98016&r1=98015&r2=98016&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 18:31:02 2010
@@ -1405,8 +1405,9 @@
   addDIEEntry(ScopeDIE, dwarf::DW_AT_abstract_origin,
               dwarf::DW_FORM_ref4, OriginDIE);
 
-  addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           getDWLabel("label", StartID));
+  MCSymbol *StartLabel = getDWLabel("label", StartID);
+  
+  addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr, StartLabel);
   addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
            getDWLabel("label", EndID));
 
@@ -1417,11 +1418,11 @@
     I = InlineInfo.find(InlinedSP.getNode());
 
   if (I == InlineInfo.end()) {
-    InlineInfo[InlinedSP.getNode()].push_back(std::make_pair(StartID,
+    InlineInfo[InlinedSP.getNode()].push_back(std::make_pair(StartLabel,
                                                              ScopeDIE));
     InlinedSPNodes.push_back(InlinedSP.getNode());
   } else
-    I->second.push_back(std::make_pair(StartID, ScopeDIE));
+    I->second.push_back(std::make_pair(StartLabel, ScopeDIE));
 
   StringPool.insert(InlinedSP.getName());
   StringPool.insert(getRealLinkageName(InlinedSP.getLinkageName()));
@@ -2965,13 +2966,12 @@
 
     for (SmallVector<InlineInfoLabels, 4>::iterator LI = Labels.begin(),
            LE = Labels.end(); LI != LE; ++LI) {
-      DIE *SP = LI->second;
-      Asm->EmitInt32(SP->getOffset()); EOL("DIE offset");
+      if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DIE offset");
+      Asm->EmitInt32(LI->second->getOffset());
 
-      // FIXME: "Labels" should hold MCSymbol*'s
-      MCSymbol *L = getDWLabel("label", LI->first);
       if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("low_pc");
-      Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(L, Asm->OutContext),
+      Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(LI->first,
+                                                         Asm->OutContext),
                                  TD->getPointerSize(), 0/*AddrSpace*/);
     }
   }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98016&r1=98015&r2=98016&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 18:31:02 2010
@@ -167,8 +167,8 @@
 
   /// InlineInfo - Keep track of inlined functions and their location.  This
   /// information is used to populate debug_inlined section.
-  typedef std::pair<unsigned, DIE *> InlineInfoLabels;
-  DenseMap<MDNode *, SmallVector<InlineInfoLabels, 4> > InlineInfo;
+  typedef std::pair<MCSymbol*, DIE *> InlineInfoLabels;
+  DenseMap<MDNode*, SmallVector<InlineInfoLabels, 4> > InlineInfo;
   SmallVector<MDNode *, 4> InlinedSPNodes;
 
   /// CompileUnitOffsets - A vector of the offsets of the compile units. This is




From clattner at apple.com  Mon Mar  8 18:35:51 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 8 Mar 2010 16:35:51 -0800
Subject: [llvm-commits] [llvm] r98013 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
In-Reply-To: <20100309000736.537482A6C12C@llvm.org>
References: <20100309000736.537482A6C12C@llvm.org>
Message-ID: <5A53462C-9733-436F-9E24-F0960384AC05@apple.com>


On Mar 8, 2010, at 4:07 PM, Dan Gohman wrote:

> Author: djg
> Date: Mon Mar  8 18:07:36 2010
> New Revision: 98013
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98013&view=rev
> Log:
> Print the correct index in the "match failed at index" message.

This isn't correct for multi-byte opcodes.  Please capture the MatcherIndex before dispatching on the opcode and print that value.

-Chris

> 
> Modified:
>    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=98013&r1=98012&r2=98013&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Mon Mar  8 18:07:36 2010
> @@ -2680,7 +2680,7 @@
>       NodeStack.append(LastScope.NodeStack.begin(), LastScope.NodeStack.end());
>       N = NodeStack.back();
> 
> -      DEBUG(errs() << "  Match failed at index " << MatcherIndex
> +      DEBUG(errs() << "  Match failed at index " << (MatcherIndex-1)
>                    << " continuing at " << LastScope.FailIndex << "\n");
> 
>       if (LastScope.NumMatchedMemRefs != MatchedMemRefs.size())
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From sabre at nondot.org  Mon Mar  8 18:39:25 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 00:39:25 -0000
Subject: [llvm-commits] [llvm] r98017 - in /llvm/trunk:
 include/llvm/MC/MCStreamer.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp lib/MC/MCStreamer.cpp
Message-ID: <20100309003925.365BE2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 18:39:24 2010
New Revision: 98017

URL: http://llvm.org/viewvc/llvm-project?rev=98017&view=rev
Log:
add a EmitSymbolValue convenience method to MCStreamer.

Modified:
    llvm/trunk/include/llvm/MC/MCStreamer.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/MC/MCStreamer.cpp

Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=98017&r1=98016&r2=98017&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Mon Mar  8 18:39:24 2010
@@ -191,6 +191,11 @@
     /// EmitIntValue - Special case of EmitValue that avoids the client having
     /// to pass in a MCExpr for constant integers.
     virtual void EmitIntValue(uint64_t Value, unsigned Size,unsigned AddrSpace);
+
+    /// EmitSymbolValue - Special case of EmitValue that avoids the client
+    /// having to pass in a MCExpr for MCSymbols.
+    virtual void EmitSymbolValue(const MCSymbol *Sym, unsigned Size,
+                                 unsigned AddrSpace);
     
     /// EmitGPRel32Value - Emit the expression @p Value into the output as a
     /// gprel32 (32-bit GP relative) value.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98017&r1=98016&r2=98017&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 18:39:24 2010
@@ -914,9 +914,7 @@
     OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) + "set" +
                                  Twine(SetCounter++));
   OutStreamer.EmitAssignment(SetLabel, Diff);
-  
-  OutStreamer.EmitValue(MCSymbolRefExpr::Create(SetLabel, OutContext),
-                        Size, 0/*AddrSpace*/);
+  OutStreamer.EmitSymbolValue(SetLabel, Size, 0/*AddrSpace*/);
 }
 
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98017&r1=98016&r2=98017&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 18:39:24 2010
@@ -10,12 +10,12 @@
 // This file contains support for writing dwarf debug info into asm files.
 //
 //===----------------------------------------------------------------------===//
+
 #define DEBUG_TYPE "dwarfdebug"
 #include "DwarfDebug.h"
 #include "llvm/Module.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
-#include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCAsmInfo.h"
@@ -2970,9 +2970,7 @@
       Asm->EmitInt32(LI->second->getOffset());
 
       if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("low_pc");
-      Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(LI->first,
-                                                         Asm->OutContext),
-                                 TD->getPointerSize(), 0/*AddrSpace*/);
+      Asm->OutStreamer.EmitSymbolValue(LI->first, TD->getPointerSize(), 0);
     }
   }
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98017&r1=98016&r2=98017&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 18:39:24 2010
@@ -258,8 +258,7 @@
     O << SecOffDir << Label->getName();
   else {
     unsigned Size = IsSmall ? 4 : TD->getPointerSize();
-    Asm->OutStreamer.EmitValue(MCSymbolRefExpr::Create(Label, Asm->OutContext),
-                               Size, 0/*AddrSpace*/);
+    Asm->OutStreamer.EmitSymbolValue(Label, Size, 0/*AddrSpace*/);
   }
 }
 

Modified: llvm/trunk/lib/MC/MCStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCStreamer.cpp?rev=98017&r1=98016&r2=98017&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCStreamer.cpp Mon Mar  8 18:39:24 2010
@@ -31,6 +31,11 @@
   EmitValue(MCConstantExpr::Create(Value, getContext()), Size, AddrSpace);
 }
 
+void MCStreamer::EmitSymbolValue(const MCSymbol *Sym, unsigned Size,
+                                 unsigned AddrSpace) {
+  EmitValue(MCSymbolRefExpr::Create(Sym, getContext()), Size, AddrSpace);
+}
+
 /// EmitFill - Emit NumBytes bytes worth of the value specified by
 /// FillValue.  This implements directives such as '.space'.
 void MCStreamer::EmitFill(uint64_t NumBytes, uint8_t FillValue,




From isanbard at gmail.com  Mon Mar  8 18:40:17 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 00:40:17 -0000
Subject: [llvm-commits] [llvm] r98018 - in /llvm/trunk/lib/Target:
 ARM/AsmPrinter/ARMAsmPrinter.cpp X86/AsmPrinter/X86AsmPrinter.cpp
Message-ID: <20100309004017.5267B2A6C12C@llvm.org>

Author: void
Date: Mon Mar  8 18:40:17 2010
New Revision: 98018

URL: http://llvm.org/viewvc/llvm-project?rev=98018&view=rev
Log:
MC-ize the stub printing in ARM.

Modified:
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98018&r1=98017&r2=98018&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Mon Mar  8 18:40:17 2010
@@ -33,6 +33,7 @@
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCStreamer.h"
@@ -1128,17 +1129,29 @@
       OutStreamer.SwitchSection(TLOFMacho.getNonLazySymbolPointerSection());
       EmitAlignment(2);
       for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
-        O << *Stubs[i].first << ":\n\t.indirect_symbol ";
-        O << *Stubs[i].second << "\n\t.long\t0\n";
+        // L_foo$stub:
+        OutStreamer.EmitLabel(Stubs[i].first);
+        //   .indirect_symbol _foo
+        OutStreamer.EmitSymbolAttribute(Stubs[i].second, MCSA_IndirectSymbol);
+        OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
       }
+
+      Stubs.clear();
+      OutStreamer.AddBlankLine();
     }
 
     Stubs = MMIMacho.GetHiddenGVStubList();
     if (!Stubs.empty()) {
       OutStreamer.SwitchSection(getObjFileLowering().getDataSection());
       EmitAlignment(2);
-      for (unsigned i = 0, e = Stubs.size(); i != e; ++i)
-        O << *Stubs[i].first << ":\n\t.long " << *Stubs[i].second << "\n";
+      for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
+        // L_foo$stub:
+        OutStreamer.EmitLabel(Stubs[i].first);
+        //   .long _foo
+        OutStreamer.EmitValue(MCSymbolRefExpr::Create(Stubs[i].second,
+                                                      OutContext),
+                              4/*size*/, 0/*addrspace*/);
+      }
     }
 
     // Funny Darwin hack: This flag tells the linker that no global symbols

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98018&r1=98017&r2=98018&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Mon Mar  8 18:40:17 2010
@@ -26,6 +26,7 @@
 #include "llvm/Module.h"
 #include "llvm/Type.h"
 #include "llvm/Assembly/Writer.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSectionMachO.h"
@@ -36,7 +37,6 @@
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/FormattedStream.h"
-#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/Target/Mangler.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/Target/TargetRegistry.h"




From isanbard at gmail.com  Mon Mar  8 18:43:34 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 00:43:34 -0000
Subject: [llvm-commits] [llvm] r98019 -
	/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
Message-ID: <20100309004334.EADFB2A6C12C@llvm.org>

Author: void
Date: Mon Mar  8 18:43:34 2010
New Revision: 98019

URL: http://llvm.org/viewvc/llvm-project?rev=98019&view=rev
Log:
Print blank line and clear stubs vector.

Modified:
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98019&r1=98018&r2=98019&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Mon Mar  8 18:43:34 2010
@@ -1132,8 +1132,16 @@
         // L_foo$stub:
         OutStreamer.EmitLabel(Stubs[i].first);
         //   .indirect_symbol _foo
-        OutStreamer.EmitSymbolAttribute(Stubs[i].second, MCSA_IndirectSymbol);
-        OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
+        MCSymbol *MCSym = Stubs[i].second;
+        OutStreamer.EmitSymbolAttribute(MCSym, MCSA_IndirectSymbol);
+
+        if (MCSym->isUndefined())
+          // External to current translation unit.
+          OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
+        else
+          // Internal to current translation unit.
+          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym, OutContext),
+                                4/*size*/, 0/*addrspace*/);
       }
 
       Stubs.clear();
@@ -1152,6 +1160,9 @@
                                                       OutContext),
                               4/*size*/, 0/*addrspace*/);
       }
+
+      Stubs.clear();
+      OutStreamer.AddBlankLine();
     }
 
     // Funny Darwin hack: This flag tells the linker that no global symbols




From dpatel at apple.com  Mon Mar  8 18:44:10 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 09 Mar 2010 00:44:10 -0000
Subject: [llvm-commits] [llvm] r98020 - in /llvm/trunk:
 docs/SourceLevelDebugging.html include/llvm/Analysis/DebugInfo.h
 include/llvm/Support/Dwarf.h lib/Analysis/DebugInfo.cpp
 test/FrontendC++/2010-02-17-DbgArtificialArg.cpp
Message-ID: <20100309004410.6D7912A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 18:44:10 2010
New Revision: 98020

URL: http://llvm.org/viewvc/llvm-project?rev=98020&view=rev
Log:
Start using DIFile.  See updated SourceLevelDebugging.html for more information.

This patch updates LLVMDebugVersion to 8.
Debug info descriptors encoded using LLVMDebugVersion 7 is supported.

Corresponding llvmgcc and clang FE commits are required.


Modified:
    llvm/trunk/docs/SourceLevelDebugging.html
    llvm/trunk/include/llvm/Analysis/DebugInfo.h
    llvm/trunk/include/llvm/Support/Dwarf.h
    llvm/trunk/lib/Analysis/DebugInfo.cpp
    llvm/trunk/test/FrontendC++/2010-02-17-DbgArtificialArg.cpp

Modified: llvm/trunk/docs/SourceLevelDebugging.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/SourceLevelDebugging.html?rev=98020&r1=98019&r2=98020&view=diff
==============================================================================
--- llvm/trunk/docs/SourceLevelDebugging.html (original)
+++ llvm/trunk/docs/SourceLevelDebugging.html Mon Mar  8 18:44:10 2010
@@ -25,6 +25,7 @@
     <li><a href="#debug_info_descriptors">Debug information descriptors</a>
     <ul>
       <li><a href="#format_compile_units">Compile unit descriptors</a></li>
+      <li><a href="#format_files">File descriptors</a></li>
       <li><a href="#format_global_variables">Global variable descriptors</a></li>
       <li><a href="#format_subprograms">Subprogram descriptors</a></li>
       <li><a href="#format_blocks">Block descriptors</a></li>
@@ -255,7 +256,7 @@
 <p>To provide basic functionality, the LLVM debugger does have to make some
    assumptions about the source-level language being debugged, though it keeps
    these to a minimum.  The only common features that the LLVM debugger assumes
-   exist are <a href="#format_compile_units">source files</a>,
+   exist are <a href="#format_files">source files</a>,
    and <a href="#format_global_variables">program objects</a>.  These abstract
    objects are used by a debugger to form stack traces, show information about
    local variables, etc.</p>
@@ -307,8 +308,8 @@
    of tags are loosely bound to the tag values of DWARF information entries.
    However, that does not restrict the use of the information supplied to DWARF
    targets.  To facilitate versioning of debug information, the tag is augmented
-   with the current debug version (LLVMDebugVersion = 7 << 16 or 0x70000 or
-   458752.)</a></p>
+   with the current debug version (LLVMDebugVersion = 8 << 16 or 0x80000 or
+   524288.)</a></p>
 
 <p>The details of the various descriptors follow.</p>  
 
@@ -346,18 +347,36 @@
    that produced it.</p>
 
 <p>Compile unit descriptors provide the root context for objects declared in a
-   specific source file.  Global variables and top level functions would be
-   defined using this context. Compile unit descriptors also provide context
-   for source line correspondence.</p>
-
-<p>Each input file is encoded as a separate compile unit in LLVM debugging
-   information output. However, many target specific tool chains prefer to
-   encode only one compile unit in an object file. In this situation, the LLVM
-   code generator will include debugging information entities in the compile
-   unit that is marked as main compile unit. The code generator accepts maximum
-   one main compile unit per module. If a module does not contain any main
-   compile unit then the code generator will emit multiple compile units in the
-   output object file.</p>
+   specific compilation unit. File descriptors are defined using this context.</p>
+
+</div>
+
+<!-- ======================================================================= -->
+<div class="doc_subsubsection">
+  <a name="format_files">File descriptors</a>
+</div>
+
+<div class="doc_text">
+
+<div class="doc_code">
+<pre>
+!0 = metadata !{
+  i32,       ;; Tag = 41 + <a href="#LLVMDebugVersion">LLVMDebugVersion</a> 
+             ;; (DW_TAG_file_type)
+  metadata,  ;; Source file name
+  metadata,  ;; Source file directory (includes trailing slash)
+  metadata   ;; Reference to compile unit where defined
+}
+</pre>
+</div>
+
+<p>These descriptors contain informations for a file. Global variables and top
+   level functions would be defined using this context.k File descriptors also
+   provide context for source line correspondence. </p>
+
+<p>Each input file is encoded as a separate file descriptor in LLVM debugging
+   information output. Each file descriptor would be defined using a 
+   compile unit. </p>
 
 </div>
 
@@ -378,7 +397,7 @@
   metadata, ;; Name
   metadata, ;; Display name (fully qualified C++ name)
   metadata, ;; MIPS linkage name (for C++)
-  metadata, ;; Reference to compile unit where defined
+  metadata, ;; Reference to file where defined
   i32,      ;; Line number where defined
   metadata, ;; Reference to type descriptor
   i1,       ;; True if the global is local to compile unit (static)
@@ -410,7 +429,7 @@
   metadata, ;; Name
   metadata, ;; Display name (fully qualified C++ name)
   metadata, ;; MIPS linkage name (for C++)
-  metadata, ;; Reference to compile unit where defined
+  metadata, ;; Reference to file where defined
   i32,      ;; Line number where defined
   metadata, ;; Reference to type descriptor
   i1,       ;; True if the global is local to compile unit (static)
@@ -461,7 +480,7 @@
             ;; (DW_TAG_base_type)
   metadata, ;; Reference to context (typically a compile unit)
   metadata, ;; Name (may be "" for anonymous types)
-  metadata, ;; Reference to compile unit where defined (may be NULL)
+  metadata, ;; Reference to file where defined (may be NULL)
   i32,      ;; Line number where defined (may be 0)
   i64,      ;; Size in bits
   i64,      ;; Alignment in bits
@@ -512,7 +531,7 @@
   i32,      ;; Tag (see below)
   metadata, ;; Reference to context
   metadata, ;; Name (may be "" for anonymous types)
-  metadata, ;; Reference to compile unit where defined (may be NULL)
+  metadata, ;; Reference to file where defined (may be NULL)
   i32,      ;; Line number where defined (may be 0)
   i32,      ;; Size in bits
   i32,      ;; Alignment in bits
@@ -580,7 +599,7 @@
   i32,      ;; Tag (see below)
   metadata, ;; Reference to context
   metadata, ;; Name (may be "" for anonymous types)
-  metadata, ;; Reference to compile unit where defined (may be NULL)
+  metadata, ;; Reference to file where defined (may be NULL)
   i32,      ;; Line number where defined (may be 0)
   i64,      ;; Size in bits
   i64,      ;; Alignment in bits
@@ -715,7 +734,7 @@
   i32,      ;; Tag (see below)
   metadata, ;; Context
   metadata, ;; Name
-  metadata, ;; Reference to compile unit where defined
+  metadata, ;; Reference to file where defined
   i32,      ;; Line number where defined
   metadata  ;; Type descriptor
 }
@@ -1012,10 +1031,10 @@
 <pre>
 ...
 ;;
-;; Define the compile unit for the source file "/Users/mine/sources/MySource.cpp".
+;; Define the compile unit for the main source file "/Users/mine/sources/MySource.cpp".
 ;;
-!3 = metadata !{
-  i32 458769,    ;; Tag
+!2 = metadata !{
+  i32 524305,    ;; Tag
   i32 0,         ;; Unused
   i32 4,         ;; Language Id
   metadata !"MySource.cpp", 
@@ -1027,19 +1046,24 @@
   i32 0}         ;; Runtime version
 
 ;;
-;; Define the compile unit for the header file "/Users/mine/sources/MyHeader.h".
+;; Define the file for the file "/Users/mine/sources/MySource.cpp".
 ;;
 !1 = metadata !{
-  i32 458769,    ;; Tag
-  i32 0,         ;; Unused
-  i32 4,         ;; Language Id
-  metadata !"MyHeader.h", 
+  i32 524329,    ;; Tag
+  metadata !"MySource.cpp", 
   metadata !"/Users/mine/sources", 
-  metadata !"4.2.1 (Based on Apple Inc. build 5649) (LLVM build 00)", 
-  i1 false,      ;; Main Compile Unit
-  i1 false,      ;; Optimized compile unit
-  metadata !"",  ;; Compiler flags
-  i32 0}         ;; Runtime version
+  metadata !3    ;; Compile unit
+}
+
+;;
+;; Define the file for the file "/Users/mine/sources/Myheader.h"
+;;
+!3 = metadata !{
+  i32 524329,    ;; Tag
+  metadata !"Myheader.h"
+  metadata !"/Users/mine/sources", 
+  metadata !3    ;; Compile unit
+}
 
 ...
 </pre>
@@ -1081,15 +1105,15 @@
 ;; variable anchor and the global variable itself.
 ;;
 !0 = metadata !{
-  i32 458804,              ;; Tag
+  i32 524340,              ;; Tag
   i32 0,                   ;; Unused
   metadata !1,             ;; Context
   metadata !"MyGlobal",    ;; Name
   metadata !"MyGlobal",    ;; Display Name
   metadata !"MyGlobal",    ;; Linkage Name
-  metadata !1,             ;; Compile Unit
+  metadata !3,             ;; Compile Unit
   i32 1,                   ;; Line Number
-  metadata !2,             ;; Type
+  metadata !4,             ;; Type
   i1 false,                ;; Is a local variable
   i1 true,                 ;; Is this a definition
   i32* @MyGlobal           ;; The global variable
@@ -1099,11 +1123,11 @@
 ;; Define the basic type of 32 bit signed integer.  Note that since int is an
 ;; intrinsic type the source file is NULL and line 0.
 ;;    
-!2 = metadata !{
-  i32 458788,              ;; Tag
+!4 = metadata !{
+  i32 524324,              ;; Tag
   metadata !1,             ;; Context
   metadata !"int",         ;; Name
-  metadata !1,             ;; Compile Unit
+  metadata !1,             ;; File
   i32 0,                   ;; Line number
   i64 32,                  ;; Size in Bits
   i64 32,                  ;; Align in Bits
@@ -1143,16 +1167,16 @@
 ;; anchor is 46, which is the same as the tag for subprograms
 ;; (46 = DW_TAG_subprogram.)
 ;;
-!0 = metadata !{
-  i32 458798,        ;; Tag
+!6 = metadata !{
+  i32 524334,        ;; Tag
   i32 0,             ;; Unused
   metadata !1,       ;; Context
   metadata !"main",  ;; Name
   metadata !"main",  ;; Display name
   metadata !"main",  ;; Linkage name
-  metadata !1,       ;; Compile unit
+  metadata !1,       ;; File
   i32 1,             ;; Line number
-  metadata !2,       ;; Type
+  metadata !4,       ;; Type
   i1 false,          ;; Is local 
   i1 true            ;; Is definition
 }
@@ -1188,10 +1212,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"bool",  ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 8,             ;; Size in Bits
   i64 8,             ;; Align in Bits
@@ -1214,10 +1238,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"char",  ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 8,             ;; Size in Bits
   i64 8,             ;; Align in Bits
@@ -1240,10 +1264,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"unsigned char", 
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 8,             ;; Size in Bits
   i64 8,             ;; Align in Bits
@@ -1266,10 +1290,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"short int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 16,            ;; Size in Bits
   i64 16,            ;; Align in Bits
@@ -1292,10 +1316,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"short unsigned int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 16,            ;; Size in Bits
   i64 16,            ;; Align in Bits
@@ -1318,10 +1342,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"int",   ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 32,            ;; Size in Bits
   i64 32,            ;; Align in Bits
@@ -1343,10 +1367,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"unsigned int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 32,            ;; Size in Bits
   i64 32,            ;; Align in Bits
@@ -1369,10 +1393,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"long long int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 64,            ;; Size in Bits
   i64 64,            ;; Align in Bits
@@ -1395,10 +1419,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"long long unsigned int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 64,            ;; Size in Bits
   i64 64,            ;; Align in Bits
@@ -1421,10 +1445,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"float",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 32,            ;; Size in Bits
   i64 32,            ;; Align in Bits
@@ -1447,10 +1471,10 @@
 <div class="doc_code">
 <pre>
 !2 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"double",;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 64,            ;; Size in Bits
   i64 64,            ;; Align in Bits
@@ -1486,10 +1510,10 @@
 ;; Define the typedef "IntPtr".
 ;;
 !2 = metadata !{
-  i32 458774,          ;; Tag
+  i32 524310,          ;; Tag
   metadata !1,         ;; Context
   metadata !"IntPtr",  ;; Name
-  metadata !3,         ;; Compile unit
+  metadata !3,         ;; File
   i32 0,               ;; Line number
   i64 0,               ;; Size in bits
   i64 0,               ;; Align in bits
@@ -1502,10 +1526,10 @@
 ;; Define the pointer type.
 ;;
 !4 = metadata !{
-  i32 458767,          ;; Tag
+  i32 524303,          ;; Tag
   metadata !1,         ;; Context
   metadata !"",        ;; Name
-  metadata !1,         ;; Compile unit
+  metadata !1,         ;; File
   i32 0,               ;; Line number
   i64 64,              ;; Size in bits
   i64 64,              ;; Align in bits
@@ -1517,10 +1541,10 @@
 ;; Define the const type.
 ;;
 !5 = metadata !{
-  i32 458790,          ;; Tag
+  i32 524326,          ;; Tag
   metadata !1,         ;; Context
   metadata !"",        ;; Name
-  metadata !1,         ;; Compile unit
+  metadata !1,         ;; File
   i32 0,               ;; Line number
   i64 32,              ;; Size in bits
   i64 32,              ;; Align in bits
@@ -1532,10 +1556,10 @@
 ;; Define the int type.
 ;;
 !6 = metadata !{
-  i32 458788,          ;; Tag
+  i32 524324,          ;; Tag
   metadata !1,         ;; Context
   metadata !"int",     ;; Name
-  metadata !1,         ;; Compile unit
+  metadata !1,         ;; File
   i32 0,               ;; Line number
   i64 32,              ;; Size in bits
   i64 32,              ;; Align in bits
@@ -1575,10 +1599,10 @@
 ;; Define basic type for unsigned int.
 ;;
 !5 = metadata !{
-  i32 458788,        ;; Tag
+  i32 524324,        ;; Tag
   metadata !1,       ;; Context
   metadata !"unsigned int",
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 0,             ;; Line number
   i64 32,            ;; Size in Bits
   i64 32,            ;; Align in Bits
@@ -1590,7 +1614,7 @@
 ;; Define composite type for struct Color.
 ;;
 !2 = metadata !{
-  i32 458771,        ;; Tag
+  i32 524307,        ;; Tag
   metadata !1,       ;; Context
   metadata !"Color", ;; Name
   metadata !1,       ;; Compile unit
@@ -1608,10 +1632,10 @@
 ;; Define the Red field.
 ;;
 !4 = metadata !{
-  i32 458765,        ;; Tag
+  i32 524301,        ;; Tag
   metadata !1,       ;; Context
   metadata !"Red",   ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 2,             ;; Line number
   i64 32,            ;; Size in bits
   i64 32,            ;; Align in bits
@@ -1624,10 +1648,10 @@
 ;; Define the Green field.
 ;;
 !6 = metadata !{
-  i32 458765,        ;; Tag
+  i32 524301,        ;; Tag
   metadata !1,       ;; Context
   metadata !"Green", ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 3,             ;; Line number
   i64 32,            ;; Size in bits
   i64 32,            ;; Align in bits
@@ -1640,10 +1664,10 @@
 ;; Define the Blue field.
 ;;
 !7 = metadata !{
-  i32 458765,        ;; Tag
+  i32 524301,        ;; Tag
   metadata !1,       ;; Context
   metadata !"Blue",  ;; Name
-  metadata !1,       ;; Compile Unit
+  metadata !1,       ;; File
   i32 4,             ;; Line number
   i64 32,            ;; Size in bits
   i64 32,            ;; Align in bits
@@ -1688,10 +1712,10 @@
 ;; Define composite type for enum Trees
 ;;
 !2 = metadata !{
-  i32 458756,        ;; Tag
+  i32 524292,        ;; Tag
   metadata !1,       ;; Context
   metadata !"Trees", ;; Name
-  metadata !1,       ;; Compile unit
+  metadata !1,       ;; File
   i32 1,             ;; Line number
   i64 32,            ;; Size in bits
   i64 32,            ;; Align in bits
@@ -1710,17 +1734,17 @@
 ;;
 ;; Define Spruce enumerator.
 ;;
-!4 = metadata !{i32 458792, metadata !"Spruce", i64 100}
+!4 = metadata !{i32 524328, metadata !"Spruce", i64 100}
 
 ;;
 ;; Define Oak enumerator.
 ;;
-!5 = metadata !{i32 458792, metadata !"Oak", i64 200}
+!5 = metadata !{i32 524328, metadata !"Oak", i64 200}
 
 ;;
 ;; Define Maple enumerator.
 ;;
-!6 = metadata !{i32 458792, metadata !"Maple", i64 300}
+!6 = metadata !{i32 524328, metadata !"Maple", i64 300}
 
 </pre>
 </div>

Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=98020&r1=98019&r2=98020&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar  8 18:44:10 2010
@@ -214,7 +214,13 @@
 
     DIScope getContext() const          { return getFieldAs<DIScope>(1); }
     StringRef getName() const           { return getStringField(2);     }
-    DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
+    DICompileUnit getCompileUnit() const{ 
+      if (getVersion() == llvm::LLVMDebugVersion7)
+        return getFieldAs<DICompileUnit>(3);
+
+      DIFile F = getFieldAs<DIFile>(3);
+      return F.getCompileUnit();
+    }
     unsigned getLineNumber() const      { return getUnsignedField(4); }
     uint64_t getSizeInBits() const      { return getUInt64Field(5); }
     uint64_t getAlignInBits() const     { return getUInt64Field(6); }
@@ -324,7 +330,14 @@
     StringRef getName() const         { return getStringField(3); }
     StringRef getDisplayName() const  { return getStringField(4); }
     StringRef getLinkageName() const  { return getStringField(5); }
-    DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(6); }
+    DICompileUnit getCompileUnit() const{ 
+      if (getVersion() == llvm::LLVMDebugVersion7)
+        return getFieldAs<DICompileUnit>(6);
+
+      DIFile F = getFieldAs<DIFile>(6); 
+      return F.getCompileUnit();
+    }
+
     unsigned getLineNumber() const      { return getUnsignedField(7); }
     DIType getType() const              { return getFieldAs<DIType>(8); }
 
@@ -346,7 +359,13 @@
     StringRef getName() const         { return getStringField(3); }
     StringRef getDisplayName() const  { return getStringField(4); }
     StringRef getLinkageName() const  { return getStringField(5); }
-    DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(6); }
+    DICompileUnit getCompileUnit() const{ 
+      if (getVersion() == llvm::LLVMDebugVersion7)
+        return getFieldAs<DICompileUnit>(6);
+
+      DIFile F = getFieldAs<DIFile>(6); 
+      return F.getCompileUnit();
+    }
     unsigned getLineNumber() const      { return getUnsignedField(7); }
     DICompositeType getType() const { return getFieldAs<DICompositeType>(8); }
 
@@ -413,7 +432,13 @@
 
     DIScope getContext() const          { return getFieldAs<DIScope>(1); }
     StringRef getName() const           { return getStringField(2);     }
-    DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
+    DICompileUnit getCompileUnit() const{ 
+      if (getVersion() == llvm::LLVMDebugVersion7)
+        return getFieldAs<DICompileUnit>(3);
+
+      DIFile F = getFieldAs<DIFile>(3); 
+      return F.getCompileUnit();
+    }
     unsigned getLineNumber() const      { return getUnsignedField(4); }
     DIType getType() const              { return getFieldAs<DIType>(5); }
 
@@ -461,7 +486,13 @@
     StringRef getName() const      { return getStringField(2);           }
     StringRef getDirectory() const { return getContext().getDirectory(); }
     StringRef getFilename() const  { return getContext().getFilename();  }
-    DICompileUnit getCompileUnit() const { return getFieldAs<DICompileUnit>(3);}
+    DICompileUnit getCompileUnit() const{ 
+      if (getVersion() == llvm::LLVMDebugVersion7)
+        return getFieldAs<DICompileUnit>(3);
+
+      DIFile F = getFieldAs<DIFile>(3); 
+      return F.getCompileUnit();
+    }
     unsigned getLineNumber() const { return getUnsignedField(4);         }
   };
 
@@ -523,14 +554,14 @@
 
     /// CreateBasicType - Create a basic type like int, float, etc.
     DIBasicType CreateBasicType(DIDescriptor Context, StringRef Name,
-                                DICompileUnit CompileUnit, unsigned LineNumber,
+                                DIFile F, unsigned LineNumber,
                                 uint64_t SizeInBits, uint64_t AlignInBits,
                                 uint64_t OffsetInBits, unsigned Flags,
                                 unsigned Encoding);
 
     /// CreateBasicType - Create a basic type like int, float, etc.
     DIBasicType CreateBasicTypeEx(DIDescriptor Context, StringRef Name,
-                                DICompileUnit CompileUnit, unsigned LineNumber,
+                                DIFile F, unsigned LineNumber,
                                 Constant *SizeInBits, Constant *AlignInBits,
                                 Constant *OffsetInBits, unsigned Flags,
                                 unsigned Encoding);
@@ -539,7 +570,7 @@
     /// pointer, typedef, etc.
     DIDerivedType CreateDerivedType(unsigned Tag, DIDescriptor Context,
                                     StringRef Name,
-                                    DICompileUnit CompileUnit,
+                                    DIFile F,
                                     unsigned LineNumber,
                                     uint64_t SizeInBits, uint64_t AlignInBits,
                                     uint64_t OffsetInBits, unsigned Flags,
@@ -548,17 +579,18 @@
     /// CreateDerivedType - Create a derived type like const qualified type,
     /// pointer, typedef, etc.
     DIDerivedType CreateDerivedTypeEx(unsigned Tag, DIDescriptor Context,
-                                        StringRef Name,
-                                    DICompileUnit CompileUnit,
-                                    unsigned LineNumber,
-                                    Constant *SizeInBits, Constant *AlignInBits,
-                                    Constant *OffsetInBits, unsigned Flags,
-                                    DIType DerivedFrom);
+                                      StringRef Name,
+                                      DIFile F,
+                                      unsigned LineNumber,
+                                      Constant *SizeInBits, 
+                                      Constant *AlignInBits,
+                                      Constant *OffsetInBits, unsigned Flags,
+                                      DIType DerivedFrom);
 
     /// CreateCompositeType - Create a composite type like array, struct, etc.
     DICompositeType CreateCompositeType(unsigned Tag, DIDescriptor Context,
                                         StringRef Name,
-                                        DICompileUnit CompileUnit,
+                                        DIFile F,
                                         unsigned LineNumber,
                                         uint64_t SizeInBits,
                                         uint64_t AlignInBits,
@@ -573,22 +605,23 @@
 
     /// CreateCompositeType - Create a composite type like array, struct, etc.
     DICompositeType CreateCompositeTypeEx(unsigned Tag, DIDescriptor Context,
-                                        StringRef Name,
-                                        DICompileUnit CompileUnit,
-                                        unsigned LineNumber,
-                                        Constant *SizeInBits,
-                                        Constant *AlignInBits,
-                                        Constant *OffsetInBits, unsigned Flags,
-                                        DIType DerivedFrom,
-                                        DIArray Elements,
-                                        unsigned RunTimeLang = 0);
+                                          StringRef Name,
+                                          DIFile F,
+                                          unsigned LineNumber,
+                                          Constant *SizeInBits,
+                                          Constant *AlignInBits,
+                                          Constant *OffsetInBits, 
+                                          unsigned Flags,
+                                          DIType DerivedFrom,
+                                          DIArray Elements,
+                                          unsigned RunTimeLang = 0);
 
     /// CreateSubprogram - Create a new descriptor for the specified subprogram.
     /// See comments in DISubprogram for descriptions of these fields.
     DISubprogram CreateSubprogram(DIDescriptor Context, StringRef Name,
                                   StringRef DisplayName,
                                   StringRef LinkageName,
-                                  DICompileUnit CompileUnit, unsigned LineNo,
+                                  DIFile F, unsigned LineNo,
                                   DIType Ty, bool isLocalToUnit,
                                   bool isDefinition,
                                   unsigned VK = 0,
@@ -605,21 +638,21 @@
     CreateGlobalVariable(DIDescriptor Context, StringRef Name,
                          StringRef DisplayName,
                          StringRef LinkageName,
-                         DICompileUnit CompileUnit,
+                         DIFile F,
                          unsigned LineNo, DIType Ty, bool isLocalToUnit,
                          bool isDefinition, llvm::GlobalVariable *GV);
 
     /// CreateVariable - Create a new descriptor for the specified variable.
     DIVariable CreateVariable(unsigned Tag, DIDescriptor Context,
                               StringRef Name,
-                              DICompileUnit CompileUnit, unsigned LineNo,
+                              DIFile F, unsigned LineNo,
                               DIType Ty);
 
     /// CreateComplexVariable - Create a new descriptor for the specified
     /// variable which has a complex address expression for its address.
     DIVariable CreateComplexVariable(unsigned Tag, DIDescriptor Context,
                                      const std::string &Name,
-                                     DICompileUnit CompileUnit, unsigned LineNo,
+                                     DIFile F, unsigned LineNo,
                                      DIType Ty,
                                      SmallVector<Value *, 9> &addr);
 
@@ -631,7 +664,7 @@
     /// CreateNameSpace - This creates new descriptor for a namespace
     /// with the specified parent context.
     DINameSpace CreateNameSpace(DIDescriptor Context, StringRef Name,
-                                DICompileUnit CU, unsigned LineNo);
+                                DIFile F, unsigned LineNo);
 
     /// CreateLocation - Creates a debug info location.
     DILocation CreateLocation(unsigned LineNo, unsigned ColumnNo,

Modified: llvm/trunk/include/llvm/Support/Dwarf.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/Dwarf.h?rev=98020&r1=98019&r2=98020&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/Dwarf.h (original)
+++ llvm/trunk/include/llvm/Support/Dwarf.h Mon Mar  8 18:44:10 2010
@@ -22,7 +22,8 @@
 // Debug info constants.
 
 enum {
-  LLVMDebugVersion = (7 << 16),         // Current version of debug information.
+  LLVMDebugVersion = (8 << 16),         // Current version of debug information.
+  LLVMDebugVersion7 = (7 << 16),        // Constant for version 7.
   LLVMDebugVersion6 = (6 << 16),        // Constant for version 6.
   LLVMDebugVersion5 = (5 << 16),        // Constant for version 5.
   LLVMDebugVersion4 = (4 << 16),        // Constant for version 4.

Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=98020&r1=98019&r2=98020&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar  8 18:44:10 2010
@@ -41,9 +41,9 @@
 
   DIDescriptor DI(N);
 
-  // Check current version. Allow Version6 for now.
+  // Check current version. Allow Version7 for now.
   unsigned Version = DI.getVersion();
-  if (Version != LLVMDebugVersion && Version != LLVMDebugVersion6)
+  if (Version != LLVMDebugVersion && Version != LLVMDebugVersion7)
     return false;
 
   switch (DI.getTag()) {
@@ -698,7 +698,7 @@
 /// CreateBasicType - Create a basic type like int, float, etc.
 DIBasicType DIFactory::CreateBasicType(DIDescriptor Context,
                                        StringRef Name,
-                                       DICompileUnit CompileUnit,
+                                       DIFile F,
                                        unsigned LineNumber,
                                        uint64_t SizeInBits,
                                        uint64_t AlignInBits,
@@ -708,7 +708,7 @@
     GetTagConstant(dwarf::DW_TAG_base_type),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     ConstantInt::get(Type::getInt64Ty(VMContext), SizeInBits),
     ConstantInt::get(Type::getInt64Ty(VMContext), AlignInBits),
@@ -723,7 +723,7 @@
 /// CreateBasicType - Create a basic type like int, float, etc.
 DIBasicType DIFactory::CreateBasicTypeEx(DIDescriptor Context,
                                          StringRef Name,
-                                         DICompileUnit CompileUnit,
+                                         DIFile F,
                                          unsigned LineNumber,
                                          Constant *SizeInBits,
                                          Constant *AlignInBits,
@@ -733,7 +733,7 @@
     GetTagConstant(dwarf::DW_TAG_base_type),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     SizeInBits,
     AlignInBits,
@@ -773,7 +773,7 @@
 DIDerivedType DIFactory::CreateDerivedType(unsigned Tag,
                                            DIDescriptor Context,
                                            StringRef Name,
-                                           DICompileUnit CompileUnit,
+                                           DIFile F,
                                            unsigned LineNumber,
                                            uint64_t SizeInBits,
                                            uint64_t AlignInBits,
@@ -784,7 +784,7 @@
     GetTagConstant(Tag),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     ConstantInt::get(Type::getInt64Ty(VMContext), SizeInBits),
     ConstantInt::get(Type::getInt64Ty(VMContext), AlignInBits),
@@ -801,7 +801,7 @@
 DIDerivedType DIFactory::CreateDerivedTypeEx(unsigned Tag,
                                              DIDescriptor Context,
                                              StringRef Name,
-                                             DICompileUnit CompileUnit,
+                                             DIFile F,
                                              unsigned LineNumber,
                                              Constant *SizeInBits,
                                              Constant *AlignInBits,
@@ -812,7 +812,7 @@
     GetTagConstant(Tag),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     SizeInBits,
     AlignInBits,
@@ -828,7 +828,7 @@
 DICompositeType DIFactory::CreateCompositeType(unsigned Tag,
                                                DIDescriptor Context,
                                                StringRef Name,
-                                               DICompileUnit CompileUnit,
+                                               DIFile F,
                                                unsigned LineNumber,
                                                uint64_t SizeInBits,
                                                uint64_t AlignInBits,
@@ -843,7 +843,7 @@
     GetTagConstant(Tag),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     ConstantInt::get(Type::getInt64Ty(VMContext), SizeInBits),
     ConstantInt::get(Type::getInt64Ty(VMContext), AlignInBits),
@@ -862,7 +862,7 @@
 DICompositeType DIFactory::CreateCompositeTypeEx(unsigned Tag,
                                                  DIDescriptor Context,
                                                  StringRef Name,
-                                                 DICompileUnit CompileUnit,
+                                                 DIFile F,
                                                  unsigned LineNumber,
                                                  Constant *SizeInBits,
                                                  Constant *AlignInBits,
@@ -876,7 +876,7 @@
     GetTagConstant(Tag),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNumber),
     SizeInBits,
     AlignInBits,
@@ -897,7 +897,7 @@
                                          StringRef Name,
                                          StringRef DisplayName,
                                          StringRef LinkageName,
-                                         DICompileUnit CompileUnit,
+                                         DIFile F,
                                          unsigned LineNo, DIType Ty,
                                          bool isLocalToUnit,
                                          bool isDefinition,
@@ -912,7 +912,7 @@
     MDString::get(VMContext, Name),
     MDString::get(VMContext, DisplayName),
     MDString::get(VMContext, LinkageName),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNo),
     Ty.getNode(),
     ConstantInt::get(Type::getInt1Ty(VMContext), isLocalToUnit),
@@ -957,7 +957,7 @@
 DIFactory::CreateGlobalVariable(DIDescriptor Context, StringRef Name,
                                 StringRef DisplayName,
                                 StringRef LinkageName,
-                                DICompileUnit CompileUnit,
+                                DIFile F,
                                 unsigned LineNo, DIType Ty,bool isLocalToUnit,
                                 bool isDefinition, llvm::GlobalVariable *Val) {
   Value *Elts[] = {
@@ -967,7 +967,7 @@
     MDString::get(VMContext, Name),
     MDString::get(VMContext, DisplayName),
     MDString::get(VMContext, LinkageName),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNo),
     Ty.getNode(),
     ConstantInt::get(Type::getInt1Ty(VMContext), isLocalToUnit),
@@ -989,13 +989,14 @@
 /// CreateVariable - Create a new descriptor for the specified variable.
 DIVariable DIFactory::CreateVariable(unsigned Tag, DIDescriptor Context,
                                      StringRef Name,
-                                     DICompileUnit CompileUnit, unsigned LineNo,
+                                     DIFile F,
+                                     unsigned LineNo,
                                      DIType Ty) {
   Value *Elts[] = {
     GetTagConstant(Tag),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNo),
     Ty.getNode(),
   };
@@ -1007,7 +1008,7 @@
 /// which has a complex address expression for its address.
 DIVariable DIFactory::CreateComplexVariable(unsigned Tag, DIDescriptor Context,
                                             const std::string &Name,
-                                            DICompileUnit CompileUnit,
+                                            DIFile F,
                                             unsigned LineNo,
                                             DIType Ty, 
                                             SmallVector<Value *, 9> &addr) {
@@ -1015,7 +1016,7 @@
   Elts.push_back(GetTagConstant(Tag));
   Elts.push_back(Context.getNode());
   Elts.push_back(MDString::get(VMContext, Name));
-  Elts.push_back(CompileUnit.getNode());
+  Elts.push_back(F.getNode());
   Elts.push_back(ConstantInt::get(Type::getInt32Ty(VMContext), LineNo));
   Elts.push_back(Ty.getNode());
   Elts.insert(Elts.end(), addr.begin(), addr.end());
@@ -1040,13 +1041,13 @@
 /// CreateNameSpace - This creates new descriptor for a namespace
 /// with the specified parent context.
 DINameSpace DIFactory::CreateNameSpace(DIDescriptor Context, StringRef Name,
-                                       DICompileUnit CompileUnit, 
+                                       DIFile F,
                                        unsigned LineNo) {
   Value *Elts[] = {
     GetTagConstant(dwarf::DW_TAG_namespace),
     Context.getNode(),
     MDString::get(VMContext, Name),
-    CompileUnit.getNode(),
+    F.getNode(),
     ConstantInt::get(Type::getInt32Ty(VMContext), LineNo)
   };
   return DINameSpace(MDNode::get(VMContext, &Elts[0], 5));

Modified: llvm/trunk/test/FrontendC++/2010-02-17-DbgArtificialArg.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC%2B%2B/2010-02-17-DbgArtificialArg.cpp?rev=98020&r1=98019&r2=98020&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC++/2010-02-17-DbgArtificialArg.cpp (original)
+++ llvm/trunk/test/FrontendC++/2010-02-17-DbgArtificialArg.cpp Mon Mar  8 18:44:10 2010
@@ -1,4 +1,4 @@
-// RUN: %llvmgcc -g -S %s -o - | grep DW_TAG_pointer_type |  grep "i32 458767, metadata .., metadata ..., metadata .., i32 ., i64 .., i64 .., i64 0, i32 64, metadata ..."
+// RUN: %llvmgcc -g -S %s -o - | grep DW_TAG_pointer_type |  grep "i32 524303, metadata .., metadata ..., metadata .., i32 ., i64 .., i64 .., i64 0, i32 64, metadata ..."
 // Here, second to last argument "i32 64" indicates that artificial type is set.                                               
 // Test to artificial attribute attahed to "this" pointer type.
 // Radar 7655792 and 7655002




From dpatel at apple.com  Mon Mar  8 18:45:32 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 09 Mar 2010 00:45:32 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r98022 - in /llvm-gcc-4.2/trunk/gcc:
 llvm-debug.cpp llvm-debug.h
Message-ID: <20100309004532.15AA72A6C12C@llvm.org>

Author: dpatel
Date: Mon Mar  8 18:45:31 2010
New Revision: 98022

URL: http://llvm.org/viewvc/llvm-project?rev=98022&view=rev
Log:
Start using DIFile. Corresponding llvm patch is r98020.

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
    llvm-gcc-4.2/trunk/gcc/llvm-debug.h

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=98022&r1=98021&r2=98022&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Mon Mar  8 18:45:31 2010
@@ -319,7 +319,7 @@
     ArtificialFnWithAbstractOrigin = true;
 
   DIDescriptor SPContext = ArtificialFnWithAbstractOrigin ?
-    getOrCreateCompileUnit(main_input_filename) :
+    getOrCreateFile(main_input_filename) :
     findRegion (DECL_CONTEXT(FnDecl));
 
   // Creating context may have triggered creation of this SP descriptor. So
@@ -361,7 +361,7 @@
     DebugFactory.CreateSubprogram(SPContext,
                                   FnName, FnName,
                                   LinkageName,
-                                  getOrCreateCompileUnit(Loc.file), lineno,
+                                  getOrCreateFile(Loc.file), lineno,
                                   FNType,
                                   Fn->hasInternalLinkage(),
                                   true /*definition*/,
@@ -385,7 +385,7 @@
   expanded_location Loc = GetNodeLocation(Node, false);
   DINameSpace DNS =
     DebugFactory.CreateNameSpace(Context, GetNodeName(Node),
-                                 getOrCreateCompileUnit(Loc.file), Loc.line);
+                                 getOrCreateFile(Loc.file), Loc.line);
 
   NameSpaceCache[Node] = WeakVH(DNS.getNode());
   return DNS;
@@ -394,7 +394,7 @@
 /// findRegion - Find tree_node N's region.
 DIDescriptor DebugInfo::findRegion(tree Node) {
   if (Node == NULL_TREE)
-    return getOrCreateCompileUnit(main_input_filename);
+    return getOrCreateFile(main_input_filename);
 
   std::map<tree_node *, WeakVH>::iterator I = RegionMap.find(Node);
   if (I != RegionMap.end())
@@ -422,7 +422,7 @@
   }
 
   // Otherwise main compile unit covers everything.
-  return getOrCreateCompileUnit(main_input_filename);
+  return getOrCreateFile(main_input_filename);
 }
 
 /// EmitFunctionEnd - Constructs the debug code for exiting a declarative
@@ -464,7 +464,7 @@
       Ty = DebugFactory.CreateArtificialType(Ty);
   llvm::DIVariable D =
     DebugFactory.CreateVariable(Tag, VarScope,
-                                Name, getOrCreateCompileUnit(Loc.file),
+                                Name, getOrCreateFile(Loc.file),
                                 Loc.line, Ty);
 
   // Insert an llvm.dbg.declare into the current block.
@@ -553,7 +553,7 @@
       LinkageName = GV->getName();
   DebugFactory.CreateGlobalVariable(findRegion(DECL_CONTEXT(decl)),
                                     DispName, DispName, LinkageName,
-                                    getOrCreateCompileUnit(Loc.file), Loc.line,
+                                    getOrCreateFile(Loc.file), Loc.line,
                                     TyD, GV->hasInternalLinkage(),
                                     true/*definition*/, GV);
 }
@@ -600,9 +600,9 @@
   }
 
   return 
-    DebugFactory.CreateBasicType(getOrCreateCompileUnit(main_input_filename),
+    DebugFactory.CreateBasicType(getOrCreateFile(main_input_filename),
                                  TypeName, 
-                                 getOrCreateCompileUnit(main_input_filename),
+                                 getOrCreateFile(main_input_filename),
                                  0, Size, Align,
                                  0, 0, Encoding);
 }
@@ -631,9 +631,9 @@
   sprintf(FwdTypeName, "fwd.type.%d", FwdTypeCount++);
   llvm::DIType FwdType = 
     DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_subroutine_type,
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      FwdTypeName,
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      0, 0, 0, 0, 0,
                                      llvm::DIType(), llvm::DIArray());
   llvm::TrackingVH<llvm::MDNode> FwdTypeNode = FwdType.getNode();
@@ -674,7 +674,7 @@
     DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_subroutine_type,
                                      findRegion(TYPE_CONTEXT(type)), 
                                      StringRef(),
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      0, 0, 0, 0, 0,
                                      llvm::DIType(), EltTypeArray);
 
@@ -706,7 +706,7 @@
       DIType Ty = 
         DebugFactory.CreateDerivedType(Tag, findRegion(DECL_CONTEXT(TyName)),
                                        GetNodeName(TyName), 
-                                       getOrCreateCompileUnit(TypeNameLoc.file),
+                                       getOrCreateFile(TypeNameLoc.file),
                                        TypeNameLoc.line,
                                        0 /*size*/,
                                        0 /*align*/,
@@ -722,7 +722,7 @@
     DebugFactory.CreateDerivedType(Tag, findRegion(TYPE_CONTEXT(type)), 
                                    Tag == DW_TAG_pointer_type ? 
                                    StringRef() : PName,
-                                   getOrCreateCompileUnit(NULL), 
+                                   getOrCreateFile(main_input_filename),
                                    0 /*line no*/, 
                                    NodeSizeInBits(type),
                                    NodeAlignInBits(type),
@@ -784,7 +784,7 @@
   return DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_array_type,
                                           findRegion(TYPE_CONTEXT(type)), 
                                           StringRef(),
-                                          getOrCreateCompileUnit(Loc.file), 0, 
+                                          getOrCreateFile(Loc.file), 0, 
                                           NodeSizeInBits(type), 
                                           NodeAlignInBits(type), 0, 0,
                                           getOrCreateType(EltTy),
@@ -816,7 +816,7 @@
   return DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_enumeration_type,
                                           findRegion(TYPE_CONTEXT(type)), 
                                           GetNodeName(type), 
-                                          getOrCreateCompileUnit(Loc.file), 
+                                          getOrCreateFile(Loc.file), 
                                           Loc.line,
                                           NodeSizeInBits(type), 
                                           NodeAlignInBits(type), 0, 0,
@@ -834,8 +834,7 @@
   if (TYPE_LANG_SPECIFIC (type)
       && lang_hooks.types.is_runtime_specific_type (type))
     {
-      DICompileUnit CU = getOrCreateCompileUnit(main_input_filename);
-      unsigned CULang = CU.getLanguage();
+      unsigned CULang = TheCU.getLanguage();
       switch (CULang) {
       case DW_LANG_ObjC_plus_plus :
         RunTimeLang = DW_LANG_ObjC_plus_plus;
@@ -890,7 +889,7 @@
     DebugFactory.CreateCompositeType(Tag, 
                                      TyContext,
                                      FwdName.c_str(),
-                                     getOrCreateCompileUnit(Loc.file), 
+                                     getOrCreateFile(Loc.file), 
                                      Loc.line, 
                                      0, 0, 0, SFlags | llvm::DIType::FlagFwdDecl,
                                      llvm::DIType(), llvm::DIArray(),
@@ -940,7 +939,7 @@
       DIType DTy = 
         DebugFactory.CreateDerivedType(DW_TAG_inheritance, 
                                        findRegion(TYPE_CONTEXT(type)), StringRef(),
-                                       llvm::DICompileUnit(), 0,0,0, 
+                                       llvm::DIFile(), 0,0,0, 
                                        Offset,
                                        BFlags, BaseClass);
       EltTys.push_back(DTy);
@@ -986,7 +985,7 @@
       DebugFactory.CreateDerivedType(DW_TAG_member, 
                                      findRegion(DECL_CONTEXT(Member)),
                                      MemberName, 
-                                     getOrCreateCompileUnit(MemLoc.file),
+                                     getOrCreateFile(MemLoc.file),
                                      MemLoc.line, NodeSizeInBits(Member),
                                      NodeAlignInBits(FieldNodeType),
                                      int_bit_position(Member), 
@@ -1025,7 +1024,7 @@
         DebugFactory.CreateSubprogram(findRegion(DECL_CONTEXT(Member)), 
                                       MemberName, MemberName,
                                       LinkageName, 
-                                      getOrCreateCompileUnit(MemLoc.file),
+                                      getOrCreateFile(MemLoc.file),
                                       MemLoc.line, SPTy, false, false,
                                       Virtuality, VIndex, ContainingType,
                                       DECL_ARTIFICIAL (Member));
@@ -1050,7 +1049,7 @@
   llvm::DICompositeType RealDecl =
     DebugFactory.CreateCompositeType(Tag, findRegion(TYPE_CONTEXT(type)),
                                      GetNodeName(type),
-                                     getOrCreateCompileUnit(Loc.file),
+                                     getOrCreateFile(Loc.file),
                                      Loc.line, 
                                      NodeSizeInBits(type), NodeAlignInBits(type),
                                      0, SFlags, llvm::DIType(), Elements,
@@ -1078,7 +1077,7 @@
       Ty = DebugFactory.CreateDerivedType(DW_TAG_typedef, 
                                           findRegion(DECL_CONTEXT(TyDef)),
                                           GetNodeName(TyDef), 
-                                          getOrCreateCompileUnit(TypeDefLoc.file),
+                                          getOrCreateFile(TypeDefLoc.file),
                                           TypeDefLoc.line,
                                           0 /*size*/,
                                           0 /*align*/,
@@ -1094,7 +1093,7 @@
     Ty = DebugFactory.CreateDerivedType(DW_TAG_volatile_type, 
                                         findRegion(TYPE_CONTEXT(type)), 
                                         StringRef(),
-                                        getOrCreateCompileUnit(NULL), 
+                                        getOrCreateFile(main_input_filename),
                                         0 /*line no*/, 
                                         NodeSizeInBits(type),
                                         NodeAlignInBits(type),
@@ -1108,7 +1107,7 @@
     Ty =  DebugFactory.CreateDerivedType(DW_TAG_const_type, 
                                          findRegion(TYPE_CONTEXT(type)), 
                                          StringRef(),
-                                         getOrCreateCompileUnit(NULL), 
+                                         getOrCreateFile(main_input_filename),
                                          0 /*line no*/, 
                                          NodeSizeInBits(type),
                                          NodeAlignInBits(type),
@@ -1227,9 +1226,9 @@
   // module does not contain any main compile unit then the code generator 
   // will emit multiple compile units in the output object file.
   if (!strcmp (main_input_filename, ""))
-    getOrCreateCompileUnit("<stdin>", true);
+    TheCU = getOrCreateCompileUnit("<stdin>", true);
   else
-    getOrCreateCompileUnit(main_input_filename, true);
+    TheCU = getOrCreateCompileUnit(main_input_filename, true);
 }
 
 /// getOrCreateCompileUnit - Get the compile unit from the cache or 
@@ -1242,10 +1241,6 @@
     else
       FullPath = main_input_filename;
   }
-  std::map<std::string, WeakVH >::iterator I = CUCache.find(FullPath);
-  if (I != CUCache.end())
-    if (Value *M = I->second)
-      return DICompileUnit(cast<MDNode>(M));
 
   // Get source file information.
   std::string Directory;
@@ -1286,13 +1281,27 @@
   unsigned ObjcRunTimeVer = 0;
   if (flag_objc_abi != 0 && flag_objc_abi != -1)
     ObjcRunTimeVer = flag_objc_abi;
-  DICompileUnit NewCU = DebugFactory.CreateCompileUnit(LangTag, FileName.c_str(),
-                                                       Directory.c_str(),
-                                                       version_string, isMain,
-                                                       optimize, Flags,
-                                                       ObjcRunTimeVer);
-  CUCache[FullPath] = WeakVH(NewCU.getNode());
-  return NewCU;
+  return DebugFactory.CreateCompileUnit(LangTag, FileName.c_str(),
+                                        Directory.c_str(),
+                                        version_string, isMain,
+                                        optimize, Flags,
+                                        ObjcRunTimeVer);
+}
+
+/// getOrCreateFile - Get DIFile descriptor.
+DIFile DebugInfo::getOrCreateFile(const char *FullPath) {
+  if (!FullPath) {
+    if (!strcmp (main_input_filename, ""))
+      FullPath = "<stdin>";
+    else
+      FullPath = main_input_filename;
+  }
+
+  // Get source file information.
+  std::string Directory;
+  std::string FileName;
+  DirectoryAndFile(FullPath, Directory, FileName);
+  return DebugFactory.CreateFile(FileName, Directory, TheCU);
 }
 
 /* LLVM LOCAL end (ENTIRE FILE!)  */

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.h
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.h?rev=98022&r1=98021&r2=98022&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.h (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.h Mon Mar  8 18:45:31 2010
@@ -62,7 +62,8 @@
   const char *PrevFullPath;             // Previous location file encountered.
   int PrevLineNo;                       // Previous location line# encountered.
   BasicBlock *PrevBB;                   // Last basic block encountered.
-  
+  DICompileUnit TheCU;                  // The compile unit.
+
   // Current GCC lexical block (or enclosing FUNCTION_DECL).
   tree_node *CurrentGCCLexicalBlock;	
   
@@ -71,7 +72,6 @@
   // debug info (through MDNodes) is not shared accidently.
   unsigned FwdTypeCount;
 
-  std::map<std::string, WeakVH > CUCache;
   std::map<tree_node *, WeakVH > TypeCache;
                                         // Cache of previously constructed 
                                         // Types.
@@ -168,6 +168,9 @@
   DICompileUnit getOrCreateCompileUnit(const char *FullPath,
                                        bool isMain = false);
 
+  /// getOrCreateFile - Get DIFile descriptor.
+  DIFile getOrCreateFile(const char *FullPath);
+
   /// findRegion - Find tree_node N's region.
   DIDescriptor findRegion(tree_node *n);
   




From wendling at apple.com  Mon Mar  8 18:46:30 2010
From: wendling at apple.com (Bill Wendling)
Date: Mon, 8 Mar 2010 16:46:30 -0800
Subject: [llvm-commits] [llvm] r98019 -
	/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
In-Reply-To: <20100309004334.EADFB2A6C12C@llvm.org>
References: <20100309004334.EADFB2A6C12C@llvm.org>
Message-ID: <A1761E01-44EE-4E4C-AB17-0CC468DBD2A0@apple.com>

Um...this also has a change of behavior (it snuck in there). If a symbol that's non-lazy is pointing to a symbol that's defined in the translation unit, then output a reference to it in the non-lazy pointer itself.

-bw

On Mar 8, 2010, at 4:43 PM, Bill Wendling wrote:

> Author: void
> Date: Mon Mar  8 18:43:34 2010
> New Revision: 98019
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98019&view=rev
> Log:
> Print blank line and clear stubs vector.
> 
> Modified:
>    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
> 
> Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98019&r1=98018&r2=98019&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Mon Mar  8 18:43:34 2010
> @@ -1132,8 +1132,16 @@
>         // L_foo$stub:
>         OutStreamer.EmitLabel(Stubs[i].first);
>         //   .indirect_symbol _foo
> -        OutStreamer.EmitSymbolAttribute(Stubs[i].second, MCSA_IndirectSymbol);
> -        OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
> +        MCSymbol *MCSym = Stubs[i].second;
> +        OutStreamer.EmitSymbolAttribute(MCSym, MCSA_IndirectSymbol);
> +
> +        if (MCSym->isUndefined())
> +          // External to current translation unit.
> +          OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
> +        else
> +          // Internal to current translation unit.
> +          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym, OutContext),
> +                                4/*size*/, 0/*addrspace*/);
>       }
> 
>       Stubs.clear();
> @@ -1152,6 +1160,9 @@
>                                                       OutContext),
>                               4/*size*/, 0/*addrspace*/);
>       }
> +
> +      Stubs.clear();
> +      OutStreamer.AddBlankLine();
>     }
> 
>     // Funny Darwin hack: This flag tells the linker that no global symbols
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From stoklund at 2pi.dk  Mon Mar  8 18:57:29 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 16:57:29 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <20100308232808.91F8F2A6C12C@llvm.org>
References: <20100308232808.91F8F2A6C12C@llvm.org>
Message-ID: <12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>


On Mar 8, 2010, at 3:28 PM, Evan Cheng wrote:

> Author: evancheng
> Date: Mon Mar  8 17:28:08 2010
> New Revision: 98007
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98007&view=rev
> Log:
> Restrict machine cse to really trivial coalescing. Leave the heavy lifting to a real coalescer.
> 
> Modified:
>    llvm/trunk/lib/CodeGen/MachineCSE.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98007&r1=98006&r2=98007&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
> +++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 17:28:08 2010
> @@ -91,7 +91,10 @@
>     unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
>     if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
>         TargetRegisterInfo::isVirtualRegister(SrcReg) &&
> +        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
>         !SrcSubIdx && !DstSubIdx) {
> +      DEBUG(dbgs() << "Coalescing: " << *DefMI);
> +      DEBUG(dbgs() << "*** to: " << *MI);
>       MO.setReg(SrcReg);
>       DefMI->eraseFromParent();
>       ++NumCoalesces;

You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.




From stoklund at 2pi.dk  Mon Mar  8 18:59:48 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 00:59:48 -0000
Subject: [llvm-commits] [llvm] r98023 -
	/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
Message-ID: <20100309005948.483A02A6C12C@llvm.org>

Author: stoklund
Date: Mon Mar  8 18:59:48 2010
New Revision: 98023

URL: http://llvm.org/viewvc/llvm-project?rev=98023&view=rev
Log:
Disable physical register coalescing when the number of live ranges for the
physreg becomes ridiculously high.

std::upper_bound may be log(N), but for sufficiently large live intervals, it
becomes log(N)*cachemiss = a long long time.

This patch improves coalescer time by 4500x for a function with 20000
function calls. The generated code is different, but not significantly worse -
the allocator hints are almost as good as physreg coalescing anyway.

Modified:
    llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp

Modified: llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp?rev=98023&r1=98022&r2=98023&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp (original)
+++ llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp Mon Mar  8 18:59:48 2010
@@ -1671,8 +1671,20 @@
         // density, do not join them, instead mark the physical register as its
         // allocation preference.
         LiveInterval &JoinVInt = SrcIsPhys ? DstInt : SrcInt;
+        LiveInterval &JoinPInt = SrcIsPhys ? SrcInt : DstInt;
         unsigned JoinVReg = SrcIsPhys ? DstReg : SrcReg;
         unsigned JoinPReg = SrcIsPhys ? SrcReg : DstReg;
+
+        // Don't join with physregs that have a ridiculous number of live
+        // ranges. The data structure performance is really bad when that
+        // happens.
+        if (JoinPInt.ranges.size() > 1000) {
+          mri_->setRegAllocationHint(JoinVInt.reg, 0, JoinPReg);
+          ++numAborts;
+          DEBUG(dbgs() << "\tPhysical register too complicated, abort!\n");
+          return false;
+        }
+
         const TargetRegisterClass *RC = mri_->getRegClass(JoinVReg);
         unsigned Threshold = allocatableRCRegs_[RC].count() * 2;
         unsigned Length = li_->getApproximateInstructionCount(JoinVInt);




From stoklund at 2pi.dk  Mon Mar  8 18:59:53 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 00:59:53 -0000
Subject: [llvm-commits] [llvm] r98024 -
	/llvm/trunk/lib/Transforms/IPO/Inliner.cpp
Message-ID: <20100309005953.5BD432A6C12D@llvm.org>

Author: stoklund
Date: Mon Mar  8 18:59:53 2010
New Revision: 98024

URL: http://llvm.org/viewvc/llvm-project?rev=98024&view=rev
Log:
Add inlining threshold to log output.

Modified:
    llvm/trunk/lib/Transforms/IPO/Inliner.cpp

Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=98024&r1=98023&r2=98024&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Mon Mar  8 18:59:53 2010
@@ -219,8 +219,10 @@
   Function *Caller = CS.getCaller();
   int CurrentThreshold = getInlineThreshold(CS);
   float FudgeFactor = getInlineFudgeFactor(CS);
-  if (Cost >= (int)(CurrentThreshold * FudgeFactor)) {
+  int AdjThreshold = (int)(CurrentThreshold * FudgeFactor);
+  if (Cost >= AdjThreshold) {
     DEBUG(dbgs() << "    NOT Inlining: cost=" << Cost
+          << ", thres=" << AdjThreshold
           << ", Call: " << *CS.getInstruction() << "\n");
     return false;
   }
@@ -285,6 +287,7 @@
   }
 
   DEBUG(dbgs() << "    Inlining: cost=" << Cost
+        << ", thres=" << AdjThreshold
         << ", Call: " << *CS.getInstruction() << '\n');
   return true;
 }




From sabre at nondot.org  Mon Mar  8 19:02:30 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 01:02:30 -0000
Subject: [llvm-commits] [llvm] r98025 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: AsmPrinter.cpp DwarfDebug.cpp
Message-ID: <20100309010230.90A922A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 19:02:30 2010
New Revision: 98025

URL: http://llvm.org/viewvc/llvm-project?rev=98025&view=rev
Log:
mcstreamerize AsmPrinter::printLabel.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98025&r1=98024&r2=98025&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 19:02:30 2010
@@ -1323,7 +1323,6 @@
                                       CurDLT.getColumnNumber(),
                                       CurDLT.getScope().getNode());
     printLabel(L);
-    O << '\n';
     DW->BeginScope(MI, L);
     PrevDLT = CurDLT.getNode();
   }
@@ -1554,12 +1553,17 @@
 /// printLabel - This method prints a local label used by debug and
 /// exception handling tables.
 void AsmPrinter::printLabelInst(const MachineInstr *MI) const {
-  printLabel(MI->getOperand(0).getImm());
-  OutStreamer.AddBlankLine();
+  MCSymbol *Sym = 
+    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) +
+                                 "label" + Twine(MI->getOperand(0).getImm()));
+  OutStreamer.EmitLabel(Sym);
 }
 
 void AsmPrinter::printLabel(unsigned Id) const {
-  O << MAI->getPrivateGlobalPrefix() << "label" << Id << ':';
+  MCSymbol *Sym = 
+    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) +
+                                 "label" + Twine(Id));
+  OutStreamer.EmitLabel(Sym);
 }
 
 /// PrintAsmOperand - Print the specified operand of MI, an INLINEASM

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98025&r1=98024&r2=98025&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 19:02:30 2010
@@ -1972,7 +1972,6 @@
 
   unsigned Label = MMI->NextLabelID();
   Asm->printLabel(Label);
-  O << '\n';
 
   SmallVector<DbgScope *, 2> &SD = I->second;
   for (SmallVector<DbgScope *, 2>::iterator SDI = SD.begin(), SDE = SD.end();
@@ -2131,7 +2130,6 @@
                                  DLT.getColumnNumber(), 
                                  DLT.getScope().getNode());
     Asm->printLabel(LabelID);
-    O << '\n';
   }
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();




From xuzhongxing at gmail.com  Mon Mar  8 19:05:35 2010
From: xuzhongxing at gmail.com (Zhongxing Xu)
Date: Tue, 9 Mar 2010 09:05:35 +0800
Subject: [llvm-commits] [PATCH] remove redundant 'static'
In-Reply-To: <F18CF852-A0B8-45FC-BD20-B4B312A1EC86@apple.com>
References: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>
	<F18CF852-A0B8-45FC-BD20-B4B312A1EC86@apple.com>
Message-ID: <5400aeb81003081705v1480ab04i1c8be003d2738b2e@mail.gmail.com>

2010/3/9 Chris Lattner <clattner at apple.com>:
>
> On Mar 8, 2010, at 4:45 AM, Zhongxing Xu wrote:
>
>> 'static' is redundant in this context:
>
> Um, no it's not. ?It prevents exporting the symbol from the .o file. ?What are you trying to fix?
>

'const' in namespace scope implies internal linkage. Is that
sufficient to prevent exporting the symbol from the .o file?

> -Chris
>
>>
>> Index: utils/TableGen/RegisterInfoEmitter.cpp
>> ===================================================================
>> --- utils/TableGen/RegisterInfoEmitter.cpp ? ?(?? 97892)
>> +++ utils/TableGen/RegisterInfoEmitter.cpp ? ?(????)
>> @@ -201,7 +201,7 @@
>>
>> ? ? // Emit the register list now.
>> ? ? OS << " ?// " << Name << " Register Class...\n"
>> - ? ? ? << " ?static const unsigned " << Name
>> + ? ? ? << " ?const unsigned " << Name
>> ? ? ? ?<< "[] = {\n ? ?";
>> ? ? for (unsigned i = 0, e = RC.Elements.size(); i != e; ++i) {
>> ? ? ? Record *Reg = RC.Elements[i];
>> @@ -223,7 +223,7 @@
>> ? ? // Emit the register list now.
>> ? ? OS << " ?// " << Name
>> ? ? ? ?<< " Register Class Value Types...\n"
>> - ? ? ? << " ?static const EVT " << Name
>> + ? ? ? << " ?const EVT " << Name
>> ? ? ? ?<< "[] = {\n ? ?";
>> ? ? for (unsigned i = 0, e = RC.VTs.size(); i != e; ++i)
>> ? ? ? OS << getEnumName(RC.VTs[i]) << ", ";
>> @@ -252,7 +252,7 @@
>>
>> ? ? ? OS << " ?// " << Name
>> ? ? ? ? ?<< " Sub-register Classes...\n"
>> - ? ? ? ? << " ?static const TargetRegisterClass* const "
>> + ? ? ? ? << " ?const TargetRegisterClass* const "
>> ? ? ? ? ?<< Name << "SubRegClasses[] = {\n ? ?";
>>
>> ? ? ? bool Empty = true;
>> @@ -298,7 +298,7 @@
>>
>> ? ? ? OS << " ?// " << Name
>> ? ? ? ? ?<< " Super-register Classes...\n"
>> - ? ? ? ? << " ?static const TargetRegisterClass* const "
>> + ? ? ? ? << " ?const TargetRegisterClass* const "
>> ? ? ? ? ?<< Name << "SuperRegClasses[] = {\n ? ?";
>>
>> ? ? ? bool Empty = true;
>> @@ -334,7 +334,7 @@
>>
>> ? ? ? OS << " ?// " << Name
>> ? ? ? ? ?<< " Register Class sub-classes...\n"
>> - ? ? ? ? << " ?static const TargetRegisterClass* const "
>> + ? ? ? ? << " ?const TargetRegisterClass* const "
>> ? ? ? ? ?<< Name << "Subclasses[] = {\n ? ?";
>>
>> ? ? ? bool Empty = true;
>> @@ -382,7 +382,7 @@
>>
>> ? ? ? OS << " ?// " << Name
>> ? ? ? ? ?<< " Register Class super-classes...\n"
>> - ? ? ? ? << " ?static const TargetRegisterClass* const "
>> + ? ? ? ? << " ?const TargetRegisterClass* const "
>> ? ? ? ? ?<< Name << "Superclasses[] = {\n ? ?";
>>
>> ? ? ? bool Empty = true;
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



From dalej at apple.com  Mon Mar  8 19:08:11 2010
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 09 Mar 2010 01:08:11 -0000
Subject: [llvm-commits] [llvm] r98026 -
	/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
Message-ID: <20100309010811.AE1252A6C12C@llvm.org>

Author: johannes
Date: Mon Mar  8 19:08:11 2010
New Revision: 98026

URL: http://llvm.org/viewvc/llvm-project?rev=98026&view=rev
Log:
Another place where debug info affected codegen.


Modified:
    llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp?rev=98026&r1=98025&r2=98026&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp Mon Mar  8 19:08:11 2010
@@ -1286,6 +1286,8 @@
       // there) so that it is guaranteed to dominate any user inside the loop.
       if (L && S->hasComputableLoopEvolution(L) && L != PostIncLoop)
         InsertPt = L->getHeader()->getFirstNonPHI();
+      while (isa<DbgInfoIntrinsic>(InsertPt))
+        InsertPt = llvm::next(BasicBlock::iterator(InsertPt));
       while (isInsertedInstruction(InsertPt))
         InsertPt = llvm::next(BasicBlock::iterator(InsertPt));
       break;




From daniel at zuster.org  Mon Mar  8 19:12:20 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 01:12:20 -0000
Subject: [llvm-commits] [llvm] r98027 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100309011221.024702A6C12C@llvm.org>

Author: ddunbar
Date: Mon Mar  8 19:12:20 2010
New Revision: 98027

URL: http://llvm.org/viewvc/llvm-project?rev=98027&view=rev
Log:
MC/Mach-O: Don't adjust section sizes when aligning zero fill sections, just pad the address.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98027&r1=98026&r2=98027&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 19:12:20 2010
@@ -1281,17 +1281,13 @@
 
     // Align this section if necessary by adding padding bytes to the previous
     // section.
-    if (uint64_t Pad = OffsetToAlignment(Address, it->getAlignment())) {
-      assert(Prev && "Missing prev section!");
-      Prev->setFileSize(Prev->getFileSize() + Pad);
+    if (uint64_t Pad = OffsetToAlignment(Address, it->getAlignment()))
       Address += Pad;
-    }
 
     SD.setAddress(Address);
     LayoutSection(SD);
     Address += SD.getSize();
 
-    Prev = &SD;
   }
 
   DEBUG_WITH_TYPE("mc-dump", {




From daniel at zuster.org  Mon Mar  8 19:12:23 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 01:12:23 -0000
Subject: [llvm-commits] [llvm] r98028 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100309011223.C0F302A6C12D@llvm.org>

Author: ddunbar
Date: Mon Mar  8 19:12:23 2010
New Revision: 98028

URL: http://llvm.org/viewvc/llvm-project?rev=98028&view=rev
Log:
MC/Mach-O: Tweak .dump() formatting.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98028&r1=98027&r2=98028&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Mon Mar  8 19:12:23 2010
@@ -1351,7 +1351,7 @@
     OS << ",\n       ";
     OS << " Fixups:[";
     for (fixup_iterator it = fixup_begin(), ie = fixup_end(); it != ie; ++it) {
-      if (it != fixup_begin()) OS << ",\n            ";
+      if (it != fixup_begin()) OS << ",\n                ";
       OS << *it;
     }
     OS << "]";
@@ -1394,7 +1394,7 @@
   OS << "<MCSectionData";
   OS << " Alignment:" << getAlignment() << " Address:" << Address
      << " Size:" << Size << " FileSize:" << FileSize
-     << " Fragments:[";
+     << " Fragments:[\n      ";
   for (iterator it = begin(), ie = end(); it != ie; ++it) {
     if (it != begin()) OS << ",\n      ";
     it->dump();
@@ -1422,7 +1422,7 @@
   raw_ostream &OS = llvm::errs();
 
   OS << "<MCAssembler\n";
-  OS << "  Sections:[";
+  OS << "  Sections:[\n    ";
   for (iterator it = begin(), ie = end(); it != ie; ++it) {
     if (it != begin()) OS << ",\n    ";
     it->dump();
@@ -1431,7 +1431,7 @@
   OS << "  Symbols:[";
 
   for (symbol_iterator it = symbol_begin(), ie = symbol_end(); it != ie; ++it) {
-    if (it != symbol_begin()) OS << ",\n    ";
+    if (it != symbol_begin()) OS << ",\n           ";
     it->dump();
   }
   OS << "]>\n";




From clattner at apple.com  Mon Mar  8 19:17:26 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 8 Mar 2010 17:17:26 -0800
Subject: [llvm-commits] [PATCH] remove redundant 'static'
In-Reply-To: <5400aeb81003081705v1480ab04i1c8be003d2738b2e@mail.gmail.com>
References: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>
	<F18CF852-A0B8-45FC-BD20-B4B312A1EC86@apple.com>
	<5400aeb81003081705v1480ab04i1c8be003d2738b2e@mail.gmail.com>
Message-ID: <D744B814-17A7-4262-9361-C99D7244AC18@apple.com>


On Mar 8, 2010, at 5:05 PM, Zhongxing Xu wrote:

> 2010/3/9 Chris Lattner <clattner at apple.com>:
>> 
>> On Mar 8, 2010, at 4:45 AM, Zhongxing Xu wrote:
>> 
>>> 'static' is redundant in this context:
>> 
>> Um, no it's not.  It prevents exporting the symbol from the .o file.  What are you trying to fix?
>> 
> 
> 'const' in namespace scope implies internal linkage. Is that
> sufficient to prevent exporting the symbol from the .o file?

It could be, but why do this?  What does it fix?  IMO, static is a lot less subtle.

-Chris


From evan.cheng at apple.com  Mon Mar  8 19:23:20 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 8 Mar 2010 17:23:20 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
Message-ID: <20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>


On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:

> 
> On Mar 8, 2010, at 3:28 PM, Evan Cheng wrote:
> 
>> Author: evancheng
>> Date: Mon Mar  8 17:28:08 2010
>> New Revision: 98007
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=98007&view=rev
>> Log:
>> Restrict machine cse to really trivial coalescing. Leave the heavy lifting to a real coalescer.
>> 
>> Modified:
>>   llvm/trunk/lib/CodeGen/MachineCSE.cpp
>> 
>> Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98007&r1=98006&r2=98007&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 17:28:08 2010
>> @@ -91,7 +91,10 @@
>>    unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
>>    if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
>>        TargetRegisterInfo::isVirtualRegister(SrcReg) &&
>> +        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
>>        !SrcSubIdx && !DstSubIdx) {
>> +      DEBUG(dbgs() << "Coalescing: " << *DefMI);
>> +      DEBUG(dbgs() << "*** to: " << *MI);
>>      MO.setReg(SrcReg);
>>      DefMI->eraseFromParent();
>>      ++NumCoalesces;
> 
> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
> 

I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.

Evan


From evan.cheng at apple.com  Mon Mar  8 19:23:49 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 8 Mar 2010 17:23:49 -0800
Subject: [llvm-commits] [llvm] r98023 -
	/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
In-Reply-To: <20100309005948.483A02A6C12C@llvm.org>
References: <20100309005948.483A02A6C12C@llvm.org>
Message-ID: <F579DF8B-D3A1-43B1-9E7B-A66EE315EA23@apple.com>

Is 1000 scientific? :-)

Evan

On Mar 8, 2010, at 4:59 PM, Jakob Stoklund Olesen wrote:

> Author: stoklund
> Date: Mon Mar  8 18:59:48 2010
> New Revision: 98023
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98023&view=rev
> Log:
> Disable physical register coalescing when the number of live ranges for the
> physreg becomes ridiculously high.
> 
> std::upper_bound may be log(N), but for sufficiently large live intervals, it
> becomes log(N)*cachemiss = a long long time.
> 
> This patch improves coalescer time by 4500x for a function with 20000
> function calls. The generated code is different, but not significantly worse -
> the allocator hints are almost as good as physreg coalescing anyway.
> 
> Modified:
>    llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp?rev=98023&r1=98022&r2=98023&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp Mon Mar  8 18:59:48 2010
> @@ -1671,8 +1671,20 @@
>         // density, do not join them, instead mark the physical register as its
>         // allocation preference.
>         LiveInterval &JoinVInt = SrcIsPhys ? DstInt : SrcInt;
> +        LiveInterval &JoinPInt = SrcIsPhys ? SrcInt : DstInt;
>         unsigned JoinVReg = SrcIsPhys ? DstReg : SrcReg;
>         unsigned JoinPReg = SrcIsPhys ? SrcReg : DstReg;
> +
> +        // Don't join with physregs that have a ridiculous number of live
> +        // ranges. The data structure performance is really bad when that
> +        // happens.
> +        if (JoinPInt.ranges.size() > 1000) {
> +          mri_->setRegAllocationHint(JoinVInt.reg, 0, JoinPReg);
> +          ++numAborts;
> +          DEBUG(dbgs() << "\tPhysical register too complicated, abort!\n");
> +          return false;
> +        }
> +
>         const TargetRegisterClass *RC = mri_->getRegClass(JoinVReg);
>         unsigned Threshold = allocatableRCRegs_[RC].count() * 2;
>         unsigned Length = li_->getApproximateInstructionCount(JoinVInt);
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From sabre at nondot.org  Mon Mar  8 19:29:59 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 01:29:59 -0000
Subject: [llvm-commits] [llvm] r98029 -
	/llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
Message-ID: <20100309012959.A1D2B2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 19:29:59 2010
New Revision: 98029

URL: http://llvm.org/viewvc/llvm-project?rev=98029&view=rev
Log:
inline RemapLabel into its only caller and simplify.

Modified:
    llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h

Modified: llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h?rev=98029&r1=98028&r2=98029&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h Mon Mar  8 19:29:59 2010
@@ -217,17 +217,9 @@
   /// MachineModuleInfo, for example because the code was deleted.
   void InvalidateLabel(unsigned LabelID) {
     // Remap to zero to indicate deletion.
-    RemapLabel(LabelID, 0);
-  }
-
-  /// RemapLabel - Indicate that a label has been merged into another.
-  ///
-  void RemapLabel(unsigned OldLabelID, unsigned NewLabelID) {
-    assert(0 < OldLabelID && OldLabelID <= LabelIDList.size() &&
-          "Old label ID out of range.");
-    assert(NewLabelID <= LabelIDList.size() &&
-          "New label ID out of range.");
-    LabelIDList[OldLabelID - 1] = NewLabelID;
+    assert(0 < LabelID && LabelID <= LabelIDList.size() &&
+           "Old label ID out of range.");
+    LabelIDList[LabelID - 1] = 0;
   }
   
   /// MappedLabel - Find out the label's final ID.  Zero indicates deletion.




From stoklund at 2pi.dk  Mon Mar  8 19:37:52 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 17:37:52 -0800
Subject: [llvm-commits] [llvm] r98023 -
	/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
In-Reply-To: <F579DF8B-D3A1-43B1-9E7B-A66EE315EA23@apple.com>
References: <20100309005948.483A02A6C12C@llvm.org>
	<F579DF8B-D3A1-43B1-9E7B-A66EE315EA23@apple.com>
Message-ID: <59424131-A98B-428A-9BE7-75786CE199C4@2pi.dk>


On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:

> Is 1000 scientific? :-)

No. That would require measuring when the LiveInterval size where it outgrows cache. I didn't do that :-(

I have also seen cases where a virtual register LiveInterval becomes too big. That is a lot harder to fix. I think the real problem is the extra indirection in SlotIndex. If SlotIndex were just an unsigned in disguise, each LiveRange would be just 16 bytes, and all ranges for a register would be in contiguous memory. 

> On Mar 8, 2010, at 4:59 PM, Jakob Stoklund Olesen wrote:
> 
>> Author: stoklund
>> Date: Mon Mar  8 18:59:48 2010
>> New Revision: 98023
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=98023&view=rev
>> Log:
>> Disable physical register coalescing when the number of live ranges for the
>> physreg becomes ridiculously high.
>> 
>> std::upper_bound may be log(N), but for sufficiently large live intervals, it
>> becomes log(N)*cachemiss = a long long time.
>> 
>> This patch improves coalescer time by 4500x for a function with 20000
>> function calls. The generated code is different, but not significantly worse -
>> the allocator hints are almost as good as physreg coalescing anyway.
>> 
>> Modified:
>>   llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
>> 
>> Modified: llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp?rev=98023&r1=98022&r2=98023&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp Mon Mar  8 18:59:48 2010
>> @@ -1671,8 +1671,20 @@
>>        // density, do not join them, instead mark the physical register as its
>>        // allocation preference.
>>        LiveInterval &JoinVInt = SrcIsPhys ? DstInt : SrcInt;
>> +        LiveInterval &JoinPInt = SrcIsPhys ? SrcInt : DstInt;
>>        unsigned JoinVReg = SrcIsPhys ? DstReg : SrcReg;
>>        unsigned JoinPReg = SrcIsPhys ? SrcReg : DstReg;
>> +
>> +        // Don't join with physregs that have a ridiculous number of live
>> +        // ranges. The data structure performance is really bad when that
>> +        // happens.
>> +        if (JoinPInt.ranges.size() > 1000) {
>> +          mri_->setRegAllocationHint(JoinVInt.reg, 0, JoinPReg);
>> +          ++numAborts;
>> +          DEBUG(dbgs() << "\tPhysical register too complicated, abort!\n");
>> +          return false;
>> +        }
>> +
>>        const TargetRegisterClass *RC = mri_->getRegClass(JoinVReg);
>>        unsigned Threshold = allocatableRCRegs_[RC].count() * 2;
>>        unsigned Length = li_->getApproximateInstructionCount(JoinVInt);
>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 




From sabre at nondot.org  Mon Mar  8 19:51:43 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 01:51:43 -0000
Subject: [llvm-commits] [llvm] r98031 - in /llvm/trunk:
 include/llvm/CodeGen/MachineModuleInfo.h
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp lib/CodeGen/MachineModuleInfo.cpp
 lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp
Message-ID: <20100309015143.92FA52A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 19:51:43 2010
New Revision: 98031

URL: http://llvm.org/viewvc/llvm-project?rev=98031&view=rev
Log:
strength reduce MMI::MappedLabel to MMI::isLabelDeleted,
and add a FIXME about how we are eventually going to zap this
lookup table once mc world domination is complete.

Modified:
    llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/MachineModuleInfo.cpp
    llvm/trunk/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp

Modified: llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h?rev=98031&r1=98030&r2=98031&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h Mon Mar  8 19:51:43 2010
@@ -222,13 +222,14 @@
     LabelIDList[LabelID - 1] = 0;
   }
   
-  /// MappedLabel - Find out the label's final ID.  Zero indicates deletion.
-  /// ID != Mapped ID indicates that the label was folded into another label.
-  unsigned MappedLabel(unsigned LabelID) const {
+  /// isLabelDeleted - Return true if the label was deleted.
+  /// FIXME: This should eventually be eliminated and use the 'is emitted' bit
+  /// on MCSymbol.
+  bool isLabelDeleted(unsigned LabelID) const {
     assert(LabelID <= LabelIDList.size() && "Debug label ID out of range.");
-    return LabelID ? LabelIDList[LabelID - 1] : 0;
+    return LabelID == 0 || LabelIDList[LabelID - 1] == 0;
   }
-
+  
   /// getFrameMoves - Returns a reference to a list of moves done in the current
   /// function's prologue.  Used to construct frame maps for debug and exception
   /// handling comsumers.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98031&r1=98030&r2=98031&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 19:51:43 2010
@@ -1362,9 +1362,14 @@
 /// constructLexicalScope - Construct new DW_TAG_lexical_block
 /// for this scope and attach DW_AT_low_pc/DW_AT_high_pc labels.
 DIE *DwarfDebug::constructLexicalScopeDIE(DbgScope *Scope) {
-  unsigned StartID = MMI->MappedLabel(Scope->getStartLabelID());
-  unsigned EndID = MMI->MappedLabel(Scope->getEndLabelID());
+  unsigned StartID = Scope->getStartLabelID();
+  unsigned EndID = Scope->getEndLabelID();
 
+  assert(!MMI->isLabelDeleted(StartID) &&
+         "Invalid starting label for an inlined scope!");
+  assert(!MMI->isLabelDeleted(EndID) &&
+         "Invalid end label for an inlined scope!");
+  
   // Ignore empty scopes.
   if (StartID == EndID && StartID != 0)
     return NULL;
@@ -1387,12 +1392,14 @@
 /// a function. Construct DIE to represent this concrete inlined copy
 /// of the function.
 DIE *DwarfDebug::constructInlinedScopeDIE(DbgScope *Scope) {
-  unsigned StartID = MMI->MappedLabel(Scope->getStartLabelID());
-  unsigned EndID = MMI->MappedLabel(Scope->getEndLabelID());
-  assert (StartID && "Invalid starting label for an inlined scope!");
-  assert (EndID && "Invalid end label for an inlined scope!");
+  unsigned StartID = Scope->getStartLabelID();
+  unsigned EndID = Scope->getEndLabelID();
+  assert(!MMI->isLabelDeleted(StartID) &&
+         "Invalid starting label for an inlined scope!");
+  assert(!MMI->isLabelDeleted(EndID) &&
+         "Invalid end label for an inlined scope!");
   // Ignore empty scopes.
-  if (StartID == EndID && StartID != 0)
+  if (StartID == EndID)
     return NULL;
   if (!Scope->getScopeNode())
     return NULL;
@@ -2589,8 +2596,8 @@
     // Construct rows of the address, source, line, column matrix.
     for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
       const SrcLineInfo &LineInfo = LineInfos[i];
-      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
-      if (!LabelID) continue;
+      unsigned LabelID = LineInfo.getLabelID();
+      if (MMI->isLabelDeleted(LabelID)) continue;
 
       if (LineInfo.getLine() == 0) continue;
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98031&r1=98030&r2=98031&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Mon Mar  8 19:51:43 2010
@@ -277,12 +277,9 @@
     const MachineMove &Move = Moves[i];
     unsigned LabelID = Move.getLabelID();
 
-    if (LabelID) {
-      LabelID = MMI->MappedLabel(LabelID);
-
-      // Throw out move if the label is invalid.
-      if (!LabelID) continue;
-    }
+    // Throw out move if the label is invalid.
+    if (LabelID && MMI->isLabelDeleted(LabelID))
+      continue;
 
     const MachineLocation &Dst = Move.getDestination();
     const MachineLocation &Src = Move.getSource();

Modified: llvm/trunk/lib/CodeGen/MachineModuleInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineModuleInfo.cpp?rev=98031&r1=98030&r2=98031&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineModuleInfo.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineModuleInfo.cpp Mon Mar  8 19:51:43 2010
@@ -185,7 +185,8 @@
 void MachineModuleInfo::TidyLandingPads() {
   for (unsigned i = 0; i != LandingPads.size(); ) {
     LandingPadInfo &LandingPad = LandingPads[i];
-    LandingPad.LandingPadLabel = MappedLabel(LandingPad.LandingPadLabel);
+    if (isLabelDeleted(LandingPad.LandingPadLabel))
+      LandingPad.LandingPadLabel = 0;
 
     // Special case: we *should* emit LPs with null LP MBB. This indicates
     // "nounwind" case.
@@ -195,17 +196,14 @@
     }
 
     for (unsigned j=0; j != LandingPads[i].BeginLabels.size(); ) {
-      unsigned BeginLabel = MappedLabel(LandingPad.BeginLabels[j]);
-      unsigned EndLabel = MappedLabel(LandingPad.EndLabels[j]);
-
-      if (!BeginLabel || !EndLabel) {
+      unsigned BeginLabel = LandingPad.BeginLabels[j];
+      unsigned EndLabel = LandingPad.EndLabels[j];
+      if (isLabelDeleted(BeginLabel) || isLabelDeleted(EndLabel)) {
         LandingPad.BeginLabels.erase(LandingPad.BeginLabels.begin() + j);
         LandingPad.EndLabels.erase(LandingPad.EndLabels.begin() + j);
         continue;
       }
 
-      LandingPad.BeginLabels[j] = BeginLabel;
-      LandingPad.EndLabels[j] = EndLabel;
       ++j;
     }
 

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp?rev=98031&r1=98030&r2=98031&view=diff
==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JITDwarfEmitter.cpp Mon Mar  8 19:51:43 2010
@@ -75,10 +75,9 @@
     unsigned LabelID = Move.getLabelID();
     
     if (LabelID) {
-      LabelID = MMI->MappedLabel(LabelID);
-    
       // Throw out move if the label is invalid.
-      if (!LabelID) continue;
+      if (MMI->isLabelDeleted(LabelID))
+        continue;
     }
     
     intptr_t LabelPtr = 0;
@@ -722,10 +721,9 @@
     unsigned LabelID = Move.getLabelID();
     
     if (LabelID) {
-      LabelID = MMI->MappedLabel(LabelID);
-    
       // Throw out move if the label is invalid.
-      if (!LabelID) continue;
+      if (MMI->isLabelDeleted(LabelID))
+        continue;
     }
     
     intptr_t LabelPtr = 0;




From sabre at nondot.org  Mon Mar  8 19:52:43 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 01:52:43 -0000
Subject: [llvm-commits] [llvm] r98032 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100309015243.9E41A2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 19:52:43 2010
New Revision: 98032

URL: http://llvm.org/viewvc/llvm-project?rev=98032&view=rev
Log:
remove a useless optimization: now that label replacement never
happens, the start/end of a scope can never be the same.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98032&r1=98031&r2=98032&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 19:52:43 2010
@@ -1370,10 +1370,6 @@
   assert(!MMI->isLabelDeleted(EndID) &&
          "Invalid end label for an inlined scope!");
   
-  // Ignore empty scopes.
-  if (StartID == EndID && StartID != 0)
-    return NULL;
-
   DIE *ScopeDIE = new DIE(dwarf::DW_TAG_lexical_block);
   if (Scope->isAbstractScope())
     return ScopeDIE;
@@ -1398,9 +1394,6 @@
          "Invalid starting label for an inlined scope!");
   assert(!MMI->isLabelDeleted(EndID) &&
          "Invalid end label for an inlined scope!");
-  // Ignore empty scopes.
-  if (StartID == EndID)
-    return NULL;
   if (!Scope->getScopeNode())
     return NULL;
   DIScope DS(Scope->getScopeNode());




From gohman at apple.com  Mon Mar  8 19:53:33 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 09 Mar 2010 01:53:33 -0000
Subject: [llvm-commits] [llvm] r98033 - in /llvm/trunk:
 lib/Analysis/LoopInfo.cpp test/Transforms/LCSSA/unreachable-use.ll
Message-ID: <20100309015333.B19B12A6C12C@llvm.org>

Author: djg
Date: Mon Mar  8 19:53:33 2010
New Revision: 98033

URL: http://llvm.org/viewvc/llvm-project?rev=98033&view=rev
Log:
Make isLCSSA ignore uses in blocks not reachable from the entry block,
as LCSSA no longer transforms such uses.

Added:
    llvm/trunk/test/Transforms/LCSSA/unreachable-use.ll
Modified:
    llvm/trunk/lib/Analysis/LoopInfo.cpp

Modified: llvm/trunk/lib/Analysis/LoopInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LoopInfo.cpp?rev=98033&r1=98032&r2=98033&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LoopInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LoopInfo.cpp Mon Mar  8 19:53:33 2010
@@ -264,6 +264,13 @@
 
 /// isLCSSAForm - Return true if the Loop is in LCSSA form
 bool Loop::isLCSSAForm() const {
+  // Collect all the reachable blocks in the function, for fast lookups.
+  SmallPtrSet<BasicBlock *, 32> ReachableBBs;
+  BasicBlock *EntryBB = getHeader()->getParent()->begin();
+  for (df_iterator<BasicBlock *> NI = df_begin(EntryBB),
+       NE = df_end(EntryBB); NI != NE; ++NI)
+    ReachableBBs.insert(*NI);
+
   // Sort the blocks vector so that we can use binary search to do quick
   // lookups.
   SmallPtrSet<BasicBlock *, 16> LoopBBs(block_begin(), block_end());
@@ -277,9 +284,13 @@
         if (PHINode *P = dyn_cast<PHINode>(*UI))
           UserBB = P->getIncomingBlock(UI);
 
-        // Check the current block, as a fast-path.  Most values are used in
-        // the same block they are defined in.
-        if (UserBB != BB && !LoopBBs.count(UserBB))
+        // Check the current block, as a fast-path, before checking whether
+        // the use is anywhere in the loop.  Most values are used in the same
+        // block they are defined in.  Also, blocks not reachable from the
+        // entry are special; uses in them don't need to go through PHIs.
+        if (UserBB != BB &&
+            !LoopBBs.count(UserBB) &&
+            ReachableBBs.count(UserBB))
           return false;
       }
   }

Added: llvm/trunk/test/Transforms/LCSSA/unreachable-use.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LCSSA/unreachable-use.ll?rev=98033&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LCSSA/unreachable-use.ll (added)
+++ llvm/trunk/test/Transforms/LCSSA/unreachable-use.ll Mon Mar  8 19:53:33 2010
@@ -0,0 +1,27 @@
+; RUN: opt < %s -lcssa -S -verify-loop-info | grep {\[%\]tmp33 = load i1\\*\\* \[%\]tmp}
+; PR6546
+
+; LCSSA doesn't need to transform uses in blocks not reachable
+; from the entry block.
+
+define fastcc void @dfs() nounwind {
+bb:
+  br label %bb44
+
+bb44:
+  br i1 undef, label %bb7, label %bb45
+
+bb7:
+  %tmp = bitcast i1** undef to i1**
+  br label %bb15
+
+bb15:
+  br label %bb44
+
+bb32:
+  %tmp33 = load i1** %tmp, align 8
+  br label %bb45
+
+bb45:
+  unreachable
+}




From sabre at nondot.org  Mon Mar  8 19:58:53 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 01:58:53 -0000
Subject: [llvm-commits] [llvm] r98034 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100309015853.E61222A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 19:58:53 2010
New Revision: 98034

URL: http://llvm.org/viewvc/llvm-project?rev=98034&view=rev
Log:
change DbgScope to keep track of the start/end label as MCSymbol* 
now that the dependence on ID is removed from MMI.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98034&r1=98033&r2=98034&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 19:58:53 2010
@@ -16,9 +16,10 @@
 #include "llvm/Module.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/MachineModuleInfo.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCStreamer.h"
-#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCSymbol.h"
 #include "llvm/Target/Mangler.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetFrameInfo.h"
@@ -171,8 +172,8 @@
   // Location at which this scope is inlined.
   AssertingVH<MDNode> InlinedAtLocation;  
   bool AbstractScope;                 // Abstract Scope
-  unsigned StartLabelID;              // Label ID of the beginning of scope.
-  unsigned EndLabelID;                // Label ID of the end of scope.
+  MCSymbol *StartLabel;               // Label ID of the beginning of scope.
+  MCSymbol *EndLabel;                 // Label ID of the end of scope.
   const MachineInstr *LastInsn;       // Last instruction of this scope.
   const MachineInstr *FirstInsn;      // First instruction of this scope.
   SmallVector<DbgScope *, 4> Scopes;  // Scopes defined in scope.
@@ -183,7 +184,7 @@
 public:
   DbgScope(DbgScope *P, DIDescriptor D, MDNode *I = 0)
     : Parent(P), Desc(D), InlinedAtLocation(I), AbstractScope(false),
-      StartLabelID(0), EndLabelID(0),
+      StartLabel(0), EndLabel(0),
       LastInsn(0), FirstInsn(0), IndentLevel(0) {}
   virtual ~DbgScope();
 
@@ -195,12 +196,12 @@
     return InlinedAtLocation;
   }
   MDNode *getScopeNode()         const { return Desc.getNode(); }
-  unsigned getStartLabelID()     const { return StartLabelID; }
-  unsigned getEndLabelID()       const { return EndLabelID; }
+  MCSymbol *getStartLabel()      const { return StartLabel; }
+  MCSymbol *getEndLabel()        const { return EndLabel; }
   SmallVector<DbgScope *, 4> &getScopes() { return Scopes; }
   SmallVector<DbgVariable *, 8> &getVariables() { return Variables; }
-  void setStartLabelID(unsigned S) { StartLabelID = S; }
-  void setEndLabelID(unsigned E)   { EndLabelID = E; }
+  void setStartLabel(MCSymbol *S) { StartLabel = S; }
+  void setEndLabel(MCSymbol *E)   { EndLabel = E; }
   void setLastInsn(const MachineInstr *MI) { LastInsn = MI; }
   const MachineInstr *getLastInsn()      { return LastInsn; }
   void setFirstInsn(const MachineInstr *MI) { FirstInsn = MI; }
@@ -264,7 +265,7 @@
   err.indent(IndentLevel);
   MDNode *N = Desc.getNode();
   N->dump();
-  err << " [" << StartLabelID << ", " << EndLabelID << "]\n";
+  err << " [" << StartLabel << ", " << EndLabel << "]\n";
   if (AbstractScope)
     err << "Abstract Scope\n";
 
@@ -1362,24 +1363,20 @@
 /// constructLexicalScope - Construct new DW_TAG_lexical_block
 /// for this scope and attach DW_AT_low_pc/DW_AT_high_pc labels.
 DIE *DwarfDebug::constructLexicalScopeDIE(DbgScope *Scope) {
-  unsigned StartID = Scope->getStartLabelID();
-  unsigned EndID = Scope->getEndLabelID();
+  MCSymbol *Start = Scope->getStartLabel();
+  MCSymbol *End = Scope->getEndLabel();
 
-  assert(!MMI->isLabelDeleted(StartID) &&
-         "Invalid starting label for an inlined scope!");
-  assert(!MMI->isLabelDeleted(EndID) &&
-         "Invalid end label for an inlined scope!");
+  assert(Start->isDefined() && "Invalid starting label for an inlined scope!");
+  assert(End->isDefined() && "Invalid end label for an inlined scope!");
   
   DIE *ScopeDIE = new DIE(dwarf::DW_TAG_lexical_block);
   if (Scope->isAbstractScope())
     return ScopeDIE;
 
   addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           StartID ? getDWLabel("label", StartID)
-                   : getDWLabel("func_begin", SubprogramCount));
+           Start ? Start : getDWLabel("func_begin", SubprogramCount));
   addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           EndID ? getDWLabel("label", EndID)
-                 : getDWLabel("func_end", SubprogramCount));
+           End ? End : getDWLabel("func_end", SubprogramCount));
 
   return ScopeDIE;
 }
@@ -1388,11 +1385,11 @@
 /// a function. Construct DIE to represent this concrete inlined copy
 /// of the function.
 DIE *DwarfDebug::constructInlinedScopeDIE(DbgScope *Scope) {
-  unsigned StartID = Scope->getStartLabelID();
-  unsigned EndID = Scope->getEndLabelID();
-  assert(!MMI->isLabelDeleted(StartID) &&
+  MCSymbol *StartLabel = Scope->getStartLabel();
+  MCSymbol *EndLabel = Scope->getEndLabel();
+  assert(StartLabel->isDefined() &&
          "Invalid starting label for an inlined scope!");
-  assert(!MMI->isLabelDeleted(EndID) &&
+  assert(EndLabel->isDefined() &&
          "Invalid end label for an inlined scope!");
   if (!Scope->getScopeNode())
     return NULL;
@@ -1405,11 +1402,8 @@
   addDIEEntry(ScopeDIE, dwarf::DW_AT_abstract_origin,
               dwarf::DW_FORM_ref4, OriginDIE);
 
-  MCSymbol *StartLabel = getDWLabel("label", StartID);
-  
   addLabel(ScopeDIE, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr, StartLabel);
-  addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           getDWLabel("label", EndID));
+  addLabel(ScopeDIE, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr, EndLabel);
 
   InlinedSubprogramDIEs.insert(OriginDIE);
 
@@ -1961,7 +1955,7 @@
   ScopeVector &SD = I->second;
   for (ScopeVector::iterator SDI = SD.begin(), SDE = SD.end();
        SDI != SDE; ++SDI)
-    (*SDI)->setStartLabelID(Label);
+    (*SDI)->setStartLabel(getDWLabel("label", Label));
 }
 
 /// endScope - Process end of a scope.
@@ -1970,13 +1964,13 @@
   if (I == DbgScopeEndMap.end())
     return;
 
-  unsigned Label = MMI->NextLabelID();
-  Asm->printLabel(Label);
+  MCSymbol *Label = getDWLabel("label", MMI->NextLabelID());
+  Asm->OutStreamer.EmitLabel(Label);
 
-  SmallVector<DbgScope *, 2> &SD = I->second;
+  SmallVector<DbgScope*, 2> &SD = I->second;
   for (SmallVector<DbgScope *, 2>::iterator SDI = SD.begin(), SDE = SD.end();
        SDI != SDE; ++SDI)
-    (*SDI)->setEndLabelID(Label);
+    (*SDI)->setEndLabel(Label);
   return;
 }
 
@@ -2120,16 +2114,17 @@
   DebugLoc FDL = MF->getDefaultDebugLoc();
   if (!FDL.isUnknown()) {
     DILocation DLT = MF->getDILocation(FDL);
-    unsigned LabelID = 0;
     DISubprogram SP = getDISubprogram(DLT.getScope().getNode());
-    if (SP.Verify())
-      LabelID = recordSourceLine(SP.getLineNumber(), 0, 
-                                 DLT.getScope().getNode());
-    else
-      LabelID = recordSourceLine(DLT.getLineNumber(), 
-                                 DLT.getColumnNumber(), 
-                                 DLT.getScope().getNode());
-    Asm->printLabel(LabelID);
+    unsigned Line, Col;
+    if (SP.Verify()) {
+      Line = SP.getLineNumber();
+      Col = 0;
+    } else {
+      Line = DLT.getLineNumber();
+      Col = DLT.getColumnNumber();
+    }
+    
+    Asm->printLabel(recordSourceLine(Line, Col, DLT.getScope().getNode()));
   }
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();




From sabre at nondot.org  Mon Mar  8 20:08:02 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 02:08:02 -0000
Subject: [llvm-commits] [llvm] r98035 - in /llvm/trunk:
 include/llvm/CodeGen/DwarfWriter.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
 lib/CodeGen/AsmPrinter/DwarfWriter.cpp
Message-ID: <20100309020802.A8D122A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 20:08:02 2010
New Revision: 98035

URL: http://llvm.org/viewvc/llvm-project?rev=98035&view=rev
Log:
Now that setStartLabel takes an MCSymbol, we can de-ID'ize
beginScope and RecordSourceLine.


Modified:
    llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp

Modified: llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DwarfWriter.h?rev=98035&r1=98034&r2=98035&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/DwarfWriter.h (original)
+++ llvm/trunk/include/llvm/CodeGen/DwarfWriter.h Mon Mar  8 20:08:02 2010
@@ -35,6 +35,7 @@
 class Module;
 class MDNode;
 class MCAsmInfo;
+class MCSymbol;
 class raw_ostream;
 class Instruction;
 class DICompileUnit;
@@ -82,10 +83,10 @@
   ///
   void EndFunction(const MachineFunction *MF);
 
-  /// RecordSourceLine - Register a source line with debug info. Returns a
-  /// unique label ID used to generate a label and provide correspondence to
+  /// RecordSourceLine - Register a source line with debug info. Returns the
+  /// unique label that was emitted and which provides correspondence to
   /// the source line list.
-  unsigned RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  MCSymbol *RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getRecordSourceLineCount - Count source lines.
   unsigned getRecordSourceLineCount();
@@ -94,7 +95,7 @@
   /// be emitted.
   bool ShouldEmitDwarfDebug() const;
 
-  void BeginScope(const MachineInstr *MI, unsigned Label);
+  void BeginScope(const MachineInstr *MI, MCSymbol *Label);
   void EndScope(const MachineInstr *MI);
 };
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98035&r1=98034&r2=98035&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 20:08:02 2010
@@ -1319,10 +1319,9 @@
     // After printing instruction
     DW->EndScope(MI);
   } else if (CurDLT.getNode() != PrevDLT) {
-    unsigned L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
-                                      CurDLT.getColumnNumber(),
-                                      CurDLT.getScope().getNode());
-    printLabel(L);
+    MCSymbol *L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
+                                       CurDLT.getColumnNumber(),
+                                       CurDLT.getScope().getNode());
     DW->BeginScope(MI, L);
     PrevDLT = CurDLT.getNode();
   }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98035&r1=98034&r2=98035&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 20:08:02 2010
@@ -1948,14 +1948,14 @@
 }
 
 /// beginScope - Process beginning of a scope starting at Label.
-void DwarfDebug::beginScope(const MachineInstr *MI, unsigned Label) {
+void DwarfDebug::beginScope(const MachineInstr *MI, MCSymbol *Label) {
   InsnToDbgScopeMapTy::iterator I = DbgScopeBeginMap.find(MI);
   if (I == DbgScopeBeginMap.end())
     return;
   ScopeVector &SD = I->second;
   for (ScopeVector::iterator SDI = SD.begin(), SDE = SD.end();
        SDI != SDE; ++SDI)
-    (*SDI)->setStartLabel(getDWLabel("label", Label));
+    (*SDI)->setStartLabel(Label);
 }
 
 /// endScope - Process end of a scope.
@@ -2124,7 +2124,7 @@
       Col = DLT.getColumnNumber();
     }
     
-    Asm->printLabel(recordSourceLine(Line, Col, DLT.getScope().getNode()));
+    recordSourceLine(Line, Col, DLT.getScope().getNode());
   }
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
@@ -2180,11 +2180,10 @@
     DebugTimer->stopTimer();
 }
 
-/// recordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::recordSourceLine(unsigned Line, unsigned Col,
-                                      MDNode *S) {
+/// recordSourceLine - Register a source line with debug info. Returns the
+/// unique label that was emitted and which provides correspondence to
+/// the source line list.
+MCSymbol *DwarfDebug::recordSourceLine(unsigned Line, unsigned Col, MDNode *S) {
   if (!MMI)
     return 0;
 
@@ -2217,7 +2216,9 @@
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
 
-  return ID;
+  MCSymbol *Label = getDWLabel("label", ID);
+  Asm->OutStreamer.EmitLabel(Label);
+  return Label;
 }
 
 /// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98035&r1=98034&r2=98035&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 20:08:02 2010
@@ -512,10 +512,10 @@
   ///
   void endFunction(const MachineFunction *MF);
 
-  /// recordSourceLine - Records location information and associates it with a 
-  /// label. Returns a unique label ID used to generate a label and provide
-  /// correspondence to the source line list.
-  unsigned recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  /// recordSourceLine - Register a source line with debug info. Returns the
+  /// unique label that was emitted and which provides correspondence to
+  /// the source line list.
+  MCSymbol *recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getSourceLineCount - Return the number of source lines in the debug
   /// info.
@@ -539,7 +539,7 @@
   void collectVariableInfo();
 
   /// beginScope - Process beginning of a scope starting at Label.
-  void beginScope(const MachineInstr *MI, unsigned Label);
+  void beginScope(const MachineInstr *MI, MCSymbol *Label);
 
   /// endScope - Prcess end of a scope.
   void endScope(const MachineInstr *MI);

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp?rev=98035&r1=98034&r2=98035&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp Mon Mar  8 20:08:02 2010
@@ -73,11 +73,11 @@
     MMI->EndFunction();
 }
 
-/// RecordSourceLine - Records location information and associates it with a 
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
-                                       MDNode *Scope) {
+/// RecordSourceLine - Register a source line with debug info. Returns the
+/// unique label that was emitted and which provides correspondence to
+/// the source line list.
+MCSymbol *DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
+                                        MDNode *Scope) {
   return DD->recordSourceLine(Line, Col, Scope);
 }
 
@@ -92,7 +92,7 @@
   return DD && DD->ShouldEmitDwarfDebug();
 }
 
-void DwarfWriter::BeginScope(const MachineInstr *MI, unsigned L) {
+void DwarfWriter::BeginScope(const MachineInstr *MI, MCSymbol *L) {
   DD->beginScope(MI, L);
 }
 void DwarfWriter::EndScope(const MachineInstr *MI) {




From gohman at apple.com  Mon Mar  8 20:15:05 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 09 Mar 2010 02:15:05 -0000
Subject: [llvm-commits] [llvm] r98036 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
Message-ID: <20100309021505.D57DC2A6C12C@llvm.org>

Author: djg
Date: Mon Mar  8 20:15:05 2010
New Revision: 98036

URL: http://llvm.org/viewvc/llvm-project?rev=98036&view=rev
Log:
Attempt to make this debug output meaningful, both in the case of
multibyte opcodes and in the case of multiple scopes.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=98036&r1=98035&r2=98036&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Mon Mar  8 20:15:05 2010
@@ -2141,6 +2141,9 @@
   
   while (1) {
     assert(MatcherIndex < TableSize && "Invalid index");
+#ifndef NDEBUG
+    unsigned CurrentOpcodeIndex = MatcherIndex;
+#endif
     BuiltinOpcodes Opcode = (BuiltinOpcodes)MatcherTable[MatcherIndex++];
     switch (Opcode) {
     case OPC_Scope: {
@@ -2666,6 +2669,7 @@
     // If the code reached this point, then the match failed.  See if there is
     // another child to try in the current 'Scope', otherwise pop it until we
     // find a case to check.
+    DEBUG(errs() << "  Match failed at index " << CurrentOpcodeIndex << "\n");
     while (1) {
       if (MatchScopes.empty()) {
         CannotYetSelect(NodeToMatch);
@@ -2680,13 +2684,12 @@
       NodeStack.append(LastScope.NodeStack.begin(), LastScope.NodeStack.end());
       N = NodeStack.back();
 
-      DEBUG(errs() << "  Match failed at index " << (MatcherIndex-1)
-                   << " continuing at " << LastScope.FailIndex << "\n");
-    
       if (LastScope.NumMatchedMemRefs != MatchedMemRefs.size())
         MatchedMemRefs.resize(LastScope.NumMatchedMemRefs);
       MatcherIndex = LastScope.FailIndex;
       
+      DEBUG(errs() << "  Continuing at " << MatcherIndex << "\n");
+    
       InputChain = LastScope.InputChain;
       InputFlag = LastScope.InputFlag;
       if (!LastScope.HasChainNodesMatched)




From echristo at apple.com  Mon Mar  8 20:36:31 2010
From: echristo at apple.com (Eric Christopher)
Date: Tue, 09 Mar 2010 02:36:31 -0000
Subject: [llvm-commits] [llvm] r98039 - in /llvm/trunk:
 include/llvm/CodeGen/DwarfWriter.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
 lib/CodeGen/AsmPrinter/DwarfWriter.cpp
Message-ID: <20100309023631.A699B2A6C12C@llvm.org>

Author: echristo
Date: Mon Mar  8 20:36:31 2010
New Revision: 98039

URL: http://llvm.org/viewvc/llvm-project?rev=98039&view=rev
Log:
Speculatively revert r98035. It appears to have caused a set of buildbot
failures.

Modified:
    llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp

Modified: llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DwarfWriter.h?rev=98039&r1=98038&r2=98039&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/DwarfWriter.h (original)
+++ llvm/trunk/include/llvm/CodeGen/DwarfWriter.h Mon Mar  8 20:36:31 2010
@@ -35,7 +35,6 @@
 class Module;
 class MDNode;
 class MCAsmInfo;
-class MCSymbol;
 class raw_ostream;
 class Instruction;
 class DICompileUnit;
@@ -83,10 +82,10 @@
   ///
   void EndFunction(const MachineFunction *MF);
 
-  /// RecordSourceLine - Register a source line with debug info. Returns the
-  /// unique label that was emitted and which provides correspondence to
+  /// RecordSourceLine - Register a source line with debug info. Returns a
+  /// unique label ID used to generate a label and provide correspondence to
   /// the source line list.
-  MCSymbol *RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  unsigned RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getRecordSourceLineCount - Count source lines.
   unsigned getRecordSourceLineCount();
@@ -95,7 +94,7 @@
   /// be emitted.
   bool ShouldEmitDwarfDebug() const;
 
-  void BeginScope(const MachineInstr *MI, MCSymbol *Label);
+  void BeginScope(const MachineInstr *MI, unsigned Label);
   void EndScope(const MachineInstr *MI);
 };
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98039&r1=98038&r2=98039&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 20:36:31 2010
@@ -1319,9 +1319,10 @@
     // After printing instruction
     DW->EndScope(MI);
   } else if (CurDLT.getNode() != PrevDLT) {
-    MCSymbol *L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
-                                       CurDLT.getColumnNumber(),
-                                       CurDLT.getScope().getNode());
+    unsigned L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
+                                      CurDLT.getColumnNumber(),
+                                      CurDLT.getScope().getNode());
+    printLabel(L);
     DW->BeginScope(MI, L);
     PrevDLT = CurDLT.getNode();
   }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98039&r1=98038&r2=98039&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 20:36:31 2010
@@ -1948,14 +1948,14 @@
 }
 
 /// beginScope - Process beginning of a scope starting at Label.
-void DwarfDebug::beginScope(const MachineInstr *MI, MCSymbol *Label) {
+void DwarfDebug::beginScope(const MachineInstr *MI, unsigned Label) {
   InsnToDbgScopeMapTy::iterator I = DbgScopeBeginMap.find(MI);
   if (I == DbgScopeBeginMap.end())
     return;
   ScopeVector &SD = I->second;
   for (ScopeVector::iterator SDI = SD.begin(), SDE = SD.end();
        SDI != SDE; ++SDI)
-    (*SDI)->setStartLabel(Label);
+    (*SDI)->setStartLabel(getDWLabel("label", Label));
 }
 
 /// endScope - Process end of a scope.
@@ -2124,7 +2124,7 @@
       Col = DLT.getColumnNumber();
     }
     
-    recordSourceLine(Line, Col, DLT.getScope().getNode());
+    Asm->printLabel(recordSourceLine(Line, Col, DLT.getScope().getNode()));
   }
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
@@ -2180,10 +2180,11 @@
     DebugTimer->stopTimer();
 }
 
-/// recordSourceLine - Register a source line with debug info. Returns the
-/// unique label that was emitted and which provides correspondence to
-/// the source line list.
-MCSymbol *DwarfDebug::recordSourceLine(unsigned Line, unsigned Col, MDNode *S) {
+/// recordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::recordSourceLine(unsigned Line, unsigned Col,
+                                      MDNode *S) {
   if (!MMI)
     return 0;
 
@@ -2216,9 +2217,7 @@
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
 
-  MCSymbol *Label = getDWLabel("label", ID);
-  Asm->OutStreamer.EmitLabel(Label);
-  return Label;
+  return ID;
 }
 
 /// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98039&r1=98038&r2=98039&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 20:36:31 2010
@@ -512,10 +512,10 @@
   ///
   void endFunction(const MachineFunction *MF);
 
-  /// recordSourceLine - Register a source line with debug info. Returns the
-  /// unique label that was emitted and which provides correspondence to
-  /// the source line list.
-  MCSymbol *recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  /// recordSourceLine - Records location information and associates it with a 
+  /// label. Returns a unique label ID used to generate a label and provide
+  /// correspondence to the source line list.
+  unsigned recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getSourceLineCount - Return the number of source lines in the debug
   /// info.
@@ -539,7 +539,7 @@
   void collectVariableInfo();
 
   /// beginScope - Process beginning of a scope starting at Label.
-  void beginScope(const MachineInstr *MI, MCSymbol *Label);
+  void beginScope(const MachineInstr *MI, unsigned Label);
 
   /// endScope - Prcess end of a scope.
   void endScope(const MachineInstr *MI);

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp?rev=98039&r1=98038&r2=98039&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp Mon Mar  8 20:36:31 2010
@@ -73,11 +73,11 @@
     MMI->EndFunction();
 }
 
-/// RecordSourceLine - Register a source line with debug info. Returns the
-/// unique label that was emitted and which provides correspondence to
-/// the source line list.
-MCSymbol *DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
-                                        MDNode *Scope) {
+/// RecordSourceLine - Records location information and associates it with a 
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
+                                       MDNode *Scope) {
   return DD->recordSourceLine(Line, Col, Scope);
 }
 
@@ -92,7 +92,7 @@
   return DD && DD->ShouldEmitDwarfDebug();
 }
 
-void DwarfWriter::BeginScope(const MachineInstr *MI, MCSymbol *L) {
+void DwarfWriter::BeginScope(const MachineInstr *MI, unsigned L) {
   DD->beginScope(MI, L);
 }
 void DwarfWriter::EndScope(const MachineInstr *MI) {




From isanbard at gmail.com  Mon Mar  8 20:46:12 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 02:46:12 -0000
Subject: [llvm-commits] [llvm] r98040 -
	/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
Message-ID: <20100309024612.ED6212A6C12C@llvm.org>

Author: void
Date: Mon Mar  8 20:46:12 2010
New Revision: 98040

URL: http://llvm.org/viewvc/llvm-project?rev=98040&view=rev
Log:
This is part of an LLC-beta test used to test <rdar://problem/6804645>. Please
bear with the awful code. It won't last in its current state beyond tonight.

Modified:
    llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=98040&r1=98039&r2=98040&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Mon Mar  8 20:46:12 2010
@@ -37,14 +37,20 @@
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/PseudoSourceValue.h"
 #include "llvm/CodeGen/SelectionDAG.h"
+#include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/ADT/VectorExtras.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Dwarf.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 #include <sstream>
 using namespace llvm;
+using namespace dwarf;
+
+static cl::opt<bool> EnableARMEHTest("enable-arm-eh-test", cl::Hidden,
+    cl::desc("Enable ARM EH Test"));
 
 static bool CC_ARM_APCS_Custom_f64(unsigned &ValNo, EVT &ValVT, EVT &LocVT,
                                    CCValAssign::LocInfo &LocInfo,
@@ -128,9 +134,33 @@
   addTypeForNEON(VT, MVT::v2f64, MVT::v4i32);
 }
 
+namespace llvm {
+
+  // FIXME: This is a test of <rdar://problem/6804645>.
+  class ARMMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
+  public:
+    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
+      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
+
+      // Exception Handling.
+      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
+                                    SectionKind::getReadOnlyWithRel());
+    }
+
+    virtual unsigned getTTypeEncoding() const {
+      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
+    }
+  };
+
+}
+
 static TargetLoweringObjectFile *createTLOF(TargetMachine &TM) {
   if (TM.getSubtarget<ARMSubtarget>().isTargetDarwin())
-    return new TargetLoweringObjectFileMachO();
+    if (EnableARMEHTest)
+      return new ARMMachOTargetObjectFile();
+    else
+      return new TargetLoweringObjectFileMachO();
+
   return new ARMElfTargetObjectFile();
 }
 




From isanbard at gmail.com  Mon Mar  8 20:46:47 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 02:46:47 -0000
Subject: [llvm-commits] [test-suite] r98041 -
	/test-suite/trunk/Makefile.programs
Message-ID: <20100309024647.9DD282A6C12C@llvm.org>

Author: void
Date: Mon Mar  8 20:46:47 2010
New Revision: 98041

URL: http://llvm.org/viewvc/llvm-project?rev=98041&view=rev
Log:
Enable LLC-beta test for ARM EH test.

Modified:
    test-suite/trunk/Makefile.programs

Modified: test-suite/trunk/Makefile.programs
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/Makefile.programs?rev=98041&r1=98040&r2=98041&view=diff
==============================================================================
--- test-suite/trunk/Makefile.programs (original)
+++ test-suite/trunk/Makefile.programs Mon Mar  8 20:46:47 2010
@@ -241,7 +241,8 @@
 LLCBETAOPTION := -enable-sparc-v9-insts
 endif
 ifeq ($(ARCH),ARM)
-LLCBETAOPTION := -enable-machine-cse
+LLCBETAOPTION := -enable-arm-eh-test
+# -enable-machine-cse
 endif
 ifeq ($(ARCH),THUMB)
 LLCBETAOPTION := -enable-machine-cse




From evan.cheng at apple.com  Mon Mar  8 20:53:20 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 8 Mar 2010 18:53:20 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
Message-ID: <DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>


On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:

> 
> On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:
> 
>> 
>> On Mar 8, 2010, at 3:28 PM, Evan Cheng wrote:
>> 
>>> Author: evancheng
>>> Date: Mon Mar  8 17:28:08 2010
>>> New Revision: 98007
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=98007&view=rev
>>> Log:
>>> Restrict machine cse to really trivial coalescing. Leave the heavy lifting to a real coalescer.
>>> 
>>> Modified:
>>>  llvm/trunk/lib/CodeGen/MachineCSE.cpp
>>> 
>>> Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98007&r1=98006&r2=98007&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
>>> +++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 17:28:08 2010
>>> @@ -91,7 +91,10 @@
>>>   unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
>>>   if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
>>>       TargetRegisterInfo::isVirtualRegister(SrcReg) &&
>>> +        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
>>>       !SrcSubIdx && !DstSubIdx) {
>>> +      DEBUG(dbgs() << "Coalescing: " << *DefMI);
>>> +      DEBUG(dbgs() << "*** to: " << *MI);
>>>     MO.setReg(SrcReg);
>>>     DefMI->eraseFromParent();
>>>     ++NumCoalesces;
>> 
>> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
>> 
> 
> I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.

Actually I do need to make it more complicated. But then BlackFin hates me (I wonder if it's related to its author).

Evan

> 
> Evan
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From xuzhongxing at gmail.com  Mon Mar  8 20:59:31 2010
From: xuzhongxing at gmail.com (Zhongxing Xu)
Date: Tue, 9 Mar 2010 10:59:31 +0800
Subject: [llvm-commits] [PATCH] remove redundant 'static'
In-Reply-To: <D744B814-17A7-4262-9361-C99D7244AC18@apple.com>
References: <5400aeb81003080445r3184c915j3c1471351f67d438@mail.gmail.com>
	<F18CF852-A0B8-45FC-BD20-B4B312A1EC86@apple.com>
	<5400aeb81003081705v1480ab04i1c8be003d2738b2e@mail.gmail.com>
	<D744B814-17A7-4262-9361-C99D7244AC18@apple.com>
Message-ID: <5400aeb81003081859m7ede3cbdr333ef6069ed533cc@mail.gmail.com>

2010/3/9 Chris Lattner <clattner at apple.com>:
>
> On Mar 8, 2010, at 5:05 PM, Zhongxing Xu wrote:
>
>> 2010/3/9 Chris Lattner <clattner at apple.com>:
>>>
>>> On Mar 8, 2010, at 4:45 AM, Zhongxing Xu wrote:
>>>
>>>> 'static' is redundant in this context:
>>>
>>> Um, no it's not. ?It prevents exporting the symbol from the .o file. ?What are you trying to fix?
>>>
>>
>> 'const' in namespace scope implies internal linkage. Is that
>> sufficient to prevent exporting the symbol from the .o file?
>
> It could be, but why do this? ?What does it fix? ?IMO, static is a lot less subtle.

I think it makes the code more concise. But never mind. It does not
fix anything real.



From gohman at apple.com  Mon Mar  8 21:01:40 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 09 Mar 2010 03:01:40 -0000
Subject: [llvm-commits] [llvm] r98042 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
Message-ID: <20100309030140.66F632A6C12C@llvm.org>

Author: djg
Date: Mon Mar  8 21:01:40 2010
New Revision: 98042

URL: http://llvm.org/viewvc/llvm-project?rev=98042&view=rev
Log:
Don't try to fold V_SET0 and V_SETALLONES to loads in medium and
large code models.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.cpp

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=98042&r1=98041&r2=98042&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Mon Mar  8 21:01:40 2010
@@ -2525,6 +2525,11 @@
     // Folding a V_SET0 or V_SETALLONES as a load, to ease register pressure.
     // Create a constant-pool entry and operands to load from it.
 
+    // Medium and large mode can't fold loads this way.
+    if (TM.getCodeModel() != CodeModel::Small &&
+        TM.getCodeModel() != CodeModel::Kernel)
+      return NULL;
+
     // x86-32 PIC requires a PIC base register for constant pools.
     unsigned PICBase = 0;
     if (TM.getRelocationModel() == Reloc::PIC_) {




From stoklund at 2pi.dk  Mon Mar  8 21:14:31 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 19:14:31 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
	<DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
Message-ID: <FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>


On Mar 8, 2010, at 6:53 PM, Evan Cheng wrote:

> 
> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
> 
>> 
>> On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:
>> 
>>> 
>>> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
>>> 
>> 
>> I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.
> 
> Actually I do need to make it more complicated. But then BlackFin hates me (I wonder if it's related to its author).

Hehe. Blackfin not hating you is a good sign that you got your register classes right ;-)

Basic register class algebra: If an instruction operand specifies register class X, a virtual register must belong to X or a subclass of X.

If you want to eliminate a copy "%x = mov %y", the resulting register must make the users of %x and %y happy at the same time. That means the new register must belong to the intersection of X and Y, the register classes of %x and %y.

The function getCommonSubClass(X, Y) finds the largest existing register class that is a subclass of both X and Y. If getCommonSubClass returns NULL, the classes are disjoint, and you cannot eliminate the copy - no register would be legal.

I think Blackfin is the only target with disjoint register classes - maybe that is why it hates you.





From evan.cheng at apple.com  Mon Mar  8 21:17:25 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 8 Mar 2010 19:17:25 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
	<DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
	<FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>
Message-ID: <B80A4BC9-3B21-45B4-8A49-4B8C93382697@apple.com>


On Mar 8, 2010, at 7:14 PM, Jakob Stoklund Olesen wrote:

> 
> On Mar 8, 2010, at 6:53 PM, Evan Cheng wrote:
> 
>> 
>> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
>> 
>>> 
>>> On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:
>>> 
>>>> 
>>>> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
>>>> 
>>> 
>>> I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.
>> 
>> Actually I do need to make it more complicated. But then BlackFin hates me (I wonder if it's related to its author).
> 
> Hehe. Blackfin not hating you is a good sign that you got your register classes right ;-)
> 
> Basic register class algebra: If an instruction operand specifies register class X, a virtual register must belong to X or a subclass of X.
> 
> If you want to eliminate a copy "%x = mov %y", the resulting register must make the users of %x and %y happy at the same time. That means the new register must belong to the intersection of X and Y, the register classes of %x and %y.
> 
> The function getCommonSubClass(X, Y) finds the largest existing register class that is a subclass of both X and Y. If getCommonSubClass returns NULL, the classes are disjoint, and you cannot eliminate the copy - no register would be legal.
> 
> I think Blackfin is the only target with disjoint register classes - maybe that is why it hates you.

I don't think that's the issue here. For example, 2009-08-15-LiveIn-SubReg.ll:
Coalescing: %reg1031<def> = MOVE %reg1028
*** to: %reg1030<def> = SETEQdd %reg1027, %reg1031

*** Bad machine code: Illegal virtual register for instruction ***
- function:    foo
- basic block: entry 0x2036ee4 (BB#0)
- instruction: %reg1030<def> = SETEQdd %reg1027, %reg1028
- operand 2:   %reg1028
Expected a D register, but got a GR register

I'm going to commit my changes to machine cse (which is not turned on yet). Can you take a look at it? Thanks.

Evan

> 
> 




From evan.cheng at apple.com  Mon Mar  8 21:21:12 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 09 Mar 2010 03:21:12 -0000
Subject: [llvm-commits] [llvm] r98043 - in /llvm/trunk/lib/CodeGen:
 LLVMTargetMachine.cpp MachineCSE.cpp
Message-ID: <20100309032112.894672A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 21:21:12 2010
New Revision: 98043

URL: http://llvm.org/viewvc/llvm-project?rev=98043&view=rev
Log:
- Make the machine cse dumb coalescer (as opposed to the more awesome simple
  coalescer) handle sub-register classes.
- Add heuristics to avoid non-profitable cse. Given the current lack of live
  range splitting, avoid cse when an expression has PHI use and the would be
  new use is in a BB where the expression wasn't already being used.

Modified:
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98043&r1=98042&r2=98043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Mon Mar  8 21:21:12 2010
@@ -328,11 +328,11 @@
     PM.add(createOptimizeExtsPass());
     if (!DisableMachineLICM)
       PM.add(createMachineLICMPass());
-    if (EnableMachineCSE)
+    //if (EnableMachineCSE)
       PM.add(createMachineCSEPass());
     if (!DisableMachineSink)
       PM.add(createMachineSinkingPass());
-    printAndVerify(PM, "After MachineLICM and MachineSinking",
+    printAndVerify(PM, "After Machine LICM, CSE and Sinking passes",
                    /* allowDoubleDefs= */ true);
   }
 

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98043&r1=98042&r2=98043&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 21:21:12 2010
@@ -33,9 +33,9 @@
   class MachineCSE : public MachineFunctionPass {
     const TargetInstrInfo *TII;
     const TargetRegisterInfo *TRI;
-    MachineRegisterInfo  *MRI;
-    MachineDominatorTree *DT;
     AliasAnalysis *AA;
+    MachineDominatorTree *DT;
+    MachineRegisterInfo *MRI;
   public:
     static char ID; // Pass identification
     MachineCSE() : MachineFunctionPass(&ID), CurrVN(0) {}
@@ -61,6 +61,7 @@
                                 MachineBasicBlock::const_iterator E);
     bool hasLivePhysRegDefUse(MachineInstr *MI, MachineBasicBlock *MBB);
     bool isCSECandidate(MachineInstr *MI);
+    bool isProfitableToCSE(unsigned Reg, MachineInstr *MI);
     bool ProcessBlock(MachineDomTreeNode *Node);
   };
 } // end anonymous namespace
@@ -91,14 +92,17 @@
     unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
     if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
         TargetRegisterInfo::isVirtualRegister(SrcReg) &&
-        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
         !SrcSubIdx && !DstSubIdx) {
-      DEBUG(dbgs() << "Coalescing: " << *DefMI);
-      DEBUG(dbgs() << "*** to: " << *MI);
-      MO.setReg(SrcReg);
-      DefMI->eraseFromParent();
-      ++NumCoalesces;
-      Changed = true;
+      const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
+      const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
+      if (SRC == RC || SRC->hasSubClass(RC) || RC->hasSubClass(SRC)) {
+        DEBUG(dbgs() << "Coalescing: " << *DefMI);
+        DEBUG(dbgs() << "*** to: " << *MI);
+        MO.setReg(SrcReg);
+        DefMI->eraseFromParent();
+        ++NumCoalesces;
+        Changed = true;
+      }
     }
   }
 
@@ -201,9 +205,31 @@
   return true;
 }
 
+/// isProfitableToCSE - Return true if it's profitable to eliminate MI with a
+/// common expression that defines Reg.
+bool MachineCSE::isProfitableToCSE(unsigned Reg, MachineInstr *MI) {
+  // FIXME: This "heuristic" works around the lack the live range splitting.
+  // If the common subexpression is used by PHIs, do not reuse it unless the
+  // defined value is already used in the BB of the new use.
+  bool HasPHI = false;
+  SmallPtrSet<MachineBasicBlock*, 4> CSBBs;
+  for (MachineRegisterInfo::use_nodbg_iterator I = 
+       MRI->use_nodbg_begin(Reg),
+       E = MRI->use_nodbg_end(); I != E; ++I) {
+    MachineInstr *Use = &*I;
+    HasPHI |= Use->isPHI();
+    CSBBs.insert(Use->getParent());
+  }
+
+  if (!HasPHI)
+    return true;
+  return CSBBs.count(MI->getParent());
+}
+
 bool MachineCSE::ProcessBlock(MachineDomTreeNode *Node) {
   bool Changed = false;
 
+  SmallVector<std::pair<unsigned, unsigned>, 8> CSEPairs;
   ScopedHashTableScope<MachineInstr*, unsigned,
     MachineInstrExpressionTrait> VNTS(VNT);
   MachineBasicBlock *MBB = Node->getBlock();
@@ -238,6 +264,9 @@
     MachineInstr *CSMI = Exps[CSVN];
     DEBUG(dbgs() << "Examining: " << *MI);
     DEBUG(dbgs() << "*** Found a common subexpression: " << *CSMI);
+
+    // Check if it's profitable to perform this CSE.
+    bool DoCSE = true;
     unsigned NumDefs = MI->getDesc().getNumDefs();
     for (unsigned i = 0, e = MI->getNumOperands(); NumDefs && i != e; ++i) {
       MachineOperand &MO = MI->getOperand(i);
@@ -250,11 +279,26 @@
       assert(TargetRegisterInfo::isVirtualRegister(OldReg) &&
              TargetRegisterInfo::isVirtualRegister(NewReg) &&
              "Do not CSE physical register defs!");
-      MRI->replaceRegWith(OldReg, NewReg);
+      if (!isProfitableToCSE(NewReg, MI)) {
+        DoCSE = false;
+        break;
+      }
+      CSEPairs.push_back(std::make_pair(OldReg, NewReg));
       --NumDefs;
     }
-    MI->eraseFromParent();
-    ++NumCSEs;
+
+    // Actually perform the elimination.
+    if (DoCSE) {
+      for (unsigned i = 0, e = CSEPairs.size(); i != e; ++i)
+        MRI->replaceRegWith(CSEPairs[i].first, CSEPairs[i].second);
+      MI->eraseFromParent();
+      ++NumCSEs;
+    } else {
+      DEBUG(dbgs() << "*** Not profitable, avoid CSE!\n");
+      VNT.insert(MI, CurrVN++);
+      Exps.push_back(MI);
+    }
+    CSEPairs.clear();
   }
 
   // Recursively call ProcessBlock with childred.
@@ -269,7 +313,7 @@
   TII = MF.getTarget().getInstrInfo();
   TRI = MF.getTarget().getRegisterInfo();
   MRI = &MF.getRegInfo();
-  DT = &getAnalysis<MachineDominatorTree>();
   AA = &getAnalysis<AliasAnalysis>();
+  DT = &getAnalysis<MachineDominatorTree>();
   return ProcessBlock(DT->getRootNode());
 }




From stoklund at 2pi.dk  Mon Mar  8 21:29:56 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 19:29:56 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <B80A4BC9-3B21-45B4-8A49-4B8C93382697@apple.com>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
	<DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
	<FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>
	<B80A4BC9-3B21-45B4-8A49-4B8C93382697@apple.com>
Message-ID: <C15ABE15-BA3E-4FBC-BBC0-8A7C0AC6CC5A@2pi.dk>


On Mar 8, 2010, at 7:17 PM, Evan Cheng wrote:

> 
> On Mar 8, 2010, at 7:14 PM, Jakob Stoklund Olesen wrote:
> 
>> 
>> On Mar 8, 2010, at 6:53 PM, Evan Cheng wrote:
>> 
>>> 
>>> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
>>> 
>>>> 
>>>> On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:
>>>> 
>>>>> 
>>>>> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
>>>>> 
>>>> 
>>>> I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.
>>> 
>>> Actually I do need to make it more complicated. But then BlackFin hates me (I wonder if it's related to its author).
>> 
>> Hehe. Blackfin not hating you is a good sign that you got your register classes right ;-)
>> 
>> Basic register class algebra: If an instruction operand specifies register class X, a virtual register must belong to X or a subclass of X.
>> 
>> If you want to eliminate a copy "%x = mov %y", the resulting register must make the users of %x and %y happy at the same time. That means the new register must belong to the intersection of X and Y, the register classes of %x and %y.
>> 
>> The function getCommonSubClass(X, Y) finds the largest existing register class that is a subclass of both X and Y. If getCommonSubClass returns NULL, the classes are disjoint, and you cannot eliminate the copy - no register would be legal.
>> 
>> I think Blackfin is the only target with disjoint register classes - maybe that is why it hates you.
> 
> I don't think that's the issue here. For example, 2009-08-15-LiveIn-SubReg.ll:
> Coalescing: %reg1031<def> = MOVE %reg1028
> *** to: %reg1030<def> = SETEQdd %reg1027, %reg1031
> 
> *** Bad machine code: Illegal virtual register for instruction ***
> - function:    foo
> - basic block: entry 0x2036ee4 (BB#0)
> - instruction: %reg1030<def> = SETEQdd %reg1027, %reg1028
> - operand 2:   %reg1028
> Expected a D register, but got a GR register

Sure. D is a subclass of GR.

%reg1028 is GR
%reg1031 is D

When you coalesce the resulting register class should be getCommonSubClass(GR, D) = D, but you are using %reg1028 which is GR - too big.

The copy was probably created by InstrEmitter::AddRegisterOperand to make the register classes match.

> I'm going to commit my changes to machine cse (which is not turned on yet). Can you take a look at it? Thanks.

Sure.





From stoklund at 2pi.dk  Mon Mar  8 21:41:25 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 19:41:25 -0800
Subject: [llvm-commits] [llvm] r98043 - in /llvm/trunk/lib/CodeGen:
	LLVMTargetMachine.cpp MachineCSE.cpp
In-Reply-To: <20100309032112.894672A6C12C@llvm.org>
References: <20100309032112.894672A6C12C@llvm.org>
Message-ID: <E20714B1-7B11-4FAA-AF3A-B984F6535E15@2pi.dk>


On Mar 8, 2010, at 7:21 PM, Evan Cheng wrote:

> @@ -91,14 +92,17 @@
>     unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
>     if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
>         TargetRegisterInfo::isVirtualRegister(SrcReg) &&
> -        MRI->getRegClass(SrcReg) == MRI->getRegClass(Reg) &&
>         !SrcSubIdx && !DstSubIdx) {
> -      DEBUG(dbgs() << "Coalescing: " << *DefMI);
> -      DEBUG(dbgs() << "*** to: " << *MI);
> -      MO.setReg(SrcReg);
> -      DefMI->eraseFromParent();
> -      ++NumCoalesces;
> -      Changed = true;
> +      const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
> +      const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
> +      if (SRC == RC || SRC->hasSubClass(RC) || RC->hasSubClass(SRC)) {
> +        DEBUG(dbgs() << "Coalescing: " << *DefMI);
> +        DEBUG(dbgs() << "*** to: " << *MI);
> +        MO.setReg(SrcReg);
> +        DefMI->eraseFromParent();
> +        ++NumCoalesces;
> +        Changed = true;
> +      }
>     }
>   }

The case "SRC->hasSubClass(RC)" is not valid. MO requires an RC operand, and SRC is a larger register class. It may contain registers that are illegal for MO.

If you don't want to change the register class of SrcReg, you have to make do with:

> +      if (SRC == RC || RC->hasSubClass(SRC)) {

If you are not afraid of switching register classes, you can go:

> +      const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
> +      const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
> +      const TargetRegisterClass *NewRC  = getCommonSubClass(SRC, RC);
> +      if (NewRC) {
> +        DEBUG(dbgs() << "Coalescing: " << *DefMI);
> +        DEBUG(dbgs() << "*** to: " << *MI);
> +        MO.setReg(SrcReg);
> +        if (SRC != NewRC)
> +          MRI->setRegClass(SrcReg, NewRC);

The problem with both approaches is that you are producing a longer live range of a smaller register class: Higher register pressure.

/jakob




From stoklund at 2pi.dk  Mon Mar  8 21:56:06 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 03:56:06 -0000
Subject: [llvm-commits] [llvm] r98044 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100309035606.8AB922A6C12C@llvm.org>

Author: stoklund
Date: Mon Mar  8 21:56:06 2010
New Revision: 98044

URL: http://llvm.org/viewvc/llvm-project?rev=98044&view=rev
Log:
Don't do illegal cross-class coalescing.

Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98044&r1=98043&r2=98044&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 21:56:06 2010
@@ -95,7 +95,7 @@
         !SrcSubIdx && !DstSubIdx) {
       const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
       const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
-      if (SRC == RC || SRC->hasSubClass(RC) || RC->hasSubClass(SRC)) {
+      if (SRC == RC || RC->hasSubClass(SRC)) {
         DEBUG(dbgs() << "Coalescing: " << *DefMI);
         DEBUG(dbgs() << "*** to: " << *MI);
         MO.setReg(SrcReg);




From evan.cheng at apple.com  Mon Mar  8 22:04:38 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 09 Mar 2010 04:04:38 -0000
Subject: [llvm-commits] [llvm] r98045 -
	/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
Message-ID: <20100309040438.D51592A6C12C@llvm.org>

Author: evancheng
Date: Mon Mar  8 22:04:38 2010
New Revision: 98045

URL: http://llvm.org/viewvc/llvm-project?rev=98045&view=rev
Log:
Revert accidental commit.

Modified:
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98045&r1=98044&r2=98045&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Mon Mar  8 22:04:38 2010
@@ -328,7 +328,7 @@
     PM.add(createOptimizeExtsPass());
     if (!DisableMachineLICM)
       PM.add(createMachineLICMPass());
-    //if (EnableMachineCSE)
+    if (EnableMachineCSE)
       PM.add(createMachineCSEPass());
     if (!DisableMachineSink)
       PM.add(createMachineSinkingPass());




From stoklund at 2pi.dk  Mon Mar  8 22:10:06 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 20:10:06 -0800
Subject: [llvm-commits] [llvm] r98044 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <20100309035606.8AB922A6C12C@llvm.org>
References: <20100309035606.8AB922A6C12C@llvm.org>
Message-ID: <2C469486-A4A0-41C8-AE44-B5DBCD776042@2pi.dk>


On Mar 8, 2010, at 7:56 PM, Jakob Stoklund Olesen wrote:

> Author: stoklund
> Date: Mon Mar  8 21:56:06 2010
> New Revision: 98044
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98044&view=rev
> Log:
> Don't do illegal cross-class coalescing.
> 
> Modified:
>    llvm/trunk/lib/CodeGen/MachineCSE.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98044&r1=98043&r2=98044&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
> +++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Mon Mar  8 21:56:06 2010
> @@ -95,7 +95,7 @@
>         !SrcSubIdx && !DstSubIdx) {
>       const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
>       const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
> -      if (SRC == RC || SRC->hasSubClass(RC) || RC->hasSubClass(SRC)) {
> +      if (SRC == RC || RC->hasSubClass(SRC)) {
>         DEBUG(dbgs() << "Coalescing: " << *DefMI);
>         DEBUG(dbgs() << "*** to: " << *MI);
>         MO.setReg(SrcReg);

That fixed the Blackfin failures. These remain:

LLVM :: CodeGen/ARM/indirectbr.ll
LLVM :: CodeGen/PowerPC/indirectbr.ll
LLVM :: CodeGen/Thumb/2009-08-20-ISelBug.ll
LLVM :: CodeGen/Thumb2/machine-licm.ll
LLVM :: CodeGen/X86/2007-10-16-IllegalAsm.ll
LLVM :: CodeGen/X86/coalesce-esp.ll
LLVM :: CodeGen/X86/pre-split6.ll
LLVM :: CodeGen/X86/stack-color-with-reg.ll





From evan.cheng at apple.com  Mon Mar  8 22:14:57 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 8 Mar 2010 20:14:57 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <C15ABE15-BA3E-4FBC-BBC0-8A7C0AC6CC5A@2pi.dk>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
	<DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
	<FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>
	<B80A4BC9-3B21-45B4-8A49-4B8C93382697@apple.com>
	<C15ABE15-BA3E-4FBC-BBC0-8A7C0AC6CC5A@2pi.dk>
Message-ID: <AA1C26BD-5241-4614-A16C-504589FA2A8B@apple.com>


On Mar 8, 2010, at 7:29 PM, Jakob Stoklund Olesen wrote:

> 
> On Mar 8, 2010, at 7:17 PM, Evan Cheng wrote:
> 
>> 
>> On Mar 8, 2010, at 7:14 PM, Jakob Stoklund Olesen wrote:
>> 
>>> 
>>> On Mar 8, 2010, at 6:53 PM, Evan Cheng wrote:
>>> 
>>>> 
>>>> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
>>>> 
>>>>> 
>>>>> On Mar 8, 2010, at 4:57 PM, Jakob Stoklund Olesen wrote:
>>>>> 
>>>>>> 
>>>>>> You could add "|| MRI->getRegClass(Reg)->hasSubClass(MRI->getRegClass(SrcReg))" without getting too complicated. That could help a bit with the few x86_64 subclasses as well.
>>>>>> 
>>>>> 
>>>>> I don't think it's necessary here. It's doing local coalescing here to avoid missing some obvious cse opportunities. If sub-reg class is involved, then it's already too complicated.
>>>> 
>>>> Actually I do need to make it more complicated. But then BlackFin hates me (I wonder if it's related to its author).
>>> 
>>> Hehe. Blackfin not hating you is a good sign that you got your register classes right ;-)
>>> 
>>> Basic register class algebra: If an instruction operand specifies register class X, a virtual register must belong to X or a subclass of X.
>>> 
>>> If you want to eliminate a copy "%x = mov %y", the resulting register must make the users of %x and %y happy at the same time. That means the new register must belong to the intersection of X and Y, the register classes of %x and %y.
>>> 
>>> The function getCommonSubClass(X, Y) finds the largest existing register class that is a subclass of both X and Y. If getCommonSubClass returns NULL, the classes are disjoint, and you cannot eliminate the copy - no register would be legal.
>>> 
>>> I think Blackfin is the only target with disjoint register classes - maybe that is why it hates you.
>> 
>> I don't think that's the issue here. For example, 2009-08-15-LiveIn-SubReg.ll:
>> Coalescing: %reg1031<def> = MOVE %reg1028
>> *** to: %reg1030<def> = SETEQdd %reg1027, %reg1031
>> 
>> *** Bad machine code: Illegal virtual register for instruction ***
>> - function:    foo
>> - basic block: entry 0x2036ee4 (BB#0)
>> - instruction: %reg1030<def> = SETEQdd %reg1027, %reg1028
>> - operand 2:   %reg1028
>> Expected a D register, but got a GR register
> 
> Sure. D is a subclass of GR.
> 
> %reg1028 is GR
> %reg1031 is D
> 
> When you coalesce the resulting register class should be getCommonSubClass(GR, D) = D, but you are using %reg1028 which is GR - too big.
> 
> The copy was probably created by InstrEmitter::AddRegisterOperand to make the register classes match.
> 
>> I'm going to commit my changes to machine cse (which is not turned on yet). Can you take a look at it? Thanks.
> 
> Sure.

Ok, I think the issue is more subtle. Your patch disabled some critical coalescing that's required. It probably should be allowed but then changes the register class.

Evan

> 
> 




From stoklund at 2pi.dk  Mon Mar  8 22:21:05 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 8 Mar 2010 20:21:05 -0800
Subject: [llvm-commits] [llvm] r98007 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
In-Reply-To: <AA1C26BD-5241-4614-A16C-504589FA2A8B@apple.com>
References: <20100308232808.91F8F2A6C12C@llvm.org>
	<12195ADE-5620-4756-8425-E09F73A9646D@2pi.dk>
	<20FB2BBD-9306-4DD8-961E-A627EE6A2FD2@apple.com>
	<DD1D3DE0-2CE7-4703-AED0-E1639C76721F@apple.com>
	<FC9CAC8F-AB01-4382-8E1E-9D2E20D9D26C@2pi.dk>
	<B80A4BC9-3B21-45B4-8A49-4B8C93382697@apple.com>
	<C15ABE15-BA3E-4FBC-BBC0-8A7C0AC6CC5A@2pi.dk>
	<AA1C26BD-5241-4614-A16C-504589FA2A8B@apple.com>
Message-ID: <4AE8D63B-1C47-44CF-849D-FE7AB42B8390@2pi.dk>


On Mar 8, 2010, at 8:14 PM, Evan Cheng wrote:
> 
> Ok, I think the issue is more subtle. Your patch disabled some critical coalescing that's required. It probably should be allowed but then changes the register class.

Switching the register class to a subclass should not give any problems besides increased register pressure.





From clattner at apple.com  Mon Mar  8 22:30:25 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 8 Mar 2010 20:30:25 -0800
Subject: [llvm-commits] [llvm] r98036 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
In-Reply-To: <20100309021505.D57DC2A6C12C@llvm.org>
References: <20100309021505.D57DC2A6C12C@llvm.org>
Message-ID: <AC7BB171-0C02-4C59-922C-66C3E352BE58@apple.com>

Looks great, thanks Dan.

-Chris

On Mar 8, 2010, at 6:15 PM, Dan Gohman wrote:

> Author: djg
> Date: Mon Mar  8 20:15:05 2010
> New Revision: 98036
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98036&view=rev
> Log:
> Attempt to make this debug output meaningful, both in the case of
> multibyte opcodes and in the case of multiple scopes.
> 
> Modified:
>    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=98036&r1=98035&r2=98036&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Mon Mar  8 20:15:05 2010
> @@ -2141,6 +2141,9 @@
> 
>   while (1) {
>     assert(MatcherIndex < TableSize && "Invalid index");
> +#ifndef NDEBUG
> +    unsigned CurrentOpcodeIndex = MatcherIndex;
> +#endif
>     BuiltinOpcodes Opcode = (BuiltinOpcodes)MatcherTable[MatcherIndex++];
>     switch (Opcode) {
>     case OPC_Scope: {
> @@ -2666,6 +2669,7 @@
>     // If the code reached this point, then the match failed.  See if there is
>     // another child to try in the current 'Scope', otherwise pop it until we
>     // find a case to check.
> +    DEBUG(errs() << "  Match failed at index " << CurrentOpcodeIndex << "\n");
>     while (1) {
>       if (MatchScopes.empty()) {
>         CannotYetSelect(NodeToMatch);
> @@ -2680,13 +2684,12 @@
>       NodeStack.append(LastScope.NodeStack.begin(), LastScope.NodeStack.end());
>       N = NodeStack.back();
> 
> -      DEBUG(errs() << "  Match failed at index " << (MatcherIndex-1)
> -                   << " continuing at " << LastScope.FailIndex << "\n");
> -    
>       if (LastScope.NumMatchedMemRefs != MatchedMemRefs.size())
>         MatchedMemRefs.resize(LastScope.NumMatchedMemRefs);
>       MatcherIndex = LastScope.FailIndex;
> 
> +      DEBUG(errs() << "  Continuing at " << MatcherIndex << "\n");
> +    
>       InputChain = LastScope.InputChain;
>       InputFlag = LastScope.InputFlag;
>       if (!LastScope.HasChainNodesMatched)
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From sabre at nondot.org  Mon Mar  8 22:48:35 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 04:48:35 -0000
Subject: [llvm-commits] [llvm] r98046 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100309044835.E4F172A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 22:48:35 2010
New Revision: 98046

URL: http://llvm.org/viewvc/llvm-project?rev=98046&view=rev
Log:
add some extra checks.  I'm not sure why, but this does unbreak a
failure remaining on mainline.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98046&r1=98045&r2=98046&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 22:48:35 2010
@@ -1365,6 +1365,7 @@
 DIE *DwarfDebug::constructLexicalScopeDIE(DbgScope *Scope) {
   MCSymbol *Start = Scope->getStartLabel();
   MCSymbol *End = Scope->getEndLabel();
+  if (Start == 0) return 0;
 
   assert(Start->isDefined() && "Invalid starting label for an inlined scope!");
   assert(End->isDefined() && "Invalid end label for an inlined scope!");
@@ -1387,6 +1388,8 @@
 DIE *DwarfDebug::constructInlinedScopeDIE(DbgScope *Scope) {
   MCSymbol *StartLabel = Scope->getStartLabel();
   MCSymbol *EndLabel = Scope->getEndLabel();
+  if (StartLabel == 0) return 0;
+  
   assert(StartLabel->isDefined() &&
          "Invalid starting label for an inlined scope!");
   assert(EndLabel->isDefined() &&




From sabre at nondot.org  Mon Mar  8 22:54:43 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 04:54:43 -0000
Subject: [llvm-commits] [llvm] r98047 - in /llvm/trunk:
 include/llvm/CodeGen/DwarfWriter.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
 lib/CodeGen/AsmPrinter/DwarfWriter.cpp
Message-ID: <20100309045444.1175B2A6C12C@llvm.org>

Author: lattner
Date: Mon Mar  8 22:54:43 2010
New Revision: 98047

URL: http://llvm.org/viewvc/llvm-project?rev=98047&view=rev
Log:
reapply r98035:
Now that setStartLabel takes an MCSymbol, we can de-ID'ize
beginScope and RecordSourceLine.

Modified:
    llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp

Modified: llvm/trunk/include/llvm/CodeGen/DwarfWriter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DwarfWriter.h?rev=98047&r1=98046&r2=98047&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/DwarfWriter.h (original)
+++ llvm/trunk/include/llvm/CodeGen/DwarfWriter.h Mon Mar  8 22:54:43 2010
@@ -35,6 +35,7 @@
 class Module;
 class MDNode;
 class MCAsmInfo;
+class MCSymbol;
 class raw_ostream;
 class Instruction;
 class DICompileUnit;
@@ -82,10 +83,10 @@
   ///
   void EndFunction(const MachineFunction *MF);
 
-  /// RecordSourceLine - Register a source line with debug info. Returns a
-  /// unique label ID used to generate a label and provide correspondence to
+  /// RecordSourceLine - Register a source line with debug info. Returns the
+  /// unique label that was emitted and which provides correspondence to
   /// the source line list.
-  unsigned RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  MCSymbol *RecordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getRecordSourceLineCount - Count source lines.
   unsigned getRecordSourceLineCount();
@@ -94,7 +95,7 @@
   /// be emitted.
   bool ShouldEmitDwarfDebug() const;
 
-  void BeginScope(const MachineInstr *MI, unsigned Label);
+  void BeginScope(const MachineInstr *MI, MCSymbol *Label);
   void EndScope(const MachineInstr *MI);
 };
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98047&r1=98046&r2=98047&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Mon Mar  8 22:54:43 2010
@@ -1319,10 +1319,9 @@
     // After printing instruction
     DW->EndScope(MI);
   } else if (CurDLT.getNode() != PrevDLT) {
-    unsigned L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
-                                      CurDLT.getColumnNumber(),
-                                      CurDLT.getScope().getNode());
-    printLabel(L);
+    MCSymbol *L = DW->RecordSourceLine(CurDLT.getLineNumber(), 
+                                       CurDLT.getColumnNumber(),
+                                       CurDLT.getScope().getNode());
     DW->BeginScope(MI, L);
     PrevDLT = CurDLT.getNode();
   }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98047&r1=98046&r2=98047&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar  8 22:54:43 2010
@@ -1951,14 +1951,14 @@
 }
 
 /// beginScope - Process beginning of a scope starting at Label.
-void DwarfDebug::beginScope(const MachineInstr *MI, unsigned Label) {
+void DwarfDebug::beginScope(const MachineInstr *MI, MCSymbol *Label) {
   InsnToDbgScopeMapTy::iterator I = DbgScopeBeginMap.find(MI);
   if (I == DbgScopeBeginMap.end())
     return;
   ScopeVector &SD = I->second;
   for (ScopeVector::iterator SDI = SD.begin(), SDE = SD.end();
        SDI != SDE; ++SDI)
-    (*SDI)->setStartLabel(getDWLabel("label", Label));
+    (*SDI)->setStartLabel(Label);
 }
 
 /// endScope - Process end of a scope.
@@ -2127,7 +2127,7 @@
       Col = DLT.getColumnNumber();
     }
     
-    Asm->printLabel(recordSourceLine(Line, Col, DLT.getScope().getNode()));
+    recordSourceLine(Line, Col, DLT.getScope().getNode());
   }
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
@@ -2183,11 +2183,10 @@
     DebugTimer->stopTimer();
 }
 
-/// recordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::recordSourceLine(unsigned Line, unsigned Col,
-                                      MDNode *S) {
+/// recordSourceLine - Register a source line with debug info. Returns the
+/// unique label that was emitted and which provides correspondence to
+/// the source line list.
+MCSymbol *DwarfDebug::recordSourceLine(unsigned Line, unsigned Col, MDNode *S) {
   if (!MMI)
     return 0;
 
@@ -2220,7 +2219,9 @@
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
 
-  return ID;
+  MCSymbol *Label = getDWLabel("label", ID);
+  Asm->OutStreamer.EmitLabel(Label);
+  return Label;
 }
 
 /// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98047&r1=98046&r2=98047&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar  8 22:54:43 2010
@@ -512,10 +512,10 @@
   ///
   void endFunction(const MachineFunction *MF);
 
-  /// recordSourceLine - Records location information and associates it with a 
-  /// label. Returns a unique label ID used to generate a label and provide
-  /// correspondence to the source line list.
-  unsigned recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
+  /// recordSourceLine - Register a source line with debug info. Returns the
+  /// unique label that was emitted and which provides correspondence to
+  /// the source line list.
+  MCSymbol *recordSourceLine(unsigned Line, unsigned Col, MDNode *Scope);
 
   /// getSourceLineCount - Return the number of source lines in the debug
   /// info.
@@ -539,7 +539,7 @@
   void collectVariableInfo();
 
   /// beginScope - Process beginning of a scope starting at Label.
-  void beginScope(const MachineInstr *MI, unsigned Label);
+  void beginScope(const MachineInstr *MI, MCSymbol *Label);
 
   /// endScope - Prcess end of a scope.
   void endScope(const MachineInstr *MI);

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp?rev=98047&r1=98046&r2=98047&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp Mon Mar  8 22:54:43 2010
@@ -73,11 +73,11 @@
     MMI->EndFunction();
 }
 
-/// RecordSourceLine - Records location information and associates it with a 
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
-                                       MDNode *Scope) {
+/// RecordSourceLine - Register a source line with debug info. Returns the
+/// unique label that was emitted and which provides correspondence to
+/// the source line list.
+MCSymbol *DwarfWriter::RecordSourceLine(unsigned Line, unsigned Col, 
+                                        MDNode *Scope) {
   return DD->recordSourceLine(Line, Col, Scope);
 }
 
@@ -92,7 +92,7 @@
   return DD && DD->ShouldEmitDwarfDebug();
 }
 
-void DwarfWriter::BeginScope(const MachineInstr *MI, unsigned L) {
+void DwarfWriter::BeginScope(const MachineInstr *MI, MCSymbol *L) {
   DD->beginScope(MI, L);
 }
 void DwarfWriter::EndScope(const MachineInstr *MI) {




From evan.cheng at apple.com  Tue Mar  9 00:38:18 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 09 Mar 2010 06:38:18 -0000
Subject: [llvm-commits] [llvm] r98048 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100309063818.0E9A22A6C12C@llvm.org>

Author: evancheng
Date: Tue Mar  9 00:38:17 2010
New Revision: 98048

URL: http://llvm.org/viewvc/llvm-project?rev=98048&view=rev
Log:
Allow more cross-rc coalescing.

Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98048&r1=98047&r2=98048&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Tue Mar  9 00:38:17 2010
@@ -93,16 +93,19 @@
     if (TII->isMoveInstr(*DefMI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) &&
         TargetRegisterInfo::isVirtualRegister(SrcReg) &&
         !SrcSubIdx && !DstSubIdx) {
-      const TargetRegisterClass *SRC = MRI->getRegClass(SrcReg);
-      const TargetRegisterClass *RC  = MRI->getRegClass(Reg);
-      if (SRC == RC || RC->hasSubClass(SRC)) {
-        DEBUG(dbgs() << "Coalescing: " << *DefMI);
-        DEBUG(dbgs() << "*** to: " << *MI);
-        MO.setReg(SrcReg);
-        DefMI->eraseFromParent();
-        ++NumCoalesces;
-        Changed = true;
-      }
+      const TargetRegisterClass *SRC   = MRI->getRegClass(SrcReg);
+      const TargetRegisterClass *RC    = MRI->getRegClass(Reg);
+      const TargetRegisterClass *NewRC = getCommonSubClass(RC, SRC);
+      if (!NewRC)
+        continue;
+      DEBUG(dbgs() << "Coalescing: " << *DefMI);
+      DEBUG(dbgs() << "*** to: " << *MI);
+      MO.setReg(SrcReg);
+      if (NewRC != SRC)
+        MRI->setRegClass(SrcReg, NewRC);
+      DefMI->eraseFromParent();
+      ++NumCoalesces;
+      Changed = true;
     }
   }
 




From baldrick at free.fr  Tue Mar  9 01:06:28 2010
From: baldrick at free.fr (Duncan Sands)
Date: Tue, 09 Mar 2010 08:06:28 +0100
Subject: [llvm-commits] [llvm-gcc-4.2] r97989
	-	/llvm-gcc-4.2/trunk/gcc/llvm-types.cpp
In-Reply-To: <20100308213302.618102A6C12C@llvm.org>
References: <20100308213302.618102A6C12C@llvm.org>
Message-ID: <4B95F374.4070505@free.fr>

Hi Dan,

>     // If this function returns via a shadow argument, the dest loc is passed
> -  // in as a pointer.  Mark that pointer as struct-ret and noalias.
> +  // in as a pointer.  Mark that pointer as struct-ret.
> +  //
> +  // It's tempting to want NoAlias here too, however even though llvm-gcc
> +  // itself currently always passes a dedicated alloca as the actual argument,
> +  // this isn't mandated by the ABI. There are other compilers which don't
> +  // always pass a dedicated alloca. Using NoAlias here would make code which
> +  // isn't interoperable with that of other compilers.
>     if (ABIConverter.isShadowReturn())
>       Attrs.push_back(AttributeWithIndex::get(ArgTys.size(),
> -                                    Attribute::StructRet | Attribute::NoAlias));
> +                                    Attribute::StructRet));

what does the ABI say exactly?  Is it really valid to pass something which may
get "accidentally" altered when the callee (for example) writes to a global or
to some other argument?

Ciao,

Duncan.


From baldrick at free.fr  Tue Mar  9 03:03:21 2010
From: baldrick at free.fr (Duncan Sands)
Date: Tue, 09 Mar 2010 09:03:21 -0000
Subject: [llvm-commits] [llvm] r98054 - /llvm/trunk/tools/Makefile
Message-ID: <20100309090321.8856A2A6C12C@llvm.org>

Author: baldrick
Date: Tue Mar  9 03:03:21 2010
New Revision: 98054

URL: http://llvm.org/viewvc/llvm-project?rev=98054&view=rev
Log:
Attempt to fix random build failures seen when doing highly
parallel builds: the gold plugin fails to link because the lto
library is in the middle of being written out by the linker.

Modified:
    llvm/trunk/tools/Makefile

Modified: llvm/trunk/tools/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/Makefile?rev=98054&r1=98053&r2=98054&view=diff
==============================================================================
--- llvm/trunk/tools/Makefile (original)
+++ llvm/trunk/tools/Makefile Tue Mar  9 03:03:21 2010
@@ -42,8 +42,7 @@
     # gold only builds if binutils is around.  It requires "lto" to build before
     # it so it is added to DIRS.
     ifdef BINUTILS_INCDIR
-      PARALLEL_DIRS += gold
-      DIRS += lto
+      DIRS += lto gold
     else
       PARALLEL_DIRS += lto
     endif




From richard at xmos.com  Tue Mar  9 10:07:47 2010
From: richard at xmos.com (Richard Osborne)
Date: Tue, 09 Mar 2010 16:07:47 -0000
Subject: [llvm-commits] [llvm] r98057 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp lib/Target/XCore/XCoreISelLowering.h
 test/CodeGen/XCore/ladd_lsub_combine.ll
Message-ID: <20100309160747.883462A6C12C@llvm.org>

Author: friedgold
Date: Tue Mar  9 10:07:47 2010
New Revision: 98057

URL: http://llvm.org/viewvc/llvm-project?rev=98057&view=rev
Log:
Add DAG combine for ladd / lsub.

Added:
    llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll
Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98057&r1=98056&r2=98057&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Tue Mar  9 10:07:47 2010
@@ -1097,6 +1097,48 @@
   DebugLoc dl = N->getDebugLoc();
   switch (N->getOpcode()) {
   default: break;
+  case XCoreISD::LADD: {
+    SDValue N0 = N->getOperand(0);
+    SDValue N1 = N->getOperand(1);
+    SDValue N2 = N->getOperand(2);
+    ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
+    ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
+    EVT VT = N0.getValueType();
+
+    // fold (ladd 0, 0, x) -> 0, x & 1
+    if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) {
+      SDValue Carry = DAG.getConstant(0, VT);
+      SDValue Result = DAG.getNode(ISD::AND, dl, VT, N2,
+                                   DAG.getConstant(1, VT));
+      SDValue Ops [] = { Carry, Result };
+      return DAG.getMergeValues(Ops, 2, dl);
+    }
+  }
+  break;
+  case XCoreISD::LSUB: {
+    SDValue N0 = N->getOperand(0);
+    SDValue N1 = N->getOperand(1);
+    SDValue N2 = N->getOperand(2);
+    ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
+    ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
+    EVT VT = N0.getValueType();
+
+    // fold (lsub 0, 0, x) -> x, -x iff x has only the low bit set
+    if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) {   
+      APInt KnownZero, KnownOne;
+      APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
+                                         VT.getSizeInBits() - 1);
+      DAG.ComputeMaskedBits(N2, Mask, KnownZero, KnownOne);
+      if (KnownZero == Mask) {
+        SDValue Borrow = N2;
+        SDValue Result = DAG.getNode(ISD::SUB, dl, VT,
+                                     DAG.getConstant(0, VT), N2);
+        SDValue Ops [] = { Borrow, Result };
+        return DAG.getMergeValues(Ops, 2, dl);
+      }
+    }
+  }
+  break;
   case ISD::STORE: {
     // Replace unaligned store of unaligned load with memmove.
     StoreSDNode *ST  = cast<StoreSDNode>(N);
@@ -1137,6 +1179,27 @@
   return SDValue();
 }
 
+void XCoreTargetLowering::computeMaskedBitsForTargetNode(const SDValue Op,
+                                                         const APInt &Mask,
+                                                         APInt &KnownZero,
+                                                         APInt &KnownOne,
+                                                         const SelectionDAG &DAG,
+                                                         unsigned Depth) const {
+  KnownZero = KnownOne = APInt(Mask.getBitWidth(), 0);
+  switch (Op.getOpcode()) {
+  default: break;
+  case XCoreISD::LADD:
+  case XCoreISD::LSUB:
+    if (Op.getResNo() == 0) {
+      // Top bits of carry / borrow are clear.
+      KnownZero = APInt::getHighBitsSet(Mask.getBitWidth(),
+                                        Mask.getBitWidth() - 1);
+      KnownZero &= Mask;
+    }
+    break;
+  }
+}
+
 //===----------------------------------------------------------------------===//
 //  Addressing mode description hooks
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98057&r1=98056&r2=98057&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Tue Mar  9 10:07:47 2010
@@ -144,6 +144,13 @@
 
     virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
 
+    virtual void computeMaskedBitsForTargetNode(const SDValue Op,
+                                                const APInt &Mask,
+                                                APInt &KnownZero,
+                                                APInt &KnownOne,
+                                                const SelectionDAG &DAG,
+                                                unsigned Depth = 0) const;
+
     virtual SDValue
       LowerFormalArguments(SDValue Chain,
                            CallingConv::ID CallConv,

Added: llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll?rev=98057&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll (added)
+++ llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll Tue Mar  9 10:07:47 2010
@@ -0,0 +1,28 @@
+; RUN: llvm-as < %s | llc -march=xcore | FileCheck %s
+
+; Only needs one ladd
+define i64 @f1(i32 %x, i32 %y) nounwind {
+entry:
+	%0 = zext i32 %x to i64		; <i64> [#uses=1]
+	%1 = zext i32 %y to i64		; <i64> [#uses=1]
+	%2 = add i64 %1, %0		; <i64> [#uses=1]
+	ret i64 %2
+}
+; CHECK: f1:
+; CHECK: ldc r2, 0
+; CHECK-NEXT: ladd r1, r0, r1, r0, r2
+; CHECK-NEXT: retsp 0
+
+; Only needs one lsub and one neg
+define i64 @f2(i32 %x, i32 %y) nounwind {
+entry:
+	%0 = zext i32 %x to i64		; <i64> [#uses=1]
+	%1 = zext i32 %y to i64		; <i64> [#uses=1]
+	%2 = sub i64 %1, %0		; <i64> [#uses=1]
+	ret i64 %2
+}
+; CHECK: f2:
+; CHECK: ldc r2, 0
+; CHECK-NEXT: lsub r1, r0, r1, r0, r2
+; CHECK-NEXT: neg r1, r1
+; CHECK-NEXT: retsp 0




From richard at xmos.com  Tue Mar  9 10:13:58 2010
From: richard at xmos.com (Richard Osborne)
Date: Tue, 09 Mar 2010 16:13:58 -0000
Subject: [llvm-commits] [llvm] r98058 -
	/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
Message-ID: <20100309161358.0E3772A6C12C@llvm.org>

Author: friedgold
Date: Tue Mar  9 10:13:57 2010
New Revision: 98058

URL: http://llvm.org/viewvc/llvm-project?rev=98058&view=rev
Log:
Canonicalize ladd constant to RHS.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98058&r1=98057&r2=98058&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Tue Mar  9 10:13:57 2010
@@ -1105,6 +1105,10 @@
     ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
     EVT VT = N0.getValueType();
 
+    // canonicalize constant to RHS
+    if (N0C && !N1C)
+      return DAG.getNode(XCoreISD::LADD, dl, DAG.getVTList(VT, VT), N1, N0, N2);
+
     // fold (ladd 0, 0, x) -> 0, x & 1
     if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) {
       SDValue Carry = DAG.getConstant(0, VT);




From richard at xmos.com  Tue Mar  9 10:34:25 2010
From: richard at xmos.com (Richard Osborne)
Date: Tue, 09 Mar 2010 16:34:25 -0000
Subject: [llvm-commits] [llvm] r98059 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp
 test/CodeGen/XCore/ladd_lsub_combine.ll
Message-ID: <20100309163425.EEBFF2A6C12C@llvm.org>

Author: friedgold
Date: Tue Mar  9 10:34:25 2010
New Revision: 98059

URL: http://llvm.org/viewvc/llvm-project?rev=98059&view=rev
Log:
In cases where the carry / borrow unused converted ladd / lsub
to an add or a sub.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98059&r1=98058&r2=98059&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Tue Mar  9 10:34:25 2010
@@ -1117,6 +1117,21 @@
       SDValue Ops [] = { Carry, Result };
       return DAG.getMergeValues(Ops, 2, dl);
     }
+
+    // fold (ladd x, 0, y) -> 0, add x, y iff carry is unused and y has only the
+    // low bit set
+    if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { 
+      APInt KnownZero, KnownOne;
+      APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
+                                         VT.getSizeInBits() - 1);
+      DAG.ComputeMaskedBits(N2, Mask, KnownZero, KnownOne);
+      if (KnownZero == Mask) {
+        SDValue Carry = DAG.getConstant(0, VT);
+        SDValue Result = DAG.getNode(ISD::ADD, dl, VT, N0, N2);
+        SDValue Ops [] = { Carry, Result };
+        return DAG.getMergeValues(Ops, 2, dl);
+      }
+    }
   }
   break;
   case XCoreISD::LSUB: {
@@ -1141,6 +1156,21 @@
         return DAG.getMergeValues(Ops, 2, dl);
       }
     }
+
+    // fold (lsub x, 0, y) -> 0, sub x, y iff borrow is unused and y has only the
+    // low bit set
+    if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { 
+      APInt KnownZero, KnownOne;
+      APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(),
+                                         VT.getSizeInBits() - 1);
+      DAG.ComputeMaskedBits(N2, Mask, KnownZero, KnownOne);
+      if (KnownZero == Mask) {
+        SDValue Borrow = DAG.getConstant(0, VT);
+        SDValue Result = DAG.getNode(ISD::SUB, dl, VT, N0, N2);
+        SDValue Ops [] = { Borrow, Result };
+        return DAG.getMergeValues(Ops, 2, dl);
+      }
+    }
   }
   break;
   case ISD::STORE: {

Modified: llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll?rev=98059&r1=98058&r2=98059&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/ladd_lsub_combine.ll Tue Mar  9 10:34:25 2010
@@ -26,3 +26,42 @@
 ; CHECK-NEXT: lsub r1, r0, r1, r0, r2
 ; CHECK-NEXT: neg r1, r1
 ; CHECK-NEXT: retsp 0
+
+; Should compile to one ladd and one add
+define i64 @f3(i64 %x, i32 %y) nounwind {
+entry:
+	%0 = zext i32 %y to i64		; <i64> [#uses=1]
+	%1 = add i64 %x, %0		; <i64> [#uses=1]
+	ret i64 %1
+}
+; CHECK: f3:
+; CHECK: ldc r3, 0
+; CHECK-NEXT: ladd r2, r0, r0, r2, r3
+; CHECK-NEXT: add r1, r1, r2
+; CHECK-NEXT: retsp 0
+
+; Should compile to one ladd and one add
+define i64 @f4(i32 %x, i64 %y) nounwind {
+entry:
+	%0 = zext i32 %x to i64		; <i64> [#uses=1]
+	%1 = add i64 %0, %y		; <i64> [#uses=1]
+	ret i64 %1
+}
+; CHECK: f4:
+; CHECK: ldc r3, 0
+; CHECK-NEXT: ladd r1, r0, r0, r1, r3
+; CHECK-NEXT: add r1, r2, r1
+; CHECK-NEXT: retsp 0
+
+; Should compile to one lsub and one sub
+define i64 @f5(i64 %x, i32 %y) nounwind {
+entry:
+	%0 = zext i32 %y to i64		; <i64> [#uses=1]
+	%1 = sub i64 %x, %0		; <i64> [#uses=1]
+	ret i64 %1
+}
+; CHECK: f5:
+; CHECK: ldc r3, 0
+; CHECK-NEXT: lsub r2, r0, r0, r2, r3
+; CHECK-NEXT: sub r1, r1, r2
+; CHECK-NEXT: retsp 0




From alenhar2 at llvm.org  Tue Mar  9 10:42:42 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Tue, 09 Mar 2010 16:42:42 -0000
Subject: [llvm-commits] [poolalloc] r98060 -
	/poolalloc/trunk/include/LinkDSA.h
Message-ID: <20100309164242.5D45F2A6C12C@llvm.org>

Author: alenhar2
Date: Tue Mar  9 10:42:42 2010
New Revision: 98060

URL: http://llvm.org/viewvc/llvm-project?rev=98060&view=rev
Log:
A file to pull to link dsa into a tool

Added:
    poolalloc/trunk/include/LinkDSA.h

Added: poolalloc/trunk/include/LinkDSA.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/LinkDSA.h?rev=98060&view=auto
==============================================================================
--- poolalloc/trunk/include/LinkDSA.h (added)
+++ poolalloc/trunk/include/LinkDSA.h Tue Mar  9 10:42:42 2010
@@ -0,0 +1,23 @@
+namespace {
+  struct ForceDSALinking {
+    ForceDSALinking() {
+      // We must reference the passes in such a way that compilers will not
+      // delete it all as dead code, even with whole program optimization,
+      // yet is effectively a NO-OP. As the compiler isn't smart enough
+      // to know that getenv() never returns -1, this will do the job.
+      if (std::getenv("bar") != (char*) -1)
+        return;
+
+      (void)new llvm::BasicDataStructures();
+      (void)new llvm::LocalDataStructures();
+      (void)new llvm::StdLibDataStructures();
+      (void)new llvm::BUDataStructures();
+      (void)new llvm::CompleteBUDataStructures();
+      (void)new llvm::EquivBUDataStructures();
+      (void)new llvm::TDDataStructures();
+      (void)new llvm::EQTDDataStructures();
+      (void)new llvm::SteensgaardDataStructures();
+      (void)new llvm::RTAssociate();
+    }
+  } ForceDSALinking; // Force link by creating a global definition.
+}




From alenhar2 at llvm.org  Tue Mar  9 10:43:02 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Tue, 09 Mar 2010 16:43:02 -0000
Subject: [llvm-commits] [poolalloc] r98061 - in /poolalloc/trunk: README
	lib/DSA/README
Message-ID: <20100309164302.3F9A62A6C12C@llvm.org>

Author: alenhar2
Date: Tue Mar  9 10:43:02 2010
New Revision: 98061

URL: http://llvm.org/viewvc/llvm-project?rev=98061&view=rev
Log:
Update readme with known issues

Added:
    poolalloc/trunk/lib/DSA/README
Modified:
    poolalloc/trunk/README

Modified: poolalloc/trunk/README
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/README?rev=98061&r1=98060&r2=98061&view=diff
==============================================================================
--- poolalloc/trunk/README (original)
+++ poolalloc/trunk/README Tue Mar  9 10:43:02 2010
@@ -5,6 +5,20 @@
 Before using the Automatic Pool Allocator, you should read the pre-release
 license in LICENSE.TXT.
 
+KNOWN ISSUES:
+=============
+
+Automatic Pool Allocation is currently broken on 2.7 due to malloc and
+free instructions being removed.  Use the release_26 branch of llvm
+and poolalloc for a working poolalloc module.
+
+DSA is undergoing significant changes and may not be entirely stable or 
+correct.  See lib/DSA/README.
+
+Windows does not support loadable modules in llvm, so poolalloc must
+be linked into opt.
+
+
 BUILDING:
 =========
 To build the Automatic Pool Allocator, you will need to have installed and

Added: poolalloc/trunk/lib/DSA/README
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/README?rev=98061&view=auto
==============================================================================
--- poolalloc/trunk/lib/DSA/README (added)
+++ poolalloc/trunk/lib/DSA/README Tue Mar  9 10:43:02 2010
@@ -0,0 +1,15 @@
+DSA is changed since PLDI07 in (at least) the following ways:
+
+1) DSA tracks types per offset.
+2) DSA does not assume that all clients will want to collapse a node if types 
+   conflict.  Many clients don't care that an offset is used as an int or a 
+   double, they just care about the points-to result.
+
+In Progress:
+
+1) DSA handles multiple entry points in a module.
+2) Optional assumption that only legal targets are called at indirect call sites
+3) Positional arguments, rather than pointer relative arguments.  This handles
+   the case where pointer and ints are assumed compatible and function pointers
+   containing them are cast freely.
+




From alenhar2 at llvm.org  Tue Mar  9 10:48:35 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Tue, 09 Mar 2010 16:48:35 -0000
Subject: [llvm-commits] [poolalloc] r98062 - in /poolalloc/trunk:
 include/dsa/DSGraph.h include/dsa/DSGraphTraits.h include/dsa/DSNode.h
 include/dsa/DSSupport.h lib/DSA/Basic.cpp lib/DSA/BottomUpClosure.cpp
 lib/DSA/DataStructure.cpp lib/DSA/Local.cpp lib/DSA/Printer.cpp
 lib/DSA/TopDownClosure.cpp
Message-ID: <20100309164836.1E11F2A6C12C@llvm.org>

Author: alenhar2
Date: Tue Mar  9 10:48:35 2010
New Revision: 98062

URL: http://llvm.org/viewvc/llvm-project?rev=98062&view=rev
Log:
Add byte offset based sets of types
Add byte offset links
add multiple entry points, WIP
get rid of stl/ext hash_
fix graph printer
api update to 2.7
handle malloc, sometimes (TODO: free, generalized allocator framework)
remove cruft
factor common operations out


Modified:
    poolalloc/trunk/include/dsa/DSGraph.h
    poolalloc/trunk/include/dsa/DSGraphTraits.h
    poolalloc/trunk/include/dsa/DSNode.h
    poolalloc/trunk/include/dsa/DSSupport.h
    poolalloc/trunk/lib/DSA/Basic.cpp
    poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
    poolalloc/trunk/lib/DSA/DataStructure.cpp
    poolalloc/trunk/lib/DSA/Local.cpp
    poolalloc/trunk/lib/DSA/Printer.cpp
    poolalloc/trunk/lib/DSA/TopDownClosure.cpp

Modified: poolalloc/trunk/include/dsa/DSGraph.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSGraph.h?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSGraph.h (original)
+++ poolalloc/trunk/include/dsa/DSGraph.h Tue Mar  9 10:48:35 2010
@@ -24,7 +24,7 @@
 
 namespace llvm {
 
-
+class TargetData;
 class GlobalValue;
 
 //===----------------------------------------------------------------------===//

Modified: poolalloc/trunk/include/dsa/DSGraphTraits.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSGraphTraits.h?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSGraphTraits.h (original)
+++ poolalloc/trunk/include/dsa/DSGraphTraits.h Tue Mar  9 10:48:35 2010
@@ -34,7 +34,7 @@
   DSNodeIterator(NodeTy *N) : Node(N), Offset(0) {}   // begin iterator
   DSNodeIterator(NodeTy *N, bool) : Node(N) {         // Create end iterator
     if (N != 0) {
-      Offset = N->getNumLinks();
+      Offset = N->getSize();
       if (Offset == 0 && Node->getForwardNode() &&
           Node->isDeadNode())        // Model Forward link
         Offset += 1;

Modified: poolalloc/trunk/include/dsa/DSNode.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSNode.h?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSNode.h (original)
+++ poolalloc/trunk/include/dsa/DSNode.h Tue Mar  9 10:48:35 2010
@@ -15,6 +15,8 @@
 #define LLVM_ANALYSIS_DSNODE_H
 
 #include "dsa/DSSupport.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/ilist_node.h"
 
 #include <map>
@@ -24,7 +26,6 @@
 
 template<typename BaseType>
 class DSNodeIterator;          // Data structure graph traversal iterator
-class TargetData;
 
 //===----------------------------------------------------------------------===//
 /// DSNode - Data structure node class
@@ -34,9 +35,14 @@
 /// different types represented in this object.
 ///
 class DSNode : public ilist_node<DSNode> {
+public:
+  typedef std::map<unsigned, std::set<const Type*> > TyMapTy;
+  typedef std::map<unsigned, DSNodeHandle> LinkMapTy;
+
+private:
   friend struct ilist_sentinel_traits<DSNode>;
   //Sentinel
-  DSNode() : NumReferrers(0), Size(0), Ty(0), NodeType(0) {}
+  DSNode() : NumReferrers(0), Size(0), NodeType(0) {}
   
   /// NumReferrers - The number of DSNodeHandles pointing to this node... if
   /// this is a forwarding node, then this is the number of node handles which
@@ -51,12 +57,6 @@
   ///
   DSNodeHandle ForwardNH;
 
-  /// Next, Prev - These instance variables are used to keep the node on a
-  /// doubly-linked ilist in the DSGraph.
-  ///
-  //DSNode *Next, *Prev;
-  //friend struct ilist_traits<DSNode>;
-
   /// Size - The current size of the node.  This should be equal to the size of
   /// the current type record.
   ///
@@ -66,18 +66,14 @@
   ///
   DSGraph *ParentGraph;
 
-  /// Ty - Keep track of the current outer most type of this object, in addition
-  /// to whether or not it has been indexed like an array or not.  If the
-  /// isArray bit is set, the node cannot grow.
-  ///
-  const Type *Ty;                 // The type itself...
-
-  /// Links - Contains one entry for every sizeof(void*) bytes in this memory
-  /// object.  Note that if the node is not a multiple of size(void*) bytes
-  /// large, that there is an extra entry for the "remainder" of the node as
-  /// well.  For this reason, nodes of 1 byte in size do have one link.
+  /// TyMap - Keep track of the loadable types and offsets those types are seen
+  // at.
+  TyMapTy TyMap;
+
+  /// Links - Contains one entry for every byte in this memory
+  /// object.  
   ///
-  std::vector<DSNodeHandle> Links;
+  LinkMapTy Links;
 
   /// Globals - The list of global values that are merged into this node.
   ///
@@ -100,13 +96,14 @@
     ReadNode        = 1 << 8,   // This node is read in this context
 
     ArrayNode       = 1 << 9,   // This node is treated like an array
-    ExternalNode    = 1 << 10,  // This node comes from an external source
-    IntToPtrNode    = 1 << 11,   // This node comes from an int cast
-    PtrToIntNode    = 1 << 12,  // This node excapes to an int cast
-    VAStartNode     = 1 << 13,  // This node excapes to an int cast
+    CollapsedNode   = 1 << 10,  // This node is collapsed
+    ExternalNode    = 1 << 11,  // This node comes from an external source
+    IntToPtrNode    = 1 << 12,   // This node comes from an int cast
+    PtrToIntNode    = 1 << 13,  // This node excapes to an int cast
+    VAStartNode     = 1 << 14,  // This node excapes to an int cast
 
     //#ifndef NDEBUG
-    DeadNode        = 1 << 14,   // This node is dead and should not be pointed to
+    DeadNode        = 1 << 15,   // This node is dead and should not be pointed to
     //#endif
 
     Composition = AllocaNode | HeapNode | GlobalNode | UnknownNode
@@ -123,7 +120,7 @@
   /// DSNode ctor - Create a node of the specified type, inserting it into the
   /// specified graph.
   ///
-  DSNode(const Type *T, DSGraph *G);
+  explicit DSNode(DSGraph *G);
 
   /// DSNode "copy ctor" - Copy the specified node, inserting it into the
   /// specified graph.  If NullLinks is true, then null out all of the links,
@@ -149,6 +146,21 @@
   inline const_iterator begin() const;
   inline const_iterator end() const;
 
+  /// type_* - Provide iterators for accessing types.  Some types may be null
+  TyMapTy::iterator type_begin() { return TyMap.begin(); }
+  TyMapTy::iterator type_end()   { return TyMap.end(); }
+  TyMapTy::const_iterator type_begin() const { return TyMap.begin(); }
+  TyMapTy::const_iterator type_end()   const { return TyMap.end(); }
+
+  /// edge_* - Provide iterators for accessing outgoing edges.  Some outgoing
+  /// edges may be null.
+  typedef LinkMapTy::iterator edge_iterator;
+  typedef LinkMapTy::const_iterator const_edge_iterator;
+  edge_iterator edge_begin() { return Links.begin(); }
+  edge_iterator edge_end() { return Links.end(); }
+  const_edge_iterator edge_begin() const { return Links.begin(); }
+  const_edge_iterator edge_end() const { return Links.end(); }
+
   //===--------------------------------------------------
   // Accessors
 
@@ -156,12 +168,6 @@
   ///
   unsigned getSize() const { return Size; }
 
-  /// getType - Return the node type of this object...
-  ///
-  const Type *getType() const { return Ty; }
-
-  bool isArray() const { return NodeType & ArrayNode; }
-
   /// hasNoReferrers - Return true if nothing is pointing to this node at all.
   ///
   bool hasNoReferrers() const { return getNumReferrers() == 0; }
@@ -171,14 +177,10 @@
   /// return the number of nodes forwarding over the node!
   unsigned getNumReferrers() const { return NumReferrers; }
 
+
   DSGraph *getParentGraph() const { return ParentGraph; }
   void setParentGraph(DSGraph *G) { ParentGraph = G; }
 
-
-  /// getTargetData - Get the target data object used to construct this node.
-  ///
-  const TargetData &getTargetData() const;
-
   /// getForwardNode - This method returns the node that this node is forwarded
   /// to, if any.
   ///
@@ -200,38 +202,30 @@
     delete this;
   }
 
+  void growSize(unsigned NSize) {
+    assert(NSize > Size && "Cannot shrink");
+    assert(!isCollapsedNode() && "growing a collapsed node");
+    Size = NSize;
+  }
+
   /// hasLink - Return true if this memory object has a link in slot LinkNo
   ///
   bool hasLink(unsigned Offset) const {
-    assert(Offset < Links.size() && "Link index is out of range!");
-    return Links[Offset].getNode();
+    assert(Offset < getSize() && "Link index is out of range!");
+    return Links.find(Offset) != Links.end();
   }
 
   /// getLink - Return the link at the specified offset.
   ///
   DSNodeHandle &getLink(unsigned Offset) {
-    assert(Offset < Links.size() && "Link index is out of range!");
+    assert(Offset < getSize() && "Link index is out of range!");
     return Links[Offset];
   }
   const DSNodeHandle &getLink(unsigned Offset) const {
-    assert(Offset < Links.size() && "Link index is out of range!");
-    return Links[Offset];
+    assert(hasLink(Offset) && "No Link");
+    return Links.find(Offset)->second;
   }
 
-  /// getNumLinks - Return the number of links in a node...
-  ///
-  unsigned getNumLinks() const { return Links.size(); }
-
-  /// edge_* - Provide iterators for accessing outgoing edges.  Some outgoing
-  /// edges may be null.
-  typedef std::vector<DSNodeHandle>::iterator edge_iterator;
-  typedef std::vector<DSNodeHandle>::const_iterator const_edge_iterator;
-  edge_iterator edge_begin() { return Links.begin(); }
-  edge_iterator edge_end() { return Links.end(); }
-  const_edge_iterator edge_begin() const { return Links.begin(); }
-  const_edge_iterator edge_end() const { return Links.end(); }
-
-
   /// mergeTypeInfo - This method merges the specified type into the current
   /// node at the specified offset.  This may update the current node's type
   /// record if this gives more information to the node, it may do nothing to
@@ -239,12 +233,10 @@
   /// completely (and return true) if the information is incompatible with what
   /// is already known.
   ///
-  /// This method returns true if the node is completely folded, otherwise
-  /// false.
-  ///
-  bool mergeTypeInfo(const Type *Ty, unsigned Offset,
-                     bool FoldIfIncompatible = true);
-
+  /// FIXME: description
+  void mergeTypeInfo(const Type *Ty, unsigned Offset);
+  void mergeTypeInfo(TyMapTy::const_iterator TyIt, unsigned Offset = 0);
+ 
   /// foldNodeCompletely - If we determine that this node has some funny
   /// behavior happening to it that we cannot represent, we fold it down to a
   /// single, completely pessimistic, node.  This node is represented as a
@@ -263,7 +255,7 @@
   /// instead one of the higher level methods should be used, below.
   ///
   void setLink(unsigned Offset, const DSNodeHandle &NH) {
-    assert(Offset < Links.size() && "Link index is out of range!");
+    assert(Offset < getSize() && "Link index is out of range!");
     Links[Offset] = NH;
   }
 
@@ -341,6 +333,7 @@
   bool isModifiedNode()   const { return NodeType & ModifiedNode;  }
   bool isReadNode()       const { return NodeType & ReadNode;      }
   bool isArrayNode()      const { return NodeType & ArrayNode;     }
+  bool isCollapsedNode()  const { return NodeType & CollapsedNode; }
   bool isIncompleteNode() const { return NodeType & IncompleteNode;}
   bool isCompleteNode()   const { return !isIncompleteNode();      }
   bool isDeadNode()       const { return NodeType & DeadNode;      }
@@ -358,6 +351,7 @@
   DSNode* setModifiedMarker()   { NodeType |= ModifiedNode;   return this; }
   DSNode* setReadMarker()       { NodeType |= ReadNode;       return this; }
   DSNode* setArrayMarker()      { NodeType |= ArrayNode;      return this; }
+  DSNode* setCollapsedMarker()  { NodeType |= CollapsedNode;  return this; }
   DSNode* setIncompleteMarker() { NodeType |= IncompleteNode; return this; }
   DSNode* setExternalMarker()   { NodeType |= ExternalNode;   return this; }
   DSNode* setIntToPtrMarker()   { NodeType |= IntToPtrNode;   return this; }
@@ -395,7 +389,7 @@
   /// DSNodes, marking any nodes which are reachable.  All reachable nodes it
   /// adds to the set, which allows it to only traverse visited nodes once.
   ///
-  void markReachableNodes(std::set<const DSNode*> &ReachableNodes) const;
+  void markReachableNodes(llvm::DenseSet<const DSNode*> &ReachableNodes) const;
 
 private:
   friend class DSNodeHandle;

Modified: poolalloc/trunk/include/dsa/DSSupport.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSSupport.h?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSSupport.h (original)
+++ poolalloc/trunk/include/dsa/DSSupport.h Tue Mar  9 10:48:35 2010
@@ -19,6 +19,7 @@
 #include <map>
 #include <set>
 
+#include "llvm/ADT/DenseSet.h"
 #include "llvm/Support/CallSite.h"
 
 namespace llvm {
@@ -54,8 +55,11 @@
   mutable unsigned Offset;
   void operator==(const DSNode *N);  // DISALLOW, use to promote N to nodehandle
 public:
+
+  DSNodeHandle() : N(0), Offset(0) {}
+
   // Allow construction, destruction, and assignment...
-  DSNodeHandle(DSNode *n = 0, unsigned offs = 0) : N(0), Offset(0) {
+  DSNodeHandle(DSNode *n, unsigned offs = 0) : N(0), Offset(0) {
     setTo(n, offs);
   }
   DSNodeHandle(const DSNodeHandle &H) : N(0), Offset(0) {
@@ -292,7 +296,7 @@
   /// DSNodes, marking any nodes which are reachable.  All reachable nodes it
   /// adds to the set, which allows it to only traverse visited nodes once.
   ///
-  void markReachableNodes(std::set<const DSNode*> &Nodes) const;
+  void markReachableNodes(DenseSet<const DSNode*> &Nodes) const;
 
   bool operator<(const DSCallSite &CS) const {
     if (isDirectCall()) {      // This must sort by callee first!

Modified: poolalloc/trunk/lib/DSA/Basic.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Basic.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Basic.cpp (original)
+++ poolalloc/trunk/lib/DSA/Basic.cpp Tue Mar  9 10:48:35 2010
@@ -38,11 +38,9 @@
   //
   // Create a void pointer type.  This is simply a pointer to an 8 bit value.
   //
-  const IntegerType * IT = IntegerType::getInt8Ty(M.getContext());
-  const PointerType * VoidPtrTy = PointerType::getUnqual(IT);
 
-  DSNode * GVNodeInternal = new DSNode(VoidPtrTy, GlobalsGraph);
-  DSNode * GVNodeExternal = new DSNode(VoidPtrTy, GlobalsGraph);
+  DSNode * GVNodeInternal = new DSNode(GlobalsGraph);
+  DSNode * GVNodeExternal = new DSNode(GlobalsGraph);
   for (Module::global_iterator I = M.global_begin(), E = M.global_end();
        I != E; ++I) {
     if (I->isDeclaration()) {
@@ -64,7 +62,7 @@
   for (Module::iterator F = M.begin(), E = M.end(); F != E; ++F) {
     if (!F->isDeclaration()) {
       DSGraph* G = new DSGraph(GlobalECs, getTargetData(), GlobalsGraph);
-      DSNode * Node = new DSNode(VoidPtrTy, G);
+      DSNode * Node = new DSNode(G);
           
       if (!F->hasInternalLinkage())
         Node->setExternalMarker();

Modified: poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/BottomUpClosure.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/BottomUpClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/BottomUpClosure.cpp Tue Mar  9 10:48:35 2010
@@ -56,15 +56,13 @@
 
   for (std::vector<const Function*>::iterator ii = EntryPoints.begin(),
           ee = EntryPoints.end(); ii != ee; ++ii)
-    if (!hasDSGraph(**ii))
+    if (!hasDSGraph(**ii)) {
+      errs() << debugname << ": Main Function: " << (*ii)->getName() << "\n";
       calculateGraphs(*ii, Stack, NextID, ValMap);
+      CloneAuxIntoGlobal(getDSGraph(**ii));
+    }
 
-   for (std::vector<const Function*>::iterator ii = EntryPoints.begin(),
-          ee = EntryPoints.end(); ii != ee; ++ii) {
-    cloneGlobalsInto(getDSGraph(**ii));
-    calculateGraph(getDSGraph(**ii));
-    CloneAuxIntoGlobal(getDSGraph(**ii));
-  }
+  //errs() << "done main Funcs\n";
 
   // Calculate the graphs for any functions that are unreachable from main...
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
@@ -78,6 +76,8 @@
       CloneAuxIntoGlobal(getDSGraph(*I));
     }
 
+  //errs() << "done unreachable Funcs\n";
+
   // If we computed any temporary indcallgraphs, free them now.
   for (std::map<std::vector<const Function*>,
          std::pair<DSGraph*, std::vector<DSNodeHandle> > >::iterator I =
@@ -133,7 +133,7 @@
       BadCalls.insert(*ii);
     else
       GoodCalls.insert(*ii);
-  std::set<const DSNode*> reachable;
+  DenseSet<const DSNode*> reachable;
   for (std::set<DSCallSite>::iterator ii = BadCalls.begin(),
          ee = BadCalls.end(); ii != ee; ++ii) {
     ii->getRetVal().getNode()->markReachableNodes(reachable);
@@ -142,7 +142,7 @@
   }
   for (std::set<DSCallSite>::iterator ii = GoodCalls.begin(),
          ee = GoodCalls.end(); ii != ee; ++ii)
-    if (reachable.find(ii->getCalleeNode()) == reachable.end())
+    if (reachable.count(ii->getCalleeNode()))
       GlobalsGraph->getAuxFunctionCalls()
         .erase(std::find(GlobalsGraph->getAuxFunctionCalls().begin(),
                          GlobalsGraph->getAuxFunctionCalls().end(),
@@ -341,6 +341,25 @@
 void BUDataStructures::calculateGraph(DSGraph* Graph) {
   DEBUG(Graph->AssertGraphOK(); Graph->getGlobalsGraph()->AssertGraphOK());
 
+  // If this graph contains the main function, clone the globals graph into this
+  // graph before we inline callees and other fun stuff.
+  bool ContainsMain = false;
+  DSGraph::ReturnNodesTy &ReturnNodes = Graph->getReturnNodes();
+
+  for (DSGraph::ReturnNodesTy::iterator I = ReturnNodes.begin(),
+       E = ReturnNodes.end(); I != E; ++I)
+    if (EP->isEntryPoint(I->first)) {
+      ContainsMain = true;
+      break;
+    }
+
+   // If this graph contains main, copy the contents of the globals graph over.
+   // Note that this is *required* for correctness.  If a callee contains a use
+   // of a global, we have to make sure to link up nodes due to global-argument
+   // bindings.
+   if (ContainsMain)
+     cloneGlobalsInto(Graph);
+
   // Move our call site list into TempFCs so that inline call sites go into the
   // new call site list and doesn't invalidate our iterators!
   std::list<DSCallSite> TempFCs;

Modified: poolalloc/trunk/lib/DSA/DataStructure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DataStructure.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DataStructure.cpp Tue Mar  9 10:48:35 2010
@@ -15,6 +15,7 @@
 #include "dsa/DataStructure.h"
 #include "dsa/DSGraph.h"
 #include "dsa/DSSupport.h"
+#include "dsa/DSNode.h"
 #include "llvm/Constants.h"
 #include "llvm/Function.h"
 #include "llvm/GlobalVariable.h"
@@ -43,9 +44,6 @@
   STATISTIC (NumDNE            , "Number of nodes removed by reachability");
 //  STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
 //  STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
-#ifdef LLVA_KERNEL
-  STATISTIC (LostPools         , "Number of pools lost to DSNode Merge");
-#endif
   static cl::opt<unsigned>
   DSAFieldLimit("dsa-field-limit", cl::Hidden,
                 cl::desc("Number of fields to track before collapsing a node"),
@@ -82,8 +80,12 @@
 
   N = Next;
   N->NumReferrers++;
-  if (N->Size <= Offset) {
-    assert(N->Size <= 1 && "Forwarded to shrunk but not collapsed node?");
+
+  if (N->isArrayNode())
+    Offset %= N->getSize();
+
+  if (N->getSize() <= Offset) {
+    assert(N->getSize() <= 1 && "Forwarded to shrunk but not collapsed node?");
     Offset = 0;
   }
   return N;
@@ -120,22 +122,20 @@
 // DSNode Implementation
 //===----------------------------------------------------------------------===//
 
-DSNode::DSNode(const Type *T, DSGraph *G)
-: NumReferrers(0), Size(0), ParentGraph(G), Ty(0), NodeType(0) {
+DSNode::DSNode(DSGraph *G)
+: NumReferrers(0), Size(0), ParentGraph(G), NodeType(0) {
   // Add the type entry if it is specified...
-  if (T) mergeTypeInfo(T, 0);
   if (G) G->addNode(this);
   ++NumNodeAllocated;
 }
 
 // DSNode copy constructor... do not copy over the referrers list!
 DSNode::DSNode(const DSNode &N, DSGraph *G, bool NullLinks)
-  : NumReferrers(0), Size(N.Size), ParentGraph(G),
-    Ty(N.Ty), Globals(N.Globals), NodeType(N.NodeType) {
+  : NumReferrers(0), Size(N.Size), ParentGraph(G), TyMap(N.TyMap),
+    Globals(N.Globals), NodeType(N.NodeType) {
   if (!NullLinks) {
     Links = N.Links;
-  } else
-    Links.resize(N.Links.size()); // Create the appropriate number of null links
+  }
   G->addNode(this);
   ++NumNodeAllocated;
 }
@@ -143,30 +143,13 @@
 DSNode::~DSNode() {
   dropAllReferences();
   assert(hasNoReferrers() && "Referrers to dead node exist!");
-
-#ifdef LLVA_KERNEL
-  //
-  // Remove all references to this node from the Pool Descriptor Map.
-  //
-  DEBUG(errs() << "LLVA: Removing " << this << "\n");
-  if (ParentGraph) {
-    hash_map<const DSNode *, MetaPoolHandle*> &pdm=ParentGraph->getPoolDescriptorsMap();
-    pdm.erase (this);
-  }
-#endif
-}
-
-/// getTargetData - Get the target data object used to construct this node.
-///
-const TargetData &DSNode::getTargetData() const {
-  return ParentGraph->getTargetData();
 }
 
 void DSNode::assertOK() const {
-  assert(((Ty && Ty->getTypeID() != Type::VoidTyID) ||
-         ((!Ty || Ty->getTypeID() == Type::VoidTyID) && (Size == 0 ||
-         (NodeType & DSNode::ArrayNode)))) &&
-         "Node not OK!");
+//  assert(((Ty && Ty->getTypeID() != Type::VoidTyID) ||
+//         ((!Ty || Ty->getTypeID() == Type::VoidTyID) && (Size == 0 ||
+//         (NodeType & DSNode::ArrayNode)))) &&
+//         "Node not OK!");
 
   assert(ParentGraph && "Node has no parent?");
   const DSScalarMap &SM = ParentGraph->getScalarMap();
@@ -188,7 +171,6 @@
   ForwardNH.setTo(To, Offset);
   NodeType = DeadNode;
   Size = 0;
-  Ty = 0;
 
   // Remove this node from the parent graph's Nodes list.
   ParentGraph->unlinkNode(this);
@@ -230,55 +212,45 @@
 void DSNode::foldNodeCompletely() {
   if (isNodeCompletelyFolded()) return;  // If this node is already folded...
 
+//  assert(0 && "Folding is happening");
+
   ++NumFolds;
 
   // If this node has a size that is <= 1, we don't need to create a forwarding
   // node.
   if (getSize() <= 1) {
-    NodeType |= DSNode::ArrayNode;
-    Ty = 0;
+    setCollapsedMarker();
     Size = 1;
     assert(Links.size() <= 1 && "Size is 1, but has more links?");
-    Links.resize(1);
   } else {
     // Create the node we are going to forward to.  This is required because
     // some referrers may have an offset that is > 0.  By forcing them to
     // forward, the forwarder has the opportunity to correct the offset.
-    DSNode *DestNode = new DSNode(0, ParentGraph);
-    DestNode->NodeType = NodeType|DSNode::ArrayNode;
-    DestNode->Ty = 0;
+    DSNode *DestNode = new DSNode(ParentGraph);
+    DestNode->NodeType = NodeType;
+    DestNode->setCollapsedMarker();
     DestNode->Size = 1;
     DestNode->Globals.swap(Globals);
-    
-    //    DOUT << "LLVA: foldNode: " << this << " becomes " << DestNode << "\n";
-#ifdef LLVA_KERNEL
-    //Again we have created a new DSNode, we need to fill in the
-    // pool desc map appropriately
-    assert(ParentGraph && "parent graph is not null \n"); 
-    hash_map<const DSNode *, MetaPoolHandle*> &pdm = ParentGraph->getPoolDescriptorsMap();
-    if (pdm.count(this) > 0) {
-      pdm[DestNode] = pdm[this];
-    } else {
-      //do nothing 
-    }
-#endif    
 
+    DSNodeHandle NH(DestNode);
+    
     // Start forwarding to the destination node...
     forwardNode(DestNode, 0);
 
-
     if (!Links.empty()) {
-      DestNode->Links.reserve(1);
-
-      DSNodeHandle NH(DestNode);
-      DestNode->Links.push_back(Links[0]);
 
       // If we have links, merge all of our outgoing links together...
-      for (unsigned i = Links.size()-1; i != 0; --i)
-        NH.getNode()->Links[0].mergeWith(Links[i]);
+      for (LinkMapTy::iterator ii = edge_begin(), ee = edge_end();
+           ii != ee; ++ii)
+        NH.getNode()->Links[0].mergeWith(ii->second);
       Links.clear();
-    } else {
-      DestNode->Links.resize(1);
+    }
+    //Merge types
+    if (!TyMap.empty()) {
+      for (TyMapTy::iterator ii = type_begin(), ee = type_end();
+           ii != ee; ++ii)
+        NH.getNode()->mergeTypeInfo(ii);
+      TyMap.clear();
     }
   }
 }
@@ -288,8 +260,7 @@
 /// all of the field sensitivity that may be present in the node.
 ///
 bool DSNode::isNodeCompletelyFolded() const {
-  return getSize() == 1 && (!Ty || Ty->getTypeID() == Type::VoidTyID)
-          && isArray();
+  return  isCollapsedNode();
 }
 
 /// addFullGlobalsList - Compute the full set of global values that are
@@ -330,179 +301,6 @@
   }
 }
 
-namespace {
-  /// TypeElementWalker Class - Used for implementation of physical subtyping...
-  ///
-  class TypeElementWalker {
-    struct StackState {
-      const Type *Ty;
-      unsigned Offset;
-      unsigned Idx;
-      StackState(const Type *T, unsigned Off = 0)
-        : Ty(T), Offset(Off), Idx(0) {}
-    };
-
-    std::vector<StackState> Stack;
-    const TargetData &TD;
-  public:
-    TypeElementWalker(const Type *T, const TargetData &td) : TD(td) {
-      Stack.push_back(T);
-      StepToLeaf();
-    }
-
-    bool isDone() const { return Stack.empty(); }
-    const Type *getCurrentType()   const { return Stack.back().Ty;     }
-    unsigned    getCurrentOffset() const { return Stack.back().Offset; }
-
-    void StepToNextType() {
-      PopStackAndAdvance();
-      StepToLeaf();
-    }
-
-  private:
-    /// PopStackAndAdvance - Pop the current element off of the stack and
-    /// advance the underlying element to the next contained member.
-    void PopStackAndAdvance() {
-      assert(!Stack.empty() && "Cannot pop an empty stack!");
-      Stack.pop_back();
-      while (!Stack.empty()) {
-        StackState &SS = Stack.back();
-        if (const StructType *ST = dyn_cast<StructType>(SS.Ty)) {
-          ++SS.Idx;
-          if (SS.Idx != ST->getNumElements()) {
-            const StructLayout *SL = TD.getStructLayout(ST);
-            SS.Offset +=
-               unsigned(SL->getElementOffset(SS.Idx)-SL->getElementOffset(SS.Idx-1));
-            return;
-          }
-          Stack.pop_back();  // At the end of the structure
-        } else {
-          const ArrayType *AT = cast<ArrayType>(SS.Ty);
-          ++SS.Idx;
-          if (SS.Idx != AT->getNumElements()) {
-            SS.Offset += unsigned(TD.getTypeAllocSize(AT->getElementType()));
-            return;
-          }
-          Stack.pop_back();  // At the end of the array
-        }
-      }
-    }
-
-    /// StepToLeaf - Used by physical subtyping to move to the first leaf node
-    /// on the type stack.
-    void StepToLeaf() {
-      if (Stack.empty()) return;
-      while (!Stack.empty() && !Stack.back().Ty->isFirstClassType()) {
-        StackState &SS = Stack.back();
-        if (const StructType *ST = dyn_cast<StructType>(SS.Ty)) {
-          if (ST->getNumElements() == 0) {
-            assert(SS.Idx == 0);
-            PopStackAndAdvance();
-          } else {
-            // Step into the structure...
-            assert(SS.Idx < ST->getNumElements());
-            const StructLayout *SL = TD.getStructLayout(ST);
-            Stack.push_back(StackState(ST->getElementType(SS.Idx),
-                            SS.Offset+unsigned(SL->getElementOffset(SS.Idx))));
-          }
-        } else {
-          const ArrayType *AT = cast<ArrayType>(SS.Ty);
-          if (AT->getNumElements() == 0) {
-            assert(SS.Idx == 0);
-            PopStackAndAdvance();
-          } else {
-            // Step into the array...
-            assert(SS.Idx < AT->getNumElements());
-            Stack.push_back(StackState(AT->getElementType(),
-                                       SS.Offset+SS.Idx*
-                             unsigned(TD.getTypeAllocSize(AT->getElementType()))));
-          }
-        }
-      }
-    }
-  };
-} // end anonymous namespace
-
-/// ElementTypesAreCompatible - Check to see if the specified types are
-/// "physically" compatible.  If so, return true, else return false.  We only
-/// have to check the fields in T1: T2 may be larger than T1.  If AllowLargerT1
-/// is true, then we also allow a larger T1.
-///
-static bool ElementTypesAreCompatible(const Type *T1, const Type *T2,
-                                      bool AllowLargerT1, const TargetData &TD){
-  TypeElementWalker T1W(T1, TD), T2W(T2, TD);
-
-  while (!T1W.isDone() && !T2W.isDone()) {
-    if (T1W.getCurrentOffset() != T2W.getCurrentOffset())
-      return false;
-
-    const Type *T1 = T1W.getCurrentType();
-    const Type *T2 = T2W.getCurrentType();
-    if (T1 != T2 && !T1->canLosslesslyBitCastTo(T2))
-      return false;
-
-    T1W.StepToNextType();
-    T2W.StepToNextType();
-  }
-
-  return AllowLargerT1 || T1W.isDone();
-}
-
-//
-// Function: getElementAtOffsetWithPadding()
-//
-// Description:
-//  Take a byte offset into a structure and return the type at that byte
-//  offset that *includes* any padding that occurs after the structure element.
-//
-static inline const Type *
-getElementAtOffsetWithPadding (const TargetData & TD,
-                               const StructLayout & SL,
-                               const StructType * STy,
-                               unsigned index) {
-  //
-  // Get the type of the element at the specified type index.
-  //
-  const Type * SubType = STy->getElementType(index);
-
-  //
-  // If this is the last element in the structure, just return the subtype;
-  // there is no padding after the last element.
-  //
-  if (index == (STy->getNumElements() - 1))
-    return SubType;
-
-  //
-  // Get the byte offset of this element and its successor element.
-  //
-  unsigned int thisOffset = (unsigned) SL.getElementOffset(index);
-  unsigned int nextOffset = (unsigned) SL.getElementOffset(index + 1);
-
-  //
-  // If the size of the current element is less than the distance between
-  // the offset at this index and the offset at the next index, then there is
-  // some padding in between this element and the next.  Create a structure
-  // type that contains this element *and* the padding.
-  //
-  if ((TD.getTypeAllocSize (SubType)) < (nextOffset - thisOffset)) {
-    //
-    // Create an array type that fits the padding size.
-    //
-    const Type * Int8Type = IntegerType::getInt8Ty(SubType->getContext());
-    const Type * paddingType = ArrayType::get (Int8Type, nextOffset-thisOffset);
-
-    //
-    // Create a structure type that contains the element and the padding array.
-    //
-    std::vector<const Type *> elementTypes;
-    elementTypes.push_back (SubType);
-    elementTypes.push_back (paddingType);
-    return (StructType::get (SubType->getContext(), elementTypes, true));
-  } else {
-    return SubType;
-  }
-}
-
 /// mergeTypeInfo - This method merges the specified type into the current node
 /// at the specified offset.  This may update the current node's type record if
 /// this gives more information to the node, it may do nothing to the node if
@@ -511,322 +309,28 @@
 ///
 /// This method returns true if the node is completely folded, otherwise false.
 ///
-bool DSNode::mergeTypeInfo(const Type *NewTy, unsigned Offset,
-                           bool FoldIfIncompatible) {
-  //DOUT << "merging " << *NewTy << " at " << Offset << " with " << *Ty << "\n";
-  if (!Ty) Ty = Type::getVoidTy(NewTy->getContext());
-
-  const TargetData &TD = getTargetData();
-  // Check to make sure the Size member is up-to-date.  Size can be one of the
-  // following:
-  //  Size = 0, Ty = Void: Nothing is known about this node.
-  //  Size = 0, Ty = FnTy: FunctionPtr doesn't have a size, so we use zero
-  //  Size = 1, Ty = Void, Array = 1: The node is collapsed
-  //  Otherwise, sizeof(Ty) = Size
-  //
-  const Type * VoidType = Type::getVoidTy(NewTy->getContext());
-  assert(((Size == 0 && Ty == VoidType && !isArray()) ||
-          (Size == 0 && !Ty->isSized() && !isArray()) ||
-          (Size == 1 && Ty == VoidType && isArray()) ||
-          (Size == 0 && !Ty->isSized() && !isArray()) ||
-          (TD.getTypeAllocSize(Ty) == Size)) &&
-         "Size member of DSNode doesn't match the type structure!");
-  assert(NewTy != VoidType && "Cannot merge void type into DSNode!");
-
-  if (Offset == 0 && NewTy == Ty)
-    return false;  // This should be a common case, handle it efficiently
-
-  // Return true immediately if the node is completely folded.
-  if (isNodeCompletelyFolded()) return true;
-
-  // If this is an array type, eliminate the outside arrays because they won't
-  // be used anyway.  This greatly reduces the size of large static arrays used
-  // as global variables, for example.
-  //
-  bool WillBeArray = false;
-  while (const ArrayType *AT = dyn_cast<ArrayType>(NewTy)) {
-    // FIXME: we might want to keep small arrays, but must be careful about
-    // things like: [2 x [10000 x int*]]
-    NewTy = AT->getElementType();
-    WillBeArray = true;
-  }
-
-  // Figure out how big the new type we're merging in is...
-  unsigned NewTySize = NewTy->isSized() ? (unsigned)TD.getTypeAllocSize(NewTy) : 0;
-
-  // Otherwise check to see if we can fold this type into the current node.  If
-  // we can't, we fold the node completely, if we can, we potentially update our
-  // internal state.
-  //
-  if (Ty == VoidType) {
-    // If this is the first type that this node has seen, just accept it without
-    // question....
-    assert(Offset == 0 && !isArray() &&
-           "Cannot have an offset into a void node!");
-
-    // If this node would have to have an unreasonable number of fields, just
-    // collapse it.  This can occur for fortran common blocks, which have stupid
-    // things like { [100000000 x double], [1000000 x double] }.
-    unsigned NumFields = NewTySize;
-    if (NumFields > DSAFieldLimit) {
-      foldNodeCompletely();
-      return true;
-    }
-
-    Ty = NewTy;
-    NodeType &= ~ArrayNode;
-    if (WillBeArray) NodeType |= ArrayNode;
-    Size = NewTySize;
-
-    // Calculate the number of outgoing links from this node.
-    Links.resize(NumFields);
-    return false;
-  }
-
-  // Handle node expansion case here...
-  if (Offset+NewTySize > Size) {
-    // It is illegal to grow this node if we have treated it as an array of
-    // objects...
-    if (isArray()) {
-      if (FoldIfIncompatible) foldNodeCompletely();
-      return true;
-    }
-
-    // If this node would have to have an unreasonable number of fields, just
-    // collapse it.  This can occur for fortran common blocks, which have stupid
-    // things like { [100000000 x double], [1000000 x double] }.
-    unsigned NumFields = NewTySize+Offset;
-    if (NumFields > DSAFieldLimit) {
-      foldNodeCompletely();
-      return true;
-    }
-
-    if (Offset) {
-      //handle some common cases:
-      // Ty:    struct { t1, t2, t3, t4, ..., tn}
-      // NewTy: struct { offset, stuff...}
-      // try merge with NewTy: struct {t1, t2, stuff...} if offset lands exactly
-      // on a field in Ty
-      if (isa<StructType>(NewTy) && isa<StructType>(Ty)) {
-        DEBUG(errs() << "Ty: " << *Ty << "\nNewTy: " << *NewTy << "@" << Offset << "\n");
-        const StructType *STy = cast<StructType>(Ty);
-        const StructLayout &SL = *TD.getStructLayout(STy);
-        unsigned i = SL.getElementContainingOffset(Offset);
-        //Either we hit it exactly or give up
-        if (SL.getElementOffset(i) != Offset) {
-          if (FoldIfIncompatible) foldNodeCompletely();
-          return true;
-        }
-        std::vector<const Type*> nt;
-        for (unsigned x = 0; x < i; ++x)
-          nt.push_back(STy->getElementType(x));
-        STy = cast<StructType>(NewTy);
-        nt.insert(nt.end(), STy->element_begin(), STy->element_end());
-        //and merge
-        STy = StructType::get(STy->getContext(), nt);
-        DEBUG(errs() << "Trying with: " << *STy << "\n");
-        return mergeTypeInfo(STy, 0);
-      }
-
-      //Ty: struct { t1, t2, t3 ... tn}
-      //NewTy T offset x
-      //try merge with NewTy: struct : {t1, t2, T} if offset lands on a field
-      //in Ty
-      if (isa<StructType>(Ty)) {
-        DEBUG(errs() << "Ty: " << *Ty << "\nNewTy: " << *NewTy << "@" << Offset << "\n");
-        const StructType *STy = cast<StructType>(Ty);
-        const StructLayout &SL = *TD.getStructLayout(STy);
-        unsigned i = SL.getElementContainingOffset(Offset);
-        //Either we hit it exactly or give up
-        if (SL.getElementOffset(i) != Offset) {
-          if (FoldIfIncompatible) foldNodeCompletely();
-          return true;
-        }
-        std::vector<const Type*> nt;
-        for (unsigned x = 0; x < i; ++x)
-          nt.push_back(STy->getElementType(x));
-        nt.push_back(NewTy);
-        //and merge
-        STy = StructType::get(STy->getContext(), nt);
-        DEBUG(errs() << "Trying with: " << *STy << "\n");
-        return mergeTypeInfo(STy, 0);
-      }
-
-      assert(0 &&
-             "UNIMP: Trying to merge a growth type into "
-             "offset != 0: Collapsing!");
-      abort();
-      if (FoldIfIncompatible) foldNodeCompletely();
-      return true;
-
-    }
-
-
-    // Okay, the situation is nice and simple, we are trying to merge a type in
-    // at offset 0 that is bigger than our current type.  Implement this by
-    // switching to the new type and then merge in the smaller one, which should
-    // hit the other code path here.  If the other code path decides it's not
-    // ok, it will collapse the node as appropriate.
-    //
-
-    const Type *OldTy = Ty;
-    Ty = NewTy;
-    NodeType &= ~ArrayNode;
-    if (WillBeArray) NodeType |= ArrayNode;
-    Size = NewTySize;
-
-    // Must grow links to be the appropriate size...
-    Links.resize(NumFields);
-
-    // Merge in the old type now... which is guaranteed to be smaller than the
-    // "current" type.
-    return mergeTypeInfo(OldTy, 0);
-  }
-
-  assert(Offset <= Size &&
-         "Cannot merge something into a part of our type that doesn't exist!");
-
-  // Find the section of Ty that NewTy overlaps with... first we find the
-  // type that starts at offset Offset.
-  //
-  unsigned O = 0;
-  const Type *SubType = Ty;
-  while (O < Offset) {
-    assert(Offset-O < TD.getTypeAllocSize(SubType) && "Offset out of range!");
-
-    switch (SubType->getTypeID()) {
-    case Type::StructTyID: {
-      const StructType *STy = cast<StructType>(SubType);
-      const StructLayout &SL = *TD.getStructLayout(STy);
-
-      unsigned i = SL.getElementContainingOffset(Offset-O);
-
-      //
-      // The offset we are looking for must be in the i'th element...
-      // However, be careful!  It is possible that the offset lands in a
-      // padding area of the structure.  Use a special methods that will either
-      // return us the type of the element we seek or a structure that contains
-      // an explicit element representing the padding.
-      //
-      SubType = getElementAtOffsetWithPadding (TD, SL, STy, i);
-      O += (unsigned)SL.getElementOffset(i);
-      break;
-    }
-    case Type::ArrayTyID: {
-      SubType = cast<ArrayType>(SubType)->getElementType();
-      unsigned ElSize = (unsigned)TD.getTypeAllocSize(SubType);
-      unsigned Remainder = (Offset-O) % ElSize;
-      O = Offset-Remainder;
-      break;
-    }
-    default:
-      if (FoldIfIncompatible) foldNodeCompletely();
-      return true;
-    }
-  }
-
-  assert(O == Offset && "Could not achieve the correct offset!");
-
-  // If we found our type exactly, early exit
-  if (SubType == NewTy) return false;
+void DSNode::mergeTypeInfo(const Type *NewTy, unsigned Offset) {
 
-  // Differing function types don't require us to merge.  They are not values
-  // anyway.
-  if (isa<FunctionType>(SubType) &&
-      isa<FunctionType>(NewTy)) return false;
-
-  unsigned SubTypeSize = SubType->isSized() ?
-       (unsigned)TD.getTypeAllocSize(SubType) : 0;
-
-  // Ok, we are getting desperate now.  Check for physical subtyping, where we
-  // just require each element in the node to be compatible.
-  if (NewTySize <= SubTypeSize && NewTySize && NewTySize < 256 &&
-      SubTypeSize && SubTypeSize < 256 &&
-      ElementTypesAreCompatible(NewTy, SubType, !isArray(), TD))
-    return false;
-
-  // Okay, so we found the leader type at the offset requested.  Search the list
-  // of types that starts at this offset.  If SubType is currently an array or
-  // structure, the type desired may actually be the first element of the
-  // composite type...
-  //
-  unsigned PadSize = SubTypeSize; // Size, including pad memory which is ignored
-  while (SubType != NewTy) {
-    const Type *NextSubType = 0;
-    unsigned NextSubTypeSize = 0;
-    unsigned NextPadSize = 0;
-    switch (SubType->getTypeID()) {
-    case Type::StructTyID: {
-      const StructType *STy = cast<StructType>(SubType);
-      const StructLayout &SL = *TD.getStructLayout(STy);
-      if (STy->getNumElements() > 1)
-        NextPadSize = (unsigned)SL.getElementOffset(1);
-      else
-        NextPadSize = SubTypeSize;
-      NextSubType = STy->getElementType(0);
-      NextSubTypeSize = (unsigned)TD.getTypeAllocSize(NextSubType);
-      break;
-    }
-    case Type::ArrayTyID:
-      NextSubType = cast<ArrayType>(SubType)->getElementType();
-      NextSubTypeSize = (unsigned)TD.getTypeAllocSize(NextSubType);
-      NextPadSize = NextSubTypeSize;
-      break;
-    default: ;
-      // fall out
-    }
+  if (!NewTy || NewTy->isVoidTy()) return;
 
-    if (NextSubType == 0)
-      break;   // In the default case, break out of the loop
+  if (isCollapsedNode()) Offset = 0;
+  if (isArrayNode()) Offset %= getSize();
 
-    if (NextPadSize < NewTySize)
-      break;   // Don't allow shrinking to a smaller type than NewTySize
-    SubType = NextSubType;
-    SubTypeSize = NextSubTypeSize;
-    PadSize = NextPadSize;
-  }
+  if (Offset >= getSize()) growSize(Offset+1);
 
-  // If we found the type exactly, return it...
-  if (SubType == NewTy)
-    return false;
+  TyMap[Offset].insert(NewTy);
+}
 
-  // Check to see if we have a compatible, but different type...
-  if (NewTySize == SubTypeSize) {
-    // Check to see if this type is obviously convertible... int -> uint f.e.
-    if (NewTy->canLosslesslyBitCastTo(SubType))
-      return false;
+void DSNode::mergeTypeInfo(TyMapTy::const_iterator TyIt, unsigned Offset) {
+  Offset += TyIt->first;
+  if (isCollapsedNode()) Offset = 0;
+  if (isArrayNode()) Offset %= getSize();
 
-    // Check to see if we have a pointer & integer mismatch going on here,
-    // loading a pointer as a long, for example.
-    //
-    if ((SubType->isIntegerTy() && NewTy->isPointerTy()) ||
-        (NewTy->isIntegerTy() && SubType->isPointerTy()))
-      return false;
-  } else if (NewTySize > SubTypeSize && NewTySize <= PadSize) {
-    // We are accessing the field, plus some structure padding.  Ignore the
-    // structure padding.
-    return false;
-  }
-
-  const Module *M = 0;
-  if (getParentGraph()->retnodes_begin() != getParentGraph()->retnodes_end())
-    M = getParentGraph()->retnodes_begin()->first->getParent();
-
-  DEBUG(errs() << "MergeTypeInfo Folding OrigTy: ");
-  raw_stderr_ostream stream;
-  DEBUG(WriteTypeSymbolic(stream, Ty, M);
-        stream << "\n due to:";
-        WriteTypeSymbolic(stream, NewTy, M);
-        stream << " @ " << Offset << "!\n" << "SubType: ";
-        WriteTypeSymbolic(stream, SubType, M);
-        stream << "\n\n");
+  if (Offset >= getSize()) growSize(Offset+1);
 
-  if (FoldIfIncompatible) foldNodeCompletely();
-  return true;
+  TyMap[Offset].insert(TyIt->second.begin(), TyIt->second.end());
 }
 
-
-
 /// addEdgeTo - Add an edge from the current node to the specified node.  This
 /// can cause merging of nodes in the graph.
 ///
@@ -855,6 +359,16 @@
   Globals.swap(Temp);
 }
 
+static unsigned gcd(unsigned m, unsigned n) {
+  if (m < n) return gcd(n,m);
+  unsigned r = m % n;
+  if (r == 0) {
+    return n;
+  } else {
+    return gcd(n, r);
+  }
+}
+
 // MergeNodes - Helper function for DSNode::mergeWith().
 // This function does the hard work of merging two nodes, CurNodeH
 // and NH after filtering out trivial cases and making sure that
@@ -891,62 +405,11 @@
     }
 #endif
   }
-#ifdef LLVA_KERNEL
-  DSNode *currNode  = CurNodeH.getNode();
-  DSNode *NNode  = NH.getNode();
-  DSGraph *pGraph =  currNode->getParentGraph();
-  assert((pGraph == NNode->getParentGraph()) && "LLVA_KERNEL : merging nodes in two different graphs?");
-  //get the pooldescriptor map
-  hash_map<const DSNode *, MetaPoolHandle*> &pdm = pGraph->getPoolDescriptorsMap();
-  if (pdm.count(currNode) == 0) {
-    if (pdm.count(NNode) == 0) {
-      //do nothing  (common case)
-    } else {
-      if (pdm[NNode]) {
-        DEBUG(errs() << "LLVA: 1: currNode (" << currNode << ") becomes " << pdm[NNode]->getName() << "(" << NNode << ")\n");
-        pdm[currNode] = pdm[NNode];
-      }
-    }
-  } else {
-    if (pdm.count(NNode) == 0) {
-#if 1
-      //
-      // FIXME:
-      //  Verify that this is correct.  I believe it is; it seems to make sense
-      //  since either node can be used after the merge.
-      //
-      DEBUG(errs() << "LLVA: MergeNodes: currnode has something, newnode has nothing\n"
-            << "LLVA: 2: currNode (" << currNode << ") becomes <no name>"
-            << "(" << NNode << ")\n");
-      pdm[NNode] = pdm[currNode];
-#endif
-      //do nothing 
-    } else {
-      if (pdm[currNode] != pdm[NNode]) {
-	//The following is commented because pdm[..] could be null!
-	//std::cerr << "LLVA: OldPool: " << pdm[currNode]->getName() << "("
-        //                       << pdm[currNode] << ") "
-	//          << " NewPool: "      << pdm[NNode]->getName() << "("
-	//                               << pdm[NNode] << ")" << std::endl;
-        pdm[NNode]->merge(pdm[currNode]);
-        /*
-        Value *currN = pdm[currNode]->getMetaPoolValue();
-        Value *NN = pdm[NNode]->getMetaPoolValue();
-        if (currN != NN) {
-          std::cerr << "LLVA: Two Pools for one DSNode\n";
-          currN->replaceAllUsesWith(NN);
-          pdm[currNode]->merge(pdm[NNode]);
-        } else {
-          //The nodes are same
-        }
-        */
-      }
-    }
-  }
-#endif  
+
   // Merge the type entries of the two nodes together...
-  if (NH.getNode()->Ty && NH.getNode()->Ty->getTypeID() != Type::VoidTyID)
-    CurNodeH.getNode()->mergeTypeInfo(NH.getNode()->Ty, NOffset);
+  for (TyMapTy::iterator ii = NH.getNode()->TyMap.begin(),
+      ee = NH.getNode()->TyMap.end(); ii != ee; ++ii)
+    CurNodeH.getNode()->mergeTypeInfo(ii, NOffset);
   assert(!CurNodeH.getNode()->isDeadNode());
 
   // If we are merging a node with a completely folded node, then both nodes are
@@ -970,6 +433,8 @@
     assert(NOffset == 0 && NSize == 1);
   }
 
+
+
   DSNode *N = NH.getNode();
   if (CurNodeH.getNode() == N || N == 0) return;
   assert(!CurNodeH.getNode()->isDeadNode());
@@ -983,9 +448,9 @@
 
   // Make all of the outgoing links of N now be outgoing links of CurNodeH.
   //
-  for (unsigned i = 0; i < N->getNumLinks(); ++i) {
-    DSNodeHandle &Link = N->getLink(i);
-    if (Link.getNode()) {
+  for (LinkMapTy::iterator ii = N->Links.begin(), ee = N->Links.end();
+       ii != ee; ++ii)
+    if (ii->second.getNode()) {
       // Compute the offset into the current node at which to
       // merge this link.  In the common case, this is a linear
       // relation to the offset in the original node (with
@@ -995,10 +460,9 @@
       unsigned MergeOffset = 0;
       DSNode *CN = CurNodeH.getNode();
       if (CN->Size != 1)
-        MergeOffset = (i+NOffset) % CN->getSize();
-      CN->addEdgeTo(MergeOffset, Link);
+        MergeOffset = (ii->first+NOffset) % CN->getSize();
+      CN->addEdgeTo(MergeOffset, ii->second);
     }
-  }
 
   // Now that there are no outgoing edges, all of the Links are dead.
   N->Links.clear();
@@ -1103,26 +567,15 @@
   DN->maskNodeTypes(BitsToKeep);
   NH = DN;
   //DOUT << "getClonedNH: " << SN << " becomes " << DN << "\n";
-#if 1
-#ifdef LLVA_KERNEL
-    //Again we have created a new DSNode, we need to fill in the
-    // pool desc map appropriately
-    hash_map<const DSNode *, MetaPoolHandle*> &pdm = Dest.getPoolDescriptorsMap();
-    if (pdm.count(SN) > 0) {
-      pdm[DN] = pdm[SN];
-    } else {
-      //do nothing 
-    }
-#endif    
-#endif
 
   // Next, recursively clone all outgoing links as necessary.  Note that
   // adding these links can cause the node to collapse itself at any time, and
   // the current node may be merged with arbitrary other nodes.  For this
   // reason, we must always go through NH.
   DN = 0;
-  for (unsigned i = 0, e = SN->getNumLinks(); i != e; ++i) {
-    const DSNodeHandle &SrcEdge = SN->getLink(i);
+  for (DSNode::LinkMapTy::const_iterator ii = SN->edge_begin(),
+          ee = SN->edge_end(); ii != ee; ++ii) {
+    const DSNodeHandle &SrcEdge = ii->second;
     if (!SrcEdge.isNull()) {
       const DSNodeHandle &DestEdge = getClonedNH(SrcEdge);
       // Compute the offset into the current node at which to
@@ -1134,7 +587,7 @@
       unsigned MergeOffset = 0;
       DSNode *CN = NH.getNode();
       if (CN->getSize() != 1)
-        MergeOffset = (i + NH.getOffset()) % CN->getSize();
+        MergeOffset = (ii->first + NH.getOffset()) % CN->getSize();
       CN->addEdgeTo(MergeOffset, DestEdge);
     }
   }
@@ -1206,12 +659,15 @@
 #endif
       }
 
+      if (DN->getSize() < SN->getSize())
+        DN->growSize(SN->getSize());
+
       // Merge the type entries of the two nodes together...
-      if ((SN->getType() && SN->getType()->getTypeID() != Type::VoidTyID)
-              && !DN->isNodeCompletelyFolded()) {
-        DN->mergeTypeInfo(SN->getType(), NH.getOffset() - SrcNH.getOffset());
-        DN = NH.getNode();
-      }
+      if (!DN->isNodeCompletelyFolded())
+        for (DSNode::TyMapTy::const_iterator ii = SN->type_begin(),
+             ee = SN->type_end(); ii != ee; ++ii)
+          DN->mergeTypeInfo(ii, NH.getOffset() - SrcNH.getOffset());
+
     }
 
     assert(!DN->isDeadNode());
@@ -1276,59 +732,14 @@
 
   //  DOUT << "LLVA: mergeWith: " << SN << " becomes " << DN << "\n";
 
-#ifdef LLVA_KERNEL
-  //Here some merge is going on just like in DSNode::merge
-  //I think because of the inplace merging we don't update the pool desc maps
-  //This is modification from DSNode::MergeNodes
-  //Here DN and SN may belong to different graphs
- DN = NH.getNode(); 
-#if 0
-  DSGraph *destGraph =  DN->getParentGraph();
-  DSGraph *srcGraph =  SN->getParentGraph();
-#else
-  DSGraph *destGraph = NH.getNode()->getParentGraph();
-  DSGraph *srcGraph =  SN->getParentGraph();
-#endif
-  if (destGraph && srcGraph) {
-    //get the pooldescriptor map
-    hash_map<const DSNode *, MetaPoolHandle*> &destpdm = destGraph->getPoolDescriptorsMap();
-    hash_map<const DSNode *, MetaPoolHandle*> &srcpdm = srcGraph->getPoolDescriptorsMap();
-    if (destpdm.count(DN) == 0) {
-      if (srcpdm.count(SN) == 0) {
-        //do nothing  (common case)
-      } else {
-        if (srcpdm[SN]) {
-          DEBUG(errs() << "DN becomes " << srcpdm[SN]->getName() << std::endl);
-          destpdm[DN] = srcpdm[SN];
-        }
-      }
-    } else {
-      if (srcpdm.count(SN) == 0) {
-        srcpdm[SN] = destpdm[DN];
-      } else {
-        if (destpdm[DN] != srcpdm[SN]) {
-          srcpdm[SN]->merge(destpdm[DN]);
-          /*
-          Value *dnv = destpdm[DN]->getMetaPoolValue();
-          Value *snv = srcpdm[SN]->getMetaPoolValue();
-          if (dnv != snv) {
-            DEBUG(std::cerr << "LLVA: Two Pools for one DSNode\n");
-            dnv->replaceAllUsesWith(snv);
-            destpdm[DN]->setMetaPoolValue(snv);
-          }
-          */
-        }
-      }
-    }
-  }
-#endif  
   // Next, recursively merge all outgoing links as necessary.  Note that
   // adding these links can cause the destination node to collapse itself at
   // any time, and the current node may be merged with arbitrary other nodes.
   // For this reason, we must always go through NH.
   DN = 0;
-  for (unsigned i = 0, e = SN->getNumLinks(); i != e; ++i) {
-    const DSNodeHandle &SrcEdge = SN->getLink(i);
+  for (DSNode::LinkMapTy::const_iterator ii = SN->edge_begin(),
+          ee = SN->edge_end(); ii != ee; ++ii) {
+    const DSNodeHandle &SrcEdge = ii->second;
     if (!SrcEdge.isNull()) {
       // Compute the offset into the current node at which to
       // merge this link.  In the common case, this is a linear
@@ -1337,8 +748,7 @@
       // recursive merging, we must make sure to merge in all remaining
       // links at offset zero.
       DSNode *CN = SCNH.getNode();
-      unsigned MergeOffset =
-        (i+SCNH.getOffset()) % CN->getSize();
+      unsigned MergeOffset = (ii->first+SCNH.getOffset()) % CN->getSize();
 
       DSNodeHandle Tmp = CN->getLink(MergeOffset);
       if (!Tmp.isNull()) {
@@ -1353,7 +763,7 @@
 
         unsigned MergeOffset = 0;
         CN = SCNH.getNode();
-        MergeOffset = (i + SCNH.getOffset()) %CN->getSize();
+        MergeOffset = (ii->first + SCNH.getOffset()) % CN->getSize();
         CN->getLink(MergeOffset).mergeWith(Tmp);
       }
     }
@@ -1405,53 +815,6 @@
   NH = RC.getClonedNH(Src);
 }
 
-#ifdef LLVA_KERNEL
-// MetaPoolHandle Implementation
-  //The following should go in a cpp file later
-   MetaPoolHandle::MetaPoolHandle(MetaPool *mp, Instruction * Maker) {
-    Rep = mp;
-    Rep->insert(this);
-    Creator = Maker;
-  }
-  const std::string& MetaPoolHandle::getName() {
-    assert(Rep != 0 && "Null meta pool ??\n");
-    return Rep->getName();
-  }
-  Value *MetaPoolHandle::getMetaPoolValue() {
-    assert(Rep != 0 && "Null meta pool ??\n");
-    return Rep->getMetaPoolValue();
-  }
-  void MetaPoolHandle::merge(MetaPoolHandle *other) {
-    //after this operation other points to what this points to .
-    //first replace all uses 
-     Value *dest = getMetaPoolValue();
-     Value *curr = other->getMetaPoolValue();
-     if (dest != curr) {
-       std::cerr << "LLVA: Merging metapools: " << this->Creator->getParent()->getParent()->getName() << " : " << other->Creator->getParent()->getParent()->getName() << "\n"
-                 << "LLVA:   " << *(this->Creator) << "\n"
-                 << "LLVA:   " << *(other->Creator) << "\n";
-       curr->replaceAllUsesWith(dest);
-     }
-   
-     //merge the hash sets in to other
-     hash_set<MetaPoolHandle *> &otherHandleSet = other->getMetaPool()->getHandleSet();
-     hash_set<MetaPoolHandle *>::iterator ohsI = otherHandleSet.begin(), ohsE = otherHandleSet.end();
-     for (; ohsI != ohsE; ++ohsI) {
-     MetaPoolHandle *omph = *ohsI;
-     //now make sure that this omph points to what we point to
-     omph->setMetaPool(Rep);
-     Rep->insert(omph);
-     }
-
-     //now delete others MetaPool
-     //gd     delete other->getMetaPool(); 
-
-     //Assign our metapool to other 
-     other->setMetaPool(Rep);
-}
-
-#endif
-
 //===----------------------------------------------------------------------===//
 // DSGraph Implementation
 //===----------------------------------------------------------------------===//
@@ -1502,12 +865,13 @@
 /// specified mapping.
 ///
 void DSNode::remapLinks(DSGraph::NodeMapTy &OldNodeMap) {
-  for (unsigned i = 0, e = Links.size(); i != e; ++i)
-    if (DSNode *N = Links[i].getNode()) {
+  for (LinkMapTy::iterator ii = edge_begin(), ee = edge_end();
+       ii != ee; ++ii)
+    if (DSNode *N = ii->second.getNode()) {
       DSGraph::NodeMapTy::const_iterator ONMI = OldNodeMap.find(N);
       if (ONMI != OldNodeMap.end()) {
         DSNode *ONMIN = ONMI->second.getNode();
-        Links[i].setTo(ONMIN, Links[i].getOffset()+ONMI->second.getOffset());
+        ii->second.setTo(ONMIN, ii->second.getOffset()+ONMI->second.getOffset());
       }
     }
 }
@@ -1534,8 +898,7 @@
 /// and does not point to any other objects in the graph.
 DSNode *DSGraph::addObjectToGraph(Value *Ptr, bool UseDeclaredType) {
   assert(isa<PointerType>(Ptr->getType()) && "Ptr is not a pointer!");
-  const Type *Ty = cast<PointerType>(Ptr->getType())->getElementType();
-  DSNode *N = new DSNode(UseDeclaredType ? Ty : 0, this);
+  DSNode *N = new DSNode(this);
   assert(ScalarMap[Ptr].isNull() && "Object already in this graph!");
   ScalarMap[Ptr] = N;
 
@@ -1758,7 +1121,7 @@
   bool AnyDirectSuccessorsReachClonedNodes = false;
   for (DSNode::const_edge_iterator EI = N->edge_begin(), EE = N->edge_end();
        EI != EE; ++EI)
-    if (DSNode *Succ = EI->getNode()) {
+    if (DSNode *Succ = EI->second.getNode()) {
       std::pair<unsigned, bool> &SuccInfo = VisitForSCCs(Succ);
       if (SuccInfo.first < Min) Min = SuccInfo.first;
       AnyDirectSuccessorsReachClonedNodes |= SuccInfo.second;
@@ -1778,7 +1141,7 @@
     for (unsigned i = SCCStack.size()-1; SCCStack[i] != N; --i)
       for (DSNode::const_edge_iterator EI = N->edge_begin(), EE = N->edge_end();
            EI != EE; ++EI)
-        if (DSNode *N = EI->getNode())
+        if (DSNode *N = EI->second.getNode())
           if (NodeInfo[N].second) {
             AnyDirectSuccessorsReachClonedNodes = true;
             goto OutOfLoop;
@@ -1872,7 +1235,7 @@
     DSNode *GlobalNode = Graph.getNodeForValue(*GI).getNode();
     for (DSNode::edge_iterator EI = GlobalNode->edge_begin(),
            EE = GlobalNode->edge_end(); EI != EE; ++EI)
-      if (SCCFinder.PathExistsToClonedNode(EI->getNode())) {
+      if (SCCFinder.PathExistsToClonedNode(EI->second.getNode())) {
         GlobalsToCopy.push_back(*GI);
         break;
       }
@@ -1961,7 +1324,7 @@
 
   // Recursively process children...
   for (DSNode::edge_iterator I = N->edge_begin(),E = N->edge_end(); I != E; ++I)
-    if (DSNode *DSN = I->getNode())
+    if (DSNode *DSN = I->second.getNode())
       markIncompleteNode(DSN);
 }
 
@@ -2037,9 +1400,9 @@
   if (DSNode * N = Edge.getNode()) // Is there an edge?
     if (N->getNumReferrers() == 1)  // Does it point to a lonely node?
       // No interesting info?
-      if ((N->getNodeFlags() & ~DSNode::IncompleteNode) == 0 &&
-          (!N->getType() || N->getType()->getTypeID() == Type::VoidTyID)
-              && !N->isNodeCompletelyFolded())
+      if ((N->getNodeFlags() & ~DSNode::IncompleteNode) == 0
+          && N->type_begin() == N->type_end()
+          && !N->isNodeCompletelyFolded())
         Edge.setTo(0, 0);  // Kill the edge!
 }
 
@@ -2288,16 +1651,16 @@
 /// DSNodes, marking any nodes which are reachable.  All reachable nodes it adds
 /// to the set, which allows it to only traverse visited nodes once.
 ///
-void DSNode::markReachableNodes(std::set<const DSNode*> &ReachableNodes) const {
+void DSNode::markReachableNodes(DenseSet<const DSNode*> &ReachableNodes) const {
   if (this == 0) return;
   assert(getForwardNode() == 0 && "Cannot mark a forwarded node!");
   if (ReachableNodes.insert(this).second)        // Is newly reachable?
     for (DSNode::const_edge_iterator I = edge_begin(), E = edge_end();
          I != E; ++I)
-      I->getNode()->markReachableNodes(ReachableNodes);
+      I->second.getNode()->markReachableNodes(ReachableNodes);
 }
 
-void DSCallSite::markReachableNodes(std::set<const DSNode*> &Nodes) const {
+void DSCallSite::markReachableNodes(DenseSet<const DSNode*> &Nodes) const {
   getRetVal().getNode()->markReachableNodes(Nodes);
   if (isIndirectCall()) getCalleeNode()->markReachableNodes(Nodes);
 
@@ -2310,8 +1673,8 @@
 // true, otherwise return false.  If an alive node is reachable, this node is
 // marked as alive...
 //
-static bool CanReachAliveNodes(DSNode *N, std::set<const DSNode*> &Alive,
-                               std::set<const DSNode*> &Visited,
+static bool CanReachAliveNodes(DSNode *N, DenseSet<const DSNode*> &Alive,
+                               DenseSet<const DSNode*> &Visited,
                                bool IgnoreGlobals) {
   if (N == 0) return false;
   assert(N->getForwardNode() == 0 && "Cannot mark a forwarded node!");
@@ -2329,7 +1692,7 @@
   Visited.insert(N);   // No recursion, insert into Visited...
 
   for (DSNode::edge_iterator I = N->edge_begin(),E = N->edge_end(); I != E; ++I)
-    if (CanReachAliveNodes(I->getNode(), Alive, Visited, IgnoreGlobals)) {
+    if (CanReachAliveNodes(I->second.getNode(), Alive, Visited, IgnoreGlobals)) {
       N->markReachableNodes(Alive);
       return true;
     }
@@ -2340,8 +1703,8 @@
 // alive nodes.
 //
 static bool CallSiteUsesAliveArgs(const DSCallSite &CS,
-                                  std::set<const DSNode*> &Alive,
-                                  std::set<const DSNode*> &Visited,
+                                  DenseSet<const DSNode*> &Alive,
+                                  DenseSet<const DSNode*> &Visited,
                                   bool IgnoreGlobals) {
   if (CanReachAliveNodes(CS.getRetVal().getNode(), Alive, Visited,
                          IgnoreGlobals))
@@ -2372,7 +1735,7 @@
   // FIXME: Merge non-trivially identical call nodes...
 
   // Alive - a set that holds all nodes found to be reachable/alive.
-  std::set<const DSNode*> Alive;
+  DenseSet<const DSNode*> Alive;
   std::vector<std::pair<const Value*, DSNode*> > GlobalNodes;
 
   // Copy and merge all information about globals to the GlobalsGraph if this is
@@ -2420,7 +1783,7 @@
   // value (which makes them live in turn), and continue till no more are found.
   //
   bool Iterate;
-  std::set<const DSNode*> Visited;
+  DenseSet<const DSNode*> Visited;
   std::set<const DSCallSite*> AuxFCallsAlive;
   do {
     Visited.clear();
@@ -2849,7 +2212,7 @@
   bool MadeChange = false;
   for (DSScalarMap::global_iterator GI = SM.global_begin(), E = SM.global_end();
        GI != E; ) {
-    const GlobalValue *GV = *GI++;
+    const GlobalValue *GV = *GI; ++GI;
     if (!ECGlobals.count(GV)) continue;
 
     const DSNodeHandle &GVNH = SM[GV];

Modified: poolalloc/trunk/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Local.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Local.cpp (original)
+++ poolalloc/trunk/lib/DSA/Local.cpp Tue Mar  9 10:48:35 2010
@@ -18,6 +18,7 @@
 #include "llvm/DerivedTypes.h"
 #include "llvm/Instructions.h"
 #include "llvm/Intrinsics.h"
+#include "llvm/InlineAsm.h"
 #include "llvm/Support/GetElementPtrTypeIterator.h"
 #include "llvm/Support/InstVisitor.h"
 #include "llvm/Target/TargetData.h"
@@ -28,7 +29,6 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/DenseSet.h"
 
-#include <iostream>
 #include <fstream>
 
 // FIXME: This should eventually be a FunctionPass that is automatically
@@ -56,6 +56,7 @@
     DSGraph &G;
     Function* FB;
     DataStructures* DS;
+    const TargetData& TD;
 
     ////////////////////////////////////////////////////////////////////////////
     // Helper functions used to implement the visitation functions...
@@ -65,9 +66,9 @@
     /// createNode - Create a new DSNode, ensuring that it is properly added to
     /// the graph.
     ///
-    DSNode *createNode(const Type *Ty = 0) 
+    DSNode *createNode() 
     {   
-      DSNode* ret = new DSNode(Ty, &G);
+      DSNode* ret = new DSNode(&G);
       assert(ret->getParentGraph() && "No parent?");
       return ret;
     }
@@ -92,6 +93,17 @@
     // Visitor functions, used to handle each instruction type we encounter...
     friend class InstVisitor<GraphBuilder>;
 
+    //TODO: generalize
+    bool visitAllocation(CallSite CS) {
+      if (Function* F = CS.getCalledFunction()) {
+        if (F->hasName() && F->getName() == "malloc") {
+          setDestTo(*CS.getInstruction(), createNode()->setHeapMarker());
+          return true;
+        }
+      }
+      return false;
+    }
+
     //FIXME: implement in stdlib pass
 //    void visitMallocInst(MallocInst &MI)
 //    { setDestTo(MI, createNode()->setHeapMarker()); }
@@ -130,7 +142,7 @@
 
   public:
     GraphBuilder(Function &f, DSGraph &g, DataStructures& DSi)
-      : G(g), FB(&f), DS(&DSi) {
+      : G(g), FB(&f), DS(&DSi), TD(g.getTargetData()) {
       // Create scalar nodes for all pointer arguments...
       for (Function::arg_iterator I = f.arg_begin(), E = f.arg_end();
            I != E; ++I) {
@@ -172,7 +184,7 @@
 
     // GraphBuilder ctor for working on the globals graph
     explicit GraphBuilder(DSGraph& g)
-      :G(g), FB(0)
+      :G(g), FB(0), TD(g.getTargetData())
     {}
 
     void mergeInGlobalInitializer(GlobalVariable *GV);
@@ -196,8 +208,8 @@
       visited.insert(N);
       N->setUnknownMarker();
       for (DSNode::edge_iterator I = N->edge_begin(), E = N->edge_end(); I != E; ++I)
-        if (!I->isNull())
-          workList.push_back(I->getNode());
+        if (!I->second.isNull())
+          workList.push_back(I->second.getNode());
     }
   }
 }
@@ -224,7 +236,7 @@
   DSNode* N;
   if (GlobalVariable* GV = dyn_cast<GlobalVariable>(V)) {
     // Create a new global node for this global variable.
-    N = createNode(GV->getType()->getElementType());
+    N = createNode();
     N->addGlobal(GV);
     if (GV->isDeclaration())
       N->setExternGlobalMarker();
@@ -247,7 +259,7 @@
         NH = G.getNodeForValue(CE);
       } else {
         // This returns a conservative unknown node for any unhandled ConstExpr
-        return NH = createNode()->setUnknownMarker();
+        NH = createNode()->setUnknownMarker();
       }
       if (NH.isNull()) {  // (getelementptr null, X) returns null
         G.eraseNodeForValue(V);
@@ -263,7 +275,7 @@
       // not handle the aliases of parameters correctly. Here is only a quick
       // fix for some special cases.
       NH = getValueDest(cast<GlobalAlias>(C)->getAliasee());
-      return 0;
+      return NH;
     } else {
       errs() << "Unknown constant: " << *C << "\n";
       assert(0 && "Unknown constant type!");
@@ -340,7 +352,7 @@
   Ptr.getNode()->setReadMarker();
 
   // Ensure a typerecord exists...
-  Ptr.getNode()->mergeTypeInfo(LI.getType(), Ptr.getOffset(), false);
+  Ptr.getNode()->mergeTypeInfo(LI.getType(), Ptr.getOffset());
 
   if (isa<PointerType>(LI.getType()))
     setDestTo(LI, getLink(Ptr));
@@ -377,7 +389,7 @@
 
   // Ensure a type record exists.
   DSNode *PtrN = Ptr.getNode();
-  PtrN->mergeTypeInfo(I.getType(), Ptr.getOffset(), false);
+  PtrN->mergeTypeInfo(I.getType(), Ptr.getOffset());
 
   if (isa<PointerType>(I.getType()))
     setDestTo(I, getLink(Ptr));
@@ -433,7 +445,7 @@
 
   // Ensure a typerecord exists...
   // FIXME: calculate offset
-  Ptr.getNode()->mergeTypeInfo(I.getType(), 0, false);
+  Ptr.getNode()->mergeTypeInfo(I.getType(), 0);
 
   if (isa<PointerType>(I.getType()))
     setDestTo(I, getLink(Ptr));
@@ -464,18 +476,11 @@
     return;
   }
 
-
-  const PointerType *PTy = cast<PointerType>(GEP.getOperand(0)->getType());
+  //Make sure the uncollapsed node has a large enough size for the struct type
+  const PointerType *PTy = cast<PointerType > (GEP.getOperand(0)->getType());
   const Type *CurTy = PTy->getElementType();
-
-  if (Value.getNode()->mergeTypeInfo(CurTy, Value.getOffset())) {
-    // If the node had to be folded... exit quickly
-    setDestTo(GEP, Value);  // GEP result points to folded node
-
-    return;
-  }
-
-  const TargetData &TD = Value.getNode()->getTargetData();
+  if (TD.getTypeAllocSize(CurTy) + Value.getOffset() > Value.getNode()->getSize())
+    Value.getNode()->growSize(TD.getTypeAllocSize(CurTy) + Value.getOffset());
 
 #if 0
   // Handle the pointer index specially...
@@ -511,12 +516,7 @@
        I != E; ++I)
     if (const StructType *STy = dyn_cast<StructType>(*I)) {
       const ConstantInt* CUI = cast<ConstantInt>(I.getOperand());
-#if 0
-      unsigned FieldNo = 
-        CUI->getType()->isSigned() ? CUI->getSExtValue() : CUI->getZExtValue();
-#else
       int FieldNo = CUI->getSExtValue();
-#endif
       Offset += (unsigned)TD.getStructLayout(STy)->getElementOffset(FieldNo);
     } else if (isa<PointerType>(*I)) {
       if (!isa<Constant>(I.getOperand()) ||
@@ -556,7 +556,7 @@
 #endif
 
   // Add in the offset calculated...
-  Value.setOffset(Value.getOffset()+Offset);
+ Value.setOffset(Value.getOffset()+Offset);
 
   // Check the offset
   DSNode *N = Value.getNode();
@@ -577,20 +577,6 @@
 
   // Value is now the pointer we want to GEP to be...  
   setDestTo(GEP, Value);
-#if 0
-  if (debug && (isa<Instruction>(GEP))) {
-    Instruction * IGEP = (Instruction *)(&GEP);
-    DSNode * N = Value.getNode();
-    if (IGEP->getParent()->getParent()->getName() == "alloc_vfsmnt")
-    {
-      if (G.getPoolDescriptorsMap().count(N) != 0)
-        if (G.getPoolDescriptorsMap()[N]) {
-	  DEBUG(errs() << "LLVA: GEP[" << 0 << "]: Pool for " << GEP.getName() << " is " << G.getPoolDescriptorsMap()[N]->getName() << "\n");
-	}
-    }
-  }
-#endif
-
 }
 
 
@@ -753,6 +739,16 @@
     if (F->isIntrinsic() && visitIntrinsic(CS, F))
       return;
 
+  if (visitAllocation(CS))
+    return;
+
+  if (isa<InlineAsm>(CS.getCalledValue())) return;
+//  if (InlineAsm* IASM = dyn_cast<InlineAsm>(CS.getCalledValue())) {
+//    if (IASM->hasSideEffects())
+//      errs() << ASM w/ Side Effects\n";
+//    return;
+//  }
+
   // Set up the return value...
   DSNodeHandle RetVal;
   Instruction *I = CS.getInstruction();
@@ -826,8 +822,6 @@
 
   if (Ty->isIntOrIntVectorTy() || Ty->isFPOrFPVectorTy()) return;
 
-  const TargetData &TD = NH.getNode()->getTargetData();
-
   if (ConstantArray *CA = dyn_cast<ConstantArray>(C)) {
     for (unsigned i = 0, e = CA->getNumOperands(); i != e; ++i)
       // We don't currently do any indexing for arrays...

Modified: poolalloc/trunk/lib/DSA/Printer.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Printer.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Printer.cpp (original)
+++ poolalloc/trunk/lib/DSA/Printer.cpp Tue Mar  9 10:48:35 2010
@@ -60,11 +60,20 @@
   if (N->isNodeCompletelyFolded())
     OS << "COLLAPSED";
   else {
-    if (N->getType())
-      WriteTypeSymbolic(OS, N->getType(), M);
+    if (N->type_begin() != N->type_end())
+      for (DSNode::TyMapTy::const_iterator ii = N->type_begin(),
+           ee = N->type_end(); ii != ee; ++ii) {
+        OS << ii->first << ": ";
+        for (DSNode::TyMapTy::mapped_type::iterator ni = ii->second.begin(),
+                ne = ii->second.end(); ni != ne; ++ni) {
+          WriteTypeSymbolic(OS, *ni, M);
+          OS << ", ";
+        }
+        OS << " ";
+      }
     else
       OS << "VOID";
-    if (N->isArray())
+    if (N->isArrayNode())
       OS << " array";
   }
   if (unsigned NodeType = N->getNodeFlags()) {
@@ -137,6 +146,11 @@
     return O != 0;
   }
 
+  static std::string getEdgeSourceLabel(const void* Node,
+                                        DSNode::const_iterator I) {
+    return "*";
+  }
+
   static DSNode::const_iterator getEdgeTarget(const DSNode *Node,
                                               DSNode::const_iterator I) {
     unsigned O = I.getNode()->getLink(I.getOffset()).getOffset();

Modified: poolalloc/trunk/lib/DSA/TopDownClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/TopDownClosure.cpp?rev=98062&r1=98061&r2=98062&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/TopDownClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/TopDownClosure.cpp Tue Mar  9 10:48:35 2010
@@ -58,8 +58,9 @@
   if (!N || Visited.count(N)) return;
   Visited.insert(N);
 
-  for (unsigned i = 0, e = N->getNumLinks(); i != e; ++i) {
-    DSNodeHandle &NH = N->getLink(i);
+  for (DSNode::LinkMapTy::iterator ii = N->edge_begin(),
+          ee = N->edge_end(); ii != ee; ++ii) {
+    DSNodeHandle &NH = ii->second;
     if (DSNode *NN = NH.getNode()) {
       std::vector<const Function*> Functions;
       NN->addFullFunctionList(Functions);




From alenhar2 at llvm.org  Tue Mar  9 10:50:15 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Tue, 09 Mar 2010 16:50:15 -0000
Subject: [llvm-commits] [poolalloc] r98063 - in /poolalloc/trunk:
 include/poolalloc/PoolAllocate.h lib/PoolAllocate/Heuristic.cpp
 lib/PoolAllocate/PointerCompress.cpp lib/PoolAllocate/PoolAllocate.cpp
 lib/PoolAllocate/RunTimeAssociate.cpp
 lib/PoolAllocate/TransformFunctionBody.cpp
Message-ID: <20100309165015.4273A2A6C12C@llvm.org>

Author: alenhar2
Date: Tue Mar  9 10:50:15 2010
New Revision: 98063

URL: http://llvm.org/viewvc/llvm-project?rev=98063&view=rev
Log:
Keep poolalloc compiling, but it is broken.
The FIXME for the dsa type changes need to be dealt with
The malloc/free situation needs to be dealt with.


Modified:
    poolalloc/trunk/include/poolalloc/PoolAllocate.h
    poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp
    poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp
    poolalloc/trunk/lib/PoolAllocate/PoolAllocate.cpp
    poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp
    poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp

Modified: poolalloc/trunk/include/poolalloc/PoolAllocate.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/poolalloc/PoolAllocate.h?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/include/poolalloc/PoolAllocate.h (original)
+++ poolalloc/trunk/include/poolalloc/PoolAllocate.h Tue Mar  9 10:50:15 2010
@@ -31,7 +31,6 @@
 #include "dsa/DataStructure.h"
 #include "poolalloc/Config/config.h"
 
-#include <set>
 #include <utility>
 
 namespace llvm {
@@ -59,7 +58,7 @@
     /// MarkedNodes - The set of nodes which are not locally pool allocatable in
     /// the current function.
     ///
-    std::set<const DSNode*> MarkedNodes;
+    DenseSet<const DSNode*> MarkedNodes;
 
     /// F - The function this FuncInfo corresponds to.
     ///

Modified: poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp Tue Mar  9 10:50:15 2010
@@ -57,9 +57,10 @@
 
 unsigned Heuristic::getRecommendedSize(const DSNode *N) {
   unsigned PoolSize = 0;
-  if (!N->isArray() && N->getType()->isSized()) {
-    PoolSize = N->getParentGraph()->getTargetData().getTypeAllocSize(N->getType());
-  }
+  //FIXME: types
+//  if (!N->isArray() && N->getType()->isSized()) {
+//    PoolSize = N->getParentGraph()->getTargetData().getTypeAllocSize(N->getType());
+//  }
   if (PoolSize == 1) PoolSize = 0;
   return PoolSize;
 }
@@ -121,14 +122,15 @@
 /// DSNode.
 ///
 unsigned Heuristic::getRecommendedAlignment(const DSNode *N) {
-  if (!N->getType() || N->getType()->isVoidTy())  // Is this void or collapsed?
+  //FIXME: Type
+  //if (!N->getType() || N->getType()->isVoidTy())  // Is this void or collapsed?
     return 0;  // No known alignment, let runtime decide.
 
-  const TargetData &TD = N->getParentGraph()->getTargetData();
+  //const TargetData &TD = N->getParentGraph()->getTargetData();
 
   // If there are no doubles on an 8-byte boundary in this structure, there is
   // no reason to 8-byte align objects in the pool.
-  return Wants8ByteAlignment(N->getType(), 0, TD) ? 8 : 4;
+  //return Wants8ByteAlignment(N->getType(), 0, TD) ? 8 : 4;
 }
  
 
@@ -231,7 +233,7 @@
     if (F == 0) {
       for (unsigned i = 0, e = NodesToPA.size(); i != e; ++i) {
         const DSNode *Node = NodesToPA[i];
-        if ((Node->isNodeCompletelyFolded() || !Node->isArray()) &&
+        if ((Node->isNodeCompletelyFolded() || !Node->isArrayNode()) &&
             NodeExistsInCycle(Node))
           ResultPools.push_back(OnePool(Node));
       }

Modified: poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp Tue Mar  9 10:50:15 2010
@@ -228,10 +228,13 @@
                                     const TargetData &TD) {
   // First step, compute the type of the compressed node.  This basically
   // replaces all pointers to compressed pools with uints.
-  NewTy = ComputeCompressedType(Pool->getType(), 0, Nodes);
+//FIXME: TYPE
+  //  NewTy = ComputeCompressedType(Pool->getType(), 0, Nodes);
+  NewTy = 0;
 
   // Get the compressed type size.
-  NewSize = NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0;
+  //NewSize = NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0;
+  NewSize = 0;
 }
 
 
@@ -320,14 +323,15 @@
 /// dump - Emit a debugging dump for this pool info.
 ///
 void CompressedPoolInfo::dump() const {
-  const TargetData &TD = getNode()->getParentGraph()->getTargetData();
-  errs() << "  From size: "
-            << (getNode()->getType()->isSized() ? 
-                        TD.getTypeAllocSize(getNode()->getType()) : 0)
-            << "  To size: "
-            << (NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0) << "\n";
-  errs() << "Node: "; getNode()->dump();
-  errs() << "New Type: " << *NewTy << "\n";
+//  const TargetData &TD = getNode()->getParentGraph()->getTargetData();
+//FIXME: type
+  //errs() << "  From size: "
+//            << (getNode()->getType()->isSized() ?
+//                        TD.getTypeAllocSize(getNode()->getType()) : 0)
+//            << "  To size: "
+//            << (NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0) << "\n";
+//  errs() << "Node: "; getNode()->dump();
+//  errs() << "New Type: " << *NewTy << "\n";
 }
 
 
@@ -704,9 +708,10 @@
     const Type* PT =
       cast<PointerType>(GEPI.getOperand(0)->getType())->getElementType();
     if(isa<ArrayType>(PT)) {
-      if (cast<ArrayType>(PT)->getElementType() == PI->getNode()->getType())
-        NTy = PointerType::getUnqual(ArrayType::get(PI->getNewType(),
-                                              cast<ArrayType>(PT)->getNumElements()));
+      //FIXME: TYPE
+//      if (cast<ArrayType>(PT)->getElementType() == PI->getNode()->getType())
+//        NTy = PointerType::getUnqual(ArrayType::get(PI->getNewType(),
+//                                              cast<ArrayType>(PT)->getNumElements()));
     }
   }
 

Modified: poolalloc/trunk/lib/PoolAllocate/PoolAllocate.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PoolAllocate.cpp?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PoolAllocate.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PoolAllocate.cpp Tue Mar  9 10:50:15 2010
@@ -349,13 +349,13 @@
 
 
 static void GetNodesReachableFromGlobals(DSGraph* G,
-                                  std::set<const DSNode*> &NodesFromGlobals) {
+                                  DenseSet<const DSNode*> &NodesFromGlobals) {
   for (DSScalarMap::global_iterator I = G->getScalarMap().global_begin(), 
          E = G->getScalarMap().global_end(); I != E; ++I)
     G->getNodeForValue(*I).getNode()->markReachableNodes(NodesFromGlobals);
 }
 
-static void MarkNodesWhichMustBePassedIn(std::set<const DSNode*> &MarkedNodes,
+static void MarkNodesWhichMustBePassedIn(DenseSet<const DSNode*> &MarkedNodes,
                                          Function &F, DSGraph* G,
                                          bool PassAllArguments) {
   // Mark globals and incomplete nodes as live... (this handles arguments)
@@ -377,16 +377,16 @@
   // Calculate which DSNodes are reachable from globals.  If a node is reachable
   // from a global, we will create a global pool for it, so no argument passage
   // is required.
-  std::set<const DSNode*> NodesFromGlobals;
+  DenseSet<const DSNode*> NodesFromGlobals;
   GetNodesReachableFromGlobals(G, NodesFromGlobals);
 
   // Remove any nodes reachable from a global.  These nodes will be put into
   // global pools, which do not require arguments to be passed in.  Also, erase
   // any marked node that is not a heap node.  Since no allocations or frees
   // will be done with it, it needs no argument.
-  for (std::set<const DSNode*>::iterator I = MarkedNodes.begin(),
+  for (DenseSet<const DSNode*>::iterator I = MarkedNodes.begin(),
          E = MarkedNodes.end(); I != E; ) {
-    const DSNode *N = *I++;
+    const DSNode *N = *I; ++I;
     if ((!(1 || N->isHeapNode()) && !PassAllArguments) || NodesFromGlobals.count(N))
       MarkedNodes.erase(N);
   }
@@ -402,7 +402,7 @@
   // Create a new entry for F.
   FuncInfo &FI =
     FunctionInfo.insert(std::make_pair(&F, FuncInfo(F))).first->second;
-  std::set<const DSNode*> &MarkedNodes = FI.MarkedNodes;
+  DenseSet<const DSNode*> &MarkedNodes = FI.MarkedNodes;
 
   if (G->node_begin() == G->node_end())
     return;  // No memory activity, nothing is required
@@ -411,8 +411,13 @@
   // current function.  This set will contain all of the DSNodes which require
   // pools to be passed in from outside of the function.
   MarkNodesWhichMustBePassedIn(MarkedNodes, F, G, PassAllArguments);
-  
-  FI.ArgNodes.insert(FI.ArgNodes.end(), MarkedNodes.begin(), MarkedNodes.end());
+
+
+  //FI.ArgNodes.insert(FI.ArgNodes.end(), MarkedNodes.begin(), MarkedNodes.end());
+  //Work around DenseSet not having iterator traits
+  for (DenseSet<const DSNode*>::iterator ii = MarkedNodes.begin(),
+       ee = MarkedNodes.end(); ii != ee; ++ii)
+    FI.ArgNodes.insert(FI.ArgNodes.end(), *ii);
 }
 
 // MakeFunctionClone - If the specified function needs to be modified for pool
@@ -527,13 +532,13 @@
   DSGraph* GG = Graphs->getGlobalsGraph();
 
   // Get all of the nodes reachable from globals.
-  std::set<const DSNode*> GlobalHeapNodes;
+  DenseSet<const DSNode*> GlobalHeapNodes;
   GetNodesReachableFromGlobals(GG, GlobalHeapNodes);
 
   // Filter out all nodes which have no heap allocations merged into them.
-  for (std::set<const DSNode*>::iterator I = GlobalHeapNodes.begin(),
+  for (DenseSet<const DSNode*>::iterator I = GlobalHeapNodes.begin(),
          E = GlobalHeapNodes.end(); I != E; ) {
-    std::set<const DSNode*>::iterator Last = I++;
+    DenseSet<const DSNode*>::iterator Last = I; ++I;
 
 #if 0
     //
@@ -547,10 +552,11 @@
     }
 #endif
 
+    //FIXME: erase on a densemap invalidates all iterators
     const DSNode *tmp = *Last;
     //    errs() << "test \n";
-    if (!(tmp->isHeapNode() || tmp->isArray()))
-      GlobalHeapNodes.erase(Last);
+    if (!(tmp->isHeapNode() || tmp->isArrayNode()))
+      GlobalHeapNodes.erase(tmp);
   }
   
   // Otherwise get the main function to insert the poolinit calls.
@@ -565,8 +571,14 @@
             << " global nodes!\n";
 
 
-  std::vector<const DSNode*> NodesToPA(GlobalHeapNodes.begin(),
-                                       GlobalHeapNodes.end());
+  //std::vector<const DSNode*> NodesToPA(GlobalHeapNodes.begin(),
+  //                                     GlobalHeapNodes.end());
+  //DenseSet Doesn't have polite iterators
+  std::vector<const DSNode*> NodesToPA;
+  for (DenseSet<const DSNode*>::iterator ii = GlobalHeapNodes.begin(),
+         ee = GlobalHeapNodes.end(); ii != ee; ++ii)
+    NodesToPA.push_back(*ii);
+
   std::vector<Heuristic::OnePool> ResultPools;
   CurHeuristic->AssignToPools(NodesToPA, 0, GG, ResultPools);
 
@@ -590,7 +602,7 @@
   }
 
   // Any unallocated DSNodes get null pool descriptor pointers.
-  for (std::set<const DSNode*>::iterator I = GlobalHeapNodes.begin(),
+  for (DenseSet<const DSNode*>::iterator I = GlobalHeapNodes.begin(),
          E = GlobalHeapNodes.end(); I != E; ++I) {
     GlobalNodes[*I] = ConstantPointerNull::get(PointerType::getUnqual(PoolDescType));
     ++NumNonprofit;
@@ -704,7 +716,7 @@
   if (G->node_begin() == G->node_end()) return;  // Quick exit if nothing to do.
   
   FuncInfo &FI = *getFuncInfo(F);
-  std::set<const DSNode*> &MarkedNodes = FI.MarkedNodes;
+  DenseSet<const DSNode*> &MarkedNodes = FI.MarkedNodes;
 
   // Calculate which DSNodes are reachable from globals.  If a node is reachable
   // from a global, we will create a global pool for it, so no argument passage
@@ -721,7 +733,7 @@
   for (DSGraph::node_iterator I = G->node_begin(), E = G->node_end(); I != E;++I){
     // We only need to make a pool if there is a heap object in it...
     DSNode *N = I;
-    if ((N->isHeapNode()) || (BoundsChecksEnabled && (N->isArray()))) {
+    if ((N->isHeapNode()) || (BoundsChecksEnabled && (N->isArrayNode()))) {
       if (GlobalsGraphNodeMapping.count(N)) {
         // If it is a global pool, set up the pool descriptor appropriately.
         DSNode *GGN = GlobalsGraphNodeMapping[N].getNode();

Modified: poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp Tue Mar  9 10:50:15 2010
@@ -56,13 +56,13 @@
 ////////////////////////////////////////////////////////////////////////////////
 
 static void GetNodesReachableFromGlobals(DSGraph* G,
-                                         std::set<const DSNode*> &NodesFromGlobals) {
+                                         DenseSet<const DSNode*> &NodesFromGlobals) {
   for (DSScalarMap::global_iterator I = G->getScalarMap().global_begin(),
           E = G->getScalarMap().global_end(); I != E; ++I)
     G->getNodeForValue(*I).getNode()->markReachableNodes(NodesFromGlobals);
 }
 
-static void MarkNodesWhichMustBePassedIn(std::set<const DSNode*> &MarkedNodes,
+static void MarkNodesWhichMustBePassedIn(DenseSet<const DSNode*> &MarkedNodes,
                                          Function &F, DSGraph* G,
                                          EntryPointAnalysis* EPA) {
   // All DSNodes reachable from arguments must be passed in...
@@ -84,13 +84,13 @@
   // Calculate which DSNodes are reachable from globals.  If a node is reachable
   // from a global, we will create a global pool for it, so no argument passage
   // is required.
-  std::set<const DSNode*> NodesFromGlobals;
+  DenseSet<const DSNode*> NodesFromGlobals;
   GetNodesReachableFromGlobals(G, NodesFromGlobals);
 
   // Remove any nodes reachable from a global.  These nodes will be put into
   // global pools, which do not require arguments to be passed in.
 
-  for (std::set<const DSNode*>::iterator I = NodesFromGlobals.begin(),
+  for (DenseSet<const DSNode*>::iterator I = NodesFromGlobals.begin(),
           E = NodesFromGlobals.end(); I != E; ++I)
     MarkedNodes.erase(*I);
 }
@@ -101,7 +101,7 @@
 /// map and recording this info in the ArgNodes set.
 static void FindFunctionPoolArgs(Function &F, FuncInfo& FI,
                                  EntryPointAnalysis* EPA) {
-  std::set<const DSNode*> MarkedNodes;
+  DenseSet<const DSNode*> MarkedNodes;
 
   if (FI.G->node_begin() == FI.G->node_end())
     return; // No memory activity, nothing is required
@@ -111,13 +111,15 @@
   // pools to be passed in from outside of the function.
   MarkNodesWhichMustBePassedIn(MarkedNodes, F, FI.G,EPA);
 
-  FI.ArgNodes.insert(FI.ArgNodes.end(), MarkedNodes.begin(), MarkedNodes.end());
+  //FI.ArgNodes.insert(FI.ArgNodes.end(), MarkedNodes.begin(), MarkedNodes.end());
+  //Work around DenseSet not having iterator traits
+  for (DenseSet<const DSNode*>::iterator ii = MarkedNodes.begin(),
+       ee = MarkedNodes.end(); ii != ee; ++ii)
+    FI.ArgNodes.insert(FI.ArgNodes.end(), *ii);
 }
 
 
 
-
-
 ////////////////////////////////////////////////////////////////////////////////
 // RTAssociate
 ////////////////////////////////////////////////////////////////////////////////
@@ -286,7 +288,7 @@
   // DSGraph* GG = Graphs->getGlobalsGraph();
 
   // Get all of the nodes reachable from globals.
-  std::set<const DSNode*> GlobalHeapNodes;
+  DenseSet<const DSNode*> GlobalHeapNodes;
   GetNodesReachableFromGlobals(GG, GlobalHeapNodes);
 
   errs() << "Pool allocating " << GlobalHeapNodes.size()

Modified: poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp?rev=98063&r1=98062&r2=98063&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp Tue Mar  9 10:50:15 2010
@@ -271,7 +271,7 @@
 
   // Get the pool handle for the node that this contributes to...
   DSNode *Node = getDSNodeHFor(&MI).getNode();
-  if (Node->isArray()) {
+  if (Node->isArrayNode()) {
     Value *PH = getPoolHandle(&MI);
     if (PH == 0 || isa<ConstantPointerNull>(PH)) return;
     TargetData &TD = PAInfo.getAnalysis<TargetData>();
@@ -751,7 +751,7 @@
         if (FI.PoolDescriptors.count(LocalNode))
           ArgVal = FI.PoolDescriptors.find(LocalNode)->second;
     if (isa<Constant>(ArgVal) && cast<Constant>(ArgVal)->isNullValue()) {
-      if ((!(PAInfo.BoundsChecksEnabled)) || (ArgNodes[i]->isArray())) {
+      if ((!(PAInfo.BoundsChecksEnabled)) || (ArgNodes[i]->isArrayNode())) {
         if (!isa<InvokeInst>(TheCall)) {
           // Dinakar: We need pooldescriptors for allocas in the callee if it
           //          escapes




From criswell at uiuc.edu  Tue Mar  9 11:01:39 2010
From: criswell at uiuc.edu (John Criswell)
Date: Tue, 09 Mar 2010 17:01:39 -0000
Subject: [llvm-commits] [poolalloc] r98064 - in
 /poolalloc/branches/release_26/lib: AssistDS/Makefile DSA/Makefile
 Macroscopic/Makefile PoolAllocate/Makefile
Message-ID: <20100309170139.911A92A6C12C@llvm.org>

Author: criswell
Date: Tue Mar  9 11:01:39 2010
New Revision: 98064

URL: http://llvm.org/viewvc/llvm-project?rev=98064&view=rev
Log:
Do not build shared libraries on Mingw or Cygwin.  These platforms do not
currently support LLVM shared libraries.

Modified:
    poolalloc/branches/release_26/lib/AssistDS/Makefile
    poolalloc/branches/release_26/lib/DSA/Makefile
    poolalloc/branches/release_26/lib/Macroscopic/Makefile
    poolalloc/branches/release_26/lib/PoolAllocate/Makefile

Modified: poolalloc/branches/release_26/lib/AssistDS/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/lib/AssistDS/Makefile?rev=98064&r1=98063&r2=98064&view=diff
==============================================================================
--- poolalloc/branches/release_26/lib/AssistDS/Makefile (original)
+++ poolalloc/branches/release_26/lib/AssistDS/Makefile Tue Mar  9 11:01:39 2010
@@ -8,7 +8,11 @@
 ##===----------------------------------------------------------------------===##
 
 LEVEL = ../..
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
 SHARED_LIBRARY=1
+endif
+endif
 LIBRARYNAME = AssistDS
 
 include $(LEVEL)/Makefile.common

Modified: poolalloc/branches/release_26/lib/DSA/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/lib/DSA/Makefile?rev=98064&r1=98063&r2=98064&view=diff
==============================================================================
--- poolalloc/branches/release_26/lib/DSA/Makefile (original)
+++ poolalloc/branches/release_26/lib/DSA/Makefile Tue Mar  9 11:01:39 2010
@@ -10,8 +10,12 @@
 LEVEL = ../..
 LIBRARYNAME = LLVMDataStructure
 BUILD_ARCHIVE := 1
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
 SHARED_LIBRARY := 1
 #LOADABLE_MODULE := 1
+endif
+endif
 
 include $(LEVEL)/Makefile.common
 

Modified: poolalloc/branches/release_26/lib/Macroscopic/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/lib/Macroscopic/Makefile?rev=98064&r1=98063&r2=98064&view=diff
==============================================================================
--- poolalloc/branches/release_26/lib/Macroscopic/Makefile (original)
+++ poolalloc/branches/release_26/lib/Macroscopic/Makefile Tue Mar  9 11:01:39 2010
@@ -2,8 +2,12 @@
 LEVEL=../..
 
 # Give the name of a library.  This will build a dynamic version.
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
 SHARED_LIBRARY=1
 LOADABLE_MODULE = 1
+endif
+endif
 DONT_BUILD_RELINKED=1
 LIBRARYNAME=macroscopic
 

Modified: poolalloc/branches/release_26/lib/PoolAllocate/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/lib/PoolAllocate/Makefile?rev=98064&r1=98063&r2=98064&view=diff
==============================================================================
--- poolalloc/branches/release_26/lib/PoolAllocate/Makefile (original)
+++ poolalloc/branches/release_26/lib/PoolAllocate/Makefile Tue Mar  9 11:01:39 2010
@@ -8,8 +8,12 @@
 #
 LIBRARYNAME=poolalloc
 BUILD_ARCHIVE := 1
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
 SHARED_LIBRARY := 1
 #LOADABLE_MODULE := 1
+endif
+endif
 
 #
 # Include Makefile.common so we know what to do.




From criswell at uiuc.edu  Tue Mar  9 11:05:54 2010
From: criswell at uiuc.edu (John Criswell)
Date: Tue, 09 Mar 2010 17:05:54 -0000
Subject: [llvm-commits] [poolalloc] r98065 - in
 /poolalloc/branches/release_26/runtime: FL2Allocator/Makefile
 FreeListAllocator/Makefile HeapFrag/Makefile PoolAllocator/Makefile
 PreRT/Makefile
Message-ID: <20100309170554.E5E312A6C12C@llvm.org>

Author: criswell
Date: Tue Mar  9 11:05:54 2010
New Revision: 98065

URL: http://llvm.org/viewvc/llvm-project?rev=98065&view=rev
Log:
Moved the library name to be the first thing listed after the LEVEL variable.
Disabled shared library support on Mingw and Cygwin since it isn't supported
at the moment.

Modified:
    poolalloc/branches/release_26/runtime/FL2Allocator/Makefile
    poolalloc/branches/release_26/runtime/FreeListAllocator/Makefile
    poolalloc/branches/release_26/runtime/HeapFrag/Makefile
    poolalloc/branches/release_26/runtime/PoolAllocator/Makefile
    poolalloc/branches/release_26/runtime/PreRT/Makefile

Modified: poolalloc/branches/release_26/runtime/FL2Allocator/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/runtime/FL2Allocator/Makefile?rev=98065&r1=98064&r2=98065&view=diff
==============================================================================
--- poolalloc/branches/release_26/runtime/FL2Allocator/Makefile (original)
+++ poolalloc/branches/release_26/runtime/FL2Allocator/Makefile Tue Mar  9 11:05:54 2010
@@ -1,7 +1,16 @@
 LEVEL = ../..
-SHARED_LIBRARY=1
 LIBRARYNAME=poolalloc_rt
 
+#
+# Build shared libraries on all platforms except Cygwin and MingW (which do
+# not support them).
+#
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
+SHARED_LIBRARY=1
+endif
+endif
+
 ifdef ENABLE_OPTIMIZED
 CXXFLAGS += -DNDEBUG=1
 endif

Modified: poolalloc/branches/release_26/runtime/FreeListAllocator/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/runtime/FreeListAllocator/Makefile?rev=98065&r1=98064&r2=98065&view=diff
==============================================================================
--- poolalloc/branches/release_26/runtime/FreeListAllocator/Makefile (original)
+++ poolalloc/branches/release_26/runtime/FreeListAllocator/Makefile Tue Mar  9 11:05:54 2010
@@ -1,5 +1,14 @@
 LEVEL = ../..
-SHARED_LIBRARY=1
 LIBRARYNAME=poolalloc_fl_rt
 
+#
+# Build shared libraries on all platforms except Cygwin and MingW (which do
+# not support them).
+#
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
+SHARED_LIBRARY=1
+endif
+endif
+
 include $(LEVEL)/Makefile.common

Modified: poolalloc/branches/release_26/runtime/HeapFrag/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/runtime/HeapFrag/Makefile?rev=98065&r1=98064&r2=98065&view=diff
==============================================================================
--- poolalloc/branches/release_26/runtime/HeapFrag/Makefile (original)
+++ poolalloc/branches/release_26/runtime/HeapFrag/Makefile Tue Mar  9 11:05:54 2010
@@ -1,7 +1,17 @@
 LEVEL = ../..
+LIBRARYNAME=heapfrag
+
 #BYTECODE_LIBRARY=1
+
+#
+# Build shared libraries on all platforms except Cygwin and MingW (which do
+# not support them).
+#
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
 #SHARED_LIBRARY=1
-LIBRARYNAME=heapfrag
+endif
+endif
 
 include $(LEVEL)/Makefile.common
 

Modified: poolalloc/branches/release_26/runtime/PoolAllocator/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/runtime/PoolAllocator/Makefile?rev=98065&r1=98064&r2=98065&view=diff
==============================================================================
--- poolalloc/branches/release_26/runtime/PoolAllocator/Makefile (original)
+++ poolalloc/branches/release_26/runtime/PoolAllocator/Makefile Tue Mar  9 11:05:54 2010
@@ -1,7 +1,16 @@
 LEVEL = ../..
-SHARED_LIBRARY=1
 LIBRARYNAME=poolalloc_rt_old
 
+#
+# Build shared libraries on all platforms except Cygwin and MingW (which do
+# not support them).
+#
+ifneq ($(OS),Cygwin)
+ifneq ($(OS),MingW)
+SHARED_LIBRARY=1
+endif
+endif
+
 include $(LEVEL)/Makefile.common
 
 # Always build optimized and debug versions

Modified: poolalloc/branches/release_26/runtime/PreRT/Makefile
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/release_26/runtime/PreRT/Makefile?rev=98065&r1=98064&r2=98065&view=diff
==============================================================================
--- poolalloc/branches/release_26/runtime/PreRT/Makefile (original)
+++ poolalloc/branches/release_26/runtime/PreRT/Makefile Tue Mar  9 11:05:54 2010
@@ -1,6 +1,6 @@
 LEVEL = ../..
-BYTECODE_LIBRARY=1
 LIBRARYNAME=pa_pre_rt
+BYTECODE_LIBRARY=1
 
 include $(LEVEL)/Makefile.common
 




From stoklund at 2pi.dk  Tue Mar  9 11:41:10 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 9 Mar 2010 09:41:10 -0800
Subject: [llvm-commits] [llvm] r98023 -
	/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
In-Reply-To: <59424131-A98B-428A-9BE7-75786CE199C4@2pi.dk>
References: <20100309005948.483A02A6C12C@llvm.org>
	<F579DF8B-D3A1-43B1-9E7B-A66EE315EA23@apple.com>
	<59424131-A98B-428A-9BE7-75786CE199C4@2pi.dk>
Message-ID: <B1330B22-1E2D-4AD4-9E4F-57543A722BD7@2pi.dk>


On Mar 8, 2010, at 5:37 PM, Jakob Stoklund Olesen wrote:

> 
> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
> 
>> Is 1000 scientific? :-)
> 
> No. That would require measuring when the LiveInterval size where it outgrows cache. I didn't do that :-(
> 
> I have also seen cases where a virtual register LiveInterval becomes too big. That is a lot harder to fix. I think the real problem is the extra indirection in SlotIndex. If SlotIndex were just an unsigned in disguise, each LiveRange would be just 16 bytes, and all ranges for a register would be in contiguous memory. 

OK, I did the science now ;-)

I was wrong.

I disabled the inliner and compiled this function with varying array lengths:

#include <utility>
typedef std::pair<int,int> P;
P f(int n) {
    static const P t[] = {
        P(84, 89),
        P(39, 44),
        P(79, 84),
        P(80, 85),
        P(92, 97),
        P(19, 24),
        P(33, 38),
        P(77, 82),
        P(28, 33),
        P(55, 60),
    };
    return t[n];
}

It produces lots of constructor calls like this:

	leaq	__ZZ1fiE1t+184(%rip), %rdi
	leaq	-196(%rbp), %rsi
	leaq	-200(%rbp), %rdx
	callq	__ZNSt4pairIiiEC1ERKiS2_

The coalescer timing formed a perfect parabola with t=4.1s at N=2000. This is not a caching effect - the coalescer is quadratic in the number of function calls!





From stoklund at 2pi.dk  Tue Mar  9 12:15:18 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 9 Mar 2010 10:15:18 -0800
Subject: [llvm-commits] [llvm] r98023 -
	/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
In-Reply-To: <B1330B22-1E2D-4AD4-9E4F-57543A722BD7@2pi.dk>
References: <20100309005948.483A02A6C12C@llvm.org>
	<F579DF8B-D3A1-43B1-9E7B-A66EE315EA23@apple.com>
	<59424131-A98B-428A-9BE7-75786CE199C4@2pi.dk>
	<B1330B22-1E2D-4AD4-9E4F-57543A722BD7@2pi.dk>
Message-ID: <E6E3C024-0BF7-4159-8723-3DF880F720F7@2pi.dk>


On Mar 9, 2010, at 9:41 AM, Jakob Stoklund Olesen wrote:

> 
> On Mar 8, 2010, at 5:37 PM, Jakob Stoklund Olesen wrote:
> 
>> 
>> On Mar 8, 2010, at 5:23 PM, Evan Cheng wrote:
>> 
>>> Is 1000 scientific? :-)
>> 
> 
> The coalescer timing formed a perfect parabola with t=4.1s at N=2000. This is not a caching effect - the coalescer is quadratic in the number of function calls!

The attached graph shows the coalescer runtime as a function of the array length in my tiny test function (inlining disabled). The patch kicks in at N=500 and reduces runtime to nothing.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: graph.svg
Type: image/svg+xml
Size: 8481 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100309/a29a1d4b/attachment.bin 

From dalej at apple.com  Tue Mar  9 12:19:45 2010
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 9 Mar 2010 10:19:45 -0800
Subject: [llvm-commits] [llvm-gcc-4.2] r97989	-
	/llvm-gcc-4.2/trunk/gcc/llvm-types.cpp
In-Reply-To: <4B95F374.4070505@free.fr>
References: <20100308213302.618102A6C12C@llvm.org> <4B95F374.4070505@free.fr>
Message-ID: <AC57E01F-6595-4090-BC80-B6130191E85D@apple.com>


On Mar 8, 2010, at 11:06 PMPST, Duncan Sands wrote:

> Hi Dan,
> 
>>    // If this function returns via a shadow argument, the dest loc is passed
>> -  // in as a pointer.  Mark that pointer as struct-ret and noalias.
>> +  // in as a pointer.  Mark that pointer as struct-ret.
>> +  //
>> +  // It's tempting to want NoAlias here too, however even though llvm-gcc
>> +  // itself currently always passes a dedicated alloca as the actual argument,
>> +  // this isn't mandated by the ABI. There are other compilers which don't
>> +  // always pass a dedicated alloca. Using NoAlias here would make code which
>> +  // isn't interoperable with that of other compilers.
>>    if (ABIConverter.isShadowReturn())
>>      Attrs.push_back(AttributeWithIndex::get(ArgTys.size(),
>> -                                    Attribute::StructRet | Attribute::NoAlias));
>> +                                    Attribute::StructRet));
> 
> what does the ABI say exactly?  Is it really valid to pass something which may
> get "accidentally" altered when the callee (for example) writes to a global or
> to some other argument?


The ABI does NOT say that the shadow return value must be a  dedicated stack-allocated object, and therefore it doesn't have to be if we see the same result under the as-if rule.  I also have seen other compilers do this (not just clang).  I believe it's safe in some circumstances, but you have to be careful.




From isanbard at gmail.com  Tue Mar  9 12:31:07 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 18:31:07 -0000
Subject: [llvm-commits] [llvm] r98068 - in /llvm/trunk/lib/Target/ARM:
 ARMISelLowering.cpp ARMTargetObjectFile.cpp ARMTargetObjectFile.h
 CMakeLists.txt
Message-ID: <20100309183107.671BC2A6C12C@llvm.org>

Author: void
Date: Tue Mar  9 12:31:07 2010
New Revision: 98068

URL: http://llvm.org/viewvc/llvm-project?rev=98068&view=rev
Log:
The ARM EH experiment worked!

Place the LSDA into the TEXT section for ARM platforms. This involves making the
encoding indirect, pcrel, and sdata4 instead of an absolute pointer. The
references to the type infos are then non-lazy pointers. Revision 98019 changed
the encoding of non-lazy pointers to add the symbol to the non-lazy pointer
definition if it's a local symbol (otherwise, it's external and set to '0' so
that the loader can adjust it to the real value). This paved the way for this
change to work on ARM.

Added:
    llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp
Modified:
    llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
    llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h
    llvm/trunk/lib/Target/ARM/CMakeLists.txt

Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=98068&r1=98067&r2=98068&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Tue Mar  9 12:31:07 2010
@@ -40,17 +40,11 @@
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/Target/TargetOptions.h"
 #include "llvm/ADT/VectorExtras.h"
-#include "llvm/Support/CommandLine.h"
-#include "llvm/Support/Dwarf.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 #include <sstream>
 using namespace llvm;
-using namespace dwarf;
-
-static cl::opt<bool> EnableARMEHTest("enable-arm-eh-test", cl::Hidden,
-    cl::desc("Enable ARM EH Test"));
 
 static bool CC_ARM_APCS_Custom_f64(unsigned &ValNo, EVT &ValVT, EVT &LocVT,
                                    CCValAssign::LocInfo &LocInfo,
@@ -134,32 +128,9 @@
   addTypeForNEON(VT, MVT::v2f64, MVT::v4i32);
 }
 
-namespace llvm {
-
-  // FIXME: This is a test of <rdar://problem/6804645>.
-  class ARMMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
-  public:
-    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
-      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
-
-      // Exception Handling.
-      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
-                                    SectionKind::getReadOnlyWithRel());
-    }
-
-    virtual unsigned getTTypeEncoding() const {
-      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
-    }
-  };
-
-}
-
 static TargetLoweringObjectFile *createTLOF(TargetMachine &TM) {
   if (TM.getSubtarget<ARMSubtarget>().isTargetDarwin())
-    if (EnableARMEHTest)
-      return new ARMMachOTargetObjectFile();
-    else
-      return new TargetLoweringObjectFileMachO();
+    return new ARMMachOTargetObjectFile();
 
   return new ARMElfTargetObjectFile();
 }

Added: llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp?rev=98068&view=auto
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp (added)
+++ llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp Tue Mar  9 12:31:07 2010
@@ -0,0 +1,54 @@
+//===-- llvm/Target/ARMTargetObjectFile.cpp - ARM Object Info Impl --------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "ARMTargetObjectFile.h"
+#include "ARMSubtarget.h"
+#include "llvm/MC/MCSectionELF.h"
+#include "llvm/MC/MCSectionMachO.h"
+#include "llvm/Support/Dwarf.h"
+#include "llvm/Target/TargetMachine.h"
+using namespace llvm;
+using namespace dwarf;
+
+//===----------------------------------------------------------------------===//
+//                               ELF Target
+//===----------------------------------------------------------------------===//
+
+void ARMElfTargetObjectFile::Initialize(MCContext &Ctx,
+                                        const TargetMachine &TM) {
+  TargetLoweringObjectFileELF::Initialize(Ctx, TM);
+
+  if (TM.getSubtarget<ARMSubtarget>().isAAPCS_ABI()) {
+    StaticCtorSection =
+      getELFSection(".init_array", MCSectionELF::SHT_INIT_ARRAY,
+                    MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
+                    SectionKind::getDataRel());
+    StaticDtorSection =
+      getELFSection(".fini_array", MCSectionELF::SHT_FINI_ARRAY,
+                    MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
+                    SectionKind::getDataRel());
+  }
+}
+
+//===----------------------------------------------------------------------===//
+//                              Mach-O Target
+//===----------------------------------------------------------------------===//
+
+void ARMMachOTargetObjectFile::Initialize(MCContext &Ctx,
+                                          const TargetMachine &TM) {
+  TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
+
+  // Exception Handling.
+  LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
+                                SectionKind::getReadOnlyWithRel());
+}
+
+unsigned ARMMachOTargetObjectFile::getTTypeEncoding() const {
+  return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
+}

Modified: llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h?rev=98068&r1=98067&r2=98068&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h (original)
+++ llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h Tue Mar  9 12:31:07 2010
@@ -11,29 +11,31 @@
 #define LLVM_TARGET_ARM_TARGETOBJECTFILE_H
 
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
-#include "llvm/MC/MCSectionELF.h"
 
 namespace llvm {
 
-  class ARMElfTargetObjectFile : public TargetLoweringObjectFileELF {
-  public:
-    ARMElfTargetObjectFile() : TargetLoweringObjectFileELF() {}
-
-    void Initialize(MCContext &Ctx, const TargetMachine &TM) {
-      TargetLoweringObjectFileELF::Initialize(Ctx, TM);
-
-      if (TM.getSubtarget<ARMSubtarget>().isAAPCS_ABI()) {
-        StaticCtorSection =
-          getELFSection(".init_array", MCSectionELF::SHT_INIT_ARRAY,
-                        MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
-                        SectionKind::getDataRel());
-        StaticDtorSection =
-          getELFSection(".fini_array", MCSectionELF::SHT_FINI_ARRAY,
-                        MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
-                        SectionKind::getDataRel());
-      }
-    }
-  };
+class MCContext;
+class TargetMachine;
+
+class ARMElfTargetObjectFile : public TargetLoweringObjectFileELF {
+public:
+  ARMElfTargetObjectFile() : TargetLoweringObjectFileELF() {}
+
+  virtual void Initialize(MCContext &Ctx, const TargetMachine &TM);
+};
+
+// FIXME: This subclass isn't 100% necessary. It will become obsolete once we
+//        can place all LSDAs into the TEXT section. See
+//        <rdar://problem/6804645>.
+class ARMMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
+public:
+  ARMMachOTargetObjectFile() : TargetLoweringObjectFileMachO() {}
+
+  virtual void Initialize(MCContext &Ctx, const TargetMachine &TM);
+
+  virtual unsigned getTTypeEncoding() const;
+};
+
 } // end namespace llvm
 
 #endif

Modified: llvm/trunk/lib/Target/ARM/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/CMakeLists.txt?rev=98068&r1=98067&r2=98068&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/CMakeLists.txt (original)
+++ llvm/trunk/lib/Target/ARM/CMakeLists.txt Tue Mar  9 12:31:07 2010
@@ -27,6 +27,7 @@
   ARMRegisterInfo.cpp
   ARMSubtarget.cpp
   ARMTargetMachine.cpp
+  ARMTargetObjectFile.cpp
   NEONMoveFix.cpp
   NEONPreAllocPass.cpp
   Thumb1InstrInfo.cpp




From isanbard at gmail.com  Tue Mar  9 12:31:40 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 09 Mar 2010 18:31:40 -0000
Subject: [llvm-commits] [test-suite] r98069 -
	/test-suite/trunk/Makefile.programs
Message-ID: <20100309183140.468652A6C12C@llvm.org>

Author: void
Date: Tue Mar  9 12:31:40 2010
New Revision: 98069

URL: http://llvm.org/viewvc/llvm-project?rev=98069&view=rev
Log:
Experiment over. Revert flag.

Modified:
    test-suite/trunk/Makefile.programs

Modified: test-suite/trunk/Makefile.programs
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/Makefile.programs?rev=98069&r1=98068&r2=98069&view=diff
==============================================================================
--- test-suite/trunk/Makefile.programs (original)
+++ test-suite/trunk/Makefile.programs Tue Mar  9 12:31:40 2010
@@ -241,8 +241,7 @@
 LLCBETAOPTION := -enable-sparc-v9-insts
 endif
 ifeq ($(ARCH),ARM)
-LLCBETAOPTION := -enable-arm-eh-test
-# -enable-machine-cse
+LLCBETAOPTION := -enable-machine-cse
 endif
 ifeq ($(ARCH),THUMB)
 LLCBETAOPTION := -enable-machine-cse




From jyasskin at google.com  Tue Mar  9 12:54:07 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Tue, 9 Mar 2010 10:54:07 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <20100308220250.5FC1F2A6C12E@llvm.org>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
Message-ID: <bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>

Now that nothing uses the result of constructCompileUnit() unless it
assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
avoid new'ing the DIE and CompileUnit? Then, if we delete ModuleCU in
endModule(), this memory leak would go away.

On Mon, Mar 8, 2010 at 2:02 PM, Devang Patel <dpatel at apple.com> wrote:
> Author: dpatel
> Date: Mon Mar ?8 16:02:50 2010
> New Revision: 97990
>
> URL: http://llvm.org/viewvc/llvm-project?rev=97990&view=rev
> Log:
> Derive DIType from DIScope. This simplifies getContext() where for members the context is a type. This also eliminates need of CompileUnitMaps maintained by dwarf writer.
>
> Modified:
> ? ?llvm/trunk/include/llvm/Analysis/DebugInfo.h
> ? ?llvm/trunk/lib/Analysis/DebugInfo.cpp
> ? ?llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
> ? ?llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
>
> Modified: llvm/trunk/include/llvm/Analysis/DebugInfo.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/DebugInfo.h?rev=97990&r1=97989&r2=97990&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/DebugInfo.h (original)
> +++ llvm/trunk/include/llvm/Analysis/DebugInfo.h Mon Mar ?8 16:02:50 2010
> @@ -172,7 +172,7 @@
> ? /// DIType - This is a wrapper for a type.
> ? /// FIXME: Types should be factored much better so that CV qualifiers and
> ? /// others do not require a huge and empty descriptor full of zeros.
> - ?class DIType : public DIDescriptor {
> + ?class DIType : public DIScope {
> ? public:
> ? ? enum {
> ? ? ? FlagPrivate ? ? ? ? ?= 1 << 0,
> @@ -188,7 +188,7 @@
> ? protected:
> ? ? // This ctor is used when the Tag has already been validated by a derived
> ? ? // ctor.
> - ? ?DIType(MDNode *N, bool, bool) : DIDescriptor(N) {}
> + ? ?DIType(MDNode *N, bool, bool) : DIScope(N) {}
>
> ? public:
>
> @@ -199,7 +199,7 @@
> ? ? explicit DIType() {}
> ? ? virtual ~DIType() {}
>
> - ? ?DIDescriptor getContext() const ? ? { return getDescriptorField(1); }
> + ? ?DIScope getContext() const ? ? ? ? ?{ return getFieldAs<DIScope>(1); }
> ? ? StringRef getName() const ? ? ? ? ? { return getStringField(2); ? ? }
> ? ? DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
> ? ? unsigned getLineNumber() const ? ? ?{ return getUnsignedField(4); }
> @@ -234,6 +234,8 @@
> ? ? bool isValid() const {
> ? ? ? return DbgNode && (isBasicType() || isDerivedType() || isCompositeType());
> ? ? }
> + ? ?StringRef getFilename() const ? ?{ return getCompileUnit().getFilename();}
> + ? ?StringRef getDirectory() const ? { return getCompileUnit().getDirectory();}
> ? ? /// dump - print type.
> ? ? void dump() const;
> ? };
> @@ -305,7 +307,7 @@
> ? public:
> ? ? virtual ~DIGlobal() {}
>
> - ? ?DIDescriptor getContext() const ? ? { return getDescriptorField(2); }
> + ? ?DIScope getContext() const ? ? ? ? ?{ return getFieldAs<DIScope>(2); }
> ? ? StringRef getName() const ? ? ? ? { return getStringField(3); }
> ? ? StringRef getDisplayName() const ?{ return getStringField(4); }
> ? ? StringRef getLinkageName() const ?{ return getStringField(5); }
> @@ -327,7 +329,7 @@
> ? public:
> ? ? explicit DISubprogram(MDNode *N = 0) : DIScope(N) {}
>
> - ? ?DIDescriptor getContext() const ? ? { return getDescriptorField(2); }
> + ? ?DIScope getContext() const ? ? ? ? ?{ return getFieldAs<DIScope>(2); }
> ? ? StringRef getName() const ? ? ? ? { return getStringField(3); }
> ? ? StringRef getDisplayName() const ?{ return getStringField(4); }
> ? ? StringRef getLinkageName() const ?{ return getStringField(5); }
> @@ -396,8 +398,8 @@
> ? ? explicit DIVariable(MDNode *N = 0)
> ? ? ? : DIDescriptor(N) {}
>
> - ? ?DIDescriptor getContext() const { return getDescriptorField(1); }
> - ? ?StringRef getName() const ? ? { return getStringField(2); ? ? }
> + ? ?DIScope getContext() const ? ? ? ? ?{ return getFieldAs<DIScope>(1); }
> + ? ?StringRef getName() const ? ? ? ? ? { return getStringField(2); ? ? }
> ? ? DICompileUnit getCompileUnit() const{ return getFieldAs<DICompileUnit>(3); }
> ? ? unsigned getLineNumber() const ? ? ?{ return getUnsignedField(4); }
> ? ? DIType getType() const ? ? ? ? ? ? ?{ return getFieldAs<DIType>(5); }
>
> Modified: llvm/trunk/lib/Analysis/DebugInfo.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/DebugInfo.cpp?rev=97990&r1=97989&r2=97990&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Analysis/DebugInfo.cpp (original)
> +++ llvm/trunk/lib/Analysis/DebugInfo.cpp Mon Mar ?8 16:02:50 2010
> @@ -243,7 +243,7 @@
> ?// Simple Descriptor Constructors and other Methods
> ?//===----------------------------------------------------------------------===//
>
> -DIType::DIType(MDNode *N) : DIDescriptor(N) {
> +DIType::DIType(MDNode *N) : DIScope(N) {
> ? if (!N) return;
> ? if (!isBasicType() && !isDerivedType() && !isCompositeType()) {
> ? ? DbgNode = 0;
> @@ -412,6 +412,8 @@
> ?}
>
> ?StringRef DIScope::getFilename() const {
> + ?if (!DbgNode)
> + ? ?return StringRef();
> ? if (isLexicalBlock())
> ? ? return DILexicalBlock(DbgNode).getFilename();
> ? if (isSubprogram())
> @@ -420,11 +422,15 @@
> ? ? return DICompileUnit(DbgNode).getFilename();
> ? if (isNameSpace())
> ? ? return DINameSpace(DbgNode).getFilename();
> + ?if (isType())
> + ? ?return DIType(DbgNode).getFilename();
> ? assert(0 && "Invalid DIScope!");
> ? return StringRef();
> ?}
>
> ?StringRef DIScope::getDirectory() const {
> + ?if (!DbgNode)
> + ? ?return StringRef();
> ? if (isLexicalBlock())
> ? ? return DILexicalBlock(DbgNode).getDirectory();
> ? if (isSubprogram())
> @@ -433,6 +439,8 @@
> ? ? return DICompileUnit(DbgNode).getDirectory();
> ? if (isNameSpace())
> ? ? return DINameSpace(DbgNode).getDirectory();
> + ?if (isType())
> + ? ?return DIType(DbgNode).getDirectory();
> ? assert(0 && "Invalid DIScope!");
> ? return StringRef();
> ?}
>
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=97990&r1=97989&r2=97990&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Mar ?8 16:02:50 2010
> @@ -417,7 +417,8 @@
> ? ? return;
>
> ? unsigned Line = V->getLineNumber();
> - ?unsigned FileID = findCompileUnit(V->getCompileUnit())->getID();
> + ?unsigned FileID = GetOrCreateSourceID(V->getContext().getDirectory(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?V->getContext().getFilename());
> ? assert(FileID && "Invalid file id");
> ? addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
> ? addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
> @@ -431,7 +432,8 @@
> ? ? return;
>
> ? unsigned Line = G->getLineNumber();
> - ?unsigned FileID = findCompileUnit(G->getCompileUnit())->getID();
> + ?unsigned FileID = GetOrCreateSourceID(G->getContext().getDirectory(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?G->getContext().getFilename());
> ? assert(FileID && "Invalid file id");
> ? addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
> ? addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
> @@ -447,9 +449,11 @@
> ? if (SP->getLineNumber() == 0)
> ? ? return;
>
> -
> ? unsigned Line = SP->getLineNumber();
> - ?unsigned FileID = findCompileUnit(SP->getCompileUnit())->getID();
> + ?if (!SP->getContext().Verify())
> + ? ?return;
> + ?unsigned FileID = GetOrCreateSourceID(SP->getContext().getDirectory(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SP->getContext().getFilename());
> ? assert(FileID && "Invalid file id");
> ? addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
> ? addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
> @@ -464,7 +468,10 @@
> ? ? return;
>
> ? unsigned Line = Ty->getLineNumber();
> - ?unsigned FileID = findCompileUnit(CU)->getID();
> + ?if (!Ty->getContext().Verify())
> + ? ?return;
> + ?unsigned FileID = GetOrCreateSourceID(Ty->getContext().getDirectory(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Ty->getContext().getFilename());
> ? assert(FileID && "Invalid file id");
> ? addUInt(Die, dwarf::DW_AT_decl_file, 0, FileID);
> ? addUInt(Die, dwarf::DW_AT_decl_line, 0, Line);
> @@ -1233,16 +1240,6 @@
> ? return SPDie;
> ?}
>
> -/// findCompileUnit - Get the compile unit for the given descriptor.
> -///
> -CompileUnit *DwarfDebug::findCompileUnit(DICompileUnit Unit) {
> - ?DenseMap<Value *, CompileUnit *>::const_iterator I =
> - ? ?CompileUnitMap.find(Unit.getNode());
> - ?if (I == CompileUnitMap.end())
> - ? ?return constructCompileUnit(Unit.getNode());
> - ?return I->second;
> -}
> -
> ?/// getUpdatedDbgScope - Find or create DbgScope assicated with the instruction.
> ?/// Initialize scope and update scope hierarchy.
> ?DbgScope *DwarfDebug::getUpdatedDbgScope(MDNode *N, const MachineInstr *MI,
> @@ -1676,8 +1673,6 @@
> ? ? ModuleCU = Unit;
> ? }
>
> - ?CompileUnitMap[DIUnit.getNode()] = Unit;
> - ?CompileUnits.push_back(Unit);
> ? return Unit;
> ?}
>
> @@ -1783,17 +1778,8 @@
> ? ? ? ? ?E = DbgFinder.compile_unit_end(); I != E; ++I)
> ? ? constructCompileUnit(*I);
>
> - ?if (CompileUnits.empty()) {
> - ? ?if (TimePassesIsEnabled)
> - ? ? ?DebugTimer->stopTimer();
> -
> - ? ?return;
> - ?}
> -
> - ?// If main compile unit for this module is not seen than randomly
> - ?// select first compile unit.
> ? if (!ModuleCU)
> - ? ?ModuleCU = CompileUnits[0];
> + ? ?return;
>
> ? // Create DIEs for each subprogram.
> ? for (DebugInfoFinder::iterator I = DbgFinder.subprogram_begin(),
>
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=97990&r1=97989&r2=97990&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon Mar ?8 16:02:50 2010
> @@ -62,14 +62,6 @@
> ? // Attributes used to construct specific Dwarf sections.
> ? //
>
> - ?/// CompileUnitMap - A map of global variables representing compile units to
> - ?/// compile units.
> - ?DenseMap<Value *, CompileUnit *> CompileUnitMap;
> -
> - ?/// CompileUnits - All the compile units in this module.
> - ?///
> - ?SmallVector<CompileUnit *, 8> CompileUnits;
> -
> ? /// ModuleCU - All DIEs are inserted in ModuleCU.
> ? CompileUnit *ModuleCU;
>
> @@ -357,10 +349,6 @@
> ? /// createSubprogramDIE - Create new DIE using SP.
> ? DIE *createSubprogramDIE(const DISubprogram &SP, bool MakeDecl = false);
>
> - ?/// findCompileUnit - Get the compile unit for the given descriptor.
> - ?///
> - ?CompileUnit *findCompileUnit(DICompileUnit Unit);
> -
> ? /// getUpdatedDbgScope - Find or create DbgScope assicated with
> ? /// the instruction. Initialize scope and update scope hierarchy.
> ? DbgScope *getUpdatedDbgScope(MDNode *N, const MachineInstr *MI, MDNode *InlinedAt);
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From grosbach at apple.com  Tue Mar  9 13:07:29 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 09 Mar 2010 19:07:29 -0000
Subject: [llvm-commits] [llvm] r98071 - in /llvm/trunk/lib/Target/ARM:
 ARMBaseRegisterInfo.cpp Thumb1RegisterInfo.cpp
Message-ID: <20100309190729.28CE92A6C12C@llvm.org>

Author: grosbach
Date: Tue Mar  9 13:07:28 2010
New Revision: 98071

URL: http://llvm.org/viewvc/llvm-project?rev=98071&view=rev
Log:
scavenged frame index value re-use gets confused when more than one base
register is involved for thumb1. Work around this for the moment by only
re-using SP-relative offsets. This is temporary 'til the code can distinguish
multiple base registers.

Modified:
    llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
    llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp?rev=98071&r1=98070&r2=98071&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp Tue Mar  9 13:07:28 2010
@@ -40,7 +40,7 @@
 #include "llvm/Support/CommandLine.h"
 using namespace llvm;
 
-static cl::opt<bool>
+cl::opt<bool>
 ReuseFrameIndexVals("arm-reuse-frame-index-vals", cl::Hidden, cl::init(true),
           cl::desc("Reuse repeated frame index values"));
 

Modified: llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp?rev=98071&r1=98070&r2=98071&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp Tue Mar  9 13:07:28 2010
@@ -33,10 +33,13 @@
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/ADT/BitVector.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/Support/CommandLine.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;
 
+extern cl::opt<bool> ReuseFrameIndexVals;
+
 Thumb1RegisterInfo::Thumb1RegisterInfo(const ARMBaseInstrInfo &tii,
                                        const ARMSubtarget &sti)
   : ARMBaseRegisterInfo(tii, sti) {
@@ -640,6 +643,7 @@
       assert (Value && "Frame index virtual allocated, but Value arg is NULL!");
       *Value = Offset;
       bool UseRR = false;
+      bool TrackVReg = FrameReg == ARM::SP;
 
       if (Opcode == ARM::tSpill) {
         if (FrameReg == ARM::SP)
@@ -648,6 +652,7 @@
         else {
           emitLoadConstPool(MBB, II, dl, VReg, 0, Offset);
           UseRR = true;
+          TrackVReg = false;
         }
       } else
         emitThumbRegPlusImmediate(MBB, II, VReg, FrameReg, Offset, TII,
@@ -658,6 +663,8 @@
         MI.addOperand(MachineOperand::CreateReg(FrameReg, false));
       else // tSTR has an extra register operand.
         MI.addOperand(MachineOperand::CreateReg(0, false));
+      if (!ReuseFrameIndexVals || !TrackVReg)
+        VReg = 0;
   } else
     assert(false && "Unexpected opcode!");
 




From grosbach at apple.com  Tue Mar  9 13:10:55 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 9 Mar 2010 11:10:55 -0800
Subject: [llvm-commits] [llvm] r98068 - in /llvm/trunk/lib/Target/ARM:
	ARMISelLowering.cpp ARMTargetObjectFile.cpp
	ARMTargetObjectFile.h CMakeLists.txt
In-Reply-To: <20100309183107.671BC2A6C12C@llvm.org>
References: <20100309183107.671BC2A6C12C@llvm.org>
Message-ID: <5D3A7E42-5190-4588-82BD-D0BECEBCB5A7@apple.com>

Very nice! Thanks, Bill.

-jim

On Mar 9, 2010, at 10:31 AM, Bill Wendling wrote:

> Author: void
> Date: Tue Mar  9 12:31:07 2010
> New Revision: 98068
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98068&view=rev
> Log:
> The ARM EH experiment worked!
> 
> Place the LSDA into the TEXT section for ARM platforms. This involves making the
> encoding indirect, pcrel, and sdata4 instead of an absolute pointer. The
> references to the type infos are then non-lazy pointers. Revision 98019 changed
> the encoding of non-lazy pointers to add the symbol to the non-lazy pointer
> definition if it's a local symbol (otherwise, it's external and set to '0' so
> that the loader can adjust it to the real value). This paved the way for this
> change to work on ARM.
> 
> Added:
>    llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp
> Modified:
>    llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
>    llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h
>    llvm/trunk/lib/Target/ARM/CMakeLists.txt
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=98068&r1=98067&r2=98068&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Tue Mar  9 12:31:07 2010
> @@ -40,17 +40,11 @@
> #include "llvm/MC/MCSectionMachO.h"
> #include "llvm/Target/TargetOptions.h"
> #include "llvm/ADT/VectorExtras.h"
> -#include "llvm/Support/CommandLine.h"
> -#include "llvm/Support/Dwarf.h"
> #include "llvm/Support/ErrorHandling.h"
> #include "llvm/Support/MathExtras.h"
> #include "llvm/Support/raw_ostream.h"
> #include <sstream>
> using namespace llvm;
> -using namespace dwarf;
> -
> -static cl::opt<bool> EnableARMEHTest("enable-arm-eh-test", cl::Hidden,
> -    cl::desc("Enable ARM EH Test"));
> 
> static bool CC_ARM_APCS_Custom_f64(unsigned &ValNo, EVT &ValVT, EVT &LocVT,
>                                    CCValAssign::LocInfo &LocInfo,
> @@ -134,32 +128,9 @@
>   addTypeForNEON(VT, MVT::v2f64, MVT::v4i32);
> }
> 
> -namespace llvm {
> -
> -  // FIXME: This is a test of <rdar://problem/6804645>.
> -  class ARMMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
> -  public:
> -    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
> -      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
> -
> -      // Exception Handling.
> -      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
> -                                    SectionKind::getReadOnlyWithRel());
> -    }
> -
> -    virtual unsigned getTTypeEncoding() const {
> -      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
> -    }
> -  };
> -
> -}
> -
> static TargetLoweringObjectFile *createTLOF(TargetMachine &TM) {
>   if (TM.getSubtarget<ARMSubtarget>().isTargetDarwin())
> -    if (EnableARMEHTest)
> -      return new ARMMachOTargetObjectFile();
> -    else
> -      return new TargetLoweringObjectFileMachO();
> +    return new ARMMachOTargetObjectFile();
> 
>   return new ARMElfTargetObjectFile();
> }
> 
> Added: llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp?rev=98068&view=auto
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp (added)
> +++ llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.cpp Tue Mar  9 12:31:07 2010
> @@ -0,0 +1,54 @@
> +//===-- llvm/Target/ARMTargetObjectFile.cpp - ARM Object Info Impl --------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#include "ARMTargetObjectFile.h"
> +#include "ARMSubtarget.h"
> +#include "llvm/MC/MCSectionELF.h"
> +#include "llvm/MC/MCSectionMachO.h"
> +#include "llvm/Support/Dwarf.h"
> +#include "llvm/Target/TargetMachine.h"
> +using namespace llvm;
> +using namespace dwarf;
> +
> +//===----------------------------------------------------------------------===//
> +//                               ELF Target
> +//===----------------------------------------------------------------------===//
> +
> +void ARMElfTargetObjectFile::Initialize(MCContext &Ctx,
> +                                        const TargetMachine &TM) {
> +  TargetLoweringObjectFileELF::Initialize(Ctx, TM);
> +
> +  if (TM.getSubtarget<ARMSubtarget>().isAAPCS_ABI()) {
> +    StaticCtorSection =
> +      getELFSection(".init_array", MCSectionELF::SHT_INIT_ARRAY,
> +                    MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
> +                    SectionKind::getDataRel());
> +    StaticDtorSection =
> +      getELFSection(".fini_array", MCSectionELF::SHT_FINI_ARRAY,
> +                    MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
> +                    SectionKind::getDataRel());
> +  }
> +}
> +
> +//===----------------------------------------------------------------------===//
> +//                              Mach-O Target
> +//===----------------------------------------------------------------------===//
> +
> +void ARMMachOTargetObjectFile::Initialize(MCContext &Ctx,
> +                                          const TargetMachine &TM) {
> +  TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
> +
> +  // Exception Handling.
> +  LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
> +                                SectionKind::getReadOnlyWithRel());
> +}
> +
> +unsigned ARMMachOTargetObjectFile::getTTypeEncoding() const {
> +  return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
> +}
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h?rev=98068&r1=98067&r2=98068&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h (original)
> +++ llvm/trunk/lib/Target/ARM/ARMTargetObjectFile.h Tue Mar  9 12:31:07 2010
> @@ -11,29 +11,31 @@
> #define LLVM_TARGET_ARM_TARGETOBJECTFILE_H
> 
> #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
> -#include "llvm/MC/MCSectionELF.h"
> 
> namespace llvm {
> 
> -  class ARMElfTargetObjectFile : public TargetLoweringObjectFileELF {
> -  public:
> -    ARMElfTargetObjectFile() : TargetLoweringObjectFileELF() {}
> -
> -    void Initialize(MCContext &Ctx, const TargetMachine &TM) {
> -      TargetLoweringObjectFileELF::Initialize(Ctx, TM);
> -
> -      if (TM.getSubtarget<ARMSubtarget>().isAAPCS_ABI()) {
> -        StaticCtorSection =
> -          getELFSection(".init_array", MCSectionELF::SHT_INIT_ARRAY,
> -                        MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
> -                        SectionKind::getDataRel());
> -        StaticDtorSection =
> -          getELFSection(".fini_array", MCSectionELF::SHT_FINI_ARRAY,
> -                        MCSectionELF::SHF_WRITE | MCSectionELF::SHF_ALLOC,
> -                        SectionKind::getDataRel());
> -      }
> -    }
> -  };
> +class MCContext;
> +class TargetMachine;
> +
> +class ARMElfTargetObjectFile : public TargetLoweringObjectFileELF {
> +public:
> +  ARMElfTargetObjectFile() : TargetLoweringObjectFileELF() {}
> +
> +  virtual void Initialize(MCContext &Ctx, const TargetMachine &TM);
> +};
> +
> +// FIXME: This subclass isn't 100% necessary. It will become obsolete once we
> +//        can place all LSDAs into the TEXT section. See
> +//        <rdar://problem/6804645>.
> +class ARMMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
> +public:
> +  ARMMachOTargetObjectFile() : TargetLoweringObjectFileMachO() {}
> +
> +  virtual void Initialize(MCContext &Ctx, const TargetMachine &TM);
> +
> +  virtual unsigned getTTypeEncoding() const;
> +};
> +
> } // end namespace llvm
> 
> #endif
> 
> Modified: llvm/trunk/lib/Target/ARM/CMakeLists.txt
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/CMakeLists.txt?rev=98068&r1=98067&r2=98068&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/CMakeLists.txt (original)
> +++ llvm/trunk/lib/Target/ARM/CMakeLists.txt Tue Mar  9 12:31:07 2010
> @@ -27,6 +27,7 @@
>   ARMRegisterInfo.cpp
>   ARMSubtarget.cpp
>   ARMTargetMachine.cpp
> +  ARMTargetObjectFile.cpp
>   NEONMoveFix.cpp
>   NEONPreAllocPass.cpp
>   Thumb1InstrInfo.cpp
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From chandlerc at gmail.com  Tue Mar  9 13:24:49 2010
From: chandlerc at gmail.com (Chandler Carruth)
Date: Tue, 09 Mar 2010 19:24:49 -0000
Subject: [llvm-commits] [llvm] r98074 - in /llvm/trunk:
 unittests/Makefile.unittest utils/unittest/googletest/Makefile
Message-ID: <20100309192449.C8DD22A6C12C@llvm.org>

Author: chandlerc
Date: Tue Mar  9 13:24:49 2010
New Revision: 98074

URL: http://llvm.org/viewvc/llvm-project?rev=98074&view=rev
Log:
Consolidate GoogleTest make options and duplicate them to its own makefile.

Modified:
    llvm/trunk/unittests/Makefile.unittest
    llvm/trunk/utils/unittest/googletest/Makefile

Modified: llvm/trunk/unittests/Makefile.unittest
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile.unittest?rev=98074&r1=98073&r2=98074&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile.unittest (original)
+++ llvm/trunk/unittests/Makefile.unittest Tue Mar  9 13:24:49 2010
@@ -14,18 +14,20 @@
 # Set up variables for building a unit test.
 ifdef TESTNAME
 
-CPP.Flags += -DGTEST_HAS_RTTI=0
-# gcc's TR1 <tuple> header depends on RTTI, so force googletest to use
-# its own tuple implementation.  When we import googletest >=1.4.0, we
-# can drop this line.
-CPP.Flags += -DGTEST_HAS_TR1_TUPLE=0
-
 include $(LEVEL)/Makefile.common
 
 LLVMUnitTestExe = $(BuildMode)/$(TESTNAME)Tests$(EXEEXT)
 
-CPP.Flags += -I$(LLVM_SRC_ROOT)/utils/unittest/googletest/include/
-CPP.Flags += $(NO_VARIADIC_MACROS)
+# Note that these flags are duplicated when building GoogleTest itself in
+# utils/unittest/googletest/Makefile; ensure that any changes are made to both.
+CPP.Flags += -I$(LLVM_SRC_ROOT)/utils/unittest/googletest/include
+CPP.Flags += $(NO_MISSING_FIELD_INITIALIZERS) $(NO_VARIADIC_MACROS)
+CPP.Flags += -DGTEST_HAS_RTTI=0
+# libstdc++'s TR1 <tuple> header depends on RTTI and uses C++'0x features not
+# supported by Clang, so force googletest to use its own tuple implementation.
+# When we import googletest >=1.4.0, we can drop this line.
+CPP.Flags += -DGTEST_HAS_TR1_TUPLE=0
+
 TESTLIBS = -lGoogleTest -lUnitTestMain
 
 ifeq ($(ENABLE_SHARED), 1)

Modified: llvm/trunk/utils/unittest/googletest/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/unittest/googletest/Makefile?rev=98074&r1=98073&r2=98074&view=diff
==============================================================================
--- llvm/trunk/utils/unittest/googletest/Makefile (original)
+++ llvm/trunk/utils/unittest/googletest/Makefile Tue Mar  9 13:24:49 2010
@@ -14,8 +14,17 @@
 LIBRARYNAME = GoogleTest
 BUILD_ARCHIVE = 1
 REQUIRES_RTTI = 1
+
+# Note that these flags are duplicated when building individual tests in
+# unittests/Makefile.unittest; ensure that any changes are made to both.
 CPP.Flags += -I$(LLVM_SRC_ROOT)/utils/unittest/googletest/include
 CPP.Flags += $(NO_MISSING_FIELD_INITIALIZERS) $(NO_VARIADIC_MACROS)
+CPP.Flags += -DGTEST_HAS_RTTI=0
+# libstdc++'s TR1 <tuple> header depends on RTTI and uses C++'0x features not
+# supported by Clang, so force googletest to use its own tuple implementation.
+# When we import googletest >=1.4.0, we can drop this line.
+CPP.Flags += -DGTEST_HAS_TR1_TUPLE=0
+
 
 ifeq ($(HOST_OS),MingW)
   CPP.Flags += -DGTEST_OS_WINDOWS=1




From baldrick at free.fr  Tue Mar  9 13:40:17 2010
From: baldrick at free.fr (Duncan Sands)
Date: Tue, 09 Mar 2010 20:40:17 +0100
Subject: [llvm-commits] [llvm-gcc-4.2] r97989	-
	/llvm-gcc-4.2/trunk/gcc/llvm-types.cpp
In-Reply-To: <AC57E01F-6595-4090-BC80-B6130191E85D@apple.com>
References: <20100308213302.618102A6C12C@llvm.org> <4B95F374.4070505@free.fr>
	<AC57E01F-6595-4090-BC80-B6130191E85D@apple.com>
Message-ID: <4B96A421.5070909@free.fr>

Hi Dale,

>>>     // If this function returns via a shadow argument, the dest loc is passed
>>> -  // in as a pointer.  Mark that pointer as struct-ret and noalias.
>>> +  // in as a pointer.  Mark that pointer as struct-ret.
>>> +  //
>>> +  // It's tempting to want NoAlias here too, however even though llvm-gcc
>>> +  // itself currently always passes a dedicated alloca as the actual argument,
>>> +  // this isn't mandated by the ABI. There are other compilers which don't
>>> +  // always pass a dedicated alloca. Using NoAlias here would make code which
>>> +  // isn't interoperable with that of other compilers.
>>>     if (ABIConverter.isShadowReturn())
>>>       Attrs.push_back(AttributeWithIndex::get(ArgTys.size(),
>>> -                                    Attribute::StructRet | Attribute::NoAlias));
>>> +                                    Attribute::StructRet));
>>
>> what does the ABI say exactly?  Is it really valid to pass something which may
>> get "accidentally" altered when the callee (for example) writes to a global or
>> to some other argument?
>
>
> The ABI does NOT say that the shadow return value must be a  dedicated stack-allocated object, and therefore it doesn't have to be if we see the same result under the as-if rule.  I also have seen other compilers do this (not just clang).  I believe it's safe in some circumstances, but you have to be careful.

if the callee can't tell that it's not being passed a dedicated stack-object,
then noalias should be fine.

Ciao,

Duncan.


From dalej at apple.com  Tue Mar  9 14:15:43 2010
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 09 Mar 2010 20:15:43 -0000
Subject: [llvm-commits] [llvm] r98077 - in /llvm/trunk:
 lib/Target/PowerPC/PPCISelLowering.cpp
 test/CodeGen/PowerPC/2010-03-09-indirect-call.ll
Message-ID: <20100309201543.2755A2A6C12C@llvm.org>

Author: johannes
Date: Tue Mar  9 14:15:42 2010
New Revision: 98077

URL: http://llvm.org/viewvc/llvm-project?rev=98077&view=rev
Log:
The address of an indirect call must be in R12 on Darwin.
Make it so.  (This patch is in LowerCall_Darwin, which seems
to be used by SVR4 code as well; since that doesn't belong here,
I haven't worried about this case.)


Added:
    llvm/trunk/test/CodeGen/PowerPC/2010-03-09-indirect-call.ll
Modified:
    llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp

Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=98077&r1=98076&r2=98077&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Tue Mar  9 14:15:42 2010
@@ -3258,6 +3258,16 @@
                          false, false, 0);
   }
 
+  // On Darwin, R12 must contain the address of an indirect callee.  This does
+  // not mean the MTCTR instruction must use R12; it's easier to model this as
+  // an extra parameter, so do that.
+  if (!isTailCall && 
+      !dyn_cast<GlobalAddressSDNode>(Callee) &&
+      !dyn_cast<ExternalSymbolSDNode>(Callee) &&
+      !isBLACompatibleAddress(Callee, DAG))
+    RegsToPass.push_back(std::make_pair((unsigned)(isPPC64 ? PPC::X12 :
+                                                   PPC::R12), Callee));
+
   // Build a sequence of copy-to-reg nodes chained together with token chain
   // and flag operands which copy the outgoing args into the appropriate regs.
   SDValue InFlag;

Added: llvm/trunk/test/CodeGen/PowerPC/2010-03-09-indirect-call.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/2010-03-09-indirect-call.ll?rev=98077&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/2010-03-09-indirect-call.ll (added)
+++ llvm/trunk/test/CodeGen/PowerPC/2010-03-09-indirect-call.ll Tue Mar  9 14:15:42 2010
@@ -0,0 +1,19 @@
+; RUN: llc < %s -march=ppc32 -mcpu=g5 -mtriple=powerpc-apple-darwin10.0 | FileCheck %s
+; ModuleID = 'nn.c'
+target datalayout = "E-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f128:64:128"
+target triple = "powerpc-apple-darwin11.0"
+; Indirect calls must use R12 on Darwin (i.e., R12 must contain the address of
+; the function being called; the mtctr is not required to use it).
+
+ at p = external global void (...)*                  ; <void (...)**> [#uses=1]
+
+define void @foo() nounwind ssp {
+entry:
+; CHECK: mtctr r12
+  %0 = load void (...)** @p, align 4              ; <void (...)*> [#uses=1]
+  call void (...)* %0() nounwind
+  br label %return
+
+return:                                           ; preds = %entry
+  ret void
+}




From dpatel at apple.com  Tue Mar  9 14:22:31 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 9 Mar 2010 12:22:31 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
Message-ID: <2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>


On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:

> Now that nothing uses the result of constructCompileUnit() unless it
> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
> avoid new'ing the DIE and CompileUnit?

With the introduction of DIFile, we are moving away from "isMain" field.  

> Then, if we delete ModuleCU in
> endModule(), this memory leak would go away.

We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.
Thanks,
-
Devang





From daniel at zuster.org  Tue Mar  9 15:27:31 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 21:27:31 -0000
Subject: [llvm-commits] [llvm] r98080 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100309212731.10C362A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 15:27:30 2010
New Revision: 98080

URL: http://llvm.org/viewvc/llvm-project?rev=98080&view=rev
Log:
MC/Mach-O: Don't generate relocations for PCrel fixups to local labels.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98080&r1=98079&r2=98080&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Tue Mar  9 15:27:30 2010
@@ -440,7 +440,7 @@
                              DenseMap<const MCSymbol*,MCSymbolData*> &SymbolMap,
                                      std::vector<MachRelocationEntry> &Relocs) {
     uint32_t Address = Fragment.getOffset() + Fixup.Offset;
-    unsigned IsPCRel = 0;
+    unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
     unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
     unsigned Type = RIT_Vanilla;
 
@@ -468,10 +468,16 @@
 
     // The value which goes in the fixup is current value of the expression.
     Fixup.FixedValue = Value - Value2 + Target.getConstant();
-    if (isFixupKindPCRel(Fixup.Kind)) {
+    if (IsPCRel)
       Fixup.FixedValue -= Address;
-      IsPCRel = 1;
-    }
+
+    // If this fixup is a vanilla PC relative relocation for a local label, we
+    // don't need a relocation.
+    //
+    // FIXME: Implement proper atom support.
+    if (IsPCRel && Target.getSymA() && Target.getSymA()->isTemporary() &&
+        !Target.getSymB())
+      return;
 
     MachRelocationEntry MRE;
     MRE.Word0 = ((Address   <<  0) |
@@ -516,7 +522,7 @@
     uint32_t Address = Fragment.getOffset() + Fixup.Offset;
     uint32_t Value = 0;
     unsigned Index = 0;
-    unsigned IsPCRel = 0;
+    unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
     unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
     unsigned IsExtern = 0;
     unsigned Type = 0;
@@ -554,11 +560,15 @@
 
     // The value which goes in the fixup is current value of the expression.
     Fixup.FixedValue = Value + Target.getConstant();
-
-    if (isFixupKindPCRel(Fixup.Kind)) {
+    if (IsPCRel)
       Fixup.FixedValue -= Address;
-      IsPCRel = 1;
-    }
+
+    // If this fixup is a vanilla PC relative relocation for a local label, we
+    // don't need a relocation.
+    //
+    // FIXME: Implement proper atom support.
+    if (IsPCRel && Target.getSymA() && Target.getSymA()->isTemporary())
+      return;
 
     // struct relocation_info (8 bytes)
     MachRelocationEntry MRE;




From daniel at zuster.org  Tue Mar  9 15:27:47 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 21:27:47 -0000
Subject: [llvm-commits] [llvm] r98081 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100309212747.F366A2A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 15:27:47 2010
New Revision: 98081

URL: http://llvm.org/viewvc/llvm-project?rev=98081&view=rev
Log:
MC/Mach-O: Also set the PCrel bit in the second half of paired relocation entries.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98081&r1=98080&r2=98081&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Tue Mar  9 15:27:47 2010
@@ -489,13 +489,11 @@
     Relocs.push_back(MRE);
 
     if (Type == RIT_LocalDifference) {
-      Type = RIT_Pair;
-
       MachRelocationEntry MRE;
       MRE.Word0 = ((0         <<  0) |
-                   (Type      << 24) |
+                   (RIT_Pair  << 24) |
                    (Log2Size  << 28) |
-                   (0   << 30) |
+                   (IsPCRel   << 30) |
                    RF_Scattered);
       MRE.Word1 = Value2;
       Relocs.push_back(MRE);




From daniel at zuster.org  Tue Mar  9 15:27:58 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 21:27:58 -0000
Subject: [llvm-commits] [llvm] r98082 - in /llvm/trunk:
 lib/MC/MCAssembler.cpp test/MC/MachO/reloc-pcrel.s
Message-ID: <20100309212758.BF0FB2A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 15:27:58 2010
New Revision: 98082

URL: http://llvm.org/viewvc/llvm-project?rev=98082&view=rev
Log:
MC/Mach-O: For PCrel relocations, we need to compensate for the PCrel adjustment when determining if we need a scattered relocation.

Added:
    llvm/trunk/test/MC/MachO/reloc-pcrel.s
Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98082&r1=98081&r2=98082&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Tue Mar  9 15:27:58 2010
@@ -504,15 +504,21 @@
                              MCAsmFixup &Fixup,
                              DenseMap<const MCSymbol*,MCSymbolData*> &SymbolMap,
                              std::vector<MachRelocationEntry> &Relocs) {
+    unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
+    unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
+
     MCValue Target;
     if (!Fixup.Value->EvaluateAsRelocatable(Target))
       llvm_report_error("expected relocatable expression");
 
-    // If this is a difference or a local symbol plus an offset, then we need a
-    // scattered relocation entry.
+    // If this is a difference or a defined symbol plus an offset, then we need
+    // a scattered relocation entry.
+    uint32_t Offset = Target.getConstant();
+    if (IsPCRel)
+      Offset += 1 << Log2Size;
     if (Target.getSymB() ||
         (Target.getSymA() && !Target.getSymA()->isUndefined() &&
-         Target.getConstant()))
+         Offset))
       return ComputeScatteredRelocationInfo(Asm, Fragment, Fixup, Target,
                                             SymbolMap, Relocs);
 
@@ -520,8 +526,6 @@
     uint32_t Address = Fragment.getOffset() + Fixup.Offset;
     uint32_t Value = 0;
     unsigned Index = 0;
-    unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
-    unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
     unsigned IsExtern = 0;
     unsigned Type = 0;
 

Added: llvm/trunk/test/MC/MachO/reloc-pcrel.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/reloc-pcrel.s?rev=98082&view=auto
==============================================================================
--- llvm/trunk/test/MC/MachO/reloc-pcrel.s (added)
+++ llvm/trunk/test/MC/MachO/reloc-pcrel.s Tue Mar  9 15:27:58 2010
@@ -0,0 +1,62 @@
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump | FileCheck %s
+
+// CHECK: # Relocation 0
+// CHECK: (('word-0', 0xe4000045),
+// CHECK:  ('word-1', 0x4)),
+// CHECK: # Relocation 1
+// CHECK: (('word-0', 0xe1000000),
+// CHECK:  ('word-1', 0x6)),
+// CHECK: # Relocation 2
+// CHECK: (('word-0', 0x40),
+// CHECK:  ('word-1', 0xd000002)),
+// CHECK: # Relocation 3
+// CHECK: (('word-0', 0x3b),
+// CHECK:  ('word-1', 0xd000002)),
+// CHECK: # Relocation 4
+// CHECK: (('word-0', 0x36),
+// CHECK:  ('word-1', 0xd000002)),
+// CHECK: # Relocation 5
+// CHECK: (('word-0', 0xe0000031),
+// CHECK:  ('word-1', 0x4)),
+// CHECK: # Relocation 6
+// CHECK: (('word-0', 0xe000002c),
+// CHECK:  ('word-1', 0x4)),
+// CHECK: # Relocation 7
+// CHECK: (('word-0', 0x27),
+// CHECK:  ('word-1', 0x5000001)),
+// CHECK: # Relocation 8
+// CHECK: (('word-0', 0xe0000022),
+// CHECK:  ('word-1', 0x2)),
+// CHECK: # Relocation 9
+// CHECK: (('word-0', 0xe000001d),
+// CHECK:  ('word-1', 0x2)),
+// CHECK: # Relocation 10
+// CHECK: (('word-0', 0x18),
+// CHECK:  ('word-1', 0x5000001)),
+// CHECK-NEXT: ])
+
+        xorl %eax,%eax
+        
+        .globl _a
+_a:
+        xorl %eax,%eax
+_b:
+        xorl %eax,%eax
+L0:
+        xorl %eax,%eax
+L1:     
+
+        call L0
+        call L0 - 1
+        call L0 + 1
+        call _a
+        call _a - 1
+        call _a + 1
+        call _b
+        call _b - 1
+        call _b + 1
+        call _c
+        call _c - 1
+        call _c + 1
+//        call _a - L0
+        call _b - L0




From johnny.chen at apple.com  Tue Mar  9 15:39:34 2010
From: johnny.chen at apple.com (Johnny Chen)
Date: Tue, 09 Mar 2010 21:39:34 -0000
Subject: [llvm-commits] [llvm] r98085 - in /llvm/trunk/lib/Target/ARM:
 ARMInstrInfo.td ARMInstrThumb2.td
Message-ID: <20100309213934.9CEAD2A6C12C@llvm.org>

Author: johnny
Date: Tue Mar  9 15:39:34 2010
New Revision: 98085

URL: http://llvm.org/viewvc/llvm-project?rev=98085&view=rev
Log:
MSR (Move to Special Register from ARM core register) requires a mask to specify
what fields of the CPSR or SPSR are affected.

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=98085&r1=98084&r2=98085&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Tue Mar  9 15:39:34 2010
@@ -2844,29 +2844,29 @@
   let Inst{7-4} = 0b0000;
 }
 
-// FIXME: mask is ignored for the time being.
-def MSR : ABI<0b0001,(outs),(ins GPR:$src), NoItinerary, "msr", "\tcpsr, $src",
+def MSR : ABI<0b0001, (outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
+              "\tcpsr${mask:msr}, $src",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0010;
   let Inst{7-4} = 0b0000;
 }
 
-// FIXME: mask is ignored for the time being.
-def MSRi : ABI<0b0011,(outs),(ins so_imm:$a), NoItinerary, "msr", "\tcpsr, $a",
+def MSRi : ABI<0b0011, (outs), (ins so_imm:$a, i32imm:$mask), NoItinerary,"msr",
+              "\tcpsr${mask:msr}, $a",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0010;
   let Inst{7-4} = 0b0000;
 }
 
-// FIXME: mask is ignored for the time being.
-def MSRsys : ABI<0b0001,(outs),(ins GPR:$src),NoItinerary,"msr","\tspsr, $src",
+def MSRsys : ABI<0b0001, (outs), (ins GPR:$src, i32imm:$mask),NoItinerary,"msr",
+              "\tspsr${mask:msr}, $src",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0110;
   let Inst{7-4} = 0b0000;
 }
 
-// FIXME: mask is ignored for the time being.
-def MSRsysi : ABI<0b0011,(outs),(ins so_imm:$a),NoItinerary,"msr","\tspsr, $a",
+def MSRsysi : ABI<0b0011,(outs),(ins so_imm:$a, i32imm:$mask),NoItinerary,"msr",
+              "\tspsr${mask:msr}, $a",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0110;
   let Inst{7-4} = 0b0000;

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=98085&r1=98084&r2=98085&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Tue Mar  9 15:39:34 2010
@@ -2637,9 +2637,9 @@
   let Inst{12} = 0;
 }
 
-// FIXME: mask is ignored for the time being.
 // Rn = Inst{19-16}
-def t2MSR : T2I<(outs), (ins GPR:$src), NoItinerary, "msr", "\tcpsr, $src",
+def t2MSR : T2I<(outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
+                "\tcpsr${mask:msr}, $src",
                 [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2649,9 +2649,9 @@
   let Inst{12} = 0;
 }
 
-// FIXME: mask is ignored for the time being.
 // Rn = Inst{19-16}
-def t2MSRsys : T2I<(outs), (ins GPR:$src), NoItinerary, "msr", "\tspsr, $src",
+def t2MSRsys : T2I<(outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
+                   "\tspsr${mask:msr}, $src",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;




From grosbach at apple.com  Tue Mar  9 15:45:49 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 09 Mar 2010 21:45:49 -0000
Subject: [llvm-commits] [llvm] r98086 - in /llvm/trunk: include/llvm/Target/
 lib/CodeGen/ lib/Target/ARM/ lib/Target/Alpha/ lib/Target/Blackfin/
 lib/Target/CellSPU/ lib/Target/MBlaze/ lib/Target/MSP430/ lib/Target/Mips/
 lib/Target/PIC16/ lib/Target/PowerPC/ lib/Target/Sparc/ lib/Target/SystemZ/
 lib/Target/X86/ lib/Target/XCore/
Message-ID: <20100309214549.B2B612A6C12C@llvm.org>

Author: grosbach
Date: Tue Mar  9 15:45:49 2010
New Revision: 98086

URL: http://llvm.org/viewvc/llvm-project?rev=98086&view=rev
Log:
Change the Value argument to eliminateFrameIndex to a type-tagged value. This
is preparatory to having PEI's scavenged frame index value reuse logic
properly distinguish types of frame values (e.g., whether the value is
stack-pointer relative or frame-pointer relative).

No functionality change.

Modified:
    llvm/trunk/include/llvm/Target/TargetRegisterInfo.h
    llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
    llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
    llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.h
    llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
    llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.h
    llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.cpp
    llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.h
    llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.cpp
    llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.h
    llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.cpp
    llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.h
    llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.cpp
    llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.h
    llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.cpp
    llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.h
    llvm/trunk/lib/Target/Mips/MipsRegisterInfo.cpp
    llvm/trunk/lib/Target/Mips/MipsRegisterInfo.h
    llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.cpp
    llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.h
    llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp
    llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.h
    llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.cpp
    llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.h
    llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.cpp
    llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.h
    llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp
    llvm/trunk/lib/Target/X86/X86RegisterInfo.h
    llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.cpp
    llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.h

Modified: llvm/trunk/include/llvm/Target/TargetRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetRegisterInfo.h (original)
+++ llvm/trunk/include/llvm/Target/TargetRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -682,8 +682,9 @@
   /// When -enable-frame-index-scavenging is enabled, the virtual register
   /// allocated for this frame index is returned and its value is stored in
   /// *Value.
+  typedef std::pair<unsigned, int> FrameIndexValue;
   virtual unsigned eliminateFrameIndex(MachineBasicBlock::iterator MI,
-                                       int SPAdj, int *Value = NULL,
+                                       int SPAdj, FrameIndexValue *Value = NULL,
                                        RegScavenger *RS=NULL) const = 0;
 
   /// emitProlog/emitEpilog - These methods insert prolog and epilog code into

Modified: llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp (original)
+++ llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp Tue Mar  9 15:45:49 2010
@@ -685,7 +685,7 @@
           // If this instruction has a FrameIndex operand, we need to
           // use that target machine register info object to eliminate
           // it.
-          int Value;
+          TargetRegisterInfo::FrameIndexValue Value;
           unsigned VReg =
             TRI.eliminateFrameIndex(MI, SPAdj, &Value,
                                     FrameIndexVirtualScavenging ?  NULL : RS);
@@ -693,7 +693,8 @@
             assert (FrameIndexVirtualScavenging &&
                     "Not scavenging, but virtual returned from "
                     "eliminateFrameIndex()!");
-            FrameConstantRegMap[VReg] = FrameConstantEntry(Value, SPAdj);
+            FrameConstantRegMap[VReg] = FrameConstantEntry(Value.second,
+                                                           SPAdj);
           }
 
           // Reset the iterator if we were at the beginning of the BB.

Modified: llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -1153,7 +1153,7 @@
 
 unsigned
 ARMBaseRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                         int SPAdj, int *Value,
+                                         int SPAdj, FrameIndexValue *Value,
                                          RegScavenger *RS) const {
   unsigned i = 0;
   MachineInstr &MI = *II;
@@ -1205,7 +1205,10 @@
     MI.getOperand(i).ChangeToRegister(FrameReg, false, false, false);
   else {
     ScratchReg = MF.getRegInfo().createVirtualRegister(ARM::GPRRegisterClass);
-    if (Value) *Value = Offset;
+    if (Value) {
+      Value->first = FrameReg; // use the frame register as a kind indicator
+      Value->second = Offset;
+    }
     if (!AFI->isThumbFunction())
       emitARMRegPlusImmediate(MBB, II, MI.getDebugLoc(), ScratchReg, FrameReg,
                               Offset, Pred, PredReg, TII);

Modified: llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -145,7 +145,7 @@
                                              MachineBasicBlock::iterator I) const;
 
   virtual unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                       int SPAdj, int *Value = NULL,
+                                       int SPAdj, FrameIndexValue *Value = NULL,
                                        RegScavenger *RS = NULL) const;
 
   virtual void emitPrologue(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -429,7 +429,7 @@
 
 unsigned
 Thumb1RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                        int SPAdj, int *Value,
+                                        int SPAdj, FrameIndexValue *Value,
                                         RegScavenger *RS) const{
   unsigned VReg = 0;
   unsigned i = 0;
@@ -641,9 +641,10 @@
   } else if (Desc.mayStore()) {
       VReg = MF.getRegInfo().createVirtualRegister(ARM::tGPRRegisterClass);
       assert (Value && "Frame index virtual allocated, but Value arg is NULL!");
-      *Value = Offset;
       bool UseRR = false;
       bool TrackVReg = FrameReg == ARM::SP;
+      Value->first = FrameReg; // use the frame register as a kind indicator
+      Value->second = Offset;
 
       if (Opcode == ARM::tSpill) {
         if (FrameReg == ARM::SP)

Modified: llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.h (original)
+++ llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -59,7 +59,7 @@
                              const TargetRegisterClass *RC,
                              unsigned Reg) const;
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void emitPrologue(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -153,7 +153,7 @@
 
 unsigned
 AlphaRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                       int SPAdj, int *Value,
+                                       int SPAdj, FrameIndexValue *Value,
                                        RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unexpected");
 

Modified: llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/Alpha/AlphaRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -42,7 +42,7 @@
                                      MachineBasicBlock::iterator I) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   //void processFunctionBeforeFrameFinalized(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -221,7 +221,7 @@
 
 unsigned
 BlackfinRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                          int SPAdj, int *Value,
+                                          int SPAdj, FrameIndexValue *Value,
                                           RegScavenger *RS) const {
   MachineInstr &MI = *II;
   MachineBasicBlock &MBB = *MI.getParent();

Modified: llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/Blackfin/BlackfinRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -65,7 +65,7 @@
                                        MachineBasicBlock::iterator I) const;
 
     unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                 int SPAdj, int *Value = NULL,
+                                 int SPAdj, FrameIndexValue *Value = NULL,
                                  RegScavenger *RS = NULL) const;
 
     void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,

Modified: llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -328,7 +328,8 @@
 
 unsigned
 SPURegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj,
-                                     int *Value, RegScavenger *RS) const
+                                     FrameIndexValue *Value,
+                                     RegScavenger *RS) const
 {
   unsigned i = 0;
   MachineInstr &MI = *II;

Modified: llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.h (original)
+++ llvm/trunk/lib/Target/CellSPU/SPURegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -64,7 +64,7 @@
                                        MachineBasicBlock::iterator I) const;
     //! Convert frame indicies into machine operands
     unsigned eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj,
-                                 int *Value = NULL,
+                                 FrameIndexValue *Value = NULL,
                                  RegScavenger *RS = NULL) const;
     //! Determine the frame's layour
     void determineFrameLayout(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -260,7 +260,7 @@
 // direct reference.
 unsigned MBlazeRegisterInfo::
 eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj,
-                    int *Value, RegScavenger *RS) const {
+                    FrameIndexValue *Value, RegScavenger *RS) const {
   MachineInstr &MI = *II;
   MachineFunction &MF = *MI.getParent()->getParent();
 

Modified: llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/MBlaze/MBlazeRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -67,7 +67,7 @@
 
   /// Stack Frame Processing Methods
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void processFunctionBeforeFrameFinalized(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -207,7 +207,7 @@
 
 unsigned
 MSP430RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                        int SPAdj, int *Value,
+                                        int SPAdj, FrameIndexValue *Value,
                                         RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unexpected");
 

Modified: llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.h (original)
+++ llvm/trunk/lib/Target/MSP430/MSP430RegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -50,7 +50,7 @@
                                      MachineBasicBlock::iterator I) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void emitPrologue(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/Mips/MipsRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mips/MipsRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mips/MipsRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/Mips/MipsRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -355,7 +355,7 @@
 // direct reference.
 unsigned MipsRegisterInfo::
 eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj,
-                    int *Value, RegScavenger *RS) const
+                    FrameIndexValue *Value, RegScavenger *RS) const
 {
   MachineInstr &MI = *II;
   MachineFunction &MF = *MI.getParent()->getParent();

Modified: llvm/trunk/lib/Target/Mips/MipsRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mips/MipsRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mips/MipsRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/Mips/MipsRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -64,7 +64,7 @@
 
   /// Stack Frame Processing Methods
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void processFunctionBeforeFrameFinalized(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -53,7 +53,7 @@
 
 unsigned PIC16RegisterInfo::
 eliminateFrameIndex(MachineBasicBlock::iterator II, int SPAdj,
-                    int *Value, RegScavenger *RS) const
+                    FrameIndexValue *Value, RegScavenger *RS) const
 {
   /* NOT YET IMPLEMENTED */
   return 0;

Modified: llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.h (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16RegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -49,7 +49,7 @@
   virtual bool hasFP(const MachineFunction &MF) const;
 
   virtual unsigned eliminateFrameIndex(MachineBasicBlock::iterator MI,
-                                       int SPAdj, int *Value = NULL,
+                                       int SPAdj, FrameIndexValue *Value = NULL,
                                        RegScavenger *RS=NULL) const;
 
   void eliminateCallFramePseudoInstr(MachineFunction &MF,

Modified: llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -713,7 +713,7 @@
 
 unsigned
 PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                     int SPAdj, int *Value,
+                                     int SPAdj, FrameIndexValue *Value,
                                      RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unexpected");
 

Modified: llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -67,7 +67,7 @@
   void lowerCRSpilling(MachineBasicBlock::iterator II, unsigned FrameIndex,
                        int SPAdj, RegScavenger *RS) const;
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   /// determineFrameLayout - Determine the size of the frame and maximum call

Modified: llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -78,7 +78,7 @@
 
 unsigned
 SparcRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                       int SPAdj, int *Value,
+                                       int SPAdj, FrameIndexValue *Value,
                                        RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unexpected");
 

Modified: llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/Sparc/SparcRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -44,7 +44,7 @@
                                      MachineBasicBlock::iterator I) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void processFunctionBeforeFrameFinalized(MachineFunction &MF) const;

Modified: llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -110,7 +110,7 @@
 
 unsigned
 SystemZRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                         int SPAdj, int *Value,
+                                         int SPAdj, FrameIndexValue *Value,
                                          RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unxpected");
 

Modified: llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/SystemZ/SystemZRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -56,7 +56,7 @@
                                      MachineBasicBlock::iterator I) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
 

Modified: llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -589,7 +589,7 @@
 
 unsigned
 X86RegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                     int SPAdj, int *Value,
+                                     int SPAdj, FrameIndexValue *Value,
                                      RegScavenger *RS) const{
   assert(SPAdj == 0 && "Unexpected");
 

Modified: llvm/trunk/lib/Target/X86/X86RegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86RegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86RegisterInfo.h (original)
+++ llvm/trunk/lib/Target/X86/X86RegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -143,7 +143,7 @@
                                      MachineBasicBlock::iterator MI) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator MI,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,

Modified: llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.cpp?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.cpp Tue Mar  9 15:45:49 2010
@@ -173,7 +173,7 @@
 
 unsigned
 XCoreRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
-                                       int SPAdj, int *Value,
+                                       int SPAdj, FrameIndexValue *Value,
                                        RegScavenger *RS) const {
   assert(SPAdj == 0 && "Unexpected");
   MachineInstr &MI = *II;

Modified: llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.h?rev=98086&r1=98085&r2=98086&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreRegisterInfo.h Tue Mar  9 15:45:49 2010
@@ -58,7 +58,7 @@
                                      MachineBasicBlock::iterator I) const;
 
   unsigned eliminateFrameIndex(MachineBasicBlock::iterator II,
-                               int SPAdj, int *Value = NULL,
+                               int SPAdj, FrameIndexValue *Value = NULL,
                                RegScavenger *RS = NULL) const;
 
   void processFunctionBeforeCalleeSavedScan(MachineFunction &MF,




From stoklund at 2pi.dk  Tue Mar  9 16:17:07 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 22:17:07 -0000
Subject: [llvm-commits] [llvm] r98088 -
	/llvm/trunk/lib/Analysis/InlineCost.cpp
Message-ID: <20100309221707.0C6E82A6C12C@llvm.org>

Author: stoklund
Date: Tue Mar  9 16:17:06 2010
New Revision: 98088

URL: http://llvm.org/viewvc/llvm-project?rev=98088&view=rev
Log:
Permit inlining into huge functions. This heuristic is ancient, and inlining
can sometimes help reduce function size.

Modified:
    llvm/trunk/lib/Analysis/InlineCost.cpp

Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=98088&r1=98087&r2=98088&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
+++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Mar  9 16:17:06 2010
@@ -352,11 +352,6 @@
   // Calls usually take a long time, so they make the inlining gain smaller.
   InlineCost += CalleeFI.Metrics.NumCalls * InlineConstants::CallPenalty;
 
-  // Don't inline into something too big, which would make it bigger.
-  // "size" here is the number of basic blocks, not instructions.
-  //
-  InlineCost += Caller->size()/15;
-  
   // Look at the size of the callee. Each instruction counts as 5.
   InlineCost += CalleeFI.Metrics.NumInsts*InlineConstants::InstrCost;
 




From stoklund at 2pi.dk  Tue Mar  9 16:17:11 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 22:17:11 -0000
Subject: [llvm-commits] [llvm] r98089 - in /llvm/trunk:
 include/llvm/Analysis/InlineCost.h
 include/llvm/Transforms/IPO/InlinerPass.h lib/Analysis/InlineCost.cpp
 lib/Transforms/IPO/InlineAlways.cpp lib/Transforms/IPO/InlineSimple.cpp
 lib/Transforms/IPO/Inliner.cpp
Message-ID: <20100309221711.3A2962A6C12D@llvm.org>

Author: stoklund
Date: Tue Mar  9 16:17:11 2010
New Revision: 98089

URL: http://llvm.org/viewvc/llvm-project?rev=98089&view=rev
Log:
Try to keep the cached inliner costs around for a bit longer for big functions.

The Caller cost info would be reset everytime a callee was inlined. If the
caller has lots of calls and there is some mutual recursion going on, the
caller cost info could be calculated many times.

This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
small function calls.

Modified:
    llvm/trunk/include/llvm/Analysis/InlineCost.h
    llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
    llvm/trunk/lib/Analysis/InlineCost.cpp
    llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
    llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
    llvm/trunk/lib/Transforms/IPO/Inliner.cpp

Modified: llvm/trunk/include/llvm/Analysis/InlineCost.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/InlineCost.h?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/InlineCost.h (original)
+++ llvm/trunk/include/llvm/Analysis/InlineCost.h Tue Mar  9 16:17:11 2010
@@ -179,6 +179,11 @@
     void resetCachedCostInfo(Function* Caller) {
       CachedFunctionInfo[Caller] = FunctionInfo();
     }
+
+    /// growCachedCostInfo - update the cached cost info for Caller after Callee
+    /// has been inlined. If Callee is NULL it means a dead call has been
+    /// eliminated.
+    void growCachedCostInfo(Function* Caller, Function* Callee);
   };
 }
 

Modified: llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h (original)
+++ llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h Tue Mar  9 16:17:11 2010
@@ -75,6 +75,10 @@
   /// 
   virtual void resetCachedCostInfo(Function* Caller) = 0;
 
+  /// growCachedCostInfo - update the cached cost info for Caller after Callee
+  /// has been inlined.
+  virtual void growCachedCostInfo(Function* Caller, Function* Callee) = 0;
+
   /// removeDeadFunctions - Remove dead functions that are not included in
   /// DNR (Do Not Remove) list.
   bool removeDeadFunctions(CallGraph &CG, 

Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
+++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Mar  9 16:17:11 2010
@@ -383,3 +383,39 @@
     Factor += 1.5f;
   return Factor;
 }
+
+/// growCachedCostInfo - update the cached cost info for Caller after Callee has
+/// been inlined.
+void
+InlineCostAnalyzer::growCachedCostInfo(Function* Caller, Function* Callee) {
+  FunctionInfo &CallerFI = CachedFunctionInfo[Caller];
+
+  // For small functions we prefer to recalculate the cost for better accuracy.
+  if (!CallerFI.Metrics.NumBlocks || CallerFI.Metrics.NumInsts < 100) {
+    resetCachedCostInfo(Caller);
+    return;
+  }
+
+  // For large functions, we can save a lot of computation time by skipping
+  // recalculations.
+  if (CallerFI.Metrics.NumCalls > 0)
+    --CallerFI.Metrics.NumCalls;
+
+  if (Callee) {
+    FunctionInfo &CalleeFI = CachedFunctionInfo[Callee];
+    if (!CalleeFI.Metrics.NumBlocks) {
+      resetCachedCostInfo(Caller);
+      return;
+    }
+    CallerFI.Metrics.NeverInline |= CalleeFI.Metrics.NeverInline;
+    CallerFI.Metrics.usesDynamicAlloca |= CalleeFI.Metrics.usesDynamicAlloca;
+
+    CallerFI.Metrics.NumInsts += CalleeFI.Metrics.NumInsts;
+    CallerFI.Metrics.NumBlocks += CalleeFI.Metrics.NumBlocks;
+    CallerFI.Metrics.NumCalls += CalleeFI.Metrics.NumCalls;
+    CallerFI.Metrics.NumVectorInsts += CalleeFI.Metrics.NumVectorInsts;
+    CallerFI.Metrics.NumRets += CalleeFI.Metrics.NumRets;
+  }
+  // We are not updating the argumentweights. We have already determined that
+  // Caller is a fairly large function, so we accept the loss of precision.
+}

Modified: llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp Tue Mar  9 16:17:11 2010
@@ -45,7 +45,10 @@
       return CA.getInlineFudgeFactor(CS);
     }
     void resetCachedCostInfo(Function *Caller) {
-      return CA.resetCachedCostInfo(Caller);
+      CA.resetCachedCostInfo(Caller);
+    }
+    void growCachedCostInfo(Function* Caller, Function* Callee) {
+      CA.growCachedCostInfo(Caller, Callee);
     }
     virtual bool doFinalization(CallGraph &CG) { 
       return removeDeadFunctions(CG, &NeverInline); 

Modified: llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp Tue Mar  9 16:17:11 2010
@@ -45,6 +45,9 @@
     void resetCachedCostInfo(Function *Caller) {
       CA.resetCachedCostInfo(Caller);
     }
+    void growCachedCostInfo(Function* Caller, Function* Callee) {
+      CA.growCachedCostInfo(Caller, Callee);
+    }
     virtual bool doInitialization(CallGraph &CG);
   };
 }

Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=98089&r1=98088&r2=98089&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Tue Mar  9 16:17:11 2010
@@ -369,6 +369,8 @@
         CG[Caller]->removeCallEdgeFor(CS);
         CS.getInstruction()->eraseFromParent();
         ++NumCallsDeleted;
+        // Update the cached cost info with the missing call
+        growCachedCostInfo(Caller, NULL);
       } else {
         // We can only inline direct calls to non-declarations.
         if (Callee == 0 || Callee->isDeclaration()) continue;
@@ -382,6 +384,9 @@
         if (!InlineCallIfPossible(CS, CG, TD, InlinedArrayAllocas))
           continue;
         ++NumInlined;
+
+        // Update the cached cost info with the inlined call.
+        growCachedCostInfo(Caller, Callee);
       }
       
       // If we inlined or deleted the last possible call site to the function,
@@ -407,11 +412,6 @@
         delete CG.removeFunctionFromModule(CalleeNode);
         ++NumDeleted;
       }
-      
-      // Remove any cached cost info for this caller, as inlining the
-      // callee has increased the size of the caller (which may be the
-      // same as the callee).
-      resetCachedCostInfo(Caller);
 
       // Remove this call site from the list.  If possible, use 
       // swap/pop_back for efficiency, but do not use it if doing so would




From stoklund at 2pi.dk  Tue Mar  9 16:43:37 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 22:43:37 -0000
Subject: [llvm-commits] [llvm] r98094 - in /llvm/trunk:
 include/llvm/Analysis/InlineCost.h
 include/llvm/Transforms/IPO/InlinerPass.h lib/Analysis/InlineCost.cpp
 lib/Transforms/IPO/InlineAlways.cpp lib/Transforms/IPO/InlineSimple.cpp
 lib/Transforms/IPO/Inliner.cpp
Message-ID: <20100309224337.E10F42A6C12C@llvm.org>

Author: stoklund
Date: Tue Mar  9 16:43:37 2010
New Revision: 98094

URL: http://llvm.org/viewvc/llvm-project?rev=98094&view=rev
Log:
Revert r98089, it was breaking a clang test.

Modified:
    llvm/trunk/include/llvm/Analysis/InlineCost.h
    llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
    llvm/trunk/lib/Analysis/InlineCost.cpp
    llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
    llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
    llvm/trunk/lib/Transforms/IPO/Inliner.cpp

Modified: llvm/trunk/include/llvm/Analysis/InlineCost.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/InlineCost.h?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/InlineCost.h (original)
+++ llvm/trunk/include/llvm/Analysis/InlineCost.h Tue Mar  9 16:43:37 2010
@@ -179,11 +179,6 @@
     void resetCachedCostInfo(Function* Caller) {
       CachedFunctionInfo[Caller] = FunctionInfo();
     }
-
-    /// growCachedCostInfo - update the cached cost info for Caller after Callee
-    /// has been inlined. If Callee is NULL it means a dead call has been
-    /// eliminated.
-    void growCachedCostInfo(Function* Caller, Function* Callee);
   };
 }
 

Modified: llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h (original)
+++ llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h Tue Mar  9 16:43:37 2010
@@ -75,10 +75,6 @@
   /// 
   virtual void resetCachedCostInfo(Function* Caller) = 0;
 
-  /// growCachedCostInfo - update the cached cost info for Caller after Callee
-  /// has been inlined.
-  virtual void growCachedCostInfo(Function* Caller, Function* Callee) = 0;
-
   /// removeDeadFunctions - Remove dead functions that are not included in
   /// DNR (Do Not Remove) list.
   bool removeDeadFunctions(CallGraph &CG, 

Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
+++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Mar  9 16:43:37 2010
@@ -383,39 +383,3 @@
     Factor += 1.5f;
   return Factor;
 }
-
-/// growCachedCostInfo - update the cached cost info for Caller after Callee has
-/// been inlined.
-void
-InlineCostAnalyzer::growCachedCostInfo(Function* Caller, Function* Callee) {
-  FunctionInfo &CallerFI = CachedFunctionInfo[Caller];
-
-  // For small functions we prefer to recalculate the cost for better accuracy.
-  if (!CallerFI.Metrics.NumBlocks || CallerFI.Metrics.NumInsts < 100) {
-    resetCachedCostInfo(Caller);
-    return;
-  }
-
-  // For large functions, we can save a lot of computation time by skipping
-  // recalculations.
-  if (CallerFI.Metrics.NumCalls > 0)
-    --CallerFI.Metrics.NumCalls;
-
-  if (Callee) {
-    FunctionInfo &CalleeFI = CachedFunctionInfo[Callee];
-    if (!CalleeFI.Metrics.NumBlocks) {
-      resetCachedCostInfo(Caller);
-      return;
-    }
-    CallerFI.Metrics.NeverInline |= CalleeFI.Metrics.NeverInline;
-    CallerFI.Metrics.usesDynamicAlloca |= CalleeFI.Metrics.usesDynamicAlloca;
-
-    CallerFI.Metrics.NumInsts += CalleeFI.Metrics.NumInsts;
-    CallerFI.Metrics.NumBlocks += CalleeFI.Metrics.NumBlocks;
-    CallerFI.Metrics.NumCalls += CalleeFI.Metrics.NumCalls;
-    CallerFI.Metrics.NumVectorInsts += CalleeFI.Metrics.NumVectorInsts;
-    CallerFI.Metrics.NumRets += CalleeFI.Metrics.NumRets;
-  }
-  // We are not updating the argumentweights. We have already determined that
-  // Caller is a fairly large function, so we accept the loss of precision.
-}

Modified: llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp Tue Mar  9 16:43:37 2010
@@ -45,10 +45,7 @@
       return CA.getInlineFudgeFactor(CS);
     }
     void resetCachedCostInfo(Function *Caller) {
-      CA.resetCachedCostInfo(Caller);
-    }
-    void growCachedCostInfo(Function* Caller, Function* Callee) {
-      CA.growCachedCostInfo(Caller, Callee);
+      return CA.resetCachedCostInfo(Caller);
     }
     virtual bool doFinalization(CallGraph &CG) { 
       return removeDeadFunctions(CG, &NeverInline); 

Modified: llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp Tue Mar  9 16:43:37 2010
@@ -45,9 +45,6 @@
     void resetCachedCostInfo(Function *Caller) {
       CA.resetCachedCostInfo(Caller);
     }
-    void growCachedCostInfo(Function* Caller, Function* Callee) {
-      CA.growCachedCostInfo(Caller, Callee);
-    }
     virtual bool doInitialization(CallGraph &CG);
   };
 }

Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=98094&r1=98093&r2=98094&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Tue Mar  9 16:43:37 2010
@@ -369,8 +369,6 @@
         CG[Caller]->removeCallEdgeFor(CS);
         CS.getInstruction()->eraseFromParent();
         ++NumCallsDeleted;
-        // Update the cached cost info with the missing call
-        growCachedCostInfo(Caller, NULL);
       } else {
         // We can only inline direct calls to non-declarations.
         if (Callee == 0 || Callee->isDeclaration()) continue;
@@ -384,9 +382,6 @@
         if (!InlineCallIfPossible(CS, CG, TD, InlinedArrayAllocas))
           continue;
         ++NumInlined;
-
-        // Update the cached cost info with the inlined call.
-        growCachedCostInfo(Caller, Callee);
       }
       
       // If we inlined or deleted the last possible call site to the function,
@@ -412,6 +407,11 @@
         delete CG.removeFunctionFromModule(CalleeNode);
         ++NumDeleted;
       }
+      
+      // Remove any cached cost info for this caller, as inlining the
+      // callee has increased the size of the caller (which may be the
+      // same as the callee).
+      resetCachedCostInfo(Caller);
 
       // Remove this call site from the list.  If possible, use 
       // swap/pop_back for efficiency, but do not use it if doing so would




From chandlerc at gmail.com  Tue Mar  9 16:45:10 2010
From: chandlerc at gmail.com (Chandler Carruth)
Date: Tue, 09 Mar 2010 22:45:10 -0000
Subject: [llvm-commits] [llvm] r98095 - in /llvm/trunk/utils/unittest:
 UnitTestMain/Makefile googletest/Makefile
Message-ID: <20100309224510.D00332A6C12C@llvm.org>

Author: chandlerc
Date: Tue Mar  9 16:45:10 2010
New Revision: 98095

URL: http://llvm.org/viewvc/llvm-project?rev=98095&view=rev
Log:
Continue propagating the GoogleTest flags until we can update our version to
eliminate this problem. This will hopefully let us make progress on Linux
bootstrapping.

Modified:
    llvm/trunk/utils/unittest/UnitTestMain/Makefile
    llvm/trunk/utils/unittest/googletest/Makefile

Modified: llvm/trunk/utils/unittest/UnitTestMain/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/unittest/UnitTestMain/Makefile?rev=98095&r1=98094&r2=98095&view=diff
==============================================================================
--- llvm/trunk/utils/unittest/UnitTestMain/Makefile (original)
+++ llvm/trunk/utils/unittest/UnitTestMain/Makefile Tue Mar  9 16:45:10 2010
@@ -14,7 +14,13 @@
 LIBRARYNAME = UnitTestMain
 BUILD_ARCHIVE = 1
 REQUIRES_RTTI = 1
+
 CPP.Flags += -I$(LLVM_SRC_ROOT)/utils/unittest/googletest/include
 CPP.Flags += $(NO_MISSING_FIELD_INITIALIZERS) $(NO_VARIADIC_MACROS)
+CPP.Flags += -DGTEST_HAS_RTTI=0
+# libstdc++'s TR1 <tuple> header depends on RTTI and uses C++'0x features not
+# supported by Clang, so force googletest to use its own tuple implementation.
+# When we import googletest >=1.4.0, we can drop this line.
+CPP.Flags += -DGTEST_HAS_TR1_TUPLE=0
 
 include $(LEVEL)/Makefile.common

Modified: llvm/trunk/utils/unittest/googletest/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/unittest/googletest/Makefile?rev=98095&r1=98094&r2=98095&view=diff
==============================================================================
--- llvm/trunk/utils/unittest/googletest/Makefile (original)
+++ llvm/trunk/utils/unittest/googletest/Makefile Tue Mar  9 16:45:10 2010
@@ -16,7 +16,8 @@
 REQUIRES_RTTI = 1
 
 # Note that these flags are duplicated when building individual tests in
-# unittests/Makefile.unittest; ensure that any changes are made to both.
+# unittests/Makefile.unittest and ../UnitTestMain/Makefile; ensure that any
+# changes are made to both.
 CPP.Flags += -I$(LLVM_SRC_ROOT)/utils/unittest/googletest/include
 CPP.Flags += $(NO_MISSING_FIELD_INITIALIZERS) $(NO_VARIADIC_MACROS)
 CPP.Flags += -DGTEST_HAS_RTTI=0




From daniel at zuster.org  Tue Mar  9 16:50:40 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 22:50:40 -0000
Subject: [llvm-commits] [llvm] r98097 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td test/MC/AsmParser/X86/x86_32-new-encoder.s
Message-ID: <20100309225041.17BB52A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 16:50:40 2010
New Revision: 98097

URL: http://llvm.org/viewvc/llvm-project?rev=98097&view=rev
Log:
MC/X86: Rename alternate spellings of CMP{8,16,32} and mark as "code gen only" so they don't get selected by the asm matcher.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=98097&r1=98096&r2=98097&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Tue Mar  9 16:50:40 2010
@@ -3453,12 +3453,18 @@
                 "cmp{l}\t{$src2, $src1|$src1, $src2}",
                 [(X86cmp GR32:$src1, (loadi32 addr:$src2)),
                  (implicit EFLAGS)]>;
-def CMP8mrmrr : I<0x3A, MRMSrcReg, (outs), (ins GR8:$src1, GR8:$src2),
-                  "cmp{b}\t{$src2, $src1|$src1, $src2}", []>;
-def CMP16mrmrr : I<0x3B, MRMSrcReg, (outs), (ins GR16:$src1, GR16:$src2),
-                   "cmp{w}\t{$src2, $src1|$src1, $src2}", []>, OpSize;
-def CMP32mrmrr : I<0x3B, MRMSrcReg, (outs), (ins GR32:$src1, GR32:$src2),
-                   "cmp{l}\t{$src2, $src1|$src1, $src2}", []>;
+
+// These are alternate spellings for use by the disassembler, we mark them as
+// code gen only to ensure they aren't matched by the assembler.
+let isCodeGenOnly = 1 in {
+  def CMP8rr_alt : I<0x3A, MRMSrcReg, (outs), (ins GR8:$src1, GR8:$src2),
+                    "cmp{b}\t{$src2, $src1|$src1, $src2}", []>;
+  def CMP16rr_alt : I<0x3B, MRMSrcReg, (outs), (ins GR16:$src1, GR16:$src2),
+                     "cmp{w}\t{$src2, $src1|$src1, $src2}", []>, OpSize;
+  def CMP32rr_alt : I<0x3B, MRMSrcReg, (outs), (ins GR32:$src1, GR32:$src2),
+                     "cmp{l}\t{$src2, $src1|$src1, $src2}", []>;
+}
+
 def CMP8ri  : Ii8<0x80, MRM7r,
                   (outs), (ins GR8:$src1, i8imm:$src2),
                   "cmp{b}\t{$src2, $src1|$src1, $src2}",

Modified: llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s?rev=98097&r1=98096&r2=98097&view=diff
==============================================================================
--- llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s (original)
+++ llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s Tue Mar  9 16:50:40 2010
@@ -41,3 +41,6 @@
 
 // CHECK: testb	%bl, %cl                # encoding: [0x84,0xcb]
         testb %bl, %cl
+
+// CHECK: cmpl	%eax, %ebx              # encoding: [0x39,0xc3]
+        cmpl %eax, %ebx




From daniel at zuster.org  Tue Mar  9 16:50:46 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 09 Mar 2010 22:50:46 -0000
Subject: [llvm-commits] [llvm] r98098 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td test/MC/AsmParser/X86/x86_32-new-encoder.s
Message-ID: <20100309225046.D89CF2A6C12D@llvm.org>

Author: ddunbar
Date: Tue Mar  9 16:50:46 2010
New Revision: 98098

URL: http://llvm.org/viewvc/llvm-project?rev=98098&view=rev
Log:
MC/X86: Rename alternate spellings of ADD{8,16,32} and mark as "code gen only" so they don't get selected by the asm matcher.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=98098&r1=98097&r2=98098&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Tue Mar  9 16:50:46 2010
@@ -2647,6 +2647,17 @@
 } // end isConvertibleToThreeAddress
 } // end isCommutable
 
+// These are alternate spellings for use by the disassembler, we mark them as
+// code gen only to ensure they aren't matched by the assembler.
+let isCodeGenOnly = 1 in {
+  def ADD8rr_alt: I<0x02, MRMSrcReg, (outs GR8:$dst), (ins GR8:$src1, GR8:$src2),
+                   "add{b}\t{$src2, $dst|$dst, $src2}", []>;
+  def ADD16rr_alt: I<0x03, MRMSrcReg,(outs GR16:$dst),(ins GR16:$src1, GR16:$src2),
+                    "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize;
+  def ADD32rr_alt: I<0x03, MRMSrcReg,(outs GR16:$dst),(ins GR16:$src1, GR16:$src2),
+                    "add{l}\t{$src2, $dst|$dst, $src2}", []>;
+}
+
 // Register-Memory Addition
 def ADD8rm   : I<0x02, MRMSrcMem, (outs GR8 :$dst),
                                   (ins GR8 :$src1, i8mem :$src2),
@@ -2664,15 +2675,6 @@
                  [(set GR32:$dst, (add GR32:$src1, (load addr:$src2))),
                   (implicit EFLAGS)]>;
                   
-// Register-Register Addition - Equivalent to the normal rr forms (ADD8rr, 
-//   ADD16rr, and ADD32rr), but differently encoded.
-def ADD8mrmrr: I<0x02, MRMSrcReg, (outs GR8:$dst), (ins GR8:$src1, GR8:$src2),
-                 "add{b}\t{$src2, $dst|$dst, $src2}", []>;
-def ADD16mrmrr: I<0x03, MRMSrcReg,(outs GR16:$dst),(ins GR16:$src1, GR16:$src2),
-                  "add{w}\t{$src2, $dst|$dst, $src2}", []>, OpSize;
-def ADD32mrmrr: I<0x03, MRMSrcReg,(outs GR16:$dst),(ins GR16:$src1, GR16:$src2),
-                  "add{l}\t{$src2, $dst|$dst, $src2}", []>;
-
 // Register-Integer Addition
 def ADD8ri    : Ii8<0x80, MRM0r, (outs GR8:$dst), (ins GR8:$src1, i8imm:$src2),
                     "add{b}\t{$src2, $dst|$dst, $src2}",

Modified: llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s?rev=98098&r1=98097&r2=98098&view=diff
==============================================================================
--- llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s (original)
+++ llvm/trunk/test/MC/AsmParser/X86/x86_32-new-encoder.s Tue Mar  9 16:50:46 2010
@@ -44,3 +44,6 @@
 
 // CHECK: cmpl	%eax, %ebx              # encoding: [0x39,0xc3]
         cmpl %eax, %ebx
+
+// CHECK: addw	%ax, %ax                # encoding: [0x66,0x01,0xc0]
+        addw %ax, %ax




From stoklund at 2pi.dk  Tue Mar  9 17:02:17 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 09 Mar 2010 23:02:17 -0000
Subject: [llvm-commits] [llvm] r98099 - in /llvm/trunk:
 include/llvm/Analysis/InlineCost.h
 include/llvm/Transforms/IPO/InlinerPass.h lib/Analysis/InlineCost.cpp
 lib/Transforms/IPO/InlineAlways.cpp lib/Transforms/IPO/InlineSimple.cpp
 lib/Transforms/IPO/Inliner.cpp
Message-ID: <20100309230217.AC9E12A6C12C@llvm.org>

Author: stoklund
Date: Tue Mar  9 17:02:17 2010
New Revision: 98099

URL: http://llvm.org/viewvc/llvm-project?rev=98099&view=rev
Log:
Try to keep the cached inliner costs around for a bit longer for big functions.

The Caller cost info would be reset everytime a callee was inlined. If the
caller has lots of calls and there is some mutual recursion going on, the
caller cost info could be calculated many times.

This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
small function calls.

This is a more conservative version of r98089 that doesn't break the clang
test CodeGenCXX/temp-order.cpp. That test relies on rather extreme inlining
for constant folding.

Modified:
    llvm/trunk/include/llvm/Analysis/InlineCost.h
    llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
    llvm/trunk/lib/Analysis/InlineCost.cpp
    llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
    llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
    llvm/trunk/lib/Transforms/IPO/Inliner.cpp

Modified: llvm/trunk/include/llvm/Analysis/InlineCost.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/InlineCost.h?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/InlineCost.h (original)
+++ llvm/trunk/include/llvm/Analysis/InlineCost.h Tue Mar  9 17:02:17 2010
@@ -179,6 +179,11 @@
     void resetCachedCostInfo(Function* Caller) {
       CachedFunctionInfo[Caller] = FunctionInfo();
     }
+
+    /// growCachedCostInfo - update the cached cost info for Caller after Callee
+    /// has been inlined. If Callee is NULL it means a dead call has been
+    /// eliminated.
+    void growCachedCostInfo(Function* Caller, Function* Callee);
   };
 }
 

Modified: llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h (original)
+++ llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h Tue Mar  9 17:02:17 2010
@@ -75,6 +75,10 @@
   /// 
   virtual void resetCachedCostInfo(Function* Caller) = 0;
 
+  /// growCachedCostInfo - update the cached cost info for Caller after Callee
+  /// has been inlined.
+  virtual void growCachedCostInfo(Function* Caller, Function* Callee) = 0;
+
   /// removeDeadFunctions - Remove dead functions that are not included in
   /// DNR (Do Not Remove) list.
   bool removeDeadFunctions(CallGraph &CG, 

Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
+++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Mar  9 17:02:17 2010
@@ -383,3 +383,45 @@
     Factor += 1.5f;
   return Factor;
 }
+
+/// growCachedCostInfo - update the cached cost info for Caller after Callee has
+/// been inlined.
+void
+InlineCostAnalyzer::growCachedCostInfo(Function* Caller, Function* Callee) {
+  FunctionInfo &CallerFI = CachedFunctionInfo[Caller];
+
+  // For small functions we prefer to recalculate the cost for better accuracy.
+  if (CallerFI.Metrics.NumBlocks < 10 || CallerFI.Metrics.NumInsts < 1000) {
+    resetCachedCostInfo(Caller);
+    return;
+  }
+
+  // For large functions, we can save a lot of computation time by skipping
+  // recalculations.
+  if (CallerFI.Metrics.NumCalls > 0)
+    --CallerFI.Metrics.NumCalls;
+
+  if (Callee) {
+    FunctionInfo &CalleeFI = CachedFunctionInfo[Callee];
+    if (!CalleeFI.Metrics.NumBlocks) {
+      resetCachedCostInfo(Caller);
+      return;
+    }
+    CallerFI.Metrics.NeverInline |= CalleeFI.Metrics.NeverInline;
+    CallerFI.Metrics.usesDynamicAlloca |= CalleeFI.Metrics.usesDynamicAlloca;
+
+    CallerFI.Metrics.NumInsts += CalleeFI.Metrics.NumInsts;
+    CallerFI.Metrics.NumBlocks += CalleeFI.Metrics.NumBlocks;
+    CallerFI.Metrics.NumCalls += CalleeFI.Metrics.NumCalls;
+    CallerFI.Metrics.NumVectorInsts += CalleeFI.Metrics.NumVectorInsts;
+    CallerFI.Metrics.NumRets += CalleeFI.Metrics.NumRets;
+
+    // analyzeBasicBlock counts each function argument as an inst.
+    if (CallerFI.Metrics.NumInsts >= Callee->arg_size())
+      CallerFI.Metrics.NumInsts -= Callee->arg_size();
+    else
+      CallerFI.Metrics.NumInsts = 0;
+  }
+  // We are not updating the argumentweights. We have already determined that
+  // Caller is a fairly large function, so we accept the loss of precision.
+}

Modified: llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp Tue Mar  9 17:02:17 2010
@@ -45,7 +45,10 @@
       return CA.getInlineFudgeFactor(CS);
     }
     void resetCachedCostInfo(Function *Caller) {
-      return CA.resetCachedCostInfo(Caller);
+      CA.resetCachedCostInfo(Caller);
+    }
+    void growCachedCostInfo(Function* Caller, Function* Callee) {
+      CA.growCachedCostInfo(Caller, Callee);
     }
     virtual bool doFinalization(CallGraph &CG) { 
       return removeDeadFunctions(CG, &NeverInline); 

Modified: llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp Tue Mar  9 17:02:17 2010
@@ -45,6 +45,9 @@
     void resetCachedCostInfo(Function *Caller) {
       CA.resetCachedCostInfo(Caller);
     }
+    void growCachedCostInfo(Function* Caller, Function* Callee) {
+      CA.growCachedCostInfo(Caller, Callee);
+    }
     virtual bool doInitialization(CallGraph &CG);
   };
 }

Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=98099&r1=98098&r2=98099&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Tue Mar  9 17:02:17 2010
@@ -369,6 +369,8 @@
         CG[Caller]->removeCallEdgeFor(CS);
         CS.getInstruction()->eraseFromParent();
         ++NumCallsDeleted;
+        // Update the cached cost info with the missing call
+        growCachedCostInfo(Caller, NULL);
       } else {
         // We can only inline direct calls to non-declarations.
         if (Callee == 0 || Callee->isDeclaration()) continue;
@@ -382,6 +384,9 @@
         if (!InlineCallIfPossible(CS, CG, TD, InlinedArrayAllocas))
           continue;
         ++NumInlined;
+
+        // Update the cached cost info with the inlined call.
+        growCachedCostInfo(Caller, Callee);
       }
       
       // If we inlined or deleted the last possible call site to the function,
@@ -407,11 +412,6 @@
         delete CG.removeFunctionFromModule(CalleeNode);
         ++NumDeleted;
       }
-      
-      // Remove any cached cost info for this caller, as inlining the
-      // callee has increased the size of the caller (which may be the
-      // same as the callee).
-      resetCachedCostInfo(Caller);
 
       // Remove this call site from the list.  If possible, use 
       // swap/pop_back for efficiency, but do not use it if doing so would




From sabre at nondot.org  Tue Mar  9 17:12:18 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 23:12:18 -0000
Subject: [llvm-commits] [llvm] r98100 - /llvm/trunk/lib/MC/MCNullStreamer.cpp
Message-ID: <20100309231218.38B4A2A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 17:12:18 2010
New Revision: 98100

URL: http://llvm.org/viewvc/llvm-project?rev=98100&view=rev
Log:
make the NullStreamer set the section on a label when emitted so that isDefined() works.

Modified:
    llvm/trunk/lib/MC/MCNullStreamer.cpp

Modified: llvm/trunk/lib/MC/MCNullStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCNullStreamer.cpp?rev=98100&r1=98099&r2=98100&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCNullStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCNullStreamer.cpp Tue Mar  9 17:12:18 2010
@@ -29,7 +29,11 @@
       CurSection = Section;
     }
 
-    virtual void EmitLabel(MCSymbol *Symbol) {}
+    virtual void EmitLabel(MCSymbol *Symbol) {
+      assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
+      assert(CurSection && "Cannot emit before setting section!");
+      Symbol->setSection(*CurSection);
+    }
 
     virtual void EmitAssemblerFlag(MCAssemblerFlag Flag) {}
 




From sabre at nondot.org  Tue Mar  9 17:19:16 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 23:19:16 -0000
Subject: [llvm-commits] [llvm] r98101 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
Message-ID: <20100309231916.0B5BE2A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 17:19:15 2010
New Revision: 98101

URL: http://llvm.org/viewvc/llvm-project?rev=98101&view=rev
Log:
convert the non-"ispcrel" case of EmitReference to MC,
significant debug info testcases are now all going through
MCStreamer, though they print a lot of extraneous newlines to "O".

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98101&r1=98100&r2=98101&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 17:19:15 2010
@@ -213,6 +213,14 @@
 ///
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, bool IsPCRelative,
                                  bool Force32Bit) const {
+  unsigned Size = Force32Bit ? 4 : TD->getPointerSize();
+  
+  if (!IsPCRelative) {
+    Asm->OutStreamer.EmitSymbolValue(Sym, Size, 0/*AddrSpace*/);
+    return;
+  }
+  
+  // FIXME: Need an MCExpr for ".".
   PrintRelDirective(Force32Bit);
   O << *Sym;
   if (IsPCRelative) O << "-" << MAI->getPCSymbol();




From sabre at nondot.org  Tue Mar  9 17:38:24 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 23:38:24 -0000
Subject: [llvm-commits] [llvm] r98102 - in
	/llvm/trunk/lib/CodeGen/AsmPrinter: DIE.cpp DwarfDebug.cpp
Message-ID: <20100309233824.190212A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 17:38:23 2010
New Revision: 98102

URL: http://llvm.org/viewvc/llvm-project?rev=98102&view=rev
Log:
eliminate a bunch of \n's that are being printed to O.  Next up is to kill
off "EOL".

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp?rev=98102&r1=98101&r2=98102&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp Tue Mar  9 17:38:23 2010
@@ -131,17 +131,15 @@
       << "Die: "
       << format("0x%lx", (long)(intptr_t)this)
       << ", Offset: " << Offset
-      << ", Size: " << Size
-      << "\n";
+      << ", Size: " << Size << "\n";
 
     O << Indent
       << dwarf::TagString(Abbrev.getTag())
       << " "
-      << dwarf::ChildrenString(Abbrev.getChildrenFlag());
+      << dwarf::ChildrenString(Abbrev.getChildrenFlag()) << "\n";
   } else {
-    O << "Size: " << Size;
+    O << "Size: " << Size << "\n";
   }
-  O << "\n";
 
   const SmallVector<DIEAbbrevData, 8> &Data = Abbrev.getData();
 
@@ -286,7 +284,6 @@
 void DIESectionOffset::EmitValue(DwarfPrinter *D, unsigned Form) const {
   bool IsSmall = Form == dwarf::DW_FORM_data4;
   D->EmitSectionOffset(Label, Section, IsSmall, IsEH);
-  D->getAsm()->O << '\n'; // FIXME: Necesssary?
 }
 
 /// SizeOf - Determine size of delta value in bytes.
@@ -367,15 +364,13 @@
   case dwarf::DW_FORM_block1: Asm->EmitInt8(Size);         break;
   case dwarf::DW_FORM_block2: Asm->EmitInt16(Size);        break;
   case dwarf::DW_FORM_block4: Asm->EmitInt32(Size);        break;
-  case dwarf::DW_FORM_block:  D->EmitULEB128(Size); break;
-  default: llvm_unreachable("Improper form for block");         break;
+  case dwarf::DW_FORM_block:  D->EmitULEB128(Size);        break;
+  default: llvm_unreachable("Improper form for block");    break;
   }
 
   const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev.getData();
-  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
-    Asm->O << '\n';
+  for (unsigned i = 0, N = Values.size(); i < N; ++i)
     Values[i]->EmitValue(D, AbbrevData[i].getForm());
-  }
 }
 
 /// SizeOf - Determine size of block data in bytes.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98102&r1=98101&r2=98102&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Mar  9 17:38:23 2010
@@ -2360,8 +2360,6 @@
   unsigned AbbrevNumber = Die->getAbbrevNumber();
   const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
 
-  Asm->O << '\n';
-
   // Emit the code (index) for the abbreviation.
   if (Asm->VerboseAsm)
     Asm->OutStreamer.AddComment("Abbrev [" + Twine(AbbrevNumber) + "] 0x" +
@@ -2396,7 +2394,6 @@
     default:
       // Emit an attribute using the defined form.
       Values[i]->EmitValue(this, Form);
-      O << "\n"; // REMOVE This once all EmitValue impls emit their own newline.
       break;
     }
   }
@@ -2444,7 +2441,6 @@
   Asm->EmitInt8(0); EOL("Extra Pad For GDB");
   Asm->EmitInt8(0); EOL("Extra Pad For GDB");
   Asm->OutStreamer.EmitLabel(getDWLabel("info_end", ModuleCU->getID()));
-  Asm->O << '\n';
 }
 
 /// emitAbbreviations - Emit the abbreviation section.
@@ -2468,7 +2464,6 @@
 
       // Emit the abbreviations data.
       Abbrev->Emit(this);
-      Asm->O << '\n';
     }
 
     // Mark end of abbreviations.
@@ -2579,7 +2574,6 @@
       O << '\t' << MAI->getCommentString() << " Section"
         << S->getName() << '\n';
     }*/
-    Asm->O << '\n';
 
     // Dwarf assumes we start with first line of first source file.
     unsigned Source = 1;
@@ -2593,9 +2587,7 @@
 
       if (LineInfo.getLine() == 0) continue;
 
-      if (!Asm->isVerbose())
-        Asm->O << '\n';
-      else {
+      if (Asm->isVerbose()) {
         std::pair<unsigned, unsigned> SourceID =
           getSourceDirectoryAndFileIds(LineInfo.getSourceID());
         O << '\t' << MAI->getCommentString() << ' '
@@ -2817,23 +2809,20 @@
 ///
 void DwarfDebug::emitDebugStr() {
   // Check to see if it is worth the effort.
-  if (!StringPool.empty()) {
-    // Start the dwarf str section.
-    Asm->OutStreamer.SwitchSection(
+  if (StringPool.empty()) return;
+  
+  // Start the dwarf str section.
+  Asm->OutStreamer.SwitchSection(
                                 Asm->getObjFileLowering().getDwarfStrSection());
 
-    // For each of strings in the string pool.
-    for (unsigned StringID = 1, N = StringPool.size();
-         StringID <= N; ++StringID) {
-      // Emit a label for reference from debug information entries.
-      Asm->OutStreamer.EmitLabel(getDWLabel("string", StringID));
-
-      // Emit the string itself.
-      const std::string &String = StringPool[StringID];
-      Asm->OutStreamer.EmitBytes(StringRef(String.c_str(), String.size()+1), 0);
-    }
-
-    Asm->O << '\n';
+  // For each of strings in the string pool.
+  for (unsigned StringID = 1, N = StringPool.size(); StringID <= N; ++StringID){
+    // Emit a label for reference from debug information entries.
+    Asm->OutStreamer.EmitLabel(getDWLabel("string", StringID));
+    
+    // Emit the string itself.
+    const std::string &String = StringPool[StringID];
+    Asm->OutStreamer.EmitBytes(StringRef(String.c_str(), String.size()+1), 0);
   }
 }
 




From gohman at apple.com  Tue Mar  9 17:46:50 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 09 Mar 2010 23:46:50 -0000
Subject: [llvm-commits] [llvm] r98103 - in /llvm/trunk:
 lib/Analysis/ScalarEvolution.cpp
 test/Analysis/ScalarEvolution/unreachable-code.ll
Message-ID: <20100309234650.C59422A6C12C@llvm.org>

Author: djg
Date: Tue Mar  9 17:46:50 2010
New Revision: 98103

URL: http://llvm.org/viewvc/llvm-project?rev=98103&view=rev
Log:
Avoid analyzing instructions in blocks not reachable from the entry block.
They are lots of trouble, and they don't matter. This fixes PR6559.

Added:
    llvm/trunk/test/Analysis/ScalarEvolution/unreachable-code.ll
Modified:
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=98103&r1=98102&r2=98103&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue Mar  9 17:46:50 2010
@@ -3101,9 +3101,16 @@
     return getUnknown(V);
 
   unsigned Opcode = Instruction::UserOp1;
-  if (Instruction *I = dyn_cast<Instruction>(V))
+  if (Instruction *I = dyn_cast<Instruction>(V)) {
     Opcode = I->getOpcode();
-  else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
+
+    // Don't attempt to analyze instructions in blocks that aren't
+    // reachable. Such instructions don't matter, and they aren't required
+    // to obey basic rules for definitions dominating uses which this
+    // analysis depends on.
+    if (!DT->isReachableFromEntry(I->getParent()))
+      return getUnknown(V);
+  } else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
     Opcode = CE->getOpcode();
   else if (ConstantInt *CI = dyn_cast<ConstantInt>(V))
     return getConstant(CI);

Added: llvm/trunk/test/Analysis/ScalarEvolution/unreachable-code.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/ScalarEvolution/unreachable-code.ll?rev=98103&view=auto
==============================================================================
--- llvm/trunk/test/Analysis/ScalarEvolution/unreachable-code.ll (added)
+++ llvm/trunk/test/Analysis/ScalarEvolution/unreachable-code.ll Tue Mar  9 17:46:50 2010
@@ -0,0 +1,13 @@
+; RUN: opt < %s -analyze -scalar-evolution | FileCheck %s
+
+; CHECK: %t = add i64 %t, 1
+; CHECK: -->  %t
+
+define void @foo() {
+entry:
+  ret void
+
+dead:
+  %t = add i64 %t, 1
+  ret void
+}




From daniel at zuster.org  Tue Mar  9 17:52:31 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 9 Mar 2010 15:52:31 -0800
Subject: [llvm-commits] [llvm] r98099 - in /llvm/trunk:
	include/llvm/Analysis/InlineCost.h
	include/llvm/Transforms/IPO/InlinerPass.h
	lib/Analysis/InlineCost.cpp lib/Transforms/IPO/InlineAlways.cpp
	lib/Transforms/IPO/InlineSimple.cpp lib/Transforms/IPO/Inl
Message-ID: <6a8523d61003091552h3171643hbc650c11250033f9@mail.gmail.com>

Hi Jakob,

On Tue, Mar 9, 2010 at 3:02 PM, Jakob Stoklund Olesen <stoklund at 2pi.dk> wrote:
> Author: stoklund
> Date: Tue Mar ?9 17:02:17 2010
> New Revision: 98099
>
> URL: http://llvm.org/viewvc/llvm-project?rev=98099&view=rev
> Log:
> Try to keep the cached inliner costs around for a bit longer for big functions.
>
> The Caller cost info would be reset everytime a callee was inlined. If the
> caller has lots of calls and there is some mutual recursion going on, the
> caller cost info could be calculated many times.
>
> This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
> small function calls.
>
> This is a more conservative version of r98089 that doesn't break the clang
> test CodeGenCXX/temp-order.cpp. That test relies on rather extreme inlining
> for constant folding.

That test doesn't *have* to pass, and we can change it if need be.
However, it sounds like this should just be a performance improvement
not a functionality change, so I wouldn't expect it to break the test.

 - Daniel

> Modified:
> ? ?llvm/trunk/include/llvm/Analysis/InlineCost.h
> ? ?llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
> ? ?llvm/trunk/lib/Analysis/InlineCost.cpp
> ? ?llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
> ? ?llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
> ? ?llvm/trunk/lib/Transforms/IPO/Inliner.cpp
>
> Modified: llvm/trunk/include/llvm/Analysis/InlineCost.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/InlineCost.h?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/InlineCost.h (original)
> +++ llvm/trunk/include/llvm/Analysis/InlineCost.h Tue Mar ?9 17:02:17 2010
> @@ -179,6 +179,11 @@
> ? ? void resetCachedCostInfo(Function* Caller) {
> ? ? ? CachedFunctionInfo[Caller] = FunctionInfo();
> ? ? }
> +
> + ? ?/// growCachedCostInfo - update the cached cost info for Caller after Callee
> + ? ?/// has been inlined. If Callee is NULL it means a dead call has been
> + ? ?/// eliminated.
> + ? ?void growCachedCostInfo(Function* Caller, Function* Callee);
> ? };
> ?}
>
>
> Modified: llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h (original)
> +++ llvm/trunk/include/llvm/Transforms/IPO/InlinerPass.h Tue Mar ?9 17:02:17 2010
> @@ -75,6 +75,10 @@
> ? ///
> ? virtual void resetCachedCostInfo(Function* Caller) = 0;
>
> + ?/// growCachedCostInfo - update the cached cost info for Caller after Callee
> + ?/// has been inlined.
> + ?virtual void growCachedCostInfo(Function* Caller, Function* Callee) = 0;
> +
> ? /// removeDeadFunctions - Remove dead functions that are not included in
> ? /// DNR (Do Not Remove) list.
> ? bool removeDeadFunctions(CallGraph &CG,
>
> Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
> +++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Mar ?9 17:02:17 2010
> @@ -383,3 +383,45 @@
> ? ? Factor += 1.5f;
> ? return Factor;
> ?}
> +
> +/// growCachedCostInfo - update the cached cost info for Caller after Callee has
> +/// been inlined.
> +void
> +InlineCostAnalyzer::growCachedCostInfo(Function* Caller, Function* Callee) {
> + ?FunctionInfo &CallerFI = CachedFunctionInfo[Caller];
> +
> + ?// For small functions we prefer to recalculate the cost for better accuracy.
> + ?if (CallerFI.Metrics.NumBlocks < 10 || CallerFI.Metrics.NumInsts < 1000) {
> + ? ?resetCachedCostInfo(Caller);
> + ? ?return;
> + ?}
> +
> + ?// For large functions, we can save a lot of computation time by skipping
> + ?// recalculations.
> + ?if (CallerFI.Metrics.NumCalls > 0)
> + ? ?--CallerFI.Metrics.NumCalls;
> +
> + ?if (Callee) {
> + ? ?FunctionInfo &CalleeFI = CachedFunctionInfo[Callee];
> + ? ?if (!CalleeFI.Metrics.NumBlocks) {
> + ? ? ?resetCachedCostInfo(Caller);
> + ? ? ?return;
> + ? ?}
> + ? ?CallerFI.Metrics.NeverInline |= CalleeFI.Metrics.NeverInline;
> + ? ?CallerFI.Metrics.usesDynamicAlloca |= CalleeFI.Metrics.usesDynamicAlloca;
> +
> + ? ?CallerFI.Metrics.NumInsts += CalleeFI.Metrics.NumInsts;
> + ? ?CallerFI.Metrics.NumBlocks += CalleeFI.Metrics.NumBlocks;
> + ? ?CallerFI.Metrics.NumCalls += CalleeFI.Metrics.NumCalls;
> + ? ?CallerFI.Metrics.NumVectorInsts += CalleeFI.Metrics.NumVectorInsts;
> + ? ?CallerFI.Metrics.NumRets += CalleeFI.Metrics.NumRets;
> +
> + ? ?// analyzeBasicBlock counts each function argument as an inst.
> + ? ?if (CallerFI.Metrics.NumInsts >= Callee->arg_size())
> + ? ? ?CallerFI.Metrics.NumInsts -= Callee->arg_size();
> + ? ?else
> + ? ? ?CallerFI.Metrics.NumInsts = 0;
> + ?}
> + ?// We are not updating the argumentweights. We have already determined that
> + ?// Caller is a fairly large function, so we accept the loss of precision.
> +}
>
> Modified: llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp (original)
> +++ llvm/trunk/lib/Transforms/IPO/InlineAlways.cpp Tue Mar ?9 17:02:17 2010
> @@ -45,7 +45,10 @@
> ? ? ? return CA.getInlineFudgeFactor(CS);
> ? ? }
> ? ? void resetCachedCostInfo(Function *Caller) {
> - ? ? ?return CA.resetCachedCostInfo(Caller);
> + ? ? ?CA.resetCachedCostInfo(Caller);
> + ? ?}
> + ? ?void growCachedCostInfo(Function* Caller, Function* Callee) {
> + ? ? ?CA.growCachedCostInfo(Caller, Callee);
> ? ? }
> ? ? virtual bool doFinalization(CallGraph &CG) {
> ? ? ? return removeDeadFunctions(CG, &NeverInline);
>
> Modified: llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp (original)
> +++ llvm/trunk/lib/Transforms/IPO/InlineSimple.cpp Tue Mar ?9 17:02:17 2010
> @@ -45,6 +45,9 @@
> ? ? void resetCachedCostInfo(Function *Caller) {
> ? ? ? CA.resetCachedCostInfo(Caller);
> ? ? }
> + ? ?void growCachedCostInfo(Function* Caller, Function* Callee) {
> + ? ? ?CA.growCachedCostInfo(Caller, Callee);
> + ? ?}
> ? ? virtual bool doInitialization(CallGraph &CG);
> ? };
> ?}
>
> Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=98099&r1=98098&r2=98099&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
> +++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Tue Mar ?9 17:02:17 2010
> @@ -369,6 +369,8 @@
> ? ? ? ? CG[Caller]->removeCallEdgeFor(CS);
> ? ? ? ? CS.getInstruction()->eraseFromParent();
> ? ? ? ? ++NumCallsDeleted;
> + ? ? ? ?// Update the cached cost info with the missing call
> + ? ? ? ?growCachedCostInfo(Caller, NULL);
> ? ? ? } else {
> ? ? ? ? // We can only inline direct calls to non-declarations.
> ? ? ? ? if (Callee == 0 || Callee->isDeclaration()) continue;
> @@ -382,6 +384,9 @@
> ? ? ? ? if (!InlineCallIfPossible(CS, CG, TD, InlinedArrayAllocas))
> ? ? ? ? ? continue;
> ? ? ? ? ++NumInlined;
> +
> + ? ? ? ?// Update the cached cost info with the inlined call.
> + ? ? ? ?growCachedCostInfo(Caller, Callee);
> ? ? ? }
>
> ? ? ? // If we inlined or deleted the last possible call site to the function,
> @@ -407,11 +412,6 @@
> ? ? ? ? delete CG.removeFunctionFromModule(CalleeNode);
> ? ? ? ? ++NumDeleted;
> ? ? ? }
> -
> - ? ? ?// Remove any cached cost info for this caller, as inlining the
> - ? ? ?// callee has increased the size of the caller (which may be the
> - ? ? ?// same as the callee).
> - ? ? ?resetCachedCostInfo(Caller);
>
> ? ? ? // Remove this call site from the list. ?If possible, use
> ? ? ? // swap/pop_back for efficiency, but do not use it if doing so would
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From dalej at apple.com  Tue Mar  9 17:52:37 2010
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 09 Mar 2010 23:52:37 -0000
Subject: [llvm-commits] [llvm] r98104 -
	/llvm/trunk/lib/CodeGen/BranchFolding.cpp
Message-ID: <20100309235237.F10272A6C12C@llvm.org>

Author: johannes
Date: Tue Mar  9 17:52:37 2010
New Revision: 98104

URL: http://llvm.org/viewvc/llvm-project?rev=98104&view=rev
Log:
Ever more complicated DEBUG_VALUE fixes for branch folding.


Modified:
    llvm/trunk/lib/CodeGen/BranchFolding.cpp

Modified: llvm/trunk/lib/CodeGen/BranchFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/BranchFolding.cpp?rev=98104&r1=98103&r2=98104&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/BranchFolding.cpp (original)
+++ llvm/trunk/lib/CodeGen/BranchFolding.cpp Tue Mar  9 17:52:37 2010
@@ -347,15 +347,29 @@
     --I1; --I2;
     // Skip debugging pseudos; necessary to avoid changing the code.
     while (I1->isDebugValue()) {
-      if (I1==MBB1->begin())
+      if (I1==MBB1->begin()) {
+        while (I2->isDebugValue()) {
+          if (I2==MBB2->begin())
+            // I1==DBG at begin; I2==DBG at begin
+            return TailLen;
+          --I2;
+        }
+        ++I2;
+        // I1==DBG at begin; I2==non-DBG, or first of DBGs not at begin
         return TailLen;
+      }
       --I1;
     }
+    // I1==first (untested) non-DBG preceding known match
     while (I2->isDebugValue()) {
-      if (I2==MBB2->begin())
+      if (I2==MBB2->begin()) {
+        ++I1;
+        // I1==non-DBG, or first of DBGs not at begin; I2==DBG at begin
         return TailLen;
+      }
       --I2;
     }
+    // I1, I2==first (untested) non-DBGs preceding known match
     if (!I1->isIdenticalTo(I2) ||
         // FIXME: This check is dubious. It's used to get around a problem where
         // people incorrectly expect inline asm directives to remain in the same
@@ -368,6 +382,29 @@
     }
     ++TailLen;
   }
+  // Back past possible debugging pseudos at beginning of block.  This matters
+  // when one block differs from the other only by whether debugging pseudos
+  // are present at the beginning.  (This way, the various checks later for
+  // I1==MBB1->begin() work as expected.)
+  if (I1 == MBB1->begin() && I2 != MBB2->begin()) {
+    --I2;
+    while (I2->isDebugValue()) {
+      if (I2 == MBB2->begin()) {
+        return TailLen;
+        }
+      --I2;
+    }
+    ++I2;
+  }
+  if (I2 == MBB2->begin() && I1 != MBB1->begin()) {
+    --I1;
+    while (I1->isDebugValue()) {
+      if (I1 == MBB1->begin())
+        return TailLen;
+      --I1;
+    }
+    ++I1;
+  }
   return TailLen;
 }
 
@@ -1163,7 +1200,23 @@
       // be 'non-branch terminators' in the block, try removing the branch and
       // then seeing if the block is empty.
       TII->RemoveBranch(*MBB);
-
+      // If the only things remaining in the block are debug info, remove these
+      // as well, so this will behave the same as an empty block in non-debug
+      // mode.
+      if (!MBB->empty()) {
+        bool NonDebugInfoFound = false;
+        for (MachineBasicBlock::iterator I = MBB->begin(), E = MBB->end();
+             I != E; ++I) {
+          if (!I->isDebugValue()) {
+            NonDebugInfoFound = true;
+            break;
+          }
+        }
+        if (!NonDebugInfoFound)
+          // Make the block empty, losing the debug info (we could probably
+          // improve this in some cases.)
+          MBB->erase(MBB->begin(), MBB->end());
+      }
       // If this block is just an unconditional branch to CurTBB, we can
       // usually completely eliminate the block.  The only case we cannot
       // completely eliminate the block is when the block before this one




From sabre at nondot.org  Tue Mar  9 17:52:58 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 23:52:58 -0000
Subject: [llvm-commits] [llvm] r98105 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfException.cpp
 DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100309235258.660C22A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 17:52:58 2010
New Revision: 98105

URL: http://llvm.org/viewvc/llvm-project?rev=98105&view=rev
Log:
eliminate EOL, adding all comments with the OutStreamer.AddComment 
method.  With this, comments should end up on the same lines as the .byte
directives (for example) and we now get no output with:

$ llc CodeGen/X86/2009-02-12-DebugInfoVLA.ll -o - -filetype=null -asm-verbose

woot.


Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98105&r1=98104&r2=98105&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Mar  9 17:52:58 2010
@@ -2405,7 +2405,9 @@
     for (unsigned j = 0, M = Children.size(); j < M; ++j)
       emitDIE(Children[j]);
 
-    Asm->EmitInt8(0); EOL("End Of Children Mark");
+    if (Asm->VerboseAsm)
+      Asm->OutStreamer.AddComment("End Of Children Mark");
+    Asm->EmitInt8(0);
   }
 }
 
@@ -2427,19 +2429,23 @@
     sizeof(int8_t) +  // Pointer Size (in bytes)
     sizeof(int32_t);  // FIXME - extra pad for gdb bug.
 
-  Asm->EmitInt32(ContentSize);  EOL("Length of Compilation Unit Info");
-  Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF version number");
+  Asm->OutStreamer.AddComment("Length of Compilation Unit Info");
+  Asm->EmitInt32(ContentSize);
+  Asm->OutStreamer.AddComment("DWARF version number");
+  Asm->EmitInt16(dwarf::DWARF_VERSION);
+  Asm->OutStreamer.AddComment("Offset Into Abbrev. Section");
   EmitSectionOffset(getTempLabel("abbrev_begin"),getTempLabel("section_abbrev"),
                     true, false);
-  EOL("Offset Into Abbrev. Section");
-  Asm->EmitInt8(TD->getPointerSize()); EOL("Address Size (in bytes)");
+  Asm->OutStreamer.AddComment("Address Size (in bytes)");
+  Asm->EmitInt8(TD->getPointerSize());
 
   emitDIE(Die);
   // FIXME - extra padding for gdb bug.
-  Asm->EmitInt8(0); EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); EOL("Extra Pad For GDB");
+  Asm->OutStreamer.AddComment("4 extra padding bytes for GDB");
+  Asm->EmitInt8(0);
+  Asm->EmitInt8(0);
+  Asm->EmitInt8(0);
+  Asm->EmitInt8(0);
   Asm->OutStreamer.EmitLabel(getDWLabel("info_end", ModuleCU->getID()));
 }
 
@@ -2478,14 +2484,20 @@
 ///
 void DwarfDebug::emitEndOfLineMatrix(unsigned SectionEnd) {
   // Define last address of section.
-  Asm->EmitInt8(0); EOL("Extended Op");
-  Asm->EmitInt8(TD->getPointerSize() + 1); EOL("Op size");
-  Asm->EmitInt8(dwarf::DW_LNE_set_address); EOL("DW_LNE_set_address");
+  Asm->OutStreamer.AddComment("Extended Op");
+  Asm->EmitInt8(0);
+  
+  Asm->OutStreamer.AddComment("Op size");
+  Asm->EmitInt8(TD->getPointerSize() + 1);
+  Asm->OutStreamer.AddComment("DW_LNE_set_address");
+  Asm->EmitInt8(dwarf::DW_LNE_set_address);
+
+  Asm->OutStreamer.AddComment("Section end label");
   EmitReference(getDWLabel("section_end", SectionEnd));
-  EOL("Section end label");
 
   // Mark end of matrix.
-  Asm->EmitInt8(0); EOL("DW_LNE_end_sequence");
+  Asm->OutStreamer.AddComment("DW_LNE_end_sequence");
+  Asm->EmitInt8(0);
   Asm->EmitInt8(1);
   Asm->EmitInt8(1);
 }
@@ -2508,33 +2520,48 @@
                             Asm->getObjFileLowering().getDwarfLineSection());
 
   // Construct the section header.
+  Asm->OutStreamer.AddComment("Length of Source Line Info");
   EmitDifference(getTempLabel("line_end"), getTempLabel("line_begin"), true);
-  EOL("Length of Source Line Info");
   Asm->OutStreamer.EmitLabel(getTempLabel("line_begin"));
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF version number");
+  Asm->OutStreamer.AddComment("DWARF version number");
+  Asm->EmitInt16(dwarf::DWARF_VERSION); 
 
+  Asm->OutStreamer.AddComment("Prolog Length");
   EmitDifference(getTempLabel("line_prolog_end"),
                  getTempLabel("line_prolog_begin"), true);
-  EOL("Prolog Length");
   Asm->OutStreamer.EmitLabel(getTempLabel("line_prolog_begin"));
 
-  Asm->EmitInt8(1); EOL("Minimum Instruction Length");
-  Asm->EmitInt8(1); EOL("Default is_stmt_start flag");
-  Asm->EmitInt8(MinLineDelta); EOL("Line Base Value (Special Opcodes)");
-  Asm->EmitInt8(MaxLineDelta); EOL("Line Range Value (Special Opcodes)");
-  Asm->EmitInt8(-MinLineDelta); EOL("Special Opcode Base");
+  Asm->OutStreamer.AddComment("Minimum Instruction Length");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("Default is_stmt_start flag");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("Line Base Value (Special Opcodes)");
+  Asm->EmitInt8(MinLineDelta);
+  Asm->OutStreamer.AddComment("Line Range Value (Special Opcodes)");
+  Asm->EmitInt8(MaxLineDelta);
+  Asm->OutStreamer.AddComment("Special Opcode Base");
+  Asm->EmitInt8(-MinLineDelta);
 
   // Line number standard opcode encodings argument count
-  Asm->EmitInt8(0); EOL("DW_LNS_copy arg count");
-  Asm->EmitInt8(1); EOL("DW_LNS_advance_pc arg count");
-  Asm->EmitInt8(1); EOL("DW_LNS_advance_line arg count");
-  Asm->EmitInt8(1); EOL("DW_LNS_set_file arg count");
-  Asm->EmitInt8(1); EOL("DW_LNS_set_column arg count");
-  Asm->EmitInt8(0); EOL("DW_LNS_negate_stmt arg count");
-  Asm->EmitInt8(0); EOL("DW_LNS_set_basic_block arg count");
-  Asm->EmitInt8(0); EOL("DW_LNS_const_add_pc arg count");
-  Asm->EmitInt8(1); EOL("DW_LNS_fixed_advance_pc arg count");
+  Asm->OutStreamer.AddComment("DW_LNS_copy arg count");
+  Asm->EmitInt8(0);
+  Asm->OutStreamer.AddComment("DW_LNS_advance_pc arg count");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("DW_LNS_advance_line arg count");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("DW_LNS_set_file arg count");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("DW_LNS_set_column arg count");
+  Asm->EmitInt8(1);
+  Asm->OutStreamer.AddComment("DW_LNS_negate_stmt arg count");
+  Asm->EmitInt8(0);
+  Asm->OutStreamer.AddComment("DW_LNS_set_basic_block arg count");
+  Asm->EmitInt8(0);
+  Asm->OutStreamer.AddComment("DW_LNS_const_add_pc arg count");
+  Asm->EmitInt8(0);
+  Asm->OutStreamer.AddComment("DW_LNS_fixed_advance_pc arg count");
+  Asm->EmitInt8(1);
 
   // Emit directories.
   for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
@@ -2543,7 +2570,8 @@
     Asm->OutStreamer.EmitBytes(StringRef(Dir.c_str(), Dir.size()+1), 0);
   }
 
-  Asm->EmitInt8(0); EOL("End of directories");
+  Asm->OutStreamer.AddComment("End of directories");
+  Asm->EmitInt8(0);
 
   // Emit files.
   for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
@@ -2558,7 +2586,8 @@
     EmitULEB128(0, "File size");
   }
 
-  Asm->EmitInt8(0); EOL("End of files");
+  Asm->OutStreamer.AddComment("End of files");
+  Asm->EmitInt8(0);
 
   Asm->OutStreamer.EmitLabel(getTempLabel("line_prolog_end"));
 
@@ -2597,15 +2626,22 @@
       }
 
       // Define the line address.
-      Asm->EmitInt8(0); EOL("Extended Op");
-      Asm->EmitInt8(TD->getPointerSize() + 1); EOL("Op size");
-      Asm->EmitInt8(dwarf::DW_LNE_set_address); EOL("DW_LNE_set_address");
-      EmitReference(getDWLabel("label", LabelID)); EOL("Location label");
+      Asm->OutStreamer.AddComment("Extended Op");
+      Asm->EmitInt8(0);
+      Asm->OutStreamer.AddComment("Op size");
+      Asm->EmitInt8(TD->getPointerSize() + 1);
+
+      Asm->OutStreamer.AddComment("DW_LNE_set_address");
+      Asm->EmitInt8(dwarf::DW_LNE_set_address); 
+
+      Asm->OutStreamer.AddComment("Location label");
+      EmitReference(getDWLabel("label", LabelID)); 
 
       // If change of source, then switch to the new source.
       if (Source != LineInfo.getSourceID()) {
         Source = LineInfo.getSourceID();
-        Asm->EmitInt8(dwarf::DW_LNS_set_file); EOL("DW_LNS_set_file");
+        Asm->OutStreamer.AddComment("DW_LNS_set_file");
+        Asm->EmitInt8(dwarf::DW_LNS_set_file); 
         EmitULEB128(Source, "New Source");
       }
 
@@ -2621,17 +2657,20 @@
         // If delta is small enough and in range...
         if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
           // ... then use fast opcode.
-          Asm->EmitInt8(Delta - MinLineDelta); EOL("Line Delta");
+          Asm->OutStreamer.AddComment("Line Delta");
+          Asm->EmitInt8(Delta - MinLineDelta);
         } else {
           // ... otherwise use long hand.
+          Asm->OutStreamer.AddComment("DW_LNS_advance_line");
           Asm->EmitInt8(dwarf::DW_LNS_advance_line);
-          EOL("DW_LNS_advance_line");
           EmitSLEB128(Offset, "Line Offset");
-          Asm->EmitInt8(dwarf::DW_LNS_copy); EOL("DW_LNS_copy");
+          Asm->OutStreamer.AddComment("DW_LNS_copy");
+          Asm->EmitInt8(dwarf::DW_LNS_copy);
         }
       } else {
         // Copy the previous row (different address or source)
-        Asm->EmitInt8(dwarf::DW_LNS_copy); EOL("DW_LNS_copy");
+        Asm->OutStreamer.AddComment("DW_LNS_copy");
+        Asm->EmitInt8(dwarf::DW_LNS_copy);
       }
     }
 
@@ -2663,21 +2702,21 @@
                               Asm->getObjFileLowering().getDwarfFrameSection());
 
   Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common"));
+  Asm->OutStreamer.AddComment("Length of Common Information Entry");
   EmitDifference(getTempLabel("debug_frame_common_end"),
                  getTempLabel("debug_frame_common_begin"), true);
-  EOL("Length of Common Information Entry");
 
   Asm->OutStreamer.EmitLabel(getTempLabel("debug_frame_common_begin"));
+  Asm->OutStreamer.AddComment("CIE Identifier Tag");
   Asm->EmitInt32((int)dwarf::DW_CIE_ID);
-  EOL("CIE Identifier Tag");
+  Asm->OutStreamer.AddComment("CIE Version");
   Asm->EmitInt8(dwarf::DW_CIE_VERSION);
-  EOL("CIE Version");
+  Asm->OutStreamer.AddComment("CIE Augmentation");
   Asm->OutStreamer.EmitIntValue(0, 1, /*addrspace*/0); // nul terminator.
-  EOL("CIE Augmentation");
   EmitULEB128(1, "CIE Code Alignment Factor");
   EmitSLEB128(stackGrowth, "CIE Data Alignment Factor");
+  Asm->OutStreamer.AddComment("CIE RA Column");
   Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
-  EOL("CIE RA Column");
 
   std::vector<MachineMove> Moves;
   RI->getInitialFrameState(Moves);
@@ -2699,22 +2738,22 @@
   Asm->OutStreamer.SwitchSection(
                               Asm->getObjFileLowering().getDwarfFrameSection());
 
+  Asm->OutStreamer.AddComment("Length of Frame Information Entry");
   EmitDifference(getDWLabel("debug_frame_end", DebugFrameInfo.Number),
                  getDWLabel("debug_frame_begin", DebugFrameInfo.Number), true);
-  EOL("Length of Frame Information Entry");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("debug_frame_begin",
                                         DebugFrameInfo.Number));
 
+  Asm->OutStreamer.AddComment("FDE CIE offset");
   EmitSectionOffset(getTempLabel("debug_frame_common"),
                     getTempLabel("section_debug_frame"), true, false);
-  EOL("FDE CIE offset");
 
+  Asm->OutStreamer.AddComment("FDE initial location");
   EmitReference(getDWLabel("func_begin", DebugFrameInfo.Number));
-  EOL("FDE initial location");
+  Asm->OutStreamer.AddComment("FDE address range");
   EmitDifference(getDWLabel("func_end", DebugFrameInfo.Number),
                  getDWLabel("func_begin", DebugFrameInfo.Number));
-  EOL("FDE address range");
 
   EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
                  false);
@@ -2731,38 +2770,41 @@
   Asm->OutStreamer.SwitchSection(
                           Asm->getObjFileLowering().getDwarfPubNamesSection());
 
+  Asm->OutStreamer.AddComment("Length of Public Names Info");
   EmitDifference(getDWLabel("pubnames_end", ModuleCU->getID()),
                  getDWLabel("pubnames_begin", ModuleCU->getID()), true);
-  EOL("Length of Public Names Info");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("pubnames_begin", ModuleCU->getID()));
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("DWARF Version");
+  Asm->OutStreamer.AddComment("DWARF Version");
+  Asm->EmitInt16(dwarf::DWARF_VERSION); 
 
+  Asm->OutStreamer.AddComment("Offset of Compilation Unit Info");
   EmitSectionOffset(getDWLabel("info_begin", ModuleCU->getID()), 
                     getTempLabel("section_info"),
                     true, false);
-  EOL("Offset of Compilation Unit Info");
 
+  Asm->OutStreamer.AddComment("Compilation Unit Length");
   EmitDifference(getDWLabel("info_end", ModuleCU->getID()),
                  getDWLabel("info_begin", ModuleCU->getID()),
                  true);
-  EOL("Compilation Unit Length");
 
   const StringMap<DIE*> &Globals = ModuleCU->getGlobals();
   for (StringMap<DIE*>::const_iterator
          GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
     const char *Name = GI->getKeyData();
-    DIE * Entity = GI->second;
+    DIE *Entity = GI->second;
 
-    Asm->EmitInt32(Entity->getOffset()); EOL("DIE offset");
+    Asm->OutStreamer.AddComment("DIE offset");
+    Asm->EmitInt32(Entity->getOffset());
     
     if (Asm->VerboseAsm)
       Asm->OutStreamer.AddComment("External Name");
     Asm->OutStreamer.EmitBytes(StringRef(Name, strlen(Name)+1), 0);
   }
 
-  Asm->EmitInt32(0); EOL("End Mark");
+  Asm->OutStreamer.AddComment("End Mark");
+  Asm->EmitInt32(0);
   Asm->OutStreamer.EmitLabel(getDWLabel("pubnames_end", ModuleCU->getID()));
 }
 
@@ -2770,23 +2812,23 @@
   // Start the dwarf pubnames section.
   Asm->OutStreamer.SwitchSection(
                           Asm->getObjFileLowering().getDwarfPubTypesSection());
+  Asm->OutStreamer.AddComment("Length of Public Types Info");
   EmitDifference(getDWLabel("pubtypes_end", ModuleCU->getID()),
                  getDWLabel("pubtypes_begin", ModuleCU->getID()), true);
-  EOL("Length of Public Types Info");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("pubtypes_begin", ModuleCU->getID()));
 
   if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DWARF Version");
   Asm->EmitInt16(dwarf::DWARF_VERSION);
 
+  Asm->OutStreamer.AddComment("Offset of Compilation ModuleCU Info");
   EmitSectionOffset(getDWLabel("info_begin", ModuleCU->getID()),
                     getTempLabel("section_info"), true, false);
-  EOL("Offset of Compilation ModuleCU Info");
 
+  Asm->OutStreamer.AddComment("Compilation ModuleCU Length");
   EmitDifference(getDWLabel("info_end", ModuleCU->getID()),
                  getDWLabel("info_begin", ModuleCU->getID()),
                  true);
-  EOL("Compilation ModuleCU Length");
 
   const StringMap<DIE*> &Globals = ModuleCU->getGlobalTypes();
   for (StringMap<DIE*>::const_iterator
@@ -2801,7 +2843,8 @@
     Asm->OutStreamer.EmitBytes(StringRef(Name, GI->getKeyLength()+1), 0);
   }
 
-  Asm->EmitInt32(0); EOL("End Mark");
+  Asm->OutStreamer.AddComment("End Mark");
+  Asm->EmitInt32(0); 
   Asm->OutStreamer.EmitLabel(getDWLabel("pubtypes_end", ModuleCU->getID()));
 }
 
@@ -2840,34 +2883,6 @@
   // Start the dwarf aranges section.
   Asm->OutStreamer.SwitchSection(
                           Asm->getObjFileLowering().getDwarfARangesSection());
-
-  // FIXME - Mock up
-#if 0
-  CompileUnit *Unit = GetBaseCompileUnit();
-
-  // Don't include size of length
-  Asm->EmitInt32(0x1c); EOL("Length of Address Ranges Info");
-
-  Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("Dwarf Version");
-
-  EmitReference("info_begin", Unit->getID());
-  EOL("Offset of Compilation Unit Info");
-
-  Asm->EmitInt8(TD->getPointerSize()); EOL("Size of Address");
-
-  Asm->EmitInt8(0); EOL("Size of Segment Descriptor");
-
-  Asm->EmitInt16(0);  EOL("Pad (1)");
-  Asm->EmitInt16(0);  EOL("Pad (2)");
-
-  // Range 1
-  EmitReference("text_begin", 0); EOL("Address");
-  EmitDifference(getTempLabel("text_end"), getTempLabel("text_begin"),
-                 true); EOL("Length");
-
-  Asm->EmitInt32(0); EOL("EOM (1)");
-  Asm->EmitInt32(0); EOL("EOM (2)");
-#endif
 }
 
 /// emitDebugRanges - Emit visible names into a debug ranges section.
@@ -2916,14 +2931,16 @@
   Asm->OutStreamer.SwitchSection(
                         Asm->getObjFileLowering().getDwarfDebugInlineSection());
 
+  Asm->OutStreamer.AddComment("Length of Debug Inlined Information Entry");
   EmitDifference(getDWLabel("debug_inlined_end", 1),
                  getDWLabel("debug_inlined_begin", 1), true);
-  EOL("Length of Debug Inlined Information Entry");
 
   Asm->OutStreamer.EmitLabel(getDWLabel("debug_inlined_begin", 1));
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); EOL("Dwarf Version");
-  Asm->EmitInt8(TD->getPointerSize()); EOL("Address Size (in bytes)");
+  Asm->OutStreamer.AddComment("Dwarf Version");
+  Asm->EmitInt16(dwarf::DWARF_VERSION);
+  Asm->OutStreamer.AddComment("Address Size (in bytes)");
+  Asm->EmitInt8(TD->getPointerSize());
 
   for (SmallVector<MDNode *, 4>::iterator I = InlinedSPNodes.begin(),
          E = InlinedSPNodes.end(); I != E; ++I) {
@@ -2936,6 +2953,7 @@
     StringRef LName = SP.getLinkageName();
     StringRef Name = SP.getName();
 
+    Asm->OutStreamer.AddComment("MIPS linkage name");
     if (LName.empty()) {
       Asm->OutStreamer.EmitBytes(Name, 0);
       Asm->OutStreamer.EmitIntValue(0, 1, 0); // nul terminator.
@@ -2944,10 +2962,9 @@
                                    StringPool.idFor(getRealLinkageName(LName))),
                         getTempLabel("section_str"), true);
 
-    EOL("MIPS linkage name");
+    Asm->OutStreamer.AddComment("Function name");
     EmitSectionOffset(getDWLabel("string", StringPool.idFor(Name)),
                       getTempLabel("section_str"), false, true);
-    EOL("Function name");
     EmitULEB128(Labels.size(), "Inline count");
 
     for (SmallVector<InlineInfoLabels, 4>::iterator LI = Labels.begin(),

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98105&r1=98104&r2=98105&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Mar  9 17:52:58 2010
@@ -94,15 +94,15 @@
   Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common", Index));
 
   // Define the eh frame length.
+  Asm->OutStreamer.AddComment("Length of Common Information Entry");
   EmitDifference(getDWLabel("eh_frame_common_end", Index),
                  getDWLabel("eh_frame_common_begin", Index), true);
-  EOL("Length of Common Information Entry");
 
   // EH frame header.
   Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_common_begin", Index));
-  if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("CIE Identifier Tag");
+  Asm->OutStreamer.AddComment("CIE Identifier Tag");
   Asm->OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
-  if (Asm->VerboseAsm) Asm->OutStreamer.AddComment("DW_CIE_VERSION");
+  Asm->OutStreamer.AddComment("DW_CIE_VERSION");
   Asm->OutStreamer.EmitIntValue(dwarf::DW_CIE_VERSION, 1/*size*/, 0/*addr*/);
 
   // The personality presence indicates that language specific information will
@@ -138,14 +138,14 @@
   if (APtr != Augmentation + 1)
     Augmentation[0] = 'z';
 
+  Asm->OutStreamer.AddComment("CIE Augmentation");
   Asm->OutStreamer.EmitBytes(StringRef(Augmentation, strlen(Augmentation)+1),0);
-  EOL("CIE Augmentation");
 
   // Round out reader.
   EmitULEB128(1, "CIE Code Alignment Factor");
   EmitSLEB128(stackGrowth, "CIE Data Alignment Factor");
+  Asm->OutStreamer.AddComment("CIE Return Address Column");
   Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), true));
-  EOL("CIE Return Address Column");
 
   if (Augmentation[0]) {
     EmitULEB128(AugmentationSize, "Augmentation Size");
@@ -153,8 +153,8 @@
     // If there is a personality, we need to indicate the function's location.
     if (PersonalityFn) {
       EmitEncodingByte(PerEncoding, "Personality");
+      Asm->OutStreamer.AddComment("Personality");
       EmitReference(PersonalityFn, PerEncoding);
-      EOL("Personality");
     }
     if (UsesLSDA[Index])
       EmitEncodingByte(LSDAEncoding, "LSDA");
@@ -222,26 +222,26 @@
     O << *EHFrameInfo.FunctionEHSym << ":\n";
 
     // EH frame header.
+    Asm->OutStreamer.AddComment("Length of Frame Information Entry");
     EmitDifference(getDWLabel("eh_frame_end", EHFrameInfo.Number),
                    getDWLabel("eh_frame_begin", EHFrameInfo.Number),
                    true);
-    EOL("Length of Frame Information Entry");
 
     Asm->OutStreamer.EmitLabel(getDWLabel("eh_frame_begin",EHFrameInfo.Number));
 
+    Asm->OutStreamer.AddComment("FDE CIE offset");
     EmitSectionOffset(getDWLabel("eh_frame_begin", EHFrameInfo.Number),
                       getDWLabel("eh_frame_common",
                                  EHFrameInfo.PersonalityIndex),
                       true, true);
 
-    EOL("FDE CIE offset");
 
+    Asm->OutStreamer.AddComment("FDE initial location");
     EmitReference(getDWLabel("eh_func_begin", EHFrameInfo.Number), FDEEncoding);
-    EOL("FDE initial location");
+    Asm->OutStreamer.AddComment("FDE address range");
     EmitDifference(getDWLabel("eh_func_end", EHFrameInfo.Number),
                    getDWLabel("eh_func_begin", EHFrameInfo.Number),
                    SizeOfEncodedValue(FDEEncoding) == 4);
-    EOL("FDE address range");
 
     // If there is a personality and landing pads then point to the language
     // specific data area in the exception table.
@@ -249,12 +249,12 @@
       unsigned Size = SizeOfEncodedValue(LSDAEncoding);
 
       EmitULEB128(Size, "Augmentation size");
+      Asm->OutStreamer.AddComment("Language Specific Data Area");
       if (EHFrameInfo.hasLandingPads)
         EmitReference(getDWLabel("exception", EHFrameInfo.Number),LSDAEncoding);
       else
         Asm->OutStreamer.EmitIntValue(0, Size/*size*/, 0/*addrspace*/);
 
-      EOL("Language Specific Data Area");
     } else {
       EmitULEB128(0, "Augmentation size");
     }
@@ -819,11 +819,12 @@
       // Offset of the call site relative to the previous call site, counted in
       // number of 16-byte bundles. The first call site is counted relative to
       // the start of the procedure fragment.
+      Asm->OutStreamer.AddComment("Region start");
       EmitSectionOffset(getDWLabel(BeginTag, BeginNumber),
                         getDWLabel("eh_func_begin", SubprogramCount),
                         true, true);
-      EOL("Region start");
 
+      Asm->OutStreamer.AddComment("Region length");
       if (!S.EndLabel)
         EmitDifference(getDWLabel("eh_func_end", SubprogramCount),
                        getDWLabel(BeginTag, BeginNumber),
@@ -832,18 +833,16 @@
         EmitDifference(getDWLabel("label", S.EndLabel), 
                        getDWLabel(BeginTag, BeginNumber), true);
 
-      EOL("Region length");
 
       // Offset of the landing pad, counted in 16-byte bundles relative to the
       // @LPStart address.
+      Asm->OutStreamer.AddComment("Landing pad");
       if (!S.PadLabel) {
-        Asm->OutStreamer.AddComment("Landing pad");
         Asm->OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
       } else {
         EmitSectionOffset(getDWLabel("label", S.PadLabel),
                           getDWLabel("eh_func_begin", SubprogramCount),
                           true, true);
-        EOL("Landing pad");
       }
 
       // Offset of the first associated action record, relative to the start of
@@ -854,11 +853,16 @@
   }
 
   // Emit the Action Table.
-  if (Actions.size() != 0) EOL("-- Action Record Table --");
+  if (Actions.size() != 0) {
+    Asm->OutStreamer.AddComment("-- Action Record Table --");
+    Asm->OutStreamer.AddBlankLine();
+  }
+  
   for (SmallVectorImpl<ActionEntry>::const_iterator
          I = Actions.begin(), E = Actions.end(); I != E; ++I) {
     const ActionEntry &Action = *I;
-    EOL("Action Record:");
+    Asm->OutStreamer.AddComment("Action Record");
+    Asm->OutStreamer.AddBlankLine();
 
     // Type Filter
     //
@@ -874,23 +878,28 @@
   }
 
   // Emit the Catch TypeInfos.
-  if (!TypeInfos.empty()) EOL("-- Catch TypeInfos --");
+  if (!TypeInfos.empty()) {
+    Asm->OutStreamer.AddComment("-- Catch TypeInfos --");
+    Asm->OutStreamer.AddBlankLine();
+  }
   for (std::vector<GlobalVariable *>::const_reverse_iterator
          I = TypeInfos.rbegin(), E = TypeInfos.rend(); I != E; ++I) {
     const GlobalVariable *GV = *I;
 
     if (GV) {
+      Asm->OutStreamer.AddComment("TypeInfo");
       EmitReference(GV, TTypeEncoding);
-      EOL("TypeInfo");
     } else {
       PrintRelDirective(TTypeEncoding);
-      O << "0x0";
-      EOL("");
+      O << "0x0\n";
     }
   }
 
   // Emit the Exception Specifications.
-  if (!FilterIds.empty()) EOL("-- Filter IDs --");
+  if (!FilterIds.empty()) {
+    Asm->OutStreamer.AddComment("-- Filter IDs --");
+    Asm->OutStreamer.AddBlankLine();
+  }
   for (std::vector<unsigned>::const_iterator
          I = FilterIds.begin(), E = FilterIds.end(); I < E; ++I) {
     unsigned TypeID = *I;

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98105&r1=98104&r2=98105&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 17:52:58 2010
@@ -92,16 +92,6 @@
         MAI->getData32bitsDirective() : MAI->getData64bitsDirective());
 }
 
-/// EOL - Print a newline character to asm stream.  If a comment is present
-/// then it will be printed first.  Comments should not contain '\n'.
-void DwarfPrinter::EOL(const Twine &Comment) const {
-  if (Asm->VerboseAsm && !Comment.isTriviallyEmpty()) {
-    Asm->O.PadToColumn(MAI->getCommentColumn());
-    Asm->O << Asm->MAI->getCommentString() << ' ' << Comment;
-  }
-  Asm->O << '\n';
-}
-
 static const char *DecodeDWARFEncoding(unsigned Encoding) {
   switch (Encoding) {
   case dwarf::DW_EH_PE_absptr: return "absptr";

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98105&r1=98104&r2=98105&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Tue Mar  9 17:52:58 2010
@@ -91,10 +91,6 @@
   void PrintRelDirective(unsigned Encoding) const;
   void PrintRelDirective(bool Force32Bit = false) const;
 
-  /// EOL - Print a newline character to asm stream.  If a comment is present
-  /// then it will be printed first.  Comments should not contain '\n'.
-  void EOL(const Twine &Comment) const;
-  
   /// EmitEncodingByte - Emit a .byte 42 directive that corresponds to an
   /// encoding.  If verbose assembly output is enabled, we output comments
   /// describing the encoding.  Desc is a string saying what the encoding is




From sabre at nondot.org  Tue Mar  9 17:54:52 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 09 Mar 2010 23:54:52 -0000
Subject: [llvm-commits] [llvm] r98106 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfPrinter.cpp DwarfPrinter.h
Message-ID: <20100309235452.8454C2A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 17:54:52 2010
New Revision: 98106

URL: http://llvm.org/viewvc/llvm-project?rev=98106&view=rev
Log:
inline the bool form of PrintRelDirective away, leaving just the unsigned form.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98106&r1=98105&r2=98106&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 17:54:52 2010
@@ -77,13 +77,6 @@
   return 0;
 }
 
-void DwarfPrinter::PrintRelDirective(bool Force32Bit) const {
-  if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
-    O << MAI->getData32bitsDirective();
-  else
-    O << MAI->getData64bitsDirective();
-}
-
 void DwarfPrinter::PrintRelDirective(unsigned Encoding) const {
   unsigned Size = SizeOfEncodedValue(Encoding);
   assert((Size == 4 || Size == 8) && "Do not support other types or rels!");
@@ -211,7 +204,10 @@
   }
   
   // FIXME: Need an MCExpr for ".".
-  PrintRelDirective(Force32Bit);
+  if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
+    O << MAI->getData32bitsDirective();
+  else
+    O << MAI->getData64bitsDirective();
   O << *Sym;
   if (IsPCRelative) O << "-" << MAI->getPCSymbol();
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98106&r1=98105&r2=98106&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Tue Mar  9 17:54:52 2010
@@ -89,7 +89,6 @@
   unsigned SizeOfEncodedValue(unsigned Encoding) const;
 
   void PrintRelDirective(unsigned Encoding) const;
-  void PrintRelDirective(bool Force32Bit = false) const;
 
   /// EmitEncodingByte - Emit a .byte 42 directive that corresponds to an
   /// encoding.  If verbose assembly output is enabled, we output comments




From stoklund at 2pi.dk  Tue Mar  9 18:05:13 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 9 Mar 2010 16:05:13 -0800
Subject: [llvm-commits] [llvm] r98099 - in /llvm/trunk:
	include/llvm/Analysis/InlineCost.h
	include/llvm/Transforms/IPO/InlinerPass.h
	lib/Analysis/InlineCost.cpp lib/Transforms/IPO/InlineAlways.cpp
	lib/Transforms/IPO/InlineSimple.cpp lib/Transforms/IPO/Inl
In-Reply-To: <6a8523d61003091552h3171643hbc650c11250033f9@mail.gmail.com>
References: <6a8523d61003091552h3171643hbc650c11250033f9@mail.gmail.com>
Message-ID: <3A41FB74-153B-489B-B844-BB138D356016@2pi.dk>


On Mar 9, 2010, at 3:52 PM, Daniel Dunbar wrote:

> Hi Jakob,
> 
> On Tue, Mar 9, 2010 at 3:02 PM, Jakob Stoklund Olesen <stoklund at 2pi.dk> wrote:
>> Author: stoklund
>> Date: Tue Mar  9 17:02:17 2010
>> New Revision: 98099
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=98099&view=rev
>> Log:
>> Try to keep the cached inliner costs around for a bit longer for big functions.
>> 
>> The Caller cost info would be reset everytime a callee was inlined. If the
>> caller has lots of calls and there is some mutual recursion going on, the
>> caller cost info could be calculated many times.
>> 
>> This patch reduces inliner runtime from 240s to 0.5s for a function with 20000
>> small function calls.
>> 
>> This is a more conservative version of r98089 that doesn't break the clang
>> test CodeGenCXX/temp-order.cpp. That test relies on rather extreme inlining
>> for constant folding.
> 
> That test doesn't *have* to pass, and we can change it if need be.
> However, it sounds like this should just be a performance improvement
> not a functionality change, so I wouldn't expect it to break the test.

Yes, it caught me by surprise too.

That test relies on inlining and constant folding to reduce function calls to just a constant. My "performance improvement" didn't expect a function to be significantly smaller after inlining, and so it cached a very outdated cost.

It is possible that my patch will change the generated code in other cases that rely on massive constant folding. I hope that it will only do that in degenerate cases. The more conservative version only cheats for functions with more than 10 basic blocks and 1000 instructions. I would not normally expect such a big function to collapse into something that can be considered for inlining.

/jakob




From sabre at nondot.org  Tue Mar  9 18:09:21 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 00:09:21 -0000
Subject: [llvm-commits] [llvm] r98107 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfException.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100310000921.97B092A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 18:09:21 2010
New Revision: 98107

URL: http://llvm.org/viewvc/llvm-project?rev=98107&view=rev
Log:
mcize uses of PrintRelDirective and eliminate it.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98107&r1=98106&r2=98107&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Mar  9 18:09:21 2010
@@ -886,13 +886,11 @@
          I = TypeInfos.rbegin(), E = TypeInfos.rend(); I != E; ++I) {
     const GlobalVariable *GV = *I;
 
-    if (GV) {
-      Asm->OutStreamer.AddComment("TypeInfo");
+    Asm->OutStreamer.AddComment("TypeInfo");
+    if (GV)
       EmitReference(GV, TTypeEncoding);
-    } else {
-      PrintRelDirective(TTypeEncoding);
-      O << "0x0\n";
-    }
+    else
+      Asm->OutStreamer.EmitIntValue(0, SizeOfEncodedValue(TTypeEncoding), 0);
   }
 
   // Emit the Exception Specifications.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98107&r1=98106&r2=98107&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 18:09:21 2010
@@ -77,14 +77,6 @@
   return 0;
 }
 
-void DwarfPrinter::PrintRelDirective(unsigned Encoding) const {
-  unsigned Size = SizeOfEncodedValue(Encoding);
-  assert((Size == 4 || Size == 8) && "Do not support other types or rels!");
-
-  O << (Size == 4 ?
-        MAI->getData32bitsDirective() : MAI->getData64bitsDirective());
-}
-
 static const char *DecodeDWARFEncoding(unsigned Encoding) {
   switch (Encoding) {
   case dwarf::DW_EH_PE_absptr: return "absptr";
@@ -215,16 +207,16 @@
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
-  PrintRelDirective(Encoding);
-  O << *TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);
+  const MCExpr *Exp = TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);
+  Asm->OutStreamer.EmitValue(Exp, SizeOfEncodedValue(Encoding), /*addrspace*/0);
 }
 
 void DwarfPrinter::EmitReference(const GlobalValue *GV, unsigned Encoding)const{
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
-  PrintRelDirective(Encoding);
-  O << *TLOF.getSymbolForDwarfGlobalReference(GV, Asm->Mang,
-                                              Asm->MMI, Encoding);
+  const MCExpr *Exp =
+    TLOF.getSymbolForDwarfGlobalReference(GV, Asm->Mang, Asm->MMI, Encoding);
+  Asm->OutStreamer.EmitValue(Exp, SizeOfEncodedValue(Encoding), /*addrspace*/0);
 }
 
 /// EmitDifference - Emit the difference between two labels.  If this assembler

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98107&r1=98106&r2=98107&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Tue Mar  9 18:09:21 2010
@@ -88,8 +88,6 @@
   /// SizeOfEncodedValue - Return the size of the encoding in bytes.
   unsigned SizeOfEncodedValue(unsigned Encoding) const;
 
-  void PrintRelDirective(unsigned Encoding) const;
-
   /// EmitEncodingByte - Emit a .byte 42 directive that corresponds to an
   /// encoding.  If verbose assembly output is enabled, we output comments
   /// describing the encoding.  Desc is a string saying what the encoding is




From dalej at apple.com  Tue Mar  9 18:11:34 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 00:11:34 -0000
Subject: [llvm-commits] [llvm] r98108 -
	/llvm/trunk/lib/CodeGen/BranchFolding.cpp
Message-ID: <20100310001134.BC8E22A6C12C@llvm.org>

Author: johannes
Date: Tue Mar  9 18:11:34 2010
New Revision: 98108

URL: http://llvm.org/viewvc/llvm-project?rev=98108&view=rev
Log:
Speculatively revert 98104; could be what's causing crashes


Modified:
    llvm/trunk/lib/CodeGen/BranchFolding.cpp

Modified: llvm/trunk/lib/CodeGen/BranchFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/BranchFolding.cpp?rev=98108&r1=98107&r2=98108&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/BranchFolding.cpp (original)
+++ llvm/trunk/lib/CodeGen/BranchFolding.cpp Tue Mar  9 18:11:34 2010
@@ -347,29 +347,15 @@
     --I1; --I2;
     // Skip debugging pseudos; necessary to avoid changing the code.
     while (I1->isDebugValue()) {
-      if (I1==MBB1->begin()) {
-        while (I2->isDebugValue()) {
-          if (I2==MBB2->begin())
-            // I1==DBG at begin; I2==DBG at begin
-            return TailLen;
-          --I2;
-        }
-        ++I2;
-        // I1==DBG at begin; I2==non-DBG, or first of DBGs not at begin
+      if (I1==MBB1->begin())
         return TailLen;
-      }
       --I1;
     }
-    // I1==first (untested) non-DBG preceding known match
     while (I2->isDebugValue()) {
-      if (I2==MBB2->begin()) {
-        ++I1;
-        // I1==non-DBG, or first of DBGs not at begin; I2==DBG at begin
+      if (I2==MBB2->begin())
         return TailLen;
-      }
       --I2;
     }
-    // I1, I2==first (untested) non-DBGs preceding known match
     if (!I1->isIdenticalTo(I2) ||
         // FIXME: This check is dubious. It's used to get around a problem where
         // people incorrectly expect inline asm directives to remain in the same
@@ -382,29 +368,6 @@
     }
     ++TailLen;
   }
-  // Back past possible debugging pseudos at beginning of block.  This matters
-  // when one block differs from the other only by whether debugging pseudos
-  // are present at the beginning.  (This way, the various checks later for
-  // I1==MBB1->begin() work as expected.)
-  if (I1 == MBB1->begin() && I2 != MBB2->begin()) {
-    --I2;
-    while (I2->isDebugValue()) {
-      if (I2 == MBB2->begin()) {
-        return TailLen;
-        }
-      --I2;
-    }
-    ++I2;
-  }
-  if (I2 == MBB2->begin() && I1 != MBB1->begin()) {
-    --I1;
-    while (I1->isDebugValue()) {
-      if (I1 == MBB1->begin())
-        return TailLen;
-      --I1;
-    }
-    ++I1;
-  }
   return TailLen;
 }
 
@@ -1200,23 +1163,7 @@
       // be 'non-branch terminators' in the block, try removing the branch and
       // then seeing if the block is empty.
       TII->RemoveBranch(*MBB);
-      // If the only things remaining in the block are debug info, remove these
-      // as well, so this will behave the same as an empty block in non-debug
-      // mode.
-      if (!MBB->empty()) {
-        bool NonDebugInfoFound = false;
-        for (MachineBasicBlock::iterator I = MBB->begin(), E = MBB->end();
-             I != E; ++I) {
-          if (!I->isDebugValue()) {
-            NonDebugInfoFound = true;
-            break;
-          }
-        }
-        if (!NonDebugInfoFound)
-          // Make the block empty, losing the debug info (we could probably
-          // improve this in some cases.)
-          MBB->erase(MBB->begin(), MBB->end());
-      }
+
       // If this block is just an unconditional branch to CurTBB, we can
       // usually completely eliminate the block.  The only case we cannot
       // completely eliminate the block is when the block before this one




From grosbach at apple.com  Tue Mar  9 18:13:42 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 10 Mar 2010 00:13:42 -0000
Subject: [llvm-commits] [llvm] r98109 - in /llvm/trunk/lib:
 CodeGen/PrologEpilogInserter.cpp CodeGen/PrologEpilogInserter.h
 Target/ARM/Thumb1RegisterInfo.cpp
Message-ID: <20100310001342.598522A6C12C@llvm.org>

Author: grosbach
Date: Tue Mar  9 18:13:42 2010
New Revision: 98109

URL: http://llvm.org/viewvc/llvm-project?rev=98109&view=rev
Log:
Clear up the last (famous last words) frame index value reuse issues for Thumb1.

Modified:
    llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
    llvm/trunk/lib/CodeGen/PrologEpilogInserter.h
    llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp

Modified: llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp?rev=98109&r1=98108&r2=98109&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp (original)
+++ llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp Tue Mar  9 18:13:42 2010
@@ -57,6 +57,7 @@
   const TargetRegisterInfo *TRI = Fn.getTarget().getRegisterInfo();
   RS = TRI->requiresRegisterScavenging(Fn) ? new RegScavenger() : NULL;
   FrameIndexVirtualScavenging = TRI->requiresFrameIndexScavenging(Fn);
+  FrameConstantRegMap.clear();
 
   // Get MachineModuleInfo so that we can track the construction of the
   // frame.
@@ -693,8 +694,7 @@
             assert (FrameIndexVirtualScavenging &&
                     "Not scavenging, but virtual returned from "
                     "eliminateFrameIndex()!");
-            FrameConstantRegMap[VReg] = FrameConstantEntry(Value.second,
-                                                           SPAdj);
+            FrameConstantRegMap[VReg] = FrameConstantEntry(Value, SPAdj);
           }
 
           // Reset the iterator if we were at the beginning of the BB.
@@ -765,12 +765,12 @@
     unsigned CurrentVirtReg = 0;
     unsigned CurrentScratchReg = 0;
     bool havePrevValue = false;
-    int PrevValue = 0;
+    TargetRegisterInfo::FrameIndexValue PrevValue(0,0);
+    TargetRegisterInfo::FrameIndexValue Value(0,0);
     MachineInstr *PrevLastUseMI = NULL;
     unsigned PrevLastUseOp = 0;
     bool trackingCurrentValue = false;
     int SPAdj = 0;
-    int Value = 0;
 
     // The instruction stream may change in the loop, so check BB->end()
     // directly.
@@ -827,8 +827,11 @@
             if (trackingCurrentValue) {
               SPAdj = (*Entry).second.second;
               Value = (*Entry).second.first;
-            } else
-              SPAdj = Value = 0;
+            } else {
+              SPAdj = 0;
+              Value.first = 0;
+              Value.second = 0;
+            }
 
             // If the scratch register from the last allocation is still
             // available, see if the value matches. If it does, just re-use it.

Modified: llvm/trunk/lib/CodeGen/PrologEpilogInserter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/PrologEpilogInserter.h?rev=98109&r1=98108&r2=98109&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/PrologEpilogInserter.h (original)
+++ llvm/trunk/lib/CodeGen/PrologEpilogInserter.h Tue Mar  9 18:13:42 2010
@@ -102,7 +102,8 @@
     // When using the scavenger post-pass to resolve frame reference
     // materialization registers, maintain a map of the registers to
     // the constant value and SP adjustment associated with it.
-    typedef std::pair<int, int> FrameConstantEntry;
+    typedef std::pair<TargetRegisterInfo::FrameIndexValue, int>
+      FrameConstantEntry;
     DenseMap<unsigned, FrameConstantEntry> FrameConstantRegMap;
 
 #ifndef NDEBUG

Modified: llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp?rev=98109&r1=98108&r2=98109&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp Tue Mar  9 18:13:42 2010
@@ -642,7 +642,7 @@
       VReg = MF.getRegInfo().createVirtualRegister(ARM::tGPRRegisterClass);
       assert (Value && "Frame index virtual allocated, but Value arg is NULL!");
       bool UseRR = false;
-      bool TrackVReg = FrameReg == ARM::SP;
+      bool TrackVReg = true;
       Value->first = FrameReg; // use the frame register as a kind indicator
       Value->second = Offset;
 




From daniel at zuster.org  Tue Mar  9 18:58:25 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 00:58:25 -0000
Subject: [llvm-commits] [llvm] r98115 - in /llvm/trunk:
 lib/MC/MCAssembler.cpp test/MC/MachO/reloc-diff.s
Message-ID: <20100310005825.8534A2A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 18:58:25 2010
New Revision: 98115

URL: http://llvm.org/viewvc/llvm-project?rev=98115&view=rev
Log:
MC/Mach-O: Use the SECTDIFF relocation type for (A - B + constant) where A is external.
 - I'm not sure why, but this is what 'as' does.

Added:
    llvm/trunk/test/MC/MachO/reloc-diff.s
Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98115&r1=98114&r2=98115&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Tue Mar  9 18:58:25 2010
@@ -446,24 +446,27 @@
 
     // See <reloc.h>.
     const MCSymbol *A = Target.getSymA();
-    MCSymbolData *SD = SymbolMap.lookup(A);
+    MCSymbolData *A_SD = SymbolMap.lookup(A);
 
-    if (!SD->getFragment())
+    if (!A_SD->getFragment())
       llvm_report_error("symbol '" + A->getName() +
                         "' can not be undefined in a subtraction expression");
 
-    uint32_t Value = SD->getFragment()->getAddress() + SD->getOffset();
+    uint32_t Value = A_SD->getFragment()->getAddress() + A_SD->getOffset();
     uint32_t Value2 = 0;
 
     if (const MCSymbol *B = Target.getSymB()) {
-      MCSymbolData *SD = SymbolMap.lookup(B);
+      MCSymbolData *B_SD = SymbolMap.lookup(B);
 
-      if (!SD->getFragment())
+      if (!B_SD->getFragment())
         llvm_report_error("symbol '" + B->getName() +
                           "' can not be undefined in a subtraction expression");
 
-      Type = RIT_LocalDifference;
-      Value2 = SD->getFragment()->getAddress() + SD->getOffset();
+      // FIXME: This change of type based on the external bit doesn't make much
+      // sense, it seems to be redundant with the other information in the
+      // relocation entry.
+      Type = A_SD->isExternal() ? RIT_Difference : RIT_LocalDifference;
+      Value2 = B_SD->getFragment()->getAddress() + B_SD->getOffset();
     }
 
     // The value which goes in the fixup is current value of the expression.
@@ -488,7 +491,7 @@
     MRE.Word1 = Value;
     Relocs.push_back(MRE);
 
-    if (Type == RIT_LocalDifference) {
+    if (Type == RIT_Difference || Type == RIT_LocalDifference) {
       MachRelocationEntry MRE;
       MRE.Word0 = ((0         <<  0) |
                    (RIT_Pair  << 24) |

Added: llvm/trunk/test/MC/MachO/reloc-diff.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/reloc-diff.s?rev=98115&view=auto
==============================================================================
--- llvm/trunk/test/MC/MachO/reloc-diff.s (added)
+++ llvm/trunk/test/MC/MachO/reloc-diff.s Tue Mar  9 18:58:25 2010
@@ -0,0 +1,55 @@
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump | FileCheck %s
+
+// CHECK: # Relocation 0
+// CHECK: (('word-0', 0xa2000014),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 1
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 2
+// CHECK: (('word-0', 0xa4000010),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 3
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 4
+// CHECK: (('word-0', 0xa400000c),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 5
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 6
+// CHECK: (('word-0', 0xa4000008),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 7
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 8
+// CHECK: (('word-0', 0xa4000004),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 9
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 10
+// CHECK: (('word-0', 0xa2000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK: # Relocation 11
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x0)),
+// CHECK-NEXT: ])
+
+_local_def:
+        .globl _external_def
+_external_def:
+Ltemp:
+        ret
+        
+        .data
+        .long _external_def - _local_def
+        .long Ltemp - _local_def
+
+        .long _local_def - _external_def
+        .long Ltemp - _external_def
+
+        .long _local_def - Ltemp
+        .long _external_def - Ltemp




From sabre at nondot.org  Tue Mar  9 19:04:13 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 01:04:13 -0000
Subject: [llvm-commits] [llvm] r98116 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfException.cpp
 DwarfPrinter.cpp
Message-ID: <20100310010413.D9C932A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 19:04:13 2010
New Revision: 98116

URL: http://llvm.org/viewvc/llvm-project?rev=98116&view=rev
Log:
add some fixme's for MCizing.  EH still has a few things that
need to be MCized, but the last debug info thing are LEB and
cygwin specific (which the MC api doesn't support yet) and
one specific form of EmitReference which I'll tackle next.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98116&r1=98115&r2=98116&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Mar  9 19:04:13 2010
@@ -2598,12 +2598,6 @@
     // Isolate current sections line info.
     const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
 
-    /*if (Asm->isVerbose()) {
-      const MCSection *S = SectionMap[j + 1];
-      O << '\t' << MAI->getCommentString() << " Section"
-        << S->getName() << '\n';
-    }*/
-
     // Dwarf assumes we start with first line of first source file.
     unsigned Source = 1;
     unsigned Line = 1;
@@ -2617,12 +2611,10 @@
       if (LineInfo.getLine() == 0) continue;
 
       if (Asm->isVerbose()) {
-        std::pair<unsigned, unsigned> SourceID =
+        std::pair<unsigned, unsigned> SrcID =
           getSourceDirectoryAndFileIds(LineInfo.getSourceID());
-        O << '\t' << MAI->getCommentString() << ' '
-          << getSourceDirectoryName(SourceID.first) << '/'
-          << getSourceFileName(SourceID.second)
-          << ':' << utostr_32(LineInfo.getLine()) << '\n';
+        Asm->OutStreamer.AddComment(Twine(getSourceDirectoryName(SrcID.first)) +
+                                    ":" + Twine(LineInfo.getLine()));
       }
 
       // Define the line address.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98116&r1=98115&r2=98116&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Mar  9 19:04:13 2010
@@ -280,7 +280,7 @@
         Asm->OutStreamer.EmitSymbolAttribute(EHFrameInfo.FunctionEHSym,
                                              MCSA_NoDeadStrip);
   }
-  Asm->O << '\n';
+  Asm->OutStreamer.AddBlankLine();
 }
 
 /// SharedTypeIds - How many leading type ids two landing pads have in common.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98116&r1=98115&r2=98116&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 19:04:13 2010
@@ -138,6 +138,7 @@
     Asm->OutStreamer.AddComment(Desc);
     
   if (MAI->hasLEB128()) {
+    // FIXME: MCize.
     O << "\t.sleb128\t" << Value;
     Asm->OutStreamer.AddBlankLine();
     return;
@@ -163,6 +164,7 @@
     Asm->OutStreamer.AddComment(Desc);
  
   if (MAI->hasLEB128() && PadTo == 0) {
+    // FIXME: MCize.
     O << "\t.uleb128\t" << Value;
     Asm->OutStreamer.AddBlankLine();
     return;
@@ -196,6 +198,7 @@
   }
   
   // FIXME: Need an MCExpr for ".".
+  // FIXME: MCize.
   if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
     O << MAI->getData32bitsDirective();
   else
@@ -241,7 +244,8 @@
   
   // On COFF targets, we have to emit the weird .secrel32 directive.
   if (const char *SecOffDir = MAI->getDwarfSectionOffsetDirective())
-    O << SecOffDir << Label->getName();
+    // FIXME: MCize.
+    Asm->O << SecOffDir << Label->getName();
   else {
     unsigned Size = IsSmall ? 4 : TD->getPointerSize();
     Asm->OutStreamer.EmitSymbolValue(Label, Size, 0/*AddrSpace*/);




From sabre at nondot.org  Tue Mar  9 19:17:49 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 01:17:49 -0000
Subject: [llvm-commits] [llvm] r98117 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DIE.cpp DwarfDebug.cpp DwarfPrinter.cpp
 DwarfPrinter.h
Message-ID: <20100310011749.7E4C82A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 19:17:49 2010
New Revision: 98117

URL: http://llvm.org/viewvc/llvm-project?rev=98117&view=rev
Log:
inline away a form of IsPCRelative, eliminating the 
dead IsPCRel argument.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp?rev=98117&r1=98116&r2=98117&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.cpp Tue Mar  9 19:17:49 2010
@@ -259,7 +259,8 @@
 ///
 void DIELabel::EmitValue(DwarfPrinter *D, unsigned Form) const {
   bool IsSmall = Form == dwarf::DW_FORM_data4;
-  D->EmitReference(Label, false, IsSmall);
+  unsigned Size = IsSmall ? 4 : D->getTargetData()->getPointerSize();
+  D->getAsm()->OutStreamer.EmitSymbolValue(Label, Size, 0/*AddrSpace*/);
 }
 
 /// SizeOf - Determine size of label value in bytes.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98117&r1=98116&r2=98117&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Mar  9 19:17:49 2010
@@ -2493,7 +2493,9 @@
   Asm->EmitInt8(dwarf::DW_LNE_set_address);
 
   Asm->OutStreamer.AddComment("Section end label");
-  EmitReference(getDWLabel("section_end", SectionEnd));
+
+  Asm->OutStreamer.EmitSymbolValue(getDWLabel("section_end", SectionEnd),
+                                   TD->getPointerSize(), 0/*AddrSpace*/);
 
   // Mark end of matrix.
   Asm->OutStreamer.AddComment("DW_LNE_end_sequence");
@@ -2627,8 +2629,9 @@
       Asm->EmitInt8(dwarf::DW_LNE_set_address); 
 
       Asm->OutStreamer.AddComment("Location label");
-      EmitReference(getDWLabel("label", LabelID)); 
-
+      Asm->OutStreamer.EmitSymbolValue(getDWLabel("label", LabelID),
+                                       TD->getPointerSize(), 0/*AddrSpace*/);
+      
       // If change of source, then switch to the new source.
       if (Source != LineInfo.getSourceID()) {
         Source = LineInfo.getSourceID();
@@ -2742,7 +2745,12 @@
                     getTempLabel("section_debug_frame"), true, false);
 
   Asm->OutStreamer.AddComment("FDE initial location");
-  EmitReference(getDWLabel("func_begin", DebugFrameInfo.Number));
+  Asm->OutStreamer.EmitSymbolValue(getDWLabel("func_begin",
+                                              DebugFrameInfo.Number),
+                                   TD->getPointerSize(), 0/*AddrSpace*/);
+  
+  
+  
   Asm->OutStreamer.AddComment("FDE address range");
   EmitDifference(getDWLabel("func_end", DebugFrameInfo.Number),
                  getDWLabel("func_begin", DebugFrameInfo.Number));

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98117&r1=98116&r2=98117&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 19:17:49 2010
@@ -186,27 +186,6 @@
 }
 
 
-/// EmitReference - Emit a reference to a label.
-///
-void DwarfPrinter::EmitReference(const MCSymbol *Sym, bool IsPCRelative,
-                                 bool Force32Bit) const {
-  unsigned Size = Force32Bit ? 4 : TD->getPointerSize();
-  
-  if (!IsPCRelative) {
-    Asm->OutStreamer.EmitSymbolValue(Sym, Size, 0/*AddrSpace*/);
-    return;
-  }
-  
-  // FIXME: Need an MCExpr for ".".
-  // FIXME: MCize.
-  if (Force32Bit || TD->getPointerSize() == sizeof(int32_t))
-    O << MAI->getData32bitsDirective();
-  else
-    O << MAI->getData64bitsDirective();
-  O << *Sym;
-  if (IsPCRelative) O << "-" << MAI->getPCSymbol();
-}
-
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h?rev=98117&r1=98116&r2=98117&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.h Tue Mar  9 19:17:49 2010
@@ -108,9 +108,6 @@
   
   /// EmitReference - Emit a reference to a label.
   ///
-  void EmitReference(const MCSymbol *Label, bool IsPCRelative = false,
-                     bool Force32Bit = false) const;
-  
   void EmitReference(const MCSymbol *Sym, unsigned Encoding) const;
   void EmitReference(const GlobalValue *GV, unsigned Encoding) const;
 




From sabre at nondot.org  Tue Mar  9 19:29:27 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 01:29:27 -0000
Subject: [llvm-commits] [llvm] r98118 - in /llvm/trunk:
 include/llvm/MC/MCContext.h lib/MC/MCContext.cpp
 lib/MC/MCParser/AsmParser.cpp
Message-ID: <20100310012927.5DDCE2A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 19:29:27 2010
New Revision: 98118

URL: http://llvm.org/viewvc/llvm-project?rev=98118&view=rev
Log:
eliminate MCContext::CreateSymbol and CreateTemporarySymbol.
Add a new GetOrCreateTemporarySymbol method and a version that
takes a twine.

Modified:
    llvm/trunk/include/llvm/MC/MCContext.h
    llvm/trunk/lib/MC/MCContext.cpp
    llvm/trunk/lib/MC/MCParser/AsmParser.cpp

Modified: llvm/trunk/include/llvm/MC/MCContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCContext.h?rev=98118&r1=98117&r2=98118&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCContext.h (original)
+++ llvm/trunk/include/llvm/MC/MCContext.h Tue Mar  9 19:29:27 2010
@@ -46,11 +46,6 @@
     /// @name Symbol Managment
     /// @{
 
-    /// CreateSymbol - Create a new symbol with the specified @p Name.
-    ///
-    /// @param Name - The symbol name, which must be unique across all symbols.
-    MCSymbol *CreateSymbol(StringRef Name);
-
     /// GetOrCreateSymbol - Lookup the symbol inside with the specified
     /// @p Name.  If it exists, return it.  If not, create a forward
     /// reference and return it.
@@ -59,13 +54,15 @@
     MCSymbol *GetOrCreateSymbol(StringRef Name);
     MCSymbol *GetOrCreateSymbol(const Twine &Name);
 
-    /// CreateTemporarySymbol - Create a new temporary symbol with the specified
-    /// @p Name.
+    /// GetOrCreateTemporarySymbol - Create a new assembler temporary symbol
+    /// with the specified @p Name if it doesn't exist or return the existing
+    /// one if it does.
     ///
     /// @param Name - The symbol name, for debugging purposes only, temporary
     /// symbols do not surive assembly. If non-empty the name must be unique
     /// across all symbols.
-    MCSymbol *CreateTemporarySymbol(StringRef Name = "");
+    MCSymbol *GetOrCreateTemporarySymbol(StringRef Name = "");
+    MCSymbol *GetOrCreateTemporarySymbol(const Twine &Name);
 
     /// LookupSymbol - Get the symbol for \p Name, or null.
     MCSymbol *LookupSymbol(StringRef Name) const;

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=98118&r1=98117&r2=98118&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Tue Mar  9 19:29:27 2010
@@ -23,16 +23,8 @@
   // we don't need to free them here.
 }
 
-MCSymbol *MCContext::CreateSymbol(StringRef Name) {
-  assert(Name[0] != '\0' && "Normal symbols cannot be unnamed!");
-
-  // Create and bind the symbol, and ensure that names are unique.
-  MCSymbol *&Entry = Symbols[Name];
-  assert(!Entry && "Duplicate symbol definition!");
-  return Entry = new (*this) MCSymbol(Name, false);
-}
-
 MCSymbol *MCContext::GetOrCreateSymbol(StringRef Name) {
+  assert(!Name.empty() && "Normal symbols cannot be unnamed!");
   MCSymbol *&Entry = Symbols[Name];
   if (Entry) return Entry;
 
@@ -46,17 +38,24 @@
 }
 
 
-MCSymbol *MCContext::CreateTemporarySymbol(StringRef Name) {
+MCSymbol *MCContext::GetOrCreateTemporarySymbol(StringRef Name) {
   // If unnamed, just create a symbol.
   if (Name.empty())
     new (*this) MCSymbol("", true);
     
   // Otherwise create as usual.
   MCSymbol *&Entry = Symbols[Name];
-  assert(!Entry && "Duplicate symbol definition!");
+  if (Entry) return Entry;
   return Entry = new (*this) MCSymbol(Name, true);
 }
 
+MCSymbol *MCContext::GetOrCreateTemporarySymbol(const Twine &Name) {
+  SmallString<128> NameSV;
+  Name.toVector(NameSV);
+  return GetOrCreateTemporarySymbol(NameSV.str());
+}
+
+
 MCSymbol *MCContext::LookupSymbol(StringRef Name) const {
   return Symbols.lookup(Name);
 }

Modified: llvm/trunk/lib/MC/MCParser/AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCParser/AsmParser.cpp?rev=98118&r1=98117&r2=98118&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCParser/AsmParser.cpp (original)
+++ llvm/trunk/lib/MC/MCParser/AsmParser.cpp Tue Mar  9 19:29:27 2010
@@ -240,14 +240,10 @@
 }
 
 MCSymbol *AsmParser::CreateSymbol(StringRef Name) {
-  if (MCSymbol *S = Ctx.LookupSymbol(Name))
-    return S;
-
   // If the label starts with L it is an assembler temporary label.
   if (Name.startswith("L"))
-    return Ctx.CreateTemporarySymbol(Name);
-
-  return Ctx.CreateSymbol(Name);
+    return Ctx.GetOrCreateTemporarySymbol(Name);
+  return Ctx.GetOrCreateSymbol(Name);
 }
 
 /// ParsePrimaryExpr - Parse a primary expression and return it.




From gohman at apple.com  Tue Mar  9 19:29:40 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 01:29:40 -0000
Subject: [llvm-commits] [llvm] r98119 -
	/llvm/trunk/docs/WritingAnLLVMPass.html
Message-ID: <20100310012940.32CE82A6C12C@llvm.org>

Author: djg
Date: Tue Mar  9 19:29:39 2010
New Revision: 98119

URL: http://llvm.org/viewvc/llvm-project?rev=98119&view=rev
Log:
Clarify the documentation for MachineFunctionPasses.

Modified:
    llvm/trunk/docs/WritingAnLLVMPass.html

Modified: llvm/trunk/docs/WritingAnLLVMPass.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/WritingAnLLVMPass.html?rev=98119&r1=98118&r2=98119&view=diff
==============================================================================
--- llvm/trunk/docs/WritingAnLLVMPass.html (original)
+++ llvm/trunk/docs/WritingAnLLVMPass.html Tue Mar  9 19:29:39 2010
@@ -909,16 +909,22 @@
 
 <p>A <tt>MachineFunctionPass</tt> is a part of the LLVM code generator that
 executes on the machine-dependent representation of each LLVM function in the
-program.  A <tt>MachineFunctionPass</tt> is also a <tt>FunctionPass</tt>, so all
+program.</p>
+
+<p>Code generator passes are registered and initialized specially by
+<tt>TargetMachine::addPassesToEmitFile</tt> and similar routines, so they
+cannot generally be run from the <tt>opt</tt> or <tt>bugpoint</tt>
+commands.</p>
+
+<p>A <tt>MachineFunctionPass</tt> is also a <tt>FunctionPass</tt>, so all
 the restrictions that apply to a <tt>FunctionPass</tt> also apply to it.
 <tt>MachineFunctionPass</tt>es also have additional restrictions. In particular,
 <tt>MachineFunctionPass</tt>es are not allowed to do any of the following:</p>
 
 <ol>
-<li>Modify any LLVM Instructions, BasicBlocks or Functions.</li>
+<li>Modify or create any LLVM IR Instructions, BasicBlocks, Arguments,
+    Functions, GlobalVariables, GlobalAliases, or Modules.</li>
 <li>Modify a MachineFunction other than the one currently being processed.</li>
-<li>Add or remove MachineFunctions from the current Module.</li>
-<li>Add or remove global variables from the current Module.</li>
 <li>Maintain state across invocations of <a
 href="#runOnMachineFunction"><tt>runOnMachineFunction</tt></a> (including global
 data)</li>




From eli.friedman at gmail.com  Tue Mar  9 19:45:34 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Tue, 9 Mar 2010 17:45:34 -0800
Subject: [llvm-commits] [llvm] r98116 - in
	/llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp
	DwarfException.cpp DwarfPrinter.cpp
In-Reply-To: <20100310010413.D9C932A6C12C@llvm.org>
References: <20100310010413.D9C932A6C12C@llvm.org>
Message-ID: <f5aa3e9b1003091745i5184ffaemff06cb06451296cb@mail.gmail.com>

On Tue, Mar 9, 2010 at 5:04 PM, Chris Lattner <sabre at nondot.org> wrote:
> @@ -2617,12 +2611,10 @@
> ? ? ? if (LineInfo.getLine() == 0) continue;
>
> ? ? ? if (Asm->isVerbose()) {
> - ? ? ? ?std::pair<unsigned, unsigned> SourceID =
> + ? ? ? ?std::pair<unsigned, unsigned> SrcID =
> ? ? ? ? ? getSourceDirectoryAndFileIds(LineInfo.getSourceID());
> - ? ? ? ?O << '\t' << MAI->getCommentString() << ' '
> - ? ? ? ? ?<< getSourceDirectoryName(SourceID.first) << '/'
> - ? ? ? ? ?<< getSourceFileName(SourceID.second)
> - ? ? ? ? ?<< ':' << utostr_32(LineInfo.getLine()) << '\n';
> + ? ? ? ?Asm->OutStreamer.AddComment(Twine(getSourceDirectoryName(SrcID.first)) +
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?":" + Twine(LineInfo.getLine()));

Did you mean to eliminate the call to getSourceFileName()?

-Eli



From daniel at zuster.org  Tue Mar  9 20:10:29 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 02:10:29 -0000
Subject: [llvm-commits] [llvm] r98120 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100310021029.785272A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 20:10:29 2010
New Revision: 98120

URL: http://llvm.org/viewvc/llvm-project?rev=98120&view=rev
Log:
MC/Mach-O: Resolve a FIXME; these relocation types are no longer semanticaly different.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98120&r1=98119&r2=98120&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Tue Mar  9 20:10:29 2010
@@ -462,9 +462,11 @@
         llvm_report_error("symbol '" + B->getName() +
                           "' can not be undefined in a subtraction expression");
 
-      // FIXME: This change of type based on the external bit doesn't make much
-      // sense, it seems to be redundant with the other information in the
-      // relocation entry.
+      // Select the appropriate difference relocation type.
+      //
+      // Note that there is no longer any semantic difference between these two
+      // relocation types from the linkers point of view, this is done solely
+      // for pedantic compatibility with 'as'.
       Type = A_SD->isExternal() ? RIT_Difference : RIT_LocalDifference;
       Value2 = B_SD->getFragment()->getAddress() + B_SD->getOffset();
     }




From evan.cheng at apple.com  Tue Mar  9 20:12:03 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 02:12:03 -0000
Subject: [llvm-commits] [llvm] r98121 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100310021203.527752A6C12C@llvm.org>

Author: evancheng
Date: Tue Mar  9 20:12:03 2010
New Revision: 98121

URL: http://llvm.org/viewvc/llvm-project?rev=98121&view=rev
Log:
Add a couple more heuristics to neuter machine cse some more.

1. Be careful with cse "cheap" expressions. e.g. constant materialization. Only cse them when the common expression is local or in a direct predecessor. We don't want cse of cheap instruction causing other expressions to be spilled.
2. Watch out for the case where the expression doesn't itself uses a virtual register. e.g. lea of frame object. If the common expression itself is used by copies (common for passing addresses to function calls), don't perform the cse. Since these expressions do not use a register, it creates a live range but doesn't close any, we want to be very careful with increasing register pressure.

Note these are heuristics so machine cse doesn't make register allocator unhappy. Once we have proper live range splitting and re-materialization support in place, these should be evaluated again.

Now machine cse is almost always a win on llvm nightly tests on x86 and x86_64.

Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98121&r1=98120&r2=98121&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Tue Mar  9 20:12:03 2010
@@ -61,7 +61,8 @@
                                 MachineBasicBlock::const_iterator E);
     bool hasLivePhysRegDefUse(MachineInstr *MI, MachineBasicBlock *MBB);
     bool isCSECandidate(MachineInstr *MI);
-    bool isProfitableToCSE(unsigned Reg, MachineInstr *MI);
+    bool isProfitableToCSE(unsigned CSReg, unsigned Reg,
+                           MachineInstr *CSMI, MachineInstr *MI);
     bool ProcessBlock(MachineDomTreeNode *Node);
   };
 } // end anonymous namespace
@@ -143,6 +144,8 @@
   return false;
 }
 
+/// hasLivePhysRegDefUse - Return true if the specified instruction read / write
+/// physical registers (except for dead defs of physical registers).
 bool MachineCSE::hasLivePhysRegDefUse(MachineInstr *MI, MachineBasicBlock *MBB){
   unsigned PhysDef = 0;
   for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
@@ -177,16 +180,19 @@
   return false;
 }
 
+static bool isCopy(const MachineInstr *MI, const TargetInstrInfo *TII) {
+  unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
+  return TII->isMoveInstr(*MI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) ||
+    MI->isExtractSubreg() || MI->isInsertSubreg() || MI->isSubregToReg();
+}
+
 bool MachineCSE::isCSECandidate(MachineInstr *MI) {
   if (MI->isLabel() || MI->isPHI() || MI->isImplicitDef() ||
       MI->isKill() || MI->isInlineAsm())
     return false;
 
-  // Ignore copies or instructions that read / write physical registers
-  // (except for dead defs of physical registers).
-  unsigned SrcReg, DstReg, SrcSubIdx, DstSubIdx;
-  if (TII->isMoveInstr(*MI, SrcReg, DstReg, SrcSubIdx, DstSubIdx) ||
-      MI->isExtractSubreg() || MI->isInsertSubreg() || MI->isSubregToReg())
+  // Ignore copies.
+  if (isCopy(MI, TII))
     return false;
 
   // Ignore stuff that we obviously can't move.
@@ -210,14 +216,52 @@
 
 /// isProfitableToCSE - Return true if it's profitable to eliminate MI with a
 /// common expression that defines Reg.
-bool MachineCSE::isProfitableToCSE(unsigned Reg, MachineInstr *MI) {
-  // FIXME: This "heuristic" works around the lack the live range splitting.
-  // If the common subexpression is used by PHIs, do not reuse it unless the
-  // defined value is already used in the BB of the new use.
+bool MachineCSE::isProfitableToCSE(unsigned CSReg, unsigned Reg,
+                                   MachineInstr *CSMI, MachineInstr *MI) {
+  // FIXME: Heuristics that works around the lack the live range splitting.
+
+  // Heuristics #1: Don't cse "cheap" computating if the def is not local or in an
+  // immediate predecessor. We don't want to increase register pressure and end up
+  // causing other computation to be spilled.
+  if (MI->getDesc().isAsCheapAsAMove()) {
+    MachineBasicBlock *CSBB = CSMI->getParent();
+    MachineBasicBlock *BB = MI->getParent();
+    if (CSBB != BB && 
+        find(CSBB->succ_begin(), CSBB->succ_end(), BB) == CSBB->succ_end())
+      return false;
+  }
+
+  // Heuristics #2: If the expression doesn't not use a vr and the only use
+  // of the redundant computation are copies, do not cse.
+  bool HasVRegUse = false;
+  for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
+    const MachineOperand &MO = MI->getOperand(i);
+    if (MO.isReg() && MO.isUse() && MO.getReg() &&
+        TargetRegisterInfo::isVirtualRegister(MO.getReg())) {
+      HasVRegUse = true;
+      break;
+    }
+  }
+  if (!HasVRegUse) {
+    bool HasNonCopyUse = false;
+    for (MachineRegisterInfo::use_nodbg_iterator I =  MRI->use_nodbg_begin(Reg),
+           E = MRI->use_nodbg_end(); I != E; ++I) {
+      MachineInstr *Use = &*I;
+      // Ignore copies.
+      if (!isCopy(Use, TII)) {
+        HasNonCopyUse = true;
+        break;
+      }
+    }
+    if (!HasNonCopyUse)
+      return false;
+  }
+
+  // Heuristics #3: If the common subexpression is used by PHIs, do not reuse
+  // it unless the defined value is already used in the BB of the new use.
   bool HasPHI = false;
   SmallPtrSet<MachineBasicBlock*, 4> CSBBs;
-  for (MachineRegisterInfo::use_nodbg_iterator I = 
-       MRI->use_nodbg_begin(Reg),
+  for (MachineRegisterInfo::use_nodbg_iterator I =  MRI->use_nodbg_begin(CSReg),
        E = MRI->use_nodbg_end(); I != E; ++I) {
     MachineInstr *Use = &*I;
     HasPHI |= Use->isPHI();
@@ -282,7 +326,7 @@
       assert(TargetRegisterInfo::isVirtualRegister(OldReg) &&
              TargetRegisterInfo::isVirtualRegister(NewReg) &&
              "Do not CSE physical register defs!");
-      if (!isProfitableToCSE(NewReg, MI)) {
+      if (!isProfitableToCSE(NewReg, OldReg, CSMI, MI)) {
         DoCSE = false;
         break;
       }




From gohman at apple.com  Tue Mar  9 20:18:48 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 02:18:48 -0000
Subject: [llvm-commits] [llvm] r98122 -
	/llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp
Message-ID: <20100310021848.7C45B2A6C12C@llvm.org>

Author: djg
Date: Tue Mar  9 20:18:48 2010
New Revision: 98122

URL: http://llvm.org/viewvc/llvm-project?rev=98122&view=rev
Log:
Fix a comment.

Modified:
    llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp

Modified: llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp?rev=98122&r1=98121&r2=98122&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp Tue Mar  9 20:18:48 2010
@@ -132,7 +132,7 @@
   bool Changed = false;
 ReprocessLoop:
 
-  // Check to see that no blocks (other than the header) in this loop that has
+  // Check to see that no blocks (other than the header) in this loop have
   // predecessors that are not in the loop.  This is not valid for natural
   // loops, but can occur if the blocks are unreachable.  Since they are
   // unreachable we can just shamelessly delete those CFG edges!




From sabre at nondot.org  Tue Mar  9 20:25:11 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 02:25:11 -0000
Subject: [llvm-commits] [llvm] r98124 - in /llvm/trunk: include/llvm/MC/
 lib/CodeGen/ lib/CodeGen/AsmPrinter/ lib/MC/ lib/Target/
 lib/Target/ARM/AsmPrinter/ lib/Target/MSP430/AsmPrinter/
 lib/Target/PowerPC/AsmPrinter/ lib/Target/X86/ lib/Target/X86/AsmPrinter/
Message-ID: <20100310022512.11AE92A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 20:25:11 2010
New Revision: 98124

URL: http://llvm.org/viewvc/llvm-project?rev=98124&view=rev
Log:
set the temporary bit on MCSymbols correctly.

Modified:
    llvm/trunk/include/llvm/MC/MCExpr.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp
    llvm/trunk/lib/CodeGen/MachineFunction.cpp
    llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/trunk/lib/MC/MCExpr.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMMCInstLower.cpp
    llvm/trunk/lib/Target/MSP430/AsmPrinter/MSP430MCInstLower.cpp
    llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp

Modified: llvm/trunk/include/llvm/MC/MCExpr.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCExpr.h?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCExpr.h (original)
+++ llvm/trunk/include/llvm/MC/MCExpr.h Tue Mar  9 20:25:11 2010
@@ -127,6 +127,10 @@
 
   static const MCSymbolRefExpr *Create(const MCSymbol *Symbol, MCContext &Ctx);
   static const MCSymbolRefExpr *Create(StringRef Name, MCContext &Ctx);
+  
+  /// CreateTemp - Create a reference to an assembler temporary label with the
+  /// specified name.
+  static const MCSymbolRefExpr *CreateTemp(StringRef Name, MCContext &Ctx);
 
   /// @}
   /// @name Accessors

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Tue Mar  9 20:25:11 2010
@@ -911,8 +911,8 @@
 
   // Otherwise, emit with .set (aka assignment).
   MCSymbol *SetLabel =
-    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) + "set" +
-                                 Twine(SetCounter++));
+    OutContext.GetOrCreateTemporarySymbol(Twine(MAI->getPrivateGlobalPrefix()) +
+                                          "set" + Twine(SetCounter++));
   OutStreamer.EmitAssignment(SetLabel, Diff);
   OutStreamer.EmitSymbolValue(SetLabel, Size, 0/*AddrSpace*/);
 }
@@ -1553,15 +1553,15 @@
 /// exception handling tables.
 void AsmPrinter::printLabelInst(const MachineInstr *MI) const {
   MCSymbol *Sym = 
-    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) +
+    OutContext.GetOrCreateTemporarySymbol(Twine(MAI->getPrivateGlobalPrefix()) +
                                  "label" + Twine(MI->getOperand(0).getImm()));
   OutStreamer.EmitLabel(Sym);
 }
 
 void AsmPrinter::printLabel(unsigned Id) const {
   MCSymbol *Sym = 
-    OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix()) +
-                                 "label" + Twine(Id));
+    OutContext.GetOrCreateTemporarySymbol(Twine(MAI->getPrivateGlobalPrefix()) +
+                                          "label" + Twine(Id));
   OutStreamer.EmitLabel(Sym);
 }
 
@@ -1603,15 +1603,14 @@
                           "_" + FnName.str() + "_" + BB->getName(), 
                           Mangler::Private);
 
-  return OutContext.GetOrCreateSymbol(NameResult.str());
+  return OutContext.GetOrCreateTemporarySymbol(NameResult.str());
 }
 
 /// GetCPISymbol - Return the symbol for the specified constant pool entry.
 MCSymbol *AsmPrinter::GetCPISymbol(unsigned CPID) const {
-  SmallString<60> Name;
-  raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix() << "CPI"
-    << getFunctionNumber() << '_' << CPID;
-  return OutContext.GetOrCreateSymbol(Name.str());
+  return OutContext.GetOrCreateTemporarySymbol
+    (Twine(MAI->getPrivateGlobalPrefix()) + "CPI" + Twine(getFunctionNumber())
+     + "_" + Twine(CPID));
 }
 
 /// GetJTISymbol - Return the symbol for the specified jump table entry.
@@ -1622,10 +1621,9 @@
 /// GetJTSetSymbol - Return the symbol for the specified jump table .set
 /// FIXME: privatize to AsmPrinter.
 MCSymbol *AsmPrinter::GetJTSetSymbol(unsigned UID, unsigned MBBID) const {
-  SmallString<60> Name;
-  raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix()
-    << getFunctionNumber() << '_' << UID << "_set_" << MBBID;
-  return OutContext.GetOrCreateSymbol(Name.str());
+  return OutContext.GetOrCreateTemporarySymbol
+  (Twine(MAI->getPrivateGlobalPrefix()) + Twine(getFunctionNumber()) + "_" +
+   Twine(UID) + "_set_" + Twine(MBBID));
 }
 
 /// GetGlobalValueSymbol - Return the MCSymbol for the specified global
@@ -1633,7 +1631,10 @@
 MCSymbol *AsmPrinter::GetGlobalValueSymbol(const GlobalValue *GV) const {
   SmallString<60> NameStr;
   Mang->getNameWithPrefix(NameStr, GV, false);
-  return OutContext.GetOrCreateSymbol(NameStr.str());
+  
+  if (!GV->hasPrivateLinkage())
+    return OutContext.GetOrCreateSymbol(NameStr.str());
+  return OutContext.GetOrCreateTemporarySymbol(NameStr.str());
 }
 
 /// GetSymbolWithGlobalValueBase - Return the MCSymbol for a symbol with
@@ -1645,7 +1646,9 @@
   SmallString<60> NameStr;
   Mang->getNameWithPrefix(NameStr, GV, ForcePrivate);
   NameStr.append(Suffix.begin(), Suffix.end());
-  return OutContext.GetOrCreateSymbol(NameStr.str());
+  if (!GV->hasPrivateLinkage() && !ForcePrivate)
+    return OutContext.GetOrCreateSymbol(NameStr.str());
+  return OutContext.GetOrCreateTemporarySymbol(NameStr.str());
 }
 
 /// GetExternalSymbolSymbol - Return the MCSymbol for the specified

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Mar  9 20:25:11 2010
@@ -60,7 +60,7 @@
   raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix()
                             << LabelName << Asm->getFunctionNumber()
                             << "_" << Index;
-  MCSymbol *DotSym = Asm->OutContext.GetOrCreateSymbol(Name.str());
+  MCSymbol *DotSym = Asm->OutContext.GetOrCreateTemporarySymbol(Name.str());
   Asm->OutStreamer.EmitLabel(DotSym);
 
   return MCBinaryExpr::CreateSub(ExprRef,

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Tue Mar  9 20:25:11 2010
@@ -45,15 +45,15 @@
   
   //assert(ID && "Should use getTempLabel if no ID");
   if (ID == 0) return getTempLabel(Name);
-  return Asm->OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix())
-                                           + Twine(Name) + Twine(ID));
+  return Asm->OutContext.GetOrCreateTemporarySymbol
+        (Twine(MAI->getPrivateGlobalPrefix()) + Twine(Name) + Twine(ID));
 }
 
 /// getTempLabel - Return the MCSymbol corresponding to the assembler temporary
 /// label with the specified name.
 MCSymbol *DwarfPrinter::getTempLabel(const char *Name) const {
-  return Asm->OutContext.GetOrCreateSymbol(Twine(MAI->getPrivateGlobalPrefix())
-                                           + Name);
+  return Asm->OutContext.GetOrCreateTemporarySymbol
+     (Twine(MAI->getPrivateGlobalPrefix()) + Name);
 }
 
 

Modified: llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineBasicBlock.cpp Tue Mar  9 20:25:11 2010
@@ -42,12 +42,11 @@
 /// getSymbol - Return the MCSymbol for this basic block.
 ///
 MCSymbol *MachineBasicBlock::getSymbol(MCContext &Ctx) const {
-  SmallString<60> Name;
   const MachineFunction *MF = getParent();
-  raw_svector_ostream(Name)
-    << MF->getTarget().getMCAsmInfo()->getPrivateGlobalPrefix() << "BB"
-    << MF->getFunctionNumber() << '_' << getNumber();
-  return Ctx.GetOrCreateSymbol(Name.str());
+  const char *Prefix = MF->getTarget().getMCAsmInfo()->getPrivateGlobalPrefix();
+  return Ctx.GetOrCreateTemporarySymbol(Twine(Prefix) + "BB" +
+                                        Twine(MF->getFunctionNumber()) + "_" +
+                                        Twine(getNumber()));
 }
 
 

Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Tue Mar  9 20:25:11 2010
@@ -464,7 +464,9 @@
   SmallString<60> Name;
   raw_svector_ostream(Name)
     << Prefix << "JTI" << getFunctionNumber() << '_' << JTI;
-  return Ctx.GetOrCreateSymbol(Name.str());
+  if (isLinkerPrivate)
+    return Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateTemporarySymbol(Name.str());
 }
 
 

Modified: llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Tue Mar  9 20:25:11 2010
@@ -403,12 +403,15 @@
 
     // Add information about the stub reference to ELFMMI so that the stub
     // gets emitted by the asmprinter.
-    MCSymbol *Sym = getContext().GetOrCreateSymbol(Name.str());
+    MCSymbol *Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
     MCSymbol *&StubSym = ELFMMI.getGVStubEntry(Sym);
     if (StubSym == 0) {
       Name.clear();
       Mang->getNameWithPrefix(Name, GV, false);
-      StubSym = getContext().GetOrCreateSymbol(Name.str());
+      if (GV->hasPrivateLinkage())
+        StubSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+      else
+        StubSym = getContext().GetOrCreateSymbol(Name.str());
     }
 
     return TargetLoweringObjectFile::
@@ -749,12 +752,15 @@
 
     // Add information about the stub reference to MachOMMI so that the stub
     // gets emitted by the asmprinter.
-    MCSymbol *Sym = getContext().GetOrCreateSymbol(Name.str());
+    MCSymbol *Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
     MCSymbol *&StubSym = MachOMMI.getGVStubEntry(Sym);
     if (StubSym == 0) {
       Name.clear();
       Mang->getNameWithPrefix(Name, GV, false);
-      StubSym = getContext().GetOrCreateSymbol(Name.str());
+      if (GV->hasPrivateLinkage())
+        StubSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+      else
+        StubSym = getContext().GetOrCreateSymbol(Name.str());
     }
 
     return TargetLoweringObjectFile::

Modified: llvm/trunk/lib/MC/MCExpr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCExpr.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCExpr.cpp (original)
+++ llvm/trunk/lib/MC/MCExpr.cpp Tue Mar  9 20:25:11 2010
@@ -133,6 +133,11 @@
   return Create(Ctx.GetOrCreateSymbol(Name), Ctx);
 }
 
+const MCSymbolRefExpr *MCSymbolRefExpr::CreateTemp(StringRef Name,
+                                                   MCContext &Ctx) {
+  return Create(Ctx.GetOrCreateTemporarySymbol(Name), Ctx);
+}
+
 void MCTargetExpr::Anchor() {}
 
 /* *** */

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Tue Mar  9 20:25:11 2010
@@ -849,7 +849,7 @@
   raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix()
     << getFunctionNumber() << '_' << uid << '_' << uid2
     << "_set_" << MBB->getNumber();
-  return OutContext.GetOrCreateSymbol(Name.str());
+  return OutContext.GetOrCreateTemporarySymbol(Name.str());
 }
 
 MCSymbol *ARMAsmPrinter::
@@ -857,7 +857,7 @@
   SmallString<60> Name;
   raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix() << "JTI"
     << getFunctionNumber() << '_' << uid << '_' << uid2;
-  return OutContext.GetOrCreateSymbol(Name.str());
+  return OutContext.GetOrCreateTemporarySymbol(Name.str());
 }
 
 void ARMAsmPrinter::printJTBlockOperand(const MachineInstr *MI, int OpNum) {
@@ -1192,7 +1192,7 @@
     // FIXME: MOVE TO SHARED PLACE.
     unsigned Id = (unsigned)MI->getOperand(2).getImm();
     const char *Prefix = MAI->getPrivateGlobalPrefix();
-    MCSymbol *Label =OutContext.GetOrCreateSymbol(Twine(Prefix)
+    MCSymbol *Label =OutContext.GetOrCreateTemporarySymbol(Twine(Prefix)
                          + "PC" + Twine(getFunctionNumber()) + "_" + Twine(Id));
     OutStreamer.EmitLabel(Label);
     

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMMCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMMCInstLower.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMMCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMMCInstLower.cpp Tue Mar  9 20:25:11 2010
@@ -74,7 +74,7 @@
 #endif
   
   // Create a symbol for the name.
-  return Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateTemporarySymbol(Name.str());
 }
 
 MCSymbol *ARMMCInstLower::
@@ -90,7 +90,7 @@
 #endif
   
   // Create a symbol for the name.
-  return Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateTemporarySymbol(Name.str());
 }
   
 MCOperand ARMMCInstLower::

Modified: llvm/trunk/lib/Target/MSP430/AsmPrinter/MSP430MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/MSP430/AsmPrinter/MSP430MCInstLower.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/MSP430/AsmPrinter/MSP430MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/MSP430/AsmPrinter/MSP430MCInstLower.cpp Tue Mar  9 20:25:11 2010
@@ -58,7 +58,7 @@
   }
 
   // Create a symbol for the name.
-  return Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateTemporarySymbol(Name.str());
 }
 
 MCSymbol *MSP430MCInstLower::
@@ -74,7 +74,7 @@
   }
 
   // Create a symbol for the name.
-  return Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateTemporarySymbol(Name.str());
 }
 
 MCOperand MSP430MCInstLower::

Modified: llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp Tue Mar  9 20:25:11 2010
@@ -309,8 +309,8 @@
       const MCSymbol *&TOCEntry = TOC[Sym];
       if (TOCEntry == 0)
         TOCEntry = OutContext.
-          GetOrCreateSymbol(StringRef(MAI->getPrivateGlobalPrefix()) + "C" +
-                            Twine(LabelID++));
+          GetOrCreateTemporarySymbol(StringRef(MAI->getPrivateGlobalPrefix()) +
+                                     "C" + Twine(LabelID++));
 
       O << *TOCEntry << "@toc";
     }
@@ -672,14 +672,14 @@
   // Remove $stub suffix, add $lazy_ptr.
   SmallString<128> TmpStr(Sym->getName().begin(), Sym->getName().end()-5);
   TmpStr += "$lazy_ptr";
-  return Ctx.GetOrCreateSymbol(TmpStr.str());
+  return Ctx.GetOrCreateTemporarySymbol(TmpStr.str());
 }
 
 static const MCSymbol *GetAnonSym(const MCSymbol *Sym, MCContext &Ctx) {
   // Add $tmp suffix to $stub, yielding $stub$tmp.
   SmallString<128> TmpStr(Sym->getName().begin(), Sym->getName().end());
   TmpStr += "$tmp";
-  return Ctx.GetOrCreateSymbol(TmpStr.str());
+  return Ctx.GetOrCreateTemporarySymbol(TmpStr.str());
 }
 
 void PPCDarwinAsmPrinter::

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Tue Mar  9 20:25:11 2010
@@ -295,7 +295,12 @@
   // FIXME: Use GetGlobalValueSymbol.
   SmallString<128> Name;
   Mang->getNameWithPrefix(Name, GV, false);
-  const MCSymbol *Sym = getContext().GetOrCreateSymbol(Name.str());
+  const MCSymbol *Sym;
+  
+  if (GV->hasPrivateLinkage())
+    Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
+  else
+    Sym = getContext().GetOrCreateSymbol(Name.str());
 
   return getSymbolForDwarfReference(Sym, MMI, Encoding);
 }

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Tue Mar  9 20:25:11 2010
@@ -56,7 +56,11 @@
 MCSymbol *X86AsmPrinter::GetGlobalValueSymbol(const GlobalValue *GV) const {
   SmallString<60> NameStr;
   Mang->getNameWithPrefix(NameStr, GV, false);
-  MCSymbol *Symb = OutContext.GetOrCreateSymbol(NameStr.str());
+  MCSymbol *Symb;
+  if (GV->hasPrivateLinkage())
+    Symb = OutContext.GetOrCreateTemporarySymbol(NameStr.str());
+  else
+    Symb = OutContext.GetOrCreateSymbol(NameStr.str());
 
   if (Subtarget->isTargetCygMing()) {
     X86COFFMachineModuleInfo &COFFMMI =

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Tue Mar  9 20:25:11 2010
@@ -89,7 +89,7 @@
   case X86II::MO_DARWIN_NONLAZY:
   case X86II::MO_DARWIN_NONLAZY_PIC_BASE: {
     Name += "$non_lazy_ptr";
-    MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str());
+    MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
 
     MCSymbol *&StubSym = getMachOMMI().getGVStubEntry(Sym);
     if (StubSym == 0) {
@@ -100,7 +100,7 @@
   }
   case X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE: {
     Name += "$non_lazy_ptr";
-    MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str());
+    MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
     MCSymbol *&StubSym = getMachOMMI().getHiddenGVStubEntry(Sym);
     if (StubSym == 0) {
       assert(MO.isGlobal() && "Extern symbol not handled yet");
@@ -110,7 +110,7 @@
   }
   case X86II::MO_DARWIN_STUB: {
     Name += "$stub";
-    MCSymbol *Sym = Ctx.GetOrCreateSymbol(Name.str());
+    MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
     MCSymbol *&StubSym = getMachOMMI().getFnStubEntry(Sym);
     if (StubSym)
       return Sym;
@@ -119,7 +119,7 @@
       StubSym = AsmPrinter.GetGlobalValueSymbol(MO.getGlobal());
     } else {
       Name.erase(Name.end()-5, Name.end());
-      StubSym = Ctx.GetOrCreateSymbol(Name.str());
+      StubSym = Ctx.GetOrCreateTemporarySymbol(Name.str());
     }
     return Sym;
   }
@@ -394,7 +394,8 @@
     // However, we can't generate a ".", so just emit a new label here and refer
     // to it.  We know that this operand flag occurs at most once per function.
     const char *Prefix = MAI->getPrivateGlobalPrefix();
-    MCSymbol *DotSym = OutContext.GetOrCreateSymbol(Twine(Prefix)+"picbaseref"+
+    MCSymbol *DotSym = OutContext.GetOrCreateTemporarySymbol(Twine(Prefix)+
+                                                             "picbaseref" +
                                                     Twine(getFunctionNumber()));
     OutStreamer.EmitLabel(DotSym);
     

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Mar  9 20:25:11 2010
@@ -1105,8 +1105,8 @@
 X86TargetLowering::getPICBaseSymbol(const MachineFunction *MF,
                                     MCContext &Ctx) const {
   const MCAsmInfo &MAI = *getTargetMachine().getMCAsmInfo();
-  return Ctx.GetOrCreateSymbol(Twine(MAI.getPrivateGlobalPrefix())+
-                               Twine(MF->getFunctionNumber())+"$pb");
+  return Ctx.GetOrCreateTemporarySymbol(Twine(MAI.getPrivateGlobalPrefix())+
+                                        Twine(MF->getFunctionNumber())+"$pb");
 }
 
 

Modified: llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp?rev=98124&r1=98123&r2=98124&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp Tue Mar  9 20:25:11 2010
@@ -20,14 +20,19 @@
 
 const MCExpr *X8664_MachoTargetObjectFile::
 getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                           MachineModuleInfo *MMI, unsigned Encoding) const {
+                                 MachineModuleInfo *MMI, 
+                                 unsigned Encoding) const {
 
   // On Darwin/X86-64, we can reference dwarf symbols with foo at GOTPCREL+4, which
   // is an indirect pc-relative reference.
   if (Encoding & (DW_EH_PE_indirect | DW_EH_PE_pcrel)) {
     SmallString<128> Name;
     Mang->getNameWithPrefix(Name, GV, false);
-    const MCSymbol *Sym = getContext().GetOrCreateSymbol(Name);
+    const MCSymbol *Sym;
+    if (GV->hasPrivateLinkage())
+      Sym = getContext().GetOrCreateTemporarySymbol(Name);
+    else
+      Sym = getContext().GetOrCreateSymbol(Name);
     const MCExpr *Res =
       X86MCTargetExpr::Create(Sym, X86MCTargetExpr::GOTPCREL, getContext());
     const MCExpr *Four = MCConstantExpr::Create(4, getContext());




From sabre at nondot.org  Tue Mar  9 20:29:31 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 02:29:31 -0000
Subject: [llvm-commits] [llvm] r98125 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100310022931.9F3DD2A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 20:29:31 2010
New Revision: 98125

URL: http://llvm.org/viewvc/llvm-project?rev=98125&view=rev
Log:
add missing filename!

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98125&r1=98124&r2=98125&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Mar  9 20:29:31 2010
@@ -2616,6 +2616,8 @@
         std::pair<unsigned, unsigned> SrcID =
           getSourceDirectoryAndFileIds(LineInfo.getSourceID());
         Asm->OutStreamer.AddComment(Twine(getSourceDirectoryName(SrcID.first)) +
+                                    "/" +
+                                    Twine(getSourceFileName(SrcID.second)) +
                                     ":" + Twine(LineInfo.getLine()));
       }
 




From clattner at apple.com  Tue Mar  9 20:30:03 2010
From: clattner at apple.com (Chris Lattner)
Date: Tue, 9 Mar 2010 18:30:03 -0800
Subject: [llvm-commits] [llvm] r98116 - in
	/llvm/trunk/lib/CodeGen/AsmPrinter:
	DwarfDebug.cpp	DwarfException.cpp DwarfPrinter.cpp
In-Reply-To: <f5aa3e9b1003091745i5184ffaemff06cb06451296cb@mail.gmail.com>
References: <20100310010413.D9C932A6C12C@llvm.org>
	<f5aa3e9b1003091745i5184ffaemff06cb06451296cb@mail.gmail.com>
Message-ID: <B1F9F17D-29B0-4E00-ABCB-D35BE14B4231@apple.com>


On Mar 9, 2010, at 5:45 PM, Eli Friedman wrote:

> On Tue, Mar 9, 2010 at 5:04 PM, Chris Lattner <sabre at nondot.org> wrote:
>> @@ -2617,12 +2611,10 @@
>>       if (LineInfo.getLine() == 0) continue;
>> 
>>       if (Asm->isVerbose()) {
>> -        std::pair<unsigned, unsigned> SourceID =
>> +        std::pair<unsigned, unsigned> SrcID =
>>           getSourceDirectoryAndFileIds(LineInfo.getSourceID());
>> -        O << '\t' << MAI->getCommentString() << ' '
>> -          << getSourceDirectoryName(SourceID.first) << '/'
>> -          << getSourceFileName(SourceID.second)
>> -          << ':' << utostr_32(LineInfo.getLine()) << '\n';
>> +        Asm->OutStreamer.AddComment(Twine(getSourceDirectoryName(SrcID.first)) +
>> +                                    ":" + Twine(LineInfo.getLine()));
> 
> Did you mean to eliminate the call to getSourceFileName()?

No I didn't, great catch, thanks!

-Chris


From daniel at zuster.org  Tue Mar  9 20:32:23 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 02:32:23 -0000
Subject: [llvm-commits] [zorg] r98126 -
	/zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py
Message-ID: <20100310023223.1DB4A2A6C12C@llvm.org>

Author: ddunbar
Date: Tue Mar  9 20:32:22 2010
New Revision: 98126

URL: http://llvm.org/viewvc/llvm-project?rev=98126&view=rev
Log:
NightlyTestCommand: Don't fail the nightly test step if 'make' fails, the
makefiles tend to fail once in a while.

Modified:
    zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py

Modified: zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py?rev=98126&r1=98125&r2=98126&view=diff
==============================================================================
--- zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py (original)
+++ zorg/trunk/zorg/buildbot/commands/NightlyTestCommand.py Tue Mar  9 20:32:22 2010
@@ -16,8 +16,10 @@
 
     def evaluateCommand(self, cmd):
         # Always fail if the command itself failed.
-        if cmd.rc != 0:
-            return builder.FAILURE
+        #
+        # Disabled for now, nightlytest is so broken.
+        #if cmd.rc != 0:
+        #    return buildbot.status.builder.FAILURE
     
         failures = {}
         xfailures = {}




From alenhar2 at llvm.org  Tue Mar  9 20:33:42 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Wed, 10 Mar 2010 02:33:42 -0000
Subject: [llvm-commits] [poolalloc] r98128 - in /poolalloc/trunk:
 include/dsa/DSNode.h lib/DSA/Local.cpp
Message-ID: <20100310023343.048CF2A6C12D@llvm.org>

Author: alenhar2
Date: Tue Mar  9 20:33:42 2010
New Revision: 98128

URL: http://llvm.org/viewvc/llvm-project?rev=98128&view=rev
Log:
Handle aliases as normal variables

Modified:
    poolalloc/trunk/include/dsa/DSNode.h
    poolalloc/trunk/lib/DSA/Local.cpp

Modified: poolalloc/trunk/include/dsa/DSNode.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSNode.h?rev=98128&r1=98127&r2=98128&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSNode.h (original)
+++ poolalloc/trunk/include/dsa/DSNode.h Tue Mar  9 20:33:42 2010
@@ -278,6 +278,8 @@
   ///
   void addGlobal(const GlobalValue *GV);
 
+  void addFunction(const Function* F);
+
   /// removeGlobal - Remove the specified global that is explicitly in the
   /// globals list.
   void removeGlobal(const GlobalValue *GV);

Modified: poolalloc/trunk/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Local.cpp?rev=98128&r1=98127&r2=98128&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Local.cpp (original)
+++ poolalloc/trunk/lib/DSA/Local.cpp Tue Mar  9 20:33:42 2010
@@ -234,19 +234,19 @@
   // Check first for constant expressions that must be traversed to
   // extract the actual value.
   DSNode* N;
-  if (GlobalVariable* GV = dyn_cast<GlobalVariable>(V)) {
+  if (Function * F = dyn_cast<Function > (V)) {
+    // Create a new global node for this function.
+    N = createNode();
+    N->addFunction(F);
+    if (F->isDeclaration())
+      N->setExternFuncMarker();
+  } else if (GlobalValue * GV = dyn_cast<GlobalValue > (V)) {
     // Create a new global node for this global variable.
     N = createNode();
     N->addGlobal(GV);
     if (GV->isDeclaration())
       N->setExternGlobalMarker();
-  } else if (Function* GV = dyn_cast<Function>(V)) {
-    // Create a new global node for this function.
-    N = createNode();
-    N->addGlobal(GV);
-    if (GV->isDeclaration())
-      N->setExternFuncMarker();
-  } else if (Constant *C = dyn_cast<Constant>(V)) {
+  } else if (Constant * C = dyn_cast<Constant > (V)) {
     if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
       if (CE->isCast()) {
         if (isa<PointerType>(CE->getOperand(0)->getType()))
@@ -269,13 +269,6 @@
     } else if (isa<UndefValue>(C)) {
       G.eraseNodeForValue(V);
       return 0;
-    } else if (isa<GlobalAlias>(C)) {
-      // XXX: Need more investigation
-      // According to Andrew, DSA is broken on global aliasing, since it does
-      // not handle the aliases of parameters correctly. Here is only a quick
-      // fix for some special cases.
-      NH = getValueDest(cast<GlobalAlias>(C)->getAliasee());
-      return NH;
     } else {
       errs() << "Unknown constant: " << *C << "\n";
       assert(0 && "Unknown constant type!");




From sabre at nondot.org  Tue Mar  9 20:48:06 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 02:48:06 -0000
Subject: [llvm-commits] [llvm] r98130 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
Message-ID: <20100310024806.B02E12A6C12C@llvm.org>

Author: lattner
Date: Tue Mar  9 20:48:06 2010
New Revision: 98130

URL: http://llvm.org/viewvc/llvm-project?rev=98130&view=rev
Log:
mcize the rest of EH emission, only one more directive missing
for darwin/x86 to be completely mcized.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98130&r1=98129&r2=98130&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Tue Mar  9 20:48:06 2010
@@ -84,9 +84,13 @@
   // Begin eh frame section.
   Asm->OutStreamer.SwitchSection(TLOF.getEHFrameSection());
 
+  MCSymbol *EHFrameSym;
   if (MAI->is_EHSymbolPrivate())
-    O << MAI->getPrivateGlobalPrefix();
-  O << "EH_frame" << Index << ":\n";
+    EHFrameSym = getDWLabel("EH_frame", Index);
+  else
+    EHFrameSym = Asm->OutContext.GetOrCreateSymbol(Twine("EH_frame") + 
+                                                   Twine(Index));
+  Asm->OutStreamer.EmitLabel(EHFrameSym);
   
   Asm->OutStreamer.EmitLabel(getDWLabel("section_eh_frame", Index));
 
@@ -195,7 +199,8 @@
 
   // If corresponding function is weak definition, this should be too.
   if (TheFunc->isWeakForLinker() && MAI->getWeakDefDirective())
-    O << MAI->getWeakDefDirective() << *EHFrameInfo.FunctionEHSym << '\n';
+    Asm->OutStreamer.EmitSymbolAttribute(EHFrameInfo.FunctionEHSym,
+                                         MCSA_WeakDefinition);
 
   // If corresponding function is hidden, this should be too.
   if (TheFunc->hasHiddenVisibility())
@@ -211,7 +216,8 @@
       (!TheFunc->isWeakForLinker() ||
        !MAI->getWeakDefDirective() ||
        MAI->getSupportsWeakOmittedEHFrame())) {
-    O << *EHFrameInfo.FunctionEHSym << " = 0\n";
+    Asm->OutStreamer.EmitAssignment(EHFrameInfo.FunctionEHSym,
+                                    MCConstantExpr::Create(0, Asm->OutContext));
     // This name has no connection to the function, so it might get
     // dead-stripped when the function is not, erroneously.  Prohibit
     // dead-stripping unconditionally.
@@ -219,7 +225,7 @@
       Asm->OutStreamer.EmitSymbolAttribute(EHFrameInfo.FunctionEHSym,
                                            MCSA_NoDeadStrip);
   } else {
-    O << *EHFrameInfo.FunctionEHSym << ":\n";
+    Asm->OutStreamer.EmitLabel(EHFrameInfo.FunctionEHSym);
 
     // EH frame header.
     Asm->OutStreamer.AddComment("Length of Frame Information Entry");
@@ -699,15 +705,14 @@
   Asm->EmitAlignment(2, 0, 0, false);
 
   // Emit the LSDA.
-  O << "GCC_except_table" << SubprogramCount << ":\n";
+  MCSymbol *GCCETSym = 
+    Asm->OutContext.GetOrCreateSymbol(Twine("GCC_except_table")+
+                                      Twine(SubprogramCount));
+  Asm->OutStreamer.EmitLabel(GCCETSym);
   Asm->OutStreamer.EmitLabel(getDWLabel("exception", SubprogramCount));
 
-  if (IsSJLJ) {
-    SmallString<16> LSDAName;
-    raw_svector_ostream(LSDAName) << MAI->getPrivateGlobalPrefix() <<
-      "_LSDA_" << Asm->getFunctionNumber();
-    O << LSDAName.str() << ":\n";
-  }
+  if (IsSJLJ)
+    Asm->OutStreamer.EmitLabel(getDWLabel("_LSDA_", Asm->getFunctionNumber()));
 
   // Emit the LSDA header.
   EmitEncodingByte(dwarf::DW_EH_PE_omit, "@LPStart");




From evan.cheng at apple.com  Tue Mar  9 21:07:42 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 03:07:42 -0000
Subject: [llvm-commits] [llvm] r98132 - in /llvm/trunk:
 lib/CodeGen/LLVMTargetMachine.cpp test/CodeGen/ARM/indirectbr.ll
 test/CodeGen/PowerPC/indirectbr.ll test/CodeGen/Thumb/2009-08-20-ISelBug.ll
 test/CodeGen/Thumb2/machine-licm.ll
 test/CodeGen/X86/2007-10-16-IllegalAsm.ll
 test/CodeGen/X86/2009-08-06-inlineasm.ll test/CodeGen/X86/machine-cse.ll
 test/CodeGen/X86/pre-split6.ll test/CodeGen/X86/stack-color-with-reg.ll
Message-ID: <20100310030742.4024A2A6C12C@llvm.org>

Author: evancheng
Date: Tue Mar  9 21:07:41 2010
New Revision: 98132

URL: http://llvm.org/viewvc/llvm-project?rev=98132&view=rev
Log:
Enable machine cse pass.

Added:
    llvm/trunk/test/CodeGen/X86/machine-cse.ll
Removed:
    llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll
Modified:
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/test/CodeGen/ARM/indirectbr.ll
    llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll
    llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll
    llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll
    llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll
    llvm/trunk/test/CodeGen/X86/pre-split6.ll
    llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Tue Mar  9 21:07:41 2010
@@ -67,9 +67,6 @@
     cl::desc("Verify generated machine code"),
     cl::init(getenv("LLVM_VERIFY_MACHINEINSTRS")!=NULL));
 
-static cl::opt<bool> EnableMachineCSE("enable-machine-cse", cl::Hidden,
-    cl::desc("Enable Machine CSE"));
-
 static cl::opt<cl::boolOrDefault>
 AsmVerbose("asm-verbose", cl::desc("Add comments to directives."),
            cl::init(cl::BOU_UNSET));
@@ -328,8 +325,7 @@
     PM.add(createOptimizeExtsPass());
     if (!DisableMachineLICM)
       PM.add(createMachineLICMPass());
-    if (EnableMachineCSE)
-      PM.add(createMachineCSEPass());
+    PM.add(createMachineCSEPass());
     if (!DisableMachineSink)
       PM.add(createMachineSinkingPass());
     printAndVerify(PM, "After Machine LICM, CSE and Sinking passes",

Modified: llvm/trunk/test/CodeGen/ARM/indirectbr.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/indirectbr.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/ARM/indirectbr.ll (original)
+++ llvm/trunk/test/CodeGen/ARM/indirectbr.ll Tue Mar  9 21:07:41 2010
@@ -15,14 +15,14 @@
 ; indirect branch gets duplicated here
 ; ARM: bx
 ; THUMB: mov pc, r1
-; THUMB2: mov pc, r1
+; THUMB2: mov pc, r2
   br i1 %1, label %bb3, label %bb2
 
 bb2:                                              ; preds = %entry, %bb3
   %gotovar.4.0 = phi i8* [ %gotovar.4.0.pre, %bb3 ], [ %0, %entry ] ; <i8*> [#uses=1]
 ; ARM: bx
 ; THUMB: mov pc, r1
-; THUMB2: mov pc, r1
+; THUMB2: mov pc, r2
   indirectbr i8* %gotovar.4.0, [label %L5, label %L4, label %L3, label %L2, label %L1]
 
 bb3:                                              ; preds = %entry

Modified: llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll (original)
+++ llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll Tue Mar  9 21:07:41 2010
@@ -43,13 +43,13 @@
 
 L1:                                               ; preds = %L2, %bb2
   %res.3 = phi i32 [ %phitmp, %L2 ], [ 2, %bb2 ]  ; <i32> [#uses=1]
-; PIC: addis r5, r4, ha16(L_BA4__foo_L5-"L1$pb")
+; PIC: addis r4, r4, ha16(L_BA4__foo_L5-"L1$pb")
 ; PIC: li r6, lo16(L_BA4__foo_L5-"L1$pb")
-; PIC: add r5, r5, r6
-; PIC: stw r5
-; STATIC: li r4, lo16(L_BA4__foo_L5)
-; STATIC: addis r4, r4, ha16(L_BA4__foo_L5)
-; STATIC: stw r4
+; PIC: add r4, r4, r6
+; PIC: stw r4
+; STATIC: li r5, lo16(L_BA4__foo_L5)
+; STATIC: addis r5, r5, ha16(L_BA4__foo_L5)
+; STATIC: stw r5
   store i8* blockaddress(@foo, %L5), i8** @nextaddr, align 4
   ret i32 %res.3
 }

Modified: llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll (original)
+++ llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll Tue Mar  9 21:07:41 2010
@@ -11,7 +11,7 @@
 
 define arm_apcscc i32 @t(%struct.asl_file_t* %s, i64 %off, i64* %out) nounwind optsize {
 ; CHECK: t:
-; CHECK: adds r3, #8
+; CHECK: adds r0, #8
 entry:
   %val = alloca i64, align 4                      ; <i64*> [#uses=3]
   %0 = icmp eq %struct.asl_file_t* %s, null       ; <i1> [#uses=1]

Modified: llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll (original)
+++ llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll Tue Mar  9 21:07:41 2010
@@ -18,9 +18,8 @@
 bb.nph:                                           ; preds = %entry
 ; CHECK: BB#1
 ; CHECK: ldr.n r2, LCPI1_0
-; CHECK: ldr r3, [r2]
-; CHECK: ldr r3, [r3]
 ; CHECK: ldr r2, [r2]
+; CHECK: ldr r3, [r2]
 ; CHECK: LBB1_2
 ; CHECK: LCPI1_0:
 ; CHECK-NOT: LCPI1_1:
@@ -29,9 +28,8 @@
 ; PIC: BB#1
 ; PIC: ldr.n r2, LCPI1_0
 ; PIC: add r2, pc
-; PIC: ldr r3, [r2]
-; PIC: ldr r3, [r3]
 ; PIC: ldr r2, [r2]
+; PIC: ldr r3, [r2]
 ; PIC: LBB1_2
 ; PIC: LCPI1_0:
 ; PIC-NOT: LCPI1_1:

Removed: llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll?rev=98131&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll (original)
+++ llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll (removed)
@@ -1,272 +0,0 @@
-; RUN: llc < %s -mtriple=x86_64-linux-gnu | grep movb | not grep x
-; PR1734
-
-	%struct.CUMULATIVE_ARGS = type { i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32 }
-	%struct.eh_status = type opaque
-	%struct.emit_status = type { i32, i32, %struct.rtx_def*, %struct.rtx_def*, %struct.sequence_stack*, i32, %struct.location_t, i32, i8*, %struct.rtx_def** }
-	%struct.expr_status = type { i32, i32, i32, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def* }
-	%struct.function = type { %struct.eh_status*, %struct.expr_status*, %struct.emit_status*, %struct.varasm_status*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.function*, i32, i32, i32, i32, %struct.rtx_def*, %struct.CUMULATIVE_ARGS, %struct.rtx_def*, %struct.rtx_def*, %struct.initial_value_struct*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, i8, i32, i64, %struct.tree_node*, %struct.tree_node*, %struct.rtx_def*, %struct.varray_head_tag*, %struct.temp_slot*, i32, %struct.var_refs_queue*, i32, i32, %struct.rtvec_def*, %struct.tree_node*, i32, i32, i32, %struct.machine_function*, i32, i32, i8, i8, %struct.language_function*, %struct.rtx_def*, i32, i32, i32, i32, %struct.location_t, %struct.varray_head_tag*, %struct.tree_node*, %struct.tree_node*, i8, i8, i8 }
-	%struct.initial_value_struct = type opaque
-	%struct.lang_decl = type opaque
-	%struct.lang_type = type opaque
-	%struct.language_function = type opaque
-	%struct.location_t = type { i8*, i32 }
-	%struct.machine_function = type { %struct.stack_local_entry*, i8*, %struct.rtx_def*, i32, i32, i32, i32, i32 }
-	%struct.rtunion = type { i8* }
-	%struct.rtvec_def = type { i32, [1 x %struct.rtx_def*] }
-	%struct.rtx_def = type { i16, i8, i8, %struct.u }
-	%struct.sequence_stack = type { %struct.rtx_def*, %struct.rtx_def*, %struct.sequence_stack* }
-	%struct.stack_local_entry = type opaque
-	%struct.temp_slot = type opaque
-	%struct.tree_common = type { %struct.tree_node*, %struct.tree_node*, %union.tree_ann_d*, i8, i8, i8, i8, i8 }
-	%struct.tree_decl = type { %struct.tree_common, %struct.location_t, i32, %struct.tree_node*, i8, i8, i8, i8, i8, i8, i8, i8, i32, %struct.tree_decl_u1, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.rtx_def*, i32, %struct.tree_decl_u2, %struct.tree_node*, %struct.tree_node*, i64, %struct.lang_decl* }
-	%struct.tree_decl_u1 = type { i64 }
-	%struct.tree_decl_u2 = type { %struct.function* }
-	%struct.tree_node = type { %struct.tree_decl }
-	%struct.tree_type = type { %struct.tree_common, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, i32, i16, i8, i8, i32, %struct.tree_node*, %struct.tree_node*, %struct.rtunion, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, i64, %struct.lang_type* }
-	%struct.u = type { [1 x %struct.rtunion] }
-	%struct.var_refs_queue = type { %struct.rtx_def*, i32, i32, %struct.var_refs_queue* }
-	%struct.varasm_status = type opaque
-	%struct.varray_data = type { [1 x i64] }
-	%struct.varray_head_tag = type { i64, i64, i32, i8*, %struct.varray_data }
-	%union.tree_ann_d = type opaque
- at .str = external constant [28 x i8]		; <[28 x i8]*> [#uses=1]
- at tree_code_type = external constant [0 x i32]		; <[0 x i32]*> [#uses=5]
- at global_trees = external global [47 x %struct.tree_node*]		; <[47 x %struct.tree_node*]*> [#uses=1]
- at mode_size = external global [48 x i8]		; <[48 x i8]*> [#uses=1]
- at __FUNCTION__.22683 = external constant [12 x i8]		; <[12 x i8]*> [#uses=1]
-
-define void @layout_type(%struct.tree_node* %type) {
-entry:
-	%tmp15 = icmp eq %struct.tree_node* %type, null		; <i1> [#uses=1]
-	br i1 %tmp15, label %cond_true, label %cond_false
-
-cond_true:		; preds = %entry
-	tail call void @fancy_abort( i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1713, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_false:		; preds = %entry
-	%tmp19 = load %struct.tree_node** getelementptr ([47 x %struct.tree_node*]* @global_trees, i32 0, i64 0), align 8		; <%struct.tree_node*> [#uses=1]
-	%tmp21 = icmp eq %struct.tree_node* %tmp19, %type		; <i1> [#uses=1]
-	br i1 %tmp21, label %UnifiedReturnBlock, label %cond_next25
-
-cond_next25:		; preds = %cond_false
-	%tmp30 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 0, i32 3		; <i8*> [#uses=1]
-	%tmp3031 = bitcast i8* %tmp30 to i32*		; <i32*> [#uses=6]
-	%tmp32 = load i32* %tmp3031, align 8		; <i32> [#uses=3]
-	%tmp3435 = trunc i32 %tmp32 to i8		; <i8> [#uses=3]
-	%tmp34353637 = zext i8 %tmp3435 to i64		; <i64> [#uses=1]
-	%tmp38 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp34353637		; <i32*> [#uses=1]
-	%tmp39 = load i32* %tmp38, align 4		; <i32> [#uses=1]
-	%tmp40 = icmp eq i32 %tmp39, 2		; <i1> [#uses=4]
-	br i1 %tmp40, label %cond_next46, label %cond_true43
-
-cond_true43:		; preds = %cond_next25
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1719, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next46:		; preds = %cond_next25
-	%tmp4950 = bitcast %struct.tree_node* %type to %struct.tree_type*		; <%struct.tree_type*> [#uses=2]
-	%tmp51 = getelementptr %struct.tree_type* %tmp4950, i32 0, i32 2		; <%struct.tree_node**> [#uses=2]
-	%tmp52 = load %struct.tree_node** %tmp51, align 8		; <%struct.tree_node*> [#uses=1]
-	%tmp53 = icmp eq %struct.tree_node* %tmp52, null		; <i1> [#uses=1]
-	br i1 %tmp53, label %cond_next57, label %UnifiedReturnBlock
-
-cond_next57:		; preds = %cond_next46
-	%tmp65 = and i32 %tmp32, 255		; <i32> [#uses=1]
-	switch i32 %tmp65, label %UnifiedReturnBlock [
-		 i32 6, label %bb140
-		 i32 7, label %bb69
-		 i32 8, label %bb140
-		 i32 13, label %bb478
-		 i32 23, label %bb
-	]
-
-bb:		; preds = %cond_next57
-	tail call void @fancy_abort( i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1727, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-bb69:		; preds = %cond_next57
-	br i1 %tmp40, label %cond_next91, label %cond_true88
-
-cond_true88:		; preds = %bb69
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1730, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next91:		; preds = %bb69
-	%tmp96 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 8		; <i8*> [#uses=1]
-	%tmp9697 = bitcast i8* %tmp96 to i32*		; <i32*> [#uses=2]
-	%tmp98 = load i32* %tmp9697, align 8		; <i32> [#uses=2]
-	%tmp100101552 = and i32 %tmp98, 511		; <i32> [#uses=1]
-	%tmp102 = icmp eq i32 %tmp100101552, 0		; <i1> [#uses=1]
-	br i1 %tmp102, label %cond_true105, label %bb140
-
-cond_true105:		; preds = %cond_next91
-	br i1 %tmp40, label %cond_next127, label %cond_true124
-
-cond_true124:		; preds = %cond_true105
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1731, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next127:		; preds = %cond_true105
-	%tmp136 = or i32 %tmp98, 1		; <i32> [#uses=1]
-	%tmp137 = and i32 %tmp136, -511		; <i32> [#uses=1]
-	store i32 %tmp137, i32* %tmp9697, align 8
-	br label %bb140
-
-bb140:		; preds = %cond_next127, %cond_next91, %cond_next57, %cond_next57
-	switch i8 %tmp3435, label %cond_true202 [
-		 i8 6, label %cond_next208
-		 i8 9, label %cond_next208
-		 i8 7, label %cond_next208
-		 i8 8, label %cond_next208
-		 i8 10, label %cond_next208
-	]
-
-cond_true202:		; preds = %bb140
-	tail call void (%struct.tree_node*, i8*, i32, i8*, ...)* @tree_check_failed( %struct.tree_node* %type, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1738, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0), i32 9, i32 6, i32 7, i32 8, i32 10, i32 0 )
-	unreachable
-
-cond_next208:		; preds = %bb140, %bb140, %bb140, %bb140, %bb140
-	%tmp213 = getelementptr %struct.tree_type* %tmp4950, i32 0, i32 14		; <%struct.tree_node**> [#uses=1]
-	%tmp214 = load %struct.tree_node** %tmp213, align 8		; <%struct.tree_node*> [#uses=2]
-	%tmp217 = getelementptr %struct.tree_node* %tmp214, i32 0, i32 0, i32 0, i32 3		; <i8*> [#uses=1]
-	%tmp217218 = bitcast i8* %tmp217 to i32*		; <i32*> [#uses=1]
-	%tmp219 = load i32* %tmp217218, align 8		; <i32> [#uses=1]
-	%tmp221222 = trunc i32 %tmp219 to i8		; <i8> [#uses=1]
-	%tmp223 = icmp eq i8 %tmp221222, 24		; <i1> [#uses=1]
-	br i1 %tmp223, label %cond_true226, label %cond_next340
-
-cond_true226:		; preds = %cond_next208
-	switch i8 %tmp3435, label %cond_true288 [
-		 i8 6, label %cond_next294
-		 i8 9, label %cond_next294
-		 i8 7, label %cond_next294
-		 i8 8, label %cond_next294
-		 i8 10, label %cond_next294
-	]
-
-cond_true288:		; preds = %cond_true226
-	tail call void (%struct.tree_node*, i8*, i32, i8*, ...)* @tree_check_failed( %struct.tree_node* %type, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1739, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0), i32 9, i32 6, i32 7, i32 8, i32 10, i32 0 )
-	unreachable
-
-cond_next294:		; preds = %cond_true226, %cond_true226, %cond_true226, %cond_true226, %cond_true226
-	%tmp301 = tail call i32 @tree_int_cst_sgn( %struct.tree_node* %tmp214 )		; <i32> [#uses=1]
-	%tmp302 = icmp sgt i32 %tmp301, -1		; <i1> [#uses=1]
-	br i1 %tmp302, label %cond_true305, label %cond_next340
-
-cond_true305:		; preds = %cond_next294
-	%tmp313 = load i32* %tmp3031, align 8		; <i32> [#uses=2]
-	%tmp315316 = trunc i32 %tmp313 to i8		; <i8> [#uses=1]
-	%tmp315316317318 = zext i8 %tmp315316 to i64		; <i64> [#uses=1]
-	%tmp319 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp315316317318		; <i32*> [#uses=1]
-	%tmp320 = load i32* %tmp319, align 4		; <i32> [#uses=1]
-	%tmp321 = icmp eq i32 %tmp320, 2		; <i1> [#uses=1]
-	br i1 %tmp321, label %cond_next327, label %cond_true324
-
-cond_true324:		; preds = %cond_true305
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1740, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next327:		; preds = %cond_true305
-	%tmp338 = or i32 %tmp313, 8192		; <i32> [#uses=1]
-	store i32 %tmp338, i32* %tmp3031, align 8
-	br label %cond_next340
-
-cond_next340:		; preds = %cond_next327, %cond_next294, %cond_next208
-	%tmp348 = load i32* %tmp3031, align 8		; <i32> [#uses=1]
-	%tmp350351 = trunc i32 %tmp348 to i8		; <i8> [#uses=1]
-	%tmp350351352353 = zext i8 %tmp350351 to i64		; <i64> [#uses=1]
-	%tmp354 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp350351352353		; <i32*> [#uses=1]
-	%tmp355 = load i32* %tmp354, align 4		; <i32> [#uses=1]
-	%tmp356 = icmp eq i32 %tmp355, 2		; <i1> [#uses=1]
-	br i1 %tmp356, label %cond_next385, label %cond_true359
-
-cond_true359:		; preds = %cond_next340
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1742, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next385:		; preds = %cond_next340
-	%tmp390 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 8		; <i8*> [#uses=1]
-	%tmp390391 = bitcast i8* %tmp390 to i32*		; <i32*> [#uses=3]
-	%tmp392 = load i32* %tmp390391, align 8		; <i32> [#uses=1]
-	%tmp394 = and i32 %tmp392, 511		; <i32> [#uses=1]
-	%tmp397 = tail call i32 @smallest_mode_for_size( i32 %tmp394, i32 2 )		; <i32> [#uses=1]
-	%tmp404 = load i32* %tmp390391, align 8		; <i32> [#uses=1]
-	%tmp397398405 = shl i32 %tmp397, 9		; <i32> [#uses=1]
-	%tmp407 = and i32 %tmp397398405, 65024		; <i32> [#uses=1]
-	%tmp408 = and i32 %tmp404, -65025		; <i32> [#uses=1]
-	%tmp409 = or i32 %tmp408, %tmp407		; <i32> [#uses=2]
-	store i32 %tmp409, i32* %tmp390391, align 8
-	%tmp417 = load i32* %tmp3031, align 8		; <i32> [#uses=1]
-	%tmp419420 = trunc i32 %tmp417 to i8		; <i8> [#uses=1]
-	%tmp419420421422 = zext i8 %tmp419420 to i64		; <i64> [#uses=1]
-	%tmp423 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp419420421422		; <i32*> [#uses=1]
-	%tmp424 = load i32* %tmp423, align 4		; <i32> [#uses=1]
-	%tmp425 = icmp eq i32 %tmp424, 2		; <i1> [#uses=1]
-	br i1 %tmp425, label %cond_next454, label %cond_true428
-
-cond_true428:		; preds = %cond_next385
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1744, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next454:		; preds = %cond_next385
-	lshr i32 %tmp409, 9		; <i32>:0 [#uses=1]
-	trunc i32 %0 to i8		; <i8>:1 [#uses=1]
-	%tmp463464 = and i8 %1, 127		; <i8> [#uses=1]
-	%tmp463464465466 = zext i8 %tmp463464 to i64		; <i64> [#uses=1]
-	%tmp467 = getelementptr [48 x i8]* @mode_size, i32 0, i64 %tmp463464465466		; <i8*> [#uses=1]
-	%tmp468 = load i8* %tmp467, align 1		; <i8> [#uses=1]
-	%tmp468469553 = zext i8 %tmp468 to i16		; <i16> [#uses=1]
-	%tmp470471 = shl i16 %tmp468469553, 3		; <i16> [#uses=1]
-	%tmp470471472 = zext i16 %tmp470471 to i64		; <i64> [#uses=1]
-	%tmp473 = tail call %struct.tree_node* @size_int_kind( i64 %tmp470471472, i32 2 )		; <%struct.tree_node*> [#uses=1]
-	store %struct.tree_node* %tmp473, %struct.tree_node** %tmp51, align 8
-	ret void
-
-bb478:		; preds = %cond_next57
-	br i1 %tmp40, label %cond_next500, label %cond_true497
-
-cond_true497:		; preds = %bb478
-	tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1755, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next500:		; preds = %bb478
-	%tmp506 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 0, i32 1		; <%struct.tree_node**> [#uses=1]
-	%tmp507 = load %struct.tree_node** %tmp506, align 8		; <%struct.tree_node*> [#uses=2]
-	%tmp511 = getelementptr %struct.tree_node* %tmp507, i32 0, i32 0, i32 0, i32 3		; <i8*> [#uses=1]
-	%tmp511512 = bitcast i8* %tmp511 to i32*		; <i32*> [#uses=1]
-	%tmp513 = load i32* %tmp511512, align 8		; <i32> [#uses=2]
-	%tmp515516 = trunc i32 %tmp513 to i8		; <i8> [#uses=1]
-	%tmp515516517518 = zext i8 %tmp515516 to i64		; <i64> [#uses=1]
-	%tmp519 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp515516517518		; <i32*> [#uses=1]
-	%tmp520 = load i32* %tmp519, align 4		; <i32> [#uses=1]
-	%tmp521 = icmp eq i32 %tmp520, 2		; <i1> [#uses=1]
-	br i1 %tmp521, label %cond_next527, label %cond_true524
-
-cond_true524:		; preds = %cond_next500
-	tail call void @tree_class_check_failed( %struct.tree_node* %tmp507, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1755, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
-	unreachable
-
-cond_next527:		; preds = %cond_next500
-	%tmp545 = and i32 %tmp513, 8192		; <i32> [#uses=1]
-	%tmp547 = and i32 %tmp32, -8193		; <i32> [#uses=1]
-	%tmp548 = or i32 %tmp547, %tmp545		; <i32> [#uses=1]
-	store i32 %tmp548, i32* %tmp3031, align 8
-	ret void
-
-UnifiedReturnBlock:		; preds = %cond_next57, %cond_next46, %cond_false
-	ret void
-}
-
-declare void @fancy_abort(i8*, i32, i8*)
-
-declare void @tree_class_check_failed(%struct.tree_node*, i32, i8*, i32, i8*)
-
-declare i32 @smallest_mode_for_size(i32, i32)
-
-declare %struct.tree_node* @size_int_kind(i64, i32)
-
-declare void @tree_check_failed(%struct.tree_node*, i8*, i32, i8*, ...)
-
-declare i32 @tree_int_cst_sgn(%struct.tree_node*)

Modified: llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll (original)
+++ llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll Tue Mar  9 21:07:41 2010
@@ -1,8 +1,10 @@
-; RUN: llc < %s
+; RUN: llc -mtriple=i386-pc-linux-gnu < %s
 ; PR4668
-; ModuleID = '<stdin>'
-target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32"
-target triple = "i386-pc-linux-gnu"
+; XFAIL: *
+; FIXME: If the coalescer happens to coalesce %level.1 with the copy to EAX
+; (for ret) then this will fail to compile. The fundamental problem is
+; once the coalescer fixes a virtual register to physical register we can't
+; evict it.
 
 define i32 @x(i32 %qscale) nounwind {
 entry:

Added: llvm/trunk/test/CodeGen/X86/machine-cse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/machine-cse.ll?rev=98132&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/machine-cse.ll (added)
+++ llvm/trunk/test/CodeGen/X86/machine-cse.ll Tue Mar  9 21:07:41 2010
@@ -0,0 +1,39 @@
+; RUN: llc -mtriple=x86_64-apple-darwin < %s | FileCheck %s
+; rdar://7610418
+
+%ptr = type { i8* }
+%struct.s1 = type { %ptr, %ptr }
+%struct.s2 = type { i32, i8*, i8*, [256 x %struct.s1*], [8 x i32], i64, i8*, i32, i64, i64, i32, %struct.s3*, %struct.s3*, [49 x i64] }
+%struct.s3 = type { %struct.s3*, %struct.s3*, i32, i32, i32 }
+
+define fastcc i8* @t(i64 %size) nounwind {
+entry:
+; CHECK: t:
+; CHECK: leaq (%rax,%rax,4)
+  %0 = zext i32 undef to i64
+  %1 = getelementptr inbounds %struct.s2* null, i64 %0
+  br i1 undef, label %bb1, label %bb2
+
+bb1:
+; CHECK: %bb1
+; CHECK-NOT: shlq $9
+; CHECK-NOT: leaq
+; CHECK: call
+  %2 = getelementptr inbounds %struct.s2* null, i64 %0, i32 0
+  call void @bar(i32* %2) nounwind
+  unreachable
+
+bb2:
+; CHECK: %bb2
+; CHECK-NOT: leaq
+; CHECK: callq
+  %3 = call fastcc i8* @foo(%struct.s2* %1) nounwind
+  unreachable
+
+bb3:
+  ret i8* undef
+}
+
+declare void @bar(i32*)
+
+declare fastcc i8* @foo(%struct.s2*) nounwind

Modified: llvm/trunk/test/CodeGen/X86/pre-split6.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pre-split6.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/pre-split6.ll (original)
+++ llvm/trunk/test/CodeGen/X86/pre-split6.ll Tue Mar  9 21:07:41 2010
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 -mattr=+sse2 -pre-alloc-split | grep {divsd	8} | count 1
+; RUN: llc < %s -march=x86 -mattr=+sse2 -pre-alloc-split | grep {divsd	24} | count 1
 
 @current_surfaces.b = external global i1		; <i1*> [#uses=1]
 

Modified: llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll?rev=98132&r1=98131&r2=98132&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll (original)
+++ llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll Tue Mar  9 21:07:41 2010
@@ -1,5 +1,6 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -relocation-model=pic -disable-fp-elim -color-ss-with-regs -stats -info-output-file - > %t
-; RUN:   grep stackcoloring %t | grep "stack slot refs replaced with reg refs"  | grep 8
+; RuN:   grep asm-printer %t | grep 156
+; RUN:   grep stackcoloring %t | grep "stack slot refs replaced with reg refs"  | grep 4
 
 	type { [62 x %struct.Bitvec*] }		; type %0
 	type { i8* }		; type %1




From evan.cheng at apple.com  Tue Mar  9 21:20:35 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 03:20:35 -0000
Subject: [llvm-commits] [test-suite] r98133 -
	/test-suite/trunk/Makefile.programs
Message-ID: <20100310032035.D2EEA2A6C12C@llvm.org>

Author: evancheng
Date: Tue Mar  9 21:20:35 2010
New Revision: 98133

URL: http://llvm.org/viewvc/llvm-project?rev=98133&view=rev
Log:
-enable-machine-cse is no more.

Modified:
    test-suite/trunk/Makefile.programs

Modified: test-suite/trunk/Makefile.programs
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/Makefile.programs?rev=98133&r1=98132&r2=98133&view=diff
==============================================================================
--- test-suite/trunk/Makefile.programs (original)
+++ test-suite/trunk/Makefile.programs Tue Mar  9 21:20:35 2010
@@ -232,19 +232,19 @@
 LLCBETAOPTION := -sched=simple 
 endif
 ifeq ($(ARCH),x86_64)
-LLCBETAOPTION := -enable-machine-cse
+LLCBETAOPTION :=
 endif
 ifeq ($(ARCH),x86)
-LLCBETAOPTION := -enable-machine-cse
+LLCBETAOPTION :=
 endif
 ifeq ($(ARCH),Sparc)
 LLCBETAOPTION := -enable-sparc-v9-insts
 endif
 ifeq ($(ARCH),ARM)
-LLCBETAOPTION := -enable-machine-cse
+LLCBETAOPTION :=
 endif
 ifeq ($(ARCH),THUMB)
-LLCBETAOPTION := -enable-machine-cse
+LLCBETAOPTION :=
 endif
 
 print-llcbeta-option:




From dalej at apple.com  Tue Mar  9 23:45:47 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 05:45:47 -0000
Subject: [llvm-commits] [llvm] r98137 -
	/llvm/trunk/lib/CodeGen/BranchFolding.cpp
Message-ID: <20100310054547.9ABAA2A6C12C@llvm.org>

Author: johannes
Date: Tue Mar  9 23:45:47 2010
New Revision: 98137

URL: http://llvm.org/viewvc/llvm-project?rev=98137&view=rev
Log:
This survived a bootstrap, so let's try 98104 again.


Modified:
    llvm/trunk/lib/CodeGen/BranchFolding.cpp

Modified: llvm/trunk/lib/CodeGen/BranchFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/BranchFolding.cpp?rev=98137&r1=98136&r2=98137&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/BranchFolding.cpp (original)
+++ llvm/trunk/lib/CodeGen/BranchFolding.cpp Tue Mar  9 23:45:47 2010
@@ -347,15 +347,29 @@
     --I1; --I2;
     // Skip debugging pseudos; necessary to avoid changing the code.
     while (I1->isDebugValue()) {
-      if (I1==MBB1->begin())
+      if (I1==MBB1->begin()) {
+        while (I2->isDebugValue()) {
+          if (I2==MBB2->begin())
+            // I1==DBG at begin; I2==DBG at begin
+            return TailLen;
+          --I2;
+        }
+        ++I2;
+        // I1==DBG at begin; I2==non-DBG, or first of DBGs not at begin
         return TailLen;
+      }
       --I1;
     }
+    // I1==first (untested) non-DBG preceding known match
     while (I2->isDebugValue()) {
-      if (I2==MBB2->begin())
+      if (I2==MBB2->begin()) {
+        ++I1;
+        // I1==non-DBG, or first of DBGs not at begin; I2==DBG at begin
         return TailLen;
+      }
       --I2;
     }
+    // I1, I2==first (untested) non-DBGs preceding known match
     if (!I1->isIdenticalTo(I2) ||
         // FIXME: This check is dubious. It's used to get around a problem where
         // people incorrectly expect inline asm directives to remain in the same
@@ -368,6 +382,29 @@
     }
     ++TailLen;
   }
+  // Back past possible debugging pseudos at beginning of block.  This matters
+  // when one block differs from the other only by whether debugging pseudos
+  // are present at the beginning.  (This way, the various checks later for
+  // I1==MBB1->begin() work as expected.)
+  if (I1 == MBB1->begin() && I2 != MBB2->begin()) {
+    --I2;
+    while (I2->isDebugValue()) {
+      if (I2 == MBB2->begin()) {
+        return TailLen;
+        }
+      --I2;
+    }
+    ++I2;
+  }
+  if (I2 == MBB2->begin() && I1 != MBB1->begin()) {
+    --I1;
+    while (I1->isDebugValue()) {
+      if (I1 == MBB1->begin())
+        return TailLen;
+      --I1;
+    }
+    ++I1;
+  }
   return TailLen;
 }
 
@@ -934,6 +971,18 @@
   return MadeChange;
 }
 
+// Blocks should be considered empty if they contain only debug info;
+// else the debug info would affect codegen.
+static bool IsEmptyBlock(MachineBasicBlock *MBB) {
+  if (MBB->empty())
+    return true;
+  for (MachineBasicBlock::iterator MBBI = MBB->begin(), MBBE = MBB->end();
+       MBBI!=MBBE; ++MBBI) {
+    if (!MBBI->isDebugValue())
+      return false;
+  }
+  return true;
+}
 
 /// IsBetterFallthrough - Return true if it would be clearly better to
 /// fall-through to MBB1 than to fall through into MBB2.  This has to return
@@ -971,7 +1020,7 @@
   // explicitly.  Landing pads should not do this since the landing-pad table
   // points to this block.  Blocks with their addresses taken shouldn't be
   // optimized away.
-  if (MBB->empty() && !MBB->isLandingPad() && !MBB->hasAddressTaken()) {
+  if (IsEmptyBlock(MBB) && !MBB->isLandingPad() && !MBB->hasAddressTaken()) {
     // Dead block?  Leave for cleanup later.
     if (MBB->pred_empty()) return MadeChange;
 
@@ -1163,7 +1212,23 @@
       // be 'non-branch terminators' in the block, try removing the branch and
       // then seeing if the block is empty.
       TII->RemoveBranch(*MBB);
-
+      // If the only things remaining in the block are debug info, remove these
+      // as well, so this will behave the same as an empty block in non-debug
+      // mode.
+      if (!MBB->empty()) {
+        bool NonDebugInfoFound = false;
+        for (MachineBasicBlock::iterator I = MBB->begin(), E = MBB->end();
+             I != E; ++I) {
+          if (!I->isDebugValue()) {
+            NonDebugInfoFound = true;
+            break;
+          }
+        }
+        if (!NonDebugInfoFound)
+          // Make the block empty, losing the debug info (we could probably
+          // improve this in some cases.)
+          MBB->erase(MBB->begin(), MBB->end());
+      }
       // If this block is just an unconditional branch to CurTBB, we can
       // usually completely eliminate the block.  The only case we cannot
       // completely eliminate the block is when the block before this one




From daniel at zuster.org  Wed Mar 10 00:27:59 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 9 Mar 2010 22:27:59 -0800
Subject: [llvm-commits] [llvm] r98132 - in /llvm/trunk:
	lib/CodeGen/LLVMTargetMachine.cpp test/CodeGen/ARM/indirectbr.ll
	test/CodeGen/PowerPC/indirectbr.ll
	test/CodeGen/Thumb/2009-08-20-ISelBug.ll
	test/CodeGen/Thumb2/machine-licm.ll test/CodeGen/X86/2007-
Message-ID: <6a8523d61003092227g3d7b94c4hb17e6f596c20482c@mail.gmail.com>

On Tue, Mar 9, 2010 at 7:07 PM, Evan Cheng <evan.cheng at apple.com> wrote:
> Author: evancheng
> Date: Tue Mar ?9 21:07:41 2010
> New Revision: 98132
>
> URL: http://llvm.org/viewvc/llvm-project?rev=98132&view=rev
> Log:
> Enable machine cse pass.
>
> Added:
> ? ?llvm/trunk/test/CodeGen/X86/machine-cse.ll
> Removed:
> ? ?llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll
> Modified:
> ? ?llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
> ? ?llvm/trunk/test/CodeGen/ARM/indirectbr.ll
> ? ?llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll
> ? ?llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll
> ? ?llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll
> ? ?llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll
> ? ?llvm/trunk/test/CodeGen/X86/pre-split6.ll
> ? ?llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll
>
> Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
> +++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Tue Mar ?9 21:07:41 2010
> @@ -67,9 +67,6 @@
> ? ? cl::desc("Verify generated machine code"),
> ? ? cl::init(getenv("LLVM_VERIFY_MACHINEINSTRS")!=NULL));
>
> -static cl::opt<bool> EnableMachineCSE("enable-machine-cse", cl::Hidden,
> - ? ?cl::desc("Enable Machine CSE"));
> -
> ?static cl::opt<cl::boolOrDefault>
> ?AsmVerbose("asm-verbose", cl::desc("Add comments to directives."),
> ? ? ? ? ? ?cl::init(cl::BOU_UNSET));
> @@ -328,8 +325,7 @@
> ? ? PM.add(createOptimizeExtsPass());
> ? ? if (!DisableMachineLICM)
> ? ? ? PM.add(createMachineLICMPass());
> - ? ?if (EnableMachineCSE)
> - ? ? ?PM.add(createMachineCSEPass());
> + ? ?PM.add(createMachineCSEPass());
> ? ? if (!DisableMachineSink)
> ? ? ? PM.add(createMachineSinkingPass());
> ? ? printAndVerify(PM, "After Machine LICM, CSE and Sinking passes",
>
> Modified: llvm/trunk/test/CodeGen/ARM/indirectbr.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/indirectbr.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/ARM/indirectbr.ll (original)
> +++ llvm/trunk/test/CodeGen/ARM/indirectbr.ll Tue Mar ?9 21:07:41 2010
> @@ -15,14 +15,14 @@
> ?; indirect branch gets duplicated here
> ?; ARM: bx
> ?; THUMB: mov pc, r1
> -; THUMB2: mov pc, r1
> +; THUMB2: mov pc, r2
> ? br i1 %1, label %bb3, label %bb2
>
> ?bb2: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?; preds = %entry, %bb3
> ? %gotovar.4.0 = phi i8* [ %gotovar.4.0.pre, %bb3 ], [ %0, %entry ] ; <i8*> [#uses=1]
> ?; ARM: bx
> ?; THUMB: mov pc, r1
> -; THUMB2: mov pc, r1
> +; THUMB2: mov pc, r2
> ? indirectbr i8* %gotovar.4.0, [label %L5, label %L4, label %L3, label %L2, label %L1]
>
> ?bb3: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?; preds = %entry
>
> Modified: llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll (original)
> +++ llvm/trunk/test/CodeGen/PowerPC/indirectbr.ll Tue Mar ?9 21:07:41 2010
> @@ -43,13 +43,13 @@
>
> ?L1: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ; preds = %L2, %bb2
> ? %res.3 = phi i32 [ %phitmp, %L2 ], [ 2, %bb2 ] ?; <i32> [#uses=1]
> -; PIC: addis r5, r4, ha16(L_BA4__foo_L5-"L1$pb")
> +; PIC: addis r4, r4, ha16(L_BA4__foo_L5-"L1$pb")
> ?; PIC: li r6, lo16(L_BA4__foo_L5-"L1$pb")
> -; PIC: add r5, r5, r6
> -; PIC: stw r5
> -; STATIC: li r4, lo16(L_BA4__foo_L5)
> -; STATIC: addis r4, r4, ha16(L_BA4__foo_L5)
> -; STATIC: stw r4
> +; PIC: add r4, r4, r6
> +; PIC: stw r4
> +; STATIC: li r5, lo16(L_BA4__foo_L5)
> +; STATIC: addis r5, r5, ha16(L_BA4__foo_L5)
> +; STATIC: stw r5
> ? store i8* blockaddress(@foo, %L5), i8** @nextaddr, align 4
> ? ret i32 %res.3
> ?}
>
> Modified: llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll (original)
> +++ llvm/trunk/test/CodeGen/Thumb/2009-08-20-ISelBug.ll Tue Mar ?9 21:07:41 2010
> @@ -11,7 +11,7 @@
>
> ?define arm_apcscc i32 @t(%struct.asl_file_t* %s, i64 %off, i64* %out) nounwind optsize {
> ?; CHECK: t:
> -; CHECK: adds r3, #8
> +; CHECK: adds r0, #8
> ?entry:
> ? %val = alloca i64, align 4 ? ? ? ? ? ? ? ? ? ? ?; <i64*> [#uses=3]
> ? %0 = icmp eq %struct.asl_file_t* %s, null ? ? ? ; <i1> [#uses=1]
>
> Modified: llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll (original)
> +++ llvm/trunk/test/CodeGen/Thumb2/machine-licm.ll Tue Mar ?9 21:07:41 2010
> @@ -18,9 +18,8 @@
> ?bb.nph: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ; preds = %entry
> ?; CHECK: BB#1
> ?; CHECK: ldr.n r2, LCPI1_0
> -; CHECK: ldr r3, [r2]
> -; CHECK: ldr r3, [r3]
> ?; CHECK: ldr r2, [r2]
> +; CHECK: ldr r3, [r2]
> ?; CHECK: LBB1_2
> ?; CHECK: LCPI1_0:
> ?; CHECK-NOT: LCPI1_1:
> @@ -29,9 +28,8 @@
> ?; PIC: BB#1
> ?; PIC: ldr.n r2, LCPI1_0
> ?; PIC: add r2, pc
> -; PIC: ldr r3, [r2]
> -; PIC: ldr r3, [r3]
> ?; PIC: ldr r2, [r2]
> +; PIC: ldr r3, [r2]
> ?; PIC: LBB1_2
> ?; PIC: LCPI1_0:
> ?; PIC-NOT: LCPI1_1:
>
> Removed: llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll?rev=98131&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll (original)
> +++ llvm/trunk/test/CodeGen/X86/2007-10-16-IllegalAsm.ll (removed)
> @@ -1,272 +0,0 @@
> -; RUN: llc < %s -mtriple=x86_64-linux-gnu | grep movb | not grep x
> -; PR1734
> -
> - ? ? ? %struct.CUMULATIVE_ARGS = type { i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32 }
> - ? ? ? %struct.eh_status = type opaque
> - ? ? ? %struct.emit_status = type { i32, i32, %struct.rtx_def*, %struct.rtx_def*, %struct.sequence_stack*, i32, %struct.location_t, i32, i8*, %struct.rtx_def** }
> - ? ? ? %struct.expr_status = type { i32, i32, i32, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def* }
> - ? ? ? %struct.function = type { %struct.eh_status*, %struct.expr_status*, %struct.emit_status*, %struct.varasm_status*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.function*, i32, i32, i32, i32, %struct.rtx_def*, %struct.CUMULATIVE_ARGS, %struct.rtx_def*, %struct.rtx_def*, %struct.initial_value_struct*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, %struct.rtx_def*, i8, i32, i64, %struct.tree_node*, %struct.tree_node*, %struct.rtx_def*, %struct.varray_head_tag*, %struct.temp_slot*, i32, %struct.var_refs_queue*, i32, i32, %struct.rtvec_def*, %struct.tree_node*, i32, i32, i32, %struct.machine_function*, i32, i32, i8, i8, %struct.language_function*, %struct.rtx_def*, i32, i32, i32, i32, %struct.location_t, %struct.varray_head_tag*, %struct.tree_node*, %struct.tree_node*, i8, i8, i8 }
> - ? ? ? %struct.initial_value_struct = type opaque
> - ? ? ? %struct.lang_decl = type opaque
> - ? ? ? %struct.lang_type = type opaque
> - ? ? ? %struct.language_function = type opaque
> - ? ? ? %struct.location_t = type { i8*, i32 }
> - ? ? ? %struct.machine_function = type { %struct.stack_local_entry*, i8*, %struct.rtx_def*, i32, i32, i32, i32, i32 }
> - ? ? ? %struct.rtunion = type { i8* }
> - ? ? ? %struct.rtvec_def = type { i32, [1 x %struct.rtx_def*] }
> - ? ? ? %struct.rtx_def = type { i16, i8, i8, %struct.u }
> - ? ? ? %struct.sequence_stack = type { %struct.rtx_def*, %struct.rtx_def*, %struct.sequence_stack* }
> - ? ? ? %struct.stack_local_entry = type opaque
> - ? ? ? %struct.temp_slot = type opaque
> - ? ? ? %struct.tree_common = type { %struct.tree_node*, %struct.tree_node*, %union.tree_ann_d*, i8, i8, i8, i8, i8 }
> - ? ? ? %struct.tree_decl = type { %struct.tree_common, %struct.location_t, i32, %struct.tree_node*, i8, i8, i8, i8, i8, i8, i8, i8, i32, %struct.tree_decl_u1, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.rtx_def*, i32, %struct.tree_decl_u2, %struct.tree_node*, %struct.tree_node*, i64, %struct.lang_decl* }
> - ? ? ? %struct.tree_decl_u1 = type { i64 }
> - ? ? ? %struct.tree_decl_u2 = type { %struct.function* }
> - ? ? ? %struct.tree_node = type { %struct.tree_decl }
> - ? ? ? %struct.tree_type = type { %struct.tree_common, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, i32, i16, i8, i8, i32, %struct.tree_node*, %struct.tree_node*, %struct.rtunion, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, %struct.tree_node*, i64, %struct.lang_type* }
> - ? ? ? %struct.u = type { [1 x %struct.rtunion] }
> - ? ? ? %struct.var_refs_queue = type { %struct.rtx_def*, i32, i32, %struct.var_refs_queue* }
> - ? ? ? %struct.varasm_status = type opaque
> - ? ? ? %struct.varray_data = type { [1 x i64] }
> - ? ? ? %struct.varray_head_tag = type { i64, i64, i32, i8*, %struct.varray_data }
> - ? ? ? %union.tree_ann_d = type opaque
> - at .str = external constant [28 x i8] ? ? ? ? ? ?; <[28 x i8]*> [#uses=1]
> - at tree_code_type = external constant [0 x i32] ? ? ? ? ?; <[0 x i32]*> [#uses=5]
> - at global_trees = external global [47 x %struct.tree_node*] ? ? ? ? ? ? ?; <[47 x %struct.tree_node*]*> [#uses=1]
> - at mode_size = external global [48 x i8] ? ? ? ? ; <[48 x i8]*> [#uses=1]
> - at __FUNCTION__.22683 = external constant [12 x i8] ? ? ? ? ? ? ?; <[12 x i8]*> [#uses=1]
> -
> -define void @layout_type(%struct.tree_node* %type) {
> -entry:
> - ? ? ? %tmp15 = icmp eq %struct.tree_node* %type, null ? ? ? ? ; <i1> [#uses=1]
> - ? ? ? br i1 %tmp15, label %cond_true, label %cond_false
> -
> -cond_true: ? ? ? ? ? ? ; preds = %entry
> - ? ? ? tail call void @fancy_abort( i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1713, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_false: ? ? ? ? ? ?; preds = %entry
> - ? ? ? %tmp19 = load %struct.tree_node** getelementptr ([47 x %struct.tree_node*]* @global_trees, i32 0, i64 0), align 8 ? ? ? ? ? ? ? ; <%struct.tree_node*> [#uses=1]
> - ? ? ? %tmp21 = icmp eq %struct.tree_node* %tmp19, %type ? ? ? ? ? ? ? ; <i1> [#uses=1]
> - ? ? ? br i1 %tmp21, label %UnifiedReturnBlock, label %cond_next25
> -
> -cond_next25: ? ? ? ? ? ; preds = %cond_false
> - ? ? ? %tmp30 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 0, i32 3 ? ? ? ? ? ? ; <i8*> [#uses=1]
> - ? ? ? %tmp3031 = bitcast i8* %tmp30 to i32* ? ? ? ? ? ; <i32*> [#uses=6]
> - ? ? ? %tmp32 = load i32* %tmp3031, align 8 ? ? ? ? ? ?; <i32> [#uses=3]
> - ? ? ? %tmp3435 = trunc i32 %tmp32 to i8 ? ? ? ? ? ? ? ; <i8> [#uses=3]
> - ? ? ? %tmp34353637 = zext i8 %tmp3435 to i64 ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp38 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp34353637 ? ? ? ? ? ? ?; <i32*> [#uses=1]
> - ? ? ? %tmp39 = load i32* %tmp38, align 4 ? ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp40 = icmp eq i32 %tmp39, 2 ? ? ? ? ?; <i1> [#uses=4]
> - ? ? ? br i1 %tmp40, label %cond_next46, label %cond_true43
> -
> -cond_true43: ? ? ? ? ? ; preds = %cond_next25
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1719, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next46: ? ? ? ? ? ; preds = %cond_next25
> - ? ? ? %tmp4950 = bitcast %struct.tree_node* %type to %struct.tree_type* ? ? ? ? ? ? ? ; <%struct.tree_type*> [#uses=2]
> - ? ? ? %tmp51 = getelementptr %struct.tree_type* %tmp4950, i32 0, i32 2 ? ? ? ? ? ? ? ?; <%struct.tree_node**> [#uses=2]
> - ? ? ? %tmp52 = load %struct.tree_node** %tmp51, align 8 ? ? ? ? ? ? ? ; <%struct.tree_node*> [#uses=1]
> - ? ? ? %tmp53 = icmp eq %struct.tree_node* %tmp52, null ? ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp53, label %cond_next57, label %UnifiedReturnBlock
> -
> -cond_next57: ? ? ? ? ? ; preds = %cond_next46
> - ? ? ? %tmp65 = and i32 %tmp32, 255 ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? switch i32 %tmp65, label %UnifiedReturnBlock [
> - ? ? ? ? ? ? ? ?i32 6, label %bb140
> - ? ? ? ? ? ? ? ?i32 7, label %bb69
> - ? ? ? ? ? ? ? ?i32 8, label %bb140
> - ? ? ? ? ? ? ? ?i32 13, label %bb478
> - ? ? ? ? ? ? ? ?i32 23, label %bb
> - ? ? ? ]
> -
> -bb: ? ? ? ? ? ?; preds = %cond_next57
> - ? ? ? tail call void @fancy_abort( i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1727, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -bb69: ? ? ? ? ?; preds = %cond_next57
> - ? ? ? br i1 %tmp40, label %cond_next91, label %cond_true88
> -
> -cond_true88: ? ? ? ? ? ; preds = %bb69
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1730, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next91: ? ? ? ? ? ; preds = %bb69
> - ? ? ? %tmp96 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 8 ? ? ? ? ? ?; <i8*> [#uses=1]
> - ? ? ? %tmp9697 = bitcast i8* %tmp96 to i32* ? ? ? ? ? ; <i32*> [#uses=2]
> - ? ? ? %tmp98 = load i32* %tmp9697, align 8 ? ? ? ? ? ?; <i32> [#uses=2]
> - ? ? ? %tmp100101552 = and i32 %tmp98, 511 ? ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp102 = icmp eq i32 %tmp100101552, 0 ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp102, label %cond_true105, label %bb140
> -
> -cond_true105: ? ? ? ? ?; preds = %cond_next91
> - ? ? ? br i1 %tmp40, label %cond_next127, label %cond_true124
> -
> -cond_true124: ? ? ? ? ?; preds = %cond_true105
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1731, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next127: ? ? ? ? ?; preds = %cond_true105
> - ? ? ? %tmp136 = or i32 %tmp98, 1 ? ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp137 = and i32 %tmp136, -511 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? store i32 %tmp137, i32* %tmp9697, align 8
> - ? ? ? br label %bb140
> -
> -bb140: ? ? ? ? ; preds = %cond_next127, %cond_next91, %cond_next57, %cond_next57
> - ? ? ? switch i8 %tmp3435, label %cond_true202 [
> - ? ? ? ? ? ? ? ?i8 6, label %cond_next208
> - ? ? ? ? ? ? ? ?i8 9, label %cond_next208
> - ? ? ? ? ? ? ? ?i8 7, label %cond_next208
> - ? ? ? ? ? ? ? ?i8 8, label %cond_next208
> - ? ? ? ? ? ? ? ?i8 10, label %cond_next208
> - ? ? ? ]
> -
> -cond_true202: ? ? ? ? ?; preds = %bb140
> - ? ? ? tail call void (%struct.tree_node*, i8*, i32, i8*, ...)* @tree_check_failed( %struct.tree_node* %type, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1738, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0), i32 9, i32 6, i32 7, i32 8, i32 10, i32 0 )
> - ? ? ? unreachable
> -
> -cond_next208: ? ? ? ? ?; preds = %bb140, %bb140, %bb140, %bb140, %bb140
> - ? ? ? %tmp213 = getelementptr %struct.tree_type* %tmp4950, i32 0, i32 14 ? ? ? ? ? ? ?; <%struct.tree_node**> [#uses=1]
> - ? ? ? %tmp214 = load %struct.tree_node** %tmp213, align 8 ? ? ? ? ? ? ; <%struct.tree_node*> [#uses=2]
> - ? ? ? %tmp217 = getelementptr %struct.tree_node* %tmp214, i32 0, i32 0, i32 0, i32 3 ? ? ? ? ?; <i8*> [#uses=1]
> - ? ? ? %tmp217218 = bitcast i8* %tmp217 to i32* ? ? ? ? ? ? ? ?; <i32*> [#uses=1]
> - ? ? ? %tmp219 = load i32* %tmp217218, align 8 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp221222 = trunc i32 %tmp219 to i8 ? ? ? ? ? ?; <i8> [#uses=1]
> - ? ? ? %tmp223 = icmp eq i8 %tmp221222, 24 ? ? ? ? ? ? ; <i1> [#uses=1]
> - ? ? ? br i1 %tmp223, label %cond_true226, label %cond_next340
> -
> -cond_true226: ? ? ? ? ?; preds = %cond_next208
> - ? ? ? switch i8 %tmp3435, label %cond_true288 [
> - ? ? ? ? ? ? ? ?i8 6, label %cond_next294
> - ? ? ? ? ? ? ? ?i8 9, label %cond_next294
> - ? ? ? ? ? ? ? ?i8 7, label %cond_next294
> - ? ? ? ? ? ? ? ?i8 8, label %cond_next294
> - ? ? ? ? ? ? ? ?i8 10, label %cond_next294
> - ? ? ? ]
> -
> -cond_true288: ? ? ? ? ?; preds = %cond_true226
> - ? ? ? tail call void (%struct.tree_node*, i8*, i32, i8*, ...)* @tree_check_failed( %struct.tree_node* %type, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1739, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0), i32 9, i32 6, i32 7, i32 8, i32 10, i32 0 )
> - ? ? ? unreachable
> -
> -cond_next294: ? ? ? ? ?; preds = %cond_true226, %cond_true226, %cond_true226, %cond_true226, %cond_true226
> - ? ? ? %tmp301 = tail call i32 @tree_int_cst_sgn( %struct.tree_node* %tmp214 ) ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp302 = icmp sgt i32 %tmp301, -1 ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp302, label %cond_true305, label %cond_next340
> -
> -cond_true305: ? ? ? ? ?; preds = %cond_next294
> - ? ? ? %tmp313 = load i32* %tmp3031, align 8 ? ? ? ? ? ; <i32> [#uses=2]
> - ? ? ? %tmp315316 = trunc i32 %tmp313 to i8 ? ? ? ? ? ?; <i8> [#uses=1]
> - ? ? ? %tmp315316317318 = zext i8 %tmp315316 to i64 ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp319 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp315316317318 ? ? ? ? ; <i32*> [#uses=1]
> - ? ? ? %tmp320 = load i32* %tmp319, align 4 ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp321 = icmp eq i32 %tmp320, 2 ? ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp321, label %cond_next327, label %cond_true324
> -
> -cond_true324: ? ? ? ? ?; preds = %cond_true305
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1740, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next327: ? ? ? ? ?; preds = %cond_true305
> - ? ? ? %tmp338 = or i32 %tmp313, 8192 ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? store i32 %tmp338, i32* %tmp3031, align 8
> - ? ? ? br label %cond_next340
> -
> -cond_next340: ? ? ? ? ?; preds = %cond_next327, %cond_next294, %cond_next208
> - ? ? ? %tmp348 = load i32* %tmp3031, align 8 ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp350351 = trunc i32 %tmp348 to i8 ? ? ? ? ? ?; <i8> [#uses=1]
> - ? ? ? %tmp350351352353 = zext i8 %tmp350351 to i64 ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp354 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp350351352353 ? ? ? ? ; <i32*> [#uses=1]
> - ? ? ? %tmp355 = load i32* %tmp354, align 4 ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp356 = icmp eq i32 %tmp355, 2 ? ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp356, label %cond_next385, label %cond_true359
> -
> -cond_true359: ? ? ? ? ?; preds = %cond_next340
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1742, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next385: ? ? ? ? ?; preds = %cond_next340
> - ? ? ? %tmp390 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 8 ? ? ? ? ? ; <i8*> [#uses=1]
> - ? ? ? %tmp390391 = bitcast i8* %tmp390 to i32* ? ? ? ? ? ? ? ?; <i32*> [#uses=3]
> - ? ? ? %tmp392 = load i32* %tmp390391, align 8 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp394 = and i32 %tmp392, 511 ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp397 = tail call i32 @smallest_mode_for_size( i32 %tmp394, i32 2 ) ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp404 = load i32* %tmp390391, align 8 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp397398405 = shl i32 %tmp397, 9 ? ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp407 = and i32 %tmp397398405, 65024 ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp408 = and i32 %tmp404, -65025 ? ? ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp409 = or i32 %tmp408, %tmp407 ? ? ? ? ? ? ? ; <i32> [#uses=2]
> - ? ? ? store i32 %tmp409, i32* %tmp390391, align 8
> - ? ? ? %tmp417 = load i32* %tmp3031, align 8 ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp419420 = trunc i32 %tmp417 to i8 ? ? ? ? ? ?; <i8> [#uses=1]
> - ? ? ? %tmp419420421422 = zext i8 %tmp419420 to i64 ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp423 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp419420421422 ? ? ? ? ; <i32*> [#uses=1]
> - ? ? ? %tmp424 = load i32* %tmp423, align 4 ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp425 = icmp eq i32 %tmp424, 2 ? ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp425, label %cond_next454, label %cond_true428
> -
> -cond_true428: ? ? ? ? ?; preds = %cond_next385
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1744, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next454: ? ? ? ? ?; preds = %cond_next385
> - ? ? ? lshr i32 %tmp409, 9 ? ? ? ? ? ? ; <i32>:0 [#uses=1]
> - ? ? ? trunc i32 %0 to i8 ? ? ? ? ? ? ?; <i8>:1 [#uses=1]
> - ? ? ? %tmp463464 = and i8 %1, 127 ? ? ? ? ? ? ; <i8> [#uses=1]
> - ? ? ? %tmp463464465466 = zext i8 %tmp463464 to i64 ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp467 = getelementptr [48 x i8]* @mode_size, i32 0, i64 %tmp463464465466 ? ? ? ? ? ? ?; <i8*> [#uses=1]
> - ? ? ? %tmp468 = load i8* %tmp467, align 1 ? ? ? ? ? ? ; <i8> [#uses=1]
> - ? ? ? %tmp468469553 = zext i8 %tmp468 to i16 ? ? ? ? ?; <i16> [#uses=1]
> - ? ? ? %tmp470471 = shl i16 %tmp468469553, 3 ? ? ? ? ? ; <i16> [#uses=1]
> - ? ? ? %tmp470471472 = zext i16 %tmp470471 to i64 ? ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp473 = tail call %struct.tree_node* @size_int_kind( i64 %tmp470471472, i32 2 ) ? ? ? ? ? ? ? ; <%struct.tree_node*> [#uses=1]
> - ? ? ? store %struct.tree_node* %tmp473, %struct.tree_node** %tmp51, align 8
> - ? ? ? ret void
> -
> -bb478: ? ? ? ? ; preds = %cond_next57
> - ? ? ? br i1 %tmp40, label %cond_next500, label %cond_true497
> -
> -cond_true497: ? ? ? ? ?; preds = %bb478
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %type, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1755, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next500: ? ? ? ? ?; preds = %bb478
> - ? ? ? %tmp506 = getelementptr %struct.tree_node* %type, i32 0, i32 0, i32 0, i32 1 ? ? ? ? ? ?; <%struct.tree_node**> [#uses=1]
> - ? ? ? %tmp507 = load %struct.tree_node** %tmp506, align 8 ? ? ? ? ? ? ; <%struct.tree_node*> [#uses=2]
> - ? ? ? %tmp511 = getelementptr %struct.tree_node* %tmp507, i32 0, i32 0, i32 0, i32 3 ? ? ? ? ?; <i8*> [#uses=1]
> - ? ? ? %tmp511512 = bitcast i8* %tmp511 to i32* ? ? ? ? ? ? ? ?; <i32*> [#uses=1]
> - ? ? ? %tmp513 = load i32* %tmp511512, align 8 ? ? ? ? ; <i32> [#uses=2]
> - ? ? ? %tmp515516 = trunc i32 %tmp513 to i8 ? ? ? ? ? ?; <i8> [#uses=1]
> - ? ? ? %tmp515516517518 = zext i8 %tmp515516 to i64 ? ? ? ? ? ?; <i64> [#uses=1]
> - ? ? ? %tmp519 = getelementptr [0 x i32]* @tree_code_type, i32 0, i64 %tmp515516517518 ? ? ? ? ; <i32*> [#uses=1]
> - ? ? ? %tmp520 = load i32* %tmp519, align 4 ? ? ? ? ? ?; <i32> [#uses=1]
> - ? ? ? %tmp521 = icmp eq i32 %tmp520, 2 ? ? ? ? ? ? ? ?; <i1> [#uses=1]
> - ? ? ? br i1 %tmp521, label %cond_next527, label %cond_true524
> -
> -cond_true524: ? ? ? ? ?; preds = %cond_next500
> - ? ? ? tail call void @tree_class_check_failed( %struct.tree_node* %tmp507, i32 2, i8* getelementptr ([28 x i8]* @.str, i32 0, i64 0), i32 1755, i8* getelementptr ([12 x i8]* @__FUNCTION__.22683, i32 0, i32 0) )
> - ? ? ? unreachable
> -
> -cond_next527: ? ? ? ? ?; preds = %cond_next500
> - ? ? ? %tmp545 = and i32 %tmp513, 8192 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp547 = and i32 %tmp32, -8193 ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? %tmp548 = or i32 %tmp547, %tmp545 ? ? ? ? ? ? ? ; <i32> [#uses=1]
> - ? ? ? store i32 %tmp548, i32* %tmp3031, align 8
> - ? ? ? ret void
> -
> -UnifiedReturnBlock: ? ? ? ? ? ?; preds = %cond_next57, %cond_next46, %cond_false
> - ? ? ? ret void
> -}
> -
> -declare void @fancy_abort(i8*, i32, i8*)
> -
> -declare void @tree_class_check_failed(%struct.tree_node*, i32, i8*, i32, i8*)
> -
> -declare i32 @smallest_mode_for_size(i32, i32)
> -
> -declare %struct.tree_node* @size_int_kind(i64, i32)
> -
> -declare void @tree_check_failed(%struct.tree_node*, i8*, i32, i8*, ...)
> -
> -declare i32 @tree_int_cst_sgn(%struct.tree_node*)
>
> Modified: llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll (original)
> +++ llvm/trunk/test/CodeGen/X86/2009-08-06-inlineasm.ll Tue Mar ?9 21:07:41 2010
> @@ -1,8 +1,10 @@
> -; RUN: llc < %s
> +; RUN: llc -mtriple=i386-pc-linux-gnu < %s
> ?; PR4668
> -; ModuleID = '<stdin>'
> -target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32"
> -target triple = "i386-pc-linux-gnu"
> +; XFAIL: *
> +; FIXME: If the coalescer happens to coalesce %level.1 with the copy to EAX
> +; (for ret) then this will fail to compile. The fundamental problem is
> +; once the coalescer fixes a virtual register to physical register we can't
> +; evict it.
>
> ?define i32 @x(i32 %qscale) nounwind {
> ?entry:
>
> Added: llvm/trunk/test/CodeGen/X86/machine-cse.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/machine-cse.ll?rev=98132&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/machine-cse.ll (added)
> +++ llvm/trunk/test/CodeGen/X86/machine-cse.ll Tue Mar ?9 21:07:41 2010
> @@ -0,0 +1,39 @@
> +; RUN: llc -mtriple=x86_64-apple-darwin < %s | FileCheck %s
> +; rdar://7610418
> +
> +%ptr = type { i8* }
> +%struct.s1 = type { %ptr, %ptr }
> +%struct.s2 = type { i32, i8*, i8*, [256 x %struct.s1*], [8 x i32], i64, i8*, i32, i64, i64, i32, %struct.s3*, %struct.s3*, [49 x i64] }
> +%struct.s3 = type { %struct.s3*, %struct.s3*, i32, i32, i32 }
> +
> +define fastcc i8* @t(i64 %size) nounwind {
> +entry:
> +; CHECK: t:
> +; CHECK: leaq (%rax,%rax,4)
> + ?%0 = zext i32 undef to i64
> + ?%1 = getelementptr inbounds %struct.s2* null, i64 %0
> + ?br i1 undef, label %bb1, label %bb2
> +
> +bb1:
> +; CHECK: %bb1
> +; CHECK-NOT: shlq $9
> +; CHECK-NOT: leaq
> +; CHECK: call
> + ?%2 = getelementptr inbounds %struct.s2* null, i64 %0, i32 0
> + ?call void @bar(i32* %2) nounwind
> + ?unreachable
> +
> +bb2:
> +; CHECK: %bb2
> +; CHECK-NOT: leaq
> +; CHECK: callq
> + ?%3 = call fastcc i8* @foo(%struct.s2* %1) nounwind
> + ?unreachable
> +
> +bb3:
> + ?ret i8* undef
> +}
> +
> +declare void @bar(i32*)
> +
> +declare fastcc i8* @foo(%struct.s2*) nounwind
>
> Modified: llvm/trunk/test/CodeGen/X86/pre-split6.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pre-split6.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/pre-split6.ll (original)
> +++ llvm/trunk/test/CodeGen/X86/pre-split6.ll Tue Mar ?9 21:07:41 2010
> @@ -1,4 +1,4 @@
> -; RUN: llc < %s -march=x86 -mattr=+sse2 -pre-alloc-split | grep {divsd 8} | count 1
> +; RUN: llc < %s -march=x86 -mattr=+sse2 -pre-alloc-split | grep {divsd 24} | count 1

This is failing on linux, it comes out as
	divsd 16(%esp), %xmm0         # 8-byte Folded Reload

> ?@current_surfaces.b = external global i1 ? ? ? ? ? ? ? ; <i1*> [#uses=1]
>
>
> Modified: llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll?rev=98132&r1=98131&r2=98132&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll (original)
> +++ llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll Tue Mar ?9 21:07:41 2010
> @@ -1,5 +1,6 @@
> ?; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -relocation-model=pic -disable-fp-elim -color-ss-with-regs -stats -info-output-file - > %t
> -; RUN: ? grep stackcoloring %t | grep "stack slot refs replaced with reg refs" ?| grep 8
> +; RuN: ? grep asm-printer %t | grep 156

Typo, RuN -> RUN.

 - Daniel

> +; RUN: ? grep stackcoloring %t | grep "stack slot refs replaced with reg refs" ?| grep 4
>
> ? ? ? ?type { [62 x %struct.Bitvec*] } ? ? ? ? ; type %0
> ? ? ? ?type { i8* } ? ? ? ? ? ?; type %1
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From evan.cheng at apple.com  Wed Mar 10 01:07:46 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 07:07:46 -0000
Subject: [llvm-commits] [llvm] r98141 -
	/llvm/trunk/test/CodeGen/X86/pre-split6.ll
Message-ID: <20100310070746.12A322A6C12C@llvm.org>

Author: evancheng
Date: Wed Mar 10 01:07:45 2010
New Revision: 98141

URL: http://llvm.org/viewvc/llvm-project?rev=98141&view=rev
Log:
Unbreak test on Linux.

Modified:
    llvm/trunk/test/CodeGen/X86/pre-split6.ll

Modified: llvm/trunk/test/CodeGen/X86/pre-split6.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pre-split6.ll?rev=98141&r1=98140&r2=98141&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/pre-split6.ll (original)
+++ llvm/trunk/test/CodeGen/X86/pre-split6.ll Wed Mar 10 01:07:45 2010
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 -mattr=+sse2 -pre-alloc-split | grep {divsd	24} | count 1
+; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=+sse2 -pre-alloc-split | grep {divsd	24} | count 1
 
 @current_surfaces.b = external global i1		; <i1*> [#uses=1]
 




From evan.cheng at apple.com  Wed Mar 10 01:07:55 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 07:07:55 -0000
Subject: [llvm-commits] [llvm] r98142 -
	/llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll
Message-ID: <20100310070755.C29C42A6C12C@llvm.org>

Author: evancheng
Date: Wed Mar 10 01:07:55 2010
New Revision: 98142

URL: http://llvm.org/viewvc/llvm-project?rev=98142&view=rev
Log:
Fix typo.

Modified:
    llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll

Modified: llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll?rev=98142&r1=98141&r2=98142&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll (original)
+++ llvm/trunk/test/CodeGen/X86/stack-color-with-reg.ll Wed Mar 10 01:07:55 2010
@@ -1,5 +1,5 @@
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin10 -relocation-model=pic -disable-fp-elim -color-ss-with-regs -stats -info-output-file - > %t
-; RuN:   grep asm-printer %t | grep 156
+; RUN:   grep asm-printer %t | grep 156
 ; RUN:   grep stackcoloring %t | grep "stack slot refs replaced with reg refs"  | grep 4
 
 	type { [62 x %struct.Bitvec*] }		; type %0




From sabre at nondot.org  Wed Mar 10 01:20:43 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 07:20:43 -0000
Subject: [llvm-commits] [llvm] r98145 - in /llvm/trunk:
 include/llvm/MC/MCAsmInfo.h include/llvm/Target/TargetLoweringObjectFile.h
 lib/CodeGen/AsmPrinter/DwarfException.cpp
 lib/CodeGen/TargetLoweringObjectFileImpl.cpp lib/MC/MCAsmInfo.cpp
 lib/MC/MCAsmInfoDarwin.cpp lib/Target/TargetLoweringObjectFile.cpp
Message-ID: <20100310072043.354F92A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 01:20:42 2010
New Revision: 98145

URL: http://llvm.org/viewvc/llvm-project?rev=98145&view=rev
Log:
move three lowering hooks from MAI to TLOF and make one of them
semantic instead of syntactic.  This completes MCization of
darwin/x86[-64]!

Modified:
    llvm/trunk/include/llvm/MC/MCAsmInfo.h
    llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/trunk/lib/MC/MCAsmInfo.cpp
    llvm/trunk/lib/MC/MCAsmInfoDarwin.cpp
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp

Modified: llvm/trunk/include/llvm/MC/MCAsmInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAsmInfo.h?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAsmInfo.h (original)
+++ llvm/trunk/include/llvm/MC/MCAsmInfo.h Wed Mar 10 01:20:42 2010
@@ -247,18 +247,6 @@
     /// encode inline subroutine information.
     bool DwarfUsesInlineInfoSection;         // Defaults to false.
 
-    /// Is_EHSymbolPrivate - If set, the "_foo.eh" is made private so that it
-    /// doesn't show up in the symbol table of the object file.
-    bool Is_EHSymbolPrivate;                 // Defaults to true.
-
-    /// GlobalEHDirective - This is the directive used to make exception frame
-    /// tables globally visible.
-    const char *GlobalEHDirective;           // Defaults to NULL.
-
-    /// SupportsWeakEmptyEHFrame - True if target assembler and linker will
-    /// handle a weak_definition of constant 0 for an omitted EH frame.
-    bool SupportsWeakOmittedEHFrame;         // Defaults to true.
-
     /// DwarfSectionOffsetDirective - Special section offset directive.
     const char* DwarfSectionOffsetDirective; // Defaults to NULL
     
@@ -419,15 +407,6 @@
     bool doesDwarfUsesInlineInfoSection() const {
       return DwarfUsesInlineInfoSection;
     }
-    bool is_EHSymbolPrivate() const {
-      return Is_EHSymbolPrivate;
-    }
-    const char *getGlobalEHDirective() const {
-      return GlobalEHDirective;
-    }
-    bool getSupportsWeakOmittedEHFrame() const {
-      return SupportsWeakOmittedEHFrame;
-    }
     const char *getDwarfSectionOffsetDirective() const {
       return DwarfSectionOffsetDirective;
     }

Modified: llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h (original)
+++ llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h Wed Mar 10 01:20:42 2010
@@ -87,11 +87,23 @@
   const MCSection *DwarfRangesSection;
   const MCSection *DwarfMacroInfoSection;
   
+  /// SupportsWeakEmptyEHFrame - True if target object file supports a
+  /// weak_definition of constant 0 for an omitted EH frame.
+  bool SupportsWeakOmittedEHFrame;
+  
+  /// IsFunctionEHSymbolGlobal - This flag is set to true if the ".eh" symbol
+  /// for a function should be marked .globl.
+  bool IsFunctionEHSymbolGlobal;
+  
+  /// IsFunctionEHFrameSymbolPrivate - This flag is set to true if the
+  /// "EH_frame" symbol for EH information should be an assembler temporary (aka
+  /// private linkage, aka an L or .L label) or false if it should be a normal
+  /// non-.globl label.  This defaults to true.
+  bool IsFunctionEHFrameSymbolPrivate;
 public:
   
   MCContext &getContext() const { return *Ctx; }
   
-
   virtual ~TargetLoweringObjectFile();
   
   /// Initialize - this method must be called before any actual lowering is
@@ -101,6 +113,15 @@
     Ctx = &ctx;
   }
   
+  bool isFunctionEHSymbolGlobal() const {
+    return IsFunctionEHSymbolGlobal;
+  }
+  bool isFunctionEHFrameSymbolPrivate() const {
+    return IsFunctionEHFrameSymbolPrivate;
+  }
+  bool getSupportsWeakOmittedEHFrame() const {
+    return SupportsWeakOmittedEHFrame;
+  }
   
   const MCSection *getTextSection() const { return TextSection; }
   const MCSection *getDataSection() const { return DataSection; }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Wed Mar 10 01:20:42 2010
@@ -85,7 +85,7 @@
   Asm->OutStreamer.SwitchSection(TLOF.getEHFrameSection());
 
   MCSymbol *EHFrameSym;
-  if (MAI->is_EHSymbolPrivate())
+  if (TLOF.isFunctionEHFrameSymbolPrivate())
     EHFrameSym = getDWLabel("EH_frame", Index);
   else
     EHFrameSym = Asm->OutContext.GetOrCreateSymbol(Twine("EH_frame") + 
@@ -193,9 +193,8 @@
 
   // Externally visible entry into the functions eh frame info. If the
   // corresponding function is static, this should not be externally visible.
-  if (!TheFunc->hasLocalLinkage())
-    if (const char *GlobalEHDirective = MAI->getGlobalEHDirective())
-      O << GlobalEHDirective << *EHFrameInfo.FunctionEHSym << '\n';
+  if (!TheFunc->hasLocalLinkage() && TLOF.isFunctionEHSymbolGlobal())
+    Asm->OutStreamer.EmitSymbolAttribute(EHFrameInfo.FunctionEHSym,MCSA_Global);
 
   // If corresponding function is weak definition, this should be too.
   if (TheFunc->isWeakForLinker() && MAI->getWeakDefDirective())
@@ -215,7 +214,7 @@
   if (!EHFrameInfo.hasCalls && !UnwindTablesMandatory &&
       (!TheFunc->isWeakForLinker() ||
        !MAI->getWeakDefDirective() ||
-       MAI->getSupportsWeakOmittedEHFrame())) {
+       TLOF.getSupportsWeakOmittedEHFrame())) {
     Asm->OutStreamer.EmitAssignment(EHFrameInfo.FunctionEHSym,
                                     MCConstantExpr::Create(0, Asm->OutContext));
     // This name has no connection to the function, so it might get
@@ -981,9 +980,10 @@
   Asm->OutStreamer.EmitLabel(getDWLabel("eh_func_end", SubprogramCount));
   EmitExceptionTable();
 
+  const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
   MCSymbol *FunctionEHSym =
     Asm->GetSymbolWithGlobalValueBase(MF->getFunction(), ".eh",
-                                      Asm->MAI->is_EHSymbolPrivate());
+                                      TLOF.isFunctionEHFrameSymbolPrivate());
   
   // Save EH frame information
   EHFrames.push_back(FunctionEHFrameInfo(FunctionEHSym, SubprogramCount,

Modified: llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Wed Mar 10 01:20:42 2010
@@ -466,6 +466,14 @@
 
 void TargetLoweringObjectFileMachO::Initialize(MCContext &Ctx,
                                                const TargetMachine &TM) {
+  // _foo.eh symbols are currently always exported so that the linker knows
+  // about them.  This is not necessary on 10.6 and later, but it
+  // doesn't hurt anything.
+  // FIXME: I need to get this from Triple.
+  IsFunctionEHSymbolGlobal = true;
+  IsFunctionEHFrameSymbolPrivate = false;
+  SupportsWeakOmittedEHFrame = false;
+  
   if (UniquingMap != 0)
     ((MachOUniqueMapTy*)UniquingMap)->clear();
   TargetLoweringObjectFile::Initialize(Ctx, TM);

Modified: llvm/trunk/lib/MC/MCAsmInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAsmInfo.cpp?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAsmInfo.cpp (original)
+++ llvm/trunk/lib/MC/MCAsmInfo.cpp Wed Mar 10 01:20:42 2010
@@ -68,9 +68,6 @@
   ExceptionsType = ExceptionHandling::None;
   DwarfRequiresFrameSection = true;
   DwarfUsesInlineInfoSection = false;
-  Is_EHSymbolPrivate = true;
-  GlobalEHDirective = 0;
-  SupportsWeakOmittedEHFrame = true;
   DwarfSectionOffsetDirective = 0;
 
   AsmTransCBE = 0;

Modified: llvm/trunk/lib/MC/MCAsmInfoDarwin.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAsmInfoDarwin.cpp?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAsmInfoDarwin.cpp (original)
+++ llvm/trunk/lib/MC/MCAsmInfoDarwin.cpp Wed Mar 10 01:20:42 2010
@@ -40,19 +40,8 @@
   HiddenVisibilityAttr = MCSA_PrivateExtern;
   // Doesn't support protected visibility.
   ProtectedVisibilityAttr = MCSA_Global;
-
   
   HasDotTypeDotSizeDirective = false;
   HasNoDeadStrip = true;
-  // Note: Even though darwin has the .lcomm directive, it is just a synonym for
-  // zerofill, so we prefer to use .zerofill.
-
-  // _foo.eh symbols are currently always exported so that the linker knows
-  // about them.  This is not necessary on 10.6 and later, but it
-  // doesn't hurt anything.
-  // FIXME: I need to get this from Triple.
-  Is_EHSymbolPrivate = false;
-  GlobalEHDirective = "\t.globl\t";
-  SupportsWeakOmittedEHFrame = false;
 }
 

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98145&r1=98144&r2=98145&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Wed Mar 10 01:20:42 2010
@@ -56,6 +56,10 @@
   DwarfARangesSection = 0;
   DwarfRangesSection = 0;
   DwarfMacroInfoSection = 0;
+  
+  IsFunctionEHSymbolGlobal = false;
+  IsFunctionEHFrameSymbolPrivate = true;
+  SupportsWeakOmittedEHFrame = true;
 }
 
 TargetLoweringObjectFile::~TargetLoweringObjectFile() {




From baldrick at free.fr  Wed Mar 10 03:52:07 2010
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 10 Mar 2010 09:52:07 -0000
Subject: [llvm-commits] [dragonegg] r98146 - /dragonegg/trunk/llvm-debug.cpp
Message-ID: <20100310095207.7363C2A6C12C@llvm.org>

Author: baldrick
Date: Wed Mar 10 03:52:07 2010
New Revision: 98146

URL: http://llvm.org/viewvc/llvm-project?rev=98146&view=rev
Log:
Port commits 97948, 97962 and 97977 (dpatel) from llvm-gcc:
- Avoid DIDescriptor.isNull() checks.
- Revert r97948.
- Avoid using DIDescriptor.isNull(). 
  One character "!" makes all the difference!

Modified:
    dragonegg/trunk/llvm-debug.cpp

Modified: dragonegg/trunk/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-debug.cpp?rev=98146&r1=98145&r2=98146&view=diff
==============================================================================
--- dragonegg/trunk/llvm-debug.cpp (original)
+++ dragonegg/trunk/llvm-debug.cpp Wed Mar 10 03:52:07 2010
@@ -1059,7 +1059,7 @@
     MainTy = getOrCreateType(TYPE_MAIN_VARIANT(type));
 
   DIType Ty = createVariantType(type, MainTy);
-  if (!Ty.isNull())
+  if (Ty.isValid())
     return Ty;
 
   // Work out details of type.




From baldrick at free.fr  Wed Mar 10 04:08:25 2010
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 10 Mar 2010 10:08:25 -0000
Subject: [llvm-commits] [dragonegg] r98147 - in /dragonegg/trunk:
	llvm-debug.cpp llvm-debug.h
Message-ID: <20100310100825.B37FC2A6C12C@llvm.org>

Author: baldrick
Date: Wed Mar 10 04:08:25 2010
New Revision: 98147

URL: http://llvm.org/viewvc/llvm-project?rev=98147&view=rev
Log:
Port commit 98022 (dpatel) from llvm-gcc:
Start using DIFile. Corresponding llvm patch is r98020.

Modified:
    dragonegg/trunk/llvm-debug.cpp
    dragonegg/trunk/llvm-debug.h

Modified: dragonegg/trunk/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-debug.cpp?rev=98147&r1=98146&r2=98147&view=diff
==============================================================================
--- dragonegg/trunk/llvm-debug.cpp (original)
+++ dragonegg/trunk/llvm-debug.cpp Wed Mar 10 04:08:25 2010
@@ -275,7 +275,7 @@
     ArtificialFnWithAbstractOrigin = true;
 
   DIDescriptor SPContext = ArtificialFnWithAbstractOrigin ?
-    getOrCreateCompileUnit(main_input_filename) :
+    getOrCreateFile(main_input_filename) :
     findRegion (DECL_CONTEXT(FnDecl));
 
   // Creating context may have triggered creation of this SP descriptor. So
@@ -316,7 +316,7 @@
     DebugFactory.CreateSubprogram(SPContext,
                                   FnName, FnName,
                                   LinkageName,
-                                  getOrCreateCompileUnit(Loc.file), lineno,
+                                  getOrCreateFile(Loc.file), lineno,
                                   FNType,
                                   Fn->hasInternalLinkage(),
                                   true /*definition*/,
@@ -340,7 +340,7 @@
   expanded_location Loc = GetNodeLocation(Node, false);
   DINameSpace DNS =
     DebugFactory.CreateNameSpace(Context, GetNodeName(Node),
-                                 getOrCreateCompileUnit(Loc.file), Loc.line);
+                                 getOrCreateFile(Loc.file), Loc.line);
 
   NameSpaceCache[Node] = WeakVH(DNS.getNode());
   return DNS;
@@ -349,7 +349,7 @@
 /// findRegion - Find tree_node N's region.
 DIDescriptor DebugInfo::findRegion(tree Node) {
   if (Node == NULL_TREE)
-    return getOrCreateCompileUnit(main_input_filename);
+    return getOrCreateFile(main_input_filename);
 
   std::map<tree_node *, WeakVH>::iterator I = RegionMap.find(Node);
   if (I != RegionMap.end())
@@ -369,7 +369,7 @@
   }
 
   // Otherwise main compile unit covers everything.
-  return getOrCreateCompileUnit(main_input_filename);
+  return getOrCreateFile(main_input_filename);
 }
 
 /// EmitFunctionEnd - Constructs the debug code for exiting a declarative
@@ -410,7 +410,7 @@
       Ty = DebugFactory.CreateArtificialType(Ty);
   llvm::DIVariable D =
     DebugFactory.CreateVariable(Tag, VarScope,
-                                Name, getOrCreateCompileUnit(Loc.file),
+                                Name, getOrCreateFile(Loc.file),
                                 Loc.line, Ty);
 
   // Insert an llvm.dbg.declare into the current block.
@@ -470,7 +470,7 @@
       LinkageName = GV->getName();
   DebugFactory.CreateGlobalVariable(findRegion(DECL_CONTEXT(decl)),
                                     DispName, DispName, LinkageName,
-                                    getOrCreateCompileUnit(Loc.file), Loc.line,
+                                    getOrCreateFile(Loc.file), Loc.line,
                                     TyD, GV->hasInternalLinkage(),
                                     true/*definition*/, GV);
 }
@@ -517,9 +517,9 @@
   }
 
   return 
-    DebugFactory.CreateBasicType(getOrCreateCompileUnit(main_input_filename),
+    DebugFactory.CreateBasicType(getOrCreateFile(main_input_filename),
                                  TypeName, 
-                                 getOrCreateCompileUnit(main_input_filename),
+                                 getOrCreateFile(main_input_filename),
                                  0, Size, Align,
                                  0, 0, Encoding);
 }
@@ -548,9 +548,9 @@
   sprintf(FwdTypeName, "fwd.type.%d", FwdTypeCount++);
   llvm::DIType FwdType = 
     DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_subroutine_type,
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      FwdTypeName,
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      0, 0, 0, 0, 0,
                                      llvm::DIType(), llvm::DIArray());
   llvm::TrackingVH<llvm::MDNode> FwdTypeNode = FwdType.getNode();
@@ -591,7 +591,7 @@
     DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_subroutine_type,
                                      findRegion(TYPE_CONTEXT(type)), 
                                      StringRef(),
-                                     getOrCreateCompileUnit(NULL), 
+                                     getOrCreateFile(main_input_filename),
                                      0, 0, 0, 0, 0,
                                      llvm::DIType(), EltTypeArray);
 
@@ -620,7 +620,7 @@
       DIType Ty = 
         DebugFactory.CreateDerivedType(Tag, findRegion(DECL_CONTEXT(TyName)),
                                        GetNodeName(TyName), 
-                                       getOrCreateCompileUnit(TypeNameLoc.file),
+                                       getOrCreateFile(TypeNameLoc.file),
                                        TypeNameLoc.line,
                                        0 /*size*/,
                                        0 /*align*/,
@@ -636,7 +636,7 @@
     DebugFactory.CreateDerivedType(Tag, findRegion(TYPE_CONTEXT(type)), 
                                    Tag == DW_TAG_pointer_type ? 
                                    StringRef() : PName,
-                                   getOrCreateCompileUnit(NULL), 
+                                   getOrCreateFile(main_input_filename),
                                    0 /*line no*/, 
                                    NodeSizeInBits(type),
                                    NodeAlignInBits(type),
@@ -698,7 +698,7 @@
   return DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_array_type,
                                           findRegion(TYPE_CONTEXT(type)), 
                                           StringRef(),
-                                          getOrCreateCompileUnit(Loc.file), 0, 
+                                          getOrCreateFile(Loc.file), 0, 
                                           NodeSizeInBits(type), 
                                           NodeAlignInBits(type), 0, 0,
                                           getOrCreateType(EltTy),
@@ -732,7 +732,7 @@
   return DebugFactory.CreateCompositeType(llvm::dwarf::DW_TAG_enumeration_type,
                                           findRegion(TYPE_CONTEXT(type)), 
                                           GetNodeName(type), 
-                                          getOrCreateCompileUnit(Loc.file), 
+                                          getOrCreateFile(Loc.file), 
                                           Loc.line,
                                           NodeSizeInBits(type), 
                                           NodeAlignInBits(type), 0, 0,
@@ -750,8 +750,7 @@
 //TODO  if (TYPE_LANG_SPECIFIC (type)
 //TODO      && lang_hooks.types.is_runtime_specific_type (type))
 //TODO    {
-//TODO      DICompileUnit CU = getOrCreateCompileUnit(main_input_filename);
-//TODO      unsigned CULang = CU.getLanguage();
+//TODO      unsigned CULang = TheCU.getLanguage();
 //TODO      switch (CULang) {
 //TODO      case DW_LANG_ObjC_plus_plus :
 //TODO        RunTimeLang = DW_LANG_ObjC_plus_plus;
@@ -802,7 +801,7 @@
     DebugFactory.CreateCompositeType(Tag, 
                                      TyContext,
                                      FwdName.c_str(),
-                                     getOrCreateCompileUnit(Loc.file), 
+                                     getOrCreateFile(Loc.file), 
                                      Loc.line, 
                                      0, 0, 0, SFlags | llvm::DIType::FlagFwdDecl,
                                      llvm::DIType(), llvm::DIArray(),
@@ -852,7 +851,7 @@
       DIType DTy = 
         DebugFactory.CreateDerivedType(DW_TAG_inheritance, 
                                        findRegion(TYPE_CONTEXT(type)), StringRef(),
-                                       llvm::DICompileUnit(), 0,0,0, 
+                                       llvm::DIFile(), 0,0,0, 
                                        Offset,
                                        BFlags, BaseClass);
       EltTys.push_back(DTy);
@@ -898,7 +897,7 @@
       DebugFactory.CreateDerivedType(DW_TAG_member, 
                                      findRegion(DECL_CONTEXT(Member)),
                                      MemberName, 
-                                     getOrCreateCompileUnit(MemLoc.file),
+                                     getOrCreateFile(MemLoc.file),
                                      MemLoc.line, NodeSizeInBits(Member),
                                      NodeAlignInBits(FieldNodeType),
                                      int_bit_position(Member), 
@@ -937,7 +936,7 @@
         DebugFactory.CreateSubprogram(findRegion(DECL_CONTEXT(Member)), 
                                       MemberName, MemberName,
                                       LinkageName, 
-                                      getOrCreateCompileUnit(MemLoc.file),
+                                      getOrCreateFile(MemLoc.file),
                                       MemLoc.line, SPTy, false, false,
                                       Virtuality, VIndex, ContainingType,
                                       DECL_ARTIFICIAL (Member));
@@ -962,7 +961,7 @@
   llvm::DICompositeType RealDecl =
     DebugFactory.CreateCompositeType(Tag, findRegion(TYPE_CONTEXT(type)),
                                      GetNodeName(type),
-                                     getOrCreateCompileUnit(Loc.file),
+                                     getOrCreateFile(Loc.file),
                                      Loc.line, 
                                      NodeSizeInBits(type), NodeAlignInBits(type),
                                      0, SFlags, llvm::DIType(), Elements,
@@ -990,7 +989,7 @@
       Ty = DebugFactory.CreateDerivedType(DW_TAG_typedef, 
                                           findRegion(DECL_CONTEXT(TyDef)),
                                           GetNodeName(TyDef), 
-                                          getOrCreateCompileUnit(TypeDefLoc.file),
+                                          getOrCreateFile(TypeDefLoc.file),
                                           TypeDefLoc.line,
                                           0 /*size*/,
                                           0 /*align*/,
@@ -1006,7 +1005,7 @@
     Ty = DebugFactory.CreateDerivedType(DW_TAG_volatile_type, 
                                         findRegion(TYPE_CONTEXT(type)), 
                                         StringRef(),
-                                        getOrCreateCompileUnit(NULL), 
+                                        getOrCreateFile(main_input_filename),
                                         0 /*line no*/, 
                                         NodeSizeInBits(type),
                                         NodeAlignInBits(type),
@@ -1020,7 +1019,7 @@
     Ty =  DebugFactory.CreateDerivedType(DW_TAG_const_type, 
                                          findRegion(TYPE_CONTEXT(type)), 
                                          StringRef(),
-                                         getOrCreateCompileUnit(NULL), 
+                                         getOrCreateFile(main_input_filename),
                                          0 /*line no*/, 
                                          NodeSizeInBits(type),
                                          NodeAlignInBits(type),
@@ -1131,9 +1130,9 @@
   // module does not contain any main compile unit then the code generator 
   // will emit multiple compile units in the output object file.
   if (!strcmp (main_input_filename, ""))
-    getOrCreateCompileUnit("<stdin>", true);
+    TheCU = getOrCreateCompileUnit("<stdin>", true);
   else
-    getOrCreateCompileUnit(main_input_filename, true);
+    TheCU = getOrCreateCompileUnit(main_input_filename, true);
 }
 
 /// getOrCreateCompileUnit - Get the compile unit from the cache or 
@@ -1146,10 +1145,6 @@
     else
       FullPath = main_input_filename;
   }
-  std::map<std::string, WeakVH >::iterator I = CUCache.find(FullPath);
-  if (I != CUCache.end())
-    if (Value *M = I->second)
-      return DICompileUnit(cast<MDNode>(M));
 
   // Get source file information.
   std::string Directory;
@@ -1185,11 +1180,25 @@
   unsigned ObjcRunTimeVer = 0;
 //  if (flag_objc_abi != 0 && flag_objc_abi != -1)
 //    ObjcRunTimeVer = flag_objc_abi;
-  DICompileUnit NewCU = DebugFactory.CreateCompileUnit(LangTag, FileName.c_str(), 
-                                                       Directory.c_str(),
-                                                       version_string, isMain,
-                                                       optimize, Flags,
-                                                       ObjcRunTimeVer);
-  CUCache[FullPath] = WeakVH(NewCU.getNode());
-  return NewCU;
+  return DebugFactory.CreateCompileUnit(LangTag, FileName.c_str(),
+                                        Directory.c_str(),
+                                        version_string, isMain,
+                                        optimize, Flags,
+                                        ObjcRunTimeVer);
+}
+
+/// getOrCreateFile - Get DIFile descriptor.
+DIFile DebugInfo::getOrCreateFile(const char *FullPath) {
+  if (!FullPath) {
+    if (!strcmp (main_input_filename, ""))
+      FullPath = "<stdin>";
+    else
+      FullPath = main_input_filename;
+  }
+
+  // Get source file information.
+  std::string Directory;
+  std::string FileName;
+  DirectoryAndFile(FullPath, Directory, FileName);
+  return DebugFactory.CreateFile(FileName, Directory, TheCU);
 }

Modified: dragonegg/trunk/llvm-debug.h
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-debug.h?rev=98147&r1=98146&r2=98147&view=diff
==============================================================================
--- dragonegg/trunk/llvm-debug.h (original)
+++ dragonegg/trunk/llvm-debug.h Wed Mar 10 04:08:25 2010
@@ -57,13 +57,14 @@
   const char *PrevFullPath;             // Previous location file encountered.
   int PrevLineNo;                       // Previous location line# encountered.
   BasicBlock *PrevBB;                   // Last basic block encountered.
-  
+
+  DICompileUnit TheCU;                  // The compile unit.
+
   // This counter counts debug info for forward referenced subroutine types.
   // This counter is used to create unique name for such types so that their 
   // debug info (through MDNodes) is not shared accidently.
   unsigned FwdTypeCount;
 
-  std::map<std::string, WeakVH > CUCache;
   std::map<tree_node *, WeakVH > TypeCache;
                                         // Cache of previously constructed 
                                         // Types.
@@ -145,6 +146,9 @@
   DICompileUnit getOrCreateCompileUnit(const char *FullPath,
                                        bool isMain = false);
 
+  /// getOrCreateFile - Get DIFile descriptor.
+  DIFile getOrCreateFile(const char *FullPath);
+
   /// findRegion - Find tree_node N's region.
   DIDescriptor findRegion(tree_node *n);
   




From richard at xmos.com  Wed Mar 10 05:24:03 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 11:24:03 -0000
Subject: [llvm-commits] [llvm] r98148 -
	/llvm/trunk/test/CodeGen/XCore/addsub64.ll
Message-ID: <20100310112403.4E8992A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 05:24:03 2010
New Revision: 98148

URL: http://llvm.org/viewvc/llvm-project?rev=98148&view=rev
Log:
Convert test to FileCheck.

Modified:
    llvm/trunk/test/CodeGen/XCore/addsub64.ll

Modified: llvm/trunk/test/CodeGen/XCore/addsub64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/addsub64.ll?rev=98148&r1=98147&r2=98148&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/addsub64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/addsub64.ll Wed Mar 10 05:24:03 2010
@@ -1,12 +1,20 @@
-; RUN: llc < %s -march=xcore -mcpu=xs1b-generic > %t1.s
-; RUN: grep ladd %t1.s | count 2
-; RUN: grep lsub %t1.s | count 2
+; RUN: llc < %s -march=xcore | FileCheck %s
 define i64 @add64(i64 %a, i64 %b) {
 	%result = add i64 %a, %b
 	ret i64 %result
 }
+; CHECK: add64
+; CHECK: ldc r11, 0
+; CHECK-NEXT: ladd r2, r0, r0, r2, r11
+; CHECK-NEXT: ladd r2, r1, r1, r3, r2
+; CHECK-NEXT: retsp 0
 
 define i64 @sub64(i64 %a, i64 %b) {
 	%result = sub i64 %a, %b
 	ret i64 %result
 }
+; CHECK: sub64
+; CHECK: ldc r11, 0
+; CHECK-NEXT: lsub r2, r0, r0, r2, r11
+; CHECK-NEXT: lsub r2, r1, r1, r3, r2
+; CHECK-NEXT: retsp 0




From richard at xmos.com  Wed Mar 10 05:41:08 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 11:41:08 -0000
Subject: [llvm-commits] [llvm] r98150 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelDAGToDAG.cpp
 lib/Target/XCore/XCoreISelLowering.cpp lib/Target/XCore/XCoreISelLowering.h
 test/CodeGen/XCore/addsub64.ll
Message-ID: <20100310114108.9F0B12A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 05:41:08 2010
New Revision: 98150

URL: http://llvm.org/viewvc/llvm-project?rev=98150&view=rev
Log:
Lower add (mul a, b), c into MACCU / MACCS nodes which translate
directly to the maccu / maccs instructions. We handle this in
ExpandADDSUB since after type legalisation it is messy to
recognise these operations.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
    llvm/trunk/test/CodeGen/XCore/addsub64.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp?rev=98150&r1=98149&r2=98150&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp Wed Mar 10 05:41:08 2010
@@ -208,6 +208,18 @@
         return CurDAG->getMachineNode(XCore::LSUB_l5r, dl, MVT::i32, MVT::i32,
                                       Ops, 3);
       }
+      case XCoreISD::MACCU: {
+        SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
+                          N->getOperand(2), N->getOperand(3) };
+        return CurDAG->getMachineNode(XCore::MACCU_l4r, dl, MVT::i32, MVT::i32,
+                                      Ops, 4);
+      }
+      case XCoreISD::MACCS: {
+        SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
+                          N->getOperand(2), N->getOperand(3) };
+        return CurDAG->getMachineNode(XCore::MACCS_l4r, dl, MVT::i32, MVT::i32,
+                                      Ops, 4);
+      }
       // Other cases are autogenerated.
     }
   }

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98150&r1=98149&r2=98150&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 05:41:08 2010
@@ -54,6 +54,8 @@
     case XCoreISD::RETSP             : return "XCoreISD::RETSP";
     case XCoreISD::LADD              : return "XCoreISD::LADD";
     case XCoreISD::LSUB              : return "XCoreISD::LSUB";
+    case XCoreISD::MACCU             : return "XCoreISD::MACCU";
+    case XCoreISD::MACCS             : return "XCoreISD::MACCS";
     case XCoreISD::BR_JT             : return "XCoreISD::BR_JT";
     case XCoreISD::BR_JT32           : return "XCoreISD::BR_JT32";
     default                           : return NULL;
@@ -542,11 +544,76 @@
 }
 
 SDValue XCoreTargetLowering::
+TryExpandADDSUBWithMul(SDNode *N, SelectionDAG &DAG)
+{
+  SDValue Mul;
+  SDValue Other;
+  if (N->getOperand(0).getOpcode() == ISD::MUL) {
+    Mul = N->getOperand(0);
+    Other = N->getOperand(1);
+  } else if (N->getOperand(1).getOpcode() == ISD::MUL) {
+    Mul = N->getOperand(1);
+    Other = N->getOperand(0);
+  } else {
+    return SDValue();
+  }
+  DebugLoc dl = N->getDebugLoc();
+  SDValue LL, RL, AddendL, AddendH;
+  LL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                   Mul.getOperand(0),  DAG.getConstant(0, MVT::i32));
+  RL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                   Mul.getOperand(1),  DAG.getConstant(0, MVT::i32));
+  AddendL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                        Other,  DAG.getConstant(0, MVT::i32));
+  AddendH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                        Other,  DAG.getConstant(1, MVT::i32));
+  APInt HighMask = APInt::getHighBitsSet(64, 32);
+  unsigned LHSSB = DAG.ComputeNumSignBits(Mul.getOperand(0));
+  unsigned RHSSB = DAG.ComputeNumSignBits(Mul.getOperand(1));
+  if (DAG.MaskedValueIsZero(Mul.getOperand(0), HighMask) &&
+      DAG.MaskedValueIsZero(Mul.getOperand(1), HighMask)) {
+    // The inputs are both zero-extended.
+    SDValue Hi = DAG.getNode(XCoreISD::MACCU, dl,
+                             DAG.getVTList(MVT::i32, MVT::i32), AddendH,
+                             AddendL, LL, RL);
+    SDValue Lo(Hi.getNode(), 1);
+    return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi);
+  }
+  if (LHSSB > 32 && RHSSB > 32) {
+    // The inputs are both sign-extended.
+    SDValue Hi = DAG.getNode(XCoreISD::MACCS, dl,
+                             DAG.getVTList(MVT::i32, MVT::i32), AddendH,
+                             AddendL, LL, RL);
+    SDValue Lo(Hi.getNode(), 1);
+    return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi);
+  }
+  SDValue LH, RH;
+  LH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                   Mul.getOperand(0),  DAG.getConstant(1, MVT::i32));
+  RH = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                   Mul.getOperand(1),  DAG.getConstant(1, MVT::i32));
+  SDValue Hi = DAG.getNode(XCoreISD::MACCU, dl,
+                           DAG.getVTList(MVT::i32, MVT::i32), AddendH,
+                           AddendL, LL, RL);
+  SDValue Lo(Hi.getNode(), 1);
+  RH = DAG.getNode(ISD::MUL, dl, MVT::i32, LL, RH);
+  LH = DAG.getNode(ISD::MUL, dl, MVT::i32, LH, RL);
+  Hi = DAG.getNode(ISD::ADD, dl, MVT::i32, Hi, RH);
+  Hi = DAG.getNode(ISD::ADD, dl, MVT::i32, Hi, LH);
+  return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi);
+}
+
+SDValue XCoreTargetLowering::
 ExpandADDSUB(SDNode *N, SelectionDAG &DAG)
 {
   assert(N->getValueType(0) == MVT::i64 &&
          (N->getOpcode() == ISD::ADD || N->getOpcode() == ISD::SUB) &&
         "Unknown operand to lower!");
+
+  SDValue Result = TryExpandADDSUBWithMul(N, DAG);
+  if (Result.getNode() != 0)
+    return Result;
+
   DebugLoc dl = N->getDebugLoc();
   
   // Extract components

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98150&r1=98149&r2=98150&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Wed Mar 10 05:41:08 2010
@@ -54,6 +54,12 @@
       // Corresponds to LSUB instruction
       LSUB,
 
+      // Corresponds to MACCU instruction
+      MACCU,
+
+      // Corresponds to MACCS instruction
+      MACCS,
+
       // Jumptable branch.
       BR_JT,
 
@@ -140,6 +146,7 @@
               EVT VT) const;
   
     // Expand specifics
+    SDValue TryExpandADDSUBWithMul(SDNode *Op, SelectionDAG &DAG);
     SDValue ExpandADDSUB(SDNode *Op, SelectionDAG &DAG);
 
     virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;

Modified: llvm/trunk/test/CodeGen/XCore/addsub64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/addsub64.ll?rev=98150&r1=98149&r2=98150&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/addsub64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/addsub64.ll Wed Mar 10 05:41:08 2010
@@ -18,3 +18,27 @@
 ; CHECK-NEXT: lsub r2, r0, r0, r2, r11
 ; CHECK-NEXT: lsub r2, r1, r1, r3, r2
 ; CHECK-NEXT: retsp 0
+
+define i64 @maccu(i64 %a, i32 %b, i32 %c) {
+entry:
+	%0 = zext i32 %b to i64
+	%1 = zext i32 %c to i64
+	%2 = mul i64 %1, %0
+	%3 = add i64 %2, %a
+	ret i64 %3
+}
+; CHECK: maccu:
+; CHECK: maccu r1, r0, r3, r2
+; CHECK-NEXT: retsp 0
+
+define i64 @maccs(i64 %a, i32 %b, i32 %c) {
+entry:
+	%0 = sext i32 %b to i64
+	%1 = sext i32 %c to i64
+	%2 = mul i64 %1, %0
+	%3 = add i64 %2, %a
+	ret i64 %3
+}
+; CHECK: maccs:
+; CHECK: maccs r1, r0, r3, r2
+; CHECK-NEXT: retsp 0




From richard at xmos.com  Wed Mar 10 05:42:05 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 11:42:05 -0000
Subject: [llvm-commits] [llvm] r98151 -
	/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
Message-ID: <20100310114205.E44332A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 05:42:05 2010
New Revision: 98151

URL: http://llvm.org/viewvc/llvm-project?rev=98151&view=rev
Log:
Fix indentation

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98151&r1=98150&r2=98151&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 05:42:05 2010
@@ -58,7 +58,7 @@
     case XCoreISD::MACCS             : return "XCoreISD::MACCS";
     case XCoreISD::BR_JT             : return "XCoreISD::BR_JT";
     case XCoreISD::BR_JT32           : return "XCoreISD::BR_JT32";
-    default                           : return NULL;
+    default                          : return NULL;
   }
 }
 




From richard at xmos.com  Wed Mar 10 07:20:07 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 13:20:07 -0000
Subject: [llvm-commits] [llvm] r98152 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp lib/Target/XCore/XCoreISelLowering.h
 test/CodeGen/XCore/mul64.ll
Message-ID: <20100310132008.0C7ED2A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 07:20:07 2010
New Revision: 98152

URL: http://llvm.org/viewvc/llvm-project?rev=98152&view=rev
Log:
Custom lower (S|U)MUL_LOHI -> MACC(S|U)

Added:
    llvm/trunk/test/CodeGen/XCore/mul64.ll
Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98152&r1=98151&r2=98152&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 07:20:07 2010
@@ -98,6 +98,8 @@
   // 64bit
   setOperationAction(ISD::ADD, MVT::i64, Custom);
   setOperationAction(ISD::SUB, MVT::i64, Custom);
+  setOperationAction(ISD::SMUL_LOHI, MVT::i32, Custom);
+  setOperationAction(ISD::UMUL_LOHI, MVT::i32, Custom);
   setOperationAction(ISD::MULHS, MVT::i32, Expand);
   setOperationAction(ISD::MULHU, MVT::i32, Expand);
   setOperationAction(ISD::SHL_PARTS, MVT::i32, Expand);
@@ -167,6 +169,8 @@
   case ISD::SELECT_CC:        return LowerSELECT_CC(Op, DAG);
   case ISD::VAARG:            return LowerVAARG(Op, DAG);
   case ISD::VASTART:          return LowerVASTART(Op, DAG);
+  case ISD::SMUL_LOHI:        return LowerSMUL_LOHI(Op, DAG);
+  case ISD::UMUL_LOHI:        return LowerUMUL_LOHI(Op, DAG);
   // FIXME: Remove these when LegalizeDAGTypes lands.
   case ISD::ADD:
   case ISD::SUB:              return ExpandADDSUB(Op.getNode(), DAG);
@@ -544,6 +548,40 @@
 }
 
 SDValue XCoreTargetLowering::
+LowerSMUL_LOHI(SDValue Op, SelectionDAG &DAG)
+{
+  assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::SMUL_LOHI &&
+         "Unexpected operand to lower!");
+  DebugLoc dl = Op.getDebugLoc();
+  SDValue LHS = Op.getOperand(0);
+  SDValue RHS = Op.getOperand(1);
+  SDValue Zero = DAG.getConstant(0, MVT::i32);
+  SDValue Hi = DAG.getNode(XCoreISD::MACCS, dl,
+                           DAG.getVTList(MVT::i32, MVT::i32), Zero, Zero,
+                           LHS, RHS);
+  SDValue Lo(Hi.getNode(), 1);
+  SDValue Ops[] = { Lo, Hi };
+  return DAG.getMergeValues(Ops, 2, dl);
+}
+
+SDValue XCoreTargetLowering::
+LowerUMUL_LOHI(SDValue Op, SelectionDAG &DAG)
+{
+  assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::UMUL_LOHI &&
+         "Unexpected operand to lower!");
+  DebugLoc dl = Op.getDebugLoc();
+  SDValue LHS = Op.getOperand(0);
+  SDValue RHS = Op.getOperand(1);
+  SDValue Zero = DAG.getConstant(0, MVT::i32);
+  SDValue Hi = DAG.getNode(XCoreISD::MACCU, dl,
+                           DAG.getVTList(MVT::i32, MVT::i32), Zero, Zero,
+                           LHS, RHS);
+  SDValue Lo(Hi.getNode(), 1);
+  SDValue Ops[] = { Lo, Hi };
+  return DAG.getMergeValues(Ops, 2, dl);
+}
+
+SDValue XCoreTargetLowering::
 TryExpandADDSUBWithMul(SDNode *N, SelectionDAG &DAG)
 {
   SDValue Mul;

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98152&r1=98151&r2=98152&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Wed Mar 10 07:20:07 2010
@@ -138,6 +138,8 @@
     SDValue LowerSELECT_CC(SDValue Op, SelectionDAG &DAG);
     SDValue LowerVAARG(SDValue Op, SelectionDAG &DAG);
     SDValue LowerVASTART(SDValue Op, SelectionDAG &DAG);
+    SDValue LowerUMUL_LOHI(SDValue Op, SelectionDAG &DAG);
+    SDValue LowerSMUL_LOHI(SDValue Op, SelectionDAG &DAG);
     SDValue LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG);
   
     // Inline asm support

Added: llvm/trunk/test/CodeGen/XCore/mul64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/mul64.ll?rev=98152&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/mul64.ll (added)
+++ llvm/trunk/test/CodeGen/XCore/mul64.ll Wed Mar 10 07:20:07 2010
@@ -0,0 +1,30 @@
+; RUN: llc < %s -march=xcore | FileCheck %s
+define i64 @umul_lohi(i32 %a, i32 %b) {
+entry:
+	%0 = zext i32 %a to i64
+	%1 = zext i32 %b to i64
+	%2 = mul i64 %1, %0
+	ret i64 %2
+}
+; CHECK: umul_lohi:
+; CHECK: ldc r2, 0
+; CHECK-NEXT: mov r3, r2
+; CHECK-NEXT: maccu r2, r3, r1, r0
+; CHECK-NEXT: mov r0, r3
+; CHECK-NEXT: mov r1, r2
+; CHECK-NEXT: retsp 0
+
+define i64 @smul_lohi(i32 %a, i32 %b) {
+entry:
+	%0 = sext i32 %a to i64
+	%1 = sext i32 %b to i64
+	%2 = mul i64 %1, %0
+	ret i64 %2
+}
+; CHECK: smul_lohi:
+; CHECK: ldc r2, 0
+; CHECK-NEXT: mov r3, r2
+; CHECK-NEXT: maccs r2, r3, r1, r0
+; CHECK-NEXT: mov r0, r3
+; CHECK-NEXT: mov r1, r2
+; CHECK-NEXT: retsp 0




From richard at xmos.com  Wed Mar 10 07:27:10 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 13:27:10 -0000
Subject: [llvm-commits] [llvm] r98153 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelDAGToDAG.cpp
 lib/Target/XCore/XCoreISelLowering.cpp lib/Target/XCore/XCoreISelLowering.h
 test/CodeGen/XCore/mul64.ll
Message-ID: <20100310132710.43C512A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 07:27:10 2010
New Revision: 98153

URL: http://llvm.org/viewvc/llvm-project?rev=98153&view=rev
Log:
Prefer LMUL to MACCU as LMUL has no tied operands.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
    llvm/trunk/test/CodeGen/XCore/mul64.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp?rev=98153&r1=98152&r2=98153&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp Wed Mar 10 07:27:10 2010
@@ -220,6 +220,12 @@
         return CurDAG->getMachineNode(XCore::MACCS_l4r, dl, MVT::i32, MVT::i32,
                                       Ops, 4);
       }
+      case XCoreISD::LMUL: {
+        SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
+                          N->getOperand(2), N->getOperand(3) };
+        return CurDAG->getMachineNode(XCore::LMUL_l6r, dl, MVT::i32, MVT::i32,
+                                      Ops, 4);
+      }
       // Other cases are autogenerated.
     }
   }

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98153&r1=98152&r2=98153&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 07:27:10 2010
@@ -54,6 +54,7 @@
     case XCoreISD::RETSP             : return "XCoreISD::RETSP";
     case XCoreISD::LADD              : return "XCoreISD::LADD";
     case XCoreISD::LSUB              : return "XCoreISD::LSUB";
+    case XCoreISD::LMUL              : return "XCoreISD::LMUL";
     case XCoreISD::MACCU             : return "XCoreISD::MACCU";
     case XCoreISD::MACCS             : return "XCoreISD::MACCS";
     case XCoreISD::BR_JT             : return "XCoreISD::BR_JT";
@@ -573,9 +574,9 @@
   SDValue LHS = Op.getOperand(0);
   SDValue RHS = Op.getOperand(1);
   SDValue Zero = DAG.getConstant(0, MVT::i32);
-  SDValue Hi = DAG.getNode(XCoreISD::MACCU, dl,
-                           DAG.getVTList(MVT::i32, MVT::i32), Zero, Zero,
-                           LHS, RHS);
+  SDValue Hi = DAG.getNode(XCoreISD::LMUL, dl,
+                           DAG.getVTList(MVT::i32, MVT::i32), LHS, RHS,
+                           Zero, Zero);
   SDValue Lo(Hi.getNode(), 1);
   SDValue Ops[] = { Lo, Hi };
   return DAG.getMergeValues(Ops, 2, dl);

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98153&r1=98152&r2=98153&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Wed Mar 10 07:27:10 2010
@@ -54,6 +54,9 @@
       // Corresponds to LSUB instruction
       LSUB,
 
+      // Corresponds to LMUL instruction
+      LMUL,
+
       // Corresponds to MACCU instruction
       MACCU,
 

Modified: llvm/trunk/test/CodeGen/XCore/mul64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/mul64.ll?rev=98153&r1=98152&r2=98153&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/mul64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/mul64.ll Wed Mar 10 07:27:10 2010
@@ -8,10 +8,7 @@
 }
 ; CHECK: umul_lohi:
 ; CHECK: ldc r2, 0
-; CHECK-NEXT: mov r3, r2
-; CHECK-NEXT: maccu r2, r3, r1, r0
-; CHECK-NEXT: mov r0, r3
-; CHECK-NEXT: mov r1, r2
+; CHECK-NEXT: lmul r1, r0, r1, r0, r2, r2
 ; CHECK-NEXT: retsp 0
 
 define i64 @smul_lohi(i32 %a, i32 %b) {




From dalej at apple.com  Wed Mar 10 09:06:26 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 15:06:26 -0000
Subject: [llvm-commits] [llvm] r98155 -
	/llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp
Message-ID: <20100310150626.EFAB52A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 09:06:26 2010
New Revision: 98155

URL: http://llvm.org/viewvc/llvm-project?rev=98155&view=rev
Log:
Fix a bug in DEBUG_VALUE handling Devang ran into.
I'll get this loop right yet.


Modified:
    llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp

Modified: llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp?rev=98155&r1=98154&r2=98155&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp (original)
+++ llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp Wed Mar 10 09:06:26 2010
@@ -593,13 +593,10 @@
 
   MachineBasicBlock::iterator E = MBB->end();  
   while (mi != E) {
-    if (mi->isDebugValue()) {
+    while (mi != E && mi->isDebugValue())
       ++mi;
-      if (mi != E && !mi->isDebugValue()) {
-        baseIndex = indexes_->getNextNonNullIndex(baseIndex);
-      }
-      continue;
-    }
+    if (mi == E)
+      break;
     if (mi->killsRegister(interval.reg, tri_)) {
       DEBUG(dbgs() << " killed");
       end = baseIndex.getDefIndex();




From benny.kra at googlemail.com  Wed Mar 10 10:04:20 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Wed, 10 Mar 2010 16:04:20 -0000
Subject: [llvm-commits] [llvm] r98156 - in /llvm/trunk:
 include/llvm/Metadata.h lib/VMCore/Metadata.cpp
Message-ID: <20100310160420.49ABD2A6C12C@llvm.org>

Author: d0k
Date: Wed Mar 10 10:04:20 2010
New Revision: 98156

URL: http://llvm.org/viewvc/llvm-project?rev=98156&view=rev
Log:
Remove duplicated code. No functionality change.

Modified:
    llvm/trunk/include/llvm/Metadata.h
    llvm/trunk/lib/VMCore/Metadata.cpp

Modified: llvm/trunk/include/llvm/Metadata.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Metadata.h?rev=98156&r1=98155&r2=98156&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Metadata.h (original)
+++ llvm/trunk/include/llvm/Metadata.h Wed Mar 10 10:04:20 2010
@@ -43,8 +43,10 @@
 
 public:
   static MDString *get(LLVMContext &Context, StringRef Str);
-  static MDString *get(LLVMContext &Context, const char *Str);
-  
+  static MDString *get(LLVMContext &Context, const char *Str) {
+    return get(Context, Str ? StringRef(Str) : StringRef());
+  }
+
   StringRef getString() const { return Str; }
 
   unsigned getLength() const { return (unsigned)Str.size(); }

Modified: llvm/trunk/lib/VMCore/Metadata.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Metadata.cpp?rev=98156&r1=98155&r2=98156&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Metadata.cpp (original)
+++ llvm/trunk/lib/VMCore/Metadata.cpp Wed Mar 10 10:04:20 2010
@@ -39,15 +39,6 @@
   return S;
 }
 
-MDString *MDString::get(LLVMContext &Context, const char *Str) {
-  LLVMContextImpl *pImpl = Context.pImpl;
-  StringMapEntry<MDString *> &Entry =
-    pImpl->MDStringCache.GetOrCreateValue(Str ? StringRef(Str) : StringRef());
-  MDString *&S = Entry.getValue();
-  if (!S) S = new MDString(Context, Entry.getKey());
-  return S;
-}
-
 //===----------------------------------------------------------------------===//
 // MDNodeOperand implementation.
 //




From richard at xmos.com  Wed Mar 10 10:19:31 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 16:19:31 -0000
Subject: [llvm-commits] [llvm] r98157 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp test/CodeGen/XCore/mul64.ll
Message-ID: <20100310161931.BD1062A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 10:19:31 2010
New Revision: 98157

URL: http://llvm.org/viewvc/llvm-project?rev=98157&view=rev
Log:
Fold add(add(mul(x,y),a),b) -> lmul(x,y,a,b) if the intermediate
results are unused elsewhere.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/test/CodeGen/XCore/mul64.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98157&r1=98156&r2=98157&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 10:19:31 2010
@@ -154,6 +154,7 @@
 
   // We have target-specific dag combine patterns for the following nodes:
   setTargetDAGCombine(ISD::STORE);
+  setTargetDAGCombine(ISD::ADD);
 }
 
 SDValue XCoreTargetLowering::
@@ -1279,6 +1280,61 @@
     }
   }
   break;
+  case ISD::ADD: {
+    // Fold expressions such as add(add(mul(x,y),a),b) -> lmul(x, y, a, b).
+    // This is only profitable if the intermediate results are unused
+    // elsewhere.
+    SDValue N0 = N->getOperand(0);
+    SDValue N1 = N->getOperand(1);
+    SDValue AddOp;
+    SDValue OtherOp;
+    if (N0.getOpcode() == ISD::ADD) {
+      AddOp = N0;
+      OtherOp = N1;
+    } else if (N1.getOpcode() == ISD::ADD) {
+      AddOp = N1;
+      OtherOp = N0;
+    } else {
+      break;
+    }
+    SDValue Addend0, Addend1;
+    SDValue Mul0;
+    SDValue Mul1;
+    if (OtherOp.getOpcode() == ISD::MUL) {
+      // add(add(a,b),mul(x,y))
+      if (!OtherOp.hasOneUse() || !AddOp.hasOneUse())
+        break;
+      Mul0 = OtherOp.getOperand(0);
+      Mul1 = OtherOp.getOperand(1);
+      Addend0 = AddOp.getOperand(0);
+      Addend1 = AddOp.getOperand(1);
+    } else if (AddOp.getOperand(0).getOpcode() == ISD::MUL) {
+      // add(add(mul(x,y),a),b)
+      if (!AddOp.getOperand(0).hasOneUse())
+        break;
+      Mul0 = AddOp.getOperand(0).getOperand(0);
+      Mul1 = AddOp.getOperand(0).getOperand(1);
+      Addend0 = AddOp.getOperand(1);
+      Addend1 = OtherOp;
+    } else if (AddOp.getOperand(1).getOpcode() == ISD::MUL) {
+      // add(add(a,mul(x,y)),b)
+      if (!AddOp.getOperand(1).hasOneUse())
+        break;
+      Mul0 = AddOp.getOperand(1).getOperand(0);
+      Mul1 = AddOp.getOperand(1).getOperand(1);
+      Addend0 = AddOp.getOperand(0);
+      Addend1 = OtherOp;
+    } else {
+      break;
+    }
+    SDValue Zero = DAG.getConstant(0, MVT::i32);
+    SDValue Ignored = DAG.getNode(XCoreISD::LMUL, dl,
+                                  DAG.getVTList(MVT::i32, MVT::i32), Mul0,
+                                  Mul1, Addend0, Addend1);
+    SDValue Result(Ignored.getNode(), 1);
+    return Result;
+  }
+  break;
   case ISD::STORE: {
     // Replace unaligned store of unaligned load with memmove.
     StoreSDNode *ST  = cast<StoreSDNode>(N);

Modified: llvm/trunk/test/CodeGen/XCore/mul64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/mul64.ll?rev=98157&r1=98156&r2=98157&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/mul64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/mul64.ll Wed Mar 10 10:19:31 2010
@@ -25,3 +25,15 @@
 ; CHECK-NEXT: mov r0, r3
 ; CHECK-NEXT: mov r1, r2
 ; CHECK-NEXT: retsp 0
+
+define i64 @mul64(i64 %a, i64 %b) {
+entry:
+	%0 = mul i64 %a, %b
+	ret i64 %0
+}
+; CHECK: mul64:
+; CHECK: ldc r11, 0
+; CHECK-NEXT: lmul r11, r4, r0, r2, r11, r11
+; CHECK-NEXT: mul r0, r0, r3
+; CHECK-NEXT: lmul r0, r1, r1, r2, r11, r0
+; CHECK-NEXT: mov r0, r4




From richard at xmos.com  Wed Mar 10 10:27:11 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 16:27:11 -0000
Subject: [llvm-commits] [llvm] r98158 - in /llvm/trunk/lib/Target/XCore:
 XCoreISelLowering.cpp XCoreISelLowering.h
Message-ID: <20100310162711.C88FB2A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 10:27:11 2010
New Revision: 98158

URL: http://llvm.org/viewvc/llvm-project?rev=98158&view=rev
Log:
Fix thinko.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98158&r1=98157&r2=98158&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 10:27:11 2010
@@ -584,7 +584,7 @@
 }
 
 SDValue XCoreTargetLowering::
-TryExpandADDSUBWithMul(SDNode *N, SelectionDAG &DAG)
+TryExpandADDWithMul(SDNode *N, SelectionDAG &DAG)
 {
   SDValue Mul;
   SDValue Other;
@@ -650,9 +650,11 @@
          (N->getOpcode() == ISD::ADD || N->getOpcode() == ISD::SUB) &&
         "Unknown operand to lower!");
 
-  SDValue Result = TryExpandADDSUBWithMul(N, DAG);
-  if (Result.getNode() != 0)
-    return Result;
+  if (N->getOpcode() == ISD::ADD) {
+    SDValue Result = TryExpandADDWithMul(N, DAG);
+    if (Result.getNode() != 0)
+      return Result;
+  }
 
   DebugLoc dl = N->getDebugLoc();
   

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98158&r1=98157&r2=98158&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Wed Mar 10 10:27:11 2010
@@ -151,7 +151,7 @@
               EVT VT) const;
   
     // Expand specifics
-    SDValue TryExpandADDSUBWithMul(SDNode *Op, SelectionDAG &DAG);
+    SDValue TryExpandADDWithMul(SDNode *Op, SelectionDAG &DAG);
     SDValue ExpandADDSUB(SDNode *Op, SelectionDAG &DAG);
 
     virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;




From richard at xmos.com  Wed Mar 10 11:10:35 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 17:10:35 -0000
Subject: [llvm-commits] [llvm] r98162 -
	/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
Message-ID: <20100310171035.7D7CB2A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 11:10:35 2010
New Revision: 98162

URL: http://llvm.org/viewvc/llvm-project?rev=98162&view=rev
Log:
Extract recognition of patterns such as add(add(mul(x,y),a),b)
into a seperate function.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98162&r1=98161&r2=98162&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 11:10:35 2010
@@ -583,6 +583,64 @@
   return DAG.getMergeValues(Ops, 2, dl);
 }
 
+/// isADDADDMUL - Return whether Op is in a form that is equivalent to
+/// add(add(mul(x,y),a),b). If requireIntermediatesHaveOneUse is true then
+/// each intermediate result in the calculation must also have a single use.
+/// If the Op is in the correct form the constituent parts are written to Mul0,
+/// Mul1, Addend0 and Addend1.
+static bool
+isADDADDMUL(SDValue Op, SDValue &Mul0, SDValue &Mul1, SDValue &Addend0,
+            SDValue &Addend1, bool requireIntermediatesHaveOneUse)
+{
+  if (Op.getOpcode() != ISD::ADD)
+    return false;
+  SDValue N0 = Op.getOperand(0);
+  SDValue N1 = Op.getOperand(1);
+  SDValue AddOp;
+  SDValue OtherOp;
+  if (N0.getOpcode() == ISD::ADD) {
+    AddOp = N0;
+    OtherOp = N1;
+  } else if (N1.getOpcode() == ISD::ADD) {
+    AddOp = N1;
+    OtherOp = N0;
+  } else {
+    return false;
+  }
+  if (OtherOp.getOpcode() == ISD::MUL) {
+    // add(add(a,b),mul(x,y))
+    if (requireIntermediatesHaveOneUse &&
+        (!OtherOp.hasOneUse() || !AddOp.hasOneUse()))
+      return false;
+    Mul0 = OtherOp.getOperand(0);
+    Mul1 = OtherOp.getOperand(1);
+    Addend0 = AddOp.getOperand(0);
+    Addend1 = AddOp.getOperand(1);
+    return true;
+  }
+  if (AddOp.getOperand(0).getOpcode() == ISD::MUL) {
+    // add(add(mul(x,y),a),b)
+    if (requireIntermediatesHaveOneUse && !AddOp.getOperand(0).hasOneUse())
+      return false;
+    Mul0 = AddOp.getOperand(0).getOperand(0);
+    Mul1 = AddOp.getOperand(0).getOperand(1);
+    Addend0 = AddOp.getOperand(1);
+    Addend1 = OtherOp;
+    return true;
+  }
+  if (AddOp.getOperand(1).getOpcode() == ISD::MUL) {
+    // add(add(a,mul(x,y)),b)
+    if (requireIntermediatesHaveOneUse && !AddOp.getOperand(1).hasOneUse())
+      return false;
+    Mul0 = AddOp.getOperand(1).getOperand(0);
+    Mul1 = AddOp.getOperand(1).getOperand(1);
+    Addend0 = AddOp.getOperand(0);
+    Addend1 = OtherOp;
+    return true;
+  }
+  return false;
+}
+
 SDValue XCoreTargetLowering::
 TryExpandADDWithMul(SDNode *N, SelectionDAG &DAG)
 {
@@ -1286,55 +1344,15 @@
     // Fold expressions such as add(add(mul(x,y),a),b) -> lmul(x, y, a, b).
     // This is only profitable if the intermediate results are unused
     // elsewhere.
-    SDValue N0 = N->getOperand(0);
-    SDValue N1 = N->getOperand(1);
-    SDValue AddOp;
-    SDValue OtherOp;
-    if (N0.getOpcode() == ISD::ADD) {
-      AddOp = N0;
-      OtherOp = N1;
-    } else if (N1.getOpcode() == ISD::ADD) {
-      AddOp = N1;
-      OtherOp = N0;
-    } else {
-      break;
-    }
-    SDValue Addend0, Addend1;
-    SDValue Mul0;
-    SDValue Mul1;
-    if (OtherOp.getOpcode() == ISD::MUL) {
-      // add(add(a,b),mul(x,y))
-      if (!OtherOp.hasOneUse() || !AddOp.hasOneUse())
-        break;
-      Mul0 = OtherOp.getOperand(0);
-      Mul1 = OtherOp.getOperand(1);
-      Addend0 = AddOp.getOperand(0);
-      Addend1 = AddOp.getOperand(1);
-    } else if (AddOp.getOperand(0).getOpcode() == ISD::MUL) {
-      // add(add(mul(x,y),a),b)
-      if (!AddOp.getOperand(0).hasOneUse())
-        break;
-      Mul0 = AddOp.getOperand(0).getOperand(0);
-      Mul1 = AddOp.getOperand(0).getOperand(1);
-      Addend0 = AddOp.getOperand(1);
-      Addend1 = OtherOp;
-    } else if (AddOp.getOperand(1).getOpcode() == ISD::MUL) {
-      // add(add(a,mul(x,y)),b)
-      if (!AddOp.getOperand(1).hasOneUse())
-        break;
-      Mul0 = AddOp.getOperand(1).getOperand(0);
-      Mul1 = AddOp.getOperand(1).getOperand(1);
-      Addend0 = AddOp.getOperand(0);
-      Addend1 = OtherOp;
-    } else {
-      break;
+    SDValue Mul0, Mul1, Addend0, Addend1;
+    if (isADDADDMUL(SDValue(N, 0), Mul0, Mul1, Addend0, Addend1, true)) {
+      SDValue Zero = DAG.getConstant(0, MVT::i32);
+      SDValue Ignored = DAG.getNode(XCoreISD::LMUL, dl,
+                                    DAG.getVTList(MVT::i32, MVT::i32), Mul0,
+                                    Mul1, Addend0, Addend1);
+      SDValue Result(Ignored.getNode(), 1);
+      return Result;
     }
-    SDValue Zero = DAG.getConstant(0, MVT::i32);
-    SDValue Ignored = DAG.getNode(XCoreISD::LMUL, dl,
-                                  DAG.getVTList(MVT::i32, MVT::i32), Mul0,
-                                  Mul1, Addend0, Addend1);
-    SDValue Result(Ignored.getNode(), 1);
-    return Result;
   }
   break;
   case ISD::STORE: {




From daniel at zuster.org  Wed Mar 10 11:10:43 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 17:10:43 -0000
Subject: [llvm-commits] [zorg] r98163 -
	/zorg/trunk/buildbot/osuosl/master/config/slaves.py
Message-ID: <20100310171043.D19E82A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 11:10:43 2010
New Revision: 98163

URL: http://llvm.org/viewvc/llvm-project?rev=98163&view=rev
Log:
Add new osu slaves.

Modified:
    zorg/trunk/buildbot/osuosl/master/config/slaves.py

Modified: zorg/trunk/buildbot/osuosl/master/config/slaves.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/osuosl/master/config/slaves.py?rev=98163&r1=98162&r2=98163&view=diff
==============================================================================
--- zorg/trunk/buildbot/osuosl/master/config/slaves.py (original)
+++ zorg/trunk/buildbot/osuosl/master/config/slaves.py Wed Mar 10 11:10:43 2010
@@ -55,6 +55,10 @@
         create_slave("baldrick16", properties={'jobs' : 8}, max_builds=1),
         create_slave("baldrick17", properties={'jobs' : 8}, max_builds=1),
 
+        # Debian x86_64, 2 x 6-core Opteron 2.6 GHz
+        create_slave("osu7", properties={'jobs' : 10}, max_builds=4),
+        create_slave("osu8", properties={'jobs' : 10}, max_builds=4),
+
         # Defunct.
         #create_slave("osu2", properties={'jobs' : 4}, max_builds=2),
         #create_slave("andrew1"),




From richard at xmos.com  Wed Mar 10 11:16:29 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 17:16:29 -0000
Subject: [llvm-commits] [llvm] r98164 -
	/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
Message-ID: <20100310171629.ED9BB2A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 11:16:29 2010
New Revision: 98164

URL: http://llvm.org/viewvc/llvm-project?rev=98164&view=rev
Log:
Fix checking of intermediates having one use in isADDADDMUL

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98164&r1=98163&r2=98164&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 11:16:29 2010
@@ -607,10 +607,11 @@
   } else {
     return false;
   }
+  if (requireIntermediatesHaveOneUse && !AddOp.hasOneUse())
+    return false;
   if (OtherOp.getOpcode() == ISD::MUL) {
     // add(add(a,b),mul(x,y))
-    if (requireIntermediatesHaveOneUse &&
-        (!OtherOp.hasOneUse() || !AddOp.hasOneUse()))
+    if (requireIntermediatesHaveOneUse && !OtherOp.hasOneUse())
       return false;
     Mul0 = OtherOp.getOperand(0);
     Mul1 = OtherOp.getOperand(1);




From bob.wilson at apple.com  Wed Mar 10 11:50:25 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 10 Mar 2010 17:50:25 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r98165 - in /llvm-gcc-4.2/trunk/gcc:
 config/arm/arm.h config/i386/i386.h llvm-convert.cpp
Message-ID: <20100310175025.B2DC52A6C12C@llvm.org>

Author: bwilson
Date: Wed Mar 10 11:50:25 2010
New Revision: 98165

URL: http://llvm.org/viewvc/llvm-project?rev=98165&view=rev
Log:
pr6552: Have llvm-gcc canonicalize register names in inline assembly and
asm register attributes to use the names known to llvm.  For ARM some of the
registers have different names depending on the target for which llvm-gcc is
configured.  Also use the new LLVM_GET_REG_NAME hook to replace the
LLVM_DO_NOT_USE_REG_NAMES macro.

Modified:
    llvm-gcc-4.2/trunk/gcc/config/arm/arm.h
    llvm-gcc-4.2/trunk/gcc/config/i386/i386.h
    llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp

Modified: llvm-gcc-4.2/trunk/gcc/config/arm/arm.h
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/config/arm/arm.h?rev=98165&r1=98164&r2=98165&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/config/arm/arm.h (original)
+++ llvm-gcc-4.2/trunk/gcc/config/arm/arm.h Wed Mar 10 11:50:25 2010
@@ -3548,6 +3548,16 @@
                                     DESTTY, OPS)                              \
         TargetIntrinsicLower(EXP, BUILTIN_CODE, DESTLOC, RESULT, DESTTY, OPS);
 
+/* LLVM_GET_REG_NAME - The registers known to llvm as "r10", "r11", and "r12"
+   may have different names in GCC.  Register "r12" is called "ip", and on
+   non-Darwin OSs, "r10" is "sl" and "r11" is "fp".  Translate those names,
+   and use the default register names for everything else.  */
+#define LLVM_GET_REG_NAME(REG_NAME, REG_NUM) \
+  ((REG_NUM) == 10 ? "r10" \
+   : (REG_NUM) == 11 ? "r11" \
+   : (REG_NUM) == 12 ? "r12" \
+   : reg_names[REG_NUM])
+
 #endif /* ENABLE_LLVM */
 /* LLVM LOCAL end */
 

Modified: llvm-gcc-4.2/trunk/gcc/config/i386/i386.h
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/config/i386/i386.h?rev=98165&r1=98164&r2=98165&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/config/i386/i386.h (original)
+++ llvm-gcc-4.2/trunk/gcc/config/i386/i386.h Wed Mar 10 11:50:25 2010
@@ -3923,11 +3923,12 @@
                                     DESTTY, OPS)                              \
         TargetIntrinsicLower(EXP, BUILTIN_CODE, DESTLOC, RESULT, DESTTY, OPS);
 
-/* When extracting a register name for a constraint, use the string extracted
-   from the magic symbol built for that register, rather than reg_names.
-   The latter maps both AH and AL to the same thing, which means we can't
-   distinguish them. */
-#define LLVM_DO_NOT_USE_REG_NAMES
+/* LLVM_GET_REG_NAME - When extracting a register name for a constraint, use
+   the string extracted from the magic symbol built for that register, rather
+   than reg_names.  The latter maps both AH and AL to the same thing, which
+   means we can't distinguish them. */
+#define LLVM_GET_REG_NAME(REG_NAME, REG_NUM) \
+  ((REG_NAME) + (*(REG_NAME) == '%' ? 1 : 0))
 
 /* Propagate code model setting to backend */
 #define LLVM_SET_MACHINE_OPTIONS(argvec)	   \

Modified: llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp?rev=98165&r1=98164&r2=98165&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp Wed Mar 10 11:50:25 2010
@@ -3995,6 +3995,11 @@
 //               ... Inline Assembly and Register Variables ...
 //===----------------------------------------------------------------------===//
 
+// LLVM_GET_REG_NAME - Default to use GCC's register names.  Targets may
+// override this to use different names for some registers.
+#ifndef LLVM_GET_REG_NAME
+#define LLVM_GET_REG_NAME(REG_NAME, REG_NUM) reg_names[REG_NUM]
+#endif
 
 /// Reads from register variables are handled by emitting an inline asm node
 /// that copies the value out of the specified register.
@@ -4012,7 +4017,9 @@
   // Turn this into a 'tmp = call Ty asm "", "={reg}"()'.
   FunctionType *FTy = FunctionType::get(Ty, std::vector<const Type*>(),false);
 
-  const char *Name = reg_names[decode_reg_name(extractRegisterName(decl))];
+  const char *Name = extractRegisterName(decl);
+  int RegNum = decode_reg_name(Name);
+  Name = LLVM_GET_REG_NAME(Name, RegNum);
 
   InlineAsm *IA = InlineAsm::get(FTy, "", "={"+std::string(Name)+"}", false);
   CallInst *Call = Builder.CreateCall(IA);
@@ -4032,7 +4039,9 @@
   ArgTys.push_back(ConvertType(TREE_TYPE(decl)));
   FunctionType *FTy = FunctionType::get(Type::getVoidTy(Context), ArgTys, false);
 
-  const char *Name = reg_names[decode_reg_name(extractRegisterName(decl))];
+  const char *Name = extractRegisterName(decl);
+  int RegNum = decode_reg_name(Name);
+  Name = LLVM_GET_REG_NAME(Name, RegNum);
 
   InlineAsm *IA = InlineAsm::get(FTy, "", "{"+std::string(Name)+"}", true);
   CallInst *Call = Builder.CreateCall(IA, RHS);
@@ -4403,23 +4412,6 @@
     free((char *)ReplacementStrings[i]);
 }
 
-// When extracting a register name from a DECL_HARD_REGISTER variable,
-// we normally want to look up RegNum in reg_names.  This works on most
-// targets, where ADDITIONAL_REGISTER_NAMES are true synonyms.  It does not
-// work on x86, where ADDITIONAL_REGISTER_NAMES are overlapping subregisters;
-// in particular AH and AL can't be distinguished if we go through reg_names.
-static const char* getConstraintRegNameFromGccTables(const char *RegName,
-                                                     unsigned int RegNum) {
-#ifdef LLVM_DO_NOT_USE_REG_NAMES
-  (void)RegNum;
-  if (*RegName == '%')
-    RegName++;
-  return RegName;
-#else
-  return reg_names[RegNum];
-#endif
-}
-
 Value *TreeToLLVM::EmitASM_EXPR(tree exp) {
   unsigned NumInputs = list_length(ASM_INPUTS(exp));
   unsigned NumOutputs = list_length(ASM_OUTPUTS(exp));
@@ -4537,7 +4529,7 @@
       const char* RegName = extractRegisterName(Operand);
       int RegNum = decode_reg_name(RegName);
       if (RegNum >= 0) {
-        RegName = getConstraintRegNameFromGccTables(RegName, RegNum);
+        RegName = LLVM_GET_REG_NAME(RegName, RegNum);
         unsigned RegNameLen = strlen(RegName);
         char *NewConstraint = (char*)alloca(RegNameLen+4);
         NewConstraint[0] = '=';
@@ -4702,7 +4694,7 @@
       const char *RegName = extractRegisterName(Val);
       int RegNum = decode_reg_name(RegName);
       if (RegNum >= 0) {
-        RegName = getConstraintRegNameFromGccTables(RegName, RegNum);
+        RegName = LLVM_GET_REG_NAME(RegName, RegNum);
         ConstraintStr += '{';
         ConstraintStr += RegName;
         ConstraintStr += '}';
@@ -4745,7 +4737,7 @@
       ConstraintStr += ",~{memory}";
       break;
     default:     // Normal register name.
-      RegName = getConstraintRegNameFromGccTables(RegName, RegCode);
+      RegName = LLVM_GET_REG_NAME(RegName, RegCode);
       ConstraintStr += ",~{";
       ConstraintStr += RegName;
       ConstraintStr += "}";




From bob.wilson at apple.com  Wed Mar 10 11:54:11 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 10 Mar 2010 17:54:11 -0000
Subject: [llvm-commits] [llvm] r98166 -
	/llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c
Message-ID: <20100310175411.D674F2A6C12C@llvm.org>

Author: bwilson
Date: Wed Mar 10 11:54:11 2010
New Revision: 98166

URL: http://llvm.org/viewvc/llvm-project?rev=98166&view=rev
Log:
Testcase for pr6552.  I changed the code to use "ip" instead of "fp" because
the "fp" register name is not valid on Darwin, and the "ip" register name was
broken for all ARM targets.

Added:
    llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c

Added: llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c?rev=98166&view=auto
==============================================================================
--- llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c (added)
+++ llvm/trunk/test/FrontendC/2010-03-10-arm-asmreg.c Wed Mar 10 11:54:11 2010
@@ -0,0 +1,15 @@
+// RUN: %llvmgcc %s -S -O0 -o - | FileCheck %s
+// pr6552
+
+// XFAIL: *
+// XTARGET: arm
+
+extern void bar(unsigned int ip);
+
+// CHECK: mov r0, r12
+void foo(void)
+{
+  register unsigned int ip __asm ("ip");
+  bar(ip);
+}
+




From daniel at zuster.org  Wed Mar 10 11:56:05 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 17:56:05 -0000
Subject: [llvm-commits] [llvm] r98167 - in /llvm/trunk/lib/MC: MCContext.cpp
 MCMachOStreamer.cpp MCParser/AsmParser.cpp
Message-ID: <20100310175605.425862A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 11:56:05 2010
New Revision: 98167

URL: http://llvm.org/viewvc/llvm-project?rev=98167&view=rev
Log:
Remove unneeded includes.

Modified:
    llvm/trunk/lib/MC/MCContext.cpp
    llvm/trunk/lib/MC/MCMachOStreamer.cpp
    llvm/trunk/lib/MC/MCParser/AsmParser.cpp

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=98167&r1=98166&r2=98167&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Wed Mar 10 11:56:05 2010
@@ -10,7 +10,6 @@
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCSymbol.h"
-#include "llvm/MC/MCValue.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/Twine.h"
 using namespace llvm;

Modified: llvm/trunk/lib/MC/MCMachOStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCMachOStreamer.cpp?rev=98167&r1=98166&r2=98167&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCMachOStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCMachOStreamer.cpp Wed Mar 10 11:56:05 2010
@@ -16,7 +16,6 @@
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCSymbol.h"
-#include "llvm/MC/MCValue.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;

Modified: llvm/trunk/lib/MC/MCParser/AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCParser/AsmParser.cpp?rev=98167&r1=98166&r2=98167&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCParser/AsmParser.cpp (original)
+++ llvm/trunk/lib/MC/MCParser/AsmParser.cpp Wed Mar 10 11:56:05 2010
@@ -20,7 +20,6 @@
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
-#include "llvm/MC/MCValue.h"
 #include "llvm/MC/MCParser/MCParsedAsmOperand.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/SourceMgr.h"




From baldrick at free.fr  Wed Mar 10 12:05:34 2010
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 10 Mar 2010 19:05:34 +0100
Subject: [llvm-commits] [llvm-gcc-4.2] r98165 - in
 /llvm-gcc-4.2/trunk/gcc: config/arm/arm.h config/i386/i386.h
 llvm-convert.cpp
In-Reply-To: <20100310175025.B2DC52A6C12C@llvm.org>
References: <20100310175025.B2DC52A6C12C@llvm.org>
Message-ID: <4B97DF6E.8030901@free.fr>

Hi Bob,

> pr6552: Have llvm-gcc canonicalize register names in inline assembly and
> asm register attributes to use the names known to llvm.  For ARM some of the
> registers have different names depending on the target for which llvm-gcc is
> configured.  Also use the new LLVM_GET_REG_NAME hook to replace the
> LLVM_DO_NOT_USE_REG_NAMES macro.

a testcase for the x86 changes would be nice.

Ciao,

Duncan.


From richard at xmos.com  Wed Mar 10 12:12:27 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 18:12:27 -0000
Subject: [llvm-commits] [llvm] r98168 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp test/CodeGen/XCore/addsub64.ll
Message-ID: <20100310181227.4FDFD2A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 12:12:27 2010
New Revision: 98168

URL: http://llvm.org/viewvc/llvm-project?rev=98168&view=rev
Log:
Handle MVT::i64 type in DAG combine for ISD::ADD. Fold 64 bit
expression add(add(mul(x,y),a),b) -> lmul(x,y,a,b) if all
operands are zero extended.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/test/CodeGen/XCore/addsub64.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98168&r1=98167&r2=98168&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Wed Mar 10 12:12:27 2010
@@ -1342,11 +1342,13 @@
   }
   break;
   case ISD::ADD: {
-    // Fold expressions such as add(add(mul(x,y),a),b) -> lmul(x, y, a, b).
+    // Fold 32 bit expressions such as add(add(mul(x,y),a),b) ->
+    // lmul(x, y, a, b). The high result of lmul will be ignored.
     // This is only profitable if the intermediate results are unused
     // elsewhere.
     SDValue Mul0, Mul1, Addend0, Addend1;
-    if (isADDADDMUL(SDValue(N, 0), Mul0, Mul1, Addend0, Addend1, true)) {
+    if (N->getValueType(0) == MVT::i32 &&
+        isADDADDMUL(SDValue(N, 0), Mul0, Mul1, Addend0, Addend1, true)) {
       SDValue Zero = DAG.getConstant(0, MVT::i32);
       SDValue Ignored = DAG.getNode(XCoreISD::LMUL, dl,
                                     DAG.getVTList(MVT::i32, MVT::i32), Mul0,
@@ -1354,6 +1356,31 @@
       SDValue Result(Ignored.getNode(), 1);
       return Result;
     }
+    APInt HighMask = APInt::getHighBitsSet(64, 32);
+    // Fold 64 bit expression such as add(add(mul(x,y),a),b) ->
+    // lmul(x, y, a, b) if all operands are zero-extended. We do this
+    // before type legalization as it is messy to match the operands after
+    // that.
+    if (N->getValueType(0) == MVT::i64 &&
+        isADDADDMUL(SDValue(N, 0), Mul0, Mul1, Addend0, Addend1, false) &&
+        DAG.MaskedValueIsZero(Mul0, HighMask) &&
+        DAG.MaskedValueIsZero(Mul1, HighMask) &&
+        DAG.MaskedValueIsZero(Addend0, HighMask) &&
+        DAG.MaskedValueIsZero(Addend1, HighMask)) {
+      SDValue Mul0L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                                  Mul0, DAG.getConstant(0, MVT::i32));
+      SDValue Mul1L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                                  Mul1, DAG.getConstant(0, MVT::i32));
+      SDValue Addend0L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                                     Addend0, DAG.getConstant(0, MVT::i32));
+      SDValue Addend1L = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32,
+                                     Addend1, DAG.getConstant(0, MVT::i32));
+      SDValue Hi = DAG.getNode(XCoreISD::LMUL, dl,
+                               DAG.getVTList(MVT::i32, MVT::i32), Mul0L, Mul1L,
+                               Addend0L, Addend1L);
+      SDValue Lo(Hi.getNode(), 1);
+      return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi);
+    }
   }
   break;
   case ISD::STORE: {

Modified: llvm/trunk/test/CodeGen/XCore/addsub64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/addsub64.ll?rev=98168&r1=98167&r2=98168&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/addsub64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/addsub64.ll Wed Mar 10 12:12:27 2010
@@ -42,3 +42,18 @@
 ; CHECK: maccs:
 ; CHECK: maccs r1, r0, r3, r2
 ; CHECK-NEXT: retsp 0
+
+define i64 @lmul(i32 %a, i32 %b, i32 %c, i32 %d) {
+entry:
+	%0 = zext i32 %a to i64
+	%1 = zext i32 %b to i64
+	%2 = zext i32 %c to i64
+	%3 = zext i32 %d to i64
+	%4 = mul i64 %1, %0
+	%5 = add i64 %4, %2
+	%6 = add i64 %5, %3
+	ret i64 %6
+}
+; CHECK: lmul:
+; CHECK: lmul r1, r0, r1, r0, r2, r3
+; CHECK-NEXT: retsp 0




From richard at xmos.com  Wed Mar 10 12:14:48 2010
From: richard at xmos.com (Richard Osborne)
Date: Wed, 10 Mar 2010 18:14:48 -0000
Subject: [llvm-commits] [llvm] r98169 -
	/llvm/trunk/lib/Target/XCore/README.txt
Message-ID: <20100310181448.214462A6C12C@llvm.org>

Author: friedgold
Date: Wed Mar 10 12:14:47 2010
New Revision: 98169

URL: http://llvm.org/viewvc/llvm-project?rev=98169&view=rev
Log:
The backend now makes a reasonable job of targeting lmul / macc

Modified:
    llvm/trunk/lib/Target/XCore/README.txt

Modified: llvm/trunk/lib/Target/XCore/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/README.txt?rev=98169&r1=98168&r2=98169&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/README.txt (original)
+++ llvm/trunk/lib/Target/XCore/README.txt Wed Mar 10 12:14:47 2010
@@ -5,4 +5,3 @@
 * Tailcalls
 * Investigate loop alignment
 * Add builtins
-* Make better use of lmul / macc




From grosser at fim.uni-passau.de  Wed Mar 10 12:41:59 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Wed, 10 Mar 2010 18:41:59 -0000
Subject: [llvm-commits] [llvm] r98171 - /llvm/trunk/test/CMakeLists.txt
Message-ID: <20100310184159.E39FA2A6C12C@llvm.org>

Author: grosser
Date: Wed Mar 10 12:41:59 2010
New Revision: 98171

URL: http://llvm.org/viewvc/llvm-project?rev=98171&view=rev
Log:
Fix make check with cmake/lit

PR6540: Set the newly introduced variables ENABLE_SHARED and
SHLIBPATH_VAR in lit.site.cfg not only in the autoconf build, but also
in a cmake one.

Modified:
    llvm/trunk/test/CMakeLists.txt

Modified: llvm/trunk/test/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CMakeLists.txt?rev=98171&r1=98170&r2=98171&view=diff
==============================================================================
--- llvm/trunk/test/CMakeLists.txt (original)
+++ llvm/trunk/test/CMakeLists.txt Wed Mar 10 12:41:59 2010
@@ -7,6 +7,21 @@
 set(LLVM_LIBS_DIR "${LLVM_BINARY_DIR}/lib/${CMAKE_CFG_INTDIR}")
 set(SHLIBEXT "${LTDL_SHLIB_EXT}")
 
+if(BUILD_SHARED_LIBS)
+  set(LLVM_SHARED_LIBS_ENABLED "1")
+else()
+  set(LLVM_SHARED_LIBS_ENABLED "0")
+endif(BUILD_SHARED_LIBS)
+
+if(${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
+  set(SHLIBPATH_VAR "DYLD_LIBRARY_PATH")
+else() # Default for all other unix like systems.
+  # CMake hardcodes the library locaction using rpath.
+  # Therefore LD_LIBRARY_PATH is not required to run binaries in the
+  # build dir. We pass it anyways. 
+  set(SHLIBPATH_VAR "LD_LIBRARY_PATH")
+endif()
+
 include(FindPythonInterp)
 if(PYTHONINTERP_FOUND)
   configure_file(
@@ -27,6 +42,8 @@
                 -e "s#\@LLVM_TOOLS_DIR\@#${LLVM_TOOLS_BINARY_DIR}/${CMAKE_CFG_INTDIR}#"
                 -e "s#\@LLVMGCCDIR\@##"
                 -e "s#\@LLVM_BUILD_MODE\@#${CMAKE_CFG_INTDIR}#"
+                -e "s#\@ENABLE_SHARED\@#${LLVM_SHARED_LIBS_ENABLED}#"
+                -e "s#\@SHLIBPATH_VAR\@#${SHLIBPATH_VAR}#"
                 ${CMAKE_CURRENT_SOURCE_DIR}/Unit/lit.site.cfg.in >
                 ${CMAKE_CURRENT_BINARY_DIR}/Unit/lit.site.cfg
     COMMAND ${PYTHON_EXECUTABLE}




From bob.wilson at apple.com  Wed Mar 10 12:44:55 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 10 Mar 2010 10:44:55 -0800
Subject: [llvm-commits] [llvm-gcc-4.2] r98165 - in
	/llvm-gcc-4.2/trunk/gcc: config/arm/arm.h config/i386/i386.h
	llvm-convert.cpp
In-Reply-To: <4B97DF6E.8030901@free.fr>
References: <20100310175025.B2DC52A6C12C@llvm.org> <4B97DF6E.8030901@free.fr>
Message-ID: <E382E7D0-B11E-49DF-8BDE-3DAFDEA9ECF3@apple.com>


On Mar 10, 2010, at 10:05 AM, Duncan Sands wrote:

> Hi Bob,
> 
>> pr6552: Have llvm-gcc canonicalize register names in inline assembly and
>> asm register attributes to use the names known to llvm.  For ARM some of the
>> registers have different names depending on the target for which llvm-gcc is
>> configured.  Also use the new LLVM_GET_REG_NAME hook to replace the
>> LLVM_DO_NOT_USE_REG_NAMES macro.
> 
> a testcase for the x86 changes would be nice.

What do you have in mind?  I was trying to keep the existing behavior for x86.


From dalej at apple.com  Wed Mar 10 12:47:15 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 10:47:15 -0800
Subject: [llvm-commits] [llvm-gcc-4.2] r98165 -
	in	/llvm-gcc-4.2/trunk/gcc: config/arm/arm.h
	config/i386/i386.h	llvm-convert.cpp
In-Reply-To: <E382E7D0-B11E-49DF-8BDE-3DAFDEA9ECF3@apple.com>
References: <20100310175025.B2DC52A6C12C@llvm.org> <4B97DF6E.8030901@free.fr>
	<E382E7D0-B11E-49DF-8BDE-3DAFDEA9ECF3@apple.com>
Message-ID: <E635A425-1A43-4555-B7E7-2781702AFD96@apple.com>

On Mar 10, 2010, at 10:44 AMPST, Bob Wilson wrote:
> On Mar 10, 2010, at 10:05 AM, Duncan Sands wrote:
> 
>> Hi Bob,
>> 
>>> pr6552: Have llvm-gcc canonicalize register names in inline assembly and
>>> asm register attributes to use the names known to llvm.  For ARM some of the
>>> registers have different names depending on the target for which llvm-gcc is
>>> configured.  Also use the new LLVM_GET_REG_NAME hook to replace the
>>> LLVM_DO_NOT_USE_REG_NAMES macro.
>> 
>> a testcase for the x86 changes would be nice.
> 
> What do you have in mind?  I was trying to keep the existing behavior for x86.

That seems right.  There were similar bugs on x86, but I think they were all fixed a while back.




From johnny.chen at apple.com  Wed Mar 10 12:59:38 2010
From: johnny.chen at apple.com (Johnny Chen)
Date: Wed, 10 Mar 2010 18:59:38 -0000
Subject: [llvm-commits] [llvm] r98172 - in /llvm/trunk/lib/Target/ARM:
 ARMInstrFormats.td ARMInstrInfo.td ARMInstrThumb.td ARMInstrThumb2.td
 AsmPrinter/ARMAsmPrinter.cpp AsmPrinter/ARMInstPrinter.h
Message-ID: <20100310185938.B9E272A6C12C@llvm.org>

Author: johnny
Date: Wed Mar 10 12:59:38 2010
New Revision: 98172

URL: http://llvm.org/viewvc/llvm-project?rev=98172&view=rev
Log:
Factored out the disassembly printing of CPS option, MSR mask, and Negative Zero
operands into their own PrintMethod, in order not to pollute the printOperand()
impl with disassembly only Imm modifiers.

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrFormats.td
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h

Modified: llvm/trunk/lib/Target/ARM/ARMInstrFormats.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrFormats.td?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrFormats.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrFormats.td Wed Mar 10 12:59:38 2010
@@ -144,6 +144,23 @@
   let PrintMethod = "printSBitModifierOperand";
 }
 
+// ARM special operands for disassembly only.
+//
+
+def cps_opt : Operand<i32> {
+  let PrintMethod = "printCPSOptionOperand";
+}
+
+def msr_mask : Operand<i32> {
+  let PrintMethod = "printMSRMaskOperand";
+}
+
+// A8.6.117, A8.6.118.  Different instructions are generated for #0 and #-0.
+// The neg_zero operand translates -0 to -1, -1 to -2, ..., etc.
+def neg_zero : Operand<i32> {
+  let PrintMethod = "printNegZeroOperand";
+}
+
 //===----------------------------------------------------------------------===//
 
 // ARM Instruction templates.

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Wed Mar 10 12:59:38 2010
@@ -710,7 +710,7 @@
 // opt{5} = changemode from Inst{17}
 // opt{8-6} = AIF from Inst{8-6}
 // opt{10-9} = imod from Inst{19-18} with 0b10 as enable and 0b11 as disable
-def CPS : AXI<(outs),(ins i32imm:$opt), MiscFrm, NoItinerary, "cps${opt:cps}",
+def CPS : AXI<(outs), (ins cps_opt:$opt), MiscFrm, NoItinerary, "cps$opt",
               [/* For disassembly only; pattern left blank */]>,
           Requires<[IsARM]> {
   let Inst{31-28} = 0b1111;
@@ -721,9 +721,12 @@
 
 // Preload signals the memory system of possible future data/instruction access.
 // These are for disassembly only.
+//
+// A8.6.117, A8.6.118.  Different instructions are generated for #0 and #-0.
+// The neg_zero operand translates -0 to -1, -1 to -2, ..., etc.
 multiclass APreLoad<bit data, bit read, string opc> {
 
-  def i : AXI<(outs), (ins GPR:$base, i32imm:$imm), MiscFrm, NoItinerary,
+  def i : AXI<(outs), (ins GPR:$base, neg_zero:$imm), MiscFrm, NoItinerary,
                !strconcat(opc, "\t[$base, $imm]"), []> {
     let Inst{31-26} = 0b111101;
     let Inst{25} = 0; // 0 for immediate form
@@ -2844,29 +2847,29 @@
   let Inst{7-4} = 0b0000;
 }
 
-def MSR : ABI<0b0001, (outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
-              "\tcpsr${mask:msr}, $src",
+def MSR : ABI<0b0001, (outs), (ins GPR:$src, msr_mask:$mask), NoItinerary,
+              "msr", "\tcpsr$mask, $src",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0010;
   let Inst{7-4} = 0b0000;
 }
 
-def MSRi : ABI<0b0011, (outs), (ins so_imm:$a, i32imm:$mask), NoItinerary,"msr",
-              "\tcpsr${mask:msr}, $a",
+def MSRi : ABI<0b0011, (outs), (ins so_imm:$a, msr_mask:$mask), NoItinerary,
+              "msr", "\tcpsr$mask, $a",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0010;
   let Inst{7-4} = 0b0000;
 }
 
-def MSRsys : ABI<0b0001, (outs), (ins GPR:$src, i32imm:$mask),NoItinerary,"msr",
-              "\tspsr${mask:msr}, $src",
+def MSRsys : ABI<0b0001, (outs), (ins GPR:$src, msr_mask:$mask), NoItinerary,
+              "msr", "\tspsr$mask, $src",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0110;
   let Inst{7-4} = 0b0000;
 }
 
-def MSRsysi : ABI<0b0011,(outs),(ins so_imm:$a, i32imm:$mask),NoItinerary,"msr",
-              "\tspsr${mask:msr}, $a",
+def MSRsysi : ABI<0b0011, (outs), (ins so_imm:$a, msr_mask:$mask), NoItinerary,
+              "msr", "\tspsr$mask, $a",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{23-20} = 0b0110;
   let Inst{7-4} = 0b0000;

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb.td?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb.td Wed Mar 10 12:59:38 2010
@@ -201,7 +201,7 @@
 //
 // The opt{4-0} and opt{5} sub-fields are to accommodate 32-bit Thumb and ARM
 // CPS which has more options.
-def tCPS : T1I<(outs), (ins i32imm:$opt), NoItinerary, "cps${opt:cps}",
+def tCPS : T1I<(outs), (ins cps_opt:$opt), NoItinerary, "cps$opt",
               [/* For disassembly only; pattern left blank */]>,
            T1Misc<0b0110011>;
 

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Wed Mar 10 12:59:38 2010
@@ -1109,10 +1109,13 @@
 
 // T2Ipl (Preload Data/Instruction) signals the memory system of possible future
 // data/instruction access.  These are for disassembly only.
+//
+// A8.6.117, A8.6.118.  Different instructions are generated for #0 and #-0.
+// The neg_zero operand translates -0 to -1, -1 to -2, ..., etc.
 multiclass T2Ipl<bit instr, bit write, string opc> {
 
-  def i12 : T2I<(outs), (ins t2addrmode_imm12:$addr), IIC_iLoadi, opc,
-                "\t$addr", []> {
+  def i12 : T2I<(outs), (ins GPR:$base, i32imm:$imm), IIC_iLoadi, opc,
+                "\t[$base, $imm]", []> {
     let Inst{31-25} = 0b1111100;
     let Inst{24} = instr;
     let Inst{23} = 1; // U = 1
@@ -1122,8 +1125,8 @@
     let Inst{15-12} = 0b1111;
   }
 
-  def i8 : T2I<(outs), (ins t2addrmode_imm8:$addr), IIC_iLoadi, opc,
-                "\t$addr", []> {
+  def i8 : T2I<(outs), (ins GPR:$base, neg_zero:$imm), IIC_iLoadi, opc,
+                "\t[$base, $imm]", []> {
     let Inst{31-25} = 0b1111100;
     let Inst{24} = instr;
     let Inst{23} = 0; // U = 0
@@ -1134,9 +1137,8 @@
     let Inst{11-8} = 0b1100;
   }
 
-  // A8.6.118 #0 and #-0 differs.  Translates -0 to -1, -1 to -2, ..., etc.
-  def pci : T2I<(outs), (ins GPR:$base, i32imm:$imm), IIC_iLoadi, opc,
-                "\t[pc, ${imm:negzero}]", []> {
+  def pci : T2I<(outs), (ins GPR:$base, neg_zero:$imm), IIC_iLoadi, opc,
+                "\t[pc, $imm]", []> {
     let Inst{31-25} = 0b1111100;
     let Inst{24} = instr;
     let Inst{23} = ?; // add = (U == 1)
@@ -2469,7 +2471,7 @@
 // opt{5} = changemode from Inst{17}
 // opt{8-6} = AIF from Inst{8-6}
 // opt{10-9} = imod from Inst{19-18} with 0b10 as enable and 0b11 as disable
-def t2CPS : T2XI<(outs),(ins i32imm:$opt), NoItinerary, "cps${opt:cps}",
+def t2CPS : T2XI<(outs),(ins cps_opt:$opt), NoItinerary, "cps$opt",
                  [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2638,8 +2640,8 @@
 }
 
 // Rn = Inst{19-16}
-def t2MSR : T2I<(outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
-                "\tcpsr${mask:msr}, $src",
+def t2MSR : T2I<(outs), (ins GPR:$src, msr_mask:$mask), NoItinerary, "msr",
+                "\tcpsr$mask, $src",
                 [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2650,8 +2652,8 @@
 }
 
 // Rn = Inst{19-16}
-def t2MSRsys : T2I<(outs), (ins GPR:$src, i32imm:$mask), NoItinerary, "msr",
-                   "\tspsr${mask:msr}, $src",
+def t2MSRsys : T2I<(outs), (ins GPR:$src, msr_mask:$mask), NoItinerary, "msr",
+                   "\tspsr$mask, $src",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Wed Mar 10 12:59:38 2010
@@ -122,6 +122,9 @@
     void printT2AddrModeImm8OffsetOperand(const MachineInstr *MI, int OpNum);
     void printT2AddrModeSoRegOperand(const MachineInstr *MI, int OpNum);
 
+    void printCPSOptionOperand(const MachineInstr *MI, int OpNum) {}
+    void printMSRMaskOperand(const MachineInstr *MI, int OpNum) {}
+    void printNegZeroOperand(const MachineInstr *MI, int OpNum) {}
     void printPredicateOperand(const MachineInstr *MI, int OpNum);
     void printMandatoryPredicateOperand(const MachineInstr *MI, int OpNum);
     void printSBitModifierOperand(const MachineInstr *MI, int OpNum);

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h?rev=98172&r1=98171&r2=98172&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h Wed Mar 10 12:59:38 2010
@@ -50,7 +50,7 @@
   void printAddrMode6Operand(const MCInst *MI, unsigned OpNum);
   void printAddrModePCOperand(const MCInst *MI, unsigned OpNum,
                               const char *Modifier = 0);
-    
+
   void printBitfieldInvMaskImmOperand(const MCInst *MI, unsigned OpNum);
 
   void printThumbS4ImmOperand(const MCInst *MI, unsigned OpNum);
@@ -70,6 +70,9 @@
   void printT2AddrModeImm8OffsetOperand(const MCInst *MI, unsigned OpNum) {}
   void printT2AddrModeSoRegOperand(const MCInst *MI, unsigned OpNum) {}
   
+  void printCPSOptionOperand(const MCInst *MI, unsigned OpNum) {}
+  void printMSRMaskOperand(const MCInst *MI, unsigned OpNum) {}
+  void printNegZeroOperand(const MCInst *MI, unsigned OpNum) {}
   void printPredicateOperand(const MCInst *MI, unsigned OpNum);
   void printMandatoryPredicateOperand(const MCInst *MI, unsigned OpNum);
   void printSBitModifierOperand(const MCInst *MI, unsigned OpNum);




From gohman at apple.com  Wed Mar 10 13:00:55 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 19:00:55 -0000
Subject: [llvm-commits] [llvm] r98173 -
	/llvm/trunk/test/Other/constant-fold-gep.ll
Message-ID: <20100310190055.131D42A6C12C@llvm.org>

Author: djg
Date: Wed Mar 10 13:00:54 2010
New Revision: 98173

URL: http://llvm.org/viewvc/llvm-project?rev=98173&view=rev
Log:
Fix whitespace.

Modified:
    llvm/trunk/test/Other/constant-fold-gep.ll

Modified: llvm/trunk/test/Other/constant-fold-gep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Other/constant-fold-gep.ll?rev=98173&r1=98172&r2=98173&view=diff
==============================================================================
--- llvm/trunk/test/Other/constant-fold-gep.ll (original)
+++ llvm/trunk/test/Other/constant-fold-gep.ll Wed Mar 10 13:00:54 2010
@@ -118,9 +118,9 @@
 ; TO: @N = constant i64* inttoptr (i64 8 to i64*)
 ; TO: @O = constant i64* inttoptr (i64 8 to i64*)
 
- at M = constant i64* getelementptr (i64 *null, i32 1)
- at N = constant i64* getelementptr ({ i64, i64 } *null, i32 0, i32 1)
- at O = constant i64* getelementptr ([2 x i64] *null, i32 0, i32 1)
+ at M = constant i64* getelementptr (i64* null, i32 1)
+ at N = constant i64* getelementptr ({ i64, i64 }* null, i32 0, i32 1)
+ at O = constant i64* getelementptr ([2 x i64]* null, i32 0, i32 1)
 
 ; Duplicate all of the above as function return values rather than
 ; global initializers.
@@ -457,14 +457,14 @@
 ; SCEV:   -->  sizeof(i64)
 
 define i64* @fM() nounwind {
-  %t = bitcast i64* getelementptr (i64 *null, i32 1) to i64*
+  %t = bitcast i64* getelementptr (i64* null, i32 1) to i64*
   ret i64* %t
 }
 define i64* @fN() nounwind {
-  %t = bitcast i64* getelementptr ({ i64, i64 } *null, i32 0, i32 1) to i64*
+  %t = bitcast i64* getelementptr ({ i64, i64 }* null, i32 0, i32 1) to i64*
   ret i64* %t
 }
 define i64* @fO() nounwind {
-  %t = bitcast i64* getelementptr ([2 x i64] *null, i32 0, i32 1) to i64*
+  %t = bitcast i64* getelementptr ([2 x i64]* null, i32 0, i32 1) to i64*
   ret i64* %t
 }




From devang.patel at gmail.com  Wed Mar 10 13:13:50 2010
From: devang.patel at gmail.com (Devang Patel)
Date: Wed, 10 Mar 2010 11:13:50 -0800
Subject: [llvm-commits] [llvm] r98105 - in
	/llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp
	DwarfException.cpp DwarfPrinter.cpp DwarfPrinter.h
In-Reply-To: <20100309235258.660C22A6C12C@llvm.org>
References: <20100309235258.660C22A6C12C@llvm.org>
Message-ID: <352a1fb21003101113w324485deu743048c609a07307@mail.gmail.com>

On Tue, Mar 9, 2010 at 3:52 PM, Chris Lattner <sabre at nondot.org> wrote:
> Author: lattner
> Date: Tue Mar ?9 17:52:58 2010
> New Revision: 98105
>
> URL: http://llvm.org/viewvc/llvm-project?rev=98105&view=rev
> Log:
> eliminate EOL, adding all comments with the OutStreamer.AddComment
> method. ?With this, comments should end up on the same lines as the .byte
> directives (for example) and we now get no output with:

yay!

>
> $ llc CodeGen/X86/2009-02-12-DebugInfoVLA.ll -o - -filetype=null -asm-verbose
>

-
Devang



From gohman at apple.com  Wed Mar 10 13:31:51 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 19:31:51 -0000
Subject: [llvm-commits] [llvm] r98178 - in /llvm/trunk:
 lib/Analysis/ConstantFolding.cpp test/Other/constant-fold-gep.ll
Message-ID: <20100310193151.5A9312A6C12C@llvm.org>

Author: djg
Date: Wed Mar 10 13:31:51 2010
New Revision: 98178

URL: http://llvm.org/viewvc/llvm-project?rev=98178&view=rev
Log:
Constant-fold GEP-of-GEP into a single GEP.

Modified:
    llvm/trunk/lib/Analysis/ConstantFolding.cpp
    llvm/trunk/test/Other/constant-fold-gep.ll

Modified: llvm/trunk/lib/Analysis/ConstantFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ConstantFolding.cpp?rev=98178&r1=98177&r2=98178&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ConstantFolding.cpp (original)
+++ llvm/trunk/lib/Analysis/ConstantFolding.cpp Wed Mar 10 13:31:51 2010
@@ -589,6 +589,17 @@
   APInt Offset = APInt(BitWidth,
                        TD->getIndexedOffset(Ptr->getType(),
                                             (Value**)Ops+1, NumOps-1));
+
+  // If this is a GEP of a GEP, fold it all into a single GEP.
+  while (GEPOperator *GEP = dyn_cast<GEPOperator>(Ptr)) {
+    SmallVector<Value *, 4> NestedOps(GEP->op_begin()+1, GEP->op_end());
+    Ptr = cast<Constant>(GEP->getOperand(0));
+    Offset += APInt(BitWidth,
+                    TD->getIndexedOffset(Ptr->getType(),
+                                         (Value**)NestedOps.data(),
+                                         NestedOps.size()));
+  }
+
   // If the base value for this address is a literal integer value, fold the
   // getelementptr to the resulting integer value casted to the pointer type.
   if (BaseIsInt) {

Modified: llvm/trunk/test/Other/constant-fold-gep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Other/constant-fold-gep.ll?rev=98178&r1=98177&r2=98178&view=diff
==============================================================================
--- llvm/trunk/test/Other/constant-fold-gep.ll (original)
+++ llvm/trunk/test/Other/constant-fold-gep.ll Wed Mar 10 13:31:51 2010
@@ -21,10 +21,12 @@
 ; PLAIN: %1 = type { double, float, double, double }
 ; PLAIN: %2 = type { i1, i1* }
 ; PLAIN: %3 = type { i64, i64 }
+; PLAIN: %4 = type { i32, i32 }
 ; OPT: %0 = type { i1, double }
 ; OPT: %1 = type { double, float, double, double }
 ; OPT: %2 = type { i1, i1* }
 ; OPT: %3 = type { i64, i64 }
+; OPT: %4 = type { i32, i32 }
 
 ; The automatic constant folder in opt does not have targetdata access, so
 ; it can't fold gep arithmetic, in general. However, the constant folder run
@@ -122,6 +124,16 @@
 @N = constant i64* getelementptr ({ i64, i64 }* null, i32 0, i32 1)
 @O = constant i64* getelementptr ([2 x i64]* null, i32 0, i32 1)
 
+; Fold GEP of a GEP. Theoretically some of these cases could be folded
+; without using targetdata, however that's not implemented yet.
+
+; PLAIN: @Z = global i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x %4]* @ext, i64 0, i64 1, i32 0), i64 1)
+; OPT: @Z = global i32* getelementptr (i32* getelementptr inbounds ([3 x %4]* @ext, i64 0, i64 1, i32 0), i64 1)
+; TO: @Z = global i32* getelementptr inbounds ([3 x %0]* @ext, i64 0, i64 1, i32 1)
+
+ at ext = external global [3 x { i32, i32 }]
+ at Z = global i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x { i32, i32 }]* @ext, i64 0, i64 1, i32 0), i64 1)
+
 ; Duplicate all of the above as function return values rather than
 ; global initializers.
 
@@ -468,3 +480,22 @@
   %t = bitcast i64* getelementptr ([2 x i64]* null, i32 0, i32 1) to i64*
   ret i64* %t
 }
+
+; PLAIN: define i32* @fZ() nounwind {
+; PLAIN:   %t = bitcast i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x %4]* @ext, i64 0, i64 1, i32 0), i64 1) to i32*
+; PLAIN:   ret i32* %t
+; PLAIN: }
+; OPT: define i32* @fZ() nounwind {
+; OPT:   ret i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x %4]* @ext, i64 0, i64 1, i32 0), i64 1)
+; OPT: }
+; TO: define i32* @fZ() nounwind {
+; TO:   ret i32* getelementptr inbounds ([3 x %0]* @ext, i64 0, i64 1, i32 1)
+; TO: }
+; SCEV: Classifying expressions for: @fZ
+; SCEV:   %t = bitcast i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x %4]* @ext, i64 0, i64 1, i32 0), i64 1) to i32*
+; SCEV:   -->  ((3 * sizeof(i32)) + @ext)
+
+define i32* @fZ() nounwind {
+  %t = bitcast i32* getelementptr inbounds (i32* getelementptr inbounds ([3 x { i32, i32 }]* @ext, i64 0, i64 1, i32 0), i64 1) to i32*
+  ret i32* %t
+}




From gohman at apple.com  Wed Mar 10 13:38:49 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 19:38:49 -0000
Subject: [llvm-commits] [llvm] r98179 - in /llvm/trunk:
 include/llvm/Analysis/LoopInfo.h lib/Analysis/LoopInfo.cpp
 lib/Transforms/Scalar/IndVarSimplify.cpp
 lib/Transforms/Scalar/LoopUnrollPass.cpp
 lib/Transforms/Scalar/LoopUnswitch.cpp lib/Transforms/Utils/LCSSA.cpp
 lib/Transforms/Utils/LoopUnroll.cpp
Message-ID: <20100310193849.7A5C92A6C12C@llvm.org>

Author: djg
Date: Wed Mar 10 13:38:49 2010
New Revision: 98179

URL: http://llvm.org/viewvc/llvm-project?rev=98179&view=rev
Log:
Add a DominatorTree argument to isLCSSA so that it doesn't have to
compute a set of reachable blocks for itself each time it is called, which
is fairly frequently.

Modified:
    llvm/trunk/include/llvm/Analysis/LoopInfo.h
    llvm/trunk/lib/Analysis/LoopInfo.cpp
    llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
    llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp
    llvm/trunk/lib/Transforms/Scalar/LoopUnswitch.cpp
    llvm/trunk/lib/Transforms/Utils/LCSSA.cpp
    llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp

Modified: llvm/trunk/include/llvm/Analysis/LoopInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/LoopInfo.h?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/LoopInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/LoopInfo.h Wed Mar 10 13:38:49 2010
@@ -571,7 +571,7 @@
   unsigned getSmallConstantTripMultiple() const;
 
   /// isLCSSAForm - Return true if the Loop is in LCSSA form
-  bool isLCSSAForm() const;
+  bool isLCSSAForm(DominatorTree &DT) const;
 
   /// isLoopSimplifyForm - Return true if the Loop is in the form that
   /// the LoopSimplify form transforms loops to, which is sometimes called

Modified: llvm/trunk/lib/Analysis/LoopInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LoopInfo.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LoopInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LoopInfo.cpp Wed Mar 10 13:38:49 2010
@@ -263,14 +263,7 @@
 }
 
 /// isLCSSAForm - Return true if the Loop is in LCSSA form
-bool Loop::isLCSSAForm() const {
-  // Collect all the reachable blocks in the function, for fast lookups.
-  SmallPtrSet<BasicBlock *, 32> ReachableBBs;
-  BasicBlock *EntryBB = getHeader()->getParent()->begin();
-  for (df_iterator<BasicBlock *> NI = df_begin(EntryBB),
-       NE = df_end(EntryBB); NI != NE; ++NI)
-    ReachableBBs.insert(*NI);
-
+bool Loop::isLCSSAForm(DominatorTree &DT) const {
   // Sort the blocks vector so that we can use binary search to do quick
   // lookups.
   SmallPtrSet<BasicBlock *, 16> LoopBBs(block_begin(), block_end());
@@ -290,7 +283,7 @@
         // entry are special; uses in them don't need to go through PHIs.
         if (UserBB != BB &&
             !LoopBBs.count(UserBB) &&
-            ReachableBBs.count(UserBB))
+            DT.isReachableFromEntry(UserBB))
           return false;
       }
   }

Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Wed Mar 10 13:38:49 2010
@@ -215,7 +215,7 @@
 void IndVarSimplify::RewriteLoopExitValues(Loop *L,
                                            SCEVExpander &Rewriter) {
   // Verify the input to the pass in already in LCSSA form.
-  assert(L->isLCSSAForm());
+  assert(L->isLCSSAForm(*DT));
 
   SmallVector<BasicBlock*, 8> ExitBlocks;
   L->getUniqueExitBlocks(ExitBlocks);
@@ -445,7 +445,7 @@
   // Clean up dead instructions.
   Changed |= DeleteDeadPHIs(L->getHeader());
   // Check a post-condition.
-  assert(L->isLCSSAForm() && "Indvars did not leave the loop in lcssa form!");
+  assert(L->isLCSSAForm(*DT) && "Indvars did not leave the loop in lcssa form!");
   return Changed;
 }
 

Modified: llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp Wed Mar 10 13:38:49 2010
@@ -86,7 +86,6 @@
 }
 
 bool LoopUnroll::runOnLoop(Loop *L, LPPassManager &LPM) {
-  assert(L->isLCSSAForm());
   LoopInfo *LI = &getAnalysis<LoopInfo>();
 
   BasicBlock *Header = L->getHeader();

Modified: llvm/trunk/lib/Transforms/Scalar/LoopUnswitch.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopUnswitch.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopUnswitch.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopUnswitch.cpp Wed Mar 10 13:38:49 2010
@@ -206,7 +206,7 @@
   Function *F = currentLoop->getHeader()->getParent();
   bool Changed = false;
   do {
-    assert(currentLoop->isLCSSAForm());
+    assert(currentLoop->isLCSSAForm(*DT));
     redoLoop = false;
     Changed |= processCurrentLoop();
   } while(redoLoop);

Modified: llvm/trunk/lib/Transforms/Utils/LCSSA.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LCSSA.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LCSSA.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LCSSA.cpp Wed Mar 10 13:38:49 2010
@@ -88,7 +88,7 @@
     /// verifyAnalysis() - Verify loop nest.
     virtual void verifyAnalysis() const {
       // Check the special guarantees that LCSSA makes.
-      assert(L->isLCSSAForm() && "LCSSA form not preserved!");
+      assert(L->isLCSSAForm(*DT) && "LCSSA form not preserved!");
     }
 
     /// inLoop - returns true if the given block is within the current loop
@@ -164,7 +164,7 @@
     }
   }
   
-  assert(L->isLCSSAForm());
+  assert(L->isLCSSAForm(*DT));
   PredCache.clear();
 
   return MadeChange;

Modified: llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp?rev=98179&r1=98178&r2=98179&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp Wed Mar 10 13:38:49 2010
@@ -105,8 +105,6 @@
 /// If a LoopPassManager is passed in, and the loop is fully removed, it will be
 /// removed from the LoopPassManager as well. LPM can also be NULL.
 bool llvm::UnrollLoop(Loop *L, unsigned Count, LoopInfo* LI, LPPassManager* LPM) {
-  assert(L->isLCSSAForm());
-
   BasicBlock *Preheader = L->getLoopPreheader();
   if (!Preheader) {
     DEBUG(dbgs() << "  Can't unroll; loop preheader-insertion failed.\n");
@@ -370,9 +368,5 @@
   if (CompletelyUnroll && LPM != NULL)
     LPM->deleteLoopFromQueue(L);
 
-  // If we didn't completely unroll the loop, it should still be in LCSSA form.
-  if (!CompletelyUnroll)
-    assert(L->isLCSSAForm());
-
   return true;
 }




From dalej at apple.com  Wed Mar 10 13:57:56 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 19:57:56 -0000
Subject: [llvm-commits] [llvm] r98181 -
	/llvm/trunk/lib/CodeGen/BranchFolding.cpp
Message-ID: <20100310195756.760A42A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 13:57:56 2010
New Revision: 98181

URL: http://llvm.org/viewvc/llvm-project?rev=98181&view=rev
Log:
Fix another place where DEBUG_VALUE affected codegen.


Modified:
    llvm/trunk/lib/CodeGen/BranchFolding.cpp

Modified: llvm/trunk/lib/CodeGen/BranchFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/BranchFolding.cpp?rev=98181&r1=98180&r2=98181&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/BranchFolding.cpp (original)
+++ llvm/trunk/lib/CodeGen/BranchFolding.cpp Wed Mar 10 13:57:56 2010
@@ -984,6 +984,17 @@
   return true;
 }
 
+// Blocks with only debug info and branches should be considered the same
+// as blocks with only branches.
+static bool IsBranchOnlyBlock(MachineBasicBlock *MBB) {
+  MachineBasicBlock::iterator MBBI, MBBE;
+  for (MBBI = MBB->begin(), MBBE = MBB->end(); MBBI!=MBBE; ++MBBI) {
+    if (!MBBI->isDebugValue())
+      break;
+  }
+  return (MBBI->getDesc().isBranch());
+}
+
 /// IsBetterFallthrough - Return true if it would be clearly better to
 /// fall-through to MBB1 than to fall through into MBB2.  This has to return
 /// a strict ordering, returning true for both (MBB1,MBB2) and (MBB2,MBB1) will
@@ -1206,7 +1217,7 @@
     // If this branch is the only thing in its block, see if we can forward
     // other blocks across it.
     if (CurTBB && CurCond.empty() && CurFBB == 0 &&
-        MBB->begin()->getDesc().isBranch() && CurTBB != MBB &&
+        IsBranchOnlyBlock(MBB) && CurTBB != MBB &&
         !MBB->hasAddressTaken()) {
       // This block may contain just an unconditional branch.  Because there can
       // be 'non-branch terminators' in the block, try removing the branch and




From grosbach at apple.com  Wed Mar 10 13:59:47 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 10 Mar 2010 19:59:47 -0000
Subject: [llvm-commits] [llvm] r98182 -
	/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
Message-ID: <20100310195947.D4D972A6C12C@llvm.org>

Author: grosbach
Date: Wed Mar 10 13:59:47 2010
New Revision: 98182

URL: http://llvm.org/viewvc/llvm-project?rev=98182&view=rev
Log:
comment why we use custom epilogue for t1 functions using vaargs.

Modified:
    llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp?rev=98182&r1=98181&r2=98182&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb1RegisterInfo.cpp Wed Mar 10 13:59:47 2010
@@ -854,6 +854,11 @@
   }
 
   if (VARegSaveSize) {
+    // Unlike T2 and ARM mode, the T1 pop instruction cannot restore
+    // to LR, and we can't pop the value directly to the PC since
+    // we need to update the SP after popping the value. Therefore, we
+    // pop the old LR into R3 as a temporary.
+
     // Move back past the callee-saved register restoration
     while (MBBI != MBB.end() && isCSRestore(MBBI, CSRegs))
       ++MBBI;




From grosbach at apple.com  Wed Mar 10 14:01:30 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 10 Mar 2010 20:01:30 -0000
Subject: [llvm-commits] [llvm] r98183 -
	/llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
Message-ID: <20100310200130.466F82A6C12C@llvm.org>

Author: grosbach
Date: Wed Mar 10 14:01:30 2010
New Revision: 98183

URL: http://llvm.org/viewvc/llvm-project?rev=98183&view=rev
Log:
Make sure the LR gets pushed in functions that use vaargs. This fixes
400.perlbench for the nightly tests.


Modified:
    llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp?rev=98183&r1=98182&r2=98183&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMBaseRegisterInfo.cpp Wed Mar 10 14:01:30 2010
@@ -590,6 +590,10 @@
       AFI->isThumb2Function())
     MF.getRegInfo().setPhysRegUsed(ARM::R4);
 
+  // Spill LR if Thumb1 function uses variable length argument lists.
+  if (AFI->isThumb1OnlyFunction() && AFI->getVarArgsRegSaveSize() > 0)
+    MF.getRegInfo().setPhysRegUsed(ARM::LR);
+
   // Don't spill FP if the frame can be eliminated. This is determined
   // by scanning the callee-save registers to see if any is used.
   const unsigned *CSRegs = getCalleeSavedRegs();




From baldrick at free.fr  Wed Mar 10 14:33:53 2010
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 10 Mar 2010 21:33:53 +0100
Subject: [llvm-commits] [llvm] r98178 - in /llvm/trunk:
 lib/Analysis/ConstantFolding.cpp test/Other/constant-fold-gep.ll
In-Reply-To: <20100310193151.5A9312A6C12C@llvm.org>
References: <20100310193151.5A9312A6C12C@llvm.org>
Message-ID: <4B980231.9020107@free.fr>

> Constant-fold GEP-of-GEP into a single GEP.

Woah - thanks for doing this Dan!

Ciao,

Duncan.



From daniel at zuster.org  Wed Mar 10 14:58:25 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 20:58:25 -0000
Subject: [llvm-commits] [llvm] r98185 - in
 /llvm/trunk/include/llvm/MC/MCParser: AsmParser.h MCAsmParser.h
Message-ID: <20100310205825.92F412A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 14:58:25 2010
New Revision: 98185

URL: http://llvm.org/viewvc/llvm-project?rev=98185&view=rev
Log:
Remove unneeded declarations.

Modified:
    llvm/trunk/include/llvm/MC/MCParser/AsmParser.h
    llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h

Modified: llvm/trunk/include/llvm/MC/MCParser/AsmParser.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCParser/AsmParser.h?rev=98185&r1=98184&r2=98185&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCParser/AsmParser.h (original)
+++ llvm/trunk/include/llvm/MC/MCParser/AsmParser.h Wed Mar 10 14:58:25 2010
@@ -31,7 +31,6 @@
 class MCInst;
 class MCStreamer;
 class MCAsmInfo;
-class MCValue;
 class SourceMgr;
 class TargetAsmParser;
 class Twine;

Modified: llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h?rev=98185&r1=98184&r2=98185&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h (original)
+++ llvm/trunk/include/llvm/MC/MCParser/MCAsmParser.h Wed Mar 10 14:58:25 2010
@@ -18,7 +18,6 @@
 class MCContext;
 class MCExpr;
 class MCStreamer;
-class MCValue;
 class SMLoc;
 class Twine;
 




From daniel at zuster.org  Wed Mar 10 14:58:29 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 20:58:29 -0000
Subject: [llvm-commits] [llvm] r98186 - in /llvm/trunk:
 include/llvm/MC/MCAssembler.h lib/MC/MCMachOStreamer.cpp
Message-ID: <20100310205829.31B7F2A6C12D@llvm.org>

Author: ddunbar
Date: Wed Mar 10 14:58:29 2010
New Revision: 98186

URL: http://llvm.org/viewvc/llvm-project?rev=98186&view=rev
Log:
MC: Move the backend section and symbol data maps to MCAssembler.

Modified:
    llvm/trunk/include/llvm/MC/MCAssembler.h
    llvm/trunk/lib/MC/MCMachOStreamer.cpp

Modified: llvm/trunk/include/llvm/MC/MCAssembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAssembler.h?rev=98186&r1=98185&r2=98186&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAssembler.h (original)
+++ llvm/trunk/include/llvm/MC/MCAssembler.h Wed Mar 10 14:58:29 2010
@@ -10,6 +10,7 @@
 #ifndef LLVM_MC_MCASSEMBLER_H
 #define LLVM_MC_MCASSEMBLER_H
 
+#include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/ilist.h"
 #include "llvm/ADT/ilist_node.h"
@@ -587,6 +588,16 @@
 
   iplist<MCSymbolData> Symbols;
 
+  /// The map of sections to their associated assembler backend data.
+  //
+  // FIXME: Avoid this indirection?
+  DenseMap<const MCSection*, MCSectionData*> SectionMap;
+
+  /// The map of symbols to their associated assembler backend data.
+  //
+  // FIXME: Avoid this indirection?
+  DenseMap<const MCSymbol*, MCSymbolData*> SymbolMap;
+
   std::vector<IndirectSymbolData> IndirectSymbols;
 
   unsigned SubsectionsViaSymbols : 1;
@@ -673,6 +684,44 @@
   size_t indirect_symbol_size() const { return IndirectSymbols.size(); }
 
   /// @}
+  /// @name Backend Data Access
+  /// @{
+
+  MCSectionData &getSectionData(const MCSection &Section) {
+    MCSectionData *&Entry = SectionMap[&Section];
+    assert(Entry && "Missing section data!");
+    return *Entry;
+  }
+
+  MCSectionData &getOrCreateSectionData(const MCSection &Section,
+                                        bool *Created = 0) {
+    MCSectionData *&Entry = SectionMap[&Section];
+
+    if (Created) *Created = !Entry;
+    if (!Entry)
+      Entry = new MCSectionData(Section, this);
+
+    return *Entry;
+  }
+
+  MCSymbolData &getSymbolData(const MCSymbol &Symbol) {
+    MCSymbolData *&Entry = SymbolMap[&Symbol];
+    assert(Entry && "Missing symbol data!");
+    return *Entry;
+  }
+
+  MCSymbolData &getOrCreateSymbolData(const MCSymbol &Symbol,
+                                      bool *Created = 0) {
+    MCSymbolData *&Entry = SymbolMap[&Symbol];
+
+    if (Created) *Created = !Entry;
+    if (!Entry)
+      Entry = new MCSymbolData(Symbol, 0, 0, this);
+
+    return *Entry;
+  }
+
+  /// @}
 
   void dump();
 };

Modified: llvm/trunk/lib/MC/MCMachOStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCMachOStreamer.cpp?rev=98186&r1=98185&r2=98186&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCMachOStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCMachOStreamer.cpp Wed Mar 10 14:58:29 2010
@@ -47,8 +47,6 @@
   MCAssembler Assembler;
   MCCodeEmitter *Emitter;
   MCSectionData *CurSectionData;
-  DenseMap<const MCSection*, MCSectionData*> SectionMap;
-  DenseMap<const MCSymbol*, MCSymbolData*> SymbolMap;
 
 private:
   MCFragment *getCurrentFragment() const {
@@ -60,24 +58,6 @@
     return 0;
   }
 
-  MCSectionData &getSectionData(const MCSection &Section) {
-    MCSectionData *&Entry = SectionMap[&Section];
-
-    if (!Entry)
-      Entry = new MCSectionData(Section, &Assembler);
-
-    return *Entry;
-  }
-
-  MCSymbolData &getSymbolData(const MCSymbol &Symbol) {
-    MCSymbolData *&Entry = SymbolMap[&Symbol];
-
-    if (!Entry)
-      Entry = new MCSymbolData(Symbol, 0, 0, &Assembler);
-
-    return *Entry;
-  }
-
 public:
   MCMachOStreamer(MCContext &Context, raw_ostream &_OS, MCCodeEmitter *_Emitter)
     : MCStreamer(Context), Assembler(Context, _OS), Emitter(_Emitter),
@@ -98,7 +78,8 @@
     }
 
     case MCExpr::SymbolRef:
-      getSymbolData(cast<MCSymbolRefExpr>(Value)->getSymbol());
+      Assembler.getOrCreateSymbolData(
+        cast<MCSymbolRefExpr>(Value)->getSymbol());
       break;
 
     case MCExpr::Unary:
@@ -163,7 +144,7 @@
   if (Section == CurSection) return;
 
   CurSection = Section;
-  CurSectionData = &getSectionData(*Section);
+  CurSectionData = &Assembler.getOrCreateSectionData(*Section);
 }
 
 void MCMachOStreamer::EmitLabel(MCSymbol *Symbol) {
@@ -174,7 +155,7 @@
   if (!F)
     F = new MCDataFragment(CurSectionData);
 
-  MCSymbolData &SD = getSymbolData(*Symbol);
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
   assert(!SD.getFragment() && "Unexpected fragment on symbol data!");
   SD.setFragment(F);
   SD.setOffset(F->getContents().size());
@@ -220,9 +201,9 @@
   }
 
   // Adding a symbol attribute always introduces the symbol, note that an
-  // important side effect of calling getSymbolData here is to register the
-  // symbol with the assembler.
-  MCSymbolData &SD = getSymbolData(*Symbol);
+  // important side effect of calling getOrCreateSymbolData here is to register
+  // the symbol with the assembler.
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
 
   // The implementation of symbol attributes is designed to match 'as', but it
   // leaves much to desired. It doesn't really make sense to arbitrarily add and
@@ -288,7 +269,7 @@
   // Encode the 'desc' value into the lowest implementation defined bits.
   assert(DescValue == (DescValue & SF_DescFlagsMask) && 
          "Invalid .desc value!");
-  getSymbolData(*Symbol).setFlags(DescValue & SF_DescFlagsMask);
+  Assembler.getOrCreateSymbolData(*Symbol).setFlags(DescValue&SF_DescFlagsMask);
 }
 
 void MCMachOStreamer::EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size,
@@ -296,14 +277,14 @@
   // FIXME: Darwin 'as' does appear to allow redef of a .comm by itself.
   assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
 
-  MCSymbolData &SD = getSymbolData(*Symbol);
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
   SD.setExternal(true);
   SD.setCommon(Size, ByteAlignment);
 }
 
 void MCMachOStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol,
                                    unsigned Size, unsigned ByteAlignment) {
-  MCSectionData &SectData = getSectionData(*Section);
+  MCSectionData &SectData = Assembler.getOrCreateSectionData(*Section);
 
   // The symbol may not be present, which only creates the section.
   if (!Symbol)
@@ -313,7 +294,7 @@
 
   assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
 
-  MCSymbolData &SD = getSymbolData(*Symbol);
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
 
   MCFragment *F = new MCZeroFillFragment(Size, ByteAlignment, &SectData);
   SD.setFragment(F);




From daniel at zuster.org  Wed Mar 10 14:58:31 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 10 Mar 2010 20:58:31 -0000
Subject: [llvm-commits] [llvm] r98187 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100310205831.D1BEE2A6C12E@llvm.org>

Author: ddunbar
Date: Wed Mar 10 14:58:31 2010
New Revision: 98187

URL: http://llvm.org/viewvc/llvm-project?rev=98187&view=rev
Log:
MC/Mach-O: Use the MCAssembler symbol map instead of reconstructing.

Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98187&r1=98186&r2=98187&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Wed Mar 10 14:58:31 2010
@@ -437,7 +437,6 @@
   void ComputeScatteredRelocationInfo(MCAssembler &Asm, MCFragment &Fragment,
                                       MCAsmFixup &Fixup,
                                       const MCValue &Target,
-                             DenseMap<const MCSymbol*,MCSymbolData*> &SymbolMap,
                                      std::vector<MachRelocationEntry> &Relocs) {
     uint32_t Address = Fragment.getOffset() + Fixup.Offset;
     unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
@@ -446,7 +445,7 @@
 
     // See <reloc.h>.
     const MCSymbol *A = Target.getSymA();
-    MCSymbolData *A_SD = SymbolMap.lookup(A);
+    MCSymbolData *A_SD = &Asm.getSymbolData(*A);
 
     if (!A_SD->getFragment())
       llvm_report_error("symbol '" + A->getName() +
@@ -456,7 +455,7 @@
     uint32_t Value2 = 0;
 
     if (const MCSymbol *B = Target.getSymB()) {
-      MCSymbolData *B_SD = SymbolMap.lookup(B);
+      MCSymbolData *B_SD = &Asm.getSymbolData(*B);
 
       if (!B_SD->getFragment())
         llvm_report_error("symbol '" + B->getName() +
@@ -507,7 +506,6 @@
 
   void ComputeRelocationInfo(MCAssembler &Asm, MCDataFragment &Fragment,
                              MCAsmFixup &Fixup,
-                             DenseMap<const MCSymbol*,MCSymbolData*> &SymbolMap,
                              std::vector<MachRelocationEntry> &Relocs) {
     unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
     unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
@@ -525,7 +523,7 @@
         (Target.getSymA() && !Target.getSymA()->isUndefined() &&
          Offset))
       return ComputeScatteredRelocationInfo(Asm, Fragment, Fixup, Target,
-                                            SymbolMap, Relocs);
+                                            Relocs);
 
     // See <reloc.h>.
     uint32_t Address = Fragment.getOffset() + Fixup.Offset;
@@ -543,7 +541,7 @@
       llvm_unreachable("FIXME: Not yet implemented!");
     } else {
       const MCSymbol *Symbol = Target.getSymA();
-      MCSymbolData *SD = SymbolMap.lookup(Symbol);
+      MCSymbolData *SD = &Asm.getSymbolData(*Symbol);
 
       if (Symbol->isUndefined()) {
         IsExtern = 1;
@@ -588,8 +586,7 @@
     Relocs.push_back(MRE);
   }
 
-  void BindIndirectSymbols(MCAssembler &Asm,
-                           DenseMap<const MCSymbol*,MCSymbolData*> &SymbolMap) {
+  void BindIndirectSymbols(MCAssembler &Asm) {
     // This is the point where 'as' creates actual symbols for indirect symbols
     // (in the following two passes). It would be easier for us to do this
     // sooner when we see the attribute, but that makes getting the order in the
@@ -609,9 +606,7 @@
       if (Type != MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS)
         continue;
 
-      MCSymbolData *&Entry = SymbolMap[it->Symbol];
-      if (!Entry)
-        Entry = new MCSymbolData(*it->Symbol, 0, 0, &Asm);
+      Asm.getOrCreateSymbolData(*it->Symbol);
     }
 
     // Then lazy symbol pointers and symbol stubs.
@@ -627,15 +622,13 @@
           Type != MCSectionMachO::S_SYMBOL_STUBS)
         continue;
 
-      MCSymbolData *&Entry = SymbolMap[it->Symbol];
-      if (!Entry) {
-        Entry = new MCSymbolData(*it->Symbol, 0, 0, &Asm);
-
-        // Set the symbol type to undefined lazy, but only on construction.
-        //
-        // FIXME: Do not hardcode.
-        Entry->setFlags(Entry->getFlags() | 0x0001);
-      }
+      // Set the symbol type to undefined lazy, but only on construction.
+      //
+      // FIXME: Do not hardcode.
+      bool Created;
+      MCSymbolData &Entry = Asm.getOrCreateSymbolData(*it->Symbol, &Created);
+      if (Created)
+        Entry.setFlags(Entry.getFlags() | 0x0001);
     }
   }
 
@@ -756,16 +749,8 @@
   void WriteObject(MCAssembler &Asm) {
     unsigned NumSections = Asm.size();
 
-    // Compute the symbol -> symbol data map.
-    //
-    // FIXME: This should not be here.
-    DenseMap<const MCSymbol*, MCSymbolData *> SymbolMap;
-    for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
-           ie = Asm.symbol_end(); it != ie; ++it)
-      SymbolMap[&it->getSymbol()] = it;
-
     // Create symbol data for any indirect symbols.
-    BindIndirectSymbols(Asm, SymbolMap);
+    BindIndirectSymbols(Asm);
 
     // Compute symbol table information.
     SmallString<256> StringTable;
@@ -845,7 +830,7 @@
         if (MCDataFragment *DF = dyn_cast<MCDataFragment>(&*it2))
           for (unsigned i = 0, e = DF->fixup_size(); i != e; ++i)
             ComputeRelocationInfo(Asm, *DF, DF->getFixups()[e - i - 1],
-                                  SymbolMap, RelocInfos);
+                                  RelocInfos);
 
       unsigned NumRelocs = RelocInfos.size() - NumRelocsStart;
       uint64_t SectionStart = SectionDataStart + SD.getAddress();
@@ -914,7 +899,7 @@
         if (Type == MCSectionMachO::S_NON_LAZY_SYMBOL_POINTERS) {
           // If this symbol is defined and internal, mark it as such.
           if (it->Symbol->isDefined() &&
-              !SymbolMap.lookup(it->Symbol)->isExternal()) {
+              !Asm.getSymbolData(*it->Symbol).isExternal()) {
             uint32_t Flags = ISF_Local;
             if (it->Symbol->isAbsolute())
               Flags |= ISF_Absolute;
@@ -923,7 +908,7 @@
           }
         }
 
-        Write32(SymbolMap[it->Symbol]->getIndex());
+        Write32(Asm.getSymbolData(*it->Symbol).getIndex());
       }
 
       // FIXME: Check that offsets match computed ones.




From gohman at apple.com  Wed Mar 10 15:04:53 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 10 Mar 2010 21:04:53 -0000
Subject: [llvm-commits] [llvm] r98188 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
Message-ID: <20100310210453.947552A6C12C@llvm.org>

Author: djg
Date: Wed Mar 10 15:04:53 2010
New Revision: 98188

URL: http://llvm.org/viewvc/llvm-project?rev=98188&view=rev
Log:
Fix another bitwidth calculation to handle vector types; based on a
patch by Micah Villmow for PR6572.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=98188&r1=98187&r2=98188&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Wed Mar 10 15:04:53 2010
@@ -5409,7 +5409,7 @@
     if (SimplifyDemandedBits(Value,
                              APInt::getLowBitsSet(
                                Value.getValueType().getScalarType().getSizeInBits(),
-                               ST->getMemoryVT().getSizeInBits())))
+                               ST->getMemoryVT().getScalarType().getSizeInBits())))
       return SDValue(N, 0);
   }
 




From sabre at nondot.org  Wed Mar 10 15:27:53 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 21:27:53 -0000
Subject: [llvm-commits] [llvm] r98193 - /llvm/trunk/Makefile.rules
Message-ID: <20100310212753.359762A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 15:27:53 2010
New Revision: 98193

URL: http://llvm.org/viewvc/llvm-project?rev=98193&view=rev
Log:
add DESTDIR support for TOOLALIAS, for PR6557, patch by
Matthias Klose!

Modified:
    llvm/trunk/Makefile.rules

Modified: llvm/trunk/Makefile.rules
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=98193&r1=98192&r2=98193&view=diff
==============================================================================
--- llvm/trunk/Makefile.rules (original)
+++ llvm/trunk/Makefile.rules Wed Mar 10 15:27:53 2010
@@ -1329,7 +1329,7 @@
 
 # TOOLALIAS install.
 ifdef TOOLALIAS
-DestToolAlias = $(PROJ_bindir)/$(TOOLALIAS)$(EXEEXT)
+DestToolAlias = $(DESTDIR)$(PROJ_bindir)/$(TOOLALIAS)$(EXEEXT)
 
 install-local:: $(DestToolAlias)
 




From sabre at nondot.org  Wed Mar 10 15:42:42 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 21:42:42 -0000
Subject: [llvm-commits] [llvm] r98194 - /llvm/trunk/lib/Target/README.txt
Message-ID: <20100310214242.F368F2A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 15:42:42 2010
New Revision: 98194

URL: http://llvm.org/viewvc/llvm-project?rev=98194&view=rev
Log:
move PR6576 here.

Modified:
    llvm/trunk/lib/Target/README.txt

Modified: llvm/trunk/lib/Target/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/README.txt?rev=98194&r1=98193&r2=98194&view=diff
==============================================================================
--- llvm/trunk/lib/Target/README.txt (original)
+++ llvm/trunk/lib/Target/README.txt Wed Mar 10 15:42:42 2010
@@ -1819,3 +1819,27 @@
 }
 
 //===---------------------------------------------------------------------===//
+
+The following test case (from PR6576):
+
+define i32 @mul(i32 %a, i32 %b) nounwind readnone {
+entry:
+ %cond1 = icmp eq i32 %b, 0                      ; <i1> [#uses=1]
+ br i1 %cond1, label %exit, label %bb.nph
+bb.nph:                                           ; preds = %entry
+ %tmp = mul i32 %b, %a                           ; <i32> [#uses=1]
+ ret i32 %tmp
+exit:                                             ; preds = %entry
+ ret i32 0
+}
+
+could be reduced to:
+
+define i32 @mul(i32 %a, i32 %b) nounwind readnone {
+entry:
+ %tmp = mul i32 %b, %a
+ ret i32 %tmp
+}
+
+//===---------------------------------------------------------------------===//
+




From dalej at apple.com  Wed Mar 10 16:13:48 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 22:13:48 -0000
Subject: [llvm-commits] [llvm] r98195 - in /llvm/trunk:
 include/llvm/CodeGen/ScheduleDAG.h include/llvm/CodeGen/SelectionDAG.h
 lib/CodeGen/ScheduleDAGInstrs.cpp lib/CodeGen/ScheduleDAGInstrs.h
 lib/CodeGen/SelectionDAG/InstrEmitter.cpp
 lib/CodeGen/SelectionDAG/InstrEmitter.h
 lib/CodeGen/SelectionDAG/SDDbgValue.h
 lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
 lib/CodeGen/SelectionDAG/SelectionDAG.cpp
 lib/Target/X86/X86ISelLowering.cpp
Message-ID: <20100310221348.5EC7D2A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 16:13:47 2010
New Revision: 98195

URL: http://llvm.org/viewvc/llvm-project?rev=98195&view=rev
Log:
Progress towards shepherding debug info through SelectionDAG.
No functional effect yet.  This is still evolving and should
not be viewed as final.


Modified:
    llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h
    llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
    llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp
    llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h
    llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h
    llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h
    llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h (original)
+++ llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h Wed Mar 10 16:13:47 2010
@@ -227,6 +227,7 @@
   private:
     SDNode *Node;                       // Representative node.
     MachineInstr *Instr;                // Alternatively, a MachineInstr.
+    MachineInstr *DbgInstr;             // A dbg_value referencing this.
   public:
     SUnit *OrigNode;                    // If not this, the node from which
                                         // this node was cloned.
@@ -269,10 +270,10 @@
     /// SUnit - Construct an SUnit for pre-regalloc scheduling to represent
     /// an SDNode and any nodes flagged to it.
     SUnit(SDNode *node, unsigned nodenum)
-      : Node(node), Instr(0), OrigNode(0), NodeNum(nodenum), NodeQueueId(0),
-        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
-        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
-        hasPhysRegClobbers(false),
+      : Node(node), Instr(0), DbgInstr(0), OrigNode(0), NodeNum(nodenum),
+        NodeQueueId(0),  Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
+        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
+        hasPhysRegDefs(false), hasPhysRegClobbers(false),
         isPending(false), isAvailable(false), isScheduled(false),
         isScheduleHigh(false), isCloned(false),
         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
@@ -281,10 +282,10 @@
     /// SUnit - Construct an SUnit for post-regalloc scheduling to represent
     /// a MachineInstr.
     SUnit(MachineInstr *instr, unsigned nodenum)
-      : Node(0), Instr(instr), OrigNode(0), NodeNum(nodenum), NodeQueueId(0),
-        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
-        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
-        hasPhysRegClobbers(false),
+      : Node(0), Instr(instr), DbgInstr(0), OrigNode(0), NodeNum(nodenum),
+        NodeQueueId(0), Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
+        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
+        hasPhysRegDefs(false), hasPhysRegClobbers(false),
         isPending(false), isAvailable(false), isScheduled(false),
         isScheduleHigh(false), isCloned(false),
         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
@@ -292,10 +293,10 @@
 
     /// SUnit - Construct a placeholder SUnit.
     SUnit()
-      : Node(0), Instr(0), OrigNode(0), NodeNum(~0u), NodeQueueId(0),
-        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
-        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
-        hasPhysRegClobbers(false),
+      : Node(0), Instr(0), DbgInstr(0), OrigNode(0), NodeNum(~0u),
+        NodeQueueId(0), Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
+        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
+        hasPhysRegDefs(false), hasPhysRegClobbers(false),
         isPending(false), isAvailable(false), isScheduled(false),
         isScheduleHigh(false), isCloned(false),
         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
@@ -329,6 +330,20 @@
       return Instr;
     }
 
+    /// setDbgInstr - Assign the debug instruction for the SUnit.
+    /// This may be used during post-regalloc scheduling.
+    void setDbgInstr(MachineInstr *MI) {
+      assert(!Node && "Setting debug MachineInstr of SUnit with SDNode!");
+      DbgInstr = MI;
+    }
+
+    /// getDbgInstr - Return the debug MachineInstr for this SUnit.
+    /// This may be used during post-regalloc scheduling.
+    MachineInstr *getDbgInstr() const {
+      assert(!Node && "Reading debug MachineInstr of SUnit with SDNode!");
+      return DbgInstr;
+    }
+
     /// addPred - This adds the specified edge as a pred of the current node if
     /// not already.  It also adds the current node as a successor of the
     /// specified node.

Modified: llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/SelectionDAG.h?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/SelectionDAG.h (original)
+++ llvm/trunk/include/llvm/CodeGen/SelectionDAG.h Wed Mar 10 16:13:47 2010
@@ -35,6 +35,7 @@
 class MachineFunction;
 class MachineModuleInfo;
 class SDNodeOrdering;
+class SDDbgValue;
 class TargetLowering;
 
 template<> struct ilist_traits<SDNode> : public ilist_default_traits<SDNode> {
@@ -57,6 +58,46 @@
   static void createNode(const SDNode &);
 };
 
+/// SDDbgInfo - Keeps track of dbg_value information through SDISel.  We do
+/// not build SDNodes for these so as not to perturb the generated code;
+/// instead the info is kept off to the side in this structure.  SDNodes may
+/// have an associated dbg_value entry in DbgValMap.  Debug info that is not
+/// associated with any SDNode is held in DbgConstMap.  It is possible for
+/// optimizations to change a variable to a constant, in which case the
+/// corresponding debug info is moved from the variable to the constant table
+/// (NYI).
+class SDDbgInfo {
+  DenseMap<const SDNode*, SDDbgValue*> DbgVblMap;
+  SmallVector<SDDbgValue*, 4> DbgConstMap;
+
+  void operator=(const SDDbgInfo&);   // Do not implement.
+  SDDbgInfo(const SDDbgInfo&);   // Do not implement.
+public:
+  SDDbgInfo() {}
+
+  void add(const SDNode *Node, SDDbgValue *V) {
+    DbgVblMap[Node] = V;
+  }
+  void add(SDDbgValue *V) { DbgConstMap.push_back(V); }
+  void remove(const SDNode *Node) {
+    DenseMap<const SDNode*, SDDbgValue*>::iterator Itr =
+                      DbgVblMap.find(Node);
+    if (Itr != DbgVblMap.end())
+      DbgVblMap.erase(Itr);
+  }
+  // No need to remove a constant.
+  void clear() {
+    DbgVblMap.clear();
+    DbgConstMap.clear();
+  }
+  SDDbgValue *getSDDbgValue(const SDNode *Node) {
+    return DbgVblMap[Node];
+  }
+  typedef SmallVector<SDDbgValue*, 4>::iterator ConstDbgIterator;
+  ConstDbgIterator DbgConstBegin() { return DbgConstMap.begin(); }
+  ConstDbgIterator DbgConstEnd() { return DbgConstMap.end(); }
+};
+
 enum CombineLevel {
   Unrestricted,   // Combine may create illegal operations and illegal types.
   NoIllegalTypes, // Combine may create illegal operations but no illegal types.
@@ -119,6 +160,9 @@
   /// the ordering of the original LLVM instructions.
   SDNodeOrdering *Ordering;
 
+  /// DbgInfo - Tracks dbg_value information through SDISel.
+  SDDbgInfo *DbgInfo;
+
   /// VerifyNode - Sanity check the given node.  Aborts if it is invalid.
   void VerifyNode(SDNode *N);
 
@@ -828,6 +872,20 @@
   /// GetOrdering - Get the order for the SDNode.
   unsigned GetOrdering(const SDNode *SD) const;
 
+  /// AssignDbgInfo - Assign debug info to the SDNode.
+  void AssignDbgInfo(SDNode *SD, SDDbgValue *db);
+
+  /// RememberDbgInfo - Remember debug info with no associated SDNode.
+  void RememberDbgInfo(SDDbgValue *db);
+
+  /// GetDbgInfo - Get the debug info for the SDNode.
+  SDDbgValue *GetDbgInfo(const SDNode* SD);
+
+  SDDbgInfo::ConstDbgIterator DbgConstBegin() { 
+    return DbgInfo->DbgConstBegin(); 
+  }
+  SDDbgInfo::ConstDbgIterator DbgConstEnd() { return DbgInfo->DbgConstEnd(); }
+
   void dump() const;
 
   /// CreateStackTemporary - Create a stack temporary, suitable for holding the

Modified: llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp (original)
+++ llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp Wed Mar 10 16:13:47 2010
@@ -34,6 +34,7 @@
                                      const MachineDominatorTree &mdt)
   : ScheduleDAG(mf), MLI(mli), MDT(mdt), LoopRegs(MLI, MDT) {
   MFI = mf.getFrameInfo();
+  DbgValueVec.clear();
 }
 
 /// Run - perform scheduling.
@@ -157,6 +158,10 @@
   std::map<const Value *, SUnit *> AliasMemDefs, NonAliasMemDefs;
   std::map<const Value *, std::vector<SUnit *> > AliasMemUses, NonAliasMemUses;
 
+  // Keep track of dangling debug references to registers.
+  std::pair<MachineInstr*, unsigned>
+        DanglingDebugValue[TargetRegisterInfo::FirstVirtualRegister];
+
   // Check to see if the scheduler cares about latencies.
   bool UnitLatencies = ForceUnitLatencies();
 
@@ -164,10 +169,25 @@
   const TargetSubtarget &ST = TM.getSubtarget<TargetSubtarget>();
   unsigned SpecialAddressLatency = ST.getSpecialAddressLatency();
 
+  // Remove any stale debug info; sometimes BuildSchedGraph is called again
+  // without emitting the info from the previous call.
+  DbgValueVec.clear();
+  std::memset(DanglingDebugValue, 0, sizeof(DanglingDebugValue));
+
   // Walk the list of instructions, from bottom moving up.
   for (MachineBasicBlock::iterator MII = InsertPos, MIE = Begin;
        MII != MIE; --MII) {
     MachineInstr *MI = prior(MII);
+    // DBG_VALUE does not have SUnit's built, so just remember these for later
+    // reinsertion.
+    if (MI->isDebugValue()) {
+      if (MI->getNumOperands()==3 && MI->getOperand(0).isReg() &&
+          MI->getOperand(0).getReg())
+        DanglingDebugValue[MI->getOperand(0).getReg()] =
+             std::make_pair(MI, DbgValueVec.size());
+      DbgValueVec.push_back(MI);
+      continue;
+    }
     const TargetInstrDesc &TID = MI->getDesc();
     assert(!TID.isTerminator() && !MI->isLabel() &&
            "Cannot schedule terminators or labels!");
@@ -188,6 +208,13 @@
       if (Reg == 0) continue;
 
       assert(TRI->isPhysicalRegister(Reg) && "Virtual register encountered!");
+
+      if (MO.isDef() && DanglingDebugValue[Reg].first!=0) {
+        SU->setDbgInstr(DanglingDebugValue[Reg].first);
+        DbgValueVec[DanglingDebugValue[Reg].second] = 0;
+        DanglingDebugValue[Reg] = std::make_pair((MachineInstr*)0, 0);
+      }
+
       std::vector<SUnit *> &UseList = Uses[Reg];
       std::vector<SUnit *> &DefList = Defs[Reg];
       // Optionally add output and anti dependencies. For anti
@@ -555,6 +582,14 @@
     BB->remove(I);
   }
 
+  // First reinsert any remaining debug_values; these are either constants,
+  // or refer to live-in registers.  The beginning of the block is the right
+  // place for the latter.  The former might reasonably be placed elsewhere
+  // using some kind of ordering algorithm, but right now it doesn't matter.
+  for (int i = DbgValueVec.size()-1; i>=0; --i)
+    if (DbgValueVec[i])
+      BB->insert(InsertPos, DbgValueVec[i]);
+
   // Then re-insert them according to the given schedule.
   for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
     SUnit *SU = Sequence[i];
@@ -565,12 +600,21 @@
     }
 
     BB->insert(InsertPos, SU->getInstr());
+    if (SU->getDbgInstr())
+      BB->insert(InsertPos, SU->getDbgInstr());
   }
 
   // Update the Begin iterator, as the first instruction in the block
   // may have been scheduled later.
-  if (!Sequence.empty())
+  if (!DbgValueVec.empty()) {
+    for (int i = DbgValueVec.size()-1; i>=0; --i)
+      if (DbgValueVec[i]!=0) {
+        Begin = DbgValueVec[DbgValueVec.size()-1];
+        break;
+      }
+  } else if (!Sequence.empty())
     Begin = Sequence[0]->getInstr();
 
+  DbgValueVec.clear();
   return BB;
 }

Modified: llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h (original)
+++ llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h Wed Mar 10 16:13:47 2010
@@ -106,6 +106,10 @@
     /// initialized and destructed for each block.
     std::vector<SUnit *> Defs[TargetRegisterInfo::FirstVirtualRegister];
     std::vector<SUnit *> Uses[TargetRegisterInfo::FirstVirtualRegister];
+ 
+    /// DbgValueVec - Remember DBG_VALUEs that refer to a particular
+    /// register.
+    std::vector<MachineInstr *>DbgValueVec;
 
     /// PendingLoads - Remember where unknown loads are after the most recent
     /// unknown store, as we iterate. As with Defs and Uses, this is here

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp Wed Mar 10 16:13:47 2010
@@ -508,6 +508,7 @@
     return;
   if (!sd)
     return;
+  assert(sd->getKind() == SDDbgValue::SD);
   unsigned VReg = getVR(SDValue(sd->getSDNode(), sd->getResNo()), VRBaseMap);
   const TargetInstrDesc &II = TII->get(TargetOpcode::DBG_VALUE);
   DebugLoc DL = sd->getDebugLoc();
@@ -524,26 +525,46 @@
   MBB->insert(InsertPos, MI);
 }
 
-/// EmitDbgValue - Generate constant debug info.  No SDNode is involved.
+/// EmitDbgValue - Generate debug info that does not refer to a SDNode.
 void
-InstrEmitter::EmitDbgValue(SDDbgValue *sd) {
+InstrEmitter::EmitDbgValue(SDDbgValue *sd,
+                         DenseMap<MachineBasicBlock*, MachineBasicBlock*> *EM) {
   if (!sd)
     return;
   const TargetInstrDesc &II = TII->get(TargetOpcode::DBG_VALUE);
+  uint64_t Offset = sd->getOffset();
+  MDNode* mdPtr = sd->getMDPtr();
+  SDDbgValue::DbgValueKind kind = sd->getKind();
   DebugLoc DL = sd->getDebugLoc();
-  MachineInstr *MI;
-  Value *V = sd->getConst();
-  if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
-    MI = BuildMI(*MF, DL, II).addImm(CI->getZExtValue()).
-                                   addImm(sd->getOffset()).
-                                   addMetadata(sd->getMDPtr());
-  } else if (ConstantFP *CF = dyn_cast<ConstantFP>(V)) {
-    MI = BuildMI(*MF, DL, II).addFPImm(CF).addImm(sd->getOffset()).
-                                   addMetadata(sd->getMDPtr());
+  MachineInstr* MI;
+  if (kind == SDDbgValue::CNST) {
+    Value *V = sd->getConst();
+    if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
+      MI = BuildMI(*MF, DL, II).addImm(CI->getZExtValue()).
+                                     addImm(Offset).addMetadata(mdPtr);
+    } else if (ConstantFP *CF = dyn_cast<ConstantFP>(V)) {
+      MI = BuildMI(*MF, DL, II).addFPImm(CF).
+                                     addImm(Offset).addMetadata(mdPtr);
+    } else {
+      // Could be an Undef.  In any case insert an Undef so we can see what we
+      // dropped.
+      MI = BuildMI(*MF, DL, II).addReg(0U).
+                                       addImm(Offset).addMetadata(mdPtr);
+    }
+  } else if (kind == SDDbgValue::FX) {
+    unsigned FrameIx = sd->getFrameIx();
+    // Stack address; this needs to be lowered in target-dependent fashion.
+    // FIXME test that the target supports this somehow; if not emit Undef.
+    // Create a pseudo for EmitInstrWithCustomInserter's consumption.
+    MI = BuildMI(*MF, DL, II).addImm(FrameIx).
+                                   addImm(Offset).addMetadata(mdPtr);
+    MBB = TLI->EmitInstrWithCustomInserter(MI, MBB, EM);
+    InsertPos = MBB->end();
+    return;
   } else {
     // Insert an Undef so we can see what we dropped.
-    MI = BuildMI(*MF, DL, II).addReg(0U).addImm(sd->getOffset()).
-                                    addMetadata(sd->getMDPtr());
+    MI = BuildMI(*MF, DL, II).addReg(0U).
+                                     addImm(Offset).addMetadata(mdPtr);
   }
   MBB->insert(InsertPos, MI);
 }

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h Wed Mar 10 16:13:47 2010
@@ -106,7 +106,8 @@
 
 
   /// EmitDbgValue - Generate a constant DBG_VALUE.  No node is involved.
-  void EmitDbgValue(SDDbgValue* sd);
+  void EmitDbgValue(SDDbgValue* sd,
+                DenseMap<MachineBasicBlock*, MachineBasicBlock*> *EM);
 
   /// EmitNode - Generate machine code for a node and needed dependencies.
   ///

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h Wed Mar 10 16:13:47 2010
@@ -28,9 +28,22 @@
 /// We do not use SDValue here to avoid including its header.
 
 class SDDbgValue {
-  SDNode *Node;           // valid for non-constants
-  unsigned ResNo;         // valid for non-constants
-  Value *Const;           // valid for constants
+public:
+  enum DbgValueKind {
+    SD = 0,
+    CNST = 1,
+    FX = 2
+  };
+private:
+  enum DbgValueKind kind;
+  union {
+    struct {
+      SDNode *Node;         // valid for non-constants
+      unsigned ResNo;       // valid for non-constants
+    } s;
+    Value *Const;           // valid for constants
+    unsigned FrameIx;       // valid for stack objects
+  } u;
   MDNode *mdPtr;
   uint64_t Offset;
   DebugLoc DL;
@@ -38,24 +51,43 @@
 public:
   // Constructor for non-constants.
   SDDbgValue(MDNode *mdP, SDNode *N, unsigned R, uint64_t off, DebugLoc dl,
-             unsigned O) :
-    Node(N), ResNo(R), Const(0), mdPtr(mdP), Offset(off), DL(dl), Order(O) {}
+             unsigned O) : mdPtr(mdP), Offset(off), DL(dl), Order(O) {
+    kind = SD;
+    u.s.Node = N;
+    u.s.ResNo = R;
+  }
 
   // Constructor for constants.
   SDDbgValue(MDNode *mdP, Value *C, uint64_t off, DebugLoc dl, unsigned O) : 
-    Node(0), ResNo(0), Const(C), mdPtr(mdP), Offset(off), DL(dl), Order(O) {}
+    mdPtr(mdP), Offset(off), DL(dl), Order(O) {
+    kind = CNST;
+    u.Const = C;
+  }
+
+  // Constructor for frame indices.
+  SDDbgValue(MDNode *mdP, unsigned FI, uint64_t off, DebugLoc dl, unsigned O) : 
+    mdPtr(mdP), Offset(off), DL(dl), Order(O) {
+    kind = FX;
+    u.FrameIx = FI;
+  }
+
+  // Returns the kind.
+  DbgValueKind getKind() { return kind; }
 
   // Returns the MDNode pointer.
   MDNode *getMDPtr() { return mdPtr; }
 
-  // Returns the SDNode* (valid for non-constants only).
-  SDNode *getSDNode() { assert (!Const); return Node; }
+  // Returns the SDNode* for a register ref
+  SDNode *getSDNode() { assert (kind==SD); return u.s.Node; }
+
+  // Returns the ResNo for a register ref
+  unsigned getResNo() { assert (kind==SD); return u.s.ResNo; }
 
-  // Returns the ResNo (valid for non-constants only).
-  unsigned getResNo() { assert (!Const); return ResNo; }
+  // Returns the Value* for a constant
+  Value *getConst() { assert (kind==CNST); return u.Const; }
 
-  // Returns the Value* for a constant (invalid for non-constants).
-  Value *getConst() { assert (!Node); return Const; }
+  // Returns the FrameIx for a stack object
+  unsigned getFrameIx() { assert (kind==FX); return u.FrameIx; }
 
   // Returns the offset.
   uint64_t getOffset() { return Offset; }

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp Wed Mar 10 16:13:47 2010
@@ -13,6 +13,7 @@
 //===----------------------------------------------------------------------===//
 
 #define DEBUG_TYPE "pre-RA-sched"
+#include "SDDbgValue.h"
 #include "ScheduleDAGSDNodes.h"
 #include "InstrEmitter.h"
 #include "llvm/CodeGen/SelectionDAG.h"
@@ -412,6 +413,14 @@
   InstrEmitter Emitter(BB, InsertPos);
   DenseMap<SDValue, unsigned> VRBaseMap;
   DenseMap<SUnit*, unsigned> CopyVRBaseMap;
+
+  // For now, any constant debug info nodes go at the beginning.
+  for (SDDbgInfo::ConstDbgIterator I = DAG->DbgConstBegin(),
+       E = DAG->DbgConstEnd(); I!=E; I++) {
+    Emitter.EmitDbgValue(*I, EM);
+    delete *I;
+  }
+
   for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
     SUnit *SU = Sequence[i];
     if (!SU) {
@@ -435,10 +444,20 @@
     while (!FlaggedNodes.empty()) {
       Emitter.EmitNode(FlaggedNodes.back(), SU->OrigNode != SU, SU->isCloned,
                        VRBaseMap, EM);
+      if (FlaggedNodes.back()->getHasDebugValue())
+        if (SDDbgValue *sd = DAG->GetDbgInfo(FlaggedNodes.back())) {
+          Emitter.EmitDbgValue(FlaggedNodes.back(), VRBaseMap, sd);
+          delete sd;
+        }
       FlaggedNodes.pop_back();
     }
     Emitter.EmitNode(SU->getNode(), SU->OrigNode != SU, SU->isCloned,
                      VRBaseMap, EM);
+    if (SU->getNode()->getHasDebugValue())
+      if (SDDbgValue *sd = DAG->GetDbgInfo(SU->getNode())) {
+        Emitter.EmitDbgValue(SU->getNode(), VRBaseMap, sd);
+        delete sd;
+      }
   }
 
   BB = Emitter.getBlock();

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Wed Mar 10 16:13:47 2010
@@ -13,6 +13,7 @@
 
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "SDNodeOrdering.h"
+#include "SDDbgValue.h"
 #include "llvm/Constants.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/Function.h"
@@ -596,6 +597,9 @@
 
   // Remove the ordering of this node.
   Ordering->remove(N);
+
+  // And its entry in the debug info table, if any.
+  DbgInfo->remove(N);
 }
 
 /// RemoveNodeFromCSEMaps - Take the specified node out of the CSE map that
@@ -793,6 +797,7 @@
     Root(getEntryNode()), Ordering(0) {
   AllNodes.push_back(&EntryNode);
   Ordering = new SDNodeOrdering();
+  DbgInfo = new SDDbgInfo();
 }
 
 void SelectionDAG::init(MachineFunction &mf, MachineModuleInfo *mmi,
@@ -806,6 +811,7 @@
 SelectionDAG::~SelectionDAG() {
   allnodes_clear();
   delete Ordering;
+  delete DbgInfo;
 }
 
 void SelectionDAG::allnodes_clear() {
@@ -833,6 +839,8 @@
   Root = getEntryNode();
   delete Ordering;
   Ordering = new SDNodeOrdering();
+  delete DbgInfo;
+  DbgInfo = new SDDbgInfo();
 }
 
 SDValue SelectionDAG::getSExtOrTrunc(SDValue Op, DebugLoc DL, EVT VT) {
@@ -5264,6 +5272,25 @@
   return Ordering->getOrder(SD);
 }
 
+/// AssignDbgInfo - Assign debug info to the SDNode.
+void SelectionDAG::AssignDbgInfo(SDNode* SD, SDDbgValue* db) {
+  assert(SD && "Trying to assign dbg info to a null node!");
+  DbgInfo->add(SD, db);
+  SD->setHasDebugValue(true);
+}
+
+/// RememberDbgInfo - Remember debug info which is not assigned to an SDNode.
+void SelectionDAG::RememberDbgInfo(SDDbgValue* db) {
+  DbgInfo->add(db);
+}
+
+/// GetDbgInfo - Get the debug info, if any, for the SDNode.
+SDDbgValue* SelectionDAG::GetDbgInfo(const SDNode *SD) {
+  assert(SD && "Trying to get the order of a null node!");
+  if (SD->getHasDebugValue())
+    return DbgInfo->getSDDbgValue(SD);
+  return 0;
+}
 
 //===----------------------------------------------------------------------===//
 //                              SDNode Class
@@ -5911,7 +5938,7 @@
   if (G)
     if (unsigned Order = G->GetOrdering(this))
       OS << " [ORD=" << Order << ']';
-  
+
   if (getNodeId() != -1)
     OS << " [ID=" << getNodeId() << ']';
 }

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98195&r1=98194&r2=98195&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Wed Mar 10 16:13:47 2010
@@ -8521,6 +8521,21 @@
     F->DeleteMachineInstr(MI);   // The pseudo instruction is gone now.
     return BB;
   }
+    // DBG_VALUE.  Only the frame index case is done here.
+  case X86::DBG_VALUE: {
+    const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
+    DebugLoc DL = MI->getDebugLoc();
+    X86AddressMode AM;
+    MachineFunction *F = BB->getParent();
+    AM.BaseType = X86AddressMode::FrameIndexBase;
+    AM.Base.FrameIndex = MI->getOperand(0).getImm();
+    addFullAddress(BuildMI(BB, DL, TII->get(X86::DBG_VALUE)), AM).
+      addImm(MI->getOperand(1).getImm()).
+      addMetadata(MI->getOperand(2).getMetadata());
+    F->DeleteMachineInstr(MI);      // Remove pseudo.
+    return BB;
+  }
+
     // String/text processing lowering.
   case X86::PCMPISTRM128REG:
     return EmitPCMP(MI, BB, 3, false /* in-mem */);




From isanbard at gmail.com  Wed Mar 10 16:34:10 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 10 Mar 2010 22:34:10 -0000
Subject: [llvm-commits] [llvm] r98199 - in /llvm/trunk:
 include/llvm/CodeGen/MachineModuleInfo.h
 include/llvm/CodeGen/MachineModuleInfoImpls.h
 lib/CodeGen/MachineModuleInfoImpls.cpp
 lib/CodeGen/TargetLoweringObjectFileImpl.cpp
 lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
 lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
 lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
 lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
Message-ID: <20100310223410.CD8C72A6C12C@llvm.org>

Author: void
Date: Wed Mar 10 16:34:10 2010
New Revision: 98199

URL: http://llvm.org/viewvc/llvm-project?rev=98199&view=rev
Log:
Add a bit along with the MCSymbols stored in the MachineModuleInfo maps that
indicates that an MCSymbol is external or not. (It's true if it's external.)
This will be used to specify the correct information to add to non-lazy
pointers. That will be explained further when this bit is used.

Modified:
    llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
    llvm/trunk/include/llvm/CodeGen/MachineModuleInfoImpls.h
    llvm/trunk/lib/CodeGen/MachineModuleInfoImpls.cpp
    llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp

Modified: llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineModuleInfo.h Wed Mar 10 16:34:10 2010
@@ -31,48 +31,46 @@
 #ifndef LLVM_CODEGEN_MACHINEMODULEINFO_H
 #define LLVM_CODEGEN_MACHINEMODULEINFO_H
 
-#include "llvm/Support/Dwarf.h"
-#include "llvm/System/DataTypes.h"
-#include "llvm/ADT/SmallVector.h"
+#include "llvm/GlobalValue.h"
+#include "llvm/Pass.h"
+#include "llvm/Metadata.h"
 #include "llvm/ADT/DenseMap.h"
-#include "llvm/ADT/UniqueVector.h"
+#include "llvm/ADT/PointerIntPair.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallSet.h"
+#include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringMap.h"
+#include "llvm/ADT/UniqueVector.h"
 #include "llvm/CodeGen/MachineLocation.h"
-#include "llvm/GlobalValue.h"
-#include "llvm/Pass.h"
-#include "llvm/Metadata.h"
+#include "llvm/Support/Dwarf.h"
 #include "llvm/Support/ValueHandle.h"
+#include "llvm/System/DataTypes.h"
 
 namespace llvm {
 
 //===----------------------------------------------------------------------===//
 // Forward declarations.
 class Constant;
+class GlobalVariable;
 class MCSymbol;
 class MDNode;
-class GlobalVariable;
 class MachineBasicBlock;
 class MachineFunction;
 class Module;
 class PointerType;
 class StructType;
   
-  
 /// MachineModuleInfoImpl - This class can be derived from and used by targets
 /// to hold private target-specific information for each Module.  Objects of
 /// type are accessed/created with MMI::getInfo and destroyed when the
 /// MachineModuleInfo is destroyed.
 class MachineModuleInfoImpl {
 public:
+  typedef PointerIntPair<MCSymbol*, 1, bool> StubValueTy;
   virtual ~MachineModuleInfoImpl();
-
-  typedef std::vector<std::pair<MCSymbol*, MCSymbol*> >
-      SymbolListTy;
+  typedef std::vector<std::pair<MCSymbol*, StubValueTy> > SymbolListTy;
 protected:
-    static SymbolListTy
-    GetSortedStubs(const DenseMap<MCSymbol*, MCSymbol*> &Map);
+  static SymbolListTy GetSortedStubs(const DenseMap<MCSymbol*, StubValueTy>&);
 };
   
   

Modified: llvm/trunk/include/llvm/CodeGen/MachineModuleInfoImpls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineModuleInfoImpls.h?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineModuleInfoImpls.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineModuleInfoImpls.h Wed Mar 10 16:34:10 2010
@@ -25,32 +25,34 @@
   class MachineModuleInfoMachO : public MachineModuleInfoImpl {
     /// FnStubs - Darwin '$stub' stubs.  The key is something like "Lfoo$stub",
     /// the value is something like "_foo".
-    DenseMap<MCSymbol*, MCSymbol*> FnStubs;
+    DenseMap<MCSymbol*, StubValueTy> FnStubs;
     
     /// GVStubs - Darwin '$non_lazy_ptr' stubs.  The key is something like
-    /// "Lfoo$non_lazy_ptr", the value is something like "_foo".
-    DenseMap<MCSymbol*, MCSymbol*> GVStubs;
+    /// "Lfoo$non_lazy_ptr", the value is something like "_foo". The extra bit
+    /// is true if this GV is external.
+    DenseMap<MCSymbol*, StubValueTy> GVStubs;
     
     /// HiddenGVStubs - Darwin '$non_lazy_ptr' stubs.  The key is something like
     /// "Lfoo$non_lazy_ptr", the value is something like "_foo".  Unlike GVStubs
-    /// these are for things with hidden visibility.
-    DenseMap<MCSymbol*, MCSymbol*> HiddenGVStubs;
+    /// these are for things with hidden visibility. The extra bit is true if
+    /// this GV is external.
+    DenseMap<MCSymbol*, StubValueTy> HiddenGVStubs;
     
     virtual void Anchor();  // Out of line virtual method.
   public:
     MachineModuleInfoMachO(const MachineModuleInfo &) {}
     
-    MCSymbol *&getFnStubEntry(MCSymbol *Sym) {
+    StubValueTy &getFnStubEntry(MCSymbol *Sym) {
       assert(Sym && "Key cannot be null");
       return FnStubs[Sym];
     }
 
-    MCSymbol *&getGVStubEntry(MCSymbol *Sym) {
+    StubValueTy &getGVStubEntry(MCSymbol *Sym) {
       assert(Sym && "Key cannot be null");
       return GVStubs[Sym];
     }
 
-    MCSymbol *&getHiddenGVStubEntry(MCSymbol *Sym) {
+    StubValueTy &getHiddenGVStubEntry(MCSymbol *Sym) {
       assert(Sym && "Key cannot be null");
       return HiddenGVStubs[Sym];
     }
@@ -72,13 +74,13 @@
   class MachineModuleInfoELF : public MachineModuleInfoImpl {
     /// GVStubs - These stubs are used to materialize global addresses in PIC
     /// mode.
-    DenseMap<MCSymbol*, MCSymbol*> GVStubs;
+    DenseMap<MCSymbol*, StubValueTy> GVStubs;
 
     virtual void Anchor();  // Out of line virtual method.
   public:
     MachineModuleInfoELF(const MachineModuleInfo &) {}
 
-    MCSymbol *&getGVStubEntry(MCSymbol *Sym) {
+    StubValueTy &getGVStubEntry(MCSymbol *Sym) {
       assert(Sym && "Key cannot be null");
       return GVStubs[Sym];
     }

Modified: llvm/trunk/lib/CodeGen/MachineModuleInfoImpls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineModuleInfoImpls.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineModuleInfoImpls.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineModuleInfoImpls.cpp Wed Mar 10 16:34:10 2010
@@ -25,10 +25,9 @@
 void MachineModuleInfoELF::Anchor() {}
 
 static int SortSymbolPair(const void *LHS, const void *RHS) {
-  const MCSymbol *LHSS =
-    ((const std::pair<MCSymbol*, MCSymbol*>*)LHS)->first;
-  const MCSymbol *RHSS =
-    ((const std::pair<MCSymbol*, MCSymbol*>*)RHS)->first;
+  typedef std::pair<MCSymbol*, MachineModuleInfoImpl::StubValueTy> PairTy;
+  const MCSymbol *LHSS = ((const PairTy *)LHS)->first;
+  const MCSymbol *RHSS = ((const PairTy *)RHS)->first;
   return LHSS->getName().compare(RHSS->getName());
 }
 
@@ -36,7 +35,7 @@
 /// sorted orer.
 MachineModuleInfoImpl::SymbolListTy
 MachineModuleInfoImpl::GetSortedStubs(const DenseMap<MCSymbol*,
-                                                     MCSymbol*> &Map) {
+                                      MachineModuleInfoImpl::StubValueTy>&Map) {
   MachineModuleInfoImpl::SymbolListTy List(Map.begin(), Map.end());
 
   if (!List.empty())

Modified: llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Wed Mar 10 16:34:10 2010
@@ -404,14 +404,19 @@
     // Add information about the stub reference to ELFMMI so that the stub
     // gets emitted by the asmprinter.
     MCSymbol *Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
-    MCSymbol *&StubSym = ELFMMI.getGVStubEntry(Sym);
-    if (StubSym == 0) {
+    MachineModuleInfoImpl::StubValueTy &StubSym = ELFMMI.getGVStubEntry(Sym);
+    if (StubSym.getPointer() == 0) {
       Name.clear();
       Mang->getNameWithPrefix(Name, GV, false);
+
       if (GV->hasPrivateLinkage())
-        StubSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(getContext().GetOrCreateTemporarySymbol(Name.str()),
+                      false);
       else
-        StubSym = getContext().GetOrCreateSymbol(Name.str());
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(getContext().GetOrCreateSymbol(Name.str()),
+                      !GV->hasInternalLinkage());
     }
 
     return TargetLoweringObjectFile::
@@ -761,14 +766,19 @@
     // Add information about the stub reference to MachOMMI so that the stub
     // gets emitted by the asmprinter.
     MCSymbol *Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
-    MCSymbol *&StubSym = MachOMMI.getGVStubEntry(Sym);
-    if (StubSym == 0) {
+    MachineModuleInfoImpl::StubValueTy &StubSym = MachOMMI.getGVStubEntry(Sym);
+    if (StubSym.getPointer() == 0) {
       Name.clear();
       Mang->getNameWithPrefix(Name, GV, false);
+
       if (GV->hasPrivateLinkage())
-        StubSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(getContext().GetOrCreateTemporarySymbol(Name.str()),
+                      false);
       else
-        StubSym = getContext().GetOrCreateSymbol(Name.str());
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(getContext().GetOrCreateSymbol(Name.str()),
+                      !GV->hasInternalLinkage());
     }
 
     return TargetLoweringObjectFile::

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Wed Mar 10 16:34:10 2010
@@ -205,11 +205,12 @@
           
           MachineModuleInfoMachO &MMIMachO =
             MMI->getObjFileInfo<MachineModuleInfoMachO>();
-          MCSymbol *&StubSym =
+          MachineModuleInfoImpl::StubValueTy &StubSym =
             GV->hasHiddenVisibility() ? MMIMachO.getHiddenGVStubEntry(Sym) :
                                         MMIMachO.getGVStubEntry(Sym);
-          if (StubSym == 0)
-            StubSym = GetGlobalValueSymbol(GV);
+          if (StubSym.getPointer() == 0)
+            StubSym = MachineModuleInfoImpl::
+              StubValueTy(GetGlobalValueSymbol(GV), !GV->hasInternalLinkage());
         }
       } else {
         assert(ACPV->isExtSymbol() && "unrecognized constant pool value");
@@ -1126,7 +1127,7 @@
 
     // Output non-lazy-pointers for external and common global variables.
     MachineModuleInfoMachO::SymbolListTy Stubs = MMIMacho.GetGVStubList();
-    
+
     if (!Stubs.empty()) {
       // Switch with ".non_lazy_symbol_pointer" directive.
       OutStreamer.SwitchSection(TLOFMacho.getNonLazySymbolPointerSection());
@@ -1135,7 +1136,7 @@
         // L_foo$stub:
         OutStreamer.EmitLabel(Stubs[i].first);
         //   .indirect_symbol _foo
-        MCSymbol *MCSym = Stubs[i].second;
+        MCSymbol *MCSym = Stubs[i].second.getPointer();
         OutStreamer.EmitSymbolAttribute(MCSym, MCSA_IndirectSymbol);
 
         if (MCSym->isUndefined())
@@ -1159,8 +1160,9 @@
         // L_foo$stub:
         OutStreamer.EmitLabel(Stubs[i].first);
         //   .long _foo
-        OutStreamer.EmitValue(MCSymbolRefExpr::Create(Stubs[i].second,
-                                                      OutContext),
+        OutStreamer.EmitValue(MCSymbolRefExpr::
+                              Create(Stubs[i].second.getPointer(),
+                                     OutContext),
                               4/*size*/, 0/*addrspace*/);
       }
 

Modified: llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp Wed Mar 10 16:34:10 2010
@@ -198,10 +198,11 @@
           if (GV->isDeclaration() || GV->isWeakForLinker()) {
             // Dynamically-resolved functions need a stub for the function.
             MCSymbol *Sym = GetSymbolWithGlobalValueBase(GV, "$stub");
-            MCSymbol *&StubSym =
+            MachineModuleInfoImpl::StubValueTy &StubSym =
               MMI->getObjFileInfo<MachineModuleInfoMachO>().getFnStubEntry(Sym);
-            if (StubSym == 0)
-              StubSym = GetGlobalValueSymbol(GV);
+            if (StubSym.getPointer() == 0)
+              StubSym = MachineModuleInfoImpl::
+                StubValueTy(GetGlobalValueSymbol(GV),!GV->hasInternalLinkage());
             O << *Sym;
             return;
           }
@@ -212,10 +213,11 @@
           TempNameStr += StringRef("$stub");
           
           MCSymbol *Sym = GetExternalSymbolSymbol(TempNameStr.str());
-          MCSymbol *&StubSym =
+          MachineModuleInfoImpl::StubValueTy &StubSym =
             MMI->getObjFileInfo<MachineModuleInfoMachO>().getFnStubEntry(Sym);
-          if (StubSym == 0)
-            StubSym = GetExternalSymbolSymbol(MO.getSymbolName());
+          if (StubSym.getPointer() == 0)
+            StubSym = MachineModuleInfoImpl::
+              StubValueTy(GetExternalSymbolSymbol(MO.getSymbolName()), true);
           O << *Sym;
           return;
         }
@@ -404,10 +406,11 @@
     MCSymbol *NLPSym = 
       OutContext.GetOrCreateSymbol(StringRef(MAI->getGlobalPrefix())+
                                    MO.getSymbolName()+"$non_lazy_ptr");
-    MCSymbol *&StubSym = 
+    MachineModuleInfoImpl::StubValueTy &StubSym = 
       MMI->getObjFileInfo<MachineModuleInfoMachO>().getGVStubEntry(NLPSym);
-    if (StubSym == 0)
-      StubSym = GetExternalSymbolSymbol(MO.getSymbolName());
+    if (StubSym.getPointer() == 0)
+      StubSym = MachineModuleInfoImpl::
+        StubValueTy(GetExternalSymbolSymbol(MO.getSymbolName()), true);
     
     O << *NLPSym;
     return;
@@ -422,19 +425,23 @@
         (GV->isDeclaration() || GV->isWeakForLinker())) {
       if (!GV->hasHiddenVisibility()) {
         SymToPrint = GetSymbolWithGlobalValueBase(GV, "$non_lazy_ptr");
-        MCSymbol *&StubSym = 
-       MMI->getObjFileInfo<MachineModuleInfoMachO>().getGVStubEntry(SymToPrint);
-        if (StubSym == 0)
-          StubSym = GetGlobalValueSymbol(GV);
+        MachineModuleInfoImpl::StubValueTy &StubSym = 
+          MMI->getObjFileInfo<MachineModuleInfoMachO>()
+            .getGVStubEntry(SymToPrint);
+        if (StubSym.getPointer() == 0)
+          StubSym = MachineModuleInfoImpl::
+            StubValueTy(GetGlobalValueSymbol(GV), !GV->hasInternalLinkage());
       } else if (GV->isDeclaration() || GV->hasCommonLinkage() ||
                  GV->hasAvailableExternallyLinkage()) {
         SymToPrint = GetSymbolWithGlobalValueBase(GV, "$non_lazy_ptr");
         
-        MCSymbol *&StubSym = 
+        MachineModuleInfoImpl::StubValueTy &StubSym = 
           MMI->getObjFileInfo<MachineModuleInfoMachO>().
                     getHiddenGVStubEntry(SymToPrint);
-        if (StubSym == 0)
-          StubSym = GetGlobalValueSymbol(GV);
+        if (StubSym.getPointer() == 0)
+          StubSym = MachineModuleInfoImpl::
+            StubValueTy(GetGlobalValueSymbol(GV),
+                        !GV->hasInternalLinkage());
       } else {
         SymToPrint = GetGlobalValueSymbol(GV);
       }
@@ -704,7 +711,7 @@
       EmitAlignment(4);
       
       const MCSymbol *Stub = Stubs[i].first;
-      const MCSymbol *RawSym = Stubs[i].second;
+      const MCSymbol *RawSym = Stubs[i].second.getPointer();
       const MCSymbol *LazyPtr = GetLazyPtr(Stub, OutContext);
       const MCSymbol *AnonSymbol = GetAnonSym(Stub, OutContext);
                                            
@@ -738,7 +745,7 @@
                               16, SectionKind::getText());
   for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
     const MCSymbol *Stub = Stubs[i].first;
-    const MCSymbol *RawSym = Stubs[i].second;
+    const MCSymbol *RawSym = Stubs[i].second.getPointer();
     const MCSymbol *LazyPtr = GetLazyPtr(Stub, OutContext);
 
     OutStreamer.SwitchSection(StubSection);
@@ -781,8 +788,10 @@
          E = Personalities.end(); I != E; ++I) {
       if (*I) {
         MCSymbol *NLPSym = GetSymbolWithGlobalValueBase(*I, "$non_lazy_ptr");
-        MCSymbol *&StubSym = MMIMacho.getGVStubEntry(NLPSym);
-        StubSym = GetGlobalValueSymbol(*I);
+        MachineModuleInfoImpl::StubValueTy &StubSym =
+          MMIMacho.getGVStubEntry(NLPSym);
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(GetGlobalValueSymbol(*I), true);
       }
     }
   }
@@ -798,7 +807,8 @@
     
     for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
       O << *Stubs[i].first << ":\n";
-      O << "\t.indirect_symbol " << *Stubs[i].second << '\n';
+      O << "\t.indirect_symbol " << *Stubs[i].second.getPointer() << '\n';
+      // FIXME: This should use the "GV is external" bit.
       O << (isPPC64 ? "\t.quad\t0\n" : "\t.long\t0\n");
     }
   }
@@ -810,7 +820,8 @@
     
     for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
       O << *Stubs[i].first << ":\n";
-      O << (isPPC64 ? "\t.quad\t" : "\t.long\t") << *Stubs[i].second << '\n';
+      O << (isPPC64 ? "\t.quad\t" : "\t.long\t")
+        << *Stubs[i].second.getPointer() << '\n';
     }
   }
 

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Wed Mar 10 16:34:10 2010
@@ -133,24 +133,25 @@
     if (MO.getTargetFlags() == X86II::MO_DARWIN_NONLAZY ||
         MO.getTargetFlags() == X86II::MO_DARWIN_NONLAZY_PIC_BASE) {
       MCSymbol *Sym = GetSymbolWithGlobalValueBase(GV, "$non_lazy_ptr");
-      
-      MCSymbol *&StubSym = 
+      MachineModuleInfoImpl::StubValueTy &StubSym = 
         MMI->getObjFileInfo<MachineModuleInfoMachO>().getGVStubEntry(Sym);
-      if (StubSym == 0)
-        StubSym = GetGlobalValueSymbol(GV);
-      
+      if (StubSym.getPointer() == 0)
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(GetGlobalValueSymbol(GV), !GV->hasInternalLinkage());
     } else if (MO.getTargetFlags() == X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE){
       MCSymbol *Sym = GetSymbolWithGlobalValueBase(GV, "$non_lazy_ptr");
-      MCSymbol *&StubSym =
+      MachineModuleInfoImpl::StubValueTy &StubSym =
         MMI->getObjFileInfo<MachineModuleInfoMachO>().getHiddenGVStubEntry(Sym);
-      if (StubSym == 0)
-        StubSym = GetGlobalValueSymbol(GV);
+      if (StubSym.getPointer() == 0)
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(GetGlobalValueSymbol(GV), !GV->hasInternalLinkage());
     } else if (MO.getTargetFlags() == X86II::MO_DARWIN_STUB) {
       MCSymbol *Sym = GetSymbolWithGlobalValueBase(GV, "$stub");
-      MCSymbol *&StubSym =
+      MachineModuleInfoImpl::StubValueTy &StubSym =
         MMI->getObjFileInfo<MachineModuleInfoMachO>().getFnStubEntry(Sym);
-      if (StubSym == 0)
-        StubSym = GetGlobalValueSymbol(GV);
+      if (StubSym.getPointer() == 0)
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(GetGlobalValueSymbol(GV), !GV->hasInternalLinkage());
     }
     
     // If the name begins with a dollar-sign, enclose it in parens.  We do this
@@ -170,13 +171,15 @@
       TempNameStr += StringRef("$stub");
       
       MCSymbol *Sym = GetExternalSymbolSymbol(TempNameStr.str());
-      MCSymbol *&StubSym =
+      MachineModuleInfoImpl::StubValueTy &StubSym =
         MMI->getObjFileInfo<MachineModuleInfoMachO>().getFnStubEntry(Sym);
-      if (StubSym == 0) {
+      if (StubSym.getPointer() == 0) {
         TempNameStr.erase(TempNameStr.end()-5, TempNameStr.end());
-        StubSym = OutContext.GetOrCreateSymbol(TempNameStr.str());
+        StubSym = MachineModuleInfoImpl::
+          StubValueTy(OutContext.GetOrCreateSymbol(TempNameStr.str()),
+                      true);
       }
-      SymToPrint = StubSym;
+      SymToPrint = StubSym.getPointer();
     } else {
       SymToPrint = GetExternalSymbolSymbol(MO.getSymbolName());
     }
@@ -507,7 +510,8 @@
         // L_foo$stub:
         OutStreamer.EmitLabel(Stubs[i].first);
         //   .indirect_symbol _foo
-        OutStreamer.EmitSymbolAttribute(Stubs[i].second, MCSA_IndirectSymbol);
+        OutStreamer.EmitSymbolAttribute(Stubs[i].second.getPointer(),
+                                        MCSA_IndirectSymbol);
         // hlt; hlt; hlt; hlt; hlt     hlt = 0xf4 = -12.
         const char HltInsts[] = { -12, -12, -12, -12, -12 };
         OutStreamer.EmitBytes(StringRef(HltInsts, 5), 0/*addrspace*/);
@@ -530,7 +534,8 @@
         // L_foo$non_lazy_ptr:
         OutStreamer.EmitLabel(Stubs[i].first);
         // .indirect_symbol _foo
-        OutStreamer.EmitSymbolAttribute(Stubs[i].second, MCSA_IndirectSymbol);
+        OutStreamer.EmitSymbolAttribute(Stubs[i].second.getPointer(),
+                                        MCSA_IndirectSymbol);
         // .long 0
         OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
       }
@@ -547,8 +552,9 @@
         // L_foo$non_lazy_ptr:
         OutStreamer.EmitLabel(Stubs[i].first);
         // .long _foo
-        OutStreamer.EmitValue(MCSymbolRefExpr::Create(Stubs[i].second,
-                                                      OutContext),
+        OutStreamer.EmitValue(MCSymbolRefExpr::
+                              Create(Stubs[i].second.getPointer(),
+                                     OutContext),
                               4/*size*/, 0/*addrspace*/);
       }
       Stubs.clear();
@@ -624,7 +630,7 @@
         O << *Stubs[i].first << ":\n"
           << (TD->getPointerSize() == 8 ?
               MAI->getData64bitsDirective() : MAI->getData32bitsDirective())
-          << *Stubs[i].second << '\n';
+          << *Stubs[i].second.getPointer() << '\n';
 
       Stubs.clear();
     }

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98199&r1=98198&r2=98199&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Wed Mar 10 16:34:10 2010
@@ -91,35 +91,49 @@
     Name += "$non_lazy_ptr";
     MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
 
-    MCSymbol *&StubSym = getMachOMMI().getGVStubEntry(Sym);
-    if (StubSym == 0) {
+    MachineModuleInfoImpl::StubValueTy &StubSym =
+      getMachOMMI().getGVStubEntry(Sym);
+    if (StubSym.getPointer() == 0) {
       assert(MO.isGlobal() && "Extern symbol not handled yet");
-      StubSym = AsmPrinter.GetGlobalValueSymbol(MO.getGlobal());
+      StubSym =
+        MachineModuleInfoImpl::
+        StubValueTy(AsmPrinter.GetGlobalValueSymbol(MO.getGlobal()),
+                    !MO.getGlobal()->hasInternalLinkage());
     }
     return Sym;
   }
   case X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE: {
     Name += "$non_lazy_ptr";
     MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
-    MCSymbol *&StubSym = getMachOMMI().getHiddenGVStubEntry(Sym);
-    if (StubSym == 0) {
+    MachineModuleInfoImpl::StubValueTy &StubSym =
+      getMachOMMI().getHiddenGVStubEntry(Sym);
+    if (StubSym.getPointer() == 0) {
       assert(MO.isGlobal() && "Extern symbol not handled yet");
-      StubSym = AsmPrinter.GetGlobalValueSymbol(MO.getGlobal());
+      StubSym =
+        MachineModuleInfoImpl::
+        StubValueTy(AsmPrinter.GetGlobalValueSymbol(MO.getGlobal()),
+                    !MO.getGlobal()->hasInternalLinkage());
     }
     return Sym;
   }
   case X86II::MO_DARWIN_STUB: {
     Name += "$stub";
     MCSymbol *Sym = Ctx.GetOrCreateTemporarySymbol(Name.str());
-    MCSymbol *&StubSym = getMachOMMI().getFnStubEntry(Sym);
-    if (StubSym)
+    MachineModuleInfoImpl::StubValueTy &StubSym =
+      getMachOMMI().getFnStubEntry(Sym);
+    if (StubSym.getPointer())
       return Sym;
     
     if (MO.isGlobal()) {
-      StubSym = AsmPrinter.GetGlobalValueSymbol(MO.getGlobal());
+      StubSym =
+        MachineModuleInfoImpl::
+        StubValueTy(AsmPrinter.GetGlobalValueSymbol(MO.getGlobal()),
+                    !MO.getGlobal()->hasInternalLinkage());
     } else {
       Name.erase(Name.end()-5, Name.end());
-      StubSym = Ctx.GetOrCreateTemporarySymbol(Name.str());
+      StubSym =
+        MachineModuleInfoImpl::
+        StubValueTy(Ctx.GetOrCreateTemporarySymbol(Name.str()), false);
     }
     return Sym;
   }




From bob.wilson at apple.com  Wed Mar 10 16:38:45 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 10 Mar 2010 22:38:45 -0000
Subject: [llvm-commits] [llvm] r98200 -
	/llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp
Message-ID: <20100310223845.C90DC2A6C12C@llvm.org>

Author: bwilson
Date: Wed Mar 10 16:38:45 2010
New Revision: 98200

URL: http://llvm.org/viewvc/llvm-project?rev=98200&view=rev
Log:
Fix an obvious typo in an assert.
Patch by Sean Callanan.

Modified:
    llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp?rev=98200&r1=98199&r2=98200&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp Wed Mar 10 16:38:45 2010
@@ -436,7 +436,7 @@
 void ARMCodeEmitter::emitMOVi2piecesInstruction(const MachineInstr &MI) {
   const MachineOperand &MO0 = MI.getOperand(0);
   const MachineOperand &MO1 = MI.getOperand(1);
-  assert(MO1.isImm() && ARM_AM::getSOImmVal(MO1.isImm()) != -1 &&
+  assert(MO1.isImm() && ARM_AM::getSOImmVal(MO1.getImm()) != -1 &&
                                             "Not a valid so_imm value!");
   unsigned V1 = ARM_AM::getSOImmTwoPartFirst(MO1.getImm());
   unsigned V2 = ARM_AM::getSOImmTwoPartSecond(MO1.getImm());




From alenhar2 at llvm.org  Wed Mar 10 17:24:55 2010
From: alenhar2 at llvm.org (alenhar2 at llvm.org)
Date: Wed, 10 Mar 2010 23:24:55 -0000
Subject: [llvm-commits] [poolalloc] r98201 - in
	/poolalloc/trunk/include/dsa/sv: ./ set.h super_set.h
Message-ID: <20100310232455.591492A6C12C@llvm.org>

Author: alenhar2
Date: Wed Mar 10 17:24:55 2010
New Revision: 98201

URL: http://llvm.org/viewvc/llvm-project?rev=98201&view=rev
Log:
helper ADTs for sorted vectors and stable sets of sets

Added:
    poolalloc/trunk/include/dsa/sv/
    poolalloc/trunk/include/dsa/sv/set.h
    poolalloc/trunk/include/dsa/sv/super_set.h

Added: poolalloc/trunk/include/dsa/sv/set.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/sv/set.h?rev=98201&view=auto
==============================================================================
--- poolalloc/trunk/include/dsa/sv/set.h (added)
+++ poolalloc/trunk/include/dsa/sv/set.h Wed Mar 10 17:24:55 2010
@@ -0,0 +1,228 @@
+#ifndef _SV_ORDERED_SET_HH_
+#define _SV_ORDERED_SET_HH_ 1
+
+#include <algorithm>
+
+namespace sv {
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+/// A set implemented atop a sorted vector.
+template< typename Key,
+        typename Compare = std::less<Key>,
+        typename Alloc = std::allocator<Key> >
+class set {
+  typedef std::vector<Key, Alloc> internal_type;
+
+// Types
+public:
+
+  typedef Key     key_type;
+  typedef Key     value_type;
+  typedef Compare key_compare;
+  typedef Compare value_compare;
+  typedef Alloc   allocator_type;
+  typedef typename Alloc::reference reference;
+  typedef typename Alloc::const_reference const_reference;
+  typedef typename Alloc::pointer pointer;
+  typedef typename Alloc::const_pointer const_pointer;
+  typedef typename internal_type::const_iterator const_iterator;
+  typedef typename internal_type::iterator iterator;
+  typedef typename internal_type::reverse_iterator reverse_iterator;
+  typedef typename internal_type::const_reverse_iterator const_reverse_iterator;
+  typedef typename internal_type::size_type size_type;
+  typedef typename internal_type::difference_type difference_type;
+        
+private:
+  
+  internal_type container_;
+
+  void sort_unique() {
+    std::sort(container_.begin(), container_.end());
+    iterator i = std::unique(container_.begin(), container_.end());
+    container_.erase(i, container_.end());
+  }
+
+public:
+  /// Empty constructor.
+  set()
+  : container_() { }
+
+  /// Constructor taking a range.
+  template< typename Iterator >
+  set(const Iterator& begin, const Iterator& end)
+  : container_(begin, end) {
+    sort_unique();
+  }
+  
+  /// Copy-constructor.
+  set(const set& rhs)
+  : container_(rhs.container_)
+  {}
+
+  /// Affectation of a sorted vector to another one.
+
+  set & operator=(const set& rhs) {
+    if (&rhs != this) {
+      this->container_ = rhs.container_;
+    }
+    return *this;
+  }
+
+  /// Returns the beginning of the sorted vector.
+  const_iterator begin() const {
+    return container_.begin();
+  }
+
+  /// Returns the end of the sorted vector.
+  const_iterator end() const {
+    return container_.end();
+  }
+
+  /// Returns the beginning of the sorted vector.
+  iterator begin() {
+    return container_.begin();
+  }
+
+  /// Returns the end of the sorted vector.
+  iterator end() {
+    return container_.end();
+  }
+
+  /// Returns the beginning of the sorted vector.
+  const_reverse_iterator rrbegin() const {
+    return container_.rbegin();
+  }
+
+  /// Returns the end of the sorted vector.
+  const_reverse_iterator rend() const {
+    return container_.rend();
+  }
+
+  /// Returns the beginning of the sorted vector.
+  reverse_iterator rbegin() {
+    return container_.rbegin();
+  }
+
+  /// Returns the end of the sorted vector.
+  reverse_iterator rend() {
+    return container_.rend();
+  }
+
+  bool empty() const {
+    return container_.empty();
+  }
+
+  size_type size() const {
+    return container_.size();
+  }
+
+  size_type max_size() const {
+    return container_.max_size();
+  }
+
+  /// Insert a value into the sorted vector.
+  std::pair<iterator,bool>
+  insert(const value_type& x) {
+    bool insertion = false;
+    iterator i = std::lower_bound(container_.begin(), container_.end(), x);
+    if (i == container_.end() || x < *i) {
+      i = container_.insert(i, x);
+      insertion = true;
+    }
+    return std::make_pair(i, insertion);
+  }
+
+  /// Insert a range.
+  template < typename Iterator >
+  void insert(Iterator _begin, Iterator _end) {
+    while (_begin != _end)
+      container_.push_back(*_begin++);
+    sort_unique();
+  }
+
+  void erase ( iterator position ) {
+    container_.erase(position);
+  }
+
+  size_type erase(const key_type& x) {
+    iterator i = find(x);
+    if (i != end()) {
+      erase(i);
+      return 1;
+    }
+    return 0;
+  }
+
+  void erase ( iterator first, iterator last ) {
+    container_.erase(first, last);
+  }
+
+  /// Swap the content of two sorted_vector.
+  void swap(set& s) {
+    container_.swap(s.container_);
+  }
+
+  void clear() {
+    container_.clear();
+  }
+
+  /// Find the key k.
+
+  const_iterator find(const key_type& k) const {
+    const_iterator i = std::lower_bound(container_.begin(), container_.end(), k);
+    if (*i == k) return i;
+    return container_.end();
+  }
+
+  iterator find(const key_type& k) {
+    iterator i = std::lower_bound(container_.begin(), container_.end(), k);
+    if (*i == k) return i;
+    return container_.end();
+  }
+
+  bool count(const key_type& k) const {
+    return find(k) != end();
+  }
+
+  iterator lower_bound(const key_type& x) const {
+    return std::lower_bound(container_.begin(), container_.end(), x);
+  }
+
+  iterator upper_bound(const key_type& x) const {
+    return std::upper_bound(container_.begin(), container_.end(), x);
+  }
+
+  std::pair<iterator,iterator> equal_range(const key_type& x) const {
+    return std::make_pair(lower_bound(x),upper_bound(x));
+  }
+
+
+  /// Tells if this sorted vector is equal to another one.
+  bool operator==(const set& rhs) const {
+    return container_ == rhs.container_;
+  }
+  bool operator<(const set& rhs) const {
+    if (rhs.size() < size()) return false;
+    if (size() < rhs.size()) return true;
+    const_iterator ii = begin();
+    const_iterator ee = end();
+    const_iterator rhs_ii = rhs.begin();
+    const_iterator rhs_ee = rhs.end();
+    while (ii != ee) {
+      if (*rhs_ii < *ii) return false;
+      if (*ii < *rhs_ii) return true;
+      ++ii;
+      ++rhs_ii;
+    }
+    // Equal
+    return false;
+  }
+  
+};
+
+////////////////////////////////////////////////////////////////////////////////////////////////////
+
+} // end of namespace sv
+
+#endif // _SV_ORDERED_SET_HH_

Added: poolalloc/trunk/include/dsa/sv/super_set.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/sv/super_set.h?rev=98201&view=auto
==============================================================================
--- poolalloc/trunk/include/dsa/sv/super_set.h (added)
+++ poolalloc/trunk/include/dsa/sv/super_set.h Wed Mar 10 17:24:55 2010
@@ -0,0 +1,41 @@
+/* 
+ * File:   super_set.h
+ * Author: andrew
+ *
+ * Created on March 10, 2010, 2:04 PM
+ */
+
+#ifndef _SUPER_SET_H
+#define	_SUPER_SET_H
+
+#include "./set.h"
+#include <set>
+
+// Contains stable references to a set
+// The sets can be grown.
+
+template<typename Ty>
+class SuperSet {
+  typedef sv::set<Ty> InnerSetTy;
+  typedef std::set<InnerSetTy> OuterSetTy;
+  OuterSetTy container;
+public:
+  typedef const typename OuterSetTy::value_type* setPtr;
+
+  setPtr getOrCreate(sv::set<Ty>& S) {
+    return &(*container.insert(S).first);
+  }
+
+  setPtr getOrCreate(setPtr P, Ty t) {
+    sv::set<Ty> s;
+    if (P)
+      s.insert(P->begin(), P->end());
+    s.insert(t);
+    return getOrCreate(s);
+  }
+};
+
+
+
+#endif	/* _SUPER_SET_H */
+




From evan.cheng at apple.com  Wed Mar 10 17:28:08 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 10 Mar 2010 15:28:08 -0800
Subject: [llvm-commits] [llvm] r98195 - in /llvm/trunk:
	include/llvm/CodeGen/ScheduleDAG.h
	include/llvm/CodeGen/SelectionDAG.h
	lib/CodeGen/ScheduleDAGInstrs.cpp lib/CodeGen/ScheduleDAGInstrs.h
	lib/CodeGen/SelectionDAG/InstrEmitter.cpp
	lib/CodeGen/SelectionDAG/InstrEmitter.h
	lib/CodeGen/SelectionDAG/SDDbgValue.h
	lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
	lib/CodeGen/SelectionDAG/SelectionDAG.cpp
	lib/Target/X86/X86ISelLowering.cpp
In-Reply-To: <20100310221348.5EC7D2A6C12C@llvm.org>
References: <20100310221348.5EC7D2A6C12C@llvm.org>
Message-ID: <F07388F7-7EE1-4CAD-A012-667738987F38@apple.com>

I need to run it through some examples to understand more about the patch. But I noticed this cosmetic issue:

> +  enum DbgValueKind {
> +    SD = 0,
> +    CNST = 1,
> +    FX = 2
> +  };

Can you spell out the enum so they are more obvious?

Thanks,

Evan

On Mar 10, 2010, at 2:13 PM, Dale Johannesen wrote:

> Author: johannes
> Date: Wed Mar 10 16:13:47 2010
> New Revision: 98195
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98195&view=rev
> Log:
> Progress towards shepherding debug info through SelectionDAG.
> No functional effect yet.  This is still evolving and should
> not be viewed as final.
> 
> 
> Modified:
>    llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h
>    llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
>    llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp
>    llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h
>    llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
>    llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h
>    llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h
>    llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
>    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
>    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> 
> Modified: llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/ScheduleDAG.h Wed Mar 10 16:13:47 2010
> @@ -227,6 +227,7 @@
>   private:
>     SDNode *Node;                       // Representative node.
>     MachineInstr *Instr;                // Alternatively, a MachineInstr.
> +    MachineInstr *DbgInstr;             // A dbg_value referencing this.
>   public:
>     SUnit *OrigNode;                    // If not this, the node from which
>                                         // this node was cloned.
> @@ -269,10 +270,10 @@
>     /// SUnit - Construct an SUnit for pre-regalloc scheduling to represent
>     /// an SDNode and any nodes flagged to it.
>     SUnit(SDNode *node, unsigned nodenum)
> -      : Node(node), Instr(0), OrigNode(0), NodeNum(nodenum), NodeQueueId(0),
> -        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
> -        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
> -        hasPhysRegClobbers(false),
> +      : Node(node), Instr(0), DbgInstr(0), OrigNode(0), NodeNum(nodenum),
> +        NodeQueueId(0),  Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
> +        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
> +        hasPhysRegDefs(false), hasPhysRegClobbers(false),
>         isPending(false), isAvailable(false), isScheduled(false),
>         isScheduleHigh(false), isCloned(false),
>         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
> @@ -281,10 +282,10 @@
>     /// SUnit - Construct an SUnit for post-regalloc scheduling to represent
>     /// a MachineInstr.
>     SUnit(MachineInstr *instr, unsigned nodenum)
> -      : Node(0), Instr(instr), OrigNode(0), NodeNum(nodenum), NodeQueueId(0),
> -        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
> -        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
> -        hasPhysRegClobbers(false),
> +      : Node(0), Instr(instr), DbgInstr(0), OrigNode(0), NodeNum(nodenum),
> +        NodeQueueId(0), Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
> +        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
> +        hasPhysRegDefs(false), hasPhysRegClobbers(false),
>         isPending(false), isAvailable(false), isScheduled(false),
>         isScheduleHigh(false), isCloned(false),
>         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
> @@ -292,10 +293,10 @@
> 
>     /// SUnit - Construct a placeholder SUnit.
>     SUnit()
> -      : Node(0), Instr(0), OrigNode(0), NodeNum(~0u), NodeQueueId(0),
> -        Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0), NumSuccsLeft(0),
> -        isTwoAddress(false), isCommutable(false), hasPhysRegDefs(false),
> -        hasPhysRegClobbers(false),
> +      : Node(0), Instr(0), DbgInstr(0), OrigNode(0), NodeNum(~0u),
> +        NodeQueueId(0), Latency(0), NumPreds(0), NumSuccs(0), NumPredsLeft(0),
> +        NumSuccsLeft(0), isTwoAddress(false), isCommutable(false),
> +        hasPhysRegDefs(false), hasPhysRegClobbers(false),
>         isPending(false), isAvailable(false), isScheduled(false),
>         isScheduleHigh(false), isCloned(false),
>         isDepthCurrent(false), isHeightCurrent(false), Depth(0), Height(0),
> @@ -329,6 +330,20 @@
>       return Instr;
>     }
> 
> +    /// setDbgInstr - Assign the debug instruction for the SUnit.
> +    /// This may be used during post-regalloc scheduling.
> +    void setDbgInstr(MachineInstr *MI) {
> +      assert(!Node && "Setting debug MachineInstr of SUnit with SDNode!");
> +      DbgInstr = MI;
> +    }
> +
> +    /// getDbgInstr - Return the debug MachineInstr for this SUnit.
> +    /// This may be used during post-regalloc scheduling.
> +    MachineInstr *getDbgInstr() const {
> +      assert(!Node && "Reading debug MachineInstr of SUnit with SDNode!");
> +      return DbgInstr;
> +    }
> +
>     /// addPred - This adds the specified edge as a pred of the current node if
>     /// not already.  It also adds the current node as a successor of the
>     /// specified node.
> 
> Modified: llvm/trunk/include/llvm/CodeGen/SelectionDAG.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/SelectionDAG.h?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/SelectionDAG.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/SelectionDAG.h Wed Mar 10 16:13:47 2010
> @@ -35,6 +35,7 @@
> class MachineFunction;
> class MachineModuleInfo;
> class SDNodeOrdering;
> +class SDDbgValue;
> class TargetLowering;
> 
> template<> struct ilist_traits<SDNode> : public ilist_default_traits<SDNode> {
> @@ -57,6 +58,46 @@
>   static void createNode(const SDNode &);
> };
> 
> +/// SDDbgInfo - Keeps track of dbg_value information through SDISel.  We do
> +/// not build SDNodes for these so as not to perturb the generated code;
> +/// instead the info is kept off to the side in this structure.  SDNodes may
> +/// have an associated dbg_value entry in DbgValMap.  Debug info that is not
> +/// associated with any SDNode is held in DbgConstMap.  It is possible for
> +/// optimizations to change a variable to a constant, in which case the
> +/// corresponding debug info is moved from the variable to the constant table
> +/// (NYI).
> +class SDDbgInfo {
> +  DenseMap<const SDNode*, SDDbgValue*> DbgVblMap;
> +  SmallVector<SDDbgValue*, 4> DbgConstMap;
> +
> +  void operator=(const SDDbgInfo&);   // Do not implement.
> +  SDDbgInfo(const SDDbgInfo&);   // Do not implement.
> +public:
> +  SDDbgInfo() {}
> +
> +  void add(const SDNode *Node, SDDbgValue *V) {
> +    DbgVblMap[Node] = V;
> +  }
> +  void add(SDDbgValue *V) { DbgConstMap.push_back(V); }
> +  void remove(const SDNode *Node) {
> +    DenseMap<const SDNode*, SDDbgValue*>::iterator Itr =
> +                      DbgVblMap.find(Node);
> +    if (Itr != DbgVblMap.end())
> +      DbgVblMap.erase(Itr);
> +  }
> +  // No need to remove a constant.
> +  void clear() {
> +    DbgVblMap.clear();
> +    DbgConstMap.clear();
> +  }
> +  SDDbgValue *getSDDbgValue(const SDNode *Node) {
> +    return DbgVblMap[Node];
> +  }
> +  typedef SmallVector<SDDbgValue*, 4>::iterator ConstDbgIterator;
> +  ConstDbgIterator DbgConstBegin() { return DbgConstMap.begin(); }
> +  ConstDbgIterator DbgConstEnd() { return DbgConstMap.end(); }
> +};
> +
> enum CombineLevel {
>   Unrestricted,   // Combine may create illegal operations and illegal types.
>   NoIllegalTypes, // Combine may create illegal operations but no illegal types.
> @@ -119,6 +160,9 @@
>   /// the ordering of the original LLVM instructions.
>   SDNodeOrdering *Ordering;
> 
> +  /// DbgInfo - Tracks dbg_value information through SDISel.
> +  SDDbgInfo *DbgInfo;
> +
>   /// VerifyNode - Sanity check the given node.  Aborts if it is invalid.
>   void VerifyNode(SDNode *N);
> 
> @@ -828,6 +872,20 @@
>   /// GetOrdering - Get the order for the SDNode.
>   unsigned GetOrdering(const SDNode *SD) const;
> 
> +  /// AssignDbgInfo - Assign debug info to the SDNode.
> +  void AssignDbgInfo(SDNode *SD, SDDbgValue *db);
> +
> +  /// RememberDbgInfo - Remember debug info with no associated SDNode.
> +  void RememberDbgInfo(SDDbgValue *db);
> +
> +  /// GetDbgInfo - Get the debug info for the SDNode.
> +  SDDbgValue *GetDbgInfo(const SDNode* SD);
> +
> +  SDDbgInfo::ConstDbgIterator DbgConstBegin() { 
> +    return DbgInfo->DbgConstBegin(); 
> +  }
> +  SDDbgInfo::ConstDbgIterator DbgConstEnd() { return DbgInfo->DbgConstEnd(); }
> +
>   void dump() const;
> 
>   /// CreateStackTemporary - Create a stack temporary, suitable for holding the
> 
> Modified: llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp (original)
> +++ llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.cpp Wed Mar 10 16:13:47 2010
> @@ -34,6 +34,7 @@
>                                      const MachineDominatorTree &mdt)
>   : ScheduleDAG(mf), MLI(mli), MDT(mdt), LoopRegs(MLI, MDT) {
>   MFI = mf.getFrameInfo();
> +  DbgValueVec.clear();
> }
> 
> /// Run - perform scheduling.
> @@ -157,6 +158,10 @@
>   std::map<const Value *, SUnit *> AliasMemDefs, NonAliasMemDefs;
>   std::map<const Value *, std::vector<SUnit *> > AliasMemUses, NonAliasMemUses;
> 
> +  // Keep track of dangling debug references to registers.
> +  std::pair<MachineInstr*, unsigned>
> +        DanglingDebugValue[TargetRegisterInfo::FirstVirtualRegister];
> +
>   // Check to see if the scheduler cares about latencies.
>   bool UnitLatencies = ForceUnitLatencies();
> 
> @@ -164,10 +169,25 @@
>   const TargetSubtarget &ST = TM.getSubtarget<TargetSubtarget>();
>   unsigned SpecialAddressLatency = ST.getSpecialAddressLatency();
> 
> +  // Remove any stale debug info; sometimes BuildSchedGraph is called again
> +  // without emitting the info from the previous call.
> +  DbgValueVec.clear();
> +  std::memset(DanglingDebugValue, 0, sizeof(DanglingDebugValue));
> +
>   // Walk the list of instructions, from bottom moving up.
>   for (MachineBasicBlock::iterator MII = InsertPos, MIE = Begin;
>        MII != MIE; --MII) {
>     MachineInstr *MI = prior(MII);
> +    // DBG_VALUE does not have SUnit's built, so just remember these for later
> +    // reinsertion.
> +    if (MI->isDebugValue()) {
> +      if (MI->getNumOperands()==3 && MI->getOperand(0).isReg() &&
> +          MI->getOperand(0).getReg())
> +        DanglingDebugValue[MI->getOperand(0).getReg()] =
> +             std::make_pair(MI, DbgValueVec.size());
> +      DbgValueVec.push_back(MI);
> +      continue;
> +    }
>     const TargetInstrDesc &TID = MI->getDesc();
>     assert(!TID.isTerminator() && !MI->isLabel() &&
>            "Cannot schedule terminators or labels!");
> @@ -188,6 +208,13 @@
>       if (Reg == 0) continue;
> 
>       assert(TRI->isPhysicalRegister(Reg) && "Virtual register encountered!");
> +
> +      if (MO.isDef() && DanglingDebugValue[Reg].first!=0) {
> +        SU->setDbgInstr(DanglingDebugValue[Reg].first);
> +        DbgValueVec[DanglingDebugValue[Reg].second] = 0;
> +        DanglingDebugValue[Reg] = std::make_pair((MachineInstr*)0, 0);
> +      }
> +
>       std::vector<SUnit *> &UseList = Uses[Reg];
>       std::vector<SUnit *> &DefList = Defs[Reg];
>       // Optionally add output and anti dependencies. For anti
> @@ -555,6 +582,14 @@
>     BB->remove(I);
>   }
> 
> +  // First reinsert any remaining debug_values; these are either constants,
> +  // or refer to live-in registers.  The beginning of the block is the right
> +  // place for the latter.  The former might reasonably be placed elsewhere
> +  // using some kind of ordering algorithm, but right now it doesn't matter.
> +  for (int i = DbgValueVec.size()-1; i>=0; --i)
> +    if (DbgValueVec[i])
> +      BB->insert(InsertPos, DbgValueVec[i]);
> +
>   // Then re-insert them according to the given schedule.
>   for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
>     SUnit *SU = Sequence[i];
> @@ -565,12 +600,21 @@
>     }
> 
>     BB->insert(InsertPos, SU->getInstr());
> +    if (SU->getDbgInstr())
> +      BB->insert(InsertPos, SU->getDbgInstr());
>   }
> 
>   // Update the Begin iterator, as the first instruction in the block
>   // may have been scheduled later.
> -  if (!Sequence.empty())
> +  if (!DbgValueVec.empty()) {
> +    for (int i = DbgValueVec.size()-1; i>=0; --i)
> +      if (DbgValueVec[i]!=0) {
> +        Begin = DbgValueVec[DbgValueVec.size()-1];
> +        break;
> +      }
> +  } else if (!Sequence.empty())
>     Begin = Sequence[0]->getInstr();
> 
> +  DbgValueVec.clear();
>   return BB;
> }
> 
> Modified: llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h (original)
> +++ llvm/trunk/lib/CodeGen/ScheduleDAGInstrs.h Wed Mar 10 16:13:47 2010
> @@ -106,6 +106,10 @@
>     /// initialized and destructed for each block.
>     std::vector<SUnit *> Defs[TargetRegisterInfo::FirstVirtualRegister];
>     std::vector<SUnit *> Uses[TargetRegisterInfo::FirstVirtualRegister];
> + 
> +    /// DbgValueVec - Remember DBG_VALUEs that refer to a particular
> +    /// register.
> +    std::vector<MachineInstr *>DbgValueVec;
> 
>     /// PendingLoads - Remember where unknown loads are after the most recent
>     /// unknown store, as we iterate. As with Defs and Uses, this is here
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp Wed Mar 10 16:13:47 2010
> @@ -508,6 +508,7 @@
>     return;
>   if (!sd)
>     return;
> +  assert(sd->getKind() == SDDbgValue::SD);
>   unsigned VReg = getVR(SDValue(sd->getSDNode(), sd->getResNo()), VRBaseMap);
>   const TargetInstrDesc &II = TII->get(TargetOpcode::DBG_VALUE);
>   DebugLoc DL = sd->getDebugLoc();
> @@ -524,26 +525,46 @@
>   MBB->insert(InsertPos, MI);
> }
> 
> -/// EmitDbgValue - Generate constant debug info.  No SDNode is involved.
> +/// EmitDbgValue - Generate debug info that does not refer to a SDNode.
> void
> -InstrEmitter::EmitDbgValue(SDDbgValue *sd) {
> +InstrEmitter::EmitDbgValue(SDDbgValue *sd,
> +                         DenseMap<MachineBasicBlock*, MachineBasicBlock*> *EM) {
>   if (!sd)
>     return;
>   const TargetInstrDesc &II = TII->get(TargetOpcode::DBG_VALUE);
> +  uint64_t Offset = sd->getOffset();
> +  MDNode* mdPtr = sd->getMDPtr();
> +  SDDbgValue::DbgValueKind kind = sd->getKind();
>   DebugLoc DL = sd->getDebugLoc();
> -  MachineInstr *MI;
> -  Value *V = sd->getConst();
> -  if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
> -    MI = BuildMI(*MF, DL, II).addImm(CI->getZExtValue()).
> -                                   addImm(sd->getOffset()).
> -                                   addMetadata(sd->getMDPtr());
> -  } else if (ConstantFP *CF = dyn_cast<ConstantFP>(V)) {
> -    MI = BuildMI(*MF, DL, II).addFPImm(CF).addImm(sd->getOffset()).
> -                                   addMetadata(sd->getMDPtr());
> +  MachineInstr* MI;
> +  if (kind == SDDbgValue::CNST) {
> +    Value *V = sd->getConst();
> +    if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
> +      MI = BuildMI(*MF, DL, II).addImm(CI->getZExtValue()).
> +                                     addImm(Offset).addMetadata(mdPtr);
> +    } else if (ConstantFP *CF = dyn_cast<ConstantFP>(V)) {
> +      MI = BuildMI(*MF, DL, II).addFPImm(CF).
> +                                     addImm(Offset).addMetadata(mdPtr);
> +    } else {
> +      // Could be an Undef.  In any case insert an Undef so we can see what we
> +      // dropped.
> +      MI = BuildMI(*MF, DL, II).addReg(0U).
> +                                       addImm(Offset).addMetadata(mdPtr);
> +    }
> +  } else if (kind == SDDbgValue::FX) {
> +    unsigned FrameIx = sd->getFrameIx();
> +    // Stack address; this needs to be lowered in target-dependent fashion.
> +    // FIXME test that the target supports this somehow; if not emit Undef.
> +    // Create a pseudo for EmitInstrWithCustomInserter's consumption.
> +    MI = BuildMI(*MF, DL, II).addImm(FrameIx).
> +                                   addImm(Offset).addMetadata(mdPtr);
> +    MBB = TLI->EmitInstrWithCustomInserter(MI, MBB, EM);
> +    InsertPos = MBB->end();
> +    return;
>   } else {
>     // Insert an Undef so we can see what we dropped.
> -    MI = BuildMI(*MF, DL, II).addReg(0U).addImm(sd->getOffset()).
> -                                    addMetadata(sd->getMDPtr());
> +    MI = BuildMI(*MF, DL, II).addReg(0U).
> +                                     addImm(Offset).addMetadata(mdPtr);
>   }
>   MBB->insert(InsertPos, MI);
> }
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.h Wed Mar 10 16:13:47 2010
> @@ -106,7 +106,8 @@
> 
> 
>   /// EmitDbgValue - Generate a constant DBG_VALUE.  No node is involved.
> -  void EmitDbgValue(SDDbgValue* sd);
> +  void EmitDbgValue(SDDbgValue* sd,
> +                DenseMap<MachineBasicBlock*, MachineBasicBlock*> *EM);
> 
>   /// EmitNode - Generate machine code for a node and needed dependencies.
>   ///
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h Wed Mar 10 16:13:47 2010
> @@ -28,9 +28,22 @@
> /// We do not use SDValue here to avoid including its header.
> 
> class SDDbgValue {
> -  SDNode *Node;           // valid for non-constants
> -  unsigned ResNo;         // valid for non-constants
> -  Value *Const;           // valid for constants
> +public:
> +  enum DbgValueKind {
> +    SD = 0,
> +    CNST = 1,
> +    FX = 2
> +  };
> +private:
> +  enum DbgValueKind kind;
> +  union {
> +    struct {
> +      SDNode *Node;         // valid for non-constants
> +      unsigned ResNo;       // valid for non-constants
> +    } s;
> +    Value *Const;           // valid for constants
> +    unsigned FrameIx;       // valid for stack objects
> +  } u;
>   MDNode *mdPtr;
>   uint64_t Offset;
>   DebugLoc DL;
> @@ -38,24 +51,43 @@
> public:
>   // Constructor for non-constants.
>   SDDbgValue(MDNode *mdP, SDNode *N, unsigned R, uint64_t off, DebugLoc dl,
> -             unsigned O) :
> -    Node(N), ResNo(R), Const(0), mdPtr(mdP), Offset(off), DL(dl), Order(O) {}
> +             unsigned O) : mdPtr(mdP), Offset(off), DL(dl), Order(O) {
> +    kind = SD;
> +    u.s.Node = N;
> +    u.s.ResNo = R;
> +  }
> 
>   // Constructor for constants.
>   SDDbgValue(MDNode *mdP, Value *C, uint64_t off, DebugLoc dl, unsigned O) : 
> -    Node(0), ResNo(0), Const(C), mdPtr(mdP), Offset(off), DL(dl), Order(O) {}
> +    mdPtr(mdP), Offset(off), DL(dl), Order(O) {
> +    kind = CNST;
> +    u.Const = C;
> +  }
> +
> +  // Constructor for frame indices.
> +  SDDbgValue(MDNode *mdP, unsigned FI, uint64_t off, DebugLoc dl, unsigned O) : 
> +    mdPtr(mdP), Offset(off), DL(dl), Order(O) {
> +    kind = FX;
> +    u.FrameIx = FI;
> +  }
> +
> +  // Returns the kind.
> +  DbgValueKind getKind() { return kind; }
> 
>   // Returns the MDNode pointer.
>   MDNode *getMDPtr() { return mdPtr; }
> 
> -  // Returns the SDNode* (valid for non-constants only).
> -  SDNode *getSDNode() { assert (!Const); return Node; }
> +  // Returns the SDNode* for a register ref
> +  SDNode *getSDNode() { assert (kind==SD); return u.s.Node; }
> +
> +  // Returns the ResNo for a register ref
> +  unsigned getResNo() { assert (kind==SD); return u.s.ResNo; }
> 
> -  // Returns the ResNo (valid for non-constants only).
> -  unsigned getResNo() { assert (!Const); return ResNo; }
> +  // Returns the Value* for a constant
> +  Value *getConst() { assert (kind==CNST); return u.Const; }
> 
> -  // Returns the Value* for a constant (invalid for non-constants).
> -  Value *getConst() { assert (!Node); return Const; }
> +  // Returns the FrameIx for a stack object
> +  unsigned getFrameIx() { assert (kind==FX); return u.FrameIx; }
> 
>   // Returns the offset.
>   uint64_t getOffset() { return Offset; }
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp Wed Mar 10 16:13:47 2010
> @@ -13,6 +13,7 @@
> //===----------------------------------------------------------------------===//
> 
> #define DEBUG_TYPE "pre-RA-sched"
> +#include "SDDbgValue.h"
> #include "ScheduleDAGSDNodes.h"
> #include "InstrEmitter.h"
> #include "llvm/CodeGen/SelectionDAG.h"
> @@ -412,6 +413,14 @@
>   InstrEmitter Emitter(BB, InsertPos);
>   DenseMap<SDValue, unsigned> VRBaseMap;
>   DenseMap<SUnit*, unsigned> CopyVRBaseMap;
> +
> +  // For now, any constant debug info nodes go at the beginning.
> +  for (SDDbgInfo::ConstDbgIterator I = DAG->DbgConstBegin(),
> +       E = DAG->DbgConstEnd(); I!=E; I++) {
> +    Emitter.EmitDbgValue(*I, EM);
> +    delete *I;
> +  }
> +
>   for (unsigned i = 0, e = Sequence.size(); i != e; i++) {
>     SUnit *SU = Sequence[i];
>     if (!SU) {
> @@ -435,10 +444,20 @@
>     while (!FlaggedNodes.empty()) {
>       Emitter.EmitNode(FlaggedNodes.back(), SU->OrigNode != SU, SU->isCloned,
>                        VRBaseMap, EM);
> +      if (FlaggedNodes.back()->getHasDebugValue())
> +        if (SDDbgValue *sd = DAG->GetDbgInfo(FlaggedNodes.back())) {
> +          Emitter.EmitDbgValue(FlaggedNodes.back(), VRBaseMap, sd);
> +          delete sd;
> +        }
>       FlaggedNodes.pop_back();
>     }
>     Emitter.EmitNode(SU->getNode(), SU->OrigNode != SU, SU->isCloned,
>                      VRBaseMap, EM);
> +    if (SU->getNode()->getHasDebugValue())
> +      if (SDDbgValue *sd = DAG->GetDbgInfo(SU->getNode())) {
> +        Emitter.EmitDbgValue(SU->getNode(), VRBaseMap, sd);
> +        delete sd;
> +      }
>   }
> 
>   BB = Emitter.getBlock();
> 
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAG.cpp Wed Mar 10 16:13:47 2010
> @@ -13,6 +13,7 @@
> 
> #include "llvm/CodeGen/SelectionDAG.h"
> #include "SDNodeOrdering.h"
> +#include "SDDbgValue.h"
> #include "llvm/Constants.h"
> #include "llvm/Analysis/ValueTracking.h"
> #include "llvm/Function.h"
> @@ -596,6 +597,9 @@
> 
>   // Remove the ordering of this node.
>   Ordering->remove(N);
> +
> +  // And its entry in the debug info table, if any.
> +  DbgInfo->remove(N);
> }
> 
> /// RemoveNodeFromCSEMaps - Take the specified node out of the CSE map that
> @@ -793,6 +797,7 @@
>     Root(getEntryNode()), Ordering(0) {
>   AllNodes.push_back(&EntryNode);
>   Ordering = new SDNodeOrdering();
> +  DbgInfo = new SDDbgInfo();
> }
> 
> void SelectionDAG::init(MachineFunction &mf, MachineModuleInfo *mmi,
> @@ -806,6 +811,7 @@
> SelectionDAG::~SelectionDAG() {
>   allnodes_clear();
>   delete Ordering;
> +  delete DbgInfo;
> }
> 
> void SelectionDAG::allnodes_clear() {
> @@ -833,6 +839,8 @@
>   Root = getEntryNode();
>   delete Ordering;
>   Ordering = new SDNodeOrdering();
> +  delete DbgInfo;
> +  DbgInfo = new SDDbgInfo();
> }
> 
> SDValue SelectionDAG::getSExtOrTrunc(SDValue Op, DebugLoc DL, EVT VT) {
> @@ -5264,6 +5272,25 @@
>   return Ordering->getOrder(SD);
> }
> 
> +/// AssignDbgInfo - Assign debug info to the SDNode.
> +void SelectionDAG::AssignDbgInfo(SDNode* SD, SDDbgValue* db) {
> +  assert(SD && "Trying to assign dbg info to a null node!");
> +  DbgInfo->add(SD, db);
> +  SD->setHasDebugValue(true);
> +}
> +
> +/// RememberDbgInfo - Remember debug info which is not assigned to an SDNode.
> +void SelectionDAG::RememberDbgInfo(SDDbgValue* db) {
> +  DbgInfo->add(db);
> +}
> +
> +/// GetDbgInfo - Get the debug info, if any, for the SDNode.
> +SDDbgValue* SelectionDAG::GetDbgInfo(const SDNode *SD) {
> +  assert(SD && "Trying to get the order of a null node!");
> +  if (SD->getHasDebugValue())
> +    return DbgInfo->getSDDbgValue(SD);
> +  return 0;
> +}
> 
> //===----------------------------------------------------------------------===//
> //                              SDNode Class
> @@ -5911,7 +5938,7 @@
>   if (G)
>     if (unsigned Order = G->GetOrdering(this))
>       OS << " [ORD=" << Order << ']';
> -  
> +
>   if (getNodeId() != -1)
>     OS << " [ID=" << getNodeId() << ']';
> }
> 
> Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98195&r1=98194&r2=98195&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Wed Mar 10 16:13:47 2010
> @@ -8521,6 +8521,21 @@
>     F->DeleteMachineInstr(MI);   // The pseudo instruction is gone now.
>     return BB;
>   }
> +    // DBG_VALUE.  Only the frame index case is done here.
> +  case X86::DBG_VALUE: {
> +    const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
> +    DebugLoc DL = MI->getDebugLoc();
> +    X86AddressMode AM;
> +    MachineFunction *F = BB->getParent();
> +    AM.BaseType = X86AddressMode::FrameIndexBase;
> +    AM.Base.FrameIndex = MI->getOperand(0).getImm();
> +    addFullAddress(BuildMI(BB, DL, TII->get(X86::DBG_VALUE)), AM).
> +      addImm(MI->getOperand(1).getImm()).
> +      addMetadata(MI->getOperand(2).getMetadata());
> +    F->DeleteMachineInstr(MI);      // Remove pseudo.
> +    return BB;
> +  }
> +
>     // String/text processing lowering.
>   case X86::PCMPISTRM128REG:
>     return EmitPCMP(MI, BB, 3, false /* in-mem */);
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From dalej at apple.com  Wed Mar 10 17:37:24 2010
From: dalej at apple.com (Dale Johannesen)
Date: Wed, 10 Mar 2010 23:37:24 -0000
Subject: [llvm-commits] [llvm] r98202 - in
 /llvm/trunk/lib/CodeGen/SelectionDAG: InstrEmitter.cpp SDDbgValue.h
Message-ID: <20100310233724.3E3E52A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 17:37:24 2010
New Revision: 98202

URL: http://llvm.org/viewvc/llvm-project?rev=98202&view=rev
Log:
Cosmetic: lengthen names and improve comments.


Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp?rev=98202&r1=98201&r2=98202&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/InstrEmitter.cpp Wed Mar 10 17:37:24 2010
@@ -508,7 +508,7 @@
     return;
   if (!sd)
     return;
-  assert(sd->getKind() == SDDbgValue::SD);
+  assert(sd->getKind() == SDDbgValue::SDNODE);
   unsigned VReg = getVR(SDValue(sd->getSDNode(), sd->getResNo()), VRBaseMap);
   const TargetInstrDesc &II = TII->get(TargetOpcode::DBG_VALUE);
   DebugLoc DL = sd->getDebugLoc();
@@ -537,7 +537,7 @@
   SDDbgValue::DbgValueKind kind = sd->getKind();
   DebugLoc DL = sd->getDebugLoc();
   MachineInstr* MI;
-  if (kind == SDDbgValue::CNST) {
+  if (kind == SDDbgValue::CONST) {
     Value *V = sd->getConst();
     if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
       MI = BuildMI(*MF, DL, II).addImm(CI->getZExtValue()).
@@ -551,7 +551,7 @@
       MI = BuildMI(*MF, DL, II).addReg(0U).
                                        addImm(Offset).addMetadata(mdPtr);
     }
-  } else if (kind == SDDbgValue::FX) {
+  } else if (kind == SDDbgValue::FRAMEIX) {
     unsigned FrameIx = sd->getFrameIx();
     // Stack address; this needs to be lowered in target-dependent fashion.
     // FIXME test that the target supports this somehow; if not emit Undef.

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h?rev=98202&r1=98201&r2=98202&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SDDbgValue.h Wed Mar 10 17:37:24 2010
@@ -24,22 +24,21 @@
 class Value;
 
 /// SDDbgValue - Holds the information from a dbg_value node through SDISel.
-/// Either Const or Node is nonzero, but not both.
 /// We do not use SDValue here to avoid including its header.
 
 class SDDbgValue {
 public:
   enum DbgValueKind {
-    SD = 0,
-    CNST = 1,
-    FX = 2
+    SDNODE = 0,             // value is the result of an expression
+    CONST = 1,              // value is a constant
+    FRAMEIX = 2             // value is contents of a stack location
   };
 private:
   enum DbgValueKind kind;
   union {
     struct {
-      SDNode *Node;         // valid for non-constants
-      unsigned ResNo;       // valid for non-constants
+      SDNode *Node;         // valid for expressions
+      unsigned ResNo;       // valid for expressions
     } s;
     Value *Const;           // valid for constants
     unsigned FrameIx;       // valid for stack objects
@@ -52,7 +51,7 @@
   // Constructor for non-constants.
   SDDbgValue(MDNode *mdP, SDNode *N, unsigned R, uint64_t off, DebugLoc dl,
              unsigned O) : mdPtr(mdP), Offset(off), DL(dl), Order(O) {
-    kind = SD;
+    kind = SDNODE;
     u.s.Node = N;
     u.s.ResNo = R;
   }
@@ -60,14 +59,14 @@
   // Constructor for constants.
   SDDbgValue(MDNode *mdP, Value *C, uint64_t off, DebugLoc dl, unsigned O) : 
     mdPtr(mdP), Offset(off), DL(dl), Order(O) {
-    kind = CNST;
+    kind = CONST;
     u.Const = C;
   }
 
   // Constructor for frame indices.
   SDDbgValue(MDNode *mdP, unsigned FI, uint64_t off, DebugLoc dl, unsigned O) : 
     mdPtr(mdP), Offset(off), DL(dl), Order(O) {
-    kind = FX;
+    kind = FRAMEIX;
     u.FrameIx = FI;
   }
 
@@ -78,16 +77,16 @@
   MDNode *getMDPtr() { return mdPtr; }
 
   // Returns the SDNode* for a register ref
-  SDNode *getSDNode() { assert (kind==SD); return u.s.Node; }
+  SDNode *getSDNode() { assert (kind==SDNODE); return u.s.Node; }
 
   // Returns the ResNo for a register ref
-  unsigned getResNo() { assert (kind==SD); return u.s.ResNo; }
+  unsigned getResNo() { assert (kind==SDNODE); return u.s.ResNo; }
 
   // Returns the Value* for a constant
-  Value *getConst() { assert (kind==CNST); return u.Const; }
+  Value *getConst() { assert (kind==CONST); return u.Const; }
 
   // Returns the FrameIx for a stack object
-  unsigned getFrameIx() { assert (kind==FX); return u.FrameIx; }
+  unsigned getFrameIx() { assert (kind==FRAMEIX); return u.FrameIx; }
 
   // Returns the offset.
   uint64_t getOffset() { return Offset; }




From sabre at nondot.org  Wed Mar 10 17:46:44 2010
From: sabre at nondot.org (Chris Lattner)
Date: Wed, 10 Mar 2010 23:46:44 -0000
Subject: [llvm-commits] [llvm] r98203 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/DAGCombiner.cpp test/CodeGen/X86/crash.ll
Message-ID: <20100310234644.A643F2A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 17:46:44 2010
New Revision: 98203

URL: http://llvm.org/viewvc/llvm-project?rev=98203&view=rev
Log:
fix PR6533 by updating the br(xor) code to remember the case
when it looked past a trunc.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    llvm/trunk/test/CodeGen/X86/crash.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=98203&r1=98202&r2=98203&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Wed Mar 10 17:46:44 2010
@@ -4605,7 +4605,7 @@
 
   SDNode *Trunc = 0;
   if (N1.getOpcode() == ISD::TRUNCATE && N1.hasOneUse()) {
-    // Look pass truncate.
+    // Look past truncate.
     Trunc = N1.getNode();
     N1 = N1.getOperand(0);
   }
@@ -4700,7 +4700,9 @@
           Equal = true;
         }
 
-      EVT SetCCVT = N1.getValueType();
+      SDValue NodeToReplace = Trunc ? SDValue(Trunc, 0) : N1;
+      
+      EVT SetCCVT = NodeToReplace.getValueType();
       if (LegalTypes)
         SetCCVT = TLI.getSetCCResultType(SetCCVT);
       SDValue SetCC = DAG.getSetCC(TheXor->getDebugLoc(),
@@ -4709,9 +4711,9 @@
                                    Equal ? ISD::SETEQ : ISD::SETNE);
       // Replace the uses of XOR with SETCC
       WorkListRemover DeadNodes(*this);
-      DAG.ReplaceAllUsesOfValueWith(N1, SetCC, &DeadNodes);
-      removeFromWorkList(N1.getNode());
-      DAG.DeleteNode(N1.getNode());
+      DAG.ReplaceAllUsesOfValueWith(NodeToReplace, SetCC, &DeadNodes);
+      removeFromWorkList(NodeToReplace.getNode());
+      DAG.DeleteNode(NodeToReplace.getNode());
       return DAG.getNode(ISD::BRCOND, N->getDebugLoc(),
                          MVT::Other, Chain, SetCC, N2);
     }

Modified: llvm/trunk/test/CodeGen/X86/crash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/crash.ll?rev=98203&r1=98202&r2=98203&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/crash.ll (original)
+++ llvm/trunk/test/CodeGen/X86/crash.ll Wed Mar 10 17:46:44 2010
@@ -18,3 +18,18 @@
   volatile store i32 %conv19.i, i32* undef
   ret i32 undef
 }
+
+; PR6533
+define void @test2(i1 %x, i32 %y) nounwind {
+  %land.ext = zext i1 %x to i32                   ; <i32> [#uses=1]
+  %and = and i32 %y, 1                        ; <i32> [#uses=1]
+  %xor = xor i32 %and, %land.ext                  ; <i32> [#uses=1]
+  %cmp = icmp eq i32 %xor, 1                      ; <i1> [#uses=1]
+  br i1 %cmp, label %if.end, label %if.then
+
+if.then:                                          ; preds = %land.end
+  ret void
+
+if.end:                                           ; preds = %land.end
+  ret void
+}




From jyasskin at google.com  Wed Mar 10 17:58:24 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Wed, 10 Mar 2010 15:58:24 -0800
Subject: [llvm-commits] [llvm] r86748 - in /llvm/trunk:
	include/llvm/Analysis/ include/llvm/CodeGen/
	include/llvm/Transforms/Utils/ lib/Analysis/
	lib/CodeGen/AsmPrinter/ lib/CodeGen/SelectionDAG/
	lib/Transforms/Utils/
In-Reply-To: <7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>
	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
	<7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>
Message-ID: <5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>

On Mon, Mar 8, 2010 at 10:13 AM, Devang Patel <dpatel at apple.com> wrote:
> Hi Jeffrey,
>
> I'll update the comment. AbstractScopes map is used to keep track of abstract scopes in entire module. AbstractScopesList is used to construct abstract scopes seen while processing a function. This use of list avoids randomness in DIE ordering.

Thanks.  Is there a test for this ordering stability?


From jyasskin at google.com  Wed Mar 10 18:01:38 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Wed, 10 Mar 2010 16:01:38 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
	<2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
Message-ID: <5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>

On Tue, Mar 9, 2010 at 12:22 PM, Devang Patel <dpatel at apple.com> wrote:
>
> On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:
>
>> Now that nothing uses the result of constructCompileUnit() unless it
>> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
>> avoid new'ing the DIE and CompileUnit?
>
> With the introduction of DIFile, we are moving away from "isMain" field.
>
>> Then, if we delete ModuleCU in
>> endModule(), this memory leak would go away.
>
> We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.

I can take it. I assume it's correct to simply not create the
CompileUnits and DIEs whose addresses aren't obviously passed
anywhere? Or do I need to look out for constructors or other functions
that escape their addresses surreptitiously?


From sabre at nondot.org  Wed Mar 10 18:06:20 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 00:06:20 -0000
Subject: [llvm-commits] [llvm] r98205 -
	/llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp
Message-ID: <20100311000620.237082A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 18:06:19 2010
New Revision: 98205

URL: http://llvm.org/viewvc/llvm-project?rev=98205&view=rev
Log:
Work around a bug in the openbsd assembler on i386,
which doesn't support .quad correctly because it is
"really really old".  PR6528.

Yet another reason the mc assembler should take over ;-)


Modified:
    llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp

Modified: llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp?rev=98205&r1=98204&r2=98205&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp Wed Mar 10 18:06:19 2010
@@ -70,7 +70,7 @@
   ExceptionsType = ExceptionHandling::Dwarf;
 }
 
-X86ELFMCAsmInfo::X86ELFMCAsmInfo(const Triple &Triple) {
+X86ELFMCAsmInfo::X86ELFMCAsmInfo(const Triple &T) {
   AsmTransCBE = x86_asm_table;
   AssemblerDialect = AsmWriterFlavor;
 
@@ -89,6 +89,11 @@
 
   // Exceptions handling
   ExceptionsType = ExceptionHandling::Dwarf;
+  
+  // OpenBSD has buggy support for .quad in 32-bit mode, just split into two
+  // .words.
+  if (T.getOS() == Triple::OpenBSD && T.getArch() == Triple::x86)
+    Data64bitsDirective = 0;
 }
 
 MCSection *X86ELFMCAsmInfo::getNonexecutableStackSection(MCContext &Ctx) const {




From stoklund at 2pi.dk  Wed Mar 10 18:11:33 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 11 Mar 2010 00:11:33 -0000
Subject: [llvm-commits] [llvm] r98207 -
	/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
Message-ID: <20100311001133.D6A8A2A6C12C@llvm.org>

Author: stoklund
Date: Wed Mar 10 18:11:33 2010
New Revision: 98207

URL: http://llvm.org/viewvc/llvm-project?rev=98207&view=rev
Log:
VirtRegRewriter spring cleaning. No functional change.

Move methods out of line and M-x whitespace-cleanup.
Promote common method arguments to member variables.

Modified:
    llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp

Modified: llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp?rev=98207&r1=98206&r2=98207&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp Wed Mar 10 18:11:33 2010
@@ -98,7 +98,7 @@
   bool runOnMachineFunction(MachineFunction &MF, VirtRegMap &VRM,
                             LiveIntervals* LIs) {
     DEBUG(dbgs() << "********** REWRITE MACHINE CODE **********\n");
-    DEBUG(dbgs() << "********** Function: " 
+    DEBUG(dbgs() << "********** Function: "
           << MF.getFunction()->getName() << '\n');
     DEBUG(dbgs() << "**** Machine Instrs"
           << "(NOTE! Does not include spills and reloads!) ****\n");
@@ -135,10 +135,10 @@
         changed |= !reglist.empty();
       }
     }
-    
+
     DEBUG(dbgs() << "**** Post Machine Instrs ****\n");
     DEBUG(MF.dump());
-    
+
     return changed;
   }
 
@@ -208,7 +208,7 @@
   /// in the specified physreg.  If CanClobber is true, the physreg can be
   /// modified at any time without changing the semantics of the program.
   void addAvailable(int SlotOrReMat, unsigned Reg, bool CanClobber = true) {
-    // If this stack slot is thought to be available in some other physreg, 
+    // If this stack slot is thought to be available in some other physreg,
     // remove its record.
     ModifyStackSlotOrReMat(SlotOrReMat);
 
@@ -364,7 +364,7 @@
 
   // AssignedPhysReg - The physreg that was assigned for use by the reload.
   unsigned AssignedPhysReg;
-  
+
   // VirtReg - The virtual register itself.
   unsigned VirtReg;
 
@@ -384,11 +384,11 @@
   ReuseInfo(MachineInstr &mi, const TargetRegisterInfo *tri) : MI(mi) {
     PhysRegsClobbered.resize(tri->getNumRegs());
   }
-  
+
   bool hasReuses() const {
     return !Reuses.empty();
   }
-  
+
   /// addReuse - If we choose to reuse a virtual register that is already
   /// available instead of reloading it, remember that we did so.
   void addReuse(unsigned OpNo, unsigned StackSlotOrReMat,
@@ -397,9 +397,9 @@
     // If the reload is to the assigned register anyway, no undo will be
     // required.
     if (PhysRegReused == AssignedPhysReg) return;
-    
+
     // Otherwise, remember this.
-    Reuses.push_back(ReusedOp(OpNo, StackSlotOrReMat, PhysRegReused, 
+    Reuses.push_back(ReusedOp(OpNo, StackSlotOrReMat, PhysRegReused,
                               AssignedPhysReg, VirtReg));
   }
 
@@ -410,10 +410,10 @@
   bool isClobbered(unsigned PhysReg) const {
     return PhysRegsClobbered.test(PhysReg);
   }
-  
+
   /// GetRegForReload - We are about to emit a reload into PhysReg.  If there
   /// is some other operand that is using the specified register, either pick
-  /// a new register to use, or evict the previous reload and use this reg. 
+  /// a new register to use, or evict the previous reload and use this reg.
   unsigned GetRegForReload(const TargetRegisterClass *RC, unsigned PhysReg,
                            MachineFunction &MF, MachineInstr *MI,
                            AvailableSpills &Spills,
@@ -525,7 +525,7 @@
 /// reference.
 static bool InvalidateRegDef(MachineBasicBlock::iterator I,
                              MachineInstr &NewDef, unsigned Reg,
-                             bool &HasLiveDef, 
+                             bool &HasLiveDef,
                              const TargetRegisterInfo *TRI) {
   // Due to remat, it's possible this reg isn't being reused. That is,
   // the def of this reg (by prev MI) is now dead.
@@ -579,7 +579,7 @@
     unsigned Reg = MO.getReg();
     if (Reg == 0)
       continue;
-    
+
     if (RegKills[Reg] && KillOps[Reg]->getParent() != &MI) {
       // That can't be right. Register is killed but not re-defined and it's
       // being reused. Let's fix that.
@@ -597,7 +597,7 @@
       }
     } else {
       // Check for subreg kills as well.
-      // d4 = 
+      // d4 =
       // store d4, fi#0
       // ...
       //    = s8<kill>
@@ -802,7 +802,7 @@
   if (It == SpillSlotsOrReMatsAvailable.end()) return;
   unsigned Reg = It->second >> 1;
   SpillSlotsOrReMatsAvailable.erase(It);
-  
+
   // This register may hold the value of multiple stack slots, only remove this
   // stack slot from the set of values the register contains.
   std::multimap<unsigned, int>::iterator I = PhysRegsAvailable.lower_bound(Reg);
@@ -832,7 +832,7 @@
                          VirtRegMap &VRM) {
   const TargetInstrInfo* TII = MF.getTarget().getInstrInfo();
   const TargetRegisterInfo *TRI = Spills.getRegInfo();
-  
+
   if (Reuses.empty()) return PhysReg;  // This is most often empty.
 
   for (unsigned ro = 0, e = Reuses.size(); ro != e; ++ro) {
@@ -853,7 +853,7 @@
     } else {
       // Otherwise, we might also have a problem if a previously reused
       // value aliases the new register. If so, codegen the previous reload
-      // and use this one.          
+      // and use this one.
       unsigned PRRU = Op.PhysRegReused;
       if (TRI->regsOverlap(PRRU, PhysReg)) {
         // Okay, we found out that an alias of a reused register
@@ -900,13 +900,13 @@
         if (DoReMat) {
           ReMaterialize(*MBB, InsertLoc, NewPhysReg, NewOp.VirtReg, TII,
                         TRI, VRM);
-        } else { 
+        } else {
           TII->loadRegFromStackSlot(*MBB, InsertLoc, NewPhysReg,
                                     NewOp.StackSlotOrReMat, AliasRC);
           MachineInstr *LoadMI = prior(InsertLoc);
           VRM.addSpillSlotUse(NewOp.StackSlotOrReMat, LoadMI);
           // Any stores to this stack slot are not dead anymore.
-          MaybeDeadStores[NewOp.StackSlotOrReMat] = NULL;            
+          MaybeDeadStores[NewOp.StackSlotOrReMat] = NULL;
           ++NumLoads;
         }
         Spills.ClobberPhysReg(NewPhysReg);
@@ -919,10 +919,10 @@
         Spills.addAvailable(NewOp.StackSlotOrReMat, NewPhysReg);
         UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
         DEBUG(dbgs() << '\t' << *prior(InsertLoc));
-        
+
         DEBUG(dbgs() << "Reuse undone!\n");
         --NumReused;
-        
+
         // Finally, PhysReg is now available, go ahead and use it.
         return PhysReg;
       }
@@ -1037,1410 +1037,1447 @@
 }
 
 namespace {
-  struct RefSorter {
-    bool operator()(const std::pair<MachineInstr*, int> &A,
-                    const std::pair<MachineInstr*, int> &B) {
-      return A.second < B.second;
-    }
-  };
-}
+
+struct RefSorter {
+  bool operator()(const std::pair<MachineInstr*, int> &A,
+                  const std::pair<MachineInstr*, int> &B) {
+    return A.second < B.second;
+  }
+};
 
 // ***************************** //
 // Local Spiller Implementation  //
 // ***************************** //
 
-namespace {
-
 class LocalRewriter : public VirtRegRewriter {
-  MachineRegisterInfo *RegInfo;
+  MachineRegisterInfo *MRI;
   const TargetRegisterInfo *TRI;
   const TargetInstrInfo *TII;
+  VirtRegMap *VRM;
   BitVector AllocatableRegs;
   DenseMap<MachineInstr*, unsigned> DistanceMap;
-public:
-
-  bool runOnMachineFunction(MachineFunction &MF, VirtRegMap &VRM,
-                            LiveIntervals* LIs) {
-    RegInfo = &MF.getRegInfo(); 
-    TRI = MF.getTarget().getRegisterInfo();
-    TII = MF.getTarget().getInstrInfo();
-    AllocatableRegs = TRI->getAllocatableSet(MF);
-    DEBUG(dbgs() << "\n**** Local spiller rewriting function '"
-          << MF.getFunction()->getName() << "':\n");
-    DEBUG(dbgs() << "**** Machine Instrs (NOTE! Does not include spills and"
-                    " reloads!) ****\n");
-    DEBUG(MF.dump());
-
-    // Spills - Keep track of which spilled values are available in physregs
-    // so that we can choose to reuse the physregs instead of emitting
-    // reloads. This is usually refreshed per basic block.
-    AvailableSpills Spills(TRI, TII);
-
-    // Keep track of kill information.
-    BitVector RegKills(TRI->getNumRegs());
-    std::vector<MachineOperand*> KillOps;
-    KillOps.resize(TRI->getNumRegs(), NULL);
-
-    // SingleEntrySuccs - Successor blocks which have a single predecessor.
-    SmallVector<MachineBasicBlock*, 4> SinglePredSuccs;
-    SmallPtrSet<MachineBasicBlock*,16> EarlyVisited;
-
-    // Traverse the basic blocks depth first.
-    MachineBasicBlock *Entry = MF.begin();
-    SmallPtrSet<MachineBasicBlock*,16> Visited;
-    for (df_ext_iterator<MachineBasicBlock*,
-           SmallPtrSet<MachineBasicBlock*,16> >
-           DFI = df_ext_begin(Entry, Visited), E = df_ext_end(Entry, Visited);
-         DFI != E; ++DFI) {
-      MachineBasicBlock *MBB = *DFI;
-      if (!EarlyVisited.count(MBB))
-        RewriteMBB(*MBB, VRM, LIs, Spills, RegKills, KillOps);
-
-      // If this MBB is the only predecessor of a successor. Keep the
-      // availability information and visit it next.
-      do {
-        // Keep visiting single predecessor successor as long as possible.
-        SinglePredSuccs.clear();
-        findSinglePredSuccessor(MBB, SinglePredSuccs);
-        if (SinglePredSuccs.empty())
-          MBB = 0;
-        else {
-          // FIXME: More than one successors, each of which has MBB has
-          // the only predecessor.
-          MBB = SinglePredSuccs[0];
-          if (!Visited.count(MBB) && EarlyVisited.insert(MBB)) {
-            Spills.AddAvailableRegsToLiveIn(*MBB, RegKills, KillOps);
-            RewriteMBB(*MBB, VRM, LIs, Spills, RegKills, KillOps);
-          }
-        }
-      } while (MBB);
 
-      // Clear the availability info.
-      Spills.clear();
-    }
+  MachineBasicBlock *MBB;       // Basic block currently being processed.
 
-    DEBUG(dbgs() << "**** Post Machine Instrs ****\n");
-    DEBUG(MF.dump());
-
-    // Mark unused spill slots.
-    MachineFrameInfo *MFI = MF.getFrameInfo();
-    int SS = VRM.getLowSpillSlot();
-    if (SS != VirtRegMap::NO_STACK_SLOT)
-      for (int e = VRM.getHighSpillSlot(); SS <= e; ++SS)
-        if (!VRM.isSpillSlotUsed(SS)) {
-          MFI->RemoveStackObject(SS);
-          ++NumDSS;
-        }
+public:
 
-    return true;
-  }
+  bool runOnMachineFunction(MachineFunction &MF, VirtRegMap &VRM,
+                            LiveIntervals* LIs);
 
 private:
 
-  /// OptimizeByUnfold2 - Unfold a series of load / store folding instructions if
-  /// a scratch register is available.
-  ///     xorq  %r12<kill>, %r13
-  ///     addq  %rax, -184(%rbp)
-  ///     addq  %r13, -184(%rbp)
-  /// ==>
-  ///     xorq  %r12<kill>, %r13
-  ///     movq  -184(%rbp), %r12
-  ///     addq  %rax, %r12
-  ///     addq  %r13, %r12
-  ///     movq  %r12, -184(%rbp)
   bool OptimizeByUnfold2(unsigned VirtReg, int SS,
-                         MachineBasicBlock &MBB,
                          MachineBasicBlock::iterator &MII,
                          std::vector<MachineInstr*> &MaybeDeadStores,
                          AvailableSpills &Spills,
                          BitVector &RegKills,
-                         std::vector<MachineOperand*> &KillOps,
-                         VirtRegMap &VRM) {
+                         std::vector<MachineOperand*> &KillOps);
 
-    MachineBasicBlock::iterator NextMII = llvm::next(MII);
-    if (NextMII == MBB.end())
-      return false;
+  bool OptimizeByUnfold(MachineBasicBlock::iterator &MII,
+                        std::vector<MachineInstr*> &MaybeDeadStores,
+                        AvailableSpills &Spills,
+                        BitVector &RegKills,
+                        std::vector<MachineOperand*> &KillOps);
 
-    if (TII->getOpcodeAfterMemoryUnfold(MII->getOpcode(), true, true) == 0)
-      return false;
+  bool CommuteToFoldReload(MachineBasicBlock::iterator &MII,
+                           unsigned VirtReg, unsigned SrcReg, int SS,
+                           AvailableSpills &Spills,
+                           BitVector &RegKills,
+                           std::vector<MachineOperand*> &KillOps,
+                           const TargetRegisterInfo *TRI);
 
-    // Now let's see if the last couple of instructions happens to have freed up
-    // a register.
-    const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-    unsigned PhysReg = FindFreeRegister(MII, MBB, RC, TRI, AllocatableRegs);
-    if (!PhysReg)
-      return false;
+  void SpillRegToStackSlot(MachineBasicBlock::iterator &MII,
+                           int Idx, unsigned PhysReg, int StackSlot,
+                           const TargetRegisterClass *RC,
+                           bool isAvailable, MachineInstr *&LastStore,
+                           AvailableSpills &Spills,
+                           SmallSet<MachineInstr*, 4> &ReMatDefs,
+                           BitVector &RegKills,
+                           std::vector<MachineOperand*> &KillOps);
 
-    MachineFunction &MF = *MBB.getParent();
-    TRI = MF.getTarget().getRegisterInfo();
-    MachineInstr &MI = *MII;
-    if (!FoldsStackSlotModRef(MI, SS, PhysReg, TII, TRI, VRM))
-      return false;
+  void TransferDeadness(unsigned CurDist,
+                        unsigned Reg, BitVector &RegKills,
+                        std::vector<MachineOperand*> &KillOps);
 
-    // If the next instruction also folds the same SS modref and can be unfoled,
-    // then it's worthwhile to issue a load from SS into the free register and
-    // then unfold these instructions.
-    if (!FoldsStackSlotModRef(*NextMII, SS, PhysReg, TII, TRI, VRM))
-      return false;
+  void RewriteMBB(LiveIntervals *LIs,
+                  AvailableSpills &Spills, BitVector &RegKills,
+                  std::vector<MachineOperand*> &KillOps);
+};
+}
 
-    // Back-schedule reloads and remats.
-    ComputeReloadLoc(MII, MBB.begin(), PhysReg, TRI, false, SS, TII, MF);
+bool LocalRewriter::runOnMachineFunction(MachineFunction &MF, VirtRegMap &vrm,
+                                         LiveIntervals* LIs) {
+  MRI = &MF.getRegInfo();
+  TRI = MF.getTarget().getRegisterInfo();
+  TII = MF.getTarget().getInstrInfo();
+  VRM = &vrm;
+  AllocatableRegs = TRI->getAllocatableSet(MF);
+  DEBUG(dbgs() << "\n**** Local spiller rewriting function '"
+        << MF.getFunction()->getName() << "':\n");
+  DEBUG(dbgs() << "**** Machine Instrs (NOTE! Does not include spills and"
+        " reloads!) ****\n");
+  DEBUG(MF.dump());
+
+  // Spills - Keep track of which spilled values are available in physregs
+  // so that we can choose to reuse the physregs instead of emitting
+  // reloads. This is usually refreshed per basic block.
+  AvailableSpills Spills(TRI, TII);
+
+  // Keep track of kill information.
+  BitVector RegKills(TRI->getNumRegs());
+  std::vector<MachineOperand*> KillOps;
+  KillOps.resize(TRI->getNumRegs(), NULL);
+
+  // SingleEntrySuccs - Successor blocks which have a single predecessor.
+  SmallVector<MachineBasicBlock*, 4> SinglePredSuccs;
+  SmallPtrSet<MachineBasicBlock*,16> EarlyVisited;
+
+  // Traverse the basic blocks depth first.
+  MachineBasicBlock *Entry = MF.begin();
+  SmallPtrSet<MachineBasicBlock*,16> Visited;
+  for (df_ext_iterator<MachineBasicBlock*,
+         SmallPtrSet<MachineBasicBlock*,16> >
+         DFI = df_ext_begin(Entry, Visited), E = df_ext_end(Entry, Visited);
+       DFI != E; ++DFI) {
+    MBB = *DFI;
+    if (!EarlyVisited.count(MBB))
+      RewriteMBB(LIs, Spills, RegKills, KillOps);
 
-    // Load from SS to the spare physical register.
-    TII->loadRegFromStackSlot(MBB, MII, PhysReg, SS, RC);
-    // This invalidates Phys.
-    Spills.ClobberPhysReg(PhysReg);
-    // Remember it's available.
-    Spills.addAvailable(SS, PhysReg);
-    MaybeDeadStores[SS] = NULL;
+    // If this MBB is the only predecessor of a successor. Keep the
+    // availability information and visit it next.
+    do {
+      // Keep visiting single predecessor successor as long as possible.
+      SinglePredSuccs.clear();
+      findSinglePredSuccessor(MBB, SinglePredSuccs);
+      if (SinglePredSuccs.empty())
+        MBB = 0;
+      else {
+        // FIXME: More than one successors, each of which has MBB has
+        // the only predecessor.
+        MBB = SinglePredSuccs[0];
+        if (!Visited.count(MBB) && EarlyVisited.insert(MBB)) {
+          Spills.AddAvailableRegsToLiveIn(*MBB, RegKills, KillOps);
+          RewriteMBB(LIs, Spills, RegKills, KillOps);
+        }
+      }
+    } while (MBB);
+
+    // Clear the availability info.
+    Spills.clear();
+  }
+
+  DEBUG(dbgs() << "**** Post Machine Instrs ****\n");
+  DEBUG(MF.dump());
+
+  // Mark unused spill slots.
+  MachineFrameInfo *MFI = MF.getFrameInfo();
+  int SS = VRM->getLowSpillSlot();
+  if (SS != VirtRegMap::NO_STACK_SLOT)
+    for (int e = VRM->getHighSpillSlot(); SS <= e; ++SS)
+      if (!VRM->isSpillSlotUsed(SS)) {
+        MFI->RemoveStackObject(SS);
+        ++NumDSS;
+      }
 
-    // Unfold current MI.
-    SmallVector<MachineInstr*, 4> NewMIs;
-    if (!TII->unfoldMemoryOperand(MF, &MI, VirtReg, false, false, NewMIs))
+  return true;
+}
+
+/// OptimizeByUnfold2 - Unfold a series of load / store folding instructions if
+/// a scratch register is available.
+///     xorq  %r12<kill>, %r13
+///     addq  %rax, -184(%rbp)
+///     addq  %r13, -184(%rbp)
+/// ==>
+///     xorq  %r12<kill>, %r13
+///     movq  -184(%rbp), %r12
+///     addq  %rax, %r12
+///     addq  %r13, %r12
+///     movq  %r12, -184(%rbp)
+bool LocalRewriter::
+OptimizeByUnfold2(unsigned VirtReg, int SS,
+                  MachineBasicBlock::iterator &MII,
+                  std::vector<MachineInstr*> &MaybeDeadStores,
+                  AvailableSpills &Spills,
+                  BitVector &RegKills,
+                  std::vector<MachineOperand*> &KillOps) {
+
+  MachineBasicBlock::iterator NextMII = llvm::next(MII);
+  if (NextMII == MBB->end())
+    return false;
+
+  if (TII->getOpcodeAfterMemoryUnfold(MII->getOpcode(), true, true) == 0)
+    return false;
+
+  // Now let's see if the last couple of instructions happens to have freed up
+  // a register.
+  const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+  unsigned PhysReg = FindFreeRegister(MII, *MBB, RC, TRI, AllocatableRegs);
+  if (!PhysReg)
+    return false;
+
+  MachineFunction &MF = *MBB->getParent();
+  TRI = MF.getTarget().getRegisterInfo();
+  MachineInstr &MI = *MII;
+  if (!FoldsStackSlotModRef(MI, SS, PhysReg, TII, TRI, *VRM))
+    return false;
+
+  // If the next instruction also folds the same SS modref and can be unfoled,
+  // then it's worthwhile to issue a load from SS into the free register and
+  // then unfold these instructions.
+  if (!FoldsStackSlotModRef(*NextMII, SS, PhysReg, TII, TRI, *VRM))
+    return false;
+
+  // Back-schedule reloads and remats.
+  ComputeReloadLoc(MII, MBB->begin(), PhysReg, TRI, false, SS, TII, MF);
+
+  // Load from SS to the spare physical register.
+  TII->loadRegFromStackSlot(*MBB, MII, PhysReg, SS, RC);
+  // This invalidates Phys.
+  Spills.ClobberPhysReg(PhysReg);
+  // Remember it's available.
+  Spills.addAvailable(SS, PhysReg);
+  MaybeDeadStores[SS] = NULL;
+
+  // Unfold current MI.
+  SmallVector<MachineInstr*, 4> NewMIs;
+  if (!TII->unfoldMemoryOperand(MF, &MI, VirtReg, false, false, NewMIs))
+    llvm_unreachable("Unable unfold the load / store folding instruction!");
+  assert(NewMIs.size() == 1);
+  AssignPhysToVirtReg(NewMIs[0], VirtReg, PhysReg, *TRI);
+  VRM->transferRestorePts(&MI, NewMIs[0]);
+  MII = MBB->insert(MII, NewMIs[0]);
+  InvalidateKills(MI, TRI, RegKills, KillOps);
+  VRM->RemoveMachineInstrFromMaps(&MI);
+  MBB->erase(&MI);
+  ++NumModRefUnfold;
+
+  // Unfold next instructions that fold the same SS.
+  do {
+    MachineInstr &NextMI = *NextMII;
+    NextMII = llvm::next(NextMII);
+    NewMIs.clear();
+    if (!TII->unfoldMemoryOperand(MF, &NextMI, VirtReg, false, false, NewMIs))
       llvm_unreachable("Unable unfold the load / store folding instruction!");
     assert(NewMIs.size() == 1);
     AssignPhysToVirtReg(NewMIs[0], VirtReg, PhysReg, *TRI);
-    VRM.transferRestorePts(&MI, NewMIs[0]);
-    MII = MBB.insert(MII, NewMIs[0]);
-    InvalidateKills(MI, TRI, RegKills, KillOps);
-    VRM.RemoveMachineInstrFromMaps(&MI);
-    MBB.erase(&MI);
+    VRM->transferRestorePts(&NextMI, NewMIs[0]);
+    MBB->insert(NextMII, NewMIs[0]);
+    InvalidateKills(NextMI, TRI, RegKills, KillOps);
+    VRM->RemoveMachineInstrFromMaps(&NextMI);
+    MBB->erase(&NextMI);
     ++NumModRefUnfold;
+    if (NextMII == MBB->end())
+      break;
+  } while (FoldsStackSlotModRef(*NextMII, SS, PhysReg, TII, TRI, *VRM));
 
-    // Unfold next instructions that fold the same SS.
-    do {
-      MachineInstr &NextMI = *NextMII;
-      NextMII = llvm::next(NextMII);
-      NewMIs.clear();
-      if (!TII->unfoldMemoryOperand(MF, &NextMI, VirtReg, false, false, NewMIs))
-        llvm_unreachable("Unable unfold the load / store folding instruction!");
-      assert(NewMIs.size() == 1);
-      AssignPhysToVirtReg(NewMIs[0], VirtReg, PhysReg, *TRI);
-      VRM.transferRestorePts(&NextMI, NewMIs[0]);
-      MBB.insert(NextMII, NewMIs[0]);
-      InvalidateKills(NextMI, TRI, RegKills, KillOps);
-      VRM.RemoveMachineInstrFromMaps(&NextMI);
-      MBB.erase(&NextMI);
-      ++NumModRefUnfold;
-      if (NextMII == MBB.end())
-        break;
-    } while (FoldsStackSlotModRef(*NextMII, SS, PhysReg, TII, TRI, VRM));
-
-    // Store the value back into SS.
-    TII->storeRegToStackSlot(MBB, NextMII, PhysReg, true, SS, RC);
-    MachineInstr *StoreMI = prior(NextMII);
-    VRM.addSpillSlotUse(SS, StoreMI);
-    VRM.virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
+  // Store the value back into SS.
+  TII->storeRegToStackSlot(*MBB, NextMII, PhysReg, true, SS, RC);
+  MachineInstr *StoreMI = prior(NextMII);
+  VRM->addSpillSlotUse(SS, StoreMI);
+  VRM->virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
 
-    return true;
-  }
+  return true;
+}
 
-  /// OptimizeByUnfold - Turn a store folding instruction into a load folding
-  /// instruction. e.g.
-  ///     xorl  %edi, %eax
-  ///     movl  %eax, -32(%ebp)
-  ///     movl  -36(%ebp), %eax
-  ///     orl   %eax, -32(%ebp)
-  /// ==>
-  ///     xorl  %edi, %eax
-  ///     orl   -36(%ebp), %eax
-  ///     mov   %eax, -32(%ebp)
-  /// This enables unfolding optimization for a subsequent instruction which will
-  /// also eliminate the newly introduced store instruction.
-  bool OptimizeByUnfold(MachineBasicBlock &MBB,
-                        MachineBasicBlock::iterator &MII,
-                        std::vector<MachineInstr*> &MaybeDeadStores,
-                        AvailableSpills &Spills,
-                        BitVector &RegKills,
-                        std::vector<MachineOperand*> &KillOps,
-                        VirtRegMap &VRM) {
-    MachineFunction &MF = *MBB.getParent();
-    MachineInstr &MI = *MII;
-    unsigned UnfoldedOpc = 0;
-    unsigned UnfoldPR = 0;
-    unsigned UnfoldVR = 0;
-    int FoldedSS = VirtRegMap::NO_STACK_SLOT;
-    VirtRegMap::MI2VirtMapTy::const_iterator I, End;
-    for (tie(I, End) = VRM.getFoldedVirts(&MI); I != End; ) {
-      // Only transform a MI that folds a single register.
-      if (UnfoldedOpc)
-        return false;
-      UnfoldVR = I->second.first;
-      VirtRegMap::ModRef MR = I->second.second;
-      // MI2VirtMap be can updated which invalidate the iterator.
-      // Increment the iterator first.
-      ++I; 
-      if (VRM.isAssignedReg(UnfoldVR))
+/// OptimizeByUnfold - Turn a store folding instruction into a load folding
+/// instruction. e.g.
+///     xorl  %edi, %eax
+///     movl  %eax, -32(%ebp)
+///     movl  -36(%ebp), %eax
+///     orl   %eax, -32(%ebp)
+/// ==>
+///     xorl  %edi, %eax
+///     orl   -36(%ebp), %eax
+///     mov   %eax, -32(%ebp)
+/// This enables unfolding optimization for a subsequent instruction which will
+/// also eliminate the newly introduced store instruction.
+bool LocalRewriter::
+OptimizeByUnfold(MachineBasicBlock::iterator &MII,
+                 std::vector<MachineInstr*> &MaybeDeadStores,
+                 AvailableSpills &Spills,
+                 BitVector &RegKills,
+                 std::vector<MachineOperand*> &KillOps) {
+  MachineFunction &MF = *MBB->getParent();
+  MachineInstr &MI = *MII;
+  unsigned UnfoldedOpc = 0;
+  unsigned UnfoldPR = 0;
+  unsigned UnfoldVR = 0;
+  int FoldedSS = VirtRegMap::NO_STACK_SLOT;
+  VirtRegMap::MI2VirtMapTy::const_iterator I, End;
+  for (tie(I, End) = VRM->getFoldedVirts(&MI); I != End; ) {
+    // Only transform a MI that folds a single register.
+    if (UnfoldedOpc)
+      return false;
+    UnfoldVR = I->second.first;
+    VirtRegMap::ModRef MR = I->second.second;
+    // MI2VirtMap be can updated which invalidate the iterator.
+    // Increment the iterator first.
+    ++I;
+    if (VRM->isAssignedReg(UnfoldVR))
+      continue;
+    // If this reference is not a use, any previous store is now dead.
+    // Otherwise, the store to this stack slot is not dead anymore.
+    FoldedSS = VRM->getStackSlot(UnfoldVR);
+    MachineInstr* DeadStore = MaybeDeadStores[FoldedSS];
+    if (DeadStore && (MR & VirtRegMap::isModRef)) {
+      unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(FoldedSS);
+      if (!PhysReg || !DeadStore->readsRegister(PhysReg))
         continue;
-      // If this reference is not a use, any previous store is now dead.
-      // Otherwise, the store to this stack slot is not dead anymore.
-      FoldedSS = VRM.getStackSlot(UnfoldVR);
-      MachineInstr* DeadStore = MaybeDeadStores[FoldedSS];
-      if (DeadStore && (MR & VirtRegMap::isModRef)) {
-        unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(FoldedSS);
-        if (!PhysReg || !DeadStore->readsRegister(PhysReg))
-          continue;
-        UnfoldPR = PhysReg;
-        UnfoldedOpc = TII->getOpcodeAfterMemoryUnfold(MI.getOpcode(),
-                                                      false, true);
-      }
+      UnfoldPR = PhysReg;
+      UnfoldedOpc = TII->getOpcodeAfterMemoryUnfold(MI.getOpcode(),
+                                                    false, true);
     }
+  }
 
-    if (!UnfoldedOpc) {
-      if (!UnfoldVR)
-        return false;
+  if (!UnfoldedOpc) {
+    if (!UnfoldVR)
+      return false;
 
-      // Look for other unfolding opportunities.
-      return OptimizeByUnfold2(UnfoldVR, FoldedSS, MBB, MII,
-                               MaybeDeadStores, Spills, RegKills, KillOps, VRM);
-    }
+    // Look for other unfolding opportunities.
+    return OptimizeByUnfold2(UnfoldVR, FoldedSS, MII, MaybeDeadStores, Spills,
+                             RegKills, KillOps);
+  }
 
-    for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
-      MachineOperand &MO = MI.getOperand(i);
-      if (!MO.isReg() || MO.getReg() == 0 || !MO.isUse())
-        continue;
-      unsigned VirtReg = MO.getReg();
-      if (TargetRegisterInfo::isPhysicalRegister(VirtReg) || MO.getSubReg())
-        continue;
-      if (VRM.isAssignedReg(VirtReg)) {
-        unsigned PhysReg = VRM.getPhys(VirtReg);
-        if (PhysReg && TRI->regsOverlap(PhysReg, UnfoldPR))
-          return false;
-      } else if (VRM.isReMaterialized(VirtReg))
-        continue;
-      int SS = VRM.getStackSlot(VirtReg);
-      unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
-      if (PhysReg) {
-        if (TRI->regsOverlap(PhysReg, UnfoldPR))
-          return false;
+  for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+    MachineOperand &MO = MI.getOperand(i);
+    if (!MO.isReg() || MO.getReg() == 0 || !MO.isUse())
+      continue;
+    unsigned VirtReg = MO.getReg();
+    if (TargetRegisterInfo::isPhysicalRegister(VirtReg) || MO.getSubReg())
+      continue;
+    if (VRM->isAssignedReg(VirtReg)) {
+      unsigned PhysReg = VRM->getPhys(VirtReg);
+      if (PhysReg && TRI->regsOverlap(PhysReg, UnfoldPR))
+        return false;
+    } else if (VRM->isReMaterialized(VirtReg))
+      continue;
+    int SS = VRM->getStackSlot(VirtReg);
+    unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
+    if (PhysReg) {
+      if (TRI->regsOverlap(PhysReg, UnfoldPR))
+        return false;
+      continue;
+    }
+    if (VRM->hasPhys(VirtReg)) {
+      PhysReg = VRM->getPhys(VirtReg);
+      if (!TRI->regsOverlap(PhysReg, UnfoldPR))
         continue;
-      }
-      if (VRM.hasPhys(VirtReg)) {
-        PhysReg = VRM.getPhys(VirtReg);
-        if (!TRI->regsOverlap(PhysReg, UnfoldPR))
-          continue;
-      }
+    }
 
-      // Ok, we'll need to reload the value into a register which makes
-      // it impossible to perform the store unfolding optimization later.
-      // Let's see if it is possible to fold the load if the store is
-      // unfolded. This allows us to perform the store unfolding
-      // optimization.
-      SmallVector<MachineInstr*, 4> NewMIs;
-      if (TII->unfoldMemoryOperand(MF, &MI, UnfoldVR, false, false, NewMIs)) {
-        assert(NewMIs.size() == 1);
-        MachineInstr *NewMI = NewMIs.back();
-        NewMIs.clear();
-        int Idx = NewMI->findRegisterUseOperandIdx(VirtReg, false);
-        assert(Idx != -1);
-        SmallVector<unsigned, 1> Ops;
-        Ops.push_back(Idx);
-        MachineInstr *FoldedMI = TII->foldMemoryOperand(MF, NewMI, Ops, SS);
-        if (FoldedMI) {
-          VRM.addSpillSlotUse(SS, FoldedMI);
-          if (!VRM.hasPhys(UnfoldVR))
-            VRM.assignVirt2Phys(UnfoldVR, UnfoldPR);
-          VRM.virtFolded(VirtReg, FoldedMI, VirtRegMap::isRef);
-          MII = MBB.insert(MII, FoldedMI);
-          InvalidateKills(MI, TRI, RegKills, KillOps);
-          VRM.RemoveMachineInstrFromMaps(&MI);
-          MBB.erase(&MI);
-          MF.DeleteMachineInstr(NewMI);
-          return true;
-        }
+    // Ok, we'll need to reload the value into a register which makes
+    // it impossible to perform the store unfolding optimization later.
+    // Let's see if it is possible to fold the load if the store is
+    // unfolded. This allows us to perform the store unfolding
+    // optimization.
+    SmallVector<MachineInstr*, 4> NewMIs;
+    if (TII->unfoldMemoryOperand(MF, &MI, UnfoldVR, false, false, NewMIs)) {
+      assert(NewMIs.size() == 1);
+      MachineInstr *NewMI = NewMIs.back();
+      NewMIs.clear();
+      int Idx = NewMI->findRegisterUseOperandIdx(VirtReg, false);
+      assert(Idx != -1);
+      SmallVector<unsigned, 1> Ops;
+      Ops.push_back(Idx);
+      MachineInstr *FoldedMI = TII->foldMemoryOperand(MF, NewMI, Ops, SS);
+      if (FoldedMI) {
+        VRM->addSpillSlotUse(SS, FoldedMI);
+        if (!VRM->hasPhys(UnfoldVR))
+          VRM->assignVirt2Phys(UnfoldVR, UnfoldPR);
+        VRM->virtFolded(VirtReg, FoldedMI, VirtRegMap::isRef);
+        MII = MBB->insert(MII, FoldedMI);
+        InvalidateKills(MI, TRI, RegKills, KillOps);
+        VRM->RemoveMachineInstrFromMaps(&MI);
+        MBB->erase(&MI);
         MF.DeleteMachineInstr(NewMI);
+        return true;
       }
+      MF.DeleteMachineInstr(NewMI);
     }
+  }
+
+  return false;
+}
 
+/// CommuteChangesDestination - We are looking for r0 = op r1, r2 and
+/// where SrcReg is r1 and it is tied to r0. Return true if after
+/// commuting this instruction it will be r0 = op r2, r1.
+static bool CommuteChangesDestination(MachineInstr *DefMI,
+                                      const TargetInstrDesc &TID,
+                                      unsigned SrcReg,
+                                      const TargetInstrInfo *TII,
+                                      unsigned &DstIdx) {
+  if (TID.getNumDefs() != 1 && TID.getNumOperands() != 3)
+    return false;
+  if (!DefMI->getOperand(1).isReg() ||
+      DefMI->getOperand(1).getReg() != SrcReg)
     return false;
+  unsigned DefIdx;
+  if (!DefMI->isRegTiedToDefOperand(1, &DefIdx) || DefIdx != 0)
+    return false;
+  unsigned SrcIdx1, SrcIdx2;
+  if (!TII->findCommutedOpIndices(DefMI, SrcIdx1, SrcIdx2))
+    return false;
+  if (SrcIdx1 == 1 && SrcIdx2 == 2) {
+    DstIdx = 2;
+    return true;
   }
+  return false;
+}
 
-  /// CommuteChangesDestination - We are looking for r0 = op r1, r2 and
-  /// where SrcReg is r1 and it is tied to r0. Return true if after
-  /// commuting this instruction it will be r0 = op r2, r1.
-  static bool CommuteChangesDestination(MachineInstr *DefMI,
-                                        const TargetInstrDesc &TID,
-                                        unsigned SrcReg,
-                                        const TargetInstrInfo *TII,
-                                        unsigned &DstIdx) {
-    if (TID.getNumDefs() != 1 && TID.getNumOperands() != 3)
+/// CommuteToFoldReload -
+/// Look for
+/// r1 = load fi#1
+/// r1 = op r1, r2<kill>
+/// store r1, fi#1
+///
+/// If op is commutable and r2 is killed, then we can xform these to
+/// r2 = op r2, fi#1
+/// store r2, fi#1
+bool LocalRewriter::
+CommuteToFoldReload(MachineBasicBlock::iterator &MII,
+                    unsigned VirtReg, unsigned SrcReg, int SS,
+                    AvailableSpills &Spills,
+                    BitVector &RegKills,
+                    std::vector<MachineOperand*> &KillOps,
+                    const TargetRegisterInfo *TRI) {
+  if (MII == MBB->begin() || !MII->killsRegister(SrcReg))
+    return false;
+
+  MachineFunction &MF = *MBB->getParent();
+  MachineInstr &MI = *MII;
+  MachineBasicBlock::iterator DefMII = prior(MII);
+  MachineInstr *DefMI = DefMII;
+  const TargetInstrDesc &TID = DefMI->getDesc();
+  unsigned NewDstIdx;
+  if (DefMII != MBB->begin() &&
+      TID.isCommutable() &&
+      CommuteChangesDestination(DefMI, TID, SrcReg, TII, NewDstIdx)) {
+    MachineOperand &NewDstMO = DefMI->getOperand(NewDstIdx);
+    unsigned NewReg = NewDstMO.getReg();
+    if (!NewDstMO.isKill() || TRI->regsOverlap(NewReg, SrcReg))
       return false;
-    if (!DefMI->getOperand(1).isReg() ||
-        DefMI->getOperand(1).getReg() != SrcReg)
+    MachineInstr *ReloadMI = prior(DefMII);
+    int FrameIdx;
+    unsigned DestReg = TII->isLoadFromStackSlot(ReloadMI, FrameIdx);
+    if (DestReg != SrcReg || FrameIdx != SS)
       return false;
-    unsigned DefIdx;
-    if (!DefMI->isRegTiedToDefOperand(1, &DefIdx) || DefIdx != 0)
+    int UseIdx = DefMI->findRegisterUseOperandIdx(DestReg, false);
+    if (UseIdx == -1)
       return false;
-    unsigned SrcIdx1, SrcIdx2;
-    if (!TII->findCommutedOpIndices(DefMI, SrcIdx1, SrcIdx2))
+    unsigned DefIdx;
+    if (!MI.isRegTiedToDefOperand(UseIdx, &DefIdx))
       return false;
-    if (SrcIdx1 == 1 && SrcIdx2 == 2) {
-      DstIdx = 2;
-      return true;
-    }
-    return false;
-  }
+    assert(DefMI->getOperand(DefIdx).isReg() &&
+           DefMI->getOperand(DefIdx).getReg() == SrcReg);
 
-  /// CommuteToFoldReload -
-  /// Look for
-  /// r1 = load fi#1
-  /// r1 = op r1, r2<kill>
-  /// store r1, fi#1
-  ///
-  /// If op is commutable and r2 is killed, then we can xform these to
-  /// r2 = op r2, fi#1
-  /// store r2, fi#1
-  bool CommuteToFoldReload(MachineBasicBlock &MBB,
-                           MachineBasicBlock::iterator &MII,
-                           unsigned VirtReg, unsigned SrcReg, int SS,
-                           AvailableSpills &Spills,
-                           BitVector &RegKills,
-                           std::vector<MachineOperand*> &KillOps,
-                           const TargetRegisterInfo *TRI,
-                           VirtRegMap &VRM) {
-    if (MII == MBB.begin() || !MII->killsRegister(SrcReg))
+    // Now commute def instruction.
+    MachineInstr *CommutedMI = TII->commuteInstruction(DefMI, true);
+    if (!CommutedMI)
+      return false;
+    SmallVector<unsigned, 1> Ops;
+    Ops.push_back(NewDstIdx);
+    MachineInstr *FoldedMI = TII->foldMemoryOperand(MF, CommutedMI, Ops, SS);
+    // Not needed since foldMemoryOperand returns new MI.
+    MF.DeleteMachineInstr(CommutedMI);
+    if (!FoldedMI)
       return false;
 
-    MachineFunction &MF = *MBB.getParent();
-    MachineInstr &MI = *MII;
-    MachineBasicBlock::iterator DefMII = prior(MII);
-    MachineInstr *DefMI = DefMII;
-    const TargetInstrDesc &TID = DefMI->getDesc();
-    unsigned NewDstIdx;
-    if (DefMII != MBB.begin() &&
-        TID.isCommutable() &&
-        CommuteChangesDestination(DefMI, TID, SrcReg, TII, NewDstIdx)) {
-      MachineOperand &NewDstMO = DefMI->getOperand(NewDstIdx);
-      unsigned NewReg = NewDstMO.getReg();
-      if (!NewDstMO.isKill() || TRI->regsOverlap(NewReg, SrcReg))
-        return false;
-      MachineInstr *ReloadMI = prior(DefMII);
-      int FrameIdx;
-      unsigned DestReg = TII->isLoadFromStackSlot(ReloadMI, FrameIdx);
-      if (DestReg != SrcReg || FrameIdx != SS)
-        return false;
-      int UseIdx = DefMI->findRegisterUseOperandIdx(DestReg, false);
-      if (UseIdx == -1)
-        return false;
-      unsigned DefIdx;
-      if (!MI.isRegTiedToDefOperand(UseIdx, &DefIdx))
-        return false;
-      assert(DefMI->getOperand(DefIdx).isReg() &&
-             DefMI->getOperand(DefIdx).getReg() == SrcReg);
-
-      // Now commute def instruction.
-      MachineInstr *CommutedMI = TII->commuteInstruction(DefMI, true);
-      if (!CommutedMI)
-        return false;
-      SmallVector<unsigned, 1> Ops;
-      Ops.push_back(NewDstIdx);
-      MachineInstr *FoldedMI = TII->foldMemoryOperand(MF, CommutedMI, Ops, SS);
-      // Not needed since foldMemoryOperand returns new MI.
-      MF.DeleteMachineInstr(CommutedMI);
-      if (!FoldedMI)
-        return false;
-
-      VRM.addSpillSlotUse(SS, FoldedMI);
-      VRM.virtFolded(VirtReg, FoldedMI, VirtRegMap::isRef);
-      // Insert new def MI and spill MI.
-      const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-      TII->storeRegToStackSlot(MBB, &MI, NewReg, true, SS, RC);
-      MII = prior(MII);
-      MachineInstr *StoreMI = MII;
-      VRM.addSpillSlotUse(SS, StoreMI);
-      VRM.virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
-      MII = MBB.insert(MII, FoldedMI);  // Update MII to backtrack.
-
-      // Delete all 3 old instructions.
-      InvalidateKills(*ReloadMI, TRI, RegKills, KillOps);
-      VRM.RemoveMachineInstrFromMaps(ReloadMI);
-      MBB.erase(ReloadMI);
-      InvalidateKills(*DefMI, TRI, RegKills, KillOps);
-      VRM.RemoveMachineInstrFromMaps(DefMI);
-      MBB.erase(DefMI);
-      InvalidateKills(MI, TRI, RegKills, KillOps);
-      VRM.RemoveMachineInstrFromMaps(&MI);
-      MBB.erase(&MI);
-
-      // If NewReg was previously holding value of some SS, it's now clobbered.
-      // This has to be done now because it's a physical register. When this
-      // instruction is re-visited, it's ignored.
-      Spills.ClobberPhysReg(NewReg);
+    VRM->addSpillSlotUse(SS, FoldedMI);
+    VRM->virtFolded(VirtReg, FoldedMI, VirtRegMap::isRef);
+    // Insert new def MI and spill MI.
+    const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+    TII->storeRegToStackSlot(*MBB, &MI, NewReg, true, SS, RC);
+    MII = prior(MII);
+    MachineInstr *StoreMI = MII;
+    VRM->addSpillSlotUse(SS, StoreMI);
+    VRM->virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
+    MII = MBB->insert(MII, FoldedMI);  // Update MII to backtrack.
+
+    // Delete all 3 old instructions.
+    InvalidateKills(*ReloadMI, TRI, RegKills, KillOps);
+    VRM->RemoveMachineInstrFromMaps(ReloadMI);
+    MBB->erase(ReloadMI);
+    InvalidateKills(*DefMI, TRI, RegKills, KillOps);
+    VRM->RemoveMachineInstrFromMaps(DefMI);
+    MBB->erase(DefMI);
+    InvalidateKills(MI, TRI, RegKills, KillOps);
+    VRM->RemoveMachineInstrFromMaps(&MI);
+    MBB->erase(&MI);
 
-      ++NumCommutes;
-      return true;
-    }
+    // If NewReg was previously holding value of some SS, it's now clobbered.
+    // This has to be done now because it's a physical register. When this
+    // instruction is re-visited, it's ignored.
+    Spills.ClobberPhysReg(NewReg);
 
-    return false;
+    ++NumCommutes;
+    return true;
   }
 
-  /// SpillRegToStackSlot - Spill a register to a specified stack slot. Check if
-  /// the last store to the same slot is now dead. If so, remove the last store.
-  void SpillRegToStackSlot(MachineBasicBlock &MBB,
-                           MachineBasicBlock::iterator &MII,
-                           int Idx, unsigned PhysReg, int StackSlot,
-                           const TargetRegisterClass *RC,
-                           bool isAvailable, MachineInstr *&LastStore,
-                           AvailableSpills &Spills,
-                           SmallSet<MachineInstr*, 4> &ReMatDefs,
-                           BitVector &RegKills,
-                           std::vector<MachineOperand*> &KillOps,
-                           VirtRegMap &VRM) {
+  return false;
+}
 
-    MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
-    TII->storeRegToStackSlot(MBB, llvm::next(MII), PhysReg, true, StackSlot, RC);
-    MachineInstr *StoreMI = prior(oldNextMII);
-    VRM.addSpillSlotUse(StackSlot, StoreMI);
-    DEBUG(dbgs() << "Store:\t" << *StoreMI);
-
-    // If there is a dead store to this stack slot, nuke it now.
-    if (LastStore) {
-      DEBUG(dbgs() << "Removed dead store:\t" << *LastStore);
-      ++NumDSE;
-      SmallVector<unsigned, 2> KillRegs;
-      InvalidateKills(*LastStore, TRI, RegKills, KillOps, &KillRegs);
-      MachineBasicBlock::iterator PrevMII = LastStore;
-      bool CheckDef = PrevMII != MBB.begin();
-      if (CheckDef)
-        --PrevMII;
-      VRM.RemoveMachineInstrFromMaps(LastStore);
-      MBB.erase(LastStore);
-      if (CheckDef) {
-        // Look at defs of killed registers on the store. Mark the defs
-        // as dead since the store has been deleted and they aren't
-        // being reused.
-        for (unsigned j = 0, ee = KillRegs.size(); j != ee; ++j) {
-          bool HasOtherDef = false;
-          if (InvalidateRegDef(PrevMII, *MII, KillRegs[j], HasOtherDef, TRI)) {
-            MachineInstr *DeadDef = PrevMII;
-            if (ReMatDefs.count(DeadDef) && !HasOtherDef) {
-              // FIXME: This assumes a remat def does not have side effects.
-              VRM.RemoveMachineInstrFromMaps(DeadDef);
-              MBB.erase(DeadDef);
-              ++NumDRM;
-            }
+/// SpillRegToStackSlot - Spill a register to a specified stack slot. Check if
+/// the last store to the same slot is now dead. If so, remove the last store.
+void LocalRewriter::
+SpillRegToStackSlot(MachineBasicBlock::iterator &MII,
+                    int Idx, unsigned PhysReg, int StackSlot,
+                    const TargetRegisterClass *RC,
+                    bool isAvailable, MachineInstr *&LastStore,
+                    AvailableSpills &Spills,
+                    SmallSet<MachineInstr*, 4> &ReMatDefs,
+                    BitVector &RegKills,
+                    std::vector<MachineOperand*> &KillOps) {
+
+  MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
+  TII->storeRegToStackSlot(*MBB, llvm::next(MII), PhysReg, true, StackSlot, RC);
+  MachineInstr *StoreMI = prior(oldNextMII);
+  VRM->addSpillSlotUse(StackSlot, StoreMI);
+  DEBUG(dbgs() << "Store:\t" << *StoreMI);
+
+  // If there is a dead store to this stack slot, nuke it now.
+  if (LastStore) {
+    DEBUG(dbgs() << "Removed dead store:\t" << *LastStore);
+    ++NumDSE;
+    SmallVector<unsigned, 2> KillRegs;
+    InvalidateKills(*LastStore, TRI, RegKills, KillOps, &KillRegs);
+    MachineBasicBlock::iterator PrevMII = LastStore;
+    bool CheckDef = PrevMII != MBB->begin();
+    if (CheckDef)
+      --PrevMII;
+    VRM->RemoveMachineInstrFromMaps(LastStore);
+    MBB->erase(LastStore);
+    if (CheckDef) {
+      // Look at defs of killed registers on the store. Mark the defs
+      // as dead since the store has been deleted and they aren't
+      // being reused.
+      for (unsigned j = 0, ee = KillRegs.size(); j != ee; ++j) {
+        bool HasOtherDef = false;
+        if (InvalidateRegDef(PrevMII, *MII, KillRegs[j], HasOtherDef, TRI)) {
+          MachineInstr *DeadDef = PrevMII;
+          if (ReMatDefs.count(DeadDef) && !HasOtherDef) {
+            // FIXME: This assumes a remat def does not have side effects.
+            VRM->RemoveMachineInstrFromMaps(DeadDef);
+            MBB->erase(DeadDef);
+            ++NumDRM;
           }
         }
       }
     }
+  }
 
-    // Allow for multi-instruction spill sequences, as on PPC Altivec.  Presume
-    // the last of multiple instructions is the actual store.
-    LastStore = prior(oldNextMII);
-
-    // If the stack slot value was previously available in some other
-    // register, change it now.  Otherwise, make the register available,
-    // in PhysReg.
-    Spills.ModifyStackSlotOrReMat(StackSlot);
-    Spills.ClobberPhysReg(PhysReg);
-    Spills.addAvailable(StackSlot, PhysReg, isAvailable);
-    ++NumStores;
-  }
-
-  /// isSafeToDelete - Return true if this instruction doesn't produce any side
-  /// effect and all of its defs are dead.
-  static bool isSafeToDelete(MachineInstr &MI) {
-    const TargetInstrDesc &TID = MI.getDesc();
-    if (TID.mayLoad() || TID.mayStore() || TID.isCall() || TID.isTerminator() ||
-        TID.isCall() || TID.isBarrier() || TID.isReturn() ||
-        TID.hasUnmodeledSideEffects())
+  // Allow for multi-instruction spill sequences, as on PPC Altivec.  Presume
+  // the last of multiple instructions is the actual store.
+  LastStore = prior(oldNextMII);
+
+  // If the stack slot value was previously available in some other
+  // register, change it now.  Otherwise, make the register available,
+  // in PhysReg.
+  Spills.ModifyStackSlotOrReMat(StackSlot);
+  Spills.ClobberPhysReg(PhysReg);
+  Spills.addAvailable(StackSlot, PhysReg, isAvailable);
+  ++NumStores;
+}
+
+/// isSafeToDelete - Return true if this instruction doesn't produce any side
+/// effect and all of its defs are dead.
+static bool isSafeToDelete(MachineInstr &MI) {
+  const TargetInstrDesc &TID = MI.getDesc();
+  if (TID.mayLoad() || TID.mayStore() || TID.isCall() || TID.isTerminator() ||
+      TID.isCall() || TID.isBarrier() || TID.isReturn() ||
+      TID.hasUnmodeledSideEffects())
+    return false;
+  for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+    MachineOperand &MO = MI.getOperand(i);
+    if (!MO.isReg() || !MO.getReg())
+      continue;
+    if (MO.isDef() && !MO.isDead())
+      return false;
+    if (MO.isUse() && MO.isKill())
+      // FIXME: We can't remove kill markers or else the scavenger will assert.
+      // An alternative is to add a ADD pseudo instruction to replace kill
+      // markers.
       return false;
-    for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
-      MachineOperand &MO = MI.getOperand(i);
-      if (!MO.isReg() || !MO.getReg())
-        continue;
-      if (MO.isDef() && !MO.isDead())
-        return false;
-      if (MO.isUse() && MO.isKill())
-        // FIXME: We can't remove kill markers or else the scavenger will assert.
-        // An alternative is to add a ADD pseudo instruction to replace kill
-        // markers.
-        return false;
-    }
-    return true;
   }
+  return true;
+}
 
-  /// TransferDeadness - A identity copy definition is dead and it's being
-  /// removed. Find the last def or use and mark it as dead / kill.
-  void TransferDeadness(MachineBasicBlock *MBB, unsigned CurDist,
-                        unsigned Reg, BitVector &RegKills,
-                        std::vector<MachineOperand*> &KillOps,
-                        VirtRegMap &VRM) {
-    SmallPtrSet<MachineInstr*, 4> Seens;
-    SmallVector<std::pair<MachineInstr*, int>,8> Refs;
-    for (MachineRegisterInfo::reg_iterator RI = RegInfo->reg_begin(Reg),
-           RE = RegInfo->reg_end(); RI != RE; ++RI) {
-      MachineInstr *UDMI = &*RI;
-      if (UDMI->getParent() != MBB)
-        continue;
-      DenseMap<MachineInstr*, unsigned>::iterator DI = DistanceMap.find(UDMI);
-      if (DI == DistanceMap.end() || DI->second > CurDist)
+/// TransferDeadness - A identity copy definition is dead and it's being
+/// removed. Find the last def or use and mark it as dead / kill.
+void LocalRewriter::
+TransferDeadness(unsigned CurDist,
+                 unsigned Reg, BitVector &RegKills,
+                 std::vector<MachineOperand*> &KillOps) {
+  SmallPtrSet<MachineInstr*, 4> Seens;
+  SmallVector<std::pair<MachineInstr*, int>,8> Refs;
+  for (MachineRegisterInfo::reg_iterator RI = MRI->reg_begin(Reg),
+         RE = MRI->reg_end(); RI != RE; ++RI) {
+    MachineInstr *UDMI = &*RI;
+    if (UDMI->getParent() != MBB)
+      continue;
+    DenseMap<MachineInstr*, unsigned>::iterator DI = DistanceMap.find(UDMI);
+    if (DI == DistanceMap.end() || DI->second > CurDist)
+      continue;
+    if (Seens.insert(UDMI))
+      Refs.push_back(std::make_pair(UDMI, DI->second));
+  }
+
+  if (Refs.empty())
+    return;
+  std::sort(Refs.begin(), Refs.end(), RefSorter());
+
+  while (!Refs.empty()) {
+    MachineInstr *LastUDMI = Refs.back().first;
+    Refs.pop_back();
+
+    MachineOperand *LastUD = NULL;
+    for (unsigned i = 0, e = LastUDMI->getNumOperands(); i != e; ++i) {
+      MachineOperand &MO = LastUDMI->getOperand(i);
+      if (!MO.isReg() || MO.getReg() != Reg)
         continue;
-      if (Seens.insert(UDMI))
-        Refs.push_back(std::make_pair(UDMI, DI->second));
+      if (!LastUD || (LastUD->isUse() && MO.isDef()))
+        LastUD = &MO;
+      if (LastUDMI->isRegTiedToDefOperand(i))
+        break;
     }
-
-    if (Refs.empty())
-      return;
-    std::sort(Refs.begin(), Refs.end(), RefSorter());
-
-    while (!Refs.empty()) {
-      MachineInstr *LastUDMI = Refs.back().first;
-      Refs.pop_back();
-
-      MachineOperand *LastUD = NULL;
-      for (unsigned i = 0, e = LastUDMI->getNumOperands(); i != e; ++i) {
-        MachineOperand &MO = LastUDMI->getOperand(i);
-        if (!MO.isReg() || MO.getReg() != Reg)
-          continue;
-        if (!LastUD || (LastUD->isUse() && MO.isDef()))
-          LastUD = &MO;
-        if (LastUDMI->isRegTiedToDefOperand(i))
-          break;
-      }
-      if (LastUD->isDef()) {
-        // If the instruction has no side effect, delete it and propagate
-        // backward further. Otherwise, mark is dead and we are done.
-        if (!isSafeToDelete(*LastUDMI)) {
-          LastUD->setIsDead();
-          break;
-        }
-        VRM.RemoveMachineInstrFromMaps(LastUDMI);
-        MBB->erase(LastUDMI);
-      } else {
-        LastUD->setIsKill();
-        RegKills.set(Reg);
-        KillOps[Reg] = LastUD;
+    if (LastUD->isDef()) {
+      // If the instruction has no side effect, delete it and propagate
+      // backward further. Otherwise, mark is dead and we are done.
+      if (!isSafeToDelete(*LastUDMI)) {
+        LastUD->setIsDead();
         break;
       }
+      VRM->RemoveMachineInstrFromMaps(LastUDMI);
+      MBB->erase(LastUDMI);
+    } else {
+      LastUD->setIsKill();
+      RegKills.set(Reg);
+      KillOps[Reg] = LastUD;
+      break;
     }
   }
+}
 
-  /// rewriteMBB - Keep track of which spills are available even after the
-  /// register allocator is done with them.  If possible, avid reloading vregs.
-  void RewriteMBB(MachineBasicBlock &MBB, VirtRegMap &VRM,
-                  LiveIntervals *LIs,
-                  AvailableSpills &Spills, BitVector &RegKills,
-                  std::vector<MachineOperand*> &KillOps) {
-
-    DEBUG(dbgs() << "\n**** Local spiller rewriting MBB '"
-          << MBB.getName() << "':\n");
+/// rewriteMBB - Keep track of which spills are available even after the
+/// register allocator is done with them.  If possible, avid reloading vregs.
+void
+LocalRewriter::RewriteMBB(LiveIntervals *LIs,
+                          AvailableSpills &Spills, BitVector &RegKills,
+                          std::vector<MachineOperand*> &KillOps) {
+
+  DEBUG(dbgs() << "\n**** Local spiller rewriting MBB '"
+               << MBB->getName() << "':\n");
+
+  MachineFunction &MF = *MBB->getParent();
+
+  // MaybeDeadStores - When we need to write a value back into a stack slot,
+  // keep track of the inserted store.  If the stack slot value is never read
+  // (because the value was used from some available register, for example), and
+  // subsequently stored to, the original store is dead.  This map keeps track
+  // of inserted stores that are not used.  If we see a subsequent store to the
+  // same stack slot, the original store is deleted.
+  std::vector<MachineInstr*> MaybeDeadStores;
+  MaybeDeadStores.resize(MF.getFrameInfo()->getObjectIndexEnd(), NULL);
+
+  // ReMatDefs - These are rematerializable def MIs which are not deleted.
+  SmallSet<MachineInstr*, 4> ReMatDefs;
+
+  // Clear kill info.
+  SmallSet<unsigned, 2> KilledMIRegs;
+  RegKills.reset();
+  KillOps.clear();
+  KillOps.resize(TRI->getNumRegs(), NULL);
+
+  unsigned Dist = 0;
+  DistanceMap.clear();
+  for (MachineBasicBlock::iterator MII = MBB->begin(), E = MBB->end();
+       MII != E; ) {
+    MachineBasicBlock::iterator NextMII = llvm::next(MII);
 
-    MachineFunction &MF = *MBB.getParent();
-    
-    // MaybeDeadStores - When we need to write a value back into a stack slot,
-    // keep track of the inserted store.  If the stack slot value is never read
-    // (because the value was used from some available register, for example), and
-    // subsequently stored to, the original store is dead.  This map keeps track
-    // of inserted stores that are not used.  If we see a subsequent store to the
-    // same stack slot, the original store is deleted.
-    std::vector<MachineInstr*> MaybeDeadStores;
-    MaybeDeadStores.resize(MF.getFrameInfo()->getObjectIndexEnd(), NULL);
-
-    // ReMatDefs - These are rematerializable def MIs which are not deleted.
-    SmallSet<MachineInstr*, 4> ReMatDefs;
-
-    // Clear kill info.
-    SmallSet<unsigned, 2> KilledMIRegs;
-    RegKills.reset();
-    KillOps.clear();
-    KillOps.resize(TRI->getNumRegs(), NULL);
-
-    unsigned Dist = 0;
-    DistanceMap.clear();
-    for (MachineBasicBlock::iterator MII = MBB.begin(), E = MBB.end();
-         MII != E; ) {
-      MachineBasicBlock::iterator NextMII = llvm::next(MII);
-
-      VirtRegMap::MI2VirtMapTy::const_iterator I, End;
-      bool Erased = false;
-      bool BackTracked = false;
-      if (OptimizeByUnfold(MBB, MII,
-                           MaybeDeadStores, Spills, RegKills, KillOps, VRM))
-        NextMII = llvm::next(MII);
+    VirtRegMap::MI2VirtMapTy::const_iterator I, End;
+    bool Erased = false;
+    bool BackTracked = false;
+    if (OptimizeByUnfold(MII, MaybeDeadStores, Spills, RegKills, KillOps))
+      NextMII = llvm::next(MII);
 
-      MachineInstr &MI = *MII;
+    MachineInstr &MI = *MII;
 
-      if (VRM.hasEmergencySpills(&MI)) {
-        // Spill physical register(s) in the rare case the allocator has run out
-        // of registers to allocate.
-        SmallSet<int, 4> UsedSS;
-        std::vector<unsigned> &EmSpills = VRM.getEmergencySpills(&MI);
-        for (unsigned i = 0, e = EmSpills.size(); i != e; ++i) {
-          unsigned PhysReg = EmSpills[i];
-          const TargetRegisterClass *RC =
-            TRI->getPhysicalRegisterRegClass(PhysReg);
-          assert(RC && "Unable to determine register class!");
-          int SS = VRM.getEmergencySpillSlot(RC);
-          if (UsedSS.count(SS))
-            llvm_unreachable("Need to spill more than one physical registers!");
-          UsedSS.insert(SS);
-          TII->storeRegToStackSlot(MBB, MII, PhysReg, true, SS, RC);
-          MachineInstr *StoreMI = prior(MII);
-          VRM.addSpillSlotUse(SS, StoreMI);
+    if (VRM->hasEmergencySpills(&MI)) {
+      // Spill physical register(s) in the rare case the allocator has run out
+      // of registers to allocate.
+      SmallSet<int, 4> UsedSS;
+      std::vector<unsigned> &EmSpills = VRM->getEmergencySpills(&MI);
+      for (unsigned i = 0, e = EmSpills.size(); i != e; ++i) {
+        unsigned PhysReg = EmSpills[i];
+        const TargetRegisterClass *RC =
+          TRI->getPhysicalRegisterRegClass(PhysReg);
+        assert(RC && "Unable to determine register class!");
+        int SS = VRM->getEmergencySpillSlot(RC);
+        if (UsedSS.count(SS))
+          llvm_unreachable("Need to spill more than one physical registers!");
+        UsedSS.insert(SS);
+        TII->storeRegToStackSlot(*MBB, MII, PhysReg, true, SS, RC);
+        MachineInstr *StoreMI = prior(MII);
+        VRM->addSpillSlotUse(SS, StoreMI);
 
-          // Back-schedule reloads and remats.
-          MachineBasicBlock::iterator InsertLoc =
-            ComputeReloadLoc(llvm::next(MII), MBB.begin(), PhysReg, TRI, false,
-                             SS, TII, MF);
+        // Back-schedule reloads and remats.
+        MachineBasicBlock::iterator InsertLoc =
+          ComputeReloadLoc(llvm::next(MII), MBB->begin(), PhysReg, TRI, false,
+                           SS, TII, MF);
 
-          TII->loadRegFromStackSlot(MBB, InsertLoc, PhysReg, SS, RC);
+        TII->loadRegFromStackSlot(*MBB, InsertLoc, PhysReg, SS, RC);
 
-          MachineInstr *LoadMI = prior(InsertLoc);
-          VRM.addSpillSlotUse(SS, LoadMI);
-          ++NumPSpills;
-          DistanceMap.insert(std::make_pair(LoadMI, Dist++));
-        }
-        NextMII = llvm::next(MII);
-      }
+        MachineInstr *LoadMI = prior(InsertLoc);
+        VRM->addSpillSlotUse(SS, LoadMI);
+        ++NumPSpills;
+        DistanceMap.insert(std::make_pair(LoadMI, Dist++));
+      }
+      NextMII = llvm::next(MII);
+    }
+
+    // Insert restores here if asked to.
+    if (VRM->isRestorePt(&MI)) {
+      std::vector<unsigned> &RestoreRegs = VRM->getRestorePtRestores(&MI);
+      for (unsigned i = 0, e = RestoreRegs.size(); i != e; ++i) {
+        unsigned VirtReg = RestoreRegs[e-i-1];  // Reverse order.
+        if (!VRM->getPreSplitReg(VirtReg))
+          continue; // Split interval spilled again.
+        unsigned Phys = VRM->getPhys(VirtReg);
+        MRI->setPhysRegUsed(Phys);
+
+        // Check if the value being restored if available. If so, it must be
+        // from a predecessor BB that fallthrough into this BB. We do not
+        // expect:
+        // BB1:
+        // r1 = load fi#1
+        // ...
+        //    = r1<kill>
+        // ... # r1 not clobbered
+        // ...
+        //    = load fi#1
+        bool DoReMat = VRM->isReMaterialized(VirtReg);
+        int SSorRMId = DoReMat
+          ? VRM->getReMatId(VirtReg) : VRM->getStackSlot(VirtReg);
+        const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+        unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
+        if (InReg == Phys) {
+          // If the value is already available in the expected register, save
+          // a reload / remat.
+          if (SSorRMId)
+            DEBUG(dbgs() << "Reusing RM#"
+                  << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
+          else
+            DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
+          DEBUG(dbgs() << " from physreg "
+                << TRI->getName(InReg) << " for vreg"
+                << VirtReg <<" instead of reloading into physreg "
+                << TRI->getName(Phys) << '\n');
+          ++NumOmitted;
+          continue;
+        } else if (InReg && InReg != Phys) {
+          if (SSorRMId)
+            DEBUG(dbgs() << "Reusing RM#"
+                  << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
+          else
+            DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
+          DEBUG(dbgs() << " from physreg "
+                << TRI->getName(InReg) << " for vreg"
+                << VirtReg <<" by copying it into physreg "
+                << TRI->getName(Phys) << '\n');
 
-      // Insert restores here if asked to.
-      if (VRM.isRestorePt(&MI)) {
-        std::vector<unsigned> &RestoreRegs = VRM.getRestorePtRestores(&MI);
-        for (unsigned i = 0, e = RestoreRegs.size(); i != e; ++i) {
-          unsigned VirtReg = RestoreRegs[e-i-1];  // Reverse order.
-          if (!VRM.getPreSplitReg(VirtReg))
-            continue; // Split interval spilled again.
-          unsigned Phys = VRM.getPhys(VirtReg);
-          RegInfo->setPhysRegUsed(Phys);
-
-          // Check if the value being restored if available. If so, it must be
-          // from a predecessor BB that fallthrough into this BB. We do not
-          // expect:
-          // BB1:
-          // r1 = load fi#1
-          // ...
-          //    = r1<kill>
-          // ... # r1 not clobbered
-          // ...
-          //    = load fi#1
-          bool DoReMat = VRM.isReMaterialized(VirtReg);
-          int SSorRMId = DoReMat
-            ? VRM.getReMatId(VirtReg) : VRM.getStackSlot(VirtReg);
-          const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-          unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
-          if (InReg == Phys) {
-            // If the value is already available in the expected register, save
-            // a reload / remat.
-            if (SSorRMId)
-              DEBUG(dbgs() << "Reusing RM#"
-                           << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
-            else
-              DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
-            DEBUG(dbgs() << " from physreg "
-                         << TRI->getName(InReg) << " for vreg"
-                         << VirtReg <<" instead of reloading into physreg "
-                         << TRI->getName(Phys) << '\n');
-            ++NumOmitted;
-            continue;
-          } else if (InReg && InReg != Phys) {
-            if (SSorRMId)
-              DEBUG(dbgs() << "Reusing RM#"
-                           << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
-            else
-              DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
-            DEBUG(dbgs() << " from physreg "
-                         << TRI->getName(InReg) << " for vreg"
-                         << VirtReg <<" by copying it into physreg "
-                         << TRI->getName(Phys) << '\n');
-
-            // If the reloaded / remat value is available in another register,
-            // copy it to the desired register.
-
-            // Back-schedule reloads and remats.
-            MachineBasicBlock::iterator InsertLoc =
-              ComputeReloadLoc(MII, MBB.begin(), Phys, TRI, DoReMat,
-                               SSorRMId, TII, MF);
-
-            TII->copyRegToReg(MBB, InsertLoc, Phys, InReg, RC, RC);
-
-            // This invalidates Phys.
-            Spills.ClobberPhysReg(Phys);
-            // Remember it's available.
-            Spills.addAvailable(SSorRMId, Phys);
-
-            // Mark is killed.
-            MachineInstr *CopyMI = prior(InsertLoc);
-            CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
-            MachineOperand *KillOpnd = CopyMI->findRegisterUseOperand(InReg);
-            KillOpnd->setIsKill();
-            UpdateKills(*CopyMI, TRI, RegKills, KillOps);
-
-            DEBUG(dbgs() << '\t' << *CopyMI);
-            ++NumCopified;
-            continue;
-          }
+          // If the reloaded / remat value is available in another register,
+          // copy it to the desired register.
 
           // Back-schedule reloads and remats.
           MachineBasicBlock::iterator InsertLoc =
-            ComputeReloadLoc(MII, MBB.begin(), Phys, TRI, DoReMat,
+            ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat,
                              SSorRMId, TII, MF);
 
-          if (VRM.isReMaterialized(VirtReg)) {
-            ReMaterialize(MBB, InsertLoc, Phys, VirtReg, TII, TRI, VRM);
-          } else {
-            const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-            TII->loadRegFromStackSlot(MBB, InsertLoc, Phys, SSorRMId, RC);
-            MachineInstr *LoadMI = prior(InsertLoc);
-            VRM.addSpillSlotUse(SSorRMId, LoadMI);
-            ++NumLoads;
-            DistanceMap.insert(std::make_pair(LoadMI, Dist++));
-          }
+          TII->copyRegToReg(*MBB, InsertLoc, Phys, InReg, RC, RC);
 
           // This invalidates Phys.
           Spills.ClobberPhysReg(Phys);
           // Remember it's available.
           Spills.addAvailable(SSorRMId, Phys);
 
-          UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
-          DEBUG(dbgs() << '\t' << *prior(MII));
-        }
-      }
-
-      // Insert spills here if asked to.
-      if (VRM.isSpillPt(&MI)) {
-        std::vector<std::pair<unsigned,bool> > &SpillRegs =
-          VRM.getSpillPtSpills(&MI);
-        for (unsigned i = 0, e = SpillRegs.size(); i != e; ++i) {
-          unsigned VirtReg = SpillRegs[i].first;
-          bool isKill = SpillRegs[i].second;
-          if (!VRM.getPreSplitReg(VirtReg))
-            continue; // Split interval spilled again.
-          const TargetRegisterClass *RC = RegInfo->getRegClass(VirtReg);
-          unsigned Phys = VRM.getPhys(VirtReg);
-          int StackSlot = VRM.getStackSlot(VirtReg);
-          MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
-          TII->storeRegToStackSlot(MBB, llvm::next(MII), Phys, isKill, StackSlot, RC);
-          MachineInstr *StoreMI = prior(oldNextMII);
-          VRM.addSpillSlotUse(StackSlot, StoreMI);
-          DEBUG(dbgs() << "Store:\t" << *StoreMI);
-          VRM.virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
-        }
-        NextMII = llvm::next(MII);
-      }
+          // Mark is killed.
+          MachineInstr *CopyMI = prior(InsertLoc);
+          CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
+          MachineOperand *KillOpnd = CopyMI->findRegisterUseOperand(InReg);
+          KillOpnd->setIsKill();
+          UpdateKills(*CopyMI, TRI, RegKills, KillOps);
 
-      /// ReusedOperands - Keep track of operand reuse in case we need to undo
-      /// reuse.
-      ReuseInfo ReusedOperands(MI, TRI);
-      SmallVector<unsigned, 4> VirtUseOps;
-      for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
-        MachineOperand &MO = MI.getOperand(i);
-        if (!MO.isReg() || MO.getReg() == 0)
-          continue;   // Ignore non-register operands.
-        
-        unsigned VirtReg = MO.getReg();
-        if (TargetRegisterInfo::isPhysicalRegister(VirtReg)) {
-          // Ignore physregs for spilling, but remember that it is used by this
-          // function.
-          RegInfo->setPhysRegUsed(VirtReg);
+          DEBUG(dbgs() << '\t' << *CopyMI);
+          ++NumCopified;
           continue;
         }
 
-        // We want to process implicit virtual register uses first.
-        if (MO.isImplicit())
-          // If the virtual register is implicitly defined, emit a implicit_def
-          // before so scavenger knows it's "defined".
-          // FIXME: This is a horrible hack done the by register allocator to
-          // remat a definition with virtual register operand.
-          VirtUseOps.insert(VirtUseOps.begin(), i);
-        else
-          VirtUseOps.push_back(i);
-      }
-
-      // Process all of the spilled uses and all non spilled reg references.
-      SmallVector<int, 2> PotentialDeadStoreSlots;
-      KilledMIRegs.clear();
-      for (unsigned j = 0, e = VirtUseOps.size(); j != e; ++j) {
-        unsigned i = VirtUseOps[j];
-        unsigned VirtReg = MI.getOperand(i).getReg();
-        assert(TargetRegisterInfo::isVirtualRegister(VirtReg) &&
-               "Not a virtual register?");
-
-        unsigned SubIdx = MI.getOperand(i).getSubReg();
-        if (VRM.isAssignedReg(VirtReg)) {
-          // This virtual register was assigned a physreg!
-          unsigned Phys = VRM.getPhys(VirtReg);
-          RegInfo->setPhysRegUsed(Phys);
-          if (MI.getOperand(i).isDef())
-            ReusedOperands.markClobbered(Phys);
-          substitutePhysReg(MI.getOperand(i), Phys, *TRI);
-          if (VRM.isImplicitlyDefined(VirtReg))
-            // FIXME: Is this needed?
-            BuildMI(MBB, &MI, MI.getDebugLoc(),
-                    TII->get(TargetOpcode::IMPLICIT_DEF), Phys);
-          continue;
+        // Back-schedule reloads and remats.
+        MachineBasicBlock::iterator InsertLoc =
+          ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat,
+                           SSorRMId, TII, MF);
+
+        if (VRM->isReMaterialized(VirtReg)) {
+          ReMaterialize(*MBB, InsertLoc, Phys, VirtReg, TII, TRI, *VRM);
+        } else {
+          const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+          TII->loadRegFromStackSlot(*MBB, InsertLoc, Phys, SSorRMId, RC);
+          MachineInstr *LoadMI = prior(InsertLoc);
+          VRM->addSpillSlotUse(SSorRMId, LoadMI);
+          ++NumLoads;
+          DistanceMap.insert(std::make_pair(LoadMI, Dist++));
         }
 
-        // This virtual register is now known to be a spilled value.
-        if (!MI.getOperand(i).isUse())
-          continue;  // Handle defs in the loop below (handle use&def here though)
-
-        bool AvoidReload = MI.getOperand(i).isUndef();
-        // Check if it is defined by an implicit def. It should not be spilled.
-        // Note, this is for correctness reason. e.g.
-        // 8   %reg1024<def> = IMPLICIT_DEF
-        // 12  %reg1024<def> = INSERT_SUBREG %reg1024<kill>, %reg1025, 2
-        // The live range [12, 14) are not part of the r1024 live interval since
-        // it's defined by an implicit def. It will not conflicts with live
-        // interval of r1025. Now suppose both registers are spilled, you can
-        // easily see a situation where both registers are reloaded before
-        // the INSERT_SUBREG and both target registers that would overlap.
-        bool DoReMat = VRM.isReMaterialized(VirtReg);
-        int SSorRMId = DoReMat
-          ? VRM.getReMatId(VirtReg) : VRM.getStackSlot(VirtReg);
-        int ReuseSlot = SSorRMId;
+        // This invalidates Phys.
+        Spills.ClobberPhysReg(Phys);
+        // Remember it's available.
+        Spills.addAvailable(SSorRMId, Phys);
+
+        UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
+        DEBUG(dbgs() << '\t' << *prior(MII));
+      }
+    }
 
-        // Check to see if this stack slot is available.
-        unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
+    // Insert spills here if asked to.
+    if (VRM->isSpillPt(&MI)) {
+      std::vector<std::pair<unsigned,bool> > &SpillRegs =
+        VRM->getSpillPtSpills(&MI);
+      for (unsigned i = 0, e = SpillRegs.size(); i != e; ++i) {
+        unsigned VirtReg = SpillRegs[i].first;
+        bool isKill = SpillRegs[i].second;
+        if (!VRM->getPreSplitReg(VirtReg))
+          continue; // Split interval spilled again.
+        const TargetRegisterClass *RC = MRI->getRegClass(VirtReg);
+        unsigned Phys = VRM->getPhys(VirtReg);
+        int StackSlot = VRM->getStackSlot(VirtReg);
+        MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
+        TII->storeRegToStackSlot(*MBB, llvm::next(MII), Phys, isKill, StackSlot,
+                                 RC);
+        MachineInstr *StoreMI = prior(oldNextMII);
+        VRM->addSpillSlotUse(StackSlot, StoreMI);
+        DEBUG(dbgs() << "Store:\t" << *StoreMI);
+        VRM->virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
+      }
+      NextMII = llvm::next(MII);
+    }
+
+    /// ReusedOperands - Keep track of operand reuse in case we need to undo
+    /// reuse.
+    ReuseInfo ReusedOperands(MI, TRI);
+    SmallVector<unsigned, 4> VirtUseOps;
+    for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+      MachineOperand &MO = MI.getOperand(i);
+      if (!MO.isReg() || MO.getReg() == 0)
+        continue;   // Ignore non-register operands.
 
-        // If this is a sub-register use, make sure the reuse register is in the
-        // right register class. For example, for x86 not all of the 32-bit
-        // registers have accessible sub-registers.
-        // Similarly so for EXTRACT_SUBREG. Consider this:
-        // EDI = op
-        // MOV32_mr fi#1, EDI
-        // ...
-        //       = EXTRACT_SUBREG fi#1
-        // fi#1 is available in EDI, but it cannot be reused because it's not in
-        // the right register file.
-        if (PhysReg && !AvoidReload && (SubIdx || MI.isExtractSubreg())) {
-          const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-          if (!RC->contains(PhysReg))
-            PhysReg = 0;
-        }
-
-        if (PhysReg && !AvoidReload) {
-          // This spilled operand might be part of a two-address operand.  If this
-          // is the case, then changing it will necessarily require changing the 
-          // def part of the instruction as well.  However, in some cases, we
-          // aren't allowed to modify the reused register.  If none of these cases
-          // apply, reuse it.
-          bool CanReuse = true;
-          bool isTied = MI.isRegTiedToDefOperand(i);
-          if (isTied) {
-            // Okay, we have a two address operand.  We can reuse this physreg as
-            // long as we are allowed to clobber the value and there isn't an
-            // earlier def that has already clobbered the physreg.
-            CanReuse = !ReusedOperands.isClobbered(PhysReg) &&
-              Spills.canClobberPhysReg(PhysReg);
-          }
-          
-          if (CanReuse) {
-            // If this stack slot value is already available, reuse it!
-            if (ReuseSlot > VirtRegMap::MAX_STACK_SLOT)
-              DEBUG(dbgs() << "Reusing RM#"
-                           << ReuseSlot-VirtRegMap::MAX_STACK_SLOT-1);
-            else
-              DEBUG(dbgs() << "Reusing SS#" << ReuseSlot);
-            DEBUG(dbgs() << " from physreg "
-                         << TRI->getName(PhysReg) << " for vreg"
-                         << VirtReg <<" instead of reloading into physreg "
-                         << TRI->getName(VRM.getPhys(VirtReg)) << '\n');
-            unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
-            MI.getOperand(i).setReg(RReg);
-            MI.getOperand(i).setSubReg(0);
-
-            // The only technical detail we have is that we don't know that
-            // PhysReg won't be clobbered by a reloaded stack slot that occurs
-            // later in the instruction.  In particular, consider 'op V1, V2'.
-            // If V1 is available in physreg R0, we would choose to reuse it
-            // here, instead of reloading it into the register the allocator
-            // indicated (say R1).  However, V2 might have to be reloaded
-            // later, and it might indicate that it needs to live in R0.  When
-            // this occurs, we need to have information available that
-            // indicates it is safe to use R1 for the reload instead of R0.
-            //
-            // To further complicate matters, we might conflict with an alias,
-            // or R0 and R1 might not be compatible with each other.  In this
-            // case, we actually insert a reload for V1 in R1, ensuring that
-            // we can get at R0 or its alias.
-            ReusedOperands.addReuse(i, ReuseSlot, PhysReg,
-                                    VRM.getPhys(VirtReg), VirtReg);
-            if (isTied)
-              // Only mark it clobbered if this is a use&def operand.
-              ReusedOperands.markClobbered(PhysReg);
-            ++NumReused;
-
-            if (MI.getOperand(i).isKill() &&
-                ReuseSlot <= VirtRegMap::MAX_STACK_SLOT) {
-
-              // The store of this spilled value is potentially dead, but we
-              // won't know for certain until we've confirmed that the re-use
-              // above is valid, which means waiting until the other operands
-              // are processed. For now we just track the spill slot, we'll
-              // remove it after the other operands are processed if valid.
+      unsigned VirtReg = MO.getReg();
+      if (TargetRegisterInfo::isPhysicalRegister(VirtReg)) {
+        // Ignore physregs for spilling, but remember that it is used by this
+        // function.
+        MRI->setPhysRegUsed(VirtReg);
+        continue;
+      }
 
-              PotentialDeadStoreSlots.push_back(ReuseSlot);
-            }
+      // We want to process implicit virtual register uses first.
+      if (MO.isImplicit())
+        // If the virtual register is implicitly defined, emit a implicit_def
+        // before so scavenger knows it's "defined".
+        // FIXME: This is a horrible hack done the by register allocator to
+        // remat a definition with virtual register operand.
+        VirtUseOps.insert(VirtUseOps.begin(), i);
+      else
+        VirtUseOps.push_back(i);
+    }
+
+    // Process all of the spilled uses and all non spilled reg references.
+    SmallVector<int, 2> PotentialDeadStoreSlots;
+    KilledMIRegs.clear();
+    for (unsigned j = 0, e = VirtUseOps.size(); j != e; ++j) {
+      unsigned i = VirtUseOps[j];
+      unsigned VirtReg = MI.getOperand(i).getReg();
+      assert(TargetRegisterInfo::isVirtualRegister(VirtReg) &&
+             "Not a virtual register?");
+
+      unsigned SubIdx = MI.getOperand(i).getSubReg();
+      if (VRM->isAssignedReg(VirtReg)) {
+        // This virtual register was assigned a physreg!
+        unsigned Phys = VRM->getPhys(VirtReg);
+        MRI->setPhysRegUsed(Phys);
+        if (MI.getOperand(i).isDef())
+          ReusedOperands.markClobbered(Phys);
+        substitutePhysReg(MI.getOperand(i), Phys, *TRI);
+        if (VRM->isImplicitlyDefined(VirtReg))
+          // FIXME: Is this needed?
+          BuildMI(*MBB, &MI, MI.getDebugLoc(),
+                  TII->get(TargetOpcode::IMPLICIT_DEF), Phys);
+        continue;
+      }
 
-            // Mark is isKill if it's there no other uses of the same virtual
-            // register and it's not a two-address operand. IsKill will be
-            // unset if reg is reused.
-            if (!isTied && KilledMIRegs.count(VirtReg) == 0) {
-              MI.getOperand(i).setIsKill();
-              KilledMIRegs.insert(VirtReg);
-            }
+      // This virtual register is now known to be a spilled value.
+      if (!MI.getOperand(i).isUse())
+        continue;  // Handle defs in the loop below (handle use&def here though)
+
+      bool AvoidReload = MI.getOperand(i).isUndef();
+      // Check if it is defined by an implicit def. It should not be spilled.
+      // Note, this is for correctness reason. e.g.
+      // 8   %reg1024<def> = IMPLICIT_DEF
+      // 12  %reg1024<def> = INSERT_SUBREG %reg1024<kill>, %reg1025, 2
+      // The live range [12, 14) are not part of the r1024 live interval since
+      // it's defined by an implicit def. It will not conflicts with live
+      // interval of r1025. Now suppose both registers are spilled, you can
+      // easily see a situation where both registers are reloaded before
+      // the INSERT_SUBREG and both target registers that would overlap.
+      bool DoReMat = VRM->isReMaterialized(VirtReg);
+      int SSorRMId = DoReMat
+        ? VRM->getReMatId(VirtReg) : VRM->getStackSlot(VirtReg);
+      int ReuseSlot = SSorRMId;
+
+      // Check to see if this stack slot is available.
+      unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
+
+      // If this is a sub-register use, make sure the reuse register is in the
+      // right register class. For example, for x86 not all of the 32-bit
+      // registers have accessible sub-registers.
+      // Similarly so for EXTRACT_SUBREG. Consider this:
+      // EDI = op
+      // MOV32_mr fi#1, EDI
+      // ...
+      //       = EXTRACT_SUBREG fi#1
+      // fi#1 is available in EDI, but it cannot be reused because it's not in
+      // the right register file.
+      if (PhysReg && !AvoidReload && (SubIdx || MI.isExtractSubreg())) {
+        const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+        if (!RC->contains(PhysReg))
+          PhysReg = 0;
+      }
+
+      if (PhysReg && !AvoidReload) {
+        // This spilled operand might be part of a two-address operand.  If this
+        // is the case, then changing it will necessarily require changing the
+        // def part of the instruction as well.  However, in some cases, we
+        // aren't allowed to modify the reused register.  If none of these cases
+        // apply, reuse it.
+        bool CanReuse = true;
+        bool isTied = MI.isRegTiedToDefOperand(i);
+        if (isTied) {
+          // Okay, we have a two address operand.  We can reuse this physreg as
+          // long as we are allowed to clobber the value and there isn't an
+          // earlier def that has already clobbered the physreg.
+          CanReuse = !ReusedOperands.isClobbered(PhysReg) &&
+            Spills.canClobberPhysReg(PhysReg);
+        }
+
+        if (CanReuse) {
+          // If this stack slot value is already available, reuse it!
+          if (ReuseSlot > VirtRegMap::MAX_STACK_SLOT)
+            DEBUG(dbgs() << "Reusing RM#"
+                  << ReuseSlot-VirtRegMap::MAX_STACK_SLOT-1);
+          else
+            DEBUG(dbgs() << "Reusing SS#" << ReuseSlot);
+          DEBUG(dbgs() << " from physreg "
+                << TRI->getName(PhysReg) << " for vreg"
+                << VirtReg <<" instead of reloading into physreg "
+                << TRI->getName(VRM->getPhys(VirtReg)) << '\n');
+          unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
+          MI.getOperand(i).setReg(RReg);
+          MI.getOperand(i).setSubReg(0);
 
-            continue;
-          }  // CanReuse
-          
-          // Otherwise we have a situation where we have a two-address instruction
-          // whose mod/ref operand needs to be reloaded.  This reload is already
-          // available in some register "PhysReg", but if we used PhysReg as the
-          // operand to our 2-addr instruction, the instruction would modify
-          // PhysReg.  This isn't cool if something later uses PhysReg and expects
-          // to get its initial value.
+          // The only technical detail we have is that we don't know that
+          // PhysReg won't be clobbered by a reloaded stack slot that occurs
+          // later in the instruction.  In particular, consider 'op V1, V2'.
+          // If V1 is available in physreg R0, we would choose to reuse it
+          // here, instead of reloading it into the register the allocator
+          // indicated (say R1).  However, V2 might have to be reloaded
+          // later, and it might indicate that it needs to live in R0.  When
+          // this occurs, we need to have information available that
+          // indicates it is safe to use R1 for the reload instead of R0.
           //
-          // To avoid this problem, and to avoid doing a load right after a store,
-          // we emit a copy from PhysReg into the designated register for this
-          // operand.
-          unsigned DesignatedReg = VRM.getPhys(VirtReg);
-          assert(DesignatedReg && "Must map virtreg to physreg!");
-
-          // Note that, if we reused a register for a previous operand, the
-          // register we want to reload into might not actually be
-          // available.  If this occurs, use the register indicated by the
-          // reuser.
-          if (ReusedOperands.hasReuses())
-            DesignatedReg = ReusedOperands.GetRegForReload(VirtReg,
-                                                           DesignatedReg, &MI, 
-                               Spills, MaybeDeadStores, RegKills, KillOps, VRM);
-          
-          // If the mapped designated register is actually the physreg we have
-          // incoming, we don't need to inserted a dead copy.
-          if (DesignatedReg == PhysReg) {
-            // If this stack slot value is already available, reuse it!
-            if (ReuseSlot > VirtRegMap::MAX_STACK_SLOT)
-              DEBUG(dbgs() << "Reusing RM#"
-                    << ReuseSlot-VirtRegMap::MAX_STACK_SLOT-1);
-            else
-              DEBUG(dbgs() << "Reusing SS#" << ReuseSlot);
-            DEBUG(dbgs() << " from physreg " << TRI->getName(PhysReg)
-                         << " for vreg" << VirtReg
-                         << " instead of reloading into same physreg.\n");
-            unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
-            MI.getOperand(i).setReg(RReg);
-            MI.getOperand(i).setSubReg(0);
-            ReusedOperands.markClobbered(RReg);
-            ++NumReused;
-            continue;
-          }
-          
-          const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-          RegInfo->setPhysRegUsed(DesignatedReg);
-          ReusedOperands.markClobbered(DesignatedReg);
+          // To further complicate matters, we might conflict with an alias,
+          // or R0 and R1 might not be compatible with each other.  In this
+          // case, we actually insert a reload for V1 in R1, ensuring that
+          // we can get at R0 or its alias.
+          ReusedOperands.addReuse(i, ReuseSlot, PhysReg,
+                                  VRM->getPhys(VirtReg), VirtReg);
+          if (isTied)
+            // Only mark it clobbered if this is a use&def operand.
+            ReusedOperands.markClobbered(PhysReg);
+          ++NumReused;
 
-          // Back-schedule reloads and remats.
-          MachineBasicBlock::iterator InsertLoc =
-            ComputeReloadLoc(&MI, MBB.begin(), PhysReg, TRI, DoReMat,
-                             SSorRMId, TII, MF);
+          if (MI.getOperand(i).isKill() &&
+              ReuseSlot <= VirtRegMap::MAX_STACK_SLOT) {
 
-          TII->copyRegToReg(MBB, InsertLoc, DesignatedReg, PhysReg, RC, RC);
+            // The store of this spilled value is potentially dead, but we
+            // won't know for certain until we've confirmed that the re-use
+            // above is valid, which means waiting until the other operands
+            // are processed. For now we just track the spill slot, we'll
+            // remove it after the other operands are processed if valid.
 
-          MachineInstr *CopyMI = prior(InsertLoc);
-          CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
-          UpdateKills(*CopyMI, TRI, RegKills, KillOps);
+            PotentialDeadStoreSlots.push_back(ReuseSlot);
+          }
+
+          // Mark is isKill if it's there no other uses of the same virtual
+          // register and it's not a two-address operand. IsKill will be
+          // unset if reg is reused.
+          if (!isTied && KilledMIRegs.count(VirtReg) == 0) {
+            MI.getOperand(i).setIsKill();
+            KilledMIRegs.insert(VirtReg);
+          }
 
-          // This invalidates DesignatedReg.
-          Spills.ClobberPhysReg(DesignatedReg);
-          
-          Spills.addAvailable(ReuseSlot, DesignatedReg);
-          unsigned RReg =
-            SubIdx ? TRI->getSubReg(DesignatedReg, SubIdx) : DesignatedReg;
-          MI.getOperand(i).setReg(RReg);
-          MI.getOperand(i).setSubReg(0);
-          DEBUG(dbgs() << '\t' << *prior(MII));
-          ++NumReused;
           continue;
-        } // if (PhysReg)
-        
-        // Otherwise, reload it and remember that we have it.
-        PhysReg = VRM.getPhys(VirtReg);
-        assert(PhysReg && "Must map virtreg to physreg!");
+        }  // CanReuse
+
+        // Otherwise we have a situation where we have a two-address instruction
+        // whose mod/ref operand needs to be reloaded.  This reload is already
+        // available in some register "PhysReg", but if we used PhysReg as the
+        // operand to our 2-addr instruction, the instruction would modify
+        // PhysReg.  This isn't cool if something later uses PhysReg and expects
+        // to get its initial value.
+        //
+        // To avoid this problem, and to avoid doing a load right after a store,
+        // we emit a copy from PhysReg into the designated register for this
+        // operand.
+        unsigned DesignatedReg = VRM->getPhys(VirtReg);
+        assert(DesignatedReg && "Must map virtreg to physreg!");
 
         // Note that, if we reused a register for a previous operand, the
         // register we want to reload into might not actually be
         // available.  If this occurs, use the register indicated by the
         // reuser.
         if (ReusedOperands.hasReuses())
-          PhysReg = ReusedOperands.GetRegForReload(VirtReg, PhysReg, &MI, 
-                               Spills, MaybeDeadStores, RegKills, KillOps, VRM);
-        
-        RegInfo->setPhysRegUsed(PhysReg);
-        ReusedOperands.markClobbered(PhysReg);
-        if (AvoidReload)
-          ++NumAvoided;
-        else {
-          // Back-schedule reloads and remats.
-          MachineBasicBlock::iterator InsertLoc =
-            ComputeReloadLoc(MII, MBB.begin(), PhysReg, TRI, DoReMat,
-                             SSorRMId, TII, MF);
+          DesignatedReg = ReusedOperands.
+            GetRegForReload(VirtReg, DesignatedReg, &MI, Spills,
+                            MaybeDeadStores, RegKills, KillOps, *VRM);
+
+        // If the mapped designated register is actually the physreg we have
+        // incoming, we don't need to inserted a dead copy.
+        if (DesignatedReg == PhysReg) {
+          // If this stack slot value is already available, reuse it!
+          if (ReuseSlot > VirtRegMap::MAX_STACK_SLOT)
+            DEBUG(dbgs() << "Reusing RM#"
+                  << ReuseSlot-VirtRegMap::MAX_STACK_SLOT-1);
+          else
+            DEBUG(dbgs() << "Reusing SS#" << ReuseSlot);
+          DEBUG(dbgs() << " from physreg " << TRI->getName(PhysReg)
+                << " for vreg" << VirtReg
+                << " instead of reloading into same physreg.\n");
+          unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
+          MI.getOperand(i).setReg(RReg);
+          MI.getOperand(i).setSubReg(0);
+          ReusedOperands.markClobbered(RReg);
+          ++NumReused;
+          continue;
+        }
 
-          if (DoReMat) {
-            ReMaterialize(MBB, InsertLoc, PhysReg, VirtReg, TII, TRI, VRM);
-          } else {
-            const TargetRegisterClass* RC = RegInfo->getRegClass(VirtReg);
-            TII->loadRegFromStackSlot(MBB, InsertLoc, PhysReg, SSorRMId, RC);
-            MachineInstr *LoadMI = prior(InsertLoc);
-            VRM.addSpillSlotUse(SSorRMId, LoadMI);
-            ++NumLoads;
-            DistanceMap.insert(std::make_pair(LoadMI, Dist++));
-          }
-          // This invalidates PhysReg.
-          Spills.ClobberPhysReg(PhysReg);
+        const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+        MRI->setPhysRegUsed(DesignatedReg);
+        ReusedOperands.markClobbered(DesignatedReg);
 
-          // Any stores to this stack slot are not dead anymore.
-          if (!DoReMat)
-            MaybeDeadStores[SSorRMId] = NULL;
-          Spills.addAvailable(SSorRMId, PhysReg);
-          // Assumes this is the last use. IsKill will be unset if reg is reused
-          // unless it's a two-address operand.
-          if (!MI.isRegTiedToDefOperand(i) &&
-              KilledMIRegs.count(VirtReg) == 0) {
-            MI.getOperand(i).setIsKill();
-            KilledMIRegs.insert(VirtReg);
-          }
+        // Back-schedule reloads and remats.
+        MachineBasicBlock::iterator InsertLoc =
+          ComputeReloadLoc(&MI, MBB->begin(), PhysReg, TRI, DoReMat,
+                           SSorRMId, TII, MF);
 
-          UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
-          DEBUG(dbgs() << '\t' << *prior(InsertLoc));
-        }
-        unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
+        TII->copyRegToReg(*MBB, InsertLoc, DesignatedReg, PhysReg, RC, RC);
+
+        MachineInstr *CopyMI = prior(InsertLoc);
+        CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
+        UpdateKills(*CopyMI, TRI, RegKills, KillOps);
+
+        // This invalidates DesignatedReg.
+        Spills.ClobberPhysReg(DesignatedReg);
+
+        Spills.addAvailable(ReuseSlot, DesignatedReg);
+        unsigned RReg =
+          SubIdx ? TRI->getSubReg(DesignatedReg, SubIdx) : DesignatedReg;
         MI.getOperand(i).setReg(RReg);
         MI.getOperand(i).setSubReg(0);
-      }
+        DEBUG(dbgs() << '\t' << *prior(MII));
+        ++NumReused;
+        continue;
+      } // if (PhysReg)
 
-      // Ok - now we can remove stores that have been confirmed dead.
-      for (unsigned j = 0, e = PotentialDeadStoreSlots.size(); j != e; ++j) {
-        // This was the last use and the spilled value is still available
-        // for reuse. That means the spill was unnecessary!
-        int PDSSlot = PotentialDeadStoreSlots[j];
-        MachineInstr* DeadStore = MaybeDeadStores[PDSSlot];
-        if (DeadStore) {
-          DEBUG(dbgs() << "Removed dead store:\t" << *DeadStore);
-          InvalidateKills(*DeadStore, TRI, RegKills, KillOps);
-          VRM.RemoveMachineInstrFromMaps(DeadStore);
-          MBB.erase(DeadStore);
-          MaybeDeadStores[PDSSlot] = NULL;
-          ++NumDSE;
+        // Otherwise, reload it and remember that we have it.
+      PhysReg = VRM->getPhys(VirtReg);
+      assert(PhysReg && "Must map virtreg to physreg!");
+
+      // Note that, if we reused a register for a previous operand, the
+      // register we want to reload into might not actually be
+      // available.  If this occurs, use the register indicated by the
+      // reuser.
+      if (ReusedOperands.hasReuses())
+        PhysReg = ReusedOperands.GetRegForReload(VirtReg, PhysReg, &MI,
+                    Spills, MaybeDeadStores, RegKills, KillOps, *VRM);
+
+      MRI->setPhysRegUsed(PhysReg);
+      ReusedOperands.markClobbered(PhysReg);
+      if (AvoidReload)
+        ++NumAvoided;
+      else {
+        // Back-schedule reloads and remats.
+        MachineBasicBlock::iterator InsertLoc =
+          ComputeReloadLoc(MII, MBB->begin(), PhysReg, TRI, DoReMat,
+                           SSorRMId, TII, MF);
+
+        if (DoReMat) {
+          ReMaterialize(*MBB, InsertLoc, PhysReg, VirtReg, TII, TRI, *VRM);
+        } else {
+          const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+          TII->loadRegFromStackSlot(*MBB, InsertLoc, PhysReg, SSorRMId, RC);
+          MachineInstr *LoadMI = prior(InsertLoc);
+          VRM->addSpillSlotUse(SSorRMId, LoadMI);
+          ++NumLoads;
+          DistanceMap.insert(std::make_pair(LoadMI, Dist++));
+        }
+        // This invalidates PhysReg.
+        Spills.ClobberPhysReg(PhysReg);
+
+        // Any stores to this stack slot are not dead anymore.
+        if (!DoReMat)
+          MaybeDeadStores[SSorRMId] = NULL;
+        Spills.addAvailable(SSorRMId, PhysReg);
+        // Assumes this is the last use. IsKill will be unset if reg is reused
+        // unless it's a two-address operand.
+        if (!MI.isRegTiedToDefOperand(i) &&
+            KilledMIRegs.count(VirtReg) == 0) {
+          MI.getOperand(i).setIsKill();
+          KilledMIRegs.insert(VirtReg);
         }
+
+        UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
+        DEBUG(dbgs() << '\t' << *prior(InsertLoc));
       }
+      unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
+      MI.getOperand(i).setReg(RReg);
+      MI.getOperand(i).setSubReg(0);
+    }
+
+    // Ok - now we can remove stores that have been confirmed dead.
+    for (unsigned j = 0, e = PotentialDeadStoreSlots.size(); j != e; ++j) {
+      // This was the last use and the spilled value is still available
+      // for reuse. That means the spill was unnecessary!
+      int PDSSlot = PotentialDeadStoreSlots[j];
+      MachineInstr* DeadStore = MaybeDeadStores[PDSSlot];
+      if (DeadStore) {
+        DEBUG(dbgs() << "Removed dead store:\t" << *DeadStore);
+        InvalidateKills(*DeadStore, TRI, RegKills, KillOps);
+        VRM->RemoveMachineInstrFromMaps(DeadStore);
+        MBB->erase(DeadStore);
+        MaybeDeadStores[PDSSlot] = NULL;
+        ++NumDSE;
+      }
+    }
 
 
-      DEBUG(dbgs() << '\t' << MI);
+    DEBUG(dbgs() << '\t' << MI);
 
 
-      // If we have folded references to memory operands, make sure we clear all
-      // physical registers that may contain the value of the spilled virtual
-      // register
-      SmallSet<int, 2> FoldedSS;
-      for (tie(I, End) = VRM.getFoldedVirts(&MI); I != End; ) {
-        unsigned VirtReg = I->second.first;
-        VirtRegMap::ModRef MR = I->second.second;
-        DEBUG(dbgs() << "Folded vreg: " << VirtReg << "  MR: " << MR);
+    // If we have folded references to memory operands, make sure we clear all
+    // physical registers that may contain the value of the spilled virtual
+    // register
+    SmallSet<int, 2> FoldedSS;
+    for (tie(I, End) = VRM->getFoldedVirts(&MI); I != End; ) {
+      unsigned VirtReg = I->second.first;
+      VirtRegMap::ModRef MR = I->second.second;
+      DEBUG(dbgs() << "Folded vreg: " << VirtReg << "  MR: " << MR);
 
-        // MI2VirtMap be can updated which invalidate the iterator.
-        // Increment the iterator first.
-        ++I;
-        int SS = VRM.getStackSlot(VirtReg);
-        if (SS == VirtRegMap::NO_STACK_SLOT)
-          continue;
-        FoldedSS.insert(SS);
-        DEBUG(dbgs() << " - StackSlot: " << SS << "\n");
-        
-        // If this folded instruction is just a use, check to see if it's a
-        // straight load from the virt reg slot.
-        if ((MR & VirtRegMap::isRef) && !(MR & VirtRegMap::isMod)) {
-          int FrameIdx;
-          unsigned DestReg = TII->isLoadFromStackSlot(&MI, FrameIdx);
-          if (DestReg && FrameIdx == SS) {
-            // If this spill slot is available, turn it into a copy (or nothing)
-            // instead of leaving it as a load!
-            if (unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SS)) {
-              DEBUG(dbgs() << "Promoted Load To Copy: " << MI);
-              if (DestReg != InReg) {
-                const TargetRegisterClass *RC = RegInfo->getRegClass(VirtReg);
-                TII->copyRegToReg(MBB, &MI, DestReg, InReg, RC, RC);
-                MachineOperand *DefMO = MI.findRegisterDefOperand(DestReg);
-                unsigned SubIdx = DefMO->getSubReg();
-                // Revisit the copy so we make sure to notice the effects of the
-                // operation on the destreg (either needing to RA it if it's 
-                // virtual or needing to clobber any values if it's physical).
-                NextMII = &MI;
-                --NextMII;  // backtrack to the copy.
-                NextMII->setAsmPrinterFlag(MachineInstr::ReloadReuse);
-                // Propagate the sub-register index over.
-                if (SubIdx) {
-                  DefMO = NextMII->findRegisterDefOperand(DestReg);
-                  DefMO->setSubReg(SubIdx);
-                }
-
-                // Mark is killed.
-                MachineOperand *KillOpnd = NextMII->findRegisterUseOperand(InReg);
-                KillOpnd->setIsKill();
-
-                BackTracked = true;
-              } else {
-                DEBUG(dbgs() << "Removing now-noop copy: " << MI);
-                // Unset last kill since it's being reused.
-                InvalidateKill(InReg, TRI, RegKills, KillOps);
-                Spills.disallowClobberPhysReg(InReg);
+      // MI2VirtMap be can updated which invalidate the iterator.
+      // Increment the iterator first.
+      ++I;
+      int SS = VRM->getStackSlot(VirtReg);
+      if (SS == VirtRegMap::NO_STACK_SLOT)
+        continue;
+      FoldedSS.insert(SS);
+      DEBUG(dbgs() << " - StackSlot: " << SS << "\n");
+
+      // If this folded instruction is just a use, check to see if it's a
+      // straight load from the virt reg slot.
+      if ((MR & VirtRegMap::isRef) && !(MR & VirtRegMap::isMod)) {
+        int FrameIdx;
+        unsigned DestReg = TII->isLoadFromStackSlot(&MI, FrameIdx);
+        if (DestReg && FrameIdx == SS) {
+          // If this spill slot is available, turn it into a copy (or nothing)
+          // instead of leaving it as a load!
+          if (unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SS)) {
+            DEBUG(dbgs() << "Promoted Load To Copy: " << MI);
+            if (DestReg != InReg) {
+              const TargetRegisterClass *RC = MRI->getRegClass(VirtReg);
+              TII->copyRegToReg(*MBB, &MI, DestReg, InReg, RC, RC);
+              MachineOperand *DefMO = MI.findRegisterDefOperand(DestReg);
+              unsigned SubIdx = DefMO->getSubReg();
+              // Revisit the copy so we make sure to notice the effects of the
+              // operation on the destreg (either needing to RA it if it's
+              // virtual or needing to clobber any values if it's physical).
+              NextMII = &MI;
+              --NextMII;  // backtrack to the copy.
+              NextMII->setAsmPrinterFlag(MachineInstr::ReloadReuse);
+              // Propagate the sub-register index over.
+              if (SubIdx) {
+                DefMO = NextMII->findRegisterDefOperand(DestReg);
+                DefMO->setSubReg(SubIdx);
               }
 
-              InvalidateKills(MI, TRI, RegKills, KillOps);
-              VRM.RemoveMachineInstrFromMaps(&MI);
-              MBB.erase(&MI);
-              Erased = true;
-              goto ProcessNextInst;
+              // Mark is killed.
+              MachineOperand *KillOpnd = NextMII->findRegisterUseOperand(InReg);
+              KillOpnd->setIsKill();
+
+              BackTracked = true;
+            } else {
+              DEBUG(dbgs() << "Removing now-noop copy: " << MI);
+              // Unset last kill since it's being reused.
+              InvalidateKill(InReg, TRI, RegKills, KillOps);
+              Spills.disallowClobberPhysReg(InReg);
             }
-          } else {
-            unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
-            SmallVector<MachineInstr*, 4> NewMIs;
-            if (PhysReg &&
-                TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, false, NewMIs)) {
-              MBB.insert(MII, NewMIs[0]);
+
+            InvalidateKills(MI, TRI, RegKills, KillOps);
+            VRM->RemoveMachineInstrFromMaps(&MI);
+            MBB->erase(&MI);
+            Erased = true;
+            goto ProcessNextInst;
+          }
+        } else {
+          unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
+          SmallVector<MachineInstr*, 4> NewMIs;
+          if (PhysReg &&
+              TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, false, NewMIs)) {
+            MBB->insert(MII, NewMIs[0]);
+            InvalidateKills(MI, TRI, RegKills, KillOps);
+            VRM->RemoveMachineInstrFromMaps(&MI);
+            MBB->erase(&MI);
+            Erased = true;
+            --NextMII;  // backtrack to the unfolded instruction.
+            BackTracked = true;
+            goto ProcessNextInst;
+          }
+        }
+      }
+
+      // If this reference is not a use, any previous store is now dead.
+      // Otherwise, the store to this stack slot is not dead anymore.
+      MachineInstr* DeadStore = MaybeDeadStores[SS];
+      if (DeadStore) {
+        bool isDead = !(MR & VirtRegMap::isRef);
+        MachineInstr *NewStore = NULL;
+        if (MR & VirtRegMap::isModRef) {
+          unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
+          SmallVector<MachineInstr*, 4> NewMIs;
+          // We can reuse this physreg as long as we are allowed to clobber
+          // the value and there isn't an earlier def that has already clobbered
+          // the physreg.
+          if (PhysReg &&
+              !ReusedOperands.isClobbered(PhysReg) &&
+              Spills.canClobberPhysReg(PhysReg) &&
+              !TII->isStoreToStackSlot(&MI, SS)) { // Not profitable!
+            MachineOperand *KillOpnd =
+              DeadStore->findRegisterUseOperand(PhysReg, true);
+            // Note, if the store is storing a sub-register, it's possible the
+            // super-register is needed below.
+            if (KillOpnd && !KillOpnd->getSubReg() &&
+                TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, true,NewMIs)){
+              MBB->insert(MII, NewMIs[0]);
+              NewStore = NewMIs[1];
+              MBB->insert(MII, NewStore);
+              VRM->addSpillSlotUse(SS, NewStore);
               InvalidateKills(MI, TRI, RegKills, KillOps);
-              VRM.RemoveMachineInstrFromMaps(&MI);
-              MBB.erase(&MI);
+              VRM->RemoveMachineInstrFromMaps(&MI);
+              MBB->erase(&MI);
               Erased = true;
+              --NextMII;
               --NextMII;  // backtrack to the unfolded instruction.
               BackTracked = true;
-              goto ProcessNextInst;
+              isDead = true;
+              ++NumSUnfold;
             }
           }
         }
 
-        // If this reference is not a use, any previous store is now dead.
-        // Otherwise, the store to this stack slot is not dead anymore.
-        MachineInstr* DeadStore = MaybeDeadStores[SS];
-        if (DeadStore) {
-          bool isDead = !(MR & VirtRegMap::isRef);
-          MachineInstr *NewStore = NULL;
-          if (MR & VirtRegMap::isModRef) {
-            unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
-            SmallVector<MachineInstr*, 4> NewMIs;
-            // We can reuse this physreg as long as we are allowed to clobber
-            // the value and there isn't an earlier def that has already clobbered
-            // the physreg.
-            if (PhysReg &&
-                !ReusedOperands.isClobbered(PhysReg) &&
-                Spills.canClobberPhysReg(PhysReg) &&
-                !TII->isStoreToStackSlot(&MI, SS)) { // Not profitable!
-              MachineOperand *KillOpnd =
-                DeadStore->findRegisterUseOperand(PhysReg, true);
-              // Note, if the store is storing a sub-register, it's possible the
-              // super-register is needed below.
-              if (KillOpnd && !KillOpnd->getSubReg() &&
-                  TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, true,NewMIs)){
-                MBB.insert(MII, NewMIs[0]);
-                NewStore = NewMIs[1];
-                MBB.insert(MII, NewStore);
-                VRM.addSpillSlotUse(SS, NewStore);
-                InvalidateKills(MI, TRI, RegKills, KillOps);
-                VRM.RemoveMachineInstrFromMaps(&MI);
-                MBB.erase(&MI);
-                Erased = true;
-                --NextMII;
-                --NextMII;  // backtrack to the unfolded instruction.
-                BackTracked = true;
-                isDead = true;
-                ++NumSUnfold;
-              }
+        if (isDead) {  // Previous store is dead.
+          // If we get here, the store is dead, nuke it now.
+          DEBUG(dbgs() << "Removed dead store:\t" << *DeadStore);
+          InvalidateKills(*DeadStore, TRI, RegKills, KillOps);
+          VRM->RemoveMachineInstrFromMaps(DeadStore);
+          MBB->erase(DeadStore);
+          if (!NewStore)
+            ++NumDSE;
+        }
+
+        MaybeDeadStores[SS] = NULL;
+        if (NewStore) {
+          // Treat this store as a spill merged into a copy. That makes the
+          // stack slot value available.
+          VRM->virtFolded(VirtReg, NewStore, VirtRegMap::isMod);
+          goto ProcessNextInst;
+        }
+      }
+
+      // If the spill slot value is available, and this is a new definition of
+      // the value, the value is not available anymore.
+      if (MR & VirtRegMap::isMod) {
+        // Notice that the value in this stack slot has been modified.
+        Spills.ModifyStackSlotOrReMat(SS);
+
+        // If this is *just* a mod of the value, check to see if this is just a
+        // store to the spill slot (i.e. the spill got merged into the copy). If
+        // so, realize that the vreg is available now, and add the store to the
+        // MaybeDeadStore info.
+        int StackSlot;
+        if (!(MR & VirtRegMap::isRef)) {
+          if (unsigned SrcReg = TII->isStoreToStackSlot(&MI, StackSlot)) {
+            assert(TargetRegisterInfo::isPhysicalRegister(SrcReg) &&
+                   "Src hasn't been allocated yet?");
+
+            if (CommuteToFoldReload(MII, VirtReg, SrcReg, StackSlot,
+                                    Spills, RegKills, KillOps, TRI)) {
+              NextMII = llvm::next(MII);
+              BackTracked = true;
+              goto ProcessNextInst;
             }
-          }
 
-          if (isDead) {  // Previous store is dead.
-            // If we get here, the store is dead, nuke it now.
-            DEBUG(dbgs() << "Removed dead store:\t" << *DeadStore);
-            InvalidateKills(*DeadStore, TRI, RegKills, KillOps);
-            VRM.RemoveMachineInstrFromMaps(DeadStore);
-            MBB.erase(DeadStore);
-            if (!NewStore)
-              ++NumDSE;
+            // Okay, this is certainly a store of SrcReg to [StackSlot].  Mark
+            // this as a potentially dead store in case there is a subsequent
+            // store into the stack slot without a read from it.
+            MaybeDeadStores[StackSlot] = &MI;
+
+            // If the stack slot value was previously available in some other
+            // register, change it now.  Otherwise, make the register
+            // available in PhysReg.
+            Spills.addAvailable(StackSlot, SrcReg, MI.killsRegister(SrcReg));
           }
+        }
+      }
+    }
 
-          MaybeDeadStores[SS] = NULL;
-          if (NewStore) {
-            // Treat this store as a spill merged into a copy. That makes the
-            // stack slot value available.
-            VRM.virtFolded(VirtReg, NewStore, VirtRegMap::isMod);
-            goto ProcessNextInst;
+    // Process all of the spilled defs.
+    for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+      MachineOperand &MO = MI.getOperand(i);
+      if (!(MO.isReg() && MO.getReg() && MO.isDef()))
+        continue;
+
+      unsigned VirtReg = MO.getReg();
+      if (!TargetRegisterInfo::isVirtualRegister(VirtReg)) {
+        // Check to see if this is a noop copy.  If so, eliminate the
+        // instruction before considering the dest reg to be changed.
+        // Also check if it's copying from an "undef", if so, we can't
+        // eliminate this or else the undef marker is lost and it will
+        // confuses the scavenger. This is extremely rare.
+        unsigned Src, Dst, SrcSR, DstSR;
+        if (TII->isMoveInstr(MI, Src, Dst, SrcSR, DstSR) && Src == Dst &&
+            !MI.findRegisterUseOperand(Src)->isUndef()) {
+          ++NumDCE;
+          DEBUG(dbgs() << "Removing now-noop copy: " << MI);
+          SmallVector<unsigned, 2> KillRegs;
+          InvalidateKills(MI, TRI, RegKills, KillOps, &KillRegs);
+          if (MO.isDead() && !KillRegs.empty()) {
+            // Source register or an implicit super/sub-register use is killed.
+            assert(KillRegs[0] == Dst ||
+                   TRI->isSubRegister(KillRegs[0], Dst) ||
+                   TRI->isSuperRegister(KillRegs[0], Dst));
+            // Last def is now dead.
+            TransferDeadness(Dist, Src, RegKills, KillOps);
           }
+          VRM->RemoveMachineInstrFromMaps(&MI);
+          MBB->erase(&MI);
+          Erased = true;
+          Spills.disallowClobberPhysReg(VirtReg);
+          goto ProcessNextInst;
+        }
+
+        // If it's not a no-op copy, it clobbers the value in the destreg.
+        Spills.ClobberPhysReg(VirtReg);
+        ReusedOperands.markClobbered(VirtReg);
+
+        // Check to see if this instruction is a load from a stack slot into
+        // a register.  If so, this provides the stack slot value in the reg.
+        int FrameIdx;
+        if (unsigned DestReg = TII->isLoadFromStackSlot(&MI, FrameIdx)) {
+          assert(DestReg == VirtReg && "Unknown load situation!");
+
+          // If it is a folded reference, then it's not safe to clobber.
+          bool Folded = FoldedSS.count(FrameIdx);
+          // Otherwise, if it wasn't available, remember that it is now!
+          Spills.addAvailable(FrameIdx, DestReg, !Folded);
+          goto ProcessNextInst;
         }
 
-        // If the spill slot value is available, and this is a new definition of
-        // the value, the value is not available anymore.
-        if (MR & VirtRegMap::isMod) {
-          // Notice that the value in this stack slot has been modified.
-          Spills.ModifyStackSlotOrReMat(SS);
-          
-          // If this is *just* a mod of the value, check to see if this is just a
-          // store to the spill slot (i.e. the spill got merged into the copy). If
-          // so, realize that the vreg is available now, and add the store to the
-          // MaybeDeadStore info.
-          int StackSlot;
-          if (!(MR & VirtRegMap::isRef)) {
-            if (unsigned SrcReg = TII->isStoreToStackSlot(&MI, StackSlot)) {
-              assert(TargetRegisterInfo::isPhysicalRegister(SrcReg) &&
-                     "Src hasn't been allocated yet?");
-
-              if (CommuteToFoldReload(MBB, MII, VirtReg, SrcReg, StackSlot,
-                                      Spills, RegKills, KillOps, TRI, VRM)) {
-                NextMII = llvm::next(MII);
-                BackTracked = true;
-                goto ProcessNextInst;
-              }
+        continue;
+      }
 
-              // Okay, this is certainly a store of SrcReg to [StackSlot].  Mark
-              // this as a potentially dead store in case there is a subsequent
-              // store into the stack slot without a read from it.
-              MaybeDeadStores[StackSlot] = &MI;
-
-              // If the stack slot value was previously available in some other
-              // register, change it now.  Otherwise, make the register
-              // available in PhysReg.
-              Spills.addAvailable(StackSlot, SrcReg, MI.killsRegister(SrcReg));
-            }
-          }
+      unsigned SubIdx = MO.getSubReg();
+      bool DoReMat = VRM->isReMaterialized(VirtReg);
+      if (DoReMat)
+        ReMatDefs.insert(&MI);
+
+      // The only vregs left are stack slot definitions.
+      int StackSlot = VRM->getStackSlot(VirtReg);
+      const TargetRegisterClass *RC = MRI->getRegClass(VirtReg);
+
+      // If this def is part of a two-address operand, make sure to execute
+      // the store from the correct physical register.
+      unsigned PhysReg;
+      unsigned TiedOp;
+      if (MI.isRegTiedToUseOperand(i, &TiedOp)) {
+        PhysReg = MI.getOperand(TiedOp).getReg();
+        if (SubIdx) {
+          unsigned SuperReg = findSuperReg(RC, PhysReg, SubIdx, TRI);
+          assert(SuperReg && TRI->getSubReg(SuperReg, SubIdx) == PhysReg &&
+                 "Can't find corresponding super-register!");
+          PhysReg = SuperReg;
+        }
+      } else {
+        PhysReg = VRM->getPhys(VirtReg);
+        if (ReusedOperands.isClobbered(PhysReg)) {
+          // Another def has taken the assigned physreg. It must have been a
+          // use&def which got it due to reuse. Undo the reuse!
+          PhysReg = ReusedOperands.GetRegForReload(VirtReg, PhysReg, &MI,
+                      Spills, MaybeDeadStores, RegKills, KillOps, *VRM);
         }
       }
 
-      // Process all of the spilled defs.
-      for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
-        MachineOperand &MO = MI.getOperand(i);
-        if (!(MO.isReg() && MO.getReg() && MO.isDef()))
-          continue;
+      assert(PhysReg && "VR not assigned a physical register?");
+      MRI->setPhysRegUsed(PhysReg);
+      unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
+      ReusedOperands.markClobbered(RReg);
+      MI.getOperand(i).setReg(RReg);
+      MI.getOperand(i).setSubReg(0);
+
+      if (!MO.isDead()) {
+        MachineInstr *&LastStore = MaybeDeadStores[StackSlot];
+        SpillRegToStackSlot(MII, -1, PhysReg, StackSlot, RC, true,
+          LastStore, Spills, ReMatDefs, RegKills, KillOps);
+        NextMII = llvm::next(MII);
 
-        unsigned VirtReg = MO.getReg();
-        if (!TargetRegisterInfo::isVirtualRegister(VirtReg)) {
-          // Check to see if this is a noop copy.  If so, eliminate the
-          // instruction before considering the dest reg to be changed.
-          // Also check if it's copying from an "undef", if so, we can't
-          // eliminate this or else the undef marker is lost and it will
-          // confuses the scavenger. This is extremely rare.
+        // Check to see if this is a noop copy.  If so, eliminate the
+        // instruction before considering the dest reg to be changed.
+        {
           unsigned Src, Dst, SrcSR, DstSR;
-          if (TII->isMoveInstr(MI, Src, Dst, SrcSR, DstSR) && Src == Dst &&
-              !MI.findRegisterUseOperand(Src)->isUndef()) {
+          if (TII->isMoveInstr(MI, Src, Dst, SrcSR, DstSR) && Src == Dst) {
             ++NumDCE;
             DEBUG(dbgs() << "Removing now-noop copy: " << MI);
-            SmallVector<unsigned, 2> KillRegs;
-            InvalidateKills(MI, TRI, RegKills, KillOps, &KillRegs);
-            if (MO.isDead() && !KillRegs.empty()) {
-              // Source register or an implicit super/sub-register use is killed.
-              assert(KillRegs[0] == Dst ||
-                     TRI->isSubRegister(KillRegs[0], Dst) ||
-                     TRI->isSuperRegister(KillRegs[0], Dst));
-              // Last def is now dead.
-              TransferDeadness(&MBB, Dist, Src, RegKills, KillOps, VRM);
-            }
-            VRM.RemoveMachineInstrFromMaps(&MI);
-            MBB.erase(&MI);
+            InvalidateKills(MI, TRI, RegKills, KillOps);
+            VRM->RemoveMachineInstrFromMaps(&MI);
+            MBB->erase(&MI);
             Erased = true;
-            Spills.disallowClobberPhysReg(VirtReg);
+            UpdateKills(*LastStore, TRI, RegKills, KillOps);
             goto ProcessNextInst;
           }
-
-          // If it's not a no-op copy, it clobbers the value in the destreg.
-          Spills.ClobberPhysReg(VirtReg);
-          ReusedOperands.markClobbered(VirtReg);
-   
-          // Check to see if this instruction is a load from a stack slot into
-          // a register.  If so, this provides the stack slot value in the reg.
-          int FrameIdx;
-          if (unsigned DestReg = TII->isLoadFromStackSlot(&MI, FrameIdx)) {
-            assert(DestReg == VirtReg && "Unknown load situation!");
-
-            // If it is a folded reference, then it's not safe to clobber.
-            bool Folded = FoldedSS.count(FrameIdx);
-            // Otherwise, if it wasn't available, remember that it is now!
-            Spills.addAvailable(FrameIdx, DestReg, !Folded);
-            goto ProcessNextInst;
-          }
-              
-          continue;
         }
-
-        unsigned SubIdx = MO.getSubReg();
-        bool DoReMat = VRM.isReMaterialized(VirtReg);
-        if (DoReMat)
-          ReMatDefs.insert(&MI);
-
-        // The only vregs left are stack slot definitions.
-        int StackSlot = VRM.getStackSlot(VirtReg);
-        const TargetRegisterClass *RC = RegInfo->getRegClass(VirtReg);
-
-        // If this def is part of a two-address operand, make sure to execute
-        // the store from the correct physical register.
-        unsigned PhysReg;
-        unsigned TiedOp;
-        if (MI.isRegTiedToUseOperand(i, &TiedOp)) {
-          PhysReg = MI.getOperand(TiedOp).getReg();
-          if (SubIdx) {
-            unsigned SuperReg = findSuperReg(RC, PhysReg, SubIdx, TRI);
-            assert(SuperReg && TRI->getSubReg(SuperReg, SubIdx) == PhysReg &&
-                   "Can't find corresponding super-register!");
-            PhysReg = SuperReg;
-          }
-        } else {
-          PhysReg = VRM.getPhys(VirtReg);
-          if (ReusedOperands.isClobbered(PhysReg)) {
-            // Another def has taken the assigned physreg. It must have been a
-            // use&def which got it due to reuse. Undo the reuse!
-            PhysReg = ReusedOperands.GetRegForReload(VirtReg, PhysReg, &MI, 
-                               Spills, MaybeDeadStores, RegKills, KillOps, VRM);
-          }
-        }
-
-        assert(PhysReg && "VR not assigned a physical register?");
-        RegInfo->setPhysRegUsed(PhysReg);
-        unsigned RReg = SubIdx ? TRI->getSubReg(PhysReg, SubIdx) : PhysReg;
-        ReusedOperands.markClobbered(RReg);
-        MI.getOperand(i).setReg(RReg);
-        MI.getOperand(i).setSubReg(0);
-
-        if (!MO.isDead()) {
-          MachineInstr *&LastStore = MaybeDeadStores[StackSlot];
-          SpillRegToStackSlot(MBB, MII, -1, PhysReg, StackSlot, RC, true,
-                            LastStore, Spills, ReMatDefs, RegKills, KillOps, VRM);
-          NextMII = llvm::next(MII);
-
-          // Check to see if this is a noop copy.  If so, eliminate the
-          // instruction before considering the dest reg to be changed.
-          {
-            unsigned Src, Dst, SrcSR, DstSR;
-            if (TII->isMoveInstr(MI, Src, Dst, SrcSR, DstSR) && Src == Dst) {
-              ++NumDCE;
-              DEBUG(dbgs() << "Removing now-noop copy: " << MI);
-              InvalidateKills(MI, TRI, RegKills, KillOps);
-              VRM.RemoveMachineInstrFromMaps(&MI);
-              MBB.erase(&MI);
-              Erased = true;
-              UpdateKills(*LastStore, TRI, RegKills, KillOps);
-              goto ProcessNextInst;
-            }
-          }
-        }    
       }
+    }
     ProcessNextInst:
-      // Delete dead instructions without side effects.
-      if (!Erased && !BackTracked && isSafeToDelete(MI)) {
-        InvalidateKills(MI, TRI, RegKills, KillOps);
-        VRM.RemoveMachineInstrFromMaps(&MI);
-        MBB.erase(&MI);
-        Erased = true;
-      }
-      if (!Erased)
-        DistanceMap.insert(std::make_pair(&MI, Dist++));
-      if (!Erased && !BackTracked) {
-        for (MachineBasicBlock::iterator II = &MI; II != NextMII; ++II)
-          UpdateKills(*II, TRI, RegKills, KillOps);
-      }
-      MII = NextMII;
+    // Delete dead instructions without side effects.
+    if (!Erased && !BackTracked && isSafeToDelete(MI)) {
+      InvalidateKills(MI, TRI, RegKills, KillOps);
+      VRM->RemoveMachineInstrFromMaps(&MI);
+      MBB->erase(&MI);
+      Erased = true;
+    }
+    if (!Erased)
+      DistanceMap.insert(std::make_pair(&MI, Dist++));
+    if (!Erased && !BackTracked) {
+      for (MachineBasicBlock::iterator II = &MI; II != NextMII; ++II)
+        UpdateKills(*II, TRI, RegKills, KillOps);
     }
-
+    MII = NextMII;
   }
 
-};
-
 }
 
 llvm::VirtRegRewriter* llvm::createVirtRegRewriter() {




From andrewl at lenharth.org  Wed Mar 10 18:18:04 2010
From: andrewl at lenharth.org (Andrew Lenharth)
Date: Thu, 11 Mar 2010 00:18:04 -0000
Subject: [llvm-commits] [poolalloc] r98209 - in /poolalloc/trunk:
 include/dsa/DSGraph.h include/dsa/DSGraphTraits.h include/dsa/DSNode.h
 include/dsa/DataStructure.h lib/DSA/Basic.cpp lib/DSA/BottomUpClosure.cpp
 lib/DSA/DataStructure.cpp lib/DSA/Local.cpp lib/DSA/Printer.cpp
 lib/DSA/Steensgaard.cpp lib/DSA/TopDownClosure.cpp
 lib/PoolAllocate/PASimple.cpp
Message-ID: <20100311001804.782152A6C12C@llvm.org>

Author: alenhar2
Date: Wed Mar 10 18:18:04 2010
New Revision: 98209

URL: http://llvm.org/viewvc/llvm-project?rev=98209&view=rev
Log:
Hold types in global uniqued sets, saves memory and time (despite being inefficient)
Use sorted vector sets some places to save memory and time.



Modified:
    poolalloc/trunk/include/dsa/DSGraph.h
    poolalloc/trunk/include/dsa/DSGraphTraits.h
    poolalloc/trunk/include/dsa/DSNode.h
    poolalloc/trunk/include/dsa/DataStructure.h
    poolalloc/trunk/lib/DSA/Basic.cpp
    poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
    poolalloc/trunk/lib/DSA/DataStructure.cpp
    poolalloc/trunk/lib/DSA/Local.cpp
    poolalloc/trunk/lib/DSA/Printer.cpp
    poolalloc/trunk/lib/DSA/Steensgaard.cpp
    poolalloc/trunk/lib/DSA/TopDownClosure.cpp
    poolalloc/trunk/lib/PoolAllocate/PASimple.cpp

Modified: poolalloc/trunk/include/dsa/DSGraph.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSGraph.h?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSGraph.h (original)
+++ poolalloc/trunk/include/dsa/DSGraph.h Wed Mar 10 18:18:04 2010
@@ -42,7 +42,7 @@
   typedef std::map<const Value*, DSNodeHandle> ValueMapTy;
   ValueMapTy ValueMap;
 
-  typedef std::set<const GlobalValue*> GlobalSetTy;
+  typedef sv::set<const GlobalValue*> GlobalSetTy;
   GlobalSetTy GlobalSet;
 
   EquivalenceClasses<const GlobalValue*> &GlobalECs;
@@ -232,14 +232,17 @@
   /// constructed for.
   const TargetData &TD;
 
+  SuperSet<const Type*>& TypeSS;
+
   void operator=(const DSGraph &); // DO NOT IMPLEMENT
   DSGraph(const DSGraph&);         // DO NOT IMPLEMENT
 public:
   // Create a new, empty, DSGraph.
   DSGraph(EquivalenceClasses<const GlobalValue*> &ECs, const TargetData &td,
+          SuperSet<const Type*>& tss,
           DSGraph *GG = 0) 
     :GlobalsGraph(GG), PrintAuxCalls(false), 
-     ScalarMap(ECs), TD(td)
+     ScalarMap(ECs), TD(td), TypeSS(tss)
   { }
 
   // Copy ctor - If you want to capture the node mapping between the source and
@@ -251,6 +254,7 @@
   // method.
   //
   DSGraph( DSGraph* DSG, EquivalenceClasses<const GlobalValue*> &ECs,
+          SuperSet<const Type*>& tss,
           unsigned CloneFlags = 0);
   ~DSGraph();
 
@@ -263,6 +267,10 @@
     return ScalarMap.getGlobalECs();
   }
 
+  SuperSet<const Type*>& getTypeSS() const {
+    return TypeSS;
+  }
+
   /// getTargetData - Return the TargetData object for the current target.
   ///
   const TargetData &getTargetData() const { return TD; }
@@ -558,7 +566,7 @@
   /// merged with other nodes in the graph.  This is used as the first step of
   /// removeDeadNodes.
   ///
-  void removeTriviallyDeadNodes(bool updateForwarders = false);
+  void removeTriviallyDeadNodes();
 };
 
 
@@ -581,7 +589,11 @@
 
   // NodeMap - A mapping from nodes in the source graph to the nodes that
   // represent them in the destination graph.
-  DSGraph::NodeMapTy NodeMap;
+  // We cannot use a densemap here as references into it are not stable across
+  // insertion
+  typedef std::map<const DSNode*, DSNodeHandle> RCNodeMap;
+  RCNodeMap NodeMap;
+
 public:
   ReachabilityCloner(DSGraph* dest, const DSGraph* src, unsigned cloneFlags,
                      bool _createDest = true)

Modified: poolalloc/trunk/include/dsa/DSGraphTraits.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSGraphTraits.h?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSGraphTraits.h (original)
+++ poolalloc/trunk/include/dsa/DSGraphTraits.h Wed Mar 10 18:18:04 2010
@@ -35,7 +35,7 @@
   DSNodeIterator(NodeTy *N, bool) : Node(N) {         // Create end iterator
     if (N != 0) {
       Offset = N->getSize();
-      if (Offset == 0 && Node->getForwardNode() &&
+      if (Offset == 0 && Node->isForwarding() &&
           Node->isDeadNode())        // Model Forward link
         Offset += 1;
     } else {
@@ -60,8 +60,10 @@
   pointer operator*() const {
     if (Node->isDeadNode())
       return Node->getForwardNode();
-    else
+    else if (Node->hasLink(Offset))
       return Node->getLink(Offset).getNode();
+    else
+      return 0;
   }
   pointer operator->() const { return operator*(); }
 

Modified: poolalloc/trunk/include/dsa/DSNode.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSNode.h?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSNode.h (original)
+++ poolalloc/trunk/include/dsa/DSNode.h Wed Mar 10 18:18:04 2010
@@ -21,6 +21,9 @@
 
 #include <map>
 #include <set>
+#include "dsa/sv/set.h"
+#include "dsa/sv/super_set.h"
+#include "DSGraph.h"
 
 namespace llvm {
 
@@ -36,7 +39,7 @@
 ///
 class DSNode : public ilist_node<DSNode> {
 public:
-  typedef std::map<unsigned, std::set<const Type*> > TyMapTy;
+  typedef std::map<unsigned, SuperSet<const Type*>::setPtr> TyMapTy;
   typedef std::map<unsigned, DSNodeHandle> LinkMapTy;
 
 private:
@@ -77,7 +80,7 @@
 
   /// Globals - The list of global values that are merged into this node.
   ///
-  std::vector<const GlobalValue*> Globals;
+  sv::set<const GlobalValue*> Globals;
 
   void operator=(const DSNode &); // DO NOT IMPLEMENT
   DSNode(const DSNode &);         // DO NOT IMPLEMENT
@@ -219,13 +222,20 @@
   ///
   DSNodeHandle &getLink(unsigned Offset) {
     assert(Offset < getSize() && "Link index is out of range!");
+    assert(!isForwarding() && "Link on a forwarding node");
     return Links[Offset];
-  }
+   }
   const DSNodeHandle &getLink(unsigned Offset) const {
     assert(hasLink(Offset) && "No Link");
+    assert(!isForwarding() && "Link on a forwarding node");
     return Links.find(Offset)->second;
   }
 
+  //unsigned getNumLinks() const {
+//     assert(!isForwarding() && "Link on a forwarding node");
+//    return Links.size();
+//  }
+
   /// mergeTypeInfo - This method merges the specified type into the current
   /// node at the specified offset.  This may update the current node's type
   /// record if this gives more information to the node, it may do nothing to
@@ -235,8 +245,9 @@
   ///
   /// FIXME: description
   void mergeTypeInfo(const Type *Ty, unsigned Offset);
-  void mergeTypeInfo(TyMapTy::const_iterator TyIt, unsigned Offset = 0);
- 
+  void mergeTypeInfo(const TyMapTy::mapped_type TyIt, unsigned Offset);
+  void mergeTypeInfo(const DSNode* D, unsigned Offset);
+
   /// foldNodeCompletely - If we determine that this node has some funny
   /// behavior happening to it that we cannot represent, we fold it down to a
   /// single, completely pessimistic, node.  This node is represented as a
@@ -303,7 +314,7 @@
   /// value leaders set that is merged into this node.  Like the getGlobalsList
   /// method, these iterators do not return globals that are part of the
   /// equivalence classes for globals in this node, but aren't leaders.
-  typedef std::vector<const GlobalValue*>::const_iterator globals_iterator;
+  typedef sv::set<const GlobalValue*>::const_iterator globals_iterator;
   globals_iterator globals_begin() const { return Globals.begin(); }
   globals_iterator globals_end() const { return Globals.end(); }
 
@@ -378,6 +389,7 @@
 
   void dropAllReferences() {
     Links.clear();
+    TyMap.clear();
     if (isForwarding())
       ForwardNH.setTo(0, 0);
   }

Modified: poolalloc/trunk/include/dsa/DataStructure.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DataStructure.h?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DataStructure.h (original)
+++ poolalloc/trunk/include/dsa/DataStructure.h Wed Mar 10 18:18:04 2010
@@ -20,6 +20,8 @@
 #include "llvm/ADT/EquivalenceClasses.h"
 
 #include "dsa/EntryPointAnalysis.h"
+#include "dsa/sv/set.h"
+#include "dsa/sv/super_set.h"
 
 #include <map>
 
@@ -37,10 +39,10 @@
 FunctionPass *createDataStructureGraphCheckerPass();
 
 class DataStructures : public ModulePass {
-  typedef std::map<const Instruction*, std::set<const Function*> > ActualCalleesTy;
+  typedef std::map<const Instruction*, sv::set<const Function*> > ActualCalleesTy;
   typedef std::map<const Function*, DSGraph*> DSInfoTy;
 public:
-  typedef std::set<const Function*>::const_iterator callee_iterator;
+  typedef sv::set<const Function*>::const_iterator callee_iterator;
   
 private:
   /// TargetData, comes in handy
@@ -58,9 +60,9 @@
   /// Were are DSGraphs stolen by another pass?
   bool DSGraphsStolen;
 
-  void buildGlobalECs(std::set<const GlobalValue*>& ECGlobals);
+  void buildGlobalECs(sv::set<const GlobalValue*>& ECGlobals);
 
-  void eliminateUsesOfECGlobals(DSGraph& G, const std::set<const GlobalValue*> &ECGlobals);
+  void eliminateUsesOfECGlobals(DSGraph& G, const sv::set<const GlobalValue*> &ECGlobals);
 
   // DSInfo, one graph for each function
   DSInfoTy DSInfo;
@@ -80,6 +82,7 @@
   /// with other global values in the DSGraphs.
   EquivalenceClasses<const GlobalValue*> GlobalECs;
 
+  SuperSet<const Type*>* TypeSS;
 
   void init(DataStructures* D, bool clone, bool printAuxCalls, bool copyGlobalAuxCalls, bool resetAux);
   void init(TargetData* T);
@@ -173,6 +176,8 @@
 
   TargetData& getTargetData() const { return *TD; }
 
+  SuperSet<const Type*>& getTypeSS() const { return *TypeSS; }
+  
   /// deleteValue/copyValue - Interfaces to update the DSGraphs in the program.
   /// These correspond to the interfaces defined in the AliasAnalysis class.
   void deleteValue(Value *V);
@@ -348,7 +353,7 @@
 /// by the bottom-up pass.
 ///
 class TDDataStructures : public DataStructures {
-  std::set<const Function*> ArgsRemainIncomplete;
+  sv::set<const Function*> ArgsRemainIncomplete;
 
   /// CallerCallEdges - For a particular graph, we keep a list of these records
   /// which indicates which graphs call this function and from where.
@@ -400,10 +405,10 @@
 
 private:
   void markReachableFunctionsExternallyAccessible(DSNode *N,
-                                                  std::set<DSNode*> &Visited);
+                                                  sv::set<DSNode*> &Visited);
 
   void InlineCallersIntoGraph(DSGraph* G);
-  void ComputePostOrder(const Function &F, std::set<DSGraph*> &Visited,
+  void ComputePostOrder(const Function &F, sv::set<DSGraph*> &Visited,
                         std::vector<DSGraph*> &PostOrder);
 };
 

Modified: poolalloc/trunk/lib/DSA/Basic.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Basic.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Basic.cpp (original)
+++ poolalloc/trunk/lib/DSA/Basic.cpp Wed Mar 10 18:18:04 2010
@@ -61,7 +61,7 @@
 
   for (Module::iterator F = M.begin(), E = M.end(); F != E; ++F) {
     if (!F->isDeclaration()) {
-      DSGraph* G = new DSGraph(GlobalECs, getTargetData(), GlobalsGraph);
+      DSGraph* G = new DSGraph(GlobalECs, getTargetData(), *TypeSS, GlobalsGraph);
       DSNode * Node = new DSNode(G);
           
       if (!F->hasInternalLinkage())

Modified: poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/BottomUpClosure.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/BottomUpClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/BottomUpClosure.cpp Wed Mar 10 18:18:04 2010
@@ -96,7 +96,7 @@
 
   //finalizeGlobals();
 
-  GlobalsGraph->removeTriviallyDeadNodes(true);
+  GlobalsGraph->removeTriviallyDeadNodes();
   GlobalsGraph->maskIncompleteMarkers();
 
   // Mark external globals incomplete.
@@ -449,7 +449,7 @@
                 E = CalledFuncs.end();
 
         // Start with a copy of the first graph.
-        GI = IndCallGraph.first = new DSGraph(getDSGraph(**I), GlobalECs);
+        GI = IndCallGraph.first = new DSGraph(getDSGraph(**I), GlobalECs, *TypeSS);
         GI->setGlobalsGraph(Graph->getGlobalsGraph());
         std::vector<DSNodeHandle> &Args = IndCallGraph.second;
 
@@ -537,7 +537,10 @@
   // When this graph is finalized, clone the globals in the graph into the
   // globals graph to make sure it has everything, from all graphs.
   DSScalarMap &MainSM = Graph->getScalarMap();
-  ReachabilityCloner RC(GlobalsGraph, Graph, DSGraph::StripAllocaBit);
+  ReachabilityCloner RC(GlobalsGraph, Graph,
+                        DSGraph::DontCloneCallNodes |
+                        DSGraph::DontCloneAuxCallNodes |
+                        DSGraph::StripAllocaBit);
 
   // Clone everything reachable from globals in the function graph into the
   // globals graph.

Modified: poolalloc/trunk/lib/DSA/DataStructure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DataStructure.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DataStructure.cpp Wed Mar 10 18:18:04 2010
@@ -42,8 +42,8 @@
   STATISTIC (NumCallNodesMerged, "Number of call nodes merged");
   STATISTIC (NumNodeAllocated  , "Number of nodes allocated");
   STATISTIC (NumDNE            , "Number of nodes removed by reachability");
-//  STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
-//  STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
+  STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
+  STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
   static cl::opt<unsigned>
   DSAFieldLimit("dsa-field-limit", cl::Hidden,
                 cl::desc("Number of fields to track before collapsing a node"),
@@ -61,7 +61,7 @@
   DEBUG(
         { //assert not looping
           DSNode* NH = N;
-          std::set<DSNode*> seen;
+          sv::set<DSNode*> seen;
           while(NH && NH->isForwarding()) {
             assert(seen.find(NH) == seen.end() && "Loop detected");
             seen.insert(NH);
@@ -81,9 +81,6 @@
   N = Next;
   N->NumReferrers++;
 
-  if (N->isArrayNode())
-    Offset %= N->getSize();
-
   if (N->getSize() <= Offset) {
     assert(N->getSize() <= 1 && "Forwarded to shrunk but not collapsed node?");
     Offset = 0;
@@ -133,9 +130,7 @@
 DSNode::DSNode(const DSNode &N, DSGraph *G, bool NullLinks)
   : NumReferrers(0), Size(N.Size), ParentGraph(G), TyMap(N.TyMap),
     Globals(N.Globals), NodeType(N.NodeType) {
-  if (!NullLinks) {
-    Links = N.Links;
-  }
+  if (!NullLinks) Links = N.Links;
   G->addNode(this);
   ++NumNodeAllocated;
 }
@@ -153,9 +148,10 @@
 
   assert(ParentGraph && "Node has no parent?");
   const DSScalarMap &SM = ParentGraph->getScalarMap();
-  for (unsigned i = 0, e = Globals.size(); i != e; ++i) {
-    assert(SM.global_count(Globals[i]));
-    assert(SM.find(Globals[i])->second.getNode() == this);
+  for (globals_iterator ii = globals_begin(), ee = globals_end();
+       ii != ee; ++ii) {
+    assert(SM.global_count(*ii));
+    assert(SM.find(*ii)->second.getNode() == this);
   }
 }
 
@@ -172,6 +168,24 @@
   NodeType = DeadNode;
   Size = 0;
 
+  DSNodeHandle ToNH(To,Offset);
+
+  //Move the Links
+  for (unsigned x = 0, xe = Links.size(); x != xe; ++x)
+    if (!Links[x].isNull()) {
+      // Compute the offset into the current node at which to
+      // merge this link.  In the common case, this is a linear
+      // relation to the offset in the original node (with
+      // wrapping), but if the current node gets collapsed due to
+      // recursive merging, we must make sure to merge in all remaining
+      // links at offset zero.
+      unsigned MergeOffset = 0;
+      if (ToNH.getNode()->getSize() != 1)
+        MergeOffset = (x + Offset) % ToNH.getNode()->getSize();
+      ToNH.getNode()->addEdgeTo(MergeOffset, Links[x]);
+    }
+  Links.clear();
+
   // Remove this node from the parent graph's Nodes list.
   ParentGraph->unlinkNode(this);
   ParentGraph = 0;
@@ -185,23 +199,19 @@
   // equivalence class.
   GV = getParentGraph()->getScalarMap().getLeaderForGlobal(GV);
 
-  // Keep the list sorted.
-  std::vector<const GlobalValue*>::iterator I =
-    std::lower_bound(Globals.begin(), Globals.end(), GV);
-
-  if (I == Globals.end() || *I != GV) {
-    Globals.insert(I, GV);
-    setGlobalMarker();
-  }
+  Globals.insert(GV);
+  setGlobalMarker();
+}
+
+void DSNode::addFunction(const Function* F) {
+  addGlobal(F);
 }
 
 // removeGlobal - Remove the specified global that is explicitly in the globals
 // list.
 void DSNode::removeGlobal(const GlobalValue *GV) {
-  std::vector<const GlobalValue*>::iterator I =
-    std::lower_bound(Globals.begin(), Globals.end(), GV);
-  assert(I != Globals.end() && *I == GV && "Global not in node!");
-  Globals.erase(I);
+  assert (Globals.count(GV) && "Global not in Node!");
+  Globals.erase(GV);
 }
 
 /// foldNodeCompletely - If we determine that this node has some funny
@@ -216,6 +226,9 @@
 
   ++NumFolds;
 
+  //Collapsed nodes don't really need a type
+  TyMap.clear();
+
   // If this node has a size that is <= 1, we don't need to create a forwarding
   // node.
   if (getSize() <= 1) {
@@ -232,26 +245,9 @@
     DestNode->Size = 1;
     DestNode->Globals.swap(Globals);
 
-    DSNodeHandle NH(DestNode);
-    
     // Start forwarding to the destination node...
     forwardNode(DestNode, 0);
 
-    if (!Links.empty()) {
-
-      // If we have links, merge all of our outgoing links together...
-      for (LinkMapTy::iterator ii = edge_begin(), ee = edge_end();
-           ii != ee; ++ii)
-        NH.getNode()->Links[0].mergeWith(ii->second);
-      Links.clear();
-    }
-    //Merge types
-    if (!TyMap.empty()) {
-      for (TyMapTy::iterator ii = type_begin(), ee = type_end();
-           ii != ee; ++ii)
-        NH.getNode()->mergeTypeInfo(ii);
-      TyMap.clear();
-    }
   }
 }
 
@@ -313,22 +309,36 @@
 
   if (!NewTy || NewTy->isVoidTy()) return;
 
-  if (isCollapsedNode()) Offset = 0;
+  if (isCollapsedNode()) return;
   if (isArrayNode()) Offset %= getSize();
 
   if (Offset >= getSize()) growSize(Offset+1);
 
-  TyMap[Offset].insert(NewTy);
+  TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(TyMap[Offset], NewTy);
 }
 
-void DSNode::mergeTypeInfo(TyMapTy::const_iterator TyIt, unsigned Offset) {
-  Offset += TyIt->first;
-  if (isCollapsedNode()) Offset = 0;
+void DSNode::mergeTypeInfo(const TyMapTy::mapped_type TyIt, unsigned Offset) {
+  if (isCollapsedNode()) return;
   if (isArrayNode()) Offset %= getSize();
 
-  if (Offset >= getSize()) growSize(Offset+1);
+  if (Offset >= getSize())
+    growSize(Offset + 1);
+
+  if (!TyMap[Offset])
+    TyMap[Offset] = TyIt;
+  if (TyIt) {
+    sv::set<const Type*> S(*TyMap[Offset]);
+    S.insert(TyIt->begin(), TyIt->end());
+    TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(S);
+  }
+}
+
+void DSNode::mergeTypeInfo(const DSNode* DN, unsigned Offset) {
+  if (isCollapsedNode()) return;
 
-  TyMap[Offset].insert(TyIt->second.begin(), TyIt->second.end());
+  for (TyMapTy::const_iterator ii = DN->TyMap.begin(), ee = DN->TyMap.end();
+       ii != ee; ++ii)
+    mergeTypeInfo(ii->second, ii->first + Offset);
 }
 
 /// addEdgeTo - Add an edge from the current node to the specified node.  This
@@ -351,12 +361,7 @@
 
 
 void DSNode::mergeGlobals(const DSNode &RHS) {
-  std::vector<const GlobalValue*> Temp;
-  std::back_insert_iterator< std::vector<const GlobalValue*> > back_it (Temp);
-  std::set_union(Globals.begin(), Globals.end(), 
-                 RHS.Globals.begin(), RHS.Globals.end(), 
-                 back_it);
-  Globals.swap(Temp);
+  Globals.insert(RHS.Globals.begin(), RHS.Globals.end());
 }
 
 static unsigned gcd(unsigned m, unsigned n) {
@@ -406,12 +411,6 @@
 #endif
   }
 
-  // Merge the type entries of the two nodes together...
-  for (TyMapTy::iterator ii = NH.getNode()->TyMap.begin(),
-      ee = NH.getNode()->TyMap.end(); ii != ee; ++ii)
-    CurNodeH.getNode()->mergeTypeInfo(ii, NOffset);
-  assert(!CurNodeH.getNode()->isDeadNode());
-
   // If we are merging a node with a completely folded node, then both nodes are
   // now completely folded.
   //
@@ -431,14 +430,18 @@
     NSize = NH.getNode()->getSize();
     NOffset = NH.getOffset();
     assert(NOffset == 0 && NSize == 1);
-  }
-
-
+  } 
 
   DSNode *N = NH.getNode();
   if (CurNodeH.getNode() == N || N == 0) return;
   assert(!CurNodeH.getNode()->isDeadNode());
 
+  // Merge the type entries of the two nodes together...
+  CurNodeH.getNode()->mergeTypeInfo(NH.getNode(), NOffset);
+  if (NH.getNode()->getSize() + NOffset > CurNodeH.getNode()->getSize())
+    CurNodeH.getNode()->growSize(NH.getNode()->getSize() + NOffset);
+  assert(!CurNodeH.getNode()->isDeadNode());
+
   // Merge the NodeType information.
   CurNodeH.getNode()->NodeType |= N->NodeType;
 
@@ -446,27 +449,6 @@
   N->forwardNode(CurNodeH.getNode(), NOffset);
   assert(!CurNodeH.getNode()->isDeadNode());
 
-  // Make all of the outgoing links of N now be outgoing links of CurNodeH.
-  //
-  for (LinkMapTy::iterator ii = N->Links.begin(), ee = N->Links.end();
-       ii != ee; ++ii)
-    if (ii->second.getNode()) {
-      // Compute the offset into the current node at which to
-      // merge this link.  In the common case, this is a linear
-      // relation to the offset in the original node (with
-      // wrapping), but if the current node gets collapsed due to
-      // recursive merging, we must make sure to merge in all remaining
-      // links at offset zero.
-      unsigned MergeOffset = 0;
-      DSNode *CN = CurNodeH.getNode();
-      if (CN->Size != 1)
-        MergeOffset = (ii->first+NOffset) % CN->getSize();
-      CN->addEdgeTo(MergeOffset, ii->second);
-    }
-
-  // Now that there are no outgoing edges, all of the Links are dead.
-  N->Links.clear();
-
   // Merge the globals list...
   CurNodeH.getNode()->mergeGlobals(*N);
 
@@ -538,11 +520,11 @@
   const DSNode *SN = SrcNH.getNode();
 
   DSNodeHandle &NH = NodeMap[SN];
-  if (!NH.isNull()) {   // Node already mapped?
+  if (!NH.isNull()) { // Node already mapped?
     DSNode *NHN = NH.getNode();
-    return DSNodeHandle(NHN, NH.getOffset()+SrcNH.getOffset());
+    return DSNodeHandle(NHN, NH.getOffset() + SrcNH.getOffset());
   }
-
+  
   // If SrcNH has globals and the destination graph has one of the same globals,
   // merge this node with the destination node, which is much more efficient.
   if (SN->globals_begin() != SN->globals_end()) {
@@ -573,8 +555,8 @@
   // the current node may be merged with arbitrary other nodes.  For this
   // reason, we must always go through NH.
   DN = 0;
-  for (DSNode::LinkMapTy::const_iterator ii = SN->edge_begin(),
-          ee = SN->edge_end(); ii != ee; ++ii) {
+  for (DSNode::const_edge_iterator ii = SN->edge_begin(), ee = SN->edge_end();
+       ii != ee; ++ii) {
     const DSNodeHandle &SrcEdge = ii->second;
     if (!SrcEdge.isNull()) {
       const DSNodeHandle &DestEdge = getClonedNH(SrcEdge);
@@ -658,15 +640,12 @@
         }
 #endif
       }
-
       if (DN->getSize() < SN->getSize())
         DN->growSize(SN->getSize());
 
       // Merge the type entries of the two nodes together...
       if (!DN->isNodeCompletelyFolded())
-        for (DSNode::TyMapTy::const_iterator ii = SN->type_begin(),
-             ee = SN->type_end(); ii != ee; ++ii)
-          DN->mergeTypeInfo(ii, NH.getOffset() - SrcNH.getOffset());
+          DN->mergeTypeInfo(SN, NH.getOffset() - SrcNH.getOffset());
 
     }
 
@@ -737,8 +716,8 @@
   // any time, and the current node may be merged with arbitrary other nodes.
   // For this reason, we must always go through NH.
   DN = 0;
-  for (DSNode::LinkMapTy::const_iterator ii = SN->edge_begin(),
-          ee = SN->edge_end(); ii != ee; ++ii) {
+  for (DSNode::const_edge_iterator ii = SN->edge_begin(), ee = SN->edge_end();
+       ii != ee; ++ii) {
     const DSNodeHandle &SrcEdge = ii->second;
     if (!SrcEdge.isNull()) {
       // Compute the offset into the current node at which to
@@ -837,8 +816,9 @@
 
 
 DSGraph::DSGraph(DSGraph* G, EquivalenceClasses<const GlobalValue*> &ECs,
+                 SuperSet<const Type*>& tss,
                  unsigned CloneFlags)
-  : GlobalsGraph(0), ScalarMap(ECs), TD(G->TD) {
+  : GlobalsGraph(0), ScalarMap(ECs), TD(G->TD), TypeSS(tss) {
   PrintAuxCalls = false;
   cloneInto(G, CloneFlags);
 }
@@ -1121,7 +1101,7 @@
   bool AnyDirectSuccessorsReachClonedNodes = false;
   for (DSNode::const_edge_iterator EI = N->edge_begin(), EE = N->edge_end();
        EI != EE; ++EI)
-    if (DSNode *Succ = EI->second.getNode()) {
+    if (DSNode * Succ = EI->second.getNode()) {
       std::pair<unsigned, bool> &SuccInfo = VisitForSCCs(Succ);
       if (SuccInfo.first < Min) Min = SuccInfo.first;
       AnyDirectSuccessorsReachClonedNodes |= SuccInfo.second;
@@ -1138,10 +1118,10 @@
 
   // Find out if any direct successors of any node reach cloned nodes.
   if (!AnyDirectSuccessorsReachClonedNodes)
-    for (unsigned i = SCCStack.size()-1; SCCStack[i] != N; --i)
+    for (unsigned i = SCCStack.size() - 1; SCCStack[i] != N; --i)
       for (DSNode::const_edge_iterator EI = N->edge_begin(), EE = N->edge_end();
            EI != EE; ++EI)
-        if (DSNode *N = EI->second.getNode())
+        if (DSNode * N = EI->second.getNode())
           if (NodeInfo[N].second) {
             AnyDirectSuccessorsReachClonedNodes = true;
             goto OutOfLoop;
@@ -1323,9 +1303,9 @@
   N->setIncompleteMarker();
 
   // Recursively process children...
-  for (DSNode::edge_iterator I = N->edge_begin(),E = N->edge_end(); I != E; ++I)
-    if (DSNode *DSN = I->second.getNode())
-      markIncompleteNode(DSN);
+  for (DSNode::edge_iterator ii = N->edge_begin(), ee = N->edge_end();
+       ii != ee; ++ii)
+    markIncompleteNode(ii->second.getNode());
 }
 
 static void markIncomplete(DSCallSite &Call) {
@@ -1563,33 +1543,23 @@
 // other nodes in the graph.  These nodes will all be trivially unreachable, so
 // we don't have to perform any non-trivial analysis here.
 //
-void DSGraph::removeTriviallyDeadNodes(bool updateForwarders) {
-#if 0
-  if (updateForwarders) {
-    /// NOTE: This code is disabled.  This slows down DSA on 177.mesa
-    /// substantially!
-    
-    // Loop over all of the nodes in the graph, calling getNode on each field.
-    // This will cause all nodes to update their forwarding edges, causing
-    // forwarded nodes to be delete-able.
-    for (node_iterator NI = node_begin(), E = node_end(); NI != E; ++NI) {
-      DSNode &N = *NI;
-      for (unsigned l = 0, e = N.getNumLinks(); l != e; ++l)
-        N.getLink(l).getNode();
-    }
-    
-    // NOTE: This code is disabled.  Though it should, in theory, allow us to
-    // remove more nodes down below, the scan of the scalar map is incredibly
-    // expensive for certain programs (with large SCCs).  In the future, if we can
-    // make the scalar map scan more efficient, then we can reenable this.
-
-    // Likewise, forward any edges from the scalar nodes.  While we are at it,
-    // clean house a bit.
-    for (DSScalarMap::iterator I = ScalarMap.begin(), E = ScalarMap.end(); I != E;) {
-      I->second.getNode();
-      ++I;
-    }
-  }
+void DSGraph::removeTriviallyDeadNodes() {
+  /// NOTE: This code is disabled.  This slows down DSA on 177.mesa
+  /// substantially!
+
+  // Loop over all of the nodes in the graph, calling getNode on each field.
+  // This will cause all nodes to update their forwarding edges, causing
+  // forwarded nodes to be delete-able.
+  for (node_iterator NI = node_begin(), E = node_end(); NI != E; ++NI)
+    for (DSNode::edge_iterator ii = NI->edge_begin(), ee = NI->edge_end();
+         ii != ee; ++ii)
+      ii->second.getNode();
+
+  // Likewise, forward any edges from the scalar nodes.  While we are at it,
+  // clean house a bit.
+  for (DSScalarMap::iterator I = ScalarMap.begin(), E = ScalarMap.end();
+       I != E; ++I)
+    I->second.getNode();
 
   bool isGlobalsGraph = !GlobalsGraph;
 
@@ -1641,7 +1611,7 @@
       ++NI;
     }
   }
-#endif
+
   removeIdenticalCalls(FunctionCalls);
   removeIdenticalCalls(AuxFunctionCalls);
 }
@@ -1653,8 +1623,8 @@
 ///
 void DSNode::markReachableNodes(DenseSet<const DSNode*> &ReachableNodes) const {
   if (this == 0) return;
-  assert(getForwardNode() == 0 && "Cannot mark a forwarded node!");
-  if (ReachableNodes.insert(this).second)        // Is newly reachable?
+  assert(!isForwarding() && "Cannot mark a forwarded node!");
+  if (ReachableNodes.insert(this).second) // Is newly reachable?
     for (DSNode::const_edge_iterator I = edge_begin(), E = edge_end();
          I != E; ++I)
       I->second.getNode()->markReachableNodes(ReachableNodes);
@@ -1677,7 +1647,7 @@
                                DenseSet<const DSNode*> &Visited,
                                bool IgnoreGlobals) {
   if (N == 0) return false;
-  assert(N->getForwardNode() == 0 && "Cannot mark a forwarded node!");
+  assert(N->isForwarding() == 0 && "Cannot mark a forwarded node!");
 
   // If this is a global node, it will end up in the globals graph anyway, so we
   // don't need to worry about it.
@@ -2098,7 +2068,7 @@
   if (Function *FromF = dyn_cast<Function>(From)) {
     Function *ToF = cast<Function>(To);
     assert(!DSInfo.count(ToF) && "New Function already exists!");
-    DSGraph *NG = new DSGraph(getDSGraph(*FromF), GlobalECs);
+    DSGraph *NG = new DSGraph(getDSGraph(*FromF), GlobalECs, *TypeSS);
     DSInfo[ToF] = NG;
     assert(NG->getReturnNodes().size() == 1 && "Cannot copy SCC's yet!");
     
@@ -2127,9 +2097,9 @@
     //Clone or Steal the Source Graph
     DSGraph* BaseGraph = GraphSource->getDSGraph(*F);
     if (Clone) {
-      G = new DSGraph(BaseGraph, GlobalECs, DSGraph::DontCloneAuxCallNodes);
+      G = new DSGraph(BaseGraph, GlobalECs, *TypeSS, DSGraph::DontCloneAuxCallNodes);
     } else {
-      G = new DSGraph(GlobalECs, GraphSource->getTargetData());
+      G = new DSGraph(GlobalECs, GraphSource->getTargetData(), *TypeSS);
       G->spliceFrom(BaseGraph);
       if (resetAuxCalls) 
         G->getAuxFunctionCalls() = G->getFunctionCalls();
@@ -2151,7 +2121,7 @@
 void DataStructures::formGlobalECs() {
   // Grow the equivalence classes for the globals to include anything that we
   // now know to be aliased.
-  std::set<const GlobalValue*> ECGlobals;
+  sv::set<const GlobalValue*> ECGlobals;
   buildGlobalECs(ECGlobals);
   if (!ECGlobals.empty()) {
     DEBUG(errs() << "Eliminating " << ECGlobals.size() << " EC Globals!\n");
@@ -2166,7 +2136,7 @@
 /// apart.  Instead of maintaining this information in all of the graphs
 /// throughout the entire program, store only a single global (the "leader") in
 /// the graphs, and build equivalence classes for the rest of the globals.
-void DataStructures::buildGlobalECs(std::set<const GlobalValue*> &ECGlobals) {
+void DataStructures::buildGlobalECs(sv::set<const GlobalValue*> &ECGlobals) {
   DSScalarMap &SM = GlobalsGraph->getScalarMap();
   EquivalenceClasses<const GlobalValue*> &GlobalECs = SM.getGlobalECs();
   for (DSGraph::node_iterator I = GlobalsGraph->node_begin(), 
@@ -2205,13 +2175,15 @@
 /// really just equivalent to some other globals, remove the globals from the
 /// specified DSGraph (if present), and merge any nodes with their leader nodes.
 void DataStructures::eliminateUsesOfECGlobals(DSGraph &G,
-                                              const std::set<const GlobalValue*> &ECGlobals) {
+                                              const sv::set<const GlobalValue*> &ECGlobals) {
   DSScalarMap &SM = G.getScalarMap();
   EquivalenceClasses<const GlobalValue*> &GlobalECs = SM.getGlobalECs();
 
   bool MadeChange = false;
-  for (DSScalarMap::global_iterator GI = SM.global_begin(), E = SM.global_end();
-       GI != E; ) {
+  std::vector<const GlobalValue*> SMGVV(SM.global_begin(), SM.global_end());
+
+  for (std::vector<const GlobalValue*>::iterator GI = SMGVV.begin(),
+       E = SMGVV.end(); GI != E; ) {
     const GlobalValue *GV = *GI; ++GI;
     if (!ECGlobals.count(GV)) continue;
 
@@ -2253,9 +2225,10 @@
   Clone = clone;
   resetAuxCalls = resetAux;
   TD = D->TD;
+  TypeSS = D->TypeSS;
   ActualCallees = D->ActualCallees;
   GlobalECs = D->getGlobalECs();
-  GlobalsGraph = new DSGraph(D->getGlobalsGraph(), GlobalECs, 
+  GlobalsGraph = new DSGraph(D->getGlobalsGraph(), GlobalECs, *TypeSS,
                              copyGlobalAuxCalls?0:DSGraph::DontCloneAuxCallNodes);
   if (printAuxCalls) GlobalsGraph->setPrintAuxCalls();
 
@@ -2270,7 +2243,8 @@
   GraphSource = 0;
   Clone = false;
   TD = T;
-  GlobalsGraph = new DSGraph(GlobalECs, *T);
+  TypeSS = new SuperSet<const Type*>();
+  GlobalsGraph = new DSGraph(GlobalECs, *T, *TypeSS);
 }
 
 void DataStructures::releaseMemory() {

Modified: poolalloc/trunk/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Local.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Local.cpp (original)
+++ poolalloc/trunk/lib/DSA/Local.cpp Wed Mar 10 18:18:04 2010
@@ -169,10 +169,12 @@
       // graph
        if (g.getScalarMap().global_begin() != g.getScalarMap().global_end()) {
         ReachabilityCloner RC(&g, g.getGlobalsGraph(), 0);
-        for (DSScalarMap::global_iterator I = g.getScalarMap().global_begin();
-             I != g.getScalarMap().global_end(); ++I)
+        std::vector<const GlobalValue*> GVV(g.getScalarMap().global_begin(),
+                                            g.getScalarMap().global_end());
+        for (std::vector<const GlobalValue*>::iterator I = GVV.begin();
+             I != GVV.end(); ++I)
           if (const GlobalVariable * GV = dyn_cast<GlobalVariable > (*I))
-            //if (GV->isConstant())
+            if (GV->isConstant())
               RC.merge(g.getNodeForValue(GV), g.getGlobalsGraph()->getNodeForValue(GV));
       }
 
@@ -246,7 +248,7 @@
     N->addGlobal(GV);
     if (GV->isDeclaration())
       N->setExternGlobalMarker();
-  } else if (Constant * C = dyn_cast<Constant > (V)) {
+  } else if (Constant *C = dyn_cast<Constant>(V)) {
     if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
       if (CE->isCast()) {
         if (isa<PointerType>(CE->getOperand(0)->getType()))
@@ -269,6 +271,13 @@
     } else if (isa<UndefValue>(C)) {
       G.eraseNodeForValue(V);
       return 0;
+    } else if (isa<GlobalAlias>(C)) {
+      // XXX: Need more investigation
+      // According to Andrew, DSA is broken on global aliasing, since it does
+      // not handle the aliases of parameters correctly. Here is only a quick
+      // fix for some special cases.
+      NH = getValueDest(cast<GlobalAlias>(C)->getAliasee());
+      return NH;
     } else {
       errs() << "Unknown constant: " << *C << "\n";
       assert(0 && "Unknown constant type!");
@@ -866,7 +875,7 @@
     while (count) {
       std::string section;
       msf >> section;
-      std::set<Value*> inSection;
+      sv::set<Value*> inSection;
       for (Module::iterator MI = M.begin(), ME = M.end();
               MI != ME; ++MI)
         if (MI->hasSection() && MI->getSection() == section)
@@ -882,7 +891,7 @@
         Value* V = M.getNamedValue(global);
         if (V) {
           DSNodeHandle& DHV = GlobalsGraph->getNodeForValue(V);
-          for (std::set<Value*>::iterator SI = inSection.begin(),
+          for (sv::set<Value*>::iterator SI = inSection.begin(),
                   SE = inSection.end(); SI != SE; ++SI) {
             DEBUG(errs() << "Merging " << V->getNameStr() << " with "
                     << (*SI)->getNameStr() << "\n");
@@ -931,7 +940,7 @@
   // Calculate all of the graphs...
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
     if (!I->isDeclaration()) {
-      DSGraph* G = new DSGraph(GlobalECs, getTargetData(), GlobalsGraph);
+      DSGraph* G = new DSGraph(GlobalECs, getTargetData(), *TypeSS, GlobalsGraph);
       GraphBuilder GGB(*I, *G, *this);
       G->getAuxFunctionCalls() = G->getFunctionCalls();
       setDSGraph(*I, G);

Modified: poolalloc/trunk/lib/DSA/Printer.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Printer.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Printer.cpp (original)
+++ poolalloc/trunk/lib/DSA/Printer.cpp Wed Mar 10 18:18:04 2010
@@ -64,11 +64,14 @@
       for (DSNode::TyMapTy::const_iterator ii = N->type_begin(),
            ee = N->type_end(); ii != ee; ++ii) {
         OS << ii->first << ": ";
-        for (DSNode::TyMapTy::mapped_type::iterator ni = ii->second.begin(),
-                ne = ii->second.end(); ni != ne; ++ni) {
-          WriteTypeSymbolic(OS, *ni, M);
-          OS << ", ";
-        }
+        if (ii->second)
+          for (sv::set<const Type*>::const_iterator ni = ii->second->begin(),
+               ne = ii->second->end(); ni != ne; ++ni) {
+            WriteTypeSymbolic(OS, *ni, M);
+            OS << ", ";
+          }
+        else
+          OS << "VOID";
         OS << " ";
       }
     else

Modified: poolalloc/trunk/lib/DSA/Steensgaard.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Steensgaard.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Steensgaard.cpp (original)
+++ poolalloc/trunk/lib/DSA/Steensgaard.cpp Wed Mar 10 18:18:04 2010
@@ -55,10 +55,10 @@
   
   // Get a copy for the globals graph.
   DSGraph * GG = DS->getGlobalsGraph();
-  GlobalsGraph = new DSGraph(GG, GG->getGlobalECs());
+  GlobalsGraph = new DSGraph(GG, GG->getGlobalECs(), *TypeSS);
 
   // Create a new, empty, graph...
-  ResultGraph = new DSGraph(GG->getGlobalECs(), getTargetData());
+  ResultGraph = new DSGraph(GG->getGlobalECs(), getTargetData(), *TypeSS);
   ResultGraph->setGlobalsGraph(GlobalsGraph);
   // ResultGraph->spliceFrom(DS->getGlobalsGraph());
 
@@ -132,7 +132,7 @@
 
   ResultGraph->removeDeadNodes(DSGraph::KeepUnreachableGlobals);
 
-  GlobalsGraph->removeTriviallyDeadNodes(true);
+  GlobalsGraph->removeTriviallyDeadNodes();
   GlobalsGraph->maskIncompleteMarkers();
 
   // Mark external globals incomplete.

Modified: poolalloc/trunk/lib/DSA/TopDownClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/TopDownClosure.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/TopDownClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/TopDownClosure.cpp Wed Mar 10 18:18:04 2010
@@ -54,20 +54,20 @@
 }
 
 void TDDataStructures::markReachableFunctionsExternallyAccessible(DSNode *N,
-                                                   std::set<DSNode*> &Visited) {
+                                                   sv::set<DSNode*> &Visited) {
   if (!N || Visited.count(N)) return;
   Visited.insert(N);
 
-  for (DSNode::LinkMapTy::iterator ii = N->edge_begin(),
-          ee = N->edge_end(); ii != ee; ++ii) {
-    DSNodeHandle &NH = ii->second;
-    if (DSNode *NN = NH.getNode()) {
+  for (DSNode::edge_iterator ii = N->edge_begin(),
+          ee = N->edge_end(); ii != ee; ++ii)
+    if (!ii->second.isNull()) {
+      DSNodeHandle &NH = ii->second;
+      DSNode * NN = NH.getNode();
       std::vector<const Function*> Functions;
       NN->addFullFunctionList(Functions);
       ArgsRemainIncomplete.insert(Functions.begin(), Functions.end());
       markReachableFunctionsExternallyAccessible(NN, Visited);
     }
-  }
 }
 
 
@@ -85,7 +85,7 @@
   // arguments are functions which are reachable by global variables in the
   // globals graph.
   const DSScalarMap &GGSM = GlobalsGraph->getScalarMap();
-  std::set<DSNode*> Visited;
+  sv::set<DSNode*> Visited;
   for (DSScalarMap::global_iterator I=GGSM.global_begin(), E=GGSM.global_end();
        I != E; ++I) {
     DSNode *N = GGSM.find(*I)->second.getNode();
@@ -114,7 +114,7 @@
 
   // We want to traverse the call graph in reverse post-order.  To do this, we
   // calculate a post-order traversal, then reverse it.
-  std::set<DSGraph*> VisitedGraph;
+  sv::set<DSGraph*> VisitedGraph;
   std::vector<DSGraph*> PostOrder;
 
 {TIME_REGION(XXX, "td:Compute postorder");
@@ -156,7 +156,7 @@
 
 
 void TDDataStructures::ComputePostOrder(const Function &F,
-                                        std::set<DSGraph*> &Visited,
+                                        sv::set<DSGraph*> &Visited,
                                         std::vector<DSGraph*> &PostOrder) {
   if (F.isDeclaration()) return;
   DSGraph* G = getOrCreateGraph(&F);
@@ -353,7 +353,7 @@
       IndCallGraph = IndCallRecI->second;
     } else {
       // Otherwise, create a new DSGraph to represent this.
-      IndCallGraph = new DSGraph(DSG->getGlobalECs(), DSG->getTargetData());
+      IndCallGraph = new DSGraph(DSG->getGlobalECs(), DSG->getTargetData(), *TypeSS);
       // Make a nullary dummy call site, which will eventually get some content
       // merged into it.  The actual callee function doesn't matter here, so we
       // just pass it something to keep the ctor happy.

Modified: poolalloc/trunk/lib/PoolAllocate/PASimple.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PASimple.cpp?rev=98209&r1=98208&r2=98209&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PASimple.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PASimple.cpp Wed Mar 10 18:18:04 2010
@@ -127,7 +127,7 @@
   // Merge all of the DSNodes in the DSGraphs.
   //
   GlobalECs = Graphs->getGlobalECs();
-  CombinedDSGraph = new DSGraph (GlobalECs, TD, Graphs->getGlobalsGraph());
+  CombinedDSGraph = new DSGraph (GlobalECs, TD, Graphs->getTypeSS(), Graphs->getGlobalsGraph());
   //CombinedDSGraph.cloneInto (getGlobalsGraph());
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
     if (Graphs->hasDSGraph (*I))




From bob.wilson at apple.com  Wed Mar 10 18:20:49 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 11 Mar 2010 00:20:49 -0000
Subject: [llvm-commits] [llvm] r98210 -
	/llvm/trunk/lib/Target/ARM/ARMSubtarget.h
Message-ID: <20100311002049.3A2DF2A6C12C@llvm.org>

Author: bwilson
Date: Wed Mar 10 18:20:49 2010
New Revision: 98210

URL: http://llvm.org/viewvc/llvm-project?rev=98210&view=rev
Log:
Lower small memcpys to load/stores on Thumb2.
Radar 7686922.

Modified:
    llvm/trunk/lib/Target/ARM/ARMSubtarget.h

Modified: llvm/trunk/lib/Target/ARM/ARMSubtarget.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMSubtarget.h?rev=98210&r1=98209&r2=98210&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMSubtarget.h (original)
+++ llvm/trunk/lib/Target/ARM/ARMSubtarget.h Wed Mar 10 18:20:49 2010
@@ -94,9 +94,9 @@
   /// getMaxInlineSizeThreshold - Returns the maximum memset / memcpy size
   /// that still makes it profitable to inline the call.
   unsigned getMaxInlineSizeThreshold() const {
-    // FIXME: For now, we don't lower memcpy's to loads / stores for Thumb.
-    // Change this once Thumb ldmia / stmia support is added.
-    return isThumb() ? 0 : 64;
+    // FIXME: For now, we don't lower memcpy's to loads / stores for Thumb1.
+    // Change this once Thumb1 ldmia / stmia support is added.
+    return isThumb1Only() ? 0 : 64;
   }
   /// ParseSubtargetFeatures - Parses features string setting specified
   /// subtarget options.  Definition of function is auto generated by tblgen.




From sabre at nondot.org  Wed Mar 10 18:22:57 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 00:22:57 -0000
Subject: [llvm-commits] [llvm] r98212 - in /llvm/trunk:
 docs/CodeGenerator.html docs/LangRef.html include/llvm/CallingConv.h
 lib/Target/X86/X86CallingConv.td lib/Target/X86/X86FastISel.cpp
 lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86RegisterInfo.cpp
 test/CodeGen/X86/ghc-cc.ll test/CodeGen/X86/ghc-cc64.ll
Message-ID: <20100311002258.47BE92A6C12C@llvm.org>

Author: lattner
Date: Wed Mar 10 18:22:57 2010
New Revision: 98212

URL: http://llvm.org/viewvc/llvm-project?rev=98212&view=rev
Log:
add support, testcases, and dox for the new GHC calling
convention.  Patch by David Terei!


Added:
    llvm/trunk/test/CodeGen/X86/ghc-cc.ll
    llvm/trunk/test/CodeGen/X86/ghc-cc64.ll
Modified:
    llvm/trunk/docs/CodeGenerator.html
    llvm/trunk/docs/LangRef.html
    llvm/trunk/include/llvm/CallingConv.h
    llvm/trunk/lib/Target/X86/X86CallingConv.td
    llvm/trunk/lib/Target/X86/X86FastISel.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp

Modified: llvm/trunk/docs/CodeGenerator.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/CodeGenerator.html?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/docs/CodeGenerator.html (original)
+++ llvm/trunk/docs/CodeGenerator.html Wed Mar 10 18:22:57 2010
@@ -1679,7 +1679,8 @@
    supported on x86/x86-64 and PowerPC. It is performed if:</p>
 
 <ul>
-  <li>Caller and callee have the calling convention <tt>fastcc</tt>.</li>
+  <li>Caller and callee have the calling convention <tt>fastcc</tt> or
+       <tt>cc 10</tt> (GHC call convention).</li>
 
   <li>The call is a tail call - in tail position (ret immediately follows call
       and ret uses value of call or is void).</li>

Modified: llvm/trunk/docs/LangRef.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/LangRef.html?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/docs/LangRef.html (original)
+++ llvm/trunk/docs/LangRef.html Wed Mar 10 18:22:57 2010
@@ -691,9 +691,9 @@
       target, without having to conform to an externally specified ABI
       (Application Binary Interface).
       <a href="CodeGenerator.html#tailcallopt">Tail calls can only be optimized
-      when this convention is used.</a>  This calling convention does not
-      support varargs and requires the prototype of all callees to exactly match
-      the prototype of the function definition.</dd>
+      when this or the GHC convention is used.</a>  This calling convention
+      does not support varargs and requires the prototype of all callees to
+      exactly match the prototype of the function definition.</dd>
 
   <dt><b>"<tt>coldcc</tt>" - The cold calling convention</b>:</dt>
   <dd>This calling convention attempts to make code in the caller as efficient
@@ -703,6 +703,26 @@
       does not support varargs and requires the prototype of all callees to
       exactly match the prototype of the function definition.</dd>
 
+  <dt><b>"<tt>cc <em>10</em></tt>" - GHC convention</b>:</dt>
+  <dd>This calling convention has been implemented specifically for use by the
+      <a href="http://www.haskell.org/ghc">Glasgow Haskell Compiler (GHC)</a>.
+      It passes everything in registers, going to extremes to achieve this by
+      disabling callee save registers. This calling convention should not be
+      used lightly but only for specific situations such as an alternative to
+      the <em>register pinning</em> performance technique often used when
+      implementing functional programming languages.At the moment only X86
+      supports this convention and it has the following limitations:
+      <ul>
+        <li>On <em>X86-32</em> only supports up to 4 bit type parameters. No
+            floating point types are supported.</li>
+        <li>On <em>X86-64</em> only supports up to 10 bit type parameters and
+            6 floating point parameters.</li>
+      </ul>
+      This calling convention supports
+      <a href="CodeGenerator.html#tailcallopt">tail call optimization</a> but
+      requires both the caller and callee are using it.
+  </dd>
+
   <dt><b>"<tt>cc &lt;<em>n</em>&gt;</tt>" - Numbered convention</b>:</dt>
   <dd>Any calling convention may be specified by number, allowing
       target-specific calling conventions to be used.  Target specific calling

Modified: llvm/trunk/include/llvm/CallingConv.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CallingConv.h?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CallingConv.h (original)
+++ llvm/trunk/include/llvm/CallingConv.h Wed Mar 10 18:22:57 2010
@@ -44,6 +44,9 @@
     // call does not break any live ranges in the caller side.
     Cold = 9,
 
+    // GHC - Calling convention used by the Glasgow Haskell Compiler (GHC).
+    GHC = 10,
+
     // Target - This is the start of the target-specific calling conventions,
     // e.g. fastcall and thiscall on X86.
     FirstTargetCC = 64,

Modified: llvm/trunk/lib/Target/X86/X86CallingConv.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86CallingConv.td?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86CallingConv.td (original)
+++ llvm/trunk/lib/Target/X86/X86CallingConv.td Wed Mar 10 18:22:57 2010
@@ -221,6 +221,20 @@
   CCIfType<[v8i8, v4i16, v2i32, v1i64], CCAssignToStack<8, 8>>
 ]>;
 
+def CC_X86_64_GHC : CallingConv<[
+  // Promote i8/i16/i32 arguments to i64.
+  CCIfType<[i8, i16, i32], CCPromoteToType<i64>>,
+
+  // Pass in STG registers: Base, Sp, Hp, R1, R2, R3, R4, R5, R6, SpLim
+  CCIfType<[i64],
+            CCAssignToReg<[R13, RBP, R12, RBX, R14, RSI, RDI, R8, R9, R15]>>,
+
+  // Pass in STG registers: F1, F2, F3, F4, D1, D2
+  CCIfType<[f32, f64, v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
+            CCIfSubtarget<"hasSSE1()",
+            CCAssignToReg<[XMM1, XMM2, XMM3, XMM4, XMM5, XMM6]>>>
+]>;
+
 //===----------------------------------------------------------------------===//
 // X86 C Calling Convention
 //===----------------------------------------------------------------------===//
@@ -320,3 +334,11 @@
   // Otherwise, same as everything else.
   CCDelegateTo<CC_X86_32_Common>
 ]>;
+
+def CC_X86_32_GHC : CallingConv<[
+  // Promote i8/i16 arguments to i32.
+  CCIfType<[i8, i16], CCPromoteToType<i32>>,
+
+  // Pass in STG registers: Base, Sp, Hp, R1
+  CCIfType<[i32], CCAssignToReg<[EBX, EBP, EDI, ESI]>>
+]>;

Modified: llvm/trunk/lib/Target/X86/X86FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FastISel.cpp?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FastISel.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86FastISel.cpp Wed Mar 10 18:22:57 2010
@@ -172,7 +172,9 @@
 CCAssignFn *X86FastISel::CCAssignFnForCall(CallingConv::ID CC,
                                            bool isTaillCall) {
   if (Subtarget->is64Bit()) {
-    if (Subtarget->isTargetWin64())
+    if (CC == CallingConv::GHC)
+      return CC_X86_64_GHC;
+    else if (Subtarget->isTargetWin64())
       return CC_X86_Win64_C;
     else
       return CC_X86_64_C;
@@ -182,6 +184,8 @@
     return CC_X86_32_FastCall;
   else if (CC == CallingConv::Fast)
     return CC_X86_32_FastCC;
+  else if (CC == CallingConv::GHC)
+    return CC_X86_32_GHC;
   else
     return CC_X86_32_C;
 }

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Wed Mar 10 18:22:57 2010
@@ -1378,6 +1378,8 @@
     return !Subtarget->is64Bit();
   case CallingConv::Fast:
     return GuaranteedTailCallOpt;
+  case CallingConv::GHC:
+    return GuaranteedTailCallOpt;
   }
 }
 
@@ -1385,7 +1387,9 @@
 /// given CallingConvention value.
 CCAssignFn *X86TargetLowering::CCAssignFnForNode(CallingConv::ID CC) const {
   if (Subtarget->is64Bit()) {
-    if (Subtarget->isTargetWin64())
+    if (CC == CallingConv::GHC)
+      return CC_X86_64_GHC;
+    else if (Subtarget->isTargetWin64())
       return CC_X86_Win64_C;
     else
       return CC_X86_64_C;
@@ -1395,6 +1399,8 @@
     return CC_X86_32_FastCall;
   else if (CC == CallingConv::Fast)
     return CC_X86_32_FastCC;
+  else if (CC == CallingConv::GHC)
+    return CC_X86_32_GHC;
   else
     return CC_X86_32_C;
 }
@@ -1412,10 +1418,16 @@
                        /*AlwaysInline=*/true, NULL, 0, NULL, 0);
 }
 
+/// IsTailCallConvention - Return true if the calling convention is one that
+/// supports tail call optimization.
+static bool IsTailCallConvention(CallingConv::ID CC) {
+  return (CC == CallingConv::Fast || CC == CallingConv::GHC);
+}
+
 /// FuncIsMadeTailCallSafe - Return true if the function is being made into
 /// a tailcall target by changing its ABI.
 static bool FuncIsMadeTailCallSafe(CallingConv::ID CC) {
-  return GuaranteedTailCallOpt && CC == CallingConv::Fast;
+  return GuaranteedTailCallOpt && IsTailCallConvention(CC);
 }
 
 SDValue
@@ -1479,8 +1491,8 @@
   bool Is64Bit = Subtarget->is64Bit();
   bool IsWin64 = Subtarget->isTargetWin64();
 
-  assert(!(isVarArg && CallConv == CallingConv::Fast) &&
-         "Var args not supported with calling convention fastcc");
+  assert(!(isVarArg && IsTailCallConvention(CallConv)) &&
+         "Var args not supported with calling convention fastcc or ghc");
 
   // Assign locations to all of the incoming arguments.
   SmallVector<CCValAssign, 16> ArgLocs;
@@ -1683,7 +1695,7 @@
   } else {
     BytesToPopOnReturn  = 0; // Callee pops nothing.
     // If this is an sret function, the return should pop the hidden pointer.
-    if (!Is64Bit && CallConv != CallingConv::Fast && ArgsAreStructReturn(Ins))
+    if (!Is64Bit && !IsTailCallConvention(CallConv) && ArgsAreStructReturn(Ins))
       BytesToPopOnReturn = 4;
   }
 
@@ -1779,8 +1791,8 @@
       ++NumTailCalls;
   }
 
-  assert(!(isVarArg && CallConv == CallingConv::Fast) &&
-         "Var args not supported with calling convention fastcc");
+  assert(!(isVarArg && IsTailCallConvention(CallConv)) &&
+         "Var args not supported with calling convention fastcc or ghc");
 
   // Analyze operands of the call, assigning locations to each operand.
   SmallVector<CCValAssign, 16> ArgLocs;
@@ -1794,7 +1806,7 @@
     // This is a sibcall. The memory operands are available in caller's
     // own caller's stack.
     NumBytes = 0;
-  else if (GuaranteedTailCallOpt && CallConv == CallingConv::Fast)
+  else if (GuaranteedTailCallOpt && IsTailCallConvention(CallConv))
     NumBytes = GetAlignedArgumentStackSize(NumBytes, DAG);
 
   int FPDiff = 0;
@@ -2150,7 +2162,7 @@
   unsigned NumBytesForCalleeToPush;
   if (IsCalleePop(isVarArg, CallConv))
     NumBytesForCalleeToPush = NumBytes;    // Callee pops everything
-  else if (!Is64Bit && CallConv != CallingConv::Fast && IsStructRet)
+  else if (!Is64Bit && !IsTailCallConvention(CallConv) && IsStructRet)
     // If this is a call to a struct-return function, the callee
     // pops the hidden struct pointer, so we have to push it back.
     // This is common for Darwin/X86, Linux & Mingw32 targets.
@@ -2288,14 +2300,14 @@
                                     const SmallVectorImpl<ISD::OutputArg> &Outs,
                                     const SmallVectorImpl<ISD::InputArg> &Ins,
                                                      SelectionDAG& DAG) const {
-  if (CalleeCC != CallingConv::Fast &&
+  if (!IsTailCallConvention(CalleeCC) &&
       CalleeCC != CallingConv::C)
     return false;
 
   // If -tailcallopt is specified, make fastcc functions tail-callable.
   const Function *CallerF = DAG.getMachineFunction().getFunction();
   if (GuaranteedTailCallOpt) {
-    if (CalleeCC == CallingConv::Fast &&
+    if (IsTailCallConvention(CalleeCC) &&
         CallerF->getCallingConv() == CalleeCC)
       return true;
     return false;

Modified: llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp?rev=98212&r1=98211&r2=98212&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86RegisterInfo.cpp Wed Mar 10 18:22:57 2010
@@ -294,13 +294,20 @@
 const unsigned *
 X86RegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
   bool callsEHReturn = false;
+  bool ghcCall = false;
 
   if (MF) {
     const MachineFrameInfo *MFI = MF->getFrameInfo();
     const MachineModuleInfo *MMI = MFI->getMachineModuleInfo();
     callsEHReturn = (MMI ? MMI->callsEHReturn() : false);
+    const Function *F = MF->getFunction();
+    ghcCall = (F ? F->getCallingConv() == CallingConv::GHC : false);
   }
 
+  static const unsigned GhcCalleeSavedRegs[] = {
+    0
+  };
+
   static const unsigned CalleeSavedRegs32Bit[] = {
     X86::ESI, X86::EDI, X86::EBX, X86::EBP,  0
   };
@@ -326,7 +333,9 @@
     X86::XMM14, X86::XMM15, 0
   };
 
-  if (Is64Bit) {
+  if (ghcCall) {
+    return GhcCalleeSavedRegs;
+  } else if (Is64Bit) {
     if (IsWin64)
       return CalleeSavedRegsWin64;
     else

Added: llvm/trunk/test/CodeGen/X86/ghc-cc.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/ghc-cc.ll?rev=98212&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/ghc-cc.ll (added)
+++ llvm/trunk/test/CodeGen/X86/ghc-cc.ll Wed Mar 10 18:22:57 2010
@@ -0,0 +1,45 @@
+; RUN: llc < %s -tailcallopt -mtriple=i686-linux-gnu | FileCheck %s
+
+; Test the GHC call convention works (x86-32)
+
+ at base = external global i32 ; assigned to register: EBX
+ at sp   = external global i32 ; assigned to register: EBP
+ at hp   = external global i32 ; assigned to register: EDI
+ at r1   = external global i32 ; assigned to register: ESI
+
+define void @zap(i32 %a, i32 %b) nounwind {
+entry:
+  ; CHECK: movl {{[0-9]*}}(%esp), %ebx
+  ; CHECK-NEXT: movl {{[0-9]*}}(%esp), %ebp
+  ; CHECK-NEXT: call addtwo
+  %0 = call cc 10 i32 @addtwo(i32 %a, i32 %b)
+  ; CHECK: call foo
+  call void @foo() nounwind
+  ret void
+}
+
+define cc 10 i32 @addtwo(i32 %x, i32 %y) nounwind {
+entry:
+  ; CHECK: leal (%ebx,%ebp), %eax
+  %0 = add i32 %x, %y
+  ; CHECK-NEXT: ret
+  ret i32 %0
+}
+
+define cc 10 void @foo() nounwind {
+entry:
+  ; CHECK: movl base, %ebx
+  ; CHECK-NEXT: movl sp, %ebp
+  ; CHECK-NEXT: movl hp, %edi
+  ; CHECK-NEXT: movl r1, %esi
+  %0 = load i32* @r1
+  %1 = load i32* @hp
+  %2 = load i32* @sp
+  %3 = load i32* @base
+  ; CHECK: jmp bar
+  tail call cc 10 void @bar( i32 %3, i32 %2, i32 %1, i32 %0 ) nounwind
+  ret void
+}
+
+declare cc 10 void @bar(i32, i32, i32, i32)
+

Added: llvm/trunk/test/CodeGen/X86/ghc-cc64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/ghc-cc64.ll?rev=98212&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/ghc-cc64.ll (added)
+++ llvm/trunk/test/CodeGen/X86/ghc-cc64.ll Wed Mar 10 18:22:57 2010
@@ -0,0 +1,86 @@
+; RUN: llc < %s -tailcallopt -mtriple=x86_64-linux-gnu | FileCheck %s
+
+; Check the GHC call convention works (x86-64)
+
+ at base  = external global i64 ; assigned to register: R13
+ at sp    = external global i64 ; assigned to register: RBP
+ at hp    = external global i64 ; assigned to register: R12
+ at r1    = external global i64 ; assigned to register: RBX
+ at r2    = external global i64 ; assigned to register: R14
+ at r3    = external global i64 ; assigned to register: RSI
+ at r4    = external global i64 ; assigned to register: RDI
+ at r5    = external global i64 ; assigned to register: R8
+ at r6    = external global i64 ; assigned to register: R9
+ at splim = external global i64 ; assigned to register: R15
+
+ at f1 = external global float  ; assigned to register: XMM1
+ at f2 = external global float  ; assigned to register: XMM2
+ at f3 = external global float  ; assigned to register: XMM3
+ at f4 = external global float  ; assigned to register: XMM4
+ at d1 = external global double ; assigned to register: XMM5
+ at d2 = external global double ; assigned to register: XMM6
+
+define void @zap(i64 %a, i64 %b) nounwind {
+entry:
+  ; CHECK:      movq %rdi, %r13
+  ; CHECK-NEXT: movq %rsi, %rbp
+  ; CHECK-NEXT: callq addtwo
+  %0 = call cc 10 i64 @addtwo(i64 %a, i64 %b)
+  ; CHECK:      callq foo
+  call void @foo() nounwind
+  ret void
+}
+
+define cc 10 i64 @addtwo(i64 %x, i64 %y) nounwind {
+entry:
+  ; CHECK:      leaq (%r13,%rbp), %rax
+  %0 = add i64 %x, %y
+  ; CHECK-NEXT: ret
+  ret i64 %0
+}
+
+define cc 10 void @foo() nounwind {
+entry:
+  ; CHECK: movq base(%rip), %r13
+  ; CHECK-NEXT: movq sp(%rip), %rbp
+  ; CHECK-NEXT: movq hp(%rip), %r12
+  ; CHECK-NEXT: movq r1(%rip), %rbx
+  ; CHECK-NEXT: movq r2(%rip), %r14
+  ; CHECK-NEXT: movq r3(%rip), %rsi
+  ; CHECK-NEXT: movq r4(%rip), %rdi
+  ; CHECK-NEXT: movq r5(%rip), %r8
+  ; CHECK-NEXT: movq r6(%rip), %r9
+  ; CHECK-NEXT: movq splim(%rip), %r15
+  ; CHECK-NEXT: movss f1(%rip), %xmm1
+  ; CHECK-NEXT: movss f2(%rip), %xmm2
+  ; CHECK-NEXT: movss f3(%rip), %xmm3
+  ; CHECK-NEXT: movss f4(%rip), %xmm4
+  ; CHECK-NEXT: movsd d1(%rip), %xmm5
+  ; CHECK-NEXT: movsd d2(%rip), %xmm6
+  %0 = load double* @d2
+  %1 = load double* @d1
+  %2 = load float* @f4
+  %3 = load float* @f3
+  %4 = load float* @f2
+  %5 = load float* @f1
+  %6 = load i64* @splim
+  %7 = load i64* @r6
+  %8 = load i64* @r5
+  %9 = load i64* @r4
+  %10 = load i64* @r3
+  %11 = load i64* @r2
+  %12 = load i64* @r1
+  %13 = load i64* @hp
+  %14 = load i64* @sp
+  %15 = load i64* @base
+  ; CHECK: jmp bar
+  tail call cc 10 void @bar( i64 %15, i64 %14, i64 %13, i64 %12, i64 %11,
+                             i64 %10, i64 %9, i64 %8, i64 %7, i64 %6,
+                             float %5, float %4, float %3, float %2, double %1,
+                             double %0 ) nounwind
+  ret void
+}
+
+declare cc 10 void @bar(i64, i64, i64, i64, i64, i64, i64, i64, i64, i64,
+                        float, float, float, float, double, double)
+




From dpatel at apple.com  Wed Mar 10 18:43:54 2010
From: dpatel at apple.com (Devang Patel)
Date: Wed, 10 Mar 2010 16:43:54 -0800
Subject: [llvm-commits] [llvm] r86748 - in /llvm/trunk:
	include/llvm/Analysis/ include/llvm/CodeGen/
	include/llvm/Transforms/Utils/ lib/Analysis/
	lib/CodeGen/AsmPrinter/ lib/CodeGen/SelectionDAG/
	lib/Transforms/Utils/
In-Reply-To: <5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>
	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
	<7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>
	<5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>
Message-ID: <ABB7E6F6-E1AB-48C4-ADEC-4DFE26680107@apple.com>


On Mar 10, 2010, at 3:58 PM, Jeffrey Yasskin wrote:

> On Mon, Mar 8, 2010 at 10:13 AM, Devang Patel <dpatel at apple.com> wrote:
>> Hi Jeffrey,
>> 
>> I'll update the comment. AbstractScopes map is used to keep track of abstract scopes in entire module. AbstractScopesList is used to construct abstract scopes seen while processing a function. This use of list avoids randomness in DIE ordering.
> 
> Thanks.  Is there a test for this ordering stability?

You mean test to reason why we use list instead of map because of ordering requirement ?
-
Devang


From bob.wilson at apple.com  Wed Mar 10 18:46:22 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 11 Mar 2010 00:46:22 -0000
Subject: [llvm-commits] [llvm] r98215 -
	/llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp
Message-ID: <20100311004622.64A242A6C12C@llvm.org>

Author: bwilson
Date: Wed Mar 10 18:46:22 2010
New Revision: 98215

URL: http://llvm.org/viewvc/llvm-project?rev=98215&view=rev
Log:
Fix ARM buildbot breakage.

Modified:
    llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp?rev=98215&r1=98214&r2=98215&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMCodeEmitter.cpp Wed Mar 10 18:46:22 2010
@@ -436,8 +436,8 @@
 void ARMCodeEmitter::emitMOVi2piecesInstruction(const MachineInstr &MI) {
   const MachineOperand &MO0 = MI.getOperand(0);
   const MachineOperand &MO1 = MI.getOperand(1);
-  assert(MO1.isImm() && ARM_AM::getSOImmVal(MO1.getImm()) != -1 &&
-                                            "Not a valid so_imm value!");
+  assert(MO1.isImm() && ARM_AM::isSOImmTwoPartVal(MO1.getImm()) &&
+                                                  "Not a valid so_imm value!");
   unsigned V1 = ARM_AM::getSOImmTwoPartFirst(MO1.getImm());
   unsigned V2 = ARM_AM::getSOImmTwoPartSecond(MO1.getImm());
 




From dpatel at apple.com  Wed Mar 10 18:47:20 2010
From: dpatel at apple.com (Devang Patel)
Date: Wed, 10 Mar 2010 16:47:20 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
	<2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
	<5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>
Message-ID: <13ECEBF5-92DB-43D5-9583-D606C640746B@apple.com>


On Mar 10, 2010, at 4:01 PM, Jeffrey Yasskin wrote:

> On Tue, Mar 9, 2010 at 12:22 PM, Devang Patel <dpatel at apple.com> wrote:
>> 
>> On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:
>> 
>>> Now that nothing uses the result of constructCompileUnit() unless it
>>> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
>>> avoid new'ing the DIE and CompileUnit?
>> 
>> With the introduction of DIFile, we are moving away from "isMain" field.
>> 
>>> Then, if we delete ModuleCU in
>>> endModule(), this memory leak would go away.
>> 
>> We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.
> 
> I can take it. I assume it's correct to simply not create the
> CompileUnits and DIEs whose addresses aren't obviously passed
> anywhere? Or do I need to look out for constructors or other functions
> that escape their addresses surreptitiously?

Now there is only one compile unit so it is simpler. Regarding other DIEs, it is expected that DIEs are created only if they will be emitted in the output. In other words, unused DIEs are not created. It would good to find and fix leaks in this area.
-
Devang


From dalej at apple.com  Wed Mar 10 18:52:12 2010
From: dalej at apple.com (Dale Johannesen)
Date: Thu, 11 Mar 2010 00:52:12 -0000
Subject: [llvm-commits] [llvm] r98216 -
	/llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h
Message-ID: <20100311005212.C7D462A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 18:52:12 2010
New Revision: 98216

URL: http://llvm.org/viewvc/llvm-project?rev=98216&view=rev
Log:
Make sure HasDebugValue is initialized.  This should fix
the buildbot running valgrind.


Modified:
    llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h

Modified: llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h?rev=98216&r1=98215&r2=98216&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h (original)
+++ llvm/trunk/include/llvm/CodeGen/SelectionDAGNodes.h Wed Mar 10 18:52:12 2010
@@ -1383,9 +1383,9 @@
   /// This constructor adds no operands itself; operands can be
   /// set later with InitOperands.
   SDNode(unsigned Opc, const DebugLoc dl, SDVTList VTs)
-    : NodeType(Opc), OperandsNeedDelete(false), SubclassData(0),
-      NodeId(-1), OperandList(0), ValueList(VTs.VTs), UseList(NULL),
-      NumOperands(0), NumValues(VTs.NumVTs),
+    : NodeType(Opc), OperandsNeedDelete(false), HasDebugValue(false),
+      SubclassData(0), NodeId(-1), OperandList(0), ValueList(VTs.VTs),
+      UseList(NULL), NumOperands(0), NumValues(VTs.NumVTs),
       debugLoc(dl) {}
 
   /// InitOperands - Initialize the operands list of this with 1 operand.




From jyasskin at google.com  Wed Mar 10 19:00:01 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Wed, 10 Mar 2010 17:00:01 -0800
Subject: [llvm-commits] [llvm] r86748 - in /llvm/trunk:
	include/llvm/Analysis/ include/llvm/CodeGen/
	include/llvm/Transforms/Utils/ lib/Analysis/
	lib/CodeGen/AsmPrinter/ lib/CodeGen/SelectionDAG/
	lib/Transforms/Utils/
In-Reply-To: <ABB7E6F6-E1AB-48C4-ADEC-4DFE26680107@apple.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>
	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
	<7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>
	<5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>
	<ABB7E6F6-E1AB-48C4-ADEC-4DFE26680107@apple.com>
Message-ID: <bb55930e1003101700v49166153m7d591c1dc6c66c47@mail.gmail.com>

On Wed, Mar 10, 2010 at 4:43 PM, Devang Patel <dpatel at apple.com> wrote:
>
> On Mar 10, 2010, at 3:58 PM, Jeffrey Yasskin wrote:
>
>> On Mon, Mar 8, 2010 at 10:13 AM, Devang Patel <dpatel at apple.com> wrote:
>>> Hi Jeffrey,
>>>
>>> I'll update the comment. AbstractScopes map is used to keep track of abstract scopes in entire module. AbstractScopesList is used to construct abstract scopes seen while processing a function. This use of list avoids randomness in DIE ordering.
>>
>> Thanks. ?Is there a test for this ordering stability?
>
> You mean test to reason why we use list instead of map because of ordering requirement ?

I'm not exactly sure how to parse that question, so sorry if this
isn't an appropriate answer, but I mean something under either test/
or unittests/ that would fail if anyone removed the list or used the
map where they should have used the list. For example, it could llc a
file with several scopes and CHECK that the DW_whatevers appeared in a
particular order. That would overspecify the requirement, I think, and
so would need to be updated any time the logic changed, but it would
fail if the output order depended on anything but the input.

Jeffrey



From dpatel at apple.com  Wed Mar 10 19:06:36 2010
From: dpatel at apple.com (Devang Patel)
Date: Wed, 10 Mar 2010 17:06:36 -0800
Subject: [llvm-commits] [llvm] r86748 - in /llvm/trunk:
	include/llvm/Analysis/ include/llvm/CodeGen/
	include/llvm/Transforms/Utils/ lib/Analysis/
	lib/CodeGen/AsmPrinter/ lib/CodeGen/SelectionDAG/
	lib/Transforms/Utils/
In-Reply-To: <bb55930e1003101700v49166153m7d591c1dc6c66c47@mail.gmail.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>
	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>
	<7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>
	<5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>
	<ABB7E6F6-E1AB-48C4-ADEC-4DFE26680107@apple.com>
	<bb55930e1003101700v49166153m7d591c1dc6c66c47@mail.gmail.com>
Message-ID: <34CF0D4C-9011-493D-93B1-0017CAA34A19@apple.com>


>>> Thanks.  Is there a test for this ordering stability?
>> 
>> You mean test to reason why we use list instead of map because of ordering requirement ?
> 
> I'm not exactly sure how to parse that question, so sorry if this
> isn't an appropriate answer, but I mean something under either test/
> or unittests/ that would fail if anyone removed the list or used the
> map where they should have used the list. For example, it could llc a
> file with several scopes and CHECK that the DW_whatevers appeared in a
> particular order. That would overspecify the requirement, I think, and
> so would need to be updated any time the logic changed, but it would
> fail if the output order depended on anything but the input.

No there is no such test. Such test would check retrieving objects from a pointers' map is not deterministic. And there are tons of places where llvm code base is careful, by using appropriate data structure,  to not introduce nondeterministic behavior. llvm-gcc bootstrap is very sensitive about such things and every one dreads bootstrap failures due to miscompare!

-
Devang


From johnny.chen at apple.com  Wed Mar 10 19:13:36 2010
From: johnny.chen at apple.com (Johnny Chen)
Date: Thu, 11 Mar 2010 01:13:36 -0000
Subject: [llvm-commits] [llvm] r98217 - in /llvm/trunk/lib/Target/ARM:
 ARMInstrThumb2.td AsmPrinter/ARMAsmPrinter.cpp AsmPrinter/ARMInstPrinter.h
Message-ID: <20100311011336.9CC192A6C12C@llvm.org>

Author: johnny
Date: Wed Mar 10 19:13:36 2010
New Revision: 98217

URL: http://llvm.org/viewvc/llvm-project?rev=98217&view=rev
Log:
Added Thumb2 LDRD/STRD pre/post variants for disassembly only.
Plus fixed the encoding of t2LDRDpci such that P = 1 and W = 0 (offset mode).

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=98217&r1=98216&r2=98217&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Wed Mar 10 19:13:36 2010
@@ -150,6 +150,10 @@
   let MIOperandInfo = (ops GPR:$base, i32imm:$offsimm);
 }
 
+def t2am_imm8s4_offset : Operand<i32> {
+  let PrintMethod = "printT2AddrModeImm8s4OffsetOperand";
+}
+
 // t2addrmode_so_reg  := reg + (reg << imm2)
 def t2addrmode_so_reg : Operand<i32>,
                         ComplexPattern<i32, 3, "SelectT2AddrModeSoReg", []> {
@@ -903,7 +907,7 @@
 def t2LDRDi8  : T2Ii8s4<1, 0, 1, (outs GPR:$dst1, GPR:$dst2),
                         (ins t2addrmode_imm8s4:$addr),
                         IIC_iLoadi, "ldrd", "\t$dst1, $addr", []>;
-def t2LDRDpci : T2Ii8s4<?, ?, 1, (outs GPR:$dst1, GPR:$dst2),
+def t2LDRDpci : T2Ii8s4<1, 0, 1, (outs GPR:$dst1, GPR:$dst2),
                         (ins i32imm:$addr), IIC_iLoadi,
                        "ldrd", "\t$dst1, $addr", []> {
   let Inst{19-16} = 0b1111; // Rn
@@ -1105,7 +1109,24 @@
 def t2STRBT  : T2IstT<0b00, "strbt">;
 def t2STRHT  : T2IstT<0b01, "strht">;
 
-// FIXME: ldrd / strd pre / post variants
+// ldrd / strd pre / post variants
+// For disassembly only.
+
+def t2LDRD_PRE  : T2Ii8s4<1, 1, 1, (outs GPR:$dst1, GPR:$dst2),
+                 (ins GPR:$base, t2am_imm8s4_offset:$imm), NoItinerary,
+                 "ldrd", "\t$dst1, $dst2, [$base, $imm]!", []>;
+
+def t2LDRD_POST : T2Ii8s4<0, 1, 1, (outs GPR:$dst1, GPR:$dst2),
+                 (ins GPR:$base, t2am_imm8s4_offset:$imm), NoItinerary,
+                 "ldrd", "\t$dst1, $dst2, [$base], $imm", []>;
+
+def t2STRD_PRE  : T2Ii8s4<1, 1, 0, (outs),
+                 (ins GPR:$src1, GPR:$src2, GPR:$base, t2am_imm8s4_offset:$imm),
+                 NoItinerary, "strd", "\t$src1, $src2, [$base, $imm]!", []>;
+
+def t2STRD_POST : T2Ii8s4<0, 1, 0, (outs),
+                 (ins GPR:$src1, GPR:$src2, GPR:$base, t2am_imm8s4_offset:$imm),
+                 NoItinerary, "strd", "\t$src1, $src2, [$base], $imm", []>;
 
 // T2Ipl (Preload Data/Instruction) signals the memory system of possible future
 // data/instruction access.  These are for disassembly only.

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98217&r1=98216&r2=98217&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Wed Mar 10 19:13:36 2010
@@ -120,6 +120,7 @@
     void printT2AddrModeImm8Operand(const MachineInstr *MI, int OpNum);
     void printT2AddrModeImm8s4Operand(const MachineInstr *MI, int OpNum);
     void printT2AddrModeImm8OffsetOperand(const MachineInstr *MI, int OpNum);
+    void printT2AddrModeImm8s4OffsetOperand(const MachineInstr *MI, int OpNum) {}
     void printT2AddrModeSoRegOperand(const MachineInstr *MI, int OpNum);
 
     void printCPSOptionOperand(const MachineInstr *MI, int OpNum) {}

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h?rev=98217&r1=98216&r2=98217&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMInstPrinter.h Wed Mar 10 19:13:36 2010
@@ -68,6 +68,7 @@
   void printT2AddrModeImm8Operand(const MCInst *MI, unsigned OpNum) {}
   void printT2AddrModeImm8s4Operand(const MCInst *MI, unsigned OpNum) {}
   void printT2AddrModeImm8OffsetOperand(const MCInst *MI, unsigned OpNum) {}
+  void printT2AddrModeImm8s4OffsetOperand(const MCInst *MI, unsigned OpNum) {}
   void printT2AddrModeSoRegOperand(const MCInst *MI, unsigned OpNum) {}
   
   void printCPSOptionOperand(const MCInst *MI, unsigned OpNum) {}




From isanbard at gmail.com  Wed Mar 10 19:18:13 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Thu, 11 Mar 2010 01:18:13 -0000
Subject: [llvm-commits] [llvm] r98218 -
	/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
Message-ID: <20100311011813.B71C02A6C12C@llvm.org>

Author: void
Date: Wed Mar 10 19:18:13 2010
New Revision: 98218

URL: http://llvm.org/viewvc/llvm-project?rev=98218&view=rev
Log:
When outputing a non-lazy pointer for a stub, we may need to fill in the value
for the NLP because the object it's pointing to may be internal to the file.

This seems counter-intuitive, but bear with me. When we place the LSDA into the
TEXT section, the type info pointers need to be indirect and pc-rel. We
accomplish this by using NLPs. However, sometimes the types are local to the
file. GCC gets around this by not using a NLP in this case, but a "regular"
indirection like this:

GCC_except_tbl:
  .long Lfoo-.

__ZTIA:  @ This is local
  ...

Lfoo:
  .long __ZTIA

LLVM prefers NLPs on Darwin. In fact, it's more optimal for load performance to
use them.

Modified:
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=98218&r1=98217&r2=98218&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Wed Mar 10 19:18:13 2010
@@ -1137,15 +1137,16 @@
         // L_foo$stub:
         OutStreamer.EmitLabel(Stubs[i].first);
         //   .indirect_symbol _foo
-        MCSymbol *MCSym = Stubs[i].second.getPointer();
-        OutStreamer.EmitSymbolAttribute(MCSym, MCSA_IndirectSymbol);
+        MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
+        OutStreamer.EmitSymbolAttribute(MCSym.getPointer(),MCSA_IndirectSymbol);
 
-        if (MCSym->isUndefined())
+        if (MCSym.getInt())
           // External to current translation unit.
           OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
         else
           // Internal to current translation unit.
-          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym, OutContext),
+          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym.getPointer(),
+                                                        OutContext),
                                 4/*size*/, 0/*addrspace*/);
       }
 




From echristo at apple.com  Wed Mar 10 19:25:07 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 01:25:07 -0000
Subject: [llvm-commits] [llvm] r98219 - in /llvm/trunk:
 include/llvm/Transforms/Utils/BuildLibCalls.h
 lib/Transforms/InstCombine/InstCombineCalls.cpp
 lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100311012507.61B652A6C12C@llvm.org>

Author: echristo
Date: Wed Mar 10 19:25:07 2010
New Revision: 98219

URL: http://llvm.org/viewvc/llvm-project?rev=98219&view=rev
Log:
Add strncpy libcall creator.  Use it when it should be used.

Modified:
    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h?rev=98219&r1=98218&r2=98219&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h (original)
+++ llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h Wed Mar 10 19:25:07 2010
@@ -39,6 +39,11 @@
   Value *EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
                     const TargetData *TD);
   
+  /// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
+  /// specified pointer arguments and length.
+  Value *EmitStrNCpy(Value *Dst, Value *Src, Value *Len, IRBuilder<> &B,
+                    const TargetData *TD);
+  
   /// EmitMemCpy - Emit a call to the memcpy function to the builder.  This
   /// always expects that the size has type 'intptr_t' and Dst/Src are pointers.
   Value *EmitMemCpy(Value *Dst, Value *Src, Value *Len,

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=98219&r1=98218&r2=98219&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Wed Mar 10 19:25:07 2010
@@ -850,7 +850,8 @@
       return 0;
     if (SizeCI->isAllOnesValue() ||
         SizeCI->getZExtValue() <= SizeArg->getZExtValue()) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
+                               CI->getOperand(3), B, TD);
       return ReplaceInstUsesWith(*CI, Ret);
     }
     return 0; 

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98219&r1=98218&r2=98219&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Wed Mar 10 19:25:07 2010
@@ -87,6 +87,26 @@
   return CI;
 }
 
+/// EmitStrNCpy - Emit a call to the strcpy function to the builder, for the
+/// specified pointer arguments.
+Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,
+                         IRBuilder<> &B, const TargetData *TD) {
+  Module *M = B.GetInsertBlock()->getParent()->getParent();
+  AttributeWithIndex AWI[2];
+  AWI[0] = AttributeWithIndex::get(2, Attribute::NoCapture);
+  AWI[1] = AttributeWithIndex::get(~0u, Attribute::NoUnwind);
+  const Type *I8Ptr = B.getInt8PtrTy();
+  Value *StrNCpy = M->getOrInsertFunction("strncpy", AttrListPtr::get(AWI, 2),
+                                         I8Ptr, I8Ptr, I8Ptr,
+                                         Len->getType(), NULL);
+  CallInst *CI = B.CreateCall3(StrNCpy, CastToCStr(Dst, B), CastToCStr(Src, B),
+                               Len, "strncpy");
+  if (const Function *F = dyn_cast<Function>(StrNCpy->stripPointerCasts()))
+    CI->setCallingConv(F->getCallingConv());
+  return CI;
+}
+
+
 /// EmitMemCpy - Emit a call to the memcpy function to the builder.  This always
 /// expects that the size has type 'intptr_t' and Dst/Src are pointers.
 Value *llvm::EmitMemCpy(Value *Dst, Value *Src, Value *Len,




From daniel at zuster.org  Wed Mar 10 19:34:16 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 01:34:16 -0000
Subject: [llvm-commits] [llvm] r98220 - in /llvm/trunk:
 include/llvm/Target/TargetRegistry.h lib/Target/X86/X86.h
 lib/Target/X86/X86AsmBackend.cpp
Message-ID: <20100311013416.92EDB2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 19:34:16 2010
New Revision: 98220

URL: http://llvm.org/viewvc/llvm-project?rev=98220&view=rev
Log:
MC: Provide the target triple to AsmBackend constructors.

Modified:
    llvm/trunk/include/llvm/Target/TargetRegistry.h
    llvm/trunk/lib/Target/X86/X86.h
    llvm/trunk/lib/Target/X86/X86AsmBackend.cpp

Modified: llvm/trunk/include/llvm/Target/TargetRegistry.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetRegistry.h?rev=98220&r1=98219&r2=98220&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetRegistry.h (original)
+++ llvm/trunk/include/llvm/Target/TargetRegistry.h Wed Mar 10 19:34:16 2010
@@ -66,7 +66,7 @@
                                             MCStreamer &Streamer,
                                             const MCAsmInfo *MAI);
     typedef TargetAsmBackend *(*AsmBackendCtorTy)(const Target &T,
-                                                  MCAssembler &A);
+                                                  const std::string &TT);
     typedef TargetAsmLexer *(*AsmLexerCtorTy)(const Target &T,
                                               const MCAsmInfo &MAI);
     typedef TargetAsmParser *(*AsmParserCtorTy)(const Target &T,MCAsmParser &P);
@@ -208,11 +208,12 @@
 
     /// createAsmBackend - Create a target specific assembly parser.
     ///
+    /// \arg Triple - The target triple string.
     /// \arg Backend - The target independent assembler object.
-    TargetAsmBackend *createAsmBackend(MCAssembler &Backend) const {
+    TargetAsmBackend *createAsmBackend(const std::string &Triple) const {
       if (!AsmBackendCtorFn)
         return 0;
-      return AsmBackendCtorFn(*this, Backend);
+      return AsmBackendCtorFn(*this, Triple);
     }
 
     /// createAsmLexer - Create a target specific assembly lexer.

Modified: llvm/trunk/lib/Target/X86/X86.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86.h?rev=98220&r1=98219&r2=98220&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86.h (original)
+++ llvm/trunk/lib/Target/X86/X86.h Wed Mar 10 19:34:16 2010
@@ -21,7 +21,6 @@
 
 class FunctionPass;
 class JITCodeEmitter;
-class MCAssembler;
 class MCCodeEmitter;
 class MCContext;
 class MachineCodeEmitter;
@@ -57,8 +56,8 @@
 MCCodeEmitter *createX86_64MCCodeEmitter(const Target &, TargetMachine &TM,
                                          MCContext &Ctx);
 
-TargetAsmBackend *createX86_32AsmBackend(const Target &, MCAssembler &);
-TargetAsmBackend *createX86_64AsmBackend(const Target &, MCAssembler &);
+TargetAsmBackend *createX86_32AsmBackend(const Target &, const std::string &);
+TargetAsmBackend *createX86_64AsmBackend(const Target &, const std::string &);
 
 /// createX86EmitCodeToMemory - Returns a pass that converts a register
 /// allocated function into raw machine code in a dynamically

Modified: llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86AsmBackend.cpp?rev=98220&r1=98219&r2=98220&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86AsmBackend.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86AsmBackend.cpp Wed Mar 10 19:34:16 2010
@@ -17,18 +17,18 @@
 
 class X86AsmBackend : public TargetAsmBackend {
 public:
-  X86AsmBackend(const Target &T, MCAssembler &A)
+  X86AsmBackend(const Target &T)
     : TargetAsmBackend(T) {}
 };
 
 }
 
 TargetAsmBackend *llvm::createX86_32AsmBackend(const Target &T,
-                                               MCAssembler &A) {
-  return new X86AsmBackend(T, A);
+                                               const std::string &TT) {
+  return new X86AsmBackend(T);
 }
 
 TargetAsmBackend *llvm::createX86_64AsmBackend(const Target &T,
-                                               MCAssembler &A) {
-  return new X86AsmBackend(T, A);
+                                               const std::string &TT) {
+  return new X86AsmBackend(T);
 }




From daniel at zuster.org  Wed Mar 10 19:34:21 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 01:34:21 -0000
Subject: [llvm-commits] [llvm] r98221 - in /llvm/trunk:
 include/llvm/Target/TargetAsmBackend.h lib/Target/X86/X86AsmBackend.cpp
Message-ID: <20100311013422.02A942A6C12D@llvm.org>

Author: ddunbar
Date: Wed Mar 10 19:34:21 2010
New Revision: 98221

URL: http://llvm.org/viewvc/llvm-project?rev=98221&view=rev
Log:
MC: Sketch some TargetAsmBackend hooks we are going to need.

Modified:
    llvm/trunk/include/llvm/Target/TargetAsmBackend.h
    llvm/trunk/lib/Target/X86/X86AsmBackend.cpp

Modified: llvm/trunk/include/llvm/Target/TargetAsmBackend.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetAsmBackend.h?rev=98221&r1=98220&r2=98221&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetAsmBackend.h (original)
+++ llvm/trunk/include/llvm/Target/TargetAsmBackend.h Wed Mar 10 19:34:21 2010
@@ -28,6 +28,28 @@
 
   const Target &getTarget() const { return TheTarget; }
 
+  /// hasAbsolutizedSet - Check whether this target "absolutizes"
+  /// assignments. That is, given code like:
+  ///   a:
+  ///   ...
+  ///   b:
+  ///   tmp = a - b
+  ///       .long tmp
+  /// will the value of 'tmp' be a relocatable expression, or the assembly time
+  /// value of L0 - L1. This distinction is only relevant for platforms that
+  /// support scattered symbols, since in the absence of scattered symbols (a -
+  /// b) cannot change after assembly.
+  virtual bool hasAbsolutizedSet() const { return false; }
+
+  /// hasScatteredSymbols - Check whether this target supports scattered
+  /// symbols. If so, the assembler should assume that atoms can be scattered by
+  /// the linker. In particular, this means that the offsets between symbols
+  /// which are in distinct atoms is not known at link time, and the assembler
+  /// must generate fixups and relocations appropriately.
+  ///
+  /// Note that the assembler currently does not reason about atoms, instead it
+  /// assumes all temporary symbols reside in the "current atom".
+  virtual bool hasScatteredSymbols() const { return false; }
 };
 
 } // End llvm namespace

Modified: llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86AsmBackend.cpp?rev=98221&r1=98220&r2=98221&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86AsmBackend.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86AsmBackend.cpp Wed Mar 10 19:34:21 2010
@@ -21,14 +21,34 @@
     : TargetAsmBackend(T) {}
 };
 
+class DarwinX86AsmBackend : public X86AsmBackend {
+public:
+  DarwinX86AsmBackend(const Target &T)
+    : X86AsmBackend(T) {}
+
+  virtual bool hasAbsolutizedSet() const { return true; }
+
+  virtual bool hasScatteredSymbols() const { return true; }
+};
+
 }
 
 TargetAsmBackend *llvm::createX86_32AsmBackend(const Target &T,
                                                const std::string &TT) {
-  return new X86AsmBackend(T);
+  switch (Triple(TT).getOS()) {
+  case Triple::Darwin:
+    return new DarwinX86AsmBackend(T);
+  default:
+    return new X86AsmBackend(T);
+  }
 }
 
 TargetAsmBackend *llvm::createX86_64AsmBackend(const Target &T,
                                                const std::string &TT) {
-  return new X86AsmBackend(T);
+  switch (Triple(TT).getOS()) {
+  case Triple::Darwin:
+    return new DarwinX86AsmBackend(T);
+  default:
+    return new X86AsmBackend(T);
+  }
 }




From daniel at zuster.org  Wed Mar 10 19:34:27 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 01:34:27 -0000
Subject: [llvm-commits] [llvm] r98222 - in /llvm/trunk:
 include/llvm/MC/MCAssembler.h include/llvm/MC/MCStreamer.h
 include/llvm/Target/TargetMachine.h lib/CodeGen/LLVMTargetMachine.cpp
 lib/MC/MCAssembler.cpp lib/MC/MCMachOStreamer.cpp tools/llvm-mc/llvm-mc.cpp
Message-ID: <20100311013427.945F62A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 19:34:27 2010
New Revision: 98222

URL: http://llvm.org/viewvc/llvm-project?rev=98222&view=rev
Log:
MC: Provide MCAssembler with a TargetAsmBackend.

Modified:
    llvm/trunk/include/llvm/MC/MCAssembler.h
    llvm/trunk/include/llvm/MC/MCStreamer.h
    llvm/trunk/include/llvm/Target/TargetMachine.h
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/MC/MCAssembler.cpp
    llvm/trunk/lib/MC/MCMachOStreamer.cpp
    llvm/trunk/tools/llvm-mc/llvm-mc.cpp

Modified: llvm/trunk/include/llvm/MC/MCAssembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAssembler.h?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAssembler.h (original)
+++ llvm/trunk/include/llvm/MC/MCAssembler.h Wed Mar 10 19:34:27 2010
@@ -28,6 +28,7 @@
 class MCSection;
 class MCSectionData;
 class MCSymbol;
+class TargetAsmBackend;
 
 /// MCAsmFixup - Represent a fixed size region of bytes inside some fragment
 /// which needs to be rewritten. This region will either be rewritten by the
@@ -582,6 +583,8 @@
 
   MCContext &Context;
 
+  TargetAsmBackend &Backend;
+
   raw_ostream &OS;
 
   iplist<MCSectionData> Sections;
@@ -617,7 +620,7 @@
   // concrete and require clients to pass in a target like object. The other
   // option is to make this abstract, and have targets provide concrete
   // implementations as we do with AsmParser.
-  MCAssembler(MCContext &_Context, raw_ostream &OS);
+  MCAssembler(MCContext &_Context, TargetAsmBackend &_Backend, raw_ostream &OS);
   ~MCAssembler();
 
   MCContext &getContext() const { return Context; }

Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Wed Mar 10 19:34:27 2010
@@ -27,6 +27,7 @@
   class MCSection;
   class MCSymbol;
   class StringRef;
+class TargetAsmBackend;
   class Twine;
   class raw_ostream;
   class formatted_raw_ostream;
@@ -304,18 +305,10 @@
                                 MCCodeEmitter *CE = 0,
                                 bool ShowInst = false);
 
-  // FIXME: These two may end up getting rolled into a single
-  // createObjectStreamer interface, which implements the assembler backend, and
-  // is parameterized on an output object file writer.
-
   /// createMachOStream - Create a machine code streamer which will generative
   /// Mach-O format object files.
-  MCStreamer *createMachOStreamer(MCContext &Ctx, raw_ostream &OS,
-                                  MCCodeEmitter *CE);
-
-  /// createELFStreamer - Create a machine code streamer which will generative
-  /// ELF format object files.
-  MCStreamer *createELFStreamer(MCContext &Ctx, raw_ostream &OS);
+  MCStreamer *createMachOStreamer(MCContext &Ctx, TargetAsmBackend &TAB,
+                                  raw_ostream &OS, MCCodeEmitter *CE);
 
 } // end namespace llvm
 

Modified: llvm/trunk/include/llvm/Target/TargetMachine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetMachine.h?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetMachine.h (original)
+++ llvm/trunk/include/llvm/Target/TargetMachine.h Wed Mar 10 19:34:27 2010
@@ -224,6 +224,8 @@
 /// implemented with the LLVM target-independent code generator.
 ///
 class LLVMTargetMachine : public TargetMachine {
+  std::string TargetTriple;
+
 protected: // Can only create subclasses.
   LLVMTargetMachine(const Target &T, const std::string &TargetTriple);
   

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Wed Mar 10 19:34:27 2010
@@ -94,8 +94,8 @@
     cl::desc("Split GEPs and run no-load GVN"));
 
 LLVMTargetMachine::LLVMTargetMachine(const Target &T,
-                                     const std::string &TargetTriple)
-  : TargetMachine(T) {
+                                     const std::string &Triple)
+  : TargetMachine(T), TargetTriple(Triple) {
   AsmInfo = T.createAsmInfo(TargetTriple);
 }
 
@@ -143,10 +143,11 @@
     // Create the code emitter for the target if it exists.  If not, .o file
     // emission fails.
     MCCodeEmitter *MCE = getTarget().createCodeEmitter(*this, *Context);
-    if (MCE == 0)
+    TargetAsmBackend *TAB = getTarget().createAsmBackend(TargetTriple);
+    if (MCE == 0 || TAB == 0)
       return true;
     
-    AsmStreamer.reset(createMachOStreamer(*Context, Out, MCE));
+    AsmStreamer.reset(createMachOStreamer(*Context, *TAB, Out, MCE));
     
     // Any output to the asmprinter's "O" stream is bad and needs to be fixed,
     // force it to come out stderr.

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Wed Mar 10 19:34:27 2010
@@ -991,8 +991,9 @@
 
 /* *** */
 
-MCAssembler::MCAssembler(MCContext &_Context, raw_ostream &_OS)
-  : Context(_Context), OS(_OS), SubsectionsViaSymbols(false)
+MCAssembler::MCAssembler(MCContext &_Context, TargetAsmBackend &_Backend,
+                         raw_ostream &_OS)
+  : Context(_Context), Backend(_Backend), OS(_OS), SubsectionsViaSymbols(false)
 {
 }
 

Modified: llvm/trunk/lib/MC/MCMachOStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCMachOStreamer.cpp?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCMachOStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCMachOStreamer.cpp Wed Mar 10 19:34:27 2010
@@ -59,8 +59,9 @@
   }
 
 public:
-  MCMachOStreamer(MCContext &Context, raw_ostream &_OS, MCCodeEmitter *_Emitter)
-    : MCStreamer(Context), Assembler(Context, _OS), Emitter(_Emitter),
+  MCMachOStreamer(MCContext &Context, TargetAsmBackend &TAB,
+                  raw_ostream &_OS, MCCodeEmitter *_Emitter)
+    : MCStreamer(Context), Assembler(Context, TAB, _OS), Emitter(_Emitter),
       CurSectionData(0) {}
   ~MCMachOStreamer() {}
 
@@ -398,7 +399,7 @@
   Assembler.Finish();
 }
 
-MCStreamer *llvm::createMachOStreamer(MCContext &Context, raw_ostream &OS,
-                                      MCCodeEmitter *CE) {
-  return new MCMachOStreamer(Context, OS, CE);
+MCStreamer *llvm::createMachOStreamer(MCContext &Context, TargetAsmBackend &TAB,
+                                      raw_ostream &OS, MCCodeEmitter *CE) {
+  return new MCMachOStreamer(Context, TAB, OS, CE);
 }

Modified: llvm/trunk/tools/llvm-mc/llvm-mc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-mc/llvm-mc.cpp?rev=98222&r1=98221&r2=98222&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-mc/llvm-mc.cpp (original)
+++ llvm/trunk/tools/llvm-mc/llvm-mc.cpp Wed Mar 10 19:34:27 2010
@@ -27,6 +27,7 @@
 #include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/System/Signals.h"
+#include "llvm/Target/TargetAsmBackend.h"
 #include "llvm/Target/TargetAsmParser.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetRegistry.h"
@@ -259,6 +260,7 @@
   OwningPtr<MCInstPrinter> IP;
   OwningPtr<MCCodeEmitter> CE;
   OwningPtr<MCStreamer> Str;
+  OwningPtr<TargetAsmBackend> TAB;
 
   const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName);
   assert(MAI && "Unable to create target asm info!");
@@ -274,7 +276,8 @@
   } else {
     assert(FileType == OFT_ObjectFile && "Invalid file type!");
     CE.reset(TheTarget->createCodeEmitter(*TM, Ctx));
-    Str.reset(createMachOStreamer(Ctx, *Out, CE.get()));
+    TAB.reset(TheTarget->createAsmBackend(TripleName));
+    Str.reset(createMachOStreamer(Ctx, *TAB, *Out, CE.get()));
   }
 
   AsmParser Parser(SrcMgr, Ctx, *Str.get(), *MAI);




From dalej at apple.com  Wed Mar 10 20:10:24 2010
From: dalej at apple.com (Dale Johannesen)
Date: Thu, 11 Mar 2010 02:10:24 -0000
Subject: [llvm-commits] [llvm] r98224 -
	/llvm/trunk/lib/CodeGen/MachineCSE.cpp
Message-ID: <20100311021024.66FA82A6C12C@llvm.org>

Author: johannes
Date: Wed Mar 10 20:10:24 2010
New Revision: 98224

URL: http://llvm.org/viewvc/llvm-project?rev=98224&view=rev
Log:
Fix debug_value handling.


Modified:
    llvm/trunk/lib/CodeGen/MachineCSE.cpp

Modified: llvm/trunk/lib/CodeGen/MachineCSE.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineCSE.cpp?rev=98224&r1=98223&r2=98224&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineCSE.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineCSE.cpp Wed Mar 10 20:10:24 2010
@@ -122,8 +122,12 @@
       // Reached end of block, register is obviously dead.
       return true;
 
-    if (I->isDebugValue())
+    if (I->isDebugValue()) {
+      // These must not count against the limit.
+      ++LookAheadLeft;
+      ++I;
       continue;
+    }
     bool SeenDef = false;
     for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) {
       const MachineOperand &MO = I->getOperand(i);
@@ -188,7 +192,7 @@
 
 bool MachineCSE::isCSECandidate(MachineInstr *MI) {
   if (MI->isLabel() || MI->isPHI() || MI->isImplicitDef() ||
-      MI->isKill() || MI->isInlineAsm())
+      MI->isKill() || MI->isInlineAsm() || MI->isDebugValue())
     return false;
 
   // Ignore copies.




From daniel at zuster.org  Wed Mar 10 20:28:48 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 02:28:48 -0000
Subject: [llvm-commits] [llvm] r98225 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
Message-ID: <20100311022848.6483D2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 20:28:48 2010
New Revision: 98225

URL: http://llvm.org/viewvc/llvm-project?rev=98225&view=rev
Log:
Remove dead include.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=98225&r1=98224&r2=98225&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Wed Mar 10 20:28:48 2010
@@ -53,7 +53,6 @@
 #include "llvm/Target/TargetInstrInfo.h"
 #include "llvm/Target/TargetLowering.h"
 #include "llvm/Target/TargetMachine.h"
-#include "SelectionDAGBuilder.h"
 #include "FunctionLoweringInfo.h"
 using namespace llvm;
 




From daniel at zuster.org  Wed Mar 10 20:28:52 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 02:28:52 -0000
Subject: [llvm-commits] [llvm] r98226 -
	/llvm/trunk/include/llvm/Target/TargetRegistry.h
Message-ID: <20100311022852.74B1E2A6C12D@llvm.org>

Author: ddunbar
Date: Wed Mar 10 20:28:52 2010
New Revision: 98226

URL: http://llvm.org/viewvc/llvm-project?rev=98226&view=rev
Log:
Fix (unused) RegisterAsmBackend template, clang++ isn't happy about this.

Modified:
    llvm/trunk/include/llvm/Target/TargetRegistry.h

Modified: llvm/trunk/include/llvm/Target/TargetRegistry.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetRegistry.h?rev=98226&r1=98225&r2=98226&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetRegistry.h (original)
+++ llvm/trunk/include/llvm/Target/TargetRegistry.h Wed Mar 10 20:28:52 2010
@@ -588,8 +588,9 @@
     }
 
   private:
-    static TargetAsmBackend *Allocator(const Target &T, MCAssembler &Backend) {
-      return new AsmBackendImpl(T, Backend);
+    static TargetAsmBackend *Allocator(const Target &T,
+                                       const std::string &Triple) {
+      return new AsmBackendImpl(T, Triple);
     }
   };
 




From daniel at zuster.org  Wed Mar 10 20:28:59 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 02:28:59 -0000
Subject: [llvm-commits] [llvm] r98227 - in /llvm/trunk:
 include/llvm/MC/MCAsmLayout.h include/llvm/MC/MCAssembler.h
 include/llvm/MC/MCExpr.h lib/MC/MCExpr.cpp
 lib/Target/X86/X86MCTargetExpr.cpp lib/Target/X86/X86MCTargetExpr.h
Message-ID: <20100311022859.3C1D92A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 20:28:59 2010
New Revision: 98227

URL: http://llvm.org/viewvc/llvm-project?rev=98227&view=rev
Log:
MC: Sketch initial MCAsmLayout class, which encapsulates the current layout of an assembly file. The MCAsmLayout is also available for use by MCExpr::EvaluateAs{Absolute,Relocatable}, to allow target specific hooks and "absolutizing" of symbols.

Added:
    llvm/trunk/include/llvm/MC/MCAsmLayout.h
Modified:
    llvm/trunk/include/llvm/MC/MCAssembler.h
    llvm/trunk/include/llvm/MC/MCExpr.h
    llvm/trunk/lib/MC/MCExpr.cpp
    llvm/trunk/lib/Target/X86/X86MCTargetExpr.cpp
    llvm/trunk/lib/Target/X86/X86MCTargetExpr.h

Added: llvm/trunk/include/llvm/MC/MCAsmLayout.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAsmLayout.h?rev=98227&view=auto
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAsmLayout.h (added)
+++ llvm/trunk/include/llvm/MC/MCAsmLayout.h Wed Mar 10 20:28:59 2010
@@ -0,0 +1,49 @@
+//===- MCAsmLayout.h - Assembly Layout Object -------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_MC_MCASMLAYOUT_H
+#define LLVM_MC_MCASMLAYOUT_H
+
+namespace llvm {
+class MCAssembler;
+
+/// Encapsulates the layout of an assembly file at a particular point in time.
+///
+/// Assembly may requiring compute multiple layouts for a particular assembly
+/// file as part of the relaxation process. This class encapsulates the layout
+/// at a single point in time in such a way that it is always possible to
+/// efficiently compute the exact addresses of any symbol in the assembly file,
+/// even during the relaxation process.
+class MCAsmLayout {
+private:
+  uint64_t CurrentLocation;
+
+  MCAssembler &Assembler;
+
+public:
+  MCAsmLayout(MCAssembler &_Assembler)
+    : CurrentLocation(0), Assembler(_Assember) {}
+
+  /// Get the assembler object this is a layout for.
+  MCAssembler &getAssembler() { return Assembler; }
+
+  /// Get the current location value, i.e. that value of the '.' expression.
+  uin64_t getCurrentLocation() {
+    return CurrentLocation;
+  }
+
+  /// Set the current location.
+  void setCurrentLocation(uint64_t Value) {
+    CurrentLocation = Value;
+  }
+};
+
+} // end namespace llvm
+
+#endif

Modified: llvm/trunk/include/llvm/MC/MCAssembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAssembler.h?rev=98227&r1=98226&r2=98227&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAssembler.h (original)
+++ llvm/trunk/include/llvm/MC/MCAssembler.h Wed Mar 10 20:28:59 2010
@@ -625,6 +625,8 @@
 
   MCContext &getContext() const { return Context; }
 
+  TargetAsmBackend &getBackend() const { return Backend; }
+
   /// Finish - Do final processing and write the object to the output stream.
   void Finish();
 

Modified: llvm/trunk/include/llvm/MC/MCExpr.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCExpr.h?rev=98227&r1=98226&r2=98227&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCExpr.h (original)
+++ llvm/trunk/include/llvm/MC/MCExpr.h Wed Mar 10 20:28:59 2010
@@ -15,6 +15,7 @@
 
 namespace llvm {
 class MCAsmInfo;
+class MCAsmLayout;
 class MCContext;
 class MCSymbol;
 class MCValue;
@@ -62,15 +63,19 @@
   /// EvaluateAsAbsolute - Try to evaluate the expression to an absolute value.
   ///
   /// @param Res - The absolute value, if evaluation succeeds.
+  /// @param Layout - The assembler layout object to use for evaluating symbol
+  /// values. If not given, then only non-symbolic expressions will be
+  /// evaluated.
   /// @result - True on success.
-  bool EvaluateAsAbsolute(int64_t &Res) const;
+  bool EvaluateAsAbsolute(int64_t &Res, MCAsmLayout *Layout = 0) const;
 
   /// EvaluateAsRelocatable - Try to evaluate the expression to a relocatable
   /// value, i.e. an expression of the fixed form (a - b + constant).
   ///
   /// @param Res - The relocatable value, if evaluation succeeds.
+  /// @param Layout - The assembler layout object to use for evaluating values.
   /// @result - True on success.
-  bool EvaluateAsRelocatable(MCValue &Res) const;
+  bool EvaluateAsRelocatable(MCValue &Res, MCAsmLayout *Layout = 0) const;
 
   /// @}
 
@@ -348,7 +353,8 @@
 public:
   
   virtual void PrintImpl(raw_ostream &OS) const = 0;
-  virtual bool EvaluateAsRelocatableImpl(MCValue &Res) const = 0;
+  virtual bool EvaluateAsRelocatableImpl(MCValue &Res,
+                                         MCAsmLayout *Layout) const = 0;
 
   
   static bool classof(const MCExpr *E) {

Modified: llvm/trunk/lib/MC/MCExpr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCExpr.cpp?rev=98227&r1=98226&r2=98227&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCExpr.cpp (original)
+++ llvm/trunk/lib/MC/MCExpr.cpp Wed Mar 10 20:28:59 2010
@@ -142,10 +142,10 @@
 
 /* *** */
 
-bool MCExpr::EvaluateAsAbsolute(int64_t &Res) const {
+bool MCExpr::EvaluateAsAbsolute(int64_t &Res, MCAsmLayout *Layout) const {
   MCValue Value;
   
-  if (!EvaluateAsRelocatable(Value) || !Value.isAbsolute())
+  if (!EvaluateAsRelocatable(Value, Layout) || !Value.isAbsolute())
     return false;
 
   Res = Value.getConstant();
@@ -174,10 +174,10 @@
   return true;
 }
 
-bool MCExpr::EvaluateAsRelocatable(MCValue &Res) const {
+bool MCExpr::EvaluateAsRelocatable(MCValue &Res, MCAsmLayout *Layout) const {
   switch (getKind()) {
   case Target:
-    return cast<MCTargetExpr>(this)->EvaluateAsRelocatableImpl(Res);
+    return cast<MCTargetExpr>(this)->EvaluateAsRelocatableImpl(Res, Layout);
       
   case Constant:
     Res = MCValue::get(cast<MCConstantExpr>(this)->getValue());
@@ -188,7 +188,7 @@
 
     // Evaluate recursively if this is a variable.
     if (Sym.isVariable())
-      return Sym.getValue()->EvaluateAsRelocatable(Res);
+      return Sym.getValue()->EvaluateAsRelocatable(Res, Layout);
 
     Res = MCValue::get(&Sym, 0, 0);
     return true;
@@ -198,7 +198,7 @@
     const MCUnaryExpr *AUE = cast<MCUnaryExpr>(this);
     MCValue Value;
 
-    if (!AUE->getSubExpr()->EvaluateAsRelocatable(Value))
+    if (!AUE->getSubExpr()->EvaluateAsRelocatable(Value, Layout))
       return false;
 
     switch (AUE->getOpcode()) {
@@ -231,8 +231,8 @@
     const MCBinaryExpr *ABE = cast<MCBinaryExpr>(this);
     MCValue LHSValue, RHSValue;
     
-    if (!ABE->getLHS()->EvaluateAsRelocatable(LHSValue) ||
-        !ABE->getRHS()->EvaluateAsRelocatable(RHSValue))
+    if (!ABE->getLHS()->EvaluateAsRelocatable(LHSValue, Layout) ||
+        !ABE->getRHS()->EvaluateAsRelocatable(RHSValue, Layout))
       return false;
 
     // We only support a few operations on non-constant expressions, handle

Modified: llvm/trunk/lib/Target/X86/X86MCTargetExpr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCTargetExpr.cpp?rev=98227&r1=98226&r2=98227&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCTargetExpr.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86MCTargetExpr.cpp Wed Mar 10 20:28:59 2010
@@ -36,12 +36,13 @@
   }
 }
 
-bool X86MCTargetExpr::EvaluateAsRelocatableImpl(MCValue &Res) const {
+bool X86MCTargetExpr::EvaluateAsRelocatableImpl(MCValue &Res,
+                                                MCAsmLayout *Layout) const {
   // FIXME: I don't know if this is right, it followed MCSymbolRefExpr.
   
   // Evaluate recursively if this is a variable.
   if (Sym->isVariable())
-    return Sym->getValue()->EvaluateAsRelocatable(Res);
+    return Sym->getValue()->EvaluateAsRelocatable(Res, Layout);
   
   Res = MCValue::get(Sym, 0, 0);
   return true;

Modified: llvm/trunk/lib/Target/X86/X86MCTargetExpr.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86MCTargetExpr.h?rev=98227&r1=98226&r2=98227&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86MCTargetExpr.h (original)
+++ llvm/trunk/lib/Target/X86/X86MCTargetExpr.h Wed Mar 10 20:28:59 2010
@@ -41,7 +41,7 @@
                                  MCContext &Ctx);
   
   void PrintImpl(raw_ostream &OS) const;
-  bool EvaluateAsRelocatableImpl(MCValue &Res) const;
+  bool EvaluateAsRelocatableImpl(MCValue &Res, MCAsmLayout *Layout) const;
 };
   
 } // end namespace llvm




From echristo at apple.com  Wed Mar 10 20:41:03 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 02:41:03 -0000
Subject: [llvm-commits] [llvm] r98228 -
	/llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
Message-ID: <20100311024103.9A3E42A6C12C@llvm.org>

Author: echristo
Date: Wed Mar 10 20:41:03 2010
New Revision: 98228

URL: http://llvm.org/viewvc/llvm-project?rev=98228&view=rev
Log:
Do some final lowering in CodeGenPrepare of _chk calls similar to
that in InstCombineCalls.

More call lowering needed.

Modified:
    llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp?rev=98228&r1=98227&r2=98228&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp Wed Mar 10 20:41:03 2010
@@ -28,6 +28,7 @@
 #include "llvm/Transforms/Utils/AddrModeMatcher.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
+#include "llvm/Transforms/Utils/BuildLibCalls.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/Assembly/Writer.h"
@@ -36,6 +37,7 @@
 #include "llvm/Support/GetElementPtrTypeIterator.h"
 #include "llvm/Support/PatternMatch.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/IRBuilder.h"
 using namespace llvm;
 using namespace llvm::PatternMatch;
 
@@ -72,6 +74,7 @@
                             DenseMap<Value*,Value*> &SunkAddrs);
     bool OptimizeInlineAsmInst(Instruction *I, CallSite CS,
                                DenseMap<Value*,Value*> &SunkAddrs);
+    bool OptimizeCallInst(CallInst *CI);
     bool MoveExtToFormExtLoad(Instruction *I);
     bool OptimizeExtUses(Instruction *I);
     void findLoopBackEdges(const Function &F);
@@ -537,6 +540,133 @@
   return MadeChange;
 }
 
+bool CodeGenPrepare::OptimizeCallInst(CallInst *CI) {
+  bool MadeChange = false;
+  
+  // Lower all uses of llvm.objectsize.*
+  IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI);
+  if (II && II->getIntrinsicID() == Intrinsic::objectsize) {
+    bool Min = (cast<ConstantInt>(II->getOperand(2))->getZExtValue() == 1);
+    const Type *ReturnTy = CI->getType();
+    Constant *RetVal = ConstantInt::get(ReturnTy, Min ? 0 : -1ULL);    
+    CI->replaceAllUsesWith(RetVal);
+    CI->eraseFromParent();
+    return true;
+  }
+
+  // From here on out we're working with named functions.
+  if (CI->getCalledFunction() == 0) return false;
+  
+  // We'll need TargetData from here on out.
+  const TargetData *TD = TLI ? TLI->getTargetData() : 0;
+  if (!TD) return false;
+  
+  // Lower all default uses of _chk calls.  This is code very similar
+  // to the code in InstCombineCalls, but here we are only lowering calls
+  // that have the default "don't know" as the objectsize.  Anything else
+  // should be left alone.
+  StringRef Name = CI->getCalledFunction()->getName();
+  BasicBlock *BB = CI->getParent();
+  IRBuilder<> B(CI->getParent()->getContext());
+  
+  // Set the builder to the instruction after the call.
+  B.SetInsertPoint(BB, CI);
+
+  if (Name == "__memcpy_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
+    if (!SizeCI)
+      return 0;
+    if (SizeCI->isAllOnesValue()) {
+      EmitMemCpy(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+                 1, B, TD);
+      CI->replaceAllUsesWith(CI->getOperand(1));
+      CI->eraseFromParent();
+      return true;
+    }
+    return 0;
+  }
+
+  // Should be similar to memcpy.
+  if (Name == "__mempcpy_chk") {
+    return 0;
+  }
+
+  if (Name == "__memmove_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
+    if (!SizeCI)
+      return 0;
+    if (SizeCI->isAllOnesValue()) {
+      EmitMemMove(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+                  1, B, TD);
+      CI->replaceAllUsesWith(CI->getOperand(1));
+      CI->eraseFromParent();
+      return true;
+    }
+    return 0;
+  }
+
+  if (Name == "__memset_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
+    if (!SizeCI)
+      return 0;
+    if (SizeCI->isAllOnesValue()) {
+      Value *Val = B.CreateIntCast(CI->getOperand(2), B.getInt8Ty(),
+                                   false);
+      EmitMemSet(CI->getOperand(1), Val,  CI->getOperand(3), B, TD);
+      CI->replaceAllUsesWith(CI->getOperand(1));
+      CI->eraseFromParent();
+      return true;
+    }
+    return 0;
+  }
+
+  if (Name == "__strcpy_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
+    if (!SizeCI)
+      return 0;
+    // If a) we don't have any length information, or b) we know this will
+    // fit then just lower to a plain strcpy. Otherwise we'll keep our
+    // strcpy_chk call which may fail at runtime if the size is too long.
+    // TODO: It might be nice to get a maximum length out of the possible
+    // string lengths for varying.
+    if (SizeCI->isAllOnesValue()) {
+      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      CI->replaceAllUsesWith(Ret);
+      CI->eraseFromParent();
+      return true;
+    }
+    return 0;
+  }
+
+  // Should be similar to strcpy.
+  if (Name == "__stpcpy_chk") {
+    return 0;
+  }
+
+  if (Name == "__strncpy_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
+    if (!SizeCI)
+      return 0;
+    if (SizeCI->isAllOnesValue()) {
+      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
+                               CI->getOperand(3), B, TD);
+      CI->replaceAllUsesWith(Ret);
+      CI->eraseFromParent();
+      return true;
+    }
+    return 0; 
+  }
+
+  if (Name == "__strcat_chk") {
+    return 0;
+  }
+
+  if (Name == "__strncat_chk") {
+    return 0;
+  }
+  
+  return MadeChange;
+}
 //===----------------------------------------------------------------------===//
 // Memory Optimization
 //===----------------------------------------------------------------------===//
@@ -913,6 +1043,10 @@
         } else
           // Sink address computing for memory operands into the block.
           MadeChange |= OptimizeInlineAsmInst(I, &(*CI), SunkAddrs);
+      } else {
+        // Other CallInst optimizations that don't need to muck with the
+        // enclosing iterator here.
+        MadeChange |= OptimizeCallInst(CI);
       }
     }
   }




From nicholas at mxc.ca  Wed Mar 10 23:52:35 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Wed, 10 Mar 2010 21:52:35 -0800
Subject: [llvm-commits] [llvm] r86748 - in
 /llvm/trunk:	include/llvm/Analysis/
 include/llvm/CodeGen/	include/llvm/Transforms/Utils/
 lib/Analysis/	lib/CodeGen/AsmPrinter/
 lib/CodeGen/SelectionDAG/	lib/Transforms/Utils/
In-Reply-To: <34CF0D4C-9011-493D-93B1-0017CAA34A19@apple.com>
References: <200911102306.nAAN61b2006429@zion.cs.uiuc.edu>	<bb55930e1003061440p3fb86507xc9e3810210454ea2@mail.gmail.com>	<7947A46A-DC95-4E4D-BC0E-D41600D77901@apple.com>	<5d44f72f1003101558l157368d5v470cc502f68cec87@mail.gmail.com>	<ABB7E6F6-E1AB-48C4-ADEC-4DFE26680107@apple.com>	<bb55930e1003101700v49166153m7d591c1dc6c66c47@mail.gmail.com>
	<34CF0D4C-9011-493D-93B1-0017CAA34A19@apple.com>
Message-ID: <4B988523.2000609@mxc.ca>

Devang Patel wrote:
>
>>>> Thanks.  Is there a test for this ordering stability?
>>>
>>> You mean test to reason why we use list instead of map because of ordering requirement ?
>>
>> I'm not exactly sure how to parse that question, so sorry if this
>> isn't an appropriate answer, but I mean something under either test/
>> or unittests/ that would fail if anyone removed the list or used the
>> map where they should have used the list. For example, it could llc a
>> file with several scopes and CHECK that the DW_whatevers appeared in a
>> particular order. That would overspecify the requirement, I think, and
>> so would need to be updated any time the logic changed, but it would
>> fail if the output order depended on anything but the input.
>
> No there is no such test. Such test would check retrieving objects from a pointers' map is not deterministic.

No, you've mentally placed the test at the wrong level. Obviously map 
iteration isn't deterministic while list iteration is, but what a test 
would test for is a change in behaviour of your map/list-using code.

Did the non-determinstic behaviour of map cause any bugs in debug 
emission? If so then that's what you would test for. The test should be 
something that uses the debug emission and passes when you use list but 
fails if you were to roll back to using a map.

On the other hand, if this is just a precautionary change to remove any 
chance of non-determinism, and you don't know that it made any 
externally-visible difference then I should think it'd be fine to pass 
on writing a test for that!

Nick

  And there are tons of places where llvm code base is careful, by using 
appropriate data structure,  to not introduce nondeterministic behavior. 
llvm-gcc bootstrap is very sensitive about such things and every one 
dreads bootstrap failures due to miscompare!
>
> -
> Devang
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From daniel at zuster.org  Wed Mar 10 23:53:33 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 05:53:33 -0000
Subject: [llvm-commits] [llvm] r98240 - in /llvm/trunk:
 include/llvm/MC/MCAsmLayout.h lib/MC/MCAssembler.cpp
Message-ID: <20100311055333.757CD2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Mar 10 23:53:33 2010
New Revision: 98240

URL: http://llvm.org/viewvc/llvm-project?rev=98240&view=rev
Log:
MC/Mach-O: Start passing in the basic MCAsmLayout object.
 - Also, drop the current location part of AsmLayout, I think I prefer to implement this via explicit symbols.

Modified:
    llvm/trunk/include/llvm/MC/MCAsmLayout.h
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/include/llvm/MC/MCAsmLayout.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAsmLayout.h?rev=98240&r1=98239&r2=98240&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAsmLayout.h (original)
+++ llvm/trunk/include/llvm/MC/MCAsmLayout.h Wed Mar 10 23:53:33 2010
@@ -22,26 +22,13 @@
 /// even during the relaxation process.
 class MCAsmLayout {
 private:
-  uint64_t CurrentLocation;
-
   MCAssembler &Assembler;
 
 public:
-  MCAsmLayout(MCAssembler &_Assembler)
-    : CurrentLocation(0), Assembler(_Assember) {}
+  MCAsmLayout(MCAssembler &_Assembler) : Assembler(_Assembler) {}
 
   /// Get the assembler object this is a layout for.
   MCAssembler &getAssembler() { return Assembler; }
-
-  /// Get the current location value, i.e. that value of the '.' expression.
-  uin64_t getCurrentLocation() {
-    return CurrentLocation;
-  }
-
-  /// Set the current location.
-  void setCurrentLocation(uint64_t Value) {
-    CurrentLocation = Value;
-  }
 };
 
 } // end namespace llvm

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98240&r1=98239&r2=98240&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Wed Mar 10 23:53:33 2010
@@ -9,6 +9,7 @@
 
 #define DEBUG_TYPE "assembler"
 #include "llvm/MC/MCAssembler.h"
+#include "llvm/MC/MCAsmLayout.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCSymbol.h"
@@ -510,8 +511,11 @@
     unsigned IsPCRel = isFixupKindPCRel(Fixup.Kind);
     unsigned Log2Size = getFixupKindLog2Size(Fixup.Kind);
 
+    // FIXME: Share layout object.
+    MCAsmLayout Layout(Asm);
+
     MCValue Target;
-    if (!Fixup.Value->EvaluateAsRelocatable(Target))
+    if (!Fixup.Value->EvaluateAsRelocatable(Target, &Layout))
       llvm_report_error("expected relocatable expression");
 
     // If this is a difference or a defined symbol plus an offset, then we need
@@ -1001,6 +1005,7 @@
 }
 
 void MCAssembler::LayoutSection(MCSectionData &SD) {
+  MCAsmLayout Layout(*this);
   uint64_t Address = SD.getAddress();
 
   for (MCSectionData::iterator it = SD.begin(), ie = SD.end(); it != ie; ++it) {
@@ -1029,21 +1034,17 @@
     case MCFragment::FT_Org: {
       MCOrgFragment &OF = cast<MCOrgFragment>(F);
 
-      MCValue Target;
-      if (!OF.getOffset().EvaluateAsRelocatable(Target))
-        llvm_report_error("expected relocatable expression");
-
-      if (!Target.isAbsolute())
-        llvm_unreachable("FIXME: Not yet implemented!");
-      uint64_t OrgOffset = Target.getConstant();
-      uint64_t Offset = Address - SD.getAddress();
+      int64_t TargetLocation;
+      if (!OF.getOffset().EvaluateAsAbsolute(TargetLocation, &Layout))
+        llvm_report_error("expected assembly-time absolute expression");
 
       // FIXME: We need a way to communicate this error.
-      if (OrgOffset < Offset)
-        llvm_report_error("invalid .org offset '" + Twine(OrgOffset) +
-                          "' (at offset '" + Twine(Offset) + "'");
+      int64_t Offset = TargetLocation - F.getOffset();
+      if (Offset < 0)
+        llvm_report_error("invalid .org offset '" + Twine(TargetLocation) +
+                          "' (at offset '" + Twine(F.getOffset()) + "'");
 
-      F.setFileSize(OrgOffset - Offset);
+      F.setFileSize(Offset);
       break;
     }
 




From daniel at zuster.org  Wed Mar 10 23:53:37 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 05:53:37 -0000
Subject: [llvm-commits] [llvm] r98241 - in /llvm/trunk:
 lib/MC/MCAssembler.cpp lib/MC/MCExpr.cpp test/MC/MachO/absolutize.s
Message-ID: <20100311055338.0206D2A6C12D@llvm.org>

Author: ddunbar
Date: Wed Mar 10 23:53:37 2010
New Revision: 98241

URL: http://llvm.org/viewvc/llvm-project?rev=98241&view=rev
Log:
MC/Mach-O: Implement "absolutizing" semantics of .set, by evaluating the assembly time value of variables.

Added:
    llvm/trunk/test/MC/MachO/absolutize.s
Modified:
    llvm/trunk/lib/MC/MCAssembler.cpp
    llvm/trunk/lib/MC/MCExpr.cpp

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98241&r1=98240&r2=98241&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Wed Mar 10 23:53:37 2010
@@ -539,10 +539,10 @@
     if (Target.isAbsolute()) { // constant
       // SymbolNum of 0 indicates the absolute section.
       //
-      // FIXME: When is this generated?
+      // FIXME: Currently, these are never generated (see code below). I cannot
+      // find a case where they are actually emitted.
       Type = RIT_Vanilla;
       Value = 0;
-      llvm_unreachable("FIXME: Not yet implemented!");
     } else {
       const MCSymbol *Symbol = Target.getSymA();
       MCSymbolData *SD = &Asm.getSymbolData(*Symbol);
@@ -572,6 +572,12 @@
     if (IsPCRel)
       Fixup.FixedValue -= Address;
 
+    // If the target evaluates to a constant, we don't need a relocation. This
+    // occurs with absolutized expressions which are not resolved to constants
+    // until after relaxation.
+    if (Target.isAbsolute())
+      return;
+
     // If this fixup is a vanilla PC relative relocation for a local label, we
     // don't need a relocation.
     //

Modified: llvm/trunk/lib/MC/MCExpr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCExpr.cpp?rev=98241&r1=98240&r2=98241&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCExpr.cpp (original)
+++ llvm/trunk/lib/MC/MCExpr.cpp Wed Mar 10 23:53:37 2010
@@ -8,11 +8,14 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCAsmLayout.h"
+#include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/MC/MCValue.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/Target/TargetAsmBackend.h"
 using namespace llvm;
 
 void MCExpr::print(raw_ostream &OS) const {
@@ -187,8 +190,23 @@
     const MCSymbol &Sym = cast<MCSymbolRefExpr>(this)->getSymbol();
 
     // Evaluate recursively if this is a variable.
-    if (Sym.isVariable())
-      return Sym.getValue()->EvaluateAsRelocatable(Res, Layout);
+    if (Sym.isVariable()) {
+      if (!Sym.getValue()->EvaluateAsRelocatable(Res, Layout))
+        return false;
+
+      // Absolutize symbol differences when we have a layout object and the
+      // target requests it.
+      if (Layout && Res.getSymB() &&
+          Layout->getAssembler().getBackend().hasAbsolutizedSet()) {
+        MCSymbolData &A = Layout->getAssembler().getSymbolData(*Res.getSymA());
+        MCSymbolData &B = Layout->getAssembler().getSymbolData(*Res.getSymB());
+        Res = MCValue::get(+ A.getFragment()->getAddress() + A.getOffset()
+                           - B.getFragment()->getAddress() - B.getOffset()
+                           + Res.getConstant());
+      }
+
+      return true;
+    }
 
     Res = MCValue::get(&Sym, 0, 0);
     return true;

Added: llvm/trunk/test/MC/MachO/absolutize.s
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/MachO/absolutize.s?rev=98241&view=auto
==============================================================================
--- llvm/trunk/test/MC/MachO/absolutize.s (added)
+++ llvm/trunk/test/MC/MachO/absolutize.s Wed Mar 10 23:53:37 2010
@@ -0,0 +1,71 @@
+// RUN: llvm-mc -triple i386-apple-darwin9 %s -filetype=obj -o - | macho-dump | FileCheck %s
+
+// CHECK: # Relocation 0
+// CHECK: (('word-0', 0xa0000028),
+// CHECK:  ('word-1', 0x2b)),
+// CHECK: # Relocation 1
+// CHECK: (('word-0', 0xa4000020),
+// CHECK:  ('word-1', 0x37)),
+// CHECK: # Relocation 2
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x33)),
+// CHECK: # Relocation 3
+// CHECK: (('word-0', 0xa4000018),
+// CHECK:  ('word-1', 0x33)),
+// CHECK: # Relocation 4
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x2f)),
+// CHECK: # Relocation 5
+// CHECK: (('word-0', 0xa4000010),
+// CHECK:  ('word-1', 0x2b)),
+// CHECK: # Relocation 6
+// CHECK: (('word-0', 0xa1000000),
+// CHECK:  ('word-1', 0x2f)),
+// CHECK-NEXT: ])
+
+_text_a:
+        xorl %eax,%eax
+_text_b:
+        xorl %eax,%eax
+Ltext_c:
+        xorl %eax,%eax
+Ltext_d:        
+        xorl %eax,%eax
+        
+        movl $(_text_a - _text_b), %eax
+Ltext_expr_0 = _text_a - _text_b
+        movl $(Ltext_expr_0), %eax
+
+        movl $(Ltext_c - _text_b), %eax
+Ltext_expr_1 = Ltext_c - _text_b
+        movl $(Ltext_expr_1), %eax
+
+        movl $(Ltext_d - Ltext_c), %eax
+Ltext_expr_2 = Ltext_d - Ltext_c
+        movl $(Ltext_expr_2), %eax
+
+        movl $(_text_a + Ltext_expr_0), %eax
+
+        .data
+_data_a:
+        .long 0
+_data_b:
+        .long 0
+Ldata_c:
+        .long 0
+Ldata_d:        
+        .long 0
+        
+        .long _data_a - _data_b
+Ldata_expr_0 = _data_a - _data_b
+        .long Ldata_expr_0
+
+        .long Ldata_c - _data_b
+Ldata_expr_1 = Ldata_c - _data_b
+        .long Ldata_expr_1
+
+        .long Ldata_d - Ldata_c
+Ldata_expr_2 = Ldata_d - Ldata_c
+        .long Ldata_expr_2
+
+        .long _data_a + Ldata_expr_0




From jyasskin at google.com  Thu Mar 11 00:07:16 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Wed, 10 Mar 2010 22:07:16 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <13ECEBF5-92DB-43D5-9583-D606C640746B@apple.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
	<2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
	<5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>
	<13ECEBF5-92DB-43D5-9583-D606C640746B@apple.com>
Message-ID: <bb55930e1003102207s6f914986qa8599f9a87610bef@mail.gmail.com>

On Wed, Mar 10, 2010 at 4:47 PM, Devang Patel <dpatel at apple.com> wrote:
>
> On Mar 10, 2010, at 4:01 PM, Jeffrey Yasskin wrote:
>
>> On Tue, Mar 9, 2010 at 12:22 PM, Devang Patel <dpatel at apple.com> wrote:
>>>
>>> On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:
>>>
>>>> Now that nothing uses the result of constructCompileUnit() unless it
>>>> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
>>>> avoid new'ing the DIE and CompileUnit?
>>>
>>> With the introduction of DIFile, we are moving away from "isMain" field.
>>>
>>>> Then, if we delete ModuleCU in
>>>> endModule(), this memory leak would go away.
>>>
>>> We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.
>>
>> I can take it. I assume it's correct to simply not create the
>> CompileUnits and DIEs whose addresses aren't obviously passed
>> anywhere? Or do I need to look out for constructors or other functions
>> that escape their addresses surreptitiously?
>
> Now there is only one compile unit so it is simpler. Regarding other DIEs, it is expected that DIEs are created only if they will be emitted in the output. In other words, unused DIEs are not created. It would good to find and fix leaks in this area.
> -

Here's my patch to delete the ModuleCU at the end of its Module, and
to avoid creating unused CompileUnits and DIEs. I'll commit it
tomorrow morning unless you have comments.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: free_compile_unit.patch
Type: application/octet-stream
Size: 2795 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100310/9dfb172e/attachment.obj 

From jyasskin at google.com  Thu Mar 11 00:14:32 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Thu, 11 Mar 2010 06:14:32 -0000
Subject: [llvm-commits] [llvm] r98243 - in /llvm/trunk/lib/System:
 DynamicLibrary.cpp SearchForAddressOfSpecialSymbol.cpp
Message-ID: <20100311061432.5F6AD2A6C12C@llvm.org>

Author: jyasskin
Date: Thu Mar 11 00:14:32 2010
New Revision: 98243

URL: http://llvm.org/viewvc/llvm-project?rev=98243&view=rev
Log:
Make clang bootstrap happier on OSX 10.5 by reducing the number of headers
included when using global symbols to ask the linker for the addresses of
various functions.  One of the symbols was actually getting declared by a
header included in DynamicLibrary.cpp, which conflicted with the "extern void*"
declaration in SearchForAddressOfSpecialSymbol().

Added:
    llvm/trunk/lib/System/SearchForAddressOfSpecialSymbol.cpp
Modified:
    llvm/trunk/lib/System/DynamicLibrary.cpp

Modified: llvm/trunk/lib/System/DynamicLibrary.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/DynamicLibrary.cpp?rev=98243&r1=98242&r2=98243&view=diff
==============================================================================
--- llvm/trunk/lib/System/DynamicLibrary.cpp (original)
+++ llvm/trunk/lib/System/DynamicLibrary.cpp Thu Mar 11 00:14:32 2010
@@ -69,44 +69,8 @@
   return false;
 }
 
-static void *SearchForAddressOfSpecialSymbol(const char* symbolName) {
-#define EXPLICIT_SYMBOL(SYM) \
-   extern void *SYM; if (!strcmp(symbolName, #SYM)) return &SYM
-
-  // If this is darwin, it has some funky issues, try to solve them here.  Some
-  // important symbols are marked 'private external' which doesn't allow
-  // SearchForAddressOfSymbol to find them.  As such, we special case them here,
-  // there is only a small handful of them.
-
-#ifdef __APPLE__
-  {
-    EXPLICIT_SYMBOL(__ashldi3);
-    EXPLICIT_SYMBOL(__ashrdi3);
-    EXPLICIT_SYMBOL(__cmpdi2);
-    EXPLICIT_SYMBOL(__divdi3);
-    EXPLICIT_SYMBOL(__eprintf);
-    EXPLICIT_SYMBOL(__fixdfdi);
-    EXPLICIT_SYMBOL(__fixsfdi);
-    EXPLICIT_SYMBOL(__fixunsdfdi);
-    EXPLICIT_SYMBOL(__fixunssfdi);
-    EXPLICIT_SYMBOL(__floatdidf);
-    EXPLICIT_SYMBOL(__floatdisf);
-    EXPLICIT_SYMBOL(__lshrdi3);
-    EXPLICIT_SYMBOL(__moddi3);
-    EXPLICIT_SYMBOL(__udivdi3);
-    EXPLICIT_SYMBOL(__umoddi3);
-  }
-#endif
-
-#ifdef __CYGWIN__
-  {
-    EXPLICIT_SYMBOL(_alloca);
-    EXPLICIT_SYMBOL(__main);
-  }
-#endif
-
-#undef EXPLICIT_SYMBOL
-  return 0;
+namespace llvm {
+void *SearchForAddressOfSpecialSymbol(const char* symbolName);
 }
 
 void* DynamicLibrary::SearchForAddressOfSymbol(const char* symbolName) {
@@ -132,7 +96,7 @@
     }
   }
 
-  if (void *Result = SearchForAddressOfSpecialSymbol(symbolName))
+  if (void *Result = llvm::SearchForAddressOfSpecialSymbol(symbolName))
     return Result;
 
 // This macro returns the address of a well-known, explicit symbol

Added: llvm/trunk/lib/System/SearchForAddressOfSpecialSymbol.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/SearchForAddressOfSpecialSymbol.cpp?rev=98243&view=auto
==============================================================================
--- llvm/trunk/lib/System/SearchForAddressOfSpecialSymbol.cpp (added)
+++ llvm/trunk/lib/System/SearchForAddressOfSpecialSymbol.cpp Thu Mar 11 00:14:32 2010
@@ -0,0 +1,64 @@
+//===- SearchForAddressOfSpecialSymbol.cpp - Function addresses -*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+//  This file pulls the addresses of certain symbols out of the linker.  It must
+//  include as few header files as possible because it declares the symbols as
+//  void*, which would conflict with the actual symbol type if any header
+//  declared it.
+//
+//===----------------------------------------------------------------------===//
+
+#include <string.h>
+
+// Must declare the symbols in the global namespace.
+static void *DoSearch(const char* symbolName) {
+#define EXPLICIT_SYMBOL(SYM) \
+   extern void *SYM; if (!strcmp(symbolName, #SYM)) return &SYM
+
+  // If this is darwin, it has some funky issues, try to solve them here.  Some
+  // important symbols are marked 'private external' which doesn't allow
+  // SearchForAddressOfSymbol to find them.  As such, we special case them here,
+  // there is only a small handful of them.
+
+#ifdef __APPLE__
+  {
+    EXPLICIT_SYMBOL(__ashldi3);
+    EXPLICIT_SYMBOL(__ashrdi3);
+    EXPLICIT_SYMBOL(__cmpdi2);
+    EXPLICIT_SYMBOL(__divdi3);
+    EXPLICIT_SYMBOL(__eprintf);
+    EXPLICIT_SYMBOL(__fixdfdi);
+    EXPLICIT_SYMBOL(__fixsfdi);
+    EXPLICIT_SYMBOL(__fixunsdfdi);
+    EXPLICIT_SYMBOL(__fixunssfdi);
+    EXPLICIT_SYMBOL(__floatdidf);
+    EXPLICIT_SYMBOL(__floatdisf);
+    EXPLICIT_SYMBOL(__lshrdi3);
+    EXPLICIT_SYMBOL(__moddi3);
+    EXPLICIT_SYMBOL(__udivdi3);
+    EXPLICIT_SYMBOL(__umoddi3);
+  }
+#endif
+
+#ifdef __CYGWIN__
+  {
+    EXPLICIT_SYMBOL(_alloca);
+    EXPLICIT_SYMBOL(__main);
+  }
+#endif
+
+#undef EXPLICIT_SYMBOL
+  return 0;
+}
+
+namespace llvm {
+void *SearchForAddressOfSpecialSymbol(const char* symbolName) {
+  return DoSearch(symbolName);
+}
+}  // namespace llvm




From echristo at apple.com  Thu Mar 11 00:20:22 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 06:20:22 -0000
Subject: [llvm-commits] [llvm] r98244 - in /llvm/trunk:
 lib/Target/X86/X86FastISel.cpp test/CodeGen/X86/object-size.ll
Message-ID: <20100311062022.9C03D2A6C12C@llvm.org>

Author: echristo
Date: Thu Mar 11 00:20:22 2010
New Revision: 98244

URL: http://llvm.org/viewvc/llvm-project?rev=98244&view=rev
Log:
Have fast-isel understand llvm.objectsize.  Update testcase for slightly
different codegen.

Modified:
    llvm/trunk/lib/Target/X86/X86FastISel.cpp
    llvm/trunk/test/CodeGen/X86/object-size.ll

Modified: llvm/trunk/lib/Target/X86/X86FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FastISel.cpp?rev=98244&r1=98243&r2=98244&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FastISel.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86FastISel.cpp Thu Mar 11 00:20:22 2010
@@ -1166,6 +1166,30 @@
   // FIXME: Handle more intrinsics.
   switch (I.getIntrinsicID()) {
   default: return false;
+  case Intrinsic::objectsize: {
+    ConstantInt *CI = dyn_cast<ConstantInt>(I.getOperand(2));
+    const Type *Ty = I.getCalledFunction()->getReturnType();
+    
+    assert(CI && "Non-constant type in Intrinsic::objectsize?");
+    
+    EVT VT;
+    if (!isTypeLegal(Ty, VT))
+      return false;
+    
+    unsigned OpC = 0;
+    if (VT == MVT::i32)
+      OpC = X86::MOV32ri;
+    else if (VT == MVT::i64)
+      OpC = X86::MOV64ri;
+    else
+      return false;
+    
+    unsigned ResultReg = createResultReg(TLI.getRegClassFor(VT));
+    BuildMI(MBB, DL, TII.get(OpC), ResultReg).
+                                  addImm(CI->getZExtValue() == 0 ? -1ULL : 0);
+    UpdateValueMap(&I, ResultReg);
+    return true;
+  }
   case Intrinsic::dbg_declare: {
     DbgDeclareInst *DI = cast<DbgDeclareInst>(&I);
     X86AddressMode AM;

Modified: llvm/trunk/test/CodeGen/X86/object-size.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/object-size.ll?rev=98244&r1=98243&r2=98244&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/object-size.ll (original)
+++ llvm/trunk/test/CodeGen/X86/object-size.ll Thu Mar 11 00:20:22 2010
@@ -12,7 +12,7 @@
   %tmp = load i8** @p                             ; <i8*> [#uses=1]
   %0 = call i64 @llvm.objectsize.i64(i8* %tmp, i1 0) ; <i64> [#uses=1]
   %cmp = icmp ne i64 %0, -1                       ; <i1> [#uses=1]
-; X64: movq    $-1, %rax
+; X64: movabsq $-1, %rax
 ; X64: cmpq    $-1, %rax
   br i1 %cmp, label %cond.true, label %cond.false
 




From anton at korobeynikov.info  Thu Mar 11 00:26:11 2010
From: anton at korobeynikov.info (Anton Korobeynikov)
Date: Thu, 11 Mar 2010 09:26:11 +0300
Subject: [llvm-commits] [llvm] r98205 -
	/llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp
In-Reply-To: <20100311000620.237082A6C12C@llvm.org>
References: <20100311000620.237082A6C12C@llvm.org>
Message-ID: <b087fca01003102226t3e43c88v4e89b07bcc4cd2bd@mail.gmail.com>

> + ?// OpenBSD has buggy support for .quad in 32-bit mode, just split into two
> + ?// .words.
> + ?if (T.getOS() == Triple::OpenBSD && T.getArch() == Triple::x86)
> + ? ?Data64bitsDirective = 0;
Doesn't the second condition redundant? It's X86ELFMCAsmInfo after all :)

-- 
With best regards, Anton Korobeynikov
Faculty of Mathematics and Mechanics, Saint Petersburg State University



From clattner at apple.com  Thu Mar 11 01:01:15 2010
From: clattner at apple.com (Chris Lattner)
Date: Wed, 10 Mar 2010 23:01:15 -0800
Subject: [llvm-commits] [llvm] r98205 -
	/llvm/trunk/lib/Target/X86/X86MCAsmInfo.cpp
In-Reply-To: <b087fca01003102226t3e43c88v4e89b07bcc4cd2bd@mail.gmail.com>
References: <20100311000620.237082A6C12C@llvm.org>
	<b087fca01003102226t3e43c88v4e89b07bcc4cd2bd@mail.gmail.com>
Message-ID: <5F3711C1-DA50-4B71-8208-9F6E7AFF2EE2@apple.com>


On Mar 10, 2010, at 10:26 PM, Anton Korobeynikov wrote:

>> +  // OpenBSD has buggy support for .quad in 32-bit mode, just split into two
>> +  // .words.
>> +  if (T.getOS() == Triple::OpenBSD && T.getArch() == Triple::x86)
>> +    Data64bitsDirective = 0;
> Doesn't the second condition redundant? It's X86ELFMCAsmInfo after all :)

No, x86 != x86_64

-Chris


From kremenek at apple.com  Thu Mar 11 01:51:23 2010
From: kremenek at apple.com (Ted Kremenek)
Date: Thu, 11 Mar 2010 07:51:23 -0000
Subject: [llvm-commits] [llvm] r98250 - /llvm/trunk/lib/System/CMakeLists.txt
Message-ID: <20100311075123.7E7192A6C12F@llvm.org>

Author: kremenek
Date: Thu Mar 11 01:51:23 2010
New Revision: 98250

URL: http://llvm.org/viewvc/llvm-project?rev=98250&view=rev
Log:
Update CMake build.

Modified:
    llvm/trunk/lib/System/CMakeLists.txt

Modified: llvm/trunk/lib/System/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/CMakeLists.txt?rev=98250&r1=98249&r2=98250&view=diff
==============================================================================
--- llvm/trunk/lib/System/CMakeLists.txt (original)
+++ llvm/trunk/lib/System/CMakeLists.txt Thu Mar 11 01:51:23 2010
@@ -12,6 +12,7 @@
   Process.cpp
   Program.cpp
   RWMutex.cpp
+  SearchForAddressOfSpecialSymbol.cpp
   Signals.cpp
   ThreadLocal.cpp
   Threading.cpp




From evan.cheng at apple.com  Thu Mar 11 02:20:21 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 11 Mar 2010 08:20:21 -0000
Subject: [llvm-commits] [llvm] r98251 - in /llvm/trunk:
 include/llvm/CodeGen/LiveIntervalAnalysis.h
 lib/CodeGen/LiveIntervalAnalysis.cpp
 lib/CodeGen/SimpleRegisterCoalescing.cpp
 test/CodeGen/X86/phys_subreg_coalesce-3.ll
Message-ID: <20100311082022.02CFA2A6C12C@llvm.org>

Author: evancheng
Date: Thu Mar 11 02:20:21 2010
New Revision: 98251

URL: http://llvm.org/viewvc/llvm-project?rev=98251&view=rev
Log:
The check for coalescing a virtual register to a physical register, e.g.
cl = EXTRACT_SUBREG reg1024, 1, is overly conservative. It should check
for overlaps of vr's live interval with the super registers of the
physical register (ECX in this case) and let JoinIntervals() handle checking
the coalescing feasibility against the physical register (cl in this case).

Added:
    llvm/trunk/test/CodeGen/X86/phys_subreg_coalesce-3.ll
Modified:
    llvm/trunk/include/llvm/CodeGen/LiveIntervalAnalysis.h
    llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp
    llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp

Modified: llvm/trunk/include/llvm/CodeGen/LiveIntervalAnalysis.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/LiveIntervalAnalysis.h?rev=98251&r1=98250&r2=98251&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/LiveIntervalAnalysis.h (original)
+++ llvm/trunk/include/llvm/CodeGen/LiveIntervalAnalysis.h Thu Mar 11 02:20:21 2010
@@ -127,11 +127,11 @@
     bool conflictsWithPhysReg(const LiveInterval &li, VirtRegMap &vrm,
                               unsigned reg);
 
-    /// conflictsWithPhysRegRef - Similar to conflictsWithPhysRegRef except
-    /// it can check use as well.
-    bool conflictsWithPhysRegRef(LiveInterval &li, unsigned Reg,
-                                 bool CheckUse,
-                                 SmallPtrSet<MachineInstr*,32> &JoinedCopies);
+    /// conflictsWithSubPhysRegRef - Similar to conflictsWithPhysRegRef except
+    /// it checks for sub-register reference and it can check use as well.
+    bool conflictsWithSubPhysRegRef(LiveInterval &li, unsigned Reg,
+                                    bool CheckUse,
+                                   SmallPtrSet<MachineInstr*,32> &JoinedCopies);
 
     // Interval creation
     LiveInterval &getOrCreateInterval(unsigned reg) {

Modified: llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp?rev=98251&r1=98250&r2=98251&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp (original)
+++ llvm/trunk/lib/CodeGen/LiveIntervalAnalysis.cpp Thu Mar 11 02:20:21 2010
@@ -218,9 +218,9 @@
   return false;
 }
 
-/// conflictsWithPhysRegRef - Similar to conflictsWithPhysRegRef except
-/// it can check use as well.
-bool LiveIntervals::conflictsWithPhysRegRef(LiveInterval &li,
+/// conflictsWithSubPhysRegRef - Similar to conflictsWithPhysRegRef except
+/// it checks for sub-register reference and it can check use as well.
+bool LiveIntervals::conflictsWithSubPhysRegRef(LiveInterval &li,
                                             unsigned Reg, bool CheckUse,
                                   SmallPtrSet<MachineInstr*,32> &JoinedCopies) {
   for (LiveInterval::Ranges::const_iterator

Modified: llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp?rev=98251&r1=98250&r2=98251&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp (original)
+++ llvm/trunk/lib/CodeGen/SimpleRegisterCoalescing.cpp Thu Mar 11 02:20:21 2010
@@ -1260,10 +1260,10 @@
   RealDstReg = tri_->getMatchingSuperReg(DstReg, SubIdx, RC);
   assert(RealDstReg && "Invalid extract_subreg instruction!");
 
+  LiveInterval &RHS = li_->getInterval(SrcReg);
   // For this type of EXTRACT_SUBREG, conservatively
   // check if the live interval of the source register interfere with the
   // actual super physical register we are trying to coalesce with.
-  LiveInterval &RHS = li_->getInterval(SrcReg);
   if (li_->hasInterval(RealDstReg) &&
       RHS.overlaps(li_->getInterval(RealDstReg))) {
     DEBUG({
@@ -1273,7 +1273,11 @@
     return false; // Not coalescable
   }
   for (const unsigned* SR = tri_->getSubRegisters(RealDstReg); *SR; ++SR)
-    if (li_->hasInterval(*SR) && RHS.overlaps(li_->getInterval(*SR))) {
+    // Do not check DstReg or its sub-register. JoinIntervals() will take care
+    // of that.
+    if (*SR != DstReg &&
+        !tri_->isSubRegister(DstReg, *SR) &&
+        li_->hasInterval(*SR) && RHS.overlaps(li_->getInterval(*SR))) {
       DEBUG({
           dbgs() << "Interfere with sub-register ";
           li_->getInterval(*SR).print(dbgs(), tri_);
@@ -1294,9 +1298,9 @@
   RealSrcReg = tri_->getMatchingSuperReg(SrcReg, SubIdx, RC);
   assert(RealSrcReg && "Invalid extract_subreg instruction!");
 
-  LiveInterval &RHS = li_->getInterval(DstReg);
+  LiveInterval &LHS = li_->getInterval(DstReg);
   if (li_->hasInterval(RealSrcReg) &&
-      RHS.overlaps(li_->getInterval(RealSrcReg))) {
+      LHS.overlaps(li_->getInterval(RealSrcReg))) {
     DEBUG({
         dbgs() << "Interfere with register ";
         li_->getInterval(RealSrcReg).print(dbgs(), tri_);
@@ -1304,7 +1308,11 @@
     return false; // Not coalescable
   }
   for (const unsigned* SR = tri_->getSubRegisters(RealSrcReg); *SR; ++SR)
-    if (li_->hasInterval(*SR) && RHS.overlaps(li_->getInterval(*SR))) {
+    // Do not check SrcReg or its sub-register. JoinIntervals() will take care
+    // of that.
+    if (*SR != SrcReg &&
+        !tri_->isSubRegister(SrcReg, *SR) &&
+        li_->hasInterval(*SR) && LHS.overlaps(li_->getInterval(*SR))) {
       DEBUG({
           dbgs() << "Interfere with sub-register ";
           li_->getInterval(*SR).print(dbgs(), tri_);
@@ -1476,6 +1484,9 @@
         return false; // Not coalescable.
       }
 
+      // FIXME: The following checks are somewhat conservative. Perhaps a better
+      // way to implement this is to treat this as coalescing a vr with the
+      // super physical register.
       if (isExtSubReg) {
         if (!CanJoinExtractSubRegToPhysReg(DstReg, SrcReg, SubIdx, RealDstReg))
           return false; // Not coalescable
@@ -2205,7 +2216,7 @@
         li_->intervalIsInOneMBB(RHS) &&
         li_->getApproximateInstructionCount(RHS) <= 10) {
       // Perform a more exhaustive check for some common cases.
-      if (li_->conflictsWithPhysRegRef(RHS, LHS.reg, true, JoinedCopies))
+      if (li_->conflictsWithSubPhysRegRef(RHS, LHS.reg, true, JoinedCopies))
         return false;
     } else {
       for (const unsigned* SR = tri_->getSubRegisters(LHS.reg); *SR; ++SR)
@@ -2222,7 +2233,7 @@
     if (LHS.containsOneValue() &&
         li_->getApproximateInstructionCount(LHS) <= 10) {
       // Perform a more exhaustive check for some common cases.
-      if (li_->conflictsWithPhysRegRef(LHS, RHS.reg, false, JoinedCopies))
+      if (li_->conflictsWithSubPhysRegRef(LHS, RHS.reg, false, JoinedCopies))
         return false;
     } else {
       for (const unsigned* SR = tri_->getSubRegisters(RHS.reg); *SR; ++SR)

Added: llvm/trunk/test/CodeGen/X86/phys_subreg_coalesce-3.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/phys_subreg_coalesce-3.ll?rev=98251&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/phys_subreg_coalesce-3.ll (added)
+++ llvm/trunk/test/CodeGen/X86/phys_subreg_coalesce-3.ll Thu Mar 11 02:20:21 2010
@@ -0,0 +1,35 @@
+; RUN: llc < %s -mtriple=i386-apple-darwin | FileCheck %s
+; rdar://5571034
+
+define void @foo(i32* nocapture %quadrant, i32* nocapture %ptr, i32 %bbSize, i32 %bbStart, i32 %shifts) nounwind ssp {
+; CHECK: foo:
+entry:
+  %j.03 = add i32 %bbSize, -1                     ; <i32> [#uses=2]
+  %0 = icmp sgt i32 %j.03, -1                     ; <i1> [#uses=1]
+  br i1 %0, label %bb.nph, label %return
+
+bb.nph:                                           ; preds = %entry
+  %tmp9 = add i32 %bbStart, %bbSize               ; <i32> [#uses=1]
+  %tmp10 = add i32 %tmp9, -1                      ; <i32> [#uses=1]
+  br label %bb
+
+bb:                                               ; preds = %bb, %bb.nph
+; CHECK: %bb
+; CHECK-NOT: movb {{.*}}l, %cl
+; CHECK: sarl %cl
+  %indvar = phi i32 [ 0, %bb.nph ], [ %indvar.next, %bb ] ; <i32> [#uses=3]
+  %j.06 = sub i32 %j.03, %indvar                  ; <i32> [#uses=1]
+  %tmp11 = sub i32 %tmp10, %indvar                ; <i32> [#uses=1]
+  %scevgep = getelementptr i32* %ptr, i32 %tmp11  ; <i32*> [#uses=1]
+  %1 = load i32* %scevgep, align 4                ; <i32> [#uses=1]
+  %2 = ashr i32 %j.06, %shifts                    ; <i32> [#uses=1]
+  %3 = and i32 %2, 65535                          ; <i32> [#uses=1]
+  %4 = getelementptr inbounds i32* %quadrant, i32 %1 ; <i32*> [#uses=1]
+  store i32 %3, i32* %4, align 4
+  %indvar.next = add i32 %indvar, 1               ; <i32> [#uses=2]
+  %exitcond = icmp eq i32 %indvar.next, %bbSize   ; <i1> [#uses=1]
+  br i1 %exitcond, label %return, label %bb
+
+return:                                           ; preds = %bb, %entry
+  ret void
+}




From richard at xmos.com  Thu Mar 11 08:58:16 2010
From: richard at xmos.com (Richard Osborne)
Date: Thu, 11 Mar 2010 14:58:16 -0000
Subject: [llvm-commits] [llvm] r98255 - in /llvm/trunk:
 include/llvm/CodeGen/MachineJumpTableInfo.h
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/MachineFunction.cpp
 lib/ExecutionEngine/JIT/JITEmitter.cpp
Message-ID: <20100311145816.44BE32A6C12C@llvm.org>

Author: friedgold
Date: Thu Mar 11 08:58:16 2010
New Revision: 98255

URL: http://llvm.org/viewvc/llvm-project?rev=98255&view=rev
Log:
Add a new jump table encoding to indicate jump tables entries
are inside the function by the target at the point of use.

Modified:
    llvm/trunk/include/llvm/CodeGen/MachineJumpTableInfo.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/MachineFunction.cpp
    llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp

Modified: llvm/trunk/include/llvm/CodeGen/MachineJumpTableInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineJumpTableInfo.h?rev=98255&r1=98254&r2=98255&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineJumpTableInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineJumpTableInfo.h Thu Mar 11 08:58:16 2010
@@ -61,7 +61,11 @@
     ///      .set L4_5_set_123, LBB123 - LJTI1_2
     ///      .word L4_5_set_123
     EK_LabelDifference32,
-    
+
+    /// EK_Inline - Jump table entries are emitted inline at their point of
+    /// use. It is the responsibility of the target to emit the entries.
+    EK_Inline,
+
     /// EK_Custom32 - Each entry is a 32-bit value that is custom lowered by the
     /// TargetLowering::LowerCustomJumpTableEntry hook.
     EK_Custom32

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98255&r1=98254&r2=98255&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Thu Mar 11 08:58:16 2010
@@ -651,6 +651,7 @@
 void AsmPrinter::EmitJumpTableInfo() {
   const MachineJumpTableInfo *MJTI = MF->getJumpTableInfo();
   if (MJTI == 0) return;
+  if (MJTI->getEntryKind() == MachineJumpTableInfo::EK_Inline) return;
   const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();
   if (JT.empty()) return;
 
@@ -727,6 +728,8 @@
                                     unsigned UID) const {
   const MCExpr *Value = 0;
   switch (MJTI->getEntryKind()) {
+  case MachineJumpTableInfo::EK_Inline:
+    llvm_unreachable("Cannot emit EK_Inline jump table entry"); break;
   case MachineJumpTableInfo::EK_Custom32:
     Value = TM.getTargetLowering()->LowerCustomJumpTableEntry(MJTI, MBB, UID,
                                                               OutContext);

Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=98255&r1=98254&r2=98255&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Thu Mar 11 08:58:16 2010
@@ -574,6 +574,8 @@
   case MachineJumpTableInfo::EK_LabelDifference32:
   case MachineJumpTableInfo::EK_Custom32:
     return 4;
+  case MachineJumpTableInfo::EK_Inline:
+    return 0;
   }
   assert(0 && "Unknown jump table encoding!");
   return ~0;
@@ -591,6 +593,8 @@
   case MachineJumpTableInfo::EK_LabelDifference32:
   case MachineJumpTableInfo::EK_Custom32:
     return TD.getABIIntegerTypeAlignment(32);
+  case MachineJumpTableInfo::EK_Inline:
+    return 1;
   }
   assert(0 && "Unknown jump table encoding!");
   return ~0;

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp?rev=98255&r1=98254&r2=98255&view=diff
==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp Thu Mar 11 08:58:16 2010
@@ -1393,6 +1393,8 @@
 void JITEmitter::initJumpTableInfo(MachineJumpTableInfo *MJTI) {
   if (TheJIT->getJITInfo().hasCustomJumpTables())
     return;
+  if (MJTI->getEntryKind() == MachineJumpTableInfo::EK_Inline)
+    return;
 
   const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();
   if (JT.empty()) return;
@@ -1420,6 +1422,8 @@
 
   
   switch (MJTI->getEntryKind()) {
+  case MachineJumpTableInfo::EK_Inline:
+    return;
   case MachineJumpTableInfo::EK_BlockAddress: {
     // EK_BlockAddress - Each entry is a plain address of block, e.g.:
     //     .word LBB123




From richard at xmos.com  Thu Mar 11 08:58:56 2010
From: richard at xmos.com (Richard Osborne)
Date: Thu, 11 Mar 2010 14:58:56 -0000
Subject: [llvm-commits] [llvm] r98256 - in /llvm/trunk/lib/Target/XCore:
 XCoreISelLowering.cpp XCoreISelLowering.h
Message-ID: <20100311145856.6FDFF2A6C12C@llvm.org>

Author: friedgold
Date: Thu Mar 11 08:58:56 2010
New Revision: 98256

URL: http://llvm.org/viewvc/llvm-project?rev=98256&view=rev
Log:
Switch XCore over to using inline jump table entries.

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.h

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98256&r1=98255&r2=98256&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Thu Mar 11 08:58:56 2010
@@ -324,6 +324,10 @@
   return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, Res);
 }
 
+unsigned XCoreTargetLowering::getJumpTableEncoding() const {
+  return MachineJumpTableInfo::EK_Inline;
+}
+
 SDValue XCoreTargetLowering::
 LowerBR_JT(SDValue Op, SelectionDAG &DAG)
 {

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.h?rev=98256&r1=98255&r2=98256&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.h (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.h Thu Mar 11 08:58:56 2010
@@ -80,6 +80,8 @@
 
     explicit XCoreTargetLowering(XCoreTargetMachine &TM);
 
+    virtual unsigned getJumpTableEncoding() const;
+
     /// LowerOperation - Provide custom lowering hooks for some operations.
     virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG);
 




From stoklund at 2pi.dk  Thu Mar 11 09:44:45 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 11 Mar 2010 07:44:45 -0800
Subject: [llvm-commits] [llvm] r98219 - in /llvm/trunk:
	include/llvm/Transforms/Utils/BuildLibCalls.h
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	lib/Transforms/Utils/BuildLibCalls.cpp
In-Reply-To: <20100311012507.61B652A6C12C@llvm.org>
References: <20100311012507.61B652A6C12C@llvm.org>
Message-ID: <B4BAA6F8-BFBC-4A30-9407-7499D0373BFA@2pi.dk>


On Mar 10, 2010, at 5:25 PM, Eric Christopher wrote:

> Author: echristo
> Date: Wed Mar 10 19:25:07 2010
> New Revision: 98219
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98219&view=rev
> Log:
> Add strncpy libcall creator.  Use it when it should be used.
> 
> Modified:
>    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
>    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
>    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
> Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98219&r1=98218&r2=98219&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
> +++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Wed Mar 10 19:25:07 2010
> @@ -87,6 +87,26 @@
>   return CI;
> }
> 
> +/// EmitStrNCpy - Emit a call to the strcpy function to the builder, for the
> +/// specified pointer arguments.
> +Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,
> +                         IRBuilder<> &B, const TargetData *TD) {

s/strcpy/strncpy/




From echristo at apple.com  Thu Mar 11 10:05:49 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 08:05:49 -0800
Subject: [llvm-commits] [llvm] r98219 - in /llvm/trunk:
	include/llvm/Transforms/Utils/BuildLibCalls.h
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	lib/Transforms/Utils/BuildLibCalls.cpp
In-Reply-To: <B4BAA6F8-BFBC-4A30-9407-7499D0373BFA@2pi.dk>
References: <20100311012507.61B652A6C12C@llvm.org>
	<B4BAA6F8-BFBC-4A30-9407-7499D0373BFA@2pi.dk>
Message-ID: <7708A4BA-682B-4025-8A78-74F011A1661E@apple.com>


On Mar 11, 2010, at 7:44 AM, Jakob Stoklund Olesen wrote:

> s/strcpy/strncpy/

Thanks :)

-eric


From richard at xmos.com  Thu Mar 11 10:26:35 2010
From: richard at xmos.com (Richard Osborne)
Date: Thu, 11 Mar 2010 16:26:35 -0000
Subject: [llvm-commits] [llvm] r98258 - in /llvm/trunk:
 lib/Target/XCore/XCoreISelLowering.cpp test/CodeGen/XCore/mul64.ll
Message-ID: <20100311162635.3AFBA2A6C12C@llvm.org>

Author: friedgold
Date: Thu Mar 11 10:26:35 2010
New Revision: 98258

URL: http://llvm.org/viewvc/llvm-project?rev=98258&view=rev
Log:
Add dag combine to simplify lmul(x, 0, a, b)

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
    llvm/trunk/test/CodeGen/XCore/mul64.ll

Modified: llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp?rev=98258&r1=98257&r2=98258&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelLowering.cpp Thu Mar 11 10:26:35 2010
@@ -1345,6 +1345,33 @@
     }
   }
   break;
+  case XCoreISD::LMUL: {
+    SDValue N0 = N->getOperand(0);
+    SDValue N1 = N->getOperand(1);
+    SDValue N2 = N->getOperand(2);
+    SDValue N3 = N->getOperand(3);
+    ConstantSDNode *N0C = dyn_cast<ConstantSDNode>(N0);
+    ConstantSDNode *N1C = dyn_cast<ConstantSDNode>(N1);
+    EVT VT = N0.getValueType();
+    // Canonicalize multiplicative constant to RHS. If both multiplicative
+    // operands are constant canonicalize smallest to RHS.
+    if ((N0C && !N1C) ||
+        (N0C && N1C && N0C->getZExtValue() < N1C->getZExtValue()))
+      return DAG.getNode(XCoreISD::LMUL, dl, DAG.getVTList(VT, VT), N1, N0, N2, N3);
+
+    // lmul(x, 0, a, b)
+    if (N1C && N1C->isNullValue()) {
+      // If the high result is unused fold to add(a, b)
+      if (N->hasNUsesOfValue(0, 0)) {
+        SDValue Lo = DAG.getNode(ISD::ADD, dl, VT, N2, N3);
+        SDValue Ops [] = { Lo, Lo };
+        return DAG.getMergeValues(Ops, 2, dl);
+      }
+      // Otherwise fold to ladd(a, b, 0)
+      return DAG.getNode(XCoreISD::LADD, dl, DAG.getVTList(VT, VT), N2, N3, N1);
+    }
+  }
+  break;
   case ISD::ADD: {
     // Fold 32 bit expressions such as add(add(mul(x,y),a),b) ->
     // lmul(x, y, a, b). The high result of lmul will be ignored.

Modified: llvm/trunk/test/CodeGen/XCore/mul64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/mul64.ll?rev=98258&r1=98257&r2=98258&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/XCore/mul64.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/mul64.ll Thu Mar 11 10:26:35 2010
@@ -37,3 +37,16 @@
 ; CHECK-NEXT: mul r0, r0, r3
 ; CHECK-NEXT: lmul r0, r1, r1, r2, r11, r0
 ; CHECK-NEXT: mov r0, r4
+
+define i64 @mul64_2(i64 %a, i32 %b) {
+entry:
+	%0 = zext i32 %b to i64
+	%1 = mul i64 %a, %0
+	ret i64 %1
+}
+; CHECK: mul64_2:
+; CHECK: ldc r3, 0
+; CHECK-NEXT: lmul r3, r0, r0, r2, r3, r3
+; CHECK-NEXT: mul r1, r1, r2
+; CHECK-NEXT: add r1, r3, r1
+; CHECK-NEXT: retsp 0




From dpatel at apple.com  Thu Mar 11 11:36:28 2010
From: dpatel at apple.com (Devang Patel)
Date: Thu, 11 Mar 2010 09:36:28 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <bb55930e1003102207s6f914986qa8599f9a87610bef@mail.gmail.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
	<2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
	<5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>
	<13ECEBF5-92DB-43D5-9583-D606C640746B@apple.com>
	<bb55930e1003102207s6f914986qa8599f9a87610bef@mail.gmail.com>
Message-ID: <7A67FDFA-9B81-438E-856B-97324269861D@apple.com>

On Mar 10, 2010, at 10:07 PM, Jeffrey Yasskin wrote:

> On Wed, Mar 10, 2010 at 4:47 PM, Devang Patel <dpatel at apple.com> wrote:
>> 
>> On Mar 10, 2010, at 4:01 PM, Jeffrey Yasskin wrote:
>> 
>>> On Tue, Mar 9, 2010 at 12:22 PM, Devang Patel <dpatel at apple.com> wrote:
>>>> 
>>>> On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:
>>>> 
>>>>> Now that nothing uses the result of constructCompileUnit() unless it
>>>>> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
>>>>> avoid new'ing the DIE and CompileUnit?
>>>> 
>>>> With the introduction of DIFile, we are moving away from "isMain" field.
>>>> 
>>>>> Then, if we delete ModuleCU in
>>>>> endModule(), this memory leak would go away.
>>>> 
>>>> We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.
>>> 
>>> I can take it. I assume it's correct to simply not create the
>>> CompileUnits and DIEs whose addresses aren't obviously passed
>>> anywhere? Or do I need to look out for constructors or other functions
>>> that escape their addresses surreptitiously?
>> 
>> Now there is only one compile unit so it is simpler. Regarding other DIEs, it is expected that DIEs are created only if they will be emitted in the output. In other words, unused DIEs are not created. It would good to find and fix leaks in this area.
>> -
> 
> Here's my patch to delete the ModuleCU at the end of its Module, and
> to avoid creating unused CompileUnits and DIEs. I'll commit it
> tomorrow morning unless you have comments.
> <free_compile_unit.patch>

Looks good. Please delete ModuleCU before stoping timer.

Thanks!
-
Devang




From echristo at apple.com  Thu Mar 11 11:45:38 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 17:45:38 -0000
Subject: [llvm-commits] [llvm] r98260 -
	/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100311174538.66E012A6C12C@llvm.org>

Author: echristo
Date: Thu Mar 11 11:45:38 2010
New Revision: 98260

URL: http://llvm.org/viewvc/llvm-project?rev=98260&view=rev
Log:
Fix typo.

Modified:
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98260&r1=98259&r2=98260&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Thu Mar 11 11:45:38 2010
@@ -87,7 +87,7 @@
   return CI;
 }
 
-/// EmitStrNCpy - Emit a call to the strcpy function to the builder, for the
+/// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
 /// specified pointer arguments.
 Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,
                          IRBuilder<> &B, const TargetData *TD) {




From daniel at zuster.org  Thu Mar 11 12:22:51 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 18:22:51 -0000
Subject: [llvm-commits] [llvm] r98266 - in /llvm/trunk:
 include/llvm/MC/MCAssembler.h lib/MC/MCAssembler.cpp
Message-ID: <20100311182251.5AB0A2A6C12C@llvm.org>

Author: ddunbar
Date: Thu Mar 11 12:22:51 2010
New Revision: 98266

URL: http://llvm.org/viewvc/llvm-project?rev=98266&view=rev
Log:
MC/Mach-O: Add MCSymbolData::getAddress() utility.

Modified:
    llvm/trunk/include/llvm/MC/MCAssembler.h
    llvm/trunk/lib/MC/MCAssembler.cpp

Modified: llvm/trunk/include/llvm/MC/MCAssembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAssembler.h?rev=98266&r1=98265&r2=98266&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAssembler.h (original)
+++ llvm/trunk/include/llvm/MC/MCAssembler.h Thu Mar 11 12:22:51 2010
@@ -507,6 +507,11 @@
   uint64_t getOffset() const { return Offset; }
   void setOffset(uint64_t Value) { Offset = Value; }
 
+  uint64_t getAddress() const {
+    assert(getFragment() && "Invalid getAddress() on undefined symbol!");
+    return getFragment()->getAddress() + getOffset();
+  }
+
   /// @}
   /// @name Symbol Attributes
   /// @{

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=98266&r1=98265&r2=98266&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Thu Mar 11 12:22:51 2010
@@ -400,7 +400,7 @@
       if (Symbol.isAbsolute()) {
         llvm_unreachable("FIXME: Not yet implemented!");
       } else {
-        Address = Data.getFragment()->getAddress() + Data.getOffset();
+        Address = Data.getAddress();
       }
     } else if (Data.isCommon()) {
       // Common symbols are encoded with the size in the address
@@ -452,7 +452,7 @@
       llvm_report_error("symbol '" + A->getName() +
                         "' can not be undefined in a subtraction expression");
 
-    uint32_t Value = A_SD->getFragment()->getAddress() + A_SD->getOffset();
+    uint32_t Value = A_SD->getAddress();
     uint32_t Value2 = 0;
 
     if (const MCSymbol *B = Target.getSymB()) {
@@ -468,7 +468,7 @@
       // relocation types from the linkers point of view, this is done solely
       // for pedantic compatibility with 'as'.
       Type = A_SD->isExternal() ? RIT_Difference : RIT_LocalDifference;
-      Value2 = B_SD->getFragment()->getAddress() + B_SD->getOffset();
+      Value2 = B_SD->getAddress();
     }
 
     // The value which goes in the fixup is current value of the expression.
@@ -561,7 +561,7 @@
           if (&*it == SD->getFragment()->getParent())
             break;
         assert(it != ie && "Unable to find section index!");
-        Value = SD->getFragment()->getAddress() + SD->getOffset();
+        Value = SD->getAddress();
       }
 
       Type = RIT_Vanilla;




From jyasskin at google.com  Thu Mar 11 12:29:55 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Thu, 11 Mar 2010 18:29:55 -0000
Subject: [llvm-commits] [llvm] r98268 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <20100311182955.C77AB2A6C12C@llvm.org>

Author: jyasskin
Date: Thu Mar 11 12:29:55 2010
New Revision: 98268

URL: http://llvm.org/viewvc/llvm-project?rev=98268&view=rev
Log:
Avoid leaking CompileUnits in DwarfDebug.cpp.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98268&r1=98267&r2=98268&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Thu Mar 11 12:29:55 2010
@@ -52,9 +52,9 @@
 
   /// Die - Compile unit debug information entry.
   ///
-  DIE *CUDie;
+  OwningPtr<DIE> CUDie;
 
-  /// IndexTyDie - An anonymous type for index type.
+  /// IndexTyDie - An anonymous type for index type.  Owned by CUDie.
   DIE *IndexTyDie;
 
   /// GVToDieMap - Tracks the mapping of unit level debug informaton
@@ -78,11 +78,10 @@
 public:
   CompileUnit(unsigned I, DIE *D)
     : ID(I), CUDie(D), IndexTyDie(0) {}
-  ~CompileUnit() { delete CUDie; delete IndexTyDie; }
 
   // Accessors.
   unsigned getID()                  const { return ID; }
-  DIE* getCUDie()                   const { return CUDie; }
+  DIE* getCUDie()                   const { return CUDie.get(); }
   const StringMap<DIE*> &getGlobals()     const { return Globals; }
   const StringMap<DIE*> &getGlobalTypes() const { return GlobalTypes; }
 
@@ -1622,8 +1621,12 @@
   return NDie;
 }
 
-CompileUnit *DwarfDebug::constructCompileUnit(MDNode *N) {
+void DwarfDebug::constructCompileUnit(MDNode *N) {
   DICompileUnit DIUnit(N);
+  // Use first compile unit marked as isMain as the compile unit for this
+  // module.
+  if (ModuleCU || !DIUnit.isMain())
+    return;
   StringRef FN = DIUnit.getFilename();
   StringRef Dir = DIUnit.getDirectory();
   unsigned ID = GetOrCreateSourceID(Dir, FN);
@@ -1653,14 +1656,9 @@
     addUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
             dwarf::DW_FORM_data1, RVer);
 
-  CompileUnit *Unit = new CompileUnit(ID, Die);
-  if (!ModuleCU && DIUnit.isMain()) {
-    // Use first compile unit marked as isMain as the compile unit
-    // for this module.
-    ModuleCU = Unit;
-  }
-
-  return Unit;
+  assert(!ModuleCU &&
+         "ModuleCU assigned since the top of constructCompileUnit");
+  ModuleCU = new CompileUnit(ID, Die);
 }
 
 void DwarfDebug::constructGlobalVariableDIE(MDNode *N) {
@@ -1898,6 +1896,9 @@
   // Emit inline info.
   emitDebugInlineInfo();
 
+  delete ModuleCU;
+  ModuleCU = NULL;  // Reset for the next Module, if any.
+
   if (TimePassesIsEnabled)
     DebugTimer->stopTimer();
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98268&r1=98267&r2=98268&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Thu Mar 11 12:29:55 2010
@@ -473,7 +473,7 @@
   /// as well.
   unsigned GetOrCreateSourceID(StringRef DirName, StringRef FileName);
 
-  CompileUnit *constructCompileUnit(MDNode *N);
+  void constructCompileUnit(MDNode *N);
 
   void constructGlobalVariableDIE(MDNode *N);
 




From jyasskin at google.com  Thu Mar 11 12:30:40 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Thu, 11 Mar 2010 10:30:40 -0800
Subject: [llvm-commits] [llvm] r97990 - in /llvm/trunk:
	include/llvm/Analysis/DebugInfo.h lib/Analysis/DebugInfo.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
In-Reply-To: <7A67FDFA-9B81-438E-856B-97324269861D@apple.com>
References: <20100308220250.5FC1F2A6C12E@llvm.org>
	<bb55930e1003091054n32d4637aqf9159a9dabe6e84e@mail.gmail.com>
	<2AAF81E1-7C13-4D5D-9738-A5741CB779EB@apple.com>
	<5d44f72f1003101601p77c3abedl92978203ae5a2d87@mail.gmail.com>
	<13ECEBF5-92DB-43D5-9583-D606C640746B@apple.com>
	<bb55930e1003102207s6f914986qa8599f9a87610bef@mail.gmail.com>
	<7A67FDFA-9B81-438E-856B-97324269861D@apple.com>
Message-ID: <5d44f72f1003111030j3fc63f7cvd79e9401352183e0@mail.gmail.com>

On Thu, Mar 11, 2010 at 9:36 AM, Devang Patel <dpatel at apple.com> wrote:
> On Mar 10, 2010, at 10:07 PM, Jeffrey Yasskin wrote:
>
>> On Wed, Mar 10, 2010 at 4:47 PM, Devang Patel <dpatel at apple.com> wrote:
>>>
>>> On Mar 10, 2010, at 4:01 PM, Jeffrey Yasskin wrote:
>>>
>>>> On Tue, Mar 9, 2010 at 12:22 PM, Devang Patel <dpatel at apple.com> wrote:
>>>>>
>>>>> On Mar 9, 2010, at 10:54 AM, Jeffrey Yasskin wrote:
>>>>>
>>>>>> Now that nothing uses the result of constructCompileUnit() unless it
>>>>>> assigns the ModuleCU, can we just bail early when !DIUnit.isMain() and
>>>>>> avoid new'ing the DIE and CompileUnit?
>>>>>
>>>>> With the introduction of DIFile, we are moving away from "isMain" field.
>>>>>
>>>>>> Then, if we delete ModuleCU in
>>>>>> endModule(), this memory leak would go away.
>>>>>
>>>>> We should delete ModuleCU. That's an oversight. Please fix it or let me know, I'll do it.
>>>>
>>>> I can take it. I assume it's correct to simply not create the
>>>> CompileUnits and DIEs whose addresses aren't obviously passed
>>>> anywhere? Or do I need to look out for constructors or other functions
>>>> that escape their addresses surreptitiously?
>>>
>>> Now there is only one compile unit so it is simpler. Regarding other DIEs, it is expected that DIEs are created only if they will be emitted in the output. In other words, unused DIEs are not created. It would good to find and fix leaks in this area.
>>> -
>>
>> Here's my patch to delete the ModuleCU at the end of its Module, and
>> to avoid creating unused CompileUnits and DIEs. I'll commit it
>> tomorrow morning unless you have comments.
>> <free_compile_unit.patch>
>
> Looks good. Please delete ModuleCU before stoping timer.

Done and committed. Thanks.


From richard at xmos.com  Thu Mar 11 12:38:59 2010
From: richard at xmos.com (Richard Osborne)
Date: Thu, 11 Mar 2010 18:38:59 -0000
Subject: [llvm-commits] [llvm] r98269 -
	/llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
Message-ID: <20100311183859.F09E22A6C12C@llvm.org>

Author: friedgold
Date: Thu Mar 11 12:38:59 2010
New Revision: 98269

URL: http://llvm.org/viewvc/llvm-project?rev=98269&view=rev
Log:
Remove dead code. (S|U)MUL_LO is now lowered to LMUL or MACC(S|U)

Modified:
    llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp

Modified: llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp?rev=98269&r1=98268&r2=98269&view=diff
==============================================================================
--- llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/XCore/XCoreISelDAGToDAG.cpp Thu Mar 11 12:38:59 2010
@@ -173,29 +173,6 @@
         }
         break;
       }
-      case ISD::SMUL_LOHI: {
-        // FIXME fold addition into the macc instruction
-        SDValue Zero(CurDAG->getMachineNode(XCore::LDC_ru6, dl, MVT::i32,
-                                CurDAG->getTargetConstant(0, MVT::i32)), 0);
-        SDValue Ops[] = { Zero, Zero, N->getOperand(0), N->getOperand(1) };
-        SDNode *ResNode = CurDAG->getMachineNode(XCore::MACCS_l4r, dl,
-                                                 MVT::i32, MVT::i32, Ops, 4);
-        ReplaceUses(SDValue(N, 0), SDValue(ResNode, 1));
-        ReplaceUses(SDValue(N, 1), SDValue(ResNode, 0));
-        return NULL;
-      }
-      case ISD::UMUL_LOHI: {
-        // FIXME fold addition into the macc / lmul instruction
-        SDValue Zero(CurDAG->getMachineNode(XCore::LDC_ru6, dl, MVT::i32,
-                                  CurDAG->getTargetConstant(0, MVT::i32)), 0);
-        SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
-                            Zero, Zero };
-        SDNode *ResNode = CurDAG->getMachineNode(XCore::LMUL_l6r, dl, MVT::i32,
-                                                 MVT::i32, Ops, 4);
-        ReplaceUses(SDValue(N, 0), SDValue(ResNode, 1));
-        ReplaceUses(SDValue(N, 1), SDValue(ResNode, 0));
-        return NULL;
-      }
       case XCoreISD::LADD: {
         SDValue Ops[] = { N->getOperand(0), N->getOperand(1),
                             N->getOperand(2) };




From evan.cheng at apple.com  Thu Mar 11 12:49:15 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 11 Mar 2010 18:49:15 -0000
Subject: [llvm-commits] [llvm] r98270 - in /llvm/trunk:
 lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/tailcallfp2.ll
Message-ID: <20100311184915.389552A6C12C@llvm.org>

Author: evancheng
Date: Thu Mar 11 12:49:14 2010
New Revision: 98270

URL: http://llvm.org/viewvc/llvm-project?rev=98270&view=rev
Log:
Bad bad bug. x86 force indirect tail call address into eax when it's meant to force it into a call preserved register instead. Change it to ecx for now.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/tailcallfp2.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98270&r1=98269&r2=98270&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Mar 11 12:49:14 2010
@@ -2091,7 +2091,7 @@
     // tailcall must happen after callee-saved registers are poped.
     // FIXME: Give it a special register class that contains caller-saved
     // register instead?
-    unsigned TCReg = Is64Bit ? X86::R11 : X86::EAX;
+    unsigned TCReg = Is64Bit ? X86::R11 : X86::ECX;
     Chain = DAG.getCopyToReg(Chain,  dl,
                              DAG.getRegister(TCReg, getPointerTy()),
                              Callee,InFlag);
@@ -2145,7 +2145,7 @@
     }
 
     assert(((Callee.getOpcode() == ISD::Register &&
-               (cast<RegisterSDNode>(Callee)->getReg() == X86::EAX ||
+               (cast<RegisterSDNode>(Callee)->getReg() == X86::ECX ||
                 cast<RegisterSDNode>(Callee)->getReg() == X86::R11)) ||
               Callee.getOpcode() == ISD::TargetExternalSymbol ||
               Callee.getOpcode() == ISD::TargetGlobalAddress) &&

Modified: llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcallfp2.ll?rev=98270&r1=98269&r2=98270&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/tailcallfp2.ll (original)
+++ llvm/trunk/test/CodeGen/X86/tailcallfp2.ll Thu Mar 11 12:49:14 2010
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%eax}
+; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%ecx}
 
 declare i32 @putchar(i32)
 




From evan.cheng at apple.com  Thu Mar 11 13:12:35 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 11 Mar 2010 11:12:35 -0800
Subject: [llvm-commits] [llvm] r98270 - in /llvm/trunk:
	lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/tailcallfp2.ll
In-Reply-To: <20100311184915.389552A6C12C@llvm.org>
References: <20100311184915.389552A6C12C@llvm.org>
Message-ID: <0B24D86E-C9FB-425A-A4DA-C2BFD54B8ACF@apple.com>

I should note this is only an issue when MachineModuleInfo callsEHReturn() returns true.

Evan

On Mar 11, 2010, at 10:49 AM, Evan Cheng wrote:

> Author: evancheng
> Date: Thu Mar 11 12:49:14 2010
> New Revision: 98270
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98270&view=rev
> Log:
> Bad bad bug. x86 force indirect tail call address into eax when it's meant to force it into a call preserved register instead. Change it to ecx for now.
> 
> Modified:
>    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
>    llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
> 
> Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98270&r1=98269&r2=98270&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Mar 11 12:49:14 2010
> @@ -2091,7 +2091,7 @@
>     // tailcall must happen after callee-saved registers are poped.
>     // FIXME: Give it a special register class that contains caller-saved
>     // register instead?
> -    unsigned TCReg = Is64Bit ? X86::R11 : X86::EAX;
> +    unsigned TCReg = Is64Bit ? X86::R11 : X86::ECX;
>     Chain = DAG.getCopyToReg(Chain,  dl,
>                              DAG.getRegister(TCReg, getPointerTy()),
>                              Callee,InFlag);
> @@ -2145,7 +2145,7 @@
>     }
> 
>     assert(((Callee.getOpcode() == ISD::Register &&
> -               (cast<RegisterSDNode>(Callee)->getReg() == X86::EAX ||
> +               (cast<RegisterSDNode>(Callee)->getReg() == X86::ECX ||
>                 cast<RegisterSDNode>(Callee)->getReg() == X86::R11)) ||
>               Callee.getOpcode() == ISD::TargetExternalSymbol ||
>               Callee.getOpcode() == ISD::TargetGlobalAddress) &&
> 
> Modified: llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcallfp2.ll?rev=98270&r1=98269&r2=98270&view=diff
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/tailcallfp2.ll (original)
> +++ llvm/trunk/test/CodeGen/X86/tailcallfp2.ll Thu Mar 11 12:49:14 2010
> @@ -1,4 +1,4 @@
> -; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%eax}
> +; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%ecx}
> 
> declare i32 @putchar(i32)
> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From echristo at apple.com  Thu Mar 11 13:24:34 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 19:24:34 -0000
Subject: [llvm-commits] [llvm] r98274 - in /llvm/trunk:
 include/llvm/Transforms/Utils/BuildLibCalls.h
 lib/Transforms/InstCombine/InstCombineCalls.cpp
 lib/Transforms/Scalar/CodeGenPrepare.cpp
 lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100311192434.E3BC22A6C12C@llvm.org>

Author: echristo
Date: Thu Mar 11 13:24:34 2010
New Revision: 98274

URL: http://llvm.org/viewvc/llvm-project?rev=98274&view=rev
Log:
Lower stpcpy_chk when possible.

Modified:
    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h?rev=98274&r1=98273&r2=98274&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h (original)
+++ llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h Thu Mar 11 13:24:34 2010
@@ -39,6 +39,11 @@
   Value *EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
                     const TargetData *TD);
   
+  /// EmitStpCpy - Emit a call to the stpcpy function to the builder, for the
+  /// specified pointer arguments.
+  Value *EmitStpCpy(Value *Dst, Value *Src, IRBuilder<> &B,
+                    const TargetData *TD);
+  
   /// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
   /// specified pointer arguments and length.
   Value *EmitStrNCpy(Value *Dst, Value *Src, Value *Len, IRBuilder<> &B,

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=98274&r1=98273&r2=98274&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Thu Mar 11 13:24:34 2010
@@ -838,6 +838,19 @@
 
   // Should be similar to strcpy.
   if (Name == "__stpcpy_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
+    if (!SizeCI)
+      return 0;
+    // If a) we don't have any length information, or b) we know this will
+    // fit then just lower to a plain stpcpy. Otherwise we'll keep our
+    // stpcpy_chk call which may fail at runtime if the size is too long.
+    // TODO: It might be nice to get a maximum length out of the possible
+    // string lengths for varying.
+    if (SizeCI->isAllOnesValue() ||
+        SizeCI->getZExtValue() >= GetStringLength(CI->getOperand(2))) {
+      Value *Ret = EmitStpCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      return ReplaceInstUsesWith(*CI, Ret);
+    }
     return 0;
   }
 

Modified: llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp?rev=98274&r1=98273&r2=98274&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp Thu Mar 11 13:24:34 2010
@@ -624,11 +624,6 @@
     ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
     if (!SizeCI)
       return 0;
-    // If a) we don't have any length information, or b) we know this will
-    // fit then just lower to a plain strcpy. Otherwise we'll keep our
-    // strcpy_chk call which may fail at runtime if the size is too long.
-    // TODO: It might be nice to get a maximum length out of the possible
-    // string lengths for varying.
     if (SizeCI->isAllOnesValue()) {
       Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
       CI->replaceAllUsesWith(Ret);
@@ -638,8 +633,16 @@
     return 0;
   }
 
-  // Should be similar to strcpy.
   if (Name == "__stpcpy_chk") {
+    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
+    if (!SizeCI)
+      return 0;
+    if (SizeCI->isAllOnesValue()) {
+      Value *Ret = EmitStpCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      CI->replaceAllUsesWith(Ret);
+      CI->eraseFromParent();
+      return true;
+    }
     return 0;
   }
 

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98274&r1=98273&r2=98274&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Thu Mar 11 13:24:34 2010
@@ -87,6 +87,24 @@
   return CI;
 }
 
+/// EmitStpCpy - Emit a call to the stpcpy function to the builder, for the
+/// specified pointer arguments.
+Value *llvm::EmitStpCpy(Value *Dst, Value *Src, IRBuilder<> &B,
+                        const TargetData *TD) {
+  Module *M = B.GetInsertBlock()->getParent()->getParent();
+  AttributeWithIndex AWI[2];
+  AWI[0] = AttributeWithIndex::get(2, Attribute::NoCapture);
+  AWI[1] = AttributeWithIndex::get(~0u, Attribute::NoUnwind);
+  const Type *I8Ptr = B.getInt8PtrTy();
+  Value *StpCpy = M->getOrInsertFunction("stpcpy", AttrListPtr::get(AWI, 2),
+                                         I8Ptr, I8Ptr, I8Ptr, NULL);
+  CallInst *CI = B.CreateCall2(StpCpy, CastToCStr(Dst, B), CastToCStr(Src, B),
+                               "stpcpy");
+  if (const Function *F = dyn_cast<Function>(StpCpy->stripPointerCasts()))
+    CI->setCallingConv(F->getCallingConv());
+  return CI;
+}
+
 /// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
 /// specified pointer arguments.
 Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,




From evan.cheng at apple.com  Thu Mar 11 13:38:18 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 11 Mar 2010 19:38:18 -0000
Subject: [llvm-commits] [llvm] r98277 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
Message-ID: <20100311193818.846682A6C12C@llvm.org>

Author: evancheng
Date: Thu Mar 11 13:38:18 2010
New Revision: 98277

URL: http://llvm.org/viewvc/llvm-project?rev=98277&view=rev
Log:
In case of tail call size of Ins and InVals may not match.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=98277&r1=98276&r2=98277&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Thu Mar 11 13:38:18 2010
@@ -5777,12 +5777,6 @@
          "LowerCall emitted a return value for a tail call!");
   assert((isTailCall || InVals.size() == Ins.size()) &&
          "LowerCall didn't emit the correct number of values!");
-  DEBUG(for (unsigned i = 0, e = Ins.size(); i != e; ++i) {
-          assert(InVals[i].getNode() &&
-                 "LowerCall emitted a null value!");
-          assert(Ins[i].VT == InVals[i].getValueType() &&
-                 "LowerCall emitted a value with the wrong type!");
-        });
 
   // For a tail call, the return value is merely live-out and there aren't
   // any nodes in the DAG representing it. Return a special value to
@@ -5793,6 +5787,13 @@
     return std::make_pair(SDValue(), SDValue());
   }
 
+  DEBUG(for (unsigned i = 0, e = Ins.size(); i != e; ++i) {
+          assert(InVals[i].getNode() &&
+                 "LowerCall emitted a null value!");
+          assert(Ins[i].VT == InVals[i].getValueType() &&
+                 "LowerCall emitted a value with the wrong type!");
+        });
+
   // Collect the legal value parts into potentially illegal values
   // that correspond to the original function's return values.
   ISD::NodeType AssertOp = ISD::DELETED_NODE;




From sabre at nondot.org  Thu Mar 11 13:41:58 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 19:41:58 -0000
Subject: [llvm-commits] [llvm] r98278 - in /llvm/trunk:
 include/llvm/CodeGen/TargetLoweringObjectFileImpl.h
 include/llvm/Target/TargetLoweringObjectFile.h
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
 lib/CodeGen/TargetLoweringObjectFileImpl.cpp
 lib/Target/TargetLoweringObjectFile.cpp
 lib/Target/X86/X86TargetObjectFile.cpp lib/Target/X86/X86TargetObjectFile.h
Message-ID: <20100311194159.053D92A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 13:41:58 2010
New Revision: 98278

URL: http://llvm.org/viewvc/llvm-project?rev=98278&view=rev
Log:
rename getSymbolForDwarf* to getExprForDwarf* since it returns
an MCExpr and not an MCSymbol.  Change it to take an MCStreamer,
which is currently unused.

No functionality change.


Modified:
    llvm/trunk/include/llvm/CodeGen/TargetLoweringObjectFileImpl.h
    llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
    llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp
    llvm/trunk/lib/Target/X86/X86TargetObjectFile.h

Modified: llvm/trunk/include/llvm/CodeGen/TargetLoweringObjectFileImpl.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/TargetLoweringObjectFileImpl.h?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/TargetLoweringObjectFileImpl.h (original)
+++ llvm/trunk/include/llvm/CodeGen/TargetLoweringObjectFileImpl.h Thu Mar 11 13:41:58 2010
@@ -78,12 +78,13 @@
   SelectSectionForGlobal(const GlobalValue *GV, SectionKind Kind,
                          Mangler *Mang, const TargetMachine &TM) const;
 
-  /// getSymbolForDwarfGlobalReference - Return an MCExpr to use for a reference
+  /// getExprForDwarfGlobalReference - Return an MCExpr to use for a reference
   /// to the specified global variable from exception handling information.
   ///
   virtual const MCExpr *
-  getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                              MachineModuleInfo *MMI, unsigned Encoding) const;
+  getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                                 MachineModuleInfo *MMI, unsigned Encoding,
+                                 MCStreamer &Streamer) const;
 };
 
 
@@ -166,11 +167,12 @@
     return NonLazySymbolPointerSection;
   }
 
-  /// getSymbolForDwarfGlobalReference - The mach-o version of this method
+  /// getExprForDwarfGlobalReference - The mach-o version of this method
   /// defaults to returning a stub reference.
   virtual const MCExpr *
-  getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                              MachineModuleInfo *MMI, unsigned Encoding) const;
+  getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                                 MachineModuleInfo *MMI, unsigned Encoding,
+                                 MCStreamer &Streamer) const;
 
   virtual unsigned getPersonalityEncoding() const;
   virtual unsigned getLSDAEncoding() const;

Modified: llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h (original)
+++ llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h Thu Mar 11 13:41:58 2010
@@ -22,11 +22,12 @@
   class MachineModuleInfo;
   class Mangler;
   class MCAsmInfo;
+  class MCContext;
   class MCExpr;
   class MCSection;
   class MCSectionMachO;
   class MCSymbol;
-  class MCContext;
+  class MCStreamer;
   class GlobalValue;
   class TargetMachine;
   
@@ -197,17 +198,20 @@
     return 0;
   }
   
-  /// getSymbolForDwarfGlobalReference - Return an MCExpr to use for a reference
+  /// getExprForDwarfGlobalReference - Return an MCExpr to use for a reference
   /// to the specified global variable from exception handling information.
   ///
   virtual const MCExpr *
-  getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                              MachineModuleInfo *MMI, unsigned Encoding) const;
-
-  virtual const MCExpr *
-  getSymbolForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
-                             unsigned Encoding) const;
-
+  getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                                 MachineModuleInfo *MMI, unsigned Encoding,
+                                 MCStreamer &Streamer) const;
+
+  /// 
+  const MCExpr *
+  getExprForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
+                           unsigned Encoding,
+                           MCStreamer &Streamer) const;
+  
   virtual unsigned getPersonalityEncoding() const;
   virtual unsigned getLSDAEncoding() const;
   virtual unsigned getFDEEncoding() const;

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Thu Mar 11 13:41:58 2010
@@ -189,7 +189,8 @@
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
-  const MCExpr *Exp = TLOF.getSymbolForDwarfReference(Sym, Asm->MMI, Encoding);
+  const MCExpr *Exp = TLOF.getExprForDwarfReference(Sym, Asm->MMI, Encoding,
+                                                    Asm->OutStreamer);
   Asm->OutStreamer.EmitValue(Exp, SizeOfEncodedValue(Encoding), /*addrspace*/0);
 }
 
@@ -197,7 +198,8 @@
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
   const MCExpr *Exp =
-    TLOF.getSymbolForDwarfGlobalReference(GV, Asm->Mang, Asm->MMI, Encoding);
+    TLOF.getExprForDwarfGlobalReference(GV, Asm->Mang, Asm->MMI, Encoding,
+                                        Asm->OutStreamer);
   Asm->OutStreamer.EmitValue(Exp, SizeOfEncodedValue(Encoding), /*addrspace*/0);
 }
 

Modified: llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Thu Mar 11 13:41:58 2010
@@ -391,8 +391,9 @@
 }
 
 const MCExpr *TargetLoweringObjectFileELF::
-getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                             MachineModuleInfo *MMI, unsigned Encoding) const {
+getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                               MachineModuleInfo *MMI,
+                               unsigned Encoding, MCStreamer &Streamer) const {
 
   if (Encoding & dwarf::DW_EH_PE_indirect) {
     MachineModuleInfoELF &ELFMMI = MMI->getObjFileInfo<MachineModuleInfoELF>();
@@ -420,12 +421,12 @@
     }
 
     return TargetLoweringObjectFile::
-      getSymbolForDwarfReference(Sym, MMI,
-                                 Encoding & ~dwarf::DW_EH_PE_indirect);
+      getExprForDwarfReference(Sym, MMI,
+                               Encoding & ~dwarf::DW_EH_PE_indirect, Streamer);
   }
 
   return TargetLoweringObjectFile::
-    getSymbolForDwarfGlobalReference(GV, Mang, MMI, Encoding);
+    getExprForDwarfGlobalReference(GV, Mang, MMI, Encoding, Streamer);
 }
 
 //===----------------------------------------------------------------------===//
@@ -751,8 +752,9 @@
 }
 
 const MCExpr *TargetLoweringObjectFileMachO::
-getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                             MachineModuleInfo *MMI, unsigned Encoding) const {
+getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                               MachineModuleInfo *MMI, unsigned Encoding,
+                               MCStreamer &Streamer) const {
   // The mach-o version of this method defaults to returning a stub reference.
 
   if (Encoding & DW_EH_PE_indirect) {
@@ -782,12 +784,12 @@
     }
 
     return TargetLoweringObjectFile::
-      getSymbolForDwarfReference(Sym, MMI,
-                                 Encoding & ~dwarf::DW_EH_PE_indirect);
+      getExprForDwarfReference(Sym, MMI, Encoding & ~dwarf::DW_EH_PE_indirect,
+                               Streamer);
   }
 
   return TargetLoweringObjectFile::
-    getSymbolForDwarfGlobalReference(GV, Mang, MMI, Encoding);
+    getExprForDwarfGlobalReference(GV, Mang, MMI, Encoding, Streamer);
 }
 
 unsigned TargetLoweringObjectFileMachO::getPersonalityEncoding() const {

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Thu Mar 11 13:41:58 2010
@@ -290,12 +290,13 @@
   return DataSection;
 }
 
-/// getSymbolForDwarfGlobalReference - Return an MCExpr to use for a
+/// getExprForDwarfGlobalReference - Return an MCExpr to use for a
 /// reference to the specified global variable from exception
 /// handling information.
 const MCExpr *TargetLoweringObjectFile::
-getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                             MachineModuleInfo *MMI, unsigned Encoding) const {
+getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                               MachineModuleInfo *MMI, unsigned Encoding,
+                               MCStreamer &Streamer) const {
   // FIXME: Use GetGlobalValueSymbol.
   SmallString<128> Name;
   Mang->getNameWithPrefix(Name, GV, false);
@@ -306,12 +307,12 @@
   else
     Sym = getContext().GetOrCreateSymbol(Name.str());
 
-  return getSymbolForDwarfReference(Sym, MMI, Encoding);
+  return getExprForDwarfReference(Sym, MMI, Encoding, Streamer);
 }
 
 const MCExpr *TargetLoweringObjectFile::
-getSymbolForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
-                           unsigned Encoding) const {
+getExprForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
+                         unsigned Encoding, MCStreamer &Streamer) const {
   const MCExpr *Res = MCSymbolRefExpr::Create(Sym, getContext());
 
   switch (Encoding & 0xF0) {

Modified: llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetObjectFile.cpp Thu Mar 11 13:41:58 2010
@@ -19,9 +19,9 @@
 using namespace dwarf;
 
 const MCExpr *X8664_MachoTargetObjectFile::
-getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                                 MachineModuleInfo *MMI, 
-                                 unsigned Encoding) const {
+getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                               MachineModuleInfo *MMI, unsigned Encoding,
+                               MCStreamer &Streamer) const {
 
   // On Darwin/X86-64, we can reference dwarf symbols with foo at GOTPCREL+4, which
   // is an indirect pc-relative reference.
@@ -40,7 +40,7 @@
   }
 
   return TargetLoweringObjectFileMachO::
-    getSymbolForDwarfGlobalReference(GV, Mang, MMI, Encoding);
+    getExprForDwarfGlobalReference(GV, Mang, MMI, Encoding, Streamer);
 }
 
 unsigned X8632_ELFTargetObjectFile::getPersonalityEncoding() const {

Modified: llvm/trunk/lib/Target/X86/X86TargetObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetObjectFile.h?rev=98278&r1=98277&r2=98278&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetObjectFile.h (original)
+++ llvm/trunk/lib/Target/X86/X86TargetObjectFile.h Thu Mar 11 13:41:58 2010
@@ -23,8 +23,9 @@
   public:
 
     virtual const MCExpr *
-    getSymbolForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
-                              MachineModuleInfo *MMI, unsigned Encoding) const;
+    getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
+                                   MachineModuleInfo *MMI, unsigned Encoding,
+                                   MCStreamer &Streamer) const;
   };
 
   class X8632_ELFTargetObjectFile : public TargetLoweringObjectFileELF {




From evan.cheng at apple.com  Thu Mar 11 13:44:06 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 11 Mar 2010 11:44:06 -0800
Subject: [llvm-commits] [llvm] r98270 - in /llvm/trunk:
	lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/tailcallfp2.ll
In-Reply-To: <0B24D86E-C9FB-425A-A4DA-C2BFD54B8ACF@apple.com>
References: <20100311184915.389552A6C12C@llvm.org>
	<0B24D86E-C9FB-425A-A4DA-C2BFD54B8ACF@apple.com>
Message-ID: <D15D3864-6337-4876-8DE9-FEC958446E98@apple.com>

Sorry, I think this is just my head malfunctioning. I'll back it out.

Evan

On Mar 11, 2010, at 11:12 AM, Evan Cheng wrote:

> I should note this is only an issue when MachineModuleInfo callsEHReturn() returns true.
> 
> Evan
> 
> On Mar 11, 2010, at 10:49 AM, Evan Cheng wrote:
> 
>> Author: evancheng
>> Date: Thu Mar 11 12:49:14 2010
>> New Revision: 98270
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=98270&view=rev
>> Log:
>> Bad bad bug. x86 force indirect tail call address into eax when it's meant to force it into a call preserved register instead. Change it to ecx for now.
>> 
>> Modified:
>>   llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
>>   llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
>> 
>> Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98270&r1=98269&r2=98270&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
>> +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Mar 11 12:49:14 2010
>> @@ -2091,7 +2091,7 @@
>>    // tailcall must happen after callee-saved registers are poped.
>>    // FIXME: Give it a special register class that contains caller-saved
>>    // register instead?
>> -    unsigned TCReg = Is64Bit ? X86::R11 : X86::EAX;
>> +    unsigned TCReg = Is64Bit ? X86::R11 : X86::ECX;
>>    Chain = DAG.getCopyToReg(Chain,  dl,
>>                             DAG.getRegister(TCReg, getPointerTy()),
>>                             Callee,InFlag);
>> @@ -2145,7 +2145,7 @@
>>    }
>> 
>>    assert(((Callee.getOpcode() == ISD::Register &&
>> -               (cast<RegisterSDNode>(Callee)->getReg() == X86::EAX ||
>> +               (cast<RegisterSDNode>(Callee)->getReg() == X86::ECX ||
>>                cast<RegisterSDNode>(Callee)->getReg() == X86::R11)) ||
>>              Callee.getOpcode() == ISD::TargetExternalSymbol ||
>>              Callee.getOpcode() == ISD::TargetGlobalAddress) &&
>> 
>> Modified: llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcallfp2.ll?rev=98270&r1=98269&r2=98270&view=diff
>> ==============================================================================
>> --- llvm/trunk/test/CodeGen/X86/tailcallfp2.ll (original)
>> +++ llvm/trunk/test/CodeGen/X86/tailcallfp2.ll Thu Mar 11 12:49:14 2010
>> @@ -1,4 +1,4 @@
>> -; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%eax}
>> +; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%ecx}
>> 
>> declare i32 @putchar(i32)
>> 
>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 




From isanbard at gmail.com  Thu Mar 11 13:50:32 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Thu, 11 Mar 2010 19:50:32 -0000
Subject: [llvm-commits] [llvm] r98281 - in /llvm/trunk:
 lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/tailcallfp2.ll
Message-ID: <20100311195032.3868C2A6C12C@llvm.org>

Author: void
Date: Thu Mar 11 13:50:31 2010
New Revision: 98281

URL: http://llvm.org/viewvc/llvm-project?rev=98281&view=rev
Log:
revert r98270.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/tailcallfp2.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98281&r1=98280&r2=98281&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Mar 11 13:50:31 2010
@@ -2091,7 +2091,7 @@
     // tailcall must happen after callee-saved registers are poped.
     // FIXME: Give it a special register class that contains caller-saved
     // register instead?
-    unsigned TCReg = Is64Bit ? X86::R11 : X86::ECX;
+    unsigned TCReg = Is64Bit ? X86::R11 : X86::EAX;
     Chain = DAG.getCopyToReg(Chain,  dl,
                              DAG.getRegister(TCReg, getPointerTy()),
                              Callee,InFlag);
@@ -2145,7 +2145,7 @@
     }
 
     assert(((Callee.getOpcode() == ISD::Register &&
-               (cast<RegisterSDNode>(Callee)->getReg() == X86::ECX ||
+               (cast<RegisterSDNode>(Callee)->getReg() == X86::EAX ||
                 cast<RegisterSDNode>(Callee)->getReg() == X86::R11)) ||
               Callee.getOpcode() == ISD::TargetExternalSymbol ||
               Callee.getOpcode() == ISD::TargetGlobalAddress) &&

Modified: llvm/trunk/test/CodeGen/X86/tailcallfp2.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/tailcallfp2.ll?rev=98281&r1=98280&r2=98281&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/tailcallfp2.ll (original)
+++ llvm/trunk/test/CodeGen/X86/tailcallfp2.ll Thu Mar 11 13:50:31 2010
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%ecx}
+; RUN: llc < %s -march=x86 -tailcallopt | grep {jmp} | grep {\\*%eax}
 
 declare i32 @putchar(i32)
 




From daniel at zuster.org  Thu Mar 11 13:53:22 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 11 Mar 2010 19:53:22 -0000
Subject: [llvm-commits] [zorg] r98282 -
 /zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist
Message-ID: <20100311195322.1461A2A6C12C@llvm.org>

Author: ddunbar
Date: Thu Mar 11 13:53:21 2010
New Revision: 98282

URL: http://llvm.org/viewvc/llvm-project?rev=98282&view=rev
Log:
Override the default number of processes, on OS X desktop this defaults to 512 which is low for a buildbot.

Also, fix a missing SoftResourceLimits line, which 10.5 launchd wasn't thrilled about.

Modified:
    zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist

Modified: zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist?rev=98282&r1=98281&r2=98282&view=diff
==============================================================================
--- zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist (original)
+++ zorg/trunk/buildbot/smooshlab/slave/org.llvm.smooshlab.slave.plist Thu Mar 11 13:53:21 2010
@@ -1,44 +1,49 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
 <plist version="1.0">
-<dict>
-         <key>Label</key>
-         <string>org.llvm.smooshlab.slave</string>
-
-         <!-- Change this to the user you want to run buildbot as -->
-         <key>UserName</key>
-         <string>buildslave</string>
-
-         <!-- Change this to your buildbot working directory -->
-         <key>WorkingDirectory</key>
-         <string>/Users/buildslave/zorg/buildbot/smooshlab/slave</string>
-
-         <key>ProgramArguments</key>
-         <array>
-                 <string>/usr/bin/twistd</string>
-                 <string>--nodaemon</string>
-                 <string>--python=buildbot.tac</string>
-                 <string>--logfile=twistd.log</string>
-                 <string>--prefix=smooshlab</string>
-         </array>
-
-         <key>KeepAlive</key>
-         <dict>
-                 <key>SuccessfulExit</key>
-                 <false/>
-         </dict>
-
-         <key>RunAtLoad</key>
-         <true/>
-
-        <dict>
-                <key>NumberOfFiles</key>
-                <integer>1024</integer>
-        </dict>
-	<key>HardResourceLimits</key>
-        <dict>
-                <key>NumberOfFiles</key>
-                <integer>1024</integer>
-        </dict>
-</dict>
+  <dict>
+    <key>Label</key>
+    <string>org.llvm.smooshlab.slave</string>
+
+    <!-- Change this to the user you want to run buildbot as -->
+    <key>UserName</key>
+    <string>buildslave</string>
+
+    <!-- Change this to your buildbot working directory -->
+    <key>WorkingDirectory</key>
+    <string>/Users/buildslave/zorg/buildbot/smooshlab/slave</string>
+
+    <key>ProgramArguments</key>
+    <array>
+      <string>/usr/bin/twistd</string>
+      <string>--nodaemon</string>
+      <string>--python=buildbot.tac</string>
+      <string>--logfile=twistd.log</string>
+      <string>--prefix=smooshlab</string>
+    </array>
+
+    <key>KeepAlive</key>
+    <dict>
+      <key>SuccessfulExit</key>
+      <false/>
+    </dict>
+
+    <key>RunAtLoad</key>
+    <true/>
+
+    <key>SoftResourceLimits</key>
+    <dict>
+      <key>NumberOfFiles</key>
+      <integer>1024</integer>
+      <key>NumberOfProcesses</key>
+      <integer>2048</integer>
+    </dict>
+    <key>HardResourceLimits</key>
+    <dict>
+      <key>NumberOfFiles</key>
+      <integer>1024</integer>
+      <key>NumberOfProcesses</key>
+      <integer>2048</integer>
+    </dict>
+  </dict>
 </plist>




From benny.kra at googlemail.com  Thu Mar 11 14:45:13 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 11 Mar 2010 20:45:13 -0000
Subject: [llvm-commits] [llvm] r98284 - in /llvm/trunk:
 include/llvm/Transforms/Utils/BuildLibCalls.h
 lib/Transforms/InstCombine/InstCombineCalls.cpp
 lib/Transforms/Scalar/CodeGenPrepare.cpp
 lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100311204513.D98322A6C12C@llvm.org>

Author: d0k
Date: Thu Mar 11 14:45:13 2010
New Revision: 98284

URL: http://llvm.org/viewvc/llvm-project?rev=98284&view=rev
Log:
stpcpy is so similar to strcpy, it doesn't deserve a complete copy of the __strcpy_chk -> strcpy code.

Modified:
    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h?rev=98284&r1=98283&r2=98284&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h (original)
+++ llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h Thu Mar 11 14:45:13 2010
@@ -37,13 +37,8 @@
   /// EmitStrCpy - Emit a call to the strcpy function to the builder, for the
   /// specified pointer arguments.
   Value *EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
-                    const TargetData *TD);
-  
-  /// EmitStpCpy - Emit a call to the stpcpy function to the builder, for the
-  /// specified pointer arguments.
-  Value *EmitStpCpy(Value *Dst, Value *Src, IRBuilder<> &B,
-                    const TargetData *TD);
-  
+                    const TargetData *TD, StringRef Name = "strcpy");
+
   /// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
   /// specified pointer arguments and length.
   Value *EmitStrNCpy(Value *Dst, Value *Src, Value *Len, IRBuilder<> &B,

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=98284&r1=98283&r2=98284&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Thu Mar 11 14:45:13 2010
@@ -819,36 +819,19 @@
     return 0;
   }
 
-  if (Name == "__strcpy_chk") {
+  if (Name == "__strcpy_chk" || Name == "__stpcpy_chk") {
     ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
     if (!SizeCI)
       return 0;
     // If a) we don't have any length information, or b) we know this will
-    // fit then just lower to a plain strcpy. Otherwise we'll keep our
-    // strcpy_chk call which may fail at runtime if the size is too long.
+    // fit then just lower to a plain st[rp]cpy. Otherwise we'll keep our
+    // st[rp]cpy_chk call which may fail at runtime if the size is too long.
     // TODO: It might be nice to get a maximum length out of the possible
     // string lengths for varying.
     if (SizeCI->isAllOnesValue() ||
       SizeCI->getZExtValue() >= GetStringLength(CI->getOperand(2))) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
-      return ReplaceInstUsesWith(*CI, Ret);
-    }
-    return 0;
-  }
-
-  // Should be similar to strcpy.
-  if (Name == "__stpcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeCI)
-      return 0;
-    // If a) we don't have any length information, or b) we know this will
-    // fit then just lower to a plain stpcpy. Otherwise we'll keep our
-    // stpcpy_chk call which may fail at runtime if the size is too long.
-    // TODO: It might be nice to get a maximum length out of the possible
-    // string lengths for varying.
-    if (SizeCI->isAllOnesValue() ||
-        SizeCI->getZExtValue() >= GetStringLength(CI->getOperand(2))) {
-      Value *Ret = EmitStpCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
+                              Name.substr(2, 6));
       return ReplaceInstUsesWith(*CI, Ret);
     }
     return 0;

Modified: llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp?rev=98284&r1=98283&r2=98284&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp Thu Mar 11 14:45:13 2010
@@ -620,25 +620,13 @@
     return 0;
   }
 
-  if (Name == "__strcpy_chk") {
+  if (Name == "__strcpy_chk" || Name == "__stpcpy_chk") {
     ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
     if (!SizeCI)
       return 0;
     if (SizeCI->isAllOnesValue()) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
-      CI->replaceAllUsesWith(Ret);
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0;
-  }
-
-  if (Name == "__stpcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      Value *Ret = EmitStpCpy(CI->getOperand(1), CI->getOperand(2), B, TD);
+      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
+                              Name.substr(2, 6));
       CI->replaceAllUsesWith(Ret);
       CI->eraseFromParent();
       return true;

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98284&r1=98283&r2=98284&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Thu Mar 11 14:45:13 2010
@@ -72,39 +72,21 @@
 /// EmitStrCpy - Emit a call to the strcpy function to the builder, for the
 /// specified pointer arguments.
 Value *llvm::EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
-                        const TargetData *TD) {
+                        const TargetData *TD, StringRef Name) {
   Module *M = B.GetInsertBlock()->getParent()->getParent();
   AttributeWithIndex AWI[2];
   AWI[0] = AttributeWithIndex::get(2, Attribute::NoCapture);
   AWI[1] = AttributeWithIndex::get(~0u, Attribute::NoUnwind);
   const Type *I8Ptr = B.getInt8PtrTy();
-  Value *StrCpy = M->getOrInsertFunction("strcpy", AttrListPtr::get(AWI, 2),
+  Value *StrCpy = M->getOrInsertFunction(Name, AttrListPtr::get(AWI, 2),
                                          I8Ptr, I8Ptr, I8Ptr, NULL);
   CallInst *CI = B.CreateCall2(StrCpy, CastToCStr(Dst, B), CastToCStr(Src, B),
-                               "strcpy");
+                               Name);
   if (const Function *F = dyn_cast<Function>(StrCpy->stripPointerCasts()))
     CI->setCallingConv(F->getCallingConv());
   return CI;
 }
 
-/// EmitStpCpy - Emit a call to the stpcpy function to the builder, for the
-/// specified pointer arguments.
-Value *llvm::EmitStpCpy(Value *Dst, Value *Src, IRBuilder<> &B,
-                        const TargetData *TD) {
-  Module *M = B.GetInsertBlock()->getParent()->getParent();
-  AttributeWithIndex AWI[2];
-  AWI[0] = AttributeWithIndex::get(2, Attribute::NoCapture);
-  AWI[1] = AttributeWithIndex::get(~0u, Attribute::NoUnwind);
-  const Type *I8Ptr = B.getInt8PtrTy();
-  Value *StpCpy = M->getOrInsertFunction("stpcpy", AttrListPtr::get(AWI, 2),
-                                         I8Ptr, I8Ptr, I8Ptr, NULL);
-  CallInst *CI = B.CreateCall2(StpCpy, CastToCStr(Dst, B), CastToCStr(Src, B),
-                               "stpcpy");
-  if (const Function *F = dyn_cast<Function>(StpCpy->stripPointerCasts()))
-    CI->setCallingConv(F->getCallingConv());
-  return CI;
-}
-
 /// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
 /// specified pointer arguments.
 Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,




From johnny.chen at apple.com  Thu Mar 11 15:02:50 2010
From: johnny.chen at apple.com (Johnny Chen)
Date: Thu, 11 Mar 2010 21:02:50 -0000
Subject: [llvm-commits] [llvm] r98285 -
	/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
Message-ID: <20100311210250.6B3732A6C12C@llvm.org>

Author: johnny
Date: Thu Mar 11 15:02:50 2010
New Revision: 98285

URL: http://llvm.org/viewvc/llvm-project?rev=98285&view=rev
Log:
Set the (Format)F filed of t2Int_MemBarrierV7 & t2Int_SyncBarrierV7 to ThumbFrm,
instead of Pseudo, which helps Thumb decoder to recognize them as Thumb instr.

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=98285&r1=98284&r2=98285&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Thu Mar 11 15:02:50 2010
@@ -2159,7 +2159,7 @@
 // memory barriers protect the atomic sequences
 let hasSideEffects = 1 in {
 def t2Int_MemBarrierV7 : AInoP<(outs), (ins),
-                        Pseudo, NoItinerary,
+                        ThumbFrm, NoItinerary,
                         "dmb", "",
                         [(ARMMemBarrierV7)]>,
                         Requires<[IsThumb2]> {
@@ -2169,7 +2169,7 @@
 }
 
 def t2Int_SyncBarrierV7 : AInoP<(outs), (ins),
-                        Pseudo, NoItinerary,
+                        ThumbFrm, NoItinerary,
                         "dsb", "",
                         [(ARMSyncBarrierV7)]>,
                         Requires<[IsThumb2]> {




From echristo at apple.com  Thu Mar 11 15:19:58 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 11 Mar 2010 13:19:58 -0800
Subject: [llvm-commits] [llvm] r98284 - in /llvm/trunk:
	include/llvm/Transforms/Utils/BuildLibCalls.h
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	lib/Transforms/Scalar/CodeGenPrepare.cpp
	lib/Transforms/Utils/BuildLibCalls.cpp
In-Reply-To: <20100311204513.D98322A6C12C@llvm.org>
References: <20100311204513.D98322A6C12C@llvm.org>
Message-ID: <138E917D-27CB-40EC-9AA5-D991FCADC43A@apple.com>


On Mar 11, 2010, at 12:45 PM, Benjamin Kramer wrote:

> URL: http://llvm.org/viewvc/llvm-project?rev=98284&view=rev
> Log:
> stpcpy is so similar to strcpy, it doesn't deserve a complete copy of the __strcpy_chk -> strcpy code.

Thanks, I'd been meaning to do that.

-eric


From sabre at nondot.org  Thu Mar 11 15:38:58 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 21:38:58 -0000
Subject: [llvm-commits] [llvm] r98288 - /llvm/trunk/lib/MC/MCContext.cpp
Message-ID: <20100311213858.31FE32A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 15:38:58 2010
New Revision: 98288

URL: http://llvm.org/viewvc/llvm-project?rev=98288&view=rev
Log:
empty symbols aren't possible, the mcsymbol ctor aborts on them.

Modified:
    llvm/trunk/lib/MC/MCContext.cpp

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=98288&r1=98287&r2=98288&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Thu Mar 11 15:38:58 2010
@@ -38,10 +38,6 @@
 
 
 MCSymbol *MCContext::GetOrCreateTemporarySymbol(StringRef Name) {
-  // If unnamed, just create a symbol.
-  if (Name.empty())
-    new (*this) MCSymbol("", true);
-    
   // Otherwise create as usual.
   MCSymbol *&Entry = Symbols[Name];
   if (Entry) return Entry;




From gohman at apple.com  Thu Mar 11 15:39:57 2010
From: gohman at apple.com (Dan Gohman)
Date: Thu, 11 Mar 2010 21:39:57 -0000
Subject: [llvm-commits] [llvm] r98289 - in /llvm/trunk:
 include/llvm/Target/TargetLowering.h
 lib/CodeGen/SelectionDAG/TargetLowering.cpp
 lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86ISelLowering.h
Message-ID: <20100311213957.E78CC2A6C12C@llvm.org>

Author: djg
Date: Thu Mar 11 15:39:57 2010
New Revision: 98289

URL: http://llvm.org/viewvc/llvm-project?rev=98289&view=rev
Log:
Remove getWidenVectorType, which is no longer used.

Modified:
    llvm/trunk/include/llvm/Target/TargetLowering.h
    llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.h

Modified: llvm/trunk/include/llvm/Target/TargetLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLowering.h?rev=98289&r1=98288&r2=98289&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetLowering.h (original)
+++ llvm/trunk/include/llvm/Target/TargetLowering.h Thu Mar 11 15:39:57 2010
@@ -323,13 +323,6 @@
     return false;
   }
 
-  /// getWidenVectorType: given a vector type, returns the type to widen to
-  /// (e.g., v7i8 to v8i8). If the vector type is legal, it returns itself.
-  /// If there is no vector type that we want to widen to, returns MVT::Other
-  /// When and were to widen is target dependent based on the cost of
-  /// scalarizing vs using the wider vector type.
-  virtual EVT getWidenVectorType(EVT VT) const;
-
   /// isFPImmLegal - Returns true if the target can instruction select the
   /// specified FP immediate natively. If false, the legalizer will materialize
   /// the FP immediate as a load from a constant pool.

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=98289&r1=98288&r2=98289&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Thu Mar 11 15:39:57 2010
@@ -792,20 +792,6 @@
   return 1;
 }
 
-/// getWidenVectorType: given a vector type, returns the type to widen to
-/// (e.g., v7i8 to v8i8). If the vector type is legal, it returns itself.
-/// If there is no vector type that we want to widen to, returns MVT::Other
-/// When and where to widen is target dependent based on the cost of
-/// scalarizing vs using the wider vector type.
-EVT TargetLowering::getWidenVectorType(EVT VT) const {
-  assert(VT.isVector());
-  if (isTypeLegal(VT))
-    return VT;
- 
-  // Default is not to widen until moved to LegalizeTypes
-  return MVT::Other;
-}
-
 /// getByValTypeAlignment - Return the desired alignment for ByVal aggregate
 /// function arguments in the caller parameter area.  This is the actual
 /// alignment, not its logarithm.

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98289&r1=98288&r2=98289&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Mar 11 15:39:57 2010
@@ -10292,41 +10292,3 @@
 
   return Res;
 }
-
-//===----------------------------------------------------------------------===//
-//                           X86 Widen vector type
-//===----------------------------------------------------------------------===//
-
-/// getWidenVectorType: given a vector type, returns the type to widen
-/// to (e.g., v7i8 to v8i8). If the vector type is legal, it returns itself.
-/// If there is no vector type that we want to widen to, returns MVT::Other
-/// When and where to widen is target dependent based on the cost of
-/// scalarizing vs using the wider vector type.
-
-EVT X86TargetLowering::getWidenVectorType(EVT VT) const {
-  assert(VT.isVector());
-  if (isTypeLegal(VT))
-    return VT;
-
-  // TODO: In computeRegisterProperty, we can compute the list of legal vector
-  //       type based on element type.  This would speed up our search (though
-  //       it may not be worth it since the size of the list is relatively
-  //       small).
-  EVT EltVT = VT.getVectorElementType();
-  unsigned NElts = VT.getVectorNumElements();
-
-  // On X86, it make sense to widen any vector wider than 1
-  if (NElts <= 1)
-    return MVT::Other;
-
-  for (unsigned nVT = MVT::FIRST_VECTOR_VALUETYPE;
-       nVT <= MVT::LAST_VECTOR_VALUETYPE; ++nVT) {
-    EVT SVT = (MVT::SimpleValueType)nVT;
-
-    if (isTypeLegal(SVT) &&
-        SVT.getVectorElementType() == EltVT &&
-        SVT.getVectorNumElements() > NElts)
-      return SVT;
-  }
-  return MVT::Other;
-}

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.h?rev=98289&r1=98288&r2=98289&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.h (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.h Thu Mar 11 15:39:57 2010
@@ -564,13 +564,6 @@
       (VT == MVT::f32 && X86ScalarSSEf32);   // f32 is when SSE1
     }
 
-    /// getWidenVectorType: given a vector type, returns the type to widen
-    /// to (e.g., v7i8 to v8i8). If the vector type is legal, it returns itself.
-    /// If there is no vector type that we want to widen to, returns EVT::Other
-    /// When and were to widen is target dependent based on the cost of
-    /// scalarizing vs using the wider vector type.
-    virtual EVT getWidenVectorType(EVT VT) const;
-
     /// createFastISel - This method returns a target specific FastISel object,
     /// or null if the target does not support "fast" ISel.
     virtual FastISel *




From sabre at nondot.org  Thu Mar 11 15:55:20 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 21:55:20 -0000
Subject: [llvm-commits] [llvm] r98292 - in /llvm/trunk:
 include/llvm/Target/Mangler.h
 include/llvm/Target/TargetLoweringObjectFile.h
 lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
 lib/CodeGen/TargetLoweringObjectFileImpl.cpp
 lib/Target/TargetLoweringObjectFile.cpp
Message-ID: <20100311215520.7CEA92A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 15:55:20 2010
New Revision: 98292

URL: http://llvm.org/viewvc/llvm-project?rev=98292&view=rev
Log:
fix a fixme in TargetLoweringObjectFile::getExprForDwarfReference
where we used ot create an MCSymbol for ".".  Now emit an assembler
temporary label and reference it instead of "." textually.

rdar://7739457

Modified:
    llvm/trunk/include/llvm/Target/Mangler.h
    llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
    llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp

Modified: llvm/trunk/include/llvm/Target/Mangler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Mangler.h?rev=98292&r1=98291&r2=98292&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Mangler.h (original)
+++ llvm/trunk/include/llvm/Target/Mangler.h Thu Mar 11 15:55:20 2010
@@ -51,6 +51,12 @@
   // symbols.
   Mangler(const MCAsmInfo &mai) : MAI(mai), NextAnonGlobalID(1) {}
 
+  /// getUniqueID() - Allocate and return a unique ID.
+  /// FIXME: Remove this.
+  unsigned getUniqueID() {
+    return NextAnonGlobalID++;
+  }
+  
   /// getNameWithPrefix - Fill OutName with the name of the appropriate prefix
   /// and the specified global variable's name.  If the global variable doesn't
   /// have a name, this fills in a unique name for the global.

Modified: llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h?rev=98292&r1=98291&r2=98292&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h (original)
+++ llvm/trunk/include/llvm/Target/TargetLoweringObjectFile.h Thu Mar 11 15:55:20 2010
@@ -208,8 +208,8 @@
 
   /// 
   const MCExpr *
-  getExprForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
-                           unsigned Encoding,
+  getExprForDwarfReference(const MCSymbol *Sym, Mangler *Mang,
+                           MachineModuleInfo *MMI, unsigned Encoding,
                            MCStreamer &Streamer) const;
   
   virtual unsigned getPersonalityEncoding() const;

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98292&r1=98291&r2=98292&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Thu Mar 11 15:55:20 2010
@@ -189,7 +189,8 @@
 void DwarfPrinter::EmitReference(const MCSymbol *Sym, unsigned Encoding) const {
   const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
 
-  const MCExpr *Exp = TLOF.getExprForDwarfReference(Sym, Asm->MMI, Encoding,
+  const MCExpr *Exp = TLOF.getExprForDwarfReference(Sym, Asm->Mang,
+                                                    Asm->MMI, Encoding,
                                                     Asm->OutStreamer);
   Asm->OutStreamer.EmitValue(Exp, SizeOfEncodedValue(Encoding), /*addrspace*/0);
 }

Modified: llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp?rev=98292&r1=98291&r2=98292&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp (original)
+++ llvm/trunk/lib/CodeGen/TargetLoweringObjectFileImpl.cpp Thu Mar 11 15:55:20 2010
@@ -421,7 +421,7 @@
     }
 
     return TargetLoweringObjectFile::
-      getExprForDwarfReference(Sym, MMI,
+      getExprForDwarfReference(Sym, Mang, MMI,
                                Encoding & ~dwarf::DW_EH_PE_indirect, Streamer);
   }
 
@@ -784,8 +784,8 @@
     }
 
     return TargetLoweringObjectFile::
-      getExprForDwarfReference(Sym, MMI, Encoding & ~dwarf::DW_EH_PE_indirect,
-                               Streamer);
+      getExprForDwarfReference(Sym, Mang, MMI,
+                               Encoding & ~dwarf::DW_EH_PE_indirect, Streamer);
   }
 
   return TargetLoweringObjectFile::

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98292&r1=98291&r2=98292&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Thu Mar 11 15:55:20 2010
@@ -19,6 +19,7 @@
 #include "llvm/GlobalVariable.h"
 #include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/Target/Mangler.h"
 #include "llvm/Target/TargetData.h"
@@ -307,29 +308,34 @@
   else
     Sym = getContext().GetOrCreateSymbol(Name.str());
 
-  return getExprForDwarfReference(Sym, MMI, Encoding, Streamer);
+  return getExprForDwarfReference(Sym, Mang, MMI, Encoding, Streamer);
 }
 
 const MCExpr *TargetLoweringObjectFile::
-getExprForDwarfReference(const MCSymbol *Sym, MachineModuleInfo *MMI,
-                         unsigned Encoding, MCStreamer &Streamer) const {
+getExprForDwarfReference(const MCSymbol *Sym, Mangler *Mang,
+                         MachineModuleInfo *MMI, unsigned Encoding,
+                         MCStreamer &Streamer) const {
   const MCExpr *Res = MCSymbolRefExpr::Create(Sym, getContext());
 
   switch (Encoding & 0xF0) {
   default:
     llvm_report_error("We do not support this DWARF encoding yet!");
-    break;
   case dwarf::DW_EH_PE_absptr:
     // Do nothing special
-    break;
-  case dwarf::DW_EH_PE_pcrel:
-    // FIXME: PCSymbol
-    const MCExpr *PC = MCSymbolRefExpr::Create(".", getContext());
-    Res = MCBinaryExpr::CreateSub(Res, PC, getContext());
-    break;
+    return Res;
+  case dwarf::DW_EH_PE_pcrel: {
+    // Emit a label to the streamer for the current position.  This gives us
+    // .-foo addressing.
+    SmallString<128> Name;
+    Mang->getNameWithPrefix(Name, Twine("PCtemp") + Twine(Mang->getUniqueID()),
+                            Mangler::Private);
+
+    MCSymbol *PCSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+    Streamer.EmitLabel(PCSym);
+    const MCExpr *PC = MCSymbolRefExpr::Create(PCSym, getContext());
+    return MCBinaryExpr::CreateSub(Res, PC, getContext());
+  }
   }
-
-  return Res;
 }
 
 unsigned TargetLoweringObjectFile::getPersonalityEncoding() const {




From sabre at nondot.org  Thu Mar 11 16:53:35 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 22:53:35 -0000
Subject: [llvm-commits] [llvm] r98293 - in /llvm/trunk:
 include/llvm/MC/MCContext.h lib/CodeGen/ELFWriter.cpp
 lib/CodeGen/LLVMTargetMachine.cpp lib/MC/MCContext.cpp
 tools/edis/EDDisassembler.cpp tools/llvm-mc/llvm-mc.cpp
Message-ID: <20100311225335.5F9D42A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 16:53:35 2010
New Revision: 98293

URL: http://llvm.org/viewvc/llvm-project?rev=98293&view=rev
Log:
change MCContext to always have an MCAsmInfo.

Modified:
    llvm/trunk/include/llvm/MC/MCContext.h
    llvm/trunk/lib/CodeGen/ELFWriter.cpp
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/MC/MCContext.cpp
    llvm/trunk/tools/edis/EDDisassembler.cpp
    llvm/trunk/tools/llvm-mc/llvm-mc.cpp

Modified: llvm/trunk/include/llvm/MC/MCContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCContext.h?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCContext.h (original)
+++ llvm/trunk/include/llvm/MC/MCContext.h Thu Mar 11 16:53:35 2010
@@ -15,6 +15,7 @@
 #include "llvm/Support/Allocator.h"
 
 namespace llvm {
+  class MCAsmInfo;
   class MCExpr;
   class MCSection;
   class MCSymbol;
@@ -28,20 +29,29 @@
     MCContext(const MCContext&); // DO NOT IMPLEMENT
     MCContext &operator=(const MCContext&); // DO NOT IMPLEMENT
 
+    /// The MCAsmInfo for this target.
+    const MCAsmInfo &MAI;
+    
     /// Sections - Bindings of names to allocated sections.
     StringMap<MCSection*> Sections;
 
     /// Symbols - Bindings of names to symbols.
     StringMap<MCSymbol*> Symbols;
 
+    /// NextUniqueID - The next ID to dole out to an unnamed assembler temporary
+    /// symbol.
+    unsigned NextUniqueID;
+    
     /// Allocator - Allocator object used for creating machine code objects.
     ///
     /// We use a bump pointer allocator to avoid the need to track all allocated
     /// objects.
     BumpPtrAllocator Allocator;
   public:
-    MCContext();
+    MCContext(const MCAsmInfo &MAI);
     ~MCContext();
+    
+    const MCAsmInfo &getAsmInfo() const { return MAI; }
 
     /// @name Symbol Managment
     /// @{

Modified: llvm/trunk/lib/CodeGen/ELFWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ELFWriter.cpp?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/ELFWriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/ELFWriter.cpp Thu Mar 11 16:53:35 2010
@@ -64,7 +64,7 @@
 
 ELFWriter::ELFWriter(raw_ostream &o, TargetMachine &tm)
   : MachineFunctionPass(&ID), O(o), TM(tm),
-    OutContext(*new MCContext()),
+    OutContext(*new MCContext(*TM.getMCAsmInfo())),
     TLOF(TM.getTargetLowering()->getObjFileLowering()),
     is64Bit(TM.getTargetData()->getPointerSizeInBits() == 64),
     isLittleEndian(TM.getTargetData()->isLittleEndian()),

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Thu Mar 11 16:53:35 2010
@@ -121,14 +121,14 @@
   if (addCommonCodeGenPasses(PM, OptLevel, DisableVerify))
     return true;
 
-  OwningPtr<MCContext> Context(new MCContext());
+  const MCAsmInfo &MAI = *getMCAsmInfo();
+  OwningPtr<MCContext> Context(new MCContext(MAI));
   OwningPtr<MCStreamer> AsmStreamer;
 
   formatted_raw_ostream *LegacyOutput;
   switch (FileType) {
   default: return true;
   case CGFT_AssemblyFile: {
-    const MCAsmInfo &MAI = *getMCAsmInfo();
     MCInstPrinter *InstPrinter =
       getTarget().createMCInstPrinter(MAI.getAssemblerDialect(), MAI, Out);
     AsmStreamer.reset(createAsmStreamer(*Context, Out, MAI,

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Thu Mar 11 16:53:35 2010
@@ -14,7 +14,7 @@
 #include "llvm/ADT/Twine.h"
 using namespace llvm;
 
-MCContext::MCContext() {
+MCContext::MCContext(const MCAsmInfo &mai) : MAI(mai), NextUniqueID(0) {
 }
 
 MCContext::~MCContext() {

Modified: llvm/trunk/tools/edis/EDDisassembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/edis/EDDisassembler.cpp?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/tools/edis/EDDisassembler.cpp (original)
+++ llvm/trunk/tools/edis/EDDisassembler.cpp Thu Mar 11 16:53:35 2010
@@ -341,19 +341,17 @@
   
   SourceMgr sourceMgr;
   sourceMgr.AddNewSourceBuffer(buf, SMLoc()); // ownership of buf handed over
-  MCContext context;
-  OwningPtr<MCStreamer> streamer
-    (createNullStreamer(context));
+  MCContext context(*AsmInfo);
+  OwningPtr<MCStreamer> streamer(createNullStreamer(context));
   AsmParser genericParser(sourceMgr, context, *streamer, *AsmInfo);
-  OwningPtr<TargetAsmParser> specificParser
-    (Tgt->createAsmParser(genericParser));
+  OwningPtr<TargetAsmParser> TargetParser(Tgt->createAsmParser(genericParser));
   
   AsmToken OpcodeToken = genericParser.Lex();
   
   if(OpcodeToken.is(AsmToken::Identifier)) {
     instName = OpcodeToken.getString();
     instLoc = OpcodeToken.getLoc();
-    if (specificParser->ParseInstruction(instName, instLoc, operands))
+    if (TargetParser->ParseInstruction(instName, instLoc, operands))
       ret = -1;
   }
   else {

Modified: llvm/trunk/tools/llvm-mc/llvm-mc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-mc/llvm-mc.cpp?rev=98293&r1=98292&r2=98293&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-mc/llvm-mc.cpp (original)
+++ llvm/trunk/tools/llvm-mc/llvm-mc.cpp Thu Mar 11 16:53:35 2010
@@ -242,7 +242,11 @@
   // it later.
   SrcMgr.setIncludeDirs(IncludeDirs);
   
-  MCContext Ctx;
+  
+  const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName);
+  assert(MAI && "Unable to create target asm info!");
+  
+  MCContext Ctx(*MAI);
   formatted_raw_ostream *Out = GetOutputStream();
   if (!Out)
     return 1;
@@ -262,9 +266,6 @@
   OwningPtr<MCStreamer> Str;
   OwningPtr<TargetAsmBackend> TAB;
 
-  const MCAsmInfo *MAI = TheTarget->createAsmInfo(TripleName);
-  assert(MAI && "Unable to create target asm info!");
-
   if (FileType == OFT_AssemblyFile) {
     IP.reset(TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *Out));
     if (ShowEncoding)




From sabre at nondot.org  Thu Mar 11 16:56:10 2010
From: sabre at nondot.org (Chris Lattner)
Date: Thu, 11 Mar 2010 22:56:10 -0000
Subject: [llvm-commits] [llvm] r98294 - /llvm/trunk/lib/MC/MCContext.cpp
Message-ID: <20100311225610.B653A2A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 16:56:10 2010
New Revision: 98294

URL: http://llvm.org/viewvc/llvm-project?rev=98294&view=rev
Log:
enhance MCContext::GetOrCreateTemporarySymbol() to create a new symbol
with an arbitrary unique name.

Modified:
    llvm/trunk/lib/MC/MCContext.cpp

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=98294&r1=98293&r2=98294&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Thu Mar 11 16:56:10 2010
@@ -8,6 +8,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCSection.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/ADT/SmallString.h"
@@ -38,6 +39,11 @@
 
 
 MCSymbol *MCContext::GetOrCreateTemporarySymbol(StringRef Name) {
+  // If there is no name, create a new anonymous symbol.
+  if (Name.empty())
+    return GetOrCreateTemporarySymbol(Twine(MAI.getPrivateGlobalPrefix()) +
+                                      "tmp" + Twine(NextUniqueID++));
+  
   // Otherwise create as usual.
   MCSymbol *&Entry = Symbols[Name];
   if (Entry) return Entry;




From stoklund at 2pi.dk  Thu Mar 11 17:04:34 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 11 Mar 2010 23:04:34 -0000
Subject: [llvm-commits] [llvm] r98295 -
	/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
Message-ID: <20100311230434.B7C9E2A6C12C@llvm.org>

Author: stoklund
Date: Thu Mar 11 17:04:34 2010
New Revision: 98295

URL: http://llvm.org/viewvc/llvm-project?rev=98295&view=rev
Log:
Extract methods from LocalRewriter::RewriteMBB bringing it down to 666 lines.

Modified:
    llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp

Modified: llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp?rev=98295&r1=98294&r2=98295&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp Thu Mar 11 17:04:34 2010
@@ -1095,10 +1095,18 @@
                            BitVector &RegKills,
                            std::vector<MachineOperand*> &KillOps);
 
-  void TransferDeadness(unsigned CurDist,
-                        unsigned Reg, BitVector &RegKills,
+  void TransferDeadness(unsigned Reg, BitVector &RegKills,
                         std::vector<MachineOperand*> &KillOps);
 
+  bool InsertEmergencySpills(MachineInstr *MI);
+
+  bool InsertRestores(MachineInstr *MI,
+                      AvailableSpills &Spills,
+                      BitVector &RegKills,
+                      std::vector<MachineOperand*> &KillOps);
+
+  bool InsertSpills(MachineInstr *MI);
+
   void RewriteMBB(LiveIntervals *LIs,
                   AvailableSpills &Spills, BitVector &RegKills,
                   std::vector<MachineOperand*> &KillOps);
@@ -1604,8 +1612,7 @@
 /// TransferDeadness - A identity copy definition is dead and it's being
 /// removed. Find the last def or use and mark it as dead / kill.
 void LocalRewriter::
-TransferDeadness(unsigned CurDist,
-                 unsigned Reg, BitVector &RegKills,
+TransferDeadness(unsigned Reg, BitVector &RegKills,
                  std::vector<MachineOperand*> &KillOps) {
   SmallPtrSet<MachineInstr*, 4> Seens;
   SmallVector<std::pair<MachineInstr*, int>,8> Refs;
@@ -1615,7 +1622,7 @@
     if (UDMI->getParent() != MBB)
       continue;
     DenseMap<MachineInstr*, unsigned>::iterator DI = DistanceMap.find(UDMI);
-    if (DI == DistanceMap.end() || DI->second > CurDist)
+    if (DI == DistanceMap.end())
       continue;
     if (Seens.insert(UDMI))
       Refs.push_back(std::make_pair(UDMI, DI->second));
@@ -1657,6 +1664,180 @@
   }
 }
 
+/// InsertEmergencySpills - Insert emergency spills before MI if requested by
+/// VRM. Return true if spills were inserted.
+bool LocalRewriter::InsertEmergencySpills(MachineInstr *MI) {
+  if (!VRM->hasEmergencySpills(MI))
+    return false;
+  MachineBasicBlock::iterator MII = MI;
+  SmallSet<int, 4> UsedSS;
+  std::vector<unsigned> &EmSpills = VRM->getEmergencySpills(MI);
+  for (unsigned i = 0, e = EmSpills.size(); i != e; ++i) {
+    unsigned PhysReg = EmSpills[i];
+    const TargetRegisterClass *RC = TRI->getPhysicalRegisterRegClass(PhysReg);
+    assert(RC && "Unable to determine register class!");
+    int SS = VRM->getEmergencySpillSlot(RC);
+    if (UsedSS.count(SS))
+      llvm_unreachable("Need to spill more than one physical registers!");
+    UsedSS.insert(SS);
+    TII->storeRegToStackSlot(*MBB, MII, PhysReg, true, SS, RC);
+    MachineInstr *StoreMI = prior(MII);
+    VRM->addSpillSlotUse(SS, StoreMI);
+
+    // Back-schedule reloads and remats.
+    MachineBasicBlock::iterator InsertLoc =
+      ComputeReloadLoc(llvm::next(MII), MBB->begin(), PhysReg, TRI, false, SS,
+                       TII, *MBB->getParent());
+
+    TII->loadRegFromStackSlot(*MBB, InsertLoc, PhysReg, SS, RC);
+
+    MachineInstr *LoadMI = prior(InsertLoc);
+    VRM->addSpillSlotUse(SS, LoadMI);
+    ++NumPSpills;
+    DistanceMap.insert(std::make_pair(LoadMI, DistanceMap.size()));
+  }
+  return true;
+}
+
+/// InsertRestores - Restore registers before MI is requested by VRM. Return
+/// true is any instructions were inserted.
+bool LocalRewriter::InsertRestores(MachineInstr *MI,
+                                   AvailableSpills &Spills,
+                                   BitVector &RegKills,
+                                   std::vector<MachineOperand*> &KillOps) {
+  if (!VRM->isRestorePt(MI))
+    return false;
+  MachineBasicBlock::iterator MII = MI;
+  std::vector<unsigned> &RestoreRegs = VRM->getRestorePtRestores(MI);
+  for (unsigned i = 0, e = RestoreRegs.size(); i != e; ++i) {
+    unsigned VirtReg = RestoreRegs[e-i-1];  // Reverse order.
+    if (!VRM->getPreSplitReg(VirtReg))
+      continue; // Split interval spilled again.
+    unsigned Phys = VRM->getPhys(VirtReg);
+    MRI->setPhysRegUsed(Phys);
+
+    // Check if the value being restored if available. If so, it must be
+    // from a predecessor BB that fallthrough into this BB. We do not
+    // expect:
+    // BB1:
+    // r1 = load fi#1
+    // ...
+    //    = r1<kill>
+    // ... # r1 not clobbered
+    // ...
+    //    = load fi#1
+    bool DoReMat = VRM->isReMaterialized(VirtReg);
+    int SSorRMId = DoReMat
+      ? VRM->getReMatId(VirtReg) : VRM->getStackSlot(VirtReg);
+    const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+    unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
+    if (InReg == Phys) {
+      // If the value is already available in the expected register, save
+      // a reload / remat.
+      if (SSorRMId)
+        DEBUG(dbgs() << "Reusing RM#"
+                     << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
+      else
+        DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
+      DEBUG(dbgs() << " from physreg "
+                   << TRI->getName(InReg) << " for vreg"
+                   << VirtReg <<" instead of reloading into physreg "
+                   << TRI->getName(Phys) << '\n');
+      ++NumOmitted;
+      continue;
+    } else if (InReg && InReg != Phys) {
+      if (SSorRMId)
+        DEBUG(dbgs() << "Reusing RM#"
+                     << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
+      else
+        DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
+      DEBUG(dbgs() << " from physreg "
+                   << TRI->getName(InReg) << " for vreg"
+                   << VirtReg <<" by copying it into physreg "
+                   << TRI->getName(Phys) << '\n');
+
+      // If the reloaded / remat value is available in another register,
+      // copy it to the desired register.
+
+      // Back-schedule reloads and remats.
+      MachineBasicBlock::iterator InsertLoc =
+        ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat, SSorRMId, TII,
+                         *MBB->getParent());
+
+      TII->copyRegToReg(*MBB, InsertLoc, Phys, InReg, RC, RC);
+
+      // This invalidates Phys.
+      Spills.ClobberPhysReg(Phys);
+      // Remember it's available.
+      Spills.addAvailable(SSorRMId, Phys);
+
+      // Mark is killed.
+      MachineInstr *CopyMI = prior(InsertLoc);
+      CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
+      MachineOperand *KillOpnd = CopyMI->findRegisterUseOperand(InReg);
+      KillOpnd->setIsKill();
+      UpdateKills(*CopyMI, TRI, RegKills, KillOps);
+
+      DEBUG(dbgs() << '\t' << *CopyMI);
+      ++NumCopified;
+      continue;
+    }
+
+    // Back-schedule reloads and remats.
+    MachineBasicBlock::iterator InsertLoc =
+      ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat, SSorRMId, TII,
+                       *MBB->getParent());
+
+    if (VRM->isReMaterialized(VirtReg)) {
+      ReMaterialize(*MBB, InsertLoc, Phys, VirtReg, TII, TRI, *VRM);
+    } else {
+      const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
+      TII->loadRegFromStackSlot(*MBB, InsertLoc, Phys, SSorRMId, RC);
+      MachineInstr *LoadMI = prior(InsertLoc);
+      VRM->addSpillSlotUse(SSorRMId, LoadMI);
+      ++NumLoads;
+      DistanceMap.insert(std::make_pair(LoadMI, DistanceMap.size()));
+    }
+
+    // This invalidates Phys.
+    Spills.ClobberPhysReg(Phys);
+    // Remember it's available.
+    Spills.addAvailable(SSorRMId, Phys);
+
+    UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
+    DEBUG(dbgs() << '\t' << *prior(MII));
+  }
+  return true;
+}
+
+/// InsertEmergencySpills - Insert spills after MI if requested by VRM. Return
+/// true if spills were inserted.
+bool LocalRewriter::InsertSpills(MachineInstr *MI) {
+  if (!VRM->isSpillPt(MI))
+    return false;
+  MachineBasicBlock::iterator MII = MI;
+  std::vector<std::pair<unsigned,bool> > &SpillRegs =
+    VRM->getSpillPtSpills(MI);
+  for (unsigned i = 0, e = SpillRegs.size(); i != e; ++i) {
+    unsigned VirtReg = SpillRegs[i].first;
+    bool isKill = SpillRegs[i].second;
+    if (!VRM->getPreSplitReg(VirtReg))
+      continue; // Split interval spilled again.
+    const TargetRegisterClass *RC = MRI->getRegClass(VirtReg);
+    unsigned Phys = VRM->getPhys(VirtReg);
+    int StackSlot = VRM->getStackSlot(VirtReg);
+    MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
+    TII->storeRegToStackSlot(*MBB, llvm::next(MII), Phys, isKill, StackSlot,
+                             RC);
+    MachineInstr *StoreMI = prior(oldNextMII);
+    VRM->addSpillSlotUse(StackSlot, StoreMI);
+    DEBUG(dbgs() << "Store:\t" << *StoreMI);
+    VRM->virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
+  }
+  return true;
+}
+
+
 /// rewriteMBB - Keep track of which spills are available even after the
 /// register allocator is done with them.  If possible, avid reloading vregs.
 void
@@ -1687,178 +1868,26 @@
   KillOps.clear();
   KillOps.resize(TRI->getNumRegs(), NULL);
 
-  unsigned Dist = 0;
   DistanceMap.clear();
   for (MachineBasicBlock::iterator MII = MBB->begin(), E = MBB->end();
        MII != E; ) {
     MachineBasicBlock::iterator NextMII = llvm::next(MII);
 
-    VirtRegMap::MI2VirtMapTy::const_iterator I, End;
-    bool Erased = false;
-    bool BackTracked = false;
     if (OptimizeByUnfold(MII, MaybeDeadStores, Spills, RegKills, KillOps))
       NextMII = llvm::next(MII);
 
-    MachineInstr &MI = *MII;
-
-    if (VRM->hasEmergencySpills(&MI)) {
-      // Spill physical register(s) in the rare case the allocator has run out
-      // of registers to allocate.
-      SmallSet<int, 4> UsedSS;
-      std::vector<unsigned> &EmSpills = VRM->getEmergencySpills(&MI);
-      for (unsigned i = 0, e = EmSpills.size(); i != e; ++i) {
-        unsigned PhysReg = EmSpills[i];
-        const TargetRegisterClass *RC =
-          TRI->getPhysicalRegisterRegClass(PhysReg);
-        assert(RC && "Unable to determine register class!");
-        int SS = VRM->getEmergencySpillSlot(RC);
-        if (UsedSS.count(SS))
-          llvm_unreachable("Need to spill more than one physical registers!");
-        UsedSS.insert(SS);
-        TII->storeRegToStackSlot(*MBB, MII, PhysReg, true, SS, RC);
-        MachineInstr *StoreMI = prior(MII);
-        VRM->addSpillSlotUse(SS, StoreMI);
-
-        // Back-schedule reloads and remats.
-        MachineBasicBlock::iterator InsertLoc =
-          ComputeReloadLoc(llvm::next(MII), MBB->begin(), PhysReg, TRI, false,
-                           SS, TII, MF);
-
-        TII->loadRegFromStackSlot(*MBB, InsertLoc, PhysReg, SS, RC);
-
-        MachineInstr *LoadMI = prior(InsertLoc);
-        VRM->addSpillSlotUse(SS, LoadMI);
-        ++NumPSpills;
-        DistanceMap.insert(std::make_pair(LoadMI, Dist++));
-      }
+    if (InsertEmergencySpills(MII))
       NextMII = llvm::next(MII);
-    }
-
-    // Insert restores here if asked to.
-    if (VRM->isRestorePt(&MI)) {
-      std::vector<unsigned> &RestoreRegs = VRM->getRestorePtRestores(&MI);
-      for (unsigned i = 0, e = RestoreRegs.size(); i != e; ++i) {
-        unsigned VirtReg = RestoreRegs[e-i-1];  // Reverse order.
-        if (!VRM->getPreSplitReg(VirtReg))
-          continue; // Split interval spilled again.
-        unsigned Phys = VRM->getPhys(VirtReg);
-        MRI->setPhysRegUsed(Phys);
-
-        // Check if the value being restored if available. If so, it must be
-        // from a predecessor BB that fallthrough into this BB. We do not
-        // expect:
-        // BB1:
-        // r1 = load fi#1
-        // ...
-        //    = r1<kill>
-        // ... # r1 not clobbered
-        // ...
-        //    = load fi#1
-        bool DoReMat = VRM->isReMaterialized(VirtReg);
-        int SSorRMId = DoReMat
-          ? VRM->getReMatId(VirtReg) : VRM->getStackSlot(VirtReg);
-        const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
-        unsigned InReg = Spills.getSpillSlotOrReMatPhysReg(SSorRMId);
-        if (InReg == Phys) {
-          // If the value is already available in the expected register, save
-          // a reload / remat.
-          if (SSorRMId)
-            DEBUG(dbgs() << "Reusing RM#"
-                  << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
-          else
-            DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
-          DEBUG(dbgs() << " from physreg "
-                << TRI->getName(InReg) << " for vreg"
-                << VirtReg <<" instead of reloading into physreg "
-                << TRI->getName(Phys) << '\n');
-          ++NumOmitted;
-          continue;
-        } else if (InReg && InReg != Phys) {
-          if (SSorRMId)
-            DEBUG(dbgs() << "Reusing RM#"
-                  << SSorRMId-VirtRegMap::MAX_STACK_SLOT-1);
-          else
-            DEBUG(dbgs() << "Reusing SS#" << SSorRMId);
-          DEBUG(dbgs() << " from physreg "
-                << TRI->getName(InReg) << " for vreg"
-                << VirtReg <<" by copying it into physreg "
-                << TRI->getName(Phys) << '\n');
-
-          // If the reloaded / remat value is available in another register,
-          // copy it to the desired register.
-
-          // Back-schedule reloads and remats.
-          MachineBasicBlock::iterator InsertLoc =
-            ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat,
-                             SSorRMId, TII, MF);
-
-          TII->copyRegToReg(*MBB, InsertLoc, Phys, InReg, RC, RC);
-
-          // This invalidates Phys.
-          Spills.ClobberPhysReg(Phys);
-          // Remember it's available.
-          Spills.addAvailable(SSorRMId, Phys);
-
-          // Mark is killed.
-          MachineInstr *CopyMI = prior(InsertLoc);
-          CopyMI->setAsmPrinterFlag(MachineInstr::ReloadReuse);
-          MachineOperand *KillOpnd = CopyMI->findRegisterUseOperand(InReg);
-          KillOpnd->setIsKill();
-          UpdateKills(*CopyMI, TRI, RegKills, KillOps);
 
-          DEBUG(dbgs() << '\t' << *CopyMI);
-          ++NumCopified;
-          continue;
-        }
-
-        // Back-schedule reloads and remats.
-        MachineBasicBlock::iterator InsertLoc =
-          ComputeReloadLoc(MII, MBB->begin(), Phys, TRI, DoReMat,
-                           SSorRMId, TII, MF);
+    InsertRestores(MII, Spills, RegKills, KillOps);
 
-        if (VRM->isReMaterialized(VirtReg)) {
-          ReMaterialize(*MBB, InsertLoc, Phys, VirtReg, TII, TRI, *VRM);
-        } else {
-          const TargetRegisterClass* RC = MRI->getRegClass(VirtReg);
-          TII->loadRegFromStackSlot(*MBB, InsertLoc, Phys, SSorRMId, RC);
-          MachineInstr *LoadMI = prior(InsertLoc);
-          VRM->addSpillSlotUse(SSorRMId, LoadMI);
-          ++NumLoads;
-          DistanceMap.insert(std::make_pair(LoadMI, Dist++));
-        }
-
-        // This invalidates Phys.
-        Spills.ClobberPhysReg(Phys);
-        // Remember it's available.
-        Spills.addAvailable(SSorRMId, Phys);
-
-        UpdateKills(*prior(InsertLoc), TRI, RegKills, KillOps);
-        DEBUG(dbgs() << '\t' << *prior(MII));
-      }
-    }
-
-    // Insert spills here if asked to.
-    if (VRM->isSpillPt(&MI)) {
-      std::vector<std::pair<unsigned,bool> > &SpillRegs =
-        VRM->getSpillPtSpills(&MI);
-      for (unsigned i = 0, e = SpillRegs.size(); i != e; ++i) {
-        unsigned VirtReg = SpillRegs[i].first;
-        bool isKill = SpillRegs[i].second;
-        if (!VRM->getPreSplitReg(VirtReg))
-          continue; // Split interval spilled again.
-        const TargetRegisterClass *RC = MRI->getRegClass(VirtReg);
-        unsigned Phys = VRM->getPhys(VirtReg);
-        int StackSlot = VRM->getStackSlot(VirtReg);
-        MachineBasicBlock::iterator oldNextMII = llvm::next(MII);
-        TII->storeRegToStackSlot(*MBB, llvm::next(MII), Phys, isKill, StackSlot,
-                                 RC);
-        MachineInstr *StoreMI = prior(oldNextMII);
-        VRM->addSpillSlotUse(StackSlot, StoreMI);
-        DEBUG(dbgs() << "Store:\t" << *StoreMI);
-        VRM->virtFolded(VirtReg, StoreMI, VirtRegMap::isMod);
-      }
+    if (InsertSpills(MII))
       NextMII = llvm::next(MII);
-    }
+
+    VirtRegMap::MI2VirtMapTy::const_iterator I, End;
+    bool Erased = false;
+    bool BackTracked = false;
+    MachineInstr &MI = *MII;
 
     /// ReusedOperands - Keep track of operand reuse in case we need to undo
     /// reuse.
@@ -2125,7 +2154,7 @@
           MachineInstr *LoadMI = prior(InsertLoc);
           VRM->addSpillSlotUse(SSorRMId, LoadMI);
           ++NumLoads;
-          DistanceMap.insert(std::make_pair(LoadMI, Dist++));
+          DistanceMap.insert(std::make_pair(LoadMI, DistanceMap.size()));
         }
         // This invalidates PhysReg.
         Spills.ClobberPhysReg(PhysReg);
@@ -2371,7 +2400,7 @@
                    TRI->isSubRegister(KillRegs[0], Dst) ||
                    TRI->isSuperRegister(KillRegs[0], Dst));
             // Last def is now dead.
-            TransferDeadness(Dist, Src, RegKills, KillOps);
+            TransferDeadness(Src, RegKills, KillOps);
           }
           VRM->RemoveMachineInstrFromMaps(&MI);
           MBB->erase(&MI);
@@ -2470,7 +2499,7 @@
       Erased = true;
     }
     if (!Erased)
-      DistanceMap.insert(std::make_pair(&MI, Dist++));
+      DistanceMap.insert(std::make_pair(&MI, DistanceMap.size()));
     if (!Erased && !BackTracked) {
       for (MachineBasicBlock::iterator II = &MI; II != NextMII; ++II)
         UpdateKills(*II, TRI, RegKills, KillOps);




From natebegeman at mac.com  Thu Mar 11 17:06:07 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Thu, 11 Mar 2010 23:06:07 -0000
Subject: [llvm-commits] [llvm] r98296 - in /llvm/trunk:
 include/llvm-c/Transforms/Scalar.h lib/Transforms/Scalar/Scalar.cpp
Message-ID: <20100311230607.CF01A2A6C12C@llvm.org>

Author: sampo
Date: Thu Mar 11 17:06:07 2010
New Revision: 98296

URL: http://llvm.org/viewvc/llvm-project?rev=98296&view=rev
Log:
Add a handful of additional useful pass manager things to the C API

Modified:
    llvm/trunk/include/llvm-c/Transforms/Scalar.h
    llvm/trunk/lib/Transforms/Scalar/Scalar.cpp

Modified: llvm/trunk/include/llvm-c/Transforms/Scalar.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm-c/Transforms/Scalar.h?rev=98296&r1=98295&r2=98296&view=diff
==============================================================================
--- llvm/trunk/include/llvm-c/Transforms/Scalar.h (original)
+++ llvm/trunk/include/llvm-c/Transforms/Scalar.h Thu Mar 11 17:06:07 2010
@@ -79,6 +79,10 @@
 /** See llvm::createScalarReplAggregatesPass function. */
 void LLVMAddScalarReplAggregatesPass(LLVMPassManagerRef PM);
 
+/** See llvm::createScalarReplAggregatesPass function. */
+void LLVMAddScalarReplAggregatesPassWithThreshold(LLVMPassManagerRef PM,
+                                                  int Threshold);
+
 /** See llvm::createSimplifyLibCallsPass function. */
 void LLVMAddSimplifyLibCallsPass(LLVMPassManagerRef PM);
 
@@ -91,6 +95,12 @@
 /** See llvm::demotePromoteMemoryToRegisterPass function. */
 void LLVMAddDemoteMemoryToRegisterPass(LLVMPassManagerRef PM);
 
+/** See llvm::createVerifierPass function. */
+void LLVMAddVerifierPass(LLVMPassManagerRef PM);
+
+/** PM->add(new TargetData(M)) */
+void LLVMAddTargetData(LLVMPassManagerRef PM, LLVMModuleRef M);
+
 #ifdef __cplusplus
 }
 #endif /* defined(__cplusplus) */

Modified: llvm/trunk/lib/Transforms/Scalar/Scalar.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/Scalar.cpp?rev=98296&r1=98295&r2=98296&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/Scalar.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/Scalar.cpp Thu Mar 11 17:06:07 2010
@@ -14,6 +14,8 @@
 
 #include "llvm-c/Transforms/Scalar.h"
 #include "llvm/PassManager.h"
+#include "llvm/Analysis/Verifier.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/Transforms/Scalar.h"
 
 using namespace llvm;
@@ -90,6 +92,11 @@
   unwrap(PM)->add(createScalarReplAggregatesPass());
 }
 
+void LLVMAddScalarReplAggregatesPassWithThreshold(LLVMPassManagerRef PM,
+                                                  int Threshold) {
+  unwrap(PM)->add(createScalarReplAggregatesPass(Threshold));
+}
+
 void LLVMAddSimplifyLibCallsPass(LLVMPassManagerRef PM) {
   unwrap(PM)->add(createSimplifyLibCallsPass());
 }
@@ -105,3 +112,11 @@
 void LLVMAddDemoteMemoryToRegisterPass(LLVMPassManagerRef PM) {
   unwrap(PM)->add(createDemoteRegisterToMemoryPass());
 }
+
+void LLVMAddVerifierPass(LLVMPassManagerRef PM) {
+  unwrap(PM)->add(createVerifierPass());
+}
+
+void LLVMAddTargetData(LLVMPassManagerRef PM, LLVMModuleRef M) {
+  unwrap(PM)->add(new TargetData(unwrap(M)));
+}




From natebegeman at mac.com  Thu Mar 11 17:21:19 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Thu, 11 Mar 2010 23:21:19 -0000
Subject: [llvm-commits] [llvm] r98297 - in /llvm/trunk:
 include/llvm-c/Transforms/Scalar.h lib/Transforms/Scalar/Scalar.cpp
Message-ID: <20100311232119.D2F362A6C12C@llvm.org>

Author: sampo
Date: Thu Mar 11 17:21:19 2010
New Revision: 98297

URL: http://llvm.org/viewvc/llvm-project?rev=98297&view=rev
Log:
Whoops this already existed.

Modified:
    llvm/trunk/include/llvm-c/Transforms/Scalar.h
    llvm/trunk/lib/Transforms/Scalar/Scalar.cpp

Modified: llvm/trunk/include/llvm-c/Transforms/Scalar.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm-c/Transforms/Scalar.h?rev=98297&r1=98296&r2=98297&view=diff
==============================================================================
--- llvm/trunk/include/llvm-c/Transforms/Scalar.h (original)
+++ llvm/trunk/include/llvm-c/Transforms/Scalar.h Thu Mar 11 17:21:19 2010
@@ -98,9 +98,6 @@
 /** See llvm::createVerifierPass function. */
 void LLVMAddVerifierPass(LLVMPassManagerRef PM);
 
-/** PM->add(new TargetData(M)) */
-void LLVMAddTargetData(LLVMPassManagerRef PM, LLVMModuleRef M);
-
 #ifdef __cplusplus
 }
 #endif /* defined(__cplusplus) */

Modified: llvm/trunk/lib/Transforms/Scalar/Scalar.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/Scalar.cpp?rev=98297&r1=98296&r2=98297&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/Scalar.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/Scalar.cpp Thu Mar 11 17:21:19 2010
@@ -116,7 +116,3 @@
 void LLVMAddVerifierPass(LLVMPassManagerRef PM) {
   unwrap(PM)->add(createVerifierPass());
 }
-
-void LLVMAddTargetData(LLVMPassManagerRef PM, LLVMModuleRef M) {
-  unwrap(PM)->add(new TargetData(unwrap(M)));
-}




From isanbard at gmail.com  Thu Mar 11 17:39:44 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Thu, 11 Mar 2010 23:39:44 -0000
Subject: [llvm-commits] [llvm] r98300 -
	/llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
Message-ID: <20100311233944.C63BD2A6C12C@llvm.org>

Author: void
Date: Thu Mar 11 17:39:44 2010
New Revision: 98300

URL: http://llvm.org/viewvc/llvm-project?rev=98300&view=rev
Log:
MC-ize PPC's asm printing of stubs.

Modified:
    llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp

Modified: llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp?rev=98300&r1=98299&r2=98300&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp Thu Mar 11 17:39:44 2010
@@ -34,6 +34,7 @@
 #include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
@@ -806,11 +807,16 @@
     EmitAlignment(isPPC64 ? 3 : 2);
     
     for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
-      O << *Stubs[i].first << ":\n";
-      O << "\t.indirect_symbol " << *Stubs[i].second.getPointer() << '\n';
-      // FIXME: This should use the "GV is external" bit.
-      O << (isPPC64 ? "\t.quad\t0\n" : "\t.long\t0\n");
+      // L_foo$stub:
+      OutStreamer.EmitLabel(Stubs[i].first);
+      //   .indirect_symbol _foo
+      MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
+      OutStreamer.EmitSymbolAttribute(MCSym.getPointer(), MCSA_IndirectSymbol);
+      OutStreamer.EmitIntValue(0, isPPC64 ? 8 : 4/*size*/, 0/*addrspace*/);
     }
+
+    Stubs.clear();
+    OutStreamer.AddBlankLine();
   }
 
   Stubs = MMIMacho.GetHiddenGVStubList();
@@ -819,10 +825,17 @@
     EmitAlignment(isPPC64 ? 3 : 2);
     
     for (unsigned i = 0, e = Stubs.size(); i != e; ++i) {
-      O << *Stubs[i].first << ":\n";
-      O << (isPPC64 ? "\t.quad\t" : "\t.long\t")
-        << *Stubs[i].second.getPointer() << '\n';
+      // L_foo$stub:
+      OutStreamer.EmitLabel(Stubs[i].first);
+      //   .long _foo
+      OutStreamer.EmitValue(MCSymbolRefExpr::
+                            Create(Stubs[i].second.getPointer(),
+                                   OutContext),
+                            isPPC64 ? 8 : 4/*size*/, 0/*addrspace*/);
     }
+
+    Stubs.clear();
+    OutStreamer.AddBlankLine();
   }
 
   // Funny Darwin hack: This flag tells the linker that no global symbols
@@ -835,8 +848,6 @@
   return AsmPrinter::doFinalization(M);
 }
 
-
-
 /// createPPCAsmPrinterPass - Returns a pass that prints the PPC assembly code
 /// for a MachineFunction to the given output stream, in a format that the
 /// Darwin assembler can deal with.




From dpatel at apple.com  Thu Mar 11 17:44:52 2010
From: dpatel at apple.com (Devang Patel)
Date: Thu, 11 Mar 2010 23:44:52 -0000
Subject: [llvm-commits] [llvm] r98302 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100311234452.97BEE2A6C12C@llvm.org>

Author: dpatel
Date: Thu Mar 11 17:44:52 2010
New Revision: 98302

URL: http://llvm.org/viewvc/llvm-project?rev=98302&view=rev
Log:
There is no need to create specification DIE for definitions at DIFile level.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98302&r1=98301&r2=98302&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Thu Mar 11 17:44:52 2010
@@ -1324,6 +1324,7 @@
  // expect specification DIE in parent function. So avoid creating 
  // specification DIE for a function defined inside a function.
  if (SP.isDefinition() && !SP.getContext().isCompileUnit()
+     && !SP.getContext().isFile()
      && !SP.getContext().isSubprogram()) {
    addUInt(SPDie, dwarf::DW_AT_declaration, dwarf::DW_FORM_flag, 1);
   // Add arguments. 
@@ -1684,7 +1685,7 @@
   // Do not create specification DIE if context is either compile unit
   // or a subprogram.
   if (DI_GV.isDefinition() && !GVContext.isCompileUnit()
-      && !GVContext.isSubprogram()) {
+      && !GVContext.isFile() && !GVContext.isSubprogram()) {
     // Create specification DIE.
     DIE *VariableSpecDIE = new DIE(dwarf::DW_TAG_variable);
     addDIEEntry(VariableSpecDIE, dwarf::DW_AT_specification,




From andrewl at lenharth.org  Thu Mar 11 18:16:42 2010
From: andrewl at lenharth.org (Andrew Lenharth)
Date: Fri, 12 Mar 2010 00:16:42 -0000
Subject: [llvm-commits] [poolalloc] r98315 - in /poolalloc/trunk:
 include/dsa/DSCallGraph.h include/dsa/DSNode.h include/dsa/DataStructure.h
 include/dsa/sv/super_set.h include/poolalloc/PoolAllocate.h
 lib/DSA/BottomUpClosure.cpp lib/DSA/CompleteBottomUp.cpp
 lib/DSA/DSGraph.cpp lib/DSA/DataStructure.cpp lib/DSA/Local.cpp
 lib/DSA/Printer.cpp lib/DSA/TopDownClosure.cpp
 lib/PoolAllocate/RunTimeAssociate.cpp
 lib/PoolAllocate/TransformFunctionBody.cpp
Message-ID: <20100312001642.52C932A6C12C@llvm.org>

Author: alenhar2
Date: Thu Mar 11 18:16:41 2010
New Revision: 98315

URL: http://llvm.org/viewvc/llvm-project?rev=98315&view=rev
Log:
Split out the callgraph and seperate DSGraph's impl from the big mess of stuff

Added:
    poolalloc/trunk/include/dsa/DSCallGraph.h
    poolalloc/trunk/lib/DSA/DSGraph.cpp
      - copied, changed from r98209, poolalloc/trunk/lib/DSA/DataStructure.cpp
Modified:
    poolalloc/trunk/include/dsa/DSNode.h
    poolalloc/trunk/include/dsa/DataStructure.h
    poolalloc/trunk/include/dsa/sv/super_set.h
    poolalloc/trunk/include/poolalloc/PoolAllocate.h
    poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
    poolalloc/trunk/lib/DSA/CompleteBottomUp.cpp
    poolalloc/trunk/lib/DSA/DataStructure.cpp
    poolalloc/trunk/lib/DSA/Local.cpp
    poolalloc/trunk/lib/DSA/Printer.cpp
    poolalloc/trunk/lib/DSA/TopDownClosure.cpp
    poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp
    poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp

Added: poolalloc/trunk/include/dsa/DSCallGraph.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSCallGraph.h?rev=98315&view=auto
==============================================================================
--- poolalloc/trunk/include/dsa/DSCallGraph.h (added)
+++ poolalloc/trunk/include/dsa/DSCallGraph.h Thu Mar 11 18:16:41 2010
@@ -0,0 +1,119 @@
+//===- DSCallGaph.h - Build call graphs -------------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file was developed by the LLVM research group and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Implement a detailed call graph for DSA.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_DSCALLGRAPH_H
+#define	LLVM_DSCALLGRAPH_H
+
+#include "dsa/sv/set.h"
+#include <map>
+
+#include "llvm/Support/CallSite.h"
+
+namespace llvm {
+class Function;
+}
+
+template<class Iter>
+class KeyIterator {
+  Iter I;
+public:
+  typedef typename Iter::difference_type difference_type;
+  typedef typename Iter::value_type::first_type value_type;
+  typedef typename Iter::value_type::first_type* pointer;
+  typedef typename Iter::value_type::first_type& reference;
+  typedef typename Iter::iterator_category iterator_category;
+
+  KeyIterator(Iter i) :I(i) {}
+  Iter base() const { return I; }
+  reference    operator*() const { return I->first; }
+  KeyIterator  operator+(difference_type n) const { return KeyIterator(I + n); }
+  KeyIterator& operator++() { ++I; return *this; }
+  KeyIterator  operator++(int) { Iter OI = I; ++I; return KeyIterator(OI); }
+  KeyIterator& operator+= (difference_type n) { I += n; return *this; }
+  KeyIterator  operator- (difference_type n) const { return KeyIterator(I - n); }
+  KeyIterator& operator--() { --I; return *this; }
+  KeyIterator  operator--(int) { Iter OI = I; --I; return KeyIterator(OI); }
+  KeyIterator& operator-=(difference_type n) { I -= n; return *this; }
+  pointer operator->() const { return &I->first; }
+  bool operator==(const KeyIterator& RHS) const { return I == RHS.I; }
+  bool operator!=(const KeyIterator& RHS) const { return I != RHS.I; }
+};
+
+class DSCallGraph {
+
+  typedef sv::set<const llvm::Function*> FuncSet;
+  typedef std::map<llvm::CallSite, FuncSet> ActualCalleesTy;
+
+  ActualCalleesTy ActualCallees;
+
+public:
+
+  typedef ActualCalleesTy::mapped_type::const_iterator iterator;
+  typedef KeyIterator<ActualCalleesTy::const_iterator> key_iterator;
+
+  void insert(llvm::CallSite CS, const llvm::Function* F) {
+    if (F) ActualCallees[CS].insert(F);
+  }
+
+  template<class Iterator>
+  void insert(llvm::CallSite CS, Iterator _begin, Iterator _end) {
+    for(; _begin != _end; ++_begin)
+      insert(CS, *_begin);
+  }
+
+  iterator callee_begin(llvm::CallSite CS) const {
+    ActualCalleesTy::const_iterator ii = ActualCallees.find(CS);
+    if (ii == ActualCallees.end())
+      ii = ActualCallees.find(llvm::CallSite());
+    return ii->second.begin();
+  }
+
+  iterator callee_end(llvm::CallSite CS) const {
+    ActualCalleesTy::const_iterator ii = ActualCallees.find(CS);
+    if (ii == ActualCallees.end())
+      ii = ActualCallees.find(llvm::CallSite());
+    return ii->second.end();
+  }
+
+
+  unsigned callee_size(llvm::CallSite CS) const {
+    ActualCalleesTy::const_iterator ii = ActualCallees.find(CS);
+    if (ii == ActualCallees.end())
+      return 0;
+    return ii->second.size();
+  }
+
+  unsigned size() const {
+    unsigned sum = 0;
+    for (ActualCalleesTy::const_iterator ii = ActualCallees.begin(),
+            ee = ActualCallees.end(); ii != ee; ++ii)
+      sum += ii->second.size();
+    return sum;
+  }
+
+  void clear() {
+    ActualCallees.clear();
+  }
+
+  key_iterator key_begin() const {
+    return key_iterator(ActualCallees.begin());
+  }
+
+  key_iterator key_end() const {
+    return key_iterator(ActualCallees.end());
+  }
+
+};
+
+#endif	/* LLVM_DSCALLGRAPH_H */
+

Modified: poolalloc/trunk/include/dsa/DSNode.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSNode.h?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSNode.h (original)
+++ poolalloc/trunk/include/dsa/DSNode.h Thu Mar 11 18:16:41 2010
@@ -15,7 +15,6 @@
 #define LLVM_ANALYSIS_DSNODE_H
 
 #include "dsa/DSSupport.h"
-#include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/ilist_node.h"
 
@@ -150,10 +149,12 @@
   inline const_iterator end() const;
 
   /// type_* - Provide iterators for accessing types.  Some types may be null
-  TyMapTy::iterator type_begin() { return TyMap.begin(); }
-  TyMapTy::iterator type_end()   { return TyMap.end(); }
-  TyMapTy::const_iterator type_begin() const { return TyMap.begin(); }
-  TyMapTy::const_iterator type_end()   const { return TyMap.end(); }
+  typedef TyMapTy::iterator type_iterator;
+  typedef TyMapTy::const_iterator const_type_iterator;
+  type_iterator type_begin() { return TyMap.begin(); }
+  type_iterator type_end()   { return TyMap.end(); }
+  const_type_iterator type_begin() const { return TyMap.begin(); }
+  const_type_iterator type_end()   const { return TyMap.end(); }
 
   /// edge_* - Provide iterators for accessing outgoing edges.  Some outgoing
   /// edges may be null.
@@ -248,6 +249,16 @@
   void mergeTypeInfo(const TyMapTy::mapped_type TyIt, unsigned Offset);
   void mergeTypeInfo(const DSNode* D, unsigned Offset);
 
+  // Types records might exist without types in them
+  bool hasNoType() {
+    type_iterator ii = type_begin(), ee = type_end();
+    while (ii != ee) {
+      if (ii->second) return false;
+      ++ii;
+    }
+    return true;
+  }
+
   /// foldNodeCompletely - If we determine that this node has some funny
   /// behavior happening to it that we cannot represent, we fold it down to a
   /// single, completely pessimistic, node.  This node is represented as a
@@ -382,6 +393,8 @@
   ///
   void forwardNode(DSNode *To, unsigned Offset);
 
+  void cleanEdges();
+
   void print(llvm::raw_ostream &O, const DSGraph *G) const;
   void dump() const;
 

Modified: poolalloc/trunk/include/dsa/DataStructure.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DataStructure.h?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DataStructure.h (original)
+++ poolalloc/trunk/include/dsa/DataStructure.h Thu Mar 11 18:16:41 2010
@@ -20,6 +20,7 @@
 #include "llvm/ADT/EquivalenceClasses.h"
 
 #include "dsa/EntryPointAnalysis.h"
+#include "dsa/DSCallGraph.h"
 #include "dsa/sv/set.h"
 #include "dsa/sv/super_set.h"
 
@@ -39,12 +40,8 @@
 FunctionPass *createDataStructureGraphCheckerPass();
 
 class DataStructures : public ModulePass {
-  typedef std::map<const Instruction*, sv::set<const Function*> > ActualCalleesTy;
   typedef std::map<const Function*, DSGraph*> DSInfoTy;
-public:
-  typedef sv::set<const Function*>::const_iterator callee_iterator;
-  
-private:
+
   /// TargetData, comes in handy
   TargetData* TD;
 
@@ -67,9 +64,6 @@
   // DSInfo, one graph for each function
   DSInfoTy DSInfo;
 
-  // Callgraph, as computed so far
-  ActualCalleesTy ActualCallees;
-
   // Name for printing
   const char* printname;
 
@@ -84,29 +78,16 @@
 
   SuperSet<const Type*>* TypeSS;
 
+  // Callgraph, as computed so far
+  DSCallGraph callgraph;
+
   void init(DataStructures* D, bool clone, bool printAuxCalls, bool copyGlobalAuxCalls, bool resetAux);
   void init(TargetData* T);
 
   void formGlobalECs();
 
-
-  void callee_add(const Instruction* I, const Function* F) {
-    ActualCallees[I].insert(F);
-  }
-
-  template<class Iterator>
-  void callee_add_many(const Instruction* I, Iterator _begin, Iterator _end) {
-    ActualCallees[I].insert(_begin,_end);
-//    typename ActualCalleesTy::mapped_type& S = ActualCallees[I];
-//    for (; _begin != _end; ++_begin)
-//      S.insert(*_begin);
-  }
-
   DataStructures(intptr_t id, const char* name) 
-    : ModulePass(id), TD(0), GraphSource(0), printname(name), GlobalsGraph(0) {
-    //a dummy node for empty call sites
-    ActualCallees[0];
-    
+    : ModulePass(id), TD(0), GraphSource(0), printname(name), GlobalsGraph(0) {  
     // For now, the graphs are owned by this pass
     DSGraphsStolen = false;
   }
@@ -117,39 +98,6 @@
   void print(llvm::raw_ostream &O, const Module *M) const;
   void dumpCallGraph() const;
 
-  callee_iterator callee_begin(const Instruction *I) const {
-    ActualCalleesTy::const_iterator ii = ActualCallees.find(I);
-    if (ii == ActualCallees.end())
-      ii = ActualCallees.find(0);
-    return ii->second.begin();
-  }
-
-  callee_iterator callee_end(const Instruction *I) const {
-    ActualCalleesTy::const_iterator ii = ActualCallees.find(I);
-    if (ii == ActualCallees.end())
-      ii = ActualCallees.find(0);
-    return ii->second.end();
-  }
-
-  void callee_site(const Instruction* I) {
-    ActualCallees[I];
-  }
-
-  unsigned callee_size() const {
-    unsigned sum = 0;
-    for (ActualCalleesTy::const_iterator ii = ActualCallees.begin(),
-           ee = ActualCallees.end(); ii != ee; ++ii)
-      sum += ii->second.size();
-    return sum;
-  }
-
-  void callee_get_keys(std::vector<const Instruction*>& keys) {
-    for (ActualCalleesTy::const_iterator ii = ActualCallees.begin(),
-           ee = ActualCallees.end(); ii != ee; ++ii)
-      if (ii->first)
-        keys.push_back(ii->first);
-  }
-
   virtual void releaseMemory();
 
   virtual bool hasDSGraph(const Function &F) const {
@@ -176,6 +124,8 @@
 
   TargetData& getTargetData() const { return *TD; }
 
+  const DSCallGraph& getCallGraph() const { return callgraph; }
+
   SuperSet<const Type*>& getTypeSS() const { return *TypeSS; }
   
   /// deleteValue/copyValue - Interfaces to update the DSGraphs in the program.

Modified: poolalloc/trunk/include/dsa/sv/super_set.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/sv/super_set.h?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/sv/super_set.h (original)
+++ poolalloc/trunk/include/dsa/sv/super_set.h Thu Mar 11 18:16:41 2010
@@ -16,6 +16,7 @@
 
 template<typename Ty>
 class SuperSet {
+  //std::set provides stable iterators, and that matters a lot
   typedef sv::set<Ty> InnerSetTy;
   typedef std::set<InnerSetTy> OuterSetTy;
   OuterSetTy container;
@@ -23,6 +24,7 @@
   typedef const typename OuterSetTy::value_type* setPtr;
 
   setPtr getOrCreate(sv::set<Ty>& S) {
+    if (S.empty()) return 0;
     return &(*container.insert(S).first);
   }
 

Modified: poolalloc/trunk/include/poolalloc/PoolAllocate.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/poolalloc/PoolAllocate.h?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/include/poolalloc/PoolAllocate.h (original)
+++ poolalloc/trunk/include/poolalloc/PoolAllocate.h Thu Mar 11 18:16:41 2010
@@ -147,8 +147,6 @@
 
   virtual Value * getGlobalPool (const DSNode * Node) {return 0;}
 
-  virtual DataStructures::callee_iterator callee_begin (Instruction *I) { return Graphs->callee_begin(I);}
-  virtual DataStructures::callee_iterator callee_end   (Instruction *I) { return Graphs->callee_end(I);}
 };
 
 /// PoolAllocate - The main pool allocation pass
@@ -366,16 +364,6 @@
       return I->second;
   }
 
-  virtual DataStructures::callee_iterator
-  callee_begin (Instruction * I) {
-    return Graphs->callee_begin(I);
-  }
-
-  virtual DataStructures::callee_iterator
-  callee_end (Instruction * I) {
-    return Graphs->callee_end(I);
-  }
-
 protected:
   
   /// AddPoolPrototypes - Add prototypes for the pool functions to the

Modified: poolalloc/trunk/lib/DSA/BottomUpClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/BottomUpClosure.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/BottomUpClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/BottomUpClosure.cpp Thu Mar 11 18:16:41 2010
@@ -45,6 +45,13 @@
   return runOnModuleInternal(M);
 }
 
+// BU:
+// Construct the callgraph from the local graphs
+// Find SCCs
+// inline bottum up
+//
+// We must split these out (they were merged in PLDI07) to handle multiple
+// entry-points correctly.
 bool BUDataStructures::runOnModuleInternal(Module& M) {
   std::vector<const Function*> Stack;
   std::map<const Function*, unsigned> ValMap;
@@ -59,24 +66,24 @@
     if (!hasDSGraph(**ii)) {
       errs() << debugname << ": Main Function: " << (*ii)->getName() << "\n";
       calculateGraphs(*ii, Stack, NextID, ValMap);
-      CloneAuxIntoGlobal(getDSGraph(**ii));
+      //CloneAuxIntoGlobal(getDSGraph(**ii));
     }
 
-  //errs() << "done main Funcs\n";
+  errs() << "done main Funcs\n";
 
   // Calculate the graphs for any functions that are unreachable from main...
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
     if (!I->isDeclaration() && !hasDSGraph(*I)) {
-      DEBUG(
+      //DEBUG(
             if (EntryPoints.size())
             errs() << debugname << ": Function unreachable from main: "
             << I->getName() << "\n";
-      );
+      //);
       calculateGraphs(I, Stack, NextID, ValMap); // Calculate all graphs.
-      CloneAuxIntoGlobal(getDSGraph(*I));
+      //CloneAuxIntoGlobal(getDSGraph(*I));
     }
 
-  //errs() << "done unreachable Funcs\n";
+  errs() << "done unreachable Funcs\n";
 
   // If we computed any temporary indcallgraphs, free them now.
   for (std::map<std::vector<const Function*>,
@@ -117,7 +124,7 @@
                                    DSGraph::IgnoreGlobals);
   }
 
-  NumCallEdges += callee_size();
+  NumCallEdges += callgraph.size();
 
   return false;
 }
@@ -371,7 +378,6 @@
     DEBUG(Graph->AssertGraphOK(); Graph->getGlobalsGraph()->AssertGraphOK());
     
     DSCallSite &CS = *TempFCs.begin();
-    Instruction *TheCall = CS.getCallSite().getInstruction();
 
     CalledFuncs.clear();
 
@@ -380,7 +386,7 @@
     if (CS.getRetVal().isNull() && CS.getNumPtrArgs() == 0) {
       if (!CS.isDirectCall()) {
         GetAnyCallees(CS, CalledFuncs);
-        callee_add_many(TheCall, CalledFuncs.begin(), CalledFuncs.end());
+        callgraph.insert(CS.getCallSite(), CalledFuncs.begin(), CalledFuncs.end());
       }
       TempFCs.erase(TempFCs.begin());
       continue;
@@ -392,7 +398,7 @@
       //Get any callees we do have for the callgraph
       if (!CS.isDirectCall()) {
         GetAnyCallees(CS, CalledFuncs);
-        callee_add_many(TheCall, CalledFuncs.begin(), CalledFuncs.end());
+        callgraph.insert(CS.getCallSite(), CalledFuncs.begin(), CalledFuncs.end());
       }
       // Remember that we could not resolve this yet!
       AuxCallsList.splice(AuxCallsList.end(), TempFCs, TempFCs.begin());
@@ -401,9 +407,7 @@
 
     DSGraph *GI;
     
-    for (std::vector<const Function*>::iterator ii = CalledFuncs.begin(), ee = CalledFuncs.end();
-         ii != ee; ++ii) 
-      callee_add(TheCall, *ii);
+    callgraph.insert(CS.getCallSite(), CalledFuncs.begin(), CalledFuncs.end());
 
     if (CalledFuncs.size() == 1) {
       const Function *Callee = CalledFuncs[0];
@@ -510,7 +514,7 @@
   // reach live nodes as live.
   Graph->removeDeadNodes(DSGraph::KeepUnreachableGlobals);
 
-  cloneIntoGlobals(Graph);
+//  cloneIntoGlobals(Graph);
   //Graph.writeGraphToFile(cerr, "bu_" + F.getName());
 }
 

Modified: poolalloc/trunk/lib/DSA/CompleteBottomUp.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/CompleteBottomUp.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/CompleteBottomUp.cpp (original)
+++ poolalloc/trunk/lib/DSA/CompleteBottomUp.cpp Thu Mar 11 18:16:41 2010
@@ -44,16 +44,15 @@
   // Loop over all of the indirect calls in the program.  If a call site can
   // call multiple different functions, we need to unify all of the callees into
   // the same equivalence class.
-  std::vector<const Instruction*> keys;
-  callee_get_keys(keys);
 
   DSGraph* G = getGlobalsGraph();
   DSGraph::ScalarMapTy& SM = G->getScalarMap();
 
   //mege nodes in the global graph for these functions
-  for (std::vector<const Instruction*>::iterator ii = keys.begin(), ee = keys.end();
-       ii != ee; ++ii) {
-    callee_iterator csi = callee_begin(*ii), cse = callee_end(*ii);
+  for (DSCallGraph::key_iterator ii = callgraph.key_begin(),
+       ee = callgraph.key_end(); ii != ee; ++ii) {
+    DSCallGraph::iterator csi = callgraph.callee_begin(*ii),
+            cse = callgraph.callee_end(*ii);
     if (csi != cse && SM.find(*csi) != SM.end()) {
       DSNodeHandle& SrcNH = SM.find(*csi)->second;
       ++csi;

Copied: poolalloc/trunk/lib/DSA/DSGraph.cpp (from r98209, poolalloc/trunk/lib/DSA/DataStructure.cpp)
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DSGraph.cpp?p2=poolalloc/trunk/lib/DSA/DSGraph.cpp&p1=poolalloc/trunk/lib/DSA/DataStructure.cpp&r1=98209&r2=98315&rev=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DSGraph.cpp Thu Mar 11 18:16:41 2010
@@ -38,766 +38,13 @@
 
 #define COLLAPSE_ARRAYS_AGGRESSIVELY 0
 namespace {
-  STATISTIC (NumFolds, "Number of nodes completely folded");
   STATISTIC (NumCallNodesMerged, "Number of call nodes merged");
-  STATISTIC (NumNodeAllocated  , "Number of nodes allocated");
   STATISTIC (NumDNE            , "Number of nodes removed by reachability");
   STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
   STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
   static cl::opt<unsigned>
-  DSAFieldLimit("dsa-field-limit", cl::Hidden,
-                cl::desc("Number of fields to track before collapsing a node"),
-                cl::init(256));
-}
-
-/// isForwarding - Return true if this NodeHandle is forwarding to another
-/// one.
-bool DSNodeHandle::isForwarding() const {
-  return N && N->isForwarding();
-}
-
-DSNode *DSNodeHandle::HandleForwarding() const {
-  assert(N->isForwarding() && "Can only be invoked if forwarding!");
-  DEBUG(
-        { //assert not looping
-          DSNode* NH = N;
-          sv::set<DSNode*> seen;
-          while(NH && NH->isForwarding()) {
-            assert(seen.find(NH) == seen.end() && "Loop detected");
-            seen.insert(NH);
-            NH = NH->ForwardNH.N;
-          }
-        }
-        );
-  // Handle node forwarding here!
-  DSNode *Next = N->ForwardNH.getNode();  // Cause recursive shrinkage
-  Offset += N->ForwardNH.getOffset();
-
-  if (--N->NumReferrers == 0) {
-    // Removing the last referrer to the node, sever the forwarding link
-    N->stopForwarding();
-  }
-
-  N = Next;
-  N->NumReferrers++;
-
-  if (N->getSize() <= Offset) {
-    assert(N->getSize() <= 1 && "Forwarded to shrunk but not collapsed node?");
-    Offset = 0;
-  }
-  return N;
-}
-
-//===----------------------------------------------------------------------===//
-// DSScalarMap Implementation
-//===----------------------------------------------------------------------===//
-
-DSNodeHandle &DSScalarMap::AddGlobal(const GlobalValue *GV) {
-  assert(ValueMap.count(GV) == 0 && "GV already exists!");
-
-  // If the node doesn't exist, check to see if it's a global that is
-  // equated to another global in the program.
-  EquivalenceClasses<const GlobalValue*>::iterator ECI = GlobalECs.findValue(GV);
-  if (ECI != GlobalECs.end()) {
-    const GlobalValue *Leader = *GlobalECs.findLeader(ECI);
-    if (Leader != GV) {
-      GV = Leader;
-      iterator I = ValueMap.find(GV);
-      if (I != ValueMap.end())
-        return I->second;
-    }
-  }
-
-  // Okay, this is either not an equivalenced global or it is the leader, it
-  // will be inserted into the scalar map now.
-  GlobalSet.insert(GV);
-
-  return ValueMap.insert(std::make_pair(GV, DSNodeHandle())).first->second;
-}
-
-//===----------------------------------------------------------------------===//
-// DSNode Implementation
-//===----------------------------------------------------------------------===//
-
-DSNode::DSNode(DSGraph *G)
-: NumReferrers(0), Size(0), ParentGraph(G), NodeType(0) {
-  // Add the type entry if it is specified...
-  if (G) G->addNode(this);
-  ++NumNodeAllocated;
-}
-
-// DSNode copy constructor... do not copy over the referrers list!
-DSNode::DSNode(const DSNode &N, DSGraph *G, bool NullLinks)
-  : NumReferrers(0), Size(N.Size), ParentGraph(G), TyMap(N.TyMap),
-    Globals(N.Globals), NodeType(N.NodeType) {
-  if (!NullLinks) Links = N.Links;
-  G->addNode(this);
-  ++NumNodeAllocated;
-}
-
-DSNode::~DSNode() {
-  dropAllReferences();
-  assert(hasNoReferrers() && "Referrers to dead node exist!");
-}
-
-void DSNode::assertOK() const {
-//  assert(((Ty && Ty->getTypeID() != Type::VoidTyID) ||
-//         ((!Ty || Ty->getTypeID() == Type::VoidTyID) && (Size == 0 ||
-//         (NodeType & DSNode::ArrayNode)))) &&
-//         "Node not OK!");
-
-  assert(ParentGraph && "Node has no parent?");
-  const DSScalarMap &SM = ParentGraph->getScalarMap();
-  for (globals_iterator ii = globals_begin(), ee = globals_end();
-       ii != ee; ++ii) {
-    assert(SM.global_count(*ii));
-    assert(SM.find(*ii)->second.getNode() == this);
-  }
-}
-
-/// forwardNode - Mark this node as being obsolete, and all references to it
-/// should be forwarded to the specified node and offset.
-///
-void DSNode::forwardNode(DSNode *To, unsigned Offset) {
-  assert(this != To && "Cannot forward a node to itself!");
-  assert(ForwardNH.isNull() && "Already forwarding from this node!");
-  if (To->Size <= 1) Offset = 0;
-  assert((Offset < To->Size || (Offset == To->Size && Offset == 0)) &&
-         "Forwarded offset is wrong!");
-  ForwardNH.setTo(To, Offset);
-  NodeType = DeadNode;
-  Size = 0;
-
-  DSNodeHandle ToNH(To,Offset);
-
-  //Move the Links
-  for (unsigned x = 0, xe = Links.size(); x != xe; ++x)
-    if (!Links[x].isNull()) {
-      // Compute the offset into the current node at which to
-      // merge this link.  In the common case, this is a linear
-      // relation to the offset in the original node (with
-      // wrapping), but if the current node gets collapsed due to
-      // recursive merging, we must make sure to merge in all remaining
-      // links at offset zero.
-      unsigned MergeOffset = 0;
-      if (ToNH.getNode()->getSize() != 1)
-        MergeOffset = (x + Offset) % ToNH.getNode()->getSize();
-      ToNH.getNode()->addEdgeTo(MergeOffset, Links[x]);
-    }
-  Links.clear();
-
-  // Remove this node from the parent graph's Nodes list.
-  ParentGraph->unlinkNode(this);
-  ParentGraph = 0;
-}
-
-// addGlobal - Add an entry for a global value to the Globals list.  This also
-// marks the node with the 'G' flag if it does not already have it.
-//
-void DSNode::addGlobal(const GlobalValue *GV) {
-  // First, check to make sure this is the leader if the global is in an
-  // equivalence class.
-  GV = getParentGraph()->getScalarMap().getLeaderForGlobal(GV);
-
-  Globals.insert(GV);
-  setGlobalMarker();
-}
-
-void DSNode::addFunction(const Function* F) {
-  addGlobal(F);
-}
-
-// removeGlobal - Remove the specified global that is explicitly in the globals
-// list.
-void DSNode::removeGlobal(const GlobalValue *GV) {
-  assert (Globals.count(GV) && "Global not in Node!");
-  Globals.erase(GV);
-}
-
-/// foldNodeCompletely - If we determine that this node has some funny
-/// behavior happening to it that we cannot represent, we fold it down to a
-/// single, completely pessimistic, node.  This node is represented as a
-/// single byte with a single TypeEntry of "void".
-///
-void DSNode::foldNodeCompletely() {
-  if (isNodeCompletelyFolded()) return;  // If this node is already folded...
-
-//  assert(0 && "Folding is happening");
-
-  ++NumFolds;
-
-  //Collapsed nodes don't really need a type
-  TyMap.clear();
-
-  // If this node has a size that is <= 1, we don't need to create a forwarding
-  // node.
-  if (getSize() <= 1) {
-    setCollapsedMarker();
-    Size = 1;
-    assert(Links.size() <= 1 && "Size is 1, but has more links?");
-  } else {
-    // Create the node we are going to forward to.  This is required because
-    // some referrers may have an offset that is > 0.  By forcing them to
-    // forward, the forwarder has the opportunity to correct the offset.
-    DSNode *DestNode = new DSNode(ParentGraph);
-    DestNode->NodeType = NodeType;
-    DestNode->setCollapsedMarker();
-    DestNode->Size = 1;
-    DestNode->Globals.swap(Globals);
-
-    // Start forwarding to the destination node...
-    forwardNode(DestNode, 0);
-
-  }
-}
-
-/// isNodeCompletelyFolded - Return true if this node has been completely
-/// folded down to something that can never be expanded, effectively losing
-/// all of the field sensitivity that may be present in the node.
-///
-bool DSNode::isNodeCompletelyFolded() const {
-  return  isCollapsedNode();
-}
-
-/// addFullGlobalsList - Compute the full set of global values that are
-/// represented by this node.  Unlike getGlobalsList(), this requires fair
-/// amount of work to compute, so don't treat this method call as free.
-void DSNode::addFullGlobalsList(std::vector<const GlobalValue*> &List) const {
-  if (globals_begin() == globals_end()) return;
-
-  EquivalenceClasses<const GlobalValue*> &EC = getParentGraph()->getGlobalECs();
-
-  for (globals_iterator I = globals_begin(), E = globals_end(); I != E; ++I) {
-    EquivalenceClasses<const GlobalValue*>::iterator ECI = EC.findValue(*I);
-    if (ECI == EC.end())
-      List.push_back(*I);
-    else
-      List.insert(List.end(), EC.member_begin(ECI), EC.member_end());
-  }
-}
-
-/// addFullFunctionList - Identical to addFullGlobalsList, but only return the
-/// functions in the full list.
-void DSNode::addFullFunctionList(std::vector<const Function*> &List) const {
-  if (globals_begin() == globals_end()) return;
-
-  EquivalenceClasses<const GlobalValue*> &EC = getParentGraph()->getGlobalECs();
-
-  for (globals_iterator I = globals_begin(), E = globals_end(); I != E; ++I) {
-    EquivalenceClasses<const GlobalValue*>::iterator ECI = EC.findValue(*I);
-    if (ECI == EC.end()) {
-      if (const Function *F = dyn_cast<Function>(*I))
-        List.push_back(F);
-    } else {
-      for (EquivalenceClasses<const GlobalValue*>::member_iterator MI =
-             EC.member_begin(ECI), E = EC.member_end(); MI != E; ++MI)
-        if (const Function *F = dyn_cast<Function>(*MI))
-          List.push_back(F);
-    }
-  }
-}
-
-/// mergeTypeInfo - This method merges the specified type into the current node
-/// at the specified offset.  This may update the current node's type record if
-/// this gives more information to the node, it may do nothing to the node if
-/// this information is already known, or it may merge the node completely (and
-/// return true) if the information is incompatible with what is already known.
-///
-/// This method returns true if the node is completely folded, otherwise false.
-///
-void DSNode::mergeTypeInfo(const Type *NewTy, unsigned Offset) {
-
-  if (!NewTy || NewTy->isVoidTy()) return;
-
-  if (isCollapsedNode()) return;
-  if (isArrayNode()) Offset %= getSize();
-
-  if (Offset >= getSize()) growSize(Offset+1);
-
-  TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(TyMap[Offset], NewTy);
-}
-
-void DSNode::mergeTypeInfo(const TyMapTy::mapped_type TyIt, unsigned Offset) {
-  if (isCollapsedNode()) return;
-  if (isArrayNode()) Offset %= getSize();
-
-  if (Offset >= getSize())
-    growSize(Offset + 1);
-
-  if (!TyMap[Offset])
-    TyMap[Offset] = TyIt;
-  if (TyIt) {
-    sv::set<const Type*> S(*TyMap[Offset]);
-    S.insert(TyIt->begin(), TyIt->end());
-    TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(S);
-  }
-}
-
-void DSNode::mergeTypeInfo(const DSNode* DN, unsigned Offset) {
-  if (isCollapsedNode()) return;
-
-  for (TyMapTy::const_iterator ii = DN->TyMap.begin(), ee = DN->TyMap.end();
-       ii != ee; ++ii)
-    mergeTypeInfo(ii->second, ii->first + Offset);
-}
-
-/// addEdgeTo - Add an edge from the current node to the specified node.  This
-/// can cause merging of nodes in the graph.
-///
-void DSNode::addEdgeTo(unsigned Offset, const DSNodeHandle &NH) {
-  if (NH.isNull()) return;       // Nothing to do
-
-  if (isNodeCompletelyFolded())
-    Offset = 0;
-
-  DSNodeHandle &ExistingEdge = getLink(Offset);
-  if (!ExistingEdge.isNull()) {
-    // Merge the two nodes...
-    ExistingEdge.mergeWith(NH);
-  } else {                             // No merging to perform...
-    setLink(Offset, NH);               // Just force a link in there...
-  }
-}
-
-
-void DSNode::mergeGlobals(const DSNode &RHS) {
-  Globals.insert(RHS.Globals.begin(), RHS.Globals.end());
-}
-
-static unsigned gcd(unsigned m, unsigned n) {
-  if (m < n) return gcd(n,m);
-  unsigned r = m % n;
-  if (r == 0) {
-    return n;
-  } else {
-    return gcd(n, r);
-  }
-}
-
-// MergeNodes - Helper function for DSNode::mergeWith().
-// This function does the hard work of merging two nodes, CurNodeH
-// and NH after filtering out trivial cases and making sure that
-// CurNodeH.offset >= NH.offset.
-//
-// ***WARNING***
-// Since merging may cause either node to go away, we must always
-// use the node-handles to refer to the nodes.  These node handles are
-// automatically updated during merging, so will always provide access
-// to the correct node after a merge.
-//
-void DSNode::MergeNodes(DSNodeHandle& CurNodeH, DSNodeHandle& NH) {
-  assert(CurNodeH.getOffset() >= NH.getOffset() &&
-         "This should have been enforced in the caller.");
-  assert(CurNodeH.getNode()->getParentGraph()==NH.getNode()->getParentGraph() &&
-         "Cannot merge two nodes that are not in the same graph!");
-
-  // Now we know that Offset >= NH.Offset, so convert it so our "Offset" (with
-  // respect to NH.Offset) is now zero.  NOffset is the distance from the base
-  // of our object that N starts from.
-  //
-  unsigned NOffset = CurNodeH.getOffset()-NH.getOffset();
-  unsigned NSize = NH.getNode()->getSize();
-
-  // If the two nodes are of different size, and the smaller node has the array
-  // bit set, collapse!
-  if (NSize != CurNodeH.getNode()->getSize()) {
-#if COLLAPSE_ARRAYS_AGGRESSIVELY
-    if (NSize < CurNodeH.getNode()->getSize()) {
-      if (NH.getNode()->isArray())
-        NH.getNode()->foldNodeCompletely();
-    } else if (CurNodeH.getNode()->isArray()) {
-      NH.getNode()->foldNodeCompletely();
-    }
-#endif
-  }
-
-  // If we are merging a node with a completely folded node, then both nodes are
-  // now completely folded.
-  //
-  if (CurNodeH.getNode()->isNodeCompletelyFolded()) {
-    if (!NH.getNode()->isNodeCompletelyFolded()) {
-      NH.getNode()->foldNodeCompletely();
-      assert(NH.getNode() && NH.getOffset() == 0 &&
-             "folding did not make offset 0?");
-      NOffset = NH.getOffset();
-      NSize = NH.getNode()->getSize();
-      assert(NOffset == 0 && NSize == 1);
-    }
-  } else if (NH.getNode()->isNodeCompletelyFolded()) {
-    CurNodeH.getNode()->foldNodeCompletely();
-    assert(CurNodeH.getNode() && CurNodeH.getOffset() == 0 &&
-           "folding did not make offset 0?");
-    NSize = NH.getNode()->getSize();
-    NOffset = NH.getOffset();
-    assert(NOffset == 0 && NSize == 1);
-  } 
-
-  DSNode *N = NH.getNode();
-  if (CurNodeH.getNode() == N || N == 0) return;
-  assert(!CurNodeH.getNode()->isDeadNode());
-
-  // Merge the type entries of the two nodes together...
-  CurNodeH.getNode()->mergeTypeInfo(NH.getNode(), NOffset);
-  if (NH.getNode()->getSize() + NOffset > CurNodeH.getNode()->getSize())
-    CurNodeH.getNode()->growSize(NH.getNode()->getSize() + NOffset);
-  assert(!CurNodeH.getNode()->isDeadNode());
-
-  // Merge the NodeType information.
-  CurNodeH.getNode()->NodeType |= N->NodeType;
-
-  // Start forwarding to the new node!
-  N->forwardNode(CurNodeH.getNode(), NOffset);
-  assert(!CurNodeH.getNode()->isDeadNode());
-
-  // Merge the globals list...
-  CurNodeH.getNode()->mergeGlobals(*N);
-
-  // Delete the globals from the old node...
-  N->Globals.clear();
-}
-
-
-/// mergeWith - Merge this node and the specified node, moving all links to and
-/// from the argument node into the current node, deleting the node argument.
-/// Offset indicates what offset the specified node is to be merged into the
-/// current node.
-///
-/// The specified node may be a null pointer (in which case, we update it to
-/// point to this node).
-///
-void DSNode::mergeWith(const DSNodeHandle &NH, unsigned Offset) {
-  //DOUT << "mergeWith: " << this << " becomes " << NH.getNode() << "\n";
-  DSNode *N = NH.getNode();
-  if (N == this && NH.getOffset() == Offset)
-    return;  // Noop
-
-  // If the RHS is a null node, make it point to this node!
-  if (N == 0) {
-    NH.mergeWith(DSNodeHandle(this, Offset));
-    return;
-  }
-
-  assert(!N->isDeadNode() && !isDeadNode());
-  assert(!hasNoReferrers() && "Should not try to fold a useless node!");
-
-  if (N == this) {
-    // We cannot merge two pieces of the same node together, collapse the node
-    // completely.
-    DEBUG(errs() << "Attempting to merge two chunks of the same node together!\n");
-    foldNodeCompletely();
-    return;
-  }
-
-  // If both nodes are not at offset 0, make sure that we are merging the node
-  // at an later offset into the node with the zero offset.
-  //
-  if (Offset < NH.getOffset()) {
-    N->mergeWith(DSNodeHandle(this, Offset), NH.getOffset());
-    return;
-  } else if (Offset == NH.getOffset() && getSize() < N->getSize()) {
-    // If the offsets are the same, merge the smaller node into the bigger node
-    N->mergeWith(DSNodeHandle(this, Offset), NH.getOffset());
-    return;
-  }
-
-  // Ok, now we can merge the two nodes.  Use a static helper that works with
-  // two node handles, since "this" may get merged away at intermediate steps.
-  DSNodeHandle CurNodeH(this, Offset);
-  DSNodeHandle NHCopy(NH);
-  if (CurNodeH.getOffset() >= NHCopy.getOffset())
-    DSNode::MergeNodes(CurNodeH, NHCopy);
-  else
-    DSNode::MergeNodes(NHCopy, CurNodeH);
-}
-
-
-//===----------------------------------------------------------------------===//
-// ReachabilityCloner Implementation
-//===----------------------------------------------------------------------===//
-
-DSNodeHandle ReachabilityCloner::getClonedNH(const DSNodeHandle &SrcNH) {
-  if (SrcNH.isNull()) return DSNodeHandle();
-  const DSNode *SN = SrcNH.getNode();
-
-  DSNodeHandle &NH = NodeMap[SN];
-  if (!NH.isNull()) { // Node already mapped?
-    DSNode *NHN = NH.getNode();
-    return DSNodeHandle(NHN, NH.getOffset() + SrcNH.getOffset());
-  }
-  
-  // If SrcNH has globals and the destination graph has one of the same globals,
-  // merge this node with the destination node, which is much more efficient.
-  if (SN->globals_begin() != SN->globals_end()) {
-    DSScalarMap &DestSM = Dest->getScalarMap();
-    for (DSNode::globals_iterator I = SN->globals_begin(),E = SN->globals_end();
-         I != E; ++I) {
-      const GlobalValue *GV = *I;
-      DSScalarMap::iterator GI = DestSM.find(GV);
-      if (GI != DestSM.end() && !GI->second.isNull()) {
-        // We found one, use merge instead!
-        merge(GI->second, Src->getNodeForValue(GV));
-        assert(!NH.isNull() && "Didn't merge node!");
-        DSNode *NHN = NH.getNode();
-        return DSNodeHandle(NHN, NH.getOffset()+SrcNH.getOffset());
-      }
-    }
-  }
-
-  if (!createDest) return DSNodeHandle(0,0);
-
-  DSNode *DN = new DSNode(*SN, Dest, true /* Null out all links */);
-  DN->maskNodeTypes(BitsToKeep);
-  NH = DN;
-  //DOUT << "getClonedNH: " << SN << " becomes " << DN << "\n";
-
-  // Next, recursively clone all outgoing links as necessary.  Note that
-  // adding these links can cause the node to collapse itself at any time, and
-  // the current node may be merged with arbitrary other nodes.  For this
-  // reason, we must always go through NH.
-  DN = 0;
-  for (DSNode::const_edge_iterator ii = SN->edge_begin(), ee = SN->edge_end();
-       ii != ee; ++ii) {
-    const DSNodeHandle &SrcEdge = ii->second;
-    if (!SrcEdge.isNull()) {
-      const DSNodeHandle &DestEdge = getClonedNH(SrcEdge);
-      // Compute the offset into the current node at which to
-      // merge this link.  In the common case, this is a linear
-      // relation to the offset in the original node (with
-      // wrapping), but if the current node gets collapsed due to
-      // recursive merging, we must make sure to merge in all remaining
-      // links at offset zero.
-      unsigned MergeOffset = 0;
-      DSNode *CN = NH.getNode();
-      if (CN->getSize() != 1)
-        MergeOffset = (ii->first + NH.getOffset()) % CN->getSize();
-      CN->addEdgeTo(MergeOffset, DestEdge);
-    }
-  }
-
-  // If this node contains any globals, make sure they end up in the scalar
-  // map with the correct offset.
-  for (DSNode::globals_iterator I = SN->globals_begin(), E = SN->globals_end();
-       I != E; ++I) {
-    const GlobalValue *GV = *I;
-    const DSNodeHandle &SrcGNH = Src->getNodeForValue(GV);
-    DSNodeHandle &DestGNH = NodeMap[SrcGNH.getNode()];
-    assert(DestGNH.getNode() == NH.getNode() &&"Global mapping inconsistent");
-    Dest->getNodeForValue(GV).mergeWith(DSNodeHandle(DestGNH.getNode(),
-                                        DestGNH.getOffset()+SrcGNH.getOffset()));
-  }
-  NH.getNode()->mergeGlobals(*SN);
-
-  return DSNodeHandle(NH.getNode(), NH.getOffset()+SrcNH.getOffset());
-}
-
-void ReachabilityCloner::merge(const DSNodeHandle &NH,
-                               const DSNodeHandle &SrcNH) {
-  if (SrcNH.isNull()) return;  // Noop
-  if (NH.isNull()) {
-    // If there is no destination node, just clone the source and assign the
-    // destination node to be it.
-    NH.mergeWith(getClonedNH(SrcNH));
-    return;
-  }
-
-  // Okay, at this point, we know that we have both a destination and a source
-  // node that need to be merged.  Check to see if the source node has already
-  // been cloned.
-  const DSNode *SN = SrcNH.getNode();
-  DSNodeHandle &SCNH = NodeMap[SN];  // SourceClonedNodeHandle
-  if (!SCNH.isNull()) {   // Node already cloned?
-    DSNode *SCNHN = SCNH.getNode();
-    NH.mergeWith(DSNodeHandle(SCNHN,
-                              SCNH.getOffset()+SrcNH.getOffset()));
-    return;  // Nothing to do!
-  }
-
-  // Okay, so the source node has not already been cloned.  Instead of creating
-  // a new DSNode, only to merge it into the one we already have, try to perform
-  // the merge in-place.  The only case we cannot handle here is when the offset
-  // into the existing node is less than the offset into the virtual node we are
-  // merging in.  In this case, we have to extend the existing node, which
-  // requires an allocation anyway.
-  DSNode *DN = NH.getNode();   // Make sure the Offset is up-to-date
-  if (NH.getOffset() >= SrcNH.getOffset()) {
-    if (!DN->isNodeCompletelyFolded()) {
-      // Make sure the destination node is folded if the source node is folded.
-      if (SN->isNodeCompletelyFolded()) {
-        DN->foldNodeCompletely();
-        DN = NH.getNode();
-      } else if (SN->getSize() != DN->getSize()) {
-        // If the two nodes are of different size, and the smaller node has the
-        // array bit set, collapse!
-#if COLLAPSE_ARRAYS_AGGRESSIVELY
-        if (SN->getSize() < DN->getSize()) {
-          if (SN->isArray()) {
-            DN->foldNodeCompletely();
-            DN = NH.getNode();
-          }
-        } else if (DN->isArray()) {
-          DN->foldNodeCompletely();
-          DN = NH.getNode();
-        }
-#endif
-      }
-      if (DN->getSize() < SN->getSize())
-        DN->growSize(SN->getSize());
-
-      // Merge the type entries of the two nodes together...
-      if (!DN->isNodeCompletelyFolded())
-          DN->mergeTypeInfo(SN, NH.getOffset() - SrcNH.getOffset());
-
-    }
-
-    assert(!DN->isDeadNode());
-
-    // Merge the NodeType information.
-    DN->mergeNodeFlags(SN->getNodeFlags() & BitsToKeep);
-
-    // Before we start merging outgoing links and updating the scalar map, make
-    // sure it is known that this is the representative node for the src node.
-    SCNH = DSNodeHandle(DN, NH.getOffset()-SrcNH.getOffset());
-
-    // If the source node contains any globals, make sure they end up in the
-    // scalar map with the correct offset.
-    if (SN->globals_begin() != SN->globals_end()) {
-      // Update the globals in the destination node itself.
-      DN->mergeGlobals(*SN);
-
-      // Update the scalar map for the graph we are merging the source node
-      // into.
-      for (DSNode::globals_iterator I = SN->globals_begin(),
-             E = SN->globals_end(); I != E; ++I) {
-        const GlobalValue *GV = *I;
-        const DSNodeHandle &SrcGNH = Src->getNodeForValue(GV);
-        DSNodeHandle &DestGNH = NodeMap[SrcGNH.getNode()];
-        assert(DestGNH.getNode()==NH.getNode() &&"Global mapping inconsistent");
-        Dest->getNodeForValue(GV).mergeWith(DSNodeHandle(DestGNH.getNode(),
-                                      DestGNH.getOffset()+SrcGNH.getOffset()));
-      }
-      NH.getNode()->mergeGlobals(*SN);
-    }
-  } else {
-    // We cannot handle this case without allocating a temporary node.  Fall
-    // back on being simple.
-    DSNode *NewDN = new DSNode(*SN, Dest, true /* Null out all links */);
-    NewDN->maskNodeTypes(BitsToKeep);
-
-    unsigned NHOffset = NH.getOffset();
-    NH.mergeWith(DSNodeHandle(NewDN, SrcNH.getOffset()));
-
-    assert(NH.getNode() &&
-           (NH.getOffset() > NHOffset ||
-            (NH.getOffset() == 0 && NH.getNode()->isNodeCompletelyFolded())) &&
-           "Merging did not adjust the offset!");
-
-    // Before we start merging outgoing links and updating the scalar map, make
-    // sure it is known that this is the representative node for the src node.
-    SCNH = DSNodeHandle(NH.getNode(), NH.getOffset()-SrcNH.getOffset());
-
-    // If the source node contained any globals, make sure to create entries
-    // in the scalar map for them!
-    for (DSNode::globals_iterator I = SN->globals_begin(),
-           E = SN->globals_end(); I != E; ++I) {
-      const GlobalValue *GV = *I;
-      const DSNodeHandle &SrcGNH = Src->getNodeForValue(GV);
-      DSNodeHandle &DestGNH = NodeMap[SrcGNH.getNode()];
-      assert(DestGNH.getNode()==NH.getNode() &&"Global mapping inconsistent");
-      assert(SrcGNH.getNode() == SN && "Global mapping inconsistent");
-      Dest->getNodeForValue(GV).mergeWith(DSNodeHandle(DestGNH.getNode(),
-                                    DestGNH.getOffset()+SrcGNH.getOffset()));
-    }
-  }
-
-  //  DOUT << "LLVA: mergeWith: " << SN << " becomes " << DN << "\n";
-
-  // Next, recursively merge all outgoing links as necessary.  Note that
-  // adding these links can cause the destination node to collapse itself at
-  // any time, and the current node may be merged with arbitrary other nodes.
-  // For this reason, we must always go through NH.
-  DN = 0;
-  for (DSNode::const_edge_iterator ii = SN->edge_begin(), ee = SN->edge_end();
-       ii != ee; ++ii) {
-    const DSNodeHandle &SrcEdge = ii->second;
-    if (!SrcEdge.isNull()) {
-      // Compute the offset into the current node at which to
-      // merge this link.  In the common case, this is a linear
-      // relation to the offset in the original node (with
-      // wrapping), but if the current node gets collapsed due to
-      // recursive merging, we must make sure to merge in all remaining
-      // links at offset zero.
-      DSNode *CN = SCNH.getNode();
-      unsigned MergeOffset = (ii->first+SCNH.getOffset()) % CN->getSize();
-
-      DSNodeHandle Tmp = CN->getLink(MergeOffset);
-      if (!Tmp.isNull()) {
-        // Perform the recursive merging.  Make sure to create a temporary NH,
-        // because the Link can disappear in the process of recursive merging.
-        merge(Tmp, SrcEdge);
-      } else {
-        Tmp.mergeWith(getClonedNH(SrcEdge));
-        // Merging this could cause all kinds of recursive things to happen,
-        // culminating in the current node being eliminated.  Since this is
-        // possible, make sure to reaquire the link from 'CN'.
-
-        unsigned MergeOffset = 0;
-        CN = SCNH.getNode();
-        MergeOffset = (ii->first + SCNH.getOffset()) % CN->getSize();
-        CN->getLink(MergeOffset).mergeWith(Tmp);
-      }
-    }
-  }
 }
 
-/// mergeCallSite - Merge the nodes reachable from the specified src call
-/// site into the nodes reachable from DestCS.
-void ReachabilityCloner::mergeCallSite(DSCallSite &DestCS,
-                                       const DSCallSite &SrcCS) {
-  merge(DestCS.getRetVal(), SrcCS.getRetVal());
-  unsigned MinArgs = DestCS.getNumPtrArgs();
-  if (SrcCS.getNumPtrArgs() < MinArgs) MinArgs = SrcCS.getNumPtrArgs();
-
-  for (unsigned a = 0; a != MinArgs; ++a)
-    merge(DestCS.getPtrArg(a), SrcCS.getPtrArg(a));
-
-  for (unsigned a = MinArgs, e = SrcCS.getNumPtrArgs(); a != e; ++a)
-    DestCS.addPtrArg(getClonedNH(SrcCS.getPtrArg(a)));
-}
-
-DSCallSite ReachabilityCloner::cloneCallSite(const DSCallSite& SrcCS) {
-  std::vector<DSNodeHandle> Args;
-  for(unsigned x = 0; x < SrcCS.getNumPtrArgs(); ++x)
-    Args.push_back(getClonedNH(SrcCS.getPtrArg(x)));
-  if (SrcCS.isDirectCall())
-    return DSCallSite(SrcCS.getCallSite(),
-                      getClonedNH(SrcCS.getRetVal()),
-                      SrcCS.getCalleeFunc(),
-                      Args);
-  else
-    return DSCallSite(SrcCS.getCallSite(),
-                      getClonedNH(SrcCS.getRetVal()),
-                      getClonedNH(SrcCS.getCalleeNode()).getNode(),
-                      Args);
-}
-
-//===----------------------------------------------------------------------===//
-// DSCallSite Implementation
-//===----------------------------------------------------------------------===//
-
-// Define here to avoid including iOther.h and BasicBlock.h in DSGraph.h
-const Function &DSCallSite::getCaller() const {
-  return *Site.getInstruction()->getParent()->getParent();
-}
-
-void DSCallSite::InitNH(DSNodeHandle &NH, const DSNodeHandle &Src,
-                        ReachabilityCloner &RC) {
-  NH = RC.getClonedNH(Src);
-}
-
-//===----------------------------------------------------------------------===//
-// DSGraph Implementation
-//===----------------------------------------------------------------------===//
-
 /// getFunctionNames - Return a space separated list of the name of the
 /// functions in this graph (if any)
 std::string DSGraph::getFunctionNames() const {
@@ -840,22 +87,6 @@
 // dump - Allow inspection of graph in a debugger.
 void DSGraph::dump() const { print(errs()); }
 
-
-/// remapLinks - Change all of the Links in the current node according to the
-/// specified mapping.
-///
-void DSNode::remapLinks(DSGraph::NodeMapTy &OldNodeMap) {
-  for (LinkMapTy::iterator ii = edge_begin(), ee = edge_end();
-       ii != ee; ++ii)
-    if (DSNode *N = ii->second.getNode()) {
-      DSGraph::NodeMapTy::const_iterator ONMI = OldNodeMap.find(N);
-      if (ONMI != OldNodeMap.end()) {
-        DSNode *ONMIN = ONMI->second.getNode();
-        ii->second.setTo(ONMIN, ii->second.getOffset()+ONMI->second.getOffset());
-      }
-    }
-}
-
 void DSGraph::removeFunctionCalls(Function& F) {
   for (std::list<DSCallSite>::iterator I = FunctionCalls.begin(),
          E = FunctionCalls.end(); I != E; ++I)
@@ -999,23 +230,6 @@
   ScalarMap.spliceFrom(RHS->ScalarMap);
 }
 
-/// spliceFrom - Copy all entries from RHS, then clear RHS.
-///
-void DSScalarMap::spliceFrom(DSScalarMap &RHS) {
-  // Special case if this is empty.
-  if (ValueMap.empty()) {
-    ValueMap.swap(RHS.ValueMap);
-    GlobalSet.swap(RHS.GlobalSet);
-  } else {
-    GlobalSet.insert(RHS.GlobalSet.begin(), RHS.GlobalSet.end());
-    for (ValueMapTy::iterator I = RHS.ValueMap.begin(), E = RHS.ValueMap.end();
-         I != E; ++I)
-      ValueMap[I->first].mergeWith(I->second);
-    RHS.ValueMap.clear();
-  }
-}
-
-
 /// getFunctionArgumentsForCall - Given a function that is currently in this
 /// graph, return the DSNodeHandles that correspond to the pointer-compatible
 /// function arguments.  The vector is filled in with the return value (or
@@ -1381,7 +595,7 @@
     if (N->getNumReferrers() == 1)  // Does it point to a lonely node?
       // No interesting info?
       if ((N->getNodeFlags() & ~DSNode::IncompleteNode) == 0
-          && N->type_begin() == N->type_end()
+          && N->hasNoType()
           && !N->isNodeCompletelyFolded())
         Edge.setTo(0, 0);  // Kill the edge!
 }
@@ -1549,11 +763,13 @@
 
   // Loop over all of the nodes in the graph, calling getNode on each field.
   // This will cause all nodes to update their forwarding edges, causing
-  // forwarded nodes to be delete-able.
+  // forwarded nodes to be delete-able.  Further, reclaim any memory used by
+  // useless edge or type entries
   for (node_iterator NI = node_begin(), E = node_end(); NI != E; ++NI)
     for (DSNode::edge_iterator ii = NI->edge_begin(), ee = NI->edge_end();
-         ii != ee; ++ii)
-      ii->second.getNode();
+         ii != ee; ++ii) {
+      ii->second.getNode()->cleanEdges();
+    }
 
   // Likewise, forward any edges from the scalar nodes.  While we are at it,
   // clean house a bit.
@@ -1616,28 +832,6 @@
   removeIdenticalCalls(AuxFunctionCalls);
 }
 
-
-/// markReachableNodes - This method recursively traverses the specified
-/// DSNodes, marking any nodes which are reachable.  All reachable nodes it adds
-/// to the set, which allows it to only traverse visited nodes once.
-///
-void DSNode::markReachableNodes(DenseSet<const DSNode*> &ReachableNodes) const {
-  if (this == 0) return;
-  assert(!isForwarding() && "Cannot mark a forwarded node!");
-  if (ReachableNodes.insert(this).second) // Is newly reachable?
-    for (DSNode::const_edge_iterator I = edge_begin(), E = edge_end();
-         I != E; ++I)
-      I->second.getNode()->markReachableNodes(ReachableNodes);
-}
-
-void DSCallSite::markReachableNodes(DenseSet<const DSNode*> &Nodes) const {
-  getRetVal().getNode()->markReachableNodes(Nodes);
-  if (isIndirectCall()) getCalleeNode()->markReachableNodes(Nodes);
-
-  for (unsigned i = 0, e = getNumPtrArgs(); i != e; ++i)
-    getPtrArg(i).getNode()->markReachableNodes(Nodes);
-}
-
 // CanReachAliveNodes - Simple graph walker that recursively traverses the graph
 // looking for a node that is marked alive.  If an alive node is found, return
 // true, otherwise return false.  If an alive node is reachable, this node is
@@ -2023,248 +1217,3 @@
 }
 
 
-////////////////////////////////////////////////////////////////////////////////
-//Base DataStructures impl:
-////////////////////////////////////////////////////////////////////////////////
-
-static const Function *getFnForValue(const Value *V) {
-  if (const Instruction *I = dyn_cast<Instruction>(V))
-    return I->getParent()->getParent();
-  else if (const Argument *A = dyn_cast<Argument>(V))
-    return A->getParent();
-  else if (const BasicBlock *BB = dyn_cast<BasicBlock>(V))
-    return BB->getParent();
-  return 0;
-}
-
-/// deleteValue/copyValue - Interfaces to update the DSGraphs in the program.
-/// These correspond to the interfaces defined in the AliasAnalysis class.
-void DataStructures::deleteValue(Value *V) {
-  if (const Function *F = getFnForValue(V)) {  // Function local value?
-    // If this is a function local value, just delete it from the scalar map!
-    getDSGraph(*F)->getScalarMap().eraseIfExists(V);
-    return;
-  }
-  
-  if (Function *F = dyn_cast<Function>(V)) {
-    assert(getDSGraph(*F)->getReturnNodes().size() == 1 &&
-           "cannot handle scc's");
-    delete DSInfo[F];
-    DSInfo.erase(F);
-    return;
-  }
-  
-  assert(!isa<GlobalVariable>(V) && "Do not know how to delete GV's yet!");
-}
-
-void DataStructures::copyValue(Value *From, Value *To) {
-  if (From == To) return;
-  if (const Function *F = getFnForValue(From)) {  // Function local value?
-    // If this is a function local value, just delete it from the scalar map!
-    getDSGraph(*F)->getScalarMap().copyScalarIfExists(From, To);
-    return;
-  }
-  
-  if (Function *FromF = dyn_cast<Function>(From)) {
-    Function *ToF = cast<Function>(To);
-    assert(!DSInfo.count(ToF) && "New Function already exists!");
-    DSGraph *NG = new DSGraph(getDSGraph(*FromF), GlobalECs, *TypeSS);
-    DSInfo[ToF] = NG;
-    assert(NG->getReturnNodes().size() == 1 && "Cannot copy SCC's yet!");
-    
-    // Change the Function* is the returnnodes map to the ToF.
-    DSNodeHandle Ret = NG->retnodes_begin()->second;
-    NG->getReturnNodes().clear();
-    NG->getReturnNodes()[ToF] = Ret;
-    return;
-  }
-  
-  if (const Function *F = getFnForValue(To)) {
-    getDSGraph(*F)->getScalarMap().copyScalarIfExists(From, To);
-    return;
-  }
-  
-  errs() << *From;
-  errs() << *To;
-  assert(0 && "Do not know how to copy this yet!");
-  abort();
-}
-
-DSGraph* DataStructures::getOrCreateGraph(const Function* F) {
-  assert(F && "No function");
-  DSGraph *&G = DSInfo[F];
-  if (!G) {
-    //Clone or Steal the Source Graph
-    DSGraph* BaseGraph = GraphSource->getDSGraph(*F);
-    if (Clone) {
-      G = new DSGraph(BaseGraph, GlobalECs, *TypeSS, DSGraph::DontCloneAuxCallNodes);
-    } else {
-      G = new DSGraph(GlobalECs, GraphSource->getTargetData(), *TypeSS);
-      G->spliceFrom(BaseGraph);
-      if (resetAuxCalls) 
-        G->getAuxFunctionCalls() = G->getFunctionCalls();
-    }
-    G->setPrintAuxCalls();
-    G->setGlobalsGraph(GlobalsGraph);
-    
-    // Note that this graph is the graph for ALL of the function in the SCC, not
-    // just F.
-    for (DSGraph::retnodes_iterator RI = G->retnodes_begin(),
-           E = G->retnodes_end(); RI != E; ++RI)
-      if (RI->first != F)
-        DSInfo[RI->first] = G;
-  }
-  return G;
-}
-
-
-void DataStructures::formGlobalECs() {
-  // Grow the equivalence classes for the globals to include anything that we
-  // now know to be aliased.
-  sv::set<const GlobalValue*> ECGlobals;
-  buildGlobalECs(ECGlobals);
-  if (!ECGlobals.empty()) {
-    DEBUG(errs() << "Eliminating " << ECGlobals.size() << " EC Globals!\n");
-    for (DSInfoTy::iterator I = DSInfo.begin(),
-           E = DSInfo.end(); I != E; ++I)
-      eliminateUsesOfECGlobals(*I->second, ECGlobals);
-  }
-}
-
-/// BuildGlobalECs - Look at all of the nodes in the globals graph.  If any node
-/// contains multiple globals, DSA will never, ever, be able to tell the globals
-/// apart.  Instead of maintaining this information in all of the graphs
-/// throughout the entire program, store only a single global (the "leader") in
-/// the graphs, and build equivalence classes for the rest of the globals.
-void DataStructures::buildGlobalECs(sv::set<const GlobalValue*> &ECGlobals) {
-  DSScalarMap &SM = GlobalsGraph->getScalarMap();
-  EquivalenceClasses<const GlobalValue*> &GlobalECs = SM.getGlobalECs();
-  for (DSGraph::node_iterator I = GlobalsGraph->node_begin(), 
-         E = GlobalsGraph->node_end();
-       I != E; ++I) {
-    if (I->numGlobals() <= 1) continue;
-
-    // First, build up the equivalence set for this block of globals.
-    DSNode::globals_iterator i = I->globals_begin(); 
-    const GlobalValue *First = *i;
-    if (GlobalECs.findValue(*i) != GlobalECs.end())
-      First = GlobalECs.getLeaderValue(*i);
-    if (*i == First) ++i;
-    for( ; i != I->globals_end(); ++i) {
-      GlobalECs.unionSets(First, *i);
-      ECGlobals.insert(*i);
-      if (SM.find(*i) != SM.end())
-        SM.erase(SM.find(*i));
-      else
-        errs() << "Global missing in scalar map " << (*i)->getName() << "\n";
-    }
-
-    // Next, get the leader element.
-    assert(First == GlobalECs.getLeaderValue(First) &&
-           "First did not end up being the leader?");
-
-    // Finally, change the global node to only contain the leader.
-    I->clearGlobals();
-    I->addGlobal(First);
-  }
-
-  DEBUG(GlobalsGraph->AssertGraphOK());
-}
-
-/// EliminateUsesOfECGlobals - Once we have determined that some globals are in
-/// really just equivalent to some other globals, remove the globals from the
-/// specified DSGraph (if present), and merge any nodes with their leader nodes.
-void DataStructures::eliminateUsesOfECGlobals(DSGraph &G,
-                                              const sv::set<const GlobalValue*> &ECGlobals) {
-  DSScalarMap &SM = G.getScalarMap();
-  EquivalenceClasses<const GlobalValue*> &GlobalECs = SM.getGlobalECs();
-
-  bool MadeChange = false;
-  std::vector<const GlobalValue*> SMGVV(SM.global_begin(), SM.global_end());
-
-  for (std::vector<const GlobalValue*>::iterator GI = SMGVV.begin(),
-       E = SMGVV.end(); GI != E; ) {
-    const GlobalValue *GV = *GI; ++GI;
-    if (!ECGlobals.count(GV)) continue;
-
-    const DSNodeHandle &GVNH = SM[GV];
-    assert(!GVNH.isNull() && "Global has null NH!?");
-
-    // Okay, this global is in some equivalence class.  Start by finding the
-    // leader of the class.
-    const GlobalValue *Leader = GlobalECs.getLeaderValue(GV);
-
-    // If the leader isn't already in the graph, insert it into the node
-    // corresponding to GV.
-    if (!SM.global_count(Leader)) {
-      GVNH.getNode()->addGlobal(Leader);
-      SM[Leader] = GVNH;
-    } else {
-      // Otherwise, the leader is in the graph, make sure the nodes are the
-      // merged in the specified graph.
-      const DSNodeHandle &LNH = SM[Leader];
-      if (LNH.getNode() != GVNH.getNode())
-        LNH.mergeWith(GVNH);
-    }
-
-    // Next step, remove the global from the DSNode.
-    GVNH.getNode()->removeGlobal(GV);
-
-    // Finally, remove the global from the ScalarMap.
-    SM.erase(GV);
-    MadeChange = true;
-  }
-
-  DEBUG(if(MadeChange) G.AssertGraphOK());
-}
-
-void DataStructures::init(DataStructures* D, bool clone, bool printAuxCalls, 
-                          bool copyGlobalAuxCalls, bool resetAux) {
-  assert (!GraphSource && "Already init");
-  GraphSource = D;
-  Clone = clone;
-  resetAuxCalls = resetAux;
-  TD = D->TD;
-  TypeSS = D->TypeSS;
-  ActualCallees = D->ActualCallees;
-  GlobalECs = D->getGlobalECs();
-  GlobalsGraph = new DSGraph(D->getGlobalsGraph(), GlobalECs, *TypeSS,
-                             copyGlobalAuxCalls?0:DSGraph::DontCloneAuxCallNodes);
-  if (printAuxCalls) GlobalsGraph->setPrintAuxCalls();
-
-  //
-  // Tell the other DSA pass if we're stealing its graph.
-  //
-  if (!clone) D->DSGraphsStolen = true;
-}
-
-void DataStructures::init(TargetData* T) {
-  assert (!TD && "Already init");
-  GraphSource = 0;
-  Clone = false;
-  TD = T;
-  TypeSS = new SuperSet<const Type*>();
-  GlobalsGraph = new DSGraph(GlobalECs, *T, *TypeSS);
-}
-
-void DataStructures::releaseMemory() {
-  //
-  // If the DSGraphs were stolen by another pass, free nothing.
-  //
-  if (DSGraphsStolen) return;
-
-  std::set<DSGraph*> toDelete;
-  for (DSInfoTy::iterator I = DSInfo.begin(), E = DSInfo.end(); I != E; ++I) {
-    I->second->getReturnNodes().clear();
-    toDelete.insert(I->second);
-  }
-  for (std::set<DSGraph*>::iterator I = toDelete.begin(), E = toDelete.end(); I != E; ++I)
-    delete *I;
-
-  // Empty map so next time memory is released, data structures are not
-  // re-deleted.
-  DSInfo.clear();
-  ActualCallees.clear();
-  delete GlobalsGraph;
-  GlobalsGraph = 0;
-}

Modified: poolalloc/trunk/lib/DSA/DataStructure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DataStructure.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DataStructure.cpp Thu Mar 11 18:16:41 2010
@@ -39,15 +39,8 @@
 #define COLLAPSE_ARRAYS_AGGRESSIVELY 0
 namespace {
   STATISTIC (NumFolds, "Number of nodes completely folded");
-  STATISTIC (NumCallNodesMerged, "Number of call nodes merged");
   STATISTIC (NumNodeAllocated  , "Number of nodes allocated");
-  STATISTIC (NumDNE            , "Number of nodes removed by reachability");
-  STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
-  STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
   static cl::opt<unsigned>
-  DSAFieldLimit("dsa-field-limit", cl::Hidden,
-                cl::desc("Number of fields to track before collapsing a node"),
-                cl::init(256));
 }
 
 /// isForwarding - Return true if this NodeHandle is forwarding to another
@@ -115,6 +108,22 @@
   return ValueMap.insert(std::make_pair(GV, DSNodeHandle())).first->second;
 }
 
+/// spliceFrom - Copy all entries from RHS, then clear RHS.
+///
+void DSScalarMap::spliceFrom(DSScalarMap &RHS) {
+  // Special case if this is empty.
+  if (ValueMap.empty()) {
+    ValueMap.swap(RHS.ValueMap);
+    GlobalSet.swap(RHS.GlobalSet);
+  } else {
+    GlobalSet.insert(RHS.GlobalSet.begin(), RHS.GlobalSet.end());
+    for (ValueMapTy::iterator I = RHS.ValueMap.begin(), E = RHS.ValueMap.end();
+         I != E; ++I)
+      ValueMap[I->first].mergeWith(I->second);
+    RHS.ValueMap.clear();
+  }
+}
+
 //===----------------------------------------------------------------------===//
 // DSNode Implementation
 //===----------------------------------------------------------------------===//
@@ -315,6 +324,7 @@
   if (Offset >= getSize()) growSize(Offset+1);
 
   TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(TyMap[Offset], NewTy);
+  assert(TyMap[Offset]);
 }
 
 void DSNode::mergeTypeInfo(const TyMapTy::mapped_type TyIt, unsigned Offset) {
@@ -331,6 +341,7 @@
     S.insert(TyIt->begin(), TyIt->end());
     TyMap[Offset] = getParentGraph()->getTypeSS().getOrCreate(S);
   }
+  assert(TyMap[Offset]);
 }
 
 void DSNode::mergeTypeInfo(const DSNode* DN, unsigned Offset) {
@@ -510,6 +521,29 @@
     DSNode::MergeNodes(NHCopy, CurNodeH);
 }
 
+void DSNode::cleanEdges() {
+  //get rid of any type edge pointing to the null type
+  for (type_iterator ii = type_begin(); ii != type_end(); ) {
+    if (ii->second)
+      ++ii;
+    else {
+      type_iterator backup = ii;
+      ++backup;
+      TyMap.erase(ii);
+      ii = backup;
+    }
+  }
+  //get rid of any node edge pointing to nothing
+  for (edge_iterator ii = edge_begin(); ii != edge_end(); ) {
+    if (ii->second.isNull()) {
+      edge_iterator backup = ii;
+      ++backup;
+      Links.erase(ii);
+      ii = backup;
+    } else
+      ++ii;
+  }
+}
 
 //===----------------------------------------------------------------------===//
 // ReachabilityCloner Implementation
@@ -794,53 +828,6 @@
   NH = RC.getClonedNH(Src);
 }
 
-//===----------------------------------------------------------------------===//
-// DSGraph Implementation
-//===----------------------------------------------------------------------===//
-
-/// getFunctionNames - Return a space separated list of the name of the
-/// functions in this graph (if any)
-std::string DSGraph::getFunctionNames() const {
-  switch (getReturnNodes().size()) {
-  case 0: return "Globals graph";
-  case 1: return retnodes_begin()->first->getName();
-  default:
-    std::string Return;
-    for (DSGraph::retnodes_iterator I = retnodes_begin();
-         I != retnodes_end(); ++I)
-      Return += I->first->getNameStr() + " ";
-    Return.erase(Return.end()-1, Return.end());   // Remove last space character
-    return Return;
-  }
-}
-
-
-DSGraph::DSGraph(DSGraph* G, EquivalenceClasses<const GlobalValue*> &ECs,
-                 SuperSet<const Type*>& tss,
-                 unsigned CloneFlags)
-  : GlobalsGraph(0), ScalarMap(ECs), TD(G->TD), TypeSS(tss) {
-  PrintAuxCalls = false;
-  cloneInto(G, CloneFlags);
-}
-
-DSGraph::~DSGraph() {
-  FunctionCalls.clear();
-  AuxFunctionCalls.clear();
-  ScalarMap.clear();
-  ReturnNodes.clear();
-
-  // Drop all intra-node references, so that assertions don't fail...
-  for (node_iterator NI = node_begin(), E = node_end(); NI != E; ++NI)
-    NI->dropAllReferences();
-
-  // Free all of the nodes.
-  Nodes.clear();
-}
-
-// dump - Allow inspection of graph in a debugger.
-void DSGraph::dump() const { print(errs()); }
-
-
 /// remapLinks - Change all of the Links in the current node according to the
 /// specified mapping.
 ///
@@ -856,182 +843,6 @@
     }
 }
 
-void DSGraph::removeFunctionCalls(Function& F) {
-  for (std::list<DSCallSite>::iterator I = FunctionCalls.begin(),
-         E = FunctionCalls.end(); I != E; ++I)
-    if (I->isDirectCall() && I->getCalleeFunc() == &F) {
-      FunctionCalls.erase(I);
-      break;
-    }
-
-  for (std::list<DSCallSite>::iterator I = AuxFunctionCalls.begin(),
-         E = AuxFunctionCalls.end(); I != E; ++I)
-    if (I->isDirectCall() && I->getCalleeFunc() == &F) {
-      AuxFunctionCalls.erase(I);
-      break;
-    }
-}
-
-/// addObjectToGraph - This method can be used to add global, stack, and heap
-/// objects to the graph.  This can be used when updating DSGraphs due to the
-/// introduction of new temporary objects.  The new object is not pointed to
-/// and does not point to any other objects in the graph.
-DSNode *DSGraph::addObjectToGraph(Value *Ptr, bool UseDeclaredType) {
-  assert(isa<PointerType>(Ptr->getType()) && "Ptr is not a pointer!");
-  DSNode *N = new DSNode(this);
-  assert(ScalarMap[Ptr].isNull() && "Object already in this graph!");
-  ScalarMap[Ptr] = N;
-
-  if (GlobalValue *GV = dyn_cast<GlobalValue>(Ptr)) {
-    N->addGlobal(GV);
-//  } else if (isa<MallocInst>(Ptr)) {
-//   N->setHeapMarker();
-  } else if (isa<AllocaInst>(Ptr)) {
-    N->setAllocaMarker();
-  } else {
-    assert(0 && "Illegal memory object input!");
-  }
-  return N;
-}
-
-
-/// cloneInto - Clone the specified DSGraph into the current graph.  The
-/// translated ScalarMap for the old function is filled into the ScalarMap
-/// for the graph, and the translated ReturnNodes map is returned into
-/// ReturnNodes.
-///
-/// The CloneFlags member controls various aspects of the cloning process.
-///
-void DSGraph::cloneInto( DSGraph* G, unsigned CloneFlags) {
-  assert(G != this && "Cannot clone graph into itself!");
-
-  NodeMapTy OldNodeMap;
-
-  // Remove alloca or mod/ref bits as specified...
-  unsigned BitsToClear = ((CloneFlags & StripAllocaBit)? DSNode::AllocaNode : 0)
-    | ((CloneFlags & StripModRefBits)? (DSNode::ModifiedNode | DSNode::ReadNode) : 0)
-    | ((CloneFlags & StripIncompleteBit)? DSNode::IncompleteNode : 0);
-  BitsToClear |= DSNode::DeadNode;  // Clear dead flag...
-
-  for (node_const_iterator I = G->node_begin(), E = G->node_end(); I != E; ++I) {
-    assert(!I->isForwarding() &&
-           "Forward nodes shouldn't be in node list!");
-    DSNode *New = new DSNode(*I, this);
-    New->maskNodeTypes(~BitsToClear);
-    OldNodeMap[I] = New;
-  }
-
-#ifndef NDEBUG
-  Timer::addPeakMemoryMeasurement();
-#endif
-
-  // Rewrite the links in the new nodes to point into the current graph now.
-  // Note that we don't loop over the node's list to do this.  The problem is
-  // that remaping links can cause recursive merging to happen, which means
-  // that node_iterator's can get easily invalidated!  Because of this, we
-  // loop over the OldNodeMap, which contains all of the new nodes as the
-  // .second element of the map elements.  Also note that if we remap a node
-  // more than once, we won't break anything.
-  for (NodeMapTy::iterator I = OldNodeMap.begin(), E = OldNodeMap.end();
-       I != E; ++I)
-    I->second.getNode()->remapLinks(OldNodeMap);
-
-  // Copy the scalar map... merging all of the global nodes...
-  for (DSScalarMap::const_iterator I = G->ScalarMap.begin(),
-         E = G->ScalarMap.end(); I != E; ++I) {
-    DSNodeHandle &MappedNode = OldNodeMap[I->second.getNode()];
-    DSNodeHandle &H = ScalarMap.getRawEntryRef(I->first);
-    DSNode *MappedNodeN = MappedNode.getNode();
-    H.mergeWith(DSNodeHandle(MappedNodeN,
-                             I->second.getOffset()+MappedNode.getOffset()));
-  }
-
-  if (!(CloneFlags & DontCloneCallNodes)) {
-    // Copy the function calls list.
-    for (fc_iterator I = G->fc_begin(), E = G->fc_end(); I != E; ++I)
-      FunctionCalls.push_back(DSCallSite(*I, OldNodeMap));
-  }
-
-  if (!(CloneFlags & DontCloneAuxCallNodes)) {
-    // Copy the auxiliary function calls list.
-    for (afc_iterator I = G->afc_begin(), E = G->afc_end(); I != E; ++I)
-      AuxFunctionCalls.push_back(DSCallSite(*I, OldNodeMap));
-  }
-
-  // Map the return node pointers over...
-  for (retnodes_iterator I = G->retnodes_begin(),
-         E = G->retnodes_end(); I != E; ++I) {
-    const DSNodeHandle &Ret = I->second;
-    DSNodeHandle &MappedRet = OldNodeMap[Ret.getNode()];
-    DSNode *MappedRetN = MappedRet.getNode();
-    ReturnNodes.insert(std::make_pair(I->first,
-                                      DSNodeHandle(MappedRetN,
-                                     MappedRet.getOffset()+Ret.getOffset())));
-  }
-}
-
-/// spliceFrom - Logically perform the operation of cloning the RHS graph into
-/// this graph, then clearing the RHS graph.  Instead of performing this as
-/// two seperate operations, do it as a single, much faster, one.
-///
-void DSGraph::spliceFrom(DSGraph* RHS) {
-  assert(this != RHS && "Splicing self");
-  // Change all of the nodes in RHS to think we are their parent.
-  for (NodeListTy::iterator I = RHS->Nodes.begin(), E = RHS->Nodes.end();
-       I != E; ++I)
-    I->setParentGraph(this);
-  // Take all of the nodes.
-  Nodes.splice(Nodes.end(), RHS->Nodes);
-
-  // Take all of the calls.
-  FunctionCalls.splice(FunctionCalls.end(), RHS->FunctionCalls);
-  AuxFunctionCalls.splice(AuxFunctionCalls.end(), RHS->AuxFunctionCalls);
-
-  // Take all of the return nodes.
-  if (ReturnNodes.empty()) {
-    ReturnNodes.swap(RHS->ReturnNodes);
-  } else {
-    ReturnNodes.insert(RHS->ReturnNodes.begin(), RHS->ReturnNodes.end());
-    RHS->ReturnNodes.clear();
-  }
-
- // Merge the scalar map in.
-  ScalarMap.spliceFrom(RHS->ScalarMap);
-}
-
-/// spliceFrom - Copy all entries from RHS, then clear RHS.
-///
-void DSScalarMap::spliceFrom(DSScalarMap &RHS) {
-  // Special case if this is empty.
-  if (ValueMap.empty()) {
-    ValueMap.swap(RHS.ValueMap);
-    GlobalSet.swap(RHS.GlobalSet);
-  } else {
-    GlobalSet.insert(RHS.GlobalSet.begin(), RHS.GlobalSet.end());
-    for (ValueMapTy::iterator I = RHS.ValueMap.begin(), E = RHS.ValueMap.end();
-         I != E; ++I)
-      ValueMap[I->first].mergeWith(I->second);
-    RHS.ValueMap.clear();
-  }
-}
-
-
-/// getFunctionArgumentsForCall - Given a function that is currently in this
-/// graph, return the DSNodeHandles that correspond to the pointer-compatible
-/// function arguments.  The vector is filled in with the return value (or
-/// null if it is not pointer compatible), followed by all of the
-/// pointer-compatible arguments.
-void DSGraph::getFunctionArgumentsForCall(const Function *F,
-                                       std::vector<DSNodeHandle> &Args) const {
-  Args.push_back(getReturnNodeFor(*F));
-  for (Function::const_arg_iterator AI = F->arg_begin(), E = F->arg_end();
-       AI != E; ++AI)
-    if (isa<PointerType>(AI->getType())) {
-      Args.push_back(getNodeForValue(AI));
-      assert(!Args.back().isNull() && "Pointer argument w/o scalarmap entry!?");
-    }
-}
-
 namespace {
   // HackedGraphSCCFinder - This is used to find nodes that have a path from the
   // node to a node cloned by the ReachabilityCloner object contained.  To be
@@ -1139,484 +950,6 @@
   return ThisNodeInfo;
 }
 
-/// mergeInCallFromOtherGraph - This graph merges in the minimal number of
-/// nodes from G2 into 'this' graph, merging the bindings specified by the
-/// call site (in this graph) with the bindings specified by the vector in G2.
-/// The two DSGraphs must be different.
-///
-void DSGraph::mergeInGraph(const DSCallSite &CS,
-                           std::vector<DSNodeHandle> &Args,
-                           const DSGraph &Graph, unsigned CloneFlags) {
-  assert((CloneFlags & DontCloneCallNodes) &&
-         "Doesn't support copying of call nodes!");
-
-  // If this is not a recursive call, clone the graph into this graph...
-  if (&Graph == this) {
-    // Merge the return value with the return value of the context.
-    Args[0].mergeWith(CS.getRetVal());
-
-    // Resolve all of the function arguments.
-    for (unsigned i = 0, e = CS.getNumPtrArgs(); i != e; ++i) {
-      if (i == Args.size()-1)
-        break;
-
-      // Add the link from the argument scalar to the provided value.
-      Args[i+1].mergeWith(CS.getPtrArg(i));
-    }
-    return;
-  }
-
-  // Clone the callee's graph into the current graph, keeping track of where
-  // scalars in the old graph _used_ to point, and of the new nodes matching
-  // nodes of the old graph.
-  ReachabilityCloner RC(this, &Graph, CloneFlags);
-
-  // Map the return node pointer over.
-  if (!CS.getRetVal().isNull())
-    RC.merge(CS.getRetVal(), Args[0]);
-
-  // Map over all of the arguments.
-  for (unsigned i = 0, e = CS.getNumPtrArgs(); i != e; ++i) {
-    if (i == Args.size()-1)
-      break;
-
-    // Add the link from the argument scalar to the provided value.
-    RC.merge(CS.getPtrArg(i), Args[i+1]);
-  }
-
-  // We generally don't want to copy global nodes or aux calls from the callee
-  // graph to the caller graph.  However, we have to copy them if there is a
-  // path from the node to a node we have already copied which does not go
-  // through another global.  Compute the set of node that can reach globals and
-  // aux call nodes to copy over, then do it.
-  std::vector<const DSCallSite*> AuxCallToCopy;
-  std::vector<const GlobalValue*> GlobalsToCopy;
-
-  // NodesReachCopiedNodes - Memoize results for efficiency.  Contains a
-  // true/false value for every visited node that reaches a copied node without
-  // going through a global.
-  HackedGraphSCCFinder SCCFinder(RC);
-
-  if (!(CloneFlags & DontCloneAuxCallNodes))
-    for (afc_const_iterator I = Graph.afc_begin(), E = Graph.afc_end(); I!=E; ++I)
-      if (SCCFinder.PathExistsToClonedNode(*I))
-        AuxCallToCopy.push_back(&*I);
-//       else if (I->isIndirectCall()){
-//  	//If the call node doesn't have any callees, clone it
-//  	std::vector< Function *> List;
-//  	I->getCalleeNode()->addFullFunctionList(List);
-//  	if (!List.size())
-//  	  AuxCallToCopy.push_back(&*I);
-//        }
-
-  const DSScalarMap &GSM = Graph.getScalarMap();
-  for (DSScalarMap::global_iterator GI = GSM.global_begin(),
-         E = GSM.global_end(); GI != E; ++GI) {
-    DSNode *GlobalNode = Graph.getNodeForValue(*GI).getNode();
-    for (DSNode::edge_iterator EI = GlobalNode->edge_begin(),
-           EE = GlobalNode->edge_end(); EI != EE; ++EI)
-      if (SCCFinder.PathExistsToClonedNode(EI->second.getNode())) {
-        GlobalsToCopy.push_back(*GI);
-        break;
-      }
-  }
-
-  // Copy aux calls that are needed.
-  for (unsigned i = 0, e = AuxCallToCopy.size(); i != e; ++i)
-    AuxFunctionCalls.push_back(DSCallSite(*AuxCallToCopy[i], RC));
-
-  // Copy globals that are needed.
-  for (unsigned i = 0, e = GlobalsToCopy.size(); i != e; ++i)
-    RC.getClonedNH(Graph.getNodeForValue(GlobalsToCopy[i]));
-}
-
-
-
-/// mergeInGraph - The method is used for merging graphs together.  If the
-/// argument graph is not *this, it makes a clone of the specified graph, then
-/// merges the nodes specified in the call site with the formal arguments in the
-/// graph.
-///
-void DSGraph::mergeInGraph(const DSCallSite &CS, const Function &F,
-                           const DSGraph &Graph, unsigned CloneFlags) {
-  // Set up argument bindings.
-  std::vector<DSNodeHandle> Args;
-  Graph.getFunctionArgumentsForCall(&F, Args);
-
-  mergeInGraph(CS, Args, Graph, CloneFlags);
-}
-
-/// getCallSiteForArguments - Get the arguments and return value bindings for
-/// the specified function in the current graph.
-///
-DSCallSite DSGraph::getCallSiteForArguments(const Function &F) const {
-  std::vector<DSNodeHandle> Args;
-
-  for (Function::const_arg_iterator I = F.arg_begin(), E = F.arg_end(); I != E; ++I)
-    if (isa<PointerType>(I->getType()))
-      Args.push_back(getNodeForValue(I));
-
-  return DSCallSite(CallSite(), getReturnNodeFor(F), &F, Args);
-}
-
-/// getDSCallSiteForCallSite - Given an LLVM CallSite object that is live in
-/// the context of this graph, return the DSCallSite for it.
-DSCallSite DSGraph::getDSCallSiteForCallSite(CallSite CS) const {
-  DSNodeHandle RetVal;
-  Instruction *I = CS.getInstruction();
-  if (isa<PointerType>(I->getType()))
-    RetVal = getNodeForValue(I);
-
-  std::vector<DSNodeHandle> Args;
-  Args.reserve(CS.arg_end()-CS.arg_begin());
-
-  // Calculate the arguments vector...
-  for (CallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end(); I != E; ++I)
-    if (isa<PointerType>((*I)->getType())) {
-      if (isa<ConstantPointerNull>(*I))
-        Args.push_back(DSNodeHandle());
-      else
-        Args.push_back(getNodeForValue(*I));
-    }
-
-  // Add a new function call entry...
-  if (Function *F = CS.getCalledFunction())
-    return DSCallSite(CS, RetVal, F, Args);
-  else
-    return DSCallSite(CS, RetVal,
-                      getNodeForValue(CS.getCalledValue()).getNode(), Args);
-}
-
-
-
-// markIncompleteNodes - Mark the specified node as having contents that are not
-// known with the current analysis we have performed.  Because a node makes all
-// of the nodes it can reach incomplete if the node itself is incomplete, we
-// must recursively traverse the data structure graph, marking all reachable
-// nodes as incomplete.
-//
-static void markIncompleteNode(DSNode *N) {
-  // Stop recursion if no node, or if node already marked...
-  if (N == 0 || N->isIncompleteNode()) return;
-
-  // Actually mark the node
-  N->setIncompleteMarker();
-
-  // Recursively process children...
-  for (DSNode::edge_iterator ii = N->edge_begin(), ee = N->edge_end();
-       ii != ee; ++ii)
-    markIncompleteNode(ii->second.getNode());
-}
-
-static void markIncomplete(DSCallSite &Call) {
-  // Then the return value is certainly incomplete!
-  markIncompleteNode(Call.getRetVal().getNode());
-
-  // All objects pointed to by function arguments are incomplete!
-  for (unsigned i = 0, e = Call.getNumPtrArgs(); i != e; ++i)
-    markIncompleteNode(Call.getPtrArg(i).getNode());
-}
-
-// markIncompleteNodes - Traverse the graph, identifying nodes that may be
-// modified by other functions that have not been resolved yet.  This marks
-// nodes that are reachable through three sources of "unknownness":
-//
-//  Global Variables, Function Calls, and Incoming Arguments
-//
-// For any node that may have unknown components (because something outside the
-// scope of current analysis may have modified it), the 'Incomplete' flag is
-// added to the NodeType.
-//
-void DSGraph::markIncompleteNodes(unsigned Flags) {
-  // Mark any incoming arguments as incomplete.
-  if (Flags & DSGraph::MarkFormalArgs)
-    for (ReturnNodesTy::iterator FI = ReturnNodes.begin(), E =ReturnNodes.end();
-         FI != E; ++FI) {
-      const Function &F = *FI->first;
-      for (Function::const_arg_iterator I = F.arg_begin(), E = F.arg_end();
-           I != E; ++I)
-        if (isa<PointerType>(I->getType()))
-          markIncompleteNode(getNodeForValue(I).getNode());
-      markIncompleteNode(FI->second.getNode());
-    }
-
-  // Mark stuff passed into functions calls as being incomplete.
-  if (!shouldPrintAuxCalls())
-    for (std::list<DSCallSite>::iterator I = FunctionCalls.begin(),
-           E = FunctionCalls.end(); I != E; ++I)
-      markIncomplete(*I);
-  else
-    for (std::list<DSCallSite>::iterator I = AuxFunctionCalls.begin(),
-           E = AuxFunctionCalls.end(); I != E; ++I)
-      markIncomplete(*I);
-
-#if 0
-  // Mark stuff passed into external functions as being incomplete.
-  // External functions may not appear in Aux during td, so process
-  // them specially
-  for (std::list<DSCallSite>::iterator I = FunctionCalls.begin(),
-         E = FunctionCalls.end(); I != E; ++I)
-    if(I->isDirectCall() && I->getCalleeFunc()->isDeclaration())
-      markIncomplete(*I);
-#endif
-
-  // Mark all global nodes as incomplete.
-  for (DSScalarMap::global_iterator I = ScalarMap.global_begin(),
-         E = ScalarMap.global_end(); I != E; ++I)
-    if (const GlobalVariable *GV = dyn_cast<GlobalVariable>(*I))
-      if (!GV->hasInitializer() ||    // Always mark external globals incomp.
-          (!GV->isConstant() && (Flags & DSGraph::IgnoreGlobals) == 0))
-        markIncompleteNode(ScalarMap[GV].getNode());
-
-  // Mark any node with the VAStart flag as incomplete.
-  if (Flags & DSGraph::MarkVAStart) {
-    for (node_iterator i=node_begin(); i != node_end(); ++i) {
-      markIncompleteNode(i);
-    }
-  }
-}
-
-static inline void killIfUselessEdge(DSNodeHandle &Edge) {
-  if (DSNode * N = Edge.getNode()) // Is there an edge?
-    if (N->getNumReferrers() == 1)  // Does it point to a lonely node?
-      // No interesting info?
-      if ((N->getNodeFlags() & ~DSNode::IncompleteNode) == 0
-          && N->type_begin() == N->type_end()
-          && !N->isNodeCompletelyFolded())
-        Edge.setTo(0, 0);  // Kill the edge!
-}
-
-static void removeIdenticalCalls(std::list<DSCallSite> &Calls) {
-  // Remove trivially identical function calls
-  Calls.sort();  // Sort by callee as primary key!
-
-  // Scan the call list cleaning it up as necessary...
-  DSNodeHandle LastCalleeNode;
-#if 0
-  Function *LastCalleeFunc = 0;
-  unsigned NumDuplicateCalls = 0;
-#endif
-  bool LastCalleeContainsExternalFunction = false;
-
-  unsigned NumDeleted = 0;
-  for (std::list<DSCallSite>::iterator I = Calls.begin(), E = Calls.end();
-       I != E;) {
-    DSCallSite &CS = *I;
-    std::list<DSCallSite>::iterator OldIt = I++;
-
-    if (!CS.isIndirectCall()) {
-      LastCalleeNode = 0;
-    } else {
-      DSNode *Callee = CS.getCalleeNode();
-
-      // If the Callee is a useless edge, this must be an unreachable call site,
-      // eliminate it.
-      if (Callee->getNumReferrers() == 1 && Callee->isCompleteNode() &&
-          Callee->isEmptyGlobals()) {  // No useful info?
-        DEBUG(errs() << "WARNING: Useless call site found.\n");
-        Calls.erase(OldIt);
-        ++NumDeleted;
-        continue;
-      }
-
-      // If the last call site in the list has the same callee as this one, and
-      // if the callee contains an external function, it will never be
-      // resolvable, just merge the call sites.
-      if (!LastCalleeNode.isNull() && LastCalleeNode.getNode() == Callee) {
-        LastCalleeContainsExternalFunction = Callee->isExternFuncNode();
-
-        std::list<DSCallSite>::iterator PrevIt = OldIt;
-        --PrevIt;
-        PrevIt->mergeWith(CS);
-
-        // No need to keep this call anymore.
-        Calls.erase(OldIt);
-        ++NumDeleted;
-        continue;
-      } else {
-        LastCalleeNode = Callee;
-      }
-    }
-
-    // If the return value or any arguments point to a void node with no
-    // information at all in it, and the call node is the only node to point
-    // to it, remove the edge to the node (killing the node).
-    //
-    killIfUselessEdge(CS.getRetVal());
-    for (unsigned a = 0, e = CS.getNumPtrArgs(); a != e; ++a)
-      killIfUselessEdge(CS.getPtrArg(a));
-
-#if 0
-    // If this call site calls the same function as the last call site, and if
-    // the function pointer contains an external function, this node will
-    // never be resolved.  Merge the arguments of the call node because no
-    // information will be lost.
-    //
-    if ((CS.isDirectCall()   && CS.getCalleeFunc() == LastCalleeFunc) ||
-        (CS.isIndirectCall() && CS.getCalleeNode() == LastCalleeNode)) {
-      ++NumDuplicateCalls;
-      if (NumDuplicateCalls == 1) {
-        if (LastCalleeNode)
-          LastCalleeContainsExternalFunction =
-            nodeContainsExternalFunction(LastCalleeNode);
-        else
-          LastCalleeContainsExternalFunction = LastCalleeFunc->isExternal();
-      }
-
-      // It is not clear why, but enabling this code makes DSA really
-      // sensitive to node forwarding.  Basically, with this enabled, DSA
-      // performs different number of inlinings based on which nodes are
-      // forwarding or not.  This is clearly a problem, so this code is
-      // disabled until this can be resolved.
-#if 1
-      if (LastCalleeContainsExternalFunction
-#if 0
-          ||
-          // This should be more than enough context sensitivity!
-          // FIXME: Evaluate how many times this is tripped!
-          NumDuplicateCalls > 20
-#endif
-          ) {
-
-        std::list<DSCallSite>::iterator PrevIt = OldIt;
-        --PrevIt;
-        PrevIt->mergeWith(CS);
-
-        // No need to keep this call anymore.
-        Calls.erase(OldIt);
-        ++NumDeleted;
-        continue;
-      }
-#endif
-    } else {
-      if (CS.isDirectCall()) {
-        LastCalleeFunc = CS.getCalleeFunc();
-        LastCalleeNode = 0;
-      } else {
-        LastCalleeNode = CS.getCalleeNode();
-        LastCalleeFunc = 0;
-      }
-      NumDuplicateCalls = 0;
-    }
-#endif
-
-    if (I != Calls.end() && CS == *I) {
-      LastCalleeNode = 0;
-      Calls.erase(OldIt);
-      ++NumDeleted;
-      continue;
-    }
-  }
-
-  // Resort now that we simplified things.
-  Calls.sort();
-
-  // Now that we are in sorted order, eliminate duplicates.
-  std::list<DSCallSite>::iterator CI = Calls.begin(), CE = Calls.end();
-  if (CI != CE)
-    while (1) {
-      std::list<DSCallSite>::iterator OldIt = CI++;
-      if (CI == CE) break;
-
-      // If this call site is now the same as the previous one, we can delete it
-      // as a duplicate.
-      if (*OldIt == *CI) {
-        DEBUG(errs() << "Deleteing " << CI->getCallSite().getInstruction() << "\n");
-        Calls.erase(CI);
-        CI = OldIt;
-        ++NumDeleted;
-      }
-    }
-
-  //Calls.erase(std::unique(Calls.begin(), Calls.end()), Calls.end());
-
-  // Track the number of call nodes merged away...
-  NumCallNodesMerged += NumDeleted;
-
-  if (NumDeleted)
-    DEBUG(errs() << "Merged " << NumDeleted << " call nodes.\n");
-}
-
-
-// removeTriviallyDeadNodes - After the graph has been constructed, this method
-// removes all unreachable nodes that are created because they got merged with
-// other nodes in the graph.  These nodes will all be trivially unreachable, so
-// we don't have to perform any non-trivial analysis here.
-//
-void DSGraph::removeTriviallyDeadNodes() {
-  /// NOTE: This code is disabled.  This slows down DSA on 177.mesa
-  /// substantially!
-
-  // Loop over all of the nodes in the graph, calling getNode on each field.
-  // This will cause all nodes to update their forwarding edges, causing
-  // forwarded nodes to be delete-able.
-  for (node_iterator NI = node_begin(), E = node_end(); NI != E; ++NI)
-    for (DSNode::edge_iterator ii = NI->edge_begin(), ee = NI->edge_end();
-         ii != ee; ++ii)
-      ii->second.getNode();
-
-  // Likewise, forward any edges from the scalar nodes.  While we are at it,
-  // clean house a bit.
-  for (DSScalarMap::iterator I = ScalarMap.begin(), E = ScalarMap.end();
-       I != E; ++I)
-    I->second.getNode();
-
-  bool isGlobalsGraph = !GlobalsGraph;
-
-  for (NodeListTy::iterator NI = Nodes.begin(), E = Nodes.end(); NI != E; ) {
-    DSNode &Node = *NI;
-
-    // Do not remove *any* global nodes in the globals graph.
-    // This is a special case because such nodes may not have I, M, R flags set.
-    if (Node.isGlobalNode() && isGlobalsGraph) {
-      ++NI;
-      continue;
-    }
-
-    if (Node.isCompleteNode() && !Node.isModifiedNode() && !Node.isReadNode()) {
-      // This is a useless node if it has no mod/ref info (checked above),
-      // outgoing edges (which it cannot, as it is not modified in this
-      // context), and it has no incoming edges.  If it is a global node it may
-      // have all of these properties and still have incoming edges, due to the
-      // scalar map, so we check those now.
-      //
-      if (Node.getNumReferrers() == Node.numGlobals()) {
-
-        // Loop through and make sure all of the globals are referring directly
-        // to the node...
-        for (DSNode::globals_iterator j = Node.globals_begin(), e = Node.globals_end();
-             j != e; ++j) {
-          DSNode *N = getNodeForValue(*j).getNode();
-          assert(N == &Node && "ScalarMap doesn't match globals list!");
-        }
-
-        // Make sure NumReferrers still agrees, if so, the node is truly dead.
-        if (Node.getNumReferrers() == Node.numGlobals()) {
-          for (DSNode::globals_iterator j = Node.globals_begin(), e = Node.globals_end();
-               j != e; ++j)
-            if (ScalarMap.find(*j) != ScalarMap.end())
-              ScalarMap.erase(*j);
-          Node.makeNodeDead();
-          ++NumTrivialGlobalDNE;
-        }
-      }
-    }
-
-    if ((Node.getNodeFlags() == 0 && Node.hasNoReferrers())
-        || (isGlobalsGraph && Node.hasNoReferrers() && !Node.isGlobalNode())){
-      // This node is dead!
-      NI = Nodes.erase(NI);    // Erase & remove from node list.
-      ++NumTrivialDNE;
-    } else {
-      ++NI;
-    }
-  }
-
-  removeIdenticalCalls(FunctionCalls);
-  removeIdenticalCalls(AuxFunctionCalls);
-}
-
-
 /// markReachableNodes - This method recursively traverses the specified
 /// DSNodes, marking any nodes which are reachable.  All reachable nodes it adds
 /// to the set, which allows it to only traverse visited nodes once.
@@ -1669,360 +1002,6 @@
   return false;
 }
 
-// CallSiteUsesAliveArgs - Return true if the specified call site can reach any
-// alive nodes.
-//
-static bool CallSiteUsesAliveArgs(const DSCallSite &CS,
-                                  DenseSet<const DSNode*> &Alive,
-                                  DenseSet<const DSNode*> &Visited,
-                                  bool IgnoreGlobals) {
-  if (CanReachAliveNodes(CS.getRetVal().getNode(), Alive, Visited,
-                         IgnoreGlobals))
-    return true;
-  if (CS.isIndirectCall() &&
-      CanReachAliveNodes(CS.getCalleeNode(), Alive, Visited, IgnoreGlobals))
-    return true;
-  for (unsigned i = 0, e = CS.getNumPtrArgs(); i != e; ++i)
-    if (CanReachAliveNodes(CS.getPtrArg(i).getNode(), Alive, Visited,
-                           IgnoreGlobals))
-      return true;
-  return false;
-}
-
-// removeDeadNodes - Use a more powerful reachability analysis to eliminate
-// subgraphs that are unreachable.  This often occurs because the data
-// structure doesn't "escape" into it's caller, and thus should be eliminated
-// from the caller's graph entirely.  This is only appropriate to use when
-// inlining graphs.
-//
-void DSGraph::removeDeadNodes(unsigned Flags) {
-  DEBUG(AssertGraphOK(); if (GlobalsGraph) GlobalsGraph->AssertGraphOK());
-
-  // Reduce the amount of work we have to do... remove dummy nodes left over by
-  // merging...
-  removeTriviallyDeadNodes();
-
-  // FIXME: Merge non-trivially identical call nodes...
-
-  // Alive - a set that holds all nodes found to be reachable/alive.
-  DenseSet<const DSNode*> Alive;
-  std::vector<std::pair<const Value*, DSNode*> > GlobalNodes;
-
-  // Copy and merge all information about globals to the GlobalsGraph if this is
-  // not a final pass (where unreachable globals are removed).
-  //
-  // Strip all alloca bits since the current function is only for the BU pass.
-  // Strip all incomplete bits since they are short-lived properties and they
-  // will be correctly computed when rematerializing nodes into the functions.
-  //
-  ReachabilityCloner GGCloner(GlobalsGraph, this, DSGraph::StripAllocaBit |
-                              DSGraph::StripIncompleteBit);
-
-  // Mark all nodes reachable by (non-global) scalar nodes as alive...
-  for (DSScalarMap::iterator I = ScalarMap.begin(), E = ScalarMap.end();
-          I != E; ++I)
-    if (isa<GlobalValue > (I->first)) { // Keep track of global nodes
-      assert(!I->second.isNull() && "Null global node?");
-      assert(I->second.getNode()->isGlobalNode() && "Should be a global node!");
-      GlobalNodes.push_back(std::make_pair(I->first, I->second.getNode()));
-
-      // Make sure that all globals are cloned over as roots.
-      if (!(Flags & DSGraph::RemoveUnreachableGlobals) && GlobalsGraph) {
-        DSGraph::ScalarMapTy::iterator SMI =
-          GlobalsGraph->getScalarMap().find(I->first);
-        if (SMI != GlobalsGraph->getScalarMap().end())
-          GGCloner.merge(SMI->second, I->second);
-        else
-          GGCloner.getClonedNH(I->second);
-      }
-    } else {
-      I->second.getNode()->markReachableNodes(Alive);
-    }
-
-  // The return values are alive as well.
-  for (ReturnNodesTy::iterator I = ReturnNodes.begin(), E = ReturnNodes.end();
-       I != E; ++I)
-    I->second.getNode()->markReachableNodes(Alive);
-
-  // Mark any nodes reachable by primary calls as alive...
-  for (fc_iterator I = fc_begin(), E = fc_end(); I != E; ++I)
-    I->markReachableNodes(Alive);
-
-
-  // Now find globals and aux call nodes that are already live or reach a live
-  // value (which makes them live in turn), and continue till no more are found.
-  //
-  bool Iterate;
-  DenseSet<const DSNode*> Visited;
-  std::set<const DSCallSite*> AuxFCallsAlive;
-  do {
-    Visited.clear();
-    // If any global node points to a non-global that is "alive", the global is
-    // "alive" as well...  Remove it from the GlobalNodes list so we only have
-    // unreachable globals in the list.
-    //
-    Iterate = false;
-    if (!(Flags & DSGraph::RemoveUnreachableGlobals))
-      for (unsigned i = 0; i != GlobalNodes.size(); ++i)
-        if (CanReachAliveNodes(GlobalNodes[i].second, Alive, Visited,
-                               Flags & DSGraph::RemoveUnreachableGlobals)) {
-          std::swap(GlobalNodes[i--], GlobalNodes.back()); // Move to end to...
-          GlobalNodes.pop_back();                          // erase efficiently
-          Iterate = true;
-        }
-
-    // Mark only unresolvable call nodes for moving to the GlobalsGraph since
-    // call nodes that get resolved will be difficult to remove from that graph.
-    // The final unresolved call nodes must be handled specially at the end of
-    // the BU pass (i.e., in main or other roots of the call graph).
-    for (afc_iterator CI = afc_begin(), E = afc_end(); CI != E; ++CI)
-      if (!AuxFCallsAlive.count(&*CI) &&
-          (CI->isIndirectCall()
-           || CallSiteUsesAliveArgs(*CI, Alive, Visited,
-                                  Flags & DSGraph::RemoveUnreachableGlobals))) {
-        CI->markReachableNodes(Alive);
-        AuxFCallsAlive.insert(&*CI);
-        Iterate = true;
-      }
-  } while (Iterate);
-
-  // Move dead aux function calls to the end of the list
-  for (std::list<DSCallSite>::iterator CI = AuxFunctionCalls.begin(),
-         E = AuxFunctionCalls.end(); CI != E; )
-    if (AuxFCallsAlive.count(&*CI))
-      ++CI;
-    else {
-      // Copy and merge global nodes and dead aux call nodes into the
-      // GlobalsGraph, and all nodes reachable from those nodes.  Update their
-      // target pointers using the GGCloner.
-      //
-      if (!(Flags & DSGraph::RemoveUnreachableGlobals))
-        GlobalsGraph->AuxFunctionCalls.push_back(DSCallSite(*CI, GGCloner));
-
-      AuxFunctionCalls.erase(CI++);
-    }
-
-  // We are finally done with the GGCloner so we can destroy it.
-  GGCloner.destroy();
-
-  // At this point, any nodes which are visited, but not alive, are nodes
-  // which can be removed.  Loop over all nodes, eliminating completely
-  // unreachable nodes.
-  //
-  std::vector<DSNode*> DeadNodes;
-  DeadNodes.reserve(Nodes.size());
-  for (NodeListTy::iterator NI = Nodes.begin(), E = Nodes.end(); NI != E;) {
-    DSNode *N = NI++;
-    assert(!N->isForwarding() && "Forwarded node in nodes list?");
-
-    if (!Alive.count(N)) {
-      Nodes.remove(N);
-      assert(!N->isForwarding() && "Cannot remove a forwarding node!");
-      DeadNodes.push_back(N);
-      N->dropAllReferences();
-      ++NumDNE;
-    }
-  }
-
-  // Remove all unreachable globals from the ScalarMap.
-  // If flag RemoveUnreachableGlobals is set, GlobalNodes has only dead nodes.
-  // In either case, the dead nodes will not be in the set Alive.
-  for (unsigned i = 0, e = GlobalNodes.size(); i != e; ++i)
-    if (!Alive.count(GlobalNodes[i].second))
-      ScalarMap.erase(GlobalNodes[i].first);
-    else
-      assert((Flags & DSGraph::RemoveUnreachableGlobals) && "non-dead global");
-
-  // Delete all dead nodes now since their referrer counts are zero.
-  for (unsigned i = 0, e = DeadNodes.size(); i != e; ++i)
-    delete DeadNodes[i];
-
-  DEBUG(AssertGraphOK(); GlobalsGraph->AssertGraphOK());
-}
-
-void DSGraph::AssertNodeContainsGlobal(const DSNode *N, const GlobalValue *GV) const {
-  assert(std::find(N->globals_begin(),N->globals_end(), GV) !=
-         N->globals_end() && "Global value not in node!");
-}
-
-void DSGraph::AssertCallSiteInGraph(const DSCallSite &CS) const {
-  if (CS.isIndirectCall()) {
-    AssertNodeInGraph(CS.getCalleeNode());
-#if 0
-    if (CS.getNumPtrArgs() && CS.getCalleeNode() == CS.getPtrArg(0).getNode() &&
-        CS.getCalleeNode() && CS.getCalleeNode()->getGlobals().empty())
-      DEBUG(errs() << "WARNING: WEIRD CALL SITE FOUND!\n");
-#endif
-  }
-  AssertNodeInGraph(CS.getRetVal().getNode());
-  for (unsigned j = 0, e = CS.getNumPtrArgs(); j != e; ++j)
-    AssertNodeInGraph(CS.getPtrArg(j).getNode());
-}
-
-void DSGraph::AssertCallNodesInGraph() const {
-  for (fc_iterator I = fc_begin(), E = fc_end(); I != E; ++I)
-    AssertCallSiteInGraph(*I);
-}
-void DSGraph::AssertAuxCallNodesInGraph() const {
-  for (afc_const_iterator I = afc_begin(), E = afc_end(); I != E; ++I)
-    AssertCallSiteInGraph(*I);
-}
-
-void DSGraph::AssertGraphOK() const {
-  for (node_const_iterator NI = node_begin(), E = node_end(); NI != E; ++NI)
-    NI->assertOK();
-
-  for (ScalarMapTy::const_iterator I = ScalarMap.begin(),
-         E = ScalarMap.end(); I != E; ++I) {
-    assert(!I->second.isNull() && "Null node in scalarmap!");
-    AssertNodeInGraph(I->second.getNode());
-    if (const GlobalValue *GV = dyn_cast<GlobalValue>(I->first)) {
-      assert(I->second.getNode()->isGlobalNode() &&
-             "Global points to node, but node isn't global?");
-      AssertNodeContainsGlobal(I->second.getNode(), GV);
-    }
-  }
-  AssertCallNodesInGraph();
-  AssertAuxCallNodesInGraph();
-
-  // Check that all pointer arguments to any functions in this graph have
-  // destinations.
-  for (ReturnNodesTy::const_iterator RI = ReturnNodes.begin(),
-         E = ReturnNodes.end();
-       RI != E; ++RI) {
-    const Function &F = *RI->first;
-    for (Function::const_arg_iterator AI = F.arg_begin(); AI != F.arg_end(); ++AI)
-      if (isa<PointerType>(AI->getType()))
-        assert(!getNodeForValue(AI).isNull() &&
-               "Pointer argument must be in the scalar map!");
-  }
-}
-
-/// computeNodeMapping - Given roots in two different DSGraphs, traverse the
-/// nodes reachable from the two graphs, computing the mapping of nodes from the
-/// first to the second graph.  This mapping may be many-to-one (i.e. the first
-/// graph may have multiple nodes representing one node in the second graph),
-/// but it will not work if there is a one-to-many or many-to-many mapping.
-///
-void DSGraph::computeNodeMapping(const DSNodeHandle &NH1,
-                                 const DSNodeHandle &NH2, NodeMapTy &NodeMap,
-                                 bool StrictChecking) {
-  DSNode *N1 = NH1.getNode(), *N2 = NH2.getNode();
-  if (N1 == 0 || N2 == 0) return;
-
-  DSNodeHandle &Entry = NodeMap[N1];
-  if (!Entry.isNull()) {
-    // Termination of recursion!
-    if (StrictChecking) {
-      assert(Entry.getNode() == N2 && "Inconsistent mapping detected!");
-      assert((Entry.getOffset() == (NH2.getOffset()-NH1.getOffset()) ||
-              Entry.getNode()->isNodeCompletelyFolded()) &&
-             "Inconsistent mapping detected!");
-    }
-    return;
-  }
-
-  Entry.setTo(N2, NH2.getOffset()-NH1.getOffset());
-
-  // Loop over all of the fields that N1 and N2 have in common, recursively
-  // mapping the edges together now.
-  int N2Idx = NH2.getOffset()-NH1.getOffset();
-  unsigned N2Size = N2->getSize();
-  if (N2Size == 0) return;   // No edges to map to.
-
-  for (unsigned i = 0, e = N1->getSize(); i < e; ++i) {
-    const DSNodeHandle &N1NH = N1->getLink(i);
-    // Don't call N2->getLink if not needed (avoiding crash if N2Idx is not
-    // aligned right).
-    if (!N1NH.isNull()) {
-      if (unsigned(N2Idx)+i < N2Size)
-        computeNodeMapping(N1NH, N2->getLink(N2Idx+i), NodeMap);
-      else
-        computeNodeMapping(N1NH,
-                           N2->getLink(unsigned(N2Idx+i) % N2Size), NodeMap);
-    }
-  }
-}
-
-
-/// computeGToGGMapping - Compute the mapping of nodes in the global graph to
-/// nodes in this graph.
-void DSGraph::computeGToGGMapping(NodeMapTy &NodeMap) {
-  DSGraph &GG = *getGlobalsGraph();
-
-  DSScalarMap &SM = getScalarMap();
-  for (DSScalarMap::global_iterator I = SM.global_begin(),
-         E = SM.global_end(); I != E; ++I)
-    DSGraph::computeNodeMapping(SM[*I], GG.getNodeForValue(*I), NodeMap);
-}
-
-/// computeGGToGMapping - Compute the mapping of nodes in the global graph to
-/// nodes in this graph.  Note that any uses of this method are probably bugs,
-/// unless it is known that the globals graph has been merged into this graph!
-void DSGraph::computeGGToGMapping(InvNodeMapTy &InvNodeMap) {
-  NodeMapTy NodeMap;
-  computeGToGGMapping(NodeMap);
-
-  while (!NodeMap.empty()) {
-    InvNodeMap.insert(std::make_pair(NodeMap.begin()->second,
-                                     NodeMap.begin()->first));
-    NodeMap.erase(NodeMap.begin());
-  }
-}
-
-
-/// computeCalleeCallerMapping - Given a call from a function in the current
-/// graph to the 'Callee' function (which lives in 'CalleeGraph'), compute the
-/// mapping of nodes from the callee to nodes in the caller.
-void DSGraph::computeCalleeCallerMapping(DSCallSite CS, const Function &Callee,
-                                         DSGraph &CalleeGraph,
-                                         NodeMapTy &NodeMap) {
-
-  DSCallSite CalleeArgs =
-    CalleeGraph.getCallSiteForArguments(const_cast<Function&>(Callee));
-
-  computeNodeMapping(CalleeArgs.getRetVal(), CS.getRetVal(), NodeMap);
-
-  unsigned NumArgs = CS.getNumPtrArgs();
-  if (NumArgs > CalleeArgs.getNumPtrArgs())
-    NumArgs = CalleeArgs.getNumPtrArgs();
-
-  for (unsigned i = 0; i != NumArgs; ++i)
-    computeNodeMapping(CalleeArgs.getPtrArg(i), CS.getPtrArg(i), NodeMap);
-
-  // Map the nodes that are pointed to by globals.
-  DSScalarMap &CalleeSM = CalleeGraph.getScalarMap();
-  DSScalarMap &CallerSM = getScalarMap();
-
-  if (CalleeSM.global_size() >= CallerSM.global_size()) {
-    for (DSScalarMap::global_iterator GI = CallerSM.global_begin(),
-           E = CallerSM.global_end(); GI != E; ++GI)
-      if (CalleeSM.global_count(*GI))
-        computeNodeMapping(CalleeSM[*GI], CallerSM[*GI], NodeMap);
-  } else {
-    for (DSScalarMap::global_iterator GI = CalleeSM.global_begin(),
-           E = CalleeSM.global_end(); GI != E; ++GI)
-      if (CallerSM.global_count(*GI))
-        computeNodeMapping(CalleeSM[*GI], CallerSM[*GI], NodeMap);
-  }
-}
-
-/// updateFromGlobalGraph - This function rematerializes global nodes and
-/// nodes reachable from them from the globals graph into the current graph.
-///
-void DSGraph::updateFromGlobalGraph() {
-  ReachabilityCloner RC(this, GlobalsGraph, 0);
-
-  // Clone the non-up-to-date global nodes into this graph.
-  for (DSScalarMap::global_iterator I = getScalarMap().global_begin(),
-         E = getScalarMap().global_end(); I != E; ++I) {
-    DSScalarMap::iterator It = GlobalsGraph->ScalarMap.find(*I);
-    if (It != GlobalsGraph->ScalarMap.end())
-      RC.merge(getNodeForValue(*I), It->second);
-  }
-}
-
-
 ////////////////////////////////////////////////////////////////////////////////
 //Base DataStructures impl:
 ////////////////////////////////////////////////////////////////////////////////
@@ -2226,7 +1205,7 @@
   resetAuxCalls = resetAux;
   TD = D->TD;
   TypeSS = D->TypeSS;
-  ActualCallees = D->ActualCallees;
+  callgraph = D->callgraph;
   GlobalECs = D->getGlobalECs();
   GlobalsGraph = new DSGraph(D->getGlobalsGraph(), GlobalECs, *TypeSS,
                              copyGlobalAuxCalls?0:DSGraph::DontCloneAuxCallNodes);
@@ -2264,7 +1243,7 @@
   // Empty map so next time memory is released, data structures are not
   // re-deleted.
   DSInfo.clear();
-  ActualCallees.clear();
+  callgraph.clear();
   delete GlobalsGraph;
   GlobalsGraph = 0;
 }

Modified: poolalloc/trunk/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Local.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Local.cpp (original)
+++ poolalloc/trunk/lib/DSA/Local.cpp Thu Mar 11 18:16:41 2010
@@ -782,7 +782,7 @@
   // Add a new function call entry...
   if (CalleeNode) {
     G.getFunctionCalls().push_back(DSCallSite(CS, RetVal, CalleeNode, Args));
-    DS->callee_site(CS.getInstruction());
+//    DS->callee_site(CS.getInstruction());
   }else
     G.getFunctionCalls().push_back(DSCallSite(CS, RetVal, cast<Function>(Callee),
                                               Args));

Modified: poolalloc/trunk/lib/DSA/Printer.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Printer.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/Printer.cpp (original)
+++ poolalloc/trunk/lib/DSA/Printer.cpp Thu Mar 11 18:16:41 2010
@@ -365,17 +365,15 @@
 
 
 void DataStructures::dumpCallGraph() const {
-  for(  ActualCalleesTy::const_iterator ii = ActualCallees.begin(), ee = ActualCallees.end();
-        ii != ee; ++ii) {
-    if (ii->first) errs() << ii->first->getParent()->getParent()->getName() << " ";
-    errs() << ii->first << ": [";
-    for (callee_iterator cbi = ii->second.begin(), cbe = ii->second.end();
-         cbi != cbe; ++cbi) {
-      errs() << (*cbi)->getName() << " ";
+  for (DSCallGraph::key_iterator ki = callgraph.key_begin(),
+          ke = callgraph.key_end(); ki != ke; ++ki)
+    if (*ki != CallSite()) {
+      errs() << ki->getInstruction()->getParent()->getParent()->getName() << ": [";
+      for (DSCallGraph::iterator cbi = callgraph.callee_begin(*ki),
+           cbe = callgraph.callee_end(*ki); cbi != cbe; ++cbi)
+        errs() << (*cbi)->getName() << " ";
+      errs() << "]\n";
     }
-    errs() << "]\n";
-    if (ii->first) ii->first->dump();
-  }
 }
 
 // print - Print out the analysis results...

Modified: poolalloc/trunk/lib/DSA/TopDownClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/TopDownClosure.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/TopDownClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/TopDownClosure.cpp Thu Mar 11 18:16:41 2010
@@ -164,12 +164,10 @@
   Visited.insert(G);
 
   // Recursively traverse all of the callee graphs.
-  for (DSGraph::fc_iterator CI = G->fc_begin(), CE = G->fc_end(); CI != CE; ++CI){
-    Instruction *CallI = CI->getCallSite().getInstruction();
-    for (callee_iterator I = callee_begin(CallI),
-           E = callee_end(CallI); I != E; ++I)
+  for (DSGraph::fc_iterator CI = G->fc_begin(), CE = G->fc_end(); CI != CE; ++CI)
+    for (DSCallGraph::iterator I = callgraph.callee_begin(CI->getCallSite()),
+           E = callgraph.callee_end(CI->getCallSite()); I != E; ++I)
       ComputePostOrder(**I, Visited, PostOrder);
-  }
 
   PostOrder.push_back(G);
 }
@@ -301,10 +299,9 @@
       continue;
     }
 
-    Instruction *CallI = CI->getCallSite().getInstruction();
     // For each function in the invoked function list at this call site...
-    callee_iterator IPI =
-      callee_begin(CallI), IPE = callee_end(CallI);
+    DSCallGraph::iterator IPI = callgraph.callee_begin(CI->getCallSite()),
+            IPE = callgraph.callee_end(CI->getCallSite());
 
     // Skip over all calls to this graph (SCC calls).
     while (IPI != IPE && getDSGraph(**IPI) == DSG)
@@ -335,8 +332,8 @@
     // so we build up a new, private, graph that represents the calls of all
     // calls to this set of functions.
     std::vector<const Function*> Callees;
-    for (callee_iterator I = callee_begin(CallI), E = callee_end(CallI);
-         I != E; ++I)
+    for (DSCallGraph::iterator I = callgraph.callee_begin(CI->getCallSite()),
+         E = callgraph.callee_end(CI->getCallSite()); I != E; ++I)
       if (!(*I)->isDeclaration())
         Callees.push_back(*I);
     std::sort(Callees.begin(), Callees.end());

Modified: poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/RunTimeAssociate.cpp Thu Mar 11 18:16:41 2010
@@ -461,8 +461,8 @@
     Instruction *OrigInst =
       cast<Instruction>(FI.getOldValueIfAvailable(CS.getInstruction()));
 
-    DataStructures::callee_iterator I = DS->callee_begin(OrigInst);
-    if (I != DS->callee_end(OrigInst))
+    DSCallGraph::iterator I = DS->getCallGraph().callee_begin(CS);
+    if (I != DS->getCallGraph().callee_end(CS))
       CF = *I;
 
     // If we didn't find the callee in the constructed call graph, try
@@ -508,7 +508,7 @@
 
 #ifndef NDEBUG
     // Verify that all potential callees at call site have the same DS graph.
-    DataStructures::callee_iterator E = DS->callee_end(OrigInst);
+    DSCallGraph::iterator E = DS->getCallGraph().callee_end(CS);
     for (; I != E; ++I)
       if (!(*I)->isDeclaration())
         assert(CalleeGraph == DS->getDSGraph(**I) &&

Modified: poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp?rev=98315&r1=98314&r2=98315&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp Thu Mar 11 18:16:41 2010
@@ -645,8 +645,8 @@
     Instruction *OrigInst =
       cast<Instruction>(getOldValueIfAvailable(CS.getInstruction()));
 
-    DataStructures::callee_iterator I = Graphs.callee_begin(OrigInst);
-    if (I != Graphs.callee_end(OrigInst))
+    DSCallGraph::iterator I = Graphs.getCallGraph().callee_begin(CS);
+    if (I != Graphs.getCallGraph().callee_end(CS))
       CF = *I;
     
     // If we didn't find the callee in the constructed call graph, try
@@ -685,7 +685,7 @@
     
 #ifndef NDEBUG
     // Verify that all potential callees at call site have the same DS graph.
-    DataStructures::callee_iterator E = Graphs.callee_end(OrigInst);
+    DSCallGraph::iterator E = Graphs.getCallGraph().callee_end(OrigInst);
     for (; I != E; ++I)
       if (!(*I)->isDeclaration())
         assert(CalleeGraph == Graphs.getDSGraph(**I) &&




From dpatel at apple.com  Thu Mar 11 18:38:19 2010
From: dpatel at apple.com (Devang Patel)
Date: Fri, 12 Mar 2010 00:38:19 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r98316 -
	/llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp
Message-ID: <20100312003819.155292A6C12C@llvm.org>

Author: dpatel
Date: Thu Mar 11 18:38:18 2010
New Revision: 98316

URL: http://llvm.org/viewvc/llvm-project?rev=98316&view=rev
Log:
Delete module before llvm_shutdown().
There is no need to destruct llvm symbols in orderly manner if compile time is important in release builds.

Modified:
    llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp

Modified: llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp?rev=98316&r1=98315&r2=98316&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp Thu Mar 11 18:38:18 2010
@@ -974,7 +974,10 @@
 
 // llvm_call_llvm_shutdown - Release LLVM global state.
 void llvm_call_llvm_shutdown(void) {
+#ifndef NDEBUG
+  delete TheModule;
   llvm_shutdown();
+#endif
 }
 
 /// llvm_emit_code_for_current_function - Top level interface for emitting a




From jyasskin at google.com  Thu Mar 11 18:44:55 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Thu, 11 Mar 2010 16:44:55 -0800
Subject: [llvm-commits] [llvm-gcc-4.2] r98316 -
	/llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp
In-Reply-To: <20100312003819.155292A6C12C@llvm.org>
References: <20100312003819.155292A6C12C@llvm.org>
Message-ID: <bb55930e1003111644s24868292wf7bd8e5d8bd9ecca@mail.gmail.com>

Thanks! Sorry I didn't get around to this.

On Thu, Mar 11, 2010 at 4:38 PM, Devang Patel <dpatel at apple.com> wrote:
> Author: dpatel
> Date: Thu Mar 11 18:38:18 2010
> New Revision: 98316
>
> URL: http://llvm.org/viewvc/llvm-project?rev=98316&view=rev
> Log:
> Delete module before llvm_shutdown().
> There is no need to destruct llvm symbols in orderly manner if compile time is important in release builds.
>
> Modified:
> ? ?llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp
>
> Modified: llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp?rev=98316&r1=98315&r2=98316&view=diff
> ==============================================================================
> --- llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp (original)
> +++ llvm-gcc-4.2/trunk/gcc/llvm-backend.cpp Thu Mar 11 18:38:18 2010
> @@ -974,7 +974,10 @@
>
> ?// llvm_call_llvm_shutdown - Release LLVM global state.
> ?void llvm_call_llvm_shutdown(void) {
> +#ifndef NDEBUG
> + ?delete TheModule;
> ? llvm_shutdown();
> +#endif
> ?}
>
> ?/// llvm_emit_code_for_current_function - Top level interface for emitting a
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From sabre at nondot.org  Thu Mar 11 19:56:43 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 01:56:43 -0000
Subject: [llvm-commits] [llvm] r98324 - in /llvm/trunk:
 include/llvm/Target/Mangler.h lib/Target/TargetLoweringObjectFile.cpp
Message-ID: <20100312015643.765082A6C12C@llvm.org>

Author: lattner
Date: Thu Mar 11 19:56:43 2010
New Revision: 98324

URL: http://llvm.org/viewvc/llvm-project?rev=98324&view=rev
Log:
make TargetLoweringObjectFile::getExprForDwarfReference
just make unnamed temp symbols instead of having to come
up with its own names.

Modified:
    llvm/trunk/include/llvm/Target/Mangler.h
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp

Modified: llvm/trunk/include/llvm/Target/Mangler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Mangler.h?rev=98324&r1=98323&r2=98324&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Mangler.h (original)
+++ llvm/trunk/include/llvm/Target/Mangler.h Thu Mar 11 19:56:43 2010
@@ -51,12 +51,6 @@
   // symbols.
   Mangler(const MCAsmInfo &mai) : MAI(mai), NextAnonGlobalID(1) {}
 
-  /// getUniqueID() - Allocate and return a unique ID.
-  /// FIXME: Remove this.
-  unsigned getUniqueID() {
-    return NextAnonGlobalID++;
-  }
-  
   /// getNameWithPrefix - Fill OutName with the name of the appropriate prefix
   /// and the specified global variable's name.  If the global variable doesn't
   /// have a name, this fills in a unique name for the global.

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98324&r1=98323&r2=98324&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Thu Mar 11 19:56:43 2010
@@ -326,11 +326,7 @@
   case dwarf::DW_EH_PE_pcrel: {
     // Emit a label to the streamer for the current position.  This gives us
     // .-foo addressing.
-    SmallString<128> Name;
-    Mang->getNameWithPrefix(Name, Twine("PCtemp") + Twine(Mang->getUniqueID()),
-                            Mangler::Private);
-
-    MCSymbol *PCSym = getContext().GetOrCreateTemporarySymbol(Name.str());
+    MCSymbol *PCSym = getContext().GetOrCreateTemporarySymbol();
     Streamer.EmitLabel(PCSym);
     const MCExpr *PC = MCSymbolRefExpr::Create(PCSym, getContext());
     return MCBinaryExpr::CreateSub(Res, PC, getContext());




From isanbard at gmail.com  Thu Mar 11 20:00:44 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Fri, 12 Mar 2010 02:00:44 -0000
Subject: [llvm-commits] [llvm] r98325 - in /llvm/trunk/lib/Target/PowerPC:
 AsmPrinter/PPCAsmPrinter.cpp CMakeLists.txt PPCISelLowering.cpp
 PPCTargetObjectFile.cpp PPCTargetObjectFile.h
Message-ID: <20100312020044.2520F2A6C12C@llvm.org>

Author: void
Date: Thu Mar 11 20:00:43 2010
New Revision: 98325

URL: http://llvm.org/viewvc/llvm-project?rev=98325&view=rev
Log:
The same situation that effected ARM effects PPC with regards to placing the
LSDA into the TEXT section. We need to generate non-lazy pointers to it on
Mach-O. However, the object the NLP points to may be local to the translation
unit. If so, then the NLP needs to have the value of that object specified
instead of "0", which the linker interprets as "external".

Added:
    llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.cpp
    llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.h
Modified:
    llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
    llvm/trunk/lib/Target/PowerPC/CMakeLists.txt
    llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp

Modified: llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp?rev=98325&r1=98324&r2=98325&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/AsmPrinter/PPCAsmPrinter.cpp Thu Mar 11 20:00:43 2010
@@ -812,7 +812,15 @@
       //   .indirect_symbol _foo
       MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
       OutStreamer.EmitSymbolAttribute(MCSym.getPointer(), MCSA_IndirectSymbol);
-      OutStreamer.EmitIntValue(0, isPPC64 ? 8 : 4/*size*/, 0/*addrspace*/);
+
+      if (MCSym.getInt())
+        // External to current translation unit.
+        OutStreamer.EmitIntValue(0, isPPC64 ? 8 : 4/*size*/, 0/*addrspace*/);
+      else
+        // Internal to current translation unit.
+        OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym.getPointer(),
+                                                      OutContext),
+                              isPPC64 ? 8 : 4/*size*/, 0/*addrspace*/);
     }
 
     Stubs.clear();

Modified: llvm/trunk/lib/Target/PowerPC/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/CMakeLists.txt?rev=98325&r1=98324&r2=98325&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/CMakeLists.txt (original)
+++ llvm/trunk/lib/Target/PowerPC/CMakeLists.txt Thu Mar 11 20:00:43 2010
@@ -24,6 +24,7 @@
   PPCRegisterInfo.cpp
   PPCSubtarget.cpp
   PPCTargetMachine.cpp
+  PPCTargetObjectFile.cpp
   )
 
 target_link_libraries (LLVMPowerPCCodeGen LLVMSelectionDAG)

Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=98325&r1=98324&r2=98325&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Thu Mar 11 20:00:43 2010
@@ -13,9 +13,10 @@
 
 #include "PPCISelLowering.h"
 #include "PPCMachineFunctionInfo.h"
+#include "PPCPerfectShuffle.h"
 #include "PPCPredicates.h"
 #include "PPCTargetMachine.h"
-#include "PPCPerfectShuffle.h"
+#include "PPCTargetObjectFile.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/VectorExtras.h"
 #include "llvm/CodeGen/CallingConvLower.h"
@@ -59,11 +60,11 @@
 
 static TargetLoweringObjectFile *CreateTLOF(const PPCTargetMachine &TM) {
   if (TM.getSubtargetImpl()->isDarwin())
-    return new TargetLoweringObjectFileMachO();
+    return new PPCMachOTargetObjectFile();
+
   return new TargetLoweringObjectFileELF();
 }
 
-
 PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM)
   : TargetLowering(TM, CreateTLOF(TM)), PPCSubTarget(*TM.getSubtargetImpl()) {
 

Added: llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.cpp?rev=98325&view=auto
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.cpp (added)
+++ llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.cpp Thu Mar 11 20:00:43 2010
@@ -0,0 +1,33 @@
+//===-- llvm/Target/PPCTargetObjectFile.cpp - PPC Object Info Impl --------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "PPCTargetObjectFile.h"
+#include "PPCSubtarget.h"
+#include "llvm/MC/MCSectionMachO.h"
+#include "llvm/Support/Dwarf.h"
+#include "llvm/Target/TargetMachine.h"
+using namespace llvm;
+using namespace dwarf;
+
+//===----------------------------------------------------------------------===//
+//                              Mach-O Target
+//===----------------------------------------------------------------------===//
+
+void PPCMachOTargetObjectFile::Initialize(MCContext &Ctx,
+                                          const TargetMachine &TM) {
+  TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
+
+  // Exception Handling.
+  LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
+                                SectionKind::getReadOnlyWithRel());
+}
+
+unsigned PPCMachOTargetObjectFile::getTTypeEncoding() const {
+  return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
+}

Added: llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.h?rev=98325&view=auto
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.h (added)
+++ llvm/trunk/lib/Target/PowerPC/PPCTargetObjectFile.h Thu Mar 11 20:00:43 2010
@@ -0,0 +1,34 @@
+//===-- llvm/Target/PPCTargetObjectFile.h - PowerPC Object Info -*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TARGET_PPC_TARGETOBJECTFILE_H
+#define LLVM_TARGET_PPC_TARGETOBJECTFILE_H
+
+#include "llvm/CodeGen/TargetLoweringObjectFileImpl.h"
+
+namespace llvm {
+
+class MCContext;
+class TargetMachine;
+
+// FIXME: This subclass isn't 100% necessary. It will become obsolete once we
+//        can place all LSDAs into the TEXT section. See
+//        <rdar://problem/6804645>.
+class PPCMachOTargetObjectFile : public TargetLoweringObjectFileMachO {
+public:
+  PPCMachOTargetObjectFile() : TargetLoweringObjectFileMachO() {}
+
+  virtual void Initialize(MCContext &Ctx, const TargetMachine &TM);
+
+  virtual unsigned getTTypeEncoding() const;
+};
+
+} // end namespace llvm
+
+#endif




From sabre at nondot.org  Fri Mar 12 00:32:13 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 06:32:13 -0000
Subject: [llvm-commits] [llvm] r98334 - /llvm/trunk/utils/mkpatch
Message-ID: <20100312063213.207A12A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 00:32:12 2010
New Revision: 98334

URL: http://llvm.org/viewvc/llvm-project?rev=98334&view=rev
Log:
update mkpatch for MC, patch by Aaron Gray

Modified:
    llvm/trunk/utils/mkpatch

Modified: llvm/trunk/utils/mkpatch
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/mkpatch?rev=98334&r1=98333&r2=98334&view=diff
==============================================================================
--- llvm/trunk/utils/mkpatch (original)
+++ llvm/trunk/utils/mkpatch Fri Mar 12 00:32:12 2010
@@ -23,7 +23,7 @@
 svn diff -x -u  >> "$NAME".patch.raw 2>&1 \
   autoconf docs utils include lib/System lib/Support lib/VMCore lib/AsmParser \
   lib/Bitcode lib/Analysis lib/Transforms lib/CodeGen lib/Target \
-  lib/ExecutionEngine lib/Linker \
+  lib/ExecutionEngine lib/Linker lib/MC \
   tools test unittests runtime projects examples Xcode
 
 echo "mkpatch: Removing cruft from the patch file"




From sabre at nondot.org  Fri Mar 12 01:15:36 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 07:15:36 -0000
Subject: [llvm-commits] [llvm] r98336 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp test/CodeGen/X86/crash.ll
Message-ID: <20100312071536.8DD122A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 01:15:36 2010
New Revision: 98336

URL: http://llvm.org/viewvc/llvm-project?rev=98336&view=rev
Log:
fix PR6577, a bug in sdbuilder lowering select instructions
whose true value was not Val#0.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
    llvm/trunk/test/CodeGen/X86/crash.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=98336&r1=98335&r2=98336&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Fri Mar 12 01:15:36 2010
@@ -2185,7 +2185,8 @@
 
   for (unsigned i = 0; i != NumValues; ++i)
     Values[i] = DAG.getNode(ISD::SELECT, getCurDebugLoc(),
-                            TrueVal.getNode()->getValueType(i), Cond,
+                          TrueVal.getNode()->getValueType(TrueVal.getResNo()+i),
+                            Cond,
                             SDValue(TrueVal.getNode(),
                                     TrueVal.getResNo() + i),
                             SDValue(FalseVal.getNode(),

Modified: llvm/trunk/test/CodeGen/X86/crash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/crash.ll?rev=98336&r1=98335&r2=98336&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/crash.ll (original)
+++ llvm/trunk/test/CodeGen/X86/crash.ll Fri Mar 12 01:15:36 2010
@@ -33,3 +33,17 @@
 if.end:                                           ; preds = %land.end
   ret void
 }
+
+; PR6577
+%pair = type { i64, double }
+
+define void @on4() {
+dependentGraph243.exit:
+  %subject19 = load %pair* undef                     ; <%1> [#uses=1]
+  %0 = extractvalue %pair %subject19, 1              ; <double> [#uses=2]
+  %1 = select i1 undef, double %0, double undef   ; <double> [#uses=1]
+  %2 = select i1 undef, double %1, double %0      ; <double> [#uses=1]
+  %3 = insertvalue %pair undef, double %2, 1         ; <%1> [#uses=1]
+  store %pair %3, %pair* undef
+  ret void
+}




From benny.kra at googlemail.com  Fri Mar 12 03:27:42 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Fri, 12 Mar 2010 09:27:42 -0000
Subject: [llvm-commits] [llvm] r98338 - in /llvm/trunk:
 include/llvm/Transforms/Utils/BuildLibCalls.h
 lib/Transforms/InstCombine/InstCombineCalls.cpp
 lib/Transforms/Scalar/CodeGenPrepare.cpp
 lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100312092742.2BEA32A6C12C@llvm.org>

Author: d0k
Date: Fri Mar 12 03:27:41 2010
New Revision: 98338

URL: http://llvm.org/viewvc/llvm-project?rev=98338&view=rev
Log:
Factor checked library call optimization into a common helper class and use it
to unify the almost identical code in CodeGenPrepare and InstCombineCalls.

Modified:
    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h?rev=98338&r1=98337&r2=98338&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h (original)
+++ llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h Fri Mar 12 03:27:41 2010
@@ -96,6 +96,18 @@
   /// a pointer, Size is an 'intptr_t', and File is a pointer to FILE.
   void EmitFWrite(Value *Ptr, Value *Size, Value *File, IRBuilder<> &B,
                   const TargetData *TD);
+
+  /// SimplifyFortifiedLibCalls - Helper class for folding checked library
+  /// calls (e.g. __strcpy_chk) into their unchecked counterparts.
+  class SimplifyFortifiedLibCalls {
+  protected:
+    CallInst *CI;
+    virtual void replaceCall(Value *With) = 0;
+    virtual bool isFoldable(unsigned SizeCIOp, unsigned SizeArgOp,
+                            bool isString) const = 0;
+  public:
+    bool fold(CallInst *CI, const TargetData *TD);
+  };
 }
 
 #endif

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=98338&r1=98337&r2=98338&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Fri Mar 12 03:27:41 2010
@@ -751,117 +751,41 @@
   return true;
 }
 
+namespace {
+class InstCombineFortifiedLibCalls : public SimplifyFortifiedLibCalls {
+  InstCombiner *IC;
+protected:
+  void replaceCall(Value *With) {
+    NewInstruction = IC->ReplaceInstUsesWith(*CI, With);
+  }
+  bool isFoldable(unsigned SizeCIOp, unsigned SizeArgOp, bool isString) const {
+    if (ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(SizeCIOp))) {
+      if (SizeCI->isAllOnesValue())
+        return true;
+      if (isString)
+        return SizeCI->getZExtValue() >=
+               GetStringLength(CI->getOperand(SizeArgOp));
+      if (ConstantInt *Arg = dyn_cast<ConstantInt>(CI->getOperand(SizeArgOp)))
+        return SizeCI->getZExtValue() <= Arg->getZExtValue();
+    }
+    return false;
+  }
+public:
+  InstCombineFortifiedLibCalls(InstCombiner *IC) : IC(IC), NewInstruction(0) { }
+  Instruction *NewInstruction;
+};
+} // end anonymous namespace
+
 // Try to fold some different type of calls here.
 // Currently we're only working with the checking functions, memcpy_chk, 
 // mempcpy_chk, memmove_chk, memset_chk, strcpy_chk, stpcpy_chk, strncpy_chk,
 // strcat_chk and strncat_chk.
 Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const TargetData *TD) {
   if (CI->getCalledFunction() == 0) return 0;
-  
-  StringRef Name = CI->getCalledFunction()->getName();
-  BasicBlock *BB = CI->getParent();
-  IRBuilder<> B(CI->getParent()->getContext());
-  
-  // Set the builder to the instruction after the call.
-  B.SetInsertPoint(BB, CI);
-
-  if (Name == "__memcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    ConstantInt *SizeArg = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeArg)
-      return 0;
-    if (SizeCI->isAllOnesValue() ||
-        SizeCI->getZExtValue() <= SizeArg->getZExtValue()) {
-      EmitMemCpy(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
-                 1, B, TD);
-      return ReplaceInstUsesWith(*CI, CI->getOperand(1));
-    }
-    return 0;
-  }
-
-  // Should be similar to memcpy.
-  if (Name == "__mempcpy_chk") {
-    return 0;
-  }
-
-  if (Name == "__memmove_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    ConstantInt *SizeArg = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeArg)
-      return 0;
-    if (SizeCI->isAllOnesValue() ||
-        SizeCI->getZExtValue() <= SizeArg->getZExtValue()) {
-      EmitMemMove(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
-                  1, B, TD);
-      return ReplaceInstUsesWith(*CI, CI->getOperand(1));
-    }
-    return 0;
-  }
-
-  if (Name == "__memset_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    ConstantInt *SizeArg = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeArg)
-      return 0;
-    if (SizeCI->isAllOnesValue() ||
-        SizeCI->getZExtValue() <= SizeArg->getZExtValue()) {
-      Value *Val = B.CreateIntCast(CI->getOperand(2), B.getInt8Ty(),
-                                   false);
-      EmitMemSet(CI->getOperand(1), Val,  CI->getOperand(3), B, TD);
-      return ReplaceInstUsesWith(*CI, CI->getOperand(1));
-    }
-    return 0;
-  }
-
-  if (Name == "__strcpy_chk" || Name == "__stpcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeCI)
-      return 0;
-    // If a) we don't have any length information, or b) we know this will
-    // fit then just lower to a plain st[rp]cpy. Otherwise we'll keep our
-    // st[rp]cpy_chk call which may fail at runtime if the size is too long.
-    // TODO: It might be nice to get a maximum length out of the possible
-    // string lengths for varying.
-    if (SizeCI->isAllOnesValue() ||
-      SizeCI->getZExtValue() >= GetStringLength(CI->getOperand(2))) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
-                              Name.substr(2, 6));
-      return ReplaceInstUsesWith(*CI, Ret);
-    }
-    return 0;
-  }
-
-  if (Name == "__strncpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    ConstantInt *SizeArg = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeArg)
-      return 0;
-    if (SizeCI->isAllOnesValue() ||
-        SizeCI->getZExtValue() <= SizeArg->getZExtValue()) {
-      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
-                               CI->getOperand(3), B, TD);
-      return ReplaceInstUsesWith(*CI, Ret);
-    }
-    return 0; 
-  }
-
-  if (Name == "__strcat_chk") {
-    return 0;
-  }
-
-  if (Name == "__strncat_chk") {
-    return 0;
-  }
 
-  return 0;
+  InstCombineFortifiedLibCalls Simplifier(this);
+  Simplifier.fold(CI, TD);
+  return Simplifier.NewInstruction;
 }
 
 // visitCallSite - Improvements for call and invoke instructions.

Modified: llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp?rev=98338&r1=98337&r2=98338&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/CodeGenPrepare.cpp Fri Mar 12 03:27:41 2010
@@ -540,9 +540,22 @@
   return MadeChange;
 }
 
+namespace {
+class CodeGenPrepareFortifiedLibCalls : public SimplifyFortifiedLibCalls {
+protected:
+  void replaceCall(Value *With) {
+    CI->replaceAllUsesWith(With);
+    CI->eraseFromParent();
+  }
+  bool isFoldable(unsigned SizeCIOp, unsigned, bool) const {
+    if (ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(SizeCIOp)))
+      return SizeCI->isAllOnesValue();
+    return false;
+  }
+};
+} // end anonymous namespace
+
 bool CodeGenPrepare::OptimizeCallInst(CallInst *CI) {
-  bool MadeChange = false;
-  
   // Lower all uses of llvm.objectsize.*
   IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI);
   if (II && II->getIntrinsicID() == Intrinsic::objectsize) {
@@ -561,102 +574,12 @@
   const TargetData *TD = TLI ? TLI->getTargetData() : 0;
   if (!TD) return false;
   
-  // Lower all default uses of _chk calls.  This is code very similar
-  // to the code in InstCombineCalls, but here we are only lowering calls
+  // Lower all default uses of _chk calls.  This is very similar
+  // to what InstCombineCalls does, but here we are only lowering calls
   // that have the default "don't know" as the objectsize.  Anything else
   // should be left alone.
-  StringRef Name = CI->getCalledFunction()->getName();
-  BasicBlock *BB = CI->getParent();
-  IRBuilder<> B(CI->getParent()->getContext());
-  
-  // Set the builder to the instruction after the call.
-  B.SetInsertPoint(BB, CI);
-
-  if (Name == "__memcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      EmitMemCpy(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
-                 1, B, TD);
-      CI->replaceAllUsesWith(CI->getOperand(1));
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0;
-  }
-
-  // Should be similar to memcpy.
-  if (Name == "__mempcpy_chk") {
-    return 0;
-  }
-
-  if (Name == "__memmove_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      EmitMemMove(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
-                  1, B, TD);
-      CI->replaceAllUsesWith(CI->getOperand(1));
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0;
-  }
-
-  if (Name == "__memset_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      Value *Val = B.CreateIntCast(CI->getOperand(2), B.getInt8Ty(),
-                                   false);
-      EmitMemSet(CI->getOperand(1), Val,  CI->getOperand(3), B, TD);
-      CI->replaceAllUsesWith(CI->getOperand(1));
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0;
-  }
-
-  if (Name == "__strcpy_chk" || Name == "__stpcpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(3));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
-                              Name.substr(2, 6));
-      CI->replaceAllUsesWith(Ret);
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0;
-  }
-
-  if (Name == "__strncpy_chk") {
-    ConstantInt *SizeCI = dyn_cast<ConstantInt>(CI->getOperand(4));
-    if (!SizeCI)
-      return 0;
-    if (SizeCI->isAllOnesValue()) {
-      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
-                               CI->getOperand(3), B, TD);
-      CI->replaceAllUsesWith(Ret);
-      CI->eraseFromParent();
-      return true;
-    }
-    return 0; 
-  }
-
-  if (Name == "__strcat_chk") {
-    return 0;
-  }
-
-  if (Name == "__strncat_chk") {
-    return 0;
-  }
-  
-  return MadeChange;
+  CodeGenPrepareFortifiedLibCalls Simplifier;
+  return Simplifier.fold(CI, TD);
 }
 //===----------------------------------------------------------------------===//
 // Memory Optimization

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98338&r1=98337&r2=98338&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Fri Mar 12 03:27:41 2010
@@ -342,3 +342,84 @@
   if (const Function *Fn = dyn_cast<Function>(F->stripPointerCasts()))
     CI->setCallingConv(Fn->getCallingConv());
 }
+
+bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
+  this->CI = CI;
+  StringRef Name = CI->getCalledFunction()->getName();
+  BasicBlock *BB = CI->getParent();
+  IRBuilder<> B(CI->getParent()->getContext());
+
+  // Set the builder to the instruction after the call.
+  B.SetInsertPoint(BB, CI);
+
+  if (Name == "__memcpy_chk") {
+    if (isFoldable(4, 3, false)) {
+      EmitMemCpy(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+                 1, B, TD);
+      replaceCall(CI->getOperand(1));
+      return true;
+    }
+    return false;
+  }
+
+  // Should be similar to memcpy.
+  if (Name == "__mempcpy_chk") {
+    return false;
+  }
+
+  if (Name == "__memmove_chk") {
+    if (isFoldable(4, 3, false)) {
+      EmitMemMove(CI->getOperand(1), CI->getOperand(2), CI->getOperand(3),
+                  1, B, TD);
+      replaceCall(CI->getOperand(1));
+      return true;
+    }
+    return false;
+  }
+
+  if (Name == "__memset_chk") {
+    if (isFoldable(4, 3, false)) {
+      Value *Val = B.CreateIntCast(CI->getOperand(2), B.getInt8Ty(),
+                                   false);
+      EmitMemSet(CI->getOperand(1), Val,  CI->getOperand(3), B, TD);
+      replaceCall(CI->getOperand(1));
+      return true;
+    }
+    return false;
+  }
+
+  if (Name == "__strcpy_chk" || Name == "__stpcpy_chk") {
+    // If a) we don't have any length information, or b) we know this will
+    // fit then just lower to a plain st[rp]cpy. Otherwise we'll keep our
+    // st[rp]cpy_chk call which may fail at runtime if the size is too long.
+    // TODO: It might be nice to get a maximum length out of the possible
+    // string lengths for varying.
+    if (isFoldable(3, 2, true)) {
+      Value *Ret = EmitStrCpy(CI->getOperand(1), CI->getOperand(2), B, TD,
+                              Name.substr(2, 6));
+      replaceCall(Ret);
+      return true;
+    }
+    return false;
+  }
+
+  if (Name == "__strncpy_chk") {
+    if (isFoldable(4, 3, false)) {
+      Value *Ret = EmitStrNCpy(CI->getOperand(1), CI->getOperand(2),
+                               CI->getOperand(3), B, TD);
+      replaceCall(Ret);
+      return true;
+    }
+    return false;
+  }
+
+  if (Name == "__strcat_chk") {
+    return false;
+  }
+
+  if (Name == "__strncat_chk") {
+    return false;
+  }
+
+  return false;
+}




From baldrick at free.fr  Fri Mar 12 05:45:06 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 12 Mar 2010 11:45:06 -0000
Subject: [llvm-commits] [llvm] r98346 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
Message-ID: <20100312114506.A92ED2A6C12C@llvm.org>

Author: baldrick
Date: Fri Mar 12 05:45:06 2010
New Revision: 98346

URL: http://llvm.org/viewvc/llvm-project?rev=98346&view=rev
Log:
Fix PR6522: implement copysign expansion for x86 long double
(it seems that FreeBSD doesn't have copysignl).  Done by
removing a bunch of assumptions from the code.  This may also
help with sparc 128 bit floats.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp?rev=98346&r1=98345&r2=98346&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp Fri Mar 12 05:45:06 2010
@@ -1585,35 +1585,51 @@
   DebugLoc dl = Node->getDebugLoc();
   SDValue Tmp1 = Node->getOperand(0);
   SDValue Tmp2 = Node->getOperand(1);
-  assert((Tmp2.getValueType() == MVT::f32 ||
-          Tmp2.getValueType() == MVT::f64) &&
-          "Ugly special-cased code!");
-  // Get the sign bit of the RHS.
+
+  // Get the sign bit of the RHS.  First obtain a value that has the same
+  // sign as the sign bit, i.e. negative if and only if the sign bit is 1.
   SDValue SignBit;
-  EVT IVT = Tmp2.getValueType() == MVT::f64 ? MVT::i64 : MVT::i32;
+  EVT FloatVT = Tmp2.getValueType();
+  EVT IVT = EVT::getIntegerVT(*DAG.getContext(), FloatVT.getSizeInBits());
   if (isTypeLegal(IVT)) {
+    // Convert to an integer with the same sign bit.
     SignBit = DAG.getNode(ISD::BIT_CONVERT, dl, IVT, Tmp2);
   } else {
-    assert(isTypeLegal(TLI.getPointerTy()) &&
-            (TLI.getPointerTy() == MVT::i32 || 
-            TLI.getPointerTy() == MVT::i64) &&
-            "Legal type for load?!");
-    SDValue StackPtr = DAG.CreateStackTemporary(Tmp2.getValueType());
-    SDValue StorePtr = StackPtr, LoadPtr = StackPtr;
+    // Store the float to memory, then load the sign part out as an integer.
+    MVT LoadTy = TLI.getPointerTy();
+    // First create a temporary that is aligned for both the load and store.
+    SDValue StackPtr = DAG.CreateStackTemporary(FloatVT, LoadTy);
+    // Then store the float to it.
     SDValue Ch =
-      DAG.getStore(DAG.getEntryNode(), dl, Tmp2, StorePtr, NULL, 0,
+      DAG.getStore(DAG.getEntryNode(), dl, Tmp2, StackPtr, NULL, 0,
                    false, false, 0);
-    if (Tmp2.getValueType() == MVT::f64 && TLI.isLittleEndian())
-      LoadPtr = DAG.getNode(ISD::ADD, dl, StackPtr.getValueType(),
-                            LoadPtr, DAG.getIntPtrConstant(4));
-    SignBit = DAG.getExtLoad(ISD::SEXTLOAD, dl, TLI.getPointerTy(),
-                             Ch, LoadPtr, NULL, 0, MVT::i32,
-                             false, false, 0);
-  }
-  SignBit =
-      DAG.getSetCC(dl, TLI.getSetCCResultType(SignBit.getValueType()),
-                    SignBit, DAG.getConstant(0, SignBit.getValueType()),
-                    ISD::SETLT);
+    if (TLI.isBigEndian()) {
+      assert(FloatVT.isByteSized() && "Unsupported floating point type!");
+      // Load out a legal integer with the same sign bit as the float.
+      SignBit = DAG.getLoad(LoadTy, dl, Ch, StackPtr, NULL, 0, false, false, 0);
+    } else { // Little endian
+      SDValue LoadPtr = StackPtr;
+      // The float may be wider than the integer we are going to load.  Advance
+      // the pointer so that the loaded integer will contain the sign bit.
+      unsigned Strides = (FloatVT.getSizeInBits()-1)/LoadTy.getSizeInBits();
+      unsigned ByteOffset = (Strides * LoadTy.getSizeInBits()) / 8;
+      LoadPtr = DAG.getNode(ISD::ADD, dl, LoadPtr.getValueType(),
+                            LoadPtr, DAG.getIntPtrConstant(ByteOffset));
+      // Load a legal integer containing the sign bit.
+      SignBit = DAG.getLoad(LoadTy, dl, Ch, LoadPtr, NULL, 0, false, false, 0);
+      // Move the sign bit to the top bit of the loaded integer.
+      unsigned BitShift = LoadTy.getSizeInBits() -
+        (FloatVT.getSizeInBits() - 8 * ByteOffset);
+      assert(BitShift < LoadTy.getSizeInBits() && "Pointer advanced wrong?");
+      if (BitShift)
+        SignBit = DAG.getNode(ISD::SHL, dl, LoadTy, SignBit,
+                              DAG.getConstant(BitShift,TLI.getShiftAmountTy()));
+    }
+  }
+  // Now get the sign bit proper, by seeing whether the value is negative.
+  SignBit = DAG.getSetCC(dl, TLI.getSetCCResultType(SignBit.getValueType()),
+                         SignBit, DAG.getConstant(0, SignBit.getValueType()),
+                         ISD::SETLT);
   // Get the absolute value of the result.
   SDValue AbsVal = DAG.getNode(ISD::FABS, dl, Tmp1.getValueType(), Tmp1);
   // Select between the nabs and abs value based on the sign bit of




From baldrick at free.fr  Fri Mar 12 06:13:59 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 12 Mar 2010 12:13:59 -0000
Subject: [llvm-commits] [llvm] r98348 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
Message-ID: <20100312121359.DF9932A6C12C@llvm.org>

Author: baldrick
Date: Fri Mar 12 06:13:59 2010
New Revision: 98348

URL: http://llvm.org/viewvc/llvm-project?rev=98348&view=rev
Log:
Now that it's supported, turn copysignl into a COPYSIGN node.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=98348&r1=98347&r2=98348&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Fri Mar 12 06:13:59 2010
@@ -4633,7 +4633,7 @@
     // can't be a library call.
     if (!F->hasLocalLinkage() && F->hasName()) {
       StringRef Name = F->getName();
-      if (Name == "copysign" || Name == "copysignf") {
+      if (Name == "copysign" || Name == "copysignf" || Name == "copysignl") {
         if (I.getNumOperands() == 3 &&   // Basic sanity checks.
             I.getOperand(1)->getType()->isFloatingPointTy() &&
             I.getType() == I.getOperand(1)->getType() &&




From benny.kra at googlemail.com  Fri Mar 12 07:54:59 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Fri, 12 Mar 2010 13:54:59 -0000
Subject: [llvm-commits] [llvm] r98349 -
	/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
Message-ID: <20100312135500.055722A6C12C@llvm.org>

Author: d0k
Date: Fri Mar 12 07:54:59 2010
New Revision: 98349

URL: http://llvm.org/viewvc/llvm-project?rev=98349&view=rev
Log:
Use StringRef::substr instead of std::string::substr to avoid using a free'd
string temporary. This should fix PR6590.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98349&r1=98348&r2=98349&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Mar 12 07:54:59 2010
@@ -9837,7 +9837,8 @@
         AsmPieces[2] == "${0:w}" &&
         IA->getConstraintString().compare(0, 5, "=r,0,") == 0) {
       AsmPieces.clear();
-      SplitString(IA->getConstraintString().substr(5), AsmPieces, ",");
+      const std::string &Constraints = IA->getConstraintString();
+      SplitString(StringRef(Constraints).substr(5), AsmPieces, ",");
       std::sort(AsmPieces.begin(), AsmPieces.end());
       if (AsmPieces.size() == 4 &&
           AsmPieces[0] == "~{cc}" &&




From kovarththanan.rajaratnam at gmail.com  Fri Mar 12 08:17:25 2010
From: kovarththanan.rajaratnam at gmail.com (Kovarththanan Rajaratnam)
Date: Fri, 12 Mar 2010 14:17:25 -0000
Subject: [llvm-commits] [llvm] r98350 - /llvm/trunk/lib/System/Win32/Path.inc
Message-ID: <20100312141725.200C12A6C12C@llvm.org>

Author: krj
Date: Fri Mar 12 08:17:24 2010
New Revision: 98350

URL: http://llvm.org/viewvc/llvm-project?rev=98350&view=rev
Log:
Remove superfluous NULL assignment

Modified:
    llvm/trunk/lib/System/Win32/Path.inc

Modified: llvm/trunk/lib/System/Win32/Path.inc
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Win32/Path.inc?rev=98350&r1=98349&r2=98350&view=diff
==============================================================================
--- llvm/trunk/lib/System/Win32/Path.inc (original)
+++ llvm/trunk/lib/System/Win32/Path.inc Fri Mar 12 08:17:24 2010
@@ -126,7 +126,7 @@
 }
 
 void Path::makeAbsolute() {
-  TCHAR  FullPath[MAX_PATH + 1] = {0}; 
+  TCHAR  FullPath[MAX_PATH + 1] = {0};
   LPTSTR FilePart = NULL;
 
   DWORD RetLength = ::GetFullPathNameA(path.c_str(),
@@ -161,7 +161,7 @@
   }
 }
 
-bool 
+bool
 Path::isAbsolute() const {
   // FIXME: This does not handle correctly an absolute path starting from
   // a drive letter or in UNC format.
@@ -174,9 +174,9 @@
     default:
       return path[0] == '/' || (path[1] == ':' && path[2] == '/');
   }
-} 
+}
 
-static Path *TempDirectory = NULL;
+static Path *TempDirectory;
 
 Path
 Path::GetTemporaryDirectory(std::string* ErrMsg) {
@@ -266,7 +266,7 @@
 Path::GetCurrentDirectory() {
   char pathname[MAX_PATH];
   ::GetCurrentDirectoryA(MAX_PATH,pathname);
-  return Path(pathname);  
+  return Path(pathname);
 }
 
 /// GetMainExecutable - Return the path to the main executable, given the
@@ -448,7 +448,7 @@
     MakeErrMsg(ErrMsg, path + ": can't get status of file");
     return true;
   }
-    
+
   if (!(fi.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)) {
     if (ErrMsg)
       *ErrMsg = path + ": not a directory";
@@ -617,7 +617,7 @@
       *next = 0;
       if (!CreateDirectory(pathname, NULL) &&
           GetLastError() != ERROR_ALREADY_EXISTS)
-          return MakeErrMsg(ErrMsg, 
+          return MakeErrMsg(ErrMsg,
             std::string(pathname) + ": Can't create directory: ");
       *next++ = '/';
     }
@@ -649,7 +649,7 @@
   WIN32_FILE_ATTRIBUTE_DATA fi;
   if (!GetFileAttributesEx(path.c_str(), GetFileExInfoStandard, &fi))
     return true;
-    
+
   if (fi.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) {
     // If it doesn't exist, we're done.
     if (!exists())
@@ -706,7 +706,7 @@
 
     pathname[lastchar] = 0;
     if (!RemoveDirectory(pathname))
-      return MakeErrMsg(ErrStr, 
+      return MakeErrMsg(ErrStr,
         std::string(pathname) + ": Can't destroy directory: ");
     return false;
   } else {
@@ -753,7 +753,7 @@
 bool
 Path::renamePathOnDisk(const Path& newName, std::string* ErrMsg) {
   if (!MoveFileEx(path.c_str(), newName.c_str(), MOVEFILE_REPLACE_EXISTING))
-    return MakeErrMsg(ErrMsg, "Can't move '" + path + "' to '" + newName.path 
+    return MakeErrMsg(ErrMsg, "Can't move '" + path + "' to '" + newName.path
         + "': ");
   return false;
 }
@@ -764,7 +764,7 @@
   if (!si.isFile) {
     return true;
   }
-  
+
   HANDLE h = CreateFile(path.c_str(),
                         FILE_READ_ATTRIBUTES | FILE_WRITE_ATTRIBUTES,
                         FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE,




From baldrick at free.fr  Fri Mar 12 10:14:49 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 12 Mar 2010 16:14:49 -0000
Subject: [llvm-commits] [dragonegg] r98351 -
	/dragonegg/trunk/llvm-backend.cpp
Message-ID: <20100312161449.D788F2A6C12C@llvm.org>

Author: baldrick
Date: Fri Mar 12 10:14:49 2010
New Revision: 98351

URL: http://llvm.org/viewvc/llvm-project?rev=98351&view=rev
Log:
Add notes for passes that have not been disabled.

Modified:
    dragonegg/trunk/llvm-backend.cpp

Modified: dragonegg/trunk/llvm-backend.cpp
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-backend.cpp?rev=98351&r1=98350&r2=98351&view=diff
==============================================================================
--- dragonegg/trunk/llvm-backend.cpp (original)
+++ dragonegg/trunk/llvm-backend.cpp Fri Mar 12 10:14:49 2010
@@ -2344,6 +2344,11 @@
     // Could just set optimize to zero (after taking a copy), but this would
     // also impact front-end optimizations.
 
+    // Leave pass_inline_parameters.  Otherwise our vector lowering fails since
+    // immediates have not been propagated into builtin callsites.
+
+    // Leave pass_ipa_function_and_variable_visibility.  Needed for correctness.
+
     // Turn off pass_ipa_early_inline.
     pass_info.pass = &pass_simple_ipa_null.pass;
     pass_info.reference_pass_name = "einline_ipa";
@@ -2351,13 +2356,44 @@
     pass_info.pos_op = PASS_POS_REPLACE;
     register_callback (plugin_name, PLUGIN_PASS_MANAGER_SETUP, NULL, &pass_info);
 
-    // Turn off pass_all_early_optimizations.
+    // Leave pass_ipa_free_lang_data.
+
+    // Leave pass pass_early_local_passes::pass_fixup_cfg. ???
+
+    // Leave pass pass_early_local_passes::pass_tree_profile.
+
+    // Leave pass_early_local_passes::pass_cleanup_cfg. ???
+
+    // Leave pass_early_local_passes::pass_init_datastructures. ???
+
+    // Leave pass_early_local_passes::pass_expand_omp.
+
+    // Leave pass_early_local_passes::pass_referenced_vars. ???
+
+    // Leave pass_early_local_passes::pass_build_ssa.
+
+    // Leave pass_early_local_passes::pass_early_warn_uninitialized.
+
+    // Leave pass_early_local_passes::pass_rebuild_cgraph_edges. ???
+
+    // Leave pass_early_local_passes::pass_early_inline.  Otherwise our vector
+    // lowering fails since immediates have not been propagated into builtin
+    // callsites.
+
+    // Turn off pass_early_local_passes::pass_all_early_optimizations.
     pass_info.pass = &pass_gimple_null.pass;
     pass_info.reference_pass_name = "early_optimizations";
     pass_info.ref_pass_instance_number = 0;
     pass_info.pos_op = PASS_POS_REPLACE;
     register_callback (plugin_name, PLUGIN_PASS_MANAGER_SETUP, NULL, &pass_info);
 
+    // Leave pass_early_local_passes::pass_release_ssa_names. ???
+
+    // Leave pass_early_local_passes::pass_rebuild_cgraph_edges. ???
+
+    // Leave pass_inline_parameters.  Otherwise our vector lowering fails since
+    // immediates have not been propagated into builtin callsites.
+
     // Turn off pass_ipa_increase_alignment.
     pass_info.pass = &pass_simple_ipa_null.pass;
     pass_info.reference_pass_name = "increase_alignment";
@@ -2372,6 +2408,8 @@
     pass_info.pos_op = PASS_POS_REPLACE;
     register_callback (plugin_name, PLUGIN_PASS_MANAGER_SETUP, NULL, &pass_info);
 
+    // Leave pass_ipa_whole_program_visibility. ???
+
     // Turn off pass_ipa_cp.
     pass_info.pass = &pass_ipa_null.pass;
     pass_info.reference_pass_name = "cp";




From criswell at uiuc.edu  Fri Mar 12 11:09:19 2010
From: criswell at uiuc.edu (John Criswell)
Date: Fri, 12 Mar 2010 17:09:19 -0000
Subject: [llvm-commits] [poolalloc] r98353 -
	/poolalloc/trunk/include/dsa/DSGraphTraits.h
Message-ID: <20100312170919.BF7052A6C12C@llvm.org>

Author: criswell
Date: Fri Mar 12 11:09:19 2010
New Revision: 98353

URL: http://llvm.org/viewvc/llvm-project?rev=98353&view=rev
Log:
Switched from using the old LLVM 2.6 llvm/ADT/iterator.h header file to
directly defining our iterator class in terms of the standard C++ iterator
class.  This change was modeled on how llvm/ADT/iterator.h defined
forward_iterator.
This fixes some compilation problems with LLVM 2.7 (which eliminated the
forward_iterator class).

Modified:
    poolalloc/trunk/include/dsa/DSGraphTraits.h

Modified: poolalloc/trunk/include/dsa/DSGraphTraits.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/dsa/DSGraphTraits.h?rev=98353&r1=98352&r2=98353&view=diff
==============================================================================
--- poolalloc/trunk/include/dsa/DSGraphTraits.h (original)
+++ poolalloc/trunk/include/dsa/DSGraphTraits.h Fri Mar 12 11:09:19 2010
@@ -18,13 +18,14 @@
 
 #include "dsa/DSGraph.h"
 #include "llvm/ADT/GraphTraits.h"
-#include "llvm/ADT/iterator.h"
 #include "llvm/ADT/STLExtras.h"
 
+#include <iterator>
+
 namespace llvm {
 
 template<typename NodeTy>
-class DSNodeIterator : public forward_iterator<const DSNode, ptrdiff_t> {
+class DSNodeIterator : public std::iterator<std::forward_iterator_tag, const DSNode, ptrdiff_t> {
   friend class DSNode;
   NodeTy * const Node;
   unsigned Offset;




From criswell at uiuc.edu  Fri Mar 12 11:10:10 2010
From: criswell at uiuc.edu (John Criswell)
Date: Fri, 12 Mar 2010 17:10:10 -0000
Subject: [llvm-commits] [poolalloc] r98354 -
 /poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp
Message-ID: <20100312171010.354FE2A6C12C@llvm.org>

Author: criswell
Date: Fri Mar 12 11:10:10 2010
New Revision: 98354

URL: http://llvm.org/viewvc/llvm-project?rev=98354&view=rev
Log:
Switched to using stdint.h instead of LLVM's DataTypes.h header file.
The DataTypes.h header files was removed from LLVM 2.7.

Modified:
    poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp

Modified: poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp?rev=98354&r1=98353&r2=98354&view=diff
==============================================================================
--- poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp (original)
+++ poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp Fri Mar 12 11:10:10 2010
@@ -16,10 +16,9 @@
 #include "PageManager.h"
 #include "PoolSlab.h"
 #include <assert.h>
-#include <stdlib.h>
+#include <stdint.h>
 #include <stdio.h>
-
-#include "llvm/Support/DataTypes.h"
+#include <stdlib.h>
 
 //===----------------------------------------------------------------------===//
 //




From criswell at uiuc.edu  Fri Mar 12 11:10:49 2010
From: criswell at uiuc.edu (John Criswell)
Date: Fri, 12 Mar 2010 17:10:49 -0000
Subject: [llvm-commits] [poolalloc] r98355 - in /poolalloc/trunk/lib/DSA:
 DSGraph.cpp DataStructure.cpp
Message-ID: <20100312171049.4E9C02A6C12C@llvm.org>

Author: criswell
Date: Fri Mar 12 11:10:49 2010
New Revision: 98355

URL: http://llvm.org/viewvc/llvm-project?rev=98355&view=rev
Log:
Removed incomplete command-line options that break compilation.

Modified:
    poolalloc/trunk/lib/DSA/DSGraph.cpp
    poolalloc/trunk/lib/DSA/DataStructure.cpp

Modified: poolalloc/trunk/lib/DSA/DSGraph.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DSGraph.cpp?rev=98355&r1=98354&r2=98355&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DSGraph.cpp (original)
+++ poolalloc/trunk/lib/DSA/DSGraph.cpp Fri Mar 12 11:10:49 2010
@@ -42,7 +42,6 @@
   STATISTIC (NumDNE            , "Number of nodes removed by reachability");
   STATISTIC (NumTrivialDNE     , "Number of nodes trivially removed");
   STATISTIC (NumTrivialGlobalDNE, "Number of globals trivially removed");
-  static cl::opt<unsigned>
 }
 
 /// getFunctionNames - Return a space separated list of the name of the

Modified: poolalloc/trunk/lib/DSA/DataStructure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DataStructure.cpp?rev=98355&r1=98354&r2=98355&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DataStructure.cpp Fri Mar 12 11:10:49 2010
@@ -40,7 +40,6 @@
 namespace {
   STATISTIC (NumFolds, "Number of nodes completely folded");
   STATISTIC (NumNodeAllocated  , "Number of nodes allocated");
-  static cl::opt<unsigned>
 }
 
 /// isForwarding - Return true if this NodeHandle is forwarding to another




From baldrick at free.fr  Fri Mar 12 11:41:34 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 12 Mar 2010 17:41:34 -0000
Subject: [llvm-commits] [llvm] r98357 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
Message-ID: <20100312174134.683382A6C12C@llvm.org>

Author: baldrick
Date: Fri Mar 12 11:41:34 2010
New Revision: 98357

URL: http://llvm.org/viewvc/llvm-project?rev=98357&view=rev
Log:
Revert turning copysignl into a COPYSIGN node for the moment:
ppc calls copysignl with a 128 bit ppc long double, resulting
in a node that the type legalizer doesn't know how to expand.

Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=98357&r1=98356&r2=98357&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Fri Mar 12 11:41:34 2010
@@ -4633,7 +4633,7 @@
     // can't be a library call.
     if (!F->hasLocalLinkage() && F->hasName()) {
       StringRef Name = F->getName();
-      if (Name == "copysign" || Name == "copysignf" || Name == "copysignl") {
+      if (Name == "copysign" || Name == "copysignf") {
         if (I.getNumOperands() == 3 &&   // Basic sanity checks.
             I.getOperand(1)->getType()->isFloatingPointTy() &&
             I.getType() == I.getOperand(1)->getType() &&




From benny.kra at googlemail.com  Fri Mar 12 11:43:55 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Fri, 12 Mar 2010 18:43:55 +0100
Subject: [llvm-commits] [poolalloc] r98354 -
	/poolalloc/trunk/runtime/FreeListAllocator/PoolAllocator.cpp
In-Reply-To: <20100312171010.354FE2A6C12C@llvm.org>
References: <20100312171010.354FE2A6C12C@llvm.org>
Message-ID: <68432C81-DB2B-4875-9A5E-7E386C48ABFC@gmail.com>


On 12.03.2010, at 18:10, John Criswell wrote:

> Author: criswell
> Date: Fri Mar 12 11:10:10 2010
> New Revision: 98354
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98354&view=rev
> Log:
> Switched to using stdint.h instead of LLVM's DataTypes.h header file.
> The DataTypes.h header files was removed from LLVM 2.7.

It was actually moved from llvm/Support to llvm/System. But I guess it doesn't
matter unless you want to support non-c99 compilers.




From jyasskin at google.com  Fri Mar 12 11:45:06 2010
From: jyasskin at google.com (Jeffrey Yasskin)
Date: Fri, 12 Mar 2010 17:45:06 -0000
Subject: [llvm-commits] [llvm] r98358 - in /llvm/trunk:
 include/llvm/ADT/STLExtras.h lib/CodeGen/AsmPrinter/DIE.h
 lib/CodeGen/AsmPrinter/DwarfDebug.cpp lib/CodeGen/AsmPrinter/DwarfDebug.h
Message-ID: <20100312174506.AD81E2A6C12C@llvm.org>

Author: jyasskin
Date: Fri Mar 12 11:45:06 2010
New Revision: 98358

URL: http://llvm.org/viewvc/llvm-project?rev=98358&view=rev
Log:
Free DbgScopes in DwarfDebug::endFunction().  Also increased the const-ness of
several fields to make it easier to figure out where bugs might be creeping in.

Modified:
    llvm/trunk/include/llvm/ADT/STLExtras.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h

Modified: llvm/trunk/include/llvm/ADT/STLExtras.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/STLExtras.h?rev=98358&r1=98357&r2=98358&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/STLExtras.h (original)
+++ llvm/trunk/include/llvm/ADT/STLExtras.h Fri Mar 12 11:45:06 2010
@@ -279,6 +279,28 @@
   qsort(&*Start, End-Start, sizeof(*Start), Compare);
 }
   
+//===----------------------------------------------------------------------===//
+//     Extra additions to <algorithm>
+//===----------------------------------------------------------------------===//
+
+/// For a container of pointers, deletes the pointers and then clears the
+/// container.
+template<typename Container>
+void DeleteContainerPointers(Container &C) {
+  for (typename Container::iterator I = C.begin(), E = C.end(); I != E; ++I)
+    delete *I;
+  C.clear();
+}
+
+/// In a container of pairs (usually a map) whose second element is a pointer,
+/// deletes the second elements and then clears the container.
+template<typename Container>
+void DeleteContainerSeconds(Container &C) {
+  for (typename Container::iterator I = C.begin(), E = C.end(); I != E; ++I)
+    delete I->second;
+  C.clear();
+}
+
 } // End llvm namespace
 
 #endif

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h?rev=98358&r1=98357&r2=98358&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DIE.h Fri Mar 12 11:45:06 2010
@@ -112,7 +112,6 @@
   //===--------------------------------------------------------------------===//
   /// DIE - A structured debug information entry.  Has an abbreviation which
   /// describes it's organization.
-  class CompileUnit;
   class DIEValue;
 
   class DIE {
@@ -159,7 +158,6 @@
     void setTag(unsigned Tag) { Abbrev.setTag(Tag); }
     void setOffset(unsigned O) { Offset = O; }
     void setSize(unsigned S) { Size = S; }
-    void setParent(DIE *P) { Parent = P; }
     
     /// addValue - Add a value and attributes to a DIE.
     ///
@@ -185,7 +183,7 @@
       }
       Abbrev.setChildrenFlag(dwarf::DW_CHILDREN_yes);
       Children.push_back(Child);
-      Child->setParent(this);
+      Child->Parent = this;
     }
 
 #ifndef NDEBUG
@@ -393,12 +391,11 @@
   /// this class can also be used as a proxy for a debug information entry not
   /// yet defined (ie. types.)
   class DIEEntry : public DIEValue {
-    DIE *Entry;
+    DIE *const Entry;
   public:
     explicit DIEEntry(DIE *E) : DIEValue(isEntry), Entry(E) {}
 
     DIE *getEntry() const { return Entry; }
-    void setEntry(DIE *E) { Entry = E; }
 
     /// EmitValue - Emit debug information entry offset.
     ///

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=98358&r1=98357&r2=98358&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Fri Mar 12 11:45:06 2010
@@ -25,6 +25,7 @@
 #include "llvm/Target/TargetFrameInfo.h"
 #include "llvm/Target/TargetLoweringObjectFile.h"
 #include "llvm/Target/TargetRegisterInfo.h"
+#include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
@@ -52,7 +53,7 @@
 
   /// Die - Compile unit debug information entry.
   ///
-  OwningPtr<DIE> CUDie;
+  const OwningPtr<DIE> CUDie;
 
   /// IndexTyDie - An anonymous type for index type.  Owned by CUDie.
   DIE *IndexTyDie;
@@ -147,16 +148,16 @@
 class DbgVariable {
   DIVariable Var;                    // Variable Descriptor.
   unsigned FrameIndex;               // Variable frame index.
-  DbgVariable *AbstractVar;          // Abstract variable for this variable.
+  DbgVariable *const AbstractVar;    // Abstract variable for this variable.
   DIE *TheDIE;
 public:
-  DbgVariable(DIVariable V, unsigned I)
-    : Var(V), FrameIndex(I), AbstractVar(0), TheDIE(0)  {}
+  // AbsVar may be NULL.
+  DbgVariable(DIVariable V, unsigned I, DbgVariable *AbsVar)
+    : Var(V), FrameIndex(I), AbstractVar(AbsVar), TheDIE(0)  {}
 
   // Accessors.
   DIVariable getVariable()           const { return Var; }
   unsigned getFrameIndex()           const { return FrameIndex; }
-  void setAbstractVariable(DbgVariable *V) { AbstractVar = V; }
   DbgVariable *getAbstractVariable() const { return AbstractVar; }
   void setDIE(DIE *D)                      { TheDIE = D; }
   DIE *getDIE()                      const { return TheDIE; }
@@ -175,8 +176,10 @@
   MCSymbol *EndLabel;                 // Label ID of the end of scope.
   const MachineInstr *LastInsn;       // Last instruction of this scope.
   const MachineInstr *FirstInsn;      // First instruction of this scope.
-  SmallVector<DbgScope *, 4> Scopes;  // Scopes defined in scope.
-  SmallVector<DbgVariable *, 8> Variables;// Variables declared in scope.
+  // Scopes defined in scope.  Contents not owned.
+  SmallVector<DbgScope *, 4> Scopes;
+  // Variables declared in scope.  Contents owned.
+  SmallVector<DbgVariable *, 8> Variables;
 
   // Private state for dump()
   mutable unsigned IndentLevel;
@@ -197,8 +200,8 @@
   MDNode *getScopeNode()         const { return Desc.getNode(); }
   MCSymbol *getStartLabel()      const { return StartLabel; }
   MCSymbol *getEndLabel()        const { return EndLabel; }
-  SmallVector<DbgScope *, 4> &getScopes() { return Scopes; }
-  SmallVector<DbgVariable *, 8> &getVariables() { return Variables; }
+  const SmallVector<DbgScope *, 4> &getScopes() { return Scopes; }
+  const SmallVector<DbgVariable *, 8> &getVariables() { return Variables; }
   void setStartLabel(MCSymbol *S) { StartLabel = S; }
   void setEndLabel(MCSymbol *E)   { EndLabel = E; }
   void setLastInsn(const MachineInstr *MI) { LastInsn = MI; }
@@ -221,14 +224,14 @@
     assert (getFirstInsn() && "First instruction is missing!");
     
     // Use the end of last child scope as end of this scope.
-    SmallVector<DbgScope *, 4> &Scopes = getScopes();
+    const SmallVector<DbgScope *, 4> &Scopes = getScopes();
     const MachineInstr *LastInsn = getFirstInsn();
     unsigned LIndex = 0;
     if (Scopes.empty()) {
       assert (getLastInsn() && "Inner most scope does not have last insn!");
       return;
     }
-    for (SmallVector<DbgScope *, 4>::iterator SI = Scopes.begin(),
+    for (SmallVector<DbgScope *, 4>::const_iterator SI = Scopes.begin(),
            SE = Scopes.end(); SI != SE; ++SI) {
       DbgScope *DS = *SI;
       DS->fixInstructionMarkers(MIIndexMap);
@@ -280,8 +283,6 @@
 #endif
 
 DbgScope::~DbgScope() {
-  for (unsigned i = 0, N = Scopes.size(); i < N; ++i)
-    delete Scopes[i];
   for (unsigned j = 0, M = Variables.size(); j < M; ++j)
     delete Variables[j];
 }
@@ -819,14 +820,13 @@
     return;
   }
 
-  // Set up proxy.
-  Entry = createDIEEntry();
-  ModuleCU->insertDIEEntry(Ty.getNode(), Entry);
-
   // Construct type.
   DIE *Buffer = getOrCreateTypeDIE(Ty);
 
-  Entry->setEntry(Buffer);
+  // Set up proxy.
+  Entry = createDIEEntry(Buffer);
+  ModuleCU->insertDIEEntry(Ty.getNode(), Entry);
+
   Entity->addValue(dwarf::DW_AT_type, dwarf::DW_FORM_ref4, Entry);
 }
 
@@ -1549,7 +1549,7 @@
   }
   
   // Add variables to scope.
-  SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
+  const SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
     DIE *VariableDIE = constructVariableDIE(Variables[i], Scope);
     if (VariableDIE)
@@ -1557,7 +1557,7 @@
   }
 
   // Add nested scopes.
-  SmallVector<DbgScope *, 4> &Scopes = Scope->getScopes();
+  const SmallVector<DbgScope *, 4> &Scopes = Scope->getScopes();
   for (unsigned j = 0, M = Scopes.size(); j < M; ++j) {
     // Define the Scope debug information entry.
     DIE *NestedDIE = constructScopeDIE(Scopes[j]);
@@ -1917,7 +1917,8 @@
   if (!Scope)
     return NULL;
 
-  AbsDbgVariable = new DbgVariable(Var, FrameIdx);
+  AbsDbgVariable = new DbgVariable(Var, FrameIdx,
+                                   NULL /* No more-abstract variable*/);
   Scope->addVariable(AbsDbgVariable);
   AbstractVariables[Var.getNode()] = AbsDbgVariable;
   return AbsDbgVariable;
@@ -1944,11 +1945,9 @@
     if (!Scope)
       continue;
 
-    DbgVariable *RegVar = new DbgVariable(DV, VP.first);
+    DbgVariable *AbsDbgVariable = findAbstractVariable(DV, VP.first, ScopeLoc);
+    DbgVariable *RegVar = new DbgVariable(DV, VP.first, AbsDbgVariable);
     Scope->addVariable(RegVar);
-    if (DbgVariable *AbsDbgVariable = findAbstractVariable(DV, VP.first,
-                                                           ScopeLoc))
-      RegVar->setAbstractVariable(AbsDbgVariable);
   }
 }
 
@@ -2067,9 +2066,9 @@
   while (!WorkList.empty()) {
     DbgScope *S = WorkList.back(); WorkList.pop_back();
 
-    SmallVector<DbgScope *, 4> &Children = S->getScopes();
+    const SmallVector<DbgScope *, 4> &Children = S->getScopes();
     if (!Children.empty()) 
-      for (SmallVector<DbgScope *, 4>::iterator SI = Children.begin(),
+      for (SmallVector<DbgScope *, 4>::const_iterator SI = Children.begin(),
              SE = Children.end(); SI != SE; ++SI)
         WorkList.push_back(*SI);
 
@@ -2174,11 +2173,13 @@
 
   // Clear debug info
   CurrentFnDbgScope = NULL;
-  DbgScopeMap.clear();
+  DeleteContainerSeconds(DbgScopeMap);
   DbgScopeBeginMap.clear();
   DbgScopeEndMap.clear();
   ConcreteScopes.clear();
+  DeleteContainerSeconds(AbstractScopes);
   AbstractScopesList.clear();
+  AbstractVariables.clear();
   Lines.clear();
   
   if (TimePassesIsEnabled)

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=98358&r1=98357&r2=98358&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Fri Mar 12 11:45:06 2010
@@ -126,7 +126,8 @@
   //
   DbgScope *CurrentFnDbgScope;
   
-  /// DbgScopeMap - Tracks the scopes in the current function.
+  /// DbgScopeMap - Tracks the scopes in the current function.  Owns the
+  /// contained DbgScope*s.
   ///
   DenseMap<MDNode *, DbgScope *> DbgScopeMap;
 
@@ -135,11 +136,12 @@
   DenseMap<MDNode *, DbgScope *> ConcreteScopes;
 
   /// AbstractScopes - Tracks the abstract scopes a module. These scopes are
-  /// not included DbgScopeMap.
+  /// not included DbgScopeMap.  AbstractScopes owns its DbgScope*s.
   DenseMap<MDNode *, DbgScope *> AbstractScopes;
   SmallVector<DbgScope *, 4>AbstractScopesList;
 
-  /// AbstractVariables - Collection on abstract variables.
+  /// AbstractVariables - Collection on abstract variables.  Owned by the
+  /// DbgScopes in AbstractScopes.
   DenseMap<MDNode *, DbgVariable *> AbstractVariables;
 
   /// InliendSubprogramDIEs - Collection of subprgram DIEs that are marked
@@ -225,7 +227,7 @@
 
   /// createDIEEntry - Creates a new DIEEntry to be a proxy for a debug
   /// information entry.
-  DIEEntry *createDIEEntry(DIE *Entry = NULL);
+  DIEEntry *createDIEEntry(DIE *Entry);
 
   /// addUInt - Add an unsigned integer attribute data and value.
   ///




From baldrick at free.fr  Fri Mar 12 11:55:21 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 12 Mar 2010 17:55:21 -0000
Subject: [llvm-commits] [llvm] r98359 - in /llvm/trunk:
 lib/Analysis/ConstantFolding.cpp test/Transforms/InstCombine/gepgep.ll
Message-ID: <20100312175521.204942A6C12C@llvm.org>

Author: baldrick
Date: Fri Mar 12 11:55:20 2010
New Revision: 98359

URL: http://llvm.org/viewvc/llvm-project?rev=98359&view=rev
Log:
When constant folding GEP of GEP, do not crash if an index of
the inner GEP is not a ConstantInt.

Added:
    llvm/trunk/test/Transforms/InstCombine/gepgep.ll
Modified:
    llvm/trunk/lib/Analysis/ConstantFolding.cpp

Modified: llvm/trunk/lib/Analysis/ConstantFolding.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ConstantFolding.cpp?rev=98359&r1=98358&r2=98359&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ConstantFolding.cpp (original)
+++ llvm/trunk/lib/Analysis/ConstantFolding.cpp Fri Mar 12 11:55:20 2010
@@ -589,15 +589,28 @@
   APInt Offset = APInt(BitWidth,
                        TD->getIndexedOffset(Ptr->getType(),
                                             (Value**)Ops+1, NumOps-1));
+  Ptr = cast<Constant>(Ptr->stripPointerCasts());
 
   // If this is a GEP of a GEP, fold it all into a single GEP.
   while (GEPOperator *GEP = dyn_cast<GEPOperator>(Ptr)) {
     SmallVector<Value *, 4> NestedOps(GEP->op_begin()+1, GEP->op_end());
+
+    // Do not try the incorporate the sub-GEP if some index is not a number.
+    bool AllConstantInt = true;
+    for (unsigned i = 0, e = NestedOps.size(); i != e; ++i)
+      if (!isa<ConstantInt>(NestedOps[i])) {
+        AllConstantInt = false;
+        break;
+      }
+    if (!AllConstantInt)
+      break;
+
     Ptr = cast<Constant>(GEP->getOperand(0));
     Offset += APInt(BitWidth,
                     TD->getIndexedOffset(Ptr->getType(),
                                          (Value**)NestedOps.data(),
                                          NestedOps.size()));
+    Ptr = cast<Constant>(Ptr->stripPointerCasts());
   }
 
   // If the base value for this address is a literal integer value, fold the
@@ -611,7 +624,6 @@
   // we eliminate over-indexing of the notional static type array bounds.
   // This makes it easy to determine if the getelementptr is "inbounds".
   // Also, this helps GlobalOpt do SROA on GlobalVariables.
-  Ptr = cast<Constant>(Ptr->stripPointerCasts());
   const Type *Ty = Ptr->getType();
   SmallVector<Constant*, 32> NewIdxs;
   do {

Added: llvm/trunk/test/Transforms/InstCombine/gepgep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/gepgep.ll?rev=98359&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/gepgep.ll (added)
+++ llvm/trunk/test/Transforms/InstCombine/gepgep.ll Fri Mar 12 11:55:20 2010
@@ -0,0 +1,13 @@
+; RUN: opt < %s -instcombine -disable-output
+
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at buffer = external global [64 x float]
+
+declare void @use(i8*)
+
+define void @f() {
+  call void @use(i8* getelementptr (i8* getelementptr (i8* bitcast ([64 x float]* @buffer to i8*), i64 and (i64 sub (i64 0, i64 ptrtoint ([64 x float]* @buffer to i64)), i64 63)), i64 64))
+  ret void
+}




From sabre at nondot.org  Fri Mar 12 12:10:35 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:10:35 -0000
Subject: [llvm-commits] [llvm] r98360 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
Message-ID: <20100312181035.3D6C92A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:10:35 2010
New Revision: 98360

URL: http://llvm.org/viewvc/llvm-project?rev=98360&view=rev
Log:
fix a bug emitting .secrel32 that I introduced, PR6587, patch
by A.Mazur!

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp?rev=98360&r1=98359&r2=98360&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfPrinter.cpp Fri Mar 12 12:10:35 2010
@@ -225,10 +225,11 @@
     return EmitDifference(Label, Section, IsSmall);
   
   // On COFF targets, we have to emit the weird .secrel32 directive.
-  if (const char *SecOffDir = MAI->getDwarfSectionOffsetDirective())
+  if (const char *SecOffDir = MAI->getDwarfSectionOffsetDirective()) {
     // FIXME: MCize.
     Asm->O << SecOffDir << Label->getName();
-  else {
+    Asm->OutStreamer.AddBlankLine();
+  } else {
     unsigned Size = IsSmall ? 4 : TD->getPointerSize();
     Asm->OutStreamer.EmitSymbolValue(Label, Size, 0/*AddrSpace*/);
   }




From sabre at nondot.org  Fri Mar 12 12:28:53 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:28:53 -0000
Subject: [llvm-commits] [llvm] r98361 - in /llvm/trunk:
 include/llvm/MC/MCStreamer.h lib/CodeGen/LLVMTargetMachine.cpp
 lib/MC/MCAsmStreamer.cpp tools/llvm-mc/llvm-mc.cpp
Message-ID: <20100312182853.658CE2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:28:53 2010
New Revision: 98361

URL: http://llvm.org/viewvc/llvm-project?rev=98361&view=rev
Log:
remove MAI argument from createAsmStreamer since it
can get it from the context now.

Modified:
    llvm/trunk/include/llvm/MC/MCStreamer.h
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/MC/MCAsmStreamer.cpp
    llvm/trunk/tools/llvm-mc/llvm-mc.cpp

Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=98361&r1=98360&r2=98361&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Fri Mar 12 12:28:53 2010
@@ -299,8 +299,7 @@
   /// \param ShowInst - Whether to show the MCInst representation inline with
   /// the assembly.
   MCStreamer *createAsmStreamer(MCContext &Ctx, formatted_raw_ostream &OS,
-                                const MCAsmInfo &MAI, bool isLittleEndian,
-                                bool isVerboseAsm,
+                                bool isLittleEndian, bool isVerboseAsm,
                                 MCInstPrinter *InstPrint = 0,
                                 MCCodeEmitter *CE = 0,
                                 bool ShowInst = false);

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98361&r1=98360&r2=98361&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Fri Mar 12 12:28:53 2010
@@ -131,7 +131,7 @@
   case CGFT_AssemblyFile: {
     MCInstPrinter *InstPrinter =
       getTarget().createMCInstPrinter(MAI.getAssemblerDialect(), MAI, Out);
-    AsmStreamer.reset(createAsmStreamer(*Context, Out, MAI,
+    AsmStreamer.reset(createAsmStreamer(*Context, Out,
                                         getTargetData()->isLittleEndian(),
                                         getVerboseAsm(), InstPrinter,
                                         /*codeemitter*/0));

Modified: llvm/trunk/lib/MC/MCAsmStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAsmStreamer.cpp?rev=98361&r1=98360&r2=98361&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAsmStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCAsmStreamer.cpp Fri Mar 12 12:28:53 2010
@@ -41,11 +41,10 @@
 
 public:
   MCAsmStreamer(MCContext &Context, formatted_raw_ostream &os,
-                const MCAsmInfo &mai,
                 bool isLittleEndian, bool isVerboseAsm, MCInstPrinter *printer,
                 MCCodeEmitter *emitter, bool showInst)
-    : MCStreamer(Context), OS(os), MAI(mai), InstPrinter(printer),
-      Emitter(emitter), CommentStream(CommentToEmit),
+    : MCStreamer(Context), OS(os), MAI(Context.getAsmInfo()),
+      InstPrinter(printer), Emitter(emitter), CommentStream(CommentToEmit),
       IsLittleEndian(isLittleEndian), IsVerboseAsm(isVerboseAsm),
       ShowInst(showInst) {
     if (InstPrinter && IsVerboseAsm)
@@ -654,9 +653,9 @@
 
 MCStreamer *llvm::createAsmStreamer(MCContext &Context,
                                     formatted_raw_ostream &OS,
-                                    const MCAsmInfo &MAI, bool isLittleEndian,
+                                    bool isLittleEndian,
                                     bool isVerboseAsm, MCInstPrinter *IP,
                                     MCCodeEmitter *CE, bool ShowInst) {
-  return new MCAsmStreamer(Context, OS, MAI, isLittleEndian, isVerboseAsm,
+  return new MCAsmStreamer(Context, OS, isLittleEndian, isVerboseAsm,
                            IP, CE, ShowInst);
 }

Modified: llvm/trunk/tools/llvm-mc/llvm-mc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-mc/llvm-mc.cpp?rev=98361&r1=98360&r2=98361&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-mc/llvm-mc.cpp (original)
+++ llvm/trunk/tools/llvm-mc/llvm-mc.cpp Fri Mar 12 12:28:53 2010
@@ -270,8 +270,7 @@
     IP.reset(TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI, *Out));
     if (ShowEncoding)
       CE.reset(TheTarget->createCodeEmitter(*TM, Ctx));
-    Str.reset(createAsmStreamer(Ctx, *Out, *MAI,
-                                TM->getTargetData()->isLittleEndian(),
+    Str.reset(createAsmStreamer(Ctx, *Out,TM->getTargetData()->isLittleEndian(),
                                 /*asmverbose*/true, IP.get(), CE.get(),
                                 ShowInst));
   } else {




From sabre at nondot.org  Fri Mar 12 12:44:54 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:44:54 -0000
Subject: [llvm-commits] [llvm] r98363 - in /llvm/trunk:
 include/llvm/MC/MCContext.h include/llvm/Target/Mangler.h
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/ELFWriter.cpp
 lib/CodeGen/LLVMTargetMachine.cpp lib/Target/CBackend/CBackend.cpp
 lib/Target/Mangler.cpp tools/lto/LTOCodeGenerator.cpp
 tools/lto/LTOModule.cpp
Message-ID: <20100312184454.4B4622A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:44:54 2010
New Revision: 98363

URL: http://llvm.org/viewvc/llvm-project?rev=98363&view=rev
Log:
make the mangler take an MCContext instead of an MAI.
No functionality change.

Modified:
    llvm/trunk/include/llvm/MC/MCContext.h
    llvm/trunk/include/llvm/Target/Mangler.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/ELFWriter.cpp
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/Target/CBackend/CBackend.cpp
    llvm/trunk/lib/Target/Mangler.cpp
    llvm/trunk/tools/lto/LTOCodeGenerator.cpp
    llvm/trunk/tools/lto/LTOModule.cpp

Modified: llvm/trunk/include/llvm/MC/MCContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCContext.h?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCContext.h (original)
+++ llvm/trunk/include/llvm/MC/MCContext.h Fri Mar 12 12:44:54 2010
@@ -48,7 +48,7 @@
     /// objects.
     BumpPtrAllocator Allocator;
   public:
-    MCContext(const MCAsmInfo &MAI);
+    explicit MCContext(const MCAsmInfo &MAI);
     ~MCContext();
     
     const MCAsmInfo &getAsmInfo() const { return MAI; }

Modified: llvm/trunk/include/llvm/Target/Mangler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Mangler.h?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Mangler.h (original)
+++ llvm/trunk/include/llvm/Target/Mangler.h Fri Mar 12 12:44:54 2010
@@ -23,7 +23,7 @@
 class Value;
 class GlobalValue;
 template <typename T> class SmallVectorImpl; 
-class MCAsmInfo;
+class MCContext;
 
 class Mangler {
 public:
@@ -34,7 +34,7 @@
   };
 
 private:
-  const MCAsmInfo &MAI;
+  const MCContext &Context;
 
   /// AnonGlobalIDs - We need to give global values the same name every time
   /// they are mangled.  This keeps track of the number we give to anonymous
@@ -47,9 +47,7 @@
   unsigned NextAnonGlobalID;
 
 public:
-  // Mangler ctor - if a prefix is specified, it will be prepended onto all
-  // symbols.
-  Mangler(const MCAsmInfo &mai) : MAI(mai), NextAnonGlobalID(1) {}
+  Mangler(const MCContext &context) : Context(context), NextAnonGlobalID(1) {}
 
   /// getNameWithPrefix - Fill OutName with the name of the appropriate prefix
   /// and the specified global variable's name.  If the global variable doesn't

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Fri Mar 12 12:44:54 2010
@@ -104,7 +104,7 @@
   const_cast<TargetLoweringObjectFile&>(getObjFileLowering())
     .Initialize(OutContext, TM);
   
-  Mang = new Mangler(*MAI);
+  Mang = new Mangler(OutContext);
   
   // Allow the target to emit any magic that it wants at the start of the file.
   EmitStartOfAsmFile(M);

Modified: llvm/trunk/lib/CodeGen/ELFWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/ELFWriter.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/ELFWriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/ELFWriter.cpp Fri Mar 12 12:44:54 2010
@@ -109,7 +109,7 @@
   // Initialize TargetLoweringObjectFile.
   const_cast<TargetLoweringObjectFile&>(TLOF).Initialize(OutContext, TM);
   
-  Mang = new Mangler(*MAI);
+  Mang = new Mangler(OutContext);
 
   // ELF Header
   // ----------

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Fri Mar 12 12:44:54 2010
@@ -101,14 +101,12 @@
 
 // Set the default code model for the JIT for a generic target.
 // FIXME: Is small right here? or .is64Bit() ? Large : Small?
-void
-LLVMTargetMachine::setCodeModelForJIT() {
+void LLVMTargetMachine::setCodeModelForJIT() {
   setCodeModel(CodeModel::Small);
 }
 
 // Set the default code model for static compilation for a generic target.
-void
-LLVMTargetMachine::setCodeModelForStatic() {
+void LLVMTargetMachine::setCodeModelForStatic() {
   setCodeModel(CodeModel::Small);
 }
 

Modified: llvm/trunk/lib/Target/CBackend/CBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CBackend/CBackend.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/lib/Target/CBackend/CBackend.cpp (original)
+++ llvm/trunk/lib/Target/CBackend/CBackend.cpp Fri Mar 12 12:44:54 2010
@@ -36,6 +36,7 @@
 #include "llvm/Target/Mangler.h"
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCContext.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetRegistry.h"
@@ -95,6 +96,7 @@
     LoopInfo *LI;
     const Module *TheModule;
     const MCAsmInfo* TAsm;
+    MCContext *TCtx;
     const TargetData* TD;
     std::map<const Type *, std::string> TypeNames;
     std::map<const ConstantFP *, unsigned> FPConstantMap;
@@ -1731,7 +1733,8 @@
     TAsm = Match->createAsmInfo(Triple);
 #endif    
   TAsm = new CBEMCAsmInfo();
-  Mang = new Mangler(*TAsm);
+  TCtx = new MCContext(*TAsm);
+  Mang = new Mangler(*TCtx);
 
   // Keep track of which functions are static ctors/dtors so they can have
   // an attribute added to their prototypes.

Modified: llvm/trunk/lib/Target/Mangler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mangler.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mangler.cpp (original)
+++ llvm/trunk/lib/Target/Mangler.cpp Fri Mar 12 12:44:54 2010
@@ -14,6 +14,7 @@
 #include "llvm/Target/Mangler.h"
 #include "llvm/GlobalValue.h"
 #include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCContext.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/Twine.h"
 using namespace llvm;
@@ -59,11 +60,10 @@
 /// appendMangledName - Add the specified string in mangled form if it uses
 /// any unusual characters.
 static void appendMangledName(SmallVectorImpl<char> &OutName, StringRef Str,
-                              const MCAsmInfo *MAI) {
+                              const MCAsmInfo &MAI) {
   // The first character is not allowed to be a number unless the target
   // explicitly allows it.
-  if ((MAI == 0 || !MAI->doesAllowNameToStartWithDigit()) &&
-      Str[0] >= '0' && Str[0] <= '9') {
+  if (!MAI.doesAllowNameToStartWithDigit() && Str[0] >= '0' && Str[0] <= '9') {
     MangleLetter(OutName, Str[0]);
     Str = Str.substr(1);
   }
@@ -100,6 +100,8 @@
   StringRef Name = GVName.toStringRef(TmpData);
   assert(!Name.empty() && "getNameWithPrefix requires non-empty name");
   
+  const MCAsmInfo &MAI = Context.getAsmInfo();
+  
   // If the global name is not led with \1, add the appropriate prefixes.
   if (Name[0] == '\1') {
     Name = Name.substr(1);
@@ -134,7 +136,7 @@
   // On systems that do not allow quoted names, we need to mangle most
   // strange characters.
   if (!MAI.doesAllowQuotesInName())
-    return appendMangledName(OutName, Name, &MAI);
+    return appendMangledName(OutName, Name, MAI);
   
   // Okay, the system allows quoted strings.  We can quote most anything, the
   // only characters that need escaping are " and \n.

Modified: llvm/trunk/tools/lto/LTOCodeGenerator.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/LTOCodeGenerator.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/tools/lto/LTOCodeGenerator.cpp (original)
+++ llvm/trunk/tools/lto/LTOCodeGenerator.cpp Fri Mar 12 12:44:54 2010
@@ -27,24 +27,25 @@
 #include "llvm/Analysis/LoopPass.h"
 #include "llvm/Analysis/Verifier.h"
 #include "llvm/Bitcode/ReaderWriter.h"
-#include "llvm/Support/CommandLine.h"
-#include "llvm/Support/FormattedStream.h"
-#include "llvm/Support/MemoryBuffer.h"
-#include "llvm/Support/StandardPasses.h"
-#include "llvm/Support/SystemUtils.h"
-#include "llvm/System/Host.h"
-#include "llvm/System/Program.h"
-#include "llvm/System/Signals.h"
+#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCContext.h"
 #include "llvm/Target/Mangler.h"
 #include "llvm/Target/SubtargetFeature.h"
 #include "llvm/Target/TargetOptions.h"
-#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetRegistry.h"
 #include "llvm/Target/TargetSelect.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/Scalar.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/FormattedStream.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/StandardPasses.h"
+#include "llvm/Support/SystemUtils.h"
+#include "llvm/System/Host.h"
+#include "llvm/System/Program.h"
+#include "llvm/System/Signals.h"
 #include "llvm/Config/config.h"
 #include <cstdlib>
 #include <unistd.h>
@@ -252,7 +253,8 @@
             args.push_back(arch);
         }
         // add -static to assembler command line when code model requires
-        if ( (_assemblerPath != NULL) && (_codeModel == LTO_CODEGEN_PIC_MODEL_STATIC) )
+        if ( (_assemblerPath != NULL) &&
+            (_codeModel == LTO_CODEGEN_PIC_MODEL_STATIC) )
             args.push_back("-static");
     }
     if ( needsCompilerOptions ) {
@@ -303,44 +305,44 @@
 
         // construct LTModule, hand over ownership of module and target
         const std::string FeatureStr =
-            SubtargetFeatures::getDefaultSubtargetFeatures(llvm::Triple(Triple));
+           SubtargetFeatures::getDefaultSubtargetFeatures(llvm::Triple(Triple));
         _target = march->createTargetMachine(Triple, FeatureStr);
     }
     return false;
 }
 
-void LTOCodeGenerator::applyScopeRestrictions()
-{
-    if ( !_scopeRestrictionsDone ) {
-        Module* mergedModule = _linker.getModule();
+void LTOCodeGenerator::applyScopeRestrictions() {
+  if (_scopeRestrictionsDone) return;
+  Module *mergedModule = _linker.getModule();
+
+  // Start off with a verification pass.
+  PassManager passes;
+  passes.add(createVerifierPass());
 
-        // Start off with a verification pass.
-        PassManager passes;
-        passes.add(createVerifierPass());
-
-        // mark which symbols can not be internalized 
-        if ( !_mustPreserveSymbols.empty() ) {
-            Mangler mangler(*_target->getMCAsmInfo());
-            std::vector<const char*> mustPreserveList;
-            for (Module::iterator f = mergedModule->begin(), 
-                                        e = mergedModule->end(); f != e; ++f) {
-                if ( !f->isDeclaration() 
-                  && _mustPreserveSymbols.count(mangler.getNameWithPrefix(f)) )
-                  mustPreserveList.push_back(::strdup(f->getNameStr().c_str()));
-            }
-            for (Module::global_iterator v = mergedModule->global_begin(), 
-                                 e = mergedModule->global_end(); v !=  e; ++v) {
-                if ( !v->isDeclaration()
-                  && _mustPreserveSymbols.count(mangler.getNameWithPrefix(v)) )
-                  mustPreserveList.push_back(::strdup(v->getNameStr().c_str()));
-            }
-            passes.add(createInternalizePass(mustPreserveList));
-        }
-        // apply scope restrictions
-        passes.run(*mergedModule);
-        
-        _scopeRestrictionsDone = true;
+  // mark which symbols can not be internalized 
+  if (!_mustPreserveSymbols.empty()) {
+    MCContext Context(*_target->getMCAsmInfo());
+    Mangler mangler(Context);
+    std::vector<const char*> mustPreserveList;
+    for (Module::iterator f = mergedModule->begin(),
+         e = mergedModule->end(); f != e; ++f) {
+      if (!f->isDeclaration() &&
+          _mustPreserveSymbols.count(mangler.getNameWithPrefix(f)))
+        mustPreserveList.push_back(::strdup(f->getNameStr().c_str()));
+    }
+    for (Module::global_iterator v = mergedModule->global_begin(), 
+         e = mergedModule->global_end(); v !=  e; ++v) {
+      if (v->isDeclaration() &&
+          _mustPreserveSymbols.count(mangler.getNameWithPrefix(v)))
+        mustPreserveList.push_back(::strdup(v->getNameStr().c_str()));
     }
+    passes.add(createInternalizePass(mustPreserveList));
+  }
+  
+  // apply scope restrictions
+  passes.run(*mergedModule);
+  
+  _scopeRestrictionsDone = true;
 }
 
 /// Optimize merged modules using various IPO passes

Modified: llvm/trunk/tools/lto/LTOModule.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/LTOModule.cpp?rev=98363&r1=98362&r2=98363&view=diff
==============================================================================
--- llvm/trunk/tools/lto/LTOModule.cpp (original)
+++ llvm/trunk/tools/lto/LTOModule.cpp Fri Mar 12 12:44:54 2010
@@ -1,4 +1,4 @@
-//===-LTOModule.cpp - LLVM Link Time Optimizer ----------------------------===//
+//===-- LTOModule.cpp - LLVM Link Time Optimizer --------------------------===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -29,6 +29,7 @@
 #include "llvm/Target/Mangler.h"
 #include "llvm/Target/SubtargetFeature.h"
 #include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCContext.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetRegistry.h"
 #include "llvm/Target/TargetSelect.h"
@@ -437,7 +438,8 @@
         _symbolsParsed = true;
         
         // Use mangler to add GlobalPrefix to names to match linker names.
-        Mangler mangler(*_target->getMCAsmInfo());
+        MCContext Context(*_target->getMCAsmInfo());
+        Mangler mangler(Context);
 
         // add functions
         for (Module::iterator f = _module->begin(); f != _module->end(); ++f) {




From sabre at nondot.org  Fri Mar 12 12:47:50 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:47:50 -0000
Subject: [llvm-commits] [llvm] r98364 - in /llvm/trunk:
 lib/Target/X86/AsmPrinter/X86MCInstLower.cpp test/CodeGen/X86/pic.ll
Message-ID: <20100312184750.E3A4B2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:47:50 2010
New Revision: 98364

URL: http://llvm.org/viewvc/llvm-project?rev=98364&view=rev
Log:
simplify code to use OutContext.GetOrCreateTemporarySymbol with
no arguments instead of having to come up with a unique name.
This also makes the code less fragile.

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
    llvm/trunk/test/CodeGen/X86/pic.ll

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98364&r1=98363&r2=98364&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Fri Mar 12 12:47:50 2010
@@ -406,11 +406,8 @@
     // For this, we want to print something like:
     //   MYGLOBAL + (. - PICBASE)
     // However, we can't generate a ".", so just emit a new label here and refer
-    // to it.  We know that this operand flag occurs at most once per function.
-    const char *Prefix = MAI->getPrivateGlobalPrefix();
-    MCSymbol *DotSym = OutContext.GetOrCreateTemporarySymbol(Twine(Prefix)+
-                                                             "picbaseref" +
-                                                    Twine(getFunctionNumber()));
+    // to it.
+    MCSymbol *DotSym = OutContext.GetOrCreateTemporarySymbol();
     OutStreamer.EmitLabel(DotSym);
     
     // Now that we have emitted the label, lower the complex operand expression.

Modified: llvm/trunk/test/CodeGen/X86/pic.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pic.ll?rev=98364&r1=98363&r2=98364&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/pic.ll (original)
+++ llvm/trunk/test/CodeGen/X86/pic.ll Fri Mar 12 12:47:50 2010
@@ -15,7 +15,7 @@
 ; LINUX:	call	.L1$pb
 ; LINUX-NEXT: .L1$pb:
 ; LINUX-NEXT:	popl
-; LINUX:	addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref1-.L1$pb),
+; LINUX:	addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L1$pb),
 ; LINUX:	movl	dst at GOT(%eax),
 ; LINUX:	movl	ptr at GOT(%eax),
 ; LINUX:	movl	src at GOT(%eax),
@@ -37,7 +37,7 @@
 ; LINUX:	call	.L2$pb
 ; LINUX-NEXT: .L2$pb:
 ; LINUX-NEXT:	popl
-; LINUX:	addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref2-.L2$pb), %eax
+; LINUX:	addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L2$pb), %eax
 ; LINUX:	movl	dst2 at GOT(%eax),
 ; LINUX:	movl	ptr2 at GOT(%eax),
 ; LINUX:	movl	src2 at GOT(%eax),
@@ -57,7 +57,7 @@
 ; LINUX-NEXT: 	call	.L3$pb
 ; LINUX-NEXT: .L3$pb:
 ; LINUX-NEXT: 	popl	%ebx
-; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref3-.L3$pb), %ebx
+; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L3$pb), %ebx
 ; LINUX: 	movl	$40, (%esp)
 ; LINUX: 	call	malloc at PLT
 ; LINUX: 	addl	$8, %esp
@@ -78,7 +78,7 @@
 ; LINUX: 	call	.L4$pb
 ; LINUX-NEXT: .L4$pb:
 ; LINUX: 	popl
-; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref4-.L4$pb),
+; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L4$pb),
 ; LINUX: 	movl	pfoo at GOT(%esi),
 ; LINUX: 	call	afoo at PLT
 ; LINUX: 	call	*
@@ -93,7 +93,7 @@
 ; LINUX: test5:
 ; LINUX: call	.L5$pb
 ; LINUX: popl	%ebx
-; LINUX: addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref5-.L5$pb), %ebx
+; LINUX: addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L5$pb), %ebx
 ; LINUX: call	foo at PLT
 }
 
@@ -115,7 +115,7 @@
 ; LINUX: 	call	.L6$pb
 ; LINUX-NEXT: .L6$pb:
 ; LINUX-NEXT: 	popl	%eax
-; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref6-.L6$pb), %eax
+; LINUX: 	addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L6$pb), %eax
 ; LINUX: 	leal	dst6 at GOTOFF(%eax), %ecx
 ; LINUX: 	movl	%ecx, ptr6 at GOTOFF(%eax)
 ; LINUX: 	movl	src6 at GOTOFF(%eax), %ecx
@@ -136,7 +136,7 @@
 ; LINUX: test7:
 ; LINUX:    call .L7$pb
 ; LINUX: .L7$pb:
-; LINUX:    addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref7-.L7$pb), 
+; LINUX:    addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L7$pb), 
 ; LINUX:    fldl	.LCPI7_0 at GOTOFF(
 }
 
@@ -188,7 +188,7 @@
 ; LINUX: test8:
 ; LINUX:   call	.L8$pb
 ; LINUX: .L8$pb:
-; LINUX:   addl	$_GLOBAL_OFFSET_TABLE_+(.Lpicbaseref8-.L8$pb),
+; LINUX:   addl	$_GLOBAL_OFFSET_TABLE_+(.L{{.*}}-.L8$pb),
 ; LINUX:   addl	.LJTI8_0 at GOTOFF(
 ; LINUX:   jmpl	*
 




From sabre at nondot.org  Fri Mar 12 12:49:32 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:49:32 -0000
Subject: [llvm-commits] [llvm] r98365 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfException.cpp DwarfException.h
Message-ID: <20100312184932.EB7B62A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:49:32 2010
New Revision: 98365

URL: http://llvm.org/viewvc/llvm-project?rev=98365&view=rev
Log:
remove dead code.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.h

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=98365&r1=98364&r2=98365&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Fri Mar 12 12:49:32 2010
@@ -50,25 +50,6 @@
   delete ExceptionTimer;
 }
 
-/// CreateLabelDiff - Emit a label and subtract it from the expression we
-/// already have.  This is equivalent to emitting "foo - .", but we have to emit
-/// the label for "." directly.
-const MCExpr *DwarfException::CreateLabelDiff(const MCExpr *ExprRef,
-                                              const char *LabelName,
-                                              unsigned Index) {
-  SmallString<64> Name;
-  raw_svector_ostream(Name) << MAI->getPrivateGlobalPrefix()
-                            << LabelName << Asm->getFunctionNumber()
-                            << "_" << Index;
-  MCSymbol *DotSym = Asm->OutContext.GetOrCreateTemporarySymbol(Name.str());
-  Asm->OutStreamer.EmitLabel(DotSym);
-
-  return MCBinaryExpr::CreateSub(ExprRef,
-                                 MCSymbolRefExpr::Create(DotSym,
-                                                         Asm->OutContext),
-                                 Asm->OutContext);
-}
-
 /// EmitCIE - Emit a Common Information Entry (CIE). This holds information that
 /// is shared among many Frame Description Entries.  There is at least one CIE
 /// in every non-empty .debug_frame section.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.h?rev=98365&r1=98364&r2=98365&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.h Fri Mar 12 12:49:32 2010
@@ -169,11 +169,6 @@
                             const SmallVectorImpl<unsigned> &FirstActions);
   void EmitExceptionTable();
 
-  /// CreateLabelDiff - Emit a label and subtract it from the expression we
-  /// already have.  This is equivalent to emitting "foo - .", but we have to
-  /// emit the label for "." directly.
-  const MCExpr *CreateLabelDiff(const MCExpr *ExprRef, const char *LabelName,
-                                unsigned Index);
 public:
   //===--------------------------------------------------------------------===//
   // Main entry points.




From sabre at nondot.org  Fri Mar 12 12:55:20 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 18:55:20 -0000
Subject: [llvm-commits] [llvm] r98366 - in /llvm/trunk:
 include/llvm/Target/Mangler.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/Target/Mangler.cpp lib/Target/TargetLoweringObjectFile.cpp
Message-ID: <20100312185520.90EDB2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 12:55:20 2010
New Revision: 98366

URL: http://llvm.org/viewvc/llvm-project?rev=98366&view=rev
Log:
finally give Mangler a getSymbol method, which returns an MCSymbol
for a global instead of messing around with string buffers.

Modified:
    llvm/trunk/include/llvm/Target/Mangler.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/Target/Mangler.cpp
    llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp

Modified: llvm/trunk/include/llvm/Target/Mangler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Mangler.h?rev=98366&r1=98365&r2=98366&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Mangler.h (original)
+++ llvm/trunk/include/llvm/Target/Mangler.h Fri Mar 12 12:55:20 2010
@@ -1,4 +1,4 @@
-//===-- llvm/Target/Mangler.h - Self-contained name mangler ----*- C++ -*-===//
+//===-- llvm/Target/Mangler.h - Self-contained name mangler -----*- C++ -*-===//
 //
 //                     The LLVM Compiler Infrastructure
 //
@@ -24,6 +24,7 @@
 class GlobalValue;
 template <typename T> class SmallVectorImpl; 
 class MCContext;
+class MCSymbol;
 
 class Mangler {
 public:
@@ -34,7 +35,7 @@
   };
 
 private:
-  const MCContext &Context;
+  MCContext &Context;
 
   /// AnonGlobalIDs - We need to give global values the same name every time
   /// they are mangled.  This keeps track of the number we give to anonymous
@@ -47,8 +48,13 @@
   unsigned NextAnonGlobalID;
 
 public:
-  Mangler(const MCContext &context) : Context(context), NextAnonGlobalID(1) {}
+  Mangler(MCContext &context) : Context(context), NextAnonGlobalID(1) {}
 
+  /// getSymbol - Return the MCSymbol for the specified global value.  This
+  /// symbol is the main label that is the address of the global.
+  MCSymbol *getSymbol(const GlobalValue *GV);
+
+  
   /// getNameWithPrefix - Fill OutName with the name of the appropriate prefix
   /// and the specified global variable's name.  If the global variable doesn't
   /// have a name, this fills in a unique name for the global.

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=98366&r1=98365&r2=98366&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Fri Mar 12 12:55:20 2010
@@ -1632,12 +1632,7 @@
 /// GetGlobalValueSymbol - Return the MCSymbol for the specified global
 /// value.
 MCSymbol *AsmPrinter::GetGlobalValueSymbol(const GlobalValue *GV) const {
-  SmallString<60> NameStr;
-  Mang->getNameWithPrefix(NameStr, GV, false);
-  
-  if (!GV->hasPrivateLinkage())
-    return OutContext.GetOrCreateSymbol(NameStr.str());
-  return OutContext.GetOrCreateTemporarySymbol(NameStr.str());
+  return Mang->getSymbol(GV);
 }
 
 /// GetSymbolWithGlobalValueBase - Return the MCSymbol for a symbol with

Modified: llvm/trunk/lib/Target/Mangler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Mangler.cpp?rev=98366&r1=98365&r2=98366&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Mangler.cpp (original)
+++ llvm/trunk/lib/Target/Mangler.cpp Fri Mar 12 12:55:20 2010
@@ -179,3 +179,16 @@
   getNameWithPrefix(Buf, GV, isImplicitlyPrivate);
   return std::string(Buf.begin(), Buf.end());
 }
+
+/// getSymbol - Return the MCSymbol for the specified global value.  This
+/// symbol is the main label that is the address of the global.
+MCSymbol *Mangler::getSymbol(const GlobalValue *GV) {
+  SmallString<60> NameStr;
+  getNameWithPrefix(NameStr, GV, false);
+  if (!GV->hasPrivateLinkage())
+    return Context.GetOrCreateSymbol(NameStr.str());
+  
+  return Context.GetOrCreateTemporarySymbol(NameStr.str());
+}
+
+

Modified: llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp?rev=98366&r1=98365&r2=98366&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp (original)
+++ llvm/trunk/lib/Target/TargetLoweringObjectFile.cpp Fri Mar 12 12:55:20 2010
@@ -298,16 +298,7 @@
 getExprForDwarfGlobalReference(const GlobalValue *GV, Mangler *Mang,
                                MachineModuleInfo *MMI, unsigned Encoding,
                                MCStreamer &Streamer) const {
-  // FIXME: Use GetGlobalValueSymbol.
-  SmallString<128> Name;
-  Mang->getNameWithPrefix(Name, GV, false);
-  const MCSymbol *Sym;
-  
-  if (GV->hasPrivateLinkage())
-    Sym = getContext().GetOrCreateTemporarySymbol(Name.str());
-  else
-    Sym = getContext().GetOrCreateSymbol(Name.str());
-
+  const MCSymbol *Sym = Mang->getSymbol(GV);
   return getExprForDwarfReference(Sym, Mang, MMI, Encoding, Streamer);
 }
 




From sabre at nondot.org  Fri Mar 12 13:04:15 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 19:04:15 -0000
Subject: [llvm-commits] [llvm] r98367 -
	/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
Message-ID: <20100312190415.255C52A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 13:04:14 2010
New Revision: 98367

URL: http://llvm.org/viewvc/llvm-project?rev=98367&view=rev
Log:
use Mang->getSymbol instead of duplicating the logic, reduce indentation.

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98367&r1=98366&r2=98367&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:04:14 2010
@@ -54,25 +54,19 @@
 }
 
 MCSymbol *X86AsmPrinter::GetGlobalValueSymbol(const GlobalValue *GV) const {
-  SmallString<60> NameStr;
-  Mang->getNameWithPrefix(NameStr, GV, false);
-  MCSymbol *Symb;
-  if (GV->hasPrivateLinkage())
-    Symb = OutContext.GetOrCreateTemporarySymbol(NameStr.str());
-  else
-    Symb = OutContext.GetOrCreateSymbol(NameStr.str());
-
-  if (Subtarget->isTargetCygMing()) {
-    X86COFFMachineModuleInfo &COFFMMI =
-      MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-    COFFMMI.DecorateCygMingName(Symb, OutContext, GV, *TM.getTargetData());
-
-    // Save function name for later type emission.
-    if (const Function *F = dyn_cast<Function>(GV))
-      if (F->isDeclaration())
-        COFFMMI.addExternalFunction(Symb->getName());
-
-  }
+  MCSymbol *Symb = Mang->getSymbol(GV);
+  
+  if (!Subtarget->isTargetCygMing() || !isa<Function>(GV))
+    return Symb;
+  
+  X86COFFMachineModuleInfo &COFFMMI =
+    MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
+  COFFMMI.DecorateCygMingName(Symb, OutContext, GV, *TM.getTargetData());
+
+  // Save function name for later type emission.
+  const Function *F = cast<Function>(GV);
+  if (F->isDeclaration())
+    COFFMMI.addExternalFunction(Symb->getName());
 
   return Symb;
 }




From sabre at nondot.org  Fri Mar 12 13:14:18 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 19:14:18 -0000
Subject: [llvm-commits] [llvm] r98368 -
	/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
Message-ID: <20100312191418.362752A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 13:14:18 2010
New Revision: 98368

URL: http://llvm.org/viewvc/llvm-project?rev=98368&view=rev
Log:
Remove some dead code.  This method only gets called on 
definitions.


Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98368&r1=98367&r2=98368&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:14:18 2010
@@ -63,11 +63,6 @@
     MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
   COFFMMI.DecorateCygMingName(Symb, OutContext, GV, *TM.getTargetData());
 
-  // Save function name for later type emission.
-  const Function *F = cast<Function>(GV);
-  if (F->isDeclaration())
-    COFFMMI.addExternalFunction(Symb->getName());
-
   return Symb;
 }
 




From clattner at apple.com  Fri Mar 12 13:15:50 2010
From: clattner at apple.com (Chris Lattner)
Date: Fri, 12 Mar 2010 11:15:50 -0800
Subject: [llvm-commits] [llvm] r98368 -
	/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
In-Reply-To: <20100312191418.362752A6C12C@llvm.org>
References: <20100312191418.362752A6C12C@llvm.org>
Message-ID: <E35952B6-24D5-4E0C-B478-7D7FC62D3C93@apple.com>

Anton, this means that this code is dead:

     // Emit type information for external functions
    for (X86COFFMachineModuleInfo::stub_iterator I = COFFMMI.stub_begin(),
           E = COFFMMI.stub_end(); I != E; ++I) {
      O << "\t.def\t " << I->getKeyData()
        << ";\t.scl\t" << COFF::C_EXT
        << ";\t.type\t" << (COFF::DT_FCN << COFF::N_BTSHFT)
        << ";\t.endef\n";
    }

Is this really true?  If so, we should remove the COFFMMI stub stuff.

-Chris

On Mar 12, 2010, at 11:14 AM, Chris Lattner wrote:

> Author: lattner
> Date: Fri Mar 12 13:14:18 2010
> New Revision: 98368
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98368&view=rev
> Log:
> Remove some dead code.  This method only gets called on 
> definitions.
> 
> 
> Modified:
>    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
> 
> Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98368&r1=98367&r2=98368&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
> +++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:14:18 2010
> @@ -63,11 +63,6 @@
>     MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
>   COFFMMI.DecorateCygMingName(Symb, OutContext, GV, *TM.getTargetData());
> 
> -  // Save function name for later type emission.
> -  const Function *F = cast<Function>(GV);
> -  if (F->isDeclaration())
> -    COFFMMI.addExternalFunction(Symb->getName());
> -
>   return Symb;
> }
> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From dpatel at apple.com  Fri Mar 12 13:18:30 2010
From: dpatel at apple.com (Devang Patel)
Date: Fri, 12 Mar 2010 19:18:30 -0000
Subject: [llvm-commits] [llvm] r98369 - in /llvm/trunk:
 lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
 test/DebugInfo/2010-03-12-llc-crash.ll
Message-ID: <20100312191830.E0FB42A6C12C@llvm.org>

Author: dpatel
Date: Fri Mar 12 13:18:30 2010
New Revision: 98369

URL: http://llvm.org/viewvc/llvm-project?rev=98369&view=rev
Log:
Fix llc crash on invalid input.


Added:
    llvm/trunk/test/DebugInfo/2010-03-12-llc-crash.ll
Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98369&r1=98368&r2=98369&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Fri Mar 12 13:18:30 2010
@@ -358,6 +358,13 @@
       } else
         printOperand(MI, 0);
     } else {
+      if (MI->getOperand(0).getType()==MachineOperand::MO_Register &&
+          MI->getOperand(0).getReg()==0) {
+        // Suppress offset in this case, it is not meaningful.
+        O << "undef";
+        OutStreamer.AddBlankLine();
+        return;
+      }
       // Frame address.  Currently handles register +- offset only.
       assert(MI->getOperand(0).getType()==MachineOperand::MO_Register);
       assert(MI->getOperand(3).getType()==MachineOperand::MO_Immediate);

Added: llvm/trunk/test/DebugInfo/2010-03-12-llc-crash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/DebugInfo/2010-03-12-llc-crash.ll?rev=98369&view=auto
==============================================================================
--- llvm/trunk/test/DebugInfo/2010-03-12-llc-crash.ll (added)
+++ llvm/trunk/test/DebugInfo/2010-03-12-llc-crash.ll Fri Mar 12 13:18:30 2010
@@ -0,0 +1,20 @@
+; RUN: llc -O0 < %s -o /dev/null
+; llc should not crash on this invalid input.
+; PR6588
+declare void @llvm.dbg.declare(metadata, metadata) nounwind readnone
+
+define void @foo() {
+entry:
+  call void @llvm.dbg.declare(metadata !{i32* undef}, metadata !0)
+  ret void
+}
+
+!0 = metadata !{i32 524545, metadata !1, metadata !"sy", metadata !2, i32 890, metadata !7} ; [ DW_TAG_arg_variable ]
+!1 = metadata !{i32 524334, i32 0, metadata !2, metadata !"foo", metadata !"foo", metadata !"foo", metadata !2, i32 892, metadata !4, i1 false, i1 true, i32 0, i32 0, null, i1 false} ; [ DW_TAG_subprogram ]
+!2 = metadata !{i32 524329, metadata !"qpainter.h", metadata !"QtGui", metadata !3} ; [ DW_TAG_file_type ]
+!3 = metadata !{i32 524305, i32 0, i32 4, metadata !"splineeditor.cpp", metadata !"editor", metadata !"clang 1.1", i1 true, i1 false, metadata !"", i32 0} ; [ DW_TAG_compile_unit ]
+!4 = metadata !{i32 524309, metadata !5, metadata !"", metadata !5, i32 0, i64 0, i64 0, i64 0, i32 0, null, metadata !6, i32 0, null} ; [ DW_TAG_subroutine_type ]
+!5 = metadata !{i32 524329, metadata !"splineeditor.cpp", metadata !"src", metadata !3} ; [ DW_TAG_file_type ]
+!6 = metadata !{null}
+!7 = metadata !{i32 524324, metadata !5, metadata !"int", metadata !5, i32 0, i64 32, i64 32, i64 0, i32 0, i32 5} ; [ DW_TAG_base_type ]
+




From isanbard at gmail.com  Fri Mar 12 13:20:40 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Fri, 12 Mar 2010 19:20:40 -0000
Subject: [llvm-commits] [llvm] r98370 - in /llvm/trunk/lib/Target/X86:
 AsmPrinter/X86AsmPrinter.cpp X86ISelLowering.cpp
Message-ID: <20100312192041.2EE2E2A6C12C@llvm.org>

Author: void
Date: Fri Mar 12 13:20:40 2010
New Revision: 98370

URL: http://llvm.org/viewvc/llvm-project?rev=98370&view=rev
Log:
Add a beta-test for placing the LSDA into the TEXT section on X86.

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98370&r1=98369&r2=98370&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:20:40 2010
@@ -523,10 +523,18 @@
         // L_foo$non_lazy_ptr:
         OutStreamer.EmitLabel(Stubs[i].first);
         // .indirect_symbol _foo
-        OutStreamer.EmitSymbolAttribute(Stubs[i].second.getPointer(),
+        MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
+        OutStreamer.EmitSymbolAttribute(MCSym.getPointer(),
                                         MCSA_IndirectSymbol);
         // .long 0
-        OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
+        if (MCSym.getInt())
+          // External to current translation unit.
+          OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
+        else
+          // Internal to current translation unit.
+          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym.getPointer(),
+                                                        OutContext),
+                                4/*size*/, 0/*addrspace*/);
       }
       Stubs.clear();
       OutStreamer.AddBlankLine();

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98370&r1=98369&r2=98370&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Mar 12 13:20:40 2010
@@ -37,6 +37,7 @@
 #include "llvm/CodeGen/PseudoSourceValue.h"
 #include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/ADT/BitVector.h"
 #include "llvm/ADT/SmallSet.h"
@@ -45,10 +46,12 @@
 #include "llvm/ADT/VectorExtras.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/Dwarf.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;
+using namespace dwarf;
 
 STATISTIC(NumTailCalls, "Number of tail calls");
 
@@ -67,10 +70,54 @@
 static SDValue getMOVL(SelectionDAG &DAG, DebugLoc dl, EVT VT, SDValue V1,
                        SDValue V2);
 
+// FIXME: This is for a test.
+static cl::opt<bool>
+EnableX86EHTest("enable-x86-eh-test", cl::Hidden);
+
+namespace llvm {
+  class X86_test_MachoTargetObjectFile : public TargetLoweringObjectFileMachO {
+  public:
+    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
+      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
+
+      // Exception Handling.
+      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
+                                    SectionKind::getReadOnlyWithRel());
+    }
+
+    virtual unsigned getTTypeEncoding() const {
+      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
+    }
+  };
+
+  class X8664_test_MachoTargetObjectFile : public X8664_MachoTargetObjectFile {
+  public:
+    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
+      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
+
+      // Exception Handling.
+      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
+                                    SectionKind::getReadOnlyWithRel());
+    }
+
+    virtual unsigned getTTypeEncoding() const {
+      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
+    }
+  };
+}
+
 static TargetLoweringObjectFile *createTLOF(X86TargetMachine &TM) {
   switch (TM.getSubtarget<X86Subtarget>().TargetType) {
   default: llvm_unreachable("unknown subtarget type");
   case X86Subtarget::isDarwin:
+    // FIXME: This is for an EH test.
+    if (EnableX86EHTest) {
+      if (TM.getSubtarget<X86Subtarget>().is64Bit())
+        return new X8664_test_MachoTargetObjectFile();
+      else
+        return new X86_test_MachoTargetObjectFile();
+    }
+
     if (TM.getSubtarget<X86Subtarget>().is64Bit())
       return new X8664_MachoTargetObjectFile();
     return new TargetLoweringObjectFileMachO();




From isanbard at gmail.com  Fri Mar 12 13:21:07 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Fri, 12 Mar 2010 19:21:07 -0000
Subject: [llvm-commits] [test-suite] r98371 -
	/test-suite/trunk/Makefile.programs
Message-ID: <20100312192107.A7FDB2A6C12C@llvm.org>

Author: void
Date: Fri Mar 12 13:21:07 2010
New Revision: 98371

URL: http://llvm.org/viewvc/llvm-project?rev=98371&view=rev
Log:
Enable beta test for x86 EH change.

Modified:
    test-suite/trunk/Makefile.programs

Modified: test-suite/trunk/Makefile.programs
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/Makefile.programs?rev=98371&r1=98370&r2=98371&view=diff
==============================================================================
--- test-suite/trunk/Makefile.programs (original)
+++ test-suite/trunk/Makefile.programs Fri Mar 12 13:21:07 2010
@@ -232,10 +232,10 @@
 LLCBETAOPTION := -sched=simple 
 endif
 ifeq ($(ARCH),x86_64)
-LLCBETAOPTION :=
+LLCBETAOPTION := -enable-x86-eh-test
 endif
 ifeq ($(ARCH),x86)
-LLCBETAOPTION :=
+LLCBETAOPTION := -enable-x86-eh-test
 endif
 ifeq ($(ARCH),Sparc)
 LLCBETAOPTION := -enable-sparc-v9-insts




From wendling at apple.com  Fri Mar 12 13:22:11 2010
From: wendling at apple.com (Bill Wendling)
Date: Fri, 12 Mar 2010 11:22:11 -0800
Subject: [llvm-commits] [llvm] r98370 - in /llvm/trunk/lib/Target/X86:
	AsmPrinter/X86AsmPrinter.cpp X86ISelLowering.cpp
In-Reply-To: <20100312192041.2EE2E2A6C12C@llvm.org>
References: <20100312192041.2EE2E2A6C12C@llvm.org>
Message-ID: <775087EA-5D58-49A5-A5AE-89C01E94CD42@apple.com>

This ugly code is temporary. It won't last beyond tomorrow.

-bw

On Mar 12, 2010, at 11:20 AM, Bill Wendling wrote:

> Author: void
> Date: Fri Mar 12 13:20:40 2010
> New Revision: 98370
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=98370&view=rev
> Log:
> Add a beta-test for placing the LSDA into the TEXT section on X86.
> 
> Modified:
>    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
>    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> 
> Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98370&r1=98369&r2=98370&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
> +++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:20:40 2010
> @@ -523,10 +523,18 @@
>         // L_foo$non_lazy_ptr:
>         OutStreamer.EmitLabel(Stubs[i].first);
>         // .indirect_symbol _foo
> -        OutStreamer.EmitSymbolAttribute(Stubs[i].second.getPointer(),
> +        MachineModuleInfoImpl::StubValueTy &MCSym = Stubs[i].second;
> +        OutStreamer.EmitSymbolAttribute(MCSym.getPointer(),
>                                         MCSA_IndirectSymbol);
>         // .long 0
> -        OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
> +        if (MCSym.getInt())
> +          // External to current translation unit.
> +          OutStreamer.EmitIntValue(0, 4/*size*/, 0/*addrspace*/);
> +        else
> +          // Internal to current translation unit.
> +          OutStreamer.EmitValue(MCSymbolRefExpr::Create(MCSym.getPointer(),
> +                                                        OutContext),
> +                                4/*size*/, 0/*addrspace*/);
>       }
>       Stubs.clear();
>       OutStreamer.AddBlankLine();
> 
> Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=98370&r1=98369&r2=98370&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Mar 12 13:20:40 2010
> @@ -37,6 +37,7 @@
> #include "llvm/CodeGen/PseudoSourceValue.h"
> #include "llvm/MC/MCAsmInfo.h"
> #include "llvm/MC/MCContext.h"
> +#include "llvm/MC/MCSectionMachO.h"
> #include "llvm/MC/MCSymbol.h"
> #include "llvm/ADT/BitVector.h"
> #include "llvm/ADT/SmallSet.h"
> @@ -45,10 +46,12 @@
> #include "llvm/ADT/VectorExtras.h"
> #include "llvm/Support/CommandLine.h"
> #include "llvm/Support/Debug.h"
> +#include "llvm/Support/Dwarf.h"
> #include "llvm/Support/ErrorHandling.h"
> #include "llvm/Support/MathExtras.h"
> #include "llvm/Support/raw_ostream.h"
> using namespace llvm;
> +using namespace dwarf;
> 
> STATISTIC(NumTailCalls, "Number of tail calls");
> 
> @@ -67,10 +70,54 @@
> static SDValue getMOVL(SelectionDAG &DAG, DebugLoc dl, EVT VT, SDValue V1,
>                        SDValue V2);
> 
> +// FIXME: This is for a test.
> +static cl::opt<bool>
> +EnableX86EHTest("enable-x86-eh-test", cl::Hidden);
> +
> +namespace llvm {
> +  class X86_test_MachoTargetObjectFile : public TargetLoweringObjectFileMachO {
> +  public:
> +    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
> +      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
> +
> +      // Exception Handling.
> +      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
> +                                    SectionKind::getReadOnlyWithRel());
> +    }
> +
> +    virtual unsigned getTTypeEncoding() const {
> +      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
> +    }
> +  };
> +
> +  class X8664_test_MachoTargetObjectFile : public X8664_MachoTargetObjectFile {
> +  public:
> +    virtual void Initialize(MCContext &Ctx, const TargetMachine &TM) {
> +      TargetLoweringObjectFileMachO::Initialize(Ctx, TM);
> +
> +      // Exception Handling.
> +      LSDASection = getMachOSection("__TEXT", "__gcc_except_tab", 0,
> +                                    SectionKind::getReadOnlyWithRel());
> +    }
> +
> +    virtual unsigned getTTypeEncoding() const {
> +      return DW_EH_PE_indirect | DW_EH_PE_pcrel | DW_EH_PE_sdata4;
> +    }
> +  };
> +}
> +
> static TargetLoweringObjectFile *createTLOF(X86TargetMachine &TM) {
>   switch (TM.getSubtarget<X86Subtarget>().TargetType) {
>   default: llvm_unreachable("unknown subtarget type");
>   case X86Subtarget::isDarwin:
> +    // FIXME: This is for an EH test.
> +    if (EnableX86EHTest) {
> +      if (TM.getSubtarget<X86Subtarget>().is64Bit())
> +        return new X8664_test_MachoTargetObjectFile();
> +      else
> +        return new X86_test_MachoTargetObjectFile();
> +    }
> +
>     if (TM.getSubtarget<X86Subtarget>().is64Bit())
>       return new X8664_MachoTargetObjectFile();
>     return new TargetLoweringObjectFileMachO();
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From sabre at nondot.org  Fri Mar 12 13:31:03 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 19:31:03 -0000
Subject: [llvm-commits] [llvm] r98372 - in /llvm/trunk/lib/Target/X86:
 X86COFFMachineModuleInfo.cpp X86COFFMachineModuleInfo.h
Message-ID: <20100312193103.3600E2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 13:31:03 2010
New Revision: 98372

URL: http://llvm.org/viewvc/llvm-project?rev=98372&view=rev
Log:
remove the FnArgWords cache to make way for future changes.

Modified:
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp?rev=98372&r1=98371&r2=98372&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp Fri Mar 12 13:31:03 2010
@@ -45,24 +45,19 @@
     return;
 
   unsigned ArgWords = 0;
-  DenseMap<const Function*, unsigned>::const_iterator item = FnArgWords.find(F);
-  if (item == FnArgWords.end()) {
-    // Calculate arguments sizes
-    for (Function::const_arg_iterator AI = F->arg_begin(), AE = F->arg_end();
-         AI != AE; ++AI) {
-      const Type* Ty = AI->getType();
-
-      // 'Dereference' type in case of byval parameter attribute
-      if (AI->hasByValAttr())
-        Ty = cast<PointerType>(Ty)->getElementType();
-
-      // Size should be aligned to DWORD boundary
-      ArgWords += ((TD.getTypeAllocSize(Ty) + 3)/4)*4;
-    }
-
-    FnArgWords[F] = ArgWords;
-  } else
-    ArgWords = item->second;
+  
+  // Calculate arguments sizes
+  for (Function::const_arg_iterator AI = F->arg_begin(), AE = F->arg_end();
+       AI != AE; ++AI) {
+    const Type *Ty = AI->getType();
+
+    // 'Dereference' type in case of byval parameter attribute
+    if (AI->hasByValAttr())
+      Ty = cast<PointerType>(Ty)->getElementType();
+
+    // Size should be aligned to DWORD boundary
+    ArgWords += ((TD.getTypeAllocSize(Ty) + 3)/4)*4;
+  }
 
   const FunctionType *FT = F->getFunctionType();
   // "Pure" variadic functions do not receive @0 suffix.

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h?rev=98372&r1=98371&r2=98372&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h Fri Mar 12 13:31:03 2010
@@ -26,7 +26,6 @@
 /// for X86 COFF targets.
 class X86COFFMachineModuleInfo : public MachineModuleInfoImpl {
   StringSet<> CygMingStubs;
-  DenseMap<const Function*, unsigned> FnArgWords;
 public:
   X86COFFMachineModuleInfo(const MachineModuleInfo &);
   ~X86COFFMachineModuleInfo();




From sabre at nondot.org  Fri Mar 12 13:42:40 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 19:42:40 -0000
Subject: [llvm-commits] [llvm] r98373 - in /llvm/trunk/lib/Target/X86:
 AsmPrinter/X86AsmPrinter.cpp AsmPrinter/X86MCInstLower.cpp
 X86COFFMachineModuleInfo.cpp X86COFFMachineModuleInfo.h
Message-ID: <20100312194240.AE6F12A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 13:42:40 2010
New Revision: 98373

URL: http://llvm.org/viewvc/llvm-project?rev=98373&view=rev
Log:
eliminate the string form of DecorateCygMingName

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98373&r1=98372&r2=98373&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:42:40 2010
@@ -61,7 +61,8 @@
   
   X86COFFMachineModuleInfo &COFFMMI =
     MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-  COFFMMI.DecorateCygMingName(Symb, OutContext, GV, *TM.getTargetData());
+  COFFMMI.DecorateCygMingName(Symb, OutContext, cast<Function>(GV),
+                              *TM.getTargetData());
 
   return Symb;
 }

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98373&r1=98372&r2=98373&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Fri Mar 12 13:42:40 2010
@@ -23,6 +23,7 @@
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCInst.h"
 #include "llvm/MC/MCStreamer.h"
+#include "llvm/MC/MCSymbol.h"
 #include "llvm/Target/Mangler.h"
 #include "llvm/Support/FormattedStream.h"
 #include "llvm/ADT/SmallString.h"
@@ -54,7 +55,21 @@
 
   SmallString<128> Name;
   
-  if (MO.isGlobal()) {
+  if (!MO.isGlobal()) {
+    assert(MO.isSymbol());
+    Name += AsmPrinter.MAI->getGlobalPrefix();
+    Name += MO.getSymbolName();
+  } else if (getSubtarget().isTargetCygMing() &&
+             isa<Function>(MO.getGlobal())) {
+    const GlobalValue *GV = MO.getGlobal();
+    MCSymbol *Sym = Mang->getSymbol(GV);
+    X86COFFMachineModuleInfo &COFFMMI = 
+    AsmPrinter.MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
+    COFFMMI.DecorateCygMingName(Sym, Ctx, cast<Function>(GV),
+                                *AsmPrinter.TM.getTargetData());
+    Name.append(Sym->getName().begin(), Sym->getName().end());
+  } else {    
+    const GlobalValue *GV = MO.getGlobal();
     bool isImplicitlyPrivate = false;
     if (MO.getTargetFlags() == X86II::MO_DARWIN_STUB ||
         MO.getTargetFlags() == X86II::MO_DARWIN_NONLAZY ||
@@ -62,18 +77,7 @@
         MO.getTargetFlags() == X86II::MO_DARWIN_HIDDEN_NONLAZY_PIC_BASE)
       isImplicitlyPrivate = true;
     
-    const GlobalValue *GV = MO.getGlobal();
     Mang->getNameWithPrefix(Name, GV, isImplicitlyPrivate);
-  
-    if (getSubtarget().isTargetCygMing()) {
-      X86COFFMachineModuleInfo &COFFMMI = 
-        AsmPrinter.MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-      COFFMMI.DecorateCygMingName(Name, GV, *AsmPrinter.TM.getTargetData());
-    }
-  } else {
-    assert(MO.isSymbol());
-    Name += AsmPrinter.MAI->getGlobalPrefix();
-    Name += MO.getSymbolName();
   }
 
   // If the target flags on the operand changes the name of the symbol, do that

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp?rev=98373&r1=98372&r2=98373&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp Fri Mar 12 13:42:40 2010
@@ -22,65 +22,50 @@
 #include "llvm/Support/raw_ostream.h"
 using namespace llvm;
 
-X86COFFMachineModuleInfo::X86COFFMachineModuleInfo(const MachineModuleInfo &) {
-}
-X86COFFMachineModuleInfo::~X86COFFMachineModuleInfo() {
-}
 
-void X86COFFMachineModuleInfo::addExternalFunction(const StringRef& Name) {
-  CygMingStubs.insert(Name);
+X86COFFMachineModuleInfo::~X86COFFMachineModuleInfo() {
 }
 
-/// DecorateCygMingName - Apply various name decorations if the function uses
-/// stdcall or fastcall calling convention.
-void X86COFFMachineModuleInfo::DecorateCygMingName(SmallVectorImpl<char> &Name,
-                                                   const GlobalValue *GV,
+/// DecorateCygMingName - Query FunctionInfoMap and use this information for
+/// various name decorations for Cygwin and MingW.
+void X86COFFMachineModuleInfo::DecorateCygMingName(MCSymbol *&NameSym,
+                                                   MCContext &Ctx,
+                                                   const Function *F,
                                                    const TargetData &TD) {
-  const Function *F = dyn_cast<Function>(GV);
-  if (!F) return;
-
+  SmallString<128> Name(NameSym->getName().begin(), NameSym->getName().end());
+  
   // We don't want to decorate non-stdcall or non-fastcall functions right now
   CallingConv::ID CC = F->getCallingConv();
   if (CC != CallingConv::X86_StdCall && CC != CallingConv::X86_FastCall)
     return;
-
+  
   unsigned ArgWords = 0;
   
   // Calculate arguments sizes
   for (Function::const_arg_iterator AI = F->arg_begin(), AE = F->arg_end();
        AI != AE; ++AI) {
     const Type *Ty = AI->getType();
-
+    
     // 'Dereference' type in case of byval parameter attribute
     if (AI->hasByValAttr())
       Ty = cast<PointerType>(Ty)->getElementType();
-
+    
     // Size should be aligned to DWORD boundary
     ArgWords += ((TD.getTypeAllocSize(Ty) + 3)/4)*4;
   }
-
+  
   const FunctionType *FT = F->getFunctionType();
   // "Pure" variadic functions do not receive @0 suffix.
   if (!FT->isVarArg() || FT->getNumParams() == 0 ||
       (FT->getNumParams() == 1 && F->hasStructRetAttr()))
     raw_svector_ostream(Name) << '@' << ArgWords;
-
+  
   if (CC == CallingConv::X86_FastCall) {
     if (Name[0] == '_')
       Name[0] = '@';
     else
       Name.insert(Name.begin(), '@');
   }
-}
-
-/// DecorateCygMingName - Query FunctionInfoMap and use this information for
-/// various name decorations for Cygwin and MingW.
-void X86COFFMachineModuleInfo::DecorateCygMingName(MCSymbol *&Name,
-                                                   MCContext &Ctx,
-                                                   const GlobalValue *GV,
-                                                   const TargetData &TD) {
-  SmallString<128> NameStr(Name->getName().begin(), Name->getName().end());
-  DecorateCygMingName(NameStr, GV, TD);
-
-  Name = Ctx.GetOrCreateSymbol(NameStr.str());
+  
+  NameSym = Ctx.GetOrCreateSymbol(Name.str());
 }

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h?rev=98373&r1=98372&r2=98373&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h Fri Mar 12 13:42:40 2010
@@ -27,15 +27,16 @@
 class X86COFFMachineModuleInfo : public MachineModuleInfoImpl {
   StringSet<> CygMingStubs;
 public:
-  X86COFFMachineModuleInfo(const MachineModuleInfo &);
-  ~X86COFFMachineModuleInfo();
+  X86COFFMachineModuleInfo(const MachineModuleInfo &) {}
+  virtual ~X86COFFMachineModuleInfo();
 
-  void DecorateCygMingName(MCSymbol* &Name, MCContext &Ctx,
-                           const GlobalValue *GV, const TargetData &TD);
-  void DecorateCygMingName(SmallVectorImpl<char> &Name, const GlobalValue *GV,
-                           const TargetData &TD);
+  void DecorateCygMingName(MCSymbol *&Name, MCContext &Ctx,
+                           const Function *F, const TargetData &TD);
 
-  void addExternalFunction(const StringRef& Name);
+  void addExternalFunction(StringRef Name) {
+    CygMingStubs.insert(Name);
+  }
+    
   typedef StringSet<>::const_iterator stub_iterator;
   stub_iterator stub_begin() const { return CygMingStubs.begin(); }
   stub_iterator stub_end() const { return CygMingStubs.end(); }




From sabre at nondot.org  Fri Mar 12 13:48:03 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 19:48:03 -0000
Subject: [llvm-commits] [llvm] r98374 - in /llvm/trunk/lib/Target/X86:
 AsmPrinter/X86AsmPrinter.cpp AsmPrinter/X86MCInstLower.cpp
 X86COFFMachineModuleInfo.cpp X86COFFMachineModuleInfo.h
Message-ID: <20100312194803.49A1B2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 13:48:03 2010
New Revision: 98374

URL: http://llvm.org/viewvc/llvm-project?rev=98374&view=rev
Log:
minor tidying, only do work if a function is 
actually X86_StdCall or X86_FastCall.

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98374&r1=98373&r2=98374&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 13:48:03 2010
@@ -61,10 +61,8 @@
   
   X86COFFMachineModuleInfo &COFFMMI =
     MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-  COFFMMI.DecorateCygMingName(Symb, OutContext, cast<Function>(GV),
-                              *TM.getTargetData());
-
-  return Symb;
+  return COFFMMI.DecorateCygMingName(Symb, OutContext, cast<Function>(GV),
+                                     *TM.getTargetData());
 }
 
 /// runOnMachineFunction - Emit the function body.

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98374&r1=98373&r2=98374&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Fri Mar 12 13:48:03 2010
@@ -65,8 +65,8 @@
     MCSymbol *Sym = Mang->getSymbol(GV);
     X86COFFMachineModuleInfo &COFFMMI = 
     AsmPrinter.MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-    COFFMMI.DecorateCygMingName(Sym, Ctx, cast<Function>(GV),
-                                *AsmPrinter.TM.getTargetData());
+    Sym = COFFMMI.DecorateCygMingName(Sym, Ctx, cast<Function>(GV),
+                                      *AsmPrinter.TM.getTargetData());
     Name.append(Sym->getName().begin(), Sym->getName().end());
   } else {    
     const GlobalValue *GV = MO.getGlobal();

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp?rev=98374&r1=98373&r2=98374&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.cpp Fri Mar 12 13:48:03 2010
@@ -28,16 +28,14 @@
 
 /// DecorateCygMingName - Query FunctionInfoMap and use this information for
 /// various name decorations for Cygwin and MingW.
-void X86COFFMachineModuleInfo::DecorateCygMingName(MCSymbol *&NameSym,
-                                                   MCContext &Ctx,
-                                                   const Function *F,
-                                                   const TargetData &TD) {
-  SmallString<128> Name(NameSym->getName().begin(), NameSym->getName().end());
-  
+MCSymbol *X86COFFMachineModuleInfo::DecorateCygMingName(MCSymbol *NameSym,
+                                                        MCContext &Ctx,
+                                                        const Function *F,
+                                                        const TargetData &TD) {
   // We don't want to decorate non-stdcall or non-fastcall functions right now
   CallingConv::ID CC = F->getCallingConv();
   if (CC != CallingConv::X86_StdCall && CC != CallingConv::X86_FastCall)
-    return;
+    return NameSym;
   
   unsigned ArgWords = 0;
   
@@ -55,6 +53,9 @@
   }
   
   const FunctionType *FT = F->getFunctionType();
+  
+  SmallString<128> Name(NameSym->getName().begin(), NameSym->getName().end());
+
   // "Pure" variadic functions do not receive @0 suffix.
   if (!FT->isVarArg() || FT->getNumParams() == 0 ||
       (FT->getNumParams() == 1 && F->hasStructRetAttr()))
@@ -67,5 +68,5 @@
       Name.insert(Name.begin(), '@');
   }
   
-  NameSym = Ctx.GetOrCreateSymbol(Name.str());
+  return Ctx.GetOrCreateSymbol(Name.str());
 }

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h?rev=98374&r1=98373&r2=98374&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h Fri Mar 12 13:48:03 2010
@@ -30,8 +30,8 @@
   X86COFFMachineModuleInfo(const MachineModuleInfo &) {}
   virtual ~X86COFFMachineModuleInfo();
 
-  void DecorateCygMingName(MCSymbol *&Name, MCContext &Ctx,
-                           const Function *F, const TargetData &TD);
+  MCSymbol *DecorateCygMingName(MCSymbol *Name, MCContext &Ctx,
+                                const Function *F, const TargetData &TD);
 
   void addExternalFunction(StringRef Name) {
     CygMingStubs.insert(Name);




From benny.kra at googlemail.com  Fri Mar 12 14:41:29 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Fri, 12 Mar 2010 20:41:29 -0000
Subject: [llvm-commits] [llvm] r98376 - in /llvm/trunk:
 include/llvm/Transforms/Utils/BuildLibCalls.h
 lib/Transforms/Utils/BuildLibCalls.cpp
Message-ID: <20100312204129.95C862A6C12C@llvm.org>

Author: d0k
Date: Fri Mar 12 14:41:29 2010
New Revision: 98376

URL: http://llvm.org/viewvc/llvm-project?rev=98376&view=rev
Log:
Add a virtual destructor and give vtable a home.

Modified:
    llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
    llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp

Modified: llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h?rev=98376&r1=98375&r2=98376&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h (original)
+++ llvm/trunk/include/llvm/Transforms/Utils/BuildLibCalls.h Fri Mar 12 14:41:29 2010
@@ -106,6 +106,7 @@
     virtual bool isFoldable(unsigned SizeCIOp, unsigned SizeArgOp,
                             bool isString) const = 0;
   public:
+    virtual ~SimplifyFortifiedLibCalls();
     bool fold(CallInst *CI, const TargetData *TD);
   };
 }

Modified: llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp?rev=98376&r1=98375&r2=98376&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BuildLibCalls.cpp Fri Mar 12 14:41:29 2010
@@ -343,6 +343,8 @@
     CI->setCallingConv(Fn->getCallingConv());
 }
 
+SimplifyFortifiedLibCalls::~SimplifyFortifiedLibCalls() { }
+
 bool SimplifyFortifiedLibCalls::fold(CallInst *CI, const TargetData *TD) {
   this->CI = CI;
   StringRef Name = CI->getCalledFunction()->getName();




From sabre at nondot.org  Fri Mar 12 14:43:52 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 20:43:52 -0000
Subject: [llvm-commits] [llvm] r98377 - in /llvm/trunk/lib/Target/X86:
 AsmPrinter/X86AsmPrinter.cpp AsmPrinter/X86MCInstLower.cpp
 X86COFFMachineModuleInfo.h
Message-ID: <20100312204352.5FFED2A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 14:43:52 2010
New Revision: 98377

URL: http://llvm.org/viewvc/llvm-project?rev=98377&view=rev
Log:
make DecorateCygMingName a static method.

Modified:
    llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
    llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
    llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp?rev=98377&r1=98376&r2=98377&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86AsmPrinter.cpp Fri Mar 12 14:43:52 2010
@@ -59,10 +59,9 @@
   if (!Subtarget->isTargetCygMing() || !isa<Function>(GV))
     return Symb;
   
-  X86COFFMachineModuleInfo &COFFMMI =
-    MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-  return COFFMMI.DecorateCygMingName(Symb, OutContext, cast<Function>(GV),
-                                     *TM.getTargetData());
+  return X86COFFMachineModuleInfo::
+    DecorateCygMingName(Symb, OutContext, cast<Function>(GV),
+                        *TM.getTargetData());
 }
 
 /// runOnMachineFunction - Emit the function body.

Modified: llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp?rev=98377&r1=98376&r2=98377&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp (original)
+++ llvm/trunk/lib/Target/X86/AsmPrinter/X86MCInstLower.cpp Fri Mar 12 14:43:52 2010
@@ -63,10 +63,9 @@
              isa<Function>(MO.getGlobal())) {
     const GlobalValue *GV = MO.getGlobal();
     MCSymbol *Sym = Mang->getSymbol(GV);
-    X86COFFMachineModuleInfo &COFFMMI = 
-    AsmPrinter.MMI->getObjFileInfo<X86COFFMachineModuleInfo>();
-    Sym = COFFMMI.DecorateCygMingName(Sym, Ctx, cast<Function>(GV),
-                                      *AsmPrinter.TM.getTargetData());
+    Sym = X86COFFMachineModuleInfo::
+      DecorateCygMingName(Sym, Ctx, cast<Function>(GV),
+                          *AsmPrinter.TM.getTargetData());
     Name.append(Sym->getName().begin(), Sym->getName().end());
   } else {    
     const GlobalValue *GV = MO.getGlobal();

Modified: llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h?rev=98377&r1=98376&r2=98377&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h (original)
+++ llvm/trunk/lib/Target/X86/X86COFFMachineModuleInfo.h Fri Mar 12 14:43:52 2010
@@ -30,8 +30,8 @@
   X86COFFMachineModuleInfo(const MachineModuleInfo &) {}
   virtual ~X86COFFMachineModuleInfo();
 
-  MCSymbol *DecorateCygMingName(MCSymbol *Name, MCContext &Ctx,
-                                const Function *F, const TargetData &TD);
+  static MCSymbol *DecorateCygMingName(MCSymbol *Name, MCContext &Ctx,
+                                       const Function *F, const TargetData &TD);
 
   void addExternalFunction(StringRef Name) {
     CygMingStubs.insert(Name);




From sabre at nondot.org  Fri Mar 12 14:47:28 2010
From: sabre at nondot.org (Chris Lattner)
Date: Fri, 12 Mar 2010 20:47:28 -0000
Subject: [llvm-commits] [llvm] r98378 - in /llvm/trunk:
 include/llvm/Target/Mangler.h lib/CodeGen/AsmPrinter/AsmPrinter.cpp
 lib/CodeGen/ELFWriter.cpp lib/Target/CBackend/CBackend.cpp
 tools/lto/LTOCodeGenerator.cpp tools/lto/LTOModule.cpp
Message-ID: <20100312204728.BC6B32A6C12C@llvm.org>

Author: lattner
Date: Fri Mar 12 14:47:28 2010
New Revision: 98378

URL: http://llvm.org/viewvc/llvm-project?rev=98378&view=rev
Log:
give Mangler access to TargetData.

Modified:
    llvm/trunk/include/llvm/Target/Mangler.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/ELFWriter.cpp
    llvm/trunk/lib/Target/CBackend/CBackend.cpp
    llvm/trunk/tools/lto/LTOCodeGenerator.cpp
    llvm/trunk/tools/lto/LTOModule.cpp

Modified: llvm/trunk/include/llvm/Target/Mangler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Mangler.h?rev=98378&r1=98377&r2=98378&view=diff
===========================================