From lattner at cs.uiuc.edu  Mon Apr 21 14:16:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 14:16:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Message-ID: <200304211915.OAA30485@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

PromoteMemoryToRegister.cpp updated: 1.38 -> 1.39

---
Log message:

Fix bug where use still existed in dead code


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
diff -u llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.38 llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.39
--- llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.38	Fri Apr 18 14:25:22 2003
+++ llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp	Mon Apr 21 14:15:26 2003
@@ -168,6 +168,8 @@
     //
     while (!I->use_empty()) {
       Instruction *U = cast<Instruction>(I->use_back());
+      if (!U->use_empty())  // If uses remain in dead code segment...
+        U->replaceAllUsesWith(Constant::getNullValue(U->getType()));
       U->getParent()->getInstList().erase(U);
     }
 




From lattner at cs.uiuc.edu  Mon Apr 21 14:54:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 14:54:00 2003
Subject: [llvm-commits] CVS: llvm/tools/gccld/gccld.cpp
Message-ID: <200304211953.OAA04514@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccld:

gccld.cpp updated: 1.27 -> 1.28

---
Log message:

Add support for allowing the user to extend the GCCLD searchpath


---
Diffs of the changes:

Index: llvm/tools/gccld/gccld.cpp
diff -u llvm/tools/gccld/gccld.cpp:1.27 llvm/tools/gccld/gccld.cpp:1.28
--- llvm/tools/gccld/gccld.cpp:1.27	Sat Apr 19 18:07:33 2003
+++ llvm/tools/gccld/gccld.cpp	Mon Apr 21 14:53:24 2003
@@ -294,6 +294,10 @@
   if (Composite.get() == 0)
     return PrintAndReturn(argv[0], ErrorMessage);
 
+  // If the user specied an extra search path in their environment, respect it.
+  if (char *SearchPath = getenv("LLVM_LIB_SEARCH_PATH"))
+    LibPaths.push_back(SearchPath);
+
   for (unsigned i = 1; i < InputFilenames.size(); ++i) {
     std::auto_ptr<Module> M(LoadObject(InputFilenames[i], ErrorMessage));
     if (M.get() == 0)




From lattner at cs.uiuc.edu  Mon Apr 21 16:07:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 16:07:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/2003-04-21-Linkage.ll testlink1.ll testlink2.ll
Message-ID: <200304212106.QAA12473@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

2003-04-21-Linkage.ll added (r1.1)
testlink1.ll updated: 1.6 -> 1.7
testlink2.ll updated: 1.7 -> 1.8

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Linker/2003-04-21-Linkage.ll
diff -c /dev/null llvm/test/Regression/Linker/2003-04-21-Linkage.ll:1.1
*** /dev/null	Mon Apr 21 16:06:36 2003
--- llvm/test/Regression/Linker/2003-04-21-Linkage.ll	Mon Apr 21 16:06:25 2003
***************
*** 0 ****
--- 1,15 ----
+ ; RUN: echo "%X = linkonce global int 5  implementation linkonce int %foo() { ret int 7 }" | as > Output/%s.1.bc
+ ; RUN: as < %s > Output/%s.2.bc
+ ; RUN: link Output/%s.[12].bc 
+ %X = external global int 
+ 
+ implementation
+ 
+ declare int %foo() 
+ 
+ void %bar() {
+ 	load int* %X
+ 	call int %foo()
+ 	ret void
+ }
+ 


Index: llvm/test/Regression/Linker/testlink1.ll
diff -u llvm/test/Regression/Linker/testlink1.ll:1.6 llvm/test/Regression/Linker/testlink1.ll:1.7
--- llvm/test/Regression/Linker/testlink1.ll:1.6	Wed Aug 21 18:52:57 2002
+++ llvm/test/Regression/Linker/testlink1.ll	Mon Apr 21 16:06:25 2003
@@ -2,9 +2,9 @@
 ; RUN: as < testlink2.ll > Output/testlink2.bc
 ; RUN: link Output/%s.bc Output/testlink2.bc
 
-%MyVar     = uninitialized global int
+%MyVar     = external global int
 %MyIntList = global { \2 *, int } { { \2, int }* null, int 17 }
-             uninitialized global int      ; int*:0
+             external global int      ; int*:0
 
 %AConst    = constant int 123
 
@@ -25,7 +25,7 @@
 	%v1 = load int* %MyVar
 	call void %print(int %v1)    ;; Should start out 4
 
-	%idx = getelementptr { \2 *, int }* %MyIntList, uint 0, ubyte 1
+	%idx = getelementptr { \2 *, int }* %MyIntList, long 0, ubyte 1
 	%v2 = load int* %idx
 	call void %print(int %v2)    ;; Should start out 17
 


Index: llvm/test/Regression/Linker/testlink2.ll
diff -u llvm/test/Regression/Linker/testlink2.ll:1.7 llvm/test/Regression/Linker/testlink2.ll:1.8
--- llvm/test/Regression/Linker/testlink2.ll:1.7	Thu Aug 22 11:39:02 2002
+++ llvm/test/Regression/Linker/testlink2.ll	Mon Apr 21 16:06:25 2003
@@ -3,7 +3,7 @@
 ; RUN: echo
 
 %MyVar     = global int 4
-%MyIntList = uninitialized global { \2 *, int }
+%MyIntList = external global { \2 *, int }
 
 %AConst    = constant int 123
 
@@ -20,7 +20,7 @@
 int "foo"(int %blah)
 begin
 	store int %blah, int *%MyVar
-	%idx = getelementptr { \2 *, int } * %MyIntList, uint 0, ubyte 1
+	%idx = getelementptr { \2 *, int } * %MyIntList, long 0, ubyte 1
 	store int 12, int* %idx
 
 	%ack = load int * %0   ;; Load from the unnamed constant




From lattner at cs.uiuc.edu  Mon Apr 21 16:08:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 16:08:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/Linker.cpp
Message-ID: <200304212107.QAA12507@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

Linker.cpp updated: 1.38 -> 1.39

---
Log message:

Fix linking a function with qualifiers to a external function declaration:
  Fixed bug: Linker/2003-04-21-Linkage.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/Linker.cpp
diff -u llvm/lib/Transforms/Utils/Linker.cpp:1.38 llvm/lib/Transforms/Utils/Linker.cpp:1.39
--- llvm/lib/Transforms/Utils/Linker.cpp:1.38	Wed Apr 16 15:28:43 2003
+++ llvm/lib/Transforms/Utils/Linker.cpp	Mon Apr 21 16:07:05 2003
@@ -202,6 +202,9 @@
     assert(SGV->hasInitializer() || SGV->hasExternalLinkage() &&
            "Global must either be external or have an initializer!");
 
+    bool SGExtern = SGV->isExternal();
+    bool DGExtern = DGV ? DGV->isExternal() : false;
+
     if (!DGV || DGV->hasInternalLinkage() || SGV->hasInternalLinkage()) {
       // No linking to be performed, simply create an identical version of the
       // symbol over in the dest module... the initializer will be filled in
@@ -213,7 +216,7 @@
 
       // Make sure to remember this mapping...
       ValueMap.insert(std::make_pair(SGV, DGV));
-    } else if (SGV->getLinkage() != DGV->getLinkage()) {
+    } else if (!SGExtern && !DGExtern && SGV->getLinkage() !=DGV->getLinkage()){
       return Error(Err, "Global variables named '" + SGV->getName() +
                    "' have different linkage specifiers!");
     } else if (SGV->hasExternalLinkage() || SGV->hasLinkOnceLinkage() ||
@@ -227,6 +230,9 @@
         return Error(Err, "Global Variable Collision on '" + 
                      SGV->getType()->getDescription() + "':%" + SGV->getName() +
                      " - Global variables differ in const'ness");
+      if (DGExtern)
+        DGV->setLinkage(SGV->getLinkage());
+
       // Okay, everything is cool, remember the mapping...
       ValueMap.insert(std::make_pair(SGV, DGV));
     } else {
@@ -301,20 +307,23 @@
       // 
       DF = cast_or_null<Function>(ST->lookup(SF->getType(), SF->getName()));
 
+    bool SFExtern = SF->isExternal();
+    bool DFExtern = DF ? DF->isExternal() : false;
+
     if (!DF || SF->hasInternalLinkage() || DF->hasInternalLinkage()) {
-      // Function does not already exist, simply insert an external function
-      // signature identical to SF into the dest module...
+      // Function does not already exist, simply insert an function signature
+      // identical to SF into the dest module...
       Function *DF = new Function(SF->getFunctionType(), SF->getLinkage(),
                                   SF->getName(), Dest);
 
       // ... and remember this mapping...
       ValueMap.insert(std::make_pair(SF, DF));
-    } else if (SF->getLinkage() != DF->getLinkage()) {
-      return Error(Err, "Functions named '" + SF->getName() +
-                   "' have different linkage specifiers!");
     } else if (SF->getLinkage() == GlobalValue::AppendingLinkage) {
       return Error(Err, "Functions named '" + SF->getName() +
                    "' have appending linkage!");
+    } else if (!SFExtern && !DFExtern && SF->getLinkage() != DF->getLinkage()) {
+      return Error(Err, "Functions named '" + SF->getName() +
+                   "' have different linkage specifiers!");
     } else if (SF->getLinkage() == GlobalValue::ExternalLinkage) {
       // If the function has a name, and that name is already in use in the Dest
       // module, make sure that the name is a compatible function...
@@ -324,6 +333,9 @@
         return Error(Err, "Function '" + 
                      SF->getFunctionType()->getDescription() + "':\"" + 
                      SF->getName() + "\" - Function is already defined!");
+      
+      if (DFExtern)
+        DF->setLinkage(SF->getLinkage());
       
       // Otherwise, just remember this mapping...
       ValueMap.insert(std::make_pair(SF, DF));




From lattner at cs.uiuc.edu  Mon Apr 21 16:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 16:16:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/Linker.cpp
Message-ID: <200304212115.QAA12925@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

Linker.cpp updated: 1.39 -> 1.40

---
Log message:

Fix bug: 2003-01-30-LinkerRename.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/Linker.cpp
diff -u llvm/lib/Transforms/Utils/Linker.cpp:1.39 llvm/lib/Transforms/Utils/Linker.cpp:1.40
--- llvm/lib/Transforms/Utils/Linker.cpp:1.39	Mon Apr 21 16:07:05 2003
+++ llvm/lib/Transforms/Utils/Linker.cpp	Mon Apr 21 16:15:04 2003
@@ -210,12 +210,26 @@
       // symbol over in the dest module... the initializer will be filled in
       // later by LinkGlobalInits...
       //
-      DGV = new GlobalVariable(SGV->getType()->getElementType(),
-                               SGV->isConstant(), SGV->getLinkage(), /*init*/0,
-                               SGV->getName(), Dest);
+      GlobalVariable *NewDGV =
+        new GlobalVariable(SGV->getType()->getElementType(),
+                           SGV->isConstant(), SGV->getLinkage(), /*init*/0,
+                           SGV->getName(), Dest);
+
+      // If the LLVM runtime renamed the global, but it is an externally visible
+      // symbol, DGV must be an existing global with internal linkage.  Rename
+      // it.
+      if (NewDGV->getName() != SGV->getName() && !NewDGV->hasInternalLinkage()){
+        assert(DGV && DGV->getName() == SGV->getName() &&
+               DGV->hasInternalLinkage());
+        DGV->setName("");
+        NewDGV->setName(SGV->getName());  // Force the name back
+        DGV->setName(SGV->getName());     // This will cause a renaming
+        assert(NewDGV->getName() == SGV->getName() &&
+               DGV->getName() != SGV->getName());
+      }
 
       // Make sure to remember this mapping...
-      ValueMap.insert(std::make_pair(SGV, DGV));
+      ValueMap.insert(std::make_pair(SGV, NewDGV));
     } else if (!SGExtern && !DGExtern && SGV->getLinkage() !=DGV->getLinkage()){
       return Error(Err, "Global variables named '" + SGV->getName() +
                    "' have different linkage specifiers!");
@@ -313,11 +327,23 @@
     if (!DF || SF->hasInternalLinkage() || DF->hasInternalLinkage()) {
       // Function does not already exist, simply insert an function signature
       // identical to SF into the dest module...
-      Function *DF = new Function(SF->getFunctionType(), SF->getLinkage(),
-                                  SF->getName(), Dest);
+      Function *NewDF = new Function(SF->getFunctionType(), SF->getLinkage(),
+                                     SF->getName(), Dest);
+
+      // If the LLVM runtime renamed the function, but it is an externally
+      // visible symbol, DF must be an existing function with internal linkage.
+      // Rename it.
+      if (NewDF->getName() != SF->getName() && !NewDF->hasInternalLinkage()) {
+        assert(DF && DF->getName() == SF->getName() &&DF->hasInternalLinkage());
+        DF->setName("");
+        NewDF->setName(SF->getName());  // Force the name back
+        DF->setName(SF->getName());     // This will cause a renaming
+        assert(NewDF->getName() == SF->getName() &&
+               DF->getName() != SF->getName());
+      }
 
       // ... and remember this mapping...
-      ValueMap.insert(std::make_pair(SF, DF));
+      ValueMap.insert(std::make_pair(SF, NewDF));
     } else if (SF->getLinkage() == GlobalValue::AppendingLinkage) {
       return Error(Err, "Functions named '" + SF->getName() +
                    "' have appending linkage!");




From lattner at cs.uiuc.edu  Mon Apr 21 16:34:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 16:34:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200304212133.QAA14671@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.85 -> 1.86

---
Log message:

Add support for a simple constantexpr: cast of one ptr type to another


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.85 llvm/lib/Target/X86/InstSelectSimple.cpp:1.86
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.85	Thu Jan 16 12:07:23 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Mon Apr 21 16:33:44 2003
@@ -307,6 +307,11 @@
       emitGEPOperation(MBB, IP, CE->getOperand(0),
                        CE->op_begin()+1, CE->op_end(), R);
       return;
+    } else if (CE->getOpcode() == Instruction::Cast &&
+               isa<PointerType>(CE->getType()) &&
+               isa<PointerType>(CE->getOperand(0)->getType())) {
+      copyConstantToRegister(MBB, IP, cast<Constant>(CE->getOperand(0)), R);
+      return;
     }
 
     std::cerr << "Offending expr: " << C << "\n";




From lattner at cs.uiuc.edu  Mon Apr 21 17:44:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 17:44:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304212243.RAA15821@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.76 -> 1.77

---
Log message:

Fix wierd idiom


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.76 llvm/tools/lli/Interpreter/Execution.cpp:1.77
--- llvm/tools/lli/Interpreter/Execution.cpp:1.76	Tue Feb 25 15:14:59 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Mon Apr 21 17:43:32 2003
@@ -111,7 +111,7 @@
     default:
       cerr << "Unhandled ConstantExpr: " << CE << "\n";
       abort();
-      { GenericValue V; return V; }
+      return GenericValue();
     }
   } else if (Constant *CPV = dyn_cast<Constant>(V)) {
     return TheEE->getConstantValue(CPV);




From lattner at cs.uiuc.edu  Mon Apr 21 17:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 17:45:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/ExecutionEngine.cpp
Message-ID: <200304212244.RAA15838@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

ExecutionEngine.cpp updated: 1.4 -> 1.5

---
Log message:

Make stderr, stdout, stdin work with LLI


---
Diffs of the changes:

Index: llvm/tools/lli/ExecutionEngine.cpp
diff -u llvm/tools/lli/ExecutionEngine.cpp:1.4 llvm/tools/lli/ExecutionEngine.cpp:1.5
--- llvm/tools/lli/ExecutionEngine.cpp:1.4	Sun Jan 12 18:58:06 2003
+++ llvm/tools/lli/ExecutionEngine.cpp	Mon Apr 21 17:44:36 2003
@@ -213,6 +213,12 @@
 
       DEBUG(std::cerr << "Global '" << I->getName() << "' -> "
 	              << (void*)GlobalAddress[I] << "\n");
+    } else if (I->getName() == "stdout") {
+      GlobalAddress[I] = &stdout;
+    } else if (I->getName() == "stderr") {
+      GlobalAddress[I] = &stderr;
+    } else if (I->getName() == "stdin") {
+      GlobalAddress[I] = &stdin;
     } else {
       std::cerr << "Global: " << I->getName() << "\n";
       assert(0 && "References to external globals not handled yet!");




From lattner at cs.uiuc.edu  Mon Apr 21 17:45:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Mon Apr 21 17:45:02 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304212244.RAA15828@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.43 -> 1.44

---
Log message:

fprintf CAN take exactly 2 args


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.43 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.44
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.43	Mon Mar 31 16:12:37 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Mon Apr 21 17:43:20 2003
@@ -685,7 +685,7 @@
 // int fprintf(FILE *,sbyte *, ...) - a very rough implementation to make output
 // useful.
 GenericValue lle_X_fprintf(FunctionType *M, const vector<GenericValue> &Args) {
-  assert(Args.size() > 2);
+  assert(Args.size() >= 2);
   char Buffer[10000];
   vector<GenericValue> NewArgs;
   NewArgs.push_back(PTOGV(Buffer));




From brukman at cs.uiuc.edu  Tue Apr 22 12:55:01 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Tue Apr 22 12:55:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200304221754.MAA03131@neo.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.86 -> 1.87

---
Log message:

It's "necessary" to spell "necessarily" correctly.


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.86 llvm/lib/Target/X86/InstSelectSimple.cpp:1.87
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.86	Mon Apr 21 16:33:44 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Tue Apr 22 12:54:23 2003
@@ -1209,7 +1209,7 @@
   // Emit the byte swap instruction...
   switch (Class) {
   case cByte:
-    // No byteswap neccesary for 8 bit value...
+    // No byteswap necessary for 8 bit value...
     BuildMI(BB, X86::MOVrr8, 1, DestReg).addReg(SrcReg);
     break;
   case cInt:
@@ -1219,7 +1219,7 @@
     
   case cShort:
     // For 16 bit we have to use an xchg instruction, because there is no
-    // 16-bit bswap.  XCHG is neccesarily not in SSA form, so we force things
+    // 16-bit bswap.  XCHG is necessarily not in SSA form, so we force things
     // into AX to do the xchg.
     //
     BuildMI(BB, X86::MOVrr16, 1, X86::AX).addReg(SrcReg);




From lattner at cs.uiuc.edu  Tue Apr 22 13:03:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:03:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Module.h
Message-ID: <200304221802.NAA06917@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

Module.h updated: 1.30 -> 1.31

---
Log message:

Add support for tracking whether a module is 64/32 bit and big/little endian
Also add a moduleID field which can be used for diagnostics


---
Diffs of the changes:

Index: llvm/include/llvm/Module.h
diff -u llvm/include/llvm/Module.h:1.30 llvm/include/llvm/Module.h:1.31
--- llvm/include/llvm/Module.h:1.30	Wed Nov 20 12:31:31 2002
+++ llvm/include/llvm/Module.h	Tue Apr 22 13:02:02 2003
@@ -50,16 +50,22 @@
   typedef std::reverse_iterator<iterator>             reverse_iterator;
   typedef std::reverse_iterator<const_iterator> const_reverse_iterator;
 
-private:
-  GlobalListType GlobalList;     // The Global Variables
-  FunctionListType FunctionList;     // The Functions
+  enum Endianness  { LittleEndian, BigEndian };
+  enum PointerSize { Pointer32, Pointer64 };
 
-  GlobalValueRefMap *GVRefMap;
+private:
+  GlobalListType GlobalList;     // The Global Variables in the module
+  FunctionListType FunctionList; // The Functions in the module
+  GlobalValueRefMap *GVRefMap;   // Keep track of GlobalValueRef's
+  SymbolTable *SymTab;           // Symbol Table for the module
+  std::string ModuleID;    // Human readable identifier for the module
 
-  SymbolTable *SymTab;
+  // These flags are probably not the right long-term way to handle this kind of
+  // target information, but it is sufficient for now.
+  Endianness  Endian;       // True if target is little endian
+  PointerSize PtrSize;    // True if target has 32-bit pointers (false = 64-bit)
 
-  // Accessor for the underlying GlobalValRefMap... only through the
-  // Constant class...
+  // Accessor for the underlying GVRefMap... only through the Constant class...
   friend class Constant;
   friend class ConstantPointerRef;
   void mutateConstantPointerRef(GlobalValue *OldGV, GlobalValue *NewGV);
@@ -67,8 +73,22 @@
   void destroyConstantPointerRef(ConstantPointerRef *CPR);
 
 public:
-  Module();
+  Module(const std::string &ModuleID);
   ~Module();
+
+  const std::string &getModuleIdentifier() const { return ModuleID; }
+
+  /// Target endian information...
+  bool isLittleEndian() const { return Endian == LittleEndian; }
+  bool isBigEndian() const { return Endian == BigEndian; }
+  Endianness getEndianness() const { return Endian; }
+  void setEndianness(Endianness E) { Endian = E; }
+
+  /// Target Pointer Size information...
+  bool has32BitPointers() const { return PtrSize == Pointer32; }
+  bool has64BitPointers() const { return PtrSize == Pointer64; }
+  PointerSize getPointerSize() const { return PtrSize; }
+  void setPointerSize(PointerSize PS) { PtrSize = PS; }
 
   /// getOrInsertFunction - Look up the specified function in the module symbol
   /// table.  If it does not exist, add a prototype for the function and return




From lattner at cs.uiuc.edu  Tue Apr 22 13:03:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:03:02 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/Module.cpp
Message-ID: <200304221802.NAA06924@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

Module.cpp updated: 1.34 -> 1.35

---
Log message:

Add support for tracking whether a module is 64/32 bit and big/little endian
Also add a moduleID field which can be used for diagnostics



---
Diffs of the changes:

Index: llvm/lib/VMCore/Module.cpp
diff -u llvm/lib/VMCore/Module.cpp:1.34 llvm/lib/VMCore/Module.cpp:1.35
--- llvm/lib/VMCore/Module.cpp:1.34	Wed Apr 16 15:28:45 2003
+++ llvm/lib/VMCore/Module.cpp	Tue Apr 22 13:02:04 2003
@@ -52,7 +52,8 @@
 };
 
 
-Module::Module() {
+Module::Module(const std::string &MID)
+  : ModuleID(MID), Endian(BigEndian), PtrSize(Pointer64) {
   FunctionList.setItemParent(this);
   FunctionList.setParent(this);
   GlobalList.setItemParent(this);




From lattner at cs.uiuc.edu  Tue Apr 22 13:03:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:03:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/CloneModule.cpp
Message-ID: <200304221802.NAA06937@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

CloneModule.cpp updated: 1.2 -> 1.3

---
Log message:

Preserve the new moduleID field


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/CloneModule.cpp
diff -u llvm/lib/Transforms/Utils/CloneModule.cpp:1.2 llvm/lib/Transforms/Utils/CloneModule.cpp:1.3
--- llvm/lib/Transforms/Utils/CloneModule.cpp:1.2	Wed Apr 16 15:28:43 2003
+++ llvm/lib/Transforms/Utils/CloneModule.cpp	Tue Apr 22 13:02:26 2003
@@ -18,7 +18,7 @@
 ///
 Module *CloneModule(const Module *M) {
   // First off, we need to create the new module...
-  Module *New = new Module();
+  Module *New = new Module(M->getModuleIdentifier());
 
   // Create the value map that maps things from the old module over to the new
   // module.




From lattner at cs.uiuc.edu  Tue Apr 22 13:04:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:04:00 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/llvmAsmParser.y
Message-ID: <200304221803.NAA06973@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

llvmAsmParser.y updated: 1.105 -> 1.106

---
Log message:

Preserve module source information in the ModuleID



---
Diffs of the changes:

Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.105 llvm/lib/AsmParser/llvmAsmParser.y:1.106
--- llvm/lib/AsmParser/llvmAsmParser.y:1.105	Wed Apr 16 15:28:31 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Tue Apr 22 13:02:52 2003
@@ -591,7 +591,8 @@
   CurFilename = Filename;
   llvmAsmlineno = 1;      // Reset the current line number...
 
-  CurModule.CurrentModule = new Module();  // Allocate a new module to read
+  // Allocate a new module to read
+  CurModule.CurrentModule = new Module(Filename);
   yyparse();       // Parse the file.
   Module *Result = ParserResult;
   llvmAsmin = stdin;    // F is about to go away, don't use it anymore...




From lattner at cs.uiuc.edu  Tue Apr 22 13:04:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:04:02 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Bytecode/Reader.h
Message-ID: <200304221803.NAA06986@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Bytecode:

Reader.h updated: 1.9 -> 1.10

---
Log message:

Add information about the module source


---
Diffs of the changes:

Index: llvm/include/llvm/Bytecode/Reader.h
diff -u llvm/include/llvm/Bytecode/Reader.h:1.9 llvm/include/llvm/Bytecode/Reader.h:1.10
--- llvm/include/llvm/Bytecode/Reader.h:1.9	Sat Apr 19 16:45:52 2003
+++ llvm/include/llvm/Bytecode/Reader.h	Tue Apr 22 13:03:13 2003
@@ -22,6 +22,7 @@
 Module *ParseBytecodeFile(const std::string &Filename,
                           std::string *ErrorStr = 0);
 Module *ParseBytecodeBuffer(const unsigned char *Buffer, unsigned BufferSize,
+                            const std::string &ModuleID,
                             std::string *ErrorStr = 0);
 
 // ReadArchiveFile - Read bytecode files from the specfied .a file, returning




From lattner at cs.uiuc.edu  Tue Apr 22 13:04:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:04:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Reader/ReadArchive.cpp Reader.cpp ReaderInternals.h
Message-ID: <200304221803.NAA06966@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Reader:

ReadArchive.cpp updated: 1.1 -> 1.2
Reader.cpp updated: 1.52 -> 1.53
ReaderInternals.h updated: 1.34 -> 1.35

---
Log message:

Preserve module source information in the ModuleID


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Reader/ReadArchive.cpp
diff -u llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.1 llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.2
--- llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.1	Sat Apr 19 16:45:34 2003
+++ llvm/lib/Bytecode/Reader/ReadArchive.cpp	Tue Apr 22 13:02:51 2003
@@ -82,7 +82,8 @@
 }
 
 
-static bool ReadArchiveBuffer(unsigned char *Buffer, unsigned Length,
+static bool ReadArchiveBuffer(const std::string &Filename,
+                              unsigned char *Buffer, unsigned Length,
                               std::vector<Module*> &Objects,
                               std::string *ErrorStr) {
   if (Length < 8 || memcmp(Buffer, "!<arch>\n", 8))
@@ -107,7 +108,8 @@
         return true;
       break;
     case UserObject: {
-      Module *M = ParseBytecodeBuffer(Buffer+sizeof(ar_hdr), Size, ErrorStr);
+      Module *M = ParseBytecodeBuffer(Buffer+sizeof(ar_hdr), Size,
+                                      Filename+":somefile", ErrorStr);
       if (!M) return true;
       Objects.push_back(M);
       break;
@@ -151,7 +153,7 @@
     return Error(ErrorStr, "Error mmapping file!");
   
   // Parse the archive files we mmap'ped in
-  bool Result = ReadArchiveBuffer(Buffer, Length, Objects, ErrorStr);
+  bool Result = ReadArchiveBuffer(Filename, Buffer, Length, Objects, ErrorStr);
   
   // Unmmap the archive...
   munmap((char*)Buffer, Length);


Index: llvm/lib/Bytecode/Reader/Reader.cpp
diff -u llvm/lib/Bytecode/Reader/Reader.cpp:1.52 llvm/lib/Bytecode/Reader/Reader.cpp:1.53
--- llvm/lib/Bytecode/Reader/Reader.cpp:1.52	Sat Apr 19 16:45:17 2003
+++ llvm/lib/Bytecode/Reader/Reader.cpp	Tue Apr 22 13:02:51 2003
@@ -620,14 +620,15 @@
   return 0;
 }
 
-Module *BytecodeParser::ParseBytecode(const uchar *Buf, const uchar *EndBuf) {
+Module *BytecodeParser::ParseBytecode(const uchar *Buf, const uchar *EndBuf,
+                                      const std::string &ModuleID) {
   unsigned Sig;
   // Read and check signature...
   if (read(Buf, EndBuf, Sig) ||
       Sig != ('l' | ('l' << 8) | ('v' << 16) | 'm' << 24))
     return ::Error(&Error, "Invalid bytecode signature!");
 
-  TheModule = new Module();
+  TheModule = new Module(ModuleID);
   if (ParseModule(Buf, EndBuf)) {
     delete TheModule;
     TheModule = 0;
@@ -637,7 +638,7 @@
 
 
 Module *ParseBytecodeBuffer(const unsigned char *Buffer, unsigned Length,
-                            std::string *ErrorStr) {
+                            const std::string &ModuleID, std::string *ErrorStr){
   BytecodeParser Parser;
   unsigned char *PtrToDelete = 0;
   if ((intptr_t)Buffer & 3) {         // If the buffer is not 4 byte aligned...
@@ -648,7 +649,7 @@
     Buffer = PtrToDelete+Offset;
   }
 
-  Module *R = Parser.ParseBytecode(Buffer, Buffer+Length);
+  Module *R = Parser.ParseBytecode(Buffer, Buffer+Length, ModuleID);
   if (ErrorStr) *ErrorStr = Parser.getError();
 
   delete [] PtrToDelete;   // Delete alignment buffer if neccesary
@@ -691,7 +692,7 @@
       return Error(ErrorStr, "Error mmapping file!");
 
     // Parse the bytecode we mmapped in
-    Result = ParseBytecodeBuffer(Buffer, Length, ErrorStr);
+    Result = ParseBytecodeBuffer(Buffer, Length, Filename, ErrorStr);
 
     // Unmmap the bytecode...
     munmap((char*)Buffer, Length);
@@ -721,7 +722,7 @@
     unsigned char *Buf = &FileData[0];
 #endif
 
-    Result = ParseBytecodeBuffer(Buf, FileData.size(), ErrorStr);
+    Result = ParseBytecodeBuffer(Buf, FileData.size(), "<stdin>", ErrorStr);
 
 #if ALIGN_PTRS
     munmap((char*)Buf, FileData.size());   // Free mmap'd data area


Index: llvm/lib/Bytecode/Reader/ReaderInternals.h
diff -u llvm/lib/Bytecode/Reader/ReaderInternals.h:1.34 llvm/lib/Bytecode/Reader/ReaderInternals.h:1.35
--- llvm/lib/Bytecode/Reader/ReaderInternals.h:1.34	Wed Apr 16 16:16:03 2003
+++ llvm/lib/Bytecode/Reader/ReaderInternals.h	Tue Apr 22 13:02:51 2003
@@ -52,7 +52,8 @@
     freeTable(ModuleValues);
   }
 
-  Module *ParseBytecode(const uchar *Buf, const uchar *EndBuf);
+  Module *ParseBytecode(const uchar *Buf, const uchar *EndBuf,
+                        const std::string &ModuleID);
 
   std::string getError() const { return Error; }
 




From lattner at cs.uiuc.edu  Tue Apr 22 13:11:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:11:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/lli.cpp
Message-ID: <200304221810.NAA07090@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

lli.cpp updated: 1.15 -> 1.16

---
Log message:

Build the appropriate target machine for the input pointer size and endianness


---
Diffs of the changes:

Index: llvm/tools/lli/lli.cpp
diff -u llvm/tools/lli/lli.cpp:1.15 llvm/tools/lli/lli.cpp:1.16
--- llvm/tools/lli/lli.cpp:1.15	Mon Dec 23 18:39:16 2002
+++ llvm/tools/lli/lli.cpp	Tue Apr 22 13:10:32 2003
@@ -75,10 +75,8 @@
   }
 #endif
 
-  // FIXME: This should look at the PointerSize and endianness of the bytecode
-  // file to determine the endianness and pointer size of target machine to use.
-  unsigned Config = TM::PtrSize64 | TM::BigEndian;
-
+  unsigned Config = (M->isLittleEndian()   ? TM::LittleEndian : TM::BigEndian) |
+                    (M->has32BitPointers() ? TM::PtrSize32    : TM::PtrSize64);
   ExecutionEngine *EE = 0;
 
   // If there is nothing that is forcing us to use the interpreter, make a JIT.




From lattner at cs.uiuc.edu  Tue Apr 22 13:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:16:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Reader/Reader.cpp
Message-ID: <200304221815.NAA07179@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Reader:

Reader.cpp updated: 1.53 -> 1.54

---
Log message:

Add support for reading and writing pointersize/endianness to and from bytecode


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Reader/Reader.cpp
diff -u llvm/lib/Bytecode/Reader/Reader.cpp:1.53 llvm/lib/Bytecode/Reader/Reader.cpp:1.54
--- llvm/lib/Bytecode/Reader/Reader.cpp:1.53	Tue Apr 22 13:02:51 2003
+++ llvm/lib/Bytecode/Reader/Reader.cpp	Tue Apr 22 13:15:08 2003
@@ -526,6 +526,11 @@
     return true;
   }
 
+  TheModule->setEndianness(isBigEndian ? Module::BigEndian :
+                                         Module::LittleEndian);
+  TheModule->setPointerSize(hasLongPointers ? Module::Pointer64 : 
+                                              Module::Pointer32);
+
   BCR_TRACE(1, "Bytecode Rev = " << (unsigned)RevisionNum << "\n");
   BCR_TRACE(1, "BigEndian/LongPointers = " << isBigEndian << ","
                << hasLongPointers << "\n");




From lattner at cs.uiuc.edu  Tue Apr 22 13:16:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:16:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Writer/Writer.cpp
Message-ID: <200304221815.NAA07186@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Writer:

Writer.cpp updated: 1.32 -> 1.33

---
Log message:

Add support for reading and writing pointersize/endianness to and from bytecode



---
Diffs of the changes:

Index: llvm/lib/Bytecode/Writer/Writer.cpp
diff -u llvm/lib/Bytecode/Writer/Writer.cpp:1.32 llvm/lib/Bytecode/Writer/Writer.cpp:1.33
--- llvm/lib/Bytecode/Writer/Writer.cpp:1.32	Wed Apr 16 16:16:05 2003
+++ llvm/lib/Bytecode/Writer/Writer.cpp	Tue Apr 22 13:15:10 2003
@@ -43,8 +43,8 @@
   // Emit the top level CLASS block.
   BytecodeBlock ModuleBlock(BytecodeFormat::Module, Out);
 
-  bool isBigEndian = true;
-  bool hasLongPointers = true;
+  bool isBigEndian = M->isBigEndian();
+  bool hasLongPointers = M->has64BitPointers();
 
   // Output the version identifier... we are currently on bytecode version #2
   unsigned Version = (2 << 4) | isBigEndian | (hasLongPointers << 1);




From lattner at cs.uiuc.edu  Tue Apr 22 13:43:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:43:01 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/llvmAsmParser.y
Message-ID: <200304221842.NAA07864@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

llvmAsmParser.y updated: 1.106 -> 1.107

---
Log message:

Namespacify


---
Diffs of the changes:

Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.106 llvm/lib/AsmParser/llvmAsmParser.y:1.107
--- llvm/lib/AsmParser/llvmAsmParser.y:1.106	Tue Apr 22 13:02:52 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Tue Apr 22 13:42:41 2003
@@ -17,20 +17,13 @@
 #include <list>
 #include <utility>
 #include <algorithm>
-using std::list;
-using std::vector;
-using std::pair;
-using std::map;
-using std::pair;
-using std::make_pair;
-using std::string;
 
 int yyerror(const char *ErrorMsg); // Forward declarations to prevent "implicit
 int yylex();                       // declaration" of xxx warnings.
 int yyparse();
 
 static Module *ParserResult;
-string CurFilename;
+std::string CurFilename;
 
 // DEBUG_UPREFS - Define this symbol if you want to enable debugging output
 // relating to upreferences in the input stream.
@@ -55,23 +48,24 @@
 // This contains info used when building the body of a function.  It is
 // destroyed when the function is completed.
 //
-typedef vector<Value *> ValueList;           // Numbered defs
-static void ResolveDefinitions(vector<ValueList> &LateResolvers,
-                               vector<ValueList> *FutureLateResolvers = 0);
+typedef std::vector<Value *> ValueList;           // Numbered defs
+static void ResolveDefinitions(std::vector<ValueList> &LateResolvers,
+                               std::vector<ValueList> *FutureLateResolvers = 0);
 
 static struct PerModuleInfo {
   Module *CurrentModule;
-  vector<ValueList>    Values;     // Module level numbered definitions
-  vector<ValueList>    LateResolveValues;
-  vector<PATypeHolder> Types;
-  map<ValID, PATypeHolder> LateResolveTypes;
+  std::vector<ValueList>    Values;     // Module level numbered definitions
+  std::vector<ValueList>    LateResolveValues;
+  std::vector<PATypeHolder> Types;
+  std::map<ValID, PATypeHolder> LateResolveTypes;
 
   // GlobalRefs - This maintains a mapping between <Type, ValID>'s and forward
   // references to global values.  Global values may be referenced before they
   // are defined, and if so, the temporary object that they represent is held
   // here.  This is used for forward references of ConstantPointerRefs.
   //
-  typedef map<pair<const PointerType *, ValID>, GlobalVariable*> GlobalRefsType;
+  typedef std::map<std::pair<const PointerType *,
+                             ValID>, GlobalVariable*> GlobalRefsType;
   GlobalRefsType GlobalRefs;
 
   void ModuleDone() {
@@ -85,7 +79,7 @@
     // resolved!
     //
     if (!GlobalRefs.empty()) {
-      string UndefinedReferences = "Unresolved global references exist:\n";
+      std::string UndefinedReferences = "Unresolved global references exist:\n";
       
       for (GlobalRefsType::iterator I = GlobalRefs.begin(), E =GlobalRefs.end();
            I != E; ++I) {
@@ -108,7 +102,8 @@
   void DeclareNewGlobalValue(GlobalValue *GV, ValID D) {
     // Check to see if there is a forward reference to this global variable...
     // if there is, eliminate it and patch the reference to use the new def'n.
-    GlobalRefsType::iterator I = GlobalRefs.find(make_pair(GV->getType(), D));
+    GlobalRefsType::iterator I =
+      GlobalRefs.find(std::make_pair(GV->getType(), D));
 
     if (I != GlobalRefs.end()) {
       GlobalVariable *OldGV = I->second;   // Get the placeholder...
@@ -139,10 +134,10 @@
 static struct PerFunctionInfo {
   Function *CurrentFunction;     // Pointer to current function being created
 
-  vector<ValueList> Values;      // Keep track of numbered definitions
-  vector<ValueList> LateResolveValues;
-  vector<PATypeHolder> Types;
-  map<ValID, PATypeHolder> LateResolveTypes;
+  std::vector<ValueList> Values;      // Keep track of numbered definitions
+  std::vector<ValueList> LateResolveValues;
+  std::vector<PATypeHolder> Types;
+  std::map<ValID, PATypeHolder> LateResolveTypes;
   bool isDeclare;                // Is this function a forward declararation?
 
   inline PerFunctionInfo() {
@@ -175,7 +170,8 @@
 //               Code to handle definitions of all the types
 //===----------------------------------------------------------------------===//
 
-static int InsertValue(Value *D, vector<ValueList> &ValueTab = CurMeth.Values) {
+static int InsertValue(Value *D,
+                       std::vector<ValueList> &ValueTab = CurMeth.Values) {
   if (D->hasName()) return -1;           // Is this a numbered definition?
 
   // Yes, insert the value into the value table...
@@ -188,7 +184,7 @@
 }
 
 // TODO: FIXME when Type are not const
-static void InsertType(const Type *Ty, vector<PATypeHolder> &Types) {
+static void InsertType(const Type *Ty, std::vector<PATypeHolder> &Types) {
   Types.push_back(Ty);
 }
 
@@ -209,7 +205,7 @@
     break;
   }
   case ValID::NameVal: {                // Is it a named definition?
-    string Name(D.Name);
+    std::string Name(D.Name);
     SymbolTable *SymTab = 0;
     Value *N = 0;
     if (inFunctionScope()) {
@@ -239,20 +235,20 @@
   //
   if (DoNotImprovise) return 0;  // Do we just want a null to be returned?
 
-  map<ValID, PATypeHolder> &LateResolver = inFunctionScope() ? 
+  std::map<ValID, PATypeHolder> &LateResolver = inFunctionScope() ? 
     CurMeth.LateResolveTypes : CurModule.LateResolveTypes;
   
-  map<ValID, PATypeHolder>::iterator I = LateResolver.find(D);
+  std::map<ValID, PATypeHolder>::iterator I = LateResolver.find(D);
   if (I != LateResolver.end()) {
     return I->second;
   }
 
   Type *Typ = OpaqueType::get();
-  LateResolver.insert(make_pair(D, Typ));
+  LateResolver.insert(std::make_pair(D, Typ));
   return Typ;
 }
 
-static Value *lookupInSymbolTable(const Type *Ty, const string &Name) {
+static Value *lookupInSymbolTable(const Type *Ty, const std::string &Name) {
   SymbolTable &SymTab = 
     inFunctionScope() ? CurMeth.CurrentFunction->getSymbolTable() :
                         CurModule.CurrentModule->getSymbolTable();
@@ -291,7 +287,7 @@
   }
 
   case ValID::NameVal: {                // Is it a named definition?
-    Value *N = lookupInSymbolTable(Ty, string(D.Name));
+    Value *N = lookupInSymbolTable(Ty, std::string(D.Name));
     if (N == 0) return 0;
 
     D.destroy();  // Free old strdup'd memory...
@@ -392,8 +388,8 @@
 // time (forward branches, phi functions for loops, etc...) resolve the 
 // defs now...
 //
-static void ResolveDefinitions(vector<ValueList> &LateResolvers,
-                               vector<ValueList> *FutureLateResolvers) {
+static void ResolveDefinitions(std::vector<ValueList> &LateResolvers,
+                               std::vector<ValueList> *FutureLateResolvers) {
   // Loop over LateResolveDefs fixing up stuff that couldn't be resolved
   for (unsigned ty = 0; ty < LateResolvers.size(); ty++) {
     while (!LateResolvers[ty].empty()) {
@@ -433,17 +429,17 @@
 // refering to the number can be resolved.  Do this now.
 //
 static void ResolveTypeTo(char *Name, const Type *ToTy) {
-  vector<PATypeHolder> &Types = inFunctionScope() ? 
+  std::vector<PATypeHolder> &Types = inFunctionScope() ? 
      CurMeth.Types : CurModule.Types;
 
    ValID D;
    if (Name) D = ValID::create(Name);
    else      D = ValID::create((int)Types.size());
 
-   map<ValID, PATypeHolder> &LateResolver = inFunctionScope() ? 
+   std::map<ValID, PATypeHolder> &LateResolver = inFunctionScope() ? 
      CurMeth.LateResolveTypes : CurModule.LateResolveTypes;
   
-   map<ValID, PATypeHolder>::iterator I = LateResolver.find(D);
+   std::map<ValID, PATypeHolder>::iterator I = LateResolver.find(D);
    if (I != LateResolver.end()) {
      ((DerivedType*)I->second.get())->refineAbstractTypeTo(ToTy);
      LateResolver.erase(I);
@@ -453,7 +449,7 @@
 // ResolveTypes - At this point, all types should be resolved.  Any that aren't
 // are errors.
 //
-static void ResolveTypes(map<ValID, PATypeHolder> &LateResolveTypes) {
+static void ResolveTypes(std::map<ValID, PATypeHolder> &LateResolveTypes) {
   if (!LateResolveTypes.empty()) {
     const ValID &DID = LateResolveTypes.begin()->first;
 
@@ -476,7 +472,7 @@
 static bool setValueName(Value *V, char *NameStr) {
   if (NameStr == 0) return false;
   
-  string Name(NameStr);           // Copy string
+  std::string Name(NameStr);      // Copy string
   free(NameStr);                  // Free old string
 
   if (V->getType() == Type::VoidTy) 
@@ -549,7 +545,7 @@
 }
 
 
-static vector<pair<unsigned, OpaqueType *> > UpRefs;
+static std::vector<std::pair<unsigned, OpaqueType *> > UpRefs;
 
 static PATypeHolder HandleUpRefs(const Type *ty) {
   PATypeHolder Ty(ty);
@@ -566,7 +562,7 @@
       if (Level == 0) {                     // Upreference should be resolved! 
 	UR_OUT("  * Resolving upreference for "
                << UpRefs[i].second->getDescription() << endl;
-	       string OldName = UpRefs[i].second->getDescription());
+	       std::string OldName = UpRefs[i].second->getDescription());
 	UpRefs[i].second->refineAbstractTypeTo(Ty);
 	UpRefs.erase(UpRefs.begin()+i);     // Remove from upreference list...
 	UR_OUT("  * Type '" << OldName << "' refined upreference to: "
@@ -586,7 +582,7 @@
 //            RunVMAsmParser - Define an interface to this parser
 //===----------------------------------------------------------------------===//
 //
-Module *RunVMAsmParser(const string &Filename, FILE *F) {
+Module *RunVMAsmParser(const std::string &Filename, FILE *F) {
   llvmAsmin = F;
   CurFilename = Filename;
   llvmAsmlineno = 1;      // Reset the current line number...
@@ -792,12 +788,12 @@
 UpRTypes : '\\' EUINT64VAL {                   // Type UpReference
     if ($2 > (uint64_t)INT64_MAX) ThrowException("Value out of range!");
     OpaqueType *OT = OpaqueType::get();        // Use temporary placeholder
-    UpRefs.push_back(make_pair((unsigned)$2, OT));  // Add to vector...
+    UpRefs.push_back(std::make_pair((unsigned)$2, OT));  // Add to vector...
     $$ = new PATypeHolder(OT);
     UR_OUT("New Upreference!\n");
   }
   | UpRTypesV '(' ArgTypeListI ')' {           // Function derived type?
-    vector<const Type*> Params;
+    std::vector<const Type*> Params;
     mapto($3->begin(), $3->end(), std::back_inserter(Params), 
 	  std::mem_fun_ref(&PATypeHandle<Type>::get));
     bool isVarArg = Params.size() && Params.back() == Type::VoidTy;
@@ -812,7 +808,7 @@
     delete $4;
   }
   | '{' TypeListI '}' {                        // Structure type?
-    vector<const Type*> Elements;
+    std::vector<const Type*> Elements;
     mapto($2->begin(), $2->end(), std::back_inserter(Elements), 
 	std::mem_fun_ref(&PATypeHandle<Type>::get));
 
@@ -820,7 +816,7 @@
     delete $2;
   }
   | '{' '}' {                                  // Empty structure type?
-    $$ = new PATypeHolder(StructType::get(vector<const Type*>()));
+    $$ = new PATypeHolder(StructType::get(std::vector<const Type*>()));
   }
   | UpRTypes '*' {                             // Pointer type?
     $$ = new PATypeHolder(HandleUpRefs(PointerType::get(*$1)));
@@ -831,7 +827,7 @@
 // declaration type lists
 //
 TypeListI : UpRTypes {
-    $$ = new list<PATypeHolder>();
+    $$ = new std::list<PATypeHolder>();
     $$->push_back(*$1); delete $1;
   }
   | TypeListI ',' UpRTypes {
@@ -844,10 +840,10 @@
     ($$=$1)->push_back(Type::VoidTy);
   }
   | DOTDOTDOT {
-    ($$ = new list<PATypeHolder>())->push_back(Type::VoidTy);
+    ($$ = new std::list<PATypeHolder>())->push_back(Type::VoidTy);
   }
   | /*empty*/ {
-    $$ = new list<PATypeHolder>();
+    $$ = new std::list<PATypeHolder>();
   };
 
 // ConstVal - The various declarations that go into the constant pool.  This
@@ -891,7 +887,7 @@
     if (NumElements != -1 && NumElements != 0) 
       ThrowException("Type mismatch: constant sized array initialized with 0"
 		     " arguments, but has size of " + itostr(NumElements) +"!");
-    $$ = ConstantArray::get(ATy, vector<Constant*>());
+    $$ = ConstantArray::get(ATy, std::vector<Constant*>());
     delete $1;
   }
   | Types 'c' STRINGCONSTANT {
@@ -907,7 +903,7 @@
       ThrowException("Can't build string constant of size " + 
 		     itostr((int)(EndStr-$3)) +
 		     " when array has size " + itostr(NumElements) + "!");
-    vector<Constant*> Vals;
+    std::vector<Constant*> Vals;
     if (ETy == Type::SByteTy) {
       for (char *C = $3; C != EndStr; ++C)
 	Vals.push_back(ConstantSInt::get(ETy, *C));
@@ -978,7 +974,7 @@
 
       // First check to see if the forward references value is already created!
       PerModuleInfo::GlobalRefsType::iterator I =
-	CurModule.GlobalRefs.find(make_pair(PT, $2));
+	CurModule.GlobalRefs.find(std::make_pair(PT, $2));
     
       if (I != CurModule.GlobalRefs.end()) {
 	V = I->second;             // Placeholder already exists, use it...
@@ -991,7 +987,7 @@
                                                 false,
                                                 GlobalValue::ExternalLinkage);
 	// Keep track of the fact that we have a forward ref to recycle it
-	CurModule.GlobalRefs.insert(make_pair(make_pair(PT, $2), GV));
+	CurModule.GlobalRefs.insert(std::make_pair(std::make_pair(PT, $2), GV));
 
 	// Must temporarily push this value into the module table...
 	CurModule.CurrentModule->getGlobalList().push_back(GV);
@@ -1044,7 +1040,7 @@
     if (!IdxTy)
       ThrowException("Index list invalid for constant getelementptr!");
 
-    vector<Constant*> IdxVec;
+    std::vector<Constant*> IdxVec;
     for (unsigned i = 0, e = $4->size(); i != e; ++i)
       if (Constant *C = dyn_cast<Constant>((*$4)[i]))
         IdxVec.push_back(C);
@@ -1072,7 +1068,7 @@
     ($$ = $1)->push_back($3);
   }
   | ConstVal {
-    $$ = new vector<Constant*>();
+    $$ = new std::vector<Constant*>();
     $$->push_back($1);
   };
 
@@ -1194,7 +1190,7 @@
 ArgVal : Types OptVAR_ID {
   if (*$1 == Type::VoidTy)
     ThrowException("void typed arguments are invalid!");
-  $$ = new pair<PATypeHolder*, char*>($1, $2);
+  $$ = new std::pair<PATypeHolder*, char*>($1, $2);
 };
 
 ArgListH : ArgListH ',' ArgVal {
@@ -1203,7 +1199,7 @@
     delete $3;
   }
   | ArgVal {
-    $$ = new vector<pair<PATypeHolder*,char*> >();
+    $$ = new std::vector<std::pair<PATypeHolder*,char*> >();
     $$->push_back(*$1);
     delete $1;
   };
@@ -1213,11 +1209,12 @@
   }
   | ArgListH ',' DOTDOTDOT {
     $$ = $1;
-    $$->push_back(pair<PATypeHolder*, char*>(new PATypeHolder(Type::VoidTy),0));
+    $$->push_back(std::pair<PATypeHolder*,
+                            char*>(new PATypeHolder(Type::VoidTy), 0));
   }
   | DOTDOTDOT {
-    $$ = new vector<pair<PATypeHolder*,char*> >();
-    $$->push_back(pair<PATypeHolder*, char*>(new PATypeHolder(Type::VoidTy),0));
+    $$ = new std::vector<std::pair<PATypeHolder*,char*> >();
+    $$->push_back(std::make_pair(new PATypeHolder(Type::VoidTy), (char*)0));
   }
   | /* empty */ {
     $$ = 0;
@@ -1227,11 +1224,11 @@
 
 FunctionHeaderH : TypesV FuncName '(' ArgList ')' {
   UnEscapeLexed($2);
-  string FunctionName($2);
+  std::string FunctionName($2);
   
-  vector<const Type*> ParamTypeList;
+  std::vector<const Type*> ParamTypeList;
   if ($4) {   // If there are arguments...
-    for (vector<pair<PATypeHolder*,char*> >::iterator I = $4->begin();
+    for (std::vector<std::pair<PATypeHolder*,char*> >::iterator I = $4->begin();
          I != $4->end(); ++I)
       ParamTypeList.push_back(I->first->get());
   }
@@ -1279,7 +1276,7 @@
       $4->pop_back();  // Delete the last entry
     }
     Function::aiterator ArgIt = Fn->abegin();
-    for (vector<pair<PATypeHolder*, char*> >::iterator I = $4->begin();
+    for (std::vector<std::pair<PATypeHolder*, char*> >::iterator I =$4->begin();
          I != $4->end(); ++I, ++ArgIt) {
       delete I->first;                          // Delete the typeholder...
 
@@ -1423,7 +1420,7 @@
                                    cast<BasicBlock>(getVal(Type::LabelTy, $6)));
     $$ = S;
 
-    vector<pair<Constant*,BasicBlock*> >::iterator I = $8->begin(),
+    std::vector<std::pair<Constant*,BasicBlock*> >::iterator I = $8->begin(),
       E = $8->end();
     for (; I != E; ++I)
       S->dest_push_back(I->first, I->second);
@@ -1436,9 +1433,10 @@
     if (!(PFTy = dyn_cast<PointerType>($2->get())) ||
         !(Ty = dyn_cast<FunctionType>(PFTy->getElementType()))) {
       // Pull out the types of all of the arguments...
-      vector<const Type*> ParamTypes;
+      std::vector<const Type*> ParamTypes;
       if ($5) {
-        for (vector<Value*>::iterator I = $5->begin(), E = $5->end(); I!=E; ++I)
+        for (std::vector<Value*>::iterator I = $5->begin(), E = $5->end();
+             I != E; ++I)
           ParamTypes.push_back((*I)->getType());
       }
 
@@ -1460,14 +1458,14 @@
 
     // Create the call node...
     if (!$5) {                                   // Has no arguments?
-      $$ = new InvokeInst(V, Normal, Except, vector<Value*>());
+      $$ = new InvokeInst(V, Normal, Except, std::vector<Value*>());
     } else {                                     // Has arguments?
       // Loop through FunctionType's arguments and ensure they are specified
       // correctly!
       //
       FunctionType::ParamTypes::const_iterator I = Ty->getParamTypes().begin();
       FunctionType::ParamTypes::const_iterator E = Ty->getParamTypes().end();
-      vector<Value*>::iterator ArgI = $5->begin(), ArgE = $5->end();
+      std::vector<Value*>::iterator ArgI = $5->begin(), ArgE = $5->end();
 
       for (; ArgI != ArgE && I != E; ++ArgI, ++I)
 	if ((*ArgI)->getType() != *I)
@@ -1490,16 +1488,16 @@
     if (V == 0)
       ThrowException("May only switch on a constant pool value!");
 
-    $$->push_back(make_pair(V, cast<BasicBlock>(getVal($5, $6))));
+    $$->push_back(std::make_pair(V, cast<BasicBlock>(getVal($5, $6))));
   }
   | IntType ConstValueRef ',' LABEL ValueRef {
-    $$ = new vector<pair<Constant*, BasicBlock*> >();
+    $$ = new std::vector<std::pair<Constant*, BasicBlock*> >();
     Constant *V = cast<Constant>(getValNonImprovising($1, $2));
 
     if (V == 0)
       ThrowException("May only switch on a constant pool value!");
 
-    $$->push_back(make_pair(V, cast<BasicBlock>(getVal($4, $5))));
+    $$->push_back(std::make_pair(V, cast<BasicBlock>(getVal($4, $5))));
   };
 
 Inst : OptAssign InstVal {
@@ -1510,20 +1508,20 @@
 };
 
 PHIList : Types '[' ValueRef ',' ValueRef ']' {    // Used for PHI nodes
-    $$ = new list<pair<Value*, BasicBlock*> >();
-    $$->push_back(make_pair(getVal(*$1, $3), 
-                            cast<BasicBlock>(getVal(Type::LabelTy, $5))));
+    $$ = new std::list<std::pair<Value*, BasicBlock*> >();
+    $$->push_back(std::make_pair(getVal(*$1, $3), 
+                                 cast<BasicBlock>(getVal(Type::LabelTy, $5))));
     delete $1;
   }
   | PHIList ',' '[' ValueRef ',' ValueRef ']' {
     $$ = $1;
-    $1->push_back(make_pair(getVal($1->front().first->getType(), $4),
-                            cast<BasicBlock>(getVal(Type::LabelTy, $6))));
+    $1->push_back(std::make_pair(getVal($1->front().first->getType(), $4),
+                                 cast<BasicBlock>(getVal(Type::LabelTy, $6))));
   };
 
 
 ValueRefList : ResolvedVal {    // Used for call statements, and memory insts...
-    $$ = new vector<Value*>();
+    $$ = new std::vector<Value*>();
     $$->push_back($1);
   }
   | ValueRefList ',' ResolvedVal {
@@ -1595,9 +1593,10 @@
     if (!(PFTy = dyn_cast<PointerType>($2->get())) ||
         !(Ty = dyn_cast<FunctionType>(PFTy->getElementType()))) {
       // Pull out the types of all of the arguments...
-      vector<const Type*> ParamTypes;
+      std::vector<const Type*> ParamTypes;
       if ($5) {
-        for (vector<Value*>::iterator I = $5->begin(), E = $5->end(); I!=E; ++I)
+        for (std::vector<Value*>::iterator I = $5->begin(), E = $5->end();
+             I != E; ++I)
           ParamTypes.push_back((*I)->getType());
       }
 
@@ -1618,14 +1617,14 @@
         ThrowException("No arguments passed to a function that "
                        "expects arguments!");
 
-      $$ = new CallInst(V, vector<Value*>());
+      $$ = new CallInst(V, std::vector<Value*>());
     } else {                                     // Has arguments?
       // Loop through FunctionType's arguments and ensure they are specified
       // correctly!
       //
       FunctionType::ParamTypes::const_iterator I = Ty->getParamTypes().begin();
       FunctionType::ParamTypes::const_iterator E = Ty->getParamTypes().end();
-      vector<Value*>::iterator ArgI = $5->begin(), ArgE = $5->end();
+      std::vector<Value*>::iterator ArgI = $5->begin(), ArgE = $5->end();
 
       for (; ArgI != ArgE && I != E; ++ArgI, ++I)
 	if ((*ArgI)->getType() != *I)
@@ -1648,7 +1647,7 @@
 IndexList : ',' ValueRefList { 
   $$ = $2; 
 } | /* empty */ { 
-  $$ = new vector<Value*>(); 
+  $$ = new std::vector<Value*>(); 
 };
 
 MemoryInst : MALLOC Types {
@@ -1744,13 +1743,14 @@
 
 %%
 int yyerror(const char *ErrorMsg) {
-  string where  = string((CurFilename == "-")? string("<stdin>") : CurFilename)
+  std::string where 
+    = std::string((CurFilename == "-") ? std::string("<stdin>") : CurFilename)
                   + ":" + utostr((unsigned) llvmAsmlineno) + ": ";
-  string errMsg = string(ErrorMsg) + string("\n") + where + " while reading ";
+  std::string errMsg = std::string(ErrorMsg) + "\n" + where + " while reading ";
   if (yychar == YYEMPTY)
     errMsg += "end-of-file.";
   else
-    errMsg += "token: '" + string(llvmAsmtext, llvmAsmleng) + "'";
+    errMsg += "token: '" + std::string(llvmAsmtext, llvmAsmleng) + "'";
   ThrowException(errMsg);
   return 0;
 }




From lattner at cs.uiuc.edu  Tue Apr 22 13:44:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 13:44:01 2003
Subject: [llvm-commits] CVS: llvm/test/Feature/README.txt
Message-ID: <200304221843.NAA07890@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Feature:

README.txt updated: 1.1 -> 1.2

---
Log message:

ADd newline to end of file


---
Diffs of the changes:

Index: llvm/test/Feature/README.txt
diff -u llvm/test/Feature/README.txt:1.1 llvm/test/Feature/README.txt:1.2
--- llvm/test/Feature/README.txt:1.1	Wed Jan 23 13:42:36 2002
+++ llvm/test/Feature/README.txt	Tue Apr 22 13:43:23 2003
@@ -2,4 +2,5 @@
 It is designed to make sure that the major components of LLVM support all of the
 features of LLVM, for very small examples.  Entire programs should not go here.
 
-Regression tests for individual bug fixes should go into the test/Regression dir.
\ No newline at end of file
+Regression tests for individual bug fixes should go into the test/Regression dir.
+




From lattner at cs.uiuc.edu  Tue Apr 22 14:07:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:07:01 2003
Subject: [llvm-commits] CVS: llvm/test/Feature/properties.ll
Message-ID: <200304221906.OAA08466@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Feature:

properties.ll added (r1.1)

---
Log message:

Add testcase for target information


---
Diffs of the changes:

Index: llvm/test/Feature/properties.ll
diff -c /dev/null llvm/test/Feature/properties.ll:1.1
*** /dev/null	Tue Apr 22 14:06:58 2003
--- llvm/test/Feature/properties.ll	Tue Apr 22 14:06:48 2003
***************
*** 0 ****
--- 1,4 ----
+ 
+ target endian = little
+ target pointersize = 32
+ 




From lattner at cs.uiuc.edu  Tue Apr 22 14:08:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:08:01 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/Lexer.l llvmAsmParser.y
Message-ID: <200304221907.OAA08482@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

Lexer.l updated: 1.31 -> 1.32
llvmAsmParser.y updated: 1.107 -> 1.108

---
Log message:

Allow information about the target to be specified in the .ll file


---
Diffs of the changes:

Index: llvm/lib/AsmParser/Lexer.l
diff -u llvm/lib/AsmParser/Lexer.l:1.31 llvm/lib/AsmParser/Lexer.l:1.32
--- llvm/lib/AsmParser/Lexer.l:1.31	Thu Apr 17 17:17:32 2003
+++ llvm/lib/AsmParser/Lexer.l	Tue Apr 22 14:07:06 2003
@@ -19,6 +19,7 @@
 
 %{
 #include "ParserInternals.h"
+#include "llvm/Module.h"
 #include <list>
 #include "llvmAsmParser.h"
 #include <ctype.h>
@@ -176,6 +177,11 @@
 to              { return TO; }
 except          { return EXCEPT; }
 not             { return NOT; }  /* Deprecated, turned into XOR */
+target          { return TARGET; }
+endian          { return ENDIAN; }
+pointersize     { return POINTERSIZE; }
+little          { return LITTLE; }
+big             { return BIG; }
 
 void            { llvmAsmlval.PrimType = Type::VoidTy  ; return VOID;   }
 bool            { llvmAsmlval.PrimType = Type::BoolTy  ; return BOOL;   }


Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.107 llvm/lib/AsmParser/llvmAsmParser.y:1.108
--- llvm/lib/AsmParser/llvmAsmParser.y:1.107	Tue Apr 22 13:42:41 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Tue Apr 22 14:07:06 2003
@@ -635,6 +635,7 @@
   Instruction::TermOps              TermOpVal;
   Instruction::MemoryOps            MemOpVal;
   Instruction::OtherOps             OtherOpVal;
+  Module::Endianness                Endianness;
 }
 
 %type <ModuleVal>     Module FunctionList
@@ -653,6 +654,7 @@
 %type <JumpTable>     JumpTable
 %type <BoolVal>       GlobalType                  // GLOBAL or CONSTANT?
 %type <Linkage>       OptLinkage
+%type <Endianness>    BigOrLittle
 
 // ValueRef - Unresolved reference to a definition or BB
 %type <ValIDVal>      ValueRef ConstValueRef SymbolicValueRef
@@ -683,7 +685,7 @@
 
 %token IMPLEMENTATION TRUE FALSE BEGINTOK ENDTOK DECLARE GLOBAL CONSTANT
 %token TO EXCEPT DOTDOTDOT NULL_TOK CONST INTERNAL LINKONCE APPENDING
-%token OPAQUE NOT EXTERNAL
+%token OPAQUE NOT EXTERNAL TARGET ENDIAN POINTERSIZE LITTLE BIG
 
 // Basic Block Terminating Operators 
 %token <TermOpVal> RET BR SWITCH
@@ -1177,7 +1179,26 @@
     }
     delete $5;
   }
+  | ConstPool TARGET TargetDefinition { 
+  }
   | /* empty: end of list */ { 
+  };
+
+
+
+BigOrLittle : BIG    { $$ = Module::BigEndian; };
+BigOrLittle : LITTLE { $$ = Module::LittleEndian; };
+
+TargetDefinition : ENDIAN '=' BigOrLittle {
+    CurModule.CurrentModule->setEndianness($3);
+  }
+  | POINTERSIZE '=' EUINT64VAL {
+    if ($3 == 32)
+      CurModule.CurrentModule->setPointerSize(Module::Pointer32);
+    else if ($3 == 64)
+      CurModule.CurrentModule->setPointerSize(Module::Pointer64);
+    else
+      ThrowException("Invalid pointer size: '" + utostr($3) + "'!");
   };
 
 




From lattner at cs.uiuc.edu  Tue Apr 22 14:08:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:08:03 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/AsmWriter.cpp
Message-ID: <200304221907.OAA08492@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

AsmWriter.cpp updated: 1.83 -> 1.84

---
Log message:

Emit information about target


---
Diffs of the changes:

Index: llvm/lib/VMCore/AsmWriter.cpp
diff -u llvm/lib/VMCore/AsmWriter.cpp:1.83 llvm/lib/VMCore/AsmWriter.cpp:1.84
--- llvm/lib/VMCore/AsmWriter.cpp:1.83	Wed Apr 16 15:28:45 2003
+++ llvm/lib/VMCore/AsmWriter.cpp	Tue Apr 22 14:07:19 2003
@@ -519,6 +519,9 @@
 
 
 void AssemblyWriter::printModule(const Module *M) {
+  Out << "target endian = " << (M->isLittleEndian() ? "little" : "big") << "\n";
+  Out << "target pointersize = " << (M->has32BitPointers() ? 32 : 64) << "\n";
+
   // Loop over the symbol table, emitting all named constants...
   printSymbolTable(M->getSymbolTable());
   




From lattner at cs.uiuc.edu  Tue Apr 22 14:14:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:14:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/Linker.cpp
Message-ID: <200304221913.OAA08928@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

Linker.cpp updated: 1.40 -> 1.41

---
Log message:

Add warning when linking modules with disagreeing target properties


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/Linker.cpp
diff -u llvm/lib/Transforms/Utils/Linker.cpp:1.40 llvm/lib/Transforms/Utils/Linker.cpp:1.41
--- llvm/lib/Transforms/Utils/Linker.cpp:1.40	Mon Apr 21 16:15:04 2003
+++ llvm/lib/Transforms/Utils/Linker.cpp	Tue Apr 22 14:13:20 2003
@@ -469,6 +469,10 @@
 // shouldn't be relied on to be consistent.
 //
 bool LinkModules(Module *Dest, const Module *Src, std::string *ErrorMsg) {
+  if (Dest->getEndianness() != Src->getEndianness())
+    std::cerr << "WARNING: Linking two modules of different endianness!\n";
+  if (Dest->getPointerSize() != Src->getPointerSize())
+    std::cerr << "WARNING: Linking two modules of different pointer size!\n";
 
   // LinkTypes - Go through the symbol table of the Src module and see if any
   // types are named in the src module that are not named in the Dst module.




From lattner at cs.uiuc.edu  Tue Apr 22 14:33:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:33:01 2003
Subject: [llvm-commits] CVS: llvm/utils/vim/llvm.vim
Message-ID: <200304221932.OAA10832@apoc.cs.uiuc.edu>


Changes in directory llvm/utils/vim:

llvm.vim updated: 1.2 -> 1.3

---
Log message:

Update to add new keywords


---
Diffs of the changes:

Index: llvm/utils/vim/llvm.vim
diff -u llvm/utils/vim/llvm.vim:1.2 llvm/utils/vim/llvm.vim:1.3
--- llvm/utils/vim/llvm.vim:1.2	Wed Apr 16 15:25:57 2003
+++ llvm/utils/vim/llvm.vim	Tue Apr 22 14:31:53 2003
@@ -1,7 +1,7 @@
 " Vim syntax file
 " Language:	llvm
-" Maintainer:	Nick Hildenbrandt <hldnbrnd at uiuc.edu>
-" Updated:	2002-10-07
+" Maintainer:	?
+" Updated:	2003-04-22
 
 if version < 600
   syntax clear
@@ -28,7 +28,8 @@
 syn keyword llvmStatement declare global constant const
 syn keyword llvmStatement internal uninitialized external implementation
 syn keyword llvmStatement linkonce appending
-syn keyword llvmStatement null to except not
+syn keyword llvmStatement null to except not target endian pointersize
+syn keyword llvmStatement big little
 
 "syn match llvmFunction /%[a-zA-Z\$._\-][a-zA-Z\$._\-0-9]*/
 syn match llvmNumber /\<\d\+\>/




From lattner at cs.uiuc.edu  Tue Apr 22 14:33:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 14:33:03 2003
Subject: [llvm-commits] CVS: llvm/utils/emacs/llvm-mode.el
Message-ID: <200304221932.OAA10839@apoc.cs.uiuc.edu>


Changes in directory llvm/utils/emacs:

llvm-mode.el updated: 1.2 -> 1.3

---
Log message:

Update to add new keywords



---
Diffs of the changes:

Index: llvm/utils/emacs/llvm-mode.el
diff -u llvm/utils/emacs/llvm-mode.el:1.2 llvm/utils/emacs/llvm-mode.el:1.3
--- llvm/utils/emacs/llvm-mode.el:1.2	Wed Apr 16 15:26:11 2003
+++ llvm/utils/emacs/llvm-mode.el	Tue Apr 22 14:31:55 2003
@@ -2,7 +2,7 @@
 ;; Description:
 ;;         Major mode for the LLVM assembler language.
 ;; Updated:
-;;         Oct 8, 2002
+;;         Apr 22, 2003
 ;; Code:
 
 ;; Create mode-specific tables.
@@ -28,7 +28,7 @@
    ;; Hex constants
    '("0x[0-9A-Fa-f]+" . font-lock-preprocessor-face)
    ;; Keywords
-   '("begin\\|end\\|true\\|false\\|declare\\|global\\|constant\\|const\\|internal\\|linkonce\\|appending\\|uninitialized\\|implementation\\|\\.\\.\\.\\|null\\|to\\|except\\|not" . font-lock-keyword-face)
+   '("begin\\|end\\|true\\|false\\|declare\\|global\\|constant\\|const\\|internal\\|linkonce\\|appending\\|uninitialized\\|implementation\\|\\.\\.\\.\\|null\\|to\\|except\\|not\\|target\\|endian\\|little\\|big\\|pointersize" . font-lock-keyword-face)
    ;; Types
    '("void\\|bool\\|sbyte\\|ubyte\\|u?short\\|u?int\\|u?long\\|float\\|double\\|type\\|label\\|opaque" . font-lock-type-face)
    ;; Arithmetic and Logical Operators




From lattner at cs.uiuc.edu  Tue Apr 22 15:21:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 15:21:00 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200304222020.PAA22144@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.77 -> 1.78

---
Log message:

Add support for the switch instruction to the CWriter


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.77 llvm/lib/CWriter/Writer.cpp:1.78
--- llvm/lib/CWriter/Writer.cpp:1.77	Wed Feb 12 14:45:00 2003
+++ llvm/lib/CWriter/Writer.cpp	Tue Apr 22 15:19:52 2003
@@ -7,11 +7,7 @@
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/Module.h"
-#include "llvm/iMemory.h"
-#include "llvm/iTerminators.h"
-#include "llvm/iPHINode.h"
-#include "llvm/iOther.h"
-#include "llvm/iOperators.h"
+#include "llvm/Instructions.h"
 #include "llvm/Pass.h"
 #include "llvm/SymbolTable.h"
 #include "llvm/SlotCalculator.h"
@@ -108,6 +104,7 @@
 
     void visitReturnInst(ReturnInst &I);
     void visitBranchInst(BranchInst &I);
+    void visitSwitchInst(SwitchInst &I);
 
     void visitPHINode(PHINode &I) {}
     void visitBinaryOperator(Instruction &I);
@@ -156,8 +153,8 @@
 string CWriter::getValueName(const Value *V) {
   if (V->hasName()) {              // Print out the label if it exists...
     if (isa<GlobalValue>(V) &&     // Do not mangle globals...
-        cast<GlobalValue>(V)->hasExternalLinkage())// && // Unless it's internal or
-        //!MangledGlobals.count(V))  // Unless the name would collide if we don't
+        cast<GlobalValue>(V)->hasExternalLinkage()) // Unless it's internal or
+      //!MangledGlobals.count(V))  // Unless the name would collide if we don't
       return makeNameProper(V->getName());
 
     return "l" + utostr(V->getType()->getUniqueID()) + "_" +
@@ -866,6 +863,25 @@
   }
   Out << ";\n";
 }
+
+void CWriter::visitSwitchInst(SwitchInst &SI) {
+  Out << "  switch (";
+  writeOperand(SI.getOperand(0));
+  Out << ") {\n  default: goto ";
+  writeOperand(SI.getDefaultDest());
+  Out << ";\n";
+  for (unsigned i = 2, e = SI.getNumOperands(); i != e; i += 2) {
+    Out << "  case ";
+    writeOperand(SI.getOperand(i));
+    Out << ":\n";
+    BasicBlock *Succ = cast<BasicBlock>(SI.getOperand(i+1));
+    printBranchToBlock(SI.getParent(), Succ, 2);
+    if (Succ == SI.getParent()->getNext())
+      Out << "    break;\n";
+  }
+  Out << "  }\n";
+}
+
 
 static bool isGotoCodeNeccessary(BasicBlock *From, BasicBlock *To) {
   // If PHI nodes need copies, we need the copy code...




From brukman at cs.uiuc.edu  Tue Apr 22 15:21:03 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Tue Apr 22 15:21:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86RegisterInfo.def
Message-ID: <200304222020.PAA22158@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86RegisterInfo.def updated: 1.12 -> 1.13

---
Log message:

It's "necessary"...


---
Diffs of the changes:

Index: llvm/lib/Target/X86/X86RegisterInfo.def
diff -u llvm/lib/Target/X86/X86RegisterInfo.def:1.12 llvm/lib/Target/X86/X86RegisterInfo.def:1.13
--- llvm/lib/Target/X86/X86RegisterInfo.def:1.12	Sun Jan 12 18:50:46 2003
+++ llvm/lib/Target/X86/X86RegisterInfo.def	Tue Apr 22 15:20:11 2003
@@ -45,7 +45,7 @@
 //  #3: Register Flags - A bitfield of flags or'd together from the
 //      MRegisterInfo.h file.
 //  #4: Target Specific Flags - Another bitfield containing X86 specific flags
-//      as neccesary.
+//      as necessary.
 //  #5: Alias set for registers aliased to this register (sets defined below).
 
 




From lattner at cs.uiuc.edu  Tue Apr 22 15:21:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 15:21:04 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/Lexer.l
Message-ID: <200304222020.PAA22169@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

Lexer.l updated: 1.32 -> 1.33

---
Log message:

Use a union to cast int to fp


---
Diffs of the changes:

Index: llvm/lib/AsmParser/Lexer.l
diff -u llvm/lib/AsmParser/Lexer.l:1.32 llvm/lib/AsmParser/Lexer.l:1.33
--- llvm/lib/AsmParser/Lexer.l:1.32	Tue Apr 22 14:07:06 2003
+++ llvm/lib/AsmParser/Lexer.l	Tue Apr 22 15:20:28 2003
@@ -73,15 +73,17 @@
 // point representation of it.
 //
 static double HexToFP(const char *Buffer) {
-  uint64_t Result = HexIntToVal(Buffer);
-
-  assert(sizeof(double) == sizeof(Result) &&
-         "Data sizes incompatible on this target!");
   // Behave nicely in the face of C TBAA rules... see:
   // http://www.nullstone.com/htmls/category/aliastyp.htm
-  //
-  char *ProxyPointer = (char*)&Result;
-  return *(double*)ProxyPointer;   // Cast Hex constant to double
+  union {
+    uint64_t UI;
+    double FP;
+  } UIntToFP;
+  UIntToFP.UI = HexIntToVal(Buffer);
+
+  assert(sizeof(double) == sizeof(uint64_t) &&
+         "Data sizes incompatible on this target!");
+  return UIntToFP.FP;   // Cast Hex constant to double
 }
 
 




From lattner at cs.uiuc.edu  Tue Apr 22 15:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 15:27:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c
Message-ID: <200304222026.PAA22499@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource/UnitTests:

2003-04-22-Switch.c added (r1.1)

---
Log message:

Add new testcase for switch statement/instruction


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c
diff -c /dev/null llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c:1.1
*** /dev/null	Tue Apr 22 15:26:08 2003
--- llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c	Tue Apr 22 15:25:58 2003
***************
*** 0 ****
--- 1,22 ----
+ 
+ #include <stdio.h>
+ 
+ static void func(unsigned i) {
+   switch (i) {
+   case 8: printf("C\n"); /* fall through. */
+   case 0:
+   case 3:
+   case 2: printf("A\n"); break;
+   case 1:
+   case 7:
+   case 9: printf("B\n"); break;
+   default: printf("D\n");
+   }
+ }
+ 
+ 
+ void main() {
+   unsigned i;
+   for (i = 0; i < 10; ++i)
+     func(i);
+ }




From lattner at cs.uiuc.edu  Tue Apr 22 15:35:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 15:35:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304222034.PAA22958@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.77 -> 1.78

---
Log message:

Add support to LLI for switch instruction


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.77 llvm/tools/lli/Interpreter/Execution.cpp:1.78
--- llvm/tools/lli/Interpreter/Execution.cpp:1.77	Mon Apr 21 17:43:32 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Tue Apr 22 15:34:47 2003
@@ -635,6 +635,27 @@
   SF.CurInst = SF.CurBB->begin();     // Update new instruction ptr...
 }
 
+static void executeSwitch(SwitchInst &I, ExecutionContext &SF) {
+  GenericValue CondVal = getOperandValue(I.getOperand(0), SF);
+  const Type *ElTy = I.getOperand(0)->getType();
+  SF.PrevBB = SF.CurBB;               // Update PrevBB so that PHI nodes work...
+  BasicBlock *Dest = 0;
+
+  // Check to see if any of the cases match...
+  for (unsigned i = 2, e = I.getNumOperands(); i != e; i += 2) {
+    if (executeSetEQInst(CondVal,
+                         getOperandValue(I.getOperand(i), SF),ElTy,SF).BoolVal){
+      Dest = cast<BasicBlock>(I.getOperand(i+1));
+      break;
+    }
+  }
+  
+  if (!Dest) Dest = I.getDefaultDest();   // No cases matched: use default
+  SF.CurBB = Dest;                        // Update CurBB to branch destination
+  SF.CurInst = SF.CurBB->begin();         // Update new instruction ptr...
+}
+
+
 //===----------------------------------------------------------------------===//
 //                     Memory Instruction Implementations
 //===----------------------------------------------------------------------===//
@@ -1106,6 +1127,7 @@
       // Terminators
     case Instruction::Ret:     executeRetInst  (cast<ReturnInst>(I), SF); break;
     case Instruction::Br:      executeBrInst   (cast<BranchInst>(I), SF); break;
+    case Instruction::Switch:  executeSwitch   (cast<SwitchInst>(I), SF); break;
       // Memory Instructions
     case Instruction::Alloca:
     case Instruction::Malloc:  executeAllocInst((AllocationInst&)I, SF); break;




From lattner at cs.uiuc.edu  Tue Apr 22 15:38:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 15:38:05 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304222037.PAA23062@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.78 -> 1.79

---
Log message:

Get rid of extraneous arguments to implementation functions


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.78 llvm/tools/lli/Interpreter/Execution.cpp:1.79
--- llvm/tools/lli/Interpreter/Execution.cpp:1.78	Tue Apr 22 15:34:47 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Tue Apr 22 15:37:39 2003
@@ -93,7 +93,7 @@
 static GenericValue executeCastOperation(Value *Src, const Type *DestTy,
                                          ExecutionContext &SF);
 static GenericValue executeAddInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF);
+				   const Type *Ty);
 
 
 static GenericValue getOperandValue(Value *V, ExecutionContext &SF) {
@@ -107,7 +107,7 @@
     case Instruction::Add:
       return executeAddInst(getOperandValue(CE->getOperand(0), SF),
                             getOperandValue(CE->getOperand(1), SF),
-                            CE->getType(), SF);
+                            CE->getType());
     default:
       cerr << "Unhandled ConstantExpr: " << CE << "\n";
       abort();
@@ -177,7 +177,7 @@
    case Type::TY##TyID: Dest.TY##Val = Src1.TY##Val OP Src2.TY##Val; break
 
 static GenericValue executeAddInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(+, UByte);
@@ -198,7 +198,7 @@
 }
 
 static GenericValue executeSubInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(-, UByte);
@@ -219,7 +219,7 @@
 }
 
 static GenericValue executeMulInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(*, UByte);
@@ -240,7 +240,7 @@
 }
 
 static GenericValue executeDivInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(/, UByte);
@@ -261,7 +261,7 @@
 }
 
 static GenericValue executeRemInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(%, UByte);
@@ -286,7 +286,7 @@
 }
 
 static GenericValue executeAndInst(GenericValue Src1, GenericValue Src2, 
-				   const Type *Ty, ExecutionContext &SF) {
+				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(&, UByte);
@@ -306,7 +306,7 @@
 
 
 static GenericValue executeOrInst(GenericValue Src1, GenericValue Src2, 
-                                  const Type *Ty, ExecutionContext &SF) {
+                                  const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(|, UByte);
@@ -326,7 +326,7 @@
 
 
 static GenericValue executeXorInst(GenericValue Src1, GenericValue Src2, 
-                                   const Type *Ty, ExecutionContext &SF) {
+                                   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_BINARY_OPERATOR(^, UByte);
@@ -349,7 +349,7 @@
    case Type::TY##TyID: Dest.BoolVal = Src1.TY##Val OP Src2.TY##Val; break
 
 static GenericValue executeSetEQInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(==, UByte);
@@ -370,7 +370,7 @@
 }
 
 static GenericValue executeSetNEInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(!=, UByte);
@@ -392,7 +392,7 @@
 }
 
 static GenericValue executeSetLEInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(<=, UByte);
@@ -413,7 +413,7 @@
 }
 
 static GenericValue executeSetGEInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(>=, UByte);
@@ -434,7 +434,7 @@
 }
 
 static GenericValue executeSetLTInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(<, UByte);
@@ -455,7 +455,7 @@
 }
 
 static GenericValue executeSetGTInst(GenericValue Src1, GenericValue Src2, 
-				     const Type *Ty, ExecutionContext &SF) {
+				     const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
     IMPLEMENT_SETCC(>, UByte);
@@ -482,20 +482,20 @@
   GenericValue R;   // Result
 
   switch (I.getOpcode()) {
-  case Instruction::Add:   R = executeAddInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::Sub:   R = executeSubInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::Mul:   R = executeMulInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::Div:   R = executeDivInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::Rem:   R = executeRemInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::And:   R = executeAndInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::Or:    R = executeOrInst   (Src1, Src2, Ty, SF); break;
-  case Instruction::Xor:   R = executeXorInst  (Src1, Src2, Ty, SF); break;
-  case Instruction::SetEQ: R = executeSetEQInst(Src1, Src2, Ty, SF); break;
-  case Instruction::SetNE: R = executeSetNEInst(Src1, Src2, Ty, SF); break;
-  case Instruction::SetLE: R = executeSetLEInst(Src1, Src2, Ty, SF); break;
-  case Instruction::SetGE: R = executeSetGEInst(Src1, Src2, Ty, SF); break;
-  case Instruction::SetLT: R = executeSetLTInst(Src1, Src2, Ty, SF); break;
-  case Instruction::SetGT: R = executeSetGTInst(Src1, Src2, Ty, SF); break;
+  case Instruction::Add:   R = executeAddInst  (Src1, Src2, Ty); break;
+  case Instruction::Sub:   R = executeSubInst  (Src1, Src2, Ty); break;
+  case Instruction::Mul:   R = executeMulInst  (Src1, Src2, Ty); break;
+  case Instruction::Div:   R = executeDivInst  (Src1, Src2, Ty); break;
+  case Instruction::Rem:   R = executeRemInst  (Src1, Src2, Ty); break;
+  case Instruction::And:   R = executeAndInst  (Src1, Src2, Ty); break;
+  case Instruction::Or:    R = executeOrInst   (Src1, Src2, Ty); break;
+  case Instruction::Xor:   R = executeXorInst  (Src1, Src2, Ty); break;
+  case Instruction::SetEQ: R = executeSetEQInst(Src1, Src2, Ty); break;
+  case Instruction::SetNE: R = executeSetNEInst(Src1, Src2, Ty); break;
+  case Instruction::SetLE: R = executeSetLEInst(Src1, Src2, Ty); break;
+  case Instruction::SetGE: R = executeSetGEInst(Src1, Src2, Ty); break;
+  case Instruction::SetLT: R = executeSetLTInst(Src1, Src2, Ty); break;
+  case Instruction::SetGT: R = executeSetGTInst(Src1, Src2, Ty); break;
   default:
     cout << "Don't know how to handle this binary operator!\n-->" << I;
     R = Src1;
@@ -644,7 +644,7 @@
   // Check to see if any of the cases match...
   for (unsigned i = 2, e = I.getNumOperands(); i != e; i += 2) {
     if (executeSetEQInst(CondVal,
-                         getOperandValue(I.getOperand(i), SF),ElTy,SF).BoolVal){
+                         getOperandValue(I.getOperand(i), SF), ElTy).BoolVal) {
       Dest = cast<BasicBlock>(I.getOperand(i+1));
       break;
     }




From lattner at cs.uiuc.edu  Tue Apr 22 16:02:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 16:02:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/fldry.c
Message-ID: <200304222101.QAA26026@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource:

fldry.c updated: 1.1 -> 1.2

---
Log message:

Don't divide by zero or print out argv[0]


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/fldry.c
diff -u llvm/test/Programs/SingleSource/fldry.c:1.1 llvm/test/Programs/SingleSource/fldry.c:1.2
--- llvm/test/Programs/SingleSource/fldry.c:1.1	Fri Feb 14 18:30:24 2003
+++ llvm/test/Programs/SingleSource/fldry.c	Tue Apr 22 16:01:47 2003
@@ -173,9 +173,8 @@
 int argc;
 char **argv;
 {
-	printf( "%s: calculate floating dhrystones using doubles size %d\n",
-		argv[0], sizeof(extended)
-	);
+  printf("calculate floating dhrystones using doubles size %d\n",
+         sizeof(extended));
 	//fflush( stdout);
 	Proc0();
 #if 0
@@ -293,6 +292,7 @@
 -- Stop Timer --
 *****************/
 
+#if 0
 #ifdef CLOCK
 	benchtime = ( clock() - starttime - nulltime )/ CLK_TCK;
 	printf("Dhrystone(%s) time for %ld passes = %ld\n",
@@ -307,8 +307,8 @@
 	printf("Dhrystone(%s) time for %ld passes = %ld\n",
 		Version,
 		(long) LOOPS, benchtime);
-	printf("This machine benchmarks at %ld floating dhrystones/second\n",
-		((long) LOOPS) / benchtime);
+        printf("This machine benchmarks at %ld floating dhrystones/second\n",
+               ((long) LOOPS) / benchtime);
 #endif
 #ifdef TIMES
 	times(&tms);
@@ -319,7 +319,7 @@
 	printf("This machine benchmarks at %ld floating dhrystones/second\n",
 		((long) LOOPS) * HZ / benchtime);
 #endif
-
+#endif
 }
 
 Proc1(PtrParIn)




From lattner at cs.uiuc.edu  Tue Apr 22 16:03:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 16:03:00 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/dry.c
Message-ID: <200304222102.QAA26187@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource:

dry.c updated: 1.1 -> 1.2

---
Log message:

Don't print times, which messes up diffs


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/dry.c
diff -u llvm/test/Programs/SingleSource/dry.c:1.1 llvm/test/Programs/SingleSource/dry.c:1.2
--- llvm/test/Programs/SingleSource/dry.c:1.1	Fri Feb 14 18:30:24 2003
+++ llvm/test/Programs/SingleSource/dry.c	Tue Apr 22 16:02:11 2003
@@ -284,6 +284,7 @@
 -- Stop Timer --
 *****************/
 
+#if 0
 #ifdef CLOCK
 	benchtime = ( clock() - starttime - nulltime )/ CLK_TCK;
 	printf("Dhrystone(%s) time for %ld passes = %ld\n",
@@ -310,7 +311,7 @@
 	printf("This machine benchmarks at %ld dhrystones/second\n",
 		((long) LOOPS) * HZ / benchtime);
 #endif
-
+#endif
 }
 
 Proc1(PtrParIn)




From lattner at cs.uiuc.edu  Tue Apr 22 16:17:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 16:17:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304222116.QAA27368@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.79 -> 1.80

---
Log message:

Implement cast to bool


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.79 llvm/tools/lli/Interpreter/Execution.cpp:1.80
--- llvm/tools/lli/Interpreter/Execution.cpp:1.79	Tue Apr 22 15:37:39 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Tue Apr 22 16:15:56 2003
@@ -979,8 +979,10 @@
     IMPLEMENT_CAST_CASE(Pointer, (PointerTy));
     IMPLEMENT_CAST_CASE(Float  , (float));
     IMPLEMENT_CAST_CASE(Double , (double));
+    IMPLEMENT_CAST_CASE(Bool   , (bool));
   default:
     cout << "Unhandled dest type for cast instruction: " << Ty << "\n";
+    abort();
   }
 
   return Dest;




From lattner at cs.uiuc.edu  Tue Apr 22 16:23:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 16:23:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304222122.QAA27563@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.80 -> 1.81

---
Log message:

Kill using declarations
Kill code for invalid operations on pointers


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.80 llvm/tools/lli/Interpreter/Execution.cpp:1.81
--- llvm/tools/lli/Interpreter/Execution.cpp:1.80	Tue Apr 22 16:15:56 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Tue Apr 22 16:22:33 2003
@@ -20,9 +20,6 @@
 #include <math.h>  // For fmod
 #include <signal.h>
 #include <setjmp.h>
-using std::vector;
-using std::cout;
-using std::cerr;
 
 Interpreter *TheEE = 0;
 
@@ -54,7 +51,7 @@
 ProfileStructureFields("profilestructfields", 
                        cl::desc("Profile Structure Field Accesses"));
 #include <map>
-static std::map<const StructType *, vector<unsigned> > FieldAccessCounts;
+static std::map<const StructType *, std::vector<unsigned> > FieldAccessCounts;
 #endif
 
 sigjmp_buf SignalRecoverBuffer;
@@ -109,7 +106,7 @@
                             getOperandValue(CE->getOperand(1), SF),
                             CE->getType());
     default:
-      cerr << "Unhandled ConstantExpr: " << CE << "\n";
+      std::cerr << "Unhandled ConstantExpr: " << CE << "\n";
       abort();
       return GenericValue();
     }
@@ -128,23 +125,23 @@
 
 static void printOperandInfo(Value *V, ExecutionContext &SF) {
   if (isa<Constant>(V)) {
-    cout << "Constant Pool Value\n";
+    std::cout << "Constant Pool Value\n";
   } else if (isa<GlobalValue>(V)) {
-    cout << "Global Value\n";
+    std::cout << "Global Value\n";
   } else {
     unsigned TyP  = V->getType()->getUniqueID();   // TypePlane for value
     unsigned Slot = getOperandSlot(V);
-    cout << "Value=" << (void*)V << " TypeID=" << TyP << " Slot=" << Slot
-         << " Addr=" << &SF.Values[TyP][Slot] << " SF=" << &SF
-         << " Contents=0x";
+    std::cout << "Value=" << (void*)V << " TypeID=" << TyP << " Slot=" << Slot
+              << " Addr=" << &SF.Values[TyP][Slot] << " SF=" << &SF
+              << " Contents=0x";
 
     const unsigned char *Buf = (const unsigned char*)&SF.Values[TyP][Slot];
     for (unsigned i = 0; i < sizeof(GenericValue); ++i) {
       unsigned char Cur = Buf[i];
-      cout << ( Cur     >= 160? char((Cur>>4)+'A'-10) : char((Cur>>4) + '0'))
-           << ((Cur&15) >=  10? char((Cur&15)+'A'-10) : char((Cur&15) + '0'));
+      std::cout << ( Cur     >= 160?char((Cur>>4)+'A'-10):char((Cur>>4) + '0'))
+                << ((Cur&15) >=  10?char((Cur&15)+'A'-10):char((Cur&15) + '0'));
     }
-    cout << "\n";
+    std::cout << "\n";
   }
 }
 
@@ -153,7 +150,7 @@
 static void SetValue(Value *V, GenericValue Val, ExecutionContext &SF) {
   unsigned TyP = V->getType()->getUniqueID();   // TypePlane for value
 
-  //cout << "Setting value: " << &SF.Values[TyP][getOperandSlot(V)] << "\n";
+  //std::cout << "Setting value: " << &SF.Values[TyP][getOperandSlot(V)]<< "\n";
   SF.Values[TyP][getOperandSlot(V)] = Val;
 }
 
@@ -190,9 +187,9 @@
     IMPLEMENT_BINARY_OPERATOR(+, Long);
     IMPLEMENT_BINARY_OPERATOR(+, Float);
     IMPLEMENT_BINARY_OPERATOR(+, Double);
-    IMPLEMENT_BINARY_OPERATOR(+, Pointer);
   default:
-    cout << "Unhandled type for Add instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Add instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -211,9 +208,9 @@
     IMPLEMENT_BINARY_OPERATOR(-, Long);
     IMPLEMENT_BINARY_OPERATOR(-, Float);
     IMPLEMENT_BINARY_OPERATOR(-, Double);
-    IMPLEMENT_BINARY_OPERATOR(-, Pointer);
   default:
-    cout << "Unhandled type for Sub instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Sub instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -232,9 +229,9 @@
     IMPLEMENT_BINARY_OPERATOR(*, Long);
     IMPLEMENT_BINARY_OPERATOR(*, Float);
     IMPLEMENT_BINARY_OPERATOR(*, Double);
-    IMPLEMENT_BINARY_OPERATOR(*, Pointer);
   default:
-    cout << "Unhandled type for Mul instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Mul instruction: " << Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -253,9 +250,9 @@
     IMPLEMENT_BINARY_OPERATOR(/, Long);
     IMPLEMENT_BINARY_OPERATOR(/, Float);
     IMPLEMENT_BINARY_OPERATOR(/, Double);
-    IMPLEMENT_BINARY_OPERATOR(/, Pointer);
   default:
-    cout << "Unhandled type for Div instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Div instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -272,7 +269,6 @@
     IMPLEMENT_BINARY_OPERATOR(%, Int);
     IMPLEMENT_BINARY_OPERATOR(%, ULong);
     IMPLEMENT_BINARY_OPERATOR(%, Long);
-    IMPLEMENT_BINARY_OPERATOR(%, Pointer);
   case Type::FloatTyID:
     Dest.FloatVal = fmod(Src1.FloatVal, Src2.FloatVal);
     break;
@@ -280,7 +276,8 @@
     Dest.DoubleVal = fmod(Src1.DoubleVal, Src2.DoubleVal);
     break;
   default:
-    cout << "Unhandled type for Rem instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Rem instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -297,9 +294,9 @@
     IMPLEMENT_BINARY_OPERATOR(&, Int);
     IMPLEMENT_BINARY_OPERATOR(&, ULong);
     IMPLEMENT_BINARY_OPERATOR(&, Long);
-    IMPLEMENT_BINARY_OPERATOR(&, Pointer);
   default:
-    cout << "Unhandled type for And instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for And instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -317,9 +314,9 @@
     IMPLEMENT_BINARY_OPERATOR(|, Int);
     IMPLEMENT_BINARY_OPERATOR(|, ULong);
     IMPLEMENT_BINARY_OPERATOR(|, Long);
-    IMPLEMENT_BINARY_OPERATOR(|, Pointer);
   default:
-    cout << "Unhandled type for Or instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Or instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -337,9 +334,9 @@
     IMPLEMENT_BINARY_OPERATOR(^, Int);
     IMPLEMENT_BINARY_OPERATOR(^, ULong);
     IMPLEMENT_BINARY_OPERATOR(^, Long);
-    IMPLEMENT_BINARY_OPERATOR(^, Pointer);
   default:
-    cout << "Unhandled type for Xor instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Xor instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -364,7 +361,8 @@
     IMPLEMENT_SETCC(==, Double);
     IMPLEMENT_SETCC(==, Pointer);
   default:
-    cout << "Unhandled type for SetEQ instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetEQ instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -386,7 +384,8 @@
     IMPLEMENT_SETCC(!=, Pointer);
 
   default:
-    cout << "Unhandled type for SetNE instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetNE instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -407,7 +406,8 @@
     IMPLEMENT_SETCC(<=, Double);
     IMPLEMENT_SETCC(<=, Pointer);
   default:
-    cout << "Unhandled type for SetLE instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetLE instruction: " << Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -428,7 +428,8 @@
     IMPLEMENT_SETCC(>=, Double);
     IMPLEMENT_SETCC(>=, Pointer);
   default:
-    cout << "Unhandled type for SetGE instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetGE instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -449,7 +450,8 @@
     IMPLEMENT_SETCC(<, Double);
     IMPLEMENT_SETCC(<, Pointer);
   default:
-    cout << "Unhandled type for SetLT instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetLT instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -470,7 +472,8 @@
     IMPLEMENT_SETCC(>, Double);
     IMPLEMENT_SETCC(>, Pointer);
   default:
-    cout << "Unhandled type for SetGT instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for SetGT instruction: " << *Ty << "\n";
+    abort();
   }
   return Dest;
 }
@@ -497,8 +500,8 @@
   case Instruction::SetLT: R = executeSetLTInst(Src1, Src2, Ty); break;
   case Instruction::SetGT: R = executeSetGTInst(Src1, Src2, Ty); break;
   default:
-    cout << "Don't know how to handle this binary operator!\n-->" << I;
-    R = Src1;
+    std::cout << "Don't know how to handle this binary operator!\n-->" << I;
+    abort();
   }
 
   SetValue(&I, R, SF);
@@ -513,10 +516,10 @@
   // Print out structure field accounting information...
   if (!FieldAccessCounts.empty()) {
     CW << "Profile Field Access Counts:\n";
-    std::map<const StructType *, vector<unsigned> >::iterator 
+    std::map<const StructType *, std::vector<unsigned> >::iterator 
       I = FieldAccessCounts.begin(), E = FieldAccessCounts.end();
     for (; I != E; ++I) {
-      vector<unsigned> &OfC = I->second;
+      std::vector<unsigned> &OfC = I->second;
       CW << "  '" << (Value*)I->first << "'\t- Sum=";
       
       unsigned Sum = 0;
@@ -533,9 +536,9 @@
     CW << "\n";
 
     CW << "Profile Field Access Percentages:\n";
-    cout.precision(3);
+    std::cout.precision(3);
     for (I = FieldAccessCounts.begin(); I != E; ++I) {
-      vector<unsigned> &OfC = I->second;
+      std::vector<unsigned> &OfC = I->second;
       unsigned Sum = 0;
       for (unsigned i = 0; i < OfC.size(); ++i)
         Sum += OfC[i];
@@ -556,9 +559,9 @@
 
 void Interpreter::exitCalled(GenericValue GV) {
   if (!QuietMode) {
-    cout << "Program returned ";
+    std::cout << "Program returned ";
     print(Type::IntTy, GV);
-    cout << " via 'void exit(int)'\n";
+    std::cout << " via 'void exit(int)'\n";
   }
 
   ExitCode = GV.SByteVal;
@@ -588,7 +591,7 @@
         CW << "Function " << M->getType() << " \"" << M->getName()
            << "\" returned ";
         print(RetTy, Result);
-        cout << "\n";
+        std::cout << "\n";
       }
 
       if (RetTy->isIntegral())
@@ -616,7 +619,7 @@
     CW << "Function " << M->getType() << " \"" << M->getName()
        << "\" returned ";
     print(RetTy, Result);
-    cout << "\n";
+    std::cout << "\n";
   }
 }
 
@@ -709,7 +712,7 @@
 #ifdef PROFILE_STRUCTURE_FIELDS
       if (ProfileStructureFields) {
         // Do accounting for this field...
-        vector<unsigned> &OfC = FieldAccessCounts[STy];
+        std::vector<unsigned> &OfC = FieldAccessCounts[STy];
         if (OfC.size() == 0) OfC.resize(STy->getElementTypes().size());
         OfC[Index]++;
       }
@@ -724,9 +727,9 @@
       unsigned Idx = getOperandValue(*I, SF).LongVal;
       if (const ArrayType *AT = dyn_cast<ArrayType>(ST))
         if (Idx >= AT->getNumElements() && ArrayChecksEnabled) {
-          cerr << "Out of range memory access to element #" << Idx
-               << " of a " << AT->getNumElements() << " element array."
-               << " Subscript #" << *I << "\n";
+          std::cerr << "Out of range memory access to element #" << Idx
+                    << " of a " << AT->getNumElements() << " element array."
+                    << " Subscript #" << *I << "\n";
           // Get outta here!!!
           siglongjmp(SignalRecoverBuffer, SIGTRAP);
         }
@@ -781,7 +784,8 @@
                                              ((uint64_t)Ptr->Untyped[7] << 56);
                             break;
     default:
-      cout << "Cannot load value of type " << I.getType() << "!\n";
+      std::cout << "Cannot load value of type " << *I.getType() << "!\n";
+      abort();
     }
   } else {
     switch (I.getType()->getPrimitiveID()) {
@@ -812,7 +816,8 @@
                                              ((uint64_t)Ptr->Untyped[0] << 56);
                             break;
     default:
-      cout << "Cannot load value of type " << I.getType() << "!\n";
+      std::cout << "Cannot load value of type " << *I.getType() << "!\n";
+      abort();
     }
   }
 
@@ -834,7 +839,7 @@
 
 void Interpreter::executeCallInst(CallInst &I, ExecutionContext &SF) {
   ECStack.back().Caller = &I;
-  vector<GenericValue> ArgVals;
+  std::vector<GenericValue> ArgVals;
   ArgVals.reserve(I.getNumOperands()-1);
   for (unsigned i = 1; i < I.getNumOperands(); ++i) {
     ArgVals.push_back(getOperandValue(I.getOperand(i), SF));
@@ -901,9 +906,8 @@
     IMPLEMENT_SHIFT(<<, Int);
     IMPLEMENT_SHIFT(<<, ULong);
     IMPLEMENT_SHIFT(<<, Long);
-    IMPLEMENT_SHIFT(<<, Pointer);
   default:
-    cout << "Unhandled type for Shl instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Shl instruction: " << *Ty << "\n";
   }
   SetValue(&I, Dest, SF);
 }
@@ -923,9 +927,9 @@
     IMPLEMENT_SHIFT(>>, Int);
     IMPLEMENT_SHIFT(>>, ULong);
     IMPLEMENT_SHIFT(>>, Long);
-    IMPLEMENT_SHIFT(>>, Pointer);
   default:
-    cout << "Unhandled type for Shr instruction: " << Ty << "\n";
+    std::cout << "Unhandled type for Shr instruction: " << *Ty << "\n";
+    abort();
   }
   SetValue(&I, Dest, SF);
 }
@@ -952,8 +956,8 @@
       IMPLEMENT_CAST(DESTTY, DESTCTY, Double)
 
 #define IMPLEMENT_CAST_CASE_END()    \
-    default: cout << "Unhandled cast: " << SrcTy << " to " << Ty << "\n";  \
-      break;                                    \
+    default: std::cout << "Unhandled cast: " << SrcTy << " to " << Ty << "\n"; \
+      abort();                                  \
     }                                           \
     break
 
@@ -981,7 +985,7 @@
     IMPLEMENT_CAST_CASE(Double , (double));
     IMPLEMENT_CAST_CASE(Bool   , (bool));
   default:
-    cout << "Unhandled dest type for cast instruction: " << Ty << "\n";
+    std::cout << "Unhandled dest type for cast instruction: " << *Ty << "\n";
     abort();
   }
 
@@ -1022,7 +1026,8 @@
 //===----------------------------------------------------------------------===//
 // callMethod - Execute the specified function...
 //
-void Interpreter::callMethod(Function *M, const vector<GenericValue> &ArgVals) {
+void Interpreter::callMethod(Function *M,
+                             const std::vector<GenericValue> &ArgVals) {
   assert((ECStack.empty() || ECStack.back().Caller == 0 || 
 	  ECStack.back().Caller->getNumOperands()-1 == ArgVals.size()) &&
 	 "Incorrect number of arguments passed into function call!");
@@ -1043,7 +1048,7 @@
         CW << "Function " << M->getType() << " \"" << M->getName()
            << "\" returned ";
         print(RetTy, Result); 
-        cout << "\n";
+        std::cout << "\n";
         
         if (RetTy->isIntegral())
           ExitCode = Result.IntVal;   // Capture the exit code of the program
@@ -1108,14 +1113,14 @@
   if (int SigNo = sigsetjmp(SignalRecoverBuffer, 1)) {
     --SF.CurInst;   // Back up to erroring instruction
     if (SigNo != SIGINT) {
-      cout << "EXCEPTION OCCURRED [" << strsignal(SigNo) << "]:\n";
+      std::cout << "EXCEPTION OCCURRED [" << strsignal(SigNo) << "]:\n";
       printStackTrace();
       // If -abort-on-exception was specified, terminate LLI instead of trying
       // to debug it.
       //
       if (AbortOnExceptions) exit(1);
     } else if (SigNo == SIGINT) {
-      cout << "CTRL-C Detected, execution halted.\n";
+      std::cout << "CTRL-C Detected, execution halted.\n";
     }
     InInstruction = false;
     return true;
@@ -1146,7 +1151,8 @@
     case Instruction::Shr:     executeShrInst  (cast<ShiftInst>(I), SF); break;
     case Instruction::Cast:    executeCastInst (cast<CastInst> (I), SF); break;
     default:
-      cout << "Don't know how to execute this instruction!\n-->" << I;
+      std::cout << "Don't know how to execute this instruction!\n-->" << I;
+      abort();
     }
   }
   InInstruction = false;
@@ -1162,7 +1168,7 @@
 
 void Interpreter::stepInstruction() {  // Do the 'step' command
   if (ECStack.empty()) {
-    cout << "Error: no program running, cannot step!\n";
+    std::cout << "Error: no program running, cannot step!\n";
     return;
   }
 
@@ -1176,7 +1182,7 @@
 // --- UI Stuff...
 void Interpreter::nextInstruction() {  // Do the 'next' command
   if (ECStack.empty()) {
-    cout << "Error: no program running, cannot 'next'!\n";
+    std::cout << "Error: no program running, cannot 'next'!\n";
     return;
   }
 
@@ -1187,7 +1193,7 @@
     // Step into the function...
     if (executeInstruction()) {
       // Hit a breakpoint, print current instruction, then return to user...
-      cout << "Breakpoint hit!\n";
+      std::cout << "Breakpoint hit!\n";
       printCurrentInstruction();
       return;
     }
@@ -1207,7 +1213,7 @@
 
 void Interpreter::run() {
   if (ECStack.empty()) {
-    cout << "Error: no program running, cannot run!\n";
+    std::cout << "Error: no program running, cannot run!\n";
     return;
   }
 
@@ -1217,16 +1223,16 @@
     HitBreakpoint = executeInstruction();
   }
 
-  if (HitBreakpoint) {
-    cout << "Breakpoint hit!\n";
-  }
+  if (HitBreakpoint)
+    std::cout << "Breakpoint hit!\n";
+
   // Print the next instruction to execute...
   printCurrentInstruction();
 }
 
 void Interpreter::finish() {
   if (ECStack.empty()) {
-    cout << "Error: no program running, cannot run!\n";
+    std::cout << "Error: no program running, cannot run!\n";
     return;
   }
 
@@ -1237,9 +1243,8 @@
     HitBreakpoint = executeInstruction();
   }
 
-  if (HitBreakpoint) {
-    cout << "Breakpoint hit!\n";
-  }
+  if (HitBreakpoint)
+    std::cout << "Breakpoint hit!\n";
 
   // Print the next instruction to execute...
   printCurrentInstruction();
@@ -1253,33 +1258,33 @@
 void Interpreter::printCurrentInstruction() {
   if (!ECStack.empty()) {
     if (ECStack.back().CurBB->begin() == ECStack.back().CurInst)  // print label
-      WriteAsOperand(cout, ECStack.back().CurBB) << ":\n";
+      WriteAsOperand(std::cout, ECStack.back().CurBB) << ":\n";
 
     Instruction &I = *ECStack.back().CurInst;
     InstNumber *IN = (InstNumber*)I.getAnnotation(SlotNumberAID);
     assert(IN && "Instruction has no numbering annotation!");
-    cout << "#" << IN->InstNum << I;
+    std::cout << "#" << IN->InstNum << I;
   }
 }
 
 void Interpreter::printValue(const Type *Ty, GenericValue V) {
   switch (Ty->getPrimitiveID()) {
-  case Type::BoolTyID:   cout << (V.BoolVal?"true":"false"); break;
+  case Type::BoolTyID:   std::cout << (V.BoolVal?"true":"false"); break;
   case Type::SByteTyID:
-    cout << (int)V.SByteVal << " '" << V.SByteVal << "'";  break;
+    std::cout << (int)V.SByteVal << " '" << V.SByteVal << "'";  break;
   case Type::UByteTyID:
-    cout << (unsigned)V.UByteVal << " '" << V.UByteVal << "'";  break;
-  case Type::ShortTyID:  cout << V.ShortVal;  break;
-  case Type::UShortTyID: cout << V.UShortVal; break;
-  case Type::IntTyID:    cout << V.IntVal;    break;
-  case Type::UIntTyID:   cout << V.UIntVal;   break;
-  case Type::LongTyID:   cout << (long)V.LongVal;   break;
-  case Type::ULongTyID:  cout << (unsigned long)V.ULongVal;  break;
-  case Type::FloatTyID:  cout << V.FloatVal;  break;
-  case Type::DoubleTyID: cout << V.DoubleVal; break;
-  case Type::PointerTyID:cout << (void*)GVTOP(V); break;
+    std::cout << (unsigned)V.UByteVal << " '" << V.UByteVal << "'";  break;
+  case Type::ShortTyID:  std::cout << V.ShortVal;  break;
+  case Type::UShortTyID: std::cout << V.UShortVal; break;
+  case Type::IntTyID:    std::cout << V.IntVal;    break;
+  case Type::UIntTyID:   std::cout << V.UIntVal;   break;
+  case Type::LongTyID:   std::cout << (long)V.LongVal;   break;
+  case Type::ULongTyID:  std::cout << (unsigned long)V.ULongVal;  break;
+  case Type::FloatTyID:  std::cout << V.FloatVal;  break;
+  case Type::DoubleTyID: std::cout << V.DoubleVal; break;
+  case Type::PointerTyID:std::cout << (void*)GVTOP(V); break;
   default:
-    cout << "- Don't know how to print value of this type!";
+    std::cout << "- Don't know how to print value of this type!";
     break;
   }
 }
@@ -1302,7 +1307,7 @@
   } else {      // Otherwise there should be an annotation for the slot#
     print(PickedVal->getType(), 
           getOperandValue(PickedVal, ECStack[CurFrame]));
-    cout << "\n";
+    std::cout << "\n";
   }
 }
 
@@ -1310,10 +1315,10 @@
   Value *PickedVal = ChooseOneOption(Name, LookupMatchingNames(Name));
   if (!PickedVal) return;
 
-  cout << "Value: ";
+  std::cout << "Value: ";
   print(PickedVal->getType(), 
         getOperandValue(PickedVal, ECStack[CurFrame]));
-  cout << "\n";
+  std::cout << "\n";
   printOperandInfo(PickedVal, ECStack[CurFrame]);
 }
 
@@ -1330,13 +1335,13 @@
   
   unsigned i = 0;
   for (Function::aiterator I = F->abegin(), E = F->aend(); I != E; ++I, ++i) {
-    if (i != 0) cout << ", ";
+    if (i != 0) std::cout << ", ";
     CW << *I << "=";
     
     printValue(I->getType(), getOperandValue(I, ECStack[FrameNo]));
   }
 
-  cout << ")\n";
+  std::cout << ")\n";
 
   if (FrameNo != int(ECStack.size()-1)) {
     BasicBlock::iterator I = ECStack[FrameNo].CurInst;




From lattner at cs.uiuc.edu  Tue Apr 22 17:01:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 17:01:02 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/BasicAA/2003-04-22-GEPProblem.ll
Message-ID: <200304222200.RAA30223@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/BasicAA:

2003-04-22-GEPProblem.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/BasicAA/2003-04-22-GEPProblem.ll
diff -c /dev/null llvm/test/Regression/Transforms/BasicAA/2003-04-22-GEPProblem.ll:1.1
*** /dev/null	Tue Apr 22 17:00:25 2003
--- llvm/test/Regression/Transforms/BasicAA/2003-04-22-GEPProblem.ll	Tue Apr 22 17:00:15 2003
***************
*** 0 ****
--- 1,15 ----
+ ; RUN: as < %s | opt -load-vn -gcse -instcombine | dis | grep sub
+ 
+ ; BasicAA was incorrectly concluding that P1 and P2 didn't conflict!
+ 
+ int %test(int *%Ptr, long %V) {
+ 	%P2 = getelementptr int* %Ptr, long 1
+ 	%P1 = getelementptr int* %Ptr, long %V
+ 	%X = load int* %P1
+ 	store int 5, int* %P2
+ 
+ 	%Y = load int* %P1
+ 
+ 	%Z = sub int %X, %Y
+ 	ret int %Z
+ }




From lattner at cs.uiuc.edu  Tue Apr 22 17:01:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Tue Apr 22 17:01:04 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/BasicAliasAnalysis.cpp
Message-ID: <200304222200.RAA30338@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

BasicAliasAnalysis.cpp updated: 1.6 -> 1.7

---
Log message:

Fix bug: BasicAA/2003-04-22-GEPProblem.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/BasicAliasAnalysis.cpp
diff -u llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.6 llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.7
--- llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.6	Thu Mar  6 10:37:29 2003
+++ llvm/lib/Analysis/BasicAliasAnalysis.cpp	Tue Apr 22 17:00:27 2003
@@ -145,7 +145,6 @@
   if (V1Size != ~0U && V2Size != ~0U)
     if (const GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(V1)) {
       AliasResult R = alias(GEP->getOperand(0), V1Size, V2, V2Size);
-      if (R == NoAlias) return NoAlias;
       if (R == MustAlias) {
         // If there is at least one non-zero constant index, we know they cannot
         // alias.




From brukman at cs.uiuc.edu  Tue Apr 22 18:01:01 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Tue Apr 22 18:01:01 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp ModuloSchedGraph.h
Message-ID: <200304222300.SAA02630@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen/ModuloScheduling:

ModuloSchedGraph.cpp updated: 1.4 -> 1.5
ModuloSchedGraph.h updated: 1.4 -> 1.5

---
Log message:

Just some code beautification changes I had sitting around in my tree.


---
Diffs of the changes:

Index: llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp
diff -u llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp:1.4 llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp:1.5
--- llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp:1.4	Thu Apr 10 14:19:23 2003
+++ llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.cpp	Tue Apr 22 18:00:08 2003
@@ -19,14 +19,12 @@
 #include <algorithm>
 #include <ostream>
 #include <vector>
-
 // FIXME: Should be using #include <cmath>
 #include <math.h>
 //#include <swig.h>
 
 #define UNIDELAY 1
 
-
 //*********************** Internal Data Structures *************************/
 
 // The following two types need to be classes, not typedefs, so we can use
@@ -52,7 +50,6 @@
 
 // class Modulo SchedGraphNode
 
-/*ctor*/
 ModuloSchedGraphNode::ModuloSchedGraphNode(unsigned int _nodeId,
                                            const BasicBlock * _bb,
                                            const Instruction * _inst,
@@ -113,8 +110,8 @@
   for (unsigned int i = 0; i < defVec.size(); i++) {
     for (Value::use_const_iterator I = defVec[i]->getInst()->use_begin();
          I != defVec[i]->getInst()->use_end(); I++) {
-      //for each use of a def, add a flow edge from the def instruction to the ref instruction
-
+      //for each use of a def, add a flow edge from the def instruction to the
+      //ref instruction
 
       const Instruction *value = defVec[i]->getInst();
       Instruction *inst = (Instruction *) (*I);
@@ -969,7 +966,6 @@
 
 int ModuloSchedGraph::computeRecII(const BasicBlock *bb)
 {
-
   int RecII = 0;
 
   //os<<"begining computerRecII()"<<"\n";
@@ -1269,7 +1265,6 @@
   DEBUG(std::cerr << "dumping basic block:");
   DEBUG(std::cerr << (bb->hasName()? bb->getName() : "block")
         << " (" << bb << ")" << "\n");
-
 }
 
 void ModuloSchedGraph::dump(const BasicBlock * bb, std::ostream & os)
@@ -1319,16 +1314,15 @@
   }
 }
 
-void ModuloSchedGraphSet::buildGraphsForMethod(const Function * F,
-                                               const TargetMachine &
-                                               target)
+void ModuloSchedGraphSet::buildGraphsForMethod(const Function *F,
+                                               const TargetMachine &target)
 {
   for (Function::const_iterator BI = F->begin(); BI != F->end(); ++BI)
     addGraph(new ModuloSchedGraph(BI, target));
 }
 
-std::ostream & operator<<(std::ostream & os,
-                          const ModuloSchedGraphNode & node)
+std::ostream& operator<<(std::ostream &os,
+                         const ModuloSchedGraphNode &node)
 {
   os << std::string(8, ' ')
       << "Node " << node.nodeId << " : "
@@ -1347,7 +1341,6 @@
     for (unsigned i = 0, N = node.outEdges.size(); i < N; i++)
       os << std::string(16, ' ') << *node.outEdges[i];
   }
-
 
   return os;
 }


Index: llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.h
diff -u llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.h:1.4 llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.h:1.5
--- llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.h:1.4	Thu Apr 10 14:19:23 2003
+++ llvm/lib/CodeGen/ModuloScheduling/ModuloSchedGraph.h	Tue Apr 22 18:00:08 2003
@@ -235,14 +235,15 @@
   const unsigned int getNumNodes() const {
     return size() + 2;
   }
+
   //return wether the BasicBlock 'bb' contains a loop
-  bool isLoop(const BasicBlock * bb);
+  bool isLoop(const BasicBlock *bb);
 
   //return this basibBlock contains a loop
   bool isLoop();
 
   //return the node for the input instruction
-  ModuloSchedGraphNode *getGraphNodeForInst(const Instruction * inst) const {
+  ModuloSchedGraphNode *getGraphNodeForInst(const Instruction *inst) const {
     const_iterator onePair = this->find(inst);
     return (onePair != this->end()) ? (*onePair).second : NULL;
   }
@@ -252,10 +253,10 @@
   void dump() const;
 
   // dump the basicBlock
-  void dump(const BasicBlock * bb);
+  void dump(const BasicBlock *bb);
 
   //dump the basicBlock into 'os' stream
-  void dump(const BasicBlock * bb, std::ostream & os);
+  void dump(const BasicBlock *bb, std::ostream &os);
 
   //dump the node property
   void dumpNodeProperty() const;
@@ -287,14 +288,13 @@
     (*this)[inst] = node;
   }
 
-  //Graph builder
-
+  // Graph builder
   ModuloSchedGraphNode *getNode(const unsigned nodeId) const;
 
-  //build the graph from the basicBlock
-  void buildGraph(const TargetMachine & target);
+  // Build the graph from the basicBlock
+  void buildGraph(const TargetMachine &target);
 
-  //Build nodes for BasicBlock
+  // Build nodes for BasicBlock
   void buildNodesforBB(const TargetMachine &target,
                        const BasicBlock *bb,
                        NodeVec &memNode,




From brukman at cs.uiuc.edu  Tue Apr 22 18:02:00 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Tue Apr 22 18:02:00 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
Message-ID: <200304222301.SAA02649@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Reoptimizer/BinInterface:

bitmath.h updated: 1.7 -> 1.8

---
Log message:

This include is necessary to compile.


---
Diffs of the changes:

Index: llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h
diff -u llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.7 llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.8
--- llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h:1.7	Thu Apr 10 15:32:49 2003
+++ llvm/include/llvm/Reoptimizer/BinInterface/bitmath.h	Tue Apr 22 18:01:06 2003
@@ -19,6 +19,8 @@
 #ifndef __BITMATH_H__
 #define __BITMATH_H__
 
+#include "Support/DataTypes.h"
+
 //*********************************
 //	Misc functions
 //*********************************




From brukman at cs.uiuc.edu  Tue Apr 22 22:00:01 2003
From: brukman at cs.uiuc.edu (Misha Brukman)
Date: Tue Apr 22 22:00:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Bytecode/Reader/ReadArchive.cpp
Message-ID: <200304230259.VAA04204@neo.cs.uiuc.edu>


Changes in directory llvm/lib/Bytecode/Reader:

ReadArchive.cpp updated: 1.2 -> 1.3

---
Log message:

Fixed 'prevalent'.


---
Diffs of the changes:

Index: llvm/lib/Bytecode/Reader/ReadArchive.cpp
diff -u llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.2 llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.3
--- llvm/lib/Bytecode/Reader/ReadArchive.cpp:1.2	Tue Apr 22 13:02:51 2003
+++ llvm/lib/Bytecode/Reader/ReadArchive.cpp	Tue Apr 22 21:59:05 2003
@@ -3,7 +3,7 @@
 // This file implements the ReadArchiveFile interface, which allows a linker to
 // read all of the LLVM bytecode files contained in a .a file.  This file
 // understands the standard system .a file format.  This can only handle the .a
-// variant prevelant on linux systems so far, but may be extended.  See
+// variant prevalent on Linux systems so far, but may be extended.  See
 // information in this source file for more information:
 //   http://sources.redhat.com/cgi-bin/cvsweb.cgi/src/bfd/archive.c?cvsroot=src
 //




From lattner at cs.uiuc.edu  Wed Apr 23 11:18:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:18:01 2003
Subject: [llvm-commits] CVS: llvm/include/Support/Casting.h
Message-ID: <200304231617.LAA21306@apoc.cs.uiuc.edu>


Changes in directory llvm/include/Support:

Casting.h updated: 1.3 -> 1.4

---
Log message:

Allow dyn_cast to operate on things that aren't OBVIOUSLY a pointer type.
These things can be converted to a pointer, like ilist_iterators


---
Diffs of the changes:

Index: llvm/include/Support/Casting.h
diff -u llvm/include/Support/Casting.h:1.3 llvm/include/Support/Casting.h:1.4
--- llvm/include/Support/Casting.h:1.3	Wed Jul 24 15:22:09 2002
+++ llvm/include/Support/Casting.h	Wed Apr 23 11:17:28 2003
@@ -211,8 +211,8 @@
 //
 
 template <class X, class Y>
-inline typename cast_retty<X, Y*>::ret_type dyn_cast(Y *Val) {
-  return isa<X>(Val) ? cast<X, Y*>(Val) : 0;
+inline typename cast_retty<X, Y>::ret_type dyn_cast(Y Val) {
+  return isa<X>(Val) ? cast<X, Y>(Val) : 0;
 }
 
 // dyn_cast_or_null<X> - Functionally identical to dyn_cast, except that a null




From lattner at cs.uiuc.edu  Wed Apr 23 11:19:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:19:01 2003
Subject: [llvm-commits] CVS: llvm/include/Support/ilist
Message-ID: <200304231618.LAA21317@apoc.cs.uiuc.edu>


Changes in directory llvm/include/Support:

ilist updated: 1.5 -> 1.6

---
Log message:

Allow autoconversion from ilist_iterator<T> to T* in a dyn_cast and friends


---
Diffs of the changes:

Index: llvm/include/Support/ilist
diff -u llvm/include/Support/ilist:1.5 llvm/include/Support/ilist:1.6
--- llvm/include/Support/ilist:1.5	Mon Sep 16 11:46:17 2002
+++ llvm/include/Support/ilist	Wed Apr 23 11:17:53 2003
@@ -151,6 +151,26 @@
   pointer getNodePtrUnchecked() const { return NodePtr; }
 };
 
+// Allow ilist_iterators to convert into pointers to a node automatically when
+// used by the dyn_cast, cast, isa mechanisms...
+
+template<typename From> struct simplify_type;
+
+template<typename NodeTy> struct simplify_type<ilist_iterator<NodeTy> > {
+  typedef NodeTy* SimpleType;
+  
+  static SimpleType getSimplifiedValue(const ilist_iterator<NodeTy> &Node) {
+    return &*Node;
+  }
+};
+template<typename NodeTy> struct simplify_type<const ilist_iterator<NodeTy> > {
+  typedef NodeTy* SimpleType;
+  
+  static SimpleType getSimplifiedValue(const ilist_iterator<NodeTy> &Node) {
+    return &*Node;
+  }
+};
+
 
 //===----------------------------------------------------------------------===//
 //




From lattner at cs.uiuc.edu  Wed Apr 23 11:19:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:19:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/LowerAllocations.cpp
Message-ID: <200304231618.LAA21328@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

LowerAllocations.cpp updated: 1.33 -> 1.34

---
Log message:

Minor cleanups


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/LowerAllocations.cpp
diff -u llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.33 llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.34
--- llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.33	Tue Oct  1 17:38:41 2002
+++ llvm/lib/Transforms/Scalar/LowerAllocations.cpp	Wed Apr 23 11:18:14 2003
@@ -15,8 +15,6 @@
 #include "llvm/Target/TargetData.h"
 #include "Support/Statistic.h"
 
-using std::vector;
-
 namespace {
   Statistic<> NumLowered("lowerallocs", "Number of allocations lowered");
 
@@ -62,10 +60,11 @@
 bool LowerAllocations::doInitialization(Module &M) {
   const FunctionType *MallocType = 
     FunctionType::get(PointerType::get(Type::SByteTy),
-                      vector<const Type*>(1, Type::UIntTy), false);
+                      std::vector<const Type*>(1, Type::UIntTy), false);
   const FunctionType *FreeType = 
     FunctionType::get(Type::VoidTy,
-                      vector<const Type*>(1, PointerType::get(Type::SByteTy)),
+                      std::vector<const Type*>(1,
+                                               PointerType::get(Type::SByteTy)),
                       false);
 
   MallocFunc = M.getOrInsertFunction("malloc", MallocType);
@@ -105,7 +104,7 @@
       
       // Create the call to Malloc...
       CallInst *MCall = new CallInst(MallocFunc,
-                                     vector<Value*>(1, MallocArg), "", I);
+                                     std::vector<Value*>(1, MallocArg), "", I);
       
       // Create a cast instruction to convert to the right type...
       CastInst *MCast = new CastInst(MCall, MI->getType(), "", I);
@@ -121,7 +120,8 @@
                                      PointerType::get(Type::UByteTy), "", I);
       
       // Insert a call to the free function...
-      CallInst *FCall = new CallInst(FreeFunc, vector<Value*>(1, MCast), "", I);
+      CallInst *FCall = new CallInst(FreeFunc, std::vector<Value*>(1, MCast),
+                                     "", I);
       
       // Delete the old free instruction
       I = --BBIL.erase(I);




From lattner at cs.uiuc.edu  Wed Apr 23 11:19:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:19:04 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c
Message-ID: <200304231618.LAA21339@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource/UnitTests:

2003-04-22-Switch.c updated: 1.1 -> 1.2

---
Log message:

Make testcase more interesting by making PHI nodes occur due to the switch


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c
diff -u llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c:1.1 llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c:1.2
--- llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c:1.1	Tue Apr 22 15:25:58 2003
+++ llvm/test/Programs/SingleSource/UnitTests/2003-04-22-Switch.c	Wed Apr 23 11:18:37 2003
@@ -1,17 +1,19 @@
 
 #include <stdio.h>
 
-static void func(unsigned i) {
+static int func(unsigned i) {
+  int X = 4;
   switch (i) {
-  case 8: printf("C\n"); /* fall through. */
+  case 8: printf("C\n"); /* fall through. */  X = 6;
   case 0:
-  case 3:
+  case 3: 
   case 2: printf("A\n"); break;
   case 1:
-  case 7:
+  case 7: X = 7;
   case 9: printf("B\n"); break;
-  default: printf("D\n");
+  default: printf("D\n"); X = 1;
   }
+  return X;
 }
 
 




From lattner at cs.uiuc.edu  Wed Apr 23 11:25:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:25:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/LowerSwitch.cpp
Message-ID: <200304231624.LAA21448@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

LowerSwitch.cpp added (r1.1)

---
Log message:

New pass to lower switch instructions to branch instructions


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/LowerSwitch.cpp
diff -c /dev/null llvm/lib/Transforms/Scalar/LowerSwitch.cpp:1.1
*** /dev/null	Wed Apr 23 11:24:09 2003
--- llvm/lib/Transforms/Scalar/LowerSwitch.cpp	Wed Apr 23 11:23:59 2003
***************
*** 0 ****
--- 1,102 ----
+ //===- LowerSwitch.cpp - Eliminate Switch instructions --------------------===//
+ //
+ // The LowerSwitch transformation rewrites switch statements with a sequence of
+ // branches, which allows targets to get away with not implementing the switch
+ // statement until it is convenient.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #include "llvm/Transforms/Scalar.h"
+ #include "llvm/Function.h"
+ #include "llvm/iTerminators.h"
+ #include "llvm/iOperators.h"
+ #include "llvm/iPHINode.h"
+ #include "llvm/Pass.h"
+ #include "Support/Statistic.h"
+ 
+ namespace {
+   Statistic<> NumLowered("lowerswitch", "Number of SwitchInst's replaced");
+ 
+   /// LowerSwitch Pass - Replace all SwitchInst instructions with chained branch
+   /// instructions.  Note that this cannot be a BasicBlock pass because it
+   /// modifies the CFG!
+   struct LowerSwitch : public FunctionPass {
+     bool runOnFunction(Function &F);
+     void processSwitchInst(SwitchInst *SI);
+   };
+ 
+   RegisterOpt<LowerSwitch>
+   X("lowerswitch", "Lower SwitchInst's to branches");
+ }
+ 
+ // createLowerSwitchPass - Interface to this file...
+ Pass *createLowerSwitchPass() {
+   return new LowerSwitch();
+ }
+ 
+ bool LowerSwitch::runOnFunction(Function &F) {
+   bool Changed = false;
+ 
+   for (Function::iterator I = F.begin(), E = F.end(); I != E; ) {
+     BasicBlock *Cur = I++; // Advance over block so we don't traverse new blocks
+ 
+     if (SwitchInst *SI = dyn_cast<SwitchInst>(Cur->getTerminator())) {
+       Changed = true;
+       processSwitchInst(SI);
+     }
+   }
+ 
+   return Changed;
+ }
+ 
+ // processSwitchInst - Replace the specified switch instruction with a sequence
+ // of chained basic blocks.  Right now we just insert an incredibly stupid
+ // linear sequence of branches.  It would be better to do a balanced binary
+ // search eventually.  FIXME
+ //
+ void LowerSwitch::processSwitchInst(SwitchInst *SI) {
+   BasicBlock *CurBlock = SI->getParent();
+   BasicBlock *OrigBlock = CurBlock;
+   Function *F = CurBlock->getParent();
+   Value *Val = SI->getOperand(0);  // The value we are switching on...
+ 
+   // Unlink the switch instruction from it's block.
+   CurBlock->getInstList().remove(SI);
+ 
+   // Expand comparisons for all of the non-default cases...
+   for (unsigned i = 2, e = SI->getNumOperands(); i != e; i += 2) {
+     // Insert a new basic block after the current one...
+     BasicBlock *NextBlock;
+     if (i != e-2) {
+       NextBlock = new BasicBlock("switchblock");
+       F->getBasicBlockList().insert(CurBlock->getNext(), NextBlock);
+     } else {   // Last case, if it's not the value, go to default block.
+       NextBlock = cast<BasicBlock>(SI->getDefaultDest());
+     }
+ 
+     // Make the seteq instruction...
+     Instruction *Comp = new SetCondInst(Instruction::SetEQ, Val,
+                                         SI->getOperand(i), "switchcase");
+     CurBlock->getInstList().push_back(Comp);
+ 
+     // Make the conditional branch...
+     BasicBlock *Succ = cast<BasicBlock>(SI->getOperand(i+1));
+     Instruction *Br = new BranchInst(Succ, NextBlock, Comp);
+     CurBlock->getInstList().push_back(Br);
+ 
+     // If there were any PHI nodes in this success, rewrite one entry from
+     // OrigBlock to come from CurBlock.
+     for (BasicBlock::iterator I = Succ->begin();
+          PHINode *PN = dyn_cast<PHINode>(I); ++I) {
+       int BlockIdx = PN->getBasicBlockIndex(OrigBlock);
+       assert(BlockIdx != -1 && "Switch didn't go to this successor??");
+       PN->setIncomingBlock((unsigned)BlockIdx, CurBlock);
+     }
+ 
+     CurBlock = NextBlock;  // Move on to the next condition
+   }
+ 
+ 
+   // We are now done with the switch instruction, delete it.
+   delete SI;
+ }




From lattner at cs.uiuc.edu  Wed Apr 23 11:25:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:25:03 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/Scalar.h
Message-ID: <200304231624.LAA21459@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

Scalar.h updated: 1.14 -> 1.15

---
Log message:

Add stub to create lowerSwitches pass


---
Diffs of the changes:

Index: llvm/include/llvm/Transforms/Scalar.h
diff -u llvm/include/llvm/Transforms/Scalar.h:1.14 llvm/include/llvm/Transforms/Scalar.h:1.15
--- llvm/include/llvm/Transforms/Scalar.h:1.14	Tue Oct  8 16:06:24 2002
+++ llvm/include/llvm/Transforms/Scalar.h	Wed Apr 23 11:24:19 2003
@@ -224,6 +224,11 @@
 Pass *createLowerAllocationsPass();
 Pass *createRaiseAllocationsPass();
 
+//===----------------------------------------------------------------------===//
+// This pass converts SwitchInst instructions into a sequence of chained binary
+// branch instructions.
+//
+Pass *createLowerSwitchPass();
 
 //===----------------------------------------------------------------------===//
 //




From lattner at cs.uiuc.edu  Wed Apr 23 11:26:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:26:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/Sparc/Sparc.cpp
Message-ID: <200304231625.LAA21475@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/Sparc:

Sparc.cpp updated: 1.60 -> 1.61

---
Log message:

Add support for the Switch instruction by running the lowerSwitch pass first


---
Diffs of the changes:

Index: llvm/lib/Target/Sparc/Sparc.cpp
diff -u llvm/lib/Target/Sparc/Sparc.cpp:1.60 llvm/lib/Target/Sparc/Sparc.cpp:1.61
--- llvm/lib/Target/Sparc/Sparc.cpp:1.60	Tue Jan 14 15:59:15 2003
+++ llvm/lib/Target/Sparc/Sparc.cpp	Wed Apr 23 11:24:53 2003
@@ -143,6 +143,9 @@
 //
 bool UltraSparc::addPassesToEmitAssembly(PassManager &PM, std::ostream &Out)
 {
+  // FIXME: implement the switch instruction in the instruction selector.
+  PM.add(createLowerSwitchPass());
+
   // Construct and initialize the MachineFunction object for this fn.
   PM.add(createMachineCodeConstructionPass(*this));
 




From lattner at cs.uiuc.edu  Wed Apr 23 11:26:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:26:04 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86TargetMachine.cpp
Message-ID: <200304231625.LAA21482@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86TargetMachine.cpp updated: 1.14 -> 1.15

---
Log message:

Add support for the Switch instruction by running the lowerSwitch pass first



---
Diffs of the changes:

Index: llvm/lib/Target/X86/X86TargetMachine.cpp
diff -u llvm/lib/Target/X86/X86TargetMachine.cpp:1.14 llvm/lib/Target/X86/X86TargetMachine.cpp:1.15
--- llvm/lib/Target/X86/X86TargetMachine.cpp:1.14	Wed Feb 26 13:58:15 2003
+++ llvm/lib/Target/X86/X86TargetMachine.cpp	Wed Apr 23 11:24:55 2003
@@ -6,10 +6,11 @@
 
 #include "X86TargetMachine.h"
 #include "X86.h"
+#include "llvm/PassManager.h"
 #include "llvm/Target/TargetMachineImpls.h"
 #include "llvm/CodeGen/MachineFunction.h"
 #include "llvm/CodeGen/Passes.h"
-#include "llvm/PassManager.h"
+#include "llvm/Transforms/Scalar.h"
 #include "Support/CommandLine.h"
 #include "Support/Statistic.h"
 #include <iostream>
@@ -45,6 +46,9 @@
 /// not supported for this target.
 ///
 bool X86TargetMachine::addPassesToJITCompile(PassManager &PM) {
+  // FIXME: Implement the switch instruction in the instruction selector!
+  PM.add(createLowerSwitchPass());
+
   PM.add(createSimpleX86InstructionSelector(*this));
 
   // TODO: optional optimizations go here




From lattner at cs.uiuc.edu  Wed Apr 23 11:26:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:26:06 2003
Subject: [llvm-commits] CVS: llvm/www/docs/CodingStandards.html
Message-ID: <200304231625.LAA21509@apoc.cs.uiuc.edu>


Changes in directory llvm/www/docs:

CodingStandards.html updated: 1.5 -> 1.6

---
Log message:

Properly convert some &'s to &amp;'s


---
Diffs of the changes:

Index: llvm/www/docs/CodingStandards.html
diff -u llvm/www/docs/CodingStandards.html:1.5 llvm/www/docs/CodingStandards.html:1.6
--- llvm/www/docs/CodingStandards.html:1.5	Sun Jan 20 13:01:26 2002
+++ llvm/www/docs/CodingStandards.html	Wed Apr 23 11:25:38 2003
@@ -378,9 +378,9 @@
         public std::iterator<std::forward_iterator_tag, value_type> {
           friend class container;
           public:
-            const value_type& operator*() const;
+            const value_type&amp; operator*() const;
             const value_type* operator->() const;
-            const_iterator& operator++();
+            const_iterator&amp; operator++();
             const_iterator operator++(int);
             friend bool operator==(const_iterator lhs,
                                    const_iterator rhs);
@@ -408,14 +408,14 @@
 There are normally only three member functions that need nontrivial
 implementations; the rest are just boilerplate.
 
-  const container::value_type&
+  const container::value_type&amp;
     container::const_iterator::operator*() const {
       // find the element and return a reference to it
     }
 
   const container::value_type*
     container::const_iterator::operator->() const {
-      return &**this;
+      return &amp;**this;
     }
 
 If there's an underlying real container, operator*() can just return a
@@ -431,7 +431,7 @@
 The operator->() function is just boilerplate around a call to
 operator*().
 
-  container::const_iterator&
+  container::const_iterator&amp;
     container::const_iterator::operator++() {
       // the incrementing logic goes here
       return *this;
@@ -491,9 +491,9 @@
           friend class container;
           friend class container::const_iterator;
           public:
-            value_type& operator*() const;
+            value_type&amp; operator*() const;
             value_type* operator->() const;
-            iterator& operator++();
+            iterator&amp; operator++();
             iterator operator++(int);
             friend bool operator==(iterator lhs, iterator rhs);
             friend bool operator!=(iterator lhs, iterator rhs);
@@ -505,10 +505,10 @@
           friend class container;
           public:
             const_iterator();
-            const_iterator(const iterator& i);
-            const value_type& operator*() const;
+            const_iterator(const iterator&amp; i);
+            const value_type&amp; operator*() const;
             const value_type* operator->() const;
-            const_iterator& operator++();
+            const_iterator&amp; operator++();
             const_iterator operator++(int);
             friend bool operator==(const_iterator lhs,
                                    const_iterator rhs);
@@ -537,7 +537,7 @@
     public std::iterator<std::bidirectional_iterator_tag, value_type> {
       public:
         //...
-        iterator& operator--();
+        iterator&amp; operator--();
         iterator operator--(int);
         //...
     };
@@ -551,8 +551,8 @@
     public std::iterator<std::random_access_iterator_tag, value_type> {
       public:
         //...
-        iterator& operator+=(difference_type rhs);
-        iterator& operator-=(difference_type rhs);
+        iterator&amp; operator+=(difference_type rhs);
+        iterator&amp; operator-=(difference_type rhs);
         friend iterator operator+(iterator lhs, difference_type rhs);
         friend iterator operator+(difference_type lhs, iterator rhs);
         friend iterator operator-(iterator lhs, difference_type rhs);
@@ -564,13 +564,13 @@
         //...
     };
 
-  container::iterator&
+  container::iterator&amp;
     container::iterator::operator+=(container::difference_type rhs) {
       // add rhs to iterator position
       return *this;
     }
 
-  container::iterator&
+  container::iterator&amp;
     container::iterator::operator-=(container::difference_type rhs) {
       // subtract rhs from iterator position
       return *this;
@@ -660,7 +660,7 @@
 <address><a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Jan 23 15:19:28 CST 2001 -->
 <!-- hhmts start -->
-Last modified: Sun Jan 20 13:01:02 CST 2002
+Last modified: Wed Apr 23 11:20:49 CDT 2003
 <!-- hhmts end -->
 </font>
 </body></html>




From lattner at cs.uiuc.edu  Wed Apr 23 11:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:27:01 2003
Subject: [llvm-commits] CVS: llvm/www/docs/ProgrammersManual.html
Message-ID: <200304231626.LAA06807@tank.cs.uiuc.edu>


Changes in directory llvm/www/docs:

ProgrammersManual.html updated: 1.38 -> 1.39

---
Log message:

Remove documented problem that is now fixed



---
Diffs of the changes:

Index: llvm/www/docs/ProgrammersManual.html
diff -u llvm/www/docs/ProgrammersManual.html:1.38 llvm/www/docs/ProgrammersManual.html:1.39
--- llvm/www/docs/ProgrammersManual.html:1.38	Wed Feb 26 10:38:15 2003
+++ llvm/www/docs/ProgrammersManual.html	Wed Apr 23 11:26:15 2003
@@ -272,7 +272,7 @@
 <pre>
   <i>// Loop over all of the phi nodes in a basic block</i>
   BasicBlock::iterator BBI = BB->begin();
-  for (; <a href="#PhiNode">PHINode</a> *PN = dyn_cast&lt;<a href="#PHINode">PHINode</a>&gt;(&amp;*BBI); ++BBI)
+  for (; <a href="#PhiNode">PHINode</a> *PN = dyn_cast&lt;<a href="#PHINode">PHINode</a>&gt;(BBI); ++BBI)
     cerr &lt;&lt; *PN;
 </pre><p>
 
@@ -617,16 +617,6 @@
 
 <pre>Instruction* pinst = i;</pre>
 
-<b>Caveat emptor</b>: The above syntax works <i>only</i> when you're <i>not</i>
-working with <tt>dyn_cast</tt>.  The template definition of <tt><a
-href="#isa">dyn_cast</a></tt> isn't implemented to handle this yet, so you'll
-still need the following in order for things to work properly:
-
-<pre>
-BasicBlock::iterator bbi = ...;
-<a href="#BranchInst">BranchInst</a>* b = <a href="#isa">dyn_cast</a>&lt;<a href="#BranchInst">BranchInst</a>&gt;(&amp;*bbi);
-</pre>
-
 It's also possible to turn a class pointer into the corresponding
 iterator.  Usually, this conversion is quite inexpensive.  The
 following code snippet illustrates use of the conversion constructors
@@ -1745,6 +1735,6 @@
 <a href="mailto:sabre at nondot.org">Chris Lattner</a></address>
 <!-- Created: Tue Aug  6 15:00:33 CDT 2002 -->
 <!-- hhmts start -->
-Last modified: Mon Feb 24 14:45:19 CST 2003
+Last modified: Wed Apr 23 11:21:57 CDT 2003
 <!-- hhmts end -->
 </font></body></html>




From lattner at cs.uiuc.edu  Wed Apr 23 11:36:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:36:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/InstIterator.h
Message-ID: <200304231635.LAA23867@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

InstIterator.h updated: 1.4 -> 1.5

---
Log message:

Kill unneccesary &*


---
Diffs of the changes:

Index: llvm/include/llvm/Support/InstIterator.h
diff -u llvm/include/llvm/Support/InstIterator.h:1.4 llvm/include/llvm/Support/InstIterator.h:1.5
--- llvm/include/llvm/Support/InstIterator.h:1.4	Thu Jul 25 10:00:45 2002
+++ llvm/include/llvm/Support/InstIterator.h	Wed Apr 23 11:35:30 2003
@@ -59,7 +59,7 @@
   inline BBIty &getBasicBlockIterator()  { return BB; }
   inline BIty  &getInstructionIterator() { return BI; }
   
-  inline IIty operator*()  const { return &*BI; }
+  inline IIty operator*()  const { return BI; }
   inline IIty operator->() const { return operator*(); }
   
   inline bool operator==(const InstIterator &y) const { 




From lattner at cs.uiuc.edu  Wed Apr 23 11:37:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:37:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/InductionVariable.cpp
Message-ID: <200304231636.LAA23904@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

InductionVariable.cpp updated: 1.21 -> 1.22

---
Log message:

Remove unneccesary &*


---
Diffs of the changes:

Index: llvm/lib/Analysis/InductionVariable.cpp
diff -u llvm/lib/Analysis/InductionVariable.cpp:1.21 llvm/lib/Analysis/InductionVariable.cpp:1.22
--- llvm/lib/Analysis/InductionVariable.cpp:1.21	Fri Nov  8 18:48:39 2002
+++ llvm/lib/Analysis/InductionVariable.cpp	Wed Apr 23 11:36:06 2003
@@ -213,7 +213,7 @@
     DEBUG(std::cerr << "no terminator instruction!");
     return NULL; 
   }
-  SetCondInst *SCI = dyn_cast<SetCondInst>(&*B->getCondition());
+  SetCondInst *SCI = dyn_cast<SetCondInst>(B->getCondition());
 
   if (SCI && InductionType == Cannonical) {
     DEBUG(std::cerr << "sci:" << *SCI);




From lattner at cs.uiuc.edu  Wed Apr 23 11:37:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:37:04 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200304231636.LAA23916@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.78 -> 1.79

---
Log message:

Remove unneccesary &*



---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.78 llvm/lib/CWriter/Writer.cpp:1.79
--- llvm/lib/CWriter/Writer.cpp:1.78	Tue Apr 22 15:19:52 2003
+++ llvm/lib/CWriter/Writer.cpp	Wed Apr 23 11:36:08 2003
@@ -896,7 +896,7 @@
 void CWriter::printBranchToBlock(BasicBlock *CurBB, BasicBlock *Succ,
                                            unsigned Indent) {
   for (BasicBlock::iterator I = Succ->begin();
-       PHINode *PN = dyn_cast<PHINode>(&*I); ++I) {
+       PHINode *PN = dyn_cast<PHINode>(I); ++I) {
     //  now we have to do the printing
     Out << string(Indent, ' ');
     outputLValue(PN);




From lattner at cs.uiuc.edu  Wed Apr 23 11:37:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:37:06 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/MachineFunction.cpp
Message-ID: <200304231636.LAA23923@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen:

MachineFunction.cpp updated: 1.38 -> 1.39

---
Log message:

Remove unneccesary &*


---
Diffs of the changes:

Index: llvm/lib/CodeGen/MachineFunction.cpp
diff -u llvm/lib/CodeGen/MachineFunction.cpp:1.38 llvm/lib/CodeGen/MachineFunction.cpp:1.39
--- llvm/lib/CodeGen/MachineFunction.cpp:1.38	Thu Jan 16 12:35:56 2003
+++ llvm/lib/CodeGen/MachineFunction.cpp	Wed Apr 23 11:36:09 2003
@@ -247,7 +247,7 @@
   
   for (Function::const_iterator BB = F->begin(), BBE = F->end(); BB !=BBE; ++BB)
     for (BasicBlock::const_iterator I = BB->begin(), E = BB->end(); I != E; ++I)
-      if (const CallInst *callInst = dyn_cast<CallInst>(&*I))
+      if (const CallInst *callInst = dyn_cast<CallInst>(I))
         {
           unsigned numOperands = callInst->getNumOperands() - 1;
           int numExtra = (int)numOperands-frameInfo.getNumFixedOutgoingArgs();




From lattner at cs.uiuc.edu  Wed Apr 23 11:37:08 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:37:08 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp
Message-ID: <200304231636.LAA23930@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen/InstrSelection:

InstrSelection.cpp updated: 1.56 -> 1.57

---
Log message:

Remove unneccesary &*



---
Diffs of the changes:

Index: llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp
diff -u llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp:1.56 llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp:1.57
--- llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp:1.56	Wed Jan 15 13:47:36 2003
+++ llvm/lib/CodeGen/InstrSelection/InstrSelection.cpp	Wed Apr 23 11:36:11 2003
@@ -182,7 +182,7 @@
   MachineFunction &MF = MachineFunction::get(&F);
   for (MachineFunction::iterator BB = MF.begin(); BB != MF.end(); ++BB) {
     for (BasicBlock::iterator IIt = BB->getBasicBlock()->begin();
-         PHINode *PN = dyn_cast<PHINode>(&*IIt); ++IIt) {
+         PHINode *PN = dyn_cast<PHINode>(IIt); ++IIt) {
       // FIXME: This is probably wrong...
       Value *PhiCpRes = new PHINode(PN->getType(), "PhiCp:");
 




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp
Message-ID: <200304231637.LAA24093@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/LightWtProfiling/Instrument:

InstLoops.cpp updated: 1.2 -> 1.3

---
Log message:

Remove unnecesary &*'s


---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp
diff -u llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp:1.2 llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp:1.3
--- llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp:1.2	Tue Jan 14 16:34:36 2003
+++ llvm/lib/Reoptimizer/LightWtProfiling/Instrument/InstLoops.cpp	Wed Apr 23 11:37:36 2003
@@ -133,7 +133,7 @@
         for(BasicBlock::iterator BB2Inst = BB->begin(), BBend = BB->end(); 
             BB2Inst != BBend; ++BB2Inst){
         
-          if(PHINode *phiInst=dyn_cast<PHINode>(&*BB2Inst)){
+          if (PHINode *phiInst = dyn_cast<PHINode>(BB2Inst)){
             int bbIndex = phiInst->getBasicBlockIndex(u);
             if(bbIndex>=0){
               phiInst->setIncomingBlock(bbIndex, newBB);




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:03 2003
Subject: [llvm-commits] CVS: llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp
Message-ID: <200304231637.LAA24102@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Reoptimizer/Trigger:

TriggerAuxillary.cpp updated: 1.2 -> 1.3

---
Log message:

Remove unnecesary &*'s



---
Diffs of the changes:

Index: llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp
diff -u llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp:1.2 llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp:1.3
--- llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp:1.2	Mon Feb 17 13:45:34 2003
+++ llvm/lib/Reoptimizer/Trigger/TriggerAuxillary.cpp	Wed Apr 23 11:37:37 2003
@@ -52,7 +52,7 @@
       for(BasicBlock::iterator II = BB->begin(), IE = BB->end();
           II != IE; ++II){
         //Instruction *inst = *II;
-        if(CallInst *cInst = dyn_cast<CallInst>(&*II)){
+        if(CallInst *cInst = dyn_cast<CallInst>(II)){
           //          std::cerr<<*cInst<<"\n";
           Function *calledFunction = cInst->getCalledFunction();
           if(calledFunction && calledFunction->getName() == "trigger"){




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:05 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/FunctionInlining.cpp
Message-ID: <200304231637.LAA24116@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

FunctionInlining.cpp updated: 1.41 -> 1.42

---
Log message:

Remove unnecesary &*'s


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/FunctionInlining.cpp
diff -u llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.41 llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.42
--- llvm/lib/Transforms/IPO/FunctionInlining.cpp:1.41	Sun Jan 12 18:27:23 2003
+++ llvm/lib/Transforms/IPO/FunctionInlining.cpp	Wed Apr 23 11:37:38 2003
@@ -169,7 +169,7 @@
 
 static inline bool DoFunctionInlining(BasicBlock *BB) {
   for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) {
-    if (CallInst *CI = dyn_cast<CallInst>(&*I)) {
+    if (CallInst *CI = dyn_cast<CallInst>(I)) {
       // Check to see if we should inline this function
       Function *F = CI->getCalledFunction();
       if (F && ShouldInlineFunction(CI, F)) {




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:07 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:07 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Instrumentation/TraceValues.cpp
Message-ID: <200304231637.LAA24127@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Instrumentation:

TraceValues.cpp updated: 1.55 -> 1.56

---
Log message:

Remove unnecesary &*'s



---
Diffs of the changes:

Index: llvm/lib/Transforms/Instrumentation/TraceValues.cpp
diff -u llvm/lib/Transforms/Instrumentation/TraceValues.cpp:1.55 llvm/lib/Transforms/Instrumentation/TraceValues.cpp:1.56
--- llvm/lib/Transforms/Instrumentation/TraceValues.cpp:1.55	Wed Apr 16 15:28:40 2003
+++ llvm/lib/Transforms/Instrumentation/TraceValues.cpp	Wed Apr 23 11:37:40 2003
@@ -298,9 +298,9 @@
                      ExternalFuncs& externalFuncs) {
   
   for (BasicBlock::iterator II=BB->begin(), IE = BB->end(); II != IE; ++II)
-    if (FreeInst *FI = dyn_cast<FreeInst>(&*II))
+    if (FreeInst *FI = dyn_cast<FreeInst>(II))
       InsertReleaseInst(FI->getOperand(0), BB, FI,externalFuncs.ReleasePtrFunc);
-    else if (AllocaInst *AI = dyn_cast<AllocaInst>(&*II))
+    else if (AllocaInst *AI = dyn_cast<AllocaInst>(II))
       InsertRecordInst(AI, BB, AI->getNext(), externalFuncs.RecordPtrFunc);
 }  
 
@@ -332,7 +332,7 @@
   // instruction *preceding* InsertPos to check when to terminate the loop.
   // 
   for (BasicBlock::iterator II = BB->begin(); &*II != InsertPos; ++II) {
-    if (StoreInst *SI = dyn_cast<StoreInst>(&*II)) {
+    if (StoreInst *SI = dyn_cast<StoreInst>(II)) {
       assert(valuesStoredInFunction &&
              "Should not be printing a store instruction at function exit");
       LoadInst *LI = new LoadInst(SI->getPointerOperand(), "reload." +




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:09 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:09 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp ProfilePaths.cpp RetracePath.cpp
Message-ID: <200304231637.LAA24138@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Instrumentation/ProfilePaths:

EdgeCode.cpp updated: 1.19 -> 1.20
ProfilePaths.cpp updated: 1.26 -> 1.27
RetracePath.cpp updated: 1.3 -> 1.4

---
Log message:

Remove unnecesary &*'s


---
Diffs of the changes:

Index: llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp
diff -u llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.19 llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.20
--- llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp:1.19	Fri Feb 14 14:38:10 2003
+++ llvm/lib/Transforms/Instrumentation/ProfilePaths/EdgeCode.cpp	Wed Apr 23 11:37:41 2003
@@ -386,7 +386,7 @@
   for(BasicBlock::iterator BB2Inst = BB2->begin(), BBend = BB2->end(); 
       BB2Inst != BBend; ++BB2Inst){
    
-    if(PHINode *phiInst=dyn_cast<PHINode>(&*BB2Inst)){
+    if(PHINode *phiInst=dyn_cast<PHINode>(BB2Inst)){
       int bbIndex=phiInst->getBasicBlockIndex(BB1);
       assert(bbIndex>=0);
       phiInst->setIncomingBlock(bbIndex, newBB);


Index: llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp
diff -u llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.26 llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.27
--- llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp:1.26	Wed Apr 16 15:28:42 2003
+++ llvm/lib/Transforms/Instrumentation/ProfilePaths/ProfilePaths.cpp	Wed Apr 23 11:37:41 2003
@@ -87,7 +87,7 @@
     nodes.push_back(nd); 
     if(&*BB == ExitNode)
       exitNode=nd;
-    if(&*BB==F.begin())
+    if(BB==F.begin())
       startNode=nd;
   }
 


Index: llvm/lib/Transforms/Instrumentation/ProfilePaths/RetracePath.cpp
diff -u llvm/lib/Transforms/Instrumentation/ProfilePaths/RetracePath.cpp:1.3 llvm/lib/Transforms/Instrumentation/ProfilePaths/RetracePath.cpp:1.4
--- llvm/lib/Transforms/Instrumentation/ProfilePaths/RetracePath.cpp:1.3	Fri Feb 14 14:41:53 2003
+++ llvm/lib/Transforms/Instrumentation/ProfilePaths/RetracePath.cpp	Wed Apr 23 11:37:41 2003
@@ -140,7 +140,7 @@
     BasicBlock *ExitNode = 0;
     for (Function::iterator I = M->begin(), E = M->end(); I != E; ++I){
       if (isa<ReturnInst>(I->getTerminator())) {
-        ExitNode = &*I;
+        ExitNode = I;
         break;
       }
     }
@@ -160,7 +160,7 @@
       if(BB->size()==3 || BB->size() ==2){
         for(BasicBlock::iterator II = BB->begin(), IE = BB->end();
             II != IE; ++II){
-          if(CallInst *callInst = dyn_cast<CallInst>(&*II)){
+          if(CallInst *callInst = dyn_cast<CallInst>(II)){
             //std::cerr<<*callInst;
             Function *calledFunction = callInst->getCalledFunction();
             if(calledFunction && calledFunction->getName() == "trigger"){
@@ -199,7 +199,7 @@
         continue;
       
       //if(BB->size()==3)
-      //if(CallInst *callInst = dyn_cast<CallInst>(&*BB->getInstList().begin()))
+      //if(CallInst *callInst = dyn_cast<CallInst>(BB->getInstList().begin()))
       //if(callInst->getCalledFunction()->getName() == "trigger")
       //continue;
       
@@ -216,20 +216,20 @@
       Node *nd=findBB(nodes, BB);
       assert(nd && "No node for this edge!");
       
-      for(BasicBlock::succ_iterator s=succ_begin(&*BB), se=succ_end(&*BB); 
+      for(BasicBlock::succ_iterator s=succ_begin(BB), se=succ_end(BB); 
           s!=se; ++s){
         
         if(triggerBBs[*s] == 9){
           //if(!pathReg[M]){ //Get the path register for this!
           //if(BB->size()>8)
-          //  if(LoadInst *ldInst = dyn_cast<LoadInst>(&*BB->getInstList().begin()))
+          //  if(LoadInst *ldInst = dyn_cast<LoadInst>(BB->getInstList().begin()))
           //    pathReg[M] = ldInst->getPointerOperand();
           //}
           continue;
         }
         //if((*s)->size()==3)
         //if(CallInst *callInst = 
-        //   dyn_cast<CallInst>(&*(*s)->getInstList().begin()))
+        //   dyn_cast<CallInst>((*s)->getInstList().begin()))
         //  if(callInst->getCalledFunction()->getName() == "trigger")
         //    continue;
         
@@ -284,11 +284,11 @@
         VBI != VBE; ++VBI){
       for(BasicBlock::iterator BBI = (*VBI)->begin(), BBE = (*VBI)->end();
           BBI != BBE; ++BBI){
-        if(LoadInst *ldInst = dyn_cast<LoadInst>(&*BBI)){
+        if(LoadInst *ldInst = dyn_cast<LoadInst>(BBI)){
           if(pathReg[M] == ldInst->getPointerOperand())
             instToErase.push_back(ldInst);
         }
-        else if(StoreInst *stInst = dyn_cast<StoreInst>(&*BBI)){
+        else if(StoreInst *stInst = dyn_cast<StoreInst>(BBI)){
           if(pathReg[M] == stInst->getPointerOperand())
             instToErase.push_back(stInst);
         }




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:12 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:12 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/ADCE.cpp CorrelatedExprs.cpp DecomposeMultiDimRefs.cpp IndVarSimplify.cpp LICM.cpp LoopPreheaders.cpp LowerAllocations.cpp Mem2Reg.cpp Reassociate.cpp SCCP.cpp
Message-ID: <200304231637.LAA24169@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

ADCE.cpp updated: 1.50 -> 1.51
CorrelatedExprs.cpp updated: 1.8 -> 1.9
DecomposeMultiDimRefs.cpp updated: 1.26 -> 1.27
IndVarSimplify.cpp updated: 1.32 -> 1.33
LICM.cpp updated: 1.26 -> 1.27
LoopPreheaders.cpp updated: 1.10 -> 1.11
LowerAllocations.cpp updated: 1.34 -> 1.35
Mem2Reg.cpp updated: 1.2 -> 1.3
Reassociate.cpp updated: 1.16 -> 1.17
SCCP.cpp updated: 1.64 -> 1.65

---
Log message:

Remove unnecesary &*'s



---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/ADCE.cpp
diff -u llvm/lib/Transforms/Scalar/ADCE.cpp:1.50 llvm/lib/Transforms/Scalar/ADCE.cpp:1.51
--- llvm/lib/Transforms/Scalar/ADCE.cpp:1.50	Mon Feb 24 14:48:30 2003
+++ llvm/lib/Transforms/Scalar/ADCE.cpp	Wed Apr 23 11:37:42 2003
@@ -119,7 +119,7 @@
   for (BasicBlock::iterator I = BB->begin(), E = --BB->end(); I != E; )
     if (!LiveSet.count(I)) {              // Is this instruction alive?
       I->dropAllReferences();             // Nope, drop references... 
-      if (PHINode *PN = dyn_cast<PHINode>(&*I)) {
+      if (PHINode *PN = dyn_cast<PHINode>(I)) {
         // We don't want to leave PHI nodes in the program that have
         // #arguments != #predecessors, so we remove them now.
         //
@@ -310,7 +310,7 @@
               // should be identical to the incoming values for LastDead.
               //
               for (BasicBlock::iterator II = NextAlive->begin();
-                   PHINode *PN = dyn_cast<PHINode>(&*II); ++II) {
+                   PHINode *PN = dyn_cast<PHINode>(II); ++II) {
                 // Get the incoming value for LastDead...
                 int OldIdx = PN->getBasicBlockIndex(LastDead);
                 assert(OldIdx != -1 && "LastDead is not a pred of NextAlive!");


Index: llvm/lib/Transforms/Scalar/CorrelatedExprs.cpp
diff -u llvm/lib/Transforms/Scalar/CorrelatedExprs.cpp:1.8 llvm/lib/Transforms/Scalar/CorrelatedExprs.cpp:1.9
--- llvm/lib/Transforms/Scalar/CorrelatedExprs.cpp:1.8	Fri Nov  8 17:18:37 2002
+++ llvm/lib/Transforms/Scalar/CorrelatedExprs.cpp	Wed Apr 23 11:37:42 2003
@@ -450,11 +450,11 @@
     
   // Put the newly discovered information into the RegionInfo...
   for (BasicBlock::iterator I = OldSucc->begin(), E = OldSucc->end(); I!=E; ++I)
-    if (PHINode *PN = dyn_cast<PHINode>(&*I)) {
+    if (PHINode *PN = dyn_cast<PHINode>(I)) {
       int OpNum = PN->getBasicBlockIndex(BB);
       assert(OpNum != -1 && "PHI doesn't have incoming edge for predecessor!?");
       PropagateEquality(PN, PN->getIncomingValue(OpNum), NewRI);      
-    } else if (SetCondInst *SCI = dyn_cast<SetCondInst>(&*I)) {
+    } else if (SetCondInst *SCI = dyn_cast<SetCondInst>(I)) {
       Relation::KnownResult Res = getSetCCResult(SCI, NewRI);
       if (Res == Relation::Unknown) return false;
       PropagateEquality(SCI, ConstantBool::get(Res), NewRI);
@@ -563,7 +563,7 @@
   // node with a new value.
   //
   for (BasicBlock::iterator I = OldSucc->begin();
-       PHINode *PN = dyn_cast<PHINode>(&*I); ) {
+       PHINode *PN = dyn_cast<PHINode>(I); ) {
 
     // Get the value flowing across the old edge and remove the PHI node entry
     // for this edge: we are about to remove the edge!  Don't remove the PHI
@@ -993,7 +993,7 @@
 bool CEE::SimplifyBasicBlock(BasicBlock &BB, const RegionInfo &RI) {
   bool Changed = false;
   for (BasicBlock::iterator I = BB.begin(), E = BB.end(); I != E; ) {
-    Instruction *Inst = &*I++;
+    Instruction *Inst = I++;
 
     // Convert instruction arguments to canonical forms...
     Changed |= SimplifyInstruction(Inst, RI);


Index: llvm/lib/Transforms/Scalar/DecomposeMultiDimRefs.cpp
diff -u llvm/lib/Transforms/Scalar/DecomposeMultiDimRefs.cpp:1.26 llvm/lib/Transforms/Scalar/DecomposeMultiDimRefs.cpp:1.27
--- llvm/lib/Transforms/Scalar/DecomposeMultiDimRefs.cpp:1.26	Tue Oct  1 17:38:41 2002
+++ llvm/lib/Transforms/Scalar/DecomposeMultiDimRefs.cpp	Wed Apr 23 11:37:42 2003
@@ -44,7 +44,7 @@
 {
   bool changed = false;
   for (BasicBlock::iterator II = BB.begin(); II != BB.end(); )
-    if (GetElementPtrInst *gep = dyn_cast<GetElementPtrInst>(&*II++)) // pre-inc
+    if (GetElementPtrInst *gep = dyn_cast<GetElementPtrInst>(II++)) // pre-inc
       if (gep->getNumIndices() >= 2)
         changed |= DecomposeArrayRef(gep); // always modifies II
   return changed;


Index: llvm/lib/Transforms/Scalar/IndVarSimplify.cpp
diff -u llvm/lib/Transforms/Scalar/IndVarSimplify.cpp:1.32 llvm/lib/Transforms/Scalar/IndVarSimplify.cpp:1.33
--- llvm/lib/Transforms/Scalar/IndVarSimplify.cpp:1.32	Mon Oct 21 15:00:26 2002
+++ llvm/lib/Transforms/Scalar/IndVarSimplify.cpp	Wed Apr 23 11:37:42 2003
@@ -45,7 +45,7 @@
   //
   std::vector<InductionVariable> IndVars;    // Induction variables for block
   BasicBlock::iterator AfterPHIIt = Header->begin();
-  for (; PHINode *PN = dyn_cast<PHINode>(&*AfterPHIIt); ++AfterPHIIt)
+  for (; PHINode *PN = dyn_cast<PHINode>(AfterPHIIt); ++AfterPHIIt)
     IndVars.push_back(InductionVariable(PN, Loops));
   // AfterPHIIt now points to first nonphi instruction...
 


Index: llvm/lib/Transforms/Scalar/LICM.cpp
diff -u llvm/lib/Transforms/Scalar/LICM.cpp:1.26 llvm/lib/Transforms/Scalar/LICM.cpp:1.27
--- llvm/lib/Transforms/Scalar/LICM.cpp:1.26	Mon Mar  3 17:32:45 2003
+++ llvm/lib/Transforms/Scalar/LICM.cpp	Wed Apr 23 11:37:42 2003
@@ -319,12 +319,12 @@
     // Rewrite all loads and stores in the block of the pointer...
     for (BasicBlock::iterator II = (*I)->begin(), E = (*I)->end();
          II != E; ++II) {
-      if (LoadInst *L = dyn_cast<LoadInst>(&*II)) {
+      if (LoadInst *L = dyn_cast<LoadInst>(II)) {
         std::map<Value*, AllocaInst*>::iterator
           I = ValueToAllocaMap.find(L->getOperand(0));
         if (I != ValueToAllocaMap.end())
           L->setOperand(0, I->second);    // Rewrite load instruction...
-      } else if (StoreInst *S = dyn_cast<StoreInst>(&*II)) {
+      } else if (StoreInst *S = dyn_cast<StoreInst>(II)) {
         std::map<Value*, AllocaInst*>::iterator
           I = ValueToAllocaMap.find(S->getOperand(1));
         if (I != ValueToAllocaMap.end())


Index: llvm/lib/Transforms/Scalar/LoopPreheaders.cpp
diff -u llvm/lib/Transforms/Scalar/LoopPreheaders.cpp:1.10 llvm/lib/Transforms/Scalar/LoopPreheaders.cpp:1.11
--- llvm/lib/Transforms/Scalar/LoopPreheaders.cpp:1.10	Fri Feb 28 10:54:17 2003
+++ llvm/lib/Transforms/Scalar/LoopPreheaders.cpp	Wed Apr 23 11:37:42 2003
@@ -132,7 +132,7 @@
   //
   if (!Preds.empty()) {  // Is the loop not obviously dead?
     for (BasicBlock::iterator I = BB->begin();
-         PHINode *PN = dyn_cast<PHINode>(&*I); ++I) {
+         PHINode *PN = dyn_cast<PHINode>(I); ++I) {
       
       // Create the new PHI node, insert it into NewBB at the end of the block
       PHINode *NewPHI = new PHINode(PN->getType(), PN->getName()+".ph", BI);
@@ -160,7 +160,7 @@
     
   } else {                       // Otherwise the loop is dead...
     for (BasicBlock::iterator I = BB->begin();
-         PHINode *PN = dyn_cast<PHINode>(&*I); ++I)
+         PHINode *PN = dyn_cast<PHINode>(I); ++I)
       // Insert dummy values as the incoming value...
       PN->addIncoming(Constant::getNullValue(PN->getType()), NewBB);
   }  


Index: llvm/lib/Transforms/Scalar/LowerAllocations.cpp
diff -u llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.34 llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.35
--- llvm/lib/Transforms/Scalar/LowerAllocations.cpp:1.34	Wed Apr 23 11:18:14 2003
+++ llvm/lib/Transforms/Scalar/LowerAllocations.cpp	Wed Apr 23 11:37:42 2003
@@ -85,7 +85,7 @@
 
   // Loop over all of the instructions, looking for malloc or free instructions
   for (BasicBlock::iterator I = BB.begin(), E = BB.end(); I != E; ++I) {
-    if (MallocInst *MI = dyn_cast<MallocInst>(&*I)) {
+    if (MallocInst *MI = dyn_cast<MallocInst>(I)) {
       const Type *AllocTy = MI->getType()->getElementType();
       
       // Get the number of bytes to be allocated for one element of the
@@ -114,7 +114,7 @@
       I = --BBIL.erase(I);         // remove and delete the malloc instr...
       Changed = true;
       ++NumLowered;
-    } else if (FreeInst *FI = dyn_cast<FreeInst>(&*I)) {
+    } else if (FreeInst *FI = dyn_cast<FreeInst>(I)) {
       // Cast the argument to free into a ubyte*...
       CastInst *MCast = new CastInst(FI->getOperand(0), 
                                      PointerType::get(Type::UByteTy), "", I);


Index: llvm/lib/Transforms/Scalar/Mem2Reg.cpp
diff -u llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.2 llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.3
--- llvm/lib/Transforms/Scalar/Mem2Reg.cpp:1.2	Mon Mar  3 11:25:16 2003
+++ llvm/lib/Transforms/Scalar/Mem2Reg.cpp	Wed Apr 23 11:37:42 2003
@@ -43,7 +43,7 @@
   // Find allocas that are safe to promote, by looking at all instructions in
   // the entry node
   for (BasicBlock::iterator I = BB.begin(), E = --BB.end(); I != E; ++I)
-    if (AllocaInst *AI = dyn_cast<AllocaInst>(&*I))       // Is it an alloca?
+    if (AllocaInst *AI = dyn_cast<AllocaInst>(I))       // Is it an alloca?
       if (isAllocaPromotable(AI, TD))
         Allocas.push_back(AI);
 


Index: llvm/lib/Transforms/Scalar/Reassociate.cpp
diff -u llvm/lib/Transforms/Scalar/Reassociate.cpp:1.16 llvm/lib/Transforms/Scalar/Reassociate.cpp:1.17
--- llvm/lib/Transforms/Scalar/Reassociate.cpp:1.16	Mon Feb 24 14:48:30 2003
+++ llvm/lib/Transforms/Scalar/Reassociate.cpp	Wed Apr 23 11:37:42 2003
@@ -224,7 +224,7 @@
     // the two operands are sorted incorrectly, fix it now.
     //
     if (BI->isAssociative()) {
-      BinaryOperator *I = cast<BinaryOperator>(&*BI);
+      BinaryOperator *I = cast<BinaryOperator>(BI);
       if (!I->use_empty()) {
         // Make sure that we don't have a tree-shaped computation.  If we do,
         // linearize it.  Convert (A+B)+(C+D) into ((A+B)+C)+D


Index: llvm/lib/Transforms/Scalar/SCCP.cpp
diff -u llvm/lib/Transforms/Scalar/SCCP.cpp:1.64 llvm/lib/Transforms/Scalar/SCCP.cpp:1.65
--- llvm/lib/Transforms/Scalar/SCCP.cpp:1.64	Sun Nov  3 20:54:22 2002
+++ llvm/lib/Transforms/Scalar/SCCP.cpp	Wed Apr 23 11:37:42 2003
@@ -474,7 +474,7 @@
       // constant now may not be.
       //
       for (BasicBlock::iterator I = Succ->begin();
-           PHINode *PN = dyn_cast<PHINode>(&*I); ++I)
+           PHINode *PN = dyn_cast<PHINode>(I); ++I)
         visitPHINode(*PN);
     }
 }




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:14 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp SimplifyCFG.cpp
Message-ID: <200304231637.LAA24180@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

BreakCriticalEdges.cpp updated: 1.8 -> 1.9
SimplifyCFG.cpp updated: 1.10 -> 1.11

---
Log message:

Remove unnecesary &*'s


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp
diff -u llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp:1.8 llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp:1.9
--- llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp:1.8	Wed Oct 30 20:44:36 2002
+++ llvm/lib/Transforms/Utils/BreakCriticalEdges.cpp	Wed Apr 23 11:37:43 2003
@@ -85,7 +85,7 @@
   // merge incoming values from NewBB instead of from TIBB.
   //
   for (BasicBlock::iterator I = DestBB->begin();
-       PHINode *PN = dyn_cast<PHINode>(&*I); ++I) {
+       PHINode *PN = dyn_cast<PHINode>(I); ++I) {
     // We no longer enter through TIBB, now we come in through NewBB.
     PN->replaceUsesOfWith(TIBB, NewBB);
   }


Index: llvm/lib/Transforms/Utils/SimplifyCFG.cpp
diff -u llvm/lib/Transforms/Utils/SimplifyCFG.cpp:1.10 llvm/lib/Transforms/Utils/SimplifyCFG.cpp:1.11
--- llvm/lib/Transforms/Utils/SimplifyCFG.cpp:1.10	Fri Mar  7 12:13:41 2003
+++ llvm/lib/Transforms/Utils/SimplifyCFG.cpp	Wed Apr 23 11:37:44 2003
@@ -40,7 +40,7 @@
       // Loop over all of the PHI nodes checking to see if there are
       // incompatible values coming in.
       for (BasicBlock::iterator I = Succ->begin();
-           PHINode *PN = dyn_cast<PHINode>(&*I); ++I) {
+           PHINode *PN = dyn_cast<PHINode>(I); ++I) {
         // Loop up the entries in the PHI node for BB and for *PI if the values
         // coming in are non-equal, we cannot merge these two blocks (instead we
         // should insert a conditional move or something, then merge the
@@ -56,7 +56,7 @@
 
   // Loop over all of the PHI nodes in the successor BB
   for (BasicBlock::iterator I = Succ->begin();
-       PHINode *PN = dyn_cast<PHINode>(&*I); ++I) {
+       PHINode *PN = dyn_cast<PHINode>(I); ++I) {
     Value *OldVal = PN->removeIncomingValue(BB, false);
     assert(OldVal && "No entry in PHI for Pred BB!");
 




From lattner at cs.uiuc.edu  Wed Apr 23 11:38:16 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:38:16 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/BasicBlock.cpp
Message-ID: <200304231637.LAA24187@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

BasicBlock.cpp updated: 1.29 -> 1.30

---
Log message:

Remove unnecesary &*'s



---
Diffs of the changes:

Index: llvm/lib/VMCore/BasicBlock.cpp
diff -u llvm/lib/VMCore/BasicBlock.cpp:1.29 llvm/lib/VMCore/BasicBlock.cpp:1.30
--- llvm/lib/VMCore/BasicBlock.cpp:1.29	Wed Nov 20 12:33:41 2002
+++ llvm/lib/VMCore/BasicBlock.cpp	Wed Apr 23 11:37:45 2003
@@ -196,7 +196,7 @@
   } else {
     // Okay, now we know that we need to remove predecessor #pred_idx from all
     // PHI nodes.  Iterate over each PHI node fixing them up
-    for (iterator II = begin(); PHINode *PN = dyn_cast<PHINode>(&*II); ++II)
+    for (iterator II = begin(); PHINode *PN = dyn_cast<PHINode>(II); ++II)
       PN->removeIncomingValue(Pred);
   }
 }
@@ -243,7 +243,7 @@
     // incoming values...
     BasicBlock *Successor = *I;
     for (BasicBlock::iterator II = Successor->begin();
-         PHINode *PN = dyn_cast<PHINode>(&*II); ++II) {
+         PHINode *PN = dyn_cast<PHINode>(II); ++II) {
       int IDX = PN->getBasicBlockIndex(this);
       while (IDX != -1) {
         PN->setIncomingBlock((unsigned)IDX, New);




From lattner at cs.uiuc.edu  Wed Apr 23 11:39:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:39:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/TestPasses.cpp
Message-ID: <200304231638.LAA24245@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

TestPasses.cpp updated: 1.1 -> 1.2

---
Log message:

Remove unnecessary &*


---
Diffs of the changes:

Index: llvm/tools/bugpoint/TestPasses.cpp
diff -u llvm/tools/bugpoint/TestPasses.cpp:1.1 llvm/tools/bugpoint/TestPasses.cpp:1.2
--- llvm/tools/bugpoint/TestPasses.cpp:1.1	Wed Nov 20 16:28:10 2002
+++ llvm/tools/bugpoint/TestPasses.cpp	Wed Apr 23 11:38:00 2003
@@ -39,7 +39,7 @@
   class DeleteCalls : public BasicBlockPass {
     bool runOnBasicBlock(BasicBlock &BB) {
       for (BasicBlock::iterator I = BB.begin(), E = BB.end(); I != E; ++I)
-        if (CallInst *CI = dyn_cast<CallInst>(&*I)) {
+        if (CallInst *CI = dyn_cast<CallInst>(I)) {
           if (!CI->use_empty())
             CI->replaceAllUsesWith(Constant::getNullValue(CI->getType()));
           CI->getParent()->getInstList().erase(CI);




From lattner at cs.uiuc.edu  Wed Apr 23 11:44:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 11:44:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Makefile
Message-ID: <200304231643.LAA25108@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

Makefile updated: 1.22 -> 1.23

---
Log message:

We now need to link libscalar to get the switch lowering pass


---
Diffs of the changes:

Index: llvm/tools/lli/Makefile
diff -u llvm/tools/lli/Makefile:1.22 llvm/tools/lli/Makefile:1.23
--- llvm/tools/lli/Makefile:1.22	Wed Jan 22 09:41:10 2003
+++ llvm/tools/lli/Makefile	Wed Apr 23 11:43:02 2003
@@ -3,8 +3,8 @@
 PARALLEL_DIRS = Interpreter JIT
 
 JITLIBS  = lli-jit codegen x86
-USEDLIBS = lli-interpreter $(JITLIBS) bcreader vmcore analysis.a support.a target.a 
-#transforms.a
+USEDLIBS = lli-interpreter $(JITLIBS) bcreader vmcore scalaropts.a \
+           analysis.a support.a target.a 
 
 # Have gcc tell the linker to export symbols from the program so that
 # dynamically loaded modules can be linked against them.




From lattner at cs.uiuc.edu  Wed Apr 23 12:23:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 12:23:02 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200304231722.MAA26861@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.87 -> 1.88

---
Log message:

Implement the constantexpr CAST instruction correctly


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.87 llvm/lib/Target/X86/InstSelectSimple.cpp:1.88
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.87	Tue Apr 22 12:54:23 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Wed Apr 23 12:22:12 2003
@@ -192,6 +192,11 @@
                           Value *Src, User::op_iterator IdxBegin,
                           User::op_iterator IdxEnd, unsigned TargetReg);
 
+    /// emitCastOperation - Common code shared between visitCastInst and
+    /// constant expression cast support.
+    void emitCastOperation(MachineBasicBlock *BB,MachineBasicBlock::iterator&IP,
+                           Value *Src, const Type *DestTy, unsigned TargetReg);
+
     /// copyConstantToRegister - Output the instructions required to put the
     /// specified constant into the specified register.
     ///
@@ -307,10 +312,8 @@
       emitGEPOperation(MBB, IP, CE->getOperand(0),
                        CE->op_begin()+1, CE->op_end(), R);
       return;
-    } else if (CE->getOpcode() == Instruction::Cast &&
-               isa<PointerType>(CE->getType()) &&
-               isa<PointerType>(CE->getOperand(0)->getType())) {
-      copyConstantToRegister(MBB, IP, cast<Constant>(CE->getOperand(0)), R);
+    } else if (CE->getOpcode() == Instruction::Cast) {
+      emitCastOperation(MBB, IP, CE->getOperand(0), CE->getType(), R);
       return;
     }
 
@@ -446,7 +449,7 @@
     // Loop over all of the PHI nodes in the LLVM basic block...
     unsigned NumPHIs = 0;
     for (BasicBlock::const_iterator I = BB->begin();
-         PHINode *PN = (PHINode*)dyn_cast<PHINode>(&*I); ++I) {
+         PHINode *PN = (PHINode*)dyn_cast<PHINode>(I); ++I) {
 
       // Create a new machine instr PHI node, and insert it.
       unsigned PHIReg = getReg(*PN);
@@ -1387,22 +1390,30 @@
 /// visitCastInst - Here we have various kinds of copying with or without
 /// sign extension going on.
 void ISel::visitCastInst(CastInst &CI) {
-  const Type *DestTy = CI.getType();
-  Value *Src = CI.getOperand(0);
+  unsigned DestReg = getReg(CI);
+  MachineBasicBlock::iterator MI = BB->end();
+  emitCastOperation(BB, MI, CI.getOperand(0), CI.getType(), DestReg);
+}
+
+/// emitCastOperation - Common code shared between visitCastInst and
+/// constant expression cast support.
+void ISel::emitCastOperation(MachineBasicBlock *BB,
+                             MachineBasicBlock::iterator &IP,
+                             Value *Src, const Type *DestTy,
+                             unsigned DestReg) {
   unsigned SrcReg = getReg(Src);
   const Type *SrcTy = Src->getType();
   unsigned SrcClass = getClassB(SrcTy);
-  unsigned DestReg = getReg(CI);
   unsigned DestClass = getClassB(DestTy);
 
   // Implement casts to bool by using compare on the operand followed by set if
   // not zero on the result.
   if (DestTy == Type::BoolTy) {
     if (SrcClass == cFP || SrcClass == cLong)
-      visitInstruction(CI);
+      abort();  // FIXME: implement cast (long & FP) to bool
     
-    BuildMI(BB, X86::CMPri8, 2).addReg(SrcReg).addZImm(0);
-    BuildMI(BB, X86::SETNEr, 1, DestReg);
+    BMI(BB, IP, X86::CMPri8, 2).addReg(SrcReg).addZImm(0);
+    BMI(BB, IP, X86::SETNEr, 1, DestReg);
     return;
   }
 
@@ -1414,11 +1425,11 @@
   // getClass) by using a register-to-register move.
   if (SrcClass == DestClass) {
     if (SrcClass <= cInt || (SrcClass == cFP && SrcTy == DestTy)) {
-      BuildMI(BB, RegRegMove[SrcClass], 1, DestReg).addReg(SrcReg);
+      BMI(BB, IP, RegRegMove[SrcClass], 1, DestReg).addReg(SrcReg);
     } else if (SrcClass == cFP) {
       if (SrcTy == Type::FloatTy) {  // double -> float
 	assert(DestTy == Type::DoubleTy && "Unknown cFP member!");
-	BuildMI(BB, X86::FpMOV, 1, DestReg).addReg(SrcReg);
+	BMI(BB, IP, X86::FpMOV, 1, DestReg).addReg(SrcReg);
       } else {                       // float -> double
 	assert(SrcTy == Type::DoubleTy && DestTy == Type::FloatTy &&
 	       "Unknown cFP member!");
@@ -1426,14 +1437,14 @@
 	// reading it back.
 	unsigned FltAlign = TM.getTargetData().getFloatAlignment();
         int FrameIdx = F->getFrameInfo()->CreateStackObject(4, FltAlign);
-	addFrameReference(BuildMI(BB, X86::FSTr32, 5), FrameIdx).addReg(SrcReg);
-	addFrameReference(BuildMI(BB, X86::FLDr32, 5, DestReg), FrameIdx);
+	addFrameReference(BMI(BB, IP, X86::FSTr32, 5), FrameIdx).addReg(SrcReg);
+	addFrameReference(BMI(BB, IP, X86::FLDr32, 5, DestReg), FrameIdx);
       }
     } else if (SrcClass == cLong) {
-      BuildMI(BB, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
-      BuildMI(BB, X86::MOVrr32, 1, DestReg+1).addReg(SrcReg+1);
+      BMI(BB, IP, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
+      BMI(BB, IP, X86::MOVrr32, 1, DestReg+1).addReg(SrcReg+1);
     } else {
-      visitInstruction(CI);
+      abort();
     }
     return;
   }
@@ -1451,21 +1462,21 @@
     };
     
     bool isUnsigned = SrcTy->isUnsigned();
-    BuildMI(BB, Opc[isUnsigned][SrcClass + DestClass - 1], 1,
-            DestReg).addReg(SrcReg);
+    BMI(BB, IP, Opc[isUnsigned][SrcClass + DestClass - 1], 1,
+        DestReg).addReg(SrcReg);
 
     if (isLong) {  // Handle upper 32 bits as appropriate...
       if (isUnsigned)     // Zero out top bits...
-	BuildMI(BB, X86::MOVir32, 1, DestReg+1).addZImm(0);
+	BMI(BB, IP, X86::MOVir32, 1, DestReg+1).addZImm(0);
       else                // Sign extend bottom half...
-	BuildMI(BB, X86::SARir32, 2, DestReg+1).addReg(DestReg).addZImm(31);
+	BMI(BB, IP, X86::SARir32, 2, DestReg+1).addReg(DestReg).addZImm(31);
     }
     return;
   }
 
   // Special case long -> int ...
   if (SrcClass == cLong && DestClass == cInt) {
-    BuildMI(BB, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
+    BMI(BB, IP, X86::MOVrr32, 1, DestReg).addReg(SrcReg);
     return;
   }
   
@@ -1474,8 +1485,8 @@
   if ((SrcClass <= cInt || SrcClass == cLong) && DestClass <= cInt
       && SrcClass > DestClass) {
     static const unsigned AReg[] = { X86::AL, X86::AX, X86::EAX, 0, X86::EAX };
-    BuildMI(BB, RegRegMove[SrcClass], 1, AReg[SrcClass]).addReg(SrcReg);
-    BuildMI(BB, RegRegMove[DestClass], 1, DestReg).addReg(AReg[DestClass]);
+    BMI(BB, IP, RegRegMove[SrcClass], 1, AReg[SrcClass]).addReg(SrcReg);
+    BMI(BB, IP, RegRegMove[DestClass], 1, DestReg).addReg(AReg[DestClass]);
     return;
   }
 
@@ -1484,14 +1495,14 @@
     // unsigned int -> load as 64 bit int.
     // unsigned long long -> more complex
     if (SrcTy->isUnsigned() && SrcTy != Type::UByteTy)
-      visitInstruction(CI);  // don't handle unsigned src yet!
+      abort();  // don't handle unsigned src yet!
 
     // We don't have the facilities for directly loading byte sized data from
     // memory.  Promote it to 16 bits.
     if (SrcClass == cByte) {
       unsigned TmpReg = makeAnotherReg(Type::ShortTy);
-      BuildMI(BB, SrcTy->isSigned() ? X86::MOVSXr16r8 : X86::MOVZXr16r8,
-	      1, TmpReg).addReg(SrcReg);
+      BMI(BB, IP, SrcTy->isSigned() ? X86::MOVSXr16r8 : X86::MOVZXr16r8,
+          1, TmpReg).addReg(SrcReg);
       SrcTy = Type::ShortTy;     // Pretend the short is our input now!
       SrcClass = cShort;
       SrcReg = TmpReg;
@@ -1502,18 +1513,18 @@
       F->getFrameInfo()->CreateStackObject(SrcTy, TM.getTargetData());
 
     if (SrcClass == cLong) {
-      if (SrcTy == Type::ULongTy) visitInstruction(CI);
-      addFrameReference(BuildMI(BB, X86::MOVrm32, 5), FrameIdx).addReg(SrcReg);
-      addFrameReference(BuildMI(BB, X86::MOVrm32, 5),
+      if (SrcTy == Type::ULongTy) abort();  // FIXME: Handle ulong -> FP
+      addFrameReference(BMI(BB, IP, X86::MOVrm32, 5), FrameIdx).addReg(SrcReg);
+      addFrameReference(BMI(BB, IP, X86::MOVrm32, 5),
 			FrameIdx, 4).addReg(SrcReg+1);
     } else {
       static const unsigned Op1[] = { X86::MOVrm8, X86::MOVrm16, X86::MOVrm32 };
-      addFrameReference(BuildMI(BB, Op1[SrcClass], 5), FrameIdx).addReg(SrcReg);
+      addFrameReference(BMI(BB, IP, Op1[SrcClass], 5), FrameIdx).addReg(SrcReg);
     }
 
     static const unsigned Op2[] =
       { 0, X86::FILDr16, X86::FILDr32, 0, X86::FILDr64 };
-    addFrameReference(BuildMI(BB, Op2[SrcClass], 5, DestReg), FrameIdx);
+    addFrameReference(BMI(BB, IP, Op2[SrcClass], 5, DestReg), FrameIdx);
     return;
   }
 
@@ -1523,20 +1534,20 @@
     // mode when truncating to an integer value.
     //
     int CWFrameIdx = F->getFrameInfo()->CreateStackObject(2, 2);
-    addFrameReference(BuildMI(BB, X86::FNSTCWm16, 4), CWFrameIdx);
+    addFrameReference(BMI(BB, IP, X86::FNSTCWm16, 4), CWFrameIdx);
 
     // Load the old value of the high byte of the control word...
     unsigned HighPartOfCW = makeAnotherReg(Type::UByteTy);
-    addFrameReference(BuildMI(BB, X86::MOVmr8, 4, HighPartOfCW), CWFrameIdx, 1);
+    addFrameReference(BMI(BB, IP, X86::MOVmr8, 4, HighPartOfCW), CWFrameIdx, 1);
 
     // Set the high part to be round to zero...
-    addFrameReference(BuildMI(BB, X86::MOVim8, 5), CWFrameIdx, 1).addZImm(12);
+    addFrameReference(BMI(BB, IP, X86::MOVim8, 5), CWFrameIdx, 1).addZImm(12);
 
     // Reload the modified control word now...
-    addFrameReference(BuildMI(BB, X86::FLDCWm16, 4), CWFrameIdx);
+    addFrameReference(BMI(BB, IP, X86::FLDCWm16, 4), CWFrameIdx);
     
     // Restore the memory image of control word to original value
-    addFrameReference(BuildMI(BB, X86::MOVrm8, 5),
+    addFrameReference(BMI(BB, IP, X86::MOVrm8, 5),
 		      CWFrameIdx, 1).addReg(HighPartOfCW);
 
     // We don't have the facilities for directly storing byte sized data to
@@ -1549,7 +1560,7 @@
       case cByte:  StoreTy = Type::ShortTy; StoreClass = cShort; break;
       case cShort: StoreTy = Type::IntTy;   StoreClass = cInt;   break;
       case cInt:   StoreTy = Type::LongTy;  StoreClass = cLong;  break;
-      case cLong:  visitInstruction(CI); // unsigned long long -> more complex
+      case cLong:  abort(); // FIXME: unsigned long long -> more complex
       default: assert(0 && "Unknown store class!");
       }
 
@@ -1559,23 +1570,23 @@
 
     static const unsigned Op1[] =
       { 0, X86::FISTr16, X86::FISTr32, 0, X86::FISTPr64 };
-    addFrameReference(BuildMI(BB, Op1[StoreClass], 5), FrameIdx).addReg(SrcReg);
+    addFrameReference(BMI(BB, IP, Op1[StoreClass], 5), FrameIdx).addReg(SrcReg);
 
     if (DestClass == cLong) {
-      addFrameReference(BuildMI(BB, X86::MOVmr32, 4, DestReg), FrameIdx);
-      addFrameReference(BuildMI(BB, X86::MOVmr32, 4, DestReg+1), FrameIdx, 4);
+      addFrameReference(BMI(BB, IP, X86::MOVmr32, 4, DestReg), FrameIdx);
+      addFrameReference(BMI(BB, IP, X86::MOVmr32, 4, DestReg+1), FrameIdx, 4);
     } else {
       static const unsigned Op2[] = { X86::MOVmr8, X86::MOVmr16, X86::MOVmr32 };
-      addFrameReference(BuildMI(BB, Op2[DestClass], 4, DestReg), FrameIdx);
+      addFrameReference(BMI(BB, IP, Op2[DestClass], 4, DestReg), FrameIdx);
     }
 
     // Reload the original control word now...
-    addFrameReference(BuildMI(BB, X86::FLDCWm16, 4), CWFrameIdx);
+    addFrameReference(BMI(BB, IP, X86::FLDCWm16, 4), CWFrameIdx);
     return;
   }
 
   // Anything we haven't handled already, we can't (yet) handle at all.
-  visitInstruction (CI);
+  abort();
 }
 
 // ExactLog2 - This function solves for (Val == 1 << (N-1)) and returns N.  It




From lattner at cs.uiuc.edu  Wed Apr 23 12:58:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 12:58:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200304231757.MAA27471@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.88 -> 1.89

---
Log message:

Fix iterator invalidation problem with cast instructions


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.88 llvm/lib/Target/X86/InstSelectSimple.cpp:1.89
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.88	Wed Apr 23 12:22:12 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Wed Apr 23 12:57:48 2003
@@ -1401,7 +1401,7 @@
                              MachineBasicBlock::iterator &IP,
                              Value *Src, const Type *DestTy,
                              unsigned DestReg) {
-  unsigned SrcReg = getReg(Src);
+  unsigned SrcReg = getReg(Src, BB, IP);
   const Type *SrcTy = Src->getType();
   unsigned SrcClass = getClassB(SrcTy);
   unsigned DestClass = getClassB(DestTy);




From lattner at cs.uiuc.edu  Wed Apr 23 13:39:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 13:39:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/2003-04-23-LinkOnceLost.ll
Message-ID: <200304231838.NAA30641@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

2003-04-23-LinkOnceLost.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Linker/2003-04-23-LinkOnceLost.ll
diff -c /dev/null llvm/test/Regression/Linker/2003-04-23-LinkOnceLost.ll:1.1
*** /dev/null	Wed Apr 23 13:38:34 2003
--- llvm/test/Regression/Linker/2003-04-23-LinkOnceLost.ll	Wed Apr 23 13:38:24 2003
***************
*** 0 ****
--- 1,8 ----
+ ; This fails because the linker renames the non-opaque type not the opaque 
+ ; one...
+ 
+ ; RUN: echo "implementation linkonce void %foo() { ret void } " | as > Output/%s.2.bc
+ ; RUN: as < %s > Output/%s.1.bc
+ ; RUN: link Output/%s.[12].bc | dis | grep foo | grep linkonce
+ 
+ declare void %foo()




From lattner at cs.uiuc.edu  Wed Apr 23 13:39:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 13:39:04 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/Linker.cpp
Message-ID: <200304231838.NAA30658@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

Linker.cpp updated: 1.41 -> 1.42

---
Log message:

Fix Bug: Linker/2003-04-23-LinkOnceLost.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/Linker.cpp
diff -u llvm/lib/Transforms/Utils/Linker.cpp:1.41 llvm/lib/Transforms/Utils/Linker.cpp:1.42
--- llvm/lib/Transforms/Utils/Linker.cpp:1.41	Tue Apr 22 14:13:20 2003
+++ llvm/lib/Transforms/Utils/Linker.cpp	Wed Apr 23 13:38:39 2003
@@ -230,11 +230,26 @@
 
       // Make sure to remember this mapping...
       ValueMap.insert(std::make_pair(SGV, NewDGV));
-    } else if (!SGExtern && !DGExtern && SGV->getLinkage() !=DGV->getLinkage()){
+    } else if (SGV->isExternal()) {
+      // If SGV is external or if both SGV & DGV are external..  Just link the
+      // external globals, we aren't adding anything.
+      ValueMap.insert(std::make_pair(SGV, DGV));
+
+    } else if (DGV->isExternal()) {   // If DGV is external but SGV is not...
+      ValueMap.insert(std::make_pair(SGV, DGV));
+      DGV->setLinkage(SGV->getLinkage());    // Inherit linkage!
+    } else if (SGV->getLinkage() != DGV->getLinkage()) {
       return Error(Err, "Global variables named '" + SGV->getName() +
                    "' have different linkage specifiers!");
-    } else if (SGV->hasExternalLinkage() || SGV->hasLinkOnceLinkage() ||
-               SGV->hasAppendingLinkage()) {
+    } else if (SGV->hasExternalLinkage()) {
+      // Allow linking two exactly identical external global variables...
+      if (SGV->isConstant() != DGV->isConstant() ||
+          SGV->getInitializer() != DGV->getInitializer())
+        return Error(Err, "Global Variable Collision on '" + 
+                     SGV->getType()->getDescription() + " %" + SGV->getName() +
+                     "' - Global variables differ in const'ness");
+      ValueMap.insert(std::make_pair(SGV, DGV));
+    } else if (SGV->hasLinkOnceLinkage()) {
       // If the global variable has a name, and that name is already in use in
       // the Dest module, make sure that the name is a compatible global
       // variable...
@@ -242,13 +257,13 @@
       // Check to see if the two GV's have the same Const'ness...
       if (SGV->isConstant() != DGV->isConstant())
         return Error(Err, "Global Variable Collision on '" + 
-                     SGV->getType()->getDescription() + "':%" + SGV->getName() +
-                     " - Global variables differ in const'ness");
-      if (DGExtern)
-        DGV->setLinkage(SGV->getLinkage());
+                     SGV->getType()->getDescription() + " %" + SGV->getName() +
+                     "' - Global variables differ in const'ness");
 
       // Okay, everything is cool, remember the mapping...
       ValueMap.insert(std::make_pair(SGV, DGV));
+    } else if (SGV->hasAppendingLinkage()) {
+      assert(0 && "FIXME: Appending linkage unimplemented!");
     } else {
       assert(0 && "Unknown linkage!");
     }
@@ -321,9 +336,6 @@
       // 
       DF = cast_or_null<Function>(ST->lookup(SF->getType(), SF->getName()));
 
-    bool SFExtern = SF->isExternal();
-    bool DFExtern = DF ? DF->isExternal() : false;
-
     if (!DF || SF->hasInternalLinkage() || DF->hasInternalLinkage()) {
       // Function does not already exist, simply insert an function signature
       // identical to SF into the dest module...
@@ -344,30 +356,28 @@
 
       // ... and remember this mapping...
       ValueMap.insert(std::make_pair(SF, NewDF));
-    } else if (SF->getLinkage() == GlobalValue::AppendingLinkage) {
-      return Error(Err, "Functions named '" + SF->getName() +
-                   "' have appending linkage!");
-    } else if (!SFExtern && !DFExtern && SF->getLinkage() != DF->getLinkage()) {
+    } else if (SF->isExternal()) {
+      // If SF is external or if both SF & DF are external..  Just link the
+      // external functions, we aren't adding anything.
+      ValueMap.insert(std::make_pair(SF, DF));
+    } else if (DF->isExternal()) {   // If DF is external but SF is not...
+      // Link the external functions, update linkage qualifiers
+      ValueMap.insert(std::make_pair(SF, DF));
+      DF->setLinkage(SF->getLinkage());
+
+    } else if (SF->getLinkage() != DF->getLinkage()) {
       return Error(Err, "Functions named '" + SF->getName() +
                    "' have different linkage specifiers!");
-    } else if (SF->getLinkage() == GlobalValue::ExternalLinkage) {
-      // If the function has a name, and that name is already in use in the Dest
-      // module, make sure that the name is a compatible function...
-      //
-      // Check to make sure the function is not defined in both modules...
-      if (!SF->isExternal() && !DF->isExternal())
-        return Error(Err, "Function '" + 
-                     SF->getFunctionType()->getDescription() + "':\"" + 
-                     SF->getName() + "\" - Function is already defined!");
-      
-      if (DFExtern)
-        DF->setLinkage(SF->getLinkage());
-      
-      // Otherwise, just remember this mapping...
-      ValueMap.insert(std::make_pair(SF, DF));
-    } else if (SF->getLinkage() == GlobalValue::LinkOnceLinkage) {
+    } else if (SF->hasExternalLinkage()) {
+      // The function is defined in both modules!!
+      return Error(Err, "Function '" + 
+                   SF->getFunctionType()->getDescription() + "':\"" + 
+                   SF->getName() + "\" - Function is already defined!");
+    } else if (SF->hasLinkOnceLinkage()) {
       // Completely ignore the source function.
       ValueMap.insert(std::make_pair(SF, DF));
+    } else {
+      assert(0 && "Unknown linkage configuration found!");
     }
   }
   return false;




From lattner at cs.uiuc.edu  Wed Apr 23 13:39:07 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 13:39:07 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/testlink1.ll testlink2.ll
Message-ID: <200304231838.NAA30665@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

testlink1.ll updated: 1.7 -> 1.8
testlink2.ll updated: 1.8 -> 1.9

---
Log message:

Tweak testcases to use linkonce as appropriate


---
Diffs of the changes:

Index: llvm/test/Regression/Linker/testlink1.ll
diff -u llvm/test/Regression/Linker/testlink1.ll:1.7 llvm/test/Regression/Linker/testlink1.ll:1.8
--- llvm/test/Regression/Linker/testlink1.ll:1.7	Mon Apr 21 16:06:25 2003
+++ llvm/test/Regression/Linker/testlink1.ll	Wed Apr 23 13:38:13 2003
@@ -12,7 +12,7 @@
 %Intern2   = internal constant int 792
 
 ; Initialized to point to external %MyVar
-%MyVarPtr  = global { int * }  { int * %MyVar }
+%MyVarPtr  = linkonce global { int * }  { int * %MyVar }
 
 declare int "foo"(int %blah)      ;; Declared in testlink2.ll
 


Index: llvm/test/Regression/Linker/testlink2.ll
diff -u llvm/test/Regression/Linker/testlink2.ll:1.8 llvm/test/Regression/Linker/testlink2.ll:1.9
--- llvm/test/Regression/Linker/testlink2.ll:1.8	Mon Apr 21 16:06:25 2003
+++ llvm/test/Regression/Linker/testlink2.ll	Wed Apr 23 13:38:13 2003
@@ -11,7 +11,7 @@
 %Intern2   = constant int 12345       ;; Intern in one but not in other
 
 %MyIntListPtr = constant { {\2,int}* } { {\2,int}* %MyIntList }
-%MyVarPtr  = global { int * }  { int * %MyVar }
+%MyVarPtr  = linkonce global { int * }  { int * %MyVar }
 
 constant int 412
 




From lattner at cs.uiuc.edu  Wed Apr 23 14:02:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:02:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/ExecutionEngine.cpp
Message-ID: <200304231901.OAA31362@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

ExecutionEngine.cpp updated: 1.5 -> 1.6

---
Log message:

* Implement access to external variables in LLI
* Implement GetElementPtr constant expressions when initializing global variables


---
Diffs of the changes:

Index: llvm/tools/lli/ExecutionEngine.cpp
diff -u llvm/tools/lli/ExecutionEngine.cpp:1.5 llvm/tools/lli/ExecutionEngine.cpp:1.6
--- llvm/tools/lli/ExecutionEngine.cpp:1.5	Mon Apr 21 17:44:36 2003
+++ llvm/tools/lli/ExecutionEngine.cpp	Wed Apr 23 14:01:49 2003
@@ -12,6 +12,7 @@
 #include "llvm/Module.h"
 #include "llvm/Target/TargetData.h"
 #include "Support/Statistic.h"
+#include <dlfcn.h>
 
 Statistic<> NumInitBytes("lli", "Number of bytes of global vars initialized");
 
@@ -29,10 +30,28 @@
 
 GenericValue ExecutionEngine::getConstantValue(const Constant *C) {
   GenericValue Result;
-#define GET_CONST_VAL(TY, CLASS) \
-  case Type::TY##TyID: Result.TY##Val = cast<CLASS>(C)->getValue(); break
+
+  if (ConstantExpr *CE = (ConstantExpr*)dyn_cast<ConstantExpr>(C))
+    switch (CE->getOpcode()) {
+    case Instruction::GetElementPtr: {
+      Result = getConstantValue(cast<Constant>(CE->getOperand(0)));
+      std::vector<Value*> Indexes(CE->op_begin()+1, CE->op_end());
+      uint64_t Offset =
+        TD->getIndexedOffset(CE->getOperand(0)->getType(), Indexes);
+                             
+      Result.LongVal += Offset;
+      return Result;
+    }
+
+    default:
+      std::cerr << "ConstantExpr not handled as global var init: " << *CE
+                << "\n";
+      abort();
+    }
 
   switch (C->getType()->getPrimitiveID()) {
+#define GET_CONST_VAL(TY, CLASS) \
+  case Type::TY##TyID: Result.TY##Val = cast<CLASS>(C)->getValue(); break
     GET_CONST_VAL(Bool   , ConstantBool);
     GET_CONST_VAL(UByte  , ConstantUInt);
     GET_CONST_VAL(SByte  , ConstantSInt);
@@ -57,6 +76,7 @@
     break;
   default:
     std::cout << "ERROR: Constant unimp for type: " << C->getType() << "\n";
+    abort();
   }
   return Result;
 }
@@ -213,15 +233,16 @@
 
       DEBUG(std::cerr << "Global '" << I->getName() << "' -> "
 	              << (void*)GlobalAddress[I] << "\n");
-    } else if (I->getName() == "stdout") {
-      GlobalAddress[I] = &stdout;
-    } else if (I->getName() == "stderr") {
-      GlobalAddress[I] = &stderr;
-    } else if (I->getName() == "stdin") {
-      GlobalAddress[I] = &stdin;
     } else {
-      std::cerr << "Global: " << I->getName() << "\n";
-      assert(0 && "References to external globals not handled yet!");
+      // External variable reference, try to use dlsym to get a pointer to it in
+      // the LLI image.
+      if (void *SymAddr = dlsym(0, I->getName().c_str()))
+        GlobalAddress[I] = SymAddr;
+      else {
+        std::cerr << "Could not resolve external global address: "
+                  << I->getName() << "\n";
+        abort();
+      }
     }
   
   // Now that all of the globals are set up in memory, loop through them all and




From lattner at cs.uiuc.edu  Wed Apr 23 14:10:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:10:03 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200304231909.OAA31442@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.79 -> 1.80

---
Log message:

Fix the super obnoxious "cast to pointer from integer of different size" warnings


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.79 llvm/lib/CWriter/Writer.cpp:1.80
--- llvm/lib/CWriter/Writer.cpp:1.79	Wed Apr 23 11:36:08 2003
+++ llvm/lib/CWriter/Writer.cpp	Wed Apr 23 14:09:22 2003
@@ -946,13 +946,8 @@
 
 void CWriter::visitBinaryOperator(Instruction &I) {
   // binary instructions, shift instructions, setCond instructions.
-  if (isa<PointerType>(I.getType())) {
-    Out << "(";
-    printType(Out, I.getType());
-    Out << ")";
-  }
+  assert(!isa<PointerType>(I.getType()));
       
-  if (isa<PointerType>(I.getType())) Out << "(long long)";
   writeOperand(I.getOperand(0));
 
   switch (I.getOpcode()) {
@@ -975,14 +970,20 @@
   default: std::cerr << "Invalid operator type!" << I; abort();
   }
 
-  if (isa<PointerType>(I.getType())) Out << "(long long)";
   writeOperand(I.getOperand(1));
 }
 
 void CWriter::visitCastInst(CastInst &I) {
   Out << "(";
-  printType(Out, I.getType(), string(""),/*ignoreName*/false, /*namedContext*/false);
+  printType(Out, I.getType(), string(""),/*ignoreName*/false,
+            /*namedContext*/false);
   Out << ")";
+  if (isa<PointerType>(I.getType())&&I.getOperand(0)->getType()->isIntegral() ||
+      isa<PointerType>(I.getOperand(0)->getType())&&I.getType()->isIntegral()) {
+    // Avoid "cast to pointer from integer of different size" warnings
+    Out << "(long)";  
+  }
+
   writeOperand(I.getOperand(0));
 }
 




From lattner at cs.uiuc.edu  Wed Apr 23 14:16:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:16:02 2003
Subject: [llvm-commits] CVS: llvm/lib/CWriter/Writer.cpp
Message-ID: <200304231915.OAA31623@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CWriter:

Writer.cpp updated: 1.80 -> 1.81

---
Log message:

Fix problem where labels were being incorrectly elided, and fix problem
where PHI copies where not emitted for the default label of switch insts


---
Diffs of the changes:

Index: llvm/lib/CWriter/Writer.cpp
diff -u llvm/lib/CWriter/Writer.cpp:1.80 llvm/lib/CWriter/Writer.cpp:1.81
--- llvm/lib/CWriter/Writer.cpp:1.80	Wed Apr 23 14:09:22 2003
+++ llvm/lib/CWriter/Writer.cpp	Wed Apr 23 14:15:13 2003
@@ -817,7 +817,8 @@
     for (Value::use_iterator UI = BB->use_begin(), UE = BB->use_end();
          UI != UE; ++UI)
       if (TerminatorInst *TI = dyn_cast<TerminatorInst>(*UI))
-        if (TI != Prev->getTerminator()) {
+        if (TI != Prev->getTerminator() ||
+            isa<SwitchInst>(Prev->getTerminator())) {
           NeedsLabel = true;
           break;        
         }
@@ -867,8 +868,8 @@
 void CWriter::visitSwitchInst(SwitchInst &SI) {
   Out << "  switch (";
   writeOperand(SI.getOperand(0));
-  Out << ") {\n  default: goto ";
-  writeOperand(SI.getDefaultDest());
+  Out << ") {\n  default:\n";
+  printBranchToBlock(SI.getParent(), SI.getDefaultDest(), 2);
   Out << ";\n";
   for (unsigned i = 2, e = SI.getNumOperands(); i != e; i += 2) {
     Out << "  case ";




From lattner at cs.uiuc.edu  Wed Apr 23 14:22:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:22:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304231921.OAA31774@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.44 -> 1.45

---
Log message:

Add support for _IO_getc function used on linux


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.44 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.45
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.44	Mon Apr 21 17:43:20 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Wed Apr 23 14:20:50 2003
@@ -666,6 +666,11 @@
   return GV;
 }
 
+// int _IO_getc(FILE *stream);
+GenericValue lle_X__IO_getc(FunctionType *F, const vector<GenericValue> &Args) {
+  return lle_X_getc(F, Args);
+}
+
 // int fputc(int C, FILE *stream);
 GenericValue lle_X_fputc(FunctionType *M, const vector<GenericValue> &Args) {
   assert(Args.size() == 2);
@@ -749,6 +754,7 @@
   FuncNames["lle_X_fflush"]       = lle_X_fflush;
   FuncNames["lle_X_fgetc"]        = lle_X_getc;
   FuncNames["lle_X_getc"]         = lle_X_getc;
+  FuncNames["lle_X__IO_getc"]     = lle_X__IO_getc;
   FuncNames["lle_X_fputc"]        = lle_X_fputc;
   FuncNames["lle_X_ungetc"]       = lle_X_ungetc;
   FuncNames["lle_X_fprintf"]      = lle_X_fprintf;




From lattner at cs.uiuc.edu  Wed Apr 23 14:22:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:22:05 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304231921.OAA31781@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.81 -> 1.82

---
Log message:

Implement &|^ on bool values


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.81 llvm/tools/lli/Interpreter/Execution.cpp:1.82
--- llvm/tools/lli/Interpreter/Execution.cpp:1.81	Tue Apr 22 16:22:33 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Wed Apr 23 14:21:00 2003
@@ -286,6 +286,7 @@
 				   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
+    IMPLEMENT_BINARY_OPERATOR(&, Bool);
     IMPLEMENT_BINARY_OPERATOR(&, UByte);
     IMPLEMENT_BINARY_OPERATOR(&, SByte);
     IMPLEMENT_BINARY_OPERATOR(&, UShort);
@@ -306,6 +307,7 @@
                                   const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
+    IMPLEMENT_BINARY_OPERATOR(|, Bool);
     IMPLEMENT_BINARY_OPERATOR(|, UByte);
     IMPLEMENT_BINARY_OPERATOR(|, SByte);
     IMPLEMENT_BINARY_OPERATOR(|, UShort);
@@ -326,6 +328,7 @@
                                    const Type *Ty) {
   GenericValue Dest;
   switch (Ty->getPrimitiveID()) {
+    IMPLEMENT_BINARY_OPERATOR(^, Bool);
     IMPLEMENT_BINARY_OPERATOR(^, UByte);
     IMPLEMENT_BINARY_OPERATOR(^, SByte);
     IMPLEMENT_BINARY_OPERATOR(^, UShort);




From lattner at cs.uiuc.edu  Wed Apr 23 14:56:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:56:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304231955.OAA32185@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.82 -> 1.83

---
Log message:

Fix a problem with setcc instructions and pointers


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.82 llvm/tools/lli/Interpreter/Execution.cpp:1.83
--- llvm/tools/lli/Interpreter/Execution.cpp:1.82	Wed Apr 23 14:21:00 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Wed Apr 23 14:55:35 2003
@@ -348,6 +348,15 @@
 #define IMPLEMENT_SETCC(OP, TY) \
    case Type::TY##TyID: Dest.BoolVal = Src1.TY##Val OP Src2.TY##Val; break
 
+// Handle pointers specially because they must be compared with only as much
+// width as the host has.  We _do not_ want to be comparing 64 bit values when
+// running on a 32-bit target, otherwise the upper 32 bits might mess up
+// comparisons if they contain garbage.
+#define IMPLEMENT_POINTERSETCC(OP) \
+   case Type::PointerTyID: \
+        Dest.BoolVal = (void*)(intptr_t)Src1.PointerVal OP \
+                       (void*)(intptr_t)Src2.PointerVal; break
+
 static GenericValue executeSetEQInst(GenericValue Src1, GenericValue Src2, 
 				     const Type *Ty) {
   GenericValue Dest;
@@ -362,7 +371,7 @@
     IMPLEMENT_SETCC(==, Long);
     IMPLEMENT_SETCC(==, Float);
     IMPLEMENT_SETCC(==, Double);
-    IMPLEMENT_SETCC(==, Pointer);
+    IMPLEMENT_POINTERSETCC(==);
   default:
     std::cout << "Unhandled type for SetEQ instruction: " << *Ty << "\n";
     abort();
@@ -384,7 +393,7 @@
     IMPLEMENT_SETCC(!=, Long);
     IMPLEMENT_SETCC(!=, Float);
     IMPLEMENT_SETCC(!=, Double);
-    IMPLEMENT_SETCC(!=, Pointer);
+    IMPLEMENT_POINTERSETCC(!=);
 
   default:
     std::cout << "Unhandled type for SetNE instruction: " << *Ty << "\n";
@@ -407,7 +416,7 @@
     IMPLEMENT_SETCC(<=, Long);
     IMPLEMENT_SETCC(<=, Float);
     IMPLEMENT_SETCC(<=, Double);
-    IMPLEMENT_SETCC(<=, Pointer);
+    IMPLEMENT_POINTERSETCC(<=);
   default:
     std::cout << "Unhandled type for SetLE instruction: " << Ty << "\n";
     abort();
@@ -429,7 +438,7 @@
     IMPLEMENT_SETCC(>=, Long);
     IMPLEMENT_SETCC(>=, Float);
     IMPLEMENT_SETCC(>=, Double);
-    IMPLEMENT_SETCC(>=, Pointer);
+    IMPLEMENT_POINTERSETCC(>=);
   default:
     std::cout << "Unhandled type for SetGE instruction: " << *Ty << "\n";
     abort();
@@ -451,7 +460,7 @@
     IMPLEMENT_SETCC(<, Long);
     IMPLEMENT_SETCC(<, Float);
     IMPLEMENT_SETCC(<, Double);
-    IMPLEMENT_SETCC(<, Pointer);
+    IMPLEMENT_POINTERSETCC(<);
   default:
     std::cout << "Unhandled type for SetLT instruction: " << *Ty << "\n";
     abort();
@@ -473,7 +482,7 @@
     IMPLEMENT_SETCC(>, Long);
     IMPLEMENT_SETCC(>, Float);
     IMPLEMENT_SETCC(>, Double);
-    IMPLEMENT_SETCC(>, Pointer);
+    IMPLEMENT_POINTERSETCC(>);
   default:
     std::cout << "Unhandled type for SetGT instruction: " << *Ty << "\n";
     abort();




From lattner at cs.uiuc.edu  Wed Apr 23 14:57:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 14:57:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304231956.OAA32190@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.45 -> 1.46

---
Log message:

Implement a bunch of new external functions


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.45 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.46
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.45	Wed Apr 23 14:20:50 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Wed Apr 23 14:55:24 2003
@@ -225,6 +225,12 @@
   return PTOGV(malloc(Args[0].UIntVal));
 }
 
+// void *calloc(uint, uint)
+GenericValue lle_X_calloc(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 2 && "calloc expects two arguments!");
+  return PTOGV(calloc(Args[0].UIntVal, Args[1].UIntVal));
+}
+
 // void free(void *)
 GenericValue lle_X_free(FunctionType *M, const vector<GenericValue> &Args) {
   assert(Args.size() == 1);
@@ -526,6 +532,46 @@
   return GV;
 }
 
+
+//===----------------------------------------------------------------------===//
+// String Functions...
+//===----------------------------------------------------------------------===//
+
+// int strcmp(const char *S1, const char *S2);
+GenericValue lle_X_strcmp(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 2);
+  GenericValue Ret;
+  Ret.IntVal = strcmp((char*)GVTOP(Args[0]), (char*)GVTOP(Args[1]));
+  return Ret;
+}
+
+// char *strcat(char *Dest, const char *src);
+GenericValue lle_X_strcat(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 2);
+  return PTOGV(strcat((char*)GVTOP(Args[0]), (char*)GVTOP(Args[1])));
+}
+
+// char *strcpy(char *Dest, const char *src);
+GenericValue lle_X_strcpy(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 2);
+  return PTOGV(strcpy((char*)GVTOP(Args[0]), (char*)GVTOP(Args[1])));
+}
+
+// long strlen(const char *src);
+GenericValue lle_X_strlen(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 1);
+  GenericValue Ret;
+  Ret.LongVal = strlen((char*)GVTOP(Args[0]));
+  return Ret;
+}
+
+// void *memset(void *S, int C, size_t N)
+GenericValue lle_X_memset(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 3);
+  return PTOGV(memset(GVTOP(Args[0]), Args[1].IntVal, Args[2].UIntVal));
+}
+
+
 //===----------------------------------------------------------------------===//
 // IO Functions...
 //===----------------------------------------------------------------------===//
@@ -727,6 +773,7 @@
   FuncNames["lle_X_exit"]         = lle_X_exit;
   FuncNames["lle_X_abort"]        = lle_X_abort;
   FuncNames["lle_X_malloc"]       = lle_X_malloc;
+  FuncNames["lle_X_calloc"]       = lle_X_calloc;
   FuncNames["lle_X_free"]         = lle_X_free;
   FuncNames["lle_X_atoi"]         = lle_X_atoi;
   FuncNames["lle_X_pow"]          = lle_X_pow;
@@ -745,6 +792,13 @@
   FuncNames["lle_X_sscanf"]       = lle_X_sscanf;
   FuncNames["lle_X_scanf"]        = lle_X_scanf;
   FuncNames["lle_i_clock"]        = lle_i_clock;
+
+  FuncNames["lle_X_strcmp"]       = lle_X_strcmp;
+  FuncNames["lle_X_strcat"]       = lle_X_strcat;
+  FuncNames["lle_X_strcpy"]       = lle_X_strcpy;
+  FuncNames["lle_X_strlen"]       = lle_X_strlen;
+  FuncNames["lle_X_memset"]       = lle_X_memset;
+
   FuncNames["lle_X_fopen"]        = lle_X_fopen;
   FuncNames["lle_X_fclose"]       = lle_X_fclose;
   FuncNames["lle_X_feof"]         = lle_X_feof;




From lattner at cs.uiuc.edu  Wed Apr 23 15:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 15:24:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304232023.PAA32474@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.46 -> 1.47

---
Log message:

Add memcpy


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.46 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.47
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.46	Wed Apr 23 14:55:24 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Wed Apr 23 15:23:16 2003
@@ -571,6 +571,12 @@
   return PTOGV(memset(GVTOP(Args[0]), Args[1].IntVal, Args[2].UIntVal));
 }
 
+// void *memcpy(void *Dest, void *src, size_t Size);
+GenericValue lle_X_memcpy(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 3);
+  return PTOGV(memcpy((char*)GVTOP(Args[0]), (char*)GVTOP(Args[1]),
+                      Args[2].UIntVal));
+}
 
 //===----------------------------------------------------------------------===//
 // IO Functions...
@@ -798,6 +804,7 @@
   FuncNames["lle_X_strcpy"]       = lle_X_strcpy;
   FuncNames["lle_X_strlen"]       = lle_X_strlen;
   FuncNames["lle_X_memset"]       = lle_X_memset;
+  FuncNames["lle_X_memcpy"]       = lle_X_memcpy;
 
   FuncNames["lle_X_fopen"]        = lle_X_fopen;
   FuncNames["lle_X_fclose"]       = lle_X_fclose;




From lattner at cs.uiuc.edu  Wed Apr 23 15:32:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 15:32:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ExecutionDriver.cpp
Message-ID: <200304232031.PAA32577@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ExecutionDriver.cpp updated: 1.2 -> 1.3

---
Log message:

Allow specifying an input file for the program being executed


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ExecutionDriver.cpp
diff -u llvm/tools/bugpoint/ExecutionDriver.cpp:1.2 llvm/tools/bugpoint/ExecutionDriver.cpp:1.3
--- llvm/tools/bugpoint/ExecutionDriver.cpp:1.2	Mon Dec 23 18:44:34 2002
+++ llvm/tools/bugpoint/ExecutionDriver.cpp	Wed Apr 23 15:31:37 2003
@@ -34,6 +34,10 @@
 			    clEnumValN(RunLLC, "run-llc", "Compile with LLC"),
 			    clEnumValN(RunCBE, "run-cbe", "Compile with CBE"),
 			    0));
+
+  cl::opt<std::string>
+  InputFile("input", cl::init("/dev/null"),
+            cl::desc("Filename to pipe in as stdin (default: /dev/null)"));
 }
 
 /// AbstractInterpreter Class - Subclasses of this class are used to execute
@@ -86,7 +90,7 @@
   };
   
   return RunProgramWithTimeout(LLIPath, Args,
-			       "/dev/null", OutputFile, OutputFile);
+			       InputFile, OutputFile, OutputFile);
 }
 
 




From lattner at cs.uiuc.edu  Wed Apr 23 15:41:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 15:41:01 2003
Subject: [llvm-commits] CVS: llvm/tools/gccas/gccas.cpp
Message-ID: <200304232040.PAA32737@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccas:

gccas.cpp updated: 1.58 -> 1.59

---
Log message:

The new CFrontend generates LOTs of basic blocks that just fall through and do
other funky stuff.  Clean it up early.


---
Diffs of the changes:

Index: llvm/tools/gccas/gccas.cpp
diff -u llvm/tools/gccas/gccas.cpp:1.58 llvm/tools/gccas/gccas.cpp:1.59
--- llvm/tools/gccas/gccas.cpp:1.58	Fri Apr 18 18:38:09 2003
+++ llvm/tools/gccas/gccas.cpp	Wed Apr 23 15:40:42 2003
@@ -70,6 +70,7 @@
   addPass(PM, createDeadTypeEliminationPass());  // Eliminate dead types
   addPass(PM, createConstantMergePass());        // Merge dup global constants
   addPass(PM, createVerifierPass());             // Verify that input is correct
+  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
   addPass(PM, createDeadInstEliminationPass());  // Remove Dead code/vars
   addPass(PM, createRaiseAllocationsPass());     // call %malloc -> malloc inst
   addPass(PM, createIndVarSimplifyPass());       // Simplify indvars




From lattner at cs.uiuc.edu  Wed Apr 23 15:42:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 15:42:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/ExecutionEngine.cpp
Message-ID: <200304232041.PAA32747@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli:

ExecutionEngine.cpp updated: 1.6 -> 1.7

---
Log message:

Fix bug where pointers were assumed to always be 64 bits in size!


---
Diffs of the changes:

Index: llvm/tools/lli/ExecutionEngine.cpp
diff -u llvm/tools/lli/ExecutionEngine.cpp:1.6 llvm/tools/lli/ExecutionEngine.cpp:1.7
--- llvm/tools/lli/ExecutionEngine.cpp:1.6	Wed Apr 23 14:01:49 2003
+++ llvm/tools/lli/ExecutionEngine.cpp	Wed Apr 23 15:41:01 2003
@@ -92,6 +92,7 @@
     case Type::ShortTyID:   Ptr->Untyped[0] = Val.UShortVal & 255;
                             Ptr->Untyped[1] = (Val.UShortVal >> 8) & 255;
                             break;
+    Store4BytesLittleEndian:
     case Type::FloatTyID:
     case Type::UIntTyID:
     case Type::IntTyID:     Ptr->Untyped[0] =  Val.UIntVal        & 255;
@@ -99,10 +100,11 @@
                             Ptr->Untyped[2] = (Val.UIntVal >> 16) & 255;
                             Ptr->Untyped[3] = (Val.UIntVal >> 24) & 255;
                             break;
+    case Type::PointerTyID: if (CurMod.has32BitPointers())
+                              goto Store4BytesLittleEndian;
     case Type::DoubleTyID:
     case Type::ULongTyID:
-    case Type::LongTyID:    
-    case Type::PointerTyID: Ptr->Untyped[0] =  Val.ULongVal        & 255;
+    case Type::LongTyID:    Ptr->Untyped[0] =  Val.ULongVal        & 255;
                             Ptr->Untyped[1] = (Val.ULongVal >>  8) & 255;
                             Ptr->Untyped[2] = (Val.ULongVal >> 16) & 255;
                             Ptr->Untyped[3] = (Val.ULongVal >> 24) & 255;
@@ -123,6 +125,7 @@
     case Type::ShortTyID:   Ptr->Untyped[1] = Val.UShortVal & 255;
                             Ptr->Untyped[0] = (Val.UShortVal >> 8) & 255;
                             break;
+    Store4BytesBigEndian:
     case Type::FloatTyID:
     case Type::UIntTyID:
     case Type::IntTyID:     Ptr->Untyped[3] =  Val.UIntVal        & 255;
@@ -130,10 +133,11 @@
                             Ptr->Untyped[1] = (Val.UIntVal >> 16) & 255;
                             Ptr->Untyped[0] = (Val.UIntVal >> 24) & 255;
                             break;
+    case Type::PointerTyID: if (CurMod.has32BitPointers())
+                              goto Store4BytesBigEndian;
     case Type::DoubleTyID:
     case Type::ULongTyID:
-    case Type::LongTyID:    
-    case Type::PointerTyID: Ptr->Untyped[7] =  Val.ULongVal        & 255;
+    case Type::LongTyID:    Ptr->Untyped[7] =  Val.ULongVal        & 255;
                             Ptr->Untyped[6] = (Val.ULongVal >>  8) & 255;
                             Ptr->Untyped[5] = (Val.ULongVal >> 16) & 255;
                             Ptr->Untyped[4] = (Val.ULongVal >> 24) & 255;




From lattner at cs.uiuc.edu  Wed Apr 23 15:42:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Wed Apr 23 15:42:04 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/Miscompilation.cpp
Message-ID: <200304232041.PAA32758@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

Miscompilation.cpp updated: 1.1 -> 1.2

---
Log message:

Print where reference output goes


---
Diffs of the changes:

Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.1 llvm/tools/bugpoint/Miscompilation.cpp:1.2
--- llvm/tools/bugpoint/Miscompilation.cpp:1.1	Mon Dec 23 17:50:16 2002
+++ llvm/tools/bugpoint/Miscompilation.cpp	Wed Apr 23 15:41:18 2003
@@ -43,7 +43,7 @@
     std::cout << "Generating reference output from raw program...";
     Output = executeProgram("bugpoint.reference.out");
     CreatedOutput = true;
-    std::cout << " done!\n";
+    std::cout << " done! Reference output is: bugpoint.reference.out.\n";
   } else if (diffProgram(Output)) {
     std::cout << "\n*** Input program does not match reference diff!\n"
 	      << "    Must be problem with input source!\n";




From lattner at cs.uiuc.edu  Thu Apr 24 10:55:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 10:55:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/CloneModule.cpp
Message-ID: <200304241554.KAA20751@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

CloneModule.cpp updated: 1.3 -> 1.4

---
Log message:

Make sure to preserve endiannes and pointer size when cloning modules!


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/CloneModule.cpp
diff -u llvm/lib/Transforms/Utils/CloneModule.cpp:1.3 llvm/lib/Transforms/Utils/CloneModule.cpp:1.4
--- llvm/lib/Transforms/Utils/CloneModule.cpp:1.3	Tue Apr 22 13:02:26 2003
+++ llvm/lib/Transforms/Utils/CloneModule.cpp	Thu Apr 24 10:54:40 2003
@@ -19,6 +19,8 @@
 Module *CloneModule(const Module *M) {
   // First off, we need to create the new module...
   Module *New = new Module(M->getModuleIdentifier());
+  New->setEndianness(M->getEndianness());
+  New->setPointerSize(M->getPointerSize());
 
   // Create the value map that maps things from the old module over to the new
   // module.




From lattner at cs.uiuc.edu  Thu Apr 24 12:03:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 12:03:02 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.cpp BugDriver.h CrashDebugger.cpp ExecutionDriver.cpp Miscompilation.cpp OptimizerDriver.cpp
Message-ID: <200304241702.MAA21597@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.cpp updated: 1.2 -> 1.3
BugDriver.h updated: 1.4 -> 1.5
CrashDebugger.cpp updated: 1.4 -> 1.5
ExecutionDriver.cpp updated: 1.3 -> 1.4
Miscompilation.cpp updated: 1.2 -> 1.3
OptimizerDriver.cpp updated: 1.3 -> 1.4

---
Log message:

Implement support for bugpoint to identify which FUNCTION an optimization
is miscompiling.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.cpp
diff -u llvm/tools/bugpoint/BugDriver.cpp:1.2 llvm/tools/bugpoint/BugDriver.cpp:1.3
--- llvm/tools/bugpoint/BugDriver.cpp:1.2	Mon Dec 23 17:49:59 2002
+++ llvm/tools/bugpoint/BugDriver.cpp	Thu Apr 24 12:02:17 2003
@@ -14,6 +14,20 @@
 #include "llvm/Pass.h"
 #include <memory>
 
+/// getPassesString - Turn a list of passes into a string which indicates the
+/// command line options that must be passed to add the passes.
+///
+std::string getPassesString(const std::vector<const PassInfo*> &Passes) {
+  std::string Result;
+  for (unsigned i = 0, e = Passes.size(); i != e; ++i) {
+    if (i) Result += " ";
+    Result += "-";
+    Result += Passes[i]->getPassArgument();
+  }
+  return Result;
+}
+
+
 /// ParseInputFile - Given a bytecode or assembly input filename, parse and
 /// return it, or return null if not possible.
 ///


Index: llvm/tools/bugpoint/BugDriver.h
diff -u llvm/tools/bugpoint/BugDriver.h:1.4 llvm/tools/bugpoint/BugDriver.h:1.5
--- llvm/tools/bugpoint/BugDriver.h:1.4	Fri Feb 28 10:13:20 2003
+++ llvm/tools/bugpoint/BugDriver.h	Thu Apr 24 12:02:17 2003
@@ -17,11 +17,17 @@
 class AbstractInterpreter;
 class Instruction;
 
+class ReduceMiscompilingPasses;
+class ReduceMiscompilingFunctions;
+
 class BugDriver {
   const std::string ToolName;  // Name of bugpoint
   Module *Program;             // The raw program, linked together
   std::vector<const PassInfo*> PassesToRun;
   AbstractInterpreter *Interpreter;   // How to run the program
+
+  friend class ReduceMiscompilingPasses;
+  friend class ReduceMiscompilingFunctions;
 public:
   BugDriver(const char *toolname)
     : ToolName(toolname), Program(0), Interpreter(0) {}
@@ -81,7 +87,7 @@
   /// EmitProgressBytecode - This function is used to output the current Program
   /// to a file named "bugpoing-ID.bc".
   ///
-  void EmitProgressBytecode(const PassInfo *Pass, const std::string &ID);
+  void EmitProgressBytecode(const std::string &ID, bool NoFlyer = false);
   
   /// runPasses - Run the specified passes on Program, outputting a bytecode
   /// file and writting the filename into OutputFile if successful.  If the
@@ -148,7 +154,13 @@
   /// is different, true is returned.
   ///
   bool diffProgram(const std::string &ReferenceOutputFile,
-		   const std::string &BytecodeFile = "");
+		   const std::string &BytecodeFile = "",
+                   bool RemoveBytecode = false);
 };
+
+/// getPassesString - Turn a list of passes into a string which indicates the
+/// command line options that must be passed to add the passes.
+///
+std::string getPassesString(const std::vector<const PassInfo*> &Passes);
 
 #endif


Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.4 llvm/tools/bugpoint/CrashDebugger.cpp:1.5
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.4	Fri Feb 28 10:13:20 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Apr 24 12:02:17 2003
@@ -11,6 +11,23 @@
 #include "llvm/Pass.h"
 #include <fstream>
 
+#if 0
+class DebugCrashes : public ListReducer<const PassInfo*> {
+  BugDriver &BD;
+public:
+  DebugCrashes(BugDriver &bd) : BD(bd) {}
+
+  // doTest - Return true iff running the "removed" passes succeeds, and running
+  // the "Kept" passes fail when run on the output of the "removed" passes.  If
+  // we return true, we update the current module of bugpoint.
+  //
+  virtual bool doTest(const std::vector<ElTy> &Removed,
+                      const std::vector<ElTy> &Kept) {
+    return BD.runPasses(Kept);
+  }
+};
+#endif
+
 /// debugCrash - This method is called when some pass crashes on input.  It
 /// attempts to prune down the testcase to something reasonable, and figure
 /// out exactly which pass is crashing.
@@ -18,7 +35,11 @@
 bool BugDriver::debugCrash() {
   std::cout << "\n*** Debugging optimizer crash!\n";
 
-  // Determine which pass causes the optimizer to crash... using binary search
+#if 0
+  // Reduce the list of passes which causes the optimizer to crash...
+  DebugCrashes(*this).reduceList(PassesToRun);
+#endif
+
   unsigned LastToPass = 0, LastToCrash = PassesToRun.size();
   while (LastToPass != LastToCrash) {
     unsigned Mid = (LastToCrash+LastToPass+1) / 2;
@@ -65,6 +86,9 @@
     removeFile(Filename);
   }
 
+  PassesToRun.clear();
+  PassesToRun.push_back(CrashingPass);
+
   return debugPassCrash(CrashingPass);
 }
 
@@ -83,7 +107,7 @@
 /// crashes, but it smaller.
 ///
 bool BugDriver::debugPassCrash(const PassInfo *Pass) {
-  EmitProgressBytecode(Pass, "passinput");
+  EmitProgressBytecode("passinput");
   bool Reduced = false, AnyReduction = false;
 
   if (CountFunctions(Program) > 1) {
@@ -124,7 +148,7 @@
   }
 
   if (Reduced) {
-    EmitProgressBytecode(Pass, "reduced-function");
+    EmitProgressBytecode("reduced-function");
     Reduced = false;
   }
 
@@ -196,7 +220,7 @@
   }
 
   if (Reduced) {
-    EmitProgressBytecode(Pass, "reduced-simplified");
+    EmitProgressBytecode("reduced-simplified");
     Reduced = false;
   }
 


Index: llvm/tools/bugpoint/ExecutionDriver.cpp
diff -u llvm/tools/bugpoint/ExecutionDriver.cpp:1.3 llvm/tools/bugpoint/ExecutionDriver.cpp:1.4
--- llvm/tools/bugpoint/ExecutionDriver.cpp:1.3	Wed Apr 23 15:31:37 2003
+++ llvm/tools/bugpoint/ExecutionDriver.cpp	Thu Apr 24 12:02:17 2003
@@ -164,7 +164,8 @@
 /// different, true is returned.
 ///
 bool BugDriver::diffProgram(const std::string &ReferenceOutputFile,
-			    const std::string &BytecodeFile) {
+			    const std::string &BytecodeFile,
+                            bool RemoveBytecode) {
   // Execute the program, generating an output file...
   std::string Output = executeProgram("", BytecodeFile);
 
@@ -192,5 +193,6 @@
   } while (C1 != EOF);
 
   removeFile(Output);
+  if (RemoveBytecode) removeFile(BytecodeFile);
   return FilesDifferent;
 }


Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.2 llvm/tools/bugpoint/Miscompilation.cpp:1.3
--- llvm/tools/bugpoint/Miscompilation.cpp:1.2	Wed Apr 23 15:41:18 2003
+++ llvm/tools/bugpoint/Miscompilation.cpp	Thu Apr 24 12:02:17 2003
@@ -8,6 +8,8 @@
 #include "SystemUtils.h"
 #include "llvm/Pass.h"
 #include "llvm/Module.h"
+#include "llvm/Transforms/Utils/Cloning.h"
+#include "llvm/Transforms/Utils/Linker.h"
 #include "Support/CommandLine.h"
 
 // Anonymous namespace to define command line options for miscompilation
@@ -24,6 +26,322 @@
 			    "(for miscompilation detection)"));
 }
 
+template<typename ElTy>
+struct ListReducer {
+  enum TestResult {
+    NoFailure,         // No failure of the predicate was detected
+    KeepSuffix,        // The suffix alone satisfies the predicate
+    KeepPrefix,        // The prefix alone satisfies the predicate
+  };
+
+  // doTest - This virtual function should be overriden by subclasses to
+  // implement the test desired.  The testcase is only required to test to see
+  // if the Kept list still satisfies the property, but if it is going to check
+  // the prefix anyway, it can.
+  //
+  virtual TestResult doTest(const std::vector<ElTy> &Prefix,
+                            const std::vector<ElTy> &Kept) = 0;
+
+  // reduceList - This function attempts to reduce the length of the specified
+  // list while still maintaining the "test" property.  This is the core of the
+  // "work" that bugpoint does.
+  //
+  void reduceList(std::vector<ElTy> &TheList) {
+    unsigned MidTop = TheList.size();
+    while (MidTop > 1) {
+      unsigned Mid = MidTop / 2;
+      std::vector<ElTy> Prefix(TheList.begin()+Mid, TheList.end());
+      std::vector<ElTy> Kept  (TheList.begin(), TheList.begin()+Mid);
+
+      switch (doTest(Prefix, Kept)) {
+      case KeepSuffix:
+        // The property still holds.  We can just drop the prefix elements, and
+        // shorten the list to the "kept" elements.
+        TheList.swap(Kept);
+        MidTop = TheList.size();
+        break;
+      case KeepPrefix:
+        // The predicate still holds, shorten the list to the prefix elements.
+        TheList.swap(Prefix);
+        MidTop = TheList.size();
+        break;
+      case NoFailure:
+        // Otherwise the property doesn't hold.  Some of the elements we removed
+        // must be neccesary to maintain the property.
+        MidTop = Mid;
+        break;
+      }
+    }
+  }
+};
+
+class ReduceMiscompilingPasses : public ListReducer<const PassInfo*> {
+  BugDriver &BD;
+public:
+  ReduceMiscompilingPasses(BugDriver &bd) : BD(bd) {}
+
+  virtual TestResult doTest(const std::vector<const PassInfo*> &Prefix,
+                            const std::vector<const PassInfo*> &Kept);
+};
+
+ReduceMiscompilingPasses::TestResult
+ReduceMiscompilingPasses::doTest(const std::vector<const PassInfo*> &Prefix,
+                                 const std::vector<const PassInfo*> &Kept) {
+  // First, run the program with just the Kept passes.  If it is still broken
+  // with JUST the kept passes, discard the prefix passes.
+  std::cout << "Checking to see if '" << getPassesString(Kept)
+            << "' compile correctly: ";
+
+  std::string BytecodeResult;
+  if (BD.runPasses(Kept, BytecodeResult, false/*delete*/, true/*quiet*/)) {
+    std::cerr << BD.getToolName() << ": Error running this sequence of passes"
+              << " on the input program!\n";
+    exit(1);
+  }
+
+  // Check to see if the finished program matches the reference output...
+  if (BD.diffProgram(Output, BytecodeResult, true /*delete bytecode*/)) {
+    std::cout << "nope.\n";
+    return KeepSuffix;        // Miscompilation detected!
+  }
+  std::cout << "yup.\n";      // No miscompilation!
+
+  if (Prefix.empty()) return NoFailure;
+
+  // First, run the program with just the Kept passes.  If it is still broken
+  // with JUST the kept passes, discard the prefix passes.
+  std::cout << "Checking to see if '" << getPassesString(Prefix)
+            << "' compile correctly: ";
+
+  // If it is not broken with the kept passes, it's possible that the prefix
+  // passes must be run before the kept passes to break it.  If the program
+  // WORKS after the prefix passes, but then fails if running the prefix AND
+  // kept passes, we can update our bytecode file to include the result of the
+  // prefix passes, then discard the prefix passes.
+  //
+  if (BD.runPasses(Prefix, BytecodeResult, false/*delete*/, true/*quiet*/)) {
+    std::cerr << BD.getToolName() << ": Error running this sequence of passes"
+              << " on the input program!\n";
+    exit(1);
+  }
+
+  // If the prefix maintains the predicate by itself, only keep the prefix!
+  if (BD.diffProgram(Output, BytecodeResult)) {
+    std::cout << "nope.\n";
+    removeFile(BytecodeResult);
+    return KeepPrefix;
+  }
+  std::cout << "yup.\n";      // No miscompilation!
+
+  // Ok, so now we know that the prefix passes work, try running the suffix
+  // passes on the result of the prefix passes.
+  //
+  Module *PrefixOutput = BD.ParseInputFile(BytecodeResult);
+  if (PrefixOutput == 0) {
+    std::cerr << BD.getToolName() << ": Error reading bytecode file '"
+              << BytecodeResult << "'!\n";
+    exit(1);
+  }
+  removeFile(BytecodeResult);  // No longer need the file on disk
+    
+  std::cout << "Checking to see if '" << getPassesString(Kept)
+            << "' passes compile correctly after the '"
+            << getPassesString(Prefix) << "' passes: ";
+
+  Module *OriginalInput = BD.Program;
+  BD.Program = PrefixOutput;
+  if (BD.runPasses(Kept, BytecodeResult, false/*delete*/, true/*quiet*/)) {
+    std::cerr << BD.getToolName() << ": Error running this sequence of passes"
+              << " on the input program!\n";
+    exit(1);
+  }
+
+  // Run the result...
+  if (BD.diffProgram(Output, BytecodeResult, true/*delete bytecode*/)) {
+    std::cout << "nope.\n";
+    delete OriginalInput;     // We pruned down the original input...
+    return KeepPrefix;
+  }
+
+  // Otherwise, we must not be running the bad pass anymore.
+  std::cout << "yup.\n";      // No miscompilation!
+  BD.Program = OriginalInput; // Restore original program
+  delete PrefixOutput;        // Free experiment
+  return NoFailure;
+}
+
+static void PrintFunctionList(const std::vector<Function*> &Funcs) {
+  for (unsigned i = 0, e = Funcs.size(); i != e; ++i) {
+    if (i) std::cout << ", ";
+    std::cout << Funcs[i]->getName();
+  }
+}
+
+
+class ReduceMiscompilingFunctions : public ListReducer<Function*> {
+  BugDriver &BD;
+public:
+  ReduceMiscompilingFunctions(BugDriver &bd) : BD(bd) {}
+
+  virtual TestResult doTest(const std::vector<Function*> &Prefix,
+                            const std::vector<Function*> &Kept) {
+    if (TestFuncs(Kept, false))
+      return KeepSuffix;
+    if (TestFuncs(Prefix, false))
+      return KeepPrefix;
+    return NoFailure;
+  }
+  
+  bool TestFuncs(const std::vector<Function*> &Prefix, bool EmitBytecode);
+};
+
+// DeleteFunctionBody - "Remove" the function by deleting all of it's basic
+// blocks, making it external.
+//
+static void DeleteFunctionBody(Function *F) {
+  // First, break circular use/def chain references...
+  for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I)
+    I->dropAllReferences();
+
+  // Next, delete all of the basic blocks.
+  F->getBasicBlockList().clear();
+
+  assert(F->isExternal() && "This didn't make the function external!");
+}
+
+
+bool ReduceMiscompilingFunctions::TestFuncs(const std::vector<Function*> &Funcs,
+                                            bool EmitBytecode) {
+  // Test to see if the function is misoptimized if we ONLY run it on the
+  // functions listed in Funcs.
+  if (!EmitBytecode) {
+    std::cout << "Checking to see if the program is misoptimized when these "
+              << "functions are run\nthrough the passes: ";
+    PrintFunctionList(Funcs);
+    std::cout << "\n";
+  } else {
+    std::cout <<"Outputting reduced bytecode files which expose the problem:\n";
+  }
+
+  // First step: clone the module for the two halves of the program we want.
+  Module *ToOptimize = CloneModule(BD.Program);
+
+  // Second step: Make sure functions & globals are all external so that linkage
+  // between the two modules will work.
+  for (Module::iterator I = ToOptimize->begin(), E = ToOptimize->end();I!=E;++I)
+    I->setLinkage(GlobalValue::ExternalLinkage);
+  for (Module::giterator I = ToOptimize->gbegin(), E = ToOptimize->gend();
+       I != E; ++I)
+    I->setLinkage(GlobalValue::ExternalLinkage);
+
+  // Third step: make a clone of the externalized program for the non-optimized
+  // part.
+  Module *ToNotOptimize = CloneModule(ToOptimize);
+
+  // Fourth step: Remove the test functions from the ToNotOptimize module, and
+  // all of the global variables.
+  for (unsigned i = 0, e = Funcs.size(); i != e; ++i) {
+    Function *TNOF = ToNotOptimize->getFunction(Funcs[i]->getName(),
+                                                Funcs[i]->getFunctionType());
+    assert(TNOF && "Function doesn't exist in module!");
+    DeleteFunctionBody(TNOF);       // Function is now external in this module!
+  }
+  for (Module::giterator I = ToNotOptimize->gbegin(), E = ToNotOptimize->gend();
+       I != E; ++I)
+    I->setInitializer(0);  // Delete the initializer to make it external
+
+  if (EmitBytecode) {
+    std::cout << "  Non-optimized portion: ";
+    std::swap(BD.Program, ToNotOptimize);
+    BD.EmitProgressBytecode("tonotoptimize", true);
+    std::swap(BD.Program, ToNotOptimize);
+  }
+
+  // Fifth step: Remove all functions from the ToOptimize module EXCEPT for the
+  // ones specified in Funcs.  We know which ones these are because they are
+  // non-external in ToOptimize, but external in ToNotOptimize.
+  //
+  for (Module::iterator I = ToOptimize->begin(), E = ToOptimize->end();I!=E;++I)
+    if (!I->isExternal()) {
+      Function *TNOF = ToNotOptimize->getFunction(I->getName(),
+                                                  I->getFunctionType());
+      assert(TNOF && "Function doesn't exist in ToNotOptimize module??");
+      if (!TNOF->isExternal())
+        DeleteFunctionBody(I);
+    }
+
+  if (EmitBytecode) {
+    std::cout << "  Portion that is input to optimizer: ";
+    std::swap(BD.Program, ToOptimize);
+    BD.EmitProgressBytecode("tooptimize");
+    std::swap(BD.Program, ToOptimize);
+  }
+
+  // Sixth step: Run the optimization passes on ToOptimize, producing a
+  // transformed version of the functions being tested.
+  Module *OldProgram = BD.Program;
+  BD.Program = ToOptimize;
+
+  if (!EmitBytecode)
+    std::cout << "  Optimizing functions being tested: ";
+  std::string BytecodeResult;
+  if (BD.runPasses(BD.PassesToRun, BytecodeResult, false/*delete*/,
+                   true/*quiet*/)) {
+    std::cerr << BD.getToolName() << ": Error running this sequence of passes"
+              << " on the input program!\n";
+    exit(1);
+  }
+
+  if (!EmitBytecode)
+    std::cout << "done.\n";
+
+  delete BD.Program;   // Delete the old "ToOptimize" module
+  BD.Program = BD.ParseInputFile(BytecodeResult);
+
+  if (EmitBytecode) {
+    std::cout << "  'tooptimize' after being optimized: ";
+    BD.EmitProgressBytecode("optimized", true);
+  }
+
+  if (BD.Program == 0) {
+    std::cerr << BD.getToolName() << ": Error reading bytecode file '"
+              << BytecodeResult << "'!\n";
+    exit(1);
+  }
+  removeFile(BytecodeResult);  // No longer need the file on disk
+
+  // Seventh step: Link the optimized part of the program back to the
+  // unoptimized part of the program.
+  //
+  if (LinkModules(BD.Program, ToNotOptimize, &BytecodeResult)) {
+    std::cerr << BD.getToolName() << ": Error linking modules together:"
+              << BytecodeResult << "\n";
+    exit(1);
+  }
+  delete ToNotOptimize;  // We are done with this module...
+
+  if (EmitBytecode) {
+    std::cout << "  Program as tested: ";
+    BD.EmitProgressBytecode("linked", true);
+    delete BD.Program;
+    BD.Program = OldProgram;
+    return false;   // We don't need to actually execute the program here.
+  }
+
+  std::cout << "  Checking to see if the merged program executes correctly: ";
+
+  // Eighth step: Execute the program.  If it does not match the expected
+  // output, then 'Funcs' are being misoptimized!
+  bool Broken = BD.diffProgram(Output);
+
+  delete BD.Program;  // Delete the hacked up program
+  BD.Program = OldProgram;   // Restore the original
+
+  std::cout << (Broken ? "nope.\n" : "yup.\n");
+  return Broken;
+}
+
+
 /// debugMiscompilation - This method is used when the passes selected are not
 /// crashing, but the generated output is semantically different from the
 /// input.
@@ -50,92 +368,42 @@
     return false;  // Problem found
   }
 
-  // Figure out which transformation is the first to miscompile the input
-  // program.  We do a binary search here in case there are a large number of
-  // passes involved.
-  //
-  unsigned LastGood = 0, LastBad = PassesToRun.size();
-  while (LastGood != LastBad) {
-    unsigned Mid = (LastBad+LastGood+1) / 2;
-    std::vector<const PassInfo*> P(PassesToRun.begin(),
-                                   PassesToRun.begin()+Mid);
-    std::cout << "Checking to see if the first " << Mid << " passes are ok: ";
-
-    std::string BytecodeResult;
-    if (runPasses(P, BytecodeResult, false, true)) {
-      std::cerr << ToolName << ": Error running this sequence of passes"
-		<< " on the input program!\n";
-      exit(1);
-    }
-
-    // Check to see if the finished program matches the reference output...
-    if (diffProgram(Output, BytecodeResult)) {
-      std::cout << "nope.\n";
-      LastBad = Mid-1;    // Miscompilation detected!
-    } else {
-      std::cout << "yup.\n";
-      LastGood = Mid;     // No miscompilation!
-    }
-
-    // We are now done with the optimized output... so remove it.
-    removeFile(BytecodeResult);
-  }
+  // Figure out which transformations miscompile the input program.
+  unsigned OldSize = PassesToRun.size();
+  ReduceMiscompilingPasses(*this).reduceList(PassesToRun);
 
   // Make sure something was miscompiled...
-  if (LastBad >= PassesToRun.size()) {
+  if (PassesToRun.size() == OldSize) {
     std::cerr << "*** Optimized program matches reference output!  No problem "
 	      << "detected...\nbugpoint can't help you with your problem!\n";
     return false;
   }
 
-  // Calculate which pass it is that miscompiles...
-  const PassInfo *ThePass = PassesToRun[LastBad];
-  
-  std::cout << "\n*** Found miscompiling pass '-" << ThePass->getPassArgument()
-            << "': " << ThePass->getPassName() << "\n";
-  
-  if (LastGood != 0) {
-    std::vector<const PassInfo*> P(PassesToRun.begin(), 
-                                   PassesToRun.begin()+LastGood);
-    std::string Filename;
-    std::cout << "Running good passes to get input for pass:";
-    if (runPasses(P, Filename, false, true)) {
-      std::cerr << "ERROR: Running the first " << LastGood
-                << " passes crashed!\n";
-      return true;
-    }
-    std::cout << " done!\n";
-    
-    // Assuming everything was successful, we now have a valid bytecode file in
-    // OutputName.  Use it for "Program" Instead.
-    delete Program;
-    Program = ParseInputFile(Filename);
-    
-    // Delete the file now.
-    removeFile(Filename);
-  }
+  std::cout << "\n*** Found miscompiling pass"
+            << (PassesToRun.size() == 1 ? "" : "es") << ": "
+            << getPassesString(PassesToRun) << "\n";
+  EmitProgressBytecode("passinput");
+
+
+  // Okay, now that we have reduced the list of passes which are causing the
+  // failure, see if we can pin down which functions are being
+  // miscompiled... first build a list of all of the non-external functions in
+  // the program.
+  std::vector<Function*> MiscompiledFunctions;
+  for (Module::iterator I = Program->begin(), E = Program->end(); I != E; ++I)
+    if (!I->isExternal())
+      MiscompiledFunctions.push_back(I);
+
+  // Do the reduction...
+  ReduceMiscompilingFunctions(*this).reduceList(MiscompiledFunctions);
+
+  std::cout << "\n*** The following functions are being miscompiled: ";
+  PrintFunctionList(MiscompiledFunctions);
+  std::cout << "\n";
 
-  bool Result = debugPassMiscompilation(ThePass, Output);
+  // Output a bunch of bytecode files for the user...
+  ReduceMiscompilingFunctions(*this).TestFuncs(MiscompiledFunctions, true);
 
   if (CreatedOutput) removeFile(Output);
-  return Result;
-}
-
-/// debugPassMiscompilation - This method is called when the specified pass
-/// miscompiles Program as input.  It tries to reduce the testcase to something
-/// that smaller that still miscompiles the program.  ReferenceOutput contains
-/// the filename of the file containing the output we are to match.
-///
-bool BugDriver::debugPassMiscompilation(const PassInfo *Pass,
-                                        const std::string &ReferenceOutput) {
-  EmitProgressBytecode(Pass, "passinput");
-
-  // Loop over all of the functions in the program, attempting to find one that
-  // is being miscompiled.  We do this by extracting the function into a module,
-  // running the "bad" optimization on that module, then linking it back into
-  // the program.  If the program fails the diff, the function got misoptimized.
-  //
-
-
   return false;
 }


Index: llvm/tools/bugpoint/OptimizerDriver.cpp
diff -u llvm/tools/bugpoint/OptimizerDriver.cpp:1.3 llvm/tools/bugpoint/OptimizerDriver.cpp:1.4
--- llvm/tools/bugpoint/OptimizerDriver.cpp:1.3	Wed Jan 22 17:24:11 2003
+++ llvm/tools/bugpoint/OptimizerDriver.cpp	Thu Apr 24 12:02:17 2003
@@ -35,8 +35,7 @@
 /// EmitProgressBytecode - This function is used to output the current Program
 /// to a file named "bugpoing-ID.bc".
 ///
-void BugDriver::EmitProgressBytecode(const PassInfo *Pass,
-                                     const std::string &ID) {
+void BugDriver::EmitProgressBytecode(const std::string &ID, bool NoFlyer) {
   // Output the input to the current pass to a bytecode file, emit a message
   // telling the user how to reproduce it: opt -foo blah.bc
   //
@@ -47,9 +46,13 @@
   }
 
   std::cout << "Emitted bytecode to '" << Filename << "'\n";
+  if (NoFlyer) return;
   std::cout << "\n*** You can reproduce the problem with: ";
 
-  unsigned PassType = Pass->getPassType();
+  unsigned PassType = PassesToRun[0]->getPassType();
+  for (unsigned i = 1, e = PassesToRun.size(); i != e; ++i)
+    PassType &= PassesToRun[i]->getPassType();
+
   if (PassType & PassInfo::Analysis)
     std::cout << "analyze";
   else if (PassType & PassInfo::Optimization)
@@ -58,7 +61,8 @@
     std::cout << "llc";
   else
     std::cout << "bugpoint";
-  std::cout << " " << Filename << " -" << Pass->getPassArgument() << "\n";
+  std::cout << " " << Filename << " ";
+  std::cout << getPassesString(PassesToRun) << "\n";
 }
 
 /// FIXME: This should be parameterizable!!




From lattner at cs.uiuc.edu  Thu Apr 24 12:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 12:16:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/CloneModule.cpp
Message-ID: <200304241715.MAA22019@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

CloneModule.cpp updated: 1.4 -> 1.5

---
Log message:

Make sure that the cloned module retains the type symbol table entries!


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/CloneModule.cpp
diff -u llvm/lib/Transforms/Utils/CloneModule.cpp:1.4 llvm/lib/Transforms/Utils/CloneModule.cpp:1.5
--- llvm/lib/Transforms/Utils/CloneModule.cpp:1.4	Thu Apr 24 10:54:40 2003
+++ llvm/lib/Transforms/Utils/CloneModule.cpp	Thu Apr 24 12:15:33 2003
@@ -8,6 +8,7 @@
 #include "llvm/Transforms/Utils/Cloning.h"
 #include "llvm/Module.h"
 #include "llvm/DerivedTypes.h"
+#include "llvm/SymbolTable.h"
 #include "llvm/Constant.h"
 #include "ValueMapper.h"
 
@@ -21,6 +22,14 @@
   Module *New = new Module(M->getModuleIdentifier());
   New->setEndianness(M->getEndianness());
   New->setPointerSize(M->getPointerSize());
+
+  // Copy all of the type symbol table entries over...
+  const SymbolTable &SymTab = M->getSymbolTable();
+  SymbolTable::const_iterator TypeI = SymTab.find(Type::TypeTy);
+  if (TypeI != SymTab.end())
+    for (SymbolTable::VarMap::const_iterator I = TypeI->second.begin(),
+           E = TypeI->second.end(); I != E; ++I)
+      New->addTypeName(I->first, cast<Type>(I->second));
 
   // Create the value map that maps things from the old module over to the new
   // module.




From lattner at cs.uiuc.edu  Thu Apr 24 12:53:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 12:53:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Message-ID: <200304241752.MAA24463@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

PromoteMemoryToRegister.cpp updated: 1.39 -> 1.40

---
Log message:

Fix iterator invalidation problem


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
diff -u llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.39 llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.40
--- llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.39	Mon Apr 21 14:15:26 2003
+++ llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp	Thu Apr 24 12:52:20 2003
@@ -201,8 +201,13 @@
   // because it is an unreachable predecessor), that all PHI nodes will have the
   // correct number of entries for their predecessors.
   Value *NullVal = Constant::getNullValue(PN->getType());
-  for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE; ++PI)
-    PN->addIncoming(NullVal, *PI);
+
+  // This is neccesary because adding incoming values to the PHI node adds uses
+  // to the basic blocks being used, which can invalidate the predecessor
+  // iterator!
+  std::vector<BasicBlock*> Preds(pred_begin(BB), pred_end(BB));
+  for (unsigned i = 0, e = Preds.size(); i != e; ++i)
+    PN->addIncoming(NullVal, Preds[i]);
 
   BBPNs[AllocaNo] = PN;
   PhiNodes[AllocaNo].push_back(BB);




From lattner at cs.uiuc.edu  Thu Apr 24 13:26:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:26:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/ExprTypeConvert.cpp LevelRaise.cpp TransformInternals.cpp TransformInternals.h
Message-ID: <200304241825.NAA26243@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms:

ExprTypeConvert.cpp updated: 1.69 -> 1.70
LevelRaise.cpp updated: 1.76 -> 1.77
TransformInternals.cpp updated: 1.32 -> 1.33
TransformInternals.h updated: 1.16 -> 1.17

---
Log message:

Make the levelraise pass be well behaved w.r.t the TargetData that the current
PassMAnager provides.


---
Diffs of the changes:

Index: llvm/lib/Transforms/ExprTypeConvert.cpp
diff -u llvm/lib/Transforms/ExprTypeConvert.cpp:1.69 llvm/lib/Transforms/ExprTypeConvert.cpp:1.70
--- llvm/lib/Transforms/ExprTypeConvert.cpp:1.69	Thu Feb 13 21:07:17 2003
+++ llvm/lib/Transforms/ExprTypeConvert.cpp	Thu Apr 24 13:25:27 2003
@@ -18,10 +18,11 @@
 using std::cerr;
 
 static bool OperandConvertableToType(User *U, Value *V, const Type *Ty,
-                                     ValueTypeCache &ConvertedTypes);
+                                     ValueTypeCache &ConvertedTypes,
+                                     const TargetData &TD);
 
 static void ConvertOperandToType(User *U, Value *OldVal, Value *NewVal,
-                                 ValueMapCache &VMC);
+                                 ValueMapCache &VMC, const TargetData &TD);
 
 // Peephole Malloc instructions: we take a look at the use chain of the
 // malloc instruction, and try to find out if the following conditions hold:
@@ -35,7 +36,8 @@
 // element.  TODO: This comment is out of date WRT arrays
 //
 static bool MallocConvertableToType(MallocInst *MI, const Type *Ty,
-                                    ValueTypeCache &CTMap) {
+                                    ValueTypeCache &CTMap,
+                                    const TargetData &TD) {
   if (!isa<PointerType>(Ty)) return false;   // Malloc always returns pointers
 
   // Deal with the type to allocate, not the pointer type...
@@ -73,7 +75,8 @@
 
 static Instruction *ConvertMallocToType(MallocInst *MI, const Type *Ty,
                                         const std::string &Name,
-                                        ValueMapCache &VMC){
+                                        ValueMapCache &VMC,
+                                        const TargetData &TD){
   BasicBlock *BB = MI->getParent();
   BasicBlock::iterator It = BB->end();
 
@@ -131,7 +134,7 @@
 
 // ExpressionConvertableToType - Return true if it is possible
 bool ExpressionConvertableToType(Value *V, const Type *Ty,
-                                 ValueTypeCache &CTMap) {
+                                 ValueTypeCache &CTMap, const TargetData &TD) {
   // Expression type must be holdable in a register.
   if (!Ty->isFirstClassType())
     return false;
@@ -172,8 +175,8 @@
   case Instruction::Add:
   case Instruction::Sub:
     if (!Ty->isInteger() && !Ty->isFloatingPoint()) return false;
-    if (!ExpressionConvertableToType(I->getOperand(0), Ty, CTMap) ||
-        !ExpressionConvertableToType(I->getOperand(1), Ty, CTMap))
+    if (!ExpressionConvertableToType(I->getOperand(0), Ty, CTMap, TD) ||
+        !ExpressionConvertableToType(I->getOperand(1), Ty, CTMap, TD))
       return false;
     break;
   case Instruction::Shr:
@@ -182,27 +185,27 @@
     // FALL THROUGH
   case Instruction::Shl:
     if (!Ty->isInteger()) return false;
-    if (!ExpressionConvertableToType(I->getOperand(0), Ty, CTMap))
+    if (!ExpressionConvertableToType(I->getOperand(0), Ty, CTMap, TD))
       return false;
     break;
 
   case Instruction::Load: {
     LoadInst *LI = cast<LoadInst>(I);
     if (!ExpressionConvertableToType(LI->getPointerOperand(),
-                                     PointerType::get(Ty), CTMap))
+                                     PointerType::get(Ty), CTMap, TD))
       return false;
     break;                                     
   }
   case Instruction::PHINode: {
     PHINode *PN = cast<PHINode>(I);
     for (unsigned i = 0; i < PN->getNumIncomingValues(); ++i)
-      if (!ExpressionConvertableToType(PN->getIncomingValue(i), Ty, CTMap))
+      if (!ExpressionConvertableToType(PN->getIncomingValue(i), Ty, CTMap, TD))
         return false;
     break;
   }
 
   case Instruction::Malloc:
-    if (!MallocConvertableToType(cast<MallocInst>(I), Ty, CTMap))
+    if (!MallocConvertableToType(cast<MallocInst>(I), Ty, CTMap, TD))
       return false;
     break;
 
@@ -258,10 +261,10 @@
       // the appropriate size... if so, allow it.
       //
       std::vector<Value*> Indices;
-      const Type *ElTy = ConvertableToGEP(PTy, I->getOperand(1), Indices);
+      const Type *ElTy = ConvertableToGEP(PTy, I->getOperand(1), Indices, TD);
       if (ElTy == PVTy) {
         if (!ExpressionConvertableToType(I->getOperand(0),
-                                         PointerType::get(ElTy), CTMap))
+                                         PointerType::get(ElTy), CTMap, TD))
           return false;  // Can't continue, ExConToTy might have polluted set!
         break;
       }
@@ -278,7 +281,7 @@
         TD.getTypeSize(PTy->getElementType()) == 
         TD.getTypeSize(GEP->getType()->getElementType())) {
       const PointerType *NewSrcTy = PointerType::get(PVTy);
-      if (!ExpressionConvertableToType(I->getOperand(0), NewSrcTy, CTMap))
+      if (!ExpressionConvertableToType(I->getOperand(0), NewSrcTy, CTMap, TD))
         return false;
       break;
     }
@@ -300,7 +303,7 @@
     const FunctionType *NewTy =
       FunctionType::get(Ty, ArgTys, FT->isVarArg());
     if (!ExpressionConvertableToType(I->getOperand(0),
-                                     PointerType::get(NewTy), CTMap))
+                                     PointerType::get(NewTy), CTMap, TD))
       return false;
     break;
   }
@@ -313,14 +316,15 @@
   // recursion.
   //
   for (Value::use_iterator It = I->use_begin(), E = I->use_end(); It != E; ++It)
-    if (!OperandConvertableToType(*It, I, Ty, CTMap))
+    if (!OperandConvertableToType(*It, I, Ty, CTMap, TD))
       return false;
 
   return true;
 }
 
 
-Value *ConvertExpressionToType(Value *V, const Type *Ty, ValueMapCache &VMC) {
+Value *ConvertExpressionToType(Value *V, const Type *Ty, ValueMapCache &VMC,
+                               const TargetData &TD) {
   if (V->getType() == Ty) return V;  // Already where we need to be?
 
   ValueMapCache::ExprMapTy::iterator VMCI = VMC.ExprMap.find(V);
@@ -373,8 +377,8 @@
                                  Dummy, Dummy, Name);
     VMC.ExprMap[I] = Res;   // Add node to expression eagerly
 
-    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), Ty, VMC));
-    Res->setOperand(1, ConvertExpressionToType(I->getOperand(1), Ty, VMC));
+    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), Ty, VMC, TD));
+    Res->setOperand(1, ConvertExpressionToType(I->getOperand(1), Ty, VMC, TD));
     break;
 
   case Instruction::Shl:
@@ -382,7 +386,7 @@
     Res = new ShiftInst(cast<ShiftInst>(I)->getOpcode(), Dummy,
                         I->getOperand(1), Name);
     VMC.ExprMap[I] = Res;
-    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), Ty, VMC));
+    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), Ty, VMC, TD));
     break;
 
   case Instruction::Load: {
@@ -391,7 +395,7 @@
     Res = new LoadInst(Constant::getNullValue(PointerType::get(Ty)), Name);
     VMC.ExprMap[I] = Res;
     Res->setOperand(0, ConvertExpressionToType(LI->getPointerOperand(),
-                                               PointerType::get(Ty), VMC));
+                                               PointerType::get(Ty), VMC, TD));
     assert(Res->getOperand(0)->getType() == PointerType::get(Ty));
     assert(Ty == Res->getType());
     assert(Res->getType()->isFirstClassType() && "Load of structure or array!");
@@ -408,7 +412,7 @@
       Value *OldVal = OldPN->getIncomingValue(0);
       ValueHandle OldValHandle(VMC, OldVal);
       OldPN->removeIncomingValue(BB, false);
-      Value *V = ConvertExpressionToType(OldVal, Ty, VMC);
+      Value *V = ConvertExpressionToType(OldVal, Ty, VMC, TD);
       NewPN->addIncoming(V, BB);
     }
     Res = NewPN;
@@ -416,7 +420,7 @@
   }
 
   case Instruction::Malloc: {
-    Res = ConvertMallocToType(cast<MallocInst>(I), Ty, Name, VMC);
+    Res = ConvertMallocToType(cast<MallocInst>(I), Ty, Name, VMC, TD);
     break;
   }
 
@@ -468,14 +472,14 @@
       //
       std::vector<Value*> Indices;
       const Type *ElTy = ConvertableToGEP(NewSrcTy, I->getOperand(1),
-                                          Indices, &It);
+                                          Indices, TD, &It);
       if (ElTy) {        
         assert(ElTy == PVTy && "Internal error, setup wrong!");
         Res = new GetElementPtrInst(Constant::getNullValue(NewSrcTy),
                                     Indices, Name);
         VMC.ExprMap[I] = Res;
         Res->setOperand(0, ConvertExpressionToType(I->getOperand(0),
-                                                   NewSrcTy, VMC));
+                                                   NewSrcTy, VMC, TD));
       }
     }
 
@@ -491,7 +495,7 @@
                                   Indices, Name);
       VMC.ExprMap[I] = Res;
       Res->setOperand(0, ConvertExpressionToType(I->getOperand(0),
-                                                 NewSrcTy, VMC));
+                                                 NewSrcTy, VMC, TD));
     }
 
 
@@ -519,7 +523,7 @@
                        std::vector<Value*>(I->op_begin()+1, I->op_end()),
                        Name);
     VMC.ExprMap[I] = Res;
-    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), NewPTy, VMC));
+    Res->setOperand(0, ConvertExpressionToType(I->getOperand(0),NewPTy,VMC,TD));
     break;
   }
   default:
@@ -541,7 +545,7 @@
   unsigned NumUses = I->use_size();
   for (unsigned It = 0; It < NumUses; ) {
     unsigned OldSize = NumUses;
-    ConvertOperandToType(*(I->use_begin()+It), I, Res, VMC);
+    ConvertOperandToType(*(I->use_begin()+It), I, Res, VMC, TD);
     NumUses = I->use_size();
     if (NumUses == OldSize) ++It;
   }
@@ -556,7 +560,8 @@
 
 // ValueConvertableToType - Return true if it is possible
 bool ValueConvertableToType(Value *V, const Type *Ty,
-                             ValueTypeCache &ConvertedTypes) {
+                             ValueTypeCache &ConvertedTypes,
+                            const TargetData &TD) {
   ValueTypeCache::iterator I = ConvertedTypes.find(V);
   if (I != ConvertedTypes.end()) return I->second == Ty;
   ConvertedTypes[V] = Ty;
@@ -566,7 +571,7 @@
   //
   if (V->getType() != Ty) {
     for (Value::use_iterator I = V->use_begin(), E = V->use_end(); I != E; ++I)
-      if (!OperandConvertableToType(*I, V, Ty, ConvertedTypes))
+      if (!OperandConvertableToType(*I, V, Ty, ConvertedTypes, TD))
         return false;
   }
 
@@ -584,7 +589,8 @@
 // the expression, not the static current type.
 //
 static bool OperandConvertableToType(User *U, Value *V, const Type *Ty,
-                                     ValueTypeCache &CTMap) {
+                                     ValueTypeCache &CTMap,
+                                     const TargetData &TD) {
   //  if (V->getType() == Ty) return true;   // Operand already the right type?
 
   // Expression type must be holdable in a register.
@@ -630,11 +636,11 @@
     if (isa<PointerType>(Ty)) {
       Value *IndexVal = I->getOperand(V == I->getOperand(0) ? 1 : 0);
       std::vector<Value*> Indices;
-      if (const Type *ETy = ConvertableToGEP(Ty, IndexVal, Indices)) {
+      if (const Type *ETy = ConvertableToGEP(Ty, IndexVal, Indices, TD)) {
         const Type *RetTy = PointerType::get(ETy);
 
         // Only successful if we can convert this type to the required type
-        if (ValueConvertableToType(I, RetTy, CTMap)) {
+        if (ValueConvertableToType(I, RetTy, CTMap, TD)) {
           CTMap[I] = RetTy;
           return true;
         }
@@ -648,13 +654,13 @@
     if (!Ty->isInteger() && !Ty->isFloatingPoint()) return false;
 
     Value *OtherOp = I->getOperand((V == I->getOperand(0)) ? 1 : 0);
-    return ValueConvertableToType(I, Ty, CTMap) &&
-           ExpressionConvertableToType(OtherOp, Ty, CTMap);
+    return ValueConvertableToType(I, Ty, CTMap, TD) &&
+           ExpressionConvertableToType(OtherOp, Ty, CTMap, TD);
   }
   case Instruction::SetEQ:
   case Instruction::SetNE: {
     Value *OtherOp = I->getOperand((V == I->getOperand(0)) ? 1 : 0);
-    return ExpressionConvertableToType(OtherOp, Ty, CTMap);
+    return ExpressionConvertableToType(OtherOp, Ty, CTMap, TD);
   }
   case Instruction::Shr:
     if (Ty->isSigned() != V->getType()->isSigned()) return false;
@@ -662,7 +668,7 @@
   case Instruction::Shl:
     if (I->getOperand(1) == V) return false;  // Cannot change shift amount type
     if (!Ty->isInteger()) return false;
-    return ValueConvertableToType(I, Ty, CTMap);
+    return ValueConvertableToType(I, Ty, CTMap, TD);
 
   case Instruction::Free:
     assert(I->getOperand(0) == V);
@@ -681,7 +687,7 @@
       if (const CompositeType *CT = dyn_cast<CompositeType>(LoadedTy)) {
         unsigned Offset = 0;     // No offset, get first leaf.
         std::vector<Value*> Indices;  // Discarded...
-        LoadedTy = getStructOffsetType(CT, Offset, Indices, false);
+        LoadedTy = getStructOffsetType(CT, Offset, Indices, TD, false);
         assert(Offset == 0 && "Offset changed from zero???");
       }
 
@@ -691,7 +697,7 @@
       if (TD.getTypeSize(LoadedTy) != TD.getTypeSize(LI->getType()))
         return false;
 
-      return ValueConvertableToType(LI, LoadedTy, CTMap);
+      return ValueConvertableToType(LI, LoadedTy, CTMap, TD);
     }
     return false;
 
@@ -722,7 +728,7 @@
         if (const StructType *SElTy = dyn_cast<StructType>(ElTy)) {
           unsigned Offset = 0;
           std::vector<Value*> Indices;
-          ElTy = getStructOffsetType(ElTy, Offset, Indices, false);
+          ElTy = getStructOffsetType(ElTy, Offset, Indices, TD, false);
           assert(Offset == 0 && "Offset changed!");
           if (ElTy == 0)    // Element at offset zero in struct doesn't exist!
             return false;   // Can only happen for {}*
@@ -738,7 +744,7 @@
       // Can convert the store if we can convert the pointer operand to match
       // the new  value type...
       return ExpressionConvertableToType(I->getOperand(1), PointerType::get(Ty),
-                                         CTMap);
+                                         CTMap, TD);
     } else if (const PointerType *PT = dyn_cast<PointerType>(Ty)) {
       const Type *ElTy = PT->getElementType();
       assert(V == I->getOperand(1));
@@ -749,7 +755,7 @@
         //
         unsigned Offset = 0;
         std::vector<Value*> Indices;
-        ElTy = getStructOffsetType(ElTy, Offset, Indices, false);
+        ElTy = getStructOffsetType(ElTy, Offset, Indices, TD, false);
         assert(Offset == 0 && "Offset changed!");
         if (ElTy == 0)    // Element at offset zero in struct doesn't exist!
           return false;   // Can only happen for {}*
@@ -761,7 +767,7 @@
         return false;
 
       // Can convert store if the incoming value is convertable...
-      return ExpressionConvertableToType(I->getOperand(0), ElTy, CTMap);
+      return ExpressionConvertableToType(I->getOperand(0), ElTy, CTMap, TD);
     }
     return false;
   }
@@ -795,20 +801,20 @@
       // be converted to the appropriate size... if so, allow it.
       //
       std::vector<Value*> Indices;
-      const Type *ElTy = ConvertableToGEP(Ty, Index, Indices);
+      const Type *ElTy = ConvertableToGEP(Ty, Index, Indices, TD);
       delete TempScale;   // Free our temporary multiply if we made it
 
       if (ElTy == 0) return false;  // Cannot make conversion...
-      return ValueConvertableToType(I, PointerType::get(ElTy), CTMap);
+      return ValueConvertableToType(I, PointerType::get(ElTy), CTMap, TD);
     }
     return false;
 
   case Instruction::PHINode: {
     PHINode *PN = cast<PHINode>(I);
     for (unsigned i = 0; i < PN->getNumIncomingValues(); ++i)
-      if (!ExpressionConvertableToType(PN->getIncomingValue(i), Ty, CTMap))
+      if (!ExpressionConvertableToType(PN->getIncomingValue(i), Ty, CTMap, TD))
         return false;
-    return ValueConvertableToType(PN, Ty, CTMap);
+    return ValueConvertableToType(PN, Ty, CTMap, TD);
   }
 
   case Instruction::Call: {
@@ -859,7 +865,7 @@
       // converted.  We succeed if we can change the return type if
       // neccesary...
       //
-      return ValueConvertableToType(I, FTy->getReturnType(), CTMap);
+      return ValueConvertableToType(I, FTy->getReturnType(), CTMap, TD);
     }
     
     const PointerType *MPtr = cast<PointerType>(I->getOperand(0)->getType());
@@ -879,13 +885,14 @@
 }
 
 
-void ConvertValueToNewType(Value *V, Value *NewVal, ValueMapCache &VMC) {
+void ConvertValueToNewType(Value *V, Value *NewVal, ValueMapCache &VMC,
+                           const TargetData &TD) {
   ValueHandle VH(VMC, V);
 
   unsigned NumUses = V->use_size();
   for (unsigned It = 0; It < NumUses; ) {
     unsigned OldSize = NumUses;
-    ConvertOperandToType(*(V->use_begin()+It), V, NewVal, VMC);
+    ConvertOperandToType(*(V->use_begin()+It), V, NewVal, VMC, TD);
     NumUses = V->use_size();
     if (NumUses == OldSize) ++It;
   }
@@ -894,7 +901,7 @@
 
 
 static void ConvertOperandToType(User *U, Value *OldVal, Value *NewVal,
-                                 ValueMapCache &VMC) {
+                                 ValueMapCache &VMC, const TargetData &TD) {
   if (isa<ValueHandle>(U)) return;  // Valuehandles don't let go of operands...
 
   if (VMC.OperandsMapped.count(U)) return;
@@ -944,7 +951,7 @@
       std::vector<Value*> Indices;
       BasicBlock::iterator It = I;
 
-      if (const Type *ETy = ConvertableToGEP(NewTy, IndexVal, Indices, &It)) {
+      if (const Type *ETy = ConvertableToGEP(NewTy, IndexVal, Indices, TD,&It)){
         // If successful, convert the add to a GEP
         //const Type *RetTy = PointerType::get(ETy);
         // First operand is actually the given pointer...
@@ -965,7 +972,7 @@
 
     unsigned OtherIdx = (OldVal == I->getOperand(0)) ? 1 : 0;
     Value *OtherOp    = I->getOperand(OtherIdx);
-    Value *NewOther   = ConvertExpressionToType(OtherOp, NewTy, VMC);
+    Value *NewOther   = ConvertExpressionToType(OtherOp, NewTy, VMC, TD);
 
     Res->setOperand(OtherIdx, NewOther);
     Res->setOperand(!OtherIdx, NewVal);
@@ -996,7 +1003,7 @@
       Indices.push_back(ConstantSInt::get(Type::LongTy, 0));
 
       unsigned Offset = 0;   // No offset, get first leaf.
-      LoadedTy = getStructOffsetType(CT, Offset, Indices, false);
+      LoadedTy = getStructOffsetType(CT, Offset, Indices, TD, false);
       assert(LoadedTy->isFirstClassType());
 
       if (Indices.size() != 1) {     // Do not generate load X, 0
@@ -1032,7 +1039,7 @@
           Indices.push_back(Constant::getNullValue(Type::LongTy));
 
           unsigned Offset = 0;
-          const Type *Ty = getStructOffsetType(ElTy, Offset, Indices, false);
+          const Type *Ty = getStructOffsetType(ElTy, Offset, Indices, TD,false);
           assert(Offset == 0 && "Offset changed!");
           assert(NewTy == Ty && "Did not convert to correct type!");
 
@@ -1049,7 +1056,7 @@
         Res = new StoreInst(NewVal, Constant::getNullValue(NewPT));
         VMC.ExprMap[I] = Res;
         Res->setOperand(1, ConvertExpressionToType(I->getOperand(1),
-                                                   NewPT, VMC));
+                                                   NewPT, VMC, TD));
       }
     } else {                           // Replace the source pointer
       const Type *ValTy = cast<PointerType>(NewTy)->getElementType();
@@ -1061,7 +1068,7 @@
         Indices.push_back(Constant::getNullValue(Type::LongTy));
 
         unsigned Offset = 0;
-        ValTy = getStructOffsetType(ValTy, Offset, Indices, false);
+        ValTy = getStructOffsetType(ValTy, Offset, Indices, TD, false);
 
         assert(Offset == 0 && ValTy);
 
@@ -1072,7 +1079,8 @@
 
       Res = new StoreInst(Constant::getNullValue(ValTy), SrcPtr);
       VMC.ExprMap[I] = Res;
-      Res->setOperand(0, ConvertExpressionToType(I->getOperand(0), ValTy, VMC));
+      Res->setOperand(0, ConvertExpressionToType(I->getOperand(0),
+                                                 ValTy, VMC, TD));
     }
     break;
   }
@@ -1097,7 +1105,7 @@
     // Perform the conversion now...
     //
     std::vector<Value*> Indices;
-    const Type *ElTy = ConvertableToGEP(NewVal->getType(), Index, Indices, &It);
+    const Type *ElTy = ConvertableToGEP(NewVal->getType(),Index,Indices,TD,&It);
     assert(ElTy != 0 && "GEP Conversion Failure!");
     Res = new GetElementPtrInst(NewVal, Indices, Name);
     assert(Res->getType() == PointerType::get(ElTy) &&
@@ -1115,7 +1123,7 @@
       //
       std::vector<Value*> Indices;
       const Type *ElTy = ConvertableToGEP(NewVal->getType(), I->getOperand(1),
-                                          Indices, &It);
+                                          Indices, TD, &It);
       assert(ElTy != 0 && "GEP Conversion Failure!");
       
       Res = new GetElementPtrInst(NewVal, Indices, Name);
@@ -1140,7 +1148,7 @@
       BasicBlock *BB = OldPN->getIncomingBlock(0);
       Value *OldVal = OldPN->getIncomingValue(0);
       OldPN->removeIncomingValue(BB, false);
-      Value *V = ConvertExpressionToType(OldVal, NewTy, VMC);
+      Value *V = ConvertExpressionToType(OldVal, NewTy, VMC, TD);
       NewPN->addIncoming(V, BB);
     }
     Res = NewPN;
@@ -1211,7 +1219,7 @@
   VMC.ExprMap[I] = Res;
 
   if (I->getType() != Res->getType())
-    ConvertValueToNewType(I, Res, VMC);
+    ConvertValueToNewType(I, Res, VMC, TD);
   else {
     for (unsigned It = 0; It < I->use_size(); ) {
       User *Use = *(I->use_begin()+It);


Index: llvm/lib/Transforms/LevelRaise.cpp
diff -u llvm/lib/Transforms/LevelRaise.cpp:1.76 llvm/lib/Transforms/LevelRaise.cpp:1.77
--- llvm/lib/Transforms/LevelRaise.cpp:1.76	Wed Nov 20 12:34:35 2002
+++ llvm/lib/Transforms/LevelRaise.cpp	Thu Apr 24 13:25:27 2003
@@ -62,6 +62,28 @@
   do { PRINT_PEEPHOLE(ID, 0, I1); PRINT_PEEPHOLE(ID, 1, I2); \
        PRINT_PEEPHOLE(ID, 2, I3); PRINT_PEEPHOLE(ID, 3, I4); } while (0)
 
+namespace {
+  struct RPR : public FunctionPass {
+    virtual bool runOnFunction(Function &F);
+
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+      AU.setPreservesCFG();
+      AU.addRequired<TargetData>();
+    }
+
+  private:
+    bool DoRaisePass(Function &F);
+    bool PeepholeOptimize(BasicBlock *BB, BasicBlock::iterator &BI);
+  };
+
+  RegisterOpt<RPR> X("raise", "Raise Pointer References");
+}
+
+Pass *createRaisePointerReferencesPass() {
+  return new RPR();
+}
+
+
 
 // isReinterpretingCast - Return true if the cast instruction specified will
 // cause the operand to be "reinterpreted".  A value is reinterpreted if the
@@ -80,7 +102,8 @@
 //       %t2 = cast <eltype> * %t3 to {<...>}*
 //
 static bool HandleCastToPointer(BasicBlock::iterator BI,
-                                const PointerType *DestPTy) {
+                                const PointerType *DestPTy,
+                                const TargetData &TD) {
   CastInst &CI = cast<CastInst>(*BI);
   if (CI.use_empty()) return false;
 
@@ -101,7 +124,7 @@
 
   std::vector<Value*> Indices;
   Value *Src = CI.getOperand(0);
-  const Type *Result = ConvertableToGEP(DestPTy, Src, Indices, &BI);
+  const Type *Result = ConvertableToGEP(DestPTy, Src, Indices, TD, &BI);
   if (Result == 0) return false;  // Not convertable...
 
   PRINT_PEEPHOLE2("cast-add-to-gep:in", Src, CI);
@@ -155,7 +178,8 @@
 //       %t2 = cast <eltype> * %t3 to {<...>}*
 //
 static bool PeepholeOptimizeAddCast(BasicBlock *BB, BasicBlock::iterator &BI,
-                                    Value *AddOp1, CastInst *AddOp2) {
+                                    Value *AddOp1, CastInst *AddOp2,
+                                    const TargetData &TD) {
   const CompositeType *CompTy;
   Value *OffsetVal = AddOp2->getOperand(0);
   Value *SrcPtr = 0;  // Of type pointer to struct...
@@ -172,7 +196,7 @@
     return false;
 
   std::vector<Value*> Indices;
-  if (!ConvertableToGEP(SrcPtr->getType(), OffsetVal, Indices, &BI))
+  if (!ConvertableToGEP(SrcPtr->getType(), OffsetVal, Indices, TD, &BI))
     return false;  // Not convertable... perhaps next time
 
   if (getPointedToComposite(AddOp1->getType())) {  // case 1
@@ -190,8 +214,9 @@
   return true;
 }
 
-static bool PeepholeOptimize(BasicBlock *BB, BasicBlock::iterator &BI) {
+bool RPR::PeepholeOptimize(BasicBlock *BB, BasicBlock::iterator &BI) {
   Instruction *I = BI;
+  const TargetData &TD = getAnalysis<TargetData>();
 
   if (CastInst *CI = dyn_cast<CastInst>(I)) {
     Value       *Src    = CI->getOperand(0);
@@ -230,13 +255,13 @@
       // destination type of the cast...
       //
       ConvertedTypes[CI] = CI->getType();  // Make sure the cast doesn't change
-      if (ExpressionConvertableToType(Src, DestTy, ConvertedTypes)) {
+      if (ExpressionConvertableToType(Src, DestTy, ConvertedTypes, TD)) {
         PRINT_PEEPHOLE3("CAST-SRC-EXPR-CONV:in ", Src, CI, BB->getParent());
           
         DEBUG(cerr << "\nCONVERTING SRC EXPR TYPE:\n");
         { // ValueMap must be destroyed before function verified!
           ValueMapCache ValueMap;
-          Value *E = ConvertExpressionToType(Src, DestTy, ValueMap);
+          Value *E = ConvertExpressionToType(Src, DestTy, ValueMap, TD);
 
           if (Constant *CPV = dyn_cast<Constant>(E))
             CI->replaceAllUsesWith(CPV);
@@ -258,13 +283,13 @@
       ConvertedTypes.clear();
       // Make sure the source doesn't change type
       ConvertedTypes[Src] = Src->getType();
-      if (ValueConvertableToType(CI, Src->getType(), ConvertedTypes)) {
+      if (ValueConvertableToType(CI, Src->getType(), ConvertedTypes, TD)) {
         PRINT_PEEPHOLE3("CAST-DEST-EXPR-CONV:in ", Src, CI, BB->getParent());
 
         DEBUG(cerr << "\nCONVERTING EXPR TYPE:\n");
         { // ValueMap must be destroyed before function verified!
           ValueMapCache ValueMap;
-          ConvertValueToNewType(CI, Src, ValueMap);  // This will delete CI!
+          ConvertValueToNewType(CI, Src, ValueMap, TD);  // This will delete CI!
         }
 
         PRINT_PEEPHOLE1("CAST-DEST-EXPR-CONV:out", Src);
@@ -283,7 +308,7 @@
     // so, convert the add into a getelementptr instruction...
     //
     if (const PointerType *DestPTy = dyn_cast<PointerType>(DestTy)) {
-      if (HandleCastToPointer(BI, DestPTy)) {
+      if (HandleCastToPointer(BI, DestPTy, TD)) {
         BI = BB->begin();  // Rescan basic block.  BI might be invalidated.
         ++NumGEPInstFormed;
         return true;
@@ -454,7 +479,7 @@
              isa<CastInst>(I->getOperand(1))) {
 
     if (PeepholeOptimizeAddCast(BB, BI, I->getOperand(0),
-                                cast<CastInst>(I->getOperand(1)))) {
+                                cast<CastInst>(I->getOperand(1)), TD)) {
       ++NumGEPInstFormed;
       return true;
     }
@@ -500,7 +525,7 @@
 
 
 
-static bool DoRaisePass(Function &F) {
+bool RPR::DoRaisePass(Function &F) {
   bool Changed = false;
   for (Function::iterator BB = F.begin(), BBE = F.end(); BB != BBE; ++BB)
     for (BasicBlock::iterator BI = BB->begin(); BI != BB->end();) {
@@ -520,17 +545,14 @@
 }
 
 
-// RaisePointerReferences::doit - Raise a function representation to a higher
-// level.
-//
-static bool doRPR(Function &F) {
+// runOnFunction - Raise a function representation to a higher level.
+bool RPR::runOnFunction(Function &F) {
   DEBUG(cerr << "\n\n\nStarting to work on Function '" << F.getName() << "'\n");
 
   // Insert casts for all incoming pointer pointer values that are treated as
   // arrays...
   //
   bool Changed = false, LocalChange;
-  
 
   // If the StartInst option was specified, then Peephole optimize that
   // instruction first if it occurs in this function.
@@ -559,24 +581,3 @@
 
   return Changed;
 }
-
-namespace {
-  struct RaisePointerReferences : public FunctionPass {
-
-    // FIXME: constructor should save and use target data here!!
-    RaisePointerReferences(const TargetData &TD) {}
-
-    virtual bool runOnFunction(Function &F) { return doRPR(F); }
-
-    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
-      AU.setPreservesCFG();
-    }
-  };
-}
-
-Pass *createRaisePointerReferencesPass(const TargetData &TD) {
-  return new RaisePointerReferences(TD);
-}
-
-static RegisterOpt<RaisePointerReferences>
-X("raise", "Raise Pointer References", createRaisePointerReferencesPass);


Index: llvm/lib/Transforms/TransformInternals.cpp
diff -u llvm/lib/Transforms/TransformInternals.cpp:1.32 llvm/lib/Transforms/TransformInternals.cpp:1.33
--- llvm/lib/Transforms/TransformInternals.cpp:1.32	Wed Jan 22 20:39:10 2003
+++ llvm/lib/Transforms/TransformInternals.cpp	Thu Apr 24 13:25:27 2003
@@ -11,15 +11,9 @@
 #include "llvm/Function.h"
 #include "llvm/iOther.h"
 
-// TargetData Hack: Eventually we will have annotations given to us by the
-// backend so that we know stuff about type size and alignments.  For now
-// though, just use this, because it happens to match the model that GCC uses.
-//
-const TargetData TD("LevelRaise: Should be GCC though!");
-
-
 static const Type *getStructOffsetStep(const StructType *STy, uint64_t &Offset,
-                                       std::vector<Value*> &Indices) {
+                                       std::vector<Value*> &Indices,
+                                       const TargetData &TD) {
   assert(Offset < TD.getTypeSize(STy) && "Offset not in composite!");
   const StructLayout *SL = TD.getStructLayout(STy);
 
@@ -52,7 +46,7 @@
 //
 const Type *getStructOffsetType(const Type *Ty, unsigned &Offset,
                                 std::vector<Value*> &Indices,
-                                bool StopEarly) {
+                                const TargetData &TD, bool StopEarly) {
   if (Offset == 0 && StopEarly && !Indices.empty())
     return Ty;    // Return the leaf type
 
@@ -60,7 +54,7 @@
   const Type *NextType;
   if (const StructType *STy = dyn_cast<StructType>(Ty)) {
     ThisOffset = Offset;
-    NextType = getStructOffsetStep(STy, ThisOffset, Indices);
+    NextType = getStructOffsetStep(STy, ThisOffset, Indices, TD);
   } else if (const ArrayType *ATy = dyn_cast<ArrayType>(Ty)) {
     assert(Offset < TD.getTypeSize(ATy) && "Offset not in composite!");
 
@@ -75,7 +69,7 @@
 
   unsigned SubOffs = Offset - ThisOffset;
   const Type *LeafTy = getStructOffsetType(NextType, SubOffs,
-                                           Indices, StopEarly);
+                                           Indices, TD, StopEarly);
   Offset = ThisOffset + SubOffs;
   return LeafTy;
 }
@@ -87,6 +81,7 @@
 //
 const Type *ConvertableToGEP(const Type *Ty, Value *OffsetVal,
                              std::vector<Value*> &Indices,
+                             const TargetData &TD,
                              BasicBlock::iterator *BI) {
   const CompositeType *CompTy = dyn_cast<CompositeType>(Ty);
   if (CompTy == 0) return 0;
@@ -116,7 +111,7 @@
     if (const StructType *StructTy = dyn_cast<StructType>(CompTy)) {
       // Step into the appropriate element of the structure...
       uint64_t ActualOffset = (Offset < 0) ? 0 : (uint64_t)Offset;
-      NextTy = getStructOffsetStep(StructTy, ActualOffset, Indices);
+      NextTy = getStructOffsetStep(StructTy, ActualOffset, Indices, TD);
       Offset -= ActualOffset;
     } else {
       const Type *ElTy = cast<SequentialType>(CompTy)->getElementType();


Index: llvm/lib/Transforms/TransformInternals.h
diff -u llvm/lib/Transforms/TransformInternals.h:1.16 llvm/lib/Transforms/TransformInternals.h:1.17
--- llvm/lib/Transforms/TransformInternals.h:1.16	Mon Sep 16 13:32:33 2002
+++ llvm/lib/Transforms/TransformInternals.h	Thu Apr 24 13:25:27 2003
@@ -15,14 +15,6 @@
 #include <map>
 #include <set>
 
-// TargetData Hack: Eventually we will have annotations given to us by the
-// backend so that we know stuff about type size and alignments.  For now
-// though, just use this, because it happens to match the model that GCC uses.
-//
-// FIXME: This should use annotations
-//
-extern const TargetData TD;
-
 static inline int64_t getConstantValue(const ConstantInt *CPI) {
   if (const ConstantSInt *CSI = dyn_cast<ConstantSInt>(CPI))
     return CSI->getValue();
@@ -49,6 +41,7 @@
 //
 const Type *ConvertableToGEP(const Type *Ty, Value *V,
                              std::vector<Value*> &Indices,
+                             const TargetData &TD,
                              BasicBlock::iterator *BI = 0);
 
 
@@ -112,14 +105,18 @@
 };
 
 
-bool ExpressionConvertableToType(Value *V, const Type *Ty, ValueTypeCache &Map);
-Value *ConvertExpressionToType(Value *V, const Type *Ty, ValueMapCache &VMC);
+bool ExpressionConvertableToType(Value *V, const Type *Ty, ValueTypeCache &Map,
+                                 const TargetData &TD);
+Value *ConvertExpressionToType(Value *V, const Type *Ty, ValueMapCache &VMC,
+                               const TargetData &TD);
 
 // ValueConvertableToType - Return true if it is possible
 bool ValueConvertableToType(Value *V, const Type *Ty,
-                            ValueTypeCache &ConvertedTypes);
+                            ValueTypeCache &ConvertedTypes,
+                            const TargetData &TD);
 
-void ConvertValueToNewType(Value *V, Value *NewVal, ValueMapCache &VMC);
+void ConvertValueToNewType(Value *V, Value *NewVal, ValueMapCache &VMC,
+                           const TargetData &TD);
 
 
 // getStructOffsetType - Return a vector of offsets that are to be used to index
@@ -135,6 +132,6 @@
 //
 const Type *getStructOffsetType(const Type *Ty, unsigned &Offset,
                                 std::vector<Value*> &Offsets,
-                                bool StopEarly = true);
+                                const TargetData &TD, bool StopEarly = true);
 
 #endif




From lattner at cs.uiuc.edu  Thu Apr 24 13:26:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:26:06 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Transforms/RaisePointerReferences.h
Message-ID: <200304241825.NAA26254@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Transforms:

RaisePointerReferences.h updated: 1.4 -> 1.5

---
Log message:

LevelRaise now gets TD from passmanager


---
Diffs of the changes:

Index: llvm/include/llvm/Transforms/RaisePointerReferences.h
diff -u llvm/include/llvm/Transforms/RaisePointerReferences.h:1.4 llvm/include/llvm/Transforms/RaisePointerReferences.h:1.5
--- llvm/include/llvm/Transforms/RaisePointerReferences.h:1.4	Tue Jul 23 12:58:42 2002
+++ llvm/include/llvm/Transforms/RaisePointerReferences.h	Thu Apr 24 13:25:42 2003
@@ -10,12 +10,11 @@
 #define LLVM_TRANSFORMS_LEVELCHANGE_H
 
 class Pass;
-class TargetData;
 
 // RaisePointerReferences - Try to eliminate as many pointer arithmetic
 // expressions as possible, by converting expressions to use getelementptr and
 // friends.
 //
-Pass *createRaisePointerReferencesPass(const TargetData &TD);
+Pass *createRaisePointerReferencesPass();
 
 #endif




From lattner at cs.uiuc.edu  Thu Apr 24 13:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:27:01 2003
Subject: [llvm-commits] CVS: llvm/tools/gccas/gccas.cpp
Message-ID: <200304241826.NAA26269@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccas:

gccas.cpp updated: 1.59 -> 1.60

---
Log message:

LevelRaise now gets target data from passmanager


---
Diffs of the changes:

Index: llvm/tools/gccas/gccas.cpp
diff -u llvm/tools/gccas/gccas.cpp:1.59 llvm/tools/gccas/gccas.cpp:1.60
--- llvm/tools/gccas/gccas.cpp:1.59	Wed Apr 23 15:40:42 2003
+++ llvm/tools/gccas/gccas.cpp	Thu Apr 24 13:26:03 2003
@@ -23,9 +23,6 @@
 #include <fstream>
 
 namespace {
-  // FIXME: This should eventually be parameterized...
-  TargetData TD("gccas target");
-
   cl::opt<std::string>
   InputFilename(cl::Positional,cl::desc("<input llvm assembly>"),cl::init("-"));
 
@@ -74,7 +71,7 @@
   addPass(PM, createDeadInstEliminationPass());  // Remove Dead code/vars
   addPass(PM, createRaiseAllocationsPass());     // call %malloc -> malloc inst
   addPass(PM, createIndVarSimplifyPass());       // Simplify indvars
-  addPass(PM, createRaisePointerReferencesPass(TD));// Recover type information
+  addPass(PM, createRaisePointerReferencesPass());// Recover type information
   addPass(PM, createInstructionCombiningPass()); // Combine silly seq's
   addPass(PM, createPromoteMemoryToRegister());  // Promote alloca's to regs
   addPass(PM, createReassociatePass());          // Reassociate expressions
@@ -96,6 +93,9 @@
 
 int main(int argc, char **argv) {
   cl::ParseCommandLineOptions(argc, argv, " llvm .s -> .o assembler for GCC\n");
+
+  // FIXME: This should eventually be parameterized...
+  TargetData TD("gccas target");
 
   std::auto_ptr<Module> M;
   try {




From lattner at cs.uiuc.edu  Thu Apr 24 13:37:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:37:01 2003
Subject: [llvm-commits] CVS: llvm/lib/CodeGen/PreOpts/PreSelection.cpp
Message-ID: <200304241836.NAA28071@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/CodeGen/PreOpts:

PreSelection.cpp updated: 1.8 -> 1.9

---
Log message:

Trivial cleanup


---
Diffs of the changes:

Index: llvm/lib/CodeGen/PreOpts/PreSelection.cpp
diff -u llvm/lib/CodeGen/PreOpts/PreSelection.cpp:1.8 llvm/lib/CodeGen/PreOpts/PreSelection.cpp:1.9
--- llvm/lib/CodeGen/PreOpts/PreSelection.cpp:1.8	Wed Apr 16 15:28:33 2003
+++ llvm/lib/CodeGen/PreOpts/PreSelection.cpp	Thu Apr 24 13:35:51 2003
@@ -150,13 +150,13 @@
     void visitOneOperand(Instruction &I, Constant* CV, unsigned opNum,
                          Instruction& insertBefore); // iworks on one operand
   };
-}  // end anonymous namespace
 
+  // Register the pass...
+  RegisterOpt<PreSelection> X("preselect",
+                              "Specialize LLVM code for a target machine",
+                              createPreSelectionPass);
+}  // end anonymous namespace
 
-// Register the pass...
-static RegisterOpt<PreSelection> X("preselect",
-                                   "Specialize LLVM code for a target machine",
-                                   createPreSelectionPass);
 
 //------------------------------------------------------------------------------
 // Helper functions used by methods of class PreSelection




From lattner at cs.uiuc.edu  Thu Apr 24 13:37:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:37:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/OptimizerDriver.cpp
Message-ID: <200304241836.NAA28233@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

OptimizerDriver.cpp updated: 1.4 -> 1.5

---
Log message:

Remove support for "target data" pass ctors


---
Diffs of the changes:

Index: llvm/tools/bugpoint/OptimizerDriver.cpp
diff -u llvm/tools/bugpoint/OptimizerDriver.cpp:1.4 llvm/tools/bugpoint/OptimizerDriver.cpp:1.5
--- llvm/tools/bugpoint/OptimizerDriver.cpp:1.4	Thu Apr 24 12:02:17 2003
+++ llvm/tools/bugpoint/OptimizerDriver.cpp	Thu Apr 24 13:36:36 2003
@@ -65,9 +65,6 @@
   std::cout << getPassesString(PassesToRun) << "\n";
 }
 
-/// FIXME: This should be parameterizable!!
-static TargetData TD("bugpoint target");
-
 static void RunChild(Module *Program,const std::vector<const PassInfo*> &Passes,
                      const std::string &OutFilename) {
   std::ofstream OutFile(OutFilename.c_str());
@@ -80,8 +77,6 @@
   for (unsigned i = 0, e = Passes.size(); i != e; ++i) {
     if (Passes[i]->getNormalCtor())
       PM.add(Passes[i]->getNormalCtor()());
-    else if (Passes[i]->getDataCtor())
-      PM.add(Passes[i]->getDataCtor()(TD));    // Provide dummy target data...
     else
       std::cerr << "Cannot create pass yet: " << Passes[i]->getPassName()
                 << "\n";




From lattner at cs.uiuc.edu  Thu Apr 24 13:37:08 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:37:08 2003
Subject: [llvm-commits] CVS: llvm/tools/gccas/gccas.cpp
Message-ID: <200304241836.NAA28240@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccas:

gccas.cpp updated: 1.60 -> 1.61

---
Log message:

Remove support for "target data" pass ctors



---
Diffs of the changes:

Index: llvm/tools/gccas/gccas.cpp
diff -u llvm/tools/gccas/gccas.cpp:1.60 llvm/tools/gccas/gccas.cpp:1.61
--- llvm/tools/gccas/gccas.cpp:1.60	Thu Apr 24 13:26:03 2003
+++ llvm/tools/gccas/gccas.cpp	Thu Apr 24 13:36:37 2003
@@ -94,9 +94,6 @@
 int main(int argc, char **argv) {
   cl::ParseCommandLineOptions(argc, argv, " llvm .s -> .o assembler for GCC\n");
 
-  // FIXME: This should eventually be parameterized...
-  TargetData TD("gccas target");
-
   std::auto_ptr<Module> M;
   try {
     // Parse the file now...




From lattner at cs.uiuc.edu  Thu Apr 24 13:37:12 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:37:12 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304241836.NAA28248@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.65 -> 1.66

---
Log message:

Remove support for "target data" pass ctors


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.65 llvm/tools/llc/llc.cpp:1.66
--- llvm/tools/llc/llc.cpp:1.65	Tue Jan 14 16:57:15 2003
+++ llvm/tools/llc/llc.cpp	Thu Apr 24 13:36:39 2003
@@ -200,8 +200,6 @@
         { // handle other passes as normal optimization passes
           if (Opt->getNormalCtor())
             Passes.add(Opt->getNormalCtor()());
-          else if (Opt->getDataCtor())
-            Passes.add(Opt->getDataCtor()(Target.getTargetData()));
           else if (Opt->getTargetCtor())
             Passes.add(Opt->getTargetCtor()(Target));
           else




From lattner at cs.uiuc.edu  Thu Apr 24 13:37:15 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:37:15 2003
Subject: [llvm-commits] CVS: llvm/tools/opt/opt.cpp
Message-ID: <200304241836.NAA28262@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/opt:

opt.cpp updated: 1.79 -> 1.80

---
Log message:

Remove support for "target data" pass ctors



---
Diffs of the changes:

Index: llvm/tools/opt/opt.cpp
diff -u llvm/tools/opt/opt.cpp:1.79 llvm/tools/opt/opt.cpp:1.80
--- llvm/tools/opt/opt.cpp:1.79	Wed Apr 16 18:02:16 2003
+++ llvm/tools/opt/opt.cpp	Thu Apr 24 13:36:41 2003
@@ -68,9 +68,6 @@
   cl::ParseCommandLineOptions(argc, argv,
 			      " llvm .bc -> .bc modular optimizer\n");
 
-  // FIXME: The choice of target should be controllable on the command line.
-  TargetData TD("opt target");
-
   // Allocate a full target machine description only if necessary...
   // FIXME: The choice of target should be controllable on the command line.
   std::auto_ptr<TargetMachine> target;
@@ -122,8 +119,6 @@
     
     if (Opt->getNormalCtor())
       Passes.add(Opt->getNormalCtor()());
-    else if (Opt->getDataCtor())
-      Passes.add(Opt->getDataCtor()(TD));    // Provide dummy target data...
     else if (Opt->getTargetCtor()) {
 #if 0
       if (target.get() == NULL)




From lattner at cs.uiuc.edu  Thu Apr 24 13:42:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:42:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/PassNameParser.h
Message-ID: <200304241841.NAA29166@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

PassNameParser.h updated: 1.4 -> 1.5

---
Log message:

Remove support for "targetdata pass ctors"


---
Diffs of the changes:

Index: llvm/include/llvm/Support/PassNameParser.h
diff -u llvm/include/llvm/Support/PassNameParser.h:1.4 llvm/include/llvm/Support/PassNameParser.h:1.5
--- llvm/include/llvm/Support/PassNameParser.h:1.4	Fri Dec  6 22:41:22 2002
+++ llvm/include/llvm/Support/PassNameParser.h	Thu Apr 24 13:41:13 2003
@@ -48,8 +48,7 @@
     // Ignore non-selectable and non-constructible passes!  Ignore
     // non-optimizations.
     return P->getPassArgument() == 0 || *P->getPassArgument() == 0 ||
-          (P->getNormalCtor() == 0 && P->getDataCtor() == 0 &&
-           P->getTargetCtor() == 0) ||
+          (P->getNormalCtor() == 0 && P->getTargetCtor() == 0) ||
           ignorablePassImpl(P);
   }
 




From lattner at cs.uiuc.edu  Thu Apr 24 13:42:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 13:42:06 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/PassSupport.h
Message-ID: <200304241841.NAA29179@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

PassSupport.h updated: 1.10 -> 1.11

---
Log message:

Remove support for "targetdata pass ctors"


---
Diffs of the changes:

Index: llvm/include/llvm/PassSupport.h
diff -u llvm/include/llvm/PassSupport.h:1.10 llvm/include/llvm/PassSupport.h:1.11
--- llvm/include/llvm/PassSupport.h:1.10	Mon Feb 24 18:00:50 2003
+++ llvm/include/llvm/PassSupport.h	Thu Apr 24 13:41:30 2003
@@ -16,7 +16,6 @@
 
 // No need to include Pass.h, we are being included by it!
 
-class TargetData;
 class TargetMachine;
 
 //===---------------------------------------------------------------------------
@@ -33,7 +32,6 @@
   std::vector<const PassInfo*> ItfImpl;// Interfaces implemented by this pass
 
   Pass *(*NormalCtor)();               // No argument ctor
-  Pass *(*DataCtor)(const TargetData&);// Ctor taking const TargetData object...
   Pass *(*TargetCtor)(TargetMachine&);   // Ctor taking TargetMachine object...
 
 public:
@@ -50,10 +48,9 @@
   /// through RegisterPass.
   PassInfo(const char *name, const char *arg, const std::type_info &ti, 
            unsigned pt, Pass *(*normal)() = 0,
-           Pass *(*datactor)(const TargetData &) = 0,
            Pass *(*targetctor)(TargetMachine &) = 0)
     : PassName(name), PassArgument(arg), TypeInfo(ti), PassType(pt),
-      NormalCtor(normal), DataCtor(datactor), TargetCtor(targetctor)  {
+      NormalCtor(normal), TargetCtor(targetctor)  {
   }
 
   /// getPassName - Return the friendly name for the pass, never returns null
@@ -97,14 +94,6 @@
     return NormalCtor();
   }
 
-  /// getDataCtor - Return a pointer to a function that creates an instance of
-  /// the pass and returns it.  This returns a constructor for a version of the
-  /// pass that takes a TargetData object as a parameter.
-  ///
-  Pass *(*getDataCtor() const)(const TargetData &) {
-    return DataCtor;
-  }
-
   /// getTargetCtor - Return a pointer to a function that creates an instance of
   /// the pass and returns it.  This returns a constructor for a version of the
   /// pass that takes a TargetMachine object as a parameter.
@@ -171,8 +160,6 @@
 
 template<typename PassName>
 Pass *callDefaultCtor() { return new PassName(); }
-template<typename PassName>
-Pass *callTargetDataCtor(const TargetData &TD) { return new PassName(TD); }
 
 template<typename PassName>
 struct RegisterPass : public RegisterPassBase {
@@ -189,18 +176,11 @@
     registerPass(new PassInfo(Name, PassArg, typeid(PassName), PassTy, ctor));
   }
 
-  // Register Pass using TargetData constructor...
-  RegisterPass(const char *PassArg, const char *Name, unsigned PassTy,
-               Pass *(*datactor)(const TargetData &)) {
-    registerPass(new PassInfo(Name, PassArg, typeid(PassName), PassTy,
-                              0, datactor));
-  }
-
   // Register Pass using TargetMachine constructor...
   RegisterPass(const char *PassArg, const char *Name, unsigned PassTy,
                Pass *(*targetctor)(TargetMachine &)) {
     registerPass(new PassInfo(Name, PassArg, typeid(PassName), PassTy,
-                              0, 0, targetctor));
+                              0, targetctor));
   }
 
   // Generic constructor version that has an unknown ctor type...
@@ -229,20 +209,12 @@
                               PassInfo::Optimization, ctor));
   }
 
-  /// Register Pass using TargetData constructor...
-  ///
-  RegisterOpt(const char *PassArg, const char *Name,
-               Pass *(*datactor)(const TargetData &)) {
-    registerPass(new PassInfo(Name, PassArg, typeid(PassName),
-                              PassInfo::Optimization, 0, datactor));
-  }
-
   /// Register Pass using TargetMachine constructor...
   ///
   RegisterOpt(const char *PassArg, const char *Name,
                Pass *(*targetctor)(TargetMachine &)) {
     registerPass(new PassInfo(Name, PassArg, typeid(PassName),
-                              PassInfo::Optimization, 0, 0, targetctor));
+                              PassInfo::Optimization, 0, targetctor));
   }
 };
 
@@ -280,14 +252,6 @@
   RegisterLLC(const char *PassArg, const char *Name, Pass *(*ctor)()) {
     registerPass(new PassInfo(Name, PassArg, typeid(PassName),
                               PassInfo::LLC, ctor));
-  }
-
-  /// Register Pass using TargetData constructor...
-  ///
-  RegisterLLC(const char *PassArg, const char *Name,
-               Pass *(*datactor)(const TargetData &)) {
-    registerPass(new PassInfo(Name, PassArg, typeid(PassName),
-                              PassInfo::LLC, 0, datactor));
   }
 
   /// Register Pass using TargetMachine constructor...




From lattner at cs.uiuc.edu  Thu Apr 24 14:09:02 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:09:02 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/TargetData.h
Message-ID: <200304241908.OAA00360@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

TargetData.h updated: 1.13 -> 1.14

---
Log message:

Add new targetdata ctor to create a targetdata appropriate to the module


---
Diffs of the changes:

Index: llvm/include/llvm/Target/TargetData.h
diff -u llvm/include/llvm/Target/TargetData.h:1.13 llvm/include/llvm/Target/TargetData.h:1.14
--- llvm/include/llvm/Target/TargetData.h:1.13	Tue Jan 14 15:29:37 2003
+++ llvm/include/llvm/Target/TargetData.h	Thu Apr 24 14:08:45 2003
@@ -38,7 +38,7 @@
  
   static Annotation *TypeAnFactory(AnnotationID, const Annotable *, void *);
 public:
-  TargetData(const std::string &TargetName = "SparcV9",
+  TargetData(const std::string &TargetName = "Default Target",
              bool LittleEndian = false,
              unsigned char SubWordDataSize = 1,
              unsigned char IntRegSize = 8,
@@ -47,6 +47,7 @@
 	     unsigned char FloatAl = 4, unsigned char LongAl = 8, 
 	     unsigned char IntAl = 4, unsigned char ShortAl = 2,
 	     unsigned char ByteAl = 1);
+  TargetData(const std::string &ToolName, const Module *M);
   ~TargetData();  // Not virtual, do not subclass this class
 
   /// Target endianness...




From lattner at cs.uiuc.edu  Thu Apr 24 14:10:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:10:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetData.cpp
Message-ID: <200304241909.OAA00371@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetData.cpp updated: 1.27 -> 1.28

---
Log message:

add a new targetdata ctor to create a target data appropriate to the module


---
Diffs of the changes:

Index: llvm/lib/Target/TargetData.cpp
diff -u llvm/lib/Target/TargetData.cpp:1.27 llvm/lib/Target/TargetData.cpp:1.28
--- llvm/lib/Target/TargetData.cpp:1.27	Tue Feb 25 14:27:09 2003
+++ llvm/lib/Target/TargetData.cpp	Thu Apr 24 14:09:05 2003
@@ -11,6 +11,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Target/TargetData.h"
+#include "llvm/Module.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/Constants.h"
 
@@ -101,6 +102,23 @@
   IntAlignment     = IntAl;
   ShortAlignment   = ShortAl;
   ByteAlignment    = ByteAl;
+}
+
+TargetData::TargetData(const std::string &ToolName, const Module *M)
+  : AID(AnnotationManager::getID("TargetData::" + ToolName)) {
+  AnnotationManager::registerAnnotationFactory(AID, TypeAnFactory, this);
+
+  LittleEndian     = M->isLittleEndian();
+  SubWordDataSize  = 1;
+  IntegerRegSize   = 8;
+  PointerSize      = M->has32BitPointers() ? 32 : 64;
+  PointerAlignment = PointerSize;
+  DoubleAlignment  = 8;
+  FloatAlignment   = 4;
+  LongAlignment    = 8;
+  IntAlignment     = 4;
+  ShortAlignment   = 2;
+  ByteAlignment    = 1;
 }
 
 TargetData::~TargetData() {




From lattner at cs.uiuc.edu  Thu Apr 24 14:11:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:11:05 2003
Subject: [llvm-commits] CVS: llvm/tools/analyze/analyze.cpp
Message-ID: <200304241910.OAA00399@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/analyze:

analyze.cpp updated: 1.50 -> 1.51

---
Log message:

Make sure to create a target data that matches the Module's target properties.


---
Diffs of the changes:

Index: llvm/tools/analyze/analyze.cpp
diff -u llvm/tools/analyze/analyze.cpp:1.50 llvm/tools/analyze/analyze.cpp:1.51
--- llvm/tools/analyze/analyze.cpp:1.50	Mon Feb 24 14:07:54 2003
+++ llvm/tools/analyze/analyze.cpp	Thu Apr 24 14:10:05 2003
@@ -15,6 +15,7 @@
 #include "llvm/Bytecode/Reader.h"
 #include "llvm/Assembly/Parser.h"
 #include "llvm/Analysis/Verifier.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/Support/PassNameParser.h"
 #include "Support/Timer.h"
 #include <algorithm>
@@ -125,6 +126,9 @@
   // about to build...
   //
   PassManager Passes;
+
+  // Add an appropriate TargetData instance for this module...
+  Passes.add(new TargetData("analyze", CurMod));
 
   // Make sure the input LLVM is well formed.
   Passes.add(createVerifierPass());




From lattner at cs.uiuc.edu  Thu Apr 24 14:11:11 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:11:11 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/OptimizerDriver.cpp
Message-ID: <200304241910.OAA00406@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

OptimizerDriver.cpp updated: 1.5 -> 1.6

---
Log message:

Make sure to create a target data that matches the Module's target properties.



---
Diffs of the changes:

Index: llvm/tools/bugpoint/OptimizerDriver.cpp
diff -u llvm/tools/bugpoint/OptimizerDriver.cpp:1.5 llvm/tools/bugpoint/OptimizerDriver.cpp:1.6
--- llvm/tools/bugpoint/OptimizerDriver.cpp:1.5	Thu Apr 24 13:36:36 2003
+++ llvm/tools/bugpoint/OptimizerDriver.cpp	Thu Apr 24 14:10:06 2003
@@ -74,6 +74,9 @@
   }
 
   PassManager PM;
+  // Make sure that the appropriate target data is always used...
+  PM.add(new TargetData("bugpoint", Program));
+
   for (unsigned i = 0, e = Passes.size(); i != e; ++i) {
     if (Passes[i]->getNormalCtor())
       PM.add(Passes[i]->getNormalCtor()());




From lattner at cs.uiuc.edu  Thu Apr 24 14:11:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:11:14 2003
Subject: [llvm-commits] CVS: llvm/tools/gccas/gccas.cpp
Message-ID: <200304241910.OAA00413@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccas:

gccas.cpp updated: 1.61 -> 1.62

---
Log message:

Make sure to create a target data that matches the Module's target properties.


---
Diffs of the changes:

Index: llvm/tools/gccas/gccas.cpp
diff -u llvm/tools/gccas/gccas.cpp:1.61 llvm/tools/gccas/gccas.cpp:1.62
--- llvm/tools/gccas/gccas.cpp:1.61	Thu Apr 24 13:36:37 2003
+++ llvm/tools/gccas/gccas.cpp	Thu Apr 24 14:10:07 2003
@@ -66,8 +66,8 @@
   addPass(PM, createGlobalDCEPass());            // Kill unused uinit g-vars
   addPass(PM, createDeadTypeEliminationPass());  // Eliminate dead types
   addPass(PM, createConstantMergePass());        // Merge dup global constants
-  addPass(PM, createVerifierPass());             // Verify that input is correct
   addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
+  addPass(PM, createVerifierPass());             // Verify that input is correct
   addPass(PM, createDeadInstEliminationPass());  // Remove Dead code/vars
   addPass(PM, createRaiseAllocationsPass());     // call %malloc -> malloc inst
   addPass(PM, createIndVarSimplifyPass());       // Simplify indvars
@@ -144,6 +144,9 @@
   // a little bit.  Do this now.
   //
   PassManager Passes;
+
+  // Add an appropriate TargetData instance for this module...
+  Passes.add(new TargetData("gccas", M.get()));
 
   // Add all of the transformation passes to the pass manager to do the cleanup
   // and optimization of the GCC output.




From lattner at cs.uiuc.edu  Thu Apr 24 14:11:18 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:11:18 2003
Subject: [llvm-commits] CVS: llvm/tools/opt/opt.cpp
Message-ID: <200304241910.OAA00420@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/opt:

opt.cpp updated: 1.80 -> 1.81

---
Log message:

Make sure to create a target data that matches the Module's target properties.



---
Diffs of the changes:

Index: llvm/tools/opt/opt.cpp
diff -u llvm/tools/opt/opt.cpp:1.80 llvm/tools/opt/opt.cpp:1.81
--- llvm/tools/opt/opt.cpp:1.80	Thu Apr 24 13:36:41 2003
+++ llvm/tools/opt/opt.cpp	Thu Apr 24 14:10:09 2003
@@ -113,6 +113,9 @@
   //
   PassManager Passes;
 
+  // Add an appropriate TargetData instance for this module...
+  Passes.add(new TargetData("opt", M.get()));
+
   // Create a new optimization pass for each one specified on the command line
   for (unsigned i = 0; i < OptimizationList.size(); ++i) {
     const PassInfo *Opt = OptimizationList[i];




From lattner at cs.uiuc.edu  Thu Apr 24 14:14:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:14:00 2003
Subject: [llvm-commits] CVS: llvm/tools/gccld/gccld.cpp
Message-ID: <200304241913.OAA00548@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/gccld:

gccld.cpp updated: 1.28 -> 1.29

---
Log message:

Make sure to create a target data that matches the Module's target properties.


---
Diffs of the changes:

Index: llvm/tools/gccld/gccld.cpp
diff -u llvm/tools/gccld/gccld.cpp:1.28 llvm/tools/gccld/gccld.cpp:1.29
--- llvm/tools/gccld/gccld.cpp:1.28	Mon Apr 21 14:53:24 2003
+++ llvm/tools/gccld/gccld.cpp	Thu Apr 24 14:13:02 2003
@@ -18,6 +18,7 @@
 #include "llvm/PassManager.h"
 #include "llvm/Bytecode/Reader.h"
 #include "llvm/Bytecode/WriteBytecodePass.h"
+#include "llvm/Target/TargetData.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/Scalar.h"
 #include "Support/CommandLine.h"
@@ -325,6 +326,9 @@
   // a little bit.  Do this now.
   //
   PassManager Passes;
+
+  // Add an appropriate TargetData instance for this module...
+  Passes.add(new TargetData("gccas", Composite.get()));
 
   // Linking modules together can lead to duplicated global constants, only keep
   // one copy of each constant...




From lattner at cs.uiuc.edu  Thu Apr 24 14:33:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 14:33:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/Miscompilation.cpp
Message-ID: <200304241932.OAA00900@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

Miscompilation.cpp updated: 1.3 -> 1.4

---
Log message:

Allow bugpoint to try new an different methods for pruning down lists


---
Diffs of the changes:

Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.3 llvm/tools/bugpoint/Miscompilation.cpp:1.4
--- llvm/tools/bugpoint/Miscompilation.cpp:1.3	Thu Apr 24 12:02:17 2003
+++ llvm/tools/bugpoint/Miscompilation.cpp	Thu Apr 24 14:32:42 2003
@@ -72,6 +72,30 @@
         break;
       }
     }
+
+    // Okay, we trimmed as much off the top and the bottom of the list as we
+    // could.  If there is more two elements in the list, try deleting interior
+    // elements and testing that.
+    //
+    if (TheList.size() > 2) {
+      bool Changed = true;
+      std::vector<ElTy> EmptyList;
+      while (Changed) {
+        Changed = false;
+        std::vector<ElTy> TrimmedList;
+        for (unsigned i = 1; i < TheList.size()-1; ++i) { // Check interior elts
+          std::vector<ElTy> TestList(TheList);
+          TestList.erase(TestList.begin()+i);
+
+          if (doTest(EmptyList, TestList) == KeepSuffix) {
+            // We can trim down the list!
+            TheList.swap(TestList);
+            --i;  // Don't skip an element of the list
+            Changed = true;
+          }
+        }
+      }
+    }
   }
 };
 
@@ -187,7 +211,7 @@
                             const std::vector<Function*> &Kept) {
     if (TestFuncs(Kept, false))
       return KeepSuffix;
-    if (TestFuncs(Prefix, false))
+    if (!Prefix.empty() && TestFuncs(Prefix, false))
       return KeepPrefix;
     return NoFailure;
   }




From lattner at cs.uiuc.edu  Thu Apr 24 15:08:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 15:08:01 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/PassManagerT.h
Message-ID: <200304242007.PAA01393@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

PassManagerT.h updated: 1.37 -> 1.38

---
Log message:

Fix a nasty bug where the ConstantMerge pass was invalidating the TargetData pass
even though it was immutable.  Immutable passes should never end up in CurrentAnalyses!


---
Diffs of the changes:

Index: llvm/lib/VMCore/PassManagerT.h
diff -u llvm/lib/VMCore/PassManagerT.h:1.37 llvm/lib/VMCore/PassManagerT.h:1.38
--- llvm/lib/VMCore/PassManagerT.h:1.37	Wed Feb 26 13:10:57 2003
+++ llvm/lib/VMCore/PassManagerT.h	Thu Apr 24 15:07:38 2003
@@ -332,12 +332,32 @@
     }
   }
 
+  Pass *getImmutablePassOrNull(const PassInfo *ID) const {
+    for (unsigned i = 0, e = ImmutablePasses.size(); i != e; ++i) {
+      const PassInfo *IPID = ImmutablePasses[i]->getPassInfo();
+      if (IPID == ID)
+        return ImmutablePasses[i];
+      
+      // This pass is the current implementation of all of the interfaces it
+      // implements as well.
+      //
+      const std::vector<const PassInfo*> &II =
+        IPID->getInterfacesImplemented();
+      for (unsigned j = 0, e = II.size(); j != e; ++j)
+        if (II[j] == ID) return ImmutablePasses[i];
+    }
+    return 0;
+  }
+
   Pass *getAnalysisOrNullDown(const PassInfo *ID) const {
     std::map<AnalysisID, Pass*>::const_iterator I = CurrentAnalyses.find(ID);
 
     if (I != CurrentAnalyses.end())
       return I->second;  // Found it.
 
+    if (Pass *P = getImmutablePassOrNull(ID))
+      return P;
+
     if (Batcher)
       return ((AnalysisResolver*)Batcher)->getAnalysisOrNullDown(ID);
     return 0;
@@ -350,6 +370,8 @@
 
     if (Parent)          // Try scanning...
       return Parent->getAnalysisOrNullUp(ID);
+    else if (!ImmutablePasses.empty())
+      return getImmutablePassOrNull(ID);
     return 0;
   }
 
@@ -386,7 +408,9 @@
       if (Parent) {
         Parent->markPassUsed(P, this);
       } else {
-        assert(0 && "Pass available but not found! "
+        assert(getAnalysisOrNullUp(P) && 
+               dynamic_cast<ImmutablePass*>(getAnalysisOrNullUp(P)) &&
+               "Pass available but not found! "
                "Perhaps this is a module pass requiring a function pass?");
       }
     }
@@ -556,18 +580,6 @@
     
     // Initialize the immutable pass...
     IP->initializePass();
-
-    // Add this pass to the currently available set...
-    if (const PassInfo *PI = IP->getPassInfo()) {
-      CurrentAnalyses[PI] = IP;
-
-      // This pass is the current implementation of all of the interfaces it
-      // implements as well.
-      //
-      const std::vector<const PassInfo*> &II = PI->getInterfacesImplemented();
-      for (unsigned i = 0, e = II.size(); i != e; ++i)
-        CurrentAnalyses[II[i]] = IP;
-    }
   }
 };
 




From lattner at cs.uiuc.edu  Thu Apr 24 15:17:03 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 15:17:03 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ListReducer.h Miscompilation.cpp
Message-ID: <200304242016.PAA03039@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ListReducer.h added (r1.1)
Miscompilation.cpp updated: 1.4 -> 1.5

---
Log message:

Move the ListReducer Class into it's own header file instead of living in Miscompilation.cpp


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ListReducer.h
diff -c /dev/null llvm/tools/bugpoint/ListReducer.h:1.1
*** /dev/null	Thu Apr 24 15:16:44 2003
--- llvm/tools/bugpoint/ListReducer.h	Thu Apr 24 15:16:29 2003
***************
*** 0 ****
--- 1,86 ----
+ //===- ListReducer.h - Trim down list while retaining property --*- C++ -*-===//
+ //
+ // This class is to be used as a base class for operations that want to zero in
+ // on a subset of the input which still causes the bug we are tracking.
+ //
+ //===----------------------------------------------------------------------===//
+ 
+ #ifndef BUGPOINT_LIST_REDUCER_H
+ #define BUGPOINT_LIST_REDUCER_H
+ 
+ #include <vector>
+ 
+ template<typename ElTy>
+ struct ListReducer {
+   enum TestResult {
+     NoFailure,         // No failure of the predicate was detected
+     KeepSuffix,        // The suffix alone satisfies the predicate
+     KeepPrefix,        // The prefix alone satisfies the predicate
+   };
+ 
+   // doTest - This virtual function should be overriden by subclasses to
+   // implement the test desired.  The testcase is only required to test to see
+   // if the Kept list still satisfies the property, but if it is going to check
+   // the prefix anyway, it can.
+   //
+   virtual TestResult doTest(const std::vector<ElTy> &Prefix,
+                             const std::vector<ElTy> &Kept) = 0;
+ 
+   // reduceList - This function attempts to reduce the length of the specified
+   // list while still maintaining the "test" property.  This is the core of the
+   // "work" that bugpoint does.
+   //
+   void reduceList(std::vector<ElTy> &TheList) {
+     unsigned MidTop = TheList.size();
+     while (MidTop > 1) {
+       unsigned Mid = MidTop / 2;
+       std::vector<ElTy> Prefix(TheList.begin()+Mid, TheList.end());
+       std::vector<ElTy> Kept  (TheList.begin(), TheList.begin()+Mid);
+ 
+       switch (doTest(Prefix, Kept)) {
+       case KeepSuffix:
+         // The property still holds.  We can just drop the prefix elements, and
+         // shorten the list to the "kept" elements.
+         TheList.swap(Kept);
+         MidTop = TheList.size();
+         break;
+       case KeepPrefix:
+         // The predicate still holds, shorten the list to the prefix elements.
+         TheList.swap(Prefix);
+         MidTop = TheList.size();
+         break;
+       case NoFailure:
+         // Otherwise the property doesn't hold.  Some of the elements we removed
+         // must be neccesary to maintain the property.
+         MidTop = Mid;
+         break;
+       }
+     }
+ 
+     // Okay, we trimmed as much off the top and the bottom of the list as we
+     // could.  If there is more two elements in the list, try deleting interior
+     // elements and testing that.
+     //
+     if (TheList.size() > 2) {
+       bool Changed = true;
+       std::vector<ElTy> EmptyList;
+       while (Changed) {
+         Changed = false;
+         std::vector<ElTy> TrimmedList;
+         for (unsigned i = 1; i < TheList.size()-1; ++i) { // Check interior elts
+           std::vector<ElTy> TestList(TheList);
+           TestList.erase(TestList.begin()+i);
+ 
+           if (doTest(EmptyList, TestList) == KeepSuffix) {
+             // We can trim down the list!
+             TheList.swap(TestList);
+             --i;  // Don't skip an element of the list
+             Changed = true;
+           }
+         }
+       }
+     }
+   }
+ };
+ 
+ #endif


Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.4 llvm/tools/bugpoint/Miscompilation.cpp:1.5
--- llvm/tools/bugpoint/Miscompilation.cpp:1.4	Thu Apr 24 14:32:42 2003
+++ llvm/tools/bugpoint/Miscompilation.cpp	Thu Apr 24 15:16:29 2003
@@ -6,6 +6,7 @@
 
 #include "BugDriver.h"
 #include "SystemUtils.h"
+#include "ListReducer.h"
 #include "llvm/Pass.h"
 #include "llvm/Module.h"
 #include "llvm/Transforms/Utils/Cloning.h"
@@ -25,79 +26,6 @@
   Output("output", cl::desc("Specify a reference program output "
 			    "(for miscompilation detection)"));
 }
-
-template<typename ElTy>
-struct ListReducer {
-  enum TestResult {
-    NoFailure,         // No failure of the predicate was detected
-    KeepSuffix,        // The suffix alone satisfies the predicate
-    KeepPrefix,        // The prefix alone satisfies the predicate
-  };
-
-  // doTest - This virtual function should be overriden by subclasses to
-  // implement the test desired.  The testcase is only required to test to see
-  // if the Kept list still satisfies the property, but if it is going to check
-  // the prefix anyway, it can.
-  //
-  virtual TestResult doTest(const std::vector<ElTy> &Prefix,
-                            const std::vector<ElTy> &Kept) = 0;
-
-  // reduceList - This function attempts to reduce the length of the specified
-  // list while still maintaining the "test" property.  This is the core of the
-  // "work" that bugpoint does.
-  //
-  void reduceList(std::vector<ElTy> &TheList) {
-    unsigned MidTop = TheList.size();
-    while (MidTop > 1) {
-      unsigned Mid = MidTop / 2;
-      std::vector<ElTy> Prefix(TheList.begin()+Mid, TheList.end());
-      std::vector<ElTy> Kept  (TheList.begin(), TheList.begin()+Mid);
-
-      switch (doTest(Prefix, Kept)) {
-      case KeepSuffix:
-        // The property still holds.  We can just drop the prefix elements, and
-        // shorten the list to the "kept" elements.
-        TheList.swap(Kept);
-        MidTop = TheList.size();
-        break;
-      case KeepPrefix:
-        // The predicate still holds, shorten the list to the prefix elements.
-        TheList.swap(Prefix);
-        MidTop = TheList.size();
-        break;
-      case NoFailure:
-        // Otherwise the property doesn't hold.  Some of the elements we removed
-        // must be neccesary to maintain the property.
-        MidTop = Mid;
-        break;
-      }
-    }
-
-    // Okay, we trimmed as much off the top and the bottom of the list as we
-    // could.  If there is more two elements in the list, try deleting interior
-    // elements and testing that.
-    //
-    if (TheList.size() > 2) {
-      bool Changed = true;
-      std::vector<ElTy> EmptyList;
-      while (Changed) {
-        Changed = false;
-        std::vector<ElTy> TrimmedList;
-        for (unsigned i = 1; i < TheList.size()-1; ++i) { // Check interior elts
-          std::vector<ElTy> TestList(TheList);
-          TestList.erase(TestList.begin()+i);
-
-          if (doTest(EmptyList, TestList) == KeepSuffix) {
-            // We can trim down the list!
-            TheList.swap(TestList);
-            --i;  // Don't skip an element of the list
-            Changed = true;
-          }
-        }
-      }
-    }
-  }
-};
 
 class ReduceMiscompilingPasses : public ListReducer<const PassInfo*> {
   BugDriver &BD;




From lattner at cs.uiuc.edu  Thu Apr 24 17:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:24:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.cpp
Message-ID: <200304242223.RAA05966@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.cpp updated: 1.3 -> 1.4

---
Log message:

Move function from Miscompilation.cpp


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.cpp
diff -u llvm/tools/bugpoint/BugDriver.cpp:1.3 llvm/tools/bugpoint/BugDriver.cpp:1.4
--- llvm/tools/bugpoint/BugDriver.cpp:1.3	Thu Apr 24 12:02:17 2003
+++ llvm/tools/bugpoint/BugDriver.cpp	Thu Apr 24 17:23:34 2003
@@ -27,6 +27,19 @@
   return Result;
 }
 
+// DeleteFunctionBody - "Remove" the function by deleting all of it's basic
+// blocks, making it external.
+//
+void DeleteFunctionBody(Function *F) {
+  // First, break circular use/def chain references...
+  for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I)
+    I->dropAllReferences();
+
+  // Next, delete all of the basic blocks.
+  F->getBasicBlockList().clear();
+
+  assert(F->isExternal() && "This didn't make the function external!");
+}
 
 /// ParseInputFile - Given a bytecode or assembly input filename, parse and
 /// return it, or return null if not possible.




From lattner at cs.uiuc.edu  Thu Apr 24 17:25:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:25:00 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ListReducer.h
Message-ID: <200304242224.RAA05999@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ListReducer.h updated: 1.1 -> 1.2

---
Log message:

Allow reducer interfaces to mutate the lists passed in


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ListReducer.h
diff -u llvm/tools/bugpoint/ListReducer.h:1.1 llvm/tools/bugpoint/ListReducer.h:1.2
--- llvm/tools/bugpoint/ListReducer.h:1.1	Thu Apr 24 15:16:29 2003
+++ llvm/tools/bugpoint/ListReducer.h	Thu Apr 24 17:23:56 2003
@@ -23,8 +23,8 @@
   // if the Kept list still satisfies the property, but if it is going to check
   // the prefix anyway, it can.
   //
-  virtual TestResult doTest(const std::vector<ElTy> &Prefix,
-                            const std::vector<ElTy> &Kept) = 0;
+  virtual TestResult doTest(std::vector<ElTy> &Prefix,
+                            std::vector<ElTy> &Kept) = 0;
 
   // reduceList - This function attempts to reduce the length of the specified
   // list while still maintaining the "test" property.  This is the core of the




From lattner at cs.uiuc.edu  Thu Apr 24 17:26:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:26:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.h CrashDebugger.cpp
Message-ID: <200304242225.RAA06109@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.h updated: 1.5 -> 1.6
CrashDebugger.cpp updated: 1.5 -> 1.6

---
Log message:

Use the list reducer to improve convergence speed and to support crashes that
only occur when multiple passes interact or when multiple functions exist in a module


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.h
diff -u llvm/tools/bugpoint/BugDriver.h:1.5 llvm/tools/bugpoint/BugDriver.h:1.6
--- llvm/tools/bugpoint/BugDriver.h:1.5	Thu Apr 24 12:02:17 2003
+++ llvm/tools/bugpoint/BugDriver.h	Thu Apr 24 17:24:58 2003
@@ -17,8 +17,10 @@
 class AbstractInterpreter;
 class Instruction;
 
+class DebugCrashes;
 class ReduceMiscompilingPasses;
 class ReduceMiscompilingFunctions;
+class ReduceCrashingFunctions;
 
 class BugDriver {
   const std::string ToolName;  // Name of bugpoint
@@ -26,8 +28,11 @@
   std::vector<const PassInfo*> PassesToRun;
   AbstractInterpreter *Interpreter;   // How to run the program
 
+  // FIXME: sort out public/private distinctions...
+  friend class DebugCrashes;
   friend class ReduceMiscompilingPasses;
   friend class ReduceMiscompilingFunctions;
+  friend class ReduceCrashingFunctions;
 public:
   BugDriver(const char *toolname)
     : ToolName(toolname), Program(0), Interpreter(0) {}
@@ -52,12 +57,6 @@
   ///
   bool debugCrash();
 
-  /// debugPassCrash - This method is called when the specified pass crashes on
-  /// Program as input.  It tries to reduce the testcase to something that still
-  /// crashes, but it smaller.
-  ///
-  bool debugPassCrash(const PassInfo *PI);
-
   /// debugMiscompilation - This method is used when the passes selected are not
   /// crashing, but the generated output is semantically different from the
   /// input.
@@ -162,5 +161,10 @@
 /// command line options that must be passed to add the passes.
 ///
 std::string getPassesString(const std::vector<const PassInfo*> &Passes);
+
+// DeleteFunctionBody - "Remove" the function by deleting all of it's basic
+// blocks, making it external.
+//
+void DeleteFunctionBody(Function *F);
 
 #endif


Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.5 llvm/tools/bugpoint/CrashDebugger.cpp:1.6
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.5	Thu Apr 24 12:02:17 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Apr 24 17:24:58 2003
@@ -6,12 +6,14 @@
 
 #include "BugDriver.h"
 #include "SystemUtils.h"
+#include "ListReducer.h"
 #include "llvm/Module.h"
+#include "llvm/Transforms/Utils/Cloning.h"
 #include "llvm/Bytecode/Writer.h"
 #include "llvm/Pass.h"
 #include <fstream>
+#include <set>
 
-#if 0
 class DebugCrashes : public ListReducer<const PassInfo*> {
   BugDriver &BD;
 public:
@@ -21,140 +23,148 @@
   // the "Kept" passes fail when run on the output of the "removed" passes.  If
   // we return true, we update the current module of bugpoint.
   //
-  virtual bool doTest(const std::vector<ElTy> &Removed,
-                      const std::vector<ElTy> &Kept) {
-    return BD.runPasses(Kept);
+  virtual TestResult doTest(std::vector<const PassInfo*> &Removed,
+                            std::vector<const PassInfo*> &Kept);
+};
+
+DebugCrashes::TestResult
+DebugCrashes::doTest(std::vector<const PassInfo*> &Prefix,
+                     std::vector<const PassInfo*> &Suffix) {
+  std::string PrefixOutput;
+  if (!Prefix.empty()) {
+    std::cout << "Checking to see if these passes crash: "
+              << getPassesString(Prefix) << ": ";
+    if (BD.runPasses(Prefix, PrefixOutput))
+      return KeepPrefix;
+  }
+
+  std::cout << "Checking to see if these passes crash: "
+            << getPassesString(Suffix) << ": ";
+  Module *OrigProgram = BD.Program;
+  BD.Program = BD.ParseInputFile(PrefixOutput);
+  if (BD.Program == 0) {
+    std::cerr << BD.getToolName() << ": Error reading bytecode file '"
+              << PrefixOutput << "'!\n";
+    exit(1);
+  }
+  removeFile(PrefixOutput);
+
+  if (BD.runPasses(Suffix)) {
+    delete OrigProgram;            // The suffix crashes alone...
+    return KeepSuffix;
+  }
+
+  // Nothing failed, restore state...
+  delete BD.Program;
+  BD.Program = OrigProgram;
+  return NoFailure;
+}
+
+class ReduceCrashingFunctions : public ListReducer<Function*> {
+  BugDriver &BD;
+public:
+  ReduceCrashingFunctions(BugDriver &bd) : BD(bd) {}
+
+  virtual TestResult doTest(std::vector<Function*> &Prefix,
+                            std::vector<Function*> &Kept) {
+    if (TestFuncs(Kept))
+      return KeepSuffix;
+    if (!Prefix.empty() && TestFuncs(Prefix))
+      return KeepPrefix;
+    return NoFailure;
   }
+  
+  bool TestFuncs(std::vector<Function*> &Prefix);
 };
-#endif
+
+bool ReduceCrashingFunctions::TestFuncs(std::vector<Function*> &Funcs) {
+  // Clone the program to try hacking it appart...
+  Module *M = CloneModule(BD.Program);
+  
+  // Convert list to set for fast lookup...
+  std::set<Function*> Functions;
+  for (unsigned i = 0, e = Funcs.size(); i != e; ++i) {
+    Function *CMF = M->getFunction(Funcs[i]->getName(), 
+                                   Funcs[i]->getFunctionType());
+    assert(CMF && "Function not in module?!");
+    Functions.insert(CMF);    
+  }
+
+  std::cout << "Checking for crash with only these functions:";
+  for (unsigned i = 0, e = Funcs.size(); i != e; ++i)
+    std::cout << " " << Funcs[i]->getName();
+  std::cout << ": ";
+
+  // Loop over and delete any functions which we aren't supposed to be playing
+  // with...
+  for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
+    if (I->isExternal() && !Functions.count(I))
+      DeleteFunctionBody(I);
+
+  // Try running the hacked up program...
+  std::swap(BD.Program, M);
+  if (BD.runPasses(BD.PassesToRun)) {
+    delete M;         // It crashed, keep the trimmed version...
+
+    // Make sure to use function pointers that point into the now-current
+    // module.
+    Funcs.assign(Functions.begin(), Functions.end());
+    return true;
+  }
+  delete BD.Program;  // It didn't crash, revert...
+  BD.Program = M;
+  return false;
+}
+
 
 /// debugCrash - This method is called when some pass crashes on input.  It
 /// attempts to prune down the testcase to something reasonable, and figure
 /// out exactly which pass is crashing.
 ///
 bool BugDriver::debugCrash() {
+  bool AnyReduction = false;
   std::cout << "\n*** Debugging optimizer crash!\n";
 
-#if 0
   // Reduce the list of passes which causes the optimizer to crash...
+  unsigned OldSize = PassesToRun.size();
   DebugCrashes(*this).reduceList(PassesToRun);
-#endif
-
-  unsigned LastToPass = 0, LastToCrash = PassesToRun.size();
-  while (LastToPass != LastToCrash) {
-    unsigned Mid = (LastToCrash+LastToPass+1) / 2;
-    std::vector<const PassInfo*> P(PassesToRun.begin(),
-                                   PassesToRun.begin()+Mid);
-    std::cout << "Checking to see if the first " << Mid << " passes crash: ";
-
-    if (runPasses(P))
-      LastToCrash = Mid-1;
-    else
-      LastToPass = Mid;
-  }
 
-  // Make sure something crashed.  :)
-  if (LastToCrash >= PassesToRun.size()) {
+  if (PassesToRun.size() == OldSize) { // Make sure something crashed.  :)
     std::cerr << "ERROR: No passes crashed!\n";
     return true;
   }
 
-  // Calculate which pass it is that crashes...
-  const PassInfo *CrashingPass = PassesToRun[LastToCrash];
-  
-  std::cout << "\n*** Found crashing pass '-" << CrashingPass->getPassArgument()
-            << "': " << CrashingPass->getPassName() << "\n";
+  std::cout << "\n*** Found crashing pass"
+            << (PassesToRun.size() == 1 ? ": " : "es: ")
+            << getPassesString(PassesToRun) << "\n";
 
-  // Compile the program with just the passes that don't crash.
-  if (LastToPass != 0) { // Don't bother doing this if the first pass crashes...
-    std::vector<const PassInfo*> P(PassesToRun.begin(), 
-                                   PassesToRun.begin()+LastToPass);
-    std::string Filename;
-    std::cout << "Running passes that don't crash to get input for pass: ";
-    if (runPasses(P, Filename)) {
-      std::cerr << "ERROR: Running the first " << LastToPass
-                << " passes crashed this time!\n";
-      return true;
-    }
-
-    // Assuming everything was successful, we now have a valid bytecode file in
-    // OutputName.  Use it for "Program" Instead.
-    delete Program;
-    Program = ParseInputFile(Filename);
-
-    // Delete the file now.
-    removeFile(Filename);
-  }
-
-  PassesToRun.clear();
-  PassesToRun.push_back(CrashingPass);
-
-  return debugPassCrash(CrashingPass);
-}
-
-/// CountFunctions - return the number of non-external functions defined in the
-/// module.
-static unsigned CountFunctions(Module *M) {
-  unsigned N = 0;
-  for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
-    if (!I->isExternal())
-      ++N;
-  return N;
-}
-
-/// debugPassCrash - This method is called when the specified pass crashes on
-/// Program as input.  It tries to reduce the testcase to something that still
-/// crashes, but it smaller.
-///
-bool BugDriver::debugPassCrash(const PassInfo *Pass) {
   EmitProgressBytecode("passinput");
-  bool Reduced = false, AnyReduction = false;
-
-  if (CountFunctions(Program) > 1) {
-    // Attempt to reduce the input program down to a single function that still
-    // crashes.  Do this by removing everything except for that one function...
-    //
-    std::cout << "\n*** Attempting to reduce the testcase to one function\n";
-
-    for (Module::iterator I = Program->begin(), E = Program->end(); I != E; ++I)
-      if (!I->isExternal()) {
-        // Extract one function from the module...
-        Module *M = extractFunctionFromModule(I);
-
-        // Make the function the current program...
-        std::swap(Program, M);
-        
-        // Find out if the pass still crashes on this pass...
-        std::cout << "Checking function '" << I->getName() << "': ";
-        if (runPass(Pass)) {
-          // Yup, it does, we delete the old module, and continue trying to
-          // reduce the testcase...
-          delete M;
-
-          Reduced = AnyReduction = true;
-          break;
-        }
-        
-        // This pass didn't crash on this function, try the next one.
-        delete Program;
-        Program = M;
-      }
+  
+  // Now try to reduce the number of functions in the module to something small.
+  std::vector<Function*> Functions;
+  for (Module::iterator I = Program->begin(), E = Program->end(); I != E; ++I)
+    if (!I->isExternal())
+      Functions.push_back(I);
 
-    if (CountFunctions(Program) > 1) {
-      std::cout << "\n*** Couldn't reduce testcase to one function.\n"
-                << "    Attempting to remove individual functions.\n";
-      std::cout << "XXX Individual function removal unimplemented!\n";
+  if (Functions.size() > 1) {
+    std::cout << "\n*** Attempting to reduce the number of functions "
+      "in the testcase\n";
+
+    OldSize = Functions.size();
+    ReduceCrashingFunctions(*this).reduceList(Functions);
+
+    if (Functions.size() < OldSize) {
+      EmitProgressBytecode("reduced-function");
+      AnyReduction = true;
     }
   }
 
-  if (Reduced) {
-    EmitProgressBytecode("reduced-function");
-    Reduced = false;
-  }
-
   // FIXME: This should attempt to delete entire basic blocks at a time to speed
   // up convergence...
 
+  // FIXME: This should use the list reducer to converge faster by deleting
+  // larger chunks of instructions at a time!
+  bool Reduced = false;
   unsigned Simplification = 4;
   do {
     --Simplification;
@@ -186,7 +196,7 @@
             
             // Find out if the pass still crashes on this pass...
             std::cout << "Checking instruction '" << I->getName() << "': ";
-            if (runPass(Pass)) {
+            if (runPasses(PassesToRun)) {
               // Yup, it does, we delete the old module, and continue trying to
               // reduce the testcase...
               delete M;
@@ -209,7 +219,7 @@
     std::swap(Program, M);
             
     // Find out if the pass still crashes on the cleaned up program...
-    if (runPass(Pass)) {
+    if (runPasses(PassesToRun)) {
       // Yup, it does, keep the reduced version...
       delete M;
       Reduced = AnyReduction = true;




From lattner at cs.uiuc.edu  Thu Apr 24 17:26:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:26:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/Miscompilation.cpp
Message-ID: <200304242225.RAA06094@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

Miscompilation.cpp updated: 1.5 -> 1.6

---
Log message:

Adjust to match new ListReducer interface
Move function to generic code


---
Diffs of the changes:

Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.5 llvm/tools/bugpoint/Miscompilation.cpp:1.6
--- llvm/tools/bugpoint/Miscompilation.cpp:1.5	Thu Apr 24 15:16:29 2003
+++ llvm/tools/bugpoint/Miscompilation.cpp	Thu Apr 24 17:24:22 2003
@@ -32,13 +32,13 @@
 public:
   ReduceMiscompilingPasses(BugDriver &bd) : BD(bd) {}
 
-  virtual TestResult doTest(const std::vector<const PassInfo*> &Prefix,
-                            const std::vector<const PassInfo*> &Kept);
+  virtual TestResult doTest(std::vector<const PassInfo*> &Prefix,
+                            std::vector<const PassInfo*> &Kept);
 };
 
 ReduceMiscompilingPasses::TestResult
-ReduceMiscompilingPasses::doTest(const std::vector<const PassInfo*> &Prefix,
-                                 const std::vector<const PassInfo*> &Kept) {
+ReduceMiscompilingPasses::doTest(std::vector<const PassInfo*> &Prefix,
+                                 std::vector<const PassInfo*> &Kept) {
   // First, run the program with just the Kept passes.  If it is still broken
   // with JUST the kept passes, discard the prefix passes.
   std::cout << "Checking to see if '" << getPassesString(Kept)
@@ -135,8 +135,8 @@
 public:
   ReduceMiscompilingFunctions(BugDriver &bd) : BD(bd) {}
 
-  virtual TestResult doTest(const std::vector<Function*> &Prefix,
-                            const std::vector<Function*> &Kept) {
+  virtual TestResult doTest(std::vector<Function*> &Prefix,
+                            std::vector<Function*> &Kept) {
     if (TestFuncs(Kept, false))
       return KeepSuffix;
     if (!Prefix.empty() && TestFuncs(Prefix, false))
@@ -146,21 +146,6 @@
   
   bool TestFuncs(const std::vector<Function*> &Prefix, bool EmitBytecode);
 };
-
-// DeleteFunctionBody - "Remove" the function by deleting all of it's basic
-// blocks, making it external.
-//
-static void DeleteFunctionBody(Function *F) {
-  // First, break circular use/def chain references...
-  for (Function::iterator I = F->begin(), E = F->end(); I != E; ++I)
-    I->dropAllReferences();
-
-  // Next, delete all of the basic blocks.
-  F->getBasicBlockList().clear();
-
-  assert(F->isExternal() && "This didn't make the function external!");
-}
-
 
 bool ReduceMiscompilingFunctions::TestFuncs(const std::vector<Function*> &Funcs,
                                             bool EmitBytecode) {




From lattner at cs.uiuc.edu  Thu Apr 24 17:54:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:54:00 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.h
Message-ID: <200304242253.RAA07736@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.h updated: 1.6 -> 1.7

---
Log message:

Remove dead functions


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.h
diff -u llvm/tools/bugpoint/BugDriver.h:1.6 llvm/tools/bugpoint/BugDriver.h:1.7
--- llvm/tools/bugpoint/BugDriver.h:1.6	Thu Apr 24 17:24:58 2003
+++ llvm/tools/bugpoint/BugDriver.h	Thu Apr 24 17:53:24 2003
@@ -110,19 +110,6 @@
     return runPasses(PassesToRun, Filename, DeleteOutput);
   }
 
-  /// runPass - Run only the specified pass on the program.
-  ///
-  bool runPass(const PassInfo *P, bool DeleteOutput = true) const {
-    return runPasses(std::vector<const PassInfo*>(1, P), DeleteOutput);
-  }
-  
-  /// extractFunctionFromModule - This method is used to extract the specified
-  /// (non-external) function from the current program, slim down the module,
-  /// and then return it.  This does not modify Program at all, it modifies a
-  /// copy, which it returns.
-  ///
-  Module *extractFunctionFromModule(Function *F) const;
-
   /// deleteInstructionFromProgram - This method clones the current Program and
   /// deletes the specified instruction from the cloned module.  It then runs a
   /// series of cleanup passes (ADCE and SimplifyCFG) to eliminate any code




From lattner at cs.uiuc.edu  Thu Apr 24 17:54:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:54:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.cpp
Message-ID: <200304242253.RAA07780@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.cpp updated: 1.4 -> 1.5

---
Log message:

Make sure that deleted functions have external linkage


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.cpp
diff -u llvm/tools/bugpoint/BugDriver.cpp:1.4 llvm/tools/bugpoint/BugDriver.cpp:1.5
--- llvm/tools/bugpoint/BugDriver.cpp:1.4	Thu Apr 24 17:23:34 2003
+++ llvm/tools/bugpoint/BugDriver.cpp	Thu Apr 24 17:53:01 2003
@@ -37,7 +37,7 @@
 
   // Next, delete all of the basic blocks.
   F->getBasicBlockList().clear();
-
+  F->setLinkage(GlobalValue::ExternalLinkage);
   assert(F->isExternal() && "This didn't make the function external!");
 }
 




From lattner at cs.uiuc.edu  Thu Apr 24 17:55:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:55:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/CrashDebugger.cpp
Message-ID: <200304242254.RAA07970@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

CrashDebugger.cpp updated: 1.6 -> 1.7

---
Log message:

The big fix is this change:
-    if (I->isExternal() && !Functions.count(I))
+    if (!I->isExternal() && !Functions.count(I))

We were not actually deleting any functions from the module!



---
Diffs of the changes:

Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.6 llvm/tools/bugpoint/CrashDebugger.cpp:1.7
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.6	Thu Apr 24 17:24:58 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Apr 24 17:54:06 2003
@@ -87,7 +87,7 @@
     Function *CMF = M->getFunction(Funcs[i]->getName(), 
                                    Funcs[i]->getFunctionType());
     assert(CMF && "Function not in module?!");
-    Functions.insert(CMF);    
+    Functions.insert(CMF);
   }
 
   std::cout << "Checking for crash with only these functions:";
@@ -98,7 +98,7 @@
   // Loop over and delete any functions which we aren't supposed to be playing
   // with...
   for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
-    if (I->isExternal() && !Functions.count(I))
+    if (!I->isExternal() && !Functions.count(I))
       DeleteFunctionBody(I);
 
   // Try running the hacked up program...
@@ -129,11 +129,6 @@
   unsigned OldSize = PassesToRun.size();
   DebugCrashes(*this).reduceList(PassesToRun);
 
-  if (PassesToRun.size() == OldSize) { // Make sure something crashed.  :)
-    std::cerr << "ERROR: No passes crashed!\n";
-    return true;
-  }
-
   std::cout << "\n*** Found crashing pass"
             << (PassesToRun.size() == 1 ? ": " : "es: ")
             << getPassesString(PassesToRun) << "\n";
@@ -164,7 +159,6 @@
 
   // FIXME: This should use the list reducer to converge faster by deleting
   // larger chunks of instructions at a time!
-  bool Reduced = false;
   unsigned Simplification = 4;
   do {
     --Simplification;
@@ -200,7 +194,7 @@
               // Yup, it does, we delete the old module, and continue trying to
               // reduce the testcase...
               delete M;
-              Reduced = AnyReduction = true;
+              AnyReduction = true;
               goto TryAgain;  // I wish I had a multi-level break here!
             }
             
@@ -222,17 +216,15 @@
     if (runPasses(PassesToRun)) {
       // Yup, it does, keep the reduced version...
       delete M;
-      Reduced = AnyReduction = true;
+      AnyReduction = true;
     } else {
       delete Program;   // Otherwise, restore the original module...
       Program = M;
     }
   }
 
-  if (Reduced) {
+  if (AnyReduction)
     EmitProgressBytecode("reduced-simplified");
-    Reduced = false;
-  }
 
   return false;
 }




From lattner at cs.uiuc.edu  Thu Apr 24 17:55:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 17:55:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ExtractFunction.cpp
Message-ID: <200304242254.RAA08004@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ExtractFunction.cpp updated: 1.4 -> 1.5

---
Log message:

Remove dead functions


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ExtractFunction.cpp
diff -u llvm/tools/bugpoint/ExtractFunction.cpp:1.4 llvm/tools/bugpoint/ExtractFunction.cpp:1.5
--- llvm/tools/bugpoint/ExtractFunction.cpp:1.4	Fri Mar  7 12:17:13 2003
+++ llvm/tools/bugpoint/ExtractFunction.cpp	Thu Apr 24 17:53:11 2003
@@ -15,31 +15,6 @@
 #include "llvm/Type.h"
 #include "llvm/Constant.h"
 
-/// extractFunctionFromModule - This method is used to extract the specified
-/// (non-external) function from the current program, slim down the module, and
-/// then return it.  This does not modify Program at all, it modifies a copy,
-/// which it returns.
-Module *BugDriver::extractFunctionFromModule(Function *F) const {
-  Module *Result = CloneModule(Program);
-
-  // Translate from the old module to the new copied module...
-  Module::iterator RFI = Result->begin(); // Get iterator to corresponding fn
-  std::advance(RFI, std::distance(Program->begin(), Module::iterator(F)));
-
-  // In addition to just parsing the input from GCC, we also want to spiff it up
-  // a little bit.  Do this now.
-  //
-  PassManager Passes;
-  Passes.add(createFunctionExtractionPass(RFI));  // Extract the function
-  Passes.add(createGlobalDCEPass());              // Delete unreachable globals
-  Passes.add(createFunctionResolvingPass());      // Delete prototypes
-  Passes.add(createDeadTypeEliminationPass());    // Remove dead types...
-  Passes.add(createVerifierPass());
-  Passes.run(*Result);
-  return Result;
-}
-
-
 /// deleteInstructionFromProgram - This method clones the current Program and
 /// deletes the specified instruction from the cloned module.  It then runs a
 /// series of cleanup passes (ADCE and SimplifyCFG) to eliminate any code which
@@ -69,9 +44,7 @@
   // Remove the instruction from the program.
   I->getParent()->getInstList().erase(I);
 
-  // In addition to just parsing the input from GCC, we also want to spiff it up
-  // a little bit.  Do this now.
-  //
+  // Spiff up the output a little bit.
   PassManager Passes;
   if (Simplification > 2)
     Passes.add(createAggressiveDCEPass());          // Remove dead code...




From lattner at cs.uiuc.edu  Thu Apr 24 18:52:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 18:52:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/BugDriver.h CrashDebugger.cpp
Message-ID: <200304242351.SAA09498@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

BugDriver.h updated: 1.7 -> 1.8
CrashDebugger.cpp updated: 1.7 -> 1.8

---
Log message:

Speed up convergence significantly and also reduce the size of testcases by making large portions of a function's CFG dead at a time.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/BugDriver.h
diff -u llvm/tools/bugpoint/BugDriver.h:1.7 llvm/tools/bugpoint/BugDriver.h:1.8
--- llvm/tools/bugpoint/BugDriver.h:1.7	Thu Apr 24 17:53:24 2003
+++ llvm/tools/bugpoint/BugDriver.h	Thu Apr 24 18:51:38 2003
@@ -21,6 +21,7 @@
 class ReduceMiscompilingPasses;
 class ReduceMiscompilingFunctions;
 class ReduceCrashingFunctions;
+class ReduceCrashingBlocks;
 
 class BugDriver {
   const std::string ToolName;  // Name of bugpoint
@@ -33,6 +34,7 @@
   friend class ReduceMiscompilingPasses;
   friend class ReduceMiscompilingFunctions;
   friend class ReduceCrashingFunctions;
+  friend class ReduceCrashingBlocks;
 public:
   BugDriver(const char *toolname)
     : ToolName(toolname), Program(0), Interpreter(0) {}


Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.7 llvm/tools/bugpoint/CrashDebugger.cpp:1.8
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.7	Thu Apr 24 17:54:06 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Apr 24 18:51:38 2003
@@ -8,9 +8,17 @@
 #include "SystemUtils.h"
 #include "ListReducer.h"
 #include "llvm/Module.h"
+#include "llvm/PassManager.h"
+#include "llvm/Pass.h"
+#include "llvm/Constant.h"
+#include "llvm/iTerminators.h"
+#include "llvm/Type.h"
+#include "llvm/SymbolTable.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/Analysis/Verifier.h"
+#include "llvm/Transforms/Scalar.h"
 #include "llvm/Transforms/Utils/Cloning.h"
 #include "llvm/Bytecode/Writer.h"
-#include "llvm/Pass.h"
 #include <fstream>
 #include <set>
 
@@ -117,6 +125,108 @@
 }
 
 
+/// ReduceCrashingBlocks reducer - This works by setting the terminators of all
+/// terminators except the specified basic blocks to a 'ret' instruction, then
+/// running the simplify-cfg pass.  This has the effect of chopping up the CFG
+/// really fast which can reduce large functions quickly.
+///
+class ReduceCrashingBlocks : public ListReducer<BasicBlock*> {
+  BugDriver &BD;
+public:
+  ReduceCrashingBlocks(BugDriver &bd) : BD(bd) {}
+    
+  virtual TestResult doTest(std::vector<BasicBlock*> &Prefix,
+                            std::vector<BasicBlock*> &Kept) {
+    if (TestBlocks(Kept))
+      return KeepSuffix;
+    if (!Prefix.empty() && TestBlocks(Prefix))
+      return KeepPrefix;
+    return NoFailure;
+  }
+    
+  bool TestBlocks(std::vector<BasicBlock*> &Prefix);
+};
+
+bool ReduceCrashingBlocks::TestBlocks(std::vector<BasicBlock*> &BBs) {
+  // Clone the program to try hacking it appart...
+  Module *M = CloneModule(BD.Program);
+  
+  // Convert list to set for fast lookup...
+  std::set<BasicBlock*> Blocks;
+  for (unsigned i = 0, e = BBs.size(); i != e; ++i) {
+    // Convert the basic block from the original module to the new module...
+    Function *F = BBs[i]->getParent();
+    Function *CMF = M->getFunction(F->getName(), F->getFunctionType());
+    assert(CMF && "Function not in module?!");
+
+    // Get the mapped basic block...
+    Function::iterator CBI = CMF->begin();
+    std::advance(CBI, std::distance(F->begin(), Function::iterator(BBs[i])));
+    Blocks.insert(CBI);
+  }
+
+  std::cout << "Checking for crash with only these blocks:";
+  for (unsigned i = 0, e = Blocks.size(); i != e; ++i)
+    std::cout << " " << BBs[i]->getName();
+  std::cout << ": ";
+
+  // Loop over and delete any hack up any blocks that are not listed...
+  for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
+    for (Function::iterator BB = I->begin(), E = I->end(); BB != E; ++BB)
+      if (!Blocks.count(BB) && !isa<ReturnInst>(BB->getTerminator())) {
+        // Loop over all of the successors of this block, deleting any PHI nodes
+        // that might include it.
+        for (succ_iterator SI = succ_begin(BB), E = succ_end(BB); SI != E; ++SI)
+          (*SI)->removePredecessor(BB);
+
+        // Delete the old terminator instruction...
+        BB->getInstList().pop_back();
+        
+        // Add a new return instruction of the appropriate type...
+        const Type *RetTy = BB->getParent()->getReturnType();
+        ReturnInst *RI = new ReturnInst(RetTy == Type::VoidTy ? 0 :
+                                        Constant::getNullValue(RetTy));
+        BB->getInstList().push_back(RI);
+      }
+
+  // The CFG Simplifier pass may delete one of the basic blocks we are
+  // interested in.  If it does we need to take the block out of the list.  Make
+  // a "persistent mapping" by turning basic blocks into <function, name> pairs.
+  // This won't work well if blocks are unnamed, but that is just the risk we
+  // have to take.
+  std::vector<std::pair<Function*, std::string> > BlockInfo;
+
+  for (std::set<BasicBlock*>::iterator I = Blocks.begin(), E = Blocks.end();
+       I != E; ++I)
+    BlockInfo.push_back(std::make_pair((*I)->getParent(), (*I)->getName()));
+
+  // Now run the CFG simplify pass on the function...
+  PassManager Passes;
+  Passes.add(createCFGSimplificationPass());
+  Passes.add(createVerifierPass());
+  Passes.run(*M);
+
+  // Try running on the hacked up program...
+  std::swap(BD.Program, M);
+  if (BD.runPasses(BD.PassesToRun)) {
+    delete M;         // It crashed, keep the trimmed version...
+
+    // Make sure to use basic block pointers that point into the now-current
+    // module, and that they don't include any deleted blocks.
+    BBs.clear();
+    for (unsigned i = 0, e = BlockInfo.size(); i != e; ++i) {
+      SymbolTable &ST = BlockInfo[i].first->getSymbolTable();
+      SymbolTable::iterator I = ST.find(Type::LabelTy);
+      if (I != ST.end() && I->second.count(BlockInfo[i].second))
+        BBs.push_back(cast<BasicBlock>(I->second[BlockInfo[i].second]));
+    }
+    return true;
+  }
+  delete BD.Program;  // It didn't crash, revert...
+  BD.Program = M;
+  return false;
+}
+
 /// debugCrash - This method is called when some pass crashes on input.  It
 /// attempts to prune down the testcase to something reasonable, and figure
 /// out exactly which pass is crashing.
@@ -154,8 +264,16 @@
     }
   }
 
-  // FIXME: This should attempt to delete entire basic blocks at a time to speed
-  // up convergence...
+  // Attempt to delete entire basic blocks at a time to speed up
+  // convergence... this actually works by setting the terminator of the blocks
+  // to a return instruction then running simplifycfg, which can potentially
+  // shrinks the code dramatically quickly
+  //
+  std::vector<BasicBlock*> Blocks;
+  for (Module::iterator I = Program->begin(), E = Program->end(); I != E; ++I)
+    for (Function::iterator FI = I->begin(), E = I->end(); FI != E; ++FI)
+      Blocks.push_back(FI);
+  ReduceCrashingBlocks(*this).reduceList(Blocks);
 
   // FIXME: This should use the list reducer to converge faster by deleting
   // larger chunks of instructions at a time!




From lattner at cs.uiuc.edu  Thu Apr 24 19:47:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 19:47:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/Mem2Reg/2003-04-24-MultipleIdenticalSuccessors.ll
Message-ID: <200304250046.TAA09962@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/Mem2Reg:

2003-04-24-MultipleIdenticalSuccessors.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/Mem2Reg/2003-04-24-MultipleIdenticalSuccessors.ll
diff -c /dev/null llvm/test/Regression/Transforms/Mem2Reg/2003-04-24-MultipleIdenticalSuccessors.ll:1.1
*** /dev/null	Thu Apr 24 19:46:25 2003
--- llvm/test/Regression/Transforms/Mem2Reg/2003-04-24-MultipleIdenticalSuccessors.ll	Thu Apr 24 19:46:14 2003
***************
*** 0 ****
--- 1,15 ----
+ ; Mem2reg used to only add one incoming value to a PHI node, even if it had
+ ; multiple incoming edges from a block.
+ ;
+ ; RUN: as < %s | opt -mem2reg -disable-output
+ 
+ int %test(bool %c1, bool %c2) {
+ 	%X = alloca int
+ 	br bool %c1, label %Exit, label %B2
+ B2:
+ 	store int 2, int* %X
+ 	br bool %c2, label %Exit, label %Exit
+ Exit:
+ 	%Y = load int *%X
+ 	ret int %Y
+ }




From lattner at cs.uiuc.edu  Thu Apr 24 19:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 19:54:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ExtractFunction.cpp
Message-ID: <200304250053.TAA10204@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ExtractFunction.cpp updated: 1.5 -> 1.6

---
Log message:

When cleaning up the final bytecode file, make sure to run DTE as well


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ExtractFunction.cpp
diff -u llvm/tools/bugpoint/ExtractFunction.cpp:1.5 llvm/tools/bugpoint/ExtractFunction.cpp:1.6
--- llvm/tools/bugpoint/ExtractFunction.cpp:1.5	Thu Apr 24 17:53:11 2003
+++ llvm/tools/bugpoint/ExtractFunction.cpp	Thu Apr 24 19:52:30 2003
@@ -64,11 +64,12 @@
 /// before handing it to the user...
 ///
 Module *BugDriver::performFinalCleanups() const {
+  Module *M = CloneModule(Program);
   PassManager CleanupPasses;
   CleanupPasses.add(createFunctionResolvingPass());
   CleanupPasses.add(createGlobalDCEPass());
+  CleanupPasses.add(createDeadTypeEliminationPass());
   CleanupPasses.add(createVerifierPass());
-  Module *M = CloneModule(Program);
   CleanupPasses.run(*M);
   return M;
 }




From lattner at cs.uiuc.edu  Thu Apr 24 19:54:05 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 19:54:05 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/CrashDebugger.cpp
Message-ID: <200304250053.TAA10211@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

CrashDebugger.cpp updated: 1.8 -> 1.9

---
Log message:

Big programs have tons of global variable initializers, and most passes don't care
about them.  Try to delete them if it doesn't affect the passes.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/CrashDebugger.cpp
diff -u llvm/tools/bugpoint/CrashDebugger.cpp:1.8 llvm/tools/bugpoint/CrashDebugger.cpp:1.9
--- llvm/tools/bugpoint/CrashDebugger.cpp:1.8	Thu Apr 24 18:51:38 2003
+++ llvm/tools/bugpoint/CrashDebugger.cpp	Thu Apr 24 19:53:05 2003
@@ -244,6 +244,36 @@
             << getPassesString(PassesToRun) << "\n";
 
   EmitProgressBytecode("passinput");
+
+  // See if we can get away with nuking all of the global variable initializers
+  // in the program...
+  if (Program->gbegin() != Program->gend()) {
+    Module *M = CloneModule(Program);
+    bool DeletedInit = false;
+    for (Module::giterator I = M->gbegin(), E = M->gend(); I != E; ++I)
+      if (I->hasInitializer()) {
+        I->setInitializer(0);
+        I->setLinkage(GlobalValue::ExternalLinkage);
+        DeletedInit = true;
+      }
+    
+    if (!DeletedInit) {
+      delete M;  // No change made...
+    } else {
+      // See if the program still causes a crash...
+      std::cout << "\nChecking to see if we can delete global inits: ";
+      std::swap(Program, M);
+      if (runPasses(PassesToRun)) {  // Still crashes?
+        AnyReduction = true;
+        delete M;
+        std::cout << "\n*** Able to remove all global initializers!\n";
+      } else {                       // No longer crashes?
+        delete Program;              // Restore program.
+        Program = M;
+        std::cout << "  - Removing all global inits hides problem!\n";
+      }
+    }
+  }
   
   // Now try to reduce the number of functions in the module to something small.
   std::vector<Function*> Functions;




From lattner at cs.uiuc.edu  Thu Apr 24 19:56:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 19:56:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
Message-ID: <200304250055.TAA10302@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Utils:

PromoteMemoryToRegister.cpp updated: 1.40 -> 1.41

---
Log message:

Fix bug: mem2reg/2003-04-24-MultipleIdenticalSuccessors.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp
diff -u llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.40 llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.41
--- llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp:1.40	Thu Apr 24 12:52:20 2003
+++ llvm/lib/Transforms/Utils/PromoteMemoryToRegister.cpp	Thu Apr 24 19:54:58 2003
@@ -222,13 +222,14 @@
   std::vector<PHINode *> &BBPNs = NewPhiNodes[BB];
   for (unsigned k = 0; k != BBPNs.size(); ++k)
     if (PHINode *PN = BBPNs[k]) {
-      int BBI = PN->getBasicBlockIndex(Pred);
-      assert(BBI >= 0 && "Predecessor not in basic block yet!");
-
-      // At this point we can assume that the array has phi nodes.. let's update
-      // the incoming data.
-      PN->setIncomingValue(BBI, IncomingVals[k]);
-
+      // The PHI node may have multiple entries for this predecessor.  We must
+      // make sure we update all of them.
+      for (unsigned i = 0, e = PN->getNumOperands(); i != e; i += 2) {
+        if (PN->getOperand(i+1) == Pred)
+          // At this point we can assume that the array has phi nodes.. let's
+          // update the incoming data.
+          PN->setOperand(i, IncomingVals[k]);
+      }
       // also note that the active variable IS designated by the phi node
       IncomingVals[k] = PN;
     }




From lattner at cs.uiuc.edu  Thu Apr 24 21:51:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 21:51:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/SCCP.cpp
Message-ID: <200304250250.VAA11664@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

SCCP.cpp updated: 1.65 -> 1.66

---
Log message:

Minor non-functional changes:
  * Spell propagate right
  * Improve performance of phi node handling
  * Delete using directive
  * Other minor changes



---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/SCCP.cpp
diff -u llvm/lib/Transforms/Scalar/SCCP.cpp:1.65 llvm/lib/Transforms/Scalar/SCCP.cpp:1.66
--- llvm/lib/Transforms/Scalar/SCCP.cpp:1.65	Wed Apr 23 11:37:42 2003
+++ llvm/lib/Transforms/Scalar/SCCP.cpp	Thu Apr 24 21:50:03 2003
@@ -17,17 +17,13 @@
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/ConstantHandling.h"
 #include "llvm/Function.h"
-#include "llvm/iPHINode.h"
-#include "llvm/iMemory.h"
-#include "llvm/iTerminators.h"
-#include "llvm/iOther.h"
+#include "llvm/Instructions.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/InstVisitor.h"
 #include "Support/STLExtras.h"
 #include "Support/Statistic.h"
 #include <algorithm>
 #include <set>
-using std::cerr;
 
 // InstVal class - This class represents the different lattice values that an 
 // instruction may occupy.  It is a simple class with value semantics.
@@ -111,9 +107,8 @@
   // the users of the instruction are updated later.
   //
   inline bool markConstant(Instruction *I, Constant *V) {
-    DEBUG(cerr << "markConstant: " << V << " = " << I);
-
     if (ValueState[I].markConstant(V)) {
+      DEBUG(std::cerr << "markConstant: " << V << " = " << I);
       InstWorkList.push_back(I);
       return true;
     }
@@ -127,7 +122,7 @@
   inline bool markOverdefined(Value *V) {
     if (ValueState[V].markOverdefined()) {
       if (Instruction *I = dyn_cast<Instruction>(V)) {
-	DEBUG(cerr << "markOverdefined: " << V);
+	DEBUG(std::cerr << "markOverdefined: " << V);
 	InstWorkList.push_back(I);  // Only instructions go on the work list
       }
       return true;
@@ -161,10 +156,19 @@
   // work list if it is not already executable...
   // 
   void markExecutable(BasicBlock *BB) {
-    if (BBExecutable.count(BB)) return;
-    DEBUG(cerr << "Marking BB Executable: " << *BB);
-    BBExecutable.insert(BB);   // Basic block is executable!
-    BBWorkList.push_back(BB);  // Add the block to the work list!
+    if (BBExecutable.count(BB)) {
+      // BB is already executable, but we may have just made an edge feasible
+      // that wasn't before.  Add the PHI nodes to the work list so that they
+      // can be rechecked.
+      for (BasicBlock::iterator I = BB->begin();
+           PHINode *PN = dyn_cast<PHINode>(I); ++I)
+        InstWorkList.push_back(PN);
+
+    } else {
+      DEBUG(std::cerr << "Marking BB Executable: " << *BB);
+      BBExecutable.insert(BB);   // Basic block is executable!
+      BBWorkList.push_back(BB);  // Add the block to the work list!
+    }
   }
 
 
@@ -193,7 +197,7 @@
 
   void visitInstruction(Instruction &I) {
     // If a new instruction is added to LLVM that we don't handle...
-    cerr << "SCCP: Don't know how to handle: " << I;
+    std::cerr << "SCCP: Don't know how to handle: " << I;
     markOverdefined(&I);   // Just in case
   }
 
@@ -214,12 +218,12 @@
   void OperandChangedState(User *U) {
     // Only instructions use other variable values!
     Instruction &I = cast<Instruction>(*U);
-    if (!BBExecutable.count(I.getParent())) return;// Inst not executable yet!
-    visit(I);
+    if (BBExecutable.count(I.getParent()))   // Inst is executable?
+      visit(I);
   }
 };
 
-  RegisterOpt<SCCP> X("sccp", "Sparse Conditional Constant Propogation");
+  RegisterOpt<SCCP> X("sccp", "Sparse Conditional Constant Propagation");
 } // end anonymous namespace
 
 
@@ -248,8 +252,7 @@
       Instruction *I = InstWorkList.back();
       InstWorkList.pop_back();
 
-      DEBUG(cerr << "\nPopped off I-WL: " << I);
-
+      DEBUG(std::cerr << "\nPopped off I-WL: " << I);
       
       // "I" got into the work list because it either made the transition from
       // bottom to constant, or to Overdefined.
@@ -265,13 +268,7 @@
       BasicBlock *BB = BBWorkList.back();
       BBWorkList.pop_back();
 
-      DEBUG(cerr << "\nPopped off BBWL: " << BB);
-
-      // If this block only has a single successor, mark it as executable as
-      // well... if not, terminate the do loop.
-      //
-      if (BB->getTerminator()->getNumSuccessors() == 1)
-        markExecutable(BB->getTerminator()->getSuccessor(0));
+      DEBUG(std::cerr << "\nPopped off BBWL: " << BB);
 
       // Notify all instructions in this basic block that they are newly
       // executable.
@@ -282,7 +279,7 @@
   if (DebugFlag) {
     for (Function::iterator I = F.begin(), E = F.end(); I != E; ++I)
       if (!BBExecutable.count(I))
-        cerr << "BasicBlock Dead:" << *I;
+        std::cerr << "BasicBlock Dead:" << *I;
   }
 
   // Iterate over all of the instructions in a function, replacing them with
@@ -295,7 +292,7 @@
       InstVal &IV = ValueState[&Inst];
       if (IV.isConstant()) {
         Constant *Const = IV.getConstant();
-        DEBUG(cerr << "Constant: " << Const << " = " << Inst);
+        DEBUG(std::cerr << "Constant: " << Const << " = " << Inst);
 
         // Replaces all of the uses of a variable with uses of the constant.
         Inst.replaceAllUsesWith(Const);
@@ -325,7 +322,7 @@
 // successors are reachable from a given terminator instruction.
 //
 void SCCP::getFeasibleSuccessors(TerminatorInst &TI, std::vector<bool> &Succs) {
-  assert(Succs.size() == TI.getNumSuccessors() && "Succs vector wrong size!");
+  Succs.resize(TI.getNumSuccessors());
   if (BranchInst *BI = dyn_cast<BranchInst>(&TI)) {
     if (BI->isUnconditional()) {
       Succs[0] = true;
@@ -362,7 +359,7 @@
       Succs[0] = true;
     }
   } else {
-    cerr << "SCCP: Don't know how to handle: " << TI;
+    std::cerr << "SCCP: Don't know how to handle: " << TI;
     Succs.assign(TI.getNumSuccessors(), true);
   }
 }
@@ -379,7 +376,7 @@
   
   // Check to make sure this edge itself is actually feasible now...
   TerminatorInst *FT = From->getTerminator();
-  std::vector<bool> SuccFeasible(FT->getNumSuccessors());
+  std::vector<bool> SuccFeasible;
   getFeasibleSuccessors(*FT, SuccFeasible);
 
   // Check all edges from From to To.  If any are feasible, return true.
@@ -409,10 +406,8 @@
 // 7. If a conditional branch has a value that is overdefined, make all
 //    successors executable.
 //
-
 void SCCP::visitPHINode(PHINode &PN) {
-  unsigned NumValues = PN.getNumIncomingValues(), i;
-  InstVal *OperandIV = 0;
+  if (getValueState(&PN).isOverdefined()) return;  // Quick exit
 
   // Look at all of the executable operands of the PHI node.  If any of them
   // are overdefined, the PHI becomes overdefined as well.  If they are all
@@ -420,24 +415,25 @@
   // constant.  If they are constant and don't agree, the PHI is overdefined.
   // If there are no executable operands, the PHI remains undefined.
   //
-  for (i = 0; i < NumValues; ++i) {
+  Constant *OperandVal = 0;
+  for (unsigned i = 0, e = PN.getNumIncomingValues(); i != e; ++i) {
+    InstVal &IV = getValueState(PN.getIncomingValue(i));
+    if (IV.isUndefined()) continue;  // Doesn't influence PHI node.
+    if (IV.isOverdefined()) {   // PHI node becomes overdefined!
+      markOverdefined(&PN);
+      return;
+    }
+    
     if (isEdgeFeasible(PN.getIncomingBlock(i), PN.getParent())) {
-      InstVal &IV = getValueState(PN.getIncomingValue(i));
-      if (IV.isUndefined()) continue;  // Doesn't influence PHI node.
-      if (IV.isOverdefined()) {   // PHI node becomes overdefined!
-        markOverdefined(&PN);
-        return;
-      }
-
-      if (OperandIV == 0) {   // Grab the first value...
-        OperandIV = &IV;
+      if (OperandVal == 0) {   // Grab the first value...
+        OperandVal = IV.getConstant();
       } else {                // Another value is being merged in!
         // There is already a reachable operand.  If we conflict with it,
         // then the PHI node becomes overdefined.  If we agree with it, we
         // can continue on.
-
+        
         // Check to see if there are two different constants merging...
-        if (IV.getConstant() != OperandIV->getConstant()) {
+        if (IV.getConstant() != OperandVal) {
           // Yes there is.  This means the PHI node is not constant.
           // You must be overdefined poor PHI.
           //
@@ -449,18 +445,16 @@
   }
 
   // If we exited the loop, this means that the PHI node only has constant
-  // arguments that agree with each other(and OperandIV is a pointer to one
-  // of their InstVal's) or OperandIV is null because there are no defined
-  // incoming arguments.  If this is the case, the PHI remains undefined.
-  //
-  if (OperandIV) {
-    assert(OperandIV->isConstant() && "Should only be here for constants!");
-    markConstant(&PN, OperandIV->getConstant());  // Aquire operand value
-  }
+  // arguments that agree with each other(and OperandVal is the constant) or
+  // OperandVal is null because there are no defined incoming arguments.  If
+  // this is the case, the PHI remains undefined.
+  //
+  if (OperandVal)
+    markConstant(&PN, OperandVal);      // Aquire operand value
 }
 
 void SCCP::visitTerminatorInst(TerminatorInst &TI) {
-  std::vector<bool> SuccFeasible(TI.getNumSuccessors());
+  std::vector<bool> SuccFeasible;
   getFeasibleSuccessors(TI, SuccFeasible);
 
   // Mark all feasible successors executable...
@@ -468,14 +462,6 @@
     if (SuccFeasible[i]) {
       BasicBlock *Succ = TI.getSuccessor(i);
       markExecutable(Succ);
-
-      // Visit all of the PHI nodes that merge values from this block...
-      // Because this edge may be new executable, and PHI nodes that used to be
-      // constant now may not be.
-      //
-      for (BasicBlock::iterator I = Succ->begin();
-           PHINode *PN = dyn_cast<PHINode>(I); ++I)
-        visitPHINode(*PN);
     }
 }
 




From lattner at cs.uiuc.edu  Thu Apr 24 21:51:07 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 21:51:07 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/TargetData.h
Message-ID: <200304250250.VAA11676@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

TargetData.h updated: 1.14 -> 1.15

---
Log message:

Default ctor doesn't provide name


---
Diffs of the changes:

Index: llvm/include/llvm/Target/TargetData.h
diff -u llvm/include/llvm/Target/TargetData.h:1.14 llvm/include/llvm/Target/TargetData.h:1.15
--- llvm/include/llvm/Target/TargetData.h:1.14	Thu Apr 24 14:08:45 2003
+++ llvm/include/llvm/Target/TargetData.h	Thu Apr 24 21:50:19 2003
@@ -38,7 +38,7 @@
  
   static Annotation *TypeAnFactory(AnnotationID, const Annotable *, void *);
 public:
-  TargetData(const std::string &TargetName = "Default Target",
+  TargetData(const std::string &TargetName = "",
              bool LittleEndian = false,
              unsigned char SubWordDataSize = 1,
              unsigned char IntRegSize = 8,




From lattner at cs.uiuc.edu  Thu Apr 24 21:51:11 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 21:51:11 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetData.cpp
Message-ID: <200304250250.VAA11686@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetData.cpp updated: 1.28 -> 1.29

---
Log message:

Allow for easy detection of when a "default" TargetData is created by the
passmanager, which is never the right thing to do.


---
Diffs of the changes:

Index: llvm/lib/Target/TargetData.cpp
diff -u llvm/lib/Target/TargetData.cpp:1.28 llvm/lib/Target/TargetData.cpp:1.29
--- llvm/lib/Target/TargetData.cpp:1.28	Thu Apr 24 14:09:05 2003
+++ llvm/lib/Target/TargetData.cpp	Thu Apr 24 21:50:45 2003
@@ -91,6 +91,13 @@
   : AID(AnnotationManager::getID("TargetData::" + TargetName)) {
   AnnotationManager::registerAnnotationFactory(AID, TypeAnFactory, this);
 
+  // If this assert triggers, a pass "required" TargetData information, but the
+  // top level tool did not provide once for it.  We do not want to default
+  // construct, or else we might end up using a bad endianness or pointer size!
+  //
+  assert(!TargetName.empty() &&
+         "ERROR: Tool did not specify a target data to use!");
+
   LittleEndian     = isLittleEndian;
   SubWordDataSize  = SubWordSize;
   IntegerRegSize   = IntRegSize;




From lattner at cs.uiuc.edu  Thu Apr 24 21:52:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 21:52:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/ConstantHandling.h
Message-ID: <200304250251.VAA11713@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm:

ConstantHandling.h updated: 1.24 -> 1.25

---
Log message:

*** Huge: Make constantexprs be handled correctly, conservatively


---
Diffs of the changes:

Index: llvm/include/llvm/ConstantHandling.h
diff -u llvm/include/llvm/ConstantHandling.h:1.24 llvm/include/llvm/ConstantHandling.h:1.25
--- llvm/include/llvm/ConstantHandling.h:1.24	Thu Apr 17 14:22:47 2003
+++ llvm/include/llvm/ConstantHandling.h	Thu Apr 24 21:51:46 2003
@@ -112,10 +112,13 @@
   // we just want to make sure to hit the cache instead of doing it indirectly,
   //  if possible...
   //
-  static inline ConstRules *get(const Constant &V) {
-    return (ConstRules*)V.getType()->getOrCreateAnnotation(AID);
+  static inline ConstRules *get(const Constant &V1, const Constant &V2) {
+    if (isa<ConstantExpr>(V1) || isa<ConstantExpr>(V2))
+      return getConstantExprRules();
+    return (ConstRules*)V1.getType()->getOrCreateAnnotation(AID);
   }
 private:
+  static ConstRules *getConstantExprRules();
   static Annotation *find(AnnotationID AID, const Annotable *Ty, void *);
 
   ConstRules(const ConstRules &);             // Do not implement
@@ -126,60 +129,60 @@
 // Standard binary operators...
 inline Constant *operator+(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->add(&V1, &V2);
+  return ConstRules::get(V1, V2)->add(&V1, &V2);
 }
 
 inline Constant *operator-(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->sub(&V1, &V2);
+  return ConstRules::get(V1, V2)->sub(&V1, &V2);
 }
 
 inline Constant *operator*(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->mul(&V1, &V2);
+  return ConstRules::get(V1, V2)->mul(&V1, &V2);
 }
 
 inline Constant *operator/(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->div(&V1, &V2);
+  return ConstRules::get(V1, V2)->div(&V1, &V2);
 }
 
 inline Constant *operator%(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->rem(&V1, &V2);
+  return ConstRules::get(V1, V2)->rem(&V1, &V2);
 }
 
 // Logical Operators...
 inline Constant *operator&(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->op_and(&V1, &V2);
+  return ConstRules::get(V1, V2)->op_and(&V1, &V2);
 }
 
 inline Constant *operator|(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->op_or(&V1, &V2);
+  return ConstRules::get(V1, V2)->op_or(&V1, &V2);
 }
 
 inline Constant *operator^(const Constant &V1, const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->op_xor(&V1, &V2);
+  return ConstRules::get(V1, V2)->op_xor(&V1, &V2);
 }
 
 // Shift Instructions...
 inline Constant *operator<<(const Constant &V1, const Constant &V2) {
   assert(V1.getType()->isInteger() && V2.getType() == Type::UByteTy);
-  return ConstRules::get(V1)->shl(&V1, &V2);
+  return ConstRules::get(V1, V2)->shl(&V1, &V2);
 }
 
 inline Constant *operator>>(const Constant &V1, const Constant &V2) {
   assert(V1.getType()->isInteger() && V2.getType() == Type::UByteTy);
-  return ConstRules::get(V1)->shr(&V1, &V2);
+  return ConstRules::get(V1, V2)->shr(&V1, &V2);
 }
 
 inline ConstantBool *operator<(const Constant &V1, 
                                const Constant &V2) {
   assert(V1.getType() == V2.getType() && "Constant types must be identical!");
-  return ConstRules::get(V1)->lessthan(&V1, &V2);
+  return ConstRules::get(V1, V2)->lessthan(&V1, &V2);
 }
 
 
@@ -194,12 +197,16 @@
 
 inline ConstantBool *operator>=(const Constant &V1, 
                                 const Constant &V2) {
-  return (V1 < V2)->inverted();      // !(V1 < V2)
+  if (ConstantBool *V = (V1 < V2))
+    return V->inverted();                // !(V1 < V2)
+  return 0;
 }
 
 inline ConstantBool *operator<=(const Constant &V1, 
                                 const Constant &V2) {
-  return (V1 > V2)->inverted();      // !(V1 > V2)
+  if (ConstantBool *V = (V1 > V2))
+    return V->inverted();                // !(V1 > V2)
+  return 0;
 }
 
 




From lattner at cs.uiuc.edu  Thu Apr 24 21:53:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 21:53:00 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/ConstantHandling.cpp
Message-ID: <200304250252.VAA11730@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

ConstantHandling.cpp updated: 1.25 -> 1.26

---
Log message:

** CORRECTLY, conservatively handle constant expressions


---
Diffs of the changes:

Index: llvm/lib/VMCore/ConstantHandling.cpp
diff -u llvm/lib/VMCore/ConstantHandling.cpp:1.25 llvm/lib/VMCore/ConstantHandling.cpp:1.26
--- llvm/lib/VMCore/ConstantHandling.cpp:1.25	Thu Apr 17 14:24:18 2003
+++ llvm/lib/VMCore/ConstantHandling.cpp	Thu Apr 24 21:52:06 2003
@@ -47,7 +47,7 @@
 
   switch (I->getOpcode()) {
   case Instruction::Cast:
-    return ConstRules::get(*Op0)->castTo(Op0, I->getType());
+    return ConstRules::get(*Op0, *Op0)->castTo(Op0, I->getType());
   case Instruction::Add:     return *Op0 + *Op1;
   case Instruction::Sub:     return *Op0 - *Op1;
   case Instruction::Mul:     return *Op0 * *Op1;
@@ -107,7 +107,7 @@
       }
     }
 
-  return ConstRules::get(*V)->castTo(V, DestTy);
+  return ConstRules::get(*V, *V)->castTo(V, DestTy);
 }
 
 Constant *ConstantFoldBinaryInstruction(unsigned Opcode, const Constant *V1,
@@ -554,4 +554,9 @@
   default:
     return new EmptyRules();
   }
+}
+
+ConstRules *ConstRules::getConstantExprRules() {
+  static EmptyRules CERules;
+  return &CERules;
 }




From lattner at cs.uiuc.edu  Thu Apr 24 22:17:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 22:17:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ListReducer.h
Message-ID: <200304250316.WAA12120@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ListReducer.h updated: 1.2 -> 1.3

---
Log message:

Rename Kept -> Suffix
FIX problem where we were incorrectly putting the prefix of the list into the "suffix" list.


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ListReducer.h
diff -u llvm/tools/bugpoint/ListReducer.h:1.2 llvm/tools/bugpoint/ListReducer.h:1.3
--- llvm/tools/bugpoint/ListReducer.h:1.2	Thu Apr 24 17:23:56 2003
+++ llvm/tools/bugpoint/ListReducer.h	Thu Apr 24 22:16:33 2003
@@ -34,14 +34,14 @@
     unsigned MidTop = TheList.size();
     while (MidTop > 1) {
       unsigned Mid = MidTop / 2;
-      std::vector<ElTy> Prefix(TheList.begin()+Mid, TheList.end());
-      std::vector<ElTy> Kept  (TheList.begin(), TheList.begin()+Mid);
+      std::vector<ElTy> Prefix(TheList.begin(), TheList.begin()+Mid);
+      std::vector<ElTy> Suffix(TheList.begin()+Mid, TheList.end());
 
-      switch (doTest(Prefix, Kept)) {
+      switch (doTest(Prefix, Suffix)) {
       case KeepSuffix:
         // The property still holds.  We can just drop the prefix elements, and
         // shorten the list to the "kept" elements.
-        TheList.swap(Kept);
+        TheList.swap(Suffix);
         MidTop = TheList.size();
         break;
       case KeepPrefix:




From lattner at cs.uiuc.edu  Thu Apr 24 22:17:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 22:17:06 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/Miscompilation.cpp
Message-ID: <200304250316.WAA12125@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

Miscompilation.cpp updated: 1.6 -> 1.7

---
Log message:

Rename Kept -> Suffix
Fix problem where we accidentally returned KeepPrefix instead of KeepSuffix!


---
Diffs of the changes:

Index: llvm/tools/bugpoint/Miscompilation.cpp
diff -u llvm/tools/bugpoint/Miscompilation.cpp:1.6 llvm/tools/bugpoint/Miscompilation.cpp:1.7
--- llvm/tools/bugpoint/Miscompilation.cpp:1.6	Thu Apr 24 17:24:22 2003
+++ llvm/tools/bugpoint/Miscompilation.cpp	Thu Apr 24 22:16:05 2003
@@ -38,14 +38,14 @@
 
 ReduceMiscompilingPasses::TestResult
 ReduceMiscompilingPasses::doTest(std::vector<const PassInfo*> &Prefix,
-                                 std::vector<const PassInfo*> &Kept) {
-  // First, run the program with just the Kept passes.  If it is still broken
+                                 std::vector<const PassInfo*> &Suffix) {
+  // First, run the program with just the Suffix passes.  If it is still broken
   // with JUST the kept passes, discard the prefix passes.
-  std::cout << "Checking to see if '" << getPassesString(Kept)
+  std::cout << "Checking to see if '" << getPassesString(Suffix)
             << "' compile correctly: ";
 
   std::string BytecodeResult;
-  if (BD.runPasses(Kept, BytecodeResult, false/*delete*/, true/*quiet*/)) {
+  if (BD.runPasses(Suffix, BytecodeResult, false/*delete*/, true/*quiet*/)) {
     std::cerr << BD.getToolName() << ": Error running this sequence of passes"
               << " on the input program!\n";
     exit(1);
@@ -60,8 +60,8 @@
 
   if (Prefix.empty()) return NoFailure;
 
-  // First, run the program with just the Kept passes.  If it is still broken
-  // with JUST the kept passes, discard the prefix passes.
+  // Next, see if the program is broken if we run the "prefix" passes first,
+  // then seperately run the "kept" passes.
   std::cout << "Checking to see if '" << getPassesString(Prefix)
             << "' compile correctly: ";
 
@@ -96,13 +96,13 @@
   }
   removeFile(BytecodeResult);  // No longer need the file on disk
     
-  std::cout << "Checking to see if '" << getPassesString(Kept)
+  std::cout << "Checking to see if '" << getPassesString(Suffix)
             << "' passes compile correctly after the '"
             << getPassesString(Prefix) << "' passes: ";
 
   Module *OriginalInput = BD.Program;
   BD.Program = PrefixOutput;
-  if (BD.runPasses(Kept, BytecodeResult, false/*delete*/, true/*quiet*/)) {
+  if (BD.runPasses(Suffix, BytecodeResult, false/*delete*/, true/*quiet*/)) {
     std::cerr << BD.getToolName() << ": Error running this sequence of passes"
               << " on the input program!\n";
     exit(1);
@@ -112,7 +112,7 @@
   if (BD.diffProgram(Output, BytecodeResult, true/*delete bytecode*/)) {
     std::cout << "nope.\n";
     delete OriginalInput;     // We pruned down the original input...
-    return KeepPrefix;
+    return KeepSuffix;
   }
 
   // Otherwise, we must not be running the bad pass anymore.
@@ -136,8 +136,8 @@
   ReduceMiscompilingFunctions(BugDriver &bd) : BD(bd) {}
 
   virtual TestResult doTest(std::vector<Function*> &Prefix,
-                            std::vector<Function*> &Kept) {
-    if (TestFuncs(Kept, false))
+                            std::vector<Function*> &Suffix) {
+    if (TestFuncs(Suffix, false))
       return KeepSuffix;
     if (!Prefix.empty() && TestFuncs(Prefix, false))
       return KeepPrefix;




From lattner at cs.uiuc.edu  Thu Apr 24 22:36:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 22:36:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/SCCP.cpp
Message-ID: <200304250335.WAA12336@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

SCCP.cpp updated: 1.66 -> 1.67

---
Log message:

Fix major problem with PHI node traversal: rechecking PHIs should go into
the visit function.  The worklist is for when their lattice value changes.


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/SCCP.cpp
diff -u llvm/lib/Transforms/Scalar/SCCP.cpp:1.66 llvm/lib/Transforms/Scalar/SCCP.cpp:1.67
--- llvm/lib/Transforms/Scalar/SCCP.cpp:1.66	Thu Apr 24 21:50:03 2003
+++ llvm/lib/Transforms/Scalar/SCCP.cpp	Thu Apr 24 22:35:10 2003
@@ -162,7 +162,7 @@
       // can be rechecked.
       for (BasicBlock::iterator I = BB->begin();
            PHINode *PN = dyn_cast<PHINode>(I); ++I)
-        InstWorkList.push_back(PN);
+        visitPHINode(*PN);
 
     } else {
       DEBUG(std::cerr << "Marking BB Executable: " << *BB);




From lattner at cs.uiuc.edu  Thu Apr 24 22:57:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 22:57:00 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/Makefile.programs
Message-ID: <200304250356.WAA12643@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs:

Makefile.programs updated: 1.36 -> 1.37

---
Log message:

Add targets to make it trivially easy to run bugpoint on a program


---
Diffs of the changes:

Index: llvm/test/Programs/Makefile.programs
diff -u llvm/test/Programs/Makefile.programs:1.36 llvm/test/Programs/Makefile.programs:1.37
--- llvm/test/Programs/Makefile.programs:1.36	Fri Feb 28 13:39:49 2003
+++ llvm/test/Programs/Makefile.programs	Thu Apr 24 22:56:37 2003
@@ -164,6 +164,23 @@
 Output/%.llvm: Output/%.linked.bc
 	$(LGCCLD) $(STATS) $< -lgcc -lc $(LIBS) -o Output/$*.llvm
 
+# Targets to get the pass arguments that gccas and gccld are using...
+Output/gccas-pass-args: $(LGCCAS)
+	$(LGCCAS) /dev/null -o /dev/null -debug-pass=Arguments > $@.1 2>&1
+	sed 's/Pass Arguments: //' < $@.1 > $@
+Output/gccld-pass-args: $(LGCCLD)
+	$(LGCCLD) /dev/null -o /dev/null -debug-pass=Arguments > $@.1 2>&1
+	sed 's/Pass Arguments: //' < $@.1 > $@
+
+
+# Rules to bugpoint the GCCAS or GCCLD command...
+$(PROGRAMS_TO_TEST:%=Output/%.bugpoint-gccas): \
+Output/%.bugpoint-gccas: Output/%.linked.rll $(LBUGPOINT) Output/gccas-pass-args
+	$(LBUGPOINT) $< `cat Output/gccas-pass-args` $(BUGPOINT_OPTIONS)
+$(PROGRAMS_TO_TEST:%=Output/%.bugpoint-gccld): \
+Output/%.bugpoint-gccld: Output/%.linked.bc $(LBUGPOINT) Output/gccld-pass-args
+	$(LBUGPOINT) $< `cat Output/gccld-pass-args` $(BUGPOINT_OPTIONS)
+
 #
 # Rules to compile the program for the C Back End
 #
@@ -179,6 +196,7 @@
 RUN_OPTIONS += 2>&1
 ifdef INPUT_FILENAME
 RUN_OPTIONS += < $(INPUT_FILENAME)
+BUGPOINT_OPTIONS += -input $(INPUT_FILENAME)
 endif
 
 # Rules to build the test output...




From lattner at cs.uiuc.edu  Thu Apr 24 22:58:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 22:58:00 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/MultiSource/Makefile.multisrc
Message-ID: <200304250357.WAA12655@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/MultiSource:

Makefile.multisrc updated: 1.25 -> 1.26

---
Log message:

Make it _super_ easy to run bugpoint on multisource programs


---
Diffs of the changes:

Index: llvm/test/Programs/MultiSource/Makefile.multisrc
diff -u llvm/test/Programs/MultiSource/Makefile.multisrc:1.25 llvm/test/Programs/MultiSource/Makefile.multisrc:1.26
--- llvm/test/Programs/MultiSource/Makefile.multisrc:1.25	Fri Feb 14 18:30:42 2003
+++ llvm/test/Programs/MultiSource/Makefile.multisrc	Thu Apr 24 22:57:08 2003
@@ -34,6 +34,9 @@
 Output/%.o: %.c Output/.dir
 	$(CC) $(CPPFLAGS) -O $(CFLAGS) -c $< -o $@
 
+bugpoint-gccas: Output/$(PROG).bugpoint-gccas
+bugpoint-gccld: Output/$(PROG).bugpoint-gccld
+
 # Raw bytecode files are files created by simply linking the output of the GCC
 # frontend without running gccas.
 #




From lattner at cs.uiuc.edu  Thu Apr 24 23:22:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Thu Apr 24 23:22:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Execution.cpp
Message-ID: <200304250421.XAA15150@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Execution.cpp updated: 1.83 -> 1.84

---
Log message:

Fix problem where we would read 64 bits worth of pointer information, even on 32 bit targets!


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Execution.cpp
diff -u llvm/tools/lli/Interpreter/Execution.cpp:1.83 llvm/tools/lli/Interpreter/Execution.cpp:1.84
--- llvm/tools/lli/Interpreter/Execution.cpp:1.83	Wed Apr 23 14:55:35 2003
+++ llvm/tools/lli/Interpreter/Execution.cpp	Thu Apr 24 23:21:19 2003
@@ -6,12 +6,8 @@
 
 #include "Interpreter.h"
 #include "ExecutionAnnotations.h"
-#include "llvm/GlobalVariable.h"
-#include "llvm/Function.h"
-#include "llvm/iPHINode.h"
-#include "llvm/iOther.h"
-#include "llvm/iTerminators.h"
-#include "llvm/iMemory.h"
+#include "llvm/Module.h"
+#include "llvm/Instructions.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/Constants.h"
 #include "llvm/Assembly/Writer.h"
@@ -776,6 +772,7 @@
     case Type::ShortTyID:   Result.UShortVal = (unsigned)Ptr->Untyped[0] |
                                               ((unsigned)Ptr->Untyped[1] << 8);
                             break;
+    Load4BytesLittleEndian:                            
     case Type::FloatTyID:
     case Type::UIntTyID:
     case Type::IntTyID:     Result.UIntVal = (unsigned)Ptr->Untyped[0] |
@@ -783,10 +780,11 @@
                                             ((unsigned)Ptr->Untyped[2] << 16) |
                                             ((unsigned)Ptr->Untyped[3] << 24);
                             break;
+    case Type::PointerTyID: if (getModule().has32BitPointers())
+                              goto Load4BytesLittleEndian;
     case Type::DoubleTyID:
     case Type::ULongTyID:
-    case Type::LongTyID:    
-    case Type::PointerTyID: Result.ULongVal = (uint64_t)Ptr->Untyped[0] |
+    case Type::LongTyID:    Result.ULongVal = (uint64_t)Ptr->Untyped[0] |
                                              ((uint64_t)Ptr->Untyped[1] <<  8) |
                                              ((uint64_t)Ptr->Untyped[2] << 16) |
                                              ((uint64_t)Ptr->Untyped[3] << 24) |
@@ -808,6 +806,7 @@
     case Type::ShortTyID:   Result.UShortVal = (unsigned)Ptr->Untyped[1] |
                                               ((unsigned)Ptr->Untyped[0] << 8);
                             break;
+    Load4BytesBigEndian:
     case Type::FloatTyID:
     case Type::UIntTyID:
     case Type::IntTyID:     Result.UIntVal = (unsigned)Ptr->Untyped[3] |
@@ -815,10 +814,11 @@
                                             ((unsigned)Ptr->Untyped[1] << 16) |
                                             ((unsigned)Ptr->Untyped[0] << 24);
                             break;
+    case Type::PointerTyID: if (getModule().has32BitPointers())
+                              goto Load4BytesBigEndian;
     case Type::DoubleTyID:
     case Type::ULongTyID:
-    case Type::LongTyID:    
-    case Type::PointerTyID: Result.ULongVal = (uint64_t)Ptr->Untyped[7] |
+    case Type::LongTyID:    Result.ULongVal = (uint64_t)Ptr->Untyped[7] |
                                              ((uint64_t)Ptr->Untyped[6] <<  8) |
                                              ((uint64_t)Ptr->Untyped[5] << 16) |
                                              ((uint64_t)Ptr->Untyped[4] << 24) |




From lattner at cs.uiuc.edu  Fri Apr 25 00:23:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 00:23:01 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304250522.AAA20605@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.66 -> 1.67

---
Log message:

Make sure to add a targetdata instance to the passmanager, and make it match the
one in use by the TargetMachine


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.66 llvm/tools/llc/llc.cpp:1.67
--- llvm/tools/llc/llc.cpp:1.66	Thu Apr 24 13:36:39 2003
+++ llvm/tools/llc/llc.cpp	Fri Apr 25 00:22:29 2003
@@ -173,6 +173,7 @@
   assert(target.get() && "Could not allocate target machine!");
 
   TargetMachine &Target = *target.get();
+  const TargetData &TD = Target.getTargetData();
 
   // Load the module to be compiled...
   std::auto_ptr<Module> M(ParseBytecodeFile(InputFilename));
@@ -184,6 +185,10 @@
 
   // Build up all of the passes that we want to do to the module...
   PassManager Passes;
+
+  Passes.add(new TargetData("llc", TD.isLittleEndian(), TD.getSubWordDataSize(),
+                            TD.getIntegerRegSize(), TD.getPointerSize(),
+                            TD.getPointerAlignment()));
 
   // Create a new optimization pass for each one specified on the command line
   // Deal specially with tracing passes, which must be run differently than opt.




From lattner at cs.uiuc.edu  Fri Apr 25 00:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 00:24:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/Sparc/SparcInstrInfo.cpp SparcInstrSelection.cpp
Message-ID: <200304250523.AAA20653@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/Sparc:

SparcInstrInfo.cpp updated: 1.38 -> 1.39
SparcInstrSelection.cpp updated: 1.86 -> 1.87

---
Log message:

Fix obvious type-o


---
Diffs of the changes:

Index: llvm/lib/Target/Sparc/SparcInstrInfo.cpp
diff -u llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.38 llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.39
--- llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.38	Wed Jan 15 15:36:50 2003
+++ llvm/lib/Target/Sparc/SparcInstrInfo.cpp	Fri Apr 25 00:23:09 2003
@@ -443,7 +443,7 @@
 
           if (opSize > destSize ||
               (val->getType()->isSigned()
-               && destSize < target.getTargetData().getIntegerRegize()))
+               && destSize < target.getTargetData().getIntegerRegSize()))
             { // operand is larger than dest,
               //    OR both are equal but smaller than the full register size
               //       AND operand is signed, so it may have extra sign bits:


Index: llvm/lib/Target/Sparc/SparcInstrSelection.cpp
diff -u llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.86 llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.87
--- llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.86	Wed Jan 15 15:36:50 2003
+++ llvm/lib/Target/Sparc/SparcInstrSelection.cpp	Fri Apr 25 00:23:10 2003
@@ -759,7 +759,7 @@
   Value* shiftDest = destVal;
   unsigned opSize = target.getTargetData().getTypeSize(argVal1->getType());
   if ((shiftOpCode == SLL || shiftOpCode == SLLX)
-      && opSize < target.getTargetData().getIntegerRegize())
+      && opSize < target.getTargetData().getIntegerRegSize())
     { // put SLL result into a temporary
       shiftDest = new TmpInstruction(argVal1, optArgVal2, "sllTmp");
       mcfi.addTemp(shiftDest);
@@ -2305,7 +2305,7 @@
                                  .addReg(dest, MOTy::Def);
               mvec.push_back(M);
             }
-          else if (destSize < target.getTargetData().getIntegerRegize())
+          else if (destSize < target.getTargetData().getIntegerRegSize())
             assert(0 && "Unsupported type size: 32 < size < 64 bits");
         }
     }




From lattner at cs.uiuc.edu  Fri Apr 25 00:24:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 00:24:06 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/TargetData.h
Message-ID: <200304250523.AAA20674@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

TargetData.h updated: 1.15 -> 1.16

---
Log message:

Fix method name type-o


---
Diffs of the changes:

Index: llvm/include/llvm/Target/TargetData.h
diff -u llvm/include/llvm/Target/TargetData.h:1.15 llvm/include/llvm/Target/TargetData.h:1.16
--- llvm/include/llvm/Target/TargetData.h:1.15	Thu Apr 24 21:50:19 2003
+++ llvm/include/llvm/Target/TargetData.h	Fri Apr 25 00:23:27 2003
@@ -64,7 +64,7 @@
   unsigned char getDoubleAlignment()  const { return  DoubleAlignment; }
   unsigned char getPointerAlignment() const { return PointerAlignment; }
   unsigned char getPointerSize()      const { return      PointerSize; }
-  unsigned char getIntegerRegize()    const { return   IntegerRegSize; }
+  unsigned char getIntegerRegSize()   const { return   IntegerRegSize; }
   AnnotationID  getStructLayoutAID()  const { return AID; }
 
   // getTypeSize - Return the number of bytes neccesary to hold the specified




From lattner at cs.uiuc.edu  Fri Apr 25 00:27:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 00:27:01 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304250526.AAA20725@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.67 -> 1.68

---
Log message:

Kill using declarations


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.67 llvm/tools/llc/llc.cpp:1.68
--- llvm/tools/llc/llc.cpp:1.67	Fri Apr 25 00:22:29 2003
+++ llvm/tools/llc/llc.cpp	Fri Apr 25 00:26:11 2003
@@ -21,8 +21,6 @@
 #include "Support/Signals.h"
 #include <memory>
 #include <fstream>
-using std::string;
-using std::cerr;
 
 //------------------------------------------------------------------------------
 // Option declarations for LLC.
@@ -43,10 +41,10 @@
 // within the corresponding llc passes, and target-specific options
 // and back-end code generation options are specified with the target machine.
 // 
-static cl::opt<string>
+static cl::opt<std::string>
 InputFilename(cl::Positional, cl::desc("<input bytecode>"), cl::init("-"));
 
-static cl::opt<string>
+static cl::opt<std::string>
 OutputFilename("o", cl::desc("Output filename"), cl::value_desc("filename"));
 
 static cl::opt<bool> Force("f", cl::desc("Overwrite output files"));
@@ -55,7 +53,7 @@
 DumpAsm("d", cl::desc("Print bytecode before native code generation"),
         cl::Hidden);
 
-static cl::opt<string>
+static cl::opt<std::string>
 TraceLibPath("tracelibpath", cl::desc("Path to libinstr for trace code"),
              cl::value_desc("directory"), cl::Hidden);
 
@@ -66,14 +64,14 @@
 
 
 // GetFileNameRoot - Helper function to get the basename of a filename...
-static inline string
-GetFileNameRoot(const string &InputFilename)
+static inline std::string
+GetFileNameRoot(const std::string &InputFilename)
 {
-  string IFN = InputFilename;
-  string outputFilename;
+  std::string IFN = InputFilename;
+  std::string outputFilename;
   int Len = IFN.length();
   if (IFN[Len-3] == '.' && IFN[Len-2] == 'b' && IFN[Len-1] == 'c') {
-    outputFilename = string(IFN.begin(), IFN.end()-3); // s/.bc/.s/
+    outputFilename = std::string(IFN.begin(), IFN.end()-3); // s/.bc/.s/
   } else {
     outputFilename = IFN;
   }
@@ -116,7 +114,7 @@
 
   // If we still didn't get it, cancel trying to link it in...
   if (TraceModule == 0)
-    cerr << "Warning, could not load trace routines to link into program!\n";
+    std::cerr <<"WARNING: couldn't load trace routines to link into program!\n";
   else
     {
       // Link in the trace routines... if this fails, don't panic, because the
@@ -124,23 +122,23 @@
       //
       std::auto_ptr<Module> TraceRoutines(TraceModule);
       if (LinkModules(&M, TraceRoutines.get(), &ErrorMessage))
-        cerr << "Warning: Error linking in trace routines: "
-             << ErrorMessage << "\n";
+        std::cerr << "WARNING: Error linking in trace routines: "
+                  << ErrorMessage << "\n";
     }
 
   // Write out the module with tracing code just before code generation
   assert (InputFilename != "-"
           && "Cannot write out traced bytecode when reading input from stdin");
-  string TraceFilename = GetFileNameRoot(InputFilename) + ".trace.bc";
+  std::string TraceFilename = GetFileNameRoot(InputFilename) + ".trace.bc";
 
   std::ofstream Out(TraceFilename.c_str());
   if (!Out.good())
-    cerr << "Error opening '" << TraceFilename
-         << "'!: Skipping output of trace code as bytecode\n";
+    std::cerr << "Error opening '" << TraceFilename
+              << "'!: Skipping output of trace code as bytecode\n";
   else
     {
-      cerr << "Emitting trace code to '" << TraceFilename
-           << "' for comparison...\n";
+      std::cerr << "Emitting trace code to '" << TraceFilename
+                << "' for comparison...\n";
       WriteBytecodeToFile(&M, Out);
     }
 
@@ -179,7 +177,7 @@
   std::auto_ptr<Module> M(ParseBytecodeFile(InputFilename));
   if (M.get() == 0)
     {
-      cerr << argv[0] << ": bytecode didn't read correctly.\n";
+      std::cerr << argv[0] << ": bytecode didn't read correctly.\n";
       return 1;
     }
 
@@ -208,8 +206,8 @@
           else if (Opt->getTargetCtor())
             Passes.add(Opt->getTargetCtor()(Target));
           else
-            cerr << argv[0] << ": cannot create pass: "
-                 << Opt->getPassName() << "\n";
+            std::cerr << argv[0] << ": cannot create pass: "
+                      << Opt->getPassName() << "\n";
         }
     }
 
@@ -227,7 +225,7 @@
 
   // If LLVM dumping after transformations is requested, add it to the pipeline
   if (DumpAsm)
-    Passes.add(new PrintFunctionPass("Code after xformations: \n", &cerr));
+    Passes.add(new PrintFunctionPass("Code after xformations: \n", &std::cerr));
 
   // Strip all of the symbols from the bytecode so that it will be smaller...
   Passes.add(createSymbolStrippingPass());
@@ -238,9 +236,9 @@
     {   // Specified an output filename?
       if (!Force && std::ifstream(OutputFilename.c_str())) {
         // If force is not specified, make sure not to overwrite a file!
-        cerr << argv[0] << ": error opening '" << OutputFilename
-             << "': file exists!\n"
-             << "Use -f command line argument to force output\n";
+        std::cerr << argv[0] << ": error opening '" << OutputFilename
+                  << "': file exists!\n"
+                  << "Use -f command line argument to force output\n";
         return 1;
       }
       Out = new std::ofstream(OutputFilename.c_str());
@@ -258,22 +256,23 @@
         }
       else
         {
-          string OutputFilename = GetFileNameRoot(InputFilename); 
+          std::string OutputFilename = GetFileNameRoot(InputFilename); 
           OutputFilename += ".s";
 
           if (!Force && std::ifstream(OutputFilename.c_str()))
             {
               // If force is not specified, make sure not to overwrite a file!
-              cerr << argv[0] << ": error opening '" << OutputFilename
-                   << "': file exists!\n"
-                   << "Use -f command line argument to force output\n";
+              std::cerr << argv[0] << ": error opening '" << OutputFilename
+                        << "': file exists!\n"
+                        << "Use -f command line argument to force output\n";
               return 1;
             }
 
           Out = new std::ofstream(OutputFilename.c_str());
           if (!Out->good())
             {
-              cerr << argv[0] << ": error opening " << OutputFilename << "!\n";
+              std::cerr << argv[0] << ": error opening " << OutputFilename
+                        << "!\n";
               delete Out;
               return 1;
             }
@@ -286,8 +285,8 @@
 
   // Ask the target to add backend passes as neccesary
   if (Target.addPassesToEmitAssembly(Passes, *Out)) {
-    cerr << argv[0] << ": target '" << Target.getName()
-         << " does not support static compilation!\n";
+    std::cerr << argv[0] << ": target '" << Target.getName()
+              << " does not support static compilation!\n";
   } else {
     // Run our queue of passes all at once now, efficiently.
     Passes.run(*M.get());




From lattner at cs.uiuc.edu  Fri Apr 25 01:07:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 01:07:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86TargetMachine.cpp
Message-ID: <200304250606.BAA25662@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86TargetMachine.cpp updated: 1.15 -> 1.16

---
Log message:

Fix compatibility bug: X86 aligns doubles to 4 bytes, not 8!


---
Diffs of the changes:

Index: llvm/lib/Target/X86/X86TargetMachine.cpp
diff -u llvm/lib/Target/X86/X86TargetMachine.cpp:1.15 llvm/lib/Target/X86/X86TargetMachine.cpp:1.16
--- llvm/lib/Target/X86/X86TargetMachine.cpp:1.15	Wed Apr 23 11:24:55 2003
+++ llvm/lib/Target/X86/X86TargetMachine.cpp	Fri Apr 25 01:05:57 2003
@@ -37,6 +37,7 @@
 		  (Config & TM::EndianMask) == TM::LittleEndian,
 		  1, 4, 
 		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
+		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4),
   FrameInfo(TargetFrameInfo::StackGrowsDown, 8/*16 for SSE*/, 4) {
 }




From lattner at cs.uiuc.edu  Fri Apr 25 01:07:08 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 01:07:08 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304250606.BAA25671@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.68 -> 1.69

---
Log message:

Make sure to copy double alignment as well


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.68 llvm/tools/llc/llc.cpp:1.69
--- llvm/tools/llc/llc.cpp:1.68	Fri Apr 25 00:26:11 2003
+++ llvm/tools/llc/llc.cpp	Fri Apr 25 01:06:13 2003
@@ -186,7 +186,7 @@
 
   Passes.add(new TargetData("llc", TD.isLittleEndian(), TD.getSubWordDataSize(),
                             TD.getIntegerRegSize(), TD.getPointerSize(),
-                            TD.getPointerAlignment()));
+                            TD.getPointerAlignment(), TD.getDoubleAlignment()));
 
   // Create a new optimization pass for each one specified on the command line
   // Deal specially with tracing passes, which must be run differently than opt.




From lattner at cs.uiuc.edu  Fri Apr 25 01:07:14 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 01:07:14 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetData.cpp
Message-ID: <200304250606.BAA25682@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetData.cpp updated: 1.29 -> 1.30

---
Log message:

Whoops, 32 bit targets have _4_ byte pointers, not _32_ byte pointers!


---
Diffs of the changes:

Index: llvm/lib/Target/TargetData.cpp
diff -u llvm/lib/Target/TargetData.cpp:1.29 llvm/lib/Target/TargetData.cpp:1.30
--- llvm/lib/Target/TargetData.cpp:1.29	Thu Apr 24 21:50:45 2003
+++ llvm/lib/Target/TargetData.cpp	Fri Apr 25 01:06:43 2003
@@ -104,6 +104,8 @@
   PointerSize      = PtrSize;
   PointerAlignment = PtrAl;
   DoubleAlignment  = DoubleAl;
+  assert(DoubleAlignment == PtrAl &&
+         "Double alignment and pointer alignment agree for now!");
   FloatAlignment   = FloatAl;
   LongAlignment    = LongAl;
   IntAlignment     = IntAl;
@@ -118,9 +120,9 @@
   LittleEndian     = M->isLittleEndian();
   SubWordDataSize  = 1;
   IntegerRegSize   = 8;
-  PointerSize      = M->has32BitPointers() ? 32 : 64;
+  PointerSize      = M->has32BitPointers() ? 4 : 8;
   PointerAlignment = PointerSize;
-  DoubleAlignment  = 8;
+  DoubleAlignment  = PointerSize;
   FloatAlignment   = 4;
   LongAlignment    = 8;
   IntAlignment     = 4;




From lattner at cs.uiuc.edu  Fri Apr 25 01:16:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 01:16:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Interpreter.cpp
Message-ID: <200304250615.BAA27068@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Interpreter.cpp updated: 1.1 -> 1.2

---
Log message:

MAke sure that LLI properly configures align_of(double)


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Interpreter.cpp
diff -u llvm/tools/lli/Interpreter/Interpreter.cpp:1.1 llvm/tools/lli/Interpreter/Interpreter.cpp:1.2
--- llvm/tools/lli/Interpreter/Interpreter.cpp:1.1	Mon Dec 23 18:01:02 2002
+++ llvm/tools/lli/Interpreter/Interpreter.cpp	Fri Apr 25 01:15:05 2003
@@ -27,6 +27,7 @@
     CurFrame(-1), TD("lli", (Config & TM::EndianMask) == TM::LittleEndian,
 		     1, 4,
 		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
+		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4) {
 
   setTargetData(TD);




From lattner at cs.uiuc.edu  Fri Apr 25 13:00:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:00:00 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/BasicAA/2003-04-25-GEPCrash.ll
Message-ID: <200304251759.MAA15270@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/BasicAA:

2003-04-25-GEPCrash.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/BasicAA/2003-04-25-GEPCrash.ll
diff -c /dev/null llvm/test/Regression/Transforms/BasicAA/2003-04-25-GEPCrash.ll:1.1
*** /dev/null	Fri Apr 25 12:59:12 2003
--- llvm/test/Regression/Transforms/BasicAA/2003-04-25-GEPCrash.ll	Fri Apr 25 12:59:02 2003
***************
*** 0 ****
--- 1,7 ----
+ ; RUN: as < %s | opt -basicaa -aa-eval -disable-output
+ ; Test for a bug in BasicAA which caused a crash when querying equality of P1&P2
+ void %test([17 x ushort]* %mask_bits) {
+ 	%P1 = getelementptr [17 x ushort]* %mask_bits, long 0, long 0
+ 	%P2 = getelementptr [17 x ushort]* %mask_bits, long 252645134, long 0
+ 	ret void
+ }




From lattner at cs.uiuc.edu  Fri Apr 25 13:04:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:04:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/BasicAliasAnalysis.cpp
Message-ID: <200304251803.NAA15400@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

BasicAliasAnalysis.cpp updated: 1.7 -> 1.8

---
Log message:

Fix BasicAA/2003-04-25-GEPCrash.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/BasicAliasAnalysis.cpp
diff -u llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.7 llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.8
--- llvm/lib/Analysis/BasicAliasAnalysis.cpp:1.7	Tue Apr 22 17:00:27 2003
+++ llvm/lib/Analysis/BasicAliasAnalysis.cpp	Fri Apr 25 13:03:06 2003
@@ -294,11 +294,11 @@
     }
   }
   
-  int Offset1 = getTargetData().getIndexedOffset(GEPPointerTy, Indices1);
-  int Offset2 = getTargetData().getIndexedOffset(GEPPointerTy, Indices2);
+  int64_t Offset1 = getTargetData().getIndexedOffset(GEPPointerTy, Indices1);
+  int64_t Offset2 = getTargetData().getIndexedOffset(GEPPointerTy, Indices2);
   assert(Offset1 < Offset2 &&"There is at least one different constant here!");
 
-  if ((unsigned)(Offset2-Offset1) >= SizeMax) {
+  if ((uint64_t)(Offset2-Offset1) >= SizeMax) {
     //std::cerr << "Determined that these two GEP's don't alias [" 
     //          << SizeMax << " bytes]: \n" << *GEP1 << *GEP2;
     return NoAlias;




From lattner at cs.uiuc.edu  Fri Apr 25 13:24:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:24:00 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304251823.NAA17640@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.47 -> 1.48

---
Log message:

Add __strdup


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.47 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.48
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.47	Wed Apr 23 15:23:16 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Fri Apr 25 13:23:38 2003
@@ -565,6 +565,12 @@
   return Ret;
 }
 
+// char *__strdup(const char *src);
+GenericValue lle_X___strdup(FunctionType *M, const vector<GenericValue> &Args) {
+  assert(Args.size() == 1);
+  return PTOGV(strdup((char*)GVTOP(Args[0])));
+}
+
 // void *memset(void *S, int C, size_t N)
 GenericValue lle_X_memset(FunctionType *M, const vector<GenericValue> &Args) {
   assert(Args.size() == 3);
@@ -803,6 +809,7 @@
   FuncNames["lle_X_strcat"]       = lle_X_strcat;
   FuncNames["lle_X_strcpy"]       = lle_X_strcpy;
   FuncNames["lle_X_strlen"]       = lle_X_strlen;
+  FuncNames["lle_X___strdup"]       = lle_X___strdup;
   FuncNames["lle_X_memset"]       = lle_X_memset;
   FuncNames["lle_X_memcpy"]       = lle_X_memcpy;
 




From lattner at cs.uiuc.edu  Fri Apr 25 13:29:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:29:01 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
Message-ID: <200304251828.NAA18150@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

ExternalFunctions.cpp updated: 1.48 -> 1.49

---
Log message:

Only do the %ld -> %lld promotion when running a 64 bit bytecode on a 32 bit host


---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/ExternalFunctions.cpp
diff -u llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.48 llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.49
--- llvm/tools/lli/Interpreter/ExternalFunctions.cpp:1.48	Fri Apr 25 13:23:38 2003
+++ llvm/tools/lli/Interpreter/ExternalFunctions.cpp	Fri Apr 25 13:28:44 2003
@@ -377,7 +377,8 @@
       case 'u': case 'o':
       case 'x': case 'X':
         if (HowLong >= 1) {
-          if (HowLong == 1) {
+          if (HowLong == 1 && TheInterpreter->getModule().has64BitPointers() &&
+              sizeof(long) < sizeof(long long)) {
             // Make sure we use %lld with a 64 bit argument because we might be
             // compiling LLI on a 32 bit compiler.
             unsigned Size = strlen(FmtBuf);




From lattner at cs.uiuc.edu  Fri Apr 25 13:34:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:34:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/dry.c
Message-ID: <200304251833.NAA18309@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource:

dry.c updated: 1.2 -> 1.3

---
Log message:

* Switch to use the "clock" function to get time
* Reduce the size of the testcase to run reasonably in LLI
* Print some progress messages


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/dry.c
diff -u llvm/test/Programs/SingleSource/dry.c:1.2 llvm/test/Programs/SingleSource/dry.c:1.3
--- llvm/test/Programs/SingleSource/dry.c:1.2	Tue Apr 22 16:02:11 2003
+++ llvm/test/Programs/SingleSource/dry.c	Fri Apr 25 13:33:25 2003
@@ -88,7 +88,7 @@
 /*#define MYSTRFNS */
 
 /* Accuracy of timings and human fatigue controlled by next two lines */
-#define LOOPS	300000		/* Use this for slow or 16 bit machines */
+#define LOOPS	10000		/* Use this for slow or 16 bit machines */
 /*#define LOOPS	900000		/* Use this for faster machines */
 
 /* Compiler dependent options */
@@ -96,8 +96,8 @@
 #undef	NOSTRUCTASSIGN		/* Define if compiler can't assign structures */
 
 /* define only one of the next three defines */
-/*#define CLOCK */
-#define TIMES			/* Use times(2) time function */
+#define CLOCK
+/*#define TIMES			/* Use times(2) time function */
 /*#define TIME			/* Use time(2) time function */
 
 /* define the granularity of your times(2) function (when used) */
@@ -256,6 +256,7 @@
 #endif
 	for (i = 0; i < LOOPS; ++i)
 	{
+          if ((i & 127) == 0) printf("making progress: %d\n", i);
 		Proc5();
 		Proc4();
 		IntLoc1 = 2;




From lattner at cs.uiuc.edu  Fri Apr 25 13:37:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:37:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/fldry.c
Message-ID: <200304251836.NAA18641@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource:

fldry.c updated: 1.2 -> 1.3

---
Log message:

* Reduce testcase size to work with LLI reasonably
* Use clock instead of time()
* don't use long doubles, use doubles instead


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/fldry.c
diff -u llvm/test/Programs/SingleSource/fldry.c:1.2 llvm/test/Programs/SingleSource/fldry.c:1.3
--- llvm/test/Programs/SingleSource/fldry.c:1.2	Tue Apr 22 16:01:47 2003
+++ llvm/test/Programs/SingleSource/fldry.c	Fri Apr 25 13:36:10 2003
@@ -80,7 +80,7 @@
 
 /* #define the type of float we will use:	*/
 #ifndef extended
-#define extended long double
+#define extended double
 #endif
 
 /* Define if should use homemade str* functions. */
@@ -88,7 +88,7 @@
 /*#define MYSTRFNS*/
 
 /* Accuracy of timings and human fatigue controlled by next two lines */
-#define LOOPS	150000		/* Use this for slow or 16 bit machines */
+#define LOOPS	15000		/* Use this for slow or 16 bit machines */
 //#define LOOPS	900000		/* Use this for faster machines */
 
 /* Compiler dependent options */
@@ -96,9 +96,9 @@
 #undef	NOSTRUCTASSIGN		/* Define if compiler can't assign structures */
 
 /* define only one of the next three defines */
-/*#define CLOCK*/
+#define CLOCK
 /*#define TIMES			/* Use times(2) time function */
-#define TIME			/* Use time(2) time function */
+/*#define TIME			/* Use time(2) time function */
 
 /* define the granularity of your times(2) function (when used) */
 /*#define HZ	50		/* times(2) returns 1/50 second (europe?) */
@@ -264,6 +264,7 @@
 #endif
 	for (i = 0; i < LOOPS; ++i)
 	{
+          if ((i & 255) == 0) printf("Making progress: %d\n", i);
 		Proc5();
 		Proc4();
 		IntLoc1 = 2.0;




From lattner at cs.uiuc.edu  Fri Apr 25 13:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 13:45:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c
Message-ID: <200304251844.NAA20450@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2003-02-12-NonlocalGoto.c updated: 1.1 -> 1.2

---
Log message:

Add comment


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c
diff -u llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c:1.1 llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c:1.2
--- llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c:1.1	Wed Feb 12 15:45:39 2003
+++ llvm/test/Regression/CFrontend/2003-02-12-NonlocalGoto.c	Fri Apr 25 13:44:30 2003
@@ -1,3 +1,7 @@
+
+/* It is unlikely that LLVM will ever support nested functions, but if it does,
+   here is a testcase. */
+
 main()
 {
   __label__ l;




From lattner at cs.uiuc.edu  Fri Apr 25 14:50:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 14:50:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CFrontend/2002-07-29-Casts.c
Message-ID: <200304251949.OAA23969@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CFrontend:

2002-07-29-Casts.c updated: 1.1 -> 1.2

---
Log message:

Fix bugs in testcase


---
Diffs of the changes:

Index: llvm/test/Regression/CFrontend/2002-07-29-Casts.c
diff -u llvm/test/Regression/CFrontend/2002-07-29-Casts.c:1.1 llvm/test/Regression/CFrontend/2002-07-29-Casts.c:1.2
--- llvm/test/Regression/CFrontend/2002-07-29-Casts.c:1.1	Mon Jul 29 14:39:52 2002
+++ llvm/test/Regression/CFrontend/2002-07-29-Casts.c	Fri Apr 25 14:49:36 2003
@@ -7,8 +7,8 @@
 {
   char     c1;
   short    s1, ssf1, ssd1;
-  uint8_t  ubs0;
-  int8_t   bs0;
+  unsigned char  ubs0;
+  signed char   bs0;
   unsigned char ubc0, uc2;
   unsigned short us2, usf1, usd1;
   int ic3, is3, sif1, sid1;
@@ -24,8 +24,8 @@
   s1 = (short) (argc >= 3)? atoi(argv[2]) : -769;   /* 0xf7ff = -769 */
   
   ubc0 = (unsigned char) c1;                      /* 100 = 'd' */
-  ubs0 = (uint8_t) s1;                            /* 0xff = 255 */
-  bs0  = (int8_t) s1;                             /* 0xff = -1 */
+  ubs0 = (unsigned char) s1;                            /* 0xff = 255 */
+  bs0  = (signed char) s1;                             /* 0xff = -1 */
   
   uc2 = (unsigned char) c1;                       /* 100 = 'd' */
   us2 = (unsigned short) s1;                      /* 0xf7ff = 64767 */




From lattner at cs.uiuc.edu  Fri Apr 25 15:10:04 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 15:10:04 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Assembler/2003-04-25-UnresolvedGlobalReference.ll
Message-ID: <200304252009.PAA25801@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Assembler:

2003-04-25-UnresolvedGlobalReference.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Assembler/2003-04-25-UnresolvedGlobalReference.ll
diff -c /dev/null llvm/test/Regression/Assembler/2003-04-25-UnresolvedGlobalReference.ll:1.1
*** /dev/null	Fri Apr 25 15:09:27 2003
--- llvm/test/Regression/Assembler/2003-04-25-UnresolvedGlobalReference.ll	Fri Apr 25 15:09:17 2003
***************
*** 0 ****
--- 1,7 ----
+ ; There should be absolutely no problem with this testcase.
+ 
+ implementation
+ 
+ int %test(int %arg1, int %arg2) {  
+ 	ret int cast (int (int, int)* %test to int)
+ }




From lattner at cs.uiuc.edu  Fri Apr 25 16:48:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 16:48:01 2003
Subject: [llvm-commits] CVS: llvm/lib/AsmParser/llvmAsmParser.y
Message-ID: <200304252147.QAA26549@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/AsmParser:

llvmAsmParser.y updated: 1.108 -> 1.109

---
Log message:

Fix bug: Assembler/2003-04-25-UnresolvedGlobalReference.ll


---
Diffs of the changes:

Index: llvm/lib/AsmParser/llvmAsmParser.y
diff -u llvm/lib/AsmParser/llvmAsmParser.y:1.108 llvm/lib/AsmParser/llvmAsmParser.y:1.109
--- llvm/lib/AsmParser/llvmAsmParser.y:1.108	Tue Apr 22 14:07:06 2003
+++ llvm/lib/AsmParser/llvmAsmParser.y	Fri Apr 25 16:47:33 2003
@@ -156,6 +156,25 @@
     // resolve the branches now...
     ResolveDefinitions(LateResolveValues, &CurModule.LateResolveValues);
 
+    // Make sure to resolve any constant expr references that might exist within
+    // the function we just declared itself.
+    ValID FID;
+    if (CurrentFunction->hasName()) {
+      FID = ValID::create((char*)CurrentFunction->getName().c_str());
+    } else {
+      unsigned Slot = CurrentFunction->getType()->getUniqueID();
+      assert(CurModule.Values.size() > Slot && "Function not inserted?");
+      // Figure out which slot number if is...
+      for (unsigned i = 0; ; ++i) {
+        assert(i < CurModule.Values[Slot].size() && "Function not found!");
+        if (CurModule.Values[Slot][i] == CurrentFunction) {
+          FID = ValID::create((int)i);
+          break;
+        }
+      }
+    }
+    CurModule.DeclareNewGlobalValue(CurrentFunction, FID);
+
     Values.clear();         // Clear out function local definitions
     Types.clear();
     CurrentFunction = 0;
@@ -964,7 +983,6 @@
     Value *V = getValNonImprovising(Ty, $2);
 
     CurMeth.CurrentFunction = SavedCurFn;
-
 
     // If this is an initializer for a constant pointer, which is referencing a
     // (currently) undefined variable, create a stub now that shall be replaced




From lattner at cs.uiuc.edu  Fri Apr 25 17:00:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 17:00:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/InstSelectSimple.cpp
Message-ID: <200304252159.QAA26807@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

InstSelectSimple.cpp updated: 1.89 -> 1.90

---
Log message:

Fix bogus assert


---
Diffs of the changes:

Index: llvm/lib/Target/X86/InstSelectSimple.cpp
diff -u llvm/lib/Target/X86/InstSelectSimple.cpp:1.89 llvm/lib/Target/X86/InstSelectSimple.cpp:1.90
--- llvm/lib/Target/X86/InstSelectSimple.cpp:1.89	Wed Apr 23 12:57:48 2003
+++ llvm/lib/Target/X86/InstSelectSimple.cpp	Fri Apr 25 16:58:54 2003
@@ -28,8 +28,8 @@
 #include <map>
 
 /// BMI - A special BuildMI variant that takes an iterator to insert the
-/// instruction at as well as a basic block.
-/// this is the version for when you have a destination register in mind.
+/// instruction at as well as a basic block.  This is the version for when you
+/// have a destination register in mind.
 inline static MachineInstrBuilder BMI(MachineBasicBlock *MBB,
                                       MachineBasicBlock::iterator &I,
                                       MachineOpCode Opcode,
@@ -47,7 +47,7 @@
                                       MachineBasicBlock::iterator &I,
                                       MachineOpCode Opcode,
                                       unsigned NumOperands) {
-  assert(I > MBB->begin() && I <= MBB->end() && "Bad iterator!");
+  assert(I >= MBB->begin() && I <= MBB->end() && "Bad iterator!");
   MachineInstr *MI = new MachineInstr(Opcode, NumOperands, true, true);
   I = MBB->insert(I, MI)+1;
   return MachineInstrBuilder(MI);




From lattner at cs.uiuc.edu  Fri Apr 25 17:09:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 17:09:01 2003
Subject: [llvm-commits] CVS: llvm/tools/bugpoint/ExtractFunction.cpp
Message-ID: <200304252208.RAA27958@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/bugpoint:

ExtractFunction.cpp updated: 1.6 -> 1.7

---
Log message:

Add options to disable simplification with passes, in case one of them crashes


---
Diffs of the changes:

Index: llvm/tools/bugpoint/ExtractFunction.cpp
diff -u llvm/tools/bugpoint/ExtractFunction.cpp:1.6 llvm/tools/bugpoint/ExtractFunction.cpp:1.7
--- llvm/tools/bugpoint/ExtractFunction.cpp:1.6	Thu Apr 24 19:52:30 2003
+++ llvm/tools/bugpoint/ExtractFunction.cpp	Fri Apr 25 17:08:12 2003
@@ -14,6 +14,19 @@
 #include "llvm/Analysis/Verifier.h"
 #include "llvm/Type.h"
 #include "llvm/Constant.h"
+#include "Support/CommandLine.h"
+
+namespace {
+  cl::opt<bool>
+  NoADCE("disable-adce",
+         cl::desc("Do not use the -adce pass to reduce testcases"));
+  cl::opt<bool>
+  NoDCE ("disable-dce",
+         cl::desc("Do not use the -dce pass to reduce testcases"));
+  cl::opt<bool>
+  NoSCFG("disable-simplifycfg",
+         cl::desc("Do not use the -simplifycfg pass to reduce testcases"));
+}
 
 /// deleteInstructionFromProgram - This method clones the current Program and
 /// deletes the specified instruction from the cloned module.  It then runs a
@@ -46,12 +59,12 @@
 
   // Spiff up the output a little bit.
   PassManager Passes;
-  if (Simplification > 2)
+  if (Simplification > 2 && !NoADCE)
     Passes.add(createAggressiveDCEPass());          // Remove dead code...
   //Passes.add(createInstructionCombiningPass());
-  if (Simplification > 1)
+  if (Simplification > 1 && !NoDCE)
     Passes.add(createDeadCodeEliminationPass());
-  if (Simplification)
+  if (Simplification && !NoSCFG)
     Passes.add(createCFGSimplificationPass());      // Delete dead control flow
 
   Passes.add(createVerifierPass());




From lattner at cs.uiuc.edu  Fri Apr 25 17:54:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 17:54:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/ADCE/2003-04-25-PHIPostDominateProblem.ll
Message-ID: <200304252253.RAA30532@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/ADCE:

2003-04-25-PHIPostDominateProblem.ll added (r1.1)

---
Log message:

New testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/ADCE/2003-04-25-PHIPostDominateProblem.ll
diff -c /dev/null llvm/test/Regression/Transforms/ADCE/2003-04-25-PHIPostDominateProblem.ll:1.1
*** /dev/null	Fri Apr 25 17:53:23 2003
--- llvm/test/Regression/Transforms/ADCE/2003-04-25-PHIPostDominateProblem.ll	Fri Apr 25 17:53:12 2003
***************
*** 0 ****
--- 1,37 ----
+ ; THis testcase caused an assertion failure because a PHI node did not have 
+ ; entries for it's postdominator.  But I think this can only happen when the 
+ ; PHI node is dead, so we just avoid patching up dead PHI nodes.
+ 
+ ; RUN: as < %s | opt -adce
+ 
+ target endian = little
+ target pointersize = 32
+ 
+ implementation   ; Functions:
+ 
+ void %dead_test8() {
+ entry:		; No predecessors!
+ 	br label %loopentry
+ 
+ loopentry:		; preds = %entry, %endif
+ 	%k.1 = phi int [ %k.0, %endif ], [ 0, %entry ]		; <int> [#uses=1]
+ 	br bool false, label %no_exit, label %return
+ 
+ no_exit:		; preds = %loopentry
+ 	br bool false, label %then, label %else
+ 
+ then:		; preds = %no_exit
+ 	br label %endif
+ 
+ else:		; preds = %no_exit
+ 	%dec = add int %k.1, -1		; <int> [#uses=1]
+ 	br label %endif
+ 
+ endif:		; preds = %else, %then
+ 	%k.0 = phi int [ %dec, %else ], [ 0, %then ]		; <int> [#uses=1]
+ 	store int 2, int* null
+ 	br label %loopentry
+ 
+ return:		; preds = %loopentry
+ 	ret void
+ }




From lattner at cs.uiuc.edu  Fri Apr 25 17:54:07 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 17:54:07 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/Scalar/ADCE.cpp
Message-ID: <200304252253.RAA30541@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/Scalar:

ADCE.cpp updated: 1.51 -> 1.52

---
Log message:

Fix bug: ADCE/2003-04-25-PHIPostDominateProblem.ll


---
Diffs of the changes:

Index: llvm/lib/Transforms/Scalar/ADCE.cpp
diff -u llvm/lib/Transforms/Scalar/ADCE.cpp:1.51 llvm/lib/Transforms/Scalar/ADCE.cpp:1.52
--- llvm/lib/Transforms/Scalar/ADCE.cpp:1.51	Wed Apr 23 11:37:42 2003
+++ llvm/lib/Transforms/Scalar/ADCE.cpp	Fri Apr 25 17:53:27 2003
@@ -310,15 +310,16 @@
               // should be identical to the incoming values for LastDead.
               //
               for (BasicBlock::iterator II = NextAlive->begin();
-                   PHINode *PN = dyn_cast<PHINode>(II); ++II) {
-                // Get the incoming value for LastDead...
-                int OldIdx = PN->getBasicBlockIndex(LastDead);
-                assert(OldIdx != -1 && "LastDead is not a pred of NextAlive!");
-                Value *InVal = PN->getIncomingValue(OldIdx);
-                
-                // Add an incoming value for BB now...
-                PN->addIncoming(InVal, BB);
-              }
+                   PHINode *PN = dyn_cast<PHINode>(II); ++II)
+                if (LiveSet.count(PN)) {  // Only modify live phi nodes
+                  // Get the incoming value for LastDead...
+                  int OldIdx = PN->getBasicBlockIndex(LastDead);
+                  assert(OldIdx != -1 &&"LastDead is not a pred of NextAlive!");
+                  Value *InVal = PN->getIncomingValue(OldIdx);
+                  
+                  // Add an incoming value for BB now...
+                  PN->addIncoming(InVal, BB);
+                }
             }
           }
 




From lattner at cs.uiuc.edu  Fri Apr 25 18:15:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 18:15:01 2003
Subject: [llvm-commits] CVS: llvm/lib/VMCore/BasicBlock.cpp
Message-ID: <200304252314.SAA01102@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/VMCore:

BasicBlock.cpp updated: 1.30 -> 1.31

---
Log message:

Fix a bug that occurred when removing the last predecessor INTO an 
infinite loop


---
Diffs of the changes:

Index: llvm/lib/VMCore/BasicBlock.cpp
diff -u llvm/lib/VMCore/BasicBlock.cpp:1.30 llvm/lib/VMCore/BasicBlock.cpp:1.31
--- llvm/lib/VMCore/BasicBlock.cpp:1.30	Wed Apr 23 11:37:45 2003
+++ llvm/lib/VMCore/BasicBlock.cpp	Fri Apr 25 18:14:19 2003
@@ -186,7 +186,11 @@
 
       // If the PHI _HAD_ two uses, replace PHI node with its now *single* value
       if (max_idx == 2) {
-	PN->replaceAllUsesWith(PN->getOperand(0));
+        if (PN->getOperand(0) != PN)
+          PN->replaceAllUsesWith(PN->getOperand(0));
+        else
+          // We are left with an infinite loop with no entries: kill the PHI.
+          PN->replaceAllUsesWith(Constant::getNullValue(PN->getType()));
         getInstList().pop_front();    // Remove the PHI node
       }
 




From lattner at cs.uiuc.edu  Fri Apr 25 18:28:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 18:28:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll
Message-ID: <200304252327.SAA02980@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LoopPreheaders:

2003-04-25-AssertFail.ll added (r1.1)

---
Log message:

New testcase that crashes preheaders


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll
diff -c /dev/null llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll:1.1
*** /dev/null	Fri Apr 25 18:27:02 2003
--- llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll	Fri Apr 25 18:26:52 2003
***************
*** 0 ****
--- 1,22 ----
+ ; RUN: as < %s | opt -preheaders
+ 
+ implementation   ; Functions:
+ 
+ int %yylex() {		; No predecessors!
+ 	br label %loopentry.0
+ 
+ loopentry.0:		; preds = %0, %yy_find_action, %else.4
+ 	br label %loopexit.2
+ 
+ loopexit.2:		; preds = %loopentry.0, %else.4, %loopexit.2
+ 	br bool false, label %loopexit.2, label %else.4
+ 
+ yy_find_action:		; preds = %loopexit.2, %else.4
+ 	br label %else.4
+ 
+ else.4:		; preds = %yy_find_action
+ 	switch uint 0, label %loopexit.2 [
+ 		 uint 2, label %yy_find_action
+ 		 uint 0, label %loopentry.0
+ 	]
+ }




From lattner at cs.uiuc.edu  Fri Apr 25 18:40:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Fri Apr 25 18:40:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CFG.h
Message-ID: <200304252339.SAA03198@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CFG.h updated: 1.8 -> 1.9

---
Log message:

Remove long-dead obsolete cruft


---
Diffs of the changes:

Index: llvm/include/llvm/Support/CFG.h
diff -u llvm/include/llvm/Support/CFG.h:1.8 llvm/include/llvm/Support/CFG.h:1.9
--- llvm/include/llvm/Support/CFG.h:1.8	Thu Oct 10 17:31:31 2002
+++ llvm/include/llvm/Support/CFG.h	Fri Apr 25 18:39:08 2003
@@ -26,15 +26,7 @@
   typedef PredIterator<_Ptr,_USE_iterator> _Self;
   typedef typename super::pointer pointer;
   
-  inline void advancePastConstants() {
-    // TODO: This is bad
-    // Loop to ignore constant pool references
-    while (It != BB->use_end() && !isa<TerminatorInst>(*It))
-      ++It;
-  }
-  
   inline PredIterator(_Ptr *bb) : BB(bb), It(bb->use_begin()) {
-    advancePastConstants();
   }
   inline PredIterator(_Ptr *bb, bool) : BB(bb), It(bb->use_end()) {}
     
@@ -43,13 +35,13 @@
   
   inline pointer operator*() const { 
     assert(It != BB->use_end() && "pred_iterator out of range!");
-    return cast<Instruction>(*It)->getParent(); 
+    return cast<TerminatorInst>(*It)->getParent(); 
   }
   inline pointer *operator->() const { return &(operator*()); }
   
   inline _Self& operator++() {   // Preincrement
     assert(It != BB->use_end() && "pred_iterator out of range!");
-    ++It; advancePastConstants();
+    ++It;
     return *this; 
   }
   




From lattner at cs.uiuc.edu  Sat Apr 26 12:39:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 12:39:01 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Support/CFG.h
Message-ID: <200304261738.MAA15592@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Support:

CFG.h updated: 1.9 -> 1.10

---
Log message:

Revert last checkin.  Note that PHI nodes can use basic blocks also!


---
Diffs of the changes:

Index: llvm/include/llvm/Support/CFG.h
diff -u llvm/include/llvm/Support/CFG.h:1.9 llvm/include/llvm/Support/CFG.h:1.10
--- llvm/include/llvm/Support/CFG.h:1.9	Fri Apr 25 18:39:08 2003
+++ llvm/include/llvm/Support/CFG.h	Sat Apr 26 12:38:26 2003
@@ -26,7 +26,14 @@
   typedef PredIterator<_Ptr,_USE_iterator> _Self;
   typedef typename super::pointer pointer;
   
+  inline void advancePastConstants() {
+    // Loop to ignore non terminator uses (for example PHI nodes)...
+    while (It != BB->use_end() && !isa<TerminatorInst>(*It))
+      ++It;
+  }
+  
   inline PredIterator(_Ptr *bb) : BB(bb), It(bb->use_begin()) {
+    advancePastConstants();
   }
   inline PredIterator(_Ptr *bb, bool) : BB(bb), It(bb->use_end()) {}
     
@@ -41,7 +48,7 @@
   
   inline _Self& operator++() {   // Preincrement
     assert(It != BB->use_end() && "pred_iterator out of range!");
-    ++It;
+    ++It; advancePastConstants();
     return *this; 
   }
   




From lattner at cs.uiuc.edu  Sat Apr 26 12:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 12:45:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll
Message-ID: <200304261744.MAA16347@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/CBackend:

2002-10-15-OpaqueTypeProblem.ll updated: 1.1 -> 1.2

---
Log message:

Whoever checked this testcase in apparently never tested it.


---
Diffs of the changes:

Index: llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll
diff -u llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll:1.1 llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll:1.2
--- llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll:1.1	Tue Oct 15 19:08:08 2002
+++ llvm/test/Regression/CBackend/2002-10-15-OpaqueTypeProblem.ll	Sat Apr 26 12:44:17 2003
@@ -1,4 +1,5 @@
-struct Comm;
-typedef struct Comm* MPI_Comm;
-MPI_Comm* thing;
+	%MPI_Comm = type %struct.Comm*
+	%struct.Comm = type opaque
+%thing = global %MPI_Comm* null		; <%MPI_Comm**> [#uses=0]
 
+implementation   ; Functions:




From lattner at cs.uiuc.edu  Sat Apr 26 14:35:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 14:35:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll
Message-ID: <200304261934.OAA17091@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Transforms/LoopPreheaders:

2003-04-25-AssertFail.ll updated: 1.1 -> 1.2

---
Log message:

Add comment to testcase


---
Diffs of the changes:

Index: llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll
diff -u llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll:1.1 llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll:1.2
--- llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll:1.1	Fri Apr 25 18:26:52 2003
+++ llvm/test/Regression/Transforms/LoopPreheaders/2003-04-25-AssertFail.ll	Sat Apr 26 14:33:51 2003
@@ -1,3 +1,6 @@
+; This testcase exposed a problem with the loop identification pass (LoopInfo).
+; Basically, it was incorrectly calculating the loop nesting information.
+;
 ; RUN: as < %s | opt -preheaders
 
 implementation   ; Functions:




From lattner at cs.uiuc.edu  Sat Apr 26 14:35:07 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 14:35:07 2003
Subject: [llvm-commits] CVS: llvm/lib/Analysis/LoopInfo.cpp
Message-ID: <200304261934.OAA17288@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Analysis:

LoopInfo.cpp updated: 1.35 -> 1.36

---
Log message:

Fix bug: LoopPreheaders/2003-04-25-AssertFail.ll


---
Diffs of the changes:

Index: llvm/lib/Analysis/LoopInfo.cpp
diff -u llvm/lib/Analysis/LoopInfo.cpp:1.35 llvm/lib/Analysis/LoopInfo.cpp:1.36
--- llvm/lib/Analysis/LoopInfo.cpp:1.35	Fri Feb 28 10:54:45 2003
+++ llvm/lib/Analysis/LoopInfo.cpp	Sat Apr 26 14:34:18 2003
@@ -128,12 +128,12 @@
   std::vector<BasicBlock *> TodoStack;
 
   // Scan the predecessors of BB, checking to see if BB dominates any of
-  // them.
+  // them.  This identifies backedges which target this node...
   for (pred_iterator I = pred_begin(BB), E = pred_end(BB); I != E; ++I)
     if (DS.dominates(BB, *I))   // If BB dominates it's predecessor...
       TodoStack.push_back(*I);
 
-  if (TodoStack.empty()) return 0;  // Doesn't dominate any predecessors...
+  if (TodoStack.empty()) return 0;  // No backedges to this block...
 
   // Create a new loop to represent this basic block...
   Loop *L = new Loop(BB);
@@ -144,8 +144,29 @@
     TodoStack.pop_back();
 
     if (!L->contains(X)) {         // As of yet unprocessed??
+      // Check to see if this block already belongs to a loop.  If this occurs
+      // then we have a case where a loop that is supposed to be a child of the
+      // current loop was processed before the current loop.  When this occurs,
+      // this child loop gets added to a part of the current loop, making it a
+      // sibling to the current loop.  We have to reparent this loop.
+      if (Loop *SubLoop = const_cast<Loop*>(getLoopFor(X)))
+        if (SubLoop->getHeader() == X && X != BB) {
+          // Remove the subloop from it's current parent...
+          assert(SubLoop->ParentLoop && SubLoop->ParentLoop != L);
+          Loop *SLP = SubLoop->ParentLoop;  // SubLoopParent
+          std::vector<Loop*>::iterator I =
+            std::find(SLP->SubLoops.begin(), SLP->SubLoops.end(), SubLoop);
+          assert(I != SLP->SubLoops.end() && "SubLoop not a child of parent?");
+          SLP->SubLoops.erase(I);   // Remove from parent...
+          
+          // Add the subloop to THIS loop...
+          SubLoop->ParentLoop = L;
+          L->SubLoops.push_back(SubLoop);
+        }
+
+      // Normal case, add the block to our loop...
       L->Blocks.push_back(X);
-      
+        
       // Add all of the predecessors of X to the end of the work stack...
       TodoStack.insert(TodoStack.end(), pred_begin(X), pred_end(X));
     }




From lattner at cs.uiuc.edu  Sat Apr 26 14:37:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 14:37:00 2003
Subject: [llvm-commits] CVS: llvm/www/docs/ChrisNotes.txt
Message-ID: <200304261936.OAA17373@apoc.cs.uiuc.edu>


Changes in directory llvm/www/docs:

ChrisNotes.txt updated: 1.28 -> 1.29

---
Log message:

Switch insns actually DO work.


---
Diffs of the changes:

Index: llvm/www/docs/ChrisNotes.txt
diff -u llvm/www/docs/ChrisNotes.txt:1.28 llvm/www/docs/ChrisNotes.txt:1.29
--- llvm/www/docs/ChrisNotes.txt:1.28	Tue Sep 24 11:15:36 2002
+++ llvm/www/docs/ChrisNotes.txt	Sat Apr 26 14:36:06 2003
@@ -1,14 +1,11 @@
 * Rewrite the llvm parser/lexer in http://www.antlr.org when time permits.
   They actually do C++.  Imagine that.
-* pred/succ iterators on basic blocks don't handle switch statements correctly
 * Need a way to attach bytecode block info at various levels of asm code.
 * Recognize and save comments in assembly and bytecode format
 * Encode line number table in bytecode (like #line), optional table
 
 * Encode negative relative offsets in the bytecode file
 
-* Implement switch to switch on a constant pool array of type: 
-  [{ label, int }] or [label]   (lookup vs index switch)
 * Apparently bison has a %pure_parser option.  Maybe useful for AsmParser
 
 * Implement the following derived types:




From lattner at cs.uiuc.edu  Sat Apr 26 14:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 14:45:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/Sparc/SparcInstrInfo.cpp SparcInstrSelection.cpp
Message-ID: <200304261944.OAA17418@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/Sparc:

SparcInstrInfo.cpp updated: 1.39 -> 1.40
SparcInstrSelection.cpp updated: 1.87 -> 1.88

---
Log message:

IntegerRegSize is always 8 for sparc


---
Diffs of the changes:

Index: llvm/lib/Target/Sparc/SparcInstrInfo.cpp
diff -u llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.39 llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.40
--- llvm/lib/Target/Sparc/SparcInstrInfo.cpp:1.39	Fri Apr 25 00:23:09 2003
+++ llvm/lib/Target/Sparc/SparcInstrInfo.cpp	Sat Apr 26 14:44:35 2003
@@ -441,9 +441,7 @@
           uint64_t C = GetConstantValueAsUnsignedInt(val, isValidConstant);
           assert(isValidConstant && "Unrecognized constant");
 
-          if (opSize > destSize ||
-              (val->getType()->isSigned()
-               && destSize < target.getTargetData().getIntegerRegSize()))
+          if (opSize > destSize || (val->getType()->isSigned() && destSize < 8))
             { // operand is larger than dest,
               //    OR both are equal but smaller than the full register size
               //       AND operand is signed, so it may have extra sign bits:


Index: llvm/lib/Target/Sparc/SparcInstrSelection.cpp
diff -u llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.87 llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.88
--- llvm/lib/Target/Sparc/SparcInstrSelection.cpp:1.87	Fri Apr 25 00:23:10 2003
+++ llvm/lib/Target/Sparc/SparcInstrSelection.cpp	Sat Apr 26 14:44:35 2003
@@ -758,8 +758,7 @@
   // 
   Value* shiftDest = destVal;
   unsigned opSize = target.getTargetData().getTypeSize(argVal1->getType());
-  if ((shiftOpCode == SLL || shiftOpCode == SLLX)
-      && opSize < target.getTargetData().getIntegerRegSize())
+  if ((shiftOpCode == SLL || shiftOpCode == SLLX) && opSize < 8)
     { // put SLL result into a temporary
       shiftDest = new TmpInstruction(argVal1, optArgVal2, "sllTmp");
       mcfi.addTemp(shiftDest);
@@ -2305,7 +2304,7 @@
                                  .addReg(dest, MOTy::Def);
               mvec.push_back(M);
             }
-          else if (destSize < target.getTargetData().getIntegerRegSize())
+          else if (destSize < 8)
             assert(0 && "Unsupported type size: 32 < size < 64 bits");
         }
     }




From lattner at cs.uiuc.edu  Sat Apr 26 14:48:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 14:48:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetMachine.cpp
Message-ID: <200304261947.OAA17451@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetMachine.cpp updated: 1.17 -> 1.18

---
Log message:

The promotion rules are the same for all targets, they are set by the C standard.


---
Diffs of the changes:

Index: llvm/lib/Target/TargetMachine.cpp
diff -u llvm/lib/Target/TargetMachine.cpp:1.17 llvm/lib/Target/TargetMachine.cpp:1.18
--- llvm/lib/Target/TargetMachine.cpp:1.17	Sat Dec 28 20:50:31 2002
+++ llvm/lib/Target/TargetMachine.cpp	Sat Apr 26 14:47:36 2003
@@ -20,16 +20,10 @@
 
 // function TargetMachine::findOptimalStorageSize 
 // 
-// Purpose:
-//   This default implementation assumes that all sub-word data items use
-//   space equal to optSizeForSubWordData, and all other primitive data
-//   items use space according to the type.
-//   
 unsigned TargetMachine::findOptimalStorageSize(const Type *Ty) const {
-  // Round integral values smaller than SubWordDataSize up to SubWordDataSize
-  if (Ty->isIntegral() &&
-      Ty->getPrimitiveSize() < DataLayout.getSubWordDataSize())
-    return DataLayout.getSubWordDataSize();
+  // All integer types smaller than ints promote to 4 byte integers.
+  if (Ty->isIntegral() && Ty->getPrimitiveSize() < 4)
+    return 4;
 
   return DataLayout.getTypeSize(Ty);
 }




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/Sparc/Sparc.cpp
Message-ID: <200304262011.PAA18870@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/Sparc:

Sparc.cpp updated: 1.61 -> 1.62

---
Log message:

Remove two fields from TargetData which are target specific.


---
Diffs of the changes:

Index: llvm/lib/Target/Sparc/Sparc.cpp
diff -u llvm/lib/Target/Sparc/Sparc.cpp:1.61 llvm/lib/Target/Sparc/Sparc.cpp:1.62
--- llvm/lib/Target/Sparc/Sparc.cpp:1.61	Wed Apr 23 11:24:53 2003
+++ llvm/lib/Target/Sparc/Sparc.cpp	Sat Apr 26 15:11:01 2003
@@ -129,7 +129,7 @@
 //---------------------------------------------------------------------------
 
 UltraSparc::UltraSparc()
-  : TargetMachine("UltraSparc-Native", false, 4),
+  : TargetMachine("UltraSparc-Native", false),
     schedInfo(*this),
     regInfo(*this),
     frameInfo(*this),




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:06 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:06 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/X86/X86TargetMachine.cpp
Message-ID: <200304262011.PAA18877@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target/X86:

X86TargetMachine.cpp updated: 1.16 -> 1.17

---
Log message:

Remove two fields from TargetData which are target specific.



---
Diffs of the changes:

Index: llvm/lib/Target/X86/X86TargetMachine.cpp
diff -u llvm/lib/Target/X86/X86TargetMachine.cpp:1.16 llvm/lib/Target/X86/X86TargetMachine.cpp:1.17
--- llvm/lib/Target/X86/X86TargetMachine.cpp:1.16	Fri Apr 25 01:05:57 2003
+++ llvm/lib/Target/X86/X86TargetMachine.cpp	Sat Apr 26 15:11:04 2003
@@ -35,7 +35,6 @@
 X86TargetMachine::X86TargetMachine(unsigned Config)
   : TargetMachine("X86", 
 		  (Config & TM::EndianMask) == TM::LittleEndian,
-		  1, 4, 
 		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		  (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4),




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:11 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:11 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304262011.PAA18884@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.69 -> 1.70

---
Log message:

Remove two fields from TargetData which are target specific.


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.69 llvm/tools/llc/llc.cpp:1.70
--- llvm/tools/llc/llc.cpp:1.69	Fri Apr 25 01:06:13 2003
+++ llvm/tools/llc/llc.cpp	Sat Apr 26 15:11:06 2003
@@ -184,8 +184,7 @@
   // Build up all of the passes that we want to do to the module...
   PassManager Passes;
 
-  Passes.add(new TargetData("llc", TD.isLittleEndian(), TD.getSubWordDataSize(),
-                            TD.getIntegerRegSize(), TD.getPointerSize(),
+  Passes.add(new TargetData("llc", TD.isLittleEndian(), TD.getPointerSize(),
                             TD.getPointerAlignment(), TD.getDoubleAlignment()));
 
   // Create a new optimization pass for each one specified on the command line




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:17 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:17 2003
Subject: [llvm-commits] CVS: llvm/tools/lli/Interpreter/Interpreter.cpp
Message-ID: <200304262011.PAA18891@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/lli/Interpreter:

Interpreter.cpp updated: 1.2 -> 1.3

---
Log message:

Remove two fields from TargetData which are target specific.



---
Diffs of the changes:

Index: llvm/tools/lli/Interpreter/Interpreter.cpp
diff -u llvm/tools/lli/Interpreter/Interpreter.cpp:1.2 llvm/tools/lli/Interpreter/Interpreter.cpp:1.3
--- llvm/tools/lli/Interpreter/Interpreter.cpp:1.2	Fri Apr 25 01:15:05 2003
+++ llvm/tools/lli/Interpreter/Interpreter.cpp	Sat Apr 26 15:11:09 2003
@@ -25,7 +25,6 @@
 			 bool DebugMode, bool TraceMode)
   : ExecutionEngine(M), ExitCode(0), Debug(DebugMode), Trace(TraceMode),
     CurFrame(-1), TD("lli", (Config & TM::EndianMask) == TM::LittleEndian,
-		     1, 4,
 		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4,
 		     (Config & TM::PtrSizeMask) == TM::PtrSize64 ? 8 : 4) {




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:22 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:22 2003
Subject: [llvm-commits] CVS: llvm/include/llvm/Target/TargetData.h TargetMachine.h
Message-ID: <200304262011.PAA18898@apoc.cs.uiuc.edu>


Changes in directory llvm/include/llvm/Target:

TargetData.h updated: 1.16 -> 1.17
TargetMachine.h updated: 1.29 -> 1.30

---
Log message:

Remove two fields from TargetData which are target specific.


---
Diffs of the changes:

Index: llvm/include/llvm/Target/TargetData.h
diff -u llvm/include/llvm/Target/TargetData.h:1.16 llvm/include/llvm/Target/TargetData.h:1.17
--- llvm/include/llvm/Target/TargetData.h:1.16	Fri Apr 25 00:23:27 2003
+++ llvm/include/llvm/Target/TargetData.h	Sat Apr 26 15:10:54 2003
@@ -24,7 +24,6 @@
 
 class TargetData : public ImmutablePass {
   bool          LittleEndian;          // Defaults to false
-  unsigned char SubWordDataSize;       // Defaults to 1 byte (no rounding up)
   unsigned char ByteAlignment;         // Defaults to 1 bytes
   unsigned char ShortAlignment;        // Defaults to 2 bytes
   unsigned char IntAlignment;          // Defaults to 4 bytes
@@ -32,7 +31,6 @@
   unsigned char FloatAlignment;        // Defaults to 4 bytes
   unsigned char DoubleAlignment;       // Defaults to 8 bytes
   unsigned char PointerSize;           // Defaults to 8 bytes
-  unsigned char IntegerRegSize;        // Defaults to PointerSize = 8 bytes
   unsigned char PointerAlignment;      // Defaults to 8 bytes
   AnnotationID  AID;                   // AID for structure layout annotation
  
@@ -40,8 +38,6 @@
 public:
   TargetData(const std::string &TargetName = "",
              bool LittleEndian = false,
-             unsigned char SubWordDataSize = 1,
-             unsigned char IntRegSize = 8,
              unsigned char PtrSize = 8,
 	     unsigned char PtrAl = 8, unsigned char DoubleAl = 8,
 	     unsigned char FloatAl = 4, unsigned char LongAl = 8, 
@@ -55,7 +51,6 @@
   bool          isBigEndian()         const { return    !LittleEndian; }
 
   /// Target alignment constraints
-  unsigned char getSubWordDataSize()  const { return  SubWordDataSize; }
   unsigned char getByteAlignment()    const { return    ByteAlignment; }
   unsigned char getShortAlignment()   const { return   ShortAlignment; }
   unsigned char getIntAlignment()     const { return     IntAlignment; }
@@ -64,7 +59,6 @@
   unsigned char getDoubleAlignment()  const { return  DoubleAlignment; }
   unsigned char getPointerAlignment() const { return PointerAlignment; }
   unsigned char getPointerSize()      const { return      PointerSize; }
-  unsigned char getIntegerRegSize()   const { return   IntegerRegSize; }
   AnnotationID  getStructLayoutAID()  const { return AID; }
 
   // getTypeSize - Return the number of bytes neccesary to hold the specified


Index: llvm/include/llvm/Target/TargetMachine.h
diff -u llvm/include/llvm/Target/TargetMachine.h:1.29 llvm/include/llvm/Target/TargetMachine.h:1.30
--- llvm/include/llvm/Target/TargetMachine.h:1.29	Sun Jan 12 18:21:32 2003
+++ llvm/include/llvm/Target/TargetMachine.h	Sat Apr 26 15:10:54 2003
@@ -35,12 +35,11 @@
 protected:
   TargetMachine(const std::string &name, // Can only create subclasses...
 		bool LittleEndian = false,
-                unsigned char SubWordSize = 1, unsigned char IntRegSize = 8,
 		unsigned char PtrSize = 8, unsigned char PtrAl = 8,
 		unsigned char DoubleAl = 8, unsigned char FloatAl = 4,
 		unsigned char LongAl = 8, unsigned char IntAl = 4,
 		unsigned char ShortAl = 2, unsigned char ByteAl = 1)
-    : Name(name), DataLayout(name, LittleEndian, SubWordSize, IntRegSize,
+    : Name(name), DataLayout(name, LittleEndian,
 			     PtrSize, PtrAl, DoubleAl, FloatAl, LongAl,
                              IntAl, ShortAl, ByteAl) {}
 public:




From lattner at cs.uiuc.edu  Sat Apr 26 15:12:28 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:12:28 2003
Subject: [llvm-commits] CVS: llvm/lib/Target/TargetData.cpp
Message-ID: <200304262011.PAA18904@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Target:

TargetData.cpp updated: 1.30 -> 1.31

---
Log message:

Remove two fields from TargetData which are target specific.



---
Diffs of the changes:

Index: llvm/lib/Target/TargetData.cpp
diff -u llvm/lib/Target/TargetData.cpp:1.30 llvm/lib/Target/TargetData.cpp:1.31
--- llvm/lib/Target/TargetData.cpp:1.30	Fri Apr 25 01:06:43 2003
+++ llvm/lib/Target/TargetData.cpp	Sat Apr 26 15:10:58 2003
@@ -82,8 +82,7 @@
 //===----------------------------------------------------------------------===//
 
 TargetData::TargetData(const std::string &TargetName,
-                       bool isLittleEndian, unsigned char SubWordSize,
-                       unsigned char IntRegSize, unsigned char PtrSize,
+                       bool isLittleEndian, unsigned char PtrSize,
                        unsigned char PtrAl, unsigned char DoubleAl,
                        unsigned char FloatAl, unsigned char LongAl, 
                        unsigned char IntAl, unsigned char ShortAl,
@@ -99,8 +98,6 @@
          "ERROR: Tool did not specify a target data to use!");
 
   LittleEndian     = isLittleEndian;
-  SubWordDataSize  = SubWordSize;
-  IntegerRegSize   = IntRegSize;
   PointerSize      = PtrSize;
   PointerAlignment = PtrAl;
   DoubleAlignment  = DoubleAl;
@@ -118,8 +115,6 @@
   AnnotationManager::registerAnnotationFactory(AID, TypeAnFactory, this);
 
   LittleEndian     = M->isLittleEndian();
-  SubWordDataSize  = 1;
-  IntegerRegSize   = 8;
   PointerSize      = M->has32BitPointers() ? 4 : 8;
   PointerAlignment = PointerSize;
   DoubleAlignment  = PointerSize;




From lattner at cs.uiuc.edu  Sat Apr 26 15:30:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:30:01 2003
Subject: [llvm-commits] CVS: llvm/test/Regression/Linker/2003-04-26-NullPtrLinkProblem.ll
Message-ID: <200304262029.PAA22542@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Regression/Linker:

2003-04-26-NullPtrLinkProblem.ll added (r1.1)

---
Log message:

New testcase distilled from a problem identified by sumant


---
Diffs of the changes:

Index: llvm/test/Regression/Linker/2003-04-26-NullPtrLinkProblem.ll
diff -c /dev/null llvm/test/Regression/Linker/2003-04-26-NullPtrLinkProblem.ll:1.1
*** /dev/null	Sat Apr 26 15:29:19 2003
--- llvm/test/Regression/Linker/2003-04-26-NullPtrLinkProblem.ll	Sat Apr 26 15:29:09 2003
***************
*** 0 ****
--- 1,19 ----
+ ; This one fails because the LLVM runtime is allowing two null pointers of
+ ; the same type to be created!
+ 
+ ; RUN: echo "%T = type int" | as > Output/%s.2.bc
+ ; RUN: as < %s > Output/%s.1.bc
+ ; RUN: link Output/%s.[12].bc
+ 
+ %T = type opaque
+ 
+ declare %T* %create()
+ 
+ implementation
+ 
+ void %test() {
+ 	%X = call %T* %create()
+ 	%v = seteq %T* %X, null
+ 	ret void
+ }
+ 




From lattner at cs.uiuc.edu  Sat Apr 26 15:45:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sat Apr 26 15:45:01 2003
Subject: [llvm-commits] CVS: llvm/test/Programs/SingleSource/UnitTests/2002-12-13-MishaTest.c
Message-ID: <200304262044.PAA23299@apoc.cs.uiuc.edu>


Changes in directory llvm/test/Programs/SingleSource/UnitTests:

2002-12-13-MishaTest.c added (r1.1)

---
Log message:

Check in an old testcase for a problem reported by misha


---
Diffs of the changes:

Index: llvm/test/Programs/SingleSource/UnitTests/2002-12-13-MishaTest.c
diff -c /dev/null llvm/test/Programs/SingleSource/UnitTests/2002-12-13-MishaTest.c:1.1
*** /dev/null	Sat Apr 26 15:44:42 2003
--- llvm/test/Programs/SingleSource/UnitTests/2002-12-13-MishaTest.c	Sat Apr 26 15:44:32 2003
***************
*** 0 ****
--- 1,25 ----
+ 
+ #include <stdio.h>
+ sum(to, from, count)
+      short *to, *from;
+      short count;
+ {
+   int i;
+   for (i = 0; i != count; ++i)
+     *to += *from++;
+ }
+ 
+ #define NUM 2
+ int main() {
+   short Array[NUM];
+   short Sum = 0;
+   int i;
+ 
+   for (i = 0; i != NUM; ++i)
+     Array[i] = i;
+ 
+   sum(&Sum, Array, NUM);
+ 
+   printf("Sum is %d\n", Sum);
+   return 0;
+ }




From lattner at cs.uiuc.edu  Sun Apr 27 17:57:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 17:57:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/LevelRaise.cpp
Message-ID: <200304272256.RAA16354@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms:

LevelRaise.cpp updated: 1.77 -> 1.78

---
Log message:

Avoid naming things exactly the same


---
Diffs of the changes:

Index: llvm/lib/Transforms/LevelRaise.cpp
diff -u llvm/lib/Transforms/LevelRaise.cpp:1.77 llvm/lib/Transforms/LevelRaise.cpp:1.78
--- llvm/lib/Transforms/LevelRaise.cpp:1.77	Thu Apr 24 13:25:27 2003
+++ llvm/lib/Transforms/LevelRaise.cpp	Sun Apr 27 17:55:56 2003
@@ -505,7 +505,7 @@
 
       // Create a new cast, inserting it right before the function call...
       CastInst *NewCast = new CastInst(CI->getCalledValue(), NewPFunTy,
-                                       CI->getCalledValue()->getName(), CI);
+                                       CI->getCalledValue()->getName()+"_c",CI);
 
       // Create a new call instruction...
       CallInst *NewCall = new CallInst(NewCast,




From lattner at cs.uiuc.edu  Sun Apr 27 20:24:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 20:24:01 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/IPO/FunctionResolution.cpp
Message-ID: <200304280123.UAA19652@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms/IPO:

FunctionResolution.cpp updated: 1.25 -> 1.26

---
Log message:

Fix several bugs:
  * Warnings were emitted all of the time and were really annoying
  * Functions could not be resolved unless they had external linkage.  Linkonce
    linkage was not allowed
  * ConstantPointerRef's were not handled when linking functions
    we now actually handle cast (CPR) to X -> cast (NewCPR) to X


---
Diffs of the changes:

Index: llvm/lib/Transforms/IPO/FunctionResolution.cpp
diff -u llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.25 llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.26
--- llvm/lib/Transforms/IPO/FunctionResolution.cpp:1.25	Fri Apr 18 19:15:27 2003
+++ llvm/lib/Transforms/IPO/FunctionResolution.cpp	Sun Apr 27 20:23:29 2003
@@ -128,7 +128,7 @@
       const FunctionType *OldMT = Old->getFunctionType();
       const FunctionType *ConcreteMT = Concrete->getFunctionType();
       
-      if (OldMT->getParamTypes().size() < ConcreteMT->getParamTypes().size() &&
+      if (OldMT->getParamTypes().size() > ConcreteMT->getParamTypes().size() &&
           !ConcreteMT->isVarArg())
         if (!Old->use_empty()) {
           std::cerr << "WARNING: Linking function '" << Old->getName()
@@ -155,14 +155,13 @@
             return Changed;
           }
       
-      // Attempt to convert all of the uses of the old function to the
-      // concrete form of the function.  If there is a use of the fn that
-      // we don't understand here we punt to avoid making a bad
-      // transformation.
+      // Attempt to convert all of the uses of the old function to the concrete
+      // form of the function.  If there is a use of the fn that we don't
+      // understand here we punt to avoid making a bad transformation.
       //
-      // At this point, we know that the return values are the same for
-      // our two functions and that the Old function has no varargs fns
-      // specified.  In otherwords it's just <retty> (...)
+      // At this point, we know that the return values are the same for our two
+      // functions and that the Old function has no varargs fns specified.  In
+      // otherwords it's just <retty> (...)
       //
       for (unsigned i = 0; i < Old->use_size(); ) {
         User *U = *(Old->use_begin()+i);
@@ -183,6 +182,18 @@
                       << " argument or something!" << CI;
             ++i;
           }
+        } else if (ConstantPointerRef *CPR = dyn_cast<ConstantPointerRef>(U)) {
+          if (CPR->use_size() == 1 && isa<ConstantExpr>(CPR->use_back()) &&
+              cast<ConstantExpr>(CPR->use_back())->getOpcode() == 
+                Instruction::Cast) {
+            ConstantExpr *CE = cast<ConstantExpr>(CPR->use_back());
+            Constant *NewCPR = ConstantPointerRef::get(Concrete);
+            CE->replaceAllUsesWith(ConstantExpr::getCast(NewCPR,CE->getType()));
+            CPR->destroyConstant();
+          } else {
+            std::cerr << "Cannot convert use of function: " << CPR << "\n";
+            ++i;
+          }
         } else {
           std::cerr << "Cannot convert use of function: " << U << "\n";
           ++i;
@@ -337,7 +348,7 @@
         GlobalValue *GV = cast<GlobalValue>(PI->second);
         assert(PI->first == GV->getName() &&
                "Global name and symbol table do not agree!");
-        if (GV->hasExternalLinkage())  // Only resolve decls to external fns
+        if (!GV->hasInternalLinkage())  // Only resolve decls to external fns
           Globals[PI->first].push_back(GV);
       }
     }




From lattner at cs.uiuc.edu  Sun Apr 27 20:26:00 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 20:26:00 2003
Subject: [llvm-commits] CVS: llvm/lib/Transforms/LevelRaise.cpp
Message-ID: <200304280125.UAA19666@apoc.cs.uiuc.edu>


Changes in directory llvm/lib/Transforms:

LevelRaise.cpp updated: 1.78 -> 1.79

---
Log message:

Create constant expression casts instead of constant instructions if possible


---
Diffs of the changes:

Index: llvm/lib/Transforms/LevelRaise.cpp
diff -u llvm/lib/Transforms/LevelRaise.cpp:1.78 llvm/lib/Transforms/LevelRaise.cpp:1.79
--- llvm/lib/Transforms/LevelRaise.cpp:1.78	Sun Apr 27 17:55:56 2003
+++ llvm/lib/Transforms/LevelRaise.cpp	Sun Apr 27 20:25:38 2003
@@ -504,8 +504,18 @@
       PointerType *NewPFunTy = PointerType::get(NewFT);
 
       // Create a new cast, inserting it right before the function call...
-      CastInst *NewCast = new CastInst(CI->getCalledValue(), NewPFunTy,
-                                       CI->getCalledValue()->getName()+"_c",CI);
+      Value *NewCast;
+      Constant *ConstantCallSrc = 0;
+      if (Constant *CS = dyn_cast<Constant>(CI->getCalledValue()))
+        ConstantCallSrc = CS;
+      else if (GlobalValue *GV = dyn_cast<GlobalValue>(CI->getCalledValue()))
+        ConstantCallSrc = ConstantPointerRef::get(GV);
+
+      if (ConstantCallSrc)
+        NewCast = ConstantExpr::getCast(ConstantCallSrc, NewPFunTy);
+      else
+        NewCast = new CastInst(CI->getCalledValue(), NewPFunTy,
+                               CI->getCalledValue()->getName()+"_c",CI);
 
       // Create a new call instruction...
       CallInst *NewCall = new CallInst(NewCast,




From lattner at cs.uiuc.edu  Sun Apr 27 22:30:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 22:30:01 2003
Subject: [llvm-commits] CVS: llvm/tools/llc/llc.cpp
Message-ID: <200304280329.WAA22361@apoc.cs.uiuc.edu>


Changes in directory llvm/tools/llc:

llc.cpp updated: 1.70 -> 1.71

---
Log message:

Add a new option to disable stripping of bytecode files


---
Diffs of the changes:

Index: llvm/tools/llc/llc.cpp
diff -u llvm/tools/llc/llc.cpp:1.70 llvm/tools/llc/llc.cpp:1.71
--- llvm/tools/llc/llc.cpp:1.70	Sat Apr 26 15:11:06 2003
+++ llvm/tools/llc/llc.cpp	Sun Apr 27 22:28:56 2003
@@ -50,6 +50,10 @@
 static cl::opt<bool> Force("f", cl::desc("Overwrite output files"));
 
 static cl::opt<bool>
+DisableStrip("disable-strip",
+          cl::desc("Do not strip the LLVM bytecode included in the executable"));
+
+static cl::opt<bool>
 DumpAsm("d", cl::desc("Print bytecode before native code generation"),
         cl::Hidden);
 
@@ -227,7 +231,8 @@
     Passes.add(new PrintFunctionPass("Code after xformations: \n", &std::cerr));
 
   // Strip all of the symbols from the bytecode so that it will be smaller...
-  Passes.add(createSymbolStrippingPass());
+  if (!DisableStrip)
+    Passes.add(createSymbolStrippingPass());
 
   // Figure out where we are going to send the output...
   std::ostream *Out = 0;




From lattner at cs.uiuc.edu  Sun Apr 27 22:41:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 22:41:01 2003
Subject: [llvm-commits] CVS: llvm/www/www-index.html
Message-ID: <200304280340.WAA07731@tank.cs.uiuc.edu>


Changes in directory llvm/www:

www-index.html updated: 1.19 -> 1.20

---
Log message:

Fix grammar


---
Diffs of the changes:

Index: llvm/www/www-index.html
diff -u llvm/www/www-index.html:1.19 llvm/www/www-index.html:1.20
--- llvm/www/www-index.html:1.19	Fri Mar  7 12:04:06 2003
+++ llvm/www/www-index.html	Sun Apr 27 22:40:20 2003
@@ -222,7 +222,7 @@
                         <td valign="top">-</td>
                         <td valign="top">Added a new
                              <a href="docs/OpenProjects.html">open projects</a>
-                             page, listing possibly improvements to LLVM.</td>
+                             page, listing some desired improvements to LLVM.</td>
                     </tr>
                     <tr>
                         <td valign="top"><b>Jan&nbsp;22nd,&nbsp;2003</b></td>




From lattner at cs.uiuc.edu  Sun Apr 27 22:43:01 2003
From: lattner at cs.uiuc.edu (Chris Lattner)
Date: Sun Apr 27 22:43:01 2003
Subject: [llvm-commits] CVS: llvm/www/docs/OpenProjects.html
Message-ID: <200304280342.WAA07769@tank.cs.uiuc.edu>


Changes in directory llvm/www/docs:

OpenProjects.html updated: 1.4 -> 1.5

---
Log message:

Remove implemented project, add a new one


---
Diffs of the changes:

Index: llvm/www/docs/OpenProjects.html
diff -u llvm/www/docs/OpenProjects.html:1.4 llvm/www/docs/OpenProjects.html:1.5
--- llvm/www/docs/OpenProjects.html:1.4	Sat Mar  8 17:59:05 2003
+++ llvm/www/docs/OpenProjects.html	Sun Apr 27 22:42:31 2003
@@ -149,9 +149,7 @@
 <ol>
 <li>Improve the efficiency of the bytecode loader/writer, allow streaming lazy
     loading of functions from the bytecode (for use by the JIT, for example)<br>
-<li>Rework the PassManager
-<li>Do not encode zero intializers for large arrays into the bytecode
-
+<li>Rework the PassManager to be more flexible
 <li>Some transformations and analyses only work on reducible flow graphs.  It
 would be nice to have a transformation which could be "required" by these passes
 which makes irreducible graphs reducible.  This can easily be accomplished
@@ -262,6 +260,8 @@
 <li>Implement a global register allocator
 <li>Implement a better instruction selector
 <li>Implement a static compiler in addition to the JIT (easy project)
+<li>Implement support for the "switch" instruction without requiring the 
+    lower-switches pass.
 </ol>
 
 <!-- ======================================================================= -->