From baldrick at free.fr  Mon Jul 26 02:54:17 2010
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 26 Jul 2010 07:54:17 -0000
Subject: [llvm-commits] [llvm] r109393 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
Message-ID: <20100726075418.06E0D2A6C12D@llvm.org>

Author: baldrick
Date: Mon Jul 26 02:54:17 2010
New Revision: 109393

URL: http://llvm.org/viewvc/llvm-project?rev=109393&view=rev
Log:
Pacify gcc-4.5 which wrongly thinks that RExcess (passed as the Excess parameter)
may be used uninitialized in the callers of HighRegPressure.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp?rev=109393&r1=109392&r2=109393&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp Mon Jul 26 02:54:17 2010
@@ -1192,11 +1192,12 @@
     }
 
     bool HighRegPressure(const SUnit *SU, unsigned &Excess) const {
+      Excess = 0;
+
       if (!TLI)
         return false;
 
       bool High = false;
-      Excess = 0;
       for (SUnit::const_pred_iterator I = SU->Preds.begin(),E = SU->Preds.end();
            I != E; ++I) {
         if (I->isCtrl())




From daniel at zuster.org  Mon Jul 26 10:28:59 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 26 Jul 2010 08:28:59 -0700
Subject: [llvm-commits] [llvm] r109302 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
In-Reply-To: <20100724005322.9283F2A6C12C@llvm.org>
References: <20100724005322.9283F2A6C12C@llvm.org>
Message-ID: <AANLkTin666D3Goh4w_+4DGySWBtj09skOndhZjyacBFH@mail.gmail.com>

Hi Devang,

The backend should not be introducing dependencies on the current
working directory. This kind of stuff needs to be in the frontend.

 - Daniel

On Fri, Jul 23, 2010 at 5:53 PM, Devang Patel <dpatel at apple.com> wrote:
> Author: dpatel
> Date: Fri Jul 23 19:53:22 2010
> New Revision: 109302
>
> URL: http://llvm.org/viewvc/llvm-project?rev=109302&view=rev
> Log:
> Use current working directory when Dirname is empty. This only happens when absolute source file path is used on compiler command line.
>
> Modified:
> ? ?llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=109302&r1=109301&r2=109302&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Fri Jul 23 19:53:22 2010
> @@ -1753,6 +1753,11 @@
> ?/// maps as well.
> ?unsigned DwarfDebug::GetOrCreateSourceID(StringRef DirName, StringRef FileName){
> ? unsigned DId;
> + ?if (DirName.empty()) {
> + ? ?llvm::sys::Path CWD = llvm::sys::Path::GetCurrentDirectory();
> + ? ?DirName = StringRef(CWD.c_str(), CWD.size());
> + ?}
> +
> ? StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
> ? if (DI != DirectoryIdMap.end()) {
> ? ? DId = DI->getValue();
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From gohman at apple.com  Mon Jul 26 11:01:15 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 16:01:15 -0000
Subject: [llvm-commits] [llvm] r109394 - /llvm/trunk/lib/MC/MCDisassembler/
Message-ID: <20100726160115.EA1222A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 11:01:15 2010
New Revision: 109394

URL: http://llvm.org/viewvc/llvm-project?rev=109394&view=rev
Log:
Add an svn:ignore.

Modified:
    llvm/trunk/lib/MC/MCDisassembler/   (props changed)

Propchange: llvm/trunk/lib/MC/MCDisassembler/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Mon Jul 26 11:01:15 2010
@@ -0,0 +1,9 @@
+Debug
+Debug+Checks
+Debug+Coverage
+Debug+Coverage-Asserts
+Release
+Release-Asserts
+Release+Coverage
+Debug+Asserts
+Release+Asserts




From clattner at apple.com  Mon Jul 26 12:05:48 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 26 Jul 2010 10:05:48 -0700
Subject: [llvm-commits] [llvm] r109302 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
In-Reply-To: <AANLkTin666D3Goh4w_+4DGySWBtj09skOndhZjyacBFH@mail.gmail.com>
References: <20100724005322.9283F2A6C12C@llvm.org>
	<AANLkTin666D3Goh4w_+4DGySWBtj09skOndhZjyacBFH@mail.gmail.com>
Message-ID: <8701F036-6C65-4803-AFA8-824E74CD9AB8@apple.com>

Completely agreed.  With a compiler server (for example) the individual compiler instances and the backend may be running in completely different directories.

-Chris

On Jul 26, 2010, at 8:28 AM, Daniel Dunbar wrote:

> Hi Devang,
> 
> The backend should not be introducing dependencies on the current
> working directory. This kind of stuff needs to be in the frontend.
> 
> - Daniel
> 
> On Fri, Jul 23, 2010 at 5:53 PM, Devang Patel <dpatel at apple.com> wrote:
>> Author: dpatel
>> Date: Fri Jul 23 19:53:22 2010
>> New Revision: 109302
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=109302&view=rev
>> Log:
>> Use current working directory when Dirname is empty. This only happens when absolute source file path is used on compiler command line.
>> 
>> Modified:
>>    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>> 
>> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=109302&r1=109301&r2=109302&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
>> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Fri Jul 23 19:53:22 2010
>> @@ -1753,6 +1753,11 @@
>>  /// maps as well.
>>  unsigned DwarfDebug::GetOrCreateSourceID(StringRef DirName, StringRef FileName){
>>   unsigned DId;
>> +  if (DirName.empty()) {
>> +    llvm::sys::Path CWD = llvm::sys::Path::GetCurrentDirectory();
>> +    DirName = StringRef(CWD.c_str(), CWD.size());
>> +  }
>> +
>>   StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
>>   if (DI != DirectoryIdMap.end()) {
>>     DId = DI->getValue();
>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From fjahanian at apple.com  Mon Jul 26 12:14:50 2010
From: fjahanian at apple.com (Fariborz Jahanian)
Date: Mon, 26 Jul 2010 17:14:50 -0000
Subject: [llvm-commits] [test-suite] r109397 -
 /test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp
Message-ID: <20100726171450.612D12A6C12C@llvm.org>

Author: fjahanian
Date: Mon Jul 26 12:14:50 2010
New Revision: 109397

URL: http://llvm.org/viewvc/llvm-project?rev=109397&view=rev
Log:
This test is only for when blocks are enabled.

Modified:
    test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp

Modified: test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp?rev=109397&r1=109396&r2=109397&view=diff
==============================================================================
--- test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp (original)
+++ test-suite/trunk/SingleSource/UnitTests/block-call-r7674133.cpp Mon Jul 26 12:14:50 2010
@@ -1,4 +1,4 @@
-#ifdef __clang__
+#if defined(__BLOCKS__) && defined(__clang__)
 extern "C" void abort(void);
 
 class Test {
@@ -10,7 +10,7 @@
     
     ~Test()
     {
-	i = -1;;
+	i = -1;
     }
     
     Test(const Test&r)
@@ -28,7 +28,7 @@
 
 int main (int , char * const []) 
 {
-#ifdef __clang__
+#if defined(__BLOCKS__) && defined(__clang__)
     Test t;
     ^{  t.speak();  }();
 #endif




From gohman at apple.com  Mon Jul 26 12:34:05 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:34:05 -0000
Subject: [llvm-commits] [llvm] r109398 -
	/llvm/trunk/lib/Analysis/RegionInfo.cpp
Message-ID: <20100726173405.3F8FC2A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:34:05 2010
New Revision: 109398

URL: http://llvm.org/viewvc/llvm-project?rev=109398&view=rev
Log:
Use DominatorTree::properlyDominates instead of dominates with an
explicit inequality check.

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

Modified: llvm/trunk/lib/Analysis/RegionInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionInfo.cpp?rev=109398&r1=109397&r2=109398&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/RegionInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/RegionInfo.cpp Mon Jul 26 12:34:05 2010
@@ -356,7 +356,7 @@
   // Do not allow edges pointing into the region.
   for (DST::iterator SI = exitSuccs->begin(), SE = exitSuccs->end();
        SI != SE; ++SI)
-    if (DT->dominates(entry, *SI) && *SI != entry && *SI != exit)
+    if (DT->properlyDominates(entry, *SI) && *SI != exit)
       return false;
 
 




From baldrick at free.fr  Mon Jul 26 12:34:56 2010
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 26 Jul 2010 17:34:56 -0000
Subject: [llvm-commits] [dragonegg] r109399 - /dragonegg/trunk/llvm-types.cpp
Message-ID: <20100726173456.B96322A6C12C@llvm.org>

Author: baldrick
Date: Mon Jul 26 12:34:56 2010
New Revision: 109399

URL: http://llvm.org/viewvc/llvm-project?rev=109399&view=rev
Log:
Remove extraneous "inline" keyword that was left behind when
this function was was moved out of line.

Modified:
    dragonegg/trunk/llvm-types.cpp

Modified: dragonegg/trunk/llvm-types.cpp
URL: http://llvm.org/viewvc/llvm-project/dragonegg/trunk/llvm-types.cpp?rev=109399&r1=109398&r2=109399&view=diff
==============================================================================
--- dragonegg/trunk/llvm-types.cpp (original)
+++ dragonegg/trunk/llvm-types.cpp Mon Jul 26 12:34:56 2010
@@ -225,7 +225,7 @@
 }
 
 /// getFieldOffsetInBits - Return the bit offset of a FIELD_DECL in a structure.
-inline uint64_t getFieldOffsetInBits(tree field) {
+uint64_t getFieldOffsetInBits(tree field) {
   assert(OffsetIsLLVMCompatible(field) && "Offset is not constant!");
   uint64_t Result = getInt64(DECL_FIELD_BIT_OFFSET(field), true);
   Result += getInt64(DECL_FIELD_OFFSET(field), true) * BITS_PER_UNIT;




From gohman at apple.com  Mon Jul 26 12:35:32 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:35:32 -0000
Subject: [llvm-commits] [llvm] r109400 -
	/llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp
Message-ID: <20100726173532.CC3A72A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:35:32 2010
New Revision: 109400

URL: http://llvm.org/viewvc/llvm-project?rev=109400&view=rev
Log:
A block dominates itself, by definition.

Modified:
    llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp?rev=109400&r1=109399&r2=109400&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp Mon Jul 26 12:35:32 2010
@@ -677,7 +677,7 @@
       for(pred_iterator PI = pred_begin(FrontierBB), PE = pred_end(FrontierBB);
           PI != PE; ++PI) {
         BasicBlock *P = *PI;
-        if (P == DeadBB || DT->dominates(DeadBB, P))
+        if (DT->dominates(DeadBB, P))
           PredBlocks.push_back(P);
       }
 




From gohman at apple.com  Mon Jul 26 12:37:36 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:37:36 -0000
Subject: [llvm-commits] [llvm] r109401 -
	/llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp
Message-ID: <20100726173736.ABAB42A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:37:36 2010
New Revision: 109401

URL: http://llvm.org/viewvc/llvm-project?rev=109401&view=rev
Log:
Use DominatorTree::properlyDominates instead of dominates with an
explicit inequality check.

Modified:
    llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp?rev=109401&r1=109400&r2=109401&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopIndexSplit.cpp Mon Jul 26 12:37:36 2010
@@ -799,7 +799,7 @@
   // the dominance frontiers.
   for (Loop::block_iterator I = LP->block_begin(), E = LP->block_end();
        I != E; ++I) {
-    if (*I == CondBB || !DT->dominates(CondBB, *I)) continue;
+    if (!DT->properlyDominates(CondBB, *I)) continue;
     DominanceFrontier::iterator BBDF = DF->find(*I);
     DominanceFrontier::DomSetType::iterator DomSetI = BBDF->second.begin();
     DominanceFrontier::DomSetType::iterator DomSetE = BBDF->second.end();




From gohman at apple.com  Mon Jul 26 12:38:15 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:38:15 -0000
Subject: [llvm-commits] [llvm] r109402 -
	/llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp
Message-ID: <20100726173815.398792A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:38:15 2010
New Revision: 109402

URL: http://llvm.org/viewvc/llvm-project?rev=109402&view=rev
Log:
A block dominates itself, by definition.

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

Modified: llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp?rev=109402&r1=109401&r2=109402&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp (original)
+++ llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp Mon Jul 26 12:38:15 2010
@@ -338,7 +338,7 @@
     for (SmallPtrSet<InvokeInst*, 32>::iterator
            UI = URoRInvokes.begin(), UE = URoRInvokes.end(); UI != UE; ++UI) {
       const BasicBlock *URoRBB = (*UI)->getParent();
-      if (SelBB == URoRBB || DT->dominates(SelBB, URoRBB)) {
+      if (DT->dominates(SelBB, URoRBB)) {
         SelsToConvert.insert(*SI);
         break;
       }




From gohman at apple.com  Mon Jul 26 12:39:33 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:39:33 -0000
Subject: [llvm-commits] [llvm] r109403 -
	/llvm/trunk/lib/VMCore/Dominators.cpp
Message-ID: <20100726173933.DC39A2A6C12D@llvm.org>

Author: djg
Date: Mon Jul 26 12:39:33 2010
New Revision: 109403

URL: http://llvm.org/viewvc/llvm-project?rev=109403&view=rev
Log:
Fix formatting.

Modified:
    llvm/trunk/lib/VMCore/Dominators.cpp

Modified: llvm/trunk/lib/VMCore/Dominators.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Dominators.cpp?rev=109403&r1=109402&r2=109403&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Dominators.cpp (original)
+++ llvm/trunk/lib/VMCore/Dominators.cpp Mon Jul 26 12:39:33 2010
@@ -123,8 +123,8 @@
 // NewBB is split and now it has one successor. Update dominance frontier to
 // reflect this change.
 void DominanceFrontier::splitBlock(BasicBlock *NewBB) {
-  assert(NewBB->getTerminator()->getNumSuccessors() == 1
-         && "NewBB should have a single successor!");
+  assert(NewBB->getTerminator()->getNumSuccessors() == 1 &&
+         "NewBB should have a single successor!");
   BasicBlock *NewBBSucc = NewBB->getTerminator()->getSuccessor(0);
 
   SmallVector<BasicBlock*, 8> PredBlocks;




From gohman at apple.com  Mon Jul 26 12:41:46 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:41:46 -0000
Subject: [llvm-commits] [llvm] r109404 -
	/llvm/trunk/lib/VMCore/Dominators.cpp
Message-ID: <20100726174146.285E72A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:41:45 2010
New Revision: 109404

URL: http://llvm.org/viewvc/llvm-project?rev=109404&view=rev
Log:
Exit a search loop when the search goal is found.

Modified:
    llvm/trunk/lib/VMCore/Dominators.cpp

Modified: llvm/trunk/lib/VMCore/Dominators.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Dominators.cpp?rev=109404&r1=109403&r2=109404&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Dominators.cpp (original)
+++ llvm/trunk/lib/VMCore/Dominators.cpp Mon Jul 26 12:41:45 2010
@@ -161,8 +161,10 @@
         bool DominatesPred = false;
         for (pred_iterator PI = pred_begin(*SetI), E = pred_end(*SetI);
              PI != E; ++PI)
-          if (DT.dominates(NewBB, *PI))
+          if (DT.dominates(NewBB, *PI)) {
             DominatesPred = true;
+            break;
+          }
         if (!DominatesPred)
           Set.erase(SetI++);
         else




From gohman at apple.com  Mon Jul 26 12:45:33 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:45:33 -0000
Subject: [llvm-commits] [llvm] r109405 -
	/llvm/trunk/lib/VMCore/Dominators.cpp
Message-ID: <20100726174533.5D53E2A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:45:33 2010
New Revision: 109405

URL: http://llvm.org/viewvc/llvm-project?rev=109405&view=rev
Log:
Avoid copying and recopying a std::set.

Modified:
    llvm/trunk/lib/VMCore/Dominators.cpp

Modified: llvm/trunk/lib/VMCore/Dominators.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Dominators.cpp?rev=109405&r1=109404&r2=109405&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Dominators.cpp (original)
+++ llvm/trunk/lib/VMCore/Dominators.cpp Mon Jul 26 12:45:33 2010
@@ -140,12 +140,8 @@
 
   // NewBBSucc inherits original NewBB frontier.
   DominanceFrontier::iterator NewBBI = find(NewBB);
-  if (NewBBI != end()) {
-    DominanceFrontier::DomSetType NewBBSet = NewBBI->second;
-    DominanceFrontier::DomSetType NewBBSuccSet;
-    NewBBSuccSet.insert(NewBBSet.begin(), NewBBSet.end());
-    addBasicBlock(NewBBSucc, NewBBSuccSet);
-  }
+  if (NewBBI != end())
+    addBasicBlock(NewBBSucc, NewBBI->second);
 
   // If NewBB dominates NewBBSucc, then DF(NewBB) is now going to be the
   // DF(NewBBSucc) without the stuff that the new block does not dominate




From echristo at apple.com  Mon Jul 26 12:53:07 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 17:53:07 -0000
Subject: [llvm-commits] [llvm] r109407 -
	/llvm/trunk/utils/TableGen/FastISelEmitter.cpp
Message-ID: <20100726175307.C5FC32A6C12C@llvm.org>

Author: echristo
Date: Mon Jul 26 12:53:07 2010
New Revision: 109407

URL: http://llvm.org/viewvc/llvm-project?rev=109407&view=rev
Log:
80-col.

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

Modified: llvm/trunk/utils/TableGen/FastISelEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/FastISelEmitter.cpp?rev=109407&r1=109406&r2=109407&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/FastISelEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/FastISelEmitter.cpp Mon Jul 26 12:53:07 2010
@@ -208,7 +208,8 @@
   typedef std::map<MVT::SimpleValueType, PredMap> RetPredMap;
   typedef std::map<MVT::SimpleValueType, RetPredMap> TypeRetPredMap;
   typedef std::map<std::string, TypeRetPredMap> OpcodeTypeRetPredMap;
-  typedef std::map<OperandsSignature, OpcodeTypeRetPredMap> OperandsOpcodeTypeRetPredMap;
+  typedef std::map<OperandsSignature, OpcodeTypeRetPredMap> 
+            OperandsOpcodeTypeRetPredMap;
 
   OperandsOpcodeTypeRetPredMap SimplePatterns;
 
@@ -375,7 +376,8 @@
       SubRegNo,
       PhysRegInputs
     };
-    assert(!SimplePatterns[Operands][OpcodeName][VT][RetVT].count(PredicateCheck) &&
+    assert(!SimplePatterns[Operands][OpcodeName][VT][RetVT]
+            .count(PredicateCheck) &&
            "Duplicate pattern!");
     SimplePatterns[Operands][OpcodeName][VT][RetVT][PredicateCheck] = Memo;
   }




From gohman at apple.com  Mon Jul 26 12:55:15 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 17:55:15 -0000
Subject: [llvm-commits] [llvm] r109408 -
	/llvm/trunk/lib/VMCore/Dominators.cpp
Message-ID: <20100726175515.B76F12A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 12:55:15 2010
New Revision: 109408

URL: http://llvm.org/viewvc/llvm-project?rev=109408&view=rev
Log:
Fix (at least) quadratic worst-case complexity in DominanceFrontier::splitBlock:
don't visit all blocks in the function, and don't iterate over the split blocks'
predecessor lists for each block visited.

Also, remove the special-case test for the entry block. Splitting the entry
block isn't common enough to make this worthwhile.

This fixes a major compile-time bottleneck which is exposed now that
LoopSimplify isn't being redundantly run both before and after
DominanceFrontier.

Modified:
    llvm/trunk/lib/VMCore/Dominators.cpp

Modified: llvm/trunk/lib/VMCore/Dominators.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Dominators.cpp?rev=109408&r1=109407&r2=109408&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Dominators.cpp (original)
+++ llvm/trunk/lib/VMCore/Dominators.cpp Mon Jul 26 12:55:15 2010
@@ -127,17 +127,6 @@
          "NewBB should have a single successor!");
   BasicBlock *NewBBSucc = NewBB->getTerminator()->getSuccessor(0);
 
-  SmallVector<BasicBlock*, 8> PredBlocks;
-  for (pred_iterator PI = pred_begin(NewBB), PE = pred_end(NewBB);
-       PI != PE; ++PI)
-    PredBlocks.push_back(*PI);  
-
-  if (PredBlocks.empty())
-    // If NewBB does not have any predecessors then it is a entry block.
-    // In this case, NewBB and its successor NewBBSucc dominates all
-    // other blocks.
-    return;
-
   // NewBBSucc inherits original NewBB frontier.
   DominanceFrontier::iterator NewBBI = find(NewBB);
   if (NewBBI != end())
@@ -147,7 +136,9 @@
   // DF(NewBBSucc) without the stuff that the new block does not dominate
   // a predecessor of.
   DominatorTree &DT = getAnalysis<DominatorTree>();
-  if (DT.dominates(NewBB, NewBBSucc)) {
+  DomTreeNode *NewBBNode = DT.getNode(NewBB);
+  DomTreeNode *NewBBSuccNode = DT.getNode(NewBBSucc);
+  if (DT.dominates(NewBBNode, NewBBSuccNode)) {
     DominanceFrontier::iterator DFI = find(NewBBSucc);
     if (DFI != end()) {
       DominanceFrontier::DomSetType Set = DFI->second;
@@ -157,7 +148,7 @@
         bool DominatesPred = false;
         for (pred_iterator PI = pred_begin(*SetI), E = pred_end(*SetI);
              PI != E; ++PI)
-          if (DT.dominates(NewBB, *PI)) {
+          if (DT.dominates(NewBBNode, DT.getNode(*PI))) {
             DominatesPred = true;
             break;
           }
@@ -185,50 +176,71 @@
     NewDFSet.insert(NewBBSucc);
     addBasicBlock(NewBB, NewDFSet);
   }
-  
-  // Now we must loop over all of the dominance frontiers in the function,
-  // replacing occurrences of NewBBSucc with NewBB in some cases.  All
-  // blocks that dominate a block in PredBlocks and contained NewBBSucc in
-  // their dominance frontier must be updated to contain NewBB instead.
-  //
-  for (Function::iterator FI = NewBB->getParent()->begin(),
-         FE = NewBB->getParent()->end(); FI != FE; ++FI) {
-    DominanceFrontier::iterator DFI = find(FI);
-    if (DFI == end()) continue;  // unreachable block.
-    
-    // Only consider nodes that have NewBBSucc in their dominator frontier.
-    if (!DFI->second.count(NewBBSucc)) continue;
 
-    // Verify whether this block dominates a block in predblocks.  If not, do
-    // not update it.
-    bool BlockDominatesAny = false;
-    for (SmallVectorImpl<BasicBlock*>::const_iterator BI = PredBlocks.begin(), 
-           BE = PredBlocks.end(); BI != BE; ++BI) {
-      if (DT.dominates(FI, *BI)) {
-        BlockDominatesAny = true;
+  // Now update dominance frontiers which either used to contain NewBBSucc
+  // or which now need to include NewBB.
+
+  // Collect the set of blocks which dominate a predecessor of NewBB or
+  // NewSuccBB and which don't dominate both. This is an initial
+  // approximation of the blocks whose dominance frontiers will need updates.
+  SmallVector<DomTreeNode *, 16> AllPredDoms;
+
+  // Compute the block which dominates both NewBBSucc and NewBB. This is
+  // the immediate dominator of NewBBSucc unless NewBB dominates NewBBSucc.
+  // The code below which climbs dominator trees will stop at this point,
+  // because from this point up, dominance frontiers are unaffected.
+  DomTreeNode *DominatesBoth = 0;
+  if (NewBBSuccNode) {
+    DominatesBoth = NewBBSuccNode->getIDom();
+    if (DominatesBoth == NewBBNode)
+      DominatesBoth = NewBBNode->getIDom();
+  }
+
+  // Collect the set of all blocks which dominate a predecessor of NewBB.
+  SmallPtrSet<DomTreeNode *, 8> NewBBPredDoms;
+  for (pred_iterator PI = pred_begin(NewBB), E = pred_end(NewBB); PI != E; ++PI)
+    for (DomTreeNode *DTN = DT.getNode(*PI); DTN; DTN = DTN->getIDom()) {
+      if (DTN == DominatesBoth)
         break;
-      }
+      if (!NewBBPredDoms.insert(DTN))
+        break;
+      AllPredDoms.push_back(DTN);
     }
 
-    // If NewBBSucc should not stay in our dominator frontier, remove it.
-    // We remove it unless there is a predecessor of NewBBSucc that we
-    // dominate, but we don't strictly dominate NewBBSucc.
-    bool ShouldRemove = true;
-    if ((BasicBlock*)FI == NewBBSucc || !DT.dominates(FI, NewBBSucc)) {
-      // Okay, we know that PredDom does not strictly dominate NewBBSucc.
-      // Check to see if it dominates any predecessors of NewBBSucc.
-      for (pred_iterator PI = pred_begin(NewBBSucc),
-           E = pred_end(NewBBSucc); PI != E; ++PI)
-        if (DT.dominates(FI, *PI)) {
-          ShouldRemove = false;
-          break;
-        }
+  // Collect the set of all blocks which dominate a predecessor of NewSuccBB.
+  SmallPtrSet<DomTreeNode *, 8> NewBBSuccPredDoms;
+  for (pred_iterator PI = pred_begin(NewBBSucc),
+       E = pred_end(NewBBSucc); PI != E; ++PI)
+    for (DomTreeNode *DTN = DT.getNode(*PI); DTN; DTN = DTN->getIDom()) {
+      if (DTN == DominatesBoth)
+        break;
+      if (!NewBBSuccPredDoms.insert(DTN))
+        break;
+      if (!NewBBPredDoms.count(DTN))
+        AllPredDoms.push_back(DTN);
     }
-    
-    if (ShouldRemove)
-      removeFromFrontier(DFI, NewBBSucc);
-    if (BlockDominatesAny && (&*FI == NewBB || !DT.dominates(FI, NewBB)))
+
+  // Visit all relevant dominance frontiers and make any needed updates.
+  for (SmallVectorImpl<DomTreeNode *>::const_iterator I = AllPredDoms.begin(),
+       E = AllPredDoms.end(); I != E; ++I) {
+    DomTreeNode *DTN = *I;
+    iterator DFI = find((*I)->getBlock());
+
+    // Only consider nodes that have NewBBSucc in their dominator frontier.
+    if (DFI == end() || !DFI->second.count(NewBBSucc)) continue;
+
+    // If the block dominates a predecessor of NewBB but does not properly
+    // dominate NewBB itself, add NewBB to its dominance frontier.
+    if (NewBBPredDoms.count(DTN) &&
+        !DT.properlyDominates(DTN, NewBBNode))
       addToFrontier(DFI, NewBB);
+
+    // If the block does not dominate a predecessor of NewBBSucc or
+    // properly dominates NewBBSucc itself, remove NewBBSucc from its
+    // dominance frontier.
+    if (!NewBBSuccPredDoms.count(DTN) ||
+        DT.properlyDominates(DTN, NewBBSuccNode))
+      removeFromFrontier(DFI, NewBBSucc);
   }
 }
 




From resistor at mac.com  Mon Jul 26 12:56:43 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 17:56:43 -0000
Subject: [llvm-commits] [llvm] r109409 - /llvm/trunk/README.txt
Message-ID: <20100726175643.225AA2A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 12:56:42 2010
New Revision: 109409

URL: http://llvm.org/viewvc/llvm-project?rev=109409&view=rev
Log:
Test commit for new post-commit hooks.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109409&r1=109408&r2=109409&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 12:56:42 2010
@@ -13,3 +13,4 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
+




From dpatel at apple.com  Mon Jul 26 12:57:29 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 26 Jul 2010 10:57:29 -0700
Subject: [llvm-commits] [llvm] r109302 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
In-Reply-To: <8701F036-6C65-4803-AFA8-824E74CD9AB8@apple.com>
References: <20100724005322.9283F2A6C12C@llvm.org>
	<AANLkTin666D3Goh4w_+4DGySWBtj09skOndhZjyacBFH@mail.gmail.com>
	<8701F036-6C65-4803-AFA8-824E74CD9AB8@apple.com>
Message-ID: <EDBE0A24-134F-432F-84E6-8BF07BEDF9A9@apple.com>


On Jul 26, 2010, at 10:05 AM, Chris Lattner wrote:

> Completely agreed.  With a compiler server (for example) the individual compiler instances and the backend may be running in completely different directories.

OK. I'll use another approach.
-
Devang
> 
> -Chris
> 
> On Jul 26, 2010, at 8:28 AM, Daniel Dunbar wrote:
> 
>> Hi Devang,
>> 
>> The backend should not be introducing dependencies on the current
>> working directory. This kind of stuff needs to be in the frontend.
>> 
>> - Daniel
>> 
>> On Fri, Jul 23, 2010 at 5:53 PM, Devang Patel <dpatel at apple.com> wrote:
>>> Author: dpatel
>>> Date: Fri Jul 23 19:53:22 2010
>>> New Revision: 109302
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=109302&view=rev
>>> Log:
>>> Use current working directory when Dirname is empty. This only happens when absolute source file path is used on compiler command line.
>>> 
>>> Modified:
>>>   llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>>> 
>>> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=109302&r1=109301&r2=109302&view=diff
>>> ==============================================================================
>>> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
>>> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Fri Jul 23 19:53:22 2010
>>> @@ -1753,6 +1753,11 @@
>>> /// maps as well.
>>> unsigned DwarfDebug::GetOrCreateSourceID(StringRef DirName, StringRef FileName){
>>>  unsigned DId;
>>> +  if (DirName.empty()) {
>>> +    llvm::sys::Path CWD = llvm::sys::Path::GetCurrentDirectory();
>>> +    DirName = StringRef(CWD.c_str(), CWD.size());
>>> +  }
>>> +
>>>  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
>>>  if (DI != DirectoryIdMap.end()) {
>>>    DId = DI->getValue();
>>> 
>>> 
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 




From ggreif at gmail.com  Mon Jul 26 12:58:32 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Mon, 26 Jul 2010 19:58:32 +0200
Subject: [llvm-commits] [llvm] r109220 - /llvm/trunk/include/llvm/Use.h
In-Reply-To: <CD914016-0F18-4C2E-A661-F6408D972618@apple.com>
References: <20100723083607.9183D2A6C12C@llvm.org>
	<4B1D9A59-3530-4C93-8436-F0F1A4B947F3@apple.com>
	<FAD12E3E-0D3F-4A58-9A9D-B7FD9C466923@apple.com>
	<730b40a2-8a73-4301-8f29-58b09716509a@z10g2000yqb.googlegroups.com>
	<CD914016-0F18-4C2E-A661-F6408D972618@apple.com>
Message-ID: <AANLkTi=e1Zzh3rks7CDs6unYjveFmegNb_uLWQJEfzp7@mail.gmail.com>

I could not reproduce a build failure in llvm-gcc. What error did you get?

   Gabor

On 7/23/10, Eric Christopher <echristo at apple.com> wrote:
>
> On Jul 23, 2010, at 2:08 AM, Gabor Greif wrote:
>
>> Alright, thanks!
>>
>> Sorry for the damage, I'll take care of llvm-gcc this week.
>
> Clang too please.
>
>> Will write up a small blurb for the release notes too.
>>
>
> Thanks!  I'm also tracking down a miscompile that's going on after one of
> these patches.  It's down to 109100(pass)-109104(fails) so I'm going to
> probably revert whichever ones will make my miscompile go away also.  It's a
> bad stack smashing while jitting problem and I'm on vacation for a few days
> so I want to get my builder clean before that.
>
> I'll work on a testcase of what went wrong after that if it's not clear when
> I revert :)
>
>> Btw. why is the buildbot site down?
>
> No idea :)
>
> -eric


From echristo at apple.com  Mon Jul 26 12:59:59 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 10:59:59 -0700
Subject: [llvm-commits] [llvm] r109220 - /llvm/trunk/include/llvm/Use.h
In-Reply-To: <AANLkTi=e1Zzh3rks7CDs6unYjveFmegNb_uLWQJEfzp7@mail.gmail.com>
References: <20100723083607.9183D2A6C12C@llvm.org>
	<4B1D9A59-3530-4C93-8436-F0F1A4B947F3@apple.com>
	<FAD12E3E-0D3F-4A58-9A9D-B7FD9C466923@apple.com>
	<730b40a2-8a73-4301-8f29-58b09716509a@z10g2000yqb.googlegroups.com>
	<CD914016-0F18-4C2E-A661-F6408D972618@apple.com>
	<AANLkTi=e1Zzh3rks7CDs6unYjveFmegNb_uLWQJEfzp7@mail.gmail.com>
Message-ID: <C3BB9A33-184B-42AB-A603-526FF5C70883@apple.com>


On Jul 26, 2010, at 10:58 AM, Gabor Greif wrote:

> I could not reproduce a build failure in llvm-gcc. What error did you get?

>> 
>> Thanks!  I'm also tracking down a miscompile that's going on after one of
>> these patches.  It's down to 109100(pass)-109104(fails) so I'm going to
>> probably revert whichever ones will make my miscompile go away also.  It's a
>> bad stack smashing while jitting problem and I'm on vacation for a few days
>> so I want to get my builder clean before that.

This was a different patch, it was the virtual register one (the only one in the range
you didn't do actually) and I reverted it Friday morning.

Thanks!

-eric




From gohman at apple.com  Mon Jul 26 13:00:43 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 18:00:43 -0000
Subject: [llvm-commits] [llvm] r109410 -
 /llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll
Message-ID: <20100726180043.3A9922A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 13:00:43 2010
New Revision: 109410

URL: http://llvm.org/viewvc/llvm-project?rev=109410&view=rev
Log:
Avoid depending on LCSSA implicitly pulling in LoopSimplify.

Modified:
    llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll

Modified: llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll?rev=109410&r1=109409&r2=109410&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll (original)
+++ llvm/trunk/test/Transforms/LCSSA/2006-06-03-IncorrectIDFPhis.ll Mon Jul 26 13:00:43 2010
@@ -1,6 +1,6 @@
-; RUN: opt < %s -lcssa -S | \
+; RUN: opt < %s -loopsimplify -lcssa -S | \
 ; RUN:   grep {%%SJE.0.0.lcssa = phi .struct.SetJmpMapEntry}
-; RUN: opt < %s -lcssa -S | \
+; RUN: opt < %s -loopsimplify -lcssa -S | \
 ; RUN:   grep {%%SJE.0.0.lcssa1 = phi .struct.SetJmpMapEntry}
 
         %struct.SetJmpMapEntry = type { i8*, i32, %struct.SetJmpMapEntry* }




From resistor at mac.com  Mon Jul 26 13:00:59 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 18:00:59 -0000
Subject: [llvm-commits] [llvm] r109411 - /llvm/trunk/README.txt
Message-ID: <20100726180059.AC7B02A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 13:00:59 2010
New Revision: 109411

URL: http://llvm.org/viewvc/llvm-project?rev=109411&view=rev
Log:
Another test commit.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109411&r1=109410&r2=109411&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 13:00:59 2010
@@ -13,4 +13,3 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
-




From gohman at apple.com  Mon Jul 26 13:02:06 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 18:02:06 -0000
Subject: [llvm-commits] [llvm] r109412 -
	/llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp
Message-ID: <20100726180206.9CA9E2A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 13:02:06 2010
New Revision: 109412

URL: http://llvm.org/viewvc/llvm-project?rev=109412&view=rev
Log:
Preserve ScalarEvolution in the loop unroller.

Modified:
    llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp

Modified: llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp?rev=109412&r1=109411&r2=109412&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LoopUnroll.cpp Mon Jul 26 13:02:06 2010
@@ -24,6 +24,7 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Analysis/ConstantFolding.h"
 #include "llvm/Analysis/LoopPass.h"
+#include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
@@ -127,6 +128,11 @@
     return false;
   }
 
+  // Notify ScalarEvolution that the loop will be substantially changed,
+  // if not outright eliminated.
+  if (ScalarEvolution *SE = LPM->getAnalysisIfAvailable<ScalarEvolution>())
+    SE->forgetLoop(L);
+
   // Find trip count
   unsigned TripCount = L->getSmallConstantTripCount();
   // Find trip multiple if count is not available




From gohman at apple.com  Mon Jul 26 13:11:16 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 18:11:16 -0000
Subject: [llvm-commits] [llvm] r109413 - in /llvm/trunk:
 include/llvm/Support/StandardPasses.h
 lib/Transforms/Scalar/LoopDeletion.cpp
 lib/Transforms/Scalar/LoopUnrollPass.cpp lib/Transforms/Utils/LCSSA.cpp
 lib/Transforms/Utils/LoopSimplify.cpp
Message-ID: <20100726181116.5345A2A6C131@llvm.org>

Author: djg
Date: Mon Jul 26 13:11:16 2010
New Revision: 109413

URL: http://llvm.org/viewvc/llvm-project?rev=109413&view=rev
Log:
Remove LCSSA's bogus dependence on LoopSimplify and LoopSimplify's bogus
dependence on DominanceFrontier. Instead, add an explicit DominanceFrontier
pass in StandardPasses.h to ensure that it gets scheduled at the right
time.

Declare that loop unrolling preserves ScalarEvolution, and shuffle some
getAnalysisUsages.

This eliminates one LoopSimplify and one LCCSA run in the standard
compile opts sequence.

Modified:
    llvm/trunk/include/llvm/Support/StandardPasses.h
    llvm/trunk/lib/Transforms/Scalar/LoopDeletion.cpp
    llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp
    llvm/trunk/lib/Transforms/Utils/LCSSA.cpp
    llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp

Modified: llvm/trunk/include/llvm/Support/StandardPasses.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/StandardPasses.h?rev=109413&r1=109412&r2=109413&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/StandardPasses.h (original)
+++ llvm/trunk/include/llvm/Support/StandardPasses.h Mon Jul 26 13:11:16 2010
@@ -20,6 +20,7 @@
 #define LLVM_SUPPORT_STANDARDPASSES_H
 
 #include "llvm/PassManager.h"
+#include "llvm/Analysis/Dominators.h"
 #include "llvm/Analysis/Passes.h"
 #include "llvm/Analysis/Verifier.h"
 #include "llvm/Transforms/Scalar.h"
@@ -128,6 +129,8 @@
     PM->add(createTailCallEliminationPass());   // Eliminate tail calls
     PM->add(createCFGSimplificationPass());     // Merge & remove BBs
     PM->add(createReassociatePass());           // Reassociate expressions
+    // Explicitly schedule this to ensure that it runs before any loop pass.
+    PM->add(new DominanceFrontier());           // Calculate Dominance Frontiers
     PM->add(createLoopRotatePass());            // Rotate Loop
     PM->add(createLICMPass());                  // Hoist loop invariants
     PM->add(createLoopUnswitchPass(OptimizeSize || OptimizationLevel < 3));

Modified: llvm/trunk/lib/Transforms/Scalar/LoopDeletion.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopDeletion.cpp?rev=109413&r1=109412&r2=109413&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopDeletion.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopDeletion.cpp Mon Jul 26 13:11:16 2010
@@ -38,9 +38,9 @@
                     bool &Changed, BasicBlock *Preheader);
 
     virtual void getAnalysisUsage(AnalysisUsage& AU) const {
-      AU.addRequired<ScalarEvolution>();
       AU.addRequired<DominatorTree>();
       AU.addRequired<LoopInfo>();
+      AU.addRequired<ScalarEvolution>();
       AU.addRequiredID(LoopSimplifyID);
       AU.addRequiredID(LCSSAID);
       

Modified: llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp?rev=109413&r1=109412&r2=109413&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopUnrollPass.cpp Mon Jul 26 13:11:16 2010
@@ -17,6 +17,7 @@
 #include "llvm/Transforms/Scalar.h"
 #include "llvm/Analysis/LoopPass.h"
 #include "llvm/Analysis/InlineCost.h"
+#include "llvm/Analysis/ScalarEvolution.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
@@ -55,17 +56,19 @@
     /// loop preheaders be inserted into the CFG...
     ///
     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+      AU.addRequired<LoopInfo>();
+      AU.addPreserved<LoopInfo>();
       AU.addRequiredID(LoopSimplifyID);
+      AU.addPreservedID(LoopSimplifyID);
       AU.addRequiredID(LCSSAID);
-      AU.addRequired<LoopInfo>();
       AU.addPreservedID(LCSSAID);
-      AU.addPreserved<LoopInfo>();
       // FIXME: Loop unroll requires LCSSA. And LCSSA requires dom info.
       // If loop unroll does not preserve dom info then LCSSA pass on next
       // loop will receive invalid dom info.
       // For now, recreate dom info, if loop is unrolled.
       AU.addPreserved<DominatorTree>();
       AU.addPreserved<DominanceFrontier>();
+      AU.addPreserved<ScalarEvolution>();
     }
   };
 }

Modified: llvm/trunk/lib/Transforms/Utils/LCSSA.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LCSSA.cpp?rev=109413&r1=109412&r2=109413&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LCSSA.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LCSSA.cpp Mon Jul 26 13:11:16 2010
@@ -64,17 +64,12 @@
     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
       AU.setPreservesCFG();
 
-      AU.addRequiredTransitive<DominatorTree>();
+      AU.addRequired<DominatorTree>();
       AU.addPreserved<DominatorTree>();
       AU.addPreserved<DominanceFrontier>();
-      AU.addRequiredTransitive<LoopInfo>();
+      AU.addRequired<LoopInfo>();
       AU.addPreserved<LoopInfo>();
-
-      // LCSSA doesn't actually require LoopSimplify, but the PassManager
-      // doesn't know how to schedule LoopSimplify by itself.
-      AU.addRequiredID(LoopSimplifyID);
       AU.addPreservedID(LoopSimplifyID);
-
       AU.addPreserved<ScalarEvolution>();
     }
   private:

Modified: llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp?rev=109413&r1=109412&r2=109413&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/LoopSimplify.cpp Mon Jul 26 13:11:16 2010
@@ -77,22 +77,17 @@
 
     virtual void getAnalysisUsage(AnalysisUsage &AU) const {
       // We need loop information to identify the loops...
-      AU.addRequiredTransitive<DominatorTree>();
+      AU.addRequired<DominatorTree>();
       AU.addPreserved<DominatorTree>();
 
-      // Request DominanceFrontier now, even though LoopSimplify does
-      // not use it. This allows Pass Manager to schedule Dominance
-      // Frontier early enough such that one LPPassManager can handle
-      // multiple loop transformation passes.
-      AU.addRequired<DominanceFrontier>();
-      AU.addPreserved<DominanceFrontier>();
-
-      AU.addRequiredTransitive<LoopInfo>();
+      AU.addRequired<LoopInfo>();
       AU.addPreserved<LoopInfo>();
 
       AU.addPreserved<AliasAnalysis>();
       AU.addPreserved<ScalarEvolution>();
       AU.addPreservedID(BreakCriticalEdgesID);  // No critical edges added.
+      AU.addPreserved<DominanceFrontier>();
+      AU.addPreservedID(LCSSAID);
     }
 
     /// verifyAnalysis() - Verify LoopSimplifyForm's guarantees.




From resistor at mac.com  Mon Jul 26 13:13:21 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 18:13:21 -0000
Subject: [llvm-commits] [llvm] r109414 - /llvm/trunk/README.txt
Message-ID: <20100726181321.855852A6C131@llvm.org>

Author: resistor
Date: Mon Jul 26 13:13:21 2010
New Revision: 109414

URL: http://llvm.org/viewvc/llvm-project?rev=109414&view=rev
Log:
Hopefully the last test commit.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109414&r1=109413&r2=109414&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 13:13:21 2010
@@ -13,3 +13,4 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
+




From gohman at apple.com  Mon Jul 26 13:15:42 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 18:15:42 -0000
Subject: [llvm-commits] [llvm] r109415 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
Message-ID: <20100726181542.1C0CE2A6C131@llvm.org>

Author: djg
Date: Mon Jul 26 13:15:41 2010
New Revision: 109415

URL: http://llvm.org/viewvc/llvm-project?rev=109415&view=rev
Log:
Handle Values with no value in getCopyFromRegs.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp?rev=109415&r1=109414&r2=109415&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp Mon Jul 26 13:15:41 2010
@@ -530,6 +530,10 @@
                                       FunctionLoweringInfo &FuncInfo,
                                       DebugLoc dl,
                                       SDValue &Chain, SDValue *Flag) const {
+  // A Value with type {} or [0 x %t] needs no registers.
+  if (ValueVTs.empty())
+    return SDValue();
+
   const TargetLowering &TLI = DAG.getTargetLoweringInfo();
 
   // Assemble the legal parts into the final values.




From scanon at apple.com  Mon Jul 26 13:17:00 2010
From: scanon at apple.com (Stephen Canon)
Date: Mon, 26 Jul 2010 18:17:00 -0000
Subject: [llvm-commits] [compiler-rt] r109416 -
	/compiler-rt/trunk/lib/fp_lib.h
Message-ID: <20100726181700.737F42A6C130@llvm.org>

Author: scanon
Date: Mon Jul 26 13:17:00 2010
New Revision: 109416

URL: http://llvm.org/viewvc/llvm-project?rev=109416&view=rev
Log:
Fix error in rep_clz on non-LP64 targets.  Patch by Christoph Gerum

Modified:
    compiler-rt/trunk/lib/fp_lib.h

Modified: compiler-rt/trunk/lib/fp_lib.h
URL: http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/fp_lib.h?rev=109416&r1=109415&r2=109416&view=diff
==============================================================================
--- compiler-rt/trunk/lib/fp_lib.h (original)
+++ compiler-rt/trunk/lib/fp_lib.h Mon Jul 26 13:17:00 2010
@@ -57,9 +57,9 @@
     return __builtin_clzl(a);
 #else
     if (a & REP_C(0xffffffff00000000))
-        return 32 + __builtin_clz(a >> 32);
+        return __builtin_clz(a >> 32);
     else 
-        return __builtin_clz(a & REP_C(0xffffffff));
+        return 32 + __builtin_clz(a & REP_C(0xffffffff));
 #endif
 }
 




From dpatel at apple.com  Mon Jul 26 13:23:40 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 26 Jul 2010 11:23:40 -0700
Subject: [llvm-commits] [llvm] r109413 - in /llvm/trunk:
	include/llvm/Support/StandardPasses.h
	lib/Transforms/Scalar/LoopDeletion.cpp
	lib/Transforms/Scalar/LoopUnrollPass.cpp
	lib/Transforms/Utils/LCSSA.cpp lib/Transforms/Utils/LoopSimplify.cpp
In-Reply-To: <20100726181116.5345A2A6C131@llvm.org>
References: <20100726181116.5345A2A6C131@llvm.org>
Message-ID: <DB402B96-1DAA-4A95-A2E7-587F0ECA43EF@apple.com>


On Jul 26, 2010, at 11:11 AM, Dan Gohman wrote:

> Author: djg
> Date: Mon Jul 26 13:11:16 2010
> New Revision: 109413
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109413&view=rev
> Log:
> Remove LCSSA's bogus dependence on LoopSimplify and LoopSimplify's bogus
> dependence on DominanceFrontier. Instead, add an explicit DominanceFrontier
> pass in StandardPasses.h to ensure that it gets scheduled at the right
> time.

If a Pass breaks because DominanceFrontier was not explicitly added in pass manager queue then the pass is broken. A pass is supposed to request all the required analysis itself.

$ opt -mypass 

should just work even if mypass needs DominanceFrontier. It is mypass's responsibility to request dominance frontier.

Enforcing `$ opt -domfrontier -mypass` requirement completely goes against getAnalysisUsage() mechanism.
-
Devang

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100726/45ed2411/attachment.html 

From evan.cheng at apple.com  Mon Jul 26 13:27:30 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 26 Jul 2010 11:27:30 -0700
Subject: [llvm-commits] [llvm] r109047 - in /llvm/trunk:
	lib/Target/ARM/ARMFastISel.cpp lib/Target/ARM/ARMISelLowering.cpp
	lib/Target/ARM/ARMISelLowering.h
	lib/Target/ARM/CMakeLists.txt lib/Target/ARM/Makefile
	test/CodeGen/ARM/fast-isel.ll
In-Reply-To: <20100721222611.E758A2A6C12C@llvm.org>
References: <20100721222611.E758A2A6C12C@llvm.org>
Message-ID: <795C641C-8137-416C-BEC8-5F65A16C1C8B@apple.com>

This is breaking ARM -O0. e.g.

unsigned FastEmit_ISD_Constant_MVT_i32_i(MVT RetVT, uint64_t imm0) {
  if (RetVT.SimpleTy != MVT::i32)
    return 0;
  if ((!Subtarget->isThumb()) && (Subtarget->hasV6T2Ops())) {
    return FastEmitInst_i(ARM::MOVi32imm, ARM::GPRRegisterClass, imm0);
  }
  if ((Subtarget->isThumb2())) {
=>  return FastEmitInst_i(ARM::t2MOVi32imm, ARM::GPRRegisterClass, imm0);
  }
  return 0;
}

But FastISel::FastEmitInst_i isn't creating the right instructions for ARM since it doesn't add the default predicate operands. I'm going to disable fastisel for ARM for now.

Evan

On Jul 21, 2010, at 3:26 PM, Eric Christopher wrote:

> Author: echristo
> Date: Wed Jul 21 17:26:11 2010
> New Revision: 109047
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109047&view=rev
> Log:
> Baby steps towards ARM fast-isel.
> 
> Added:
>    llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
>    llvm/trunk/test/CodeGen/ARM/fast-isel.ll
> Modified:
>    llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
>    llvm/trunk/lib/Target/ARM/ARMISelLowering.h
>    llvm/trunk/lib/Target/ARM/CMakeLists.txt
>    llvm/trunk/lib/Target/ARM/Makefile
> 
> Added: llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMFastISel.cpp?rev=109047&view=auto
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMFastISel.cpp (added)
> +++ llvm/trunk/lib/Target/ARM/ARMFastISel.cpp Wed Jul 21 17:26:11 2010
> @@ -0,0 +1,71 @@
> +//===-- ARMFastISel.cpp - ARM FastISel implementation ---------------------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +//
> +// This file defines the ARM-specific support for the FastISel class. Some
> +// of the target-specific code is generated by tablegen in the file
> +// ARMGenFastISel.inc, which is #included here.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#include "ARM.h"
> +#include "ARMRegisterInfo.h"
> +#include "ARMTargetMachine.h"
> +#include "ARMSubtarget.h"
> +#include "llvm/CallingConv.h"
> +#include "llvm/DerivedTypes.h"
> +#include "llvm/GlobalVariable.h"
> +#include "llvm/Instructions.h"
> +#include "llvm/IntrinsicInst.h"
> +#include "llvm/CodeGen/Analysis.h"
> +#include "llvm/CodeGen/FastISel.h"
> +#include "llvm/CodeGen/FunctionLoweringInfo.h"
> +#include "llvm/CodeGen/MachineConstantPool.h"
> +#include "llvm/CodeGen/MachineFrameInfo.h"
> +#include "llvm/CodeGen/MachineRegisterInfo.h"
> +#include "llvm/Support/CallSite.h"
> +#include "llvm/Support/ErrorHandling.h"
> +#include "llvm/Support/GetElementPtrTypeIterator.h"
> +#include "llvm/Target/TargetOptions.h"
> +using namespace llvm;
> +
> +namespace {
> +
> +class ARMFastISel : public FastISel {
> +
> +  /// Subtarget - Keep a pointer to the ARMSubtarget around so that we can
> +  /// make the right decision when generating code for different targets.
> +  const ARMSubtarget *Subtarget;
> +
> +  public:
> +    explicit ARMFastISel(FunctionLoweringInfo &funcInfo) : FastISel(funcInfo) {
> +      Subtarget = &TM.getSubtarget<ARMSubtarget>();
> +    }
> +
> +    virtual bool TargetSelectInstruction(const Instruction *I);
> +
> +  #include "ARMGenFastISel.inc"
> +
> +  };
> +
> +} // end anonymous namespace
> +
> +// #include "ARMGenCallingConv.inc"
> +
> +bool ARMFastISel::TargetSelectInstruction(const Instruction *I) {
> +  switch (I->getOpcode()) {
> +    default: break;
> +  }
> +  return false;
> +}
> +
> +namespace llvm {
> +  llvm::FastISel *ARM::createFastISel(FunctionLoweringInfo &funcInfo) {
> +    return new ARMFastISel(funcInfo);
> +  }
> +}
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=109047&r1=109046&r2=109047&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Wed Jul 21 17:26:11 2010
> @@ -694,6 +694,12 @@
>   return TargetLowering::getRegClassFor(VT);
> }
> 
> +// Create a fast isel object.
> +FastISel *
> +ARMTargetLowering::createFastISel(FunctionLoweringInfo &funcInfo) const {
> +  return ARM::createFastISel(funcInfo);
> +}
> +
> /// getFunctionAlignment - Return the Log2 alignment of this function.
> unsigned ARMTargetLowering::getFunctionAlignment(const Function *F) const {
>   return getTargetMachine().getSubtarget<ARMSubtarget>().isThumb() ? 1 : 2;
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.h?rev=109047&r1=109046&r2=109047&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMISelLowering.h (original)
> +++ llvm/trunk/lib/Target/ARM/ARMISelLowering.h Wed Jul 21 17:26:11 2010
> @@ -17,6 +17,7 @@
> 
> #include "ARMSubtarget.h"
> #include "llvm/Target/TargetLowering.h"
> +#include "llvm/CodeGen/FastISel.h"
> #include "llvm/CodeGen/SelectionDAG.h"
> #include "llvm/CodeGen/CallingConvLower.h"
> #include <vector>
> @@ -261,6 +262,10 @@
>     /// getFunctionAlignment - Return the Log2 alignment of this function.
>     virtual unsigned getFunctionAlignment(const Function *F) const;
> 
> +    /// createFastISel - This method returns a target specific FastISel object,
> +    /// or null if the target does not support "fast" ISel.
> +    virtual FastISel *createFastISel(FunctionLoweringInfo &funcInfo) const;
> +
>     Sched::Preference getSchedulingPreference(SDNode *N) const;
> 
>     bool isShuffleMaskLegal(const SmallVectorImpl<int> &M, EVT VT) const;
> @@ -387,6 +392,10 @@
>                                         unsigned BinOpcode) const;
> 
>   };
> +  
> +  namespace ARM {
> +    FastISel *createFastISel(FunctionLoweringInfo &funcInfo);
> +  }
> }
> 
> #endif  // ARMISELLOWERING_H
> 
> Modified: llvm/trunk/lib/Target/ARM/CMakeLists.txt
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/CMakeLists.txt?rev=109047&r1=109046&r2=109047&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/CMakeLists.txt (original)
> +++ llvm/trunk/lib/Target/ARM/CMakeLists.txt Wed Jul 21 17:26:11 2010
> @@ -11,6 +11,7 @@
> tablegen(ARMGenCallingConv.inc -gen-callingconv)
> tablegen(ARMGenSubtarget.inc -gen-subtarget)
> tablegen(ARMGenEDInfo.inc -gen-enhanced-disassembly-info)
> +tablegen(ARMFastISel.inc -gen-fast-isel)
> 
> add_llvm_target(ARMCodeGen
>   ARMAsmPrinter.cpp
> 
> Modified: llvm/trunk/lib/Target/ARM/Makefile
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Makefile?rev=109047&r1=109046&r2=109047&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/Makefile (original)
> +++ llvm/trunk/lib/Target/ARM/Makefile Wed Jul 21 17:26:11 2010
> @@ -17,7 +17,8 @@
>                 ARMGenInstrInfo.inc ARMGenAsmWriter.inc \
>                 ARMGenDAGISel.inc ARMGenSubtarget.inc \
>                 ARMGenCodeEmitter.inc ARMGenCallingConv.inc \
> -                ARMGenDecoderTables.inc ARMGenEDInfo.inc
> +                ARMGenDecoderTables.inc ARMGenEDInfo.inc \
> +                ARMGenFastISel.inc
> 
> DIRS = AsmPrinter AsmParser Disassembler TargetInfo
> 
> 
> Added: llvm/trunk/test/CodeGen/ARM/fast-isel.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/fast-isel.ll?rev=109047&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/ARM/fast-isel.ll (added)
> +++ llvm/trunk/test/CodeGen/ARM/fast-isel.ll Wed Jul 21 17:26:11 2010
> @@ -0,0 +1,15 @@
> +; RUN: llc < %s -fast-isel -fast-isel-abort -march=arm
> +
> +; Very basic fast-isel functionality.
> +
> +define i32 @add(i32 %a, i32 %b) nounwind ssp {
> +entry:
> +  %a.addr = alloca i32, align 4
> +  %b.addr = alloca i32, align 4
> +  store i32 %a, i32* %a.addr
> +  store i32 %b, i32* %b.addr
> +  %tmp = load i32* %a.addr
> +  %tmp1 = load i32* %b.addr
> +  %add = add nsw i32 %tmp, %tmp1
> +  ret i32 %add
> +}
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From gohman at apple.com  Mon Jul 26 13:28:14 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 18:28:14 -0000
Subject: [llvm-commits] [llvm] r109419 - in /llvm/trunk:
 lib/Analysis/ScalarEvolutionExpander.cpp
 test/Transforms/IndVarSimplify/crash.ll
Message-ID: <20100726182815.24A702A6C12F@llvm.org>

Author: djg
Date: Mon Jul 26 13:28:14 2010
New Revision: 109419

URL: http://llvm.org/viewvc/llvm-project?rev=109419&view=rev
Log:
Fix SCEVExpander::visitAddRecExpr so that it remembers the induction variable
it inserted rather than using LoopInfo::getCanonicalInductionVariable to
rediscover it, since that doesn't work on non-canonical loops. This fixes
infinite recurrsion on such loops; PR7562.

Modified:
    llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
    llvm/trunk/test/Transforms/IndVarSimplify/crash.ll

Modified: llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp?rev=109419&r1=109418&r2=109419&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp Mon Jul 26 13:28:14 2010
@@ -1112,21 +1112,13 @@
                                 SE.getUnknown(expand(Rest))));
   }
 
-  // {0,+,1} --> Insert a canonical induction variable into the loop!
-  if (S->isAffine() && S->getOperand(1)->isOne()) {
-    // If there's a canonical IV, just use it.
-    if (CanonicalIV) {
-      assert(Ty == SE.getEffectiveSCEVType(CanonicalIV->getType()) &&
-             "IVs with types different from the canonical IV should "
-             "already have been handled!");
-      return CanonicalIV;
-    }
-
+  // If we don't yet have a canonical IV, create one.
+  if (!CanonicalIV) {
     // Create and insert the PHI node for the induction variable in the
     // specified loop.
     BasicBlock *Header = L->getHeader();
-    PHINode *PN = PHINode::Create(Ty, "indvar", Header->begin());
-    rememberInstruction(PN);
+    CanonicalIV = PHINode::Create(Ty, "indvar", Header->begin());
+    rememberInstruction(CanonicalIV);
 
     Constant *One = ConstantInt::get(Ty, 1);
     for (pred_iterator HPI = pred_begin(Header), HPE = pred_end(Header);
@@ -1135,40 +1127,45 @@
       if (L->contains(HP)) {
         // Insert a unit add instruction right before the terminator
         // corresponding to the back-edge.
-        Instruction *Add = BinaryOperator::CreateAdd(PN, One, "indvar.next",
-                                                           HP->getTerminator());
+        Instruction *Add = BinaryOperator::CreateAdd(CanonicalIV, One,
+                                                     "indvar.next",
+                                                     HP->getTerminator());
         rememberInstruction(Add);
-        PN->addIncoming(Add, HP);
+        CanonicalIV->addIncoming(Add, HP);
       } else {
-        PN->addIncoming(Constant::getNullValue(Ty), HP);
+        CanonicalIV->addIncoming(Constant::getNullValue(Ty), HP);
       }
     }
   }
 
+  // {0,+,1} --> Insert a canonical induction variable into the loop!
+  if (S->isAffine() && S->getOperand(1)->isOne()) {
+    assert(Ty == SE.getEffectiveSCEVType(CanonicalIV->getType()) &&
+           "IVs with types different from the canonical IV should "
+           "already have been handled!");
+    return CanonicalIV;
+  }
+
   // {0,+,F} --> {0,+,1} * F
-  // Get the canonical induction variable I for this loop.
-  Value *I = CanonicalIV ?
-             CanonicalIV :
-             getOrInsertCanonicalInductionVariable(L, Ty);
 
   // If this is a simple linear addrec, emit it now as a special case.
   if (S->isAffine())    // {0,+,F} --> i*F
     return
       expand(SE.getTruncateOrNoop(
-        SE.getMulExpr(SE.getUnknown(I),
+        SE.getMulExpr(SE.getUnknown(CanonicalIV),
                       SE.getNoopOrAnyExtend(S->getOperand(1),
-                                            I->getType())),
+                                            CanonicalIV->getType())),
         Ty));
 
   // If this is a chain of recurrences, turn it into a closed form, using the
   // folders, then expandCodeFor the closed form.  This allows the folders to
   // simplify the expression without having to build a bunch of special code
   // into this folder.
-  const SCEV *IH = SE.getUnknown(I);   // Get I as a "symbolic" SCEV.
+  const SCEV *IH = SE.getUnknown(CanonicalIV);   // Get I as a "symbolic" SCEV.
 
   // Promote S up to the canonical IV type, if the cast is foldable.
   const SCEV *NewS = S;
-  const SCEV *Ext = SE.getNoopOrAnyExtend(S, I->getType());
+  const SCEV *Ext = SE.getNoopOrAnyExtend(S, CanonicalIV->getType());
   if (isa<SCEVAddRecExpr>(Ext))
     NewS = Ext;
 

Modified: llvm/trunk/test/Transforms/IndVarSimplify/crash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/crash.ll?rev=109419&r1=109418&r2=109419&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/crash.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/crash.ll Mon Jul 26 13:28:14 2010
@@ -1,4 +1,5 @@
 ; RUN: opt -indvars %s -disable-output
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
 
 declare i32 @putchar(i8) nounwind
 
@@ -17,3 +18,38 @@
 ; <label>:6                                       ; preds = %1
   ret void
 }
+
+; PR7562
+define void @fannkuch() nounwind {
+entry:                                              ; preds = %entry
+  br label %bb12
+
+bb12:                                             ; preds = %bb29, %entry
+  %i.1 = phi i32 [ undef, %entry ], [ %i.0, %bb29 ] ; <i32> [#uses=2]
+  %r.1 = phi i32 [ undef, %entry ], [ %r.0, %bb29 ] ; <i32> [#uses=2]
+  br i1 undef, label %bb13, label %bb24
+
+bb13:                                             ; preds = %bb12
+  br label %bb24
+
+bb24:                                             ; preds = %bb30, %bb13, %bb12
+  %i.2 = phi i32 [ %i.1, %bb13 ], [ %i.0, %bb30 ], [ %i.1, %bb12 ] ; <i32> [#uses=1]
+  %r.0 = phi i32 [ %r.1, %bb13 ], [ %2, %bb30 ], [ %r.1, %bb12 ] ; <i32> [#uses=3]
+  br label %bb28
+
+bb27:                                             ; preds = %bb28
+  %0 = add nsw i32 %i.0, 1                        ; <i32> [#uses=1]
+  br label %bb28
+
+bb28:                                             ; preds = %bb27, %bb26
+  %i.0 = phi i32 [ %i.2, %bb24 ], [ %0, %bb27 ]   ; <i32> [#uses=4]
+  %1 = icmp slt i32 %i.0, %r.0                    ; <i1> [#uses=1]
+  br i1 %1, label %bb27, label %bb29
+
+bb29:                                             ; preds = %bb28
+  br i1 undef, label %bb12, label %bb30
+
+bb30:                                             ; preds = %bb29
+  %2 = add nsw i32 %r.0, 1                        ; <i32> [#uses=1]
+  br label %bb24
+}




From echristo at apple.com  Mon Jul 26 13:28:36 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 11:28:36 -0700
Subject: [llvm-commits] [llvm] r109047 - in /llvm/trunk:
	lib/Target/ARM/ARMFastISel.cpp lib/Target/ARM/ARMISelLowering.cpp
	lib/Target/ARM/ARMISelLowering.h
	lib/Target/ARM/CMakeLists.txt lib/Target/ARM/Makefile
	test/CodeGen/ARM/fast-isel.ll
In-Reply-To: <795C641C-8137-416C-BEC8-5F65A16C1C8B@apple.com>
References: <20100721222611.E758A2A6C12C@llvm.org>
	<795C641C-8137-416C-BEC8-5F65A16C1C8B@apple.com>
Message-ID: <88B42A7E-617C-4203-BB03-36179988B4D2@apple.com>


On Jul 26, 2010, at 11:27 AM, Evan Cheng wrote:

> This is breaking ARM -O0. e.g.
> 
> unsigned FastEmit_ISD_Constant_MVT_i32_i(MVT RetVT, uint64_t imm0) {
>  if (RetVT.SimpleTy != MVT::i32)
>    return 0;
>  if ((!Subtarget->isThumb()) && (Subtarget->hasV6T2Ops())) {
>    return FastEmitInst_i(ARM::MOVi32imm, ARM::GPRRegisterClass, imm0);
>  }
>  if ((Subtarget->isThumb2())) {
> =>  return FastEmitInst_i(ARM::t2MOVi32imm, ARM::GPRRegisterClass, imm0);
>  }
>  return 0;
> }
> 
> But FastISel::FastEmitInst_i isn't creating the right instructions for ARM since it doesn't add the default predicate operands. I'm going to disable fastisel for ARM for now.
> 

Yep.  PR7707 :)

I was just about to turn it off myself :)

It should have just been ignoring the instructions and I'm finding out why it isn't.

-eric




From resistor at mac.com  Mon Jul 26 13:31:17 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 18:31:17 -0000
Subject: [llvm-commits] [llvm] r109420 - /llvm/trunk/README.txt
Message-ID: <20100726183117.8DFB72A6C12F@llvm.org>

Author: resistor
Date: Mon Jul 26 13:31:17 2010
New Revision: 109420

URL: http://llvm.org/viewvc/llvm-project?rev=109420&view=rev
Log:
One more test commit... 

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109420&r1=109419&r2=109420&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 13:31:17 2010
@@ -1,4 +1,4 @@
-Low Level Virtual Machine (LLVM)
+\Low Level Virtual Machine (LLVM)
 ================================
 
 This directory and its subdirectories contain source code for the Low Level
@@ -13,4 +13,3 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
-




From evan.cheng at apple.com  Mon Jul 26 13:32:55 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 26 Jul 2010 18:32:55 -0000
Subject: [llvm-commits] [llvm] r109421 -
	/llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
Message-ID: <20100726183255.7F8CA2A6C12E@llvm.org>

Author: evancheng
Date: Mon Jul 26 13:32:55 2010
New Revision: 109421

URL: http://llvm.org/viewvc/llvm-project?rev=109421&view=rev
Log:
ARM fastisel isn't ready.

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

Modified: llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMFastISel.cpp?rev=109421&r1=109420&r2=109421&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMFastISel.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMFastISel.cpp Mon Jul 26 13:32:55 2010
@@ -66,6 +66,7 @@
 
 namespace llvm {
   llvm::FastISel *ARM::createFastISel(FunctionLoweringInfo &funcInfo) {
-    return new ARMFastISel(funcInfo);
+    // Turn it off for now. It's not quite ready.
+    return 0;
   }
 }




From echristo at apple.com  Mon Jul 26 13:35:30 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 11:35:30 -0700
Subject: [llvm-commits] [llvm] r109421 -
	/llvm/trunk/lib/Target/ARM/ARMFastISel.cpp
In-Reply-To: <20100726183255.7F8CA2A6C12E@llvm.org>
References: <20100726183255.7F8CA2A6C12E@llvm.org>
Message-ID: <D9778ECA-43A0-474A-BB6A-C1629ECADF79@apple.com>


On Jul 26, 2010, at 11:32 AM, Evan Cheng wrote:

> Author: evancheng
> Date: Mon Jul 26 13:32:55 2010
> New Revision: 109421
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109421&view=rev
> Log:
> ARM fastisel isn't ready.

Thanks Evan.

-eric


From gohman at apple.com  Mon Jul 26 13:42:11 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 11:42:11 -0700
Subject: [llvm-commits] [llvm] r109413 - in /llvm/trunk:
	include/llvm/Support/StandardPasses.h
	lib/Transforms/Scalar/LoopDeletion.cpp
	lib/Transforms/Scalar/LoopUnrollPass.cpp
	lib/Transforms/Utils/LCSSA.cpp lib/Transforms/Utils/LoopSimplify.cpp
In-Reply-To: <DB402B96-1DAA-4A95-A2E7-587F0ECA43EF@apple.com>
References: <20100726181116.5345A2A6C131@llvm.org>
	<DB402B96-1DAA-4A95-A2E7-587F0ECA43EF@apple.com>
Message-ID: <D37F1A39-F8C6-4BF0-9205-1475450AA2C4@apple.com>


On Jul 26, 2010, at 11:23 AM, Devang Patel wrote:

> 
> On Jul 26, 2010, at 11:11 AM, Dan Gohman wrote:
> 
>> Author: djg
>> Date: Mon Jul 26 13:11:16 2010
>> New Revision: 109413
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=109413&view=rev
>> Log:
>> Remove LCSSA's bogus dependence on LoopSimplify and LoopSimplify's bogus
>> dependence on DominanceFrontier. Instead, add an explicit DominanceFrontier
>> pass in StandardPasses.h to ensure that it gets scheduled at the right
>> time.
> 
> If a Pass breaks because DominanceFrontier was not explicitly added in pass manager queue then the pass is broken. A pass is supposed to request all the required analysis itself.
> 
> $ opt -mypass 
> 
> should just work even if mypass needs DominanceFrontier. It is mypass's responsibility to request dominance frontier.
> 
> Enforcing `$ opt -domfrontier -mypass` requirement completely goes against getAnalysisUsage() mechanism.

The issue here is ordering, not correctness. Everything works if
DominanceFrontier is left out of the StandardCompileOpts list, however
the pass manager ends up inserting DominanceFrontier in the middle of
the loop passes, which splits them up and clobbers LoopSimplify so
that it must be re-run.

Dan




From resistor at mac.com  Mon Jul 26 13:44:56 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 18:44:56 -0000
Subject: [llvm-commits] [llvm] r109422 -
	/llvm/trunk/test/Transforms/JumpThreading/crash.ll
Message-ID: <20100726184456.49C052A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 13:44:56 2010
New Revision: 109422

URL: http://llvm.org/viewvc/llvm-project?rev=109422&view=rev
Log:
Fix a test with malformed IR.  Not sure why this didn't fail before.

Modified:
    llvm/trunk/test/Transforms/JumpThreading/crash.ll

Modified: llvm/trunk/test/Transforms/JumpThreading/crash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/JumpThreading/crash.ll?rev=109422&r1=109421&r2=109422&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/JumpThreading/crash.ll (original)
+++ llvm/trunk/test/Transforms/JumpThreading/crash.ll Mon Jul 26 13:44:56 2010
@@ -216,6 +216,9 @@
 
 ; PR5698
 define void @test7(i32 %x) {
+entry:
+  br label %tailrecurse
+
 tailrecurse:
   switch i32 %x, label %return [
     i32 2, label %bb2




From asl at math.spbu.ru  Mon Jul 26 13:45:40 2010
From: asl at math.spbu.ru (Anton Korobeynikov)
Date: Mon, 26 Jul 2010 18:45:40 -0000
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
 lib/Target/ARM/ARMGlobalMerge.cpp
 test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
Message-ID: <20100726184540.1FD682A6C12C@llvm.org>

Author: asl
Date: Mon Jul 26 13:45:39 2010
New Revision: 109423

URL: http://llvm.org/viewvc/llvm-project?rev=109423&view=rev
Log:
Currently EH lowering code expects typeinfo to be global only.
This assumption is not satisfied due to global mergeing.
Workaround the issue by temporary disablinge mergeing of const globals.
Also, ignore LLVM "special" globals. This fixes PR7716

Added:
    llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp?rev=109423&r1=109422&r2=109423&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp Mon Jul 26 13:45:39 2010
@@ -178,6 +178,11 @@
     if (I->getAlignment() != 0)
       continue;
 
+    // Ignore all 'special' globals.
+    if (I->getName().startswith("llvm.") ||
+        I->getName().startswith(".llvm."))
+      continue;
+
     if (TD->getTypeAllocSize(I->getType()) < MaxOffset) {
       if (I->isConstant())
         ConstGlobals.push_back(I);
@@ -188,8 +193,12 @@
 
   if (Globals.size() > 1)
     Changed |= doMerge(Globals, M, false);
-  if (ConstGlobals.size() > 1)
-    Changed |= doMerge(ConstGlobals, M, true);
+  // FIXME: This currently breaks the EH processing due to way how the 
+  // typeinfo detection works. We might want to detect the TIs and ignore 
+  // them in the future.
+  
+  // if (ConstGlobals.size() > 1)
+  //  Changed |= doMerge(ConstGlobals, M, true);
 
   return Changed;
 }

Added: llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll?rev=109423&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll (added)
+++ llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll Mon Jul 26 13:45:39 2010
@@ -0,0 +1,95 @@
+; RUN: llc -enable-correct-eh-support < %s
+; PR7716
+target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:32-n32"
+target triple = "thumbv7-apple-darwin10.0.0"
+
+%0 = type { i8*, i8* }
+%struct.A = type { i32 }
+
+ at d = internal global i32 0, align 4               ; <i32*> [#uses=6]
+ at _ZTVN10__cxxabiv117__class_type_infoE = external global i8* ; <i8**> [#uses=1]
+ at _ZTS1A = internal constant [3 x i8] c"1A\00"     ; <[3 x i8]*> [#uses=1]
+ at _ZTI1A = internal constant %0 { i8* bitcast (i8** getelementptr inbounds (i8** @_ZTVN10__cxxabiv117__class_type_infoE, i32 2) to i8*), i8* getelementptr inbounds ([3 x i8]* @_ZTS1A, i32 0, i32 0) } ; <%0*> [#uses=1]
+ at .str2 = private constant [18 x i8] c"c == %d, d == %d\0A\00" ; <[18 x i8]*> [#uses=1]
+ at .str3 = private constant [16 x i8] c"A(const A&) %d\0A\00" ; <[16 x i8]*> [#uses=1]
+ at .str4 = private constant [9 x i8] c"~A() %d\0A\00" ; <[9 x i8]*> [#uses=1]
+ at .str5 = private constant [8 x i8] c"A() %d\0A\00" ; <[8 x i8]*> [#uses=1]
+ at str = internal constant [14 x i8] c"Throwing 1...\00" ; <[14 x i8]*> [#uses=1]
+ at str1 = internal constant [8 x i8] c"Caught.\00"  ; <[8 x i8]*> [#uses=1]
+
+declare i32 @printf(i8* nocapture, ...) nounwind
+
+declare i8* @__cxa_allocate_exception(i32)
+
+declare i8* @llvm.eh.exception() nounwind readonly
+
+declare i32 @__gxx_personality_sj0(...)
+
+declare i32 @llvm.eh.selector(i8*, i8*, ...) nounwind
+
+declare i32 @llvm.eh.typeid.for(i8*) nounwind
+
+declare void @_Unwind_SjLj_Resume(i8*)
+
+define internal void @_ZN1AD1Ev(%struct.A* nocapture %this) nounwind ssp align 2 {
+entry:
+  %tmp.i = getelementptr inbounds %struct.A* %this, i32 0, i32 0 ; <i32*> [#uses=1]
+  %tmp2.i = load i32* %tmp.i                      ; <i32> [#uses=1]
+  %call.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([9 x i8]* @.str4, i32 0, i32 0), i32 %tmp2.i) nounwind ; <i32> [#uses=0]
+  %tmp3.i = load i32* @d                          ; <i32> [#uses=1]
+  %inc.i = add nsw i32 %tmp3.i, 1                 ; <i32> [#uses=1]
+  store i32 %inc.i, i32* @d
+  ret void
+}
+
+declare void @__cxa_throw(i8*, i8*, i8*)
+
+define i32 @main() ssp {
+entry:
+  %puts.i = tail call i32 @puts(i8* getelementptr inbounds ([14 x i8]* @str, i32 0, i32 0)) ; <i32> [#uses=0]
+  %exception.i = tail call i8* @__cxa_allocate_exception(i32 4) nounwind ; <i8*> [#uses=2]
+  %tmp2.i.i.i = bitcast i8* %exception.i to i32*  ; <i32*> [#uses=1]
+  store i32 1, i32* %tmp2.i.i.i
+  %call.i.i.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([8 x i8]* @.str5, i32 0, i32 0), i32 1) nounwind ; <i32> [#uses=0]
+  invoke void @__cxa_throw(i8* %exception.i, i8* bitcast (%0* @_ZTI1A to i8*), i8* bitcast (void (%struct.A*)* @_ZN1AD1Ev to i8*)) noreturn
+          to label %.noexc unwind label %lpad
+
+.noexc:                                           ; preds = %entry
+  unreachable
+
+try.cont:                                         ; preds = %lpad
+  %0 = tail call i8* @__cxa_get_exception_ptr(i8* %exn) nounwind ; <i8*> [#uses=0]
+  %call.i.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([16 x i8]* @.str3, i32 0, i32 0), i32 2) nounwind ; <i32> [#uses=0]
+  %1 = tail call i8* @__cxa_begin_catch(i8* %exn) nounwind ; <i8*> [#uses=0]
+  %puts = tail call i32 @puts(i8* getelementptr inbounds ([8 x i8]* @str1, i32 0, i32 0)) ; <i32> [#uses=0]
+  %call.i.i3 = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([9 x i8]* @.str4, i32 0, i32 0), i32 2) nounwind ; <i32> [#uses=0]
+  %tmp3.i.i = load i32* @d                        ; <i32> [#uses=1]
+  %inc.i.i4 = add nsw i32 %tmp3.i.i, 1            ; <i32> [#uses=1]
+  store i32 %inc.i.i4, i32* @d
+  tail call void @__cxa_end_catch()
+  %tmp13 = load i32* @d                           ; <i32> [#uses=1]
+  %call14 = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([18 x i8]* @.str2, i32 0, i32 0), i32 2, i32 %tmp13) ; <i32> [#uses=0]
+  %tmp16 = load i32* @d                           ; <i32> [#uses=1]
+  %cmp = icmp ne i32 %tmp16, 2                    ; <i1> [#uses=1]
+  %conv = zext i1 %cmp to i32                     ; <i32> [#uses=1]
+  ret i32 %conv
+
+lpad:                                             ; preds = %entry
+  %exn = tail call i8* @llvm.eh.exception() nounwind ; <i8*> [#uses=4]
+  %eh.selector = tail call i32 (i8*, i8*, ...)* @llvm.eh.selector(i8* %exn, i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*), i8* bitcast (%0* @_ZTI1A to i8*), i8* null) nounwind ; <i32> [#uses=1]
+  %2 = tail call i32 @llvm.eh.typeid.for(i8* bitcast (%0* @_ZTI1A to i8*)) nounwind ; <i32> [#uses=1]
+  %3 = icmp eq i32 %eh.selector, %2               ; <i1> [#uses=1]
+  br i1 %3, label %try.cont, label %eh.resume
+
+eh.resume:                                        ; preds = %lpad
+  tail call void @_Unwind_SjLj_Resume(i8* %exn) noreturn
+  unreachable
+}
+
+declare i8* @__cxa_get_exception_ptr(i8*)
+
+declare i8* @__cxa_begin_catch(i8*)
+
+declare void @__cxa_end_catch()
+
+declare i32 @puts(i8* nocapture) nounwind




From resistor at mac.com  Mon Jul 26 13:48:03 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 18:48:03 -0000
Subject: [llvm-commits] [llvm] r109424 - in /llvm/trunk:
 include/llvm/Analysis/LazyValueInfo.h lib/Analysis/LazyValueInfo.cpp
 lib/Transforms/Scalar/JumpThreading.cpp
Message-ID: <20100726184803.913A32A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 13:48:03 2010
New Revision: 109424

URL: http://llvm.org/viewvc/llvm-project?rev=109424&view=rev
Log:
Add an initial implementation of LazyValueInfo updating for JumpThreading.  Disabled for now.

Modified:
    llvm/trunk/include/llvm/Analysis/LazyValueInfo.h
    llvm/trunk/lib/Analysis/LazyValueInfo.cpp
    llvm/trunk/lib/Transforms/Scalar/JumpThreading.cpp

Modified: llvm/trunk/include/llvm/Analysis/LazyValueInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/LazyValueInfo.h?rev=109424&r1=109423&r2=109424&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/LazyValueInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/LazyValueInfo.h Mon Jul 26 13:48:03 2010
@@ -57,6 +57,10 @@
   /// constant on the specified edge.  Return null if not.
   Constant *getConstantOnEdge(Value *V, BasicBlock *FromBB, BasicBlock *ToBB);
   
+  /// threadEdge - Inform the analysis cache that we have threaded an edge from
+  /// PredBB to OldSucc to be from PredBB to NewSucc instead.
+  void threadEdge(BasicBlock *PredBB, BasicBlock *OldSucc, BasicBlock *NewSucc);
+  
   
   // Implementation boilerplate.
   

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109424&r1=109423&r2=109424&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Mon Jul 26 13:48:03 2010
@@ -217,6 +217,11 @@
     /// ValueCache - This is all of the cached information for all values,
     /// mapped from Value* to key information.
     DenseMap<Value*, ValueCacheEntryTy> ValueCache;
+    
+    /// OverDefinedCache - This tracks, on a per-block basis, the set of 
+    /// values that are over-defined at the end of that block.  This is required
+    /// for cache updating.
+    DenseMap<BasicBlock*, std::set<Value*> > OverDefinedCache;
   public:
     
     /// getValueInBlock - This is the query interface to determine the lattice
@@ -226,6 +231,11 @@
     /// getValueOnEdge - This is the query interface to determine the lattice
     /// value for the specified Value* that is true on the specified edge.
     LVILatticeVal getValueOnEdge(Value *V, BasicBlock *FromBB,BasicBlock *ToBB);
+    
+    /// threadEdge - This is the update interface to inform the cache that an
+    /// edge from PredBB to OldSucc has been threaded to be from PredBB to
+    /// NewSucc.
+    void threadEdge(BasicBlock *PredBB,BasicBlock *OldSucc,BasicBlock *NewSucc);
   };
 } // end anonymous namespace
 
@@ -270,12 +280,17 @@
     /// This is all of the cached information about this value.
     ValueCacheEntryTy &Cache;
     
+    /// This tracks, for each block, what values are overdefined.
+    DenseMap<BasicBlock*, std::set<Value*> > &OverDefinedCache;
+    
     ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
     /// added to cache but that are not in sorted order.
     DenseMap<BasicBlock*, LVILatticeVal> NewBlockInfo;
   public:
     
-    LVIQuery(Value *V, ValueCacheEntryTy &VC) : Val(V), Cache(VC) {
+    LVIQuery(Value *V, ValueCacheEntryTy &VC,
+             DenseMap<BasicBlock*, std::set<Value*> > &ODC)
+      : Val(V), Cache(VC), OverDefinedCache(ODC) {
     }
 
     ~LVIQuery() {
@@ -306,6 +321,11 @@
       array_pod_sort(Cache.begin(), Cache.end(),
                      BlockCacheEntryComparator::Compare);
       
+      for (DenseMap<BasicBlock*, LVILatticeVal>::iterator
+           I = NewBlockInfo.begin(), E = NewBlockInfo.end(); I != E; ++I) {
+        if (I->second.isOverdefined())
+          OverDefinedCache[I->first].insert(Val);
+      }
     }
 
     LVILatticeVal getBlockValue(BasicBlock *BB);
@@ -474,7 +494,8 @@
   DEBUG(dbgs() << "LVI Getting block end value " << *V << " at '"
         << BB->getName() << "'\n");
   
-  LVILatticeVal Result = LVIQuery(V, ValueCache[V]).getBlockValue(BB);
+  LVILatticeVal Result = LVIQuery(V, ValueCache[V], 
+                                  OverDefinedCache).getBlockValue(BB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   return Result;
@@ -488,14 +509,75 @@
   
   DEBUG(dbgs() << "LVI Getting edge value " << *V << " from '"
         << FromBB->getName() << "' to '" << ToBB->getName() << "'\n");
+  
   LVILatticeVal Result =
-    LVIQuery(V, ValueCache[V]).getEdgeValue(FromBB, ToBB);
+    LVIQuery(V, ValueCache[V],
+             OverDefinedCache).getEdgeValue(FromBB, ToBB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   
   return Result;
 }
 
+void LazyValueInfoCache::threadEdge(BasicBlock *PredBB, BasicBlock *OldSucc,
+                                    BasicBlock *NewSucc) {
+  // When an edge in the graph has been threaded, values that we could not 
+  // determine a value for before (i.e. were marked overdefined) may be possible
+  // to solve now.  We do NOT try to proactively update these values.  Instead,
+  // we clear their entries from the cache, and allow lazy updating to recompute
+  // them when needed.
+  
+  // The updating process is fairly simple: we need to dropped cached info
+  // for all values that were marked overdefined in OldSucc, and for those same
+  // values in any successor of OldSucc (except NewSucc) in which they were
+  // also marked overdefined.
+  std::vector<BasicBlock*> worklist;
+  worklist.push_back(OldSucc);
+  
+  std::set<Value*> ClearSet = OverDefinedCache[OldSucc];
+  LVILatticeVal OverDef;
+  OverDef.markOverdefined();
+  
+  // Use a worklist to perform a depth-first search of OldSucc's successors.
+  // NOTE: We do not need a visited list since any blocks we have already
+  // visited will have had their overdefined markers cleared already, and we
+  // thus won't loop to their successors.
+  while (!worklist.empty()) {
+    BasicBlock *ToUpdate = worklist.back();
+    worklist.pop_back();
+    
+    // Skip blocks only accessible through NewSucc.
+    if (ToUpdate == NewSucc) continue;
+    
+    bool changed = false;
+    std::set<Value*> &CurrentSet = OverDefinedCache[ToUpdate];
+    for (std::set<Value*>::iterator I = ClearSet.begin(),E = ClearSet.end();
+         I != E; ++I) {
+      // If a value was marked overdefined in OldSucc, and is here too...
+      if (CurrentSet.count(*I)) {
+        // Remove it from the caches.
+        ValueCacheEntryTy &Entry = ValueCache[*I];
+        ValueCacheEntryTy::iterator CI =
+          std::lower_bound(Entry.begin(), Entry.end(),
+                           std::make_pair(ToUpdate, OverDef),
+                           BlockCacheEntryComparator());
+        assert(CI != Entry.end() && "Couldn't find entry to update?");
+        Entry.erase(CI);
+
+        CurrentSet.erase(*I);
+
+        // If we removed anything, then we potentially need to update 
+        // blocks successors too.
+        changed = true;
+      }
+    }
+        
+    if (!changed) continue;
+    
+    worklist.insert(worklist.end(), succ_begin(ToUpdate), succ_end(ToUpdate));
+  }
+}
+
 //===----------------------------------------------------------------------===//
 //                            LazyValueInfo Impl
 //===----------------------------------------------------------------------===//
@@ -579,4 +661,7 @@
   return Unknown;
 }
 
-
+void LazyValueInfo::threadEdge(BasicBlock *PredBB, BasicBlock *OldSucc,
+                               BasicBlock* NewSucc) {
+  getCache(PImpl).threadEdge(PredBB, OldSucc, NewSucc);
+}

Modified: llvm/trunk/lib/Transforms/Scalar/JumpThreading.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/JumpThreading.cpp?rev=109424&r1=109423&r2=109424&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/JumpThreading.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/JumpThreading.cpp Mon Jul 26 13:48:03 2010
@@ -1314,6 +1314,9 @@
         << ", across block:\n    "
         << *BB << "\n");
   
+  if (LVI)
+    LVI->threadEdge(PredBB, BB, SuccBB);
+  
   // We are going to have to map operands from the original BB block to the new
   // copy of the block 'NewBB'.  If there are PHI nodes in BB, evaluate them to
   // account for entry from PredBB.




From ggreif at gmail.com  Mon Jul 26 13:48:07 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Mon, 26 Jul 2010 18:48:07 -0000
Subject: [llvm-commits] [llvm] r109425 - in /llvm/trunk:
	docs/ReleaseNotes.html include/llvm/Use.h
Message-ID: <20100726184807.61E492A6C12D@llvm.org>

Author: ggreif
Date: Mon Jul 26 13:48:07 2010
New Revision: 109425

URL: http://llvm.org/viewvc/llvm-project?rev=109425&view=rev
Log:
re-commit r109220, the compile error has already been fixed

Simplifying use_iterators by dereferencing
is not a good idea. The codebase does not depend
in this any more, and it may introduce hidden
runtime cost. If you get compile errors, please
dereference your iterator before passing to cast<>
(and friends).

Also: please consider caching the result of
operator* and reusing that instead of dereferencing
many times.

Modified:
    llvm/trunk/docs/ReleaseNotes.html
    llvm/trunk/include/llvm/Use.h

Modified: llvm/trunk/docs/ReleaseNotes.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/ReleaseNotes.html?rev=109425&r1=109424&r2=109425&view=diff
==============================================================================
--- llvm/trunk/docs/ReleaseNotes.html (original)
+++ llvm/trunk/docs/ReleaseNotes.html Mon Jul 26 13:48:07 2010
@@ -318,6 +318,11 @@
   To be portable across releases, resort to <tt>CallSite</tt> and the
   high-level accessors, such as <tt>getCalledValue</tt> and <tt>setUnwindDest</tt>.
 </li>
+<li>
+  You can no longer pass use_iterators directly to cast<> (and similar), because
+  these routines tend to perform costly dereference operations more than once. You
+  have to dereference the iterators yourself and pass them in.
+</li>
 
 </ul>
 

Modified: llvm/trunk/include/llvm/Use.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Use.h?rev=109425&r1=109424&r2=109425&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Use.h (original)
+++ llvm/trunk/include/llvm/Use.h Mon Jul 26 13:48:07 2010
@@ -210,30 +210,6 @@
   unsigned getOperandNo() const;
 };
 
-
-template<> struct simplify_type<value_use_iterator<User> > {
-  typedef User* SimpleType;
-  
-  static SimpleType getSimplifiedValue(const value_use_iterator<User> &Val) {
-    return *Val;
-  }
-};
-
-template<> struct simplify_type<const value_use_iterator<User> >
- : public simplify_type<value_use_iterator<User> > {};
-
-template<> struct simplify_type<value_use_iterator<const User> > {
-  typedef const User* SimpleType;
-  
-  static SimpleType getSimplifiedValue(const 
-                                       value_use_iterator<const User> &Val) {
-    return *Val;
-  }
-};
-
-template<> struct simplify_type<const value_use_iterator<const User> >
-  : public simplify_type<value_use_iterator<const User> > {};
- 
 } // End llvm namespace
 
 #endif




From dpatel at apple.com  Mon Jul 26 13:48:58 2010
From: dpatel at apple.com (Devang Patel)
Date: Mon, 26 Jul 2010 11:48:58 -0700
Subject: [llvm-commits] [llvm] r109413 - in /llvm/trunk:
	include/llvm/Support/StandardPasses.h
	lib/Transforms/Scalar/LoopDeletion.cpp
	lib/Transforms/Scalar/LoopUnrollPass.cpp
	lib/Transforms/Utils/LCSSA.cpp lib/Transforms/Utils/LoopSimplify.cpp
In-Reply-To: <D37F1A39-F8C6-4BF0-9205-1475450AA2C4@apple.com>
References: <20100726181116.5345A2A6C131@llvm.org>
	<DB402B96-1DAA-4A95-A2E7-587F0ECA43EF@apple.com>
	<D37F1A39-F8C6-4BF0-9205-1475450AA2C4@apple.com>
Message-ID: <90DAF4BE-0FCB-428F-84B2-3D614964BE92@apple.com>


On Jul 26, 2010, at 11:42 AM, Dan Gohman wrote:
> The issue here is ordering, not correctness. Everything works if
> DominanceFrontier is left out of the StandardCompileOpts list, however
> the pass manager ends up inserting DominanceFrontier in the middle of
> the loop passes, which splits them up and clobbers LoopSimplify so
> that it must be re-run.


Aha.. makes sense.
-
Devang


From matt at console-pimps.org  Mon Jul 26 15:00:18 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:18 +0100
Subject: [llvm-commits] [PATCH 0/5] ELF object support
Message-ID: <cover.1280173465.git.matt@console-pimps.org>

This series contains the patches that have been sitting in my github
tree that me and Roman Divacky have been working on for a while now. It
isn't a complete implementation but is enough to get a semi-complete
relocatable ELF file emitted.

I've tried to split the patches in a sensible way; bits required by the
ELF code but not ELF-specific come first, then the code ELF support,
then a patch to wire it up for X86.

All comments and feedback welcome.

Matt Fleming (5):
  Some ELF sections contain fixed-sized entries. Provide a way to
    record the entry size of a section.
  Record a symbol's size which is needed for ELF symbol tables.
  Introduce a helper method to add a section to the end of a layout.
    This will be used by the ELF ObjectWriter code to add the metadata
    sections (symbol table, etc) to the end of an object file.
  Add ELF ObjectWriter and Streamer support.
  Hookup the ELF support for X86.

 include/llvm/MC/ELFObjectWriter.h   |   46 ++
 include/llvm/MC/MCAssembler.h       |   32 ++
 include/llvm/MC/MCStreamer.h        |    6 +
 include/llvm/Support/ELF.h          |   14 +
 lib/MC/CMakeLists.txt               |    2 +
 lib/MC/ELFObjectWriter.cpp          |  873 +++++++++++++++++++++++++++++++++++
 lib/MC/MCAssembler.cpp              |   38 ++-
 lib/MC/MCELFStreamer.cpp            |  496 ++++++++++++++++++++
 lib/Target/X86/X86AsmBackend.cpp    |   17 +-
 lib/Target/X86/X86TargetMachine.cpp |    2 +
 10 files changed, 1521 insertions(+), 5 deletions(-)
 create mode 100644 include/llvm/MC/ELFObjectWriter.h
 create mode 100644 lib/MC/ELFObjectWriter.cpp
 create mode 100644 lib/MC/MCELFStreamer.cpp



From matt at console-pimps.org  Mon Jul 26 15:00:21 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:21 +0100
Subject: [llvm-commits] [PATCH 3/5] Introduce a helper method to add a
	section to the end of a layout. This will be used by the ELF
	ObjectWriter code to add the metadata sections (symbol table,
	etc) to the end of an object file.
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <da0bcb6698213acb180ad20d1efcef6229588a0c.1280173465.git.matt@console-pimps.org>

---
 include/llvm/MC/MCAssembler.h |    2 ++
 lib/MC/MCAssembler.cpp        |   34 ++++++++++++++++++++++++++++++++++
 2 files changed, 36 insertions(+), 0 deletions(-)

diff --git a/include/llvm/MC/MCAssembler.h b/include/llvm/MC/MCAssembler.h
index c189a1d..89d83ba 100644
--- a/include/llvm/MC/MCAssembler.h
+++ b/include/llvm/MC/MCAssembler.h
@@ -679,6 +679,8 @@ public:
   void WriteSectionData(const MCSectionData *Section, const MCAsmLayout &Layout,
                         MCObjectWriter *OW) const;
 
+  void AddSectionToTheEnd(MCSectionData &SD, MCAsmLayout &Layout);
+
 public:
   /// Construct a new assembler instance.
   ///
diff --git a/lib/MC/MCAssembler.cpp b/lib/MC/MCAssembler.cpp
index fe55e51..0444dc2 100644
--- a/lib/MC/MCAssembler.cpp
+++ b/lib/MC/MCAssembler.cpp
@@ -649,6 +649,40 @@ void MCAssembler::WriteSectionData(const MCSectionData *SD,
   assert(OW->getStream().tell() - Start == Layout.getSectionFileSize(SD));
 }
 
+void MCAssembler::AddSectionToTheEnd(MCSectionData &SD, MCAsmLayout &Layout) {
+  // Create dummy fragments and assign section ordinals.
+  unsigned SectionIndex = 0;
+  for (MCAssembler::iterator it = begin(), ie = end(); it != ie; ++it)
+    SectionIndex++;
+
+  SD.setOrdinal(SectionIndex);
+
+  // Assign layout order indices to sections and fragments.
+  unsigned FragmentIndex = 0;
+  unsigned i = 0;
+  for (unsigned e = Layout.getSectionOrder().size(); i != e; ++i) {
+    MCSectionData *SD = Layout.getSectionOrder()[i];
+
+    for (MCSectionData::iterator it2 = SD->begin(),
+           ie2 = SD->end(); it2 != ie2; ++it2)
+      FragmentIndex++;
+  }
+
+  SD.setLayoutOrder(i);
+  for (MCSectionData::iterator it2 = SD.begin(),
+         ie2 = SD.end(); it2 != ie2; ++it2) {
+    it2->setLayoutOrder(FragmentIndex++);
+  }
+  Layout.getSectionOrder().push_back(&SD);
+
+  Layout.LayoutSection(&SD);
+
+  // Layout until everything fits.
+  while (LayoutOnce(Layout))
+    continue;
+
+}
+
 void MCAssembler::Finish(MCObjectWriter *Writer) {
   DEBUG_WITH_TYPE("mc-dump", {
       llvm::errs() << "assembler backend - pre-layout\n--\n";
-- 
1.6.4.rc0



From matt at console-pimps.org  Mon Jul 26 15:00:19 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:19 +0100
Subject: [llvm-commits] [PATCH 1/5] Some ELF sections contain fixed-sized
	entries. Provide a way to record the entry size of a section.
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <c48da6bb8b6f1b0ec18e643ccc0769677f4884da.1280173465.git.matt@console-pimps.org>

---
 include/llvm/MC/MCAssembler.h |    8 ++++++++
 lib/MC/MCAssembler.cpp        |    1 +
 2 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/include/llvm/MC/MCAssembler.h b/include/llvm/MC/MCAssembler.h
index 97ab1a1..51f265d 100644
--- a/include/llvm/MC/MCAssembler.h
+++ b/include/llvm/MC/MCAssembler.h
@@ -366,6 +366,11 @@ private:
   /// Alignment - The maximum alignment seen in this section.
   unsigned Alignment;
 
+  /// EntrySize - The size of each entry in this section. This is ~0
+  /// until initialized. This size only makes sense for sections that
+  /// contain fixed-sized entries.
+  unsigned EntrySize;
+
   /// @name Assembler Backend Data
   /// @{
   //
@@ -391,6 +396,9 @@ public:
   unsigned getAlignment() const { return Alignment; }
   void setAlignment(unsigned Value) { Alignment = Value; }
 
+  unsigned getEntrySize() const { return EntrySize; }
+  void setEntrySize(unsigned Value) { EntrySize = Value; }
+
   bool hasInstructions() const { return HasInstructions; }
   void setHasInstructions(bool Value) { HasInstructions = Value; }
 
diff --git a/lib/MC/MCAssembler.cpp b/lib/MC/MCAssembler.cpp
index c1365c1..a284c2e 100644
--- a/lib/MC/MCAssembler.cpp
+++ b/lib/MC/MCAssembler.cpp
@@ -192,6 +192,7 @@ MCSectionData::MCSectionData() : Section(0) {}
 MCSectionData::MCSectionData(const MCSection &_Section, MCAssembler *A)
   : Section(&_Section),
     Alignment(1),
+    EntrySize(~UINT64_C(0)),
     Address(~UINT64_C(0)),
     HasInstructions(false)
 {
-- 
1.6.4.rc0



From matt at console-pimps.org  Mon Jul 26 15:00:20 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:20 +0100
Subject: [llvm-commits] [PATCH 2/5] Record a symbol's size which is needed
	for ELF symbol tables.
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <75e6e9d13ae15c9baf796f3d6fbbe504a2197172.1280173465.git.matt@console-pimps.org>

---
 include/llvm/MC/MCAssembler.h |   22 ++++++++++++++++++++++
 lib/MC/MCAssembler.cpp        |    3 ++-
 2 files changed, 24 insertions(+), 1 deletions(-)

diff --git a/include/llvm/MC/MCAssembler.h b/include/llvm/MC/MCAssembler.h
index 51f265d..c189a1d 100644
--- a/include/llvm/MC/MCAssembler.h
+++ b/include/llvm/MC/MCAssembler.h
@@ -24,6 +24,7 @@ namespace llvm {
 class raw_ostream;
 class MCAsmLayout;
 class MCAssembler;
+class MCBinaryExpr;
 class MCContext;
 class MCCodeEmitter;
 class MCExpr;
@@ -460,6 +461,10 @@ public:
   // common symbol can never get a definition.
   uint64_t CommonSize;
 
+  uint64_t SymbolSize;
+
+  const MCBinaryExpr *SizeSymbol;
+
   /// CommonAlign - The alignment of the symbol, if it is 'common'.
   //
   // FIXME: Pack this in with other fields?
@@ -517,6 +522,23 @@ public:
     return CommonSize;
   }
 
+  void setSize(uint64_t _SymbolSize) {
+    SymbolSize = _SymbolSize;
+  }
+
+  uint64_t getSize() {
+    return SymbolSize;
+  }
+
+  void setSizeSymbol(const MCBinaryExpr *SS) {
+    SizeSymbol = SS;
+  }
+
+  const MCBinaryExpr *getSizeSymbol() {
+    return SizeSymbol;
+  }
+
+
   /// getCommonAlignment - Return the alignment of a 'common' symbol.
   unsigned getCommonAlignment() const {
     assert(isCommon() && "Not a 'common' symbol!");
diff --git a/lib/MC/MCAssembler.cpp b/lib/MC/MCAssembler.cpp
index a284c2e..fe55e51 100644
--- a/lib/MC/MCAssembler.cpp
+++ b/lib/MC/MCAssembler.cpp
@@ -208,7 +208,8 @@ MCSymbolData::MCSymbolData(const MCSymbol &_Symbol, MCFragment *_Fragment,
                            uint64_t _Offset, MCAssembler *A)
   : Symbol(&_Symbol), Fragment(_Fragment), Offset(_Offset),
     IsExternal(false), IsPrivateExtern(false),
-    CommonSize(0), CommonAlign(0), Flags(0), Index(0)
+    CommonSize(0), SymbolSize(0), SizeSymbol(0), CommonAlign(0),
+    Flags(0), Index(0)
 {
   if (A)
     A->getSymbolList().push_back(this);
-- 
1.6.4.rc0



From matt at console-pimps.org  Mon Jul 26 15:00:23 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:23 +0100
Subject: [llvm-commits] [PATCH 5/5] Hookup the ELF support for X86.
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>

---
 lib/Target/X86/X86AsmBackend.cpp    |   17 +++++++++++++----
 lib/Target/X86/X86TargetMachine.cpp |    2 ++
 2 files changed, 15 insertions(+), 4 deletions(-)

diff --git a/lib/Target/X86/X86AsmBackend.cpp b/lib/Target/X86/X86AsmBackend.cpp
index 2cf65c1..b6a14b2 100644
--- a/lib/Target/X86/X86AsmBackend.cpp
+++ b/lib/Target/X86/X86AsmBackend.cpp
@@ -11,6 +11,7 @@
 #include "X86.h"
 #include "X86FixupKinds.h"
 #include "llvm/ADT/Twine.h"
+#include "llvm/MC/ELFObjectWriter.h"
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCObjectWriter.h"
@@ -190,10 +191,6 @@ public:
     HasScatteredSymbols = true;
   }
 
-  MCObjectWriter *createObjectWriter(raw_ostream &OS) const {
-    return 0;
-  }
-
   bool isVirtualSection(const MCSection &Section) const {
     const MCSectionELF &SE = static_cast<const MCSectionELF&>(Section);
     return SE.getType() == MCSectionELF::SHT_NOBITS;;
@@ -204,12 +201,24 @@ class ELFX86_32AsmBackend : public ELFX86AsmBackend {
 public:
   ELFX86_32AsmBackend(const Target &T)
     : ELFX86AsmBackend(T) {}
+
+  MCObjectWriter *createObjectWriter(raw_ostream &OS) const {
+    return new ELFObjectWriter(OS, /*Is64Bit=*/false,
+                               /*IsLittleEndian=*/true,
+                               /*HasRelocationAddend=*/false);
+  }
 };
 
 class ELFX86_64AsmBackend : public ELFX86AsmBackend {
 public:
   ELFX86_64AsmBackend(const Target &T)
     : ELFX86AsmBackend(T) {}
+
+  MCObjectWriter *createObjectWriter(raw_ostream &OS) const {
+    return new ELFObjectWriter(OS, /*Is64Bit=*/true,
+                               /*IsLittleEndian=*/true,
+                               /*HasRelocationAddend=*/true);
+  }
 };
 
 class DarwinX86AsmBackend : public X86AsmBackend {
diff --git a/lib/Target/X86/X86TargetMachine.cpp b/lib/Target/X86/X86TargetMachine.cpp
index 900b516..6e16083 100644
--- a/lib/Target/X86/X86TargetMachine.cpp
+++ b/lib/Target/X86/X86TargetMachine.cpp
@@ -46,6 +46,8 @@ static MCStreamer *createMCStreamer(const Target &T, const std::string &TT,
                                     bool RelaxAll) {
   Triple TheTriple(TT);
   switch (TheTriple.getOS()) {
+  case Triple::Linux:
+    return createELFStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
   default:
     return createMachOStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
   }
-- 
1.6.4.rc0



From matt at console-pimps.org  Mon Jul 26 15:00:22 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:00:22 +0100
Subject: [llvm-commits] [PATCH 4/5] Add ELF ObjectWriter and Streamer
	support.
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <cac33388f858fe37905e5eb017a333827b3ae1ba.1280173465.git.matt@console-pimps.org>

---
 include/llvm/MC/ELFObjectWriter.h |   46 ++
 include/llvm/MC/MCStreamer.h      |    6 +
 include/llvm/Support/ELF.h        |   14 +
 lib/MC/CMakeLists.txt             |    2 +
 lib/MC/ELFObjectWriter.cpp        |  873 +++++++++++++++++++++++++++++++++++++
 lib/MC/MCELFStreamer.cpp          |  496 +++++++++++++++++++++
 6 files changed, 1437 insertions(+), 0 deletions(-)
 create mode 100644 include/llvm/MC/ELFObjectWriter.h
 create mode 100644 lib/MC/ELFObjectWriter.cpp
 create mode 100644 lib/MC/MCELFStreamer.cpp

diff --git a/include/llvm/MC/ELFObjectWriter.h b/include/llvm/MC/ELFObjectWriter.h
new file mode 100644
index 0000000..3b9951f
--- /dev/null
+++ b/include/llvm/MC/ELFObjectWriter.h
@@ -0,0 +1,46 @@
+//===-- llvm/MC/ELFObjectWriter.h - ELF File Writer ---------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_MC_ELFOBJECTWRITER_H
+#define LLVM_MC_ELFOBJECTWRITER_H
+
+#include "llvm/MC/MCObjectWriter.h"
+#include "llvm/Support/raw_ostream.h"
+#include <cassert>
+
+namespace llvm {
+class MCAsmFixup;
+class MCAssembler;
+class MCFragment;
+class MCValue;
+class raw_ostream;
+
+class ELFObjectWriter : public MCObjectWriter {
+  void *Impl;
+
+public:
+  ELFObjectWriter(raw_ostream &OS, bool Is64Bit, bool IsLittleEndian = true,
+                  bool HasRelocationAddend = true);
+
+  virtual ~ELFObjectWriter();
+
+  virtual void ExecutePostLayoutBinding(MCAssembler &Asm);
+
+  virtual void RecordRelocation(const MCAssembler &Asm,
+                                const MCAsmLayout &Layout,
+                                const MCFragment *Fragment,
+                                const MCFixup &Fixup, MCValue Target,
+                                uint64_t &FixedValue);
+
+  virtual void WriteObject(const MCAssembler &Asm, const MCAsmLayout &Layout);
+};
+
+} // End llvm namespace
+
+#endif
diff --git a/include/llvm/MC/MCStreamer.h b/include/llvm/MC/MCStreamer.h
index aca7dd3..a12e92d 100644
--- a/include/llvm/MC/MCStreamer.h
+++ b/include/llvm/MC/MCStreamer.h
@@ -353,6 +353,12 @@ namespace llvm {
                                     TargetAsmBackend &TAB,
                                     MCCodeEmitter &CE, raw_ostream &OS);
 
+  /// createELFStreamer - Create a machine code streamer which will generate
+  /// ELF format object files.
+  MCStreamer *createELFStreamer(MCContext &Ctx, TargetAsmBackend &TAB,
+				raw_ostream &OS, MCCodeEmitter *CE,
+				bool RelaxAll = false);
+
   /// createLoggingStreamer - Create a machine code streamer which just logs the
   /// API calls and then dispatches to another streamer.
   ///
diff --git a/include/llvm/Support/ELF.h b/include/llvm/Support/ELF.h
index a4223cf..1888d59 100644
--- a/include/llvm/Support/ELF.h
+++ b/include/llvm/Support/ELF.h
@@ -330,6 +330,12 @@ struct Elf64_Sym {
   }
 };
 
+// The size (in bytes) of symbol table entries.
+enum {
+  SYMENTRY_SIZE32 = 16, // 32-bit symbol entry size
+  SYMENTRY_SIZE64 = 24  // 64-bit symbol entry size.
+};
+
 // Symbol bindings.
 enum {
   STB_LOCAL = 0,   // Local symbol, not visible outside obj file containing def
@@ -359,6 +365,14 @@ enum {
   STV_PROTECTED = 3   // Visible in other components but not preemptable
 };
 
+// Because all the symbol flags need to be stored in the MCSymbolData
+// 'flags' variable we need to provide shift constants per flag type.
+enum {
+  STT_SHIFT = 0, // Shift value for STT_* flags.
+  STB_SHIFT = 4, // Shift value for STB_* flags.
+  STV_SHIFT = 8  // Shift value ofr STV_* flags.
+};
+
 // Relocation entry, without explicit addend.
 struct Elf32_Rel {
   Elf32_Addr r_offset; // Location (file byte offset, or program virtual addr)
diff --git a/lib/MC/CMakeLists.txt b/lib/MC/CMakeLists.txt
index fc4f3c6..91adec9 100644
--- a/lib/MC/CMakeLists.txt
+++ b/lib/MC/CMakeLists.txt
@@ -1,4 +1,5 @@
 add_llvm_library(LLVMMC
+  ELFObjectWriter.cpp
   MCAsmInfo.cpp
   MCAsmInfoCOFF.cpp
   MCAsmInfoDarwin.cpp
@@ -7,6 +8,7 @@ add_llvm_library(LLVMMC
   MCCodeEmitter.cpp
   MCContext.cpp
   MCDisassembler.cpp
+  MCELFStreamer.cpp
   MCExpr.cpp
   MCInst.cpp
   MCInstPrinter.cpp
diff --git a/lib/MC/ELFObjectWriter.cpp b/lib/MC/ELFObjectWriter.cpp
new file mode 100644
index 0000000..a2588cb
--- /dev/null
+++ b/lib/MC/ELFObjectWriter.cpp
@@ -0,0 +1,873 @@
+//===- lib/MC/ELFObjectWriter.cpp - ELF File Writer -------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements ELF object file writer information.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/MC/ELFObjectWriter.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/StringMap.h"
+#include "llvm/ADT/Twine.h"
+#include "llvm/MC/MCAssembler.h"
+#include "llvm/MC/MCAsmLayout.h"
+#include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCObjectWriter.h"
+#include "llvm/MC/MCSectionELF.h"
+#include "llvm/MC/MCSymbol.h"
+#include "llvm/MC/MCValue.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/ELF.h"
+#include "llvm/Target/TargetAsmBackend.h"
+
+#include "../Target/X86/X86FixupKinds.h"
+
+#include <vector>
+using namespace llvm;
+
+namespace {
+
+class ELFObjectWriterImpl {
+  static bool isFixupKindPCRel(unsigned Kind) {
+    switch (Kind) {
+    default:
+      return false;
+    case X86::reloc_pcrel_1byte:
+    case X86::reloc_pcrel_4byte:
+    case X86::reloc_riprel_4byte:
+    case X86::reloc_riprel_4byte_movq_load:
+      return true;
+    }
+  }
+
+  static bool isFixupKindRIPRel(unsigned Kind) {
+    return Kind == X86::reloc_riprel_4byte ||
+      Kind == X86::reloc_riprel_4byte_movq_load;
+  }
+
+
+  /// ELFSymbolData - Helper struct for containing some precomputed information
+  /// on symbols.
+  struct ELFSymbolData {
+    MCSymbolData *SymbolData;
+    uint64_t StringIndex;
+    uint32_t SectionIndex;
+
+    // Support lexicographic sorting.
+    bool operator<(const ELFSymbolData &RHS) const {
+      const std::string &Name = SymbolData->getSymbol().getName();
+      return Name < RHS.SymbolData->getSymbol().getName();
+    }
+  };
+
+  /// @name Relocation Data
+  /// @{
+
+  struct ELFRelocationEntry {
+    // Make these big enough for both 32-bit and 64-bit
+    uint64_t r_offset;
+    uint64_t r_info;
+    uint64_t r_addend;
+
+    // Support lexicographic sorting.
+    bool operator<(const ELFRelocationEntry &RE) const {
+      return RE.r_offset < r_offset;
+    }
+  };
+
+  llvm::DenseMap<const MCSectionData*,
+                 std::vector<ELFRelocationEntry> > Relocations;
+  DenseMap<const MCSection*, uint64_t> SectionStringTableIndex;
+
+  /// @}
+  /// @name Symbol Table Data
+  /// @{
+
+  SmallString<256> StringTable;
+  std::vector<ELFSymbolData> LocalSymbolData;
+  std::vector<ELFSymbolData> ExternalSymbolData;
+  std::vector<ELFSymbolData> UndefinedSymbolData;
+
+  /// @}
+
+  ELFObjectWriter *Writer;
+
+  raw_ostream &OS;
+
+  // This holds the current offset into the object file.
+  size_t FileOff;
+
+  unsigned Is64Bit : 1;
+
+  bool HasRelocationAddend;
+
+  // This holds the symbol table index of the last local symbol.
+  unsigned LastLocalSymbolIndex;
+  // This holds the .strtab section index.
+  unsigned StringTableIndex;
+
+  unsigned ShstrtabIndex;
+
+public:
+  ELFObjectWriterImpl(ELFObjectWriter *_Writer, bool _Is64Bit,
+                      bool _HasRelAddend)
+    : Writer(_Writer), OS(Writer->getStream()), FileOff(0),
+      Is64Bit(_Is64Bit), HasRelocationAddend(_HasRelAddend) {
+  }
+
+  void Write8(uint8_t Value) { Writer->Write8(Value); }
+  void Write16(uint16_t Value) { Writer->Write16(Value); }
+  void Write32(uint32_t Value) { Writer->Write32(Value); }
+  void Write64(uint64_t Value) { Writer->Write64(Value); }
+  void WriteZeros(unsigned N) { Writer->WriteZeros(N); }
+  void WriteBytes(StringRef Str, unsigned ZeroFillSize = 0) {
+    Writer->WriteBytes(Str, ZeroFillSize);
+  }
+
+  void WriteWord(uint64_t W) {
+    if (Is64Bit) {
+      Writer->Write64(W);
+    } else {
+      Writer->Write32(W);
+    }
+  }
+
+  // Emit the ELF header.
+  void WriteHeader(uint64_t SectionDataSize, unsigned NumberOfSections) {
+
+    // ELF Header
+    // ----------
+    //
+    // Note
+    // ----
+    // emitWord method behaves differently for ELF32 and ELF64, writing
+    // 4 bytes in the former and 8 in the latter.
+
+    Write8(0x7f); // e_ident[EI_MAG0]
+    Write8('E');  // e_ident[EI_MAG1]
+    Write8('L');  // e_ident[EI_MAG2]
+    Write8('F');  // e_ident[EI_MAG3]
+
+    Write8(Is64Bit ? ELF::ELFCLASS64 : ELF::ELFCLASS32); // e_ident[EI_CLASS]
+
+    // e_ident[EI_DATA]
+    Write8(Writer->isLittleEndian() ? ELF::ELFDATA2LSB : ELF::ELFDATA2MSB);
+
+    Write8(ELF::EV_CURRENT);        // e_ident[EI_VERSION]
+    Write8(ELF::ELFOSABI_LINUX);    // e_ident[EI_OSABI]
+    Write8(0);			// e_ident[EI_ABIVERSION]
+
+    WriteZeros(ELF::EI_NIDENT - ELF::EI_PAD);
+
+    Write16(ELF::ET_REL);             // e_type
+
+    // FIXME
+    Write16(ELF::EM_X86_64); // e_machine = target
+
+    Write32(ELF::EV_CURRENT);         // e_version
+    WriteWord(0);                    // e_entry, no entry point in .o file
+    WriteWord(0);                    // e_phoff, no program header for .o
+    WriteWord(SectionDataSize + 64);  // e_shoff = sec hdr table off in bytes
+
+    // FIXME
+    Write32(0);   // e_flags = whatever the target wants
+
+    Write16(Is64Bit ? 64 : 52);  // e_ehsize = ELF header size
+    Write16(0);                  // e_phentsize = prog header entry size
+    Write16(0);                  // e_phnum = # prog header entries = 0
+
+    // e_shentsize = Section header entry size
+    Write16(Is64Bit ? 64 : 40);
+
+    // e_shnum     = # of section header ents
+    Write16(NumberOfSections);
+
+    // e_shstrndx  = Section # of '.shstrtab'
+    Write16(ShstrtabIndex);
+  }
+
+  void WriteSymbolEntry(MCDataFragment *F, uint64_t name, uint8_t info,
+                        uint64_t value, uint64_t size,
+                        uint8_t other, uint16_t shndx) {
+    if (Is64Bit) {
+      F->getContents() += StringRef((const char *)&name, 4);  // st_name
+      F->getContents() += StringRef((const char *)&info, 1);  // st_info
+      F->getContents() += StringRef((const char *)&other, 1); // st_other
+      F->getContents() += StringRef((const char *)&shndx, 2); // st_shndx
+      F->getContents() += StringRef((const char *)&value, 8); // st_value
+      // FIXME
+      F->getContents() += StringRef((const char *)&size, 8);  // st_size
+    } else {
+      F->getContents() += StringRef((const char *)&name, 4);  // st_name
+      F->getContents() += StringRef((const char *)&value, 4); // st_value
+      // FIXME
+      F->getContents() += StringRef((const char *)&size, 4);  // st_size
+      F->getContents() += StringRef((const char *)&info, 1);  // st_info
+      F->getContents() += StringRef((const char *)&other, 1); // st_other
+      F->getContents() += StringRef((const char *)&shndx, 2); // st_shndx
+    }
+  }
+
+  void WriteSymbol(MCDataFragment *F, ELFSymbolData &MSD,
+                   const MCAsmLayout &Layout) {
+    MCSymbolData &Data = *MSD.SymbolData;
+    const MCSymbol &Symbol = Data.getSymbol();
+    uint8_t Info = (Data.getFlags() & 0xff);
+    uint8_t Other = ((Data.getFlags() & 0xf00) >> ELF::STV_SHIFT);
+    uint64_t Value = 0;
+    uint64_t Size = 0;
+
+    Size = Data.getSize();
+    if (Data.isCommon())
+      Value = Data.getCommonAlignment();
+    if (Data.getSizeSymbol()) {
+      Value = Layout.getSymbolAddress(&Data);
+      MCValue Res;
+      if (Data.getSizeSymbol()->EvaluateAsRelocatable(Res, &Layout)) {
+        MCSymbolData &A =
+            Layout.getAssembler().getSymbolData(Res.getSymA()->getSymbol());
+        MCSymbolData &B =
+            Layout.getAssembler().getSymbolData(Res.getSymB()->getSymbol());
+
+	Size = Layout.getSymbolAddress(&A) - Layout.getSymbolAddress(&B);
+      }
+
+    }
+
+    // Write out the symbol table entry
+    WriteSymbolEntry(F, MSD.StringIndex, Info, Value,
+                     Size, Other, MSD.SectionIndex);
+  }
+
+  void WriteSymbolTable(MCDataFragment *F, const MCAssembler &Asm,
+                        const MCAsmLayout &Layout) {
+
+    // The string table must be emitted first because we need the index
+    // into the string table for all the symbol names.
+    assert(StringTable.size() && "Missing string table");
+
+    // FIXME: Make sure the start of the symbol table is aligned.
+
+    // The first entry is the undefined symbol entry.
+    unsigned EntrySize = Is64Bit ? ELF::SYMENTRY_SIZE64 : ELF::SYMENTRY_SIZE32;
+    for (unsigned i = 0; i < EntrySize; ++i)
+      F->getContents() += '\x00';
+
+    // Write the symbol table entries.
+    LastLocalSymbolIndex = LocalSymbolData.size() + 1;
+    for (unsigned i = 0, e = LocalSymbolData.size(); i != e; ++i) {
+      ELFSymbolData &MSD = LocalSymbolData[i];
+      WriteSymbol(F, MSD, Layout);
+    }
+
+    // Write out a symbol table entry for each section.
+    for (unsigned Index = 1; Index < Asm.size(); ++Index)
+      WriteSymbolEntry(F, 0, ELF::STT_SECTION, 0, 0, ELF::STV_DEFAULT, Index);
+    LastLocalSymbolIndex += Asm.size() - 1;
+
+    for (unsigned i = 0, e = ExternalSymbolData.size(); i != e; ++i) {
+      ELFSymbolData &MSD = ExternalSymbolData[i];
+      MCSymbolData &Data = *MSD.SymbolData;
+      assert((Data.getFlags() & (ELF::STB_GLOBAL << ELF::STB_SHIFT)) &&
+	     "External symbol requires STB_GLOBAL flag");
+      WriteSymbol(F, MSD, Layout);
+      if (Data.getFlags() & (ELF::STB_LOCAL << ELF::STB_SHIFT))
+        LastLocalSymbolIndex++;
+    }
+
+    for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i) {
+      ELFSymbolData &MSD = UndefinedSymbolData[i];
+      MCSymbolData &Data = *MSD.SymbolData;
+      Data.setFlags(Data.getFlags() | (ELF::STB_GLOBAL << ELF::STB_SHIFT));
+      WriteSymbol(F, MSD, Layout);
+      if (Data.getFlags() & (ELF::STB_LOCAL << ELF::STB_SHIFT))
+        LastLocalSymbolIndex++;
+    }
+  }
+
+  // XXX: this is currently X86_64 only
+  void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
+                        const MCFragment *Fragment, const MCFixup &Fixup,
+                        MCValue Target, uint64_t &FixedValue) {
+     unsigned IsPCRel = isFixupKindPCRel(Fixup.getKind());
+     ELFRelocationEntry ERE;
+     struct ELF::Elf64_Rela ERE64;
+
+     uint64_t FixupOffset =
+	      Layout.getFragmentOffset(Fragment) + Fixup.getOffset();
+     int64_t Value;
+     int64_t Addend = 0;
+     unsigned Index = 0;
+     unsigned Type;
+
+     Value = Target.getConstant();
+
+     if (Target.isAbsolute()) {
+       Type = ELF::R_X86_64_NONE;
+       Index = 0;
+     } else {
+       const MCSymbol *Symbol = &Target.getSymA()->getSymbol();
+       MCSymbolData &SD = Asm.getSymbolData(*Symbol);
+       const MCSymbolData *Base = Asm.getAtom(Layout, &SD);
+
+       if (Base) {
+         Index = getSymbolIndexInSymbolTable(const_cast<MCAssembler &>(Asm), &Base->getSymbol());
+         if (Base != &SD)
+           Value += Layout.getSymbolAddress(&SD) - Layout.getSymbolAddress(Base);
+         Addend = Value;
+         Value = 0;
+       } else {
+         MCFragment *F = SD.getFragment();
+         if (F) {
+           // Index of the section in .symtab against this symbol
+           // is being relocated + 2 (empty section + abs. symbols).
+           Index = SD.getFragment()->getParent()->getOrdinal() +
+             getNumOfLocalSymbols(Asm) + 1;
+
+           MCSectionData *FSD = F->getParent();
+           // Offset of the symbol in the section
+           Addend = Layout.getSymbolAddress(&SD) - Layout.getSectionAddress(FSD);
+         } else {
+           FixedValue = Value;
+           return;
+         }
+       }
+     }
+
+     // determine the type of the relocation
+     if (IsPCRel) {
+       Type = ELF::R_X86_64_PC32;
+     } else {
+       switch (Fixup.getKind()) {
+         case FK_Data_8: Type = ELF::R_X86_64_64; break;
+         case X86::reloc_pcrel_4byte:
+         case FK_Data_4:
+           long Offset;
+           Offset = Target.getConstant();
+           // check that the offset fits within a signed long
+           if (!(((long) -1 << 31) & Offset) || (((long) -1 << 31) & Offset) == ((long) -1 << 31))
+             Type = ELF::R_X86_64_32S;
+           else
+             Type = ELF::R_X86_64_32;
+           break;
+         case FK_Data_2: Type = ELF::R_X86_64_16; break;
+         case X86::reloc_pcrel_1byte:
+         case FK_Data_1:
+           Type = ELF::R_X86_64_8;
+           break;
+       }
+     }
+
+     FixedValue = Value;
+
+     ERE64.setSymbolAndType(Index, Type);
+
+     ERE.r_offset = FixupOffset;
+     ERE.r_info = ERE64.r_info;
+     if (HasRelocationAddend)
+       ERE.r_addend = Addend;
+
+     Relocations[Fragment->getParent()].push_back(ERE);
+  }
+
+  // XXX-PERF: this should be cached
+  uint64_t getNumOfLocalSymbols(const MCAssembler &Asm) {
+    std::vector<const MCSymbol*> Local;
+
+    uint64_t Index = 0;
+    for (MCAssembler::const_symbol_iterator it = Asm.symbol_begin(),
+           ie = Asm.symbol_end(); it != ie; ++it) {
+      const MCSymbol &Symbol = it->getSymbol();
+
+      // Ignore non-linker visible symbols.
+      if (!Asm.isSymbolLinkerVisible(Symbol))
+        continue;
+
+      if (it->isExternal() || Symbol.isUndefined())
+        continue;
+
+      Index++;
+    }
+
+    return Index;
+  }
+
+  // XXX-PERF: this should be cached
+  uint64_t getSymbolIndexInSymbolTable(MCAssembler &Asm, const MCSymbol *S) {
+    std::vector<ELFSymbolData> Local;
+    std::vector<ELFSymbolData> External;
+    std::vector<ELFSymbolData> Undefined;
+
+    for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
+           ie = Asm.symbol_end(); it != ie; ++it) {
+      const MCSymbol &Symbol = it->getSymbol();
+
+      // Ignore non-linker visible symbols.
+      if (!Asm.isSymbolLinkerVisible(Symbol))
+        continue;
+
+      if (it->isExternal() || Symbol.isUndefined())
+        continue;
+
+      ELFSymbolData MSD;
+      MSD.SymbolData = it;
+
+      Local.push_back(MSD);
+    }
+    for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
+           ie = Asm.symbol_end(); it != ie; ++it) {
+      const MCSymbol &Symbol = it->getSymbol();
+
+      // Ignore non-linker visible symbols.
+      if (!Asm.isSymbolLinkerVisible(Symbol))
+        continue;
+
+      if (!it->isExternal() && !Symbol.isUndefined())
+        continue;
+
+      ELFSymbolData MSD;
+      MSD.SymbolData = it;
+
+      if (Symbol.isUndefined()) {
+        Undefined.push_back(MSD);
+      } else if (Symbol.isAbsolute()) {
+        External.push_back(MSD);
+      } else if (it->isCommon()) {
+        External.push_back(MSD);
+      } else {
+        External.push_back(MSD);
+      }
+    }
+
+    array_pod_sort(Local.begin(), Local.end());
+    array_pod_sort(External.begin(), External.end());
+    array_pod_sort(Undefined.begin(), Undefined.end());
+
+    for (unsigned i = 0, e = Local.size(); i != e; ++i)
+      if (&Local[i].SymbolData->getSymbol() == S)
+        return i + /* empty symbol */ 1;
+    for (unsigned i = 0, e = External.size(); i != e; ++i)
+      if (&External[i].SymbolData->getSymbol() == S)
+        return i + Local.size() + Asm.size() + /* empty symbol */ 1 +
+	   /* .rela.text + .rela.eh_frame */ + 2;
+    for (unsigned i = 0, e = Undefined.size(); i != e; ++i)
+      if (&Undefined[i].SymbolData->getSymbol() == S)
+        return i + Local.size() + External.size() + Asm.size() + /* empty symbol */ 1 +
+	   /* .rela.text + .rela.eh_frame */ + 2;
+}
+
+  /// ComputeSymbolTable - Compute the symbol table data
+  ///
+  /// \param StringTable [out] - The string table data.
+  /// \param StringIndexMap [out] - Map from symbol names to offsets in the
+  /// string table.
+  void ComputeSymbolTable(MCAssembler &Asm) {
+    // Build section lookup table.
+    DenseMap<const MCSection*, uint8_t> SectionIndexMap;
+    unsigned Index = 1;
+    for (MCAssembler::iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it, ++Index)
+      SectionIndexMap[&it->getSection()] = Index;
+
+    // Index 0 is always the empty string.
+    StringMap<uint64_t> StringIndexMap;
+    StringTable += '\x00';
+
+    // Add the data for local symbols.
+    for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
+           ie = Asm.symbol_end(); it != ie; ++it) {
+      const MCSymbol &Symbol = it->getSymbol();
+
+      // Ignore non-linker visible symbols.
+      if (!Asm.isSymbolLinkerVisible(Symbol))
+        continue;
+
+      if (it->isExternal() || Symbol.isUndefined())
+        continue;
+
+      uint64_t &Entry = StringIndexMap[Symbol.getName()];
+      if (!Entry) {
+        Entry = StringTable.size();
+        StringTable += Symbol.getName();
+        StringTable += '\x00';
+      }
+
+      ELFSymbolData MSD;
+      MSD.SymbolData = it;
+      MSD.StringIndex = Entry;
+
+      if (Symbol.isAbsolute()) {
+        MSD.SectionIndex = ELF::SHN_ABS;
+        LocalSymbolData.push_back(MSD);
+      } else {
+        MSD.SectionIndex = SectionIndexMap.lookup(&Symbol.getSection());
+        assert(MSD.SectionIndex && "Invalid section index!");
+        LocalSymbolData.push_back(MSD);
+      }
+    }
+
+    // Now add non-local symbols.
+    for (MCAssembler::symbol_iterator it = Asm.symbol_begin(),
+           ie = Asm.symbol_end(); it != ie; ++it) {
+      const MCSymbol &Symbol = it->getSymbol();
+
+      // Ignore non-linker visible symbols.
+      if (!Asm.isSymbolLinkerVisible(Symbol))
+        continue;
+
+      if (!it->isExternal() && !Symbol.isUndefined())
+        continue;
+
+      uint64_t &Entry = StringIndexMap[Symbol.getName()];
+      if (!Entry) {
+        Entry = StringTable.size();
+        StringTable += Symbol.getName();
+        StringTable += '\x00';
+      }
+
+      ELFSymbolData MSD;
+      MSD.SymbolData = it;
+      MSD.StringIndex = Entry;
+
+      if (Symbol.isUndefined()) {
+        MSD.SectionIndex = ELF::SHN_UNDEF;
+	// XXX: for some reason we dont Emit* this
+	it->setFlags(it->getFlags() | (ELF::STB_GLOBAL << ELF::STB_SHIFT));
+        UndefinedSymbolData.push_back(MSD);
+      } else if (Symbol.isAbsolute()) {
+        MSD.SectionIndex = ELF::SHN_ABS;
+        ExternalSymbolData.push_back(MSD);
+      } else if (it->isCommon()) {
+        MSD.SectionIndex = ELF::SHN_COMMON;
+        ExternalSymbolData.push_back(MSD);
+      } else {
+        MSD.SectionIndex = SectionIndexMap.lookup(&Symbol.getSection());
+        assert(MSD.SectionIndex && "Invalid section index!");
+        ExternalSymbolData.push_back(MSD);
+      }
+    }
+
+    // Symbols are required to be in lexicographic order.
+    array_pod_sort(LocalSymbolData.begin(), LocalSymbolData.end());
+    array_pod_sort(ExternalSymbolData.begin(), ExternalSymbolData.end());
+    array_pod_sort(UndefinedSymbolData.begin(), UndefinedSymbolData.end());
+
+    // Set the symbol indices. Local symbols must come before all other
+    // symbols with non-local bindings.
+    Index = 0;
+    for (unsigned i = 0, e = LocalSymbolData.size(); i != e; ++i)
+      LocalSymbolData[i].SymbolData->setIndex(Index++);
+    for (unsigned i = 0, e = ExternalSymbolData.size(); i != e; ++i)
+      ExternalSymbolData[i].SymbolData->setIndex(Index++);
+    for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i)
+      UndefinedSymbolData[i].SymbolData->setIndex(Index++);
+  }
+
+  void WriteRelocation(MCAssembler &Asm, MCAsmLayout &Layout,
+                       const MCSectionData &SD) {
+
+    if (!Relocations[&SD].empty()) {
+      MCContext &Ctx = Asm.getContext();
+      const MCSection *RelaSection;
+      const MCSectionELF &Section =
+        static_cast<const MCSectionELF&>(SD.getSection());
+
+      const StringRef SectionName = Section.getSectionName();
+      std::string RelaSectionName = ".rela";
+
+      RelaSectionName += SectionName;
+
+      RelaSection = Ctx.getELFSection(RelaSectionName, ELF::SHT_RELA, 0,
+                                      SectionKind::getReadOnly(), false);
+
+      MCSectionData &RelaSD = Asm.getOrCreateSectionData(*RelaSection);
+
+      RelaSD.setAlignment(1);
+      RelaSD.setEntrySize(Is64Bit ? ELF::SYMENTRY_SIZE64 : ELF::SYMENTRY_SIZE32);
+
+      MCDataFragment *F = new MCDataFragment(&RelaSD);
+
+      WriteRelocationsFragment(Asm, F, &SD);
+
+      Asm.AddSectionToTheEnd(RelaSD, Layout);
+    }
+  }
+
+  void WriteRelocations(MCAssembler &Asm, MCAsmLayout &Layout) {
+    for (MCAssembler::const_iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it) {
+      const MCSectionData &SD = *it;
+
+      WriteRelocation(Asm, Layout, SD);
+    }
+  }
+
+void CreateMetadataSections(MCAssembler &Asm, MCAsmLayout &Layout) {
+    MCContext &Ctx = Asm.getContext();
+    MCDataFragment *F;
+
+    WriteRelocations(Asm, Layout);
+
+    const MCSection *SymtabSection;
+    SymtabSection = Ctx.getELFSection(".symtab", ELF::SHT_SYMTAB, 0,
+                                      SectionKind::getReadOnly(), false);
+
+    MCSectionData &SymtabSD = Asm.getOrCreateSectionData(*SymtabSection);
+
+    SymtabSD.setAlignment(Is64Bit ? 8 : 4);
+    SymtabSD.setEntrySize(Is64Bit ? ELF::SYMENTRY_SIZE64 : ELF::SYMENTRY_SIZE32);
+
+    F = new MCDataFragment(&SymtabSD);
+
+    // Symbol table
+    WriteSymbolTable(F, Asm, Layout);
+    Asm.AddSectionToTheEnd(SymtabSD, Layout);
+
+    const MCSection *StrtabSection;
+    StrtabSection = Ctx.getELFSection(".strtab", ELF::SHT_STRTAB, 0,
+				      SectionKind::getReadOnly(), false);
+
+    MCSectionData &StrtabSD = Asm.getOrCreateSectionData(*StrtabSection);
+    StrtabSD.setAlignment(1);
+
+    // FIXME: This isn't right. If the sections get rearranged this will
+    // be wrong. We need a proper lookup.
+    StringTableIndex = Asm.size();
+
+    F = new MCDataFragment(&StrtabSD);
+    F->getContents().append(StringTable.begin(), StringTable.end());
+    Asm.AddSectionToTheEnd(StrtabSD, Layout);
+
+    const MCSection *ShstrtabSection;
+    ShstrtabSection = Ctx.getELFSection(".shstrtab", ELF::SHT_STRTAB, 0,
+					SectionKind::getReadOnly(), false);
+
+    MCSectionData &ShstrtabSD = Asm.getOrCreateSectionData(*ShstrtabSection);
+    ShstrtabSD.setAlignment(1);
+
+    F = new MCDataFragment(&ShstrtabSD);
+
+    // FIXME: This isn't right. If the sections get rearranged this will
+    // be wrong. We need a proper lookup.
+    ShstrtabIndex = Asm.size();
+
+    // Section header string table.
+    //
+    // The first entry of a string table holds a null character so skip
+    // section 0.
+    uint64_t Index = 1;
+    F->getContents() += '\x00';
+
+    for (MCAssembler::const_iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it) {
+      const MCSectionData &SD = *it;
+      const MCSectionELF &Section =
+        static_cast<const MCSectionELF&>(SD.getSection());
+
+
+      // Remember the index into the string table so we can write it
+      // into the sh_name field of the section header table.
+      SectionStringTableIndex[&it->getSection()] = Index;
+
+      Index += Section.getSectionName().size() + 1;
+      F->getContents() += Section.getSectionName();
+      F->getContents() += '\x00';
+    }
+
+    Asm.AddSectionToTheEnd(ShstrtabSD, Layout);
+  }
+
+  void ExecutePostLayoutBinding(MCAssembler &Asm) {
+  }
+
+  void WriteSecHdrEntry(uint32_t Name, uint32_t Type, uint64_t Flags,
+			uint64_t Address, uint64_t Offset,
+			uint64_t Size, uint32_t Link, uint32_t Info,
+			uint64_t Alignment, uint64_t EntrySize) {
+    // Suppose that all non-initialized EntrySizes are actually a zero
+    if (EntrySize == ~0U)
+      EntrySize = 0;
+
+    Write32(Name);        // sh_name: index into string table
+    Write32(Type);        // sh_type
+    WriteWord(Flags);     // sh_flags
+    WriteWord(Address);   // sh_addr
+    WriteWord(Offset);    // sh_offset
+    WriteWord(Size);      // sh_size
+    Write32(Link);        // sh_link
+    Write32(Info);        // sh_info
+    WriteWord(Alignment); // sh_addralign
+    WriteWord(EntrySize); // sh_entsize
+  }
+
+  void WriteRelocationsFragment(const MCAssembler &Asm, MCDataFragment *F,
+    const MCSectionData *SD) {
+      std::vector<ELFRelocationEntry> &Relocs = Relocations[SD];
+      // sort by the r_offset just like gnu as does
+      array_pod_sort(Relocs.begin(), Relocs.end());
+
+      for (unsigned i = 0, e = Relocs.size(); i != e; ++i) {
+        ELFRelocationEntry entry = Relocs[e - i - 1];
+
+        if (Is64Bit) {
+	  F->getContents() +=  StringRef((const char *)&entry.r_offset, 8);
+	  F->getContents() +=  StringRef((const char *)&entry.r_info, 8);
+
+          if (HasRelocationAddend)
+	    F->getContents() +=  StringRef((const char *)&entry.r_addend, 8);
+        } else {
+	  F->getContents() +=  StringRef((const char *)&entry.r_offset, 4);
+	  F->getContents() +=  StringRef((const char *)&entry.r_info, 4);
+
+	  if (HasRelocationAddend)
+	    F->getContents() +=  StringRef((const char *)&entry.r_addend, 4);
+        }
+      }
+  }
+
+  void WriteObject(const MCAssembler &Asm, const MCAsmLayout &Layout) {
+    // Compute symbol table information.
+    ComputeSymbolTable(const_cast<MCAssembler&>(Asm));
+
+    CreateMetadataSections(const_cast<MCAssembler&>(Asm),
+                           const_cast<MCAsmLayout&>(Layout));
+
+    // Add 1 for the null section.
+    unsigned NumSections = Asm.size() + 1;
+
+    uint64_t SectionDataSize = 0;
+
+    for (MCAssembler::const_iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it) {
+      const MCSectionData &SD = *it;
+
+      // Get the size of the section in the output file (including padding).
+      uint64_t Size = Layout.getSectionFileSize(&SD);
+      SectionDataSize += Size;
+    }
+
+    // Write out the ELF header ...
+    WriteHeader(SectionDataSize, NumSections);
+    FileOff = Is64Bit ? 64 : 52;
+
+    // ... then all of the sections ...
+    DenseMap<const MCSection*, uint64_t> SectionOffsetMap;
+
+    for (MCAssembler::const_iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it) {
+      // Remember the offset into the file for this section.
+      SectionOffsetMap[&it->getSection()] = FileOff;
+
+      const MCSectionData &SD = *it;
+      FileOff += Layout.getSectionFileSize(&SD);
+
+      Asm.WriteSectionData(it, Layout, Writer);
+    }
+
+    // ... and then the section header table.
+    // Should we align the section header table?
+    //
+    // Null secton first.
+    WriteSecHdrEntry(0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+
+    for (MCAssembler::const_iterator it = Asm.begin(),
+           ie = Asm.end(); it != ie; ++it) {
+      const MCSectionData &SD = *it;
+      const MCSectionELF &Section =
+        static_cast<const MCSectionELF&>(SD.getSection());
+
+      uint64_t sh_link = 0;
+      uint64_t sh_info = 0;
+
+      switch(Section.getType()) {
+      case ELF::SHT_DYNAMIC:
+        sh_link = SectionStringTableIndex[&it->getSection()];
+        sh_info = 0;
+        break;
+
+      case ELF::SHT_REL:
+      case ELF::SHT_RELA:
+        const MCSection *SymtabSection;
+        const MCSection *InfoSection;
+        const StringRef *SectionName;
+        const MCSectionData *SymtabSD;
+        const MCSectionData *InfoSD;
+
+        SymtabSection = Asm.getContext().getELFSection(".symtab", ELF::SHT_SYMTAB, 0,
+                        SectionKind::getReadOnly(), false);
+        SymtabSD = &Asm.getSectionData(*SymtabSection);
+        // we have to count the empty section in too
+        sh_link = SymtabSD->getLayoutOrder() + 1;
+
+        SectionName = &Section.getSectionName();
+        SectionName = &SectionName->slice(5, SectionName->size());
+        InfoSection = Asm.getContext().getELFSection(*SectionName,
+          ELF::SHT_PROGBITS, 0, SectionKind::getReadOnly(), false);
+        InfoSD = &Asm.getSectionData(*InfoSection);
+        sh_info = InfoSD->getLayoutOrder() + 1;
+        break;
+
+      case ELF::SHT_SYMTAB:
+      case ELF::SHT_DYNSYM:
+        sh_link = StringTableIndex;
+        sh_info = LastLocalSymbolIndex;
+        break;
+
+      case ELF::SHT_HASH:
+      case ELF::SHT_GROUP:
+      case ELF::SHT_SYMTAB_SHNDX:
+        assert(0 && "FIXME: sh_type value not supported!");
+        break;
+      }
+
+      WriteSecHdrEntry(SectionStringTableIndex[&it->getSection()],
+		       Section.getType(), Section.getFlags(),
+		       Layout.getSectionAddress(&SD),
+		       SectionOffsetMap.lookup(&SD.getSection()),
+		       Layout.getSectionSize(&SD), sh_link,
+                       sh_info, SD.getAlignment(),
+                       SD.getEntrySize());
+    }
+  }
+};
+
+}
+
+ELFObjectWriter::ELFObjectWriter(raw_ostream &OS,
+                                 bool Is64Bit,
+                                 bool IsLittleEndian,
+                                 bool HasRelocationAddend)
+  : MCObjectWriter(OS, IsLittleEndian)
+{
+  Impl = new ELFObjectWriterImpl(this, Is64Bit, HasRelocationAddend);
+}
+
+ELFObjectWriter::~ELFObjectWriter() {
+  delete (ELFObjectWriterImpl*) Impl;
+}
+
+void ELFObjectWriter::ExecutePostLayoutBinding(MCAssembler &Asm) {
+  ((ELFObjectWriterImpl*) Impl)->ExecutePostLayoutBinding(Asm);
+}
+
+void ELFObjectWriter::RecordRelocation(const MCAssembler &Asm,
+                                        const MCAsmLayout &Layout,
+                                        const MCFragment *Fragment,
+                                        const MCFixup &Fixup, MCValue Target,
+                                        uint64_t &FixedValue) {
+  ((ELFObjectWriterImpl*) Impl)->RecordRelocation(Asm, Layout, Fragment, Fixup,
+                                                   Target, FixedValue);
+}
+
+void ELFObjectWriter::WriteObject(const MCAssembler &Asm,
+                                   const MCAsmLayout &Layout) {
+  ((ELFObjectWriterImpl*) Impl)->WriteObject(Asm, Layout);
+}
diff --git a/lib/MC/MCELFStreamer.cpp b/lib/MC/MCELFStreamer.cpp
new file mode 100644
index 0000000..291d181
--- /dev/null
+++ b/lib/MC/MCELFStreamer.cpp
@@ -0,0 +1,496 @@
+//===- lib/MC/MCELFStreamer.cpp - ELF Object Output ------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file assembles .s files and emits ELF .o object files.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/MC/MCStreamer.h"
+
+#include "llvm/MC/MCAssembler.h"
+#include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCCodeEmitter.h"
+#include "llvm/MC/MCExpr.h"
+#include "llvm/MC/MCInst.h"
+#include "llvm/MC/MCSection.h"
+#include "llvm/MC/MCSectionELF.h"
+#include "llvm/MC/MCSymbol.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/ELF.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Target/TargetAsmBackend.h"
+
+using namespace llvm;
+
+namespace {
+
+class MCELFStreamer : public MCStreamer {
+private:
+  MCAssembler Assembler;
+  MCSectionData *CurSectionData;
+
+private:
+  MCFragment *getCurrentFragment() const {
+    assert(CurSectionData && "No current section!");
+
+    if (!CurSectionData->empty())
+      return &CurSectionData->getFragmentList().back();
+
+    return 0;
+  }
+
+  /// Get a data fragment to write into, creating a new one if the current
+  /// fragment is not a data fragment.
+  MCDataFragment *getOrCreateDataFragment() const {
+    MCDataFragment *F = dyn_cast_or_null<MCDataFragment>(getCurrentFragment());
+    if (!F)
+      F = new MCDataFragment(CurSectionData);
+    return F;
+  }
+
+public:
+  MCELFStreamer(MCContext &Context, TargetAsmBackend &TAB,
+                  raw_ostream &_OS, MCCodeEmitter *_Emitter)
+    : MCStreamer(Context), Assembler(Context, TAB, *_Emitter, _OS),
+      CurSectionData(0) {}
+  ~MCELFStreamer() {}
+
+  MCAssembler &getAssembler() { return Assembler; }
+
+  const MCExpr *AddValueSymbols(const MCExpr *Value) {
+    switch (Value->getKind()) {
+    case MCExpr::Target: assert(0 && "Can't handle target exprs yet!");
+    case MCExpr::Constant:
+      break;
+
+    case MCExpr::Binary: {
+      const MCBinaryExpr *BE = cast<MCBinaryExpr>(Value);
+      AddValueSymbols(BE->getLHS());
+      AddValueSymbols(BE->getRHS());
+      break;
+    }
+
+    case MCExpr::SymbolRef:
+      Assembler.getOrCreateSymbolData(
+        cast<MCSymbolRefExpr>(Value)->getSymbol());
+      break;
+
+    case MCExpr::Unary:
+      AddValueSymbols(cast<MCUnaryExpr>(Value)->getSubExpr());
+      break;
+    }
+
+    return Value;
+  }
+
+  /// @name MCStreamer Interface
+  /// @{
+
+  virtual void SwitchSection(const MCSection *Section);
+  virtual void EmitLabel(MCSymbol *Symbol);
+  virtual void EmitAssemblerFlag(MCAssemblerFlag Flag);
+  virtual void EmitAssignment(MCSymbol *Symbol, const MCExpr *Value);
+  virtual void EmitSymbolAttribute(MCSymbol *Symbol, MCSymbolAttr Attribute);
+  virtual void EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue);
+  virtual void EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size,
+                                unsigned ByteAlignment);
+  virtual void BeginCOFFSymbolDef(const MCSymbol *Symbol) {
+    assert(0 && "ELF doesn't support this directive");
+  }
+
+  virtual void EmitCOFFSymbolStorageClass(int StorageClass) {
+    assert(0 && "ELF doesn't support this directive");
+  }
+
+  virtual void EmitCOFFSymbolType(int Type) {
+    assert(0 && "ELF doesn't support this directive");
+  }
+
+  virtual void EndCOFFSymbolDef() {
+    assert(0 && "ELF doesn't support this directive");
+  }
+
+  virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) {
+     MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+     uint64_t Size = 0;
+     const MCBinaryExpr *BE = NULL;
+
+     if (Value->getKind() == MCExpr::Constant) {
+       const MCConstantExpr *CE;
+       CE = static_cast<const MCConstantExpr *>(Value);
+       Size = CE->getValue();
+     } else if (Value->getKind() == MCExpr::Binary) {
+       BE = static_cast<const MCBinaryExpr *>(Value);
+     }
+
+     if (Size)
+       SD.setSize(Size);
+     if (BE)
+       SD.setSizeSymbol(BE);
+  }
+
+  virtual void EmitLocalCommonSymbol(MCSymbol *Symbol, uint64_t Size) {
+    assert(0 && "ELF doesn't support this directive");
+  }
+  virtual void EmitZerofill(const MCSection *Section, MCSymbol *Symbol = 0,
+                            unsigned Size = 0, unsigned ByteAlignment = 0);
+  virtual void EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol,
+			      uint64_t Size, unsigned ByteAlignment = 0);
+  virtual void EmitBytes(StringRef Data, unsigned AddrSpace);
+  virtual void EmitValue(const MCExpr *Value, unsigned Size,unsigned AddrSpace);
+  virtual void EmitGPRel32Value(const MCExpr *Value) {
+    assert(0 && "ELF doesn't support this directive");
+  }
+  virtual void EmitValueToAlignment(unsigned ByteAlignment, int64_t Value = 0,
+                                    unsigned ValueSize = 1,
+                                    unsigned MaxBytesToEmit = 0);
+  virtual void EmitCodeAlignment(unsigned ByteAlignment,
+                                 unsigned MaxBytesToEmit = 0);
+  virtual void EmitValueToOffset(const MCExpr *Offset,
+                                 unsigned char Value = 0);
+
+  virtual void EmitFileDirective(StringRef Filename);
+  virtual void EmitDwarfFileDirective(unsigned FileNo, StringRef Filename) {
+    errs() << "FIXME: MCELFStreamer:EmitDwarfFileDirective not implemented\n";
+  }
+
+  virtual void EmitInstruction(const MCInst &Inst);
+  virtual void Finish();
+
+  /// @}
+};
+
+} // end anonymous namespace.
+
+void MCELFStreamer::SwitchSection(const MCSection *Section) {
+  assert(Section && "Cannot switch to a null section!");
+
+  // If already in this section, then this is a noop.
+  if (Section == CurSection) return;
+
+  CurSection = Section;
+  CurSectionData = &Assembler.getOrCreateSectionData(*Section);
+}
+
+void MCELFStreamer::EmitLabel(MCSymbol *Symbol) {
+  assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
+
+  // FIXME: This is wasteful, we don't necessarily need to create a data
+  // fragment. Instead, we should mark the symbol as pointing into the data
+  // fragment if it exists, otherwise we should just queue the label and set its
+  // fragment pointer when we emit the next fragment.
+  MCDataFragment *F = getOrCreateDataFragment();
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+  assert(!SD.getFragment() && "Unexpected fragment on symbol data!");
+  SD.setFragment(F);
+  SD.setOffset(F->getContents().size());
+
+  Symbol->setSection(*CurSection);
+}
+
+void MCELFStreamer::EmitAssemblerFlag(MCAssemblerFlag Flag) {
+  switch (Flag) {
+  case MCAF_SubsectionsViaSymbols:
+    Assembler.setSubsectionsViaSymbols(true);
+    return;
+  }
+
+  assert(0 && "invalid assembler flag!");
+}
+
+void MCELFStreamer::EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) {
+  // Only absolute symbols can be redefined.
+  assert((Symbol->isUndefined() || Symbol->isAbsolute()) &&
+         "Cannot define a symbol twice!");
+
+  // FIXME: Lift context changes into super class.
+  // FIXME: Set associated section.
+  //  Symbol->setValue(AddValueSymbols(Value));
+}
+
+void MCELFStreamer::EmitSymbolAttribute(MCSymbol *Symbol,
+                                          MCSymbolAttr Attribute) {
+  // Indirect symbols are handled differently, to match how 'as' handles
+  // them. This makes writing matching .o files easier.
+  if (Attribute == MCSA_IndirectSymbol) {
+    // Note that we intentionally cannot use the symbol data here; this is
+    // important for matching the string table that 'as' generates.
+    IndirectSymbolData ISD;
+    ISD.Symbol = Symbol;
+    ISD.SectionData = CurSectionData;
+    Assembler.getIndirectSymbols().push_back(ISD);
+    return;
+  }
+
+  // Adding a symbol attribute always introduces the symbol, note that an
+  // important side effect of calling getOrCreateSymbolData here is to register
+  // the symbol with the assembler.
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+
+  // The implementation of symbol attributes is designed to match 'as', but it
+  // leaves much to desired. It doesn't really make sense to arbitrarily add and
+  // remove flags, but 'as' allows this (in particular, see .desc).
+  //
+  // In the future it might be worth trying to make these operations more well
+  // defined.
+  switch (Attribute) {
+  case MCSA_LazyReference:
+  case MCSA_Reference:
+  case MCSA_NoDeadStrip:
+  case MCSA_PrivateExtern:
+  case MCSA_WeakReference:
+  case MCSA_WeakDefinition:
+  case MCSA_Invalid:
+  case MCSA_ELF_TypeIndFunction:
+  case MCSA_IndirectSymbol:
+    assert(0 && "Invalid symbol attribute for ELF!");
+    break;
+
+  case MCSA_Global:
+    SD.setFlags(SD.getFlags() | (ELF::STB_GLOBAL << ELF::STB_SHIFT));
+    SD.setExternal(true);
+    break;
+
+  case MCSA_Weak:
+    SD.setFlags(SD.getFlags() | (ELF::STB_WEAK << ELF::STB_SHIFT));
+    break;
+
+  case MCSA_Local:
+    SD.setFlags(SD.getFlags() | (ELF::STB_LOCAL << ELF::STB_SHIFT));
+    break;
+
+  case MCSA_ELF_TypeFunction:
+    SD.setFlags(SD.getFlags() | (ELF::STT_FUNC << ELF::STT_SHIFT));
+    break;
+
+  case MCSA_ELF_TypeObject:
+    SD.setFlags(SD.getFlags() | (ELF::STT_OBJECT << ELF::STT_SHIFT));
+    break;
+
+  case MCSA_ELF_TypeTLS:
+    SD.setFlags(SD.getFlags() | (ELF::STT_TLS << ELF::STT_SHIFT));
+    break;
+
+  case MCSA_ELF_TypeCommon:
+    SD.setFlags(SD.getFlags() | (ELF::STT_COMMON << ELF::STT_SHIFT));
+    break;
+
+  case MCSA_ELF_TypeNoType:
+    SD.setFlags(SD.getFlags() | (ELF::STT_NOTYPE << ELF::STT_SHIFT));
+    break;
+
+  case MCSA_Protected:
+    SD.setFlags(SD.getFlags() | (ELF::STV_PROTECTED << ELF::STV_SHIFT));
+    break;
+
+  case MCSA_Hidden:
+    SD.setFlags(SD.getFlags() | (ELF::STV_HIDDEN << ELF::STV_SHIFT));
+    break;
+
+  case MCSA_Internal:
+    SD.setFlags(SD.getFlags() | (ELF::STV_INTERNAL << ELF::STV_SHIFT));
+    break;
+  }
+}
+
+void MCELFStreamer::EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) {
+}
+
+void MCELFStreamer::EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size,
+                                       unsigned ByteAlignment) {
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+
+  if ((SD.getFlags() & (0xff << ELF::STB_SHIFT)) == ELF::STB_LOCAL << ELF::STB_SHIFT) {
+    const MCSection *Section = Assembler.getContext().getELFSection(".bss",
+                                                                    MCSectionELF::SHT_NOBITS,
+                                                                    MCSectionELF::SHF_WRITE |
+                                                                    MCSectionELF::SHF_ALLOC,
+                                                                    SectionKind::getBSS());
+
+    MCSectionData &SectData = Assembler.getOrCreateSectionData(*Section);
+    MCFragment *F = new MCFillFragment(0, 0, Size, &SectData);
+    SD.setFragment(F);
+    Symbol->setSection(*Section);
+  } else {
+    SD.setExternal(true);
+  }
+
+  SD.setCommon(Size, ByteAlignment);
+}
+
+void MCELFStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol,
+                                   unsigned Size, unsigned ByteAlignment) {
+  MCSectionData &SectData = Assembler.getOrCreateSectionData(*Section);
+
+  // The symbol may not be present, which only creates the section.
+  if (!Symbol)
+    return;
+
+  // FIXME: Assert that this section has the zerofill type.
+
+  assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
+
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+
+  MCFragment *F = new MCFillFragment(0, 0, Size, &SectData);
+  SD.setFragment(F);
+
+  Symbol->setSection(*Section);
+
+  // Update the maximum alignment on the zero fill section if necessary.
+  if (ByteAlignment > SectData.getAlignment())
+    SectData.setAlignment(ByteAlignment);
+}
+
+// This should always be called with the thread local bss section.  Like the
+// .zerofill directive this doesn't actually switch sections on us.
+void MCELFStreamer::EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol,
+                                   uint64_t Size, unsigned ByteAlignment) {
+  EmitZerofill(Section, Symbol, Size, ByteAlignment);
+}
+
+void MCELFStreamer::EmitBytes(StringRef Data, unsigned AddrSpace) {
+  getOrCreateDataFragment()->getContents().append(Data.begin(), Data.end());
+}
+
+void MCELFStreamer::EmitValue(const MCExpr *Value, unsigned Size,
+                                unsigned AddrSpace) {
+  MCDataFragment *DF = getOrCreateDataFragment();
+
+  // Avoid fixups when possible.
+  int64_t AbsValue;
+  if (AddValueSymbols(Value)->EvaluateAsAbsolute(AbsValue)) {
+    // FIXME: Endianness assumption.
+    for (unsigned i = 0; i != Size; ++i)
+      DF->getContents().push_back(uint8_t(AbsValue >> (i * 8)));
+  } else {
+    DF->addFixup(MCFixup::Create(DF->getContents().size(), AddValueSymbols(Value),
+                                 MCFixup::getKindForSize(Size)));
+    DF->getContents().resize(DF->getContents().size() + Size, 0);
+  }
+}
+
+void MCELFStreamer::EmitValueToAlignment(unsigned ByteAlignment,
+                                           int64_t Value, unsigned ValueSize,
+                                           unsigned MaxBytesToEmit) {
+  if (MaxBytesToEmit == 0)
+    MaxBytesToEmit = ByteAlignment;
+  new MCAlignFragment(ByteAlignment, Value, ValueSize, MaxBytesToEmit,
+                      CurSectionData);
+
+  // Update the maximum alignment on the current section if necessary.
+  if (ByteAlignment > CurSectionData->getAlignment())
+    CurSectionData->setAlignment(ByteAlignment);
+}
+
+void MCELFStreamer::EmitCodeAlignment(unsigned ByteAlignment,
+                                        unsigned MaxBytesToEmit) {
+  if (MaxBytesToEmit == 0)
+    MaxBytesToEmit = ByteAlignment;
+  new MCAlignFragment(ByteAlignment, 0, 1, MaxBytesToEmit,
+                      CurSectionData);
+
+  // Update the maximum alignment on the current section if necessary.
+  if (ByteAlignment > CurSectionData->getAlignment())
+    CurSectionData->setAlignment(ByteAlignment);
+}
+
+void MCELFStreamer::EmitValueToOffset(const MCExpr *Offset,
+                                        unsigned char Value) {
+  new MCOrgFragment(*Offset, Value, CurSectionData);
+}
+
+// Add a symbol for the file name of this module. This is the second
+// entry in the module's symbol table (the first being the null symbol).
+void MCELFStreamer::EmitFileDirective(StringRef Filename) {
+  MCSymbol *Symbol = Assembler.getContext().GetOrCreateSymbol(Filename);
+  Symbol->setSection(*CurSection);
+  Symbol->setAbsolute();
+
+  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
+
+  SD.setFlags(ELF::STT_FILE << ELF::STT_SHIFT);
+  SD.setFlags(SD.getFlags() | (ELF::STB_LOCAL << ELF::STB_SHIFT));
+  SD.setFlags(SD.getFlags() | (ELF::STV_DEFAULT << ELF::STV_SHIFT));
+
+}
+
+void MCELFStreamer::EmitInstruction(const MCInst &Inst) {
+  // Scan for values.
+  for (unsigned i = 0; i != Inst.getNumOperands(); ++i)
+    if (Inst.getOperand(i).isExpr())
+      AddValueSymbols(Inst.getOperand(i).getExpr());
+
+  CurSectionData->setHasInstructions(true);
+
+  // FIXME-PERF: Common case is that we don't need to relax, encode directly
+  // onto the data fragments buffers.
+
+  SmallVector<MCFixup, 4> Fixups;
+  SmallString<256> Code;
+  raw_svector_ostream VecOS(Code);
+  Assembler.getEmitter().EncodeInstruction(Inst, VecOS, Fixups);
+  VecOS.flush();
+
+  // FIXME: Eliminate this copy.
+  SmallVector<MCFixup, 4> AsmFixups;
+  for (unsigned i = 0, e = Fixups.size(); i != e; ++i) {
+    MCFixup &F = Fixups[i];
+    AsmFixups.push_back(MCFixup::Create(F.getOffset(), F.getValue(),
+                                        F.getKind()));
+  }
+
+  // See if we might need to relax this instruction, if so it needs its own
+  // fragment.
+  //
+  // FIXME-PERF: Support target hook to do a fast path that avoids the encoder,
+  // when we can immediately tell that we will get something which might need
+  // relaxation (and compute its size).
+  //
+  // FIXME-PERF: We should also be smart about immediately relaxing instructions
+  // which we can already show will never possibly fit (we can also do a very
+  // good job of this before we do the first relaxation pass, because we have
+  // total knowledge about undefined symbols at that point). Even now, though,
+  // we can do a decent job, especially on Darwin where scattering means that we
+  // are going to often know that we can never fully resolve a fixup.
+  if (Assembler.getBackend().MayNeedRelaxation(Inst)) {
+    MCInstFragment *IF = new MCInstFragment(Inst, CurSectionData);
+
+    // Add the fixups and data.
+    //
+    // FIXME: Revisit this design decision when relaxation is done, we may be
+    // able to get away with not storing any extra data in the MCInst.
+    IF->getCode() = Code;
+    IF->getFixups() = AsmFixups;
+
+    return;
+  }
+
+  // Add the fixups and data.
+  MCDataFragment *DF = getOrCreateDataFragment();
+  for (unsigned i = 0, e = AsmFixups.size(); i != e; ++i) {
+    AsmFixups[i].setOffset(AsmFixups[i].getOffset() + DF->getContents().size());
+    DF->addFixup(AsmFixups[i]);
+  }
+  DF->getContents().append(Code.begin(), Code.end());
+}
+
+void MCELFStreamer::Finish() {
+  Assembler.Finish();
+}
+
+MCStreamer *llvm::createELFStreamer(MCContext &Context, TargetAsmBackend &TAB,
+                                      raw_ostream &OS, MCCodeEmitter *CE,
+                                      bool RelaxAll) {
+  MCELFStreamer *S = new MCELFStreamer(Context, TAB, OS, CE);
+  if (RelaxAll)
+    S->getAssembler().setRelaxAll(true);
+  return S;
+}
-- 
1.6.4.rc0



From rdivacky at freebsd.org  Mon Jul 26 15:04:33 2010
From: rdivacky at freebsd.org (Roman Divacky)
Date: Mon, 26 Jul 2010 22:04:33 +0200
Subject: [llvm-commits] [PATCH 0/5] ELF object support
In-Reply-To: <cover.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
Message-ID: <20100726200433.GA85270@freebsd.org>

and if you apply:

	http://lev.vlakno.cz/~rdivacky/clang-object-streamer.patch

you can use "clang -integrated-as" directly. this patch should/could be
committed regardless of the ELF patches as we got MachO and COFF.

thnx


On Mon, Jul 26, 2010 at 09:00:18PM +0100, Matt Fleming wrote:
> This series contains the patches that have been sitting in my github
> tree that me and Roman Divacky have been working on for a while now. It
> isn't a complete implementation but is enough to get a semi-complete
> relocatable ELF file emitted.
> 
> I've tried to split the patches in a sensible way; bits required by the
> ELF code but not ELF-specific come first, then the code ELF support,
> then a patch to wire it up for X86.
> 
> All comments and feedback welcome.
> 
> Matt Fleming (5):
>   Some ELF sections contain fixed-sized entries. Provide a way to
>     record the entry size of a section.
>   Record a symbol's size which is needed for ELF symbol tables.
>   Introduce a helper method to add a section to the end of a layout.
>     This will be used by the ELF ObjectWriter code to add the metadata
>     sections (symbol table, etc) to the end of an object file.
>   Add ELF ObjectWriter and Streamer support.
>   Hookup the ELF support for X86.
> 
>  include/llvm/MC/ELFObjectWriter.h   |   46 ++
>  include/llvm/MC/MCAssembler.h       |   32 ++
>  include/llvm/MC/MCStreamer.h        |    6 +
>  include/llvm/Support/ELF.h          |   14 +
>  lib/MC/CMakeLists.txt               |    2 +
>  lib/MC/ELFObjectWriter.cpp          |  873 +++++++++++++++++++++++++++++++++++
>  lib/MC/MCAssembler.cpp              |   38 ++-
>  lib/MC/MCELFStreamer.cpp            |  496 ++++++++++++++++++++
>  lib/Target/X86/X86AsmBackend.cpp    |   17 +-
>  lib/Target/X86/X86TargetMachine.cpp |    2 +
>  10 files changed, 1521 insertions(+), 5 deletions(-)
>  create mode 100644 include/llvm/MC/ELFObjectWriter.h
>  create mode 100644 lib/MC/ELFObjectWriter.cpp
>  create mode 100644 lib/MC/MCELFStreamer.cpp
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100726/7266379a/attachment.bin 

From gohman at apple.com  Mon Jul 26 15:11:39 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 20:11:39 -0000
Subject: [llvm-commits] [llvm] r109429 - in /llvm/trunk/test:
	Analysis/RegionInfo/ MC/AsmParser/ELF/
Message-ID: <20100726201139.557DD2A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 15:11:39 2010
New Revision: 109429

URL: http://llvm.org/viewvc/llvm-project?rev=109429&view=rev
Log:
Add svn:ignore.

Modified:
    llvm/trunk/test/Analysis/RegionInfo/   (props changed)
    llvm/trunk/test/MC/AsmParser/ELF/   (props changed)

Propchange: llvm/trunk/test/Analysis/RegionInfo/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Mon Jul 26 15:11:39 2010
@@ -0,0 +1 @@
+Output

Propchange: llvm/trunk/test/MC/AsmParser/ELF/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Mon Jul 26 15:11:39 2010
@@ -0,0 +1 @@
+Output




From gohman at apple.com  Mon Jul 26 15:15:47 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 20:15:47 -0000
Subject: [llvm-commits] [llvm] r109430 - /llvm/trunk/Makefile.rules
Message-ID: <20100726201547.85A972A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 15:15:47 2010
New Revision: 109430

URL: http://llvm.org/viewvc/llvm-project?rev=109430&view=rev
Log:
Fix library build messages.

Modified:
    llvm/trunk/Makefile.rules

Modified: llvm/trunk/Makefile.rules
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=109430&r1=109429&r2=109430&view=diff
==============================================================================
--- llvm/trunk/Makefile.rules (original)
+++ llvm/trunk/Makefile.rules Mon Jul 26 15:15:47 2010
@@ -1125,12 +1125,12 @@
 endif
 $(LibName.SO): $(ObjectsO) $(ProjLibsPaths) $(LLVMLibsPaths) $(LibDir)/.dir
 	$(Echo) Linking $(BuildMode) $(SharedLibKindMessage) \
-	  $(LIBRARYNAME)$(SHLIBEXT)
+	  $(notdir $@)
 	$(Verb) $(Link) $(SharedLinkOptions) -o $@ $(ObjectsO) \
 	  $(ProjLibsOptions) $(LLVMLibsOptions) $(LIBS)
 else
 $(LibName.SO): $(ObjectsO) $(LibDir)/.dir
-	$(Echo) Linking $(BuildMode) Shared Library $(basename $@)
+	$(Echo) Linking $(BuildMode) Shared Library $(notdir $@)
 	$(Verb) $(Link) $(SharedLinkOptions) -o $@ $(ObjectsO)
 endif
 




From resistor at mac.com  Mon Jul 26 15:19:04 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 20:19:04 -0000
Subject: [llvm-commits] [llvm] r109431 - /llvm/trunk/README.txt
Message-ID: <20100726201904.329262A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 15:19:04 2010
New Revision: 109431

URL: http://llvm.org/viewvc/llvm-project?rev=109431&view=rev
Log:
Test some more.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109431&r1=109430&r2=109431&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 15:19:04 2010
@@ -13,3 +13,4 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
+




From eli.friedman at gmail.com  Mon Jul 26 15:32:42 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 26 Jul 2010 13:32:42 -0700
Subject: [llvm-commits] [PATCH 5/5] Hookup the ELF support for X86.
In-Reply-To: <a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
	<a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>
Message-ID: <AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>

On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
> --- a/lib/Target/X86/X86TargetMachine.cpp
> +++ b/lib/Target/X86/X86TargetMachine.cpp
> @@ -46,6 +46,8 @@ static MCStreamer *createMCStreamer(const Target &T, const std::string &TT,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? bool RelaxAll) {
> ? Triple TheTriple(TT);
> ? switch (TheTriple.getOS()) {
> + ?case Triple::Linux:
> + ? ?return createELFStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);

+ case Triple::AuroraUX:
+ case Triple::DragonFly:
+ case Triple::NetBSD:
+ case Triple::OpenBSD:
+ case Triple::FreeBSD:
+ case Triple::Solaris:
+ case Triple::Haiku:

-Eli



From benny.kra at googlemail.com  Mon Jul 26 15:37:21 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Mon, 26 Jul 2010 22:37:21 +0200
Subject: [llvm-commits] [PATCH 5/5] Hookup the ELF support for X86.
In-Reply-To: <AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>
References: <cover.1280173465.git.matt@console-pimps.org>
	<a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>
	<AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>
Message-ID: <5E482B11-F629-405E-9317-226EB3B59515@gmail.com>


On 26.07.2010, at 22:32, Eli Friedman wrote:

> On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
>> --- a/lib/Target/X86/X86TargetMachine.cpp
>> +++ b/lib/Target/X86/X86TargetMachine.cpp
>> @@ -46,6 +46,8 @@ static MCStreamer *createMCStreamer(const Target &T, const std::string &TT,
>>                                     bool RelaxAll) {
>>   Triple TheTriple(TT);
>>   switch (TheTriple.getOS()) {
>> +  case Triple::Linux:
>> +    return createELFStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
> 
> + case Triple::AuroraUX:
> + case Triple::DragonFly:
> + case Triple::NetBSD:
> + case Triple::OpenBSD:
> + case Triple::FreeBSD:
> + case Triple::Solaris:
> + case Triple::Haiku:

Can't we just make ELF the default? There are way more ELF targets than MachO targets and we default to ELF in other places.


From eli.friedman at gmail.com  Mon Jul 26 15:43:42 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 26 Jul 2010 13:43:42 -0700
Subject: [llvm-commits] [PATCH 1/5] Some ELF sections contain
	fixed-sized entries. Provide a way to record the entry size
	of a section.
In-Reply-To: <c48da6bb8b6f1b0ec18e643ccc0769677f4884da.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
	<c48da6bb8b6f1b0ec18e643ccc0769677f4884da.1280173465.git.matt@console-pimps.org>
Message-ID: <AANLkTinOt3YxypzjLTFoKL_fnaf=SB2W__dQ=L=gy4rU@mail.gmail.com>

On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
> ---
> ?include/llvm/MC/MCAssembler.h | ? ?8 ++++++++
> ?lib/MC/MCAssembler.cpp ? ? ? ?| ? ?1 +
> ?2 files changed, 9 insertions(+), 0 deletions(-)
>
> diff --git a/include/llvm/MC/MCAssembler.h b/include/llvm/MC/MCAssembler.h
> index 97ab1a1..51f265d 100644
> --- a/include/llvm/MC/MCAssembler.h
> +++ b/include/llvm/MC/MCAssembler.h
> @@ -366,6 +366,11 @@ private:
> ? /// Alignment - The maximum alignment seen in this section.
> ? unsigned Alignment;
>
> + ?/// EntrySize - The size of each entry in this section. This is ~0
> + ?/// until initialized. This size only makes sense for sections that
> + ?/// contain fixed-sized entries.
> + ?unsigned EntrySize;

I'm not sure this is the right place for this... I'm pretty sure
getStubSize() is roughly the equivalent on Darwin, and it's in
MCSectionMacho.

-Eli



From matt at console-pimps.org  Mon Jul 26 15:50:18 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Mon, 26 Jul 2010 21:50:18 +0100
Subject: [llvm-commits] [PATCH 5/5] Hookup the ELF support for X86.
In-Reply-To: <AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>
References: <cover.1280173465.git.matt@console-pimps.org>
	<a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>
	<AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>
Message-ID: <87iq42c6id.fsf@linux-g6p1.site>

On Mon, 26 Jul 2010 13:32:42 -0700, Eli Friedman <eli.friedman at gmail.com> wrote:
> On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
> > --- a/lib/Target/X86/X86TargetMachine.cpp
> > +++ b/lib/Target/X86/X86TargetMachine.cpp
> > @@ -46,6 +46,8 @@ static MCStreamer *createMCStreamer(const Target &T, const std::string &TT,
> > ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? bool RelaxAll) {
> > ? Triple TheTriple(TT);
> > ? switch (TheTriple.getOS()) {
> > + ?case Triple::Linux:
> > + ? ?return createELFStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
> 
> + case Triple::AuroraUX:
> + case Triple::DragonFly:
> + case Triple::NetBSD:
> + case Triple::OpenBSD:
> + case Triple::FreeBSD:
> + case Triple::Solaris:
> + case Triple::Haiku:

Whoops! Makes sense. I think Roman told me to do this before but I
forgot ;-) I've folded this into the patch series.

Thanks!



From resistor at mac.com  Mon Jul 26 15:51:45 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 20:51:45 -0000
Subject: [llvm-commits] [llvm] r109433 - /llvm/trunk/README.txt
Message-ID: <20100726205145.D74662A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 15:51:45 2010
New Revision: 109433

URL: http://llvm.org/viewvc/llvm-project?rev=109433&view=rev
Log:
Testing some more.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109433&r1=109432&r2=109433&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 15:51:45 2010
@@ -13,4 +13,3 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
-




From eli.friedman at gmail.com  Mon Jul 26 15:52:56 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 26 Jul 2010 13:52:56 -0700
Subject: [llvm-commits] [PATCH 2/5] Record a symbol's size which is
	needed for ELF symbol tables.
In-Reply-To: <75e6e9d13ae15c9baf796f3d6fbbe504a2197172.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
	<75e6e9d13ae15c9baf796f3d6fbbe504a2197172.1280173465.git.matt@console-pimps.org>
Message-ID: <AANLkTi=Ng8z2PhgWokdr6gimACOp_KcQHa8emF1QTAru@mail.gmail.com>

On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
> ---
> ?include/llvm/MC/MCAssembler.h | ? 22 ++++++++++++++++++++++
> ?lib/MC/MCAssembler.cpp ? ? ? ?| ? ?3 ++-
> ?2 files changed, 24 insertions(+), 1 deletions(-)
>
> diff --git a/include/llvm/MC/MCAssembler.h b/include/llvm/MC/MCAssembler.h
> index 51f265d..c189a1d 100644
> --- a/include/llvm/MC/MCAssembler.h
> +++ b/include/llvm/MC/MCAssembler.h
> @@ -24,6 +24,7 @@ namespace llvm {
> ?class raw_ostream;
> ?class MCAsmLayout;
> ?class MCAssembler;
> +class MCBinaryExpr;
> ?class MCContext;
> ?class MCCodeEmitter;
> ?class MCExpr;
> @@ -460,6 +461,10 @@ public:
> ? // common symbol can never get a definition.
> ? uint64_t CommonSize;
>
> + ?uint64_t SymbolSize;
> +
> + ?const MCBinaryExpr *SizeSymbol;
> +
> ? /// CommonAlign - The alignment of the symbol, if it is 'common'.
> ? //
> ? // FIXME: Pack this in with other fields?
> @@ -517,6 +522,23 @@ public:
> ? ? return CommonSize;
> ? }
>
> + ?void setSize(uint64_t _SymbolSize) {
> + ? ?SymbolSize = _SymbolSize;
> + ?}
> +
> + ?uint64_t getSize() {
> + ? ?return SymbolSize;
> + ?}
> +
> + ?void setSizeSymbol(const MCBinaryExpr *SS) {
> + ? ?SizeSymbol = SS;
> + ?}
> +
> + ?const MCBinaryExpr *getSizeSymbol() {
> + ? ?return SizeSymbol;
> + ?}

Why not just "const MCExpr *getSize()"?

On a side note, I don't really follow the distinction between MCSymbol
and MCSymbolData...

-Eli



From bruno.cardoso at gmail.com  Mon Jul 26 16:01:18 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Mon, 26 Jul 2010 21:01:18 -0000
Subject: [llvm-commits] [llvm] r109434 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
Message-ID: <20100726210118.BFB9B2A6C12C@llvm.org>

Author: bruno
Date: Mon Jul 26 16:01:18 2010
New Revision: 109434

URL: http://llvm.org/viewvc/llvm-project?rev=109434&view=rev
Log:
Temporary hack to let codegen assert or generate poor code in case
we are using AVX and no AVX version of the desired intruction is present,
this is better for incremental dev (without fallbacks it's easier to spot
what's missing). Not sure this is the best hack thought (we can also disable
all HasSSE* predicates by dinamically marking them 'false' if AVX is present)


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

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=109434&r1=109433&r2=109434&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Jul 26 16:01:18 2010
@@ -348,14 +348,19 @@
 // X86 Instruction Predicate Definitions.
 def HasCMov      : Predicate<"Subtarget->hasCMov()">;
 def NoCMov       : Predicate<"!Subtarget->hasCMov()">;
-def HasMMX       : Predicate<"Subtarget->hasMMX()">;
-def HasSSE1      : Predicate<"Subtarget->hasSSE1()">;
-def HasSSE2      : Predicate<"Subtarget->hasSSE2()">;
-def HasSSE3      : Predicate<"Subtarget->hasSSE3()">;
-def HasSSSE3     : Predicate<"Subtarget->hasSSSE3()">;
-def HasSSE41     : Predicate<"Subtarget->hasSSE41()">;
-def HasSSE42     : Predicate<"Subtarget->hasSSE42()">;
-def HasSSE4A     : Predicate<"Subtarget->hasSSE4A()">;
+
+// FIXME: temporary hack to let codegen assert or generate poor code in case
+// no AVX version of the desired intructions is present, this is better for
+// incremental dev (without fallbacks it's easier to spot what's missing)
+def HasMMX       : Predicate<"Subtarget->hasMMX() & !Subtarget->hasAVX()">;
+def HasSSE1      : Predicate<"Subtarget->hasSSE1() & !Subtarget->hasAVX()">;
+def HasSSE2      : Predicate<"Subtarget->hasSSE2() & !Subtarget->hasAVX()">;
+def HasSSE3      : Predicate<"Subtarget->hasSSE3() & !Subtarget->hasAVX()">;
+def HasSSSE3     : Predicate<"Subtarget->hasSSSE3() & !Subtarget->hasAVX()">;
+def HasSSE41     : Predicate<"Subtarget->hasSSE41() & !Subtarget->hasAVX()">;
+def HasSSE42     : Predicate<"Subtarget->hasSSE42() & !Subtarget->hasAVX()">;
+def HasSSE4A     : Predicate<"Subtarget->hasSSE4A() & !Subtarget->hasAVX()">;
+
 def HasAVX       : Predicate<"Subtarget->hasAVX()">;
 def HasCLMUL     : Predicate<"Subtarget->hasCLMUL()">;
 def HasFMA3      : Predicate<"Subtarget->hasFMA3()">;




From gohman at apple.com  Mon Jul 26 16:05:37 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 21:05:37 -0000
Subject: [llvm-commits] [llvm] r109435 -
	/llvm/trunk/include/llvm/Support/Registry.h
Message-ID: <20100726210537.602E12A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 16:05:37 2010
New Revision: 109435

URL: http://llvm.org/viewvc/llvm-project?rev=109435&view=rev
Log:
Add a comment.

Modified:
    llvm/trunk/include/llvm/Support/Registry.h

Modified: llvm/trunk/include/llvm/Support/Registry.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/Registry.h?rev=109435&r1=109434&r2=109435&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/Registry.h (original)
+++ llvm/trunk/include/llvm/Support/Registry.h Mon Jul 26 16:05:37 2010
@@ -203,6 +203,8 @@
 
   };
 
+  // Since these are defined in a header file, plugins must be sure to export
+  // these symbols.
 
   template <typename T, typename U>
   typename Registry<T,U>::node *Registry<T,U>::Head;




From clattner at apple.com  Mon Jul 26 16:31:38 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 26 Jul 2010 14:31:38 -0700
Subject: [llvm-commits] [PATCH 5/5] Hookup the ELF support for X86.
In-Reply-To: <5E482B11-F629-405E-9317-226EB3B59515@gmail.com>
References: <cover.1280173465.git.matt@console-pimps.org>
	<a6d3294a6920a32e101b2694c7a291b2c6233af4.1280173465.git.matt@console-pimps.org>
	<AANLkTi=+_SzUWf+-V1kND+BO5BpRrGMLGM9UVMt9fvFr@mail.gmail.com>
	<5E482B11-F629-405E-9317-226EB3B59515@gmail.com>
Message-ID: <2CE45E08-BFE7-483E-9325-306EC766F665@apple.com>


On Jul 26, 2010, at 1:37 PM, Benjamin Kramer wrote:

>> + case Triple::DragonFly:
>> + case Triple::NetBSD:
>> + case Triple::OpenBSD:
>> + case Triple::FreeBSD:
>> + case Triple::Solaris:
>> + case Triple::Haiku:
> 
> Can't we just make ELF the default? There are way more ELF targets than MachO targets and we default to ELF in other places.

Yes, please do.

-Chris


From eli.friedman at gmail.com  Mon Jul 26 16:36:35 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 26 Jul 2010 14:36:35 -0700
Subject: [llvm-commits] [PATCH 4/5] Add ELF ObjectWriter and Streamer
	support.
In-Reply-To: <cac33388f858fe37905e5eb017a333827b3ae1ba.1280173465.git.matt@console-pimps.org>
References: <cover.1280173465.git.matt@console-pimps.org>
	<cac33388f858fe37905e5eb017a333827b3ae1ba.1280173465.git.matt@console-pimps.org>
Message-ID: <AANLkTikpCshvf6iioHhQcZtRfsno7+GKPxzFuYjc8OtC@mail.gmail.com>

On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
+// Because all the symbol flags need to be stored in the MCSymbolData
+// 'flags' variable we need to provide shift constants per flag type.
+enum {
+  STT_SHIFT = 0, // Shift value for STT_* flags.
+  STB_SHIFT = 4, // Shift value for STB_* flags.
+  STV_SHIFT = 8  // Shift value ofr STV_* flags.
+};

I don't think this belongs in ELF.h; it seems specific to the MC implementation.

> +  static bool isFixupKindPCRel(unsigned Kind) {
> +    switch (Kind) {
> +    default:
> +      return false;
> +    case X86::reloc_pcrel_1byte:
> +    case X86::reloc_pcrel_4byte:
> +    case X86::reloc_riprel_4byte:
> +    case X86::reloc_riprel_4byte_movq_load:
> +      return true;
> +    }
> +  }

Change the name to indicate that this x86-specific?

> +  static bool isFixupKindRIPRel(unsigned Kind) {
> +    return Kind == X86::reloc_riprel_4byte ||
> +      Kind == X86::reloc_riprel_4byte_movq_load;
> +  }

Same.

> +  // Emit the ELF header.
> +  void WriteHeader(uint64_t SectionDataSize, unsigned NumberOfSections) {

(General comment) Could you move the large method implementations out of the
class definition?

> +    Write8(ELF::EV_CURRENT);        // e_ident[EI_VERSION]
> +    Write8(ELF::ELFOSABI_LINUX);    // e_ident[EI_OSABI]
> +    Write8(0);                 // e_ident[EI_ABIVERSION]

Factor this out so that it's obvious something needs to change here to make
it crossplatform.

> +    // FIXME
> +    Write16(ELF::EM_X86_64); // e_machine = target

Same.

> +    Write32(ELF::EV_CURRENT);         // e_version
> +    WriteWord(0);                    // e_entry, no entry point in .o file
> +    WriteWord(0);                    // e_phoff, no program header for .o
> +    WriteWord(SectionDataSize + 64);  // e_shoff = sec hdr table off in bytes

Odd-looking indentation.

> +    // FIXME
> +    Write32(0);   // e_flags = whatever the target wants

Factor out.

> +    Write16(Is64Bit ? 64 : 52);  // e_ehsize = ELF header size

sizeof(Elf32_Ehdr)?  Or at least name these values somehow...

> +    // e_shentsize = Section header entry size
> +    Write16(Is64Bit ? 64 : 40);

See above.

> +    // e_shnum     = # of section header ents
> +    Write16(NumberOfSections);
> +
> +    // e_shstrndx  = Section # of '.shstrtab'
> +    Write16(ShstrtabIndex);
> +  }
> +
> +  void WriteSymbolEntry(MCDataFragment *F, uint64_t name, uint8_t info,
> +                        uint64_t value, uint64_t size,
> +                        uint8_t other, uint16_t shndx) {
> +    if (Is64Bit) {
> +      F->getContents() += StringRef((const char *)&name, 4);  // st_name
> +      F->getContents() += StringRef((const char *)&info, 1);  // st_info
> +      F->getContents() += StringRef((const char *)&other, 1); // st_other
> +      F->getContents() += StringRef((const char *)&shndx, 2); // st_shndx
> +      F->getContents() += StringRef((const char *)&value, 8); // st_value
> +      // FIXME
> +      F->getContents() += StringRef((const char *)&size, 8);  // st_size

FIXME what?  Same for all the unlableled FIXMEs.

> +      Value = Layout.getSymbolAddress(&Data);
> +      MCValue Res;
> +      if (Data.getSizeSymbol()->EvaluateAsRelocatable(Res, &Layout)) {
> +        MCSymbolData &A =
> +            Layout.getAssembler().getSymbolData(Res.getSymA()->getSymbol());
> +        MCSymbolData &B =
> +            Layout.getAssembler().getSymbolData(Res.getSymB()->getSymbol());
> +
> +       Size = Layout.getSymbolAddress(&A) - Layout.getSymbolAddress(&B);

How is this different from EvaluateAsAbsolute?

> +  // XXX: this is currently X86_64 only

Don't use XXX; use FIXME instead.

> +      if (Symbol.isUndefined()) {
> +        MSD.SectionIndex = ELF::SHN_UNDEF;
> +       // XXX: for some reason we dont Emit* this
> +       it->setFlags(it->getFlags() | (ELF::STB_GLOBAL << ELF::STB_SHIFT));
> +        UndefinedSymbolData.push_back(MSD);

Indentation?

> +
> +  void WriteRelocations(MCAssembler &Asm, MCAsmLayout &Layout) {
> +    for (MCAssembler::const_iterator it = Asm.begin(),
> +           ie = Asm.end(); it != ie; ++it) {
> +      const MCSectionData &SD = *it;
> +
> +      WriteRelocation(Asm, Layout, SD);

WriteRelocation(Asm, Layout, *it);

> +  void WriteSecHdrEntry(uint32_t Name, uint32_t Type, uint64_t Flags,
> +                       uint64_t Address, uint64_t Offset,
> +                       uint64_t Size, uint32_t Link, uint32_t Info,
> +                       uint64_t Alignment, uint64_t EntrySize) {
> +    // Suppose that all non-initialized EntrySizes are actually a zero
> +    if (EntrySize == ~0U)
> +      EntrySize = 0;

Messy... would it be possible to just set it correctly in the first place?

> +         F->getContents() +=  StringRef((const char *)&entry.r_offset, 8);
> +         F->getContents() +=  StringRef((const char *)&entry.r_info, 8);

Is this endian-safe?

> +    // Write out the ELF header ...
> +    WriteHeader(SectionDataSize, NumSections);
> +    FileOff = Is64Bit ? 64 : 52;

Again, names...

> +    // ... and then the section header table.
> +    // Should we align the section header table?
> +    //
> +    // Null secton first.

*section

> +      switch(Section.getType()) {
> +      case ELF::SHT_DYNAMIC:
> +        sh_link = SectionStringTableIndex[&it->getSection()];
> +        sh_info = 0;
> +        break;
> +
> +      case ELF::SHT_REL:
> +      case ELF::SHT_RELA:
> +        const MCSection *SymtabSection;
> +        const MCSection *InfoSection;
> +        const StringRef *SectionName;
> +        const MCSectionData *SymtabSD;
> +        const MCSectionData *InfoSD;
> +
> +        SymtabSection = Asm.getContext().getELFSection(".symtab", ELF::SHT_SYMTAB, 0,
> +                        SectionKind::getReadOnly(), false);
> +        SymtabSD = &Asm.getSectionData(*SymtabSection);
> +        // we have to count the empty section in too
> +        sh_link = SymtabSD->getLayoutOrder() + 1;
> +
> +        SectionName = &Section.getSectionName();
> +        SectionName = &SectionName->slice(5, SectionName->size());
> +        InfoSection = Asm.getContext().getELFSection(*SectionName,
> +          ELF::SHT_PROGBITS, 0, SectionKind::getReadOnly(), false);
> +        InfoSD = &Asm.getSectionData(*InfoSection);
> +        sh_info = InfoSD->getLayoutOrder() + 1;
> +        break;
> +
> +      case ELF::SHT_SYMTAB:
> +      case ELF::SHT_DYNSYM:
> +        sh_link = StringTableIndex;
> +        sh_info = LastLocalSymbolIndex;
> +        break;
> +
> +      case ELF::SHT_HASH:
> +      case ELF::SHT_GROUP:
> +      case ELF::SHT_SYMTAB_SHNDX:

"default:" ?

> +        assert(0 && "FIXME: sh_type value not supported!");
> +        break;
> +      }
> +
> +      WriteSecHdrEntry(SectionStringTableIndex[&it->getSection()],
> +                      Section.getType(), Section.getFlags(),
> +                      Layout.getSectionAddress(&SD),
> +                      SectionOffsetMap.lookup(&SD.getSection()),
> +                      Layout.getSectionSize(&SD), sh_link,
> +                       sh_info, SD.getAlignment(),
> +                       SD.getEntrySize());

Indentation.

+namespace {
+
+class MCELFStreamer : public MCStreamer {
+private:

Should inherit from MCObjectStreamer.

> +  MCFragment *getCurrentFragment() const {
> +    assert(CurSectionData && "No current section!");
> +
> +    if (!CurSectionData->empty())
> +      return &CurSectionData->getFragmentList().back();
> +
> +    return 0;
> +  }

Shadows MCObjectStreamer::getCurrentFragment().

> +  /// Get a data fragment to write into, creating a new one if the current
> +  /// fragment is not a data fragment.
> +  MCDataFragment *getOrCreateDataFragment() const {
> +    MCDataFragment *F = dyn_cast_or_null<MCDataFragment>(getCurrentFragment());
> +    if (!F)
> +      F = new MCDataFragment(CurSectionData);
> +    return F;
> +  }

Shadows MCObjectStreamer::getOrCreateDataFragment().

> +  MCAssembler &getAssembler() { return Assembler; }

Shadows MCObjectStreamer::getAssembler().

> +  const MCExpr *AddValueSymbols(const MCExpr *Value) {
> +    switch (Value->getKind()) {
> +    case MCExpr::Target: assert(0 && "Can't handle target exprs yet!");
> +    case MCExpr::Constant:
> +      break;
> +
> +    case MCExpr::Binary: {
> +      const MCBinaryExpr *BE = cast<MCBinaryExpr>(Value);
> +      AddValueSymbols(BE->getLHS());
> +      AddValueSymbols(BE->getRHS());
> +      break;
> +    }
> +
> +    case MCExpr::SymbolRef:
> +      Assembler.getOrCreateSymbolData(
> +        cast<MCSymbolRefExpr>(Value)->getSymbol());
> +      break;
> +
> +    case MCExpr::Unary:
> +      AddValueSymbols(cast<MCUnaryExpr>(Value)->getSubExpr());
> +      break;
> +    }
> +
> +    return Value;
> +  }

Shadows MCObjectStreamer::AddValueSymbols.

> +
> +  virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) {

Push definition outside of class defn.

> +  virtual void EmitDwarfFileDirective(unsigned FileNo, StringRef Filename) {
> +    errs() << "FIXME: MCELFStreamer:EmitDwarfFileDirective not implemented\n";
> +  }

DEBUG(dbgs() << "")?

> +  virtual void EmitInstruction(const MCInst &Inst);
> +  virtual void Finish();
> +
> +  /// @}
> +};
> +
> +} // end anonymous namespace.
> +
> +void MCELFStreamer::SwitchSection(const MCSection *Section) {
> +  assert(Section && "Cannot switch to a null section!");
> +
> +  // If already in this section, then this is a noop.
> +  if (Section == CurSection) return;
> +
> +  CurSection = Section;
> +  CurSectionData = &Assembler.getOrCreateSectionData(*Section);
> +}

Identical to MCObjectStreamer::SwitchSection.

> +void MCELFStreamer::EmitLabel(MCSymbol *Symbol) {
> +  assert(Symbol->isUndefined() && "Cannot define a symbol twice!");
> +
> +  // FIXME: This is wasteful, we don't necessarily need to create a data
> +  // fragment. Instead, we should mark the symbol as pointing into the data
> +  // fragment if it exists, otherwise we should just queue the label and set its
> +  // fragment pointer when we emit the next fragment.
> +  MCDataFragment *F = getOrCreateDataFragment();
> +  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
> +  assert(!SD.getFragment() && "Unexpected fragment on symbol data!");
> +  SD.setFragment(F);
> +  SD.setOffset(F->getContents().size());
> +
> +  Symbol->setSection(*CurSection);
> +}
> +
> +void MCELFStreamer::EmitAssemblerFlag(MCAssemblerFlag Flag) {
> +  switch (Flag) {
> +  case MCAF_SubsectionsViaSymbols:
> +    Assembler.setSubsectionsViaSymbols(true);
> +    return;
> +  }
> +
> +  assert(0 && "invalid assembler flag!");
> +}
> +
> +void MCELFStreamer::EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) {
> +  // Only absolute symbols can be redefined.
> +  assert((Symbol->isUndefined() || Symbol->isAbsolute()) &&
> +         "Cannot define a symbol twice!");
> +
> +  // FIXME: Lift context changes into super class.
> +  // FIXME: Set associated section.
> +  //  Symbol->setValue(AddValueSymbols(Value));
> +}

Should this be different from MCMachOStreamer::EmitAssignment?

> +void MCELFStreamer::EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) {
> +}

assert(0)?  Or what is this supposed to be useful for?

> +void MCELFStreamer::EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size,
> +                                       unsigned ByteAlignment) {
> +  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
> +
> +  if ((SD.getFlags() & (0xff << ELF::STB_SHIFT)) == ELF::STB_LOCAL << ELF::STB_SHIFT) {

STB_* is only 4 bits, so 0xf, no?

> +void MCELFStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol,
> +                                   unsigned Size, unsigned ByteAlignment) {

AFAIK, .zerofill is Darwin-only.

> +void MCELFStreamer::EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol,
> +                                   uint64_t Size, unsigned ByteAlignment) {

Also Darwin only, AFAIK.

> +void MCELFStreamer::EmitBytes(StringRef Data, unsigned AddrSpace) {
> +  getOrCreateDataFragment()->getContents().append(Data.begin(), Data.end());
> +}

Same as MCMachOStreamer::EmitBytes; toss in a TODO?

> +void MCELFStreamer::EmitValue(const MCExpr *Value, unsigned Size,
> +                                unsigned AddrSpace) {
> +  MCDataFragment *DF = getOrCreateDataFragment();
> +
> +  // Avoid fixups when possible.
> +  int64_t AbsValue;
> +  if (AddValueSymbols(Value)->EvaluateAsAbsolute(AbsValue)) {
> +    // FIXME: Endianness assumption.
> +    for (unsigned i = 0; i != Size; ++i)
> +      DF->getContents().push_back(uint8_t(AbsValue >> (i * 8)));
> +  } else {
> +    DF->addFixup(MCFixup::Create(DF->getContents().size(), AddValueSymbols(Value),
> +                                 MCFixup::getKindForSize(Size)));
> +    DF->getContents().resize(DF->getContents().size() + Size, 0);
> +  }
> +}

Same as MCMachOStreamer::EmitValue; toss in a TODO?

> +void MCELFStreamer::EmitValueToAlignment(unsigned ByteAlignment,
> +                                           int64_t Value, unsigned ValueSize,
> +                                           unsigned MaxBytesToEmit) {
> +  if (MaxBytesToEmit == 0)
> +    MaxBytesToEmit = ByteAlignment;
> +  new MCAlignFragment(ByteAlignment, Value, ValueSize, MaxBytesToEmit,
> +                      CurSectionData);
> +
> +  // Update the maximum alignment on the current section if necessary.
> +  if (ByteAlignment > CurSectionData->getAlignment())
> +    CurSectionData->setAlignment(ByteAlignment);
> +}

Same as MCMachOStreamer::EmitValueToAlignment; toss in a TODO?

> +void MCELFStreamer::EmitCodeAlignment(unsigned ByteAlignment,
> +                                        unsigned MaxBytesToEmit) {
> +  if (MaxBytesToEmit == 0)
> +    MaxBytesToEmit = ByteAlignment;
> +  new MCAlignFragment(ByteAlignment, 0, 1, MaxBytesToEmit,
> +                      CurSectionData);
> +
> +  // Update the maximum alignment on the current section if necessary.
> +  if (ByteAlignment > CurSectionData->getAlignment())
> +    CurSectionData->setAlignment(ByteAlignment);
> +}

Same as MCMachOStreamer::EmitCodeAlignment; toss in a TODO?

> +void MCELFStreamer::EmitValueToOffset(const MCExpr *Offset,
> +                                        unsigned char Value) {
> +  new MCOrgFragment(*Offset, Value, CurSectionData);
> +}

Same as MCMachOStreamer::EmitValueToOffset; toss in a TODO?

-Eli


From clattner at apple.com  Mon Jul 26 16:36:57 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 26 Jul 2010 14:36:57 -0700
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
	lib/Target/ARM/ARMGlobalMerge.cpp
	test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
In-Reply-To: <20100726184540.1FD682A6C12C@llvm.org>
References: <20100726184540.1FD682A6C12C@llvm.org>
Message-ID: <6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>


On Jul 26, 2010, at 11:45 AM, Anton Korobeynikov wrote:

> Author: asl
> Date: Mon Jul 26 13:45:39 2010
> New Revision: 109423
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109423&view=rev
> Log:
> Currently EH lowering code expects typeinfo to be global only.
> This assumption is not satisfied due to global mergeing.
> Workaround the issue by temporary disablinge mergeing of const globals.
> Also, ignore LLVM "special" globals. This fixes PR7716

Ok, I understand "llvm.foo", but what is ".llvm.foo" ?

-Chris

> 
> Added:
>    llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
> Modified:
>    llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp
> 
> Modified: llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp?rev=109423&r1=109422&r2=109423&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/ARMGlobalMerge.cpp Mon Jul 26 13:45:39 2010
> @@ -178,6 +178,11 @@
>     if (I->getAlignment() != 0)
>       continue;
> 
> +    // Ignore all 'special' globals.
> +    if (I->getName().startswith("llvm.") ||
> +        I->getName().startswith(".llvm."))
> +      continue;
> +
>     if (TD->getTypeAllocSize(I->getType()) < MaxOffset) {
>       if (I->isConstant())
>         ConstGlobals.push_back(I);
> @@ -188,8 +193,12 @@
> 
>   if (Globals.size() > 1)
>     Changed |= doMerge(Globals, M, false);
> -  if (ConstGlobals.size() > 1)
> -    Changed |= doMerge(ConstGlobals, M, true);
> +  // FIXME: This currently breaks the EH processing due to way how the 
> +  // typeinfo detection works. We might want to detect the TIs and ignore 
> +  // them in the future.
> +  
> +  // if (ConstGlobals.size() > 1)
> +  //  Changed |= doMerge(ConstGlobals, M, true);
> 
>   return Changed;
> }
> 
> Added: llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll?rev=109423&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll (added)
> +++ llvm/trunk/test/CodeGen/ARM/2010-07-26-GlobalMerge.ll Mon Jul 26 13:45:39 2010
> @@ -0,0 +1,95 @@
> +; RUN: llc -enable-correct-eh-support < %s
> +; PR7716
> +target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:32-n32"
> +target triple = "thumbv7-apple-darwin10.0.0"
> +
> +%0 = type { i8*, i8* }
> +%struct.A = type { i32 }
> +
> + at d = internal global i32 0, align 4               ; <i32*> [#uses=6]
> + at _ZTVN10__cxxabiv117__class_type_infoE = external global i8* ; <i8**> [#uses=1]
> + at _ZTS1A = internal constant [3 x i8] c"1A\00"     ; <[3 x i8]*> [#uses=1]
> + at _ZTI1A = internal constant %0 { i8* bitcast (i8** getelementptr inbounds (i8** @_ZTVN10__cxxabiv117__class_type_infoE, i32 2) to i8*), i8* getelementptr inbounds ([3 x i8]* @_ZTS1A, i32 0, i32 0) } ; <%0*> [#uses=1]
> + at .str2 = private constant [18 x i8] c"c == %d, d == %d\0A\00" ; <[18 x i8]*> [#uses=1]
> + at .str3 = private constant [16 x i8] c"A(const A&) %d\0A\00" ; <[16 x i8]*> [#uses=1]
> + at .str4 = private constant [9 x i8] c"~A() %d\0A\00" ; <[9 x i8]*> [#uses=1]
> + at .str5 = private constant [8 x i8] c"A() %d\0A\00" ; <[8 x i8]*> [#uses=1]
> + at str = internal constant [14 x i8] c"Throwing 1...\00" ; <[14 x i8]*> [#uses=1]
> + at str1 = internal constant [8 x i8] c"Caught.\00"  ; <[8 x i8]*> [#uses=1]
> +
> +declare i32 @printf(i8* nocapture, ...) nounwind
> +
> +declare i8* @__cxa_allocate_exception(i32)
> +
> +declare i8* @llvm.eh.exception() nounwind readonly
> +
> +declare i32 @__gxx_personality_sj0(...)
> +
> +declare i32 @llvm.eh.selector(i8*, i8*, ...) nounwind
> +
> +declare i32 @llvm.eh.typeid.for(i8*) nounwind
> +
> +declare void @_Unwind_SjLj_Resume(i8*)
> +
> +define internal void @_ZN1AD1Ev(%struct.A* nocapture %this) nounwind ssp align 2 {
> +entry:
> +  %tmp.i = getelementptr inbounds %struct.A* %this, i32 0, i32 0 ; <i32*> [#uses=1]
> +  %tmp2.i = load i32* %tmp.i                      ; <i32> [#uses=1]
> +  %call.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([9 x i8]* @.str4, i32 0, i32 0), i32 %tmp2.i) nounwind ; <i32> [#uses=0]
> +  %tmp3.i = load i32* @d                          ; <i32> [#uses=1]
> +  %inc.i = add nsw i32 %tmp3.i, 1                 ; <i32> [#uses=1]
> +  store i32 %inc.i, i32* @d
> +  ret void
> +}
> +
> +declare void @__cxa_throw(i8*, i8*, i8*)
> +
> +define i32 @main() ssp {
> +entry:
> +  %puts.i = tail call i32 @puts(i8* getelementptr inbounds ([14 x i8]* @str, i32 0, i32 0)) ; <i32> [#uses=0]
> +  %exception.i = tail call i8* @__cxa_allocate_exception(i32 4) nounwind ; <i8*> [#uses=2]
> +  %tmp2.i.i.i = bitcast i8* %exception.i to i32*  ; <i32*> [#uses=1]
> +  store i32 1, i32* %tmp2.i.i.i
> +  %call.i.i.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([8 x i8]* @.str5, i32 0, i32 0), i32 1) nounwind ; <i32> [#uses=0]
> +  invoke void @__cxa_throw(i8* %exception.i, i8* bitcast (%0* @_ZTI1A to i8*), i8* bitcast (void (%struct.A*)* @_ZN1AD1Ev to i8*)) noreturn
> +          to label %.noexc unwind label %lpad
> +
> +.noexc:                                           ; preds = %entry
> +  unreachable
> +
> +try.cont:                                         ; preds = %lpad
> +  %0 = tail call i8* @__cxa_get_exception_ptr(i8* %exn) nounwind ; <i8*> [#uses=0]
> +  %call.i.i = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([16 x i8]* @.str3, i32 0, i32 0), i32 2) nounwind ; <i32> [#uses=0]
> +  %1 = tail call i8* @__cxa_begin_catch(i8* %exn) nounwind ; <i8*> [#uses=0]
> +  %puts = tail call i32 @puts(i8* getelementptr inbounds ([8 x i8]* @str1, i32 0, i32 0)) ; <i32> [#uses=0]
> +  %call.i.i3 = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([9 x i8]* @.str4, i32 0, i32 0), i32 2) nounwind ; <i32> [#uses=0]
> +  %tmp3.i.i = load i32* @d                        ; <i32> [#uses=1]
> +  %inc.i.i4 = add nsw i32 %tmp3.i.i, 1            ; <i32> [#uses=1]
> +  store i32 %inc.i.i4, i32* @d
> +  tail call void @__cxa_end_catch()
> +  %tmp13 = load i32* @d                           ; <i32> [#uses=1]
> +  %call14 = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([18 x i8]* @.str2, i32 0, i32 0), i32 2, i32 %tmp13) ; <i32> [#uses=0]
> +  %tmp16 = load i32* @d                           ; <i32> [#uses=1]
> +  %cmp = icmp ne i32 %tmp16, 2                    ; <i1> [#uses=1]
> +  %conv = zext i1 %cmp to i32                     ; <i32> [#uses=1]
> +  ret i32 %conv
> +
> +lpad:                                             ; preds = %entry
> +  %exn = tail call i8* @llvm.eh.exception() nounwind ; <i8*> [#uses=4]
> +  %eh.selector = tail call i32 (i8*, i8*, ...)* @llvm.eh.selector(i8* %exn, i8* bitcast (i32 (...)* @__gxx_personality_sj0 to i8*), i8* bitcast (%0* @_ZTI1A to i8*), i8* null) nounwind ; <i32> [#uses=1]
> +  %2 = tail call i32 @llvm.eh.typeid.for(i8* bitcast (%0* @_ZTI1A to i8*)) nounwind ; <i32> [#uses=1]
> +  %3 = icmp eq i32 %eh.selector, %2               ; <i1> [#uses=1]
> +  br i1 %3, label %try.cont, label %eh.resume
> +
> +eh.resume:                                        ; preds = %lpad
> +  tail call void @_Unwind_SjLj_Resume(i8* %exn) noreturn
> +  unreachable
> +}
> +
> +declare i8* @__cxa_get_exception_ptr(i8*)
> +
> +declare i8* @__cxa_begin_catch(i8*)
> +
> +declare void @__cxa_end_catch()
> +
> +declare i32 @puts(i8* nocapture) nounwind
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From wendling at apple.com  Mon Jul 26 16:41:23 2010
From: wendling at apple.com (Bill Wendling)
Date: Mon, 26 Jul 2010 14:41:23 -0700
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
	lib/Target/ARM/ARMGlobalMerge.cpp	test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
In-Reply-To: <6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
References: <20100726184540.1FD682A6C12C@llvm.org>
	<6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
Message-ID: <41B34A30-4238-4252-A07F-4095444E1D1F@apple.com>

On Jul 26, 2010, at 2:36 PM, Chris Lattner wrote:

> On Jul 26, 2010, at 11:45 AM, Anton Korobeynikov wrote:
> 
>> Author: asl
>> Date: Mon Jul 26 13:45:39 2010
>> New Revision: 109423
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=109423&view=rev
>> Log:
>> Currently EH lowering code expects typeinfo to be global only.
>> This assumption is not satisfied due to global mergeing.
>> Workaround the issue by temporary disablinge mergeing of const globals.
>> Also, ignore LLVM "special" globals. This fixes PR7716
> 
> Ok, I understand "llvm.foo", but what is ".llvm.foo" ?
> 
I added a ".llvm." variable to my EH horrible hack. It can probably change to llvm... .

-bw





From asl at math.spbu.ru  Mon Jul 26 16:48:36 2010
From: asl at math.spbu.ru (Anton Korobeynikov)
Date: Mon, 26 Jul 2010 21:48:36 -0000
Subject: [llvm-commits] [llvm] r109448 -
	/llvm/trunk/lib/Target/ARM/README.txt
Message-ID: <20100726214836.2B1A52A6C12C@llvm.org>

Author: asl
Date: Mon Jul 26 16:48:35 2010
New Revision: 109448

URL: http://llvm.org/viewvc/llvm-project?rev=109448&view=rev
Log:
Add a note

Modified:
    llvm/trunk/lib/Target/ARM/README.txt

Modified: llvm/trunk/lib/Target/ARM/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/README.txt?rev=109448&r1=109447&r2=109448&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/README.txt (original)
+++ llvm/trunk/lib/Target/ARM/README.txt Mon Jul 26 16:48:35 2010
@@ -636,3 +636,24 @@
 version.
 
 //===---------------------------------------------------------------------===//
+
+Consider the following simple C code:
+
+void foo(unsigned char *a, unsigned char *b, int *c) {
+ if ((*a | *b) == 0) *c = 0;
+}
+
+currently llvm-gcc generates something like this (nice branchless code I'd say):
+
+?? ?? ?? ??ldrb ?? ??r0, [r0]
+ ?? ?? ?? ldrb ?? ??r1, [r1]
+ ?? ?? ?? orr ?? ?? r0, r1, r0
+ ?? ?? ?? tst ?? ?? r0, #255
+ ?? ?? ?? moveq ?? r0, #0
+ ?? ?? ?? streq ?? r0, [r2]
+ ?? ?? ?? bx ?? ?? ??lr
+
+Note that both "tst" and "moveq" are redundant.
+
+//===---------------------------------------------------------------------===//
+




From evan.cheng at apple.com  Mon Jul 26 16:49:07 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 26 Jul 2010 21:49:07 -0000
Subject: [llvm-commits] [llvm] r109449 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
Message-ID: <20100726214907.C13CE2A6C12C@llvm.org>

Author: evancheng
Date: Mon Jul 26 16:49:07 2010
New Revision: 109449

URL: http://llvm.org/viewvc/llvm-project?rev=109449&view=rev
Log:
The "excess register pressure" returned by HighRegPressure() is not accurate enough to factor into scheduling priority. Eliminate it and add early exits to speed up scheduling.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp?rev=109449&r1=109448&r2=109449&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp Mon Jul 26 16:49:07 2010
@@ -1191,13 +1191,10 @@
       SU->NodeQueueId = 0;
     }
 
-    bool HighRegPressure(const SUnit *SU, unsigned &Excess) const {
-      Excess = 0;
-
+    bool HighRegPressure(const SUnit *SU) const {
       if (!TLI)
         return false;
 
-      bool High = false;
       for (SUnit::const_pred_iterator I = SU->Preds.begin(),E = SU->Preds.end();
            I != E; ++I) {
         if (I->isCtrl())
@@ -1209,10 +1206,8 @@
             EVT VT = PN->getValueType(0);
             unsigned RCId = TLI->getRepRegClassFor(VT)->getID();
             unsigned Cost = TLI->getRepRegClassCostFor(VT);
-            if ((RegPressure[RCId] + Cost) >= RegLimit[RCId]) {
-              High = true;
-              Excess += (RegPressure[RCId] + Cost) - RegLimit[RCId];
-            }
+            if ((RegPressure[RCId] + Cost) >= RegLimit[RCId])
+              return true;
           }
           continue;
         }
@@ -1225,10 +1220,8 @@
           unsigned Cost = TLI->getRepRegClassCostFor(VT);
           // Check if this increases register pressure of the specific register
           // class to the point where it would cause spills.
-          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId]) {
-            High = true;
-            Excess += (RegPressure[RCId] + Cost) - RegLimit[RCId];
-          }
+          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId])
+            return true;
           continue;            
         } else if (POpc == TargetOpcode::INSERT_SUBREG ||
                    POpc == TargetOpcode::SUBREG_TO_REG) {
@@ -1237,29 +1230,27 @@
           unsigned Cost = TLI->getRepRegClassCostFor(VT);
           // Check if this increases register pressure of the specific register
           // class to the point where it would cause spills.
-          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId]) {
-            High = true;
-            Excess += (RegPressure[RCId] + Cost) - RegLimit[RCId];
-          }
+          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId])
+            return true;
           continue;
         }
         unsigned NumDefs = TII->get(PN->getMachineOpcode()).getNumDefs();
         for (unsigned i = 0; i != NumDefs; ++i) {
           EVT VT = PN->getValueType(i);
-          if (!PN->hasAnyUseOfValue(i))
-            continue;
           unsigned RCId = TLI->getRepRegClassFor(VT)->getID();
+          if (RegPressure[RCId] >= RegLimit[RCId])
+            return true; // Reg pressure already high.
           unsigned Cost = TLI->getRepRegClassCostFor(VT);
+          if (!PN->hasAnyUseOfValue(i))
+            continue;
           // Check if this increases register pressure of the specific register
           // class to the point where it would cause spills.
-          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId]) {
-            High = true;
-            Excess += (RegPressure[RCId] + Cost) - RegLimit[RCId];
-          }
+          if ((RegPressure[RCId] + Cost) >= RegLimit[RCId])
+            return true;
         }
       }
 
-      return High;
+      return false;
     }
 
     void ScheduledNode(SUnit *SU) {
@@ -1558,21 +1549,15 @@
 }
 
 bool hybrid_ls_rr_sort::operator()(const SUnit *left, const SUnit *right) const{
-  unsigned LExcess, RExcess;
-  bool LHigh = SPQ->HighRegPressure(left, LExcess);
-  bool RHigh = SPQ->HighRegPressure(right, RExcess);
+  bool LHigh = SPQ->HighRegPressure(left);
+  bool RHigh = SPQ->HighRegPressure(right);
   // Avoid causing spills. If register pressure is high, schedule for
   // register pressure reduction.
   if (LHigh && !RHigh)
     return true;
   else if (!LHigh && RHigh)
     return false;
-  else if (LHigh && RHigh) {
-    if (LExcess > RExcess)
-      return true;
-    else if (LExcess < RExcess)
-      return false;
-  } else {
+  else if (!LHigh && !RHigh) {
     // Low register pressure situation, schedule for latency if possible.
     bool LStall = left->SchedulingPref == Sched::Latency &&
       SPQ->getCurCycle() < left->getHeight();
@@ -1606,21 +1591,15 @@
 
 bool ilp_ls_rr_sort::operator()(const SUnit *left,
                                 const SUnit *right) const {
-  unsigned LExcess, RExcess;
-  bool LHigh = SPQ->HighRegPressure(left, LExcess);
-  bool RHigh = SPQ->HighRegPressure(right, RExcess);
+  bool LHigh = SPQ->HighRegPressure(left);
+  bool RHigh = SPQ->HighRegPressure(right);
   // Avoid causing spills. If register pressure is high, schedule for
   // register pressure reduction.
   if (LHigh && !RHigh)
     return true;
   else if (!LHigh && RHigh)
     return false;
-  else if (LHigh && RHigh) {
-    if (LExcess > RExcess)
-      return true;
-    else if (LExcess < RExcess)
-      return false;    
-  } else {
+  else if (!LHigh && !RHigh) {
     // Low register pressure situation, schedule to maximize instruction level
     // parallelism.
     if (left->NumPreds > right->NumPreds)




From asl at math.spbu.ru  Mon Jul 26 16:50:04 2010
From: asl at math.spbu.ru (Anton Korobeynikov)
Date: Tue, 27 Jul 2010 01:50:04 +0400
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
	lib/Target/ARM/ARMGlobalMerge.cpp
	test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
In-Reply-To: <6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
References: <20100726184540.1FD682A6C12C@llvm.org>
	<6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
Message-ID: <AANLkTinufGpBkST0POmaLAc5cmW3uB4JyqPg771QF5xA@mail.gmail.com>

> Ok, I understand "llvm.foo", but what is ".llvm.foo" ?
Ask Bill :)

-- 
With best regards, Anton Korobeynikov
Faculty of Mathematics and Mechanics, Saint Petersburg State University


From evan.cheng at apple.com  Mon Jul 26 16:50:05 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 26 Jul 2010 21:50:05 -0000
Subject: [llvm-commits] [llvm] r109450 - in /llvm/trunk/lib/Target/X86:
 X86ISelLowering.cpp X86ISelLowering.h
Message-ID: <20100726215005.648812A6C12C@llvm.org>

Author: evancheng
Date: Mon Jul 26 16:50:05 2010
New Revision: 109450

URL: http://llvm.org/viewvc/llvm-project?rev=109450&view=rev
Log:
On x86, f32 / f64 nodes share the same registers as 128-bit vector values.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.h

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109450&r1=109449&r2=109450&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Mon Jul 26 16:50:05 2010
@@ -1190,6 +1190,32 @@
   return F->hasFnAttr(Attribute::OptimizeForSize) ? 0 : 4;
 }
 
+std::pair<const TargetRegisterClass*, uint8_t>
+X86TargetLowering::findRepresentativeClass(EVT VT) const{
+  const TargetRegisterClass *RRC = 0;
+  uint8_t Cost = 1;
+  switch (VT.getSimpleVT().SimpleTy) {
+  default:
+    return TargetLowering::findRepresentativeClass(VT);
+  case MVT::i8: case MVT::i16: case MVT::i32: case MVT::i64:
+    RRC = (Subtarget->is64Bit()
+           ? X86::GR64RegisterClass : X86::GR32RegisterClass);
+    break;
+  case MVT::v8i8: case MVT::v4i16:
+  case MVT::v2i32: case MVT::v1i64: 
+    RRC = X86::VR64RegisterClass;
+    break;
+  case MVT::f32: case MVT::f64:
+  case MVT::v16i8: case MVT::v8i16: case MVT::v4i32: case MVT::v2i64:
+  case MVT::v4f32: case MVT::v2f64:
+  case MVT::v32i8: case MVT::v8i32: case MVT::v4i64: case MVT::v8f32:
+  case MVT::v4f64:
+    RRC = X86::VR128RegisterClass;
+    break;
+  }
+  return std::make_pair(RRC, Cost);
+}
+
 unsigned
 X86TargetLowering::getRegPressureLimit(const TargetRegisterClass *RC,
                                        MachineFunction &MF) const {

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.h?rev=109450&r1=109449&r2=109450&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.h (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.h Mon Jul 26 16:50:05 2010
@@ -599,6 +599,10 @@
     /// appropriate.
     virtual bool getStackCookieLocation(unsigned &AddressSpace, unsigned &Offset) const;
 
+  protected:
+    std::pair<const TargetRegisterClass*, uint8_t>
+    findRepresentativeClass(EVT VT) const;
+
   private:
     /// Subtarget - Keep a pointer to the X86Subtarget around so that we can
     /// make the right decision when generating code for different targets.




From clattner at apple.com  Mon Jul 26 16:55:24 2010
From: clattner at apple.com (Chris Lattner)
Date: Mon, 26 Jul 2010 14:55:24 -0700
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
	lib/Target/ARM/ARMGlobalMerge.cpp	test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
In-Reply-To: <41B34A30-4238-4252-A07F-4095444E1D1F@apple.com>
References: <20100726184540.1FD682A6C12C@llvm.org>
	<6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
	<41B34A30-4238-4252-A07F-4095444E1D1F@apple.com>
Message-ID: <D59DE412-0B37-4235-9A7A-F97573407036@apple.com>


On Jul 26, 2010, at 2:41 PM, Bill Wendling wrote:

> On Jul 26, 2010, at 2:36 PM, Chris Lattner wrote:
> 
>> On Jul 26, 2010, at 11:45 AM, Anton Korobeynikov wrote:
>> 
>>> Author: asl
>>> Date: Mon Jul 26 13:45:39 2010
>>> New Revision: 109423
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=109423&view=rev
>>> Log:
>>> Currently EH lowering code expects typeinfo to be global only.
>>> This assumption is not satisfied due to global mergeing.
>>> Workaround the issue by temporary disablinge mergeing of const globals.
>>> Also, ignore LLVM "special" globals. This fixes PR7716
>> 
>> Ok, I understand "llvm.foo", but what is ".llvm.foo" ?
>> 
> I added a ".llvm." variable to my EH horrible hack. It can probably change to llvm... .

yes, please do.  "llvm." is the reserved namespace for system stuff.

-Chris


From bob.wilson at apple.com  Mon Jul 26 16:58:00 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Mon, 26 Jul 2010 21:58:00 -0000
Subject: [llvm-commits] [llvm] r109451 -
	/llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp
Message-ID: <20100726215800.3B3802A6C12C@llvm.org>

Author: bwilson
Date: Mon Jul 26 16:58:00 2010
New Revision: 109451

URL: http://llvm.org/viewvc/llvm-project?rev=109451&view=rev
Log:
Don't call __register_frame from the JIT on systems that use setjmp/longjmp
exception handling.  Also fix an extra underscore typo in one instance of
"__ARM_EABI__".  Radar 8236264.

Modified:
    llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp?rev=109451&r1=109450&r2=109451&view=diff
==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp Mon Jul 26 16:58:00 2010
@@ -67,7 +67,7 @@
 }
 
 
-#if defined(__GNUC__) && !defined(__ARM__EABI__)
+#if defined(__GNUC__) && !defined(__ARM_EABI__) && !defined(__USING_SJLJ_EXCEPTIONS__)
  
 // libgcc defines the __register_frame function to dynamically register new
 // dwarf frames for exception handling. This functionality is not portable
@@ -308,7 +308,7 @@
   }
   
   // Register routine for informing unwinding runtime about new EH frames
-#if defined(__GNUC__) && !defined(__ARM_EABI__)
+#if defined(__GNUC__) && !defined(__ARM_EABI__) && !defined(__USING_SJLJ_EXCEPTIONS__)
 #if USE_KEYMGR
   struct LibgccObjectInfo* LOI = (struct LibgccObjectInfo*)
     _keymgr_get_and_lock_processwide_ptr(KEYMGR_GCC3_DW2_OBJ_LIST);




From criswell at uiuc.edu  Mon Jul 26 17:05:35 2010
From: criswell at uiuc.edu (John Criswell)
Date: Mon, 26 Jul 2010 22:05:35 -0000
Subject: [llvm-commits] [poolalloc] r109453 -
	/poolalloc/trunk/lib/DSA/SanityCheck.cpp
Message-ID: <20100726220535.16F322A6C12C@llvm.org>

Author: criswell
Date: Mon Jul 26 17:05:34 2010
New Revision: 109453

URL: http://llvm.org/viewvc/llvm-project?rev=109453&view=rev
Log:
Pass that does some sanity checking on DSA results.

Added:
    poolalloc/trunk/lib/DSA/SanityCheck.cpp

Added: poolalloc/trunk/lib/DSA/SanityCheck.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/SanityCheck.cpp?rev=109453&view=auto
==============================================================================
--- poolalloc/trunk/lib/DSA/SanityCheck.cpp (added)
+++ poolalloc/trunk/lib/DSA/SanityCheck.cpp Mon Jul 26 17:05:34 2010
@@ -0,0 +1,136 @@
+//===- SanityCheck.cpp - Check sanity and consistency of DSA results ------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file was developed by the LLVM research group and is distributed under
+// the University of Illinois Open Source License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Perform a sanity check on the DSA results.
+//
+//===----------------------------------------------------------------------===//
+
+#define DEBUG_TYPE "dsa-check"
+
+#include "dsa/DataStructure.h"
+#include "dsa/DSGraph.h"
+
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+
+using namespace llvm;
+
+//
+// Class: CheckSanity
+//
+// Description:
+//  This is an LLVM pass that will request DSA results and then do some sanity
+//  checks on the DSA results.
+//
+class SanityCheck : public ModulePass {
+  public:
+    static char ID;
+    SanityCheck() : ModulePass ((intptr_t)&ID) {
+      dsaPass = 0;
+    }
+    virtual bool runOnModule(Module &M);
+    virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+      AU.addRequired<EQTDDataStructures>();
+      AU.setPreservesAll();
+    }
+
+  protected:
+    void checkLoad (LoadInst & LI);
+
+    // The DSA pass which is to be sanitized
+    DataStructures * dsaPass;
+};
+
+// Sanity Check Pass ID
+char SanityCheck::ID = 0;
+
+static RegisterPass<SanityCheck>
+X("dsa-check", "Check consistency of Data Structure Analysis");
+
+//
+// Method: checkLoad()
+//
+// Description:
+//  This method ensures that the DSNode associated with the result of a load
+//  instruction is linked from the DSNode associated with the pointer
+//  dereferenced by the load.
+//
+void
+SanityCheck::checkLoad (LoadInst & LI) {
+  //
+  // Get the DSA Graph for the function in which the load instruction lives.
+  //
+  Function * F = LI.getParent()->getParent();
+  DSGraph * G = dsaPass->getDSGraph (*F);
+
+  //
+  // Get the DSNode information for both the result of the load instruction
+  // and its pointer operand.  If the result of the load has no DSNode, then
+  // nothing needs to be done.
+  //
+  DSNode* ResultNode = G->getNodeForValue (&LI).getNode();
+  DSNode* PtrNode    = G->getNodeForValue (LI.getPointerOperand()).getNode();
+  if (!ResultNode) {
+    return;
+  }
+  assert (PtrNode && "Load operand has no DSNode!\n");
+
+  //
+  // Scan through the links of the DSNode of the load's pointer operand; we
+  // need to determine the offset into the memory object from which the result
+  // of the load is loaded.
+  //
+  DSNode::LinkMapTy::iterator linki = PtrNode->edge_begin();
+  bool found = false;
+  for (; linki != PtrNode->edge_end(); ++linki) {
+    DSNodeHandle & LinkNode = linki->second;
+    if (LinkNode.getNode() == ResultNode) {
+      found = true;
+    }
+  }
+
+  assert (found && "Can't find link for load result!\n");
+  return;
+}
+
+//
+// Method: runOnModule()
+//
+// Description:
+//  This method is the entry point for the sanity checking pass.
+//
+// Inputs:
+//  M - The LLVM module which should be analyzed for sanity and consistency.
+//
+// Return value:
+//  false - The Module has not been modified (this pass should never modify the
+//          Module).
+//
+bool
+SanityCheck::runOnModule(Module & M) {
+  //
+  // Get a reference to the DSA pass to check.
+  //
+  dsaPass = &getAnalysis<EQTDDataStructures>();
+
+  //
+  // Scan through all instructions in the module and sanity check the load
+  // instructions.
+  //
+  for (Module::iterator F = M.begin(); F != M.end(); ++F) {
+    for (Function::iterator BB = F->begin(); BB != F->end(); ++BB) {
+      for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) {
+        if (LoadInst * LI = dyn_cast<LoadInst>(I))
+          checkLoad (*LI);
+      }
+    }
+  }
+
+  return false;
+}




From echristo at apple.com  Mon Jul 26 17:06:04 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 15:06:04 -0700
Subject: [llvm-commits] [llvm] r109449 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
In-Reply-To: <20100726214907.C13CE2A6C12C@llvm.org>
References: <20100726214907.C13CE2A6C12C@llvm.org>
Message-ID: <211D8908-7FEB-4241-8061-865E7CB7D2A1@apple.com>


On Jul 26, 2010, at 2:49 PM, Evan Cheng wrote:

> The "excess register pressure" returned by HighRegPressure() is not accurate enough to factor into scheduling priority. Eliminate it and add early exits to speed up scheduling.

Excess was "number of registers used over the number in the class"?  If so it would probably only be useful when trying to schedule for conservation of registers, or to get registers back to where we can start scheduling for ILP again.

Why wasn't it accurate?  Cost problems or...?

-eric


From asl at math.spbu.ru  Mon Jul 26 17:23:07 2010
From: asl at math.spbu.ru (Anton Korobeynikov)
Date: Mon, 26 Jul 2010 22:23:07 -0000
Subject: [llvm-commits] [llvm] r109456 -
	/llvm/trunk/lib/Target/ARM/README.txt
Message-ID: <20100726222308.091762A6C12C@llvm.org>

Author: asl
Date: Mon Jul 26 17:23:07 2010
New Revision: 109456

URL: http://llvm.org/viewvc/llvm-project?rev=109456&view=rev
Log:
Drop some non-ascii stuff

Modified:
    llvm/trunk/lib/Target/ARM/README.txt

Modified: llvm/trunk/lib/Target/ARM/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/README.txt?rev=109456&r1=109455&r2=109456&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/README.txt (original)
+++ llvm/trunk/lib/Target/ARM/README.txt Mon Jul 26 17:23:07 2010
@@ -645,13 +645,13 @@
 
 currently llvm-gcc generates something like this (nice branchless code I'd say):
 
-?? ?? ?? ??ldrb ?? ??r0, [r0]
- ?? ?? ?? ldrb ?? ??r1, [r1]
- ?? ?? ?? orr ?? ?? r0, r1, r0
- ?? ?? ?? tst ?? ?? r0, #255
- ?? ?? ?? moveq ?? r0, #0
- ?? ?? ?? streq ?? r0, [r2]
- ?? ?? ?? bx ?? ?? ??lr
+       ldrb    r0, [r0]
+       ldrb    r1, [r1]
+       orr     r0, r1, r0
+       tst     r0, #255
+       moveq   r0, #0
+       streq   r0, [r2]
+ ?? ?? ?? bx      lr
 
 Note that both "tst" and "moveq" are redundant.
 




From eli.friedman at gmail.com  Mon Jul 26 17:28:19 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 26 Jul 2010 22:28:19 -0000
Subject: [llvm-commits] [llvm] r109458 -
	/llvm/trunk/lib/Target/ARM/README.txt
Message-ID: <20100726222819.294032A6C12C@llvm.org>

Author: efriedma
Date: Mon Jul 26 17:28:18 2010
New Revision: 109458

URL: http://llvm.org/viewvc/llvm-project?rev=109458&view=rev
Log:
And a bit more non-ASCII stuff.


Modified:
    llvm/trunk/lib/Target/ARM/README.txt

Modified: llvm/trunk/lib/Target/ARM/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/README.txt?rev=109458&r1=109457&r2=109458&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/README.txt (original)
+++ llvm/trunk/lib/Target/ARM/README.txt Mon Jul 26 17:28:18 2010
@@ -651,7 +651,7 @@
        tst     r0, #255
        moveq   r0, #0
        streq   r0, [r2]
- ?? ?? ?? bx      lr
+       bx      lr
 
 Note that both "tst" and "moveq" are redundant.
 




From evan.cheng at apple.com  Mon Jul 26 17:31:03 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 26 Jul 2010 15:31:03 -0700
Subject: [llvm-commits] [llvm] r109449 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
In-Reply-To: <211D8908-7FEB-4241-8061-865E7CB7D2A1@apple.com>
References: <20100726214907.C13CE2A6C12C@llvm.org>
	<211D8908-7FEB-4241-8061-865E7CB7D2A1@apple.com>
Message-ID: <2FAF248B-8414-4F37-ACEB-C50FC86C4DA4@apple.com>

By definition register pressure tracking in the scheduler is an approximation. We're not modeling live throughs, nor are we taking into consideration special register allocation requirement, etc.

The other major issue is register pressure excess for different register classes are the same. e.g. let's say node A will cause high GPR register pressure, node B VR128. The scheduler doesn't know whether it's better to schedule A or B first.

Evan

On Jul 26, 2010, at 3:06 PM, Eric Christopher wrote:

> 
> On Jul 26, 2010, at 2:49 PM, Evan Cheng wrote:
> 
>> The "excess register pressure" returned by HighRegPressure() is not accurate enough to factor into scheduling priority. Eliminate it and add early exits to speed up scheduling.
> 
> Excess was "number of registers used over the number in the class"?  If so it would probably only be useful when trying to schedule for conservation of registers, or to get registers back to where we can start scheduling for ILP again.
> 
> Why wasn't it accurate?  Cost problems or...?
> 
> -eric




From isanbard at gmail.com  Mon Jul 26 17:36:29 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 26 Jul 2010 22:36:29 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109461 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
Message-ID: <20100726223629.930D72A6C12C@llvm.org>

Author: void
Date: Mon Jul 26 17:36:29 2010
New Revision: 109461

URL: http://llvm.org/viewvc/llvm-project?rev=109461&view=rev
Log:
Using llvm.eh.catch.all.value instead of .llvm.eh.catch.all.value.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp?rev=109461&r1=109460&r2=109461&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp Mon Jul 26 17:36:29 2010
@@ -2192,7 +2192,7 @@
 
             CatchAll = new GlobalVariable(*TheModule, Init->getType(), true,
                                           GlobalVariable::LinkOnceAnyLinkage,
-                                          Init, ".llvm.eh.catch.all.value");
+                                          Init, "llvm.eh.catch.all.value");
             CatchAll->setSection("llvm.metadata");
           }
 
@@ -2235,7 +2235,7 @@
 
             CatchAll = new GlobalVariable(*TheModule, Init->getType(), true,
                                           GlobalVariable::LinkOnceAnyLinkage,
-                                          Init, ".llvm.eh.catch.all.value");
+                                          Init, "llvm.eh.catch.all.value");
             CatchAll->setSection("llvm.metadata");
           }
 




From isanbard at gmail.com  Mon Jul 26 17:36:52 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 26 Jul 2010 22:36:52 -0000
Subject: [llvm-commits] [llvm] r109462 - in /llvm/trunk/lib/CodeGen:
	Analysis.cpp DwarfEHPrepare.cpp
Message-ID: <20100726223652.97DE22A6C12C@llvm.org>

Author: void
Date: Mon Jul 26 17:36:52 2010
New Revision: 109462

URL: http://llvm.org/viewvc/llvm-project?rev=109462&view=rev
Log:
Using llvm.eh.catch.all.value instead of .llvm.eh.catch.all.value.

Modified:
    llvm/trunk/lib/CodeGen/Analysis.cpp
    llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp

Modified: llvm/trunk/lib/CodeGen/Analysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Analysis.cpp?rev=109462&r1=109461&r2=109462&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/Analysis.cpp (original)
+++ llvm/trunk/lib/CodeGen/Analysis.cpp Mon Jul 26 17:36:52 2010
@@ -109,7 +109,7 @@
   V = V->stripPointerCasts();
   GlobalVariable *GV = dyn_cast<GlobalVariable>(V);
 
-  if (GV && GV->getName() == ".llvm.eh.catch.all.value") {
+  if (GV && GV->getName() == "llvm.eh.catch.all.value") {
     assert(GV->hasInitializer() &&
            "The EH catch-all value must have an initializer");
     Value *Init = GV->getInitializer();

Modified: llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp?rev=109462&r1=109461&r2=109462&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp (original)
+++ llvm/trunk/lib/CodeGen/DwarfEHPrepare.cpp Mon Jul 26 17:36:52 2010
@@ -86,7 +86,7 @@
     }
 
     /// CleanupSelectors - Any remaining eh.selector intrinsic calls which still
-    /// use the ".llvm.eh.catch.all.value" call need to convert to using its
+    /// use the "llvm.eh.catch.all.value" call need to convert to using its
     /// initializer instead.
     bool CleanupSelectors(SmallPtrSet<IntrinsicInst*, 32> &Sels);
 
@@ -231,7 +231,7 @@
 }
 
 /// CleanupSelectors - Any remaining eh.selector intrinsic calls which still use
-/// the ".llvm.eh.catch.all.value" call need to convert to using its
+/// the "llvm.eh.catch.all.value" call need to convert to using its
 /// initializer instead.
 bool DwarfEHPrepare::CleanupSelectors(SmallPtrSet<IntrinsicInst*, 32> &Sels) {
   if (!EHCatchAllValue) return false;
@@ -247,7 +247,7 @@
          I = Sels.begin(), E = Sels.end(); I != E; ++I) {
     IntrinsicInst *Sel = *I;
 
-    // Index of the ".llvm.eh.catch.all.value" variable.
+    // Index of the "llvm.eh.catch.all.value" variable.
     unsigned OpIdx = Sel->getNumArgOperands() - 1;
     GlobalVariable *GV = dyn_cast<GlobalVariable>(Sel->getArgOperand(OpIdx));
     if (GV != EHCatchAllValue) continue;
@@ -304,7 +304,7 @@
 bool DwarfEHPrepare::HandleURoRInvokes() {
   if (!EHCatchAllValue) {
     EHCatchAllValue =
-      F->getParent()->getNamedGlobal(".llvm.eh.catch.all.value");
+      F->getParent()->getNamedGlobal("llvm.eh.catch.all.value");
     if (!EHCatchAllValue) return false;
   }
 




From echristo at apple.com  Mon Jul 26 17:37:51 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 26 Jul 2010 15:37:51 -0700
Subject: [llvm-commits] [llvm] r109449 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGRRList.cpp
In-Reply-To: <2FAF248B-8414-4F37-ACEB-C50FC86C4DA4@apple.com>
References: <20100726214907.C13CE2A6C12C@llvm.org>
	<211D8908-7FEB-4241-8061-865E7CB7D2A1@apple.com>
	<2FAF248B-8414-4F37-ACEB-C50FC86C4DA4@apple.com>
Message-ID: <D4C4B95C-1116-4986-A411-7448BC65DC8E@apple.com>


On Jul 26, 2010, at 3:31 PM, Evan Cheng wrote:

> By definition register pressure tracking in the scheduler is an approximation. We're not modeling live throughs, nor are we taking into consideration special register allocation requirement, etc.
> 
> The other major issue is register pressure excess for different register classes are the same. e.g. let's say node A will cause high GPR register pressure, node B VR128. The scheduler doesn't know whether it's better to schedule A or B first.

Aha.  Excellent, thanks.

-eric


From wendling at apple.com  Mon Jul 26 17:37:54 2010
From: wendling at apple.com (Bill Wendling)
Date: Mon, 26 Jul 2010 15:37:54 -0700
Subject: [llvm-commits] [llvm] r109423 - in /llvm/trunk:
	lib/Target/ARM/ARMGlobalMerge.cpp	test/CodeGen/ARM/2010-07-26-GlobalMerge.ll
In-Reply-To: <D59DE412-0B37-4235-9A7A-F97573407036@apple.com>
References: <20100726184540.1FD682A6C12C@llvm.org>
	<6F1B568E-69B5-4262-B1FC-B7666254F8D0@apple.com>
	<41B34A30-4238-4252-A07F-4095444E1D1F@apple.com>
	<D59DE412-0B37-4235-9A7A-F97573407036@apple.com>
Message-ID: <D3B0F7BE-1FFA-4A6F-8E1D-82F479FFEF2B@apple.com>


On Jul 26, 2010, at 2:55 PM, Chris Lattner wrote:

> 
> On Jul 26, 2010, at 2:41 PM, Bill Wendling wrote:
> 
>> On Jul 26, 2010, at 2:36 PM, Chris Lattner wrote:
>> 
>>> On Jul 26, 2010, at 11:45 AM, Anton Korobeynikov wrote:
>>> 
>>>> Author: asl
>>>> Date: Mon Jul 26 13:45:39 2010
>>>> New Revision: 109423
>>>> 
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=109423&view=rev
>>>> Log:
>>>> Currently EH lowering code expects typeinfo to be global only.
>>>> This assumption is not satisfied due to global mergeing.
>>>> Workaround the issue by temporary disablinge mergeing of const globals.
>>>> Also, ignore LLVM "special" globals. This fixes PR7716
>>> 
>>> Ok, I understand "llvm.foo", but what is ".llvm.foo" ?
>>> 
>> I added a ".llvm." variable to my EH horrible hack. It can probably change to llvm... .
> 
> yes, please do.  "llvm." is the reserved namespace for system stuff.
> 
Done.

-bw





From wendling at apple.com  Mon Jul 26 17:38:27 2010
From: wendling at apple.com (Bill Wendling)
Date: Mon, 26 Jul 2010 15:38:27 -0700
Subject: [llvm-commits] [llvm] r109448
	-	/llvm/trunk/lib/Target/ARM/README.txt
In-Reply-To: <20100726214836.2B1A52A6C12C@llvm.org>
References: <20100726214836.2B1A52A6C12C@llvm.org>
Message-ID: <EC7B3CD0-0E3D-456C-8BDE-A627C981D53A@apple.com>

On Jul 26, 2010, at 2:48 PM, Anton Korobeynikov wrote:

> Author: asl
> Date: Mon Jul 26 16:48:35 2010
> New Revision: 109448
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109448&view=rev
> Log:
> Add a note
> 
> Modified:
>    llvm/trunk/lib/Target/ARM/README.txt
> 
> Modified: llvm/trunk/lib/Target/ARM/README.txt
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/README.txt?rev=109448&r1=109447&r2=109448&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/README.txt (original)
> +++ llvm/trunk/lib/Target/ARM/README.txt Mon Jul 26 16:48:35 2010
> @@ -636,3 +636,24 @@
> version.
> 
> //===---------------------------------------------------------------------===//
> +
> +Consider the following simple C code:
> +
> +void foo(unsigned char *a, unsigned char *b, int *c) {
> + if ((*a | *b) == 0) *c = 0;
> +}
> +
> +currently llvm-gcc generates something like this (nice branchless code I'd say):
> +
> +?  ?  ?  ? ldrb ?  ? r0, [r0]

Hi Anton,

Unprintable characters perhaps? :)

-bw

> + ?  ?  ?  ldrb ?  ? r1, [r1]
> + ?  ?  ?  orr ?  ?  r0, r1, r0
> + ?  ?  ?  tst ?  ?  r0, #255
> + ?  ?  ?  moveq ?  r0, #0
> + ?  ?  ?  streq ?  r0, [r2]
> + ?  ?  ?  bx ?  ?  ? lr
> +
> +Note that both "tst" and "moveq" are redundant.
> +
> +//===---------------------------------------------------------------------===//
> +
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From stoklund at 2pi.dk  Mon Jul 26 17:49:28 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 26 Jul 2010 15:49:28 -0700
Subject: [llvm-commits] [llvm] r109456 -
	/llvm/trunk/lib/Target/ARM/README.txt
In-Reply-To: <20100726222308.091762A6C12C@llvm.org>
References: <20100726222308.091762A6C12C@llvm.org>
Message-ID: <97E687C9-5601-4760-84BE-8E16AD9E1E77@2pi.dk>


On Jul 26, 2010, at 3:23 PM, Anton Korobeynikov wrote:

> Author: asl
> Date: Mon Jul 26 17:23:07 2010
> New Revision: 109456
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109456&view=rev
> Log:
> Drop some non-ascii stuff

> + ?  ?  ?  bx      lr

But not all ;-)





From dalej at apple.com  Mon Jul 26 17:53:50 2010
From: dalej at apple.com (Dale Johannesen)
Date: Mon, 26 Jul 2010 22:53:50 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109465 -
	/llvm-gcc-4.2/trunk/gcc/config/darwin-c.c
Message-ID: <20100726225350.89C322A6C12C@llvm.org>

Author: johannes
Date: Mon Jul 26 17:53:50 2010
New Revision: 109465

URL: http://llvm.org/viewvc/llvm-project?rev=109465&view=rev
Log:
Only warn about unknown #pragma optimization_level when
the appropriate flag is given (no warning by default).  8229854.


Modified:
    llvm-gcc-4.2/trunk/gcc/config/darwin-c.c

Modified: llvm-gcc-4.2/trunk/gcc/config/darwin-c.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/config/darwin-c.c?rev=109465&r1=109464&r2=109465&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/config/darwin-c.c (original)
+++ llvm-gcc-4.2/trunk/gcc/config/darwin-c.c Mon Jul 26 17:53:50 2010
@@ -416,14 +416,16 @@
 void
 darwin_pragma_opt_level  (cpp_reader *pfile ATTRIBUTE_UNUSED)
 {
-  warning (0, "ignoring '#pragma optimization_level'");  
+  if (warn_unknown_pragmas > in_system_header)
+    warning (OPT_Wunknown_pragmas, "ignoring #pragma optimization_level");  
   return;
 }
 
 void
 darwin_pragma_opt_size  (cpp_reader *pfile ATTRIBUTE_UNUSED)
 {
-  warning (0, "ignoring '#pragma optimize_for_size'");
+  if (warn_unknown_pragmas > in_system_header)
+    warning (OPT_Wunknown_pragmas, "ignoring #pragma optimize_for_size");
   return;
 }
 /* LLVM LOCAL end disable optimization pragmas */




From gohman at apple.com  Mon Jul 26 18:40:24 2010
From: gohman at apple.com (Dan Gohman)
Date: Mon, 26 Jul 2010 23:40:24 -0000
Subject: [llvm-commits] [llvm] r109468 -
	/llvm/trunk/lib/CodeGen/CriticalAntiDepBreaker.cpp
Message-ID: <20100726234024.2C5412A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 18:40:24 2010
New Revision: 109468

URL: http://llvm.org/viewvc/llvm-project?rev=109468&view=rev
Log:
Fix a use-after-free.

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

Modified: llvm/trunk/lib/CodeGen/CriticalAntiDepBreaker.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/CriticalAntiDepBreaker.cpp?rev=109468&r1=109467&r2=109468&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/CriticalAntiDepBreaker.cpp (original)
+++ llvm/trunk/lib/CodeGen/CriticalAntiDepBreaker.cpp Mon Jul 26 18:40:24 2010
@@ -41,8 +41,6 @@
 }
 
 void CriticalAntiDepBreaker::StartBlock(MachineBasicBlock *BB) {
-  Classes.clear();
-
   const unsigned BBSize = BB->size();
   for (unsigned i = 0, e = TRI->getNumRegs(); i != e; ++i) {
     // Clear out the register class data.




From stoklund at 2pi.dk  Mon Jul 26 18:44:11 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Mon, 26 Jul 2010 23:44:11 -0000
Subject: [llvm-commits] [llvm] r109469 - in /llvm/trunk/lib/CodeGen:
 InlineSpiller.cpp SplitKit.cpp SplitKit.h VirtRegMap.h
Message-ID: <20100726234411.F2CF42A6C12C@llvm.org>

Author: stoklund
Date: Mon Jul 26 18:44:11 2010
New Revision: 109469

URL: http://llvm.org/viewvc/llvm-project?rev=109469&view=rev
Log:
Add SplitEditor to SplitKit. This class will be used to edit live intervals and
rewrite instructions for live range splitting.

Still work in progress.

Modified:
    llvm/trunk/lib/CodeGen/InlineSpiller.cpp
    llvm/trunk/lib/CodeGen/SplitKit.cpp
    llvm/trunk/lib/CodeGen/SplitKit.h
    llvm/trunk/lib/CodeGen/VirtRegMap.h

Modified: llvm/trunk/lib/CodeGen/InlineSpiller.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/InlineSpiller.cpp?rev=109469&r1=109468&r2=109469&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/InlineSpiller.cpp (original)
+++ llvm/trunk/lib/CodeGen/InlineSpiller.cpp Mon Jul 26 18:44:11 2010
@@ -110,8 +110,8 @@
   splitAnalysis_.analyze(li_);
 
   if (const MachineLoop *loop = splitAnalysis_.getBestSplitLoop()) {
-    if (splitAroundLoop(splitAnalysis_, loop))
-      return true;
+    SplitEditor(splitAnalysis_, lis_, vrm_).splitAroundLoop(loop);
+    return true;
   }
   return false;
 }

Modified: llvm/trunk/lib/CodeGen/SplitKit.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SplitKit.cpp?rev=109469&r1=109468&r2=109469&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SplitKit.cpp (original)
+++ llvm/trunk/lib/CodeGen/SplitKit.cpp Mon Jul 26 18:44:11 2010
@@ -14,8 +14,10 @@
 
 #define DEBUG_TYPE "splitter"
 #include "SplitKit.h"
+#include "VirtRegMap.h"
 #include "llvm/CodeGen/LiveIntervalAnalysis.h"
 #include "llvm/CodeGen/MachineFunctionPass.h"
+#include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineLoopInfo.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/Support/CommandLine.h"
@@ -47,6 +49,7 @@
   usingInstrs_.clear();
   usingBlocks_.clear();
   usingLoops_.clear();
+  curli_ = 0;
 }
 
 bool SplitAnalysis::canAnalyzeBranch(const MachineBasicBlock *MBB) {
@@ -268,11 +271,216 @@
   return Best;
 }
 
+
+//===----------------------------------------------------------------------===//
+//                               Split Editor
+//===----------------------------------------------------------------------===//
+
+/// Create a new SplitEditor for editing the LiveInterval analyzed by SA.
+SplitEditor::SplitEditor(SplitAnalysis &sa, LiveIntervals &lis, VirtRegMap &vrm)
+  : sa_(sa), lis_(lis), vrm_(vrm),
+    mri_(vrm.getMachineFunction().getRegInfo()),
+    tii_(*vrm.getMachineFunction().getTarget().getInstrInfo()),
+    dupli_(0), openli_(0)
+{
+  const LiveInterval *curli = sa_.getCurLI();
+  assert(curli && "SplitEditor created from empty SplitAnalysis");
+
+  // Make sure curli is assigned a stack slot, so all our intervals get the
+  // same slot as curli.
+  if (vrm_.getStackSlot(curli->reg) == VirtRegMap::NO_STACK_SLOT)
+    vrm_.assignVirt2StackSlot(curli->reg);
+
+  // Create an interval for dupli that is a copy of curli.
+  dupli_ = createInterval();
+  dupli_->Copy(*curli, &mri_, lis_.getVNInfoAllocator());
+  DEBUG(dbgs() << "SplitEditor DupLI: " << *dupli_ << '\n');
+}
+
+LiveInterval *SplitEditor::createInterval() {
+  unsigned curli = sa_.getCurLI()->reg;
+  unsigned Reg = mri_.createVirtualRegister(mri_.getRegClass(curli));
+  LiveInterval &Intv = lis_.getOrCreateInterval(Reg);
+  vrm_.grow();
+  vrm_.assignVirt2StackSlot(Reg, vrm_.getStackSlot(curli));
+  return &Intv;
+}
+
+VNInfo *SplitEditor::mapValue(VNInfo *dupliVNI) {
+  VNInfo *&VNI = valueMap_[dupliVNI];
+  if (!VNI)
+    VNI = openli_->createValueCopy(dupliVNI, lis_.getVNInfoAllocator());
+  return VNI;
+}
+
+/// Create a new virtual register and live interval to be used by following
+/// use* and copy* calls.
+void SplitEditor::openLI() {
+  assert(!openli_ && "Previous LI not closed before openLI");
+  openli_ = createInterval();
+}
+
+/// copyToPHI - Insert a copy to openli at the end of A, and catch it with a
+/// PHI def at the beginning of the successor B. This call is ignored if dupli
+/// is not live out of A.
+void SplitEditor::copyToPHI(MachineBasicBlock &A, MachineBasicBlock &B) {
+  assert(openli_ && "openLI not called before copyToPHI");
+
+  SlotIndex EndA = lis_.getMBBEndIdx(&A);
+  VNInfo *DupVNIA = dupli_->getVNInfoAt(EndA.getPrevIndex());
+  if (!DupVNIA) {
+    DEBUG(dbgs() << "  ignoring copyToPHI, dupli not live out of BB#"
+                 << A.getNumber() << ".\n");
+    return;
+  }
+
+  // Insert the COPY instruction at the end of A.
+  MachineInstr *MI = BuildMI(A, A.getFirstTerminator(), DebugLoc(),
+                                 tii_.get(TargetOpcode::COPY), dupli_->reg)
+                           .addReg(openli_->reg);
+  SlotIndex DefIdx = lis_.InsertMachineInstrInMaps(MI).getDefIndex();
+
+  // Add a phi kill value and live range out of A.
+  VNInfo *VNIA = openli_->getNextValue(DefIdx, MI, true,
+                                       lis_.getVNInfoAllocator());
+  openli_->addRange(LiveRange(DefIdx, EndA, VNIA));
+
+  // Now look at the start of B.
+  SlotIndex StartB = lis_.getMBBStartIdx(&B);
+  SlotIndex EndB = lis_.getMBBEndIdx(&B);
+  LiveRange *DupB = dupli_->getLiveRangeContaining(StartB);
+  if (!DupB) {
+    DEBUG(dbgs() << "  copyToPHI:, dupli not live in to BB#"
+                 << B.getNumber() << ".\n");
+    return;
+  }
+
+  VNInfo *VNIB = openli_->getVNInfoAt(StartB);
+  if (!VNIB) {
+    // Create a phi value.
+    VNIB = openli_->getNextValue(SlotIndex(StartB, true), 0, false,
+                                 lis_.getVNInfoAllocator());
+    VNIB->setIsPHIDef(true);
+    // Add a minimal range for the new value.
+    openli_->addRange(LiveRange(VNIB->def, std::min(EndB, DupB->end), VNIB));
+
+    VNInfo *&mapVNI = valueMap_[DupB->valno];
+    if (mapVNI) {
+      // Multiple copies - must create PHI value.
+      abort();
+    } else {
+      // This is the first copy of dupLR. Mark the mapping.
+      mapVNI = VNIB;
+    }
+
+  }
+
+  DEBUG(dbgs() << "  copyToPHI at " << DefIdx << ": " << *openli_ << '\n');
+}
+
+/// useLI - indicate that all instructions in MBB should use openli.
+void SplitEditor::useLI(const MachineBasicBlock &MBB) {
+  useLI(lis_.getMBBStartIdx(&MBB), lis_.getMBBEndIdx(&MBB));
+}
+
+void SplitEditor::useLI(SlotIndex Start, SlotIndex End) {
+  assert(openli_ && "openLI not called before useLI");
+
+  // Map the dupli values from the interval into openli_
+  LiveInterval::const_iterator B = dupli_->begin(), E = dupli_->end();
+  LiveInterval::const_iterator I = std::lower_bound(B, E, Start);
+
+  if (I != B) {
+    --I;
+    // I begins before Start, but overlaps. openli may already have a value from
+    // copyToLI.
+    if (I->end > Start && !openli_->liveAt(Start))
+      openli_->addRange(LiveRange(Start, std::min(End, I->end),
+                        mapValue(I->valno)));
+    ++I;
+  }
+
+  // The remaining ranges begin after Start.
+  for (;I != E && I->start < End; ++I)
+    openli_->addRange(LiveRange(I->start, std::min(End, I->end),
+                                mapValue(I->valno)));
+  DEBUG(dbgs() << "  added range [" << Start << ';' << End << "): " << *openli_
+               << '\n');
+}
+
+/// copyFromLI - Insert a copy back to dupli from openli at position I.
+SlotIndex SplitEditor::copyFromLI(MachineBasicBlock &MBB, MachineBasicBlock::iterator I) {
+  assert(openli_ && "openLI not called before copyFromLI");
+
+  // Insert the COPY instruction.
+  MachineInstr *MI =
+    BuildMI(MBB, I, DebugLoc(), tii_.get(TargetOpcode::COPY), openli_->reg)
+      .addReg(dupli_->reg);
+  SlotIndex Idx = lis_.InsertMachineInstrInMaps(MI);
+
+  DEBUG(dbgs() << "  copyFromLI at " << Idx << ": " << *openli_ << '\n');
+  return Idx;
+}
+
+/// closeLI - Indicate that we are done editing the currently open
+/// LiveInterval, and ranges can be trimmed.
+void SplitEditor::closeLI() {
+  assert(openli_ && "openLI not called before closeLI");
+  openli_ = 0;
+}
+
+/// rewrite - after all the new live ranges have been created, rewrite
+/// instructions using curli to use the new intervals.
+void SplitEditor::rewrite() {
+  assert(!openli_ && "Previous LI not closed before rewrite");
+}
+
+
 //===----------------------------------------------------------------------===//
 //                               Loop Splitting
 //===----------------------------------------------------------------------===//
 
-bool llvm::splitAroundLoop(SplitAnalysis &sa, const MachineLoop *loop) {
-  return false;
+void SplitEditor::splitAroundLoop(const MachineLoop *Loop) {
+  SplitAnalysis::LoopBlocks Blocks;
+  sa_.getLoopBlocks(Loop, Blocks);
+
+  // Break critical edges as needed.
+  SplitAnalysis::BlockPtrSet CriticalExits;
+  sa_.getCriticalExits(Blocks, CriticalExits);
+  assert(CriticalExits.empty() && "Cannot break critical exits yet");
+
+  // Create new live interval for the loop.
+  openLI();
+
+  // Insert copies in the predecessors.
+  for (SplitAnalysis::BlockPtrSet::iterator I = Blocks.Preds.begin(),
+       E = Blocks.Preds.end(); I != E; ++I) {
+    MachineBasicBlock &MBB = const_cast<MachineBasicBlock&>(**I);
+    copyToPHI(MBB, *Loop->getHeader());
+  }
+
+  // Switch all loop blocks.
+  for (SplitAnalysis::BlockPtrSet::iterator I = Blocks.Loop.begin(),
+       E = Blocks.Loop.end(); I != E; ++I)
+     useLI(**I);
+
+  // Insert back copies in the exit blocks.
+  for (SplitAnalysis::BlockPtrSet::iterator I = Blocks.Exits.begin(),
+       E = Blocks.Exits.end(); I != E; ++I) {
+    MachineBasicBlock &MBB = const_cast<MachineBasicBlock&>(**I);
+    SlotIndex Start = lis_.getMBBStartIdx(&MBB);
+    VNInfo *VNI = sa_.getCurLI()->getVNInfoAt(Start);
+    // Only insert a back copy if curli is live and is either a phi or a value
+    // defined inside the loop.
+    if (!VNI) continue;
+    if (openli_->liveAt(VNI->def) ||
+        (VNI->isPHIDef() && VNI->def.getBaseIndex() == Start))
+      copyFromLI(MBB, MBB.begin());
+  }
+
+  // Done.
+  closeLI();
+  rewrite();
+  abort();
 }
 

Modified: llvm/trunk/lib/CodeGen/SplitKit.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SplitKit.h?rev=109469&r1=109468&r2=109469&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SplitKit.h (original)
+++ llvm/trunk/lib/CodeGen/SplitKit.h Mon Jul 26 18:44:11 2010
@@ -14,19 +14,22 @@
 
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/CodeGen/SlotIndexes.h"
 
 namespace llvm {
 
 class LiveInterval;
 class LiveIntervals;
-class MachineBasicBlock;
 class MachineInstr;
-class MachineFunction;
-class MachineFunctionPass;
 class MachineLoop;
 class MachineLoopInfo;
+class MachineRegisterInfo;
 class TargetInstrInfo;
+class VirtRegMap;
+class VNInfo;
 
+/// SplitAnalysis - Analyze a LiveInterval, looking for live range splitting
+/// opportunities.
 class SplitAnalysis {
   const MachineFunction &mf_;
   const LiveIntervals &lis_;
@@ -63,6 +66,8 @@
   /// split.
   void analyze(const LiveInterval *li);
 
+  const LiveInterval *getCurLI() { return curli_; }
+
   /// clear - clear all data structures so SplitAnalysis is ready to analyze a
   /// new interval.
   void clear();
@@ -113,8 +118,85 @@
   const MachineLoop *getBestSplitLoop();
 };
 
-/// splitAroundLoop - Try to split curli into a separate live interval inside
-/// the loop. Retun true on success.
-bool splitAroundLoop(SplitAnalysis&, const MachineLoop*);
+/// SplitEditor - Edit machine code and LiveIntervals for live range
+/// splitting.
+///
+/// 1. Create a SplitEditor from a SplitAnalysis. This will create a new
+///    LiveInterval, dupli, that is identical to SA.curli.
+/// 2. Start a new live interval with openLI.
+/// 3. Insert copies to the new interval with copyTo* and mark the ranges where
+///    it should be used with use*.
+/// 4. Insert back-copies with copyFromLI.
+/// 5. Finish the current LI with closeLI and repeat from 2.
+/// 6. Rewrite instructions with rewrite().
+///
+class SplitEditor {
+  SplitAnalysis &sa_;
+  LiveIntervals &lis_;
+  VirtRegMap &vrm_;
+  MachineRegisterInfo &mri_;
+  const TargetInstrInfo &tii_;
+
+  /// dupli_ - Created as a copy of sa_.curli_, ranges are carved out as new
+  /// intervals get added through openLI / closeLI.
+  LiveInterval *dupli_;
+
+  /// Currently open LiveInterval.
+  LiveInterval *openli_;
+
+  /// createInterval - Create a new virtual register and LiveInterval with same
+  /// register class and spill slot as curli.
+  LiveInterval *createInterval();
+
+	/// valueMap_ - Map values in dupli to values in openli. These are direct 1-1
+	/// mappings, and do not include values created by inserted copies.
+	DenseMap<VNInfo*,VNInfo*> valueMap_;
+
+	/// mapValue - Return the openli value that corresponds to the given dupli
+	/// value.
+	VNInfo *mapValue(VNInfo *dupliVNI);	
+
+public:
+  /// Create a new SplitEditor for editing the LiveInterval analyzed by SA.
+  SplitEditor(SplitAnalysis&, LiveIntervals&, VirtRegMap&);
+
+	/// getAnalysis - Get the corresponding analysis.
+	SplitAnalysis &getAnalysis() { return sa_; }
+
+  /// Create a new virtual register and live interval to be used by following
+  /// use* and copy* calls.
+  void openLI();
+
+  /// copyToPHI - Insert a copy to openli at the end of A, and catch it with a
+  /// PHI def at the beginning of the successor B. This call is ignored if dupli
+  /// is not live out of A.
+  void copyToPHI(MachineBasicBlock &A, MachineBasicBlock &B);
+
+  /// useLI - indicate that all instructions in MBB should use openli.
+  void useLI(const MachineBasicBlock &MBB);
+
+  /// useLI - indicate that all instructions in range should use openli.
+  void useLI(SlotIndex Start, SlotIndex End);
+
+  /// copyFromLI - Insert a copy back to dupli from openli at position I.
+	/// This also marks the remainder of MBB as not used by openli.
+  SlotIndex copyFromLI(MachineBasicBlock &MBB, MachineBasicBlock::iterator I);
+
+  /// closeLI - Indicate that we are done editing the currently open
+  /// LiveInterval, and ranges can be trimmed.
+  void closeLI();
+
+  /// rewrite - after all the new live ranges have been created, rewrite
+  /// instructions using curli to use the new intervals.
+  void rewrite();
+
+  // ===--- High level methods ---===
+
+  /// splitAroundLoop - Split curli into a separate live interval inside
+  /// the loop.
+  void splitAroundLoop(const MachineLoop*);
+
+};
+
 
 }

Modified: llvm/trunk/lib/CodeGen/VirtRegMap.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/VirtRegMap.h?rev=109469&r1=109468&r2=109469&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/VirtRegMap.h (original)
+++ llvm/trunk/lib/CodeGen/VirtRegMap.h Mon Jul 26 18:44:11 2010
@@ -152,6 +152,11 @@
       MachineFunctionPass::getAnalysisUsage(AU);
     }
 
+    MachineFunction &getMachineFunction() const {
+      assert(MF && "getMachineFunction called before runOnMAchineFunction");
+      return *MF;
+    }
+
     void grow();
 
     /// @brief returns true if the specified virtual register is




From resistor at mac.com  Mon Jul 26 18:58:46 2010
From: resistor at mac.com (Owen Anderson)
Date: Mon, 26 Jul 2010 23:58:46 -0000
Subject: [llvm-commits] [llvm] r109471 - /llvm/trunk/README.txt
Message-ID: <20100726235846.41F7A2A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 18:58:46 2010
New Revision: 109471

URL: http://llvm.org/viewvc/llvm-project?rev=109471&view=rev
Log:
Another test.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109471&r1=109470&r2=109471&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 18:58:46 2010
@@ -13,3 +13,4 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
+




From resistor at mac.com  Mon Jul 26 19:00:47 2010
From: resistor at mac.com (Owen Anderson)
Date: Tue, 27 Jul 2010 00:00:47 -0000
Subject: [llvm-commits] [llvm] r109472 - /llvm/trunk/README.txt
Message-ID: <20100727000047.D35E22A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 19:00:47 2010
New Revision: 109472

URL: http://llvm.org/viewvc/llvm-project?rev=109472&view=rev
Log:
Another test.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109472&r1=109471&r2=109472&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 19:00:47 2010
@@ -13,4 +13,3 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
-




From resistor at mac.com  Mon Jul 26 19:04:55 2010
From: resistor at mac.com (Owen Anderson)
Date: Tue, 27 Jul 2010 00:04:55 -0000
Subject: [llvm-commits] [llvm] r109473 - /llvm/trunk/README.txt
Message-ID: <20100727000455.4FFC02A6C12C@llvm.org>

Author: resistor
Date: Mon Jul 26 19:04:55 2010
New Revision: 109473

URL: http://llvm.org/viewvc/llvm-project?rev=109473&view=rev
Log:
(Hopefully) One last test.

Modified:
    llvm/trunk/README.txt

Modified: llvm/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/README.txt?rev=109473&r1=109472&r2=109473&view=diff
==============================================================================
--- llvm/trunk/README.txt (original)
+++ llvm/trunk/README.txt Mon Jul 26 19:04:55 2010
@@ -13,3 +13,4 @@
 
 If you're writing a package for LLVM, see docs/Packaging.html for our
 suggestions.
+




From gohman at apple.com  Mon Jul 26 20:19:06 2010
From: gohman at apple.com (Dan Gohman)
Date: Tue, 27 Jul 2010 01:19:06 -0000
Subject: [llvm-commits] [llvm] r109478 -
	/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h
Message-ID: <20100727011906.755292A6C12C@llvm.org>

Author: djg
Date: Mon Jul 26 20:19:06 2010
New Revision: 109478

URL: http://llvm.org/viewvc/llvm-project?rev=109478&view=rev
Log:
Use AssertingVH for InsertedValues and InsertedPostIncValues, to verify
that the values they refer to aren't being deleted underneath them.

Make sure these containters get cleared by clear(), which IndVarSimplify
and LSR both use before deleting instructions.

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

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h?rev=109478&r1=109477&r2=109478&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h Mon Jul 26 20:19:06 2010
@@ -18,6 +18,7 @@
 #include "llvm/Analysis/ScalarEvolutionNormalization.h"
 #include "llvm/Support/IRBuilder.h"
 #include "llvm/Support/TargetFolder.h"
+#include "llvm/Support/ValueHandle.h"
 #include <set>
 
 namespace llvm {
@@ -31,8 +32,8 @@
     ScalarEvolution &SE;
     std::map<std::pair<const SCEV *, Instruction *>, AssertingVH<Value> >
       InsertedExpressions;
-    std::set<Value*> InsertedValues;
-    std::set<Value*> InsertedPostIncValues;
+    std::set<AssertingVH<Value> > InsertedValues;
+    std::set<AssertingVH<Value> > InsertedPostIncValues;
 
     /// PostIncLoops - Addrecs referring to any of the given loops are expanded
     /// in post-inc mode. For example, expanding {1,+,1}<L> in post-inc mode
@@ -70,7 +71,11 @@
     /// clear - Erase the contents of the InsertedExpressions map so that users
     /// trying to expand the same expression into multiple BasicBlocks or
     /// different places within the same BasicBlock can do so.
-    void clear() { InsertedExpressions.clear(); }
+    void clear() {
+      InsertedExpressions.clear();
+      InsertedValues.clear();
+      InsertedPostIncValues.clear();
+    }
 
     /// getOrInsertCanonicalInductionVariable - This method returns the
     /// canonical induction variable of the specified type for the specified




From isanbard at gmail.com  Mon Jul 26 20:55:20 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 27 Jul 2010 01:55:20 -0000
Subject: [llvm-commits] [llvm] r109481 - in /llvm/trunk:
 include/llvm/CodeGen/MachineFrameInfo.h lib/CodeGen/MachineFunction.cpp
 lib/CodeGen/PrologEpilogInserter.cpp
 lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
Message-ID: <20100727015520.322172A6C12C@llvm.org>

Author: void
Date: Mon Jul 26 20:55:19 2010
New Revision: 109481

URL: http://llvm.org/viewvc/llvm-project?rev=109481&view=rev
Log:
It's better to have the arrays, which would trigger the creation of stack
protectors, to be near the stack protectors on the stack. Accomplish this by
tagging the stack object with a predicate that indicates that it would trigger
this. In the prolog-epilog inserter, assign these objects to the stack after the
stack protector but before the other objects.

Modified:
    llvm/trunk/include/llvm/CodeGen/MachineFrameInfo.h
    llvm/trunk/lib/CodeGen/MachineFunction.cpp
    llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp

Modified: llvm/trunk/include/llvm/CodeGen/MachineFrameInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineFrameInfo.h?rev=109481&r1=109480&r2=109481&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineFrameInfo.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineFrameInfo.h Mon Jul 26 20:55:19 2010
@@ -94,13 +94,19 @@
     // default, fixed objects are immutable unless marked otherwise.
     bool isImmutable;
 
-    // isSpillSlot - If true, the stack object is used as spill slot. It
+    // isSpillSlot - If true the stack object is used as spill slot. It
     // cannot alias any other memory objects.
     bool isSpillSlot;
 
-    StackObject(uint64_t Sz, unsigned Al, int64_t SP, bool IM, bool isSS)
+    // MayNeedSP - If true the stack object triggered the creation of the stack
+    // protector. We should allocate this object right after the stack
+    // protector.
+    bool MayNeedSP;
+
+    StackObject(uint64_t Sz, unsigned Al, int64_t SP, bool IM,
+                bool isSS, bool NSP)
       : SPOffset(SP), Size(Sz), Alignment(Al), isImmutable(IM),
-        isSpillSlot(isSS) {}
+        isSpillSlot(isSS), MayNeedSP(NSP) {}
   };
 
   /// Objects - The list of stack objects allocated...
@@ -276,6 +282,14 @@
     MaxAlignment = std::max(MaxAlignment, Align);
   }
 
+  /// NeedsStackProtector - Returns true if the object may need stack
+  /// protectors.
+  bool MayNeedStackProtector(int ObjectIdx) const {
+    assert(unsigned(ObjectIdx+NumFixedObjects) < Objects.size() &&
+           "Invalid Object Idx!");
+    return Objects[ObjectIdx+NumFixedObjects].MayNeedSP;
+  }
+
   /// getObjectOffset - Return the assigned stack offset of the specified object
   /// from the incoming stack pointer.
   ///
@@ -382,25 +396,26 @@
     return Objects[ObjectIdx+NumFixedObjects].Size == ~0ULL;
   }
 
-  /// CreateStackObject - Create a new statically sized stack object,
-  /// returning a nonnegative identifier to represent it.
+  /// CreateStackObject - Create a new statically sized stack object, returning
+  /// a nonnegative identifier to represent it.
   ///
-  int CreateStackObject(uint64_t Size, unsigned Alignment, bool isSS) {
+  int CreateStackObject(uint64_t Size, unsigned Alignment, bool isSS,
+                        bool MayNeedSP = false) {
     assert(Size != 0 && "Cannot allocate zero size stack objects!");
-    Objects.push_back(StackObject(Size, Alignment, 0, false, isSS));
-    int Index = (int)Objects.size()-NumFixedObjects-1;
+    Objects.push_back(StackObject(Size, Alignment, 0, false, isSS, MayNeedSP));
+    int Index = (int)Objects.size() - NumFixedObjects - 1;
     assert(Index >= 0 && "Bad frame index!");
     MaxAlignment = std::max(MaxAlignment, Alignment);
     return Index;
   }
 
-  /// CreateSpillStackObject - Create a new statically sized stack
-  /// object that represents a spill slot, returning a nonnegative
-  /// identifier to represent it.
+  /// CreateSpillStackObject - Create a new statically sized stack object that
+  /// represents a spill slot, returning a nonnegative identifier to represent
+  /// it.
   ///
   int CreateSpillStackObject(uint64_t Size, unsigned Alignment) {
-    CreateStackObject(Size, Alignment, true);
-    int Index = (int)Objects.size()-NumFixedObjects-1;
+    CreateStackObject(Size, Alignment, true, false);
+    int Index = (int)Objects.size() - NumFixedObjects - 1;
     MaxAlignment = std::max(MaxAlignment, Alignment);
     return Index;
   }
@@ -419,7 +434,7 @@
   ///
   int CreateVariableSizedObject(unsigned Alignment) {
     HasVarSizedObjects = true;
-    Objects.push_back(StackObject(0, Alignment, 0, false, false));
+    Objects.push_back(StackObject(0, Alignment, 0, false, false, true));
     MaxAlignment = std::max(MaxAlignment, Alignment);
     return (int)Objects.size()-NumFixedObjects-1;
   }

Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=109481&r1=109480&r2=109481&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Mon Jul 26 20:55:19 2010
@@ -446,7 +446,7 @@
   unsigned StackAlign = TFI.getStackAlignment();
   unsigned Align = MinAlign(SPOffset, StackAlign);
   Objects.insert(Objects.begin(), StackObject(Size, Align, SPOffset, Immutable,
-                                              /*isSS*/false));
+                                              /*isSS*/false, false));
   return -++NumFixedObjects;
 }
 

Modified: llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp?rev=109481&r1=109480&r2=109481&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp (original)
+++ llvm/trunk/lib/CodeGen/PrologEpilogInserter.cpp Mon Jul 26 20:55:19 2010
@@ -33,6 +33,7 @@
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/ADT/IndexedMap.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/STLExtras.h"
 #include <climits>
 
@@ -549,10 +550,29 @@
 
   // Make sure that the stack protector comes before the local variables on the
   // stack.
-  if (MFI->getStackProtectorIndex() >= 0)
+  SmallSet<int, 16> LargeStackObjs;
+  if (MFI->getStackProtectorIndex() >= 0) {
     AdjustStackOffset(MFI, MFI->getStackProtectorIndex(), StackGrowsDown,
                       Offset, MaxAlign);
 
+    // Assign large stack objects first.
+    for (unsigned i = 0, e = MFI->getObjectIndexEnd(); i != e; ++i) {
+      if (i >= MinCSFrameIndex && i <= MaxCSFrameIndex)
+        continue;
+      if (RS && (int)i == RS->getScavengingFrameIndex())
+        continue;
+      if (MFI->isDeadObjectIndex(i))
+        continue;
+      if (MFI->getStackProtectorIndex() == (int)i)
+        continue;
+      if (!MFI->MayNeedStackProtector(i))
+        continue;
+
+      AdjustStackOffset(MFI, i, StackGrowsDown, Offset, MaxAlign);
+      LargeStackObjs.insert(i);
+    }
+  }
+
   // Then assign frame offsets to stack objects that are not used to spill
   // callee saved registers.
   for (unsigned i = 0, e = MFI->getObjectIndexEnd(); i != e; ++i) {
@@ -564,6 +584,8 @@
       continue;
     if (MFI->getStackProtectorIndex() == (int)i)
       continue;
+    if (LargeStackObjs.count(i))
+      continue;
 
     AdjustStackOffset(MFI, i, StackGrowsDown, Offset, MaxAlign);
   }

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp?rev=109481&r1=109480&r2=109481&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp Mon Jul 26 20:55:19 2010
@@ -112,8 +112,16 @@
 
         TySize *= CUI->getZExtValue();   // Get total allocated size.
         if (TySize == 0) TySize = 1; // Don't create zero-sized stack objects.
+
+        // The object may need to be placed onto the stack near the stack
+        // protector if one exists. Determine here if this object is a suitable
+        // candidate. I.e., it would trigger the creation of a stack protector.
+        bool MayNeedSP =
+          (AI->isArrayAllocation() ||
+           (TySize > 8 && isa<ArrayType>(Ty) &&
+            cast<ArrayType>(Ty)->getElementType()->isIntegerTy(8)));
         StaticAllocaMap[AI] =
-          MF->getFrameInfo()->CreateStackObject(TySize, Align, false);
+          MF->getFrameInfo()->CreateStackObject(TySize, Align, false, MayNeedSP);
       }
 
   for (; BB != EB; ++BB)




From espindola at google.com  Mon Jul 26 22:27:12 2010
From: espindola at google.com (Rafael Espindola)
Date: Mon, 26 Jul 2010 23:27:12 -0400
Subject: [llvm-commits] A small cleanup in bugpoint
Message-ID: <AANLkTimrgW2f1fbJ-MfSvAdXpRi-tZ_NFCWqYuptrkKV@mail.gmail.com>

I recently had to debug bugpoint and found the code a bit confusing.
The two main things I found hard to follow were:

* The constant mutation a swapping of the "program being tested"
* The doTest functions that take a prefix and a suffix and are
expected to mutate the program

I think the first case can be avoided by moving the Program field out
of BugDriver. This also opens the possibility of implementing a
functional doTest.  I think we could implement something with the
interface:

  virtual const Module  *doTest(cost Module *M, std::vector<ElTy>
&List, std::string &Error);

The idea is that the test is pure and it should "apply" the List to M
and return the new module *if the bug is not present*. If the bug is
present, it should return NULL. A basic advantaged of this interface
is avoiding any confusion on when the program is updated or not.

Applying the list has different meanings depending on what ElTy is. It
can be running that list of passes or running a fixed list of passes
on that list of functions. A "bug being present" can be a crash or
miscompiled program.

The current Prefix + Suffix behavior of doTest would become something
like the pseudo code:


TempM = DoTest(M, Prefix) // try the prefix on M
if !TempM:
  Continue with the Prefix. It "has" the bug.
else:
  X = doTest(TempM, Suffix) // try the suffix on the result of
applying the prefix on M
  if !X:
    delete M
    M = TempM
    continue with the Suffix. It "has" the bug.
  else: // Splitting the list in these Suffix and Prefix doesn't
reproduce the bug
    delete TempM
    delete X

The attached patch is just a small part of this that is particularly
easy to fix. Instead of abusing swapProgramIn, just add a Module
argument to EmitProgressBitcode.

Is the patch OK. Are the above description and assumptions reasonable?

Cheers,
-- 
Rafael ?vila de Esp?ndola
-------------- next part --------------
A non-text attachment was scrubbed...
Name: bugpoint.patch
Type: text/x-patch
Size: 9613 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100726/8d998a30/attachment.bin 

From stoklund at 2pi.dk  Mon Jul 26 23:16:58 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 27 Jul 2010 04:16:58 -0000
Subject: [llvm-commits] [llvm] r109488 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
Message-ID: <20100727041658.B210E2A6C12C@llvm.org>

Author: stoklund
Date: Mon Jul 26 23:16:58 2010
New Revision: 109488

URL: http://llvm.org/viewvc/llvm-project?rev=109488&view=rev
Log:
Add assertions that expose the PR7713 miscompilation: Accessing a stack slot
with a too-big register class.

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

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=109488&r1=109487&r2=109488&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Mon Jul 26 23:16:58 2010
@@ -1998,6 +1998,8 @@
                                        const TargetRegisterClass *RC,
                                        const TargetRegisterInfo *TRI) const {
   const MachineFunction &MF = *MBB.getParent();
+  assert(MF.getFrameInfo()->getObjectSize(FrameIdx) >= RC->getSize() &&
+         "Stack slot too small for store");
   bool isAligned = (RI.getStackAlignment() >= 16) || RI.canRealignStack(MF);
   unsigned Opc = getStoreRegOpcode(SrcReg, RC, isAligned, TM);
   DebugLoc DL = MBB.findDebugLoc(MI);
@@ -2030,6 +2032,8 @@
                                         const TargetRegisterClass *RC,
                                         const TargetRegisterInfo *TRI) const {
   const MachineFunction &MF = *MBB.getParent();
+  assert(MF.getFrameInfo()->getObjectSize(FrameIdx) >= RC->getSize() &&
+         "Stack slot too small for load");
   bool isAligned = (RI.getStackAlignment() >= 16) || RI.canRealignStack(MF);
   unsigned Opc = getLoadRegOpcode(DestReg, RC, isAligned, TM);
   DebugLoc DL = MBB.findDebugLoc(MI);




From stoklund at 2pi.dk  Mon Jul 26 23:17:01 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 27 Jul 2010 04:17:01 -0000
Subject: [llvm-commits] [llvm] r109489 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
Message-ID: <20100727041702.0D1482A6C12D@llvm.org>

Author: stoklund
Date: Mon Jul 26 23:17:01 2010
New Revision: 109489

URL: http://llvm.org/viewvc/llvm-project?rev=109489&view=rev
Log:
The isLoadFromStackSlot and isStoreToStackSlot have no way of reporting
subregister operands like this:

%reg1040:sub_32bit<def> = MOV32rm <fi#-2>, 1, %reg0, 0, %reg0, %reg1040<imp-def>; mem:LD4[FixedStack-2](align=8)

Make them return false when subreg operands are present. VirtRegRewriter is
making bad assumptions otherwise.

This fixes PR7713.

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

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=109489&r1=109488&r2=109489&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Mon Jul 26 23:17:01 2010
@@ -787,7 +787,7 @@
 unsigned X86InstrInfo::isLoadFromStackSlot(const MachineInstr *MI, 
                                            int &FrameIndex) const {
   if (isFrameLoadOpcode(MI->getOpcode()))
-    if (isFrameOperand(MI, 1, FrameIndex))
+    if (MI->getOperand(0).getSubReg() == 0 && isFrameOperand(MI, 1, FrameIndex))
       return MI->getOperand(0).getReg();
   return 0;
 }
@@ -826,7 +826,8 @@
 unsigned X86InstrInfo::isStoreToStackSlot(const MachineInstr *MI,
                                           int &FrameIndex) const {
   if (isFrameStoreOpcode(MI->getOpcode()))
-    if (isFrameOperand(MI, 0, FrameIndex))
+    if (MI->getOperand(X86::AddrNumOperands).getSubReg() == 0 &&
+        isFrameOperand(MI, 0, FrameIndex))
       return MI->getOperand(X86::AddrNumOperands).getReg();
   return 0;
 }




From grosser at fim.uni-passau.de  Mon Jul 26 23:17:14 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Tue, 27 Jul 2010 04:17:14 -0000
Subject: [llvm-commits] [llvm] r109490 - in /llvm/trunk:
 include/llvm/Analysis/RegionInfo.h lib/Analysis/RegionInfo.cpp
Message-ID: <20100727041715.A2B3A2A6C12C@llvm.org>

Author: grosser
Date: Mon Jul 26 23:17:13 2010
New Revision: 109490

URL: http://llvm.org/viewvc/llvm-project?rev=109490&view=rev
Log:
Add function to query RegionInfo about loops.

* contains(Loop), * getOutermostLoop()
* Improve getNameStr() to return a sensible name, if basic blocks are not named.

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

Modified: llvm/trunk/include/llvm/Analysis/RegionInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/RegionInfo.h?rev=109490&r1=109489&r2=109490&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/RegionInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/RegionInfo.h Mon Jul 26 23:17:13 2010
@@ -37,6 +37,8 @@
 class Region;
 class RegionInfo;
 class raw_ostream;
+class Loop;
+class LoopInfo;
 
 /// @brief Marker class to iterate over the elements of a Region in flat mode.
 ///
@@ -287,16 +289,7 @@
 
   /// @brief Returns the name of the Region.
   /// @return The Name of the Region.
-  std::string getNameStr() const {
-    std::string exitName;
-
-    if (getExit())
-      exitName = getExit()->getNameStr();
-    else
-      exitName = "<Function Return>";
-
-    return getEntry()->getNameStr() + " => " + exitName;
-  }
+  std::string getNameStr() const;
 
   /// @brief Return the RegionInfo object, that belongs to this Region.
   RegionInfo *getRegionInfo() const {
@@ -340,6 +333,36 @@
     return contains(Inst->getParent());
   }
 
+  /// @brief Check if the region contains a loop.
+  ///
+  /// @param L The loop that might be contained in this region.
+  /// @return True if the loop is contained in the region otherwise false.
+  ///         In case a NULL pointer is passed to this function the result
+  ///         is false, except for the region that describes the whole function.
+  ///         In that case true is returned.
+  bool contains(const Loop *L) const;
+
+  /// @brief Get the outermost loop in the region that contains a loop.
+  ///
+  /// Find for a Loop L the outermost loop OuterL that is a parent loop of L
+  /// and is itself contained in the region.
+  ///
+  /// @param L The loop the lookup is started.
+  /// @return The outermost loop in the region, NULL if such a loop does not
+  ///         exist or if the region describes the whole function.
+  Loop *outermostLoopInRegion(Loop *L) const;
+
+  /// @brief Get the outermost loop in the region that contains a basic block.
+  ///
+  /// Find for a basic block BB the outermost loop L that contains BB and is
+  /// itself contained in the region.
+  ///
+  /// @param LI A pointer to a LoopInfo analysis.
+  /// @param BB The basic block surrounded by the loop.
+  /// @return The outermost loop in the region, NULL if such a loop does not
+  ///         exist or if the region describes the whole function.
+  Loop *outermostLoopInRegion(LoopInfo *LI, BasicBlock* BB) const;
+
   /// @brief Get the subregion that starts at a BasicBlock
   ///
   /// @param BB The BasicBlock the subregion should start.

Modified: llvm/trunk/lib/Analysis/RegionInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionInfo.cpp?rev=109490&r1=109489&r2=109490&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/RegionInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/RegionInfo.cpp Mon Jul 26 23:17:13 2010
@@ -17,6 +17,7 @@
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/Analysis/LoopInfo.h"
 
 #define DEBUG_TYPE "region"
 #include "llvm/Support/Debug.h"
@@ -81,6 +82,44 @@
     && !(DT->dominates(exit, BB) && DT->dominates(entry, exit)));
 }
 
+bool Region::contains(const Loop *L) const {
+  // BBs that are not part of any loop are element of the Loop
+  // described by the NULL pointer. This loop is not part of any region,
+  // except if the region describes the whole function.
+  if (L == 0)
+    return getExit() == 0;
+
+  if (!contains(L->getHeader()))
+    return false;
+
+  SmallVector<BasicBlock *, 8> ExitingBlocks;
+  L->getExitingBlocks(ExitingBlocks);
+
+  for (SmallVectorImpl<BasicBlock*>::iterator BI = ExitingBlocks.begin(),
+       BE = ExitingBlocks.end(); BI != BE; ++BI)
+    if (!contains(*BI))
+      return false;
+
+  return true;
+}
+
+Loop *Region::outermostLoopInRegion(Loop *L) const {
+  if (!contains(L))
+    return 0;
+
+  while (L && contains(L->getParentLoop())) {
+    L = L->getParentLoop();
+  }
+
+  return L;
+}
+
+Loop *Region::outermostLoopInRegion(LoopInfo *LI, BasicBlock* BB) const {
+  assert(LI && BB && "LI and BB cannot be null!");
+  Loop *L = LI->getLoopFor(BB);
+  return outermostLoopInRegion(L);
+}
+
 bool Region::isSimple() const {
   bool isSimple = true;
   bool found = false;
@@ -116,6 +155,32 @@
   return isSimple;
 }
 
+std::string Region::getNameStr() const {
+  std::string exitName;
+  std::string entryName;
+
+  if (getEntry()->getName().empty()) {
+    raw_string_ostream OS(entryName);
+
+    WriteAsOperand(OS, getEntry(), false);
+    entryName = OS.str();
+  } else
+    entryName = getEntry()->getNameStr();
+
+  if (getExit()) {
+    if (getExit()->getName().empty()) {
+      raw_string_ostream OS(exitName);
+
+      WriteAsOperand(OS, getExit(), false);
+      exitName = OS.str();
+    } else
+      exitName = getExit()->getNameStr();
+  } else
+    exitName = "<Function Return>";
+
+  return entryName + " => " + exitName;
+}
+
 void Region::verifyBBInRegion(BasicBlock *BB) const {
   if (!contains(BB))
     llvm_unreachable("Broken region found!");




From kledzik at apple.com  Tue Jul 27 01:24:32 2010
From: kledzik at apple.com (Nick Kledzik)
Date: Tue, 27 Jul 2010 06:24:32 -0000
Subject: [llvm-commits] [compiler-rt] r109492 -
	/compiler-rt/trunk/lib/arm/modsi3.S
Message-ID: <20100727062433.6AF0F2A6C12C@llvm.org>

Author: kledzik
Date: Tue Jul 27 01:24:32 2010
New Revision: 109492

URL: http://llvm.org/viewvc/llvm-project?rev=109492&view=rev
Log:
add assembly implementation of modsi3 so compiler does not have to special case a - (a / b) * b optimization

Added:
    compiler-rt/trunk/lib/arm/modsi3.S

Added: compiler-rt/trunk/lib/arm/modsi3.S
URL: http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/arm/modsi3.S?rev=109492&view=auto
==============================================================================
--- compiler-rt/trunk/lib/arm/modsi3.S (added)
+++ compiler-rt/trunk/lib/arm/modsi3.S Tue Jul 27 01:24:32 2010
@@ -0,0 +1,36 @@
+//===-------- modsi3.S - Implement modsi3 ---------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "../assembly.h"
+
+//
+// extern int32_t __modsi3(int32_t a, int32_t b);
+//
+// Returns the remainder when dividing two 32-bit signed integers.
+// Conceptually, the function is: { return a - (a / b) * b; }
+// But if you write that in C, llvm compiles it to a call to __modsi3...
+//
+	.align 2
+DEFINE_COMPILERRT_FUNCTION(__modsi3)
+	push	{r4, r5, r7, lr}
+	add	r7, sp, #8	// set stack frame
+	mov	r5, r0		// save a
+	mov	r4, r1		// save b
+	bl      ___divsi3	// compute a/b
+#if __ARM_ARCH_7A__
+	mls     r0, r4, r0, r5  // mulitple result * b and subtract from a
+#else	
+	// before armv7, does not have "mls" instruction
+	mul	r3, r0, r4	// multiple result * b
+	sub     r0, r5, r3      // a - result
+#endif
+	pop     {r4, r5, r7, pc}
+	
+
+




From kledzik at apple.com  Tue Jul 27 01:27:23 2010
From: kledzik at apple.com (Nick Kledzik)
Date: Tue, 27 Jul 2010 06:27:23 -0000
Subject: [llvm-commits] [compiler-rt] r109493 -
	/compiler-rt/tags/Apple/Libcompiler_rt-9/
Message-ID: <20100727062723.E169C2A6C12C@llvm.org>

Author: kledzik
Date: Tue Jul 27 01:27:23 2010
New Revision: 109493

URL: http://llvm.org/viewvc/llvm-project?rev=109493&view=rev
Log:
Libcompiler_rt-9

Added:
    compiler-rt/tags/Apple/Libcompiler_rt-9/
      - copied from r109492, compiler-rt/trunk/



From bigcheesegs at gmail.com  Tue Jul 27 01:46:16 2010
From: bigcheesegs at gmail.com (Michael J. Spencer)
Date: Tue, 27 Jul 2010 06:46:16 -0000
Subject: [llvm-commits] [llvm] r109494 - in /llvm/trunk:
 lib/Target/X86/X86AsmBackend.cpp lib/Target/X86/X86TargetMachine.cpp
 test/MC/COFF/ test/MC/COFF/basic-coff.ll test/MC/COFF/dg.exp
 test/Scripts/coff-dump.py test/Scripts/coff-dump.py.bat
Message-ID: <20100727064616.4BAD62A6C12C@llvm.org>

Author: mspencer
Date: Tue Jul 27 01:46:15 2010
New Revision: 109494

URL: http://llvm.org/viewvc/llvm-project?rev=109494&view=rev
Log:
Make MC use Windows COFF on Windows and add tests.

Added:
    llvm/trunk/test/MC/COFF/
    llvm/trunk/test/MC/COFF/basic-coff.ll
    llvm/trunk/test/MC/COFF/dg.exp
    llvm/trunk/test/Scripts/coff-dump.py
    llvm/trunk/test/Scripts/coff-dump.py.bat
Modified:
    llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
    llvm/trunk/lib/Target/X86/X86TargetMachine.cpp

Modified: llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86AsmBackend.cpp?rev=109494&r1=109493&r2=109494&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86AsmBackend.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86AsmBackend.cpp Tue Jul 27 01:46:15 2010
@@ -14,6 +14,7 @@
 #include "llvm/MC/MCAssembler.h"
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCObjectWriter.h"
+#include "llvm/MC/MCSectionCOFF.h"
 #include "llvm/MC/MCSectionELF.h"
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MachObjectWriter.h"
@@ -212,6 +213,24 @@
     : ELFX86AsmBackend(T) {}
 };
 
+class WindowsX86AsmBackend : public X86AsmBackend {
+public:
+  WindowsX86AsmBackend(const Target &T)
+    : X86AsmBackend(T) {
+    HasAbsolutizedSet = true;
+    HasScatteredSymbols = true;
+  }
+
+  MCObjectWriter *createObjectWriter(raw_ostream &OS) const {
+    return createWinCOFFObjectWriter (OS);
+  }
+
+  bool isVirtualSection(const MCSection &Section) const {
+    const MCSectionCOFF &SE = static_cast<const MCSectionCOFF&>(Section);
+    return SE.getCharacteristics() & COFF::IMAGE_SCN_CNT_UNINITIALIZED_DATA;
+  }
+};
+
 class DarwinX86AsmBackend : public X86AsmBackend {
 public:
   DarwinX86AsmBackend(const Target &T)
@@ -290,6 +309,8 @@
   switch (Triple(TT).getOS()) {
   case Triple::Darwin:
     return new DarwinX86_32AsmBackend(T);
+  case Triple::Win32:
+    return new WindowsX86AsmBackend(T);
   default:
     return new ELFX86_32AsmBackend(T);
   }

Modified: llvm/trunk/lib/Target/X86/X86TargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetMachine.cpp?rev=109494&r1=109493&r2=109494&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetMachine.cpp Tue Jul 27 01:46:15 2010
@@ -46,6 +46,8 @@
                                     bool RelaxAll) {
   Triple TheTriple(TT);
   switch (TheTriple.getOS()) {
+  case Triple::Win32:
+    return createWinCOFFStreamer(Ctx, TAB, *_Emitter, _OS);
   default:
     return createMachOStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
   }

Added: llvm/trunk/test/MC/COFF/basic-coff.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/COFF/basic-coff.ll?rev=109494&view=auto
==============================================================================
--- llvm/trunk/test/MC/COFF/basic-coff.ll (added)
+++ llvm/trunk/test/MC/COFF/basic-coff.ll Tue Jul 27 01:46:15 2010
@@ -0,0 +1,136 @@
+; RUN: llc -filetype=obj %s -o %t
+; RUN: coff-dump.py %abs_tmp | FileCheck %s
+
+; ModuleID = '-'
+target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f80:128:128-v64:64:64-v128:128:128-a0:0:64-f80:32:32-n8:16:32"
+target triple = "i686-pc-win32"
+
+ at .str = private constant [12 x i8] c"Hello World\00" ; <[12 x i8]*> [#uses=1]
+
+define i32 @main() nounwind {
+entry:
+  %call = tail call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([12 x i8]* @.str, i32 0, i32 0)) nounwind ; <i32> [#uses=0]
+  ret i32 0
+}
+
+declare i32 @printf(i8* nocapture, ...) nounwind
+
+; CHECK: {
+; CHECK:   MachineType              = IMAGE_FILE_MACHINE_I386 (0x14C)
+; CHECK:   NumberOfSections         = 2
+; CHECK:   TimeDateStamp            = {{[0-9]+}}
+; CHECK:   PointerToSymbolTable     = 0x99
+; CHECK:   NumberOfSymbols          = 7
+; CHECK:   SizeOfOptionalHeader     = 0
+; CHECK:   Characteristics          = 0x0
+; CHECK:   Sections                 = [
+; CHECK:     0 = {
+; CHECK:       Name                     = .text
+; CHECK:       VirtualSize              = 0
+; CHECK:       VirtualAddress           = 0
+; CHECK:       SizeOfRawData            = 21
+; CHECK:       PointerToRawData         = 0x64
+; CHECK:       PointerToRelocations     = 0x79
+; CHECK:       PointerToLineNumbers     = 0x0
+; CHECK:       NumberOfRelocations      = 2
+; CHECK:       NumberOfLineNumbers      = 0
+; CHECK:       Charateristics           = 0x60500020
+; CHECK:         IMAGE_SCN_CNT_CODE
+; CHECK:         IMAGE_SCN_ALIGN_16BYTES
+; CHECK:         IMAGE_SCN_MEM_EXECUTE
+; CHECK:         IMAGE_SCN_MEM_READ
+; CHECK:       SectionData              =
+; CHECK:         83 EC 04 C7 04 24 00 00 - 00 00 E8 00 00 00 00 31 |.....$.........1|
+; CHECK:         C0 83 C4 04 C3                                    |.....|
+; CHECK:       Relocations              = [
+; CHECK:         0 = {
+; CHECK:           VirtualAddress           = 0x6
+; CHECK:           SymbolTableIndex         = 5
+; CHECK:           Type                     = IMAGE_REL_I386_DIR32 (6)
+; CHECK:           SymbolName               = _main
+; CHECK:         }
+; CHECK:         1 = {
+; CHECK:           VirtualAddress           = 0xB
+; CHECK:           SymbolTableIndex         = 6
+; CHECK:           Type                     = IMAGE_REL_I386_REL32 (20)
+; CHECK:           SymbolName               = L_.str
+; CHECK:         }
+; CHECK:       ]
+; CHECK:     }
+; CHECK:     1 = {
+; CHECK:       Name                     = .data
+; CHECK:       VirtualSize              = 0
+; CHECK:       VirtualAddress           = 0
+; CHECK:       SizeOfRawData            = 12
+; CHECK:       PointerToRawData         = 0x8D
+; CHECK:       PointerToRelocations     = 0x0
+; CHECK:       PointerToLineNumbers     = 0x0
+; CHECK:       NumberOfRelocations      = 0
+; CHECK:       NumberOfLineNumbers      = 0
+; CHECK:       Charateristics           = 0xC0100040
+; CHECK:         IMAGE_SCN_CNT_INITIALIZED_DATA
+; CHECK:         IMAGE_SCN_ALIGN_1BYTES
+; CHECK:         IMAGE_SCN_MEM_READ
+; CHECK:         IMAGE_SCN_MEM_WRITE
+; CHECK:       SectionData              =
+; CHECK:         48 65 6C 6C 6F 20 57 6F - 72 6C 64 00             |Hello World.|
+; CHECK:       Relocations              = None
+; CHECK:     }
+; CHECK:   ]
+; CHECK:   Symbols                  = [
+; CHECK:     0 = {
+; CHECK:       Name                     = .text
+; CHECK:       Value                    = 0
+; CHECK:       SectionNumber            = 1
+; CHECK:       SimpleType               = IMAGE_SYM_TYPE_NULL (0)
+; CHECK:       ComplexType              = IMAGE_SYM_DTYPE_NULL (0)
+; CHECK:       StorageClass             = IMAGE_SYM_CLASS_STATIC (3)
+; CHECK:       NumberOfAuxSymbols       = 1
+; CHECK:       AuxillaryData            =
+; CHECK:         15 00 00 00 02 00 00 00 - 00 00 00 00 01 00 00 00 |................|
+; CHECK:         00 00                                             |..|
+; CHECK:     }
+; CHECK:     1 = {
+; CHECK:       Name                     = .data
+; CHECK:       Value                    = 0
+; CHECK:       SectionNumber            = 2
+; CHECK:       SimpleType               = IMAGE_SYM_TYPE_NULL (0)
+; CHECK:       ComplexType              = IMAGE_SYM_DTYPE_NULL (0)
+; CHECK:       StorageClass             = IMAGE_SYM_CLASS_STATIC (3)
+; CHECK:       NumberOfAuxSymbols       = 1
+; CHECK:       AuxillaryData            =
+; CHECK:         0C 00 00 00 00 00 00 00 - 00 00 00 00 02 00 00 00 |................|
+; CHECK:         00 00                                             |..|
+; CHECK:     }
+; CHECK:     2 = {
+; CHECK:       Name                     = _main
+; CHECK:       Value                    = 0
+; CHECK:       SectionNumber            = 1
+; CHECK:       SimpleType               = unknown (32)
+; CHECK:       ComplexType              = IMAGE_SYM_DTYPE_NULL (0)
+; CHECK:       StorageClass             = IMAGE_SYM_CLASS_EXTERNAL (2)
+; CHECK:       NumberOfAuxSymbols       = 0
+; CHECK:       AuxillaryData            =
+; CHECK:     }
+; CHECK:     3 = {
+; CHECK:       Name                     = L_.str
+; CHECK:       Value                    = 0
+; CHECK:       SectionNumber            = 2
+; CHECK:       SimpleType               = IMAGE_SYM_TYPE_NULL (0)
+; CHECK:       ComplexType              = IMAGE_SYM_DTYPE_NULL (0)
+; CHECK:       StorageClass             = IMAGE_SYM_CLASS_STATIC (3)
+; CHECK:       NumberOfAuxSymbols       = 0
+; CHECK:       AuxillaryData            =
+; CHECK:     }
+; CHECK:     4 = {
+; CHECK:       Name                     = _printf
+; CHECK:       Value                    = 0
+; CHECK:       SectionNumber            = 0
+; CHECK:       SimpleType               = IMAGE_SYM_TYPE_NULL (0)
+; CHECK:       ComplexType              = IMAGE_SYM_DTYPE_NULL (0)
+; CHECK:       StorageClass             = IMAGE_SYM_CLASS_EXTERNAL (2)
+; CHECK:       NumberOfAuxSymbols       = 0
+; CHECK:       AuxillaryData            =
+; CHECK:     }
+; CHECK:   ]
+; CHECK: }

Added: llvm/trunk/test/MC/COFF/dg.exp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/COFF/dg.exp?rev=109494&view=auto
==============================================================================
--- llvm/trunk/test/MC/COFF/dg.exp (added)
+++ llvm/trunk/test/MC/COFF/dg.exp Tue Jul 27 01:46:15 2010
@@ -0,0 +1,5 @@
+load_lib llvm.exp
+
+if { [llvm_supports_target X86] } {
+  RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll}]]
+}

Added: llvm/trunk/test/Scripts/coff-dump.py
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Scripts/coff-dump.py?rev=109494&view=auto
==============================================================================
--- llvm/trunk/test/Scripts/coff-dump.py (added)
+++ llvm/trunk/test/Scripts/coff-dump.py Tue Jul 27 01:46:15 2010
@@ -0,0 +1,543 @@
+#===-- coff-dump.py - COFF object file dump utility-------------------------===#
+#
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+#
+# COFF File Definition
+#
+
+def string_table_entry (offset):
+	return ('ptr', '+ + PointerToSymbolTable * NumberOfSymbols 18 %s' % offset, ('scalar', 'cstr', '%s'))
+
+def secname(value):
+	if value[0] == '/':
+		return string_table_entry (value [1:].rstrip('\0'))
+	else:
+		return '%s'
+
+def symname(value):
+	parts = struct.unpack("<2L", value)
+	if parts [0] == 0:
+		return string_table_entry (parts [1])
+	else:
+		return '%s'
+
+file = ('struct', [
+	('MachineType', ('enum', '<H', '0x%X', {
+		0x0:    'IMAGE_FILE_MACHINE_UNKNOWN',
+		0x1d3:  'IMAGE_FILE_MACHINE_AM33',
+		0x866:  'IMAGE_FILE_MACHINE_AMD64',
+		0x1c0:  'IMAGE_FILE_MACHINE_ARM',
+		0xebc:  'IMAGE_FILE_MACHINE_EBC',
+		0x14c:  'IMAGE_FILE_MACHINE_I386',
+		0x200:  'IMAGE_FILE_MACHINE_IA64',
+		0x904:  'IMAGE_FILE_MACHINE_M32R',
+		0x266:  'IMAGE_FILE_MACHINE_MIPS16',
+		0x366:  'IMAGE_FILE_MACHINE_MIPSFPU',
+		0x466:  'IMAGE_FILE_MACHINE_MIPSFPU16',
+		0x1f0:  'IMAGE_FILE_MACHINE_POWERPC',
+		0x1f1:  'IMAGE_FILE_MACHINE_POWERPCFP',
+		0x166:  'IMAGE_FILE_MACHINE_R4000',
+		0x1a2:  'IMAGE_FILE_MACHINE_SH3',
+		0x1a3:  'IMAGE_FILE_MACHINE_SH3DSP',
+		0x1a6:  'IMAGE_FILE_MACHINE_SH4',
+		0x1a8:  'IMAGE_FILE_MACHINE_SH5',
+		0x1c2:  'IMAGE_FILE_MACHINE_THUMB',
+		0x169:  'IMAGE_FILE_MACHINE_WCEMIPSV2',
+	})),
+	('NumberOfSections',     ('scalar',  '<H', '%d')),
+	('TimeDateStamp',        ('scalar',  '<L', '%d')),
+	('PointerToSymbolTable', ('scalar',  '<L', '0x%0X')),
+	('NumberOfSymbols',      ('scalar',  '<L', '%d')),
+	('SizeOfOptionalHeader', ('scalar',  '<H', '%d')),
+	('Characteristics',      ('flags',   '<H', '0x%x', [
+		(0x0001,      'IMAGE_FILE_RELOCS_STRIPPED',         ),
+		(0x0002,      'IMAGE_FILE_EXECUTABLE_IMAGE',        ),
+		(0x0004,      'IMAGE_FILE_LINE_NUMS_STRIPPED',      ),
+		(0x0008,      'IMAGE_FILE_LOCAL_SYMS_STRIPPED',     ),
+		(0x0010,      'IMAGE_FILE_AGGRESSIVE_WS_TRIM',      ),
+		(0x0020,      'IMAGE_FILE_LARGE_ADDRESS_AWARE',     ),
+		(0x0080,      'IMAGE_FILE_BYTES_REVERSED_LO',       ),
+		(0x0100,      'IMAGE_FILE_32BIT_MACHINE',           ),
+		(0x0200,      'IMAGE_FILE_DEBUG_STRIPPED',          ),
+		(0x0400,      'IMAGE_FILE_REMOVABLE_RUN_FROM_SWAP', ),
+		(0x0800,      'IMAGE_FILE_NET_RUN_FROM_SWAP',       ),
+		(0x1000,      'IMAGE_FILE_SYSTEM',                  ),
+		(0x2000,      'IMAGE_FILE_DLL',                     ),
+		(0x4000,      'IMAGE_FILE_UP_SYSTEM_ONLY',          ),
+		(0x8000,      'IMAGE_FILE_BYTES_REVERSED_HI',       ),
+	])),
+	('Sections', ('array', 'NumberOfSections', ('struct', [
+		('Name',                 ('scalar',  '<8s', secname)),
+		('VirtualSize',          ('scalar',  '<L',  '%d'   )),
+		('VirtualAddress',       ('scalar',  '<L',  '%d'   )),
+		('SizeOfRawData',        ('scalar',  '<L',  '%d'   )),
+		('PointerToRawData',     ('scalar',  '<L',  '0x%X' )),
+		('PointerToRelocations', ('scalar',  '<L',  '0x%X' )),
+		('PointerToLineNumbers', ('scalar',  '<L',  '0x%X' )),
+		('NumberOfRelocations',  ('scalar',  '<H',  '%d'   )),
+		('NumberOfLineNumbers',  ('scalar',  '<H',  '%d'   )),
+		('Charateristics',       ('flags',   '<L',  '0x%X', [
+			(0x00000008, 'IMAGE_SCN_TYPE_NO_PAD'),
+			(0x00000020, 'IMAGE_SCN_CNT_CODE'),
+			(0x00000040, 'IMAGE_SCN_CNT_INITIALIZED_DATA'),
+			(0x00000080, 'IMAGE_SCN_CNT_UNINITIALIZED_DATA'),
+			(0x00000100, 'IMAGE_SCN_LNK_OTHER'),
+			(0x00000200, 'IMAGE_SCN_LNK_INFO'),
+			(0x00000800, 'IMAGE_SCN_LNK_REMOVE'),
+			(0x00001000, 'IMAGE_SCN_LNK_COMDAT'),
+			(0x00008000, 'IMAGE_SCN_GPREL'),
+			(0x00020000, 'IMAGE_SCN_MEM_PURGEABLE'),
+			(0x00020000, 'IMAGE_SCN_MEM_16BIT'),
+			(0x00040000, 'IMAGE_SCN_MEM_LOCKED'),
+			(0x00080000, 'IMAGE_SCN_MEM_PRELOAD'),
+			(0x00F00000, 'IMAGE_SCN_ALIGN', {
+				0x00100000: 'IMAGE_SCN_ALIGN_1BYTES',
+				0x00200000: 'IMAGE_SCN_ALIGN_2BYTES',
+				0x00300000: 'IMAGE_SCN_ALIGN_4BYTES',
+				0x00400000: 'IMAGE_SCN_ALIGN_8BYTES',
+				0x00500000: 'IMAGE_SCN_ALIGN_16BYTES',
+				0x00600000: 'IMAGE_SCN_ALIGN_32BYTES',
+				0x00700000: 'IMAGE_SCN_ALIGN_64BYTES',
+				0x00800000: 'IMAGE_SCN_ALIGN_128BYTES',
+				0x00900000: 'IMAGE_SCN_ALIGN_256BYTES',
+				0x00A00000: 'IMAGE_SCN_ALIGN_512BYTES',
+				0x00B00000: 'IMAGE_SCN_ALIGN_1024BYTES',
+				0x00C00000: 'IMAGE_SCN_ALIGN_2048BYTES',
+				0x00D00000: 'IMAGE_SCN_ALIGN_4096BYTES',
+				0x00E00000: 'IMAGE_SCN_ALIGN_8192BYTES',
+			}),
+			(0x01000000, 'IMAGE_SCN_LNK_NRELOC_OVFL'),
+			(0x02000000, 'IMAGE_SCN_MEM_DISCARDABLE'),
+			(0x04000000, 'IMAGE_SCN_MEM_NOT_CACHED'),
+			(0x08000000, 'IMAGE_SCN_MEM_NOT_PAGED'),
+			(0x10000000, 'IMAGE_SCN_MEM_SHARED'),
+			(0x20000000, 'IMAGE_SCN_MEM_EXECUTE'),
+			(0x40000000, 'IMAGE_SCN_MEM_READ'),
+			(0x80000000, 'IMAGE_SCN_MEM_WRITE'),
+		])),
+		('SectionData', ('ptr', 'PointerToRawData', ('blob', 'SizeOfRawData'))),
+		('Relocations', ('ptr', 'PointerToRelocations', ('array', 'NumberOfRelocations', ('struct', [
+			('VirtualAddress',   ('scalar', '<L', '0x%X')),
+			('SymbolTableIndex', ('scalar', '<L', '%d'  )),
+			('Type',             ('enum', '<H', '%d', ('MachineType', {
+				0x14c: {
+					0x0000: 'IMAGE_REL_I386_ABSOLUTE',
+					0x0001: 'IMAGE_REL_I386_DIR16',
+					0x0002: 'IMAGE_REL_I386_REL16',
+					0x0006: 'IMAGE_REL_I386_DIR32',
+					0x0007: 'IMAGE_REL_I386_DIR32NB',
+					0x0009: 'IMAGE_REL_I386_SEG12',
+					0x000A: 'IMAGE_REL_I386_SECTION',
+					0x000B: 'IMAGE_REL_I386_SECREL',
+					0x000C: 'IMAGE_REL_I386_TOKEN',
+					0x000D: 'IMAGE_REL_I386_SECREL7',
+					0x0014: 'IMAGE_REL_I386_REL32',
+				},
+			}))),
+			('SymbolName',       ('ptr', '+ PointerToSymbolTable * - SymbolTableIndex 1 18', ('scalar',  '<8s', symname)))
+		])))),
+	]))),
+	('Symbols', ('ptr', 'PointerToSymbolTable', ('byte-array', '* NumberOfSymbols 18',  ('struct', [
+		('Name',                ('scalar',  '<8s', symname)),
+		('Value',               ('scalar',  '<L',  '%d'   )),
+		('SectionNumber',       ('scalar',  '<H',  '%d'   )),
+		('SimpleType',          ('enum',    '<B',  '%d', {
+			0: 'IMAGE_SYM_TYPE_NULL',
+			1: 'IMAGE_SYM_TYPE_VOID',
+			2: 'IMAGE_SYM_TYPE_CHAR',
+			3: 'IMAGE_SYM_TYPE_SHORT',
+			4: 'IMAGE_SYM_TYPE_INT',
+			5: 'IMAGE_SYM_TYPE_LONG',
+			6: 'IMAGE_SYM_TYPE_FLOAT',
+			7: 'IMAGE_SYM_TYPE_DOUBLE',
+			8: 'IMAGE_SYM_TYPE_STRUCT',
+			9: 'IMAGE_SYM_TYPE_UNION',
+			10: 'IMAGE_SYM_TYPE_ENUM',
+			11: 'IMAGE_SYM_TYPE_MOE',
+			12: 'IMAGE_SYM_TYPE_BYTE',
+			13: 'IMAGE_SYM_TYPE_WORD',
+			14: 'IMAGE_SYM_TYPE_UINT',
+			15: 'IMAGE_SYM_TYPE_DWORD',
+		})),
+		('ComplexType',         ('enum',    '<B',  '%d', {
+			0: 'IMAGE_SYM_DTYPE_NULL',
+			1: 'IMAGE_SYM_DTYPE_POINTER',
+			2: 'IMAGE_SYM_DTYPE_FUNCTION',
+			3: 'IMAGE_SYM_DTYPE_ARRAY',
+		})),
+		('StorageClass',        ('enum',    '<B',  '%d', {
+			-1:	'IMAGE_SYM_CLASS_END_OF_FUNCTION',
+			0: 'IMAGE_SYM_CLASS_NULL',
+			1: 'IMAGE_SYM_CLASS_AUTOMATIC',
+			2: 'IMAGE_SYM_CLASS_EXTERNAL',
+			3: 'IMAGE_SYM_CLASS_STATIC',
+			4: 'IMAGE_SYM_CLASS_REGISTER',
+			5: 'IMAGE_SYM_CLASS_EXTERNAL_DEF',
+			6: 'IMAGE_SYM_CLASS_LABEL',
+			7: 'IMAGE_SYM_CLASS_UNDEFINED_LABEL',
+			8: 'IMAGE_SYM_CLASS_MEMBER_OF_STRUCT',
+			9: 'IMAGE_SYM_CLASS_ARGUMENT',
+			10: 'IMAGE_SYM_CLASS_STRUCT_TAG',
+			11: 'IMAGE_SYM_CLASS_MEMBER_OF_UNION',
+			12: 'IMAGE_SYM_CLASS_UNION_TAG',
+			13: 'IMAGE_SYM_CLASS_TYPE_DEFINITION',
+			14: 'IMAGE_SYM_CLASS_UNDEFINED_STATIC',
+			15: 'IMAGE_SYM_CLASS_ENUM_TAG',
+			16: 'IMAGE_SYM_CLASS_MEMBER_OF_ENUM',
+			17: 'IMAGE_SYM_CLASS_REGISTER_PARAM',
+			18: 'IMAGE_SYM_CLASS_BIT_FIELD',
+			100: 'IMAGE_SYM_CLASS_BLOCK',
+			101: 'IMAGE_SYM_CLASS_FUNCTION',
+			102: 'IMAGE_SYM_CLASS_END_OF_STRUCT',
+			103: 'IMAGE_SYM_CLASS_FILE',
+			104: 'IMAGE_SYM_CLASS_SECTION',
+			105: 'IMAGE_SYM_CLASS_WEAK_EXTERNAL',
+			107: 'IMAGE_SYM_CLASS_CLR_TOKEN',
+		})),
+		('NumberOfAuxSymbols',  ('scalar',  '<B',  '%d'  )),
+		('AuxillaryData', ('blob', '* NumberOfAuxSymbols 18')),
+	])))),
+])
+
+#
+# Definition Interpreter
+#
+
+import sys, types, struct, re
+
+Input = None
+Stack = []
+Fields = {}
+
+Indent = 0
+NewLine = True
+
+def indent():
+	global Indent
+	Indent += 1
+
+def dedent():
+	global Indent
+	Indent -= 1
+
+def write(input):
+	global NewLine
+	output = ""
+	
+	for char in input:
+		
+		if NewLine:
+			output += Indent * '  '
+			NewLine = False
+			
+		output += char
+		
+		if char == '\n':
+			NewLine = True
+	
+	sys.stdout.write (output)
+
+def read(format):
+	return struct.unpack (format, Input.read(struct.calcsize(format)))
+
+def read_cstr ():
+	output = ""
+	while True:
+		char = Input.read (1)
+		if len (char) == 0:
+			raise RuntimeError ("EOF while reading cstr")
+		if char == '\0':
+			break
+		output += char
+	return output
+
+def push_pos(seek_to = None):
+	Stack [0:0] = [Input.tell ()]
+	if seek_to:
+		Input.seek (seek_to)
+
+def pop_pos():
+	assert(len (Stack) > 0)
+	Input.seek (Stack [0])
+	del Stack [0]
+
+def print_binary_data(size):
+	value = ""
+	while size > 0:
+		if size >= 16:
+			data = Input.read(16)
+			size -= 16
+		else:
+			data = Input.read(size)
+			size = 0
+		value += data
+		bytes = ""
+		text = ""
+		for index in xrange (16):
+			if index < len (data):
+				if index == 8:
+					bytes += "- "
+				ch = ord (data [index])
+				bytes += "%02X " % ch
+				if ch >= 0x20 and ch <= 0x7F:
+					text += data [index]
+				else:
+					text += "."
+			else:
+				if index == 8:
+					bytes += "  "
+				bytes += "   "
+		
+		write ("%s|%s|\n" % (bytes, text))
+	return value
+
+idlit = re.compile ("[a-zA-Z][a-zA-Z0-9_-]*")
+numlit = re.compile ("[0-9]+")
+
+def read_value(expr):
+
+	input = iter (expr.split ())
+	
+	def eval():
+		
+		token = input.next ()
+		
+		if expr == 'cstr':
+			return read_cstr ()
+		if expr == 'true':
+			return True
+		if expr == 'false':
+			return False
+		
+		if len (token) > 1 and token [0] in ('=', '@', '<', '!', '>'):
+			val = read(expr)
+			assert (len (val) == 1)
+			return val [0]
+		
+		if token == '+':
+			return eval () + eval ()
+		if token == '-':
+			return eval () - eval ()
+		if token == '*':
+			return eval () * eval ()
+		if token == '/':
+			return eval () / eval ()
+		
+		if idlit.match (token):
+			return Fields [token]
+		if numlit.match (token):
+			return int (token)
+		
+		raise RuntimeError ("unexpected token %s" % repr(token))
+	
+	value = eval ()
+	
+	try:
+		input.next ()
+	except StopIteration:
+		return value
+	raise RuntimeError("unexpected input at end of expression")
+
+def write_value(format,value):
+	format_type = type (format)
+	if format_type is types.StringType:
+		write (format%value)
+	elif format_type is types.FunctionType:
+		write_value (format (value), value)
+	elif format_type is types.TupleType:
+		Fields ['this'] = value
+		handle_element (format)
+	else:
+		raise RuntimeError("unexpected type: %s" % repr(format_type))
+
+def handle_scalar(entry):
+	iformat = entry [1]
+	oformat = entry [2]
+	
+	value = read_value (iformat)
+	
+	write_value (oformat, value)
+	
+	return value
+
+def handle_enum(entry):
+	iformat = entry [1]
+	oformat = entry [2]
+	definitions = entry [3]
+	
+	value = read_value (iformat)
+	
+	if type (definitions) is types.TupleType:
+		selector = read_value (definitions [0])
+		definitions = definitions [1] [selector]
+	
+	description = definitions[value] if value in definitions else "unknown"
+	
+	write ("%s (" % description)
+	write_value (oformat, value)
+	write (")")
+	
+	return value
+
+def handle_flags(entry):
+	iformat = entry [1]
+	oformat = entry [2]
+	definitions = entry [3]
+	
+	value = read_value (iformat)
+	
+	write_value (oformat, value)
+	
+	indent ()
+	for entry in definitions:
+		mask = entry [0]
+		name = entry [1]
+		if len (entry) == 3:
+			map = entry [2]
+			selection = value & mask
+			if selection in map:
+				write("\n%s" % map[selection])
+			else:
+				write("\n%s <%d>" % (name, selection))
+		elif len (entry) == 2:
+			if value & mask != 0:
+				write("\n%s" % name)
+	dedent ()
+	
+	return value
+
+def handle_struct(entry):
+	global Fields
+	members = entry [1]
+	
+	newFields = {}
+	
+	write ("{\n");
+	indent ()
+	
+	for member in members:
+		name = member [0]
+		type = member [1]
+		
+		write("%s = "%name.ljust(24))
+		
+		value = handle_element(type)
+		
+		write("\n")
+		
+		Fields [name] = value
+		newFields [name] = value
+	
+	dedent ()
+	write ("}")
+	
+	return newFields
+
+def handle_array(entry):
+	length = entry [1]
+	element = entry [2]
+	
+	newItems = []
+	
+	write ("[\n")
+	indent ()
+	
+	value = read_value (length)
+	
+	for index in xrange (value):
+		write ("%d = "%index)
+		value = handle_element(element)
+		write ("\n")
+		newItems.append (value)
+	
+	dedent ()
+	write ("]")
+	
+	return newItems
+
+def handle_byte_array(entry):
+	length = entry [1]
+	element = entry [2]
+	
+	newItems = []
+	
+	write ("[\n")
+	indent ()
+	
+	value = read_value (length)
+	end_of_array = Input.tell () + value
+
+	index = 0
+	while Input.tell () < end_of_array:
+		write ("%d = "%index)
+		value = handle_element(element)
+		write ("\n")
+		newItems.append (value)
+		index += 1
+	
+	dedent ()
+	write ("]")
+	
+	return newItems
+
+def handle_ptr(entry):
+	offset = entry[1]
+	element = entry [2]
+	
+	value = None
+	offset = read_value (offset)
+	
+	if offset != 0:
+		
+		push_pos (offset)
+		
+		value = handle_element (element)
+		
+		pop_pos ()
+	
+	else:
+		write ("None")
+	
+	return value
+
+def handle_blob(entry):
+	length = entry [1]
+	
+	write ("\n")
+	indent ()
+	
+	value = print_binary_data (read_value (length))
+	
+	dedent ()
+	
+	return value
+
+def handle_element(entry):
+	handlers = {
+		'struct':      handle_struct,
+		'scalar':      handle_scalar,
+		'enum':        handle_enum,
+		'flags':       handle_flags,
+		'ptr':         handle_ptr,
+		'blob':        handle_blob,
+		'array':       handle_array,
+		'byte-array':  handle_byte_array,
+	}
+	
+	if not entry [0] in handlers:
+		raise RuntimeError ("unexpected type '%s'" % str (entry[0]))
+	
+	return handlers [entry [0]] (entry)
+
+Input = open (sys.argv [1], "rb")
+try:
+	handle_element (file)
+finally:
+	Input.close ()
+	Input = None

Added: llvm/trunk/test/Scripts/coff-dump.py.bat
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Scripts/coff-dump.py.bat?rev=109494&view=auto
==============================================================================
--- llvm/trunk/test/Scripts/coff-dump.py.bat (added)
+++ llvm/trunk/test/Scripts/coff-dump.py.bat Tue Jul 27 01:46:15 2010
@@ -0,0 +1,4 @@
+ at echo off
+
+%PYTHON_EXECUTABLE% %LLVM_SRC_ROOT%\test\Scripts\coff-dump.py %1 %2 %3 %4 %5 %6 %7 %8 %9
+




From baldrick at free.fr  Tue Jul 27 01:53:14 2010
From: baldrick at free.fr (Duncan Sands)
Date: Tue, 27 Jul 2010 06:53:14 -0000
Subject: [llvm-commits] [llvm] r109495 - /llvm/trunk/lib/VMCore/Value.cpp
Message-ID: <20100727065314.D3E9E2A6C12C@llvm.org>

Author: baldrick
Date: Tue Jul 27 01:53:14 2010
New Revision: 109495

URL: http://llvm.org/viewvc/llvm-project?rev=109495&view=rev
Log:
After updating value handles for RAUW, check that no weak or tracking handles
are still on the list.  This might happen if a CallbackVH created some new value
handles for the old value when doing RAUW.  Barf if it occurs, since it is almost
certainly a mistake.

Modified:
    llvm/trunk/lib/VMCore/Value.cpp

Modified: llvm/trunk/lib/VMCore/Value.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Value.cpp?rev=109495&r1=109494&r2=109495&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Value.cpp (original)
+++ llvm/trunk/lib/VMCore/Value.cpp Tue Jul 27 01:53:14 2010
@@ -494,7 +494,7 @@
   // Note that we deliberately do not the support the case when dropping a value
   // handle results in a new value handle being permanently added to the list
   // (as might occur in theory for CallbackVH's): the new value handle will not
-  // be processed and the checking code will meet out righteous punishment if
+  // be processed and the checking code will mete out righteous punishment if
   // the handle is still present once we have finished processing all the other
   // value handles (it is fine to momentarily add then remove a value handle).
   for (ValueHandleBase Iterator(Assert, *Entry); Entry; Entry = Iterator.Next) {
@@ -577,6 +577,24 @@
       break;
     }
   }
+
+#ifndef NDEBUG
+  // If any new tracking or weak value handles were added while processing the
+  // list, then complain about it now.
+  if (Old->HasValueHandle)
+    for (Entry = pImpl->ValueHandles[Old]; Entry; Entry = Entry->Next)
+      switch (Entry->getKind()) {
+      case Tracking:
+      case Weak:
+        dbgs() << "After RAUW from " << *Old->getType() << " %"
+          << Old->getNameStr() << " to " << *New->getType() << " %"
+          << New->getNameStr() << "\n";
+        llvm_unreachable("A tracking or weak value handle still pointed to the"
+                         " old value!\n");
+      default:
+        break;
+      }
+#endif
 }
 
 /// ~CallbackVH. Empty, but defined here to avoid emitting the vtable




From evan.cheng at apple.com  Tue Jul 27 02:24:13 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 27 Jul 2010 00:24:13 -0700
Subject: [llvm-commits] [llvm] r109089 - in /llvm/trunk: docs/
 include/llvm/ include/llvm/Analysis/ include/llvm/Support/ lib/Analysis/
 test/Analysis/RegionInfo/
In-Reply-To: <20100722074632.575522A6C12D@llvm.org>
References: <20100722074632.575522A6C12D@llvm.org>
Message-ID: <B2707C0D-15AF-4D83-84DE-6D6D6EFD2496@apple.com>

Any chance you could templatify this so someone can add a MachineRegionInfo later? It would be useful for region based register allocation, etc.  

Thanks,

Evan

On Jul 22, 2010, at 12:46 AM, Tobias Grosser <grosser at fim.uni-passau.de> wrote:

> Author: grosser
> Date: Thu Jul 22 02:46:31 2010
> New Revision: 109089
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109089&view=rev
> Log:
> Add new RegionInfo pass.
> 
> The RegionInfo pass detects single entry single exit regions in a function,
> where a region is defined as any subgraph that is connected to the remaining
> graph at only two spots.
> Furthermore an hierarchical region tree is built.
> Use it by calling "opt -regions analyze" or "opt -view-regions".
> 
> Added:
>    llvm/trunk/include/llvm/Analysis/RegionInfo.h
>    llvm/trunk/include/llvm/Analysis/RegionIterator.h
>    llvm/trunk/include/llvm/Analysis/RegionPrinter.h
>    llvm/trunk/lib/Analysis/RegionInfo.cpp
>    llvm/trunk/lib/Analysis/RegionPrinter.cpp
>    llvm/trunk/test/Analysis/RegionInfo/
>    llvm/trunk/test/Analysis/RegionInfo/block_sort.ll
>    llvm/trunk/test/Analysis/RegionInfo/cond_loop.ll
>    llvm/trunk/test/Analysis/RegionInfo/condition_complicated.ll
>    llvm/trunk/test/Analysis/RegionInfo/condition_complicated_2.ll
>    llvm/trunk/test/Analysis/RegionInfo/condition_forward_edge.ll
>    llvm/trunk/test/Analysis/RegionInfo/condition_same_exit.ll
>    llvm/trunk/test/Analysis/RegionInfo/condition_simple.ll
>    llvm/trunk/test/Analysis/RegionInfo/dg.exp
>    llvm/trunk/test/Analysis/RegionInfo/exit_in_condition.ll
>    llvm/trunk/test/Analysis/RegionInfo/infinite_loop.ll
>    llvm/trunk/test/Analysis/RegionInfo/infinite_loop_2.ll
>    llvm/trunk/test/Analysis/RegionInfo/infinite_loop_3.ll
>    llvm/trunk/test/Analysis/RegionInfo/infinite_loop_4.ll
>    llvm/trunk/test/Analysis/RegionInfo/loop_with_condition.ll
>    llvm/trunk/test/Analysis/RegionInfo/loops_1.ll
>    llvm/trunk/test/Analysis/RegionInfo/loops_2.ll
>    llvm/trunk/test/Analysis/RegionInfo/mix_1.ll
>    llvm/trunk/test/Analysis/RegionInfo/multiple_exiting_edge.ll
>    llvm/trunk/test/Analysis/RegionInfo/nested_loops.ll
>    llvm/trunk/test/Analysis/RegionInfo/next.ll
>    llvm/trunk/test/Analysis/RegionInfo/paper.ll
>    llvm/trunk/test/Analysis/RegionInfo/two_loops_same_header.ll
> Modified:
>    llvm/trunk/docs/Passes.html
>    llvm/trunk/include/llvm/Analysis/Passes.h
>    llvm/trunk/include/llvm/LinkAllPasses.h
>    llvm/trunk/include/llvm/Support/GraphWriter.h
>    llvm/trunk/lib/Analysis/CMakeLists.txt
> 
> Modified: llvm/trunk/docs/Passes.html
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/Passes.html?rev=109089&r1=109088&r2=109089&view=diff
> ==============================================================================
> --- llvm/trunk/docs/Passes.html (original)
> +++ llvm/trunk/docs/Passes.html Thu Jul 22 02:46:31 2010
> @@ -120,6 +120,7 @@
> <tr><td><a href="#print-used-types">-print-used-types</a></td><td>Find Used Types</td></tr>
> <tr><td><a href="#profile-estimator">-profile-estimator</a></td><td>Estimate profiling information</td></tr>
> <tr><td><a href="#profile-loader">-profile-loader</a></td><td>Load profile information from llvmprof.out</td></tr>
> +<tr><td><a href="#regions">-regions</a></td><td>Detect single entry single exit regions in a function</td></tr>
> <tr><td><a href="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr>
> <tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr>
> <tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr>
> @@ -771,6 +772,17 @@
> <div class="doc_text">
>   <p>Pass that checks profiling information for plausibility.</p>
> </div>
> +<div class="doc_subsection">
> +  <a name="regions">-regions: Detect single entry single exit regions in a function</a>
> +</div>
> +<div class="doc_text">
> +  <p>
> +  The <code>RegionInfo</code> pass detects single entry single exit regions in a
> +  function, where a region is defined as any subgraph that is connected to the
> +  remaining graph at only two spots. Furthermore, an hierarchical region tree is
> +  built.
> +  </p>
> +</div>
> 
> <!-------------------------------------------------------------------------- -->
> <div class="doc_subsection">
> 
> Modified: llvm/trunk/include/llvm/Analysis/Passes.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/Passes.h?rev=109089&r1=109088&r2=109089&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/Passes.h (original)
> +++ llvm/trunk/include/llvm/Analysis/Passes.h Thu Jul 22 02:46:31 2010
> @@ -154,6 +154,13 @@
>   // print debug info intrinsics in human readable form
>   FunctionPass *createDbgInfoPrinterPass();
> 
> +  //===--------------------------------------------------------------------===//
> +  //
> +  // createRegionInfoPass - This pass finds all single entry single exit regions
> +  // in a function and builds the region hierarchy.
> +  //
> +  FunctionPass *createRegionInfoPass();
> +
>   // Print module-level debug info metadata in human-readable form.
>   ModulePass *createModuleDebugInfoPrinterPass();
> }
> 
> Added: llvm/trunk/include/llvm/Analysis/RegionInfo.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/RegionInfo.h?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/RegionInfo.h (added)
> +++ llvm/trunk/include/llvm/Analysis/RegionInfo.h Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,601 @@
> +//===- RegionInfo.h - SESE region analysis ----------------------*- C++ -*-===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +//
> +// Calculate a program structure tree built out of single entry single exit
> +// regions.
> +// The basic ideas are taken from "The Program Structure Tree - Richard Johnson,
> +// David Pearson, Keshav Pingali - 1994", however enriched with ideas from "The
> +// Refined Process Structure Tree - Jussi Vanhatalo, Hagen Voelyer, Jana
> +// Koehler - 2009".
> +// The algorithm to calculate these data structures however is completely
> +// different, as it takes advantage of existing information already available
> +// in (Post)dominace tree and dominance frontier passes. This leads to a simpler
> +// and in practice hopefully better performing algorithm. The runtime of the
> +// algorithms described in the papers above are both linear in graph size,
> +// O(V+E), whereas this algorithm is not, as the dominance frontier information
> +// itself is not, but in practice runtime seems to be in the order of magnitude
> +// of dominance tree calculation.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#ifndef LLVM_ANALYSIS_REGION_INFO_H
> +#define LLVM_ANALYSIS_REGION_INFO_H
> +
> +#include "llvm/ADT/PointerIntPair.h"
> +#include "llvm/Analysis/Dominators.h"
> +#include "llvm/Analysis/PostDominators.h"
> +#include "llvm/Support/Allocator.h"
> +
> +namespace llvm {
> +
> +class Region;
> +class RegionInfo;
> +class raw_ostream;
> +
> +/// @brief Marker class to iterate over the elements of a Region in flat mode.
> +///
> +/// The class is used to either iterate in Flat mode or by not using it to not
> +/// iterate in Flat mode.  During a Flat mode iteration all Regions are entered
> +/// and the iteration returns every BasicBlock.  If the Flat mode is not
> +/// selected for SubRegions just one RegionNode containing the subregion is
> +/// returned.
> +template <class GraphType>
> +class FlatIt {};
> +
> +/// @brief A RegionNode represents a subregion or a BasicBlock that is part of a
> +/// Region.
> +class RegionNode {
> +  // DO NOT IMPLEMENT
> +  RegionNode(const RegionNode &);
> +  // DO NOT IMPLEMENT
> +  const RegionNode &operator=(const RegionNode &);
> +
> +  /// This is the entry basic block that starts this region node.  If this is a
> +  /// BasicBlock RegionNode, then entry is just the basic block, that this
> +  /// RegionNode represents.  Otherwise it is the entry of this (Sub)RegionNode.
> +  ///
> +  /// In the BBtoRegionNode map of the parent of this node, BB will always map
> +  /// to this node no matter which kind of node this one is.
> +  ///
> +  /// The node can hold either a Region or a BasicBlock.
> +  /// Use one bit to save, if this RegionNode is a subregion or BasicBlock
> +  /// RegionNode.
> +  PointerIntPair<BasicBlock*, 1, bool> entry;
> +
> +protected:
> +  /// @brief The parent Region of this RegionNode.
> +  /// @see getParent()
> +  Region* parent;
> +
> +public:
> +  /// @brief Create a RegionNode.
> +  ///
> +  /// @param Parent      The parent of this RegionNode.
> +  /// @param Entry       The entry BasicBlock of the RegionNode.  If this
> +  ///                    RegionNode represents a BasicBlock, this is the
> +  ///                    BasicBlock itself.  If it represents a subregion, this
> +  ///                    is the entry BasicBlock of the subregion.
> +  /// @param isSubRegion If this RegionNode represents a SubRegion.
> +  inline RegionNode(Region* Parent, BasicBlock* Entry, bool isSubRegion = 0)
> +    : entry(Entry, isSubRegion), parent(Parent) {}
> +
> +  /// @brief Get the parent Region of this RegionNode.
> +  ///
> +  /// The parent Region is the Region this RegionNode belongs to. If for
> +  /// example a BasicBlock is element of two Regions, there exist two
> +  /// RegionNodes for this BasicBlock. Each with the getParent() function
> +  /// pointing to the Region this RegionNode belongs to.
> +  ///
> +  /// @return Get the parent Region of this RegionNode.
> +  inline Region* getParent() const { return parent; }
> +
> +  /// @brief Get the entry BasicBlock of this RegionNode.
> +  ///
> +  /// If this RegionNode represents a BasicBlock this is just the BasicBlock
> +  /// itself, otherwise we return the entry BasicBlock of the Subregion
> +  ///
> +  /// @return The entry BasicBlock of this RegionNode.
> +  inline BasicBlock* getEntry() const { return entry.getPointer(); }
> +
> +  /// @brief Get the content of this RegionNode.
> +  ///
> +  /// This can be either a BasicBlock or a subregion. Before calling getNodeAs()
> +  /// check the type of the content with the isSubRegion() function call.
> +  ///
> +  /// @return The content of this RegionNode.
> +  template<class T>
> +  inline T* getNodeAs() const;
> +
> +  /// @brief Is this RegionNode a subregion?
> +  ///
> +  /// @return True if it contains a subregion. False if it contains a
> +  ///         BasicBlock.
> +  inline bool isSubRegion() const {
> +    return entry.getInt();
> +  }
> +};
> +
> +/// Print a RegionNode.
> +inline raw_ostream &operator<<(raw_ostream &OS, const RegionNode &Node);
> +
> +template<>
> +inline BasicBlock* RegionNode::getNodeAs<BasicBlock>() const {
> +  assert(!isSubRegion() && "This is not a BasicBlock RegionNode!");
> +  return getEntry();
> +}
> +
> +template<>
> +inline Region* RegionNode::getNodeAs<Region>() const {
> +  assert(isSubRegion() && "This is not a subregion RegionNode!");
> +  return reinterpret_cast<Region*>(const_cast<RegionNode*>(this));
> +}
> +
> +//===----------------------------------------------------------------------===//
> +/// @brief A single entry single exit Region.
> +///
> +/// A Region is a connected subgraph of a control flow graph that has exactly
> +/// two connections to the remaining graph. It can be used to analyze or
> +/// optimize parts of the control flow graph.
> +///
> +/// A <em> simple Region </em> is connected to the remaing graph by just two
> +/// edges. One edge entering the Region and another one leaving the Region.
> +///
> +/// An <em> extended Region </em> (or just Region) is a subgraph that can be
> +/// transform into a simple Region. The transformation is done by adding
> +/// BasicBlocks that merge several entry or exit edges so that after the merge
> +/// just one entry and one exit edge exists.
> +///
> +/// The \e Entry of a Region is the first BasicBlock that is passed after
> +/// entering the Region. It is an element of the Region. The entry BasicBlock
> +/// dominates all BasicBlocks in the Region.
> +///
> +/// The \e Exit of a Region is the first BasicBlock that is passed after
> +/// leaving the Region. It is not an element of the Region. The exit BasicBlock,
> +/// postdominates all BasicBlocks in the Region.
> +///
> +/// A <em> canonical Region </em> cannot be constructed by combining smaller
> +/// Regions.
> +///
> +/// Region A is the \e parent of Region B, if B is completely contained in A.
> +///
> +/// Two canonical Regions either do not intersect at all or one is
> +/// the parent of the other.
> +///
> +/// The <em> Program Structure Tree</em> is a graph (V, E) where V is the set of
> +/// Regions in the control flow graph and E is the \e parent relation of these
> +/// Regions.
> +///
> +/// Example:
> +///
> +/// \verbatim
> +/// A simple control flow graph, that contains two regions.
> +///
> +///        1
> +///       / |
> +///      2   |
> +///     / \   3
> +///    4   5  |
> +///    |   |  |
> +///    6   7  8
> +///     \  | /
> +///      \ |/       Region A: 1 -> 9 {1,2,3,4,5,6,7,8}
> +///        9        Region B: 2 -> 9 {2,4,5,6,7}
> +/// \endverbatim
> +///
> +/// You can obtain more examples by either calling
> +///
> +/// <tt> "opt -regions -analyze anyprogram.ll" </tt>
> +/// or
> +/// <tt> "opt -view-regions-only anyprogram.ll" </tt>
> +///
> +/// on any LLVM file you are interested in.
> +///
> +/// The first call returns a textual representation of the program structure
> +/// tree, the second one creates a graphical representation using graphviz.
> +class Region : public RegionNode {
> +  friend class RegionInfo;
> +  // DO NOT IMPLEMENT
> +  Region(const Region &);
> +  // DO NOT IMPLEMENT
> +  const Region &operator=(const Region &);
> +
> +  // Information necessary to manage this Region.
> +  RegionInfo* RI;
> +  DominatorTree *DT;
> +
> +  // The exit BasicBlock of this region.
> +  // (The entry BasicBlock is part of RegionNode)
> +  BasicBlock *exit;
> +
> +  typedef std::vector<Region*> RegionSet;
> +
> +  // The subregions of this region.
> +  RegionSet children;
> +
> +  typedef std::map<BasicBlock*, RegionNode*> BBNodeMapT;
> +
> +  // Save the BasicBlock RegionNodes that are element of this Region.
> +  mutable BBNodeMapT BBNodeMap;
> +
> +  /// verifyBBInRegion - Check if a BB is in this Region. This check also works
> +  /// if the region is incorrectly built. (EXPENSIVE!)
> +  void verifyBBInRegion(BasicBlock* BB) const;
> +
> +  /// verifyWalk - Walk over all the BBs of the region starting from BB and
> +  /// verify that all reachable basic blocks are elements of the region.
> +  /// (EXPENSIVE!)
> +  void verifyWalk(BasicBlock* BB, std::set<BasicBlock*>* visitedBB) const;
> +
> +  /// verifyRegionNest - Verify if the region and its children are valid
> +  /// regions (EXPENSIVE!)
> +  void verifyRegionNest() const;
> +
> +public:
> +  /// @brief Create a new region.
> +  ///
> +  /// @param Entry  The entry basic block of the region.
> +  /// @param Exit   The exit basic block of the region.
> +  /// @param RI     The region info object that is managing this region.
> +  /// @param DT     The dominator tree of the current function.
> +  /// @param Parent The surrounding region or NULL if this is a top level
> +  ///               region.
> +  Region(BasicBlock *Entry, BasicBlock *Exit, RegionInfo* RI,
> +         DominatorTree *DT, Region *Parent = 0);
> +
> +  /// Delete the Region and all its subregions.
> +  ~Region();
> +
> +  /// @brief Get the entry BasicBlock of the Region.
> +  /// @return The entry BasicBlock of the region.
> +  BasicBlock *getEntry() const { return RegionNode::getEntry(); }
> +
> +  /// @brief Get the exit BasicBlock of the Region.
> +  /// @return The exit BasicBlock of the Region, NULL if this is the TopLevel
> +  ///         Region.
> +  BasicBlock *getExit() const { return exit; }
> +
> +  /// @brief Get the parent of the Region.
> +  /// @return The parent of the Region or NULL if this is a top level
> +  ///         Region.
> +  Region *getParent() const { return RegionNode::getParent(); }
> +
> +  /// @brief Get the RegionNode representing the current Region.
> +  /// @return The RegionNode representing the current Region.
> +  RegionNode* getNode() const {
> +    return const_cast<RegionNode*>(reinterpret_cast<const RegionNode*>(this));
> +  }
> +
> +  /// @brief Get the nesting level of this Region.
> +  ///
> +  /// An toplevel Region has depth 0.
> +  ///
> +  /// @return The depth of the region.
> +  unsigned getDepth() const;
> +
> +  /// @brief Is this a simple region?
> +  ///
> +  /// A region is simple if it has exactly one exit and one entry edge.
> +  ///
> +  /// @return True if the Region is simple.
> +  bool isSimple() const;
> +
> +  /// @brief Returns the name of the Region.
> +  /// @return The Name of the Region.
> +  std::string getNameStr() const {
> +    std::string exitName;
> +
> +    if (getExit())
> +      exitName = getExit()->getNameStr();
> +    else
> +      exitName = "<Function Return>";
> +
> +    return getEntry()->getNameStr() + " => " + exitName;
> +  }
> +
> +  /// @brief Return the RegionInfo object, that belongs to this Region.
> +  RegionInfo *getRegionInfo() const {
> +    return RI;
> +  }
> +
> +  /// @brief Print the region.
> +  ///
> +  /// @param OS The output stream the Region is printed to.
> +  /// @param printTree Print also the tree of subregions.
> +  /// @param level The indentation level used for printing.
> +  void print(raw_ostream& OS, bool printTree = true, unsigned level = 0) const;
> +
> +  /// @brief Print the region to stderr.
> +  void dump() const;
> +
> +  /// @brief Check if the region contains a BasicBlock.
> +  ///
> +  /// @param BB The BasicBlock that might be contained in this Region.
> +  /// @return True if the block is contained in the region otherwise false.
> +  bool contains(const BasicBlock *BB) const;
> +
> +  /// @brief Check if the region contains another region.
> +  ///
> +  /// @param SubRegion The region that might be contained in this Region.
> +  /// @return True if SubRegion is contained in the region otherwise false.
> +  bool contains(const Region *SubRegion) const {
> +    // Toplevel Region.
> +    if (!getExit())
> +      return true;
> +
> +    return contains(SubRegion->getEntry())
> +      && (contains(SubRegion->getExit()) || SubRegion->getExit() == getExit());
> +  }
> +
> +  /// @brief Check if the region contains an Instruction.
> +  ///
> +  /// @param Inst The Instruction that might be contained in this region.
> +  /// @return True if the Instruction is contained in the region otherwise false.
> +  bool contains(const Instruction *Inst) const {
> +    return contains(Inst->getParent());
> +  }
> +
> +  /// @brief Get the subregion that starts at a BasicBlock
> +  ///
> +  /// @param BB The BasicBlock the subregion should start.
> +  /// @return The Subregion if available, otherwise NULL.
> +  Region* getSubRegionNode(BasicBlock *BB) const;
> +
> +  /// @brief Get the RegionNode for a BasicBlock
> +  ///
> +  /// @param BB The BasicBlock at which the RegionNode should start.
> +  /// @return If available, the RegionNode that represents the subregion
> +  ///         starting at BB. If no subregion starts at BB, the RegionNode
> +  ///         representing BB.
> +  RegionNode* getNode(BasicBlock *BB) const;
> +
> +  /// @brief Get the BasicBlock RegionNode for a BasicBlock
> +  ///
> +  /// @param BB The BasicBlock for which the RegionNode is requested.
> +  /// @return The RegionNode representing the BB.
> +  RegionNode* getBBNode(BasicBlock *BB) const;
> +
> +  /// @brief Add a new subregion to this Region.
> +  ///
> +  /// @param SubRegion The new subregion that will be added.
> +  void addSubRegion(Region *SubRegion);
> +
> +  /// @brief Remove a subregion from this Region.
> +  ///
> +  /// The subregion is not deleted, as it will probably be inserted into another
> +  /// region.
> +  /// @param SubRegion The SubRegion that will be removed.
> +  Region *removeSubRegion(Region *SubRegion);
> +
> +  /// @brief Move all direct child nodes of this Region to another Region.
> +  ///
> +  /// @param To The Region the child nodes will be transfered to.
> +  void transferChildrenTo(Region *To);
> +
> +  /// @brief Verify if the region is a correct region.
> +  ///
> +  /// Check if this is a correctly build Region. This is an expensive check, as
> +  /// the complete CFG of the Region will be walked.
> +  void verifyRegion() const;
> +
> +  /// @brief Clear the cache for BB RegionNodes.
> +  ///
> +  /// After calling this function the BasicBlock RegionNodes will be stored at
> +  /// different memory locations. RegionNodes obtained before this function is
> +  /// called are therefore not comparable to RegionNodes abtained afterwords.
> +  void clearNodeCache();
> +
> +  /// @name Subregion Iterators
> +  ///
> +  /// These iterators iterator over all subregions of this Region.
> +  //@{
> +  typedef RegionSet::iterator iterator;
> +  typedef RegionSet::const_iterator const_iterator;
> +
> +  iterator begin() { return children.begin(); }
> +  iterator end() { return children.end(); }
> +
> +  const_iterator begin() const { return children.begin(); }
> +  const_iterator end() const { return children.end(); }
> +  //@}
> +
> +  /// @name BasicBlock Iterators
> +  ///
> +  /// These iterators iterate over all BasicBlock RegionNodes that are
> +  /// contained in this Region. The iterator also iterates over BasicBlocks
> +  /// that are elements of a subregion of this Region. It is therefore called a
> +  /// flat iterator.
> +  //@{
> +  typedef df_iterator<RegionNode*, SmallPtrSet<RegionNode*, 8>, false,
> +                      GraphTraits<FlatIt<RegionNode*> > > block_iterator;
> +
> +  typedef df_iterator<const RegionNode*, SmallPtrSet<const RegionNode*, 8>,
> +                      false, GraphTraits<FlatIt<const RegionNode*> > >
> +            const_block_iterator;
> +
> +  block_iterator block_begin();
> +  block_iterator block_end();
> +
> +  const_block_iterator block_begin() const;
> +  const_block_iterator block_end() const;
> +  //@}
> +
> +  /// @name Element Iterators
> +  ///
> +  /// These iterators iterate over all BasicBlock and subregion RegionNodes that
> +  /// are direct children of this Region. It does not iterate over any
> +  /// RegionNodes that are also element of a subregion of this Region.
> +  //@{
> +  typedef df_iterator<RegionNode*, SmallPtrSet<RegionNode*, 8>, false,
> +                      GraphTraits<RegionNode*> > element_iterator;
> +
> +  typedef df_iterator<const RegionNode*, SmallPtrSet<const RegionNode*, 8>,
> +                      false, GraphTraits<const RegionNode*> >
> +            const_element_iterator;
> +
> +  element_iterator element_begin();
> +  element_iterator element_end();
> +
> +  const_element_iterator element_begin() const;
> +  const_element_iterator element_end() const;
> +  //@}
> +};
> +
> +//===----------------------------------------------------------------------===//
> +/// @brief Analysis that detects all canonical Regions.
> +///
> +/// The RegionInfo pass detects all canonical regions in a function. The Regions
> +/// are connected using the parent relation. This builds a Program Structure
> +/// Tree.
> +class RegionInfo : public FunctionPass {
> +  typedef DenseMap<BasicBlock*,BasicBlock*> BBtoBBMap;
> +  typedef DenseMap<BasicBlock*, Region*> BBtoRegionMap;
> +  typedef SmallPtrSet<Region*, 4> RegionSet;
> +
> +  // DO NOT IMPLEMENT
> +  RegionInfo(const RegionInfo &);
> +  // DO NOT IMPLEMENT
> +  const RegionInfo &operator=(const RegionInfo &);
> +
> +  DominatorTree *DT;
> +  PostDominatorTree *PDT;
> +  DominanceFrontier *DF;
> +
> +  /// The top level region.
> +  Region *TopLevelRegion;
> +
> +  /// Map every BB to the smallest region, that contains BB.
> +  BBtoRegionMap BBtoRegion;
> +
> +  // isCommonDomFrontier - Returns true if BB is in the dominance frontier of
> +  // entry, because it was inherited from exit. In the other case there is an
> +  // edge going from entry to BB without passing exit.
> +  bool isCommonDomFrontier(BasicBlock* BB, BasicBlock* entry,
> +                           BasicBlock* exit) const;
> +
> +  // isRegion - Check if entry and exit surround a valid region, based on
> +  // dominance tree and dominance frontier.
> +  bool isRegion(BasicBlock* entry, BasicBlock* exit) const;
> +
> +  // insertShortCut - Saves a shortcut pointing from entry to exit.
> +  // This function may extend this shortcut if possible.
> +  void insertShortCut(BasicBlock* entry, BasicBlock* exit,
> +                      BBtoBBMap* ShortCut) const;
> +
> +  // getNextPostDom - Returns the next BB that postdominates N, while skipping
> +  // all post dominators that cannot finish a canonical region.
> +  DomTreeNode *getNextPostDom(DomTreeNode* N, BBtoBBMap *ShortCut) const;
> +
> +  // isTrivialRegion - A region is trivial, if it contains only one BB.
> +  bool isTrivialRegion(BasicBlock *entry, BasicBlock *exit) const;
> +
> +  // createRegion - Creates a single entry single exit region.
> +  Region *createRegion(BasicBlock *entry, BasicBlock *exit);
> +
> +  // findRegionsWithEntry - Detect all regions starting with bb 'entry'.
> +  void findRegionsWithEntry(BasicBlock *entry, BBtoBBMap *ShortCut);
> +
> +  // scanForRegions - Detects regions in F.
> +  void scanForRegions(Function &F, BBtoBBMap *ShortCut);
> +
> +  // getTopMostParent - Get the top most parent with the same entry block.
> +  Region *getTopMostParent(Region *region);
> +
> +  // buildRegionsTree - build the region hierarchy after all region detected.
> +  void buildRegionsTree(DomTreeNode *N, Region *region);
> +
> +  // Calculate - detecte all regions in function and build the region tree.
> +  void Calculate(Function& F);
> +
> +  void releaseMemory();
> +
> +  // updateStatistics - Update statistic about created regions.
> +  void updateStatistics(Region *R);
> +
> +  // isSimple - Check if a region is a simple region with exactly one entry
> +  // edge and exactly one exit edge.
> +  bool isSimple(Region* R) const;
> +
> +public:
> +  static char ID;
> +  explicit RegionInfo();
> +
> +  ~RegionInfo();
> +
> +  /// @name FunctionPass interface
> +  //@{
> +  virtual bool runOnFunction(Function &F);
> +  virtual void getAnalysisUsage(AnalysisUsage &AU) const;
> +  virtual void print(raw_ostream &OS, const Module *) const;
> +  virtual void verifyAnalysis() const;
> +  //@}
> +
> +  /// @brief Get the smallest region that contains a BasicBlock.
> +  ///
> +  /// @param BB The basic block.
> +  /// @return The smallest region, that contains BB or NULL, if there is no
> +  /// region containing BB.
> +  Region *getRegionFor(BasicBlock *BB) const;
> +
> +  /// @brief A shortcut for getRegionFor().
> +  ///
> +  /// @param BB The basic block.
> +  /// @return The smallest region, that contains BB or NULL, if there is no
> +  /// region containing BB.
> +  Region *operator[](BasicBlock *BB) const;
> +
> +  /// @brief Find the smallest region that contains two regions.
> +  ///
> +  /// @param A The first region.
> +  /// @param B The second region.
> +  /// @return The smallest region containing A and B.
> +  Region *getCommonRegion(Region* A, Region *B) const;
> +
> +  /// @brief Find the smallest region that contains two basic blocks.
> +  ///
> +  /// @param A The first basic block.
> +  /// @param B The second basic block.
> +  /// @return The smallest region that contains A and B.
> +  Region* getCommonRegion(BasicBlock* A, BasicBlock *B) const {
> +    return getCommonRegion(getRegionFor(A), getRegionFor(B));
> +  }
> +
> +  /// @brief Find the smallest region that contains a set of regions.
> +  ///
> +  /// @param Regions A vector of regions.
> +  /// @return The smallest region that contains all regions in Regions.
> +  Region* getCommonRegion(SmallVectorImpl<Region*> &Regions) const;
> +
> +  /// @brief Find the smallest region that contains a set of basic blocks.
> +  ///
> +  /// @param BBs A vector of basic blocks.
> +  /// @return The smallest region that contains all basic blocks in BBS.
> +  Region* getCommonRegion(SmallVectorImpl<BasicBlock*> &BBs) const;
> +
> +  Region *getTopLevelRegion() const {
> +    return TopLevelRegion;
> +  }
> +
> +  /// @brief Clear the Node Cache for all Regions.
> +  ///
> +  /// @see Region::clearNodeCache()
> +  void clearNodeCache() {
> +    if (TopLevelRegion)
> +      TopLevelRegion->clearNodeCache();
> +  }
> +};
> +
> +inline raw_ostream &operator<<(raw_ostream &OS, const RegionNode &Node) {
> +  if (Node.isSubRegion())
> +    return OS << Node.getNodeAs<Region>()->getNameStr();
> +  else
> +    return OS << Node.getNodeAs<BasicBlock>()->getNameStr();
> +}
> +} // End llvm namespace
> +#endif
> +
> 
> Added: llvm/trunk/include/llvm/Analysis/RegionIterator.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/RegionIterator.h?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/RegionIterator.h (added)
> +++ llvm/trunk/include/llvm/Analysis/RegionIterator.h Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,342 @@
> +//===- RegionIterator.h - Iterators to iteratate over Regions ---*- C++ -*-===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +// This file defines the iterators to iterate over the elements of a Region.
> +//===----------------------------------------------------------------------===//
> +#ifndef LLVM_ANALYSIS_REGION_ITERATOR_H
> +#define LLVM_ANALYSIS_REGION_ITERATOR_H
> +
> +#include "llvm/ADT/GraphTraits.h"
> +#include "llvm/ADT/SmallPtrSet.h"
> +#include "llvm/ADT/PointerIntPair.h"
> +#include "llvm/Analysis/RegionInfo.h"
> +#include "llvm/Support/CFG.h"
> +#include "llvm/Support/raw_ostream.h"
> +
> +namespace llvm {
> +//===----------------------------------------------------------------------===//
> +/// @brief Hierachical RegionNode successor iterator.
> +///
> +/// This iterator iterates over all successors of a RegionNode.
> +///
> +/// For a BasicBlock RegionNode it skips all BasicBlocks that are not part of
> +/// the parent Region.  Furthermore for BasicBlocks that start a subregion, a
> +/// RegionNode representing the subregion is returned.
> +///
> +/// For a subregion RegionNode there is just one successor. The RegionNode
> +/// representing the exit of the subregion.
> +template<class NodeType>
> +class RNSuccIterator : public std::iterator<std::forward_iterator_tag,
> +                                           NodeType, ptrdiff_t>
> +{
> +  typedef std::iterator<std::forward_iterator_tag, NodeType, ptrdiff_t> super;
> +  // The iterator works in two modes, bb mode or region mode.
> +  enum ItMode{
> +    // In BB mode it returns all successors of this BasicBlock as its
> +    // successors.
> +    ItBB,
> +    // In region mode there is only one successor, thats the regionnode mapping
> +    // to the exit block of the regionnode
> +    ItRgBegin, // At the beginning of the regionnode successor.
> +    ItRgEnd    // At the end of the regionnode successor.
> +  };
> +
> +  // Use two bit to represent the mode iterator.
> +  PointerIntPair<NodeType*, 2, enum ItMode> Node;
> +
> +  // The block successor iterator.
> +  succ_iterator BItor;
> +
> +  // advanceRegionSucc - A region node has only one successor. It reaches end
> +  // once we advance it.
> +  void advanceRegionSucc() {
> +    assert(Node.getInt() == ItRgBegin && "Cannot advance region successor!");
> +    Node.setInt(ItRgEnd);
> +  }
> +
> +  NodeType* getNode() const{ return Node.getPointer(); }
> +
> +  // isRegionMode - Is the current iterator in region mode?
> +  bool isRegionMode() const { return Node.getInt() != ItBB; }
> +
> +  // Get the immediate successor. This function may return a Basic Block
> +  // RegionNode or a subregion RegionNode.
> +  RegionNode* getISucc(BasicBlock* BB) const {
> +    RegionNode *succ;
> +    succ = getNode()->getParent()->getNode(BB);
> +    assert(succ && "BB not in Region or entered subregion!");
> +    return succ;
> +  }
> +
> +  // getRegionSucc - Return the successor basic block of a SubRegion RegionNode.
> +  inline BasicBlock* getRegionSucc() const {
> +    assert(Node.getInt() == ItRgBegin && "Cannot get the region successor!");
> +    return getNode()->template getNodeAs<Region>()->getExit();
> +  }
> +
> +  // isExit - Is this the exit BB of the Region?
> +  inline bool isExit(BasicBlock* BB) const {
> +    return getNode()->getParent()->getExit() == BB;
> +  }
> +public:
> +  typedef RNSuccIterator<NodeType> Self;
> +
> +  typedef typename super::pointer pointer;
> +
> +  /// @brief Create begin iterator of a RegionNode.
> +  inline RNSuccIterator(NodeType* node)
> +    : Node(node, node->isSubRegion() ? ItRgBegin : ItBB),
> +    BItor(succ_begin(node->getEntry())) {
> +
> +
> +    // Skip the exit block
> +    if (!isRegionMode())
> +      while (succ_end(node->getEntry()) != BItor && isExit(*BItor))
> +        ++BItor;
> +
> +    if (isRegionMode() && isExit(getRegionSucc()))
> +      advanceRegionSucc();
> +  }
> +
> +  /// @brief Create an end iterator.
> +  inline RNSuccIterator(NodeType* node, bool)
> +    : Node(node, node->isSubRegion() ? ItRgEnd : ItBB),
> +    BItor(succ_end(node->getEntry())) {}
> +
> +  inline bool operator==(const Self& x) const {
> +    assert(isRegionMode() == x.isRegionMode() && "Broken iterator!");
> +    if (isRegionMode())
> +      return Node.getInt() == x.Node.getInt();
> +    else
> +      return BItor == x.BItor;
> +  }
> +
> +  inline bool operator!=(const Self& x) const { return !operator==(x); }
> +
> +  inline pointer operator*() const {
> +    BasicBlock* BB = isRegionMode() ? getRegionSucc() : *BItor;
> +    assert(!isExit(BB) && "Iterator out of range!");
> +    return getISucc(BB);
> +  }
> +
> +  inline Self& operator++() {
> +    if(isRegionMode()) {
> +      // The Region only has 1 successor.
> +      advanceRegionSucc();
> +    } else {
> +      // Skip the exit.
> +      do
> +        ++BItor;
> +      while (BItor != succ_end(getNode()->getEntry())
> +          && isExit(*BItor));
> +    }
> +    return *this;
> +  }
> +
> +  inline Self operator++(int) {
> +    Self tmp = *this;
> +    ++*this;
> +    return tmp;
> +  }
> +
> +  inline const Self &operator=(const Self &I) {
> +    if (this != &I) {
> +      assert(getNode()->getParent() == I.getNode()->getParent()
> +             && "Cannot assign iterators of two different regions!");
> +      Node = I.Node;
> +      BItor = I.BItor;
> +    }
> +    return *this;
> +  }
> +};
> +
> +
> +//===----------------------------------------------------------------------===//
> +/// @brief Flat RegionNode iterator.
> +///
> +/// The Flat Region iterator will iterate over all BasicBlock RegionNodes that
> +/// are contained in the Region and its subregions. This is close to a virtual
> +/// control flow graph of the Region.
> +template<class NodeType>
> +class RNSuccIterator<FlatIt<NodeType> >
> +  : public std::iterator<std::forward_iterator_tag, NodeType, ptrdiff_t>
> +{
> +  typedef std::iterator<std::forward_iterator_tag, NodeType, ptrdiff_t> super;
> +  NodeType* Node;
> +  succ_iterator Itor;
> +
> +public:
> +  typedef RNSuccIterator<FlatIt<NodeType> > Self;
> +  typedef typename super::pointer pointer;
> +
> +  /// @brief Create the iterator from a RegionNode.
> +  ///
> +  /// Note that the incoming node must be a bb node, otherwise it will trigger
> +  /// an assertion when we try to get a BasicBlock.
> +  inline RNSuccIterator(NodeType* node) : Node(node),
> +    Itor(succ_begin(node->getEntry())) {
> +      assert(!Node->isSubRegion()
> +             && "Subregion node not allowed in flat iterating mode!");
> +      assert(Node->getParent() && "A BB node must have a parent!");
> +
> +      // Skip the exit block of the iterating region.
> +      while (succ_end(Node->getEntry()) != Itor
> +          && Node->getParent()->getExit() == *Itor)
> +        ++Itor;
> +  }
> +  /// @brief Create an end iterator
> +  inline RNSuccIterator(NodeType* node, bool) : Node(node),
> +    Itor(succ_end(node->getEntry())) {
> +      assert(!Node->isSubRegion()
> +             && "Subregion node not allowed in flat iterating mode!");
> +  }
> +
> +  inline bool operator==(const Self& x) const {
> +    assert(Node->getParent() == x.Node->getParent()
> +           && "Cannot compare iterators of different regions!");
> +
> +    return Itor == x.Itor && Node == x.Node;
> +  }
> +
> +  inline bool operator!=(const Self& x) const { return !operator==(x); }
> +
> +  inline pointer operator*() const {
> +    BasicBlock* BB = *Itor;
> +
> +    // Get the iterating region.
> +    Region* Parent = Node->getParent();
> +
> +    // The only case that the successor reaches out of the region is it reaches
> +    // the exit of the region.
> +    assert(Parent->getExit() != BB && "iterator out of range!");
> +
> +    return Parent->getBBNode(BB);
> +  }
> +
> +  inline Self& operator++() {
> +    // Skip the exit block of the iterating region.
> +    do
> +      ++Itor;
> +    while (Itor != succ_end(Node->getEntry())
> +        && Node->getParent()->getExit() == *Itor);
> +
> +    return *this;
> +  }
> +
> +  inline Self operator++(int) {
> +    Self tmp = *this;
> +    ++*this;
> +    return tmp;
> +  }
> +
> +  inline const Self &operator=(const Self &I) {
> +    if (this != &I) {
> +      assert(Node->getParent() == I.Node->getParent()
> +             && "Cannot assign iterators to two different regions!");
> +      Node = I.Node;
> +      Itor = I.Itor;
> +    }
> +    return *this;
> +  }
> +};
> +
> +template<class NodeType>
> +inline RNSuccIterator<NodeType> succ_begin(NodeType* Node) {
> +  return RNSuccIterator<NodeType>(Node);
> +}
> +
> +template<class NodeType>
> +inline RNSuccIterator<NodeType> succ_end(NodeType* Node) {
> +  return RNSuccIterator<NodeType>(Node, true);
> +}
> +
> +//===--------------------------------------------------------------------===//
> +// RegionNode GraphTraits specialization so the bbs in the region can be
> +// iterate by generic graph iterators.
> +//
> +// NodeT can either be region node or const region node, otherwise child_begin
> +// and child_end fail.
> +
> +#define RegionNodeGraphTraits(NodeT) \
> +  template<> struct GraphTraits<NodeT*> { \
> +  typedef NodeT NodeType; \
> +  typedef RNSuccIterator<NodeType> ChildIteratorType; \
> +  static NodeType *getEntryNode(NodeType* N) { return N; } \
> +  static inline ChildIteratorType child_begin(NodeType *N) { \
> +    return RNSuccIterator<NodeType>(N); \
> +  } \
> +  static inline ChildIteratorType child_end(NodeType *N) { \
> +    return RNSuccIterator<NodeType>(N, true); \
> +  } \
> +}; \
> +template<> struct GraphTraits<FlatIt<NodeT*> > { \
> +  typedef NodeT NodeType; \
> +  typedef RNSuccIterator<FlatIt<NodeT> > ChildIteratorType; \
> +  static NodeType *getEntryNode(NodeType* N) { return N; } \
> +  static inline ChildIteratorType child_begin(NodeType *N) { \
> +    return RNSuccIterator<FlatIt<NodeType> >(N); \
> +  } \
> +  static inline ChildIteratorType child_end(NodeType *N) { \
> +    return RNSuccIterator<FlatIt<NodeType> >(N, true); \
> +  } \
> +}
> +
> +#define RegionGraphTraits(RegionT, NodeT) \
> +template<> struct GraphTraits<RegionT*> \
> +  : public GraphTraits<NodeT*> { \
> +  typedef df_iterator<NodeType*> nodes_iterator; \
> +  static NodeType *getEntryNode(RegionT* R) { \
> +    return R->getNode(R->getEntry()); \
> +  } \
> +  static nodes_iterator nodes_begin(RegionT* R) { \
> +    return nodes_iterator::begin(getEntryNode(R)); \
> +  } \
> +  static nodes_iterator nodes_end(RegionT* R) { \
> +    return nodes_iterator::end(getEntryNode(R)); \
> +  } \
> +}; \
> +template<> struct GraphTraits<FlatIt<RegionT*> > \
> +  : public GraphTraits<FlatIt<NodeT*> > { \
> +  typedef df_iterator<NodeType*, SmallPtrSet<NodeType*, 8>, false, \
> +  GraphTraits<FlatIt<NodeType*> > > nodes_iterator; \
> +  static NodeType *getEntryNode(RegionT* R) { \
> +    return R->getBBNode(R->getEntry()); \
> +  } \
> +  static nodes_iterator nodes_begin(RegionT* R) { \
> +    return nodes_iterator::begin(getEntryNode(R)); \
> +  } \
> +  static nodes_iterator nodes_end(RegionT* R) { \
> +    return nodes_iterator::end(getEntryNode(R)); \
> +  } \
> +}
> +
> +RegionNodeGraphTraits(RegionNode);
> +RegionNodeGraphTraits(const RegionNode);
> +
> +RegionGraphTraits(Region, RegionNode);
> +RegionGraphTraits(const Region, const RegionNode);
> +
> +template <> struct GraphTraits<RegionInfo*>
> +  : public GraphTraits<FlatIt<RegionNode*> > {
> +  typedef df_iterator<NodeType*, SmallPtrSet<NodeType*, 8>, false,
> +                      GraphTraits<FlatIt<NodeType*> > > nodes_iterator;
> +
> +  static NodeType *getEntryNode(RegionInfo *RI) {
> +    return GraphTraits<FlatIt<Region*> >::getEntryNode(RI->getTopLevelRegion());
> +  }
> +  static nodes_iterator nodes_begin(RegionInfo* RI) {
> +    return nodes_iterator::begin(getEntryNode(RI));
> +  }
> +  static nodes_iterator nodes_end(RegionInfo *RI) {
> +    return nodes_iterator::end(getEntryNode(RI));
> +  }
> +};
> +
> +} // End namespace llvm
> +
> +#endif
> 
> Added: llvm/trunk/include/llvm/Analysis/RegionPrinter.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/RegionPrinter.h?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/RegionPrinter.h (added)
> +++ llvm/trunk/include/llvm/Analysis/RegionPrinter.h Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,26 @@
> +//===-- RegionPrinter.h - Region printer external interface -----*- C++ -*-===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +//
> +// This file defines external functions that can be called to explicitly
> +// instantiate the region printer.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#ifndef LLVM_ANALYSIS_REGIONPRINTER_H
> +#define LLVM_ANALYSIS_REGIONPRINTER_H
> +
> +namespace llvm {
> +  class FunctionPass;
> +  FunctionPass *createRegionViewerPass();
> +  FunctionPass *createRegionOnlyViewerPass();
> +  FunctionPass *createRegionPrinterPass();
> +  FunctionPass *createRegionOnlyPrinterPass();
> +} // End llvm namespace
> +
> +#endif
> 
> Modified: llvm/trunk/include/llvm/LinkAllPasses.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/LinkAllPasses.h?rev=109089&r1=109088&r2=109089&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/LinkAllPasses.h (original)
> +++ llvm/trunk/include/llvm/LinkAllPasses.h Thu Jul 22 02:46:31 2010
> @@ -22,6 +22,7 @@
> #include "llvm/Analysis/Passes.h"
> #include "llvm/Analysis/PointerTracking.h"
> #include "llvm/Analysis/PostDominators.h"
> +#include "llvm/Analysis/RegionPrinter.h"
> #include "llvm/Analysis/ScalarEvolution.h"
> #include "llvm/Analysis/Lint.h"
> #include "llvm/Assembly/PrintModulePass.h"
> @@ -106,6 +107,11 @@
>       (void) llvm::createPostDomOnlyViewerPass();
>       (void) llvm::createPostDomViewerPass();
>       (void) llvm::createReassociatePass();
> +      (void) llvm::createRegionInfoPass();
> +      (void) llvm::createRegionOnlyPrinterPass();
> +      (void) llvm::createRegionOnlyViewerPass();
> +      (void) llvm::createRegionPrinterPass();
> +      (void) llvm::createRegionViewerPass();
>       (void) llvm::createSCCPPass();
>       (void) llvm::createScalarReplAggregatesPass();
>       (void) llvm::createSimplifyLibCallsPass();
> 
> Modified: llvm/trunk/include/llvm/Support/GraphWriter.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/GraphWriter.h?rev=109089&r1=109088&r2=109089&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Support/GraphWriter.h (original)
> +++ llvm/trunk/include/llvm/Support/GraphWriter.h Thu Jul 22 02:46:31 2010
> @@ -271,6 +271,12 @@
>       O << "[" << Attrs << "]";
>     O << ";\n";
>   }
> +
> +  /// getOStream - Get the raw output stream into the graph file. Useful to
> +  /// write fancy things using addCustomGraphFeatures().
> +  raw_ostream &getOStream() {
> +    return O;
> +  }
> };
> 
> template<typename GraphType>
> 
> Modified: llvm/trunk/lib/Analysis/CMakeLists.txt
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/CMakeLists.txt?rev=109089&r1=109088&r2=109089&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Analysis/CMakeLists.txt (original)
> +++ llvm/trunk/lib/Analysis/CMakeLists.txt Thu Jul 22 02:46:31 2010
> @@ -38,6 +38,8 @@
>   ProfileInfoLoader.cpp
>   ProfileInfoLoaderPass.cpp
>   ProfileVerifierPass.cpp
> +  RegionInfo.cpp
> +  RegionPrinter.cpp
>   ScalarEvolution.cpp
>   ScalarEvolutionAliasAnalysis.cpp
>   ScalarEvolutionExpander.cpp
> 
> Added: llvm/trunk/lib/Analysis/RegionInfo.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionInfo.cpp?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/lib/Analysis/RegionInfo.cpp (added)
> +++ llvm/trunk/lib/Analysis/RegionInfo.cpp Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,637 @@
> +//===- RegionInfo.cpp - SESE region detection analysis --------------------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +// Detects single entry single exit regions in the control flow graph.
> +//===----------------------------------------------------------------------===//
> +
> +#include "llvm/Analysis/RegionInfo.h"
> +#include "llvm/Analysis/RegionIterator.h"
> +
> +#include "llvm/ADT/PostOrderIterator.h"
> +#include "llvm/ADT/Statistic.h"
> +#include "llvm/Support/CommandLine.h"
> +#include "llvm/Support/ErrorHandling.h"
> +#include "llvm/Support/raw_ostream.h"
> +
> +#define DEBUG_TYPE "region"
> +#include "llvm/Support/Debug.h"
> +
> +#include <set>
> +#include <algorithm>
> +
> +using namespace llvm;
> +
> +// Always verify if expensive checking is enabled.
> +#ifdef XDEBUG
> +bool VerifyRegionInfo = true;
> +#else
> +bool VerifyRegionInfo = false;
> +#endif
> +
> +static cl::opt<bool,true>
> +VerifyRegionInfoX("verify-region-info", cl::location(VerifyRegionInfo),
> +                cl::desc("Verify region info (time consuming)"));
> +
> +STATISTIC(numRegions,       "The # of regions");
> +STATISTIC(numSimpleRegions, "The # of simple regions");
> +
> +//===----------------------------------------------------------------------===//
> +/// PrintStyle - Print region in difference ways.
> +enum PrintStyle { PrintNone, PrintBB, PrintRN  };
> +
> +cl::opt<enum PrintStyle> printStyle("print-region-style", cl::Hidden,
> +  cl::desc("style of printing regions"),
> +  cl::values(
> +    clEnumValN(PrintNone, "none",  "print no details"),
> +    clEnumValN(PrintBB, "bb",  "print regions in detail with block_iterator"),
> +    clEnumValN(PrintRN, "rn",  "print regions in detail with element_iterator"),
> +    clEnumValEnd));
> +//===----------------------------------------------------------------------===//
> +/// Region Implementation
> +Region::Region(BasicBlock *Entry, BasicBlock *Exit, RegionInfo* RInfo,
> +               DominatorTree *dt, Region *Parent)
> +               : RegionNode(Parent, Entry, 1), RI(RInfo), DT(dt), exit(Exit) {}
> +
> +Region::~Region() {
> +  // Only clean the cache for this Region. Caches of child Regions will be
> +  // cleaned when the child Regions are deleted.
> +  BBNodeMap.clear();
> +
> +  for (iterator I = begin(), E = end(); I != E; ++I)
> +    delete *I;
> +}
> +
> +bool Region::contains(const BasicBlock *B) const {
> +  BasicBlock *BB = const_cast<BasicBlock*>(B);
> +
> +  assert(DT->getNode(BB) && "BB not part of the dominance tree");
> +
> +  BasicBlock *entry = getEntry(), *exit = getExit();
> +
> +  // Toplevel region.
> +  if (!exit)
> +    return true;
> +
> +  return (DT->dominates(entry, BB)
> +    && !(DT->dominates(exit, BB) && DT->dominates(entry, exit)));
> +}
> +
> +bool Region::isSimple() const {
> +  bool isSimple = true;
> +  bool found = false;
> +
> +  BasicBlock *entry = getEntry(), *exit = getExit();
> +
> +  // TopLevelRegion
> +  if (!exit)
> +    return false;
> +
> +  for (pred_iterator PI = pred_begin(entry), PE = pred_end(entry); PI != PE;
> +       ++PI)
> +    if (!contains(*PI)) {
> +      if (found) {
> +        isSimple = false;
> +        break;
> +      }
> +      found = true;
> +    }
> +
> +  found = false;
> +
> +  for (pred_iterator PI = pred_begin(exit), PE = pred_end(exit); PI != PE;
> +       ++PI)
> +    if (contains(*PI)) {
> +      if (found) {
> +        isSimple = false;
> +        break;
> +      }
> +      found = true;
> +    }
> +
> +  return isSimple;
> +}
> +
> +void Region::verifyBBInRegion(BasicBlock *BB) const {
> +  if (!contains(BB))
> +    llvm_unreachable("Broken region found!");
> +
> +  BasicBlock *entry = getEntry(), *exit = getExit();
> +
> +  for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE; ++SI)
> +    if (!contains(*SI) && exit != *SI)
> +      llvm_unreachable("Broken region found!");
> +
> +  if (entry != BB)
> +    for (pred_iterator SI = pred_begin(BB), SE = pred_end(BB); SI != SE; ++SI)
> +      if (!contains(*SI))
> +        llvm_unreachable("Broken region found!");
> +}
> +
> +void Region::verifyWalk(BasicBlock *BB, std::set<BasicBlock*> *visited) const {
> +  BasicBlock *exit = getExit();
> +
> +  visited->insert(BB);
> +
> +  verifyBBInRegion(BB);
> +
> +  for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE; ++SI)
> +    if (*SI != exit && visited->find(*SI) == visited->end())
> +        verifyWalk(*SI, visited);
> +}
> +
> +void Region::verifyRegion() const {
> +  // Only do verification when user wants to, otherwise this expensive
> +  // check will be invoked by PassManager.
> +  if (!VerifyRegionInfo) return;
> +
> +  std::set<BasicBlock*> visited;
> +  verifyWalk(getEntry(), &visited);
> +}
> +
> +void Region::verifyRegionNest() const {
> +  for (Region::const_iterator RI = begin(), RE = end(); RI != RE; ++RI)
> +    (*RI)->verifyRegionNest();
> +
> +  verifyRegion();
> +}
> +
> +Region::block_iterator Region::block_begin() {
> +  return GraphTraits<FlatIt<Region*> >::nodes_begin(this);
> +}
> +
> +Region::block_iterator Region::block_end() {
> +  return GraphTraits<FlatIt<Region*> >::nodes_end(this);
> +}
> +
> +Region::const_block_iterator Region::block_begin() const {
> +  return GraphTraits<FlatIt<const Region*> >::nodes_begin(this);
> +}
> +
> +Region::const_block_iterator Region::block_end() const {
> +  return GraphTraits<FlatIt<const Region*> >::nodes_end(this);
> +}
> +
> +Region::element_iterator Region::element_begin() {
> +  return GraphTraits<Region*>::nodes_begin(this);
> +}
> +
> +Region::element_iterator Region::element_end() {
> +  return GraphTraits<Region*>::nodes_end(this);
> +}
> +
> +Region::const_element_iterator Region::element_begin() const {
> +  return GraphTraits<const Region*>::nodes_begin(this);
> +}
> +
> +Region::const_element_iterator Region::element_end() const {
> +  return GraphTraits<const Region*>::nodes_end(this);
> +}
> +
> +Region* Region::getSubRegionNode(BasicBlock *BB) const {
> +  Region *R = RI->getRegionFor(BB);
> +
> +  if (!R || R == this)
> +    return 0;
> +
> +  // If we pass the BB out of this region, that means our code is broken.
> +  assert(contains(R) && "BB not in current region!");
> +
> +  while (contains(R->getParent()) && R->getParent() != this)
> +    R = R->getParent();
> +
> +  if (R->getEntry() != BB)
> +    return 0;
> +
> +  return R;
> +}
> +
> +RegionNode* Region::getBBNode(BasicBlock *BB) const {
> +  assert(contains(BB) && "Can get BB node out of this region!");
> +
> +  BBNodeMapT::const_iterator at = BBNodeMap.find(BB);
> +
> +  if (at != BBNodeMap.end())
> +    return at->second;
> +
> +  RegionNode *NewNode = new RegionNode(const_cast<Region*>(this), BB);
> +  BBNodeMap.insert(std::make_pair(BB, NewNode));
> +  return NewNode;
> +}
> +
> +RegionNode* Region::getNode(BasicBlock *BB) const {
> +  assert(contains(BB) && "Can get BB node out of this region!");
> +  if (Region* Child = getSubRegionNode(BB))
> +    return Child->getNode();
> +
> +  return getBBNode(BB);
> +}
> +
> +void Region::transferChildrenTo(Region *To) {
> +  for (iterator I = begin(), E = end(); I != E; ++I) {
> +    (*I)->parent = To;
> +    To->children.push_back(*I);
> +  }
> +  children.clear();
> +}
> +
> +void Region::addSubRegion(Region *SubRegion) {
> +  assert(SubRegion->parent == 0 && "SubRegion already has a parent!");
> +  SubRegion->parent = this;
> +  // Set up the region node.
> +  assert(std::find(children.begin(), children.end(), SubRegion) == children.end()
> +         && "Node already exist!");
> +  children.push_back(SubRegion);
> +}
> +
> +
> +Region *Region::removeSubRegion(Region *Child) {
> +  assert(Child->parent == this && "Child is not a child of this region!");
> +  Child->parent = 0;
> +  RegionSet::iterator I = std::find(children.begin(), children.end(), Child);
> +  assert(I != children.end() && "Region does not exit. Unable to remove.");
> +  children.erase(children.begin()+(I-begin()));
> +  return Child;
> +}
> +
> +unsigned Region::getDepth() const {
> +  unsigned Depth = 0;
> +
> +  for (Region *R = parent; R != 0; R = R->parent)
> +    ++Depth;
> +
> +  return Depth;
> +}
> +
> +void Region::print(raw_ostream &OS, bool print_tree, unsigned level) const {
> +  if (print_tree)
> +    OS.indent(level*2) << "[" << level << "] " << getNameStr();
> +  else
> +    OS.indent(level*2) << getNameStr();
> +
> +  OS << "\n";
> +
> +
> +  if (printStyle != PrintNone) {
> +    OS.indent(level*2) << "{\n";
> +    OS.indent(level*2 + 2);
> +
> +    if (printStyle == PrintBB) {
> +      for (const_block_iterator I = block_begin(), E = block_end(); I!=E; ++I)
> +        OS << **I << ", "; // TODO: remove the last ","
> +    } else if (printStyle == PrintRN) {
> +      for (const_element_iterator I = element_begin(), E = element_end(); I!=E; ++I)
> +        OS << **I << ", "; // TODO: remove the last ",
> +    }
> +
> +    OS << "\n";
> +  }
> +
> +  if (print_tree)
> +    for (const_iterator RI = begin(), RE = end(); RI != RE; ++RI)
> +      (*RI)->print(OS, print_tree, level+1);
> +
> +  if (printStyle != PrintNone)
> +    OS.indent(level*2) << "} \n";
> +}
> +
> +void Region::dump() const {
> +  print(dbgs(), true, getDepth());
> +}
> +
> +void Region::clearNodeCache() {
> +  BBNodeMap.clear();
> +  for (Region::iterator RI = begin(), RE = end(); RI != RE; ++RI)
> +    (*RI)->clearNodeCache();
> +}
> +
> +//===----------------------------------------------------------------------===//
> +// RegionInfo implementation
> +//
> +
> +bool RegionInfo::isCommonDomFrontier(BasicBlock *BB, BasicBlock *entry,
> +                                     BasicBlock *exit) const {
> +  for (pred_iterator PI = pred_begin(BB), PE = pred_end(BB); PI != PE; ++PI)
> +    if (DT->dominates(entry, *PI) && !DT->dominates(exit, *PI))
> +      return false;
> +
> +  return true;
> +}
> +
> +bool RegionInfo::isRegion(BasicBlock *entry, BasicBlock *exit) const {
> +  assert(entry && exit && "entry and exit must not be null!");
> +  typedef DominanceFrontier::DomSetType DST;
> +
> +  DST *entrySuccs = &(*DF->find(entry)).second;
> +
> +  // Exit is the header of a loop that contains the entry. In this case,
> +  // the dominance frontier must only contain the exit.
> +  if (!DT->dominates(entry, exit)) {
> +    for (DST::iterator SI = entrySuccs->begin(), SE = entrySuccs->end();
> +         SI != SE; ++SI)
> +      if (*SI != exit && *SI != entry)
> +        return false;
> +
> +    return true;
> +  }
> +
> +  DST *exitSuccs = &(*DF->find(exit)).second;
> +
> +  // Do not allow edges leaving the region.
> +  for (DST::iterator SI = entrySuccs->begin(), SE = entrySuccs->end();
> +       SI != SE; ++SI) {
> +    if (*SI == exit || *SI == entry)
> +      continue;
> +    if (exitSuccs->find(*SI) == exitSuccs->end())
> +      return false;
> +    if (!isCommonDomFrontier(*SI, entry, exit))
> +      return false;
> +  }
> +
> +  // Do not allow edges pointing into the region.
> +  for (DST::iterator SI = exitSuccs->begin(), SE = exitSuccs->end();
> +       SI != SE; ++SI)
> +    if (DT->dominates(entry, *SI) && *SI != entry && *SI != exit)
> +      return false;
> +
> +
> +  return true;
> +}
> +
> +void RegionInfo::insertShortCut(BasicBlock *entry, BasicBlock *exit,
> +                             BBtoBBMap *ShortCut) const {
> +  assert(entry && exit && "entry and exit must not be null!");
> +
> +  BBtoBBMap::iterator e = ShortCut->find(exit);
> +
> +  if (e == ShortCut->end())
> +    // No further region at exit available.
> +    (*ShortCut)[entry] = exit;
> +  else {
> +    // We found a region e that starts at exit. Therefore (entry, e->second)
> +    // is also a region, that is larger than (entry, exit). Insert the
> +    // larger one.
> +    BasicBlock *BB = e->second;
> +    (*ShortCut)[entry] = BB;
> +  }
> +}
> +
> +DomTreeNode* RegionInfo::getNextPostDom(DomTreeNode* N,
> +                                        BBtoBBMap *ShortCut) const {
> +  BBtoBBMap::iterator e = ShortCut->find(N->getBlock());
> +
> +  if (e == ShortCut->end())
> +    return N->getIDom();
> +
> +  return PDT->getNode(e->second)->getIDom();
> +}
> +
> +bool RegionInfo::isTrivialRegion(BasicBlock *entry, BasicBlock *exit) const {
> +  assert(entry && exit && "entry and exit must not be null!");
> +
> +  unsigned num_successors = succ_end(entry) - succ_begin(entry);
> +
> +  if (num_successors <= 1 && exit == *(succ_begin(entry)))
> +    return true;
> +
> +  return false;
> +}
> +
> +void RegionInfo::updateStatistics(Region *R) {
> +  ++numRegions;
> +
> +  // TODO: Slow. Should only be enabled if -stats is used.
> +  if (R->isSimple()) ++numSimpleRegions;
> +}
> +
> +Region *RegionInfo::createRegion(BasicBlock *entry, BasicBlock *exit) {
> +  assert(entry && exit && "entry and exit must not be null!");
> +
> +  if (isTrivialRegion(entry, exit))
> +    return 0;
> +
> +  Region *region = new Region(entry, exit, this, DT);
> +  BBtoRegion.insert(std::make_pair(entry, region));
> +
> + #ifdef XDEBUG
> +    region->verifyRegion();
> + #else
> +    DEBUG(region->verifyRegion());
> + #endif
> +
> +  updateStatistics(region);
> +  return region;
> +}
> +
> +void RegionInfo::findRegionsWithEntry(BasicBlock *entry, BBtoBBMap *ShortCut) {
> +  assert(entry);
> +
> +  DomTreeNode *N = PDT->getNode(entry);
> +
> +  if (!N)
> +    return;
> +
> +  Region *lastRegion= 0;
> +  BasicBlock *lastExit = entry;
> +
> +  // As only a BasicBlock that postdominates entry can finish a region, walk the
> +  // post dominance tree upwards.
> +  while ((N = getNextPostDom(N, ShortCut))) {
> +    BasicBlock *exit = N->getBlock();
> +
> +    if (!exit)
> +      break;
> +
> +    if (isRegion(entry, exit)) {
> +      Region *newRegion = createRegion(entry, exit);
> +
> +      if (lastRegion)
> +        newRegion->addSubRegion(lastRegion);
> +
> +      lastRegion = newRegion;
> +      lastExit = exit;
> +    }
> +
> +    // This can never be a region, so stop the search.
> +    if (!DT->dominates(entry, exit))
> +      break;
> +  }
> +
> +  // Tried to create regions from entry to lastExit.  Next time take a
> +  // shortcut from entry to lastExit.
> +  if (lastExit != entry)
> +    insertShortCut(entry, lastExit, ShortCut);
> +}
> +
> +void RegionInfo::scanForRegions(Function &F, BBtoBBMap *ShortCut) {
> +  BasicBlock *entry = &(F.getEntryBlock());
> +  DomTreeNode *N = DT->getNode(entry);
> +
> +  // Iterate over the dominance tree in post order to start with the small
> +  // regions from the bottom of the dominance tree.  If the small regions are
> +  // detected first, detection of bigger regions is faster, as we can jump
> +  // over the small regions.
> +  for (po_iterator<DomTreeNode*> FI = po_begin(N), FE = po_end(N); FI != FE;
> +    ++FI) {
> +    findRegionsWithEntry((*FI)->getBlock(), ShortCut);
> +  }
> +}
> +
> +Region *RegionInfo::getTopMostParent(Region *region) {
> +  while (region->parent)
> +    region = region->getParent();
> +
> +  return region;
> +}
> +
> +void RegionInfo::buildRegionsTree(DomTreeNode *N, Region *region) {
> +  BasicBlock *BB = N->getBlock();
> +
> +  // Passed region exit
> +  while (BB == region->getExit())
> +    region = region->getParent();
> +
> +  BBtoRegionMap::iterator it = BBtoRegion.find(BB);
> +
> +  // This basic block is a start block of a region. It is already in the
> +  // BBtoRegion relation. Only the child basic blocks have to be updated.
> +  if (it != BBtoRegion.end()) {
> +    Region *newRegion = it->second;;
> +    region->addSubRegion(getTopMostParent(newRegion));
> +    region = newRegion;
> +  } else {
> +    BBtoRegion[BB] = region;
> +  }
> +
> +  for (DomTreeNode::iterator CI = N->begin(), CE = N->end(); CI != CE; ++CI)
> +    buildRegionsTree(*CI, region);
> +}
> +
> +void RegionInfo::releaseMemory() {
> +  BBtoRegion.clear();
> +  if (TopLevelRegion)
> +    delete TopLevelRegion;
> +  TopLevelRegion = 0;
> +}
> +
> +RegionInfo::RegionInfo() : FunctionPass(&ID) {
> +  TopLevelRegion = 0;
> +}
> +
> +RegionInfo::~RegionInfo() {
> +  releaseMemory();
> +}
> +
> +void RegionInfo::Calculate(Function &F) {
> +  // ShortCut a function where for every BB the exit of the largest region
> +  // starting with BB is stored. These regions can be threated as single BBS.
> +  // This improves performance on linear CFGs.
> +  BBtoBBMap ShortCut;
> +
> +  scanForRegions(F, &ShortCut);
> +  BasicBlock *BB = &F.getEntryBlock();
> +  buildRegionsTree(DT->getNode(BB), TopLevelRegion);
> +}
> +
> +bool RegionInfo::runOnFunction(Function &F) {
> +  releaseMemory();
> +
> +  DT = &getAnalysis<DominatorTree>();
> +  PDT = &getAnalysis<PostDominatorTree>();
> +  DF = &getAnalysis<DominanceFrontier>();
> +
> +  TopLevelRegion = new Region(&F.getEntryBlock(), 0, this, DT, 0);
> +  updateStatistics(TopLevelRegion);
> +
> +  Calculate(F);
> +
> +  return false;
> +}
> +
> +void RegionInfo::getAnalysisUsage(AnalysisUsage &AU) const {
> +  AU.setPreservesAll();
> +  AU.addRequiredTransitive<DominatorTree>();
> +  AU.addRequired<PostDominatorTree>();
> +  AU.addRequired<DominanceFrontier>();
> +}
> +
> +void RegionInfo::print(raw_ostream &OS, const Module *) const {
> +  OS << "Region tree:\n";
> +  TopLevelRegion->print(OS, true, 0);
> +  OS << "End region tree\n";
> +}
> +
> +void RegionInfo::verifyAnalysis() const {
> +  // Only do verification when user wants to, otherwise this expensive check
> +  // will be invoked by PMDataManager::verifyPreservedAnalysis when
> +  // a regionpass (marked PreservedAll) finish.
> +  if (!VerifyRegionInfo) return;
> +
> +  TopLevelRegion->verifyRegionNest();
> +}
> +
> +// Region pass manager support.
> +Region *RegionInfo::getRegionFor(BasicBlock *BB) const {
> +  BBtoRegionMap::const_iterator I=
> +    BBtoRegion.find(BB);
> +  return I != BBtoRegion.end() ? I->second : 0;
> +}
> +
> +Region *RegionInfo::operator[](BasicBlock *BB) const {
> +  return getRegionFor(BB);
> +}
> +
> +Region*
> +RegionInfo::getCommonRegion(Region *A, Region *B) const {
> +  assert (A && B && "One of the Regions is NULL");
> +
> +  if (A->contains(B)) return A;
> +
> +  while (!B->contains(A))
> +    B = B->getParent();
> +
> +  return B;
> +}
> +
> +Region*
> +RegionInfo::getCommonRegion(SmallVectorImpl<Region*> &Regions) const {
> +  Region* ret = Regions.back();
> +  Regions.pop_back();
> +
> +  for (SmallVectorImpl<Region*>::const_iterator I = Regions.begin(),
> +       E = Regions.end(); I != E; ++I)
> +      ret = getCommonRegion(ret, *I);
> +
> +  return ret;
> +}
> +
> +Region*
> +RegionInfo::getCommonRegion(SmallVectorImpl<BasicBlock*> &BBs) const {
> +  Region* ret = getRegionFor(BBs.back());
> +  BBs.pop_back();
> +
> +  for (SmallVectorImpl<BasicBlock*>::const_iterator I = BBs.begin(),
> +       E = BBs.end(); I != E; ++I)
> +      ret = getCommonRegion(ret, getRegionFor(*I));
> +
> +  return ret;
> +}
> +
> +char RegionInfo::ID = 0;
> +INITIALIZE_PASS(RegionInfo, "regions",
> +                "Detect single entry single exit regions", true, true);
> +
> +// Create methods available outside of this file, to use them
> +// "include/llvm/LinkAllPasses.h". Otherwise the pass would be deleted by
> +// the link time optimization.
> +
> +namespace llvm {
> +  FunctionPass *createRegionInfoPass() {
> +    return new RegionInfo();
> +  }
> +}
> +
> 
> Added: llvm/trunk/lib/Analysis/RegionPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionPrinter.cpp?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/lib/Analysis/RegionPrinter.cpp (added)
> +++ llvm/trunk/lib/Analysis/RegionPrinter.cpp Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,182 @@
> +//===- RegionPrinter.cpp - Print regions tree pass ------------------------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +// Print out the region tree of a function using dotty/graphviz.
> +//===----------------------------------------------------------------------===//
> +
> +#include "llvm/Analysis/RegionInfo.h"
> +#include "llvm/Analysis/RegionIterator.h"
> +#include "llvm/Analysis/RegionPrinter.h"
> +#include "llvm/Analysis/Passes.h"
> +#include "llvm/Analysis/DOTGraphTraitsPass.h"
> +#include "llvm/ADT/Statistic.h"
> +#include "llvm/ADT/PostOrderIterator.h"
> +#include "llvm/ADT/DepthFirstIterator.h"
> +#include "llvm/Support/Debug.h"
> +#include "llvm/Support/CommandLine.h"
> +#include "llvm/Support/raw_ostream.h"
> +
> +using namespace llvm;
> +
> +//===----------------------------------------------------------------------===//
> +/// onlySimpleRegion - Show only the simple regions in the RegionViewer.
> +static cl::opt<bool>
> +onlySimpleRegions("only-simple-regions",
> +                  cl::desc("Show only simple regions in the graphviz viewer"),
> +                  cl::Hidden,
> +                  cl::init(false));
> +
> +namespace llvm {
> +template<>
> +struct DOTGraphTraits<RegionNode*> : public DefaultDOTGraphTraits {
> +
> +  DOTGraphTraits (bool isSimple=false)
> +    : DefaultDOTGraphTraits(isSimple) {}
> +
> +  std::string getNodeLabel(RegionNode *Node, RegionNode *Graph) {
> +
> +    if (!Node->isSubRegion()) {
> +      BasicBlock *BB = Node->getNodeAs<BasicBlock>();
> +
> +      if (isSimple())
> +        return DOTGraphTraits<const Function*>
> +          ::getSimpleNodeLabel(BB, BB->getParent());
> +      else
> +        return DOTGraphTraits<const Function*>
> +          ::getCompleteNodeLabel(BB, BB->getParent());
> +    }
> +
> +    return "Not implemented";
> +  }
> +};
> +
> +template<>
> +struct DOTGraphTraits<RegionInfo*> : public DOTGraphTraits<RegionNode*> {
> +
> +  DOTGraphTraits (bool isSimple=false)
> +    : DOTGraphTraits<RegionNode*>(isSimple) {}
> +
> +  static std::string getGraphName(RegionInfo *DT) {
> +    return "Region Graph";
> +  }
> +
> +  std::string getNodeLabel(RegionNode *Node, RegionInfo *G) {
> +    return DOTGraphTraits<RegionNode*>::getNodeLabel(Node,
> +                                                     G->getTopLevelRegion());
> +  }
> +
> +  // Print the cluster of the subregions. This groups the single basic blocks
> +  // and adds a different background color for each group.
> +  static void printRegionCluster(const Region *R, GraphWriter<RegionInfo*> &GW,
> +                                 unsigned depth = 0) {
> +    raw_ostream &O = GW.getOStream();
> +    O.indent(2 * depth) << "subgraph cluster_" << static_cast<const void*>(R)
> +      << " {\n";
> +    O.indent(2 * (depth + 1)) << "label = \"\";\n";
> +
> +    if (!onlySimpleRegions || R->isSimple()) {
> +      O.indent(2 * (depth + 1)) << "style = filled;\n";
> +      O.indent(2 * (depth + 1)) << "color = "
> +        << ((R->getDepth() * 2 % 12) + 1) << "\n";
> +
> +    } else {
> +      O.indent(2 * (depth + 1)) << "style = solid;\n";
> +      O.indent(2 * (depth + 1)) << "color = "
> +        << ((R->getDepth() * 2 % 12) + 2) << "\n";
> +    }
> +
> +    for (Region::const_iterator RI = R->begin(), RE = R->end(); RI != RE; ++RI)
> +      printRegionCluster(*RI, GW, depth + 1);
> +
> +    RegionInfo *RI = R->getRegionInfo();
> +
> +    for (Region::const_block_iterator BI = R->block_begin(),
> +         BE = R->block_end(); BI != BE; ++BI) {
> +      BasicBlock *BB = (*BI)->getNodeAs<BasicBlock>();
> +      if (RI->getRegionFor(BB) == R)
> +        O.indent(2 * (depth + 1)) << "Node"
> +          << static_cast<const void*>(RI->getTopLevelRegion()->getBBNode(BB))
> +          << ";\n";
> +    }
> +
> +    O.indent(2 * depth) << "}\n";
> +  }
> +
> +  static void addCustomGraphFeatures(const RegionInfo* RI,
> +                                     GraphWriter<RegionInfo*> &GW) {
> +    raw_ostream &O = GW.getOStream();
> +    O << "\tcolorscheme = \"paired12\"\n";
> +    printRegionCluster(RI->getTopLevelRegion(), GW, 4);
> +  }
> +};
> +} //end namespace llvm
> +
> +namespace {
> +
> +struct RegionViewer
> +  : public DOTGraphTraitsViewer<RegionInfo, false> {
> +  static char ID;
> +  RegionViewer() : DOTGraphTraitsViewer<RegionInfo, false>("reg", &ID){}
> +};
> +
> +char RegionViewer::ID = 0;
> +INITIALIZE_PASS(RegionViewer, "view-regions", "View regions of function",
> +                true, true);
> +
> +struct RegionOnlyViewer
> +  : public DOTGraphTraitsViewer<RegionInfo, true> {
> +  static char ID;
> +  RegionOnlyViewer() : DOTGraphTraitsViewer<RegionInfo, true>("regonly", &ID){}
> +};
> +
> +char RegionOnlyViewer::ID = 0;
> +INITIALIZE_PASS(RegionOnlyViewer, "view-regions-only",
> +                "View regions of function (with no function bodies)",
> +                true, true);
> +
> +struct RegionPrinter
> +  : public DOTGraphTraitsPrinter<RegionInfo, false> {
> +  static char ID;
> +  RegionPrinter() :
> +    DOTGraphTraitsPrinter<RegionInfo, false>("reg", &ID) {}
> +};
> +} //end anonymous namespace
> +
> +char RegionPrinter::ID = 0;
> +INITIALIZE_PASS(RegionPrinter, "dot-regions",
> +                "Print regions of function to 'dot' file", true, true);
> +
> +struct RegionOnlyPrinter
> +  : public DOTGraphTraitsPrinter<RegionInfo, true> {
> +  static char ID;
> +  RegionOnlyPrinter() :
> +    DOTGraphTraitsPrinter<RegionInfo, true>("reg", &ID) {}
> +};
> +
> +char RegionOnlyPrinter::ID = 0;
> +INITIALIZE_PASS(RegionOnlyPrinter, "dot-regions-only",
> +                "Print regions of function to 'dot' file "
> +                "(with no function bodies)",
> +                true, true);
> +
> +FunctionPass* llvm::createRegionViewerPass() {
> +  return new RegionViewer();
> +}
> +
> +FunctionPass* llvm::createRegionOnlyViewerPass() {
> +  return new RegionOnlyViewer();
> +}
> +
> +FunctionPass* llvm::createRegionPrinterPass() {
> +  return new RegionPrinter();
> +}
> +
> +FunctionPass* llvm::createRegionOnlyPrinterPass() {
> +  return new RegionOnlyPrinter();
> +}
> +
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/block_sort.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/block_sort.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/block_sort.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/block_sort.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,42 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats -analyze < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @BZ2_blockSort() nounwind {
> +start:
> +  br label %while
> +
> +while:
> +  br label %while.body134.i.i
> +
> +while.body134.i.i:
> +  br i1 1, label %end, label %w
> +
> +w:
> +  br label %if.end140.i.i
> +
> +if.end140.i.i:
> +  br i1 1, label %while.end186.i.i, label %if.end183.i.i
> +
> +if.end183.i.i:
> +  br label %while.body134.i.i
> +
> +while.end186.i.i:
> +  br label %while
> +
> +end:
> +  ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] start => <Function Return>
> +; CHECK: [1] while => end
> +
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: start, while, while.body134.i.i, end, w, if.end140.i.i, while.end186.i.i, if.end183.i.i,
> +; BBIT: while, while.body134.i.i, w, if.end140.i.i, while.end186.i.i, if.end183.i.i,
> +
> +; RNIT: start, while => end, end,
> +; RNIT: while, while.body134.i.i, w, if.end140.i.i, while.end186.i.i, if.end183.i.i,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/cond_loop.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/cond_loop.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/cond_loop.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/cond_loop.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,33 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +5:
> +        br label %"0"
> +
> +0:
> +	br label %"1"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	ret void
> +3:
> +	br i1 1, label %"1", label %"4"
> +4:
> +        br label %"0"
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] 5 => <Function Return>
> +; CHECK: [1] 0 => 2
> +
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 5, 0, 1, 2, 3, 4,
> +; BBIT: 0, 1, 3, 4,
> +
> +; RNIT: 5, 0 => 2, 2,
> +; RNIT: 0, 1, 3, 4,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/condition_complicated.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/condition_complicated.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/condition_complicated.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/condition_complicated.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,60 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc zeroext i8 @handle_compress() nounwind {
> +end165:
> +  br i1 1, label %false239, label %true181
> +
> +true181:
> +  br i1 1, label %then187, label %else232
> +
> +then187:
> +  br label %end265
> +
> +else232:
> +  br i1 1, label %false239, label %then245
> +
> +false239:
> +  br i1 1, label %then245, label %else259
> +
> +then245:
> +  br i1 1, label %then251, label %end253
> +
> +then251:
> +  br label %end253
> +
> +end253:
> +  br label %end265
> +
> +else259:
> +  br label %end265
> +
> +end265:
> +  br i1 1, label %then291, label %end298
> +
> +then291:
> +  br label %end298
> +
> +end298:
> +  ret i8 1
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] end165 => <Function Return>
> +; CHECK-NEXT: [1] end165 => end265
> +; CHECK-NEXT: [2] then245 => end253
> +; CHECK-NEXT: [1] end265 => end298
> +
> +; STAT: 4 region - The # of regions
> +
> +; BBIT: end165, false239, then245, then251, end253, end265, then291, end298, else259, true181, then187, else232,
> +; BBIT: end165, false239, then245, then251, end253, else259, true181, then187, else232,
> +; BBIT: then245, then251,
> +; BBIT: end265, then291,
> +
> +; RNIT: end165 => end265, end265 => end298, end298,
> +; RNIT: end165, false239, then245 => end253, end253, else259, true181, then187, else232,
> +; RNIT: then245, then251,
> +; RNIT: end265, then291,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/condition_complicated_2.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/condition_complicated_2.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/condition_complicated_2.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/condition_complicated_2.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,44 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc void @compress() nounwind {
> +end33:
> +  br i1 1, label %end124, label %lor.lhs.false95
> +
> +lor.lhs.false95:
> +  br i1 1, label %then107, label %end172
> +
> +then107:
> +  br i1 1, label %end124, label %then113
> +
> +then113:
> +  br label %end124
> +
> +end124:
> +  br label %exit
> +
> +end172:
> +  br label %exit
> +
> +
> +exit:
> +  unreachable
> +
> +
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] end33 => <Function Return>
> +; CHECK-NEXT:      [1] end33 => exit
> +; CHECK-NEXT:   [2] then107 => end124
> +
> +; STAT: 3 region - The # of regions
> +
> +; BBIT: end33, end124, exit, lor.lhs.false95, then107, then113, end172,
> +; BBIT: end33, end124, lor.lhs.false95, then107, then113, end172,
> +; BBIT: then107, then113,
> +
> +; RNIT: end33 => exit, exit,
> +; RNIT: end33, end124, lor.lhs.false95, then107 => end124, end172,
> +; RNIT: then107, then113,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/condition_forward_edge.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/condition_forward_edge.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/condition_forward_edge.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/condition_forward_edge.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,26 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"1"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"3"
> +3:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK: [1] 1 => 3
> +
> +; STAT: 2 region - The # of regions
> +
> +; BBIT: 0, 1, 2, 3,
> +; BBIT: 1, 2,
> +
> +; RNIT: 0, 1 => 3, 3,
> +; RNIT: 1, 2,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/condition_same_exit.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/condition_same_exit.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/condition_same_exit.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/condition_same_exit.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,31 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br i1 1, label %"1", label %"4"
> +
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"4"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 0 => 4
> +; CHECK-NEXT:   [2] 1 => 4
> +; STAT: 3 region - The # of regions
> +
> +; BBIT: 0, 1, 2, 4, 3,
> +; BBIT: 0, 1, 2, 3,
> +; BBIT: 1, 2, 3,
> +
> +; RNIT: 0 => 4, 4,
> +; RNIT: 0, 1 => 4,
> +; RNIT: 1, 2, 3,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/condition_simple.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/condition_simple.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/condition_simple.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/condition_simple.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,28 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"1"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"4"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 1 => 4
> +; STAT: 2 region - The # of regions
> +
> +; BBIT: 0, 1, 2, 4, 3,
> +; BBIT: 1, 2, 3,
> +
> +; RNIT: 0, 1 => 4, 4,
> +; RNIT: 1, 2, 3,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/dg.exp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/dg.exp?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/dg.exp (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/dg.exp Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,3 @@
> +load_lib llvm.exp
> +
> +RunLLVMTests [lsort [glob -nocomplain $srcdir/$subdir/*.{ll,c,cpp}]]
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/exit_in_condition.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/exit_in_condition.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/exit_in_condition.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/exit_in_condition.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,38 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc zeroext i8 @handle_compress() nounwind {
> +entry:
> +  br label %outer
> +
> +outer:
> +  br label %body
> +
> +body:
> +  br i1 1, label %body.i, label %if.end
> +
> +body.i:
> +  br i1 1, label %end, label %if.end
> +
> +if.end:
> +  br label %if.then64
> +
> +if.then64:
> +  br label %outer
> +
> +end:
> +  ret i8 1
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK-NEXT: [1] outer => end
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: entry, outer, body, body.i, end, if.end, if.then64,
> +; BBIT: outer, body, body.i, if.end, if.then64,
> +
> +; RNIT: entry, outer => end, end,
> +; RNIT: outer, body, body.i, if.end, if.then64,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/infinite_loop.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/infinite_loop.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/infinite_loop.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/infinite_loop.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,20 @@
> +; RUN: opt -regions -analyze < %s 
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"1"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"2"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK: [1] 1 => 4
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/infinite_loop_2.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/infinite_loop_2.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/infinite_loop_2.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/infinite_loop_2.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,36 @@
> +; RUN: opt -regions -analyze < %s 
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"1"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"5"
> +5:
> +	br i1 1, label %"11", label %"12"
> +11:
> +        br label %"6"
> +12:
> +        br label %"6"
> +6:
> +        br label %"2"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK: [1] 1 => 3
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 0, 1, 2, 5, 11, 6, 12, 3, 4,
> +; BBIT: 1, 2, 5, 11, 6, 12,
> +
> +; RNIT: 0, 1 => 3, 3, 4,
> +; RNIT: 1, 2, 5, 11, 6, 12,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/infinite_loop_3.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/infinite_loop_3.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/infinite_loop_3.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/infinite_loop_3.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,52 @@
> +; RUN: opt -regions -analyze < %s 
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"7"
> +7:
> +	br i1 1, label %"1", label %"8"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"5"
> +5:
> +	br i1 1, label %"11", label %"12"
> +11:
> +        br label %"6"
> +12:
> +        br label %"6"
> +6:
> +        br label %"2"
> +8:
> +	br label %"9"
> +9:
> +	br i1 1, label %"13", label %"14"
> +13:
> +        br label %"10"
> +14:
> +        br label %"10"
> +10:
> +        br label %"8"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 1 => 3
> +; CHECK-NEXT: [1] 7 => 1
> +; STAT: 3 region - The # of regions
> +; STAT: 2 region - The # of simple regions
> +
> +; BBIT: 0, 7, 1, 2, 5, 11, 6, 12, 3, 4, 8, 9, 13, 10, 14,
> +; BBIT: 7, 8, 9, 13, 10, 14,
> +; BBIT: 1, 2, 5, 11, 6, 12,
> +
> +; RNIT: 0, 7 => 1, 1 => 3, 3, 4,
> +; RNIT: 7, 8, 9, 13, 10, 14,
> +; RNIT: 1, 2, 5, 11, 6, 12,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/infinite_loop_4.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/infinite_loop_4.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/infinite_loop_4.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/infinite_loop_4.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,48 @@
> +; RUN: opt -regions -analyze < %s 
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +	br label %"7"
> +7:
> +	br i1 1, label %"1", label %"8"
> +1:
> +	br i1 1, label %"2", label %"3"
> +2:
> +	br label %"5"
> +5:
> +	br i1 1, label %"11", label %"12"
> +11:
> +        br label %"6"
> +12:
> +        br label %"6"
> +6:
> +	br i1 1, label %"2", label %"10"
> +8:
> +	br label %"9"
> +9:
> +	br i1 1, label %"13", label %"14"
> +13:
> +        br label %"10"
> +14:
> +        br label %"10"
> +10:
> +        br label %"8"
> +3:
> +	br label %"4"
> +4:
> +	ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 7 => 3
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 0, 7, 1, 2, 5, 11, 6, 10, 8, 9, 13, 14, 12, 3, 4,
> +; BBIT: 7, 1, 2, 5, 11, 6, 10, 8, 9, 13, 14, 12,
> +
> +; RNIT: 0, 7 => 3, 3, 4,
> +; RNIT: 7, 1, 2, 5, 11, 6, 10, 8, 9, 13, 14, 12,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/loop_with_condition.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/loop_with_condition.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/loop_with_condition.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/loop_with_condition.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,46 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition() nounwind {
> +0:
> +        br label %"1"
> +1:
> +	br i1 1, label %"6", label %"2"
> +2:
> +	br i1 1, label %"3", label %"4"
> +3:
> +	br label %"5"
> +4:
> +	br label %"5"
> +5:
> +        br label %"8"
> +8:
> +        br i1 1, label %"7", label %"9"
> +9:
> +        br label %"2"
> +7:
> +        br label %"6"
> +6:
> +	ret void
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 1 => 6
> +; CHECK-NEXT:   [2] 2 => 7
> +; CHECK-NEXT:     [3] 2 => 5
> +; STAT: 4 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 0, 1, 6, 2, 3, 5, 8, 7, 9, 4,
> +; BBIT: 1, 2, 3, 5, 8, 7, 9, 4,
> +; BBIT: 2, 3, 5, 8, 9, 4,
> +; BBIT: 2, 3, 4,
> +
> +; RNIT: 0, 1 => 6, 6,
> +; RNIT: 1, 2 => 7, 7,
> +; RNIT: 2 => 5, 5, 8, 9,
> +; RNIT: 2, 3, 4,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/loops_1.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/loops_1.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/loops_1.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/loops_1.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,40 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc zeroext i8 @loops_1() nounwind {
> +entry:
> +  br i1 1, label %outer , label %a
> +
> +a:
> +  br label %body
> +
> +outer:
> +  br label %body
> +
> +body:
> +  br i1 1, label %land, label %if
> +
> +land:
> +  br i1 1, label %exit, label %end
> +
> +exit:
> +  br i1 1, label %if, label %end
> +
> +if:
> +  br label %outer
> +
> +end:
> +  ret i8 1
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK-NEXT: [1] entry => end
> +; STAT: 2 region - The # of regions
> +
> +; BBIT: entry, outer, body, land, exit, if, end, a,
> +; BBIT: entry, outer, body, land, exit, if, a,
> +
> +; RNIT: entry => end, end,
> +; RNIT: entry, outer, body, land, exit, if, a,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/loops_2.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/loops_2.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/loops_2.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/loops_2.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,49 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @meread_() nounwind {
> +entry:
> +  br label %bb23
> +
> +bb23:
> +  br label %bb.i
> +
> +bb.i:                                             ; preds = %bb.i, %bb54
> +  br label %pflini_.exit
> +
> +pflini_.exit:                                     ; preds = %bb.i
> +  br label %bb58thread-split
> +
> +bb58thread-split:                                 ; preds = %bb64, %bb61, %pflini_.exit
> +  br label %bb58
> +
> +bb58:                                             ; preds = %bb60, %bb58thread-split
> +  br i1 1, label %bb59, label %bb23
> +
> +bb59:                                             ; preds = %bb58
> +  switch i32 1, label %bb60 [
> +    i32 1, label %l98
> +  ]
> +
> +bb60:                                             ; preds = %bb59
> +  br i1 1, label %bb61, label %bb58
> +
> +bb61:                                             ; preds = %bb60
> +  br label %bb58thread-split
> +
> +l98:                                   ; preds = %bb69, %bb59
> +  ret void
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK: [1] bb23 => l98
> +; STAT: 2 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: entry, bb23, bb.i, pflini_.exit, bb58thread-split, bb58, bb59, bb60, bb61, l98,
> +; BBIT: bb23, bb.i, pflini_.exit, bb58thread-split, bb58, bb59, bb60, bb61,
> +
> +; RNIT: entry, bb23 => l98, l98,
> +; RNIT: bb23, bb.i, pflini_.exit, bb58thread-split, bb58, bb59, bb60, bb61,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/mix_1.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/mix_1.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/mix_1.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/mix_1.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,69 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @a_linear_impl_fig_1() nounwind {
> +0:
> +
> +	br i1 1, label %"1", label %"15"
> +1:
> + 	switch i32 0, label %"2" [ i32 0, label %"3"
> +                                  i32 1, label %"7"]
> +2:
> +	br label %"4"
> +3:
> +	br label %"5"
> +4:
> +	br label %"6"
> +5:
> +	br label %"6"
> +6:
> +	br label %"7"
> +7:
> +	br label %"15"
> +15:
> +	br label %"8"
> +8:
> +	br label %"16"
> +16:
> +        br label %"9"
> +9:
> +	br i1 1, label %"10", label %"11"
> +11:
> +	br i1 1, label %"13", label %"12"
> +13:
> +	br label %"14"
> +12:
> +	br label %"14"
> +14:
> +	br label %"8"
> +10:
> +        br label %"17"
> +17:
> +        br label %"18"
> +18:
> +        ret void
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 0 => 15
> +; CHECK-NEXT:  [2] 1 => 7
> +; CHECK-NEXT: [1] 8 => 10
> +; CHECK-NEXT:  [2] 11 => 14
> +; STAT: 5 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 0, 1, 2, 4, 6, 7, 15, 8, 16, 9, 10, 17, 18, 11, 13, 14, 12, 3, 5,
> +; BBIT: 0, 1, 2, 4, 6, 7, 3, 5,
> +; BBIT: 1, 2, 4, 6, 3, 5,
> +; BBIT: 8, 16, 9, 11, 13, 14, 12,
> +; BBIT: 11, 13, 12,
> +
> +; RNIT: 0 => 15, 15, 8 => 10, 10, 17, 18,
> +; RNIT: 0, 1 => 7, 7,
> +; RNIT: 1, 2, 4, 6, 3, 5,
> +; RNIT: 8, 16, 9, 11 => 14, 14,
> +; RNIT: 11, 13, 12,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/multiple_exiting_edge.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/multiple_exiting_edge.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/multiple_exiting_edge.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/multiple_exiting_edge.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,38 @@
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @normal_condition_0() nounwind {
> +bb38:                                             ; preds = %bb34, %bb34, %bb37
> +  switch i32 undef, label %bb42 [
> +    i32 67, label %bb42
> +    i32 90, label %bb41
> +  ]
> +bb41:                                             ; preds = %bb38
> +  br label %bb42
> +bb42:                                             ; preds = %bb38, %bb38, %bb41
> +  ret void
> +}
> +
> +; BBIT: bb38, bb42, bb41,
> +; BBIT: bb38, bb41,
> +
> +; RNIT: bb38 => bb42, bb42,
> +; RNIT: bb38, bb41,
> +
> +define void @normal_condition_1() nounwind {
> +bb38:                                             ; preds = %bb34, %bb34, %bb37
> +  switch i32 undef, label %bb41 [
> +    i32 67, label %bb42
> +    i32 90, label %bb42
> +  ]
> +bb41:                                             ; preds = %bb38
> +  br label %bb42
> +bb42:                                             ; preds = %bb38, %bb38, %bb41
> +  ret void
> +}
> +
> +; BBIT: bb38, bb41, bb42,
> +; BBIT: bb38, bb41,
> +
> +; RNIT: bb38 => bb42, bb42,
> +; RNIT: bb38, bb41,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/nested_loops.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/nested_loops.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/nested_loops.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/nested_loops.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,33 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc zeroext i8 @handle_compress() nounwind {
> +entry:
> +  br label %outer
> +
> +outer:
> +  br label %body
> +
> +body:
> +  br i1 1, label %exit172, label %end
> +
> +exit172:
> +  br i1 1, label %end, label %outer
> +
> +end:
> +  ret i8 1
> +}
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK-NEXT: [1] outer => end
> +
> +; STAT: 2 region - The # of regions
> +
> +; BBIT: entry, outer, body, exit172, end,
> +; BBIT: outer, body, exit172,
> +
> +; RNIT: entry, outer => end, end,
> +; RNIT: outer, body, exit172,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/next.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/next.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/next.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/next.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,49 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @MAIN__() nounwind {
> +entry:
> +  br label %__label_002001.outer
> +
> +__label_002001.outer:                             ; preds = %bb236, %bb92
> +  br label %__label_002001
> +
> +__label_002001:                                   ; preds = %bb229, %__label_002001.outer
> +  br i1  1, label %bb93, label %__label_000020
> +
> +bb93:                                             ; preds = %__label_002001
> +  br i1  1, label %__label_000020, label %bb197
> +
> +bb197:                                            ; preds = %bb193
> +  br i1  1, label %bb229, label %bb224
> +
> +bb224:                                            ; preds = %bb223, %bb227
> +  br i1  1, label %bb229, label %bb224
> +
> +bb229:                                            ; preds = %bb227, %bb223
> +  br i1  1, label %__label_002001, label %__label_002001.outer
> +
> +__label_000020:                                   ; preds = %__label_002001, %bb194
> +  ret void
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK-NEXT:  [1] __label_002001.outer => __label_000020
> +; CHECK-NEXT;      [2] bb197 => bb229
> +; CHECK-NEXT;            [3] bb224 => bb229
> +
> +; STAT: 4 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: entry, __label_002001.outer, __label_002001, bb93, __label_000020, bb197, bb229, bb224,
> +; BBIT: __label_002001.outer, __label_002001, bb93, bb197, bb229, bb224,
> +; BBIT: bb197, bb224,
> +; BBIT: bb224,
> +
> +; RNIT: entry, __label_002001.outer => __label_000020, __label_000020,
> +; RNIT: __label_002001.outer, __label_002001, bb93, bb197 => bb229, bb229,
> +; RNIT: bb197, bb224 => bb229,
> +; RNIT: bb224,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/paper.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/paper.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/paper.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/paper.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,55 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define void @a_linear_impl_fig_1() nounwind {
> +0:
> +        br label %"1"
> +1:
> +	br label %"2"
> +2:
> +	br label %"3"
> +3:
> +	br i1 1, label %"13", label %"4"
> +4:
> +	br i1 1, label %"5", label %"1"
> +5:
> +	br i1 1, label %"8", label %"6"
> +6:
> +	br i1 1, label %"7", label %"4"
> +7:
> +	ret void
> +8:
> +	br i1 1, label %"9", label %"1"
> +9:
> +	br label %"10"
> +10:
> +	br i1 1, label %"12", label %"11"
> +11:
> +	br i1 1, label %"9", label %"8"
> +13:
> +	br i1 1, label %"2", label %"1"
> +12:
> + 	switch i32 0, label %"1" [ i32 0, label %"9"
> +                                  i32 1, label %"8"]
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] 0 => <Function Return>
> +; CHECK-NEXT: [1] 1 => 7
> +; CHECK-NEXT:   [2] 1 => 4
> +; CHECK-NEXT:   [2] 8 => 1
> +
> +; STAT: 4 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: 0, 1, 2, 3, 13, 4, 5, 8, 9, 10, 12, 11, 6, 7,
> +; BBIT: 1, 2, 3, 13, 4, 5, 8, 9, 10, 12, 11, 6,
> +; BBIT: 1, 2, 3, 13,
> +; BBIT: 8, 9, 10, 12, 11,
> +
> +; RNIT: 0, 1 => 7, 7,
> +; RNIT: 1 => 4, 4, 5, 8 => 1, 6,
> +; RNIT: 1, 2, 3, 13,
> +; RNIT: 8, 9, 10, 12, 11,
> 
> Added: llvm/trunk/test/Analysis/RegionInfo/two_loops_same_header.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/RegionInfo/two_loops_same_header.ll?rev=109089&view=auto
> ==============================================================================
> --- llvm/trunk/test/Analysis/RegionInfo/two_loops_same_header.ll (added)
> +++ llvm/trunk/test/Analysis/RegionInfo/two_loops_same_header.ll Thu Jul 22 02:46:31 2010
> @@ -0,0 +1,46 @@
> +; RUN: opt -regions -analyze < %s | FileCheck %s
> +; RUN: opt -regions -stats < %s |& FileCheck -check-prefix=STAT %s
> +; RUN: opt -regions -print-region-style=bb  -analyze < %s |& FileCheck -check-prefix=BBIT %s
> +; RUN: opt -regions -print-region-style=rn  -analyze < %s |& FileCheck -check-prefix=RNIT %s
> +
> +define internal fastcc zeroext i8 @handle_compress() nounwind {
> +entry:
> +  br label %outer
> +
> +outer:
> +  br label %body
> +
> +body:
> +  br i1 1, label %else, label %true77
> +
> +true77:
> +  br i1 1, label %then83, label %else
> +
> +then83:
> +  br label %outer
> +
> +else:
> +  br label %else106
> +
> +else106:
> +  br i1 1, label %end, label %outer
> +
> +end:
> +  ret i8 1
> +}
> +
> +; CHECK-NOT: =>
> +; CHECK: [0] entry => <Function Return>
> +; CHECK-NEXT: [1] outer => end
> +; CHECK-NEXT:   [2] outer => else
> +
> +; STAT: 3 region - The # of regions
> +; STAT: 1 region - The # of simple regions
> +
> +; BBIT: entry, outer, body, else, else106, end, true77, then83,
> +; BBIT: outer, body, else, else106, true77, then83,
> +; BBIT: outer, body, true77, then83,
> +
> +; RNIT: entry, outer => end, end,
> +; RNIT: outer => else, else, else106,
> +; RNIT: outer, body, true77, then83,
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits


From grosser at fim.uni-passau.de  Tue Jul 27 02:32:29 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Tue, 27 Jul 2010 09:32:29 +0200
Subject: [llvm-commits] [llvm] r109089 - in /llvm/trunk: docs/
 include/llvm/ include/llvm/Analysis/ include/llvm/Support/ lib/Analysis/
 test/Analysis/RegionInfo/
In-Reply-To: <21598_1280215528_4C4E89E8_21598_4785_1_B2707C0D-15AF-4D83-84DE-6D6D6EFD2496@apple.com>
References: <20100722074632.575522A6C12D@llvm.org>
	<21598_1280215528_4C4E89E8_21598_4785_1_B2707C0D-15AF-4D83-84DE-6D6D6EFD2496@apple.com>
Message-ID: <4C4E8B8D.9020703@fim.uni-passau.de>

On 07/27/10 09:24, Evan Cheng wrote:
> Any chance you could templatify this so someone can add a MachineRegionInfo later? It would be useful for region based register allocation, etc.  
> 
> Thanks,
> 
> Evan

Hi Evan,

ether did once a templated version and I would love to test and commit
it. I just wanted to import an untemplated version of the RegionInfo
pass first, as it was easier to review.

@ether. Do you think you could port your templated version of the
RegionInfo pass to trunk?

Tobi


From grosser at fim.uni-passau.de  Tue Jul 27 03:39:43 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Tue, 27 Jul 2010 08:39:43 -0000
Subject: [llvm-commits] [llvm] r109496 - in /llvm/trunk:
 include/llvm/Analysis/RegionInfo.h lib/Analysis/RegionInfo.cpp
Message-ID: <20100727083943.CB19D2A6C12C@llvm.org>

Author: grosser
Date: Tue Jul 27 03:39:43 2010
New Revision: 109496

URL: http://llvm.org/viewvc/llvm-project?rev=109496&view=rev
Log:
RegionInfo: Add getMaxRegionExit()

getMaxRegionExit returns the exit of the maximal refined region starting
at a specific basic block.

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

Modified: llvm/trunk/include/llvm/Analysis/RegionInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/RegionInfo.h?rev=109496&r1=109495&r2=109496&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/RegionInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/RegionInfo.h Tue Jul 27 03:39:43 2010
@@ -572,6 +572,12 @@
   /// region containing BB.
   Region *operator[](BasicBlock *BB) const;
 
+  /// @brief Return the exit of the maximal refined region, that starts at a
+  /// BasicBlock.
+  ///
+  /// @param BB The BasicBlock the refined region starts.
+  BasicBlock *getMaxRegionExit(BasicBlock *BB) const;
+
   /// @brief Find the smallest region that contains two regions.
   ///
   /// @param A The first region.

Modified: llvm/trunk/lib/Analysis/RegionInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionInfo.cpp?rev=109496&r1=109495&r2=109496&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/RegionInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/RegionInfo.cpp Tue Jul 27 03:39:43 2010
@@ -651,6 +651,45 @@
   return getRegionFor(BB);
 }
 
+
+BasicBlock *RegionInfo::getMaxRegionExit(BasicBlock *BB) const {
+  BasicBlock *Exit = NULL;
+
+  while (true) {
+    // Get largest region that starts at BB.
+    Region *R = getRegionFor(BB);
+    while (R && R->getParent() && R->getParent()->getEntry() == BB)
+      R = R->getParent();
+
+    // Get the single exit of BB.
+    if (R && R->getEntry() == BB)
+      Exit = R->getExit();
+    else if (++succ_begin(BB) == succ_end(BB))
+      Exit = *succ_begin(BB);
+    else // No single exit exists.
+      return Exit;
+
+    // Get largest region that starts at Exit.
+    Region *ExitR = getRegionFor(Exit);
+    while (ExitR && ExitR->getParent()
+           && ExitR->getParent()->getEntry() == Exit)
+      ExitR = ExitR->getParent();
+
+    for (pred_iterator PI = pred_begin(Exit), PE = pred_end(Exit); PI != PE;
+         ++PI)
+      if (!R->contains(*PI) && !ExitR->contains(*PI))
+        break;
+
+    // This stops infinite cycles.
+    if (DT->dominates(Exit, BB))
+      break;
+
+    BB = Exit;
+  }
+
+  return Exit;
+}
+
 Region*
 RegionInfo::getCommonRegion(Region *A, Region *B) const {
   assert (A && B && "One of the Regions is NULL");




From grosser at fim.uni-passau.de  Tue Jul 27 04:01:26 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Tue, 27 Jul 2010 09:01:26 -0000
Subject: [llvm-commits] [llvm] r109497 -
	/llvm/trunk/test/Scripts/coff-dump.py
Message-ID: <20100727090126.E74892A6C12C@llvm.org>

Author: grosser
Date: Tue Jul 27 04:01:26 2010
New Revision: 109497

URL: http://llvm.org/viewvc/llvm-project?rev=109497&view=rev
Log:
Make coff-dump.py executable and add python as executable for this script.

This fixes the MC/COFF/basic-coff.ll test case.

Modified:
    llvm/trunk/test/Scripts/coff-dump.py   (contents, props changed)

Modified: llvm/trunk/test/Scripts/coff-dump.py
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Scripts/coff-dump.py?rev=109497&r1=109496&r2=109497&view=diff
==============================================================================
--- llvm/trunk/test/Scripts/coff-dump.py (original)
+++ llvm/trunk/test/Scripts/coff-dump.py Tue Jul 27 04:01:26 2010
@@ -1,3 +1,4 @@
+#!/usr/bin/env python
 #===-- coff-dump.py - COFF object file dump utility-------------------------===#
 #
 #                     The LLVM Compiler Infrastructure

Propchange: llvm/trunk/test/Scripts/coff-dump.py
------------------------------------------------------------------------------
    svn:executable = *




From baldrick at free.fr  Tue Jul 27 04:09:05 2010
From: baldrick at free.fr (Duncan Sands)
Date: Tue, 27 Jul 2010 09:09:05 -0000
Subject: [llvm-commits] [llvm] r109498 -
	/llvm/trunk/include/llvm/ADT/ValueMap.h
Message-ID: <20100727090906.0CF452A6C12C@llvm.org>

Author: baldrick
Date: Tue Jul 27 04:09:05 2010
New Revision: 109498

URL: http://llvm.org/viewvc/llvm-project?rev=109498&view=rev
Log:
In commit 91421, isPod was changed from false to true for these value handles.
Change it back again: destructors and constructors need to be run.

Modified:
    llvm/trunk/include/llvm/ADT/ValueMap.h

Modified: llvm/trunk/include/llvm/ADT/ValueMap.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/ValueMap.h?rev=109498&r1=109497&r2=109498&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/ValueMap.h (original)
+++ llvm/trunk/include/llvm/ADT/ValueMap.h Tue Jul 27 04:09:05 2010
@@ -250,12 +250,6 @@
   }
 };
 
-  
-template<typename KeyT, typename ValueT, typename Config, typename ValueInfoT>
-struct isPodLike<ValueMapCallbackVH<KeyT, ValueT, Config, ValueInfoT> > {
-  static const bool value = true;
-};
-
 template<typename KeyT, typename ValueT, typename Config, typename ValueInfoT>
 struct DenseMapInfo<ValueMapCallbackVH<KeyT, ValueT, Config, ValueInfoT> > {
   typedef ValueMapCallbackVH<KeyT, ValueT, Config, ValueInfoT> VH;




From foldr at codedgers.com  Tue Jul 27 06:19:41 2010
From: foldr at codedgers.com (Mikhail Glushenkov)
Date: Tue, 27 Jul 2010 11:19:41 -0000
Subject: [llvm-commits] [llvm] r109500 -
	/llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp
Message-ID: <20100727111941.256942A6C12D@llvm.org>

Author: foldr
Date: Tue Jul 27 06:19:40 2010
New Revision: 109500

URL: http://llvm.org/viewvc/llvm-project?rev=109500&view=rev
Log:
Fix silent failure with no input files.

Modified:
    llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp

Modified: llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp?rev=109500&r1=109499&r2=109500&view=diff
==============================================================================
--- llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp (original)
+++ llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp Tue Jul 27 06:19:40 2010
@@ -355,6 +355,7 @@
 int CompilationGraph::Build (const sys::Path& TempDir,
                              const LanguageMap& LangMap) {
   InputLanguagesSet InLangs;
+  bool WasSomeActionGenerated = !InputFilenames.empty();
 
   // Traverse initial parts of the toolchains and fill in InLangs.
   if (int ret = BuildInitial(InLangs, TempDir, LangMap))
@@ -375,6 +376,7 @@
     if (JT->JoinListEmpty() && !(JT->WorksOnEmpty() && InputFilenames.empty()))
       continue;
 
+    WasSomeActionGenerated = true;
     Action CurAction;
     if (int ret = JT->GenerateAction(CurAction, CurNode->HasChildren(),
                                      TempDir, InLangs, LangMap)) {
@@ -401,6 +403,11 @@
     }
   }
 
+  if (!WasSomeActionGenerated) {
+    PrintError("no input files");
+    return 1;
+  }
+
   return 0;
 }
 




From foldr at codedgers.com  Tue Jul 27 06:19:36 2010
From: foldr at codedgers.com (Mikhail Glushenkov)
Date: Tue, 27 Jul 2010 11:19:36 -0000
Subject: [llvm-commits] [llvm] r109499 - in /llvm/trunk:
 include/llvm/CompilerDriver/Action.h
 include/llvm/CompilerDriver/CompilationGraph.h
 include/llvm/CompilerDriver/Tool.h lib/CompilerDriver/CompilationGraph.cpp
 lib/CompilerDriver/Main.cpp utils/TableGen/LLVMCConfigurationEmitter.cpp
Message-ID: <20100727111936.EC0622A6C12C@llvm.org>

Author: foldr
Date: Tue Jul 27 06:19:36 2010
New Revision: 109499

URL: http://llvm.org/viewvc/llvm-project?rev=109499&view=rev
Log:
Return -1 only on failure to execute a program.

Also fix some comments.

Modified:
    llvm/trunk/include/llvm/CompilerDriver/Action.h
    llvm/trunk/include/llvm/CompilerDriver/CompilationGraph.h
    llvm/trunk/include/llvm/CompilerDriver/Tool.h
    llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp
    llvm/trunk/lib/CompilerDriver/Main.cpp
    llvm/trunk/utils/TableGen/LLVMCConfigurationEmitter.cpp

Modified: llvm/trunk/include/llvm/CompilerDriver/Action.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CompilerDriver/Action.h?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CompilerDriver/Action.h (original)
+++ llvm/trunk/include/llvm/CompilerDriver/Action.h Tue Jul 27 06:19:36 2010
@@ -43,7 +43,7 @@
     }
     bool IsConstructed () { return (Command_.size() != 0);}
 
-    /// Execute - Executes the represented action.
+    /// Execute - Executes the command. Returns -1 on error.
     int Execute () const;
     bool StopCompilation () const { return StopCompilation_; }
     const std::string& OutFile() { return OutFile_; }

Modified: llvm/trunk/include/llvm/CompilerDriver/CompilationGraph.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CompilerDriver/CompilationGraph.h?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CompilerDriver/CompilationGraph.h (original)
+++ llvm/trunk/include/llvm/CompilerDriver/CompilationGraph.h Tue Jul 27 06:19:36 2010
@@ -132,28 +132,28 @@
     void insertNode(Tool* T);
 
     /// insertEdge - Insert a new edge into the graph. Takes ownership
-    /// of the Edge object.
+    /// of the Edge object. Returns non-zero value on error.
     int insertEdge(const std::string& A, Edge* E);
 
-    /// Build - Build target(s) from the input file set. Command-line
-    /// options are passed implicitly as global variables.
+    /// Build - Build target(s) from the input file set. Command-line options
+    /// are passed implicitly as global variables. Returns non-zero value on
+    /// error (usually the failed program's exit code).
     int Build(llvm::sys::Path const& TempDir, const LanguageMap& LangMap);
 
-    /// Check - Check the compilation graph for common errors like
-    /// cycles, input/output language mismatch and multiple default
-    /// edges. Prints error messages and in case it finds any errors.
+    /// Check - Check the compilation graph for common errors like cycles,
+    /// input/output language mismatch and multiple default edges. Prints error
+    /// messages and in case it finds any errors.
     int Check();
 
-    /// getNode - Return a reference to the node correponding to the
-    /// given tool name. Throws std::runtime_error.
+    /// getNode - Return a reference to the node corresponding to the given tool
+    /// name. Returns 0 on error.
     Node* getNode(const std::string& ToolName);
     const Node* getNode(const std::string& ToolName) const;
 
-    /// viewGraph - This function is meant for use from the debugger.
-    /// You can just say 'call G->viewGraph()' and a ghostview window
-    /// should pop up from the program, displaying the compilation
-    /// graph. This depends on there being a 'dot' and 'gv' program
-    /// in your path.
+    /// viewGraph - This function is meant for use from the debugger. You can
+    /// just say 'call G->viewGraph()' and a ghostview window should pop up from
+    /// the program, displaying the compilation graph. This depends on there
+    /// being a 'dot' and 'gv' program in your path.
     void viewGraph();
 
     /// writeGraph - Write Graphviz .dot source file to the current direcotry.
@@ -167,12 +167,11 @@
     // Helper functions.
 
     /// getToolsVector - Return a reference to the list of tool names
-    /// corresponding to the given language name. Throws
-    /// std::runtime_error.
+    /// corresponding to the given language name. Returns 0 on error.
     const tools_vector_type* getToolsVector(const std::string& LangName) const;
 
-    /// PassThroughGraph - Pass the input file through the toolchain
-    /// starting at StartNode.
+    /// PassThroughGraph - Pass the input file through the toolchain starting at
+    /// StartNode.
     int PassThroughGraph (const llvm::sys::Path& In, const Node* StartNode,
                           const InputLanguagesSet& InLangs,
                           const llvm::sys::Path& TempDir,
@@ -185,26 +184,32 @@
                               InputLanguagesSet& InLangs,
                               const LanguageMap& LangMap) const;
 
-    /// BuildInitial - Traverse the initial parts of the toolchains.
+    /// BuildInitial - Traverse the initial parts of the toolchains. Returns
+    /// non-zero value on error.
     int BuildInitial(InputLanguagesSet& InLangs,
                      const llvm::sys::Path& TempDir,
                      const LanguageMap& LangMap);
 
-    /// TopologicalSort - Sort the nodes in topological order.
+    /// TopologicalSort - Sort the nodes in topological order. Returns non-zero
+    /// value on error.
     int TopologicalSort(std::vector<const Node*>& Out);
-    /// TopologicalSortFilterJoinNodes - Call TopologicalSort and
-    /// filter the resulting list to include only Join nodes.
+    /// TopologicalSortFilterJoinNodes - Call TopologicalSort and filter the
+    /// resulting list to include only Join nodes. Returns non-zero value on
+    /// error.
     int TopologicalSortFilterJoinNodes(std::vector<const Node*>& Out);
 
     // Functions used to implement Check().
 
-    /// CheckLanguageNames - Check that output/input language names
-    /// match for all nodes.
+    /// CheckLanguageNames - Check that output/input language names match for
+    /// all nodes. Returns non-zero value on error (number of errors
+    /// encountered).
     int CheckLanguageNames() const;
-    /// CheckMultipleDefaultEdges - check that there are no multiple
-    /// default default edges.
+    /// CheckMultipleDefaultEdges - check that there are no multiple default
+    /// default edges. Returns non-zero value on error (number of errors
+    /// encountered).
     int CheckMultipleDefaultEdges() const;
-    /// CheckCycles - Check that there are no cycles in the graph.
+    /// CheckCycles - Check that there are no cycles in the graph. Returns
+    /// non-zero value on error (number of errors encountered).
     int CheckCycles();
 
   };

Modified: llvm/trunk/include/llvm/CompilerDriver/Tool.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CompilerDriver/Tool.h?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/include/llvm/CompilerDriver/Tool.h (original)
+++ llvm/trunk/include/llvm/CompilerDriver/Tool.h Tue Jul 27 06:19:36 2010
@@ -38,6 +38,8 @@
 
     virtual ~Tool() {}
 
+    /// GenerateAction - Generate an Action given particular command-line
+    /// options. Returns non-zero value on error.
     virtual int GenerateAction (Action& Out,
                                 const PathVector& inFiles,
                                 const bool HasChildren,
@@ -45,6 +47,8 @@
                                 const InputLanguagesSet& InLangs,
                                 const LanguageMap& LangMap) const = 0;
 
+    /// GenerateAction - Generate an Action given particular command-line
+    /// options. Returns non-zero value on error.
     virtual int GenerateAction (Action& Out,
                                 const llvm::sys::Path& inFile,
                                 const bool HasChildren,

Modified: llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp (original)
+++ llvm/trunk/lib/CompilerDriver/CompilationGraph.cpp Tue Jul 27 06:19:36 2010
@@ -139,7 +139,7 @@
 int CompilationGraph::insertEdge(const std::string& A, Edge* Edg) {
   Node* B = getNode(Edg->ToolName());
   if (B == 0)
-    return -1;
+    return 1;
 
   if (A == "root") {
     const char** InLangs = B->ToolPtr->InputLanguages();
@@ -150,7 +150,7 @@
   else {
     Node* N = getNode(A);
     if (N == 0)
-      return -1;
+      return 1;
 
     N->AddEdge(Edg);
   }
@@ -193,11 +193,11 @@
 
     const Edge* Edg = ChooseEdge(CurNode->OutEdges, InLangs, CurNode->Name());
     if (Edg == 0)
-      return -1;
+      return 1;
 
     CurNode = getNode(Edg->ToolName());
     if (CurNode == 0)
-      return -1;
+      return 1;
 
     In = CurAction.OutFile();
   }
@@ -295,7 +295,7 @@
     // Find the toolchain corresponding to this file.
     const Node* N = FindToolChain(In, xLanguage, InLangs, LangMap);
     if (N == 0)
-      return -1;
+      return 1;
     // Pass file through the chain starting at head.
     if (int ret = PassThroughGraph(In, N, InLangs, TempDir, LangMap))
       return ret;
@@ -310,7 +310,7 @@
 
   Node* Root = getNode("root");
   if (Root == 0)
-    return -1;
+    return 1;
 
   Q.push(Root);
 
@@ -322,7 +322,7 @@
          EB != EE; ++EB) {
       Node* B = getNode((*EB)->ToolName());
       if (B == 0)
-        return -1;
+        return 1;
 
       B->DecrInEdges();
       if (B->HasNoInEdges())
@@ -389,11 +389,11 @@
 
     const Edge* Edg = ChooseEdge(CurNode->OutEdges, InLangs, CurNode->Name());
     if (Edg == 0)
-      return -1;
+      return 1;
 
     const Node* NextNode = getNode(Edg->ToolName());
     if (NextNode == 0)
-      return -1;
+      return 1;
 
     if (int ret = PassThroughGraph(sys::Path(CurAction.OutFile()), NextNode,
                                    InLangs, TempDir, LangMap)) {
@@ -417,7 +417,7 @@
            EB != EE; ++EB) {
         const Node* N2 = this->getNode((*EB)->ToolName());
         if (N2 == 0)
-          return -1;
+          return 1;
 
         if (!N2->ToolPtr) {
           ++ret;
@@ -497,7 +497,7 @@
 
   Node* Root = getNode("root");
   if (Root == 0)
-    return -1;
+    return 1;
 
   Q.push(Root);
 
@@ -513,7 +513,7 @@
          EB != EE; ++EB) {
       Node* B = getNode((*EB)->ToolName());
       if (B == 0)
-        return -1;
+        return 1;
 
       B->DecrInEdges();
       if (B->HasNoInEdges())
@@ -539,19 +539,19 @@
   // Check that output/input language names match.
   ret = this->CheckLanguageNames();
   if (ret < 0)
-    return -1;
+    return 1;
   errs += ret;
 
   // Check for multiple default edges.
   ret = this->CheckMultipleDefaultEdges();
   if (ret < 0)
-    return -1;
+    return 1;
   errs += ret;
 
   // Check for cycles.
   ret = this->CheckCycles();
   if (ret < 0)
-    return -1;
+    return 1;
   errs += ret;
 
   return errs;
@@ -617,7 +617,7 @@
   }
   else {
     PrintError("Error opening file '" + OutputFilename + "' for writing!");
-    return -1;
+    return 1;
   }
 
   return 0;

Modified: llvm/trunk/lib/CompilerDriver/Main.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CompilerDriver/Main.cpp?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/lib/CompilerDriver/Main.cpp (original)
+++ llvm/trunk/lib/CompilerDriver/Main.cpp Tue Jul 27 06:19:36 2010
@@ -30,7 +30,9 @@
 
   std::stringstream* GlobalTimeLog;
 
-  int getTempDir(sys::Path& tempDir) {
+  /// GetTempDir - Get the temporary directory location. Returns non-zero value
+  /// on error.
+  int GetTempDir(sys::Path& tempDir) {
     // The --temp-dir option.
     if (!TempDirname.empty()) {
       tempDir = TempDirname;
@@ -53,20 +55,20 @@
       std::string ErrMsg;
       if (tempDir.createDirectoryOnDisk(true, &ErrMsg)) {
         PrintError(ErrMsg);
-        return -1;
+        return 1;
       }
     }
 
     return 0;
   }
 
-  /// BuildTargets - A small wrapper for CompilationGraph::Build.
+  /// BuildTargets - A small wrapper for CompilationGraph::Build. Returns non-zero value
   int BuildTargets(CompilationGraph& graph, const LanguageMap& langMap) {
     int ret;
     sys::Path tempDir;
     bool toDelete = (SaveTemps == SaveTempsEnum::Unset);
 
-    if (int ret = getTempDir(tempDir))
+    if (int ret = GetTempDir(tempDir))
       return ret;
 
     ret = graph.Build(tempDir, langMap);

Modified: llvm/trunk/utils/TableGen/LLVMCConfigurationEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/LLVMCConfigurationEmitter.cpp?rev=109499&r1=109498&r2=109499&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/LLVMCConfigurationEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/LLVMCConfigurationEmitter.cpp Tue Jul 27 06:19:36 2010
@@ -1960,7 +1960,7 @@
       << "PrintError(\""
       << (d.getNumArgs() >= 1 ? InitPtrToString(d.getArg(0)) : "Unknown error!")
       << "\");\n";
-    O.indent(IndentLevel) << "return -1;\n";
+    O.indent(IndentLevel) << "return 1;\n";
   }
 
   void onWarningDag(const DagInit& d,




From ggreif at gmail.com  Tue Jul 27 08:31:22 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 13:31:22 -0000
Subject: [llvm-commits] [llvm] r109502 -
	/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
Message-ID: <20100727133122.B38802A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 08:31:22 2010
New Revision: 109502

URL: http://llvm.org/viewvc/llvm-project?rev=109502&view=rev
Log:
simplify

Modified:
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=109502&r1=109501&r2=109502&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Tue Jul 27 08:31:22 2010
@@ -373,7 +373,7 @@
       // If the call is readnone, ignore it, otherwise bail out.  We don't even
       // allow readonly here because we don't want something like:
       // A[1] = 2; strlen(A); A[2] = 2; -> memcpy(A, ...); strlen(A).
-      if (AA.getModRefBehavior(CallSite::get(BI)) ==
+      if (AA.getModRefBehavior(CallSite(BI)) ==
             AliasAnalysis::DoesNotAccessMemory)
         continue;
       




From ggreif at gmail.com  Tue Jul 27 09:15:29 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 14:15:29 -0000
Subject: [llvm-commits] [llvm] r109503 -
	/llvm/trunk/lib/Analysis/InlineCost.cpp
Message-ID: <20100727141529.518652A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 09:15:29 2010
New Revision: 109503

URL: http://llvm.org/viewvc/llvm-project?rev=109503&view=rev
Log:
use ImmutableCallSite for const-corrgoodness

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

Modified: llvm/trunk/lib/Analysis/InlineCost.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/InlineCost.cpp?rev=109503&r1=109502&r2=109503&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/InlineCost.cpp (original)
+++ llvm/trunk/lib/Analysis/InlineCost.cpp Tue Jul 27 09:15:29 2010
@@ -152,14 +152,14 @@
     if (isa<CallInst>(II) || isa<InvokeInst>(II)) {
       if (isa<DbgInfoIntrinsic>(II))
         continue;  // Debug intrinsics don't count as size.
-      
-      CallSite CS = CallSite::get(const_cast<Instruction*>(&*II));
-      
+
+      ImmutableCallSite CS(cast<Instruction>(II));
+
       // If this function contains a call to setjmp or _setjmp, never inline
       // it.  This is a hack because we depend on the user marking their local
       // variables as volatile if they are live across a setjmp call, and they
       // probably won't do this in callers.
-      if (Function *F = CS.getCalledFunction()) {
+      if (const Function *F = CS.getCalledFunction()) {
         if (F->isDeclaration() && 
             (F->getName() == "setjmp" || F->getName() == "_setjmp"))
           callsSetJmp = true;




From ggreif at gmail.com  Tue Jul 27 09:38:39 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 14:38:39 -0000
Subject: [llvm-commits] [llvm] r109504 -
	/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
Message-ID: <20100727143839.19C342A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 09:38:38 2010
New Revision: 109504

URL: http://llvm.org/viewvc/llvm-project?rev=109504&view=rev
Log:
simplify

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

Modified: llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp?rev=109504&r1=109503&r2=109504&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp (original)
+++ llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp Tue Jul 27 09:38:38 2010
@@ -120,10 +120,9 @@
       Pointer = CI->getArgOperand(0);
       // calls to free() erase the entire structure
       PointerSize = ~0ULL;
-    } else if (isa<CallInst>(Inst) || isa<InvokeInst>(Inst)) {
+    } else if (CallSite InstCS = Inst) {
       // Debug intrinsics don't cause dependences.
       if (isa<DbgInfoIntrinsic>(Inst)) continue;
-      CallSite InstCS = CallSite::get(Inst);
       // If these two calls do not interfere, look past it.
       switch (AA->getModRefInfo(CS, InstCS)) {
       case AliasAnalysis::NoModRef:
@@ -387,7 +386,7 @@
       MemSize = cast<ConstantInt>(II->getArgOperand(1))->getZExtValue();
       break;
     default:
-      CallSite QueryCS = CallSite::get(QueryInst);
+      CallSite QueryCS(QueryInst);
       bool isReadOnly = AA->onlyReadsMemory(QueryCS);
       LocalCache = getCallSiteDependencyFrom(QueryCS, isReadOnly, ScanPos,
                                              QueryParent);




From ggreif at gmail.com  Tue Jul 27 10:02:37 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 15:02:37 -0000
Subject: [llvm-commits] [llvm] r109506 -
	/llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp
Message-ID: <20100727150237.4E8962A6C134@llvm.org>

Author: ggreif
Date: Tue Jul 27 10:02:37 2010
New Revision: 109506

URL: http://llvm.org/viewvc/llvm-project?rev=109506&view=rev
Log:
simplify: CallSite::get --> CallSite constructor

Modified:
    llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp

Modified: llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp?rev=109506&r1=109505&r2=109506&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/InlineFunction.cpp Tue Jul 27 10:02:37 2010
@@ -215,12 +215,12 @@
     if (I->second->getFunction() == 0)
       if (Function *F = CallSite(NewCall).getCalledFunction()) {
         // Indirect call site resolved to direct call.
-        CallerNode->addCalledFunction(CallSite::get(NewCall), CG[F]);
-        
+        CallerNode->addCalledFunction(CallSite(NewCall), CG[F]);
+
         continue;
       }
-    
-    CallerNode->addCalledFunction(CallSite::get(NewCall), I->second);
+
+    CallerNode->addCalledFunction(CallSite(NewCall), I->second);
   }
   
   // Update the call graph by deleting the edge from Callee to Caller.  We must




From ggreif at gmail.com  Tue Jul 27 10:18:11 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 15:18:11 -0000
Subject: [llvm-commits] [llvm] r109508 -
	/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
Message-ID: <20100727151811.B3A5A2A6C133@llvm.org>

Author: ggreif
Date: Tue Jul 27 10:18:11 2010
New Revision: 109508

URL: http://llvm.org/viewvc/llvm-project?rev=109508&view=rev
Log:
back out r109504, breaks the bots

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

Modified: llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp?rev=109508&r1=109507&r2=109508&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp (original)
+++ llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp Tue Jul 27 10:18:11 2010
@@ -120,9 +120,10 @@
       Pointer = CI->getArgOperand(0);
       // calls to free() erase the entire structure
       PointerSize = ~0ULL;
-    } else if (CallSite InstCS = Inst) {
+    } else if (isa<CallInst>(Inst) || isa<InvokeInst>(Inst)) {
       // Debug intrinsics don't cause dependences.
       if (isa<DbgInfoIntrinsic>(Inst)) continue;
+      CallSite InstCS = CallSite::get(Inst);
       // If these two calls do not interfere, look past it.
       switch (AA->getModRefInfo(CS, InstCS)) {
       case AliasAnalysis::NoModRef:
@@ -386,7 +387,7 @@
       MemSize = cast<ConstantInt>(II->getArgOperand(1))->getZExtValue();
       break;
     default:
-      CallSite QueryCS(QueryInst);
+      CallSite QueryCS = CallSite::get(QueryInst);
       bool isReadOnly = AA->onlyReadsMemory(QueryCS);
       LocalCache = getCallSiteDependencyFrom(QueryCS, isReadOnly, ScanPos,
                                              QueryParent);




From stoklund at 2pi.dk  Tue Jul 27 10:42:58 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 27 Jul 2010 08:42:58 -0700
Subject: [llvm-commits] [llvm] r109089 - in /llvm/trunk: docs/
	include/llvm/ include/llvm/Analysis/ include/llvm/Support/
	lib/Analysis/ test/Analysis/RegionInfo/
In-Reply-To: <B2707C0D-15AF-4D83-84DE-6D6D6EFD2496@apple.com>
References: <20100722074632.575522A6C12D@llvm.org>
	<B2707C0D-15AF-4D83-84DE-6D6D6EFD2496@apple.com>
Message-ID: <0D6B10A0-DB6F-458A-A834-54322A3E1D56@2pi.dk>


On Jul 27, 2010, at 12:24 AM, Evan Cheng wrote:

> Any chance you could templatify this so someone can add a MachineRegionInfo later? It would be useful for region based register allocation, etc.  

Hear, hear.

Look at how it's done in LoopInfo.h


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1929 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100727/1957556c/attachment.bin 

From ggreif at gmail.com  Tue Jul 27 10:56:07 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 15:56:07 -0000
Subject: [llvm-commits] [llvm] r109509 -
	/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
Message-ID: <20100727155607.8BDD92A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 10:56:07 2010
New Revision: 109509

URL: http://llvm.org/viewvc/llvm-project?rev=109509&view=rev
Log:
back out this too to restore the bots

Modified:
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=109509&r1=109508&r2=109509&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Tue Jul 27 10:56:07 2010
@@ -373,7 +373,7 @@
       // If the call is readnone, ignore it, otherwise bail out.  We don't even
       // allow readonly here because we don't want something like:
       // A[1] = 2; strlen(A); A[2] = 2; -> memcpy(A, ...); strlen(A).
-      if (AA.getModRefBehavior(CallSite(BI)) ==
+      if (AA.getModRefBehavior(CallSite::get(BI)) ==
             AliasAnalysis::DoesNotAccessMemory)
         continue;
       




From clattner at apple.com  Tue Jul 27 11:04:48 2010
From: clattner at apple.com (Chris Lattner)
Date: Tue, 27 Jul 2010 09:04:48 -0700
Subject: [llvm-commits] [LLVMdev] DominanceFrontier for MachineBB
In-Reply-To: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>
References: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>
Message-ID: <970DFC2E-6B20-42BB-824F-7582231BB176@apple.com>


On Jul 27, 2010, at 8:16 AM, ether zhhb wrote:

>> ether did once a templated version and I would love to test and commit
>> it. I just wanted to import an untemplated version of the RegionInfo
>> pass first, as it was easier to review.
>> 
>> @ether. Do you think you could port your templated version of the
>> RegionInfo pass to trunk?
> RegionInfo need PostDominatorTree and DominanceFrontier, do us have
> DominanceFrontier for MachineBB?

What does RegionInfo use DomFrontier for?  I had hoped that we would eventually eliminate it.  It's currently only being used for a couple things (like mem2reg) and is extremely expensive to compute (besides the implementation being terrible, it's an inherently N^2 data structure).

-Chris


From etherzhhb at gmail.com  Tue Jul 27 10:16:41 2010
From: etherzhhb at gmail.com (ether zhhb)
Date: Tue, 27 Jul 2010 23:16:41 +0800
Subject: [llvm-commits] DominanceFrontier for MachineBB
Message-ID: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>

hi,

On Tue, Jul 27, 2010 at 3:32 PM, Tobias Grosser
<grosser at fim.uni-passau.de> wrote:
> On 07/27/10 09:24, Evan Cheng wrote:
>> Any chance you could templatify this so someone can add a MachineRegionInfo later? It would be useful for region based register allocation, etc.
>>
>> Thanks,
>>
>> Evan
>
> Hi Evan,
>
> ether did once a templated version and I would love to test and commit
> it. I just wanted to import an untemplated version of the RegionInfo
> pass first, as it was easier to review.
>
> @ether. Do you think you could port your templated version of the
> RegionInfo pass to trunk?
RegionInfo need PostDominatorTree and DominanceFrontier, do us have
DominanceFrontier for MachineBB?
>
> Tobi
>

--
best regards
ether


From grosser at fim.uni-passau.de  Tue Jul 27 11:42:25 2010
From: grosser at fim.uni-passau.de (Tobias Grosser)
Date: Tue, 27 Jul 2010 18:42:25 +0200
Subject: [llvm-commits] [LLVMdev] DominanceFrontier for MachineBB
In-Reply-To: <15152_1280246715_4C4F03BB_15152_2275_1_970DFC2E-6B20-42BB-824F-7582231BB176@apple.com>
References: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>
	<15152_1280246715_4C4F03BB_15152_2275_1_970DFC2E-6B20-42BB-824F-7582231BB176@apple.com>
Message-ID: <4C4F0C71.10104@fim.uni-passau.de>

On 07/27/10 18:04, Chris Lattner wrote:
> 
> On Jul 27, 2010, at 8:16 AM, ether zhhb wrote:
> 
>>> ether did once a templated version and I would love to test and commit
>>> it. I just wanted to import an untemplated version of the RegionInfo
>>> pass first, as it was easier to review.
>>>
>>> @ether. Do you think you could port your templated version of the
>>> RegionInfo pass to trunk?
>> RegionInfo need PostDominatorTree and DominanceFrontier, do us have
>> DominanceFrontier for MachineBB?
> 
> What does RegionInfo use DomFrontier for?  I had hoped that we would eventually eliminate it.  It's currently only being used for a couple things (like mem2reg) and is extremely expensive to compute (besides the implementation being terrible, it's an inherently N^2 data structure).

It is currently used to check if two BasicBlocks form a region. The two
calls to DominanceFrontier are in RegionInfo::isRegion(). I do not know
of an easy way to remove the usage of this analysis, however I have to
admit that I did not spend too much time on trying this.
I know the complexity is N^2, however my tests showed that at least for
the spec 2006 and polyhedron fortran benchmarks dominance frontier was
pretty fast to calculate. Actually it was faster than all other analysis:

http://tobias.osaft.eu/regioninfo/sum_runtime.png (relative timings)

I did not spend a lot of time to optimize this. Do you know of any non
artificial test case (or even an artificial one), where dominance
frontier calculation is slowing down compilation? In case there is a
chance to improve performance of regioninfo or dominance frontier
calculation I would love to do this.

Cheers
Tobi



From ggreif at gmail.com  Tue Jul 27 11:44:23 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 16:44:23 -0000
Subject: [llvm-commits] [llvm] r109510 -
	/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
Message-ID: <20100727164423.45F652A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 11:44:23 2010
New Revision: 109510

URL: http://llvm.org/viewvc/llvm-project?rev=109510&view=rev
Log:
recommit simplification (r109502, backed out r109509); seems to innocent

Modified:
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=109510&r1=109509&r2=109510&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Tue Jul 27 11:44:23 2010
@@ -373,7 +373,7 @@
       // If the call is readnone, ignore it, otherwise bail out.  We don't even
       // allow readonly here because we don't want something like:
       // A[1] = 2; strlen(A); A[2] = 2; -> memcpy(A, ...); strlen(A).
-      if (AA.getModRefBehavior(CallSite::get(BI)) ==
+      if (AA.getModRefBehavior(CallSite(BI)) ==
             AliasAnalysis::DoesNotAccessMemory)
         continue;
       




From stoklund at 2pi.dk  Tue Jul 27 12:01:35 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Tue, 27 Jul 2010 10:01:35 -0700
Subject: [llvm-commits] [LLVMdev] DominanceFrontier for MachineBB
In-Reply-To: <4C4F0C71.10104@fim.uni-passau.de>
References: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>
	<15152_1280246715_4C4F03BB_15152_2275_1_970DFC2E-6B20-42BB-824F-7582231BB176@apple.com>
	<4C4F0C71.10104@fim.uni-passau.de>
Message-ID: <DB7EB1AF-A974-4A14-8549-FEF1EB8A18BD@2pi.dk>


On Jul 27, 2010, at 9:42 AM, Tobias Grosser wrote:

> I did not spend a lot of time to optimize this. Do you know of any non
> artificial test case (or even an artificial one), where dominance
> frontier calculation is slowing down compilation? In case there is a
> chance to improve performance of regioninfo or dominance frontier
> calculation I would love to do this.

Usually, re2c is pretty good at producing C source with an insane amount of basic blocks. People use re2c, so it is not an artificial case.

http://sourceforge.net/projects/re2c




From grosbach at apple.com  Tue Jul 27 12:14:29 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 27 Jul 2010 17:14:29 -0000
Subject: [llvm-commits] [llvm] r109511 -
	/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
Message-ID: <20100727171429.2C7C62A6C12C@llvm.org>

Author: grosbach
Date: Tue Jul 27 12:14:29 2010
New Revision: 109511

URL: http://llvm.org/viewvc/llvm-project?rev=109511&view=rev
Log:
fix typo

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

Modified: llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp?rev=109511&r1=109510&r2=109511&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp Tue Jul 27 12:14:29 2010
@@ -1864,7 +1864,7 @@
 
 
 /// rewriteMBB - Keep track of which spills are available even after the
-/// register allocator is done with them.  If possible, avid reloading vregs.
+/// register allocator is done with them.  If possible, avoid reloading vregs.
 void
 LocalRewriter::RewriteMBB(LiveIntervals *LIs,
                           AvailableSpills &Spills, BitVector &RegKills,




From grosbach at apple.com  Tue Jul 27 12:38:47 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 27 Jul 2010 17:38:47 -0000
Subject: [llvm-commits] [llvm] r109513 -
	/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
Message-ID: <20100727173847.4F8472A6C12C@llvm.org>

Author: grosbach
Date: Tue Jul 27 12:38:47 2010
New Revision: 109513

URL: http://llvm.org/viewvc/llvm-project?rev=109513&view=rev
Log:
80 column

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

Modified: llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp?rev=109513&r1=109512&r2=109513&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp (original)
+++ llvm/trunk/lib/CodeGen/VirtRegRewriter.cpp Tue Jul 27 12:38:47 2010
@@ -460,7 +460,7 @@
 /// blocks each of which is a successor of the specified BB and has no other
 /// predecessor.
 static void findSinglePredSuccessor(MachineBasicBlock *MBB,
-                                   SmallVectorImpl<MachineBasicBlock *> &Succs) {
+                                   SmallVectorImpl<MachineBasicBlock *> &Succs){
   for (MachineBasicBlock::succ_iterator SI = MBB->succ_begin(),
          SE = MBB->succ_end(); SI != SE; ++SI) {
     MachineBasicBlock *SuccMBB = *SI;
@@ -852,8 +852,8 @@
       // Yup, use the reload register that we didn't use before.
       unsigned NewReg = Op.AssignedPhysReg;
       Rejected.insert(PhysReg);
-      return GetRegForReload(RC, NewReg, MF, MI, Spills, MaybeDeadStores, Rejected,
-                             RegKills, KillOps, VRM);
+      return GetRegForReload(RC, NewReg, MF, MI, Spills, MaybeDeadStores,
+                             Rejected, RegKills, KillOps, VRM);
     } else {
       // Otherwise, we might also have a problem if a previously reused
       // value aliases the new register. If so, codegen the previous reload
@@ -2302,7 +2302,7 @@
           unsigned PhysReg = Spills.getSpillSlotOrReMatPhysReg(SS);
           SmallVector<MachineInstr*, 4> NewMIs;
           if (PhysReg &&
-              TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, false, NewMIs)) {
+              TII->unfoldMemoryOperand(MF, &MI, PhysReg, false, false, NewMIs)){
             MBB->insert(MII, NewMIs[0]);
             InvalidateKills(MI, TRI, RegKills, KillOps);
             VRM->RemoveMachineInstrFromMaps(&MI);




From natebegeman at mac.com  Tue Jul 27 13:02:19 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Tue, 27 Jul 2010 18:02:19 -0000
Subject: [llvm-commits] [llvm] r109519 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/DAGCombiner.cpp
 test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll
Message-ID: <20100727180219.288C12A6C12C@llvm.org>

Author: sampo
Date: Tue Jul 27 13:02:18 2010
New Revision: 109519

URL: http://llvm.org/viewvc/llvm-project?rev=109519&view=rev
Log:
Fix a crash in the dag combiner caused by ConstantFoldBIT_CONVERTofBUILD_VECTOR calling itself
recursively and returning a SCALAR_TO_VECTOR node, but assuming the input was always a BUILD_VECTOR.

Added:
    llvm/trunk/test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=109519&r1=109518&r2=109519&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Tue Jul 27 13:02:18 2010
@@ -4489,6 +4489,16 @@
   // If this is a conversion of N elements of one type to N elements of another
   // type, convert each element.  This handles FP<->INT cases.
   if (SrcBitSize == DstBitSize) {
+    EVT VT = EVT::getVectorVT(*DAG.getContext(), DstEltVT,
+                              BV->getValueType(0).getVectorNumElements());
+
+    // Due to the FP element handling below calling this routine recursively,
+    // we can end up with a scalar-to-vector node here.
+    if (BV->getOpcode() == ISD::SCALAR_TO_VECTOR)
+      return DAG.getNode(ISD::SCALAR_TO_VECTOR, BV->getDebugLoc(), VT, 
+                         DAG.getNode(ISD::BIT_CONVERT, BV->getDebugLoc(),
+                                     DstEltVT, BV->getOperand(0)));
+      
     SmallVector<SDValue, 8> Ops;
     for (unsigned i = 0, e = BV->getNumOperands(); i != e; ++i) {
       SDValue Op = BV->getOperand(i);
@@ -4500,8 +4510,6 @@
                                 DstEltVT, Op));
       AddToWorkList(Ops.back().getNode());
     }
-    EVT VT = EVT::getVectorVT(*DAG.getContext(), DstEltVT,
-                              BV->getValueType(0).getVectorNumElements());
     return DAG.getNode(ISD::BUILD_VECTOR, BV->getDebugLoc(), VT,
                        &Ops[0], Ops.size());
   }

Added: llvm/trunk/test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll?rev=109519&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll (added)
+++ llvm/trunk/test/CodeGen/Generic/2010-07-27-DAGCombineCrash.ll Tue Jul 27 13:02:18 2010
@@ -0,0 +1,6 @@
+; RUN: llc < %s
+
+define float @test1()
+{
+	ret float extractelement (<2 x float> bitcast (<1 x double> <double 0x3f800000> to <2 x float>), i32 1);
+}




From bob.wilson at apple.com  Tue Jul 27 13:03:46 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Tue, 27 Jul 2010 18:03:46 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109520 -
	/llvm-gcc-4.2/trunk/build_gcc
Message-ID: <20100727180346.5FB602A6C12C@llvm.org>

Author: bwilson
Date: Tue Jul 27 13:03:46 2010
New Revision: 109520

URL: http://llvm.org/viewvc/llvm-project?rev=109520&view=rev
Log:
Control llvm-gcc's --enable-checking/--disable-checking options with
ENABLE_ASSERTIONS instead of LLVM_DEBUG.  Configure with --enable-werror
regardless.

Modified:
    llvm-gcc-4.2/trunk/build_gcc

Modified: llvm-gcc-4.2/trunk/build_gcc
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/build_gcc?rev=109520&r1=109519&r2=109520&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/build_gcc (original)
+++ llvm-gcc-4.2/trunk/build_gcc Tue Jul 27 13:03:46 2010
@@ -188,13 +188,14 @@
 
 # These are the configure and build flags that are used.
 # LLVM LOCAL begin
-if [ "x$LLVM_DEBUG" == "x" ]; then
-    CHECKING_FLAGS="--disable-checking --enable-werror"
+if [ "x$LLVM_ASSERTIONS" == "x" ]; then
+    CHECKING_FLAGS="--disable-checking"
 else
     CHECKING_FLAGS="--enable-checking"
 fi
 
 CONFIGFLAGS="$CHECKING_FLAGS \
+  --enable-werror \
   --prefix=$DEST_ROOT \
   --mandir=\${prefix}/share/man \
   --enable-languages=$LANGUAGES \




From dpatel at apple.com  Tue Jul 27 13:13:53 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 27 Jul 2010 18:13:53 -0000
Subject: [llvm-commits] [llvm] r109521 - in /llvm/trunk/test:
 FrontendC++/2009-07-15-LineNumbers.cpp FrontendC++/2010-07-23-DeclLoc.cpp
 FrontendObjC/2009-08-17-DebugInfo.m
Message-ID: <20100727181353.5BDDC2A6C12C@llvm.org>

Author: dpatel
Date: Tue Jul 27 13:13:53 2010
New Revision: 109521

URL: http://llvm.org/viewvc/llvm-project?rev=109521&view=rev
Log:
Update tests to not rely on input file's absolute path.

Modified:
    llvm/trunk/test/FrontendC++/2009-07-15-LineNumbers.cpp
    llvm/trunk/test/FrontendC++/2010-07-23-DeclLoc.cpp
    llvm/trunk/test/FrontendObjC/2009-08-17-DebugInfo.m

Modified: llvm/trunk/test/FrontendC++/2009-07-15-LineNumbers.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC%2B%2B/2009-07-15-LineNumbers.cpp?rev=109521&r1=109520&r2=109521&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC++/2009-07-15-LineNumbers.cpp (original)
+++ llvm/trunk/test/FrontendC++/2009-07-15-LineNumbers.cpp Tue Jul 27 13:13:53 2010
@@ -1,7 +1,7 @@
 // This is a regression test on debug info to make sure that we can
 // print line numbers in asm.
 // RUN: %llvmgcc -S -O0 -g %s -o - | \
-// RUN:    llc --disable-fp-elim -O0 -relocation-model=pic | grep { 2009-07-15-LineNumbers.cpp:25$}
+// RUN:    llc --disable-fp-elim -O0 -relocation-model=pic | grep {2009-07-15-LineNumbers.cpp:25$}
 
 #include <stdlib.h>
 

Modified: llvm/trunk/test/FrontendC++/2010-07-23-DeclLoc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC%2B%2B/2010-07-23-DeclLoc.cpp?rev=109521&r1=109520&r2=109521&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC++/2010-07-23-DeclLoc.cpp (original)
+++ llvm/trunk/test/FrontendC++/2010-07-23-DeclLoc.cpp Tue Jul 27 13:13:53 2010
@@ -63,13 +63,13 @@
     _TEXT_text_section(0, 0, __null), _dsym_path(__null), _dylib_current_version(0), _dylib_compatibility_version(0) {}
 };
 template <typename SIZE_AND_ENDIANNESS> class TExtendedMachOHeader {};
-# 16 "/decl_should_be_here.hpp"
+# 16 "decl_should_be_here.hpp"
 template <typename SIZE_AND_ENDIANNESS> void extract_dwarf_data_from_header(TExtendedMachOHeader<SIZE_AND_ENDIANNESS>& header,
                                                                             TRawSymbolOwnerData<typename SIZE_AND_ENDIANNESS::SIZE>& symbol_owner_data,
                                                                             TAddressRelocator<typename SIZE_AND_ENDIANNESS::SIZE>* address_relocator) {}
 struct CSCppSymbolOwnerHashFunctor {
   size_t operator()(const CSCppSymbolOwner& symbol_owner) const {
-# 97 "/wrong_place_for_decl.cpp"
+# 97 "wrong_place_for_decl.cpp"
   }
 };
 template <typename SIZE_AND_ENDIANNESS> CSCppSymbolOwnerData* create_symbol_owner_data_arch_specific(CSCppSymbolOwner* symbol_owner, const char* dsym_path) {

Modified: llvm/trunk/test/FrontendObjC/2009-08-17-DebugInfo.m
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendObjC/2009-08-17-DebugInfo.m?rev=109521&r1=109520&r2=109521&view=diff
==============================================================================
--- llvm/trunk/test/FrontendObjC/2009-08-17-DebugInfo.m (original)
+++ llvm/trunk/test/FrontendObjC/2009-08-17-DebugInfo.m Tue Jul 27 13:13:53 2010
@@ -5,7 +5,7 @@
 // RUN: %link %t.o -o %t.exe -framework Foundation
 // RUN: echo {break randomFunc\n} > %t.in 
 // RUN: gdb -q -batch -n -x %t.in %t.exe | tee %t.out | \
-// RUN:   grep {Breakpoint 1 at 0x.*: file 2009-08-17-DebugInfo.m, line 21}
+// RUN:   grep {Breakpoint 1 at 0x.*: file .*2009-08-17-DebugInfo.m, line 21}
 // XTARGET: darwin
 @interface MyClass
 {




From dpatel at apple.com  Tue Jul 27 13:15:02 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 27 Jul 2010 18:15:02 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109522 - in /llvm-gcc-4.2/trunk/gcc:
 llvm-debug.cpp llvm-debug.h
Message-ID: <20100727181502.1F3472A6C12C@llvm.org>

Author: dpatel
Date: Tue Jul 27 13:15:01 2010
New Revision: 109522

URL: http://llvm.org/viewvc/llvm-project?rev=109522&view=rev
Log:
Reapply 109316.
r109521 updated tests to not rely on absolute path.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=109522&r1=109521&r2=109522&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Tue Jul 27 13:15:01 2010
@@ -57,28 +57,16 @@
 #endif
 
 
-/// DirectoryAndFile - Extract the directory and file name from a path.  If no
-/// directory is specified, then use the source working directory.
+/// DirectoryAndFile - If Filename is absolute then directory name is empty.
+/// Otherwise directory name is current working directory.
 static void DirectoryAndFile(const std::string &FullPath,
                              std::string &Directory, std::string &FileName) {
-  // Look for the directory slash.
-  size_t Slash = FullPath.rfind('/');
-  
-  // If no slash
-  if (Slash == std::string::npos) {
-    // The entire path is the file name.
+  FileName = FullPath;
+
+  if (FullPath[0] != '/')
+    Directory = std::string(get_src_pwd());
+  else
     Directory = "";
-    FileName = FullPath;
-  } else {
-    // Separate the directory from the file name.
-    Directory = FullPath.substr(0, Slash);
-    FileName = FullPath.substr(Slash + 1);
-  }
-  
-  // If no directory present then use source working directory.
-  if (Directory.empty() || Directory[0] != '/') {
-    Directory = std::string(get_src_pwd()) + "/" + Directory;
-  }
 }
 
 /// NodeSizeInBits - Returns the size in bits stored in a tree node regardless
@@ -1394,10 +1382,13 @@
   }
 
   // Get source file information.
-  std::string Directory;
-  std::string FileName;
-  DirectoryAndFile(FullPath, Directory, FileName);
-  
+  llvm::StringRef Dir;
+  if (FullPath[0] != '/') {
+    if (CWD.empty())
+      CWD = std::string(get_src_pwd());
+    Dir = llvm::StringRef(CWD);
+  }
+
   // Set up Language number.
   unsigned LangTag;
   const std::string LanguageName(lang_hooks.name);
@@ -1432,8 +1423,8 @@
   unsigned ObjcRunTimeVer = 0;
   if (flag_objc_abi != 0 && flag_objc_abi != -1)
     ObjcRunTimeVer = flag_objc_abi;
-  return DebugFactory.CreateCompileUnit(LangTag, FileName.c_str(),
-                                        Directory.c_str(),
+  return DebugFactory.CreateCompileUnit(LangTag, FullPath,
+                                        Dir,
                                         version_string, isMain,
                                         optimize, Flags,
                                         ObjcRunTimeVer);
@@ -1449,10 +1440,13 @@
   }
 
   // Get source file information.
-  std::string Directory;
-  std::string FileName;
-  DirectoryAndFile(FullPath, Directory, FileName);
-  return DebugFactory.CreateFile(FileName, Directory, TheCU);
+  llvm::StringRef Dir;
+  if (FullPath[0] != '/') {
+    if (CWD.empty())
+      CWD = std::string(get_src_pwd());
+    Dir = llvm::StringRef(CWD);
+  }
+  return DebugFactory.CreateFile(FullPath, Dir, TheCU);
 }
 
 /* LLVM LOCAL end (ENTIRE FILE!)  */

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.h
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.h?rev=109522&r1=109521&r2=109522&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.h (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.h Tue Jul 27 13:15:01 2010
@@ -63,6 +63,7 @@
   int PrevLineNo;                       // Previous location line# encountered.
   BasicBlock *PrevBB;                   // Last basic block encountered.
   DICompileUnit TheCU;                  // The compile unit.
+  std::string CWD;                      // Current working directory.
 
   // Current GCC lexical block (or enclosing FUNCTION_DECL).
   tree_node *CurrentGCCLexicalBlock;	




From clattner at apple.com  Tue Jul 27 13:22:41 2010
From: clattner at apple.com (Chris Lattner)
Date: Tue, 27 Jul 2010 11:22:41 -0700
Subject: [llvm-commits] [LLVMdev] DominanceFrontier for MachineBB
In-Reply-To: <4C4F0C71.10104@fim.uni-passau.de>
References: <AANLkTineJcbAZHPxg6TTcJr0m765zG1f3XLSB9u_n3_N@mail.gmail.com>
	<15152_1280246715_4C4F03BB_15152_2275_1_970DFC2E-6B20-42BB-824F-7582231BB176@apple.com>
	<4C4F0C71.10104@fim.uni-passau.de>
Message-ID: <AA85D57E-1A28-429F-8825-DDC07A7A502E@apple.com>

>> What does RegionInfo use DomFrontier for?  I had hoped that we would eventually eliminate it.  It's currently only being used for a couple things (like mem2reg) and is extremely expensive to compute (besides the implementation being terrible, it's an inherently N^2 data structure).
> 
> It is currently used to check if two BasicBlocks form a region. The two
> calls to DominanceFrontier are in RegionInfo::isRegion(). I do not know
> of an easy way to remove the usage of this analysis, however I have to
> admit that I did not spend too much time on trying this.
> I know the complexity is N^2, however my tests showed that at least for
> the spec 2006 and polyhedron fortran benchmarks dominance frontier was
> pretty fast to calculate. Actually it was faster than all other analysis:
> 
> http://tobias.osaft.eu/regioninfo/sum_runtime.png (relative timings)
> 
> I did not spend a lot of time to optimize this. Do you know of any non
> artificial test case (or even an artificial one), where dominance
> frontier calculation is slowing down compilation? In case there is a
> chance to improve performance of regioninfo or dominance frontier
> calculation I would love to do this.

Independently of speed, we still want to remove it. :)  It would be great if you could find a solution that doesn't depend on domfrontier.

-Chris


From grosbach at apple.com  Tue Jul 27 13:36:28 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Tue, 27 Jul 2010 18:36:28 -0000
Subject: [llvm-commits] [llvm] r109525 -
	/llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp
Message-ID: <20100727183628.238232A6C12C@llvm.org>

Author: grosbach
Date: Tue Jul 27 13:36:27 2010
New Revision: 109525

URL: http://llvm.org/viewvc/llvm-project?rev=109525&view=rev
Log:
Grammar

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

Modified: llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp?rev=109525&r1=109524&r2=109525&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp (original)
+++ llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp Tue Jul 27 13:36:27 2010
@@ -255,9 +255,9 @@
                             SmallVector<LiveInterval*, 8> &SpillIntervals);
 
     /// attemptTrivialCoalescing - If a simple interval is defined by a copy,
-    /// try allocate the definition the same register as the source register
-    /// if the register is not defined during live time of the interval. This
-    /// eliminate a copy. This is used to coalesce copies which were not
+    /// try to allocate the definition to the same register as the source,
+    /// if the register is not defined during the life time of the interval.
+    /// This eliminates a copy, and is used to coalesce copies which were not
     /// coalesced away before allocation either due to dest and src being in
     /// different register classes or because the coalescer was overly
     /// conservative.




From sabre at nondot.org  Tue Jul 27 13:59:50 2010
From: sabre at nondot.org (Chris Lattner)
Date: Tue, 27 Jul 2010 18:59:50 -0000
Subject: [llvm-commits] [llvm] r109530 -
	/llvm/trunk/include/llvm/ADT/ScopedHashTable.h
Message-ID: <20100727185950.A7C0D2A6C12C@llvm.org>

Author: lattner
Date: Tue Jul 27 13:59:50 2010
New Revision: 109530

URL: http://llvm.org/viewvc/llvm-project?rev=109530&view=rev
Log:
make lookup failures not fatal.

Modified:
    llvm/trunk/include/llvm/ADT/ScopedHashTable.h

Modified: llvm/trunk/include/llvm/ADT/ScopedHashTable.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/ScopedHashTable.h?rev=109530&r1=109529&r2=109530&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/ScopedHashTable.h (original)
+++ llvm/trunk/include/llvm/ADT/ScopedHashTable.h Tue Jul 27 13:59:50 2010
@@ -139,7 +139,12 @@
   }
 
   V lookup(const K &Key) {
-    return TopLevelMap[Key]->getValue();
+    typename DenseMap<K, ScopedHashTableVal<K, V, KInfo>*, KInfo>::iterator
+      I = TopLevelMap.find(Key);
+    if (I != TopLevelMap.end())
+      return I->second->getValue();
+      
+    return V();
   }
 
   void insert(const K &Key, const V &Val) {




From dpatel at apple.com  Tue Jul 27 15:50:43 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 27 Jul 2010 20:50:43 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109536 -
	/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
Message-ID: <20100727205043.E09CD2A6C12C@llvm.org>

Author: dpatel
Date: Tue Jul 27 15:50:43 2010
New Revision: 109536

URL: http://llvm.org/viewvc/llvm-project?rev=109536&view=rev
Log:
Always use current working directory as an DW_AT_comp_dir value.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp?rev=109536&r1=109535&r2=109536&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-debug.cpp Tue Jul 27 15:50:43 2010
@@ -1381,14 +1381,6 @@
       FullPath = main_input_filename;
   }
 
-  // Get source file information.
-  llvm::StringRef Dir;
-  if (FullPath[0] != '/') {
-    if (CWD.empty())
-      CWD = std::string(get_src_pwd());
-    Dir = llvm::StringRef(CWD);
-  }
-
   // Set up Language number.
   unsigned LangTag;
   const std::string LanguageName(lang_hooks.name);
@@ -1424,7 +1416,7 @@
   if (flag_objc_abi != 0 && flag_objc_abi != -1)
     ObjcRunTimeVer = flag_objc_abi;
   return DebugFactory.CreateCompileUnit(LangTag, FullPath,
-                                        Dir,
+                                        get_src_pwd(),
                                         version_string, isMain,
                                         optimize, Flags,
                                         ObjcRunTimeVer);
@@ -1439,14 +1431,7 @@
       FullPath = main_input_filename;
   }
 
-  // Get source file information.
-  llvm::StringRef Dir;
-  if (FullPath[0] != '/') {
-    if (CWD.empty())
-      CWD = std::string(get_src_pwd());
-    Dir = llvm::StringRef(CWD);
-  }
-  return DebugFactory.CreateFile(FullPath, Dir, TheCU);
+  return DebugFactory.CreateFile(FullPath, get_src_pwd(), TheCU);
 }
 
 /* LLVM LOCAL end (ENTIRE FILE!)  */




From dpatel at apple.com  Tue Jul 27 15:51:15 2010
From: dpatel at apple.com (Devang Patel)
Date: Tue, 27 Jul 2010 20:51:15 -0000
Subject: [llvm-commits] [llvm] r109538 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <20100727205116.02C942A6C12C@llvm.org>

Author: dpatel
Date: Tue Jul 27 15:51:15 2010
New Revision: 109538

URL: http://llvm.org/viewvc/llvm-project?rev=109538&view=rev
Log:
It is FE's responsibility to emit proper directory name.

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

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=109538&r1=109537&r2=109538&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Tue Jul 27 15:51:15 2010
@@ -1753,10 +1753,7 @@
 /// maps as well.
 unsigned DwarfDebug::GetOrCreateSourceID(StringRef DirName, StringRef FileName){
   unsigned DId;
-  if (DirName.empty()) {
-    llvm::sys::Path CWD = llvm::sys::Path::GetCurrentDirectory();
-    DirName = StringRef(CWD.c_str(), CWD.size());
-  }
+  assert (DirName.empty() == false && "Invalid directory name!");
 
   StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
   if (DI != DirectoryIdMap.end()) {




From ggreif at gmail.com  Tue Jul 27 16:46:11 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 21:46:11 -0000
Subject: [llvm-commits] [llvm] r109544 -
	/llvm/trunk/include/llvm/Support/CallSite.h
Message-ID: <20100727214612.01C762A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 16:46:11 2010
New Revision: 109544

URL: http://llvm.org/viewvc/llvm-project?rev=109544&view=rev
Log:
remove bogus assert, use static_cast for additional checking

left two new asserts commented out, because they would fire in clang, have to hunt those down first

Modified:
    llvm/trunk/include/llvm/Support/CallSite.h

Modified: llvm/trunk/include/llvm/Support/CallSite.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/CallSite.h?rev=109544&r1=109543&r2=109544&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/CallSite.h (original)
+++ llvm/trunk/include/llvm/Support/CallSite.h Tue Jul 27 16:46:11 2010
@@ -49,13 +49,12 @@
   PointerIntPair<InstrTy*, 1, bool> I;
 public:
   CallSiteBase() : I(0, false) {}
-  CallSiteBase(CallTy *CI) : I(reinterpret_cast<InstrTy*>(CI), true) {}
-  CallSiteBase(InvokeTy *II) : I(reinterpret_cast<InstrTy*>(II), false) {}
+  CallSiteBase(CallTy *CI) : I(CI, true) { /*assert(CI);*/ }
+  CallSiteBase(InvokeTy *II) : I(II, false) { /*assert(II);*/ }
   CallSiteBase(ValTy *II) { *this = get(II); }
   CallSiteBase(InstrTy *II) {
     assert(II && "Null instruction given?");
     *this = get(II);
-    assert(I.getPointer());
   }
 
   /// CallSiteBase::get - This static method is sort of like a constructor.  It
@@ -66,9 +65,9 @@
   static CallSiteBase get(ValTy *V) {
     if (InstrTy *II = dyn_cast<InstrTy>(V)) {
       if (II->getOpcode() == Instruction::Call)
-        return CallSiteBase(reinterpret_cast<CallTy*>(II));
+        return CallSiteBase(static_cast<CallTy*>(II));
       else if (II->getOpcode() == Instruction::Invoke)
-        return CallSiteBase(reinterpret_cast<InvokeTy*>(II));
+        return CallSiteBase(static_cast<InvokeTy*>(II));
     }
     return CallSiteBase();
   }




From ggreif at gmail.com  Tue Jul 27 17:02:00 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 22:02:00 -0000
Subject: [llvm-commits] [llvm] r109547 -
	/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
Message-ID: <20100727220200.F10362A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 17:02:00 2010
New Revision: 109547

URL: http://llvm.org/viewvc/llvm-project?rev=109547&view=rev
Log:
recommit simplification (originally r109504, backed out in r109508) now that problem in CallSiteBase is fixed

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

Modified: llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp?rev=109547&r1=109546&r2=109547&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp (original)
+++ llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp Tue Jul 27 17:02:00 2010
@@ -120,10 +120,9 @@
       Pointer = CI->getArgOperand(0);
       // calls to free() erase the entire structure
       PointerSize = ~0ULL;
-    } else if (isa<CallInst>(Inst) || isa<InvokeInst>(Inst)) {
+    } else if (CallSite InstCS = Inst) {
       // Debug intrinsics don't cause dependences.
       if (isa<DbgInfoIntrinsic>(Inst)) continue;
-      CallSite InstCS = CallSite::get(Inst);
       // If these two calls do not interfere, look past it.
       switch (AA->getModRefInfo(CS, InstCS)) {
       case AliasAnalysis::NoModRef:
@@ -387,7 +386,7 @@
       MemSize = cast<ConstantInt>(II->getArgOperand(1))->getZExtValue();
       break;
     default:
-      CallSite QueryCS = CallSite::get(QueryInst);
+      CallSite QueryCS(QueryInst);
       bool isReadOnly = AA->onlyReadsMemory(QueryCS);
       LocalCache = getCallSiteDependencyFrom(QueryCS, isReadOnly, ScanPos,
                                              QueryParent);




From natebegeman at mac.com  Tue Jul 27 17:37:06 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Tue, 27 Jul 2010 22:37:06 -0000
Subject: [llvm-commits] [llvm] r109549 - in /llvm/trunk:
 lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86ISelLowering.h
 test/CodeGen/X86/vec_shift4.ll
Message-ID: <20100727223706.82B7B2A6C12C@llvm.org>

Author: sampo
Date: Tue Jul 27 17:37:06 2010
New Revision: 109549

URL: http://llvm.org/viewvc/llvm-project?rev=109549&view=rev
Log:
~40% faster vector shl <4 x i32> on SSE 4.1  Larger improvements for smaller types coming in future patches.

For:

define <2 x i64> @shl(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp {
entry:
  %shl = shl <4 x i32> %r, %a                     ; <<4 x i32>> [#uses=1]
  %tmp2 = bitcast <4 x i32> %shl to <2 x i64>     ; <<2 x i64>> [#uses=1]
  ret <2 x i64> %tmp2
}

We get:

_shl:                                   ## @shl
	pslld	$23, %xmm1
	paddd	LCPI0_0, %xmm1
	cvttps2dq	%xmm1, %xmm1
	pmulld	%xmm1, %xmm0
	ret

Instead of:

_shl:                                   ## @shl
	pshufd	$3, %xmm0, %xmm2
	movd	%xmm2, %eax
	pshufd	$3, %xmm1, %xmm2
	movd	%xmm2, %ecx
	shll	%cl, %eax
	movd	%eax, %xmm2
	pshufd	$1, %xmm0, %xmm3
	movd	%xmm3, %eax
	pshufd	$1, %xmm1, %xmm3
	movd	%xmm3, %ecx
	shll	%cl, %eax
	movd	%eax, %xmm3
	punpckldq	%xmm2, %xmm3
	movd	%xmm0, %eax
	movd	%xmm1, %ecx
	shll	%cl, %eax
	movd	%eax, %xmm2
	movhlps	%xmm0, %xmm0
	movd	%xmm0, %eax
	movhlps	%xmm1, %xmm1
	movd	%xmm1, %ecx
	shll	%cl, %eax
	movd	%eax, %xmm0
	punpckldq	%xmm0, %xmm2
	movdqa	%xmm2, %xmm0
	punpckldq	%xmm3, %xmm0
	ret

Added:
    llvm/trunk/test/CodeGen/X86/vec_shift4.ll
Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86ISelLowering.h

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109549&r1=109548&r2=109549&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Jul 27 17:37:06 2010
@@ -838,6 +838,9 @@
     // FIXME: Do we need to handle scalar-to-vector here?
     setOperationAction(ISD::MUL,                MVT::v4i32, Legal);
 
+    // Can turn SHL into an integer multiply.
+    setOperationAction(ISD::SHL,                MVT::v4i32, Custom);
+
     // i8 and i16 vectors are custom , because the source register and source
     // source memory operand types are not the same width.  f32 vectors are
     // custom since the immediate controlling the insert encodes additional
@@ -7498,6 +7501,35 @@
   return Res;
 }
 
+SDValue X86TargetLowering::LowerSHL(SDValue Op, SelectionDAG &DAG) const {
+  EVT VT = Op.getValueType();
+  DebugLoc dl = Op.getDebugLoc();
+  SDValue R = Op.getOperand(0);
+
+  assert(Subtarget->hasSSE41() && "Cannot lower SHL without SSE4.1 or later");
+  assert(VT == MVT::v4i32 && "Only know how to lower v4i32");
+  
+  Op = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                   DAG.getConstant(Intrinsic::x86_sse2_pslli_d, MVT::i32),
+                   Op.getOperand(1), DAG.getConstant(23, MVT::i32));
+
+  std::vector<Constant*> CV;
+  LLVMContext *Context = DAG.getContext();
+  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
+  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
+  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
+  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
+  Constant *C = ConstantVector::get(CV);
+  SDValue CPIdx = DAG.getConstantPool(C, getPointerTy(), 16);
+  SDValue Addend = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
+                               PseudoSourceValue::getConstantPool(), 0,
+                               false, false, 16);
+
+  Op = DAG.getNode(ISD::ADD, dl, VT, Op, Addend);
+  Op = DAG.getNode(ISD::BIT_CONVERT, dl, MVT::v4f32, Op);
+  Op = DAG.getNode(ISD::FP_TO_SINT, dl, VT, Op);
+  return DAG.getNode(ISD::MUL, dl, VT, Op, R);
+}
 
 SDValue X86TargetLowering::LowerXALUO(SDValue Op, SelectionDAG &DAG) const {
   // Lower the "add/sub/mul with overflow" instruction into a regular ins plus
@@ -7730,6 +7762,7 @@
   case ISD::CTLZ:               return LowerCTLZ(Op, DAG);
   case ISD::CTTZ:               return LowerCTTZ(Op, DAG);
   case ISD::MUL:                return LowerMUL_V2I64(Op, DAG);
+  case ISD::SHL:                return LowerSHL(Op, DAG);
   case ISD::SADDO:
   case ISD::UADDO:
   case ISD::SSUBO:

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.h?rev=109549&r1=109548&r2=109549&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.h (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.h Tue Jul 27 17:37:06 2010
@@ -723,6 +723,7 @@
     SDValue LowerCTLZ(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerCTTZ(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerMUL_V2I64(SDValue Op, SelectionDAG &DAG) const;
+    SDValue LowerSHL(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerXALUO(SDValue Op, SelectionDAG &DAG) const;
 
     SDValue LowerCMP_SWAP(SDValue Op, SelectionDAG &DAG) const;

Added: llvm/trunk/test/CodeGen/X86/vec_shift4.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_shift4.ll?rev=109549&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vec_shift4.ll (added)
+++ llvm/trunk/test/CodeGen/X86/vec_shift4.ll Tue Jul 27 17:37:06 2010
@@ -0,0 +1,14 @@
+; RUN: llc < %s -march=x86 -mattr=+sse41 | FileCheck %s
+
+define <2 x i64> @shl(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp {
+entry:
+; CHECK-NOT: shll
+; CHECK: pslld
+; CHECK: paddd
+; CHECK: cvttps2dq
+; CHECK: pmulld
+
+  %shl = shl <4 x i32> %r, %a                     ; <<4 x i32>> [#uses=1]
+  %tmp2 = bitcast <4 x i32> %shl to <2 x i64>     ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %tmp2
+}




From ggreif at gmail.com  Tue Jul 27 17:53:28 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Tue, 27 Jul 2010 22:53:28 -0000
Subject: [llvm-commits] [llvm] r109553 - in /llvm/trunk:
 include/llvm/Support/CallSite.h lib/Analysis/MemoryDependenceAnalysis.cpp
Message-ID: <20100727225328.D6F552A6C12C@llvm.org>

Author: ggreif
Date: Tue Jul 27 17:53:28 2010
New Revision: 109553

URL: http://llvm.org/viewvc/llvm-project?rev=109553&view=rev
Log:
reintroduce original (asserting) semantics of CallSite(Instruction *II)
add instead a CallSite(Value* V) constructor that is consistent with ImmutableCallSize
and use that one in client code

Modified:
    llvm/trunk/include/llvm/Support/CallSite.h
    llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp

Modified: llvm/trunk/include/llvm/Support/CallSite.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/CallSite.h?rev=109553&r1=109552&r2=109553&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/CallSite.h (original)
+++ llvm/trunk/include/llvm/Support/CallSite.h Tue Jul 27 17:53:28 2010
@@ -55,6 +55,7 @@
   CallSiteBase(InstrTy *II) {
     assert(II && "Null instruction given?");
     *this = get(II);
+    assert(I.getPointer() && "Not a call?");
   }
 
   /// CallSiteBase::get - This static method is sort of like a constructor.  It
@@ -293,6 +294,7 @@
 public:
   CallSite() {}
   CallSite(Base B) : Base(B) {}
+  CallSite(Value* V) : Base(V) {}
   CallSite(CallInst *CI) : Base(CI) {}
   CallSite(InvokeInst *II) : Base(II) {}
   CallSite(Instruction *II) : Base(II) {}

Modified: llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp?rev=109553&r1=109552&r2=109553&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp (original)
+++ llvm/trunk/lib/Analysis/MemoryDependenceAnalysis.cpp Tue Jul 27 17:53:28 2010
@@ -120,7 +120,7 @@
       Pointer = CI->getArgOperand(0);
       // calls to free() erase the entire structure
       PointerSize = ~0ULL;
-    } else if (CallSite InstCS = Inst) {
+    } else if (CallSite InstCS = cast<Value>(Inst)) {
       // Debug intrinsics don't cause dependences.
       if (isa<DbgInfoIntrinsic>(Inst)) continue;
       // If these two calls do not interfere, look past it.




From stuart at apple.com  Tue Jul 27 18:15:07 2010
From: stuart at apple.com (Stuart Hastings)
Date: Tue, 27 Jul 2010 23:15:07 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109556 -
	/llvm-gcc-4.2/trunk/gcc/fold-const.c
Message-ID: <20100727231507.1E4992A6C12C@llvm.org>

Author: stuart
Date: Tue Jul 27 18:15:06 2010
New Revision: 109556

URL: http://llvm.org/viewvc/llvm-project?rev=109556&view=rev
Log:
Defeat const-multiply folding across a MIN().  Radar 8198362.

Modified:
    llvm-gcc-4.2/trunk/gcc/fold-const.c

Modified: llvm-gcc-4.2/trunk/gcc/fold-const.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/fold-const.c?rev=109556&r1=109555&r2=109556&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/fold-const.c (original)
+++ llvm-gcc-4.2/trunk/gcc/fold-const.c Tue Jul 27 18:15:06 2010
@@ -5714,6 +5714,9 @@
       if (TYPE_UNSIGNED (ctype) != TYPE_UNSIGNED (type))
 	break;
 
+      /* LLVM LOCAL 8198362 */
+      if (TYPE_UNSIGNED (ctype)) break;
+
       /* MIN (a, b) / 5 -> MIN (a / 5, b / 5)  */
       sub_strict_overflow_p = false;
       if ((t1 = extract_muldiv (op0, c, code, wide_type,




From stuart at apple.com  Tue Jul 27 18:15:26 2010
From: stuart at apple.com (Stuart Hastings)
Date: Tue, 27 Jul 2010 23:15:26 -0000
Subject: [llvm-commits] [llvm] r109557 -
	/llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c
Message-ID: <20100727231526.115CC2A6C12C@llvm.org>

Author: stuart
Date: Tue Jul 27 18:15:25 2010
New Revision: 109557

URL: http://llvm.org/viewvc/llvm-project?rev=109557&view=rev
Log:
Testcase for r109556.  Radar 8198362.

Added:
    llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c

Added: llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c?rev=109557&view=auto
==============================================================================
--- llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c (added)
+++ llvm/trunk/test/FrontendC/2010-07-27-MinNoFoldConst.c Tue Jul 27 18:15:25 2010
@@ -0,0 +1,18 @@
+// RUN: %llvmgcc -S %s -o - | FileCheck %s
+extern int printf(const char *, ...);
+static void bad(unsigned int v1, unsigned int v2) {
+  printf("%u\n", 1631381461u * (((v2 - 1273463329u <= v1 - 1273463329u) ? v2 : v1) - 1273463329u) + 121322179u);
+}
+// Radar 8198362
+// GCC FE wants to convert the above to
+//   1631381461u * MIN(v2 - 1273463329u, v1 - 1273463329u)
+// and then to
+//   MIN(1631381461u * v2 - 4047041419, 1631381461u * v1 - 4047041419)
+//
+// 1631381461u * 1273463329u = 2077504466193943669, but 32-bit overflow clips
+// this to 4047041419. This breaks the comparision implicit in the MIN().
+// Two multiply operations suggests the bad optimization is happening;
+// one multiplication, after the MIN(), is correct.
+// CHECK: mul
+// CHECK-NOT: mul
+// CHECK: ret




From resistor at mac.com  Tue Jul 27 18:55:47 2010
From: resistor at mac.com (Owen Anderson)
Date: Tue, 27 Jul 2010 23:55:47 -0000
Subject: [llvm-commits] [llvm] r109562 -
	/llvm/trunk/include/llvm/ADT/DenseSet.h
Message-ID: <20100727235547.5BF772A6C12C@llvm.org>

Author: resistor
Date: Tue Jul 27 18:55:47 2010
New Revision: 109562

URL: http://llvm.org/viewvc/llvm-project?rev=109562&view=rev
Log:
Fill out the interface of DenseSet a bit.

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

Modified: llvm/trunk/include/llvm/ADT/DenseSet.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/DenseSet.h?rev=109562&r1=109561&r2=109562&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/DenseSet.h (original)
+++ llvm/trunk/include/llvm/ADT/DenseSet.h Tue Jul 27 18:55:47 2010
@@ -58,6 +58,7 @@
 
   class Iterator {
     typename MapTy::iterator I;
+    friend class DenseSet;
   public:
     typedef typename MapTy::iterator::difference_type difference_type;
     typedef ValueT value_type;
@@ -77,6 +78,7 @@
 
   class ConstIterator {
     typename MapTy::const_iterator I;
+    friend class DenseSet;
   public:
     typedef typename MapTy::const_iterator::difference_type difference_type;
     typedef ValueT value_type;
@@ -103,6 +105,10 @@
   const_iterator begin() const { return ConstIterator(TheMap.begin()); }
   const_iterator end() const { return ConstIterator(TheMap.end()); }
 
+  iterator find(const ValueT &V) { return Iterator(TheMap.find(V)); }
+  bool erase(Iterator I) { return TheMap.erase(I.I); }
+  bool erase(ConstIterator CI) { return TheMap.erase(CI.I); }
+
   std::pair<iterator, bool> insert(const ValueT &V) {
     return TheMap.insert(std::make_pair(V, 0));
   }




From resistor at mac.com  Tue Jul 27 18:58:12 2010
From: resistor at mac.com (Owen Anderson)
Date: Tue, 27 Jul 2010 23:58:12 -0000
Subject: [llvm-commits] [llvm] r109563 -
	/llvm/trunk/lib/Analysis/LazyValueInfo.cpp
Message-ID: <20100727235812.0BCE52A6C12C@llvm.org>

Author: resistor
Date: Tue Jul 27 18:58:11 2010
New Revision: 109563

URL: http://llvm.org/viewvc/llvm-project?rev=109563&view=rev
Log:
Rearrange several datastructures in LazyValueInfo to improve compile time.
This is still not perfect, but better than it was before.

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

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109563&r1=109562&r2=109563&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Tue Jul 27 18:58:11 2010
@@ -22,6 +22,7 @@
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/PointerIntPair.h"
 #include "llvm/ADT/STLExtras.h"
 using namespace llvm;
@@ -211,7 +212,7 @@
     /// ValueCacheEntryTy - This is all of the cached block information for
     /// exactly one Value*.  The entries are sorted by the BasicBlock* of the
     /// entries, allowing us to do a lookup with a binary search.
-    typedef std::vector<BlockCacheEntryTy> ValueCacheEntryTy;
+    typedef DenseMap<BasicBlock*, LVILatticeVal> ValueCacheEntryTy;
 
   private:
     /// ValueCache - This is all of the cached information for all values,
@@ -221,7 +222,7 @@
     /// OverDefinedCache - This tracks, on a per-block basis, the set of 
     /// values that are over-defined at the end of that block.  This is required
     /// for cache updating.
-    DenseMap<BasicBlock*, std::set<Value*> > OverDefinedCache;
+    DenseSet<std::pair<BasicBlock*, Value*> > OverDefinedCache;
   public:
     
     /// getValueInBlock - This is the query interface to determine the lattice
@@ -281,15 +282,15 @@
     ValueCacheEntryTy &Cache;
     
     /// This tracks, for each block, what values are overdefined.
-    DenseMap<BasicBlock*, std::set<Value*> > &OverDefinedCache;
+    DenseSet<std::pair<BasicBlock*, Value*> > &OverDefinedCache;
     
     ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
     /// added to cache but that are not in sorted order.
-    DenseMap<BasicBlock*, LVILatticeVal> NewBlockInfo;
+    DenseSet<BasicBlock*> NewBlockInfo;
   public:
     
     LVIQuery(Value *V, ValueCacheEntryTy &VC,
-             DenseMap<BasicBlock*, std::set<Value*> > &ODC)
+             DenseSet<std::pair<BasicBlock*, Value*> > &ODC)
       : Val(V), Cache(VC), OverDefinedCache(ODC) {
     }
 
@@ -297,34 +298,11 @@
       // When the query is done, insert the newly discovered facts into the
       // cache in sorted order.
       if (NewBlockInfo.empty()) return;
-
-      // Grow the cache to exactly fit the new data.
-      Cache.reserve(Cache.size() + NewBlockInfo.size());
-      
-      // If we only have one new entry, insert it instead of doing a full-on
-      // sort.
-      if (NewBlockInfo.size() == 1) {
-        BlockCacheEntryTy Entry = *NewBlockInfo.begin();
-        ValueCacheEntryTy::iterator I =
-          std::lower_bound(Cache.begin(), Cache.end(), Entry,
-                           BlockCacheEntryComparator());
-        assert((I == Cache.end() || I->first != Entry.first) &&
-               "Entry already in map!");
-        
-        Cache.insert(I, Entry);
-        return;
-      }
-      
-      // TODO: If we only have two new elements, INSERT them both.
-      
-      Cache.insert(Cache.end(), NewBlockInfo.begin(), NewBlockInfo.end());
-      array_pod_sort(Cache.begin(), Cache.end(),
-                     BlockCacheEntryComparator::Compare);
       
-      for (DenseMap<BasicBlock*, LVILatticeVal>::iterator
-           I = NewBlockInfo.begin(), E = NewBlockInfo.end(); I != E; ++I) {
-        if (I->second.isOverdefined())
-          OverDefinedCache[I->first].insert(Val);
+      for (DenseSet<BasicBlock*>::iterator I = NewBlockInfo.begin(),
+           E = NewBlockInfo.end(); I != E; ++I) {
+        if (Cache[*I].isOverdefined())
+          OverDefinedCache.insert(std::make_pair(*I, Val));
       }
     }
 
@@ -337,23 +315,10 @@
 } // end anonymous namespace
 
 /// getCachedEntryForBlock - See if we already have a value for this block.  If
-/// so, return it, otherwise create a new entry in the NewBlockInfo map to use.
+/// so, return it, otherwise create a new entry in the Cache map to use.
 LVILatticeVal &LVIQuery::getCachedEntryForBlock(BasicBlock *BB) {
-  
-  // Do a binary search to see if we already have an entry for this block in
-  // the cache set.  If so, find it.
-  if (!Cache.empty()) {
-    ValueCacheEntryTy::iterator Entry =
-      std::lower_bound(Cache.begin(), Cache.end(),
-                       BlockCacheEntryTy(BB, LVILatticeVal()),
-                       BlockCacheEntryComparator());
-    if (Entry != Cache.end() && Entry->first == BB)
-      return Entry->second;
-  }
-  
-  // Otherwise, check to see if it's in NewBlockInfo or create a new entry if
-  // not.
-  return NewBlockInfo[BB];
+  NewBlockInfo.insert(BB);
+  return Cache[BB];
 }
 
 LVILatticeVal LVIQuery::getBlockValue(BasicBlock *BB) {
@@ -534,9 +499,12 @@
   std::vector<BasicBlock*> worklist;
   worklist.push_back(OldSucc);
   
-  std::set<Value*> ClearSet = OverDefinedCache[OldSucc];
-  LVILatticeVal OverDef;
-  OverDef.markOverdefined();
+  DenseSet<Value*> ClearSet;
+  for (DenseSet<std::pair<BasicBlock*, Value*> >::iterator
+       I = OverDefinedCache.begin(), E = OverDefinedCache.end(); I != E; ++I) {
+    if (I->first == OldSucc)
+      ClearSet.insert(I->second);
+  }
   
   // Use a worklist to perform a depth-first search of OldSucc's successors.
   // NOTE: We do not need a visited list since any blocks we have already
@@ -550,26 +518,24 @@
     if (ToUpdate == NewSucc) continue;
     
     bool changed = false;
-    std::set<Value*> &CurrentSet = OverDefinedCache[ToUpdate];
-    for (std::set<Value*>::iterator I = ClearSet.begin(),E = ClearSet.end();
+    for (DenseSet<Value*>::iterator I = ClearSet.begin(),E = ClearSet.end();
          I != E; ++I) {
       // If a value was marked overdefined in OldSucc, and is here too...
-      if (CurrentSet.count(*I)) {
-        // Remove it from the caches.
-        ValueCacheEntryTy &Entry = ValueCache[*I];
-        ValueCacheEntryTy::iterator CI =
-          std::lower_bound(Entry.begin(), Entry.end(),
-                           std::make_pair(ToUpdate, OverDef),
-                           BlockCacheEntryComparator());
-        assert(CI != Entry.end() && "Couldn't find entry to update?");
-        Entry.erase(CI);
-
-        CurrentSet.erase(*I);
-
-        // If we removed anything, then we potentially need to update 
-        // blocks successors too.
-        changed = true;
-      }
+      DenseSet<std::pair<BasicBlock*, Value*> >::iterator OI =
+        OverDefinedCache.find(std::make_pair(ToUpdate, *I));
+      if (OI == OverDefinedCache.end()) continue;
+
+      // Remove it from the caches.
+      ValueCacheEntryTy &Entry = ValueCache[*I];
+      ValueCacheEntryTy::iterator CI = Entry.find(ToUpdate);
+        
+      assert(CI != Entry.end() && "Couldn't find entry to update?");
+      Entry.erase(CI);
+      OverDefinedCache.erase(OI);
+
+      // If we removed anything, then we potentially need to update 
+      // blocks successors too.
+      changed = true;
     }
         
     if (!changed) continue;




From stoklund at 2pi.dk  Tue Jul 27 19:11:49 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Wed, 28 Jul 2010 00:11:49 -0000
Subject: [llvm-commits] [test-suite] r109564 - in
 /test-suite/trunk/MultiSource/Applications/lua: Makefile
 lua.reference_output
Message-ID: <20100728001149.ADB372A6C12C@llvm.org>

Author: stoklund
Date: Tue Jul 27 19:11:49 2010
New Revision: 109564

URL: http://llvm.org/viewvc/llvm-project?rev=109564&view=rev
Log:
Provide the full lua reference output instead of one that terminates early
because of a missing input.

The full reference output is 10MB, so hash it.

Modified:
    test-suite/trunk/MultiSource/Applications/lua/Makefile
    test-suite/trunk/MultiSource/Applications/lua/lua.reference_output

Modified: test-suite/trunk/MultiSource/Applications/lua/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lua/Makefile?rev=109564&r1=109563&r2=109564&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lua/Makefile (original)
+++ test-suite/trunk/MultiSource/Applications/lua/Makefile Tue Jul 27 19:11:49 2010
@@ -10,6 +10,7 @@
 
 PROG = lua
 RUN_OPTIONS = alltests.lua "\<in"
+HASH_PROGRAM_OUTPUT = 1
 
 include $(LEVEL)/Makefile.config
 

Modified: test-suite/trunk/MultiSource/Applications/lua/lua.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lua/lua.reference_output?rev=109564&r1=109563&r2=109564&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lua/lua.reference_output (original)
+++ test-suite/trunk/MultiSource/Applications/lua/lua.reference_output Tue Jul 27 19:11:49 2010
@@ -1,181 +1 @@
-run script: 'bench/ackermann.lua', with arg='9'
-Ack(3,9): 4093
-bench/ackermann.lua
-results = 	true
-ackermann:0.00
-run script: 'bench/ary.lua', with arg='7000'
-1000 7000000
-bench/ary.lua
-results = 	true
-ary:0.00
-run script: 'bench/binarytrees.lua', with arg='12'
-stretch tree of depth 13	 check: -1
-8192	 trees of depth 4	 check: -8192
-2048	 trees of depth 6	 check: -2048
-512	 trees of depth 8	 check: -512
-128	 trees of depth 10	 check: -128
-32	 trees of depth 12	 check: -32
-long lived tree of depth 12	 check: -1
-bench/binarytrees.lua
-results = 	true
-binarytrees:0.00
-run script: 'bench/chameneos.lua', with arg='350000'
-700000
-bench/chameneos.lua
-results = 	true
-chameneos:0.00
-run script: 'bench/except.lua', with arg='500000'
-Exceptions: HI=250000 / LO=250000
-bench/except.lua
-results = 	true
-except:0.00
-run script: 'bench/fannkuch.lua', with arg='9'
-123456789
-213456789
-231456789
-321456789
-312456789
-132456789
-234156789
-324156789
-342156789
-432156789
-423156789
-243156789
-341256789
-431256789
-413256789
-143256789
-134256789
-314256789
-412356789
-142356789
-124356789
-214356789
-241356789
-421356789
-234516789
-324516789
-342516789
-432516789
-423516789
-243516789
-Pfannkuchen(9) = 30
-bench/fannkuch.lua
-results = 	true
-fannkuch:0.00
-run script: 'bench/fibo.lua', with arg='31'
-2178309
-bench/fibo.lua
-results = 	true
-fibo:0.00
-run script: 'bench/harmonic.lua', with arg='20000000'
-17.388458521
-bench/harmonic.lua
-results = 	true
-harmonic:0.00
-run script: 'bench/hash.lua', with arg='200000'
-30999
-bench/hash.lua
-results = 	true
-hash:0.00
-run script: 'bench/hash2.lua', with arg='250'
-1 9999 250 2499750
-bench/hash2.lua
-results = 	true
-hash2:0.00
-run script: 'bench/heapsort.lua', with arg='150000'
-0.9999928555
-bench/heapsort.lua
-results = 	true
-heapsort:0.00
-run script: 'bench/hello.lua', with arg='1'
-hello world
-bench/hello.lua
-results = 	true
-hello:0.00
-run script: 'bench/knucleotide.lua', with arg='1'
-T 30.408
-A 30.305
-C 19.652
-G 19.635
-
-TT 9.247
-AT 9.244
-TA 9.230
-AA 9.152
-TC 6.013
-GA 5.975
-GT 5.972
-AG 5.963
-CA 5.948
-AC 5.946
-CT 5.945
-TG 5.917
-CG 3.880
-CC 3.879
-GG 3.875
-GC 3.813
-
-1190	GGT
-358	GGTA
-36	GGTATT
-0	GGTATTTTAATT
-0	GGTATTTTAATTTATAGT
-bench/knucleotide.lua
-results = 	true
-knucleotide:0.00
-run script: 'bench/lists.lua', with arg='35'
-10000
-bench/lists.lua
-results = 	true
-lists:0.00
-run script: 'bench/matrix.lua', with arg='250'
-270165 1061760 1453695 1856025
-bench/matrix.lua
-results = 	true
-matrix:0.00
-run script: 'bench/meteor.lua', with arg='0'
-2098 solutions found
-
-0 0 0 0 1 
- 2 2 2 0 1 
-2 6 6 1 1 
- 2 6 1 5 5 
-8 6 5 5 5 
- 8 6 3 3 3 
-4 8 8 9 3 
- 4 4 8 9 3 
-4 7 4 7 9 
- 7 7 7 9 9 
-
-9 9 9 9 8 
- 9 6 6 8 5 
-6 6 8 8 5 
- 6 8 2 5 5 
-7 7 7 2 5 
- 7 4 7 2 0 
-1 4 2 2 0 
- 1 4 4 0 3 
-1 4 0 0 3 
- 1 1 3 3 3 
-
-bench/meteor.lua
-results = 	true
-meteor:0.00
-run script: 'bench/methcall.lua', with arg='1000000'
-true
-false
-bench/methcall.lua
-results = 	true
-methcall:0.00
-run script: 'bench/moments.lua', with arg='1'
-lua: run_bench_tests.lua:55: bad argument #1 to 'input' (input/moments-input400000.txt: No such file or directory)
-stack traceback:
-	[C]: in function 'input'
-	run_bench_tests.lua:55: in function 'run_bench'
-	run_bench_tests.lua:262: in main chunk
-	[C]: in function 'dofile'
-	alltests.lua:1: in main chunk
-	[C]: ?
-exit 1
+80650b40ee6b6a44115e337b5089e026




From wendling at apple.com  Tue Jul 27 19:20:44 2010
From: wendling at apple.com (Bill Wendling)
Date: Tue, 27 Jul 2010 17:20:44 -0700
Subject: [llvm-commits] [llvm] r109562
	-	/llvm/trunk/include/llvm/ADT/DenseSet.h
In-Reply-To: <20100727235547.5BF772A6C12C@llvm.org>
References: <20100727235547.5BF772A6C12C@llvm.org>
Message-ID: <1385FBDB-84B2-40BE-B210-A8BCB438307B@apple.com>


On Jul 27, 2010, at 4:55 PM, Owen Anderson wrote:

> Author: resistor
> Date: Tue Jul 27 18:55:47 2010
> New Revision: 109562
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109562&view=rev
> Log:
> Fill out the interface of DenseSet a bit.
> 
> Modified:
>    llvm/trunk/include/llvm/ADT/DenseSet.h
> 
> Modified: llvm/trunk/include/llvm/ADT/DenseSet.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/DenseSet.h?rev=109562&r1=109561&r2=109562&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/ADT/DenseSet.h (original)
> +++ llvm/trunk/include/llvm/ADT/DenseSet.h Tue Jul 27 18:55:47 2010
> @@ -58,6 +58,7 @@
> 
>   class Iterator {
>     typename MapTy::iterator I;
> +    friend class DenseSet;

Why do you use "friend" here and in ConstIterator?

-bw

>   public:
>     typedef typename MapTy::iterator::difference_type difference_type;
>     typedef ValueT value_type;
> @@ -77,6 +78,7 @@
> 
>   class ConstIterator {
>     typename MapTy::const_iterator I;
> +    friend class DenseSet;
>   public:
>     typedef typename MapTy::const_iterator::difference_type difference_type;
>     typedef ValueT value_type;
> @@ -103,6 +105,10 @@
>   const_iterator begin() const { return ConstIterator(TheMap.begin()); }
>   const_iterator end() const { return ConstIterator(TheMap.end()); }
> 
> +  iterator find(const ValueT &V) { return Iterator(TheMap.find(V)); }
> +  bool erase(Iterator I) { return TheMap.erase(I.I); }
> +  bool erase(ConstIterator CI) { return TheMap.erase(CI.I); }
> +
>   std::pair<iterator, bool> insert(const ValueT &V) {
>     return TheMap.insert(std::make_pair(V, 0));
>   }
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From gohman at apple.com  Tue Jul 27 19:21:18 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 00:21:18 -0000
Subject: [llvm-commits] [llvm] r109565 -
 /llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
Message-ID: <20100728002118.E26452A6C12C@llvm.org>

Author: djg
Date: Tue Jul 27 19:21:18 2010
New Revision: 109565

URL: http://llvm.org/viewvc/llvm-project?rev=109565&view=rev
Log:
Add a comment.

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

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h?rev=109565&r1=109564&r2=109565&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h Tue Jul 27 19:21:18 2010
@@ -523,6 +523,8 @@
   class SCEVUnknown : public SCEV {
     friend class ScalarEvolution;
 
+    // This should be an AssertingVH, however SCEVUnknowns are allocated in a
+    // BumpPtrAllocator so their destructors are never called.
     Value *V;
     SCEVUnknown(const FoldingSetNodeIDRef ID, Value *v) :
       SCEV(ID, scUnknown), V(v) {}




From natebegeman at mac.com  Tue Jul 27 19:21:49 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Wed, 28 Jul 2010 00:21:49 -0000
Subject: [llvm-commits] [llvm] r109566 - in /llvm/trunk:
 lib/Target/X86/X86ISelLowering.cpp test/CodeGen/X86/vec_shift4.ll
Message-ID: <20100728002149.360E72A6C12C@llvm.org>

Author: sampo
Date: Tue Jul 27 19:21:48 2010
New Revision: 109566

URL: http://llvm.org/viewvc/llvm-project?rev=109566&view=rev
Log:
Implement a vectorized algorithm for <16 x i8> << <16 x i8>
This is about 4x faster and smaller than the existing scalarization.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/vec_shift4.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109566&r1=109565&r2=109566&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Jul 27 19:21:48 2010
@@ -840,6 +840,7 @@
 
     // Can turn SHL into an integer multiply.
     setOperationAction(ISD::SHL,                MVT::v4i32, Custom);
+    setOperationAction(ISD::SHL,                MVT::v16i8, Custom);
 
     // i8 and i16 vectors are custom , because the source register and source
     // source memory operand types are not the same width.  f32 vectors are
@@ -7506,29 +7507,80 @@
   DebugLoc dl = Op.getDebugLoc();
   SDValue R = Op.getOperand(0);
 
+  LLVMContext *Context = DAG.getContext();
+
   assert(Subtarget->hasSSE41() && "Cannot lower SHL without SSE4.1 or later");
-  assert(VT == MVT::v4i32 && "Only know how to lower v4i32");
-  
-  Op = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
-                   DAG.getConstant(Intrinsic::x86_sse2_pslli_d, MVT::i32),
-                   Op.getOperand(1), DAG.getConstant(23, MVT::i32));
 
-  std::vector<Constant*> CV;
-  LLVMContext *Context = DAG.getContext();
-  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
-  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
-  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
-  CV.push_back(ConstantInt::get(*Context, APInt(32, 0x3f800000U)));
-  Constant *C = ConstantVector::get(CV);
-  SDValue CPIdx = DAG.getConstantPool(C, getPointerTy(), 16);
-  SDValue Addend = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
-                               PseudoSourceValue::getConstantPool(), 0,
-                               false, false, 16);
-
-  Op = DAG.getNode(ISD::ADD, dl, VT, Op, Addend);
-  Op = DAG.getNode(ISD::BIT_CONVERT, dl, MVT::v4f32, Op);
-  Op = DAG.getNode(ISD::FP_TO_SINT, dl, VT, Op);
-  return DAG.getNode(ISD::MUL, dl, VT, Op, R);
+  if (VT == MVT::v4i32) {
+    Op = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                     DAG.getConstant(Intrinsic::x86_sse2_pslli_d, MVT::i32),
+                     Op.getOperand(1), DAG.getConstant(23, MVT::i32));
+
+    ConstantInt *CI = ConstantInt::get(*Context, APInt(32, 0x3f800000U));
+    
+    std::vector<Constant*> CV(4, CI);
+    Constant *C = ConstantVector::get(CV);
+    SDValue CPIdx = DAG.getConstantPool(C, getPointerTy(), 16);
+    SDValue Addend = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
+                                 PseudoSourceValue::getConstantPool(), 0,
+                                 false, false, 16);
+
+    Op = DAG.getNode(ISD::ADD, dl, VT, Op, Addend);
+    Op = DAG.getNode(ISD::BIT_CONVERT, dl, MVT::v4f32, Op);
+    Op = DAG.getNode(ISD::FP_TO_SINT, dl, VT, Op);
+    return DAG.getNode(ISD::MUL, dl, VT, Op, R);
+  }
+  if (VT == MVT::v16i8) {
+    // a = a << 5;
+    Op = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                     DAG.getConstant(Intrinsic::x86_sse2_pslli_w, MVT::i32),
+                     Op.getOperand(1), DAG.getConstant(5, MVT::i32));
+
+    ConstantInt *CM1 = ConstantInt::get(*Context, APInt(8, 15));
+    ConstantInt *CM2 = ConstantInt::get(*Context, APInt(8, 63));
+
+    std::vector<Constant*> CVM1(16, CM1);
+    std::vector<Constant*> CVM2(16, CM2);
+    Constant *C = ConstantVector::get(CVM1);
+    SDValue CPIdx = DAG.getConstantPool(C, getPointerTy(), 16);
+    SDValue M = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
+                            PseudoSourceValue::getConstantPool(), 0,
+                            false, false, 16);
+
+    // r = pblendv(r, psllw(r & (char16)15, 4), a);
+    M = DAG.getNode(ISD::AND, dl, VT, R, M);
+    M = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                    DAG.getConstant(Intrinsic::x86_sse2_pslli_w, MVT::i32), M,
+                    DAG.getConstant(4, MVT::i32));
+    R = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                    DAG.getConstant(Intrinsic::x86_sse41_pblendvb, MVT::i32),
+                    R, M, Op);
+    // a += a
+    Op = DAG.getNode(ISD::ADD, dl, VT, Op, Op);
+    
+    C = ConstantVector::get(CVM2);
+    CPIdx = DAG.getConstantPool(C, getPointerTy(), 16);
+    M = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
+                    PseudoSourceValue::getConstantPool(), 0, false, false, 16);
+    
+    // r = pblendv(r, psllw(r & (char16)63, 2), a);
+    M = DAG.getNode(ISD::AND, dl, VT, R, M);
+    M = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                    DAG.getConstant(Intrinsic::x86_sse2_pslli_w, MVT::i32), M,
+                    DAG.getConstant(2, MVT::i32));
+    R = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                    DAG.getConstant(Intrinsic::x86_sse41_pblendvb, MVT::i32),
+                    R, M, Op);
+    // a += a
+    Op = DAG.getNode(ISD::ADD, dl, VT, Op, Op);
+    
+    // return pblendv(r, r+r, a);
+    R = DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, VT,
+                    DAG.getConstant(Intrinsic::x86_sse41_pblendvb, MVT::i32),
+                    R, DAG.getNode(ISD::ADD, dl, VT, R, R), Op);
+    return R;
+  }
+  return SDValue();
 }
 
 SDValue X86TargetLowering::LowerXALUO(SDValue Op, SelectionDAG &DAG) const {

Modified: llvm/trunk/test/CodeGen/X86/vec_shift4.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/vec_shift4.ll?rev=109566&r1=109565&r2=109566&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/vec_shift4.ll (original)
+++ llvm/trunk/test/CodeGen/X86/vec_shift4.ll Tue Jul 27 19:21:48 2010
@@ -1,6 +1,6 @@
 ; RUN: llc < %s -march=x86 -mattr=+sse41 | FileCheck %s
 
-define <2 x i64> @shl(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp {
+define <2 x i64> @shl1(<4 x i32> %r, <4 x i32> %a) nounwind readnone ssp {
 entry:
 ; CHECK-NOT: shll
 ; CHECK: pslld
@@ -12,3 +12,14 @@
   %tmp2 = bitcast <4 x i32> %shl to <2 x i64>     ; <<2 x i64>> [#uses=1]
   ret <2 x i64> %tmp2
 }
+
+define <2 x i64> @shl2(<16 x i8> %r, <16 x i8> %a) nounwind readnone ssp {
+entry:
+; CHECK-NOT: shlb
+; CHECK: pblendvb
+; CHECK: pblendvb
+; CHECK: pblendvb
+  %shl = shl <16 x i8> %r, %a                     ; <<16 x i8>> [#uses=1]
+  %tmp2 = bitcast <16 x i8> %shl to <2 x i64>     ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %tmp2
+}




From wendling at apple.com  Tue Jul 27 19:23:34 2010
From: wendling at apple.com (Bill Wendling)
Date: Tue, 27 Jul 2010 17:23:34 -0700
Subject: [llvm-commits] [llvm] r109562	-
	/llvm/trunk/include/llvm/ADT/DenseSet.h
In-Reply-To: <1385FBDB-84B2-40BE-B210-A8BCB438307B@apple.com>
References: <20100727235547.5BF772A6C12C@llvm.org>
	<1385FBDB-84B2-40BE-B210-A8BCB438307B@apple.com>
Message-ID: <A7DDB4E9-914C-4055-9281-0E1257D61CD3@apple.com>

On Jul 27, 2010, at 5:20 PM, Bill Wendling wrote:

> On Jul 27, 2010, at 4:55 PM, Owen Anderson wrote:
> 
>> Author: resistor
>> Date: Tue Jul 27 18:55:47 2010
>> New Revision: 109562
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=109562&view=rev
>> Log:
>> Fill out the interface of DenseSet a bit.
>> 
>> Modified:
>>   llvm/trunk/include/llvm/ADT/DenseSet.h
>> 
>> Modified: llvm/trunk/include/llvm/ADT/DenseSet.h
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/DenseSet.h?rev=109562&r1=109561&r2=109562&view=diff
>> ==============================================================================
>> --- llvm/trunk/include/llvm/ADT/DenseSet.h (original)
>> +++ llvm/trunk/include/llvm/ADT/DenseSet.h Tue Jul 27 18:55:47 2010
>> @@ -58,6 +58,7 @@
>> 
>>  class Iterator {
>>    typename MapTy::iterator I;
>> +    friend class DenseSet;
> 
> Why do you use "friend" here and in ConstIterator?
> 
> -bw
> 
>>  public:
>>    typedef typename MapTy::iterator::difference_type difference_type;
>>    typedef ValueT value_type;
>> @@ -77,6 +78,7 @@
>> 
>>  class ConstIterator {
>>    typename MapTy::const_iterator I;
>> +    friend class DenseSet;
>>  public:
>>    typedef typename MapTy::const_iterator::difference_type difference_type;
>>    typedef ValueT value_type;
>> @@ -103,6 +105,10 @@
>>  const_iterator begin() const { return ConstIterator(TheMap.begin()); }
>>  const_iterator end() const { return ConstIterator(TheMap.end()); }
>> 
>> +  iterator find(const ValueT &V) { return Iterator(TheMap.find(V)); }
>> +  bool erase(Iterator I) { return TheMap.erase(I.I); }
>> +  bool erase(ConstIterator CI) { return TheMap.erase(CI.I); }
>> +

I see. It's because of these. Never mind

-bw

>>  std::pair<iterator, bool> insert(const ValueT &V) {
>>    return TheMap.insert(std::make_pair(V, 0));
>>  }
>> 
>> 
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From gohman at apple.com  Tue Jul 27 19:28:26 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 00:28:26 -0000
Subject: [llvm-commits] [llvm] r109567 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <20100728002826.2D5E42A6C12C@llvm.org>

Author: djg
Date: Tue Jul 27 19:28:25 2010
New Revision: 109567

URL: http://llvm.org/viewvc/llvm-project?rev=109567&view=rev
Log:
Make SCEVCallbackVH::allUsesReplacedWith unconditionally delete
the old value.

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

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=109567&r1=109566&r2=109567&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue Jul 27 19:28:25 2010
@@ -5699,7 +5699,6 @@
   SmallVector<User *, 16> Worklist;
   SmallPtrSet<User *, 8> Visited;
   Value *Old = getValPtr();
-  bool DeleteOld = false;
   for (Value::use_iterator UI = Old->use_begin(), UE = Old->use_end();
        UI != UE; ++UI)
     Worklist.push_back(*UI);
@@ -5707,10 +5706,8 @@
     User *U = Worklist.pop_back_val();
     // Deleting the Old value will cause this to dangle. Postpone
     // that until everything else is done.
-    if (U == Old) {
-      DeleteOld = true;
+    if (U == Old)
       continue;
-    }
     if (!Visited.insert(U))
       continue;
     if (PHINode *PN = dyn_cast<PHINode>(U))
@@ -5720,14 +5717,11 @@
          UI != UE; ++UI)
       Worklist.push_back(*UI);
   }
-  // Delete the Old value if it (indirectly) references itself.
-  if (DeleteOld) {
-    if (PHINode *PN = dyn_cast<PHINode>(Old))
-      SE->ConstantEvolutionLoopExitValue.erase(PN);
-    SE->Scalars.erase(Old);
-    // this now dangles!
-  }
-  // this may dangle!
+  // Delete the Old value.
+  if (PHINode *PN = dyn_cast<PHINode>(Old))
+    SE->ConstantEvolutionLoopExitValue.erase(PN);
+  SE->Scalars.erase(Old);
+  // this now dangles!
 }
 
 ScalarEvolution::SCEVCallbackVH::SCEVCallbackVH(Value *V, ScalarEvolution *se)




From gohman at apple.com  Tue Jul 27 20:09:07 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 01:09:07 -0000
Subject: [llvm-commits] [llvm] r109570 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <20100728010907.D9BE32A6C12C@llvm.org>

Author: djg
Date: Tue Jul 27 20:09:07 2010
New Revision: 109570

URL: http://llvm.org/viewvc/llvm-project?rev=109570&view=rev
Log:
Make SCEVCallbackVH::allUsesReplacedWith update the old SCEVUnknown
object, as it may still be referenced by SCEVs not cleaned up by the
use list traversal.

Also, in ScalarEvolution::forgetValue, only check for a SCEVUnknown
object for the original value, not for any value in the use list,
because other SCEVUnknown values aren't necessary obsolete at that
point.

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

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=109570&r1=109569&r2=109570&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue Jul 27 20:09:07 2010
@@ -3664,6 +3664,26 @@
 /// changed a value in a way that may effect its value, or which may
 /// disconnect it from a def-use chain linking it to a loop.
 void ScalarEvolution::forgetValue(Value *V) {
+  // If there's a SCEVUnknown tying this value into the SCEV
+  // space, remove it from the folding set map. The SCEVUnknown
+  // object and any other SCEV objects which reference it
+  // (transitively) remain allocated, effectively leaked until
+  // the underlying BumpPtrAllocator is freed.
+  //
+  // This permits SCEV pointers to be used as keys in maps
+  // such as the ValuesAtScopes map.
+  FoldingSetNodeID ID;
+  ID.AddInteger(scUnknown);
+  ID.AddPointer(V);
+  void *IP;
+  if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
+    UniqueSCEVs.RemoveNode(S);
+
+    // This isn't necessary, but we might as well remove the
+    // value from the ValuesAtScopes map too.
+    ValuesAtScopes.erase(S);
+  }
+
   Instruction *I = dyn_cast<Instruction>(V);
   if (!I) return;
 
@@ -3685,26 +3705,6 @@
         ConstantEvolutionLoopExitValue.erase(PN);
     }
 
-    // If there's a SCEVUnknown tying this value into the SCEV
-    // space, remove it from the folding set map. The SCEVUnknown
-    // object and any other SCEV objects which reference it
-    // (transitively) remain allocated, effectively leaked until
-    // the underlying BumpPtrAllocator is freed.
-    //
-    // This permits SCEV pointers to be used as keys in maps
-    // such as the ValuesAtScopes map.
-    FoldingSetNodeID ID;
-    ID.AddInteger(scUnknown);
-    ID.AddPointer(I);
-    void *IP;
-    if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
-      UniqueSCEVs.RemoveNode(S);
-
-      // This isn't necessary, but we might as well remove the
-      // value from the ValuesAtScopes map too.
-      ValuesAtScopes.erase(S);
-    }
-
     PushDefUseChildren(I, Worklist);
   }
 }
@@ -5690,15 +5690,32 @@
   // this now dangles!
 }
 
-void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *) {
+void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *V) {
   assert(SE && "SCEVCallbackVH called with a null ScalarEvolution!");
 
+  Value *Old = getValPtr();
+
+  // If there's a SCEVUnknown tying this value into the SCEV
+  // space, replace the SCEVUnknown's value with the new value
+  // for the benefit of any SCEVs still referencing it, and
+  // and remove it from the folding set map so that new scevs
+  // don't reference it.
+  FoldingSetNodeID ID;
+  ID.AddInteger(scUnknown);
+  ID.AddPointer(Old);
+  void *IP;
+  if (SCEVUnknown *S = cast_or_null<SCEVUnknown>(
+        SE->UniqueSCEVs.FindNodeOrInsertPos(ID, IP))) {
+    S->V = V;
+    SE->UniqueSCEVs.RemoveNode(S);
+    SE->ValuesAtScopes.erase(S);
+  }
+
   // Forget all the expressions associated with users of the old value,
   // so that future queries will recompute the expressions using the new
   // value.
   SmallVector<User *, 16> Worklist;
   SmallPtrSet<User *, 8> Visited;
-  Value *Old = getValPtr();
   for (Value::use_iterator UI = Old->use_begin(), UE = Old->use_end();
        UI != UE; ++UI)
     Worklist.push_back(*UI);




From echristo at apple.com  Tue Jul 27 20:52:23 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 01:52:23 -0000
Subject: [llvm-commits] [llvm] r109572 -
	/llvm/trunk/utils/TableGen/FastISelEmitter.cpp
Message-ID: <20100728015223.B65A92A6C12C@llvm.org>

Author: echristo
Date: Tue Jul 27 20:52:23 2010
New Revision: 109572

URL: http://llvm.org/viewvc/llvm-project?rev=109572&view=rev
Log:
Filter out patterns that have PredicateOperands.

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

Modified: llvm/trunk/utils/TableGen/FastISelEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/FastISelEmitter.cpp?rev=109572&r1=109571&r2=109572&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/FastISelEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/FastISelEmitter.cpp Tue Jul 27 20:52:23 2010
@@ -54,6 +54,7 @@
   bool initialize(TreePatternNode *InstPatNode,
                   const CodeGenTarget &Target,
                   MVT::SimpleValueType VT) {
+
     if (!InstPatNode->isLeaf()) {
       if (InstPatNode->getOperator()->getName() == "imm") {
         Operands.push_back("i");
@@ -69,6 +70,7 @@
     
     for (unsigned i = 0, e = InstPatNode->getNumChildren(); i != e; ++i) {
       TreePatternNode *Op = InstPatNode->getChild(i);
+      
       // For now, filter out any operand with a predicate.
       // For now, filter out any operand with multiple values.
       if (!Op->getPredicateFns().empty() ||
@@ -105,6 +107,7 @@
         RC = Target.getRegisterClassForRegister(OpLeafRec);
       else
         return false;
+        
       // For now, require the register operands' register classes to all
       // be the same.
       if (!RC)
@@ -262,6 +265,15 @@
     if (II.OperandList.empty())
       continue;
 
+    // For now ignore instructions that have predicate operands.
+    bool HasPredicate = false;
+    for (unsigned i = 0, e = II.OperandList.size(); i != e; ++i) {
+      if(II.OperandList[i].Rec->isSubClassOf("PredicateOperand"))
+        HasPredicate = true;
+    }
+    if (HasPredicate)
+      continue;
+      
     // For now, ignore multi-instruction patterns.
     bool MultiInsts = false;
     for (unsigned i = 0, e = Dst->getNumChildren(); i != e; ++i) {




From evan.cheng at apple.com  Tue Jul 27 21:01:40 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 27 Jul 2010 19:01:40 -0700
Subject: [llvm-commits] [llvm] r109572 -
	/llvm/trunk/utils/TableGen/FastISelEmitter.cpp
In-Reply-To: <20100728015223.B65A92A6C12C@llvm.org>
References: <20100728015223.B65A92A6C12C@llvm.org>
Message-ID: <6E54BDF5-8D67-4827-9E54-7E84AE0F36AF@apple.com>

Is that desirable? Just about every ARM instruction has PredicateOperands.

Evan

On Jul 27, 2010, at 6:52 PM, Eric Christopher wrote:

> Author: echristo
> Date: Tue Jul 27 20:52:23 2010
> New Revision: 109572
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109572&view=rev
> Log:
> Filter out patterns that have PredicateOperands.
> 
> Modified:
>    llvm/trunk/utils/TableGen/FastISelEmitter.cpp
> 
> Modified: llvm/trunk/utils/TableGen/FastISelEmitter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/FastISelEmitter.cpp?rev=109572&r1=109571&r2=109572&view=diff
> ==============================================================================
> --- llvm/trunk/utils/TableGen/FastISelEmitter.cpp (original)
> +++ llvm/trunk/utils/TableGen/FastISelEmitter.cpp Tue Jul 27 20:52:23 2010
> @@ -54,6 +54,7 @@
>   bool initialize(TreePatternNode *InstPatNode,
>                   const CodeGenTarget &Target,
>                   MVT::SimpleValueType VT) {
> +
>     if (!InstPatNode->isLeaf()) {
>       if (InstPatNode->getOperator()->getName() == "imm") {
>         Operands.push_back("i");
> @@ -69,6 +70,7 @@
> 
>     for (unsigned i = 0, e = InstPatNode->getNumChildren(); i != e; ++i) {
>       TreePatternNode *Op = InstPatNode->getChild(i);
> +      
>       // For now, filter out any operand with a predicate.
>       // For now, filter out any operand with multiple values.
>       if (!Op->getPredicateFns().empty() ||
> @@ -105,6 +107,7 @@
>         RC = Target.getRegisterClassForRegister(OpLeafRec);
>       else
>         return false;
> +        
>       // For now, require the register operands' register classes to all
>       // be the same.
>       if (!RC)
> @@ -262,6 +265,15 @@
>     if (II.OperandList.empty())
>       continue;
> 
> +    // For now ignore instructions that have predicate operands.
> +    bool HasPredicate = false;
> +    for (unsigned i = 0, e = II.OperandList.size(); i != e; ++i) {
> +      if(II.OperandList[i].Rec->isSubClassOf("PredicateOperand"))
> +        HasPredicate = true;
> +    }
> +    if (HasPredicate)
> +      continue;
> +      
>     // For now, ignore multi-instruction patterns.
>     bool MultiInsts = false;
>     for (unsigned i = 0, e = Dst->getNumChildren(); i != e; ++i) {
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From echristo at apple.com  Tue Jul 27 21:02:26 2010
From: echristo at apple.com (Eric Christopher)
Date: Tue, 27 Jul 2010 19:02:26 -0700
Subject: [llvm-commits] [llvm] r109572 -
	/llvm/trunk/utils/TableGen/FastISelEmitter.cpp
In-Reply-To: <6E54BDF5-8D67-4827-9E54-7E84AE0F36AF@apple.com>
References: <20100728015223.B65A92A6C12C@llvm.org>
	<6E54BDF5-8D67-4827-9E54-7E84AE0F36AF@apple.com>
Message-ID: <FCB49C90-98F0-4695-8A96-66D684E6F3EA@apple.com>


On Jul 27, 2010, at 7:01 PM, Evan Cheng wrote:

> Is that desirable? Just about every ARM instruction has PredicateOperands.

Temporary while I'm rewriting the operand handling, just in case something else has a predicate operand since we can't handle it at the moment :)

-eric


From daniel at zuster.org  Tue Jul 27 23:51:10 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Tue, 27 Jul 2010 21:51:10 -0700
Subject: [llvm-commits] [llvm] r109570 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
In-Reply-To: <20100728010907.D9BE32A6C12C@llvm.org>
References: <20100728010907.D9BE32A6C12C@llvm.org>
Message-ID: <AANLkTi=iOCCgoMNc6CTxYoVG491rVHGUNo_yN2iYv93p@mail.gmail.com>

Hi Dan,

On Tue, Jul 27, 2010 at 6:09 PM, Dan Gohman <gohman at apple.com> wrote:
> Author: djg
> Date: Tue Jul 27 20:09:07 2010
> New Revision: 109570
>
> URL: http://llvm.org/viewvc/llvm-project?rev=109570&view=rev
> Log:
> Make SCEVCallbackVH::allUsesReplacedWith update the old SCEVUnknown
> object, as it may still be referenced by SCEVs not cleaned up by the
> use list traversal.
>
> Also, in ScalarEvolution::forgetValue, only check for a SCEVUnknown
> object for the original value, not for any value in the use list,
> because other SCEVUnknown values aren't necessary obsolete at that
> point.
>
> Modified:
> ? ?llvm/trunk/lib/Analysis/ScalarEvolution.cpp
>
> Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=109570&r1=109569&r2=109570&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
> +++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue Jul 27 20:09:07 2010
> @@ -3664,6 +3664,26 @@
> ?/// changed a value in a way that may effect its value, or which may
> ?/// disconnect it from a def-use chain linking it to a loop.
> ?void ScalarEvolution::forgetValue(Value *V) {
> + ?// If there's a SCEVUnknown tying this value into the SCEV
> + ?// space, remove it from the folding set map. The SCEVUnknown
> + ?// object and any other SCEV objects which reference it
> + ?// (transitively) remain allocated, effectively leaked until
> + ?// the underlying BumpPtrAllocator is freed.
> + ?//
> + ?// This permits SCEV pointers to be used as keys in maps
> + ?// such as the ValuesAtScopes map.
> + ?FoldingSetNodeID ID;
> + ?ID.AddInteger(scUnknown);
> + ?ID.AddPointer(V);
> + ?void *IP;
> + ?if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
> + ? ?UniqueSCEVs.RemoveNode(S);
> +
> + ? ?// This isn't necessary, but we might as well remove the
> + ? ?// value from the ValuesAtScopes map too.
> + ? ?ValuesAtScopes.erase(S);
> + ?}
> +
> ? Instruction *I = dyn_cast<Instruction>(V);
> ? if (!I) return;
>
> @@ -3685,26 +3705,6 @@
> ? ? ? ? ConstantEvolutionLoopExitValue.erase(PN);
> ? ? }
>
> - ? ?// If there's a SCEVUnknown tying this value into the SCEV
> - ? ?// space, remove it from the folding set map. The SCEVUnknown
> - ? ?// object and any other SCEV objects which reference it
> - ? ?// (transitively) remain allocated, effectively leaked until
> - ? ?// the underlying BumpPtrAllocator is freed.
> - ? ?//
> - ? ?// This permits SCEV pointers to be used as keys in maps
> - ? ?// such as the ValuesAtScopes map.
> - ? ?FoldingSetNodeID ID;
> - ? ?ID.AddInteger(scUnknown);
> - ? ?ID.AddPointer(I);
> - ? ?void *IP;
> - ? ?if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
> - ? ? ?UniqueSCEVs.RemoveNode(S);
> -
> - ? ? ?// This isn't necessary, but we might as well remove the
> - ? ? ?// value from the ValuesAtScopes map too.
> - ? ? ?ValuesAtScopes.erase(S);
> - ? ?}
> -
> ? ? PushDefUseChildren(I, Worklist);
> ? }
> ?}
> @@ -5690,15 +5690,32 @@
> ? // this now dangles!
> ?}
>
> -void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *) {
> +void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *V) {
> ? assert(SE && "SCEVCallbackVH called with a null ScalarEvolution!");
>
> + ?Value *Old = getValPtr();
> +
> + ?// If there's a SCEVUnknown tying this value into the SCEV
> + ?// space, replace the SCEVUnknown's value with the new value
> + ?// for the benefit of any SCEVs still referencing it, and
> + ?// and remove it from the folding set map so that new scevs
> + ?// don't reference it.
> + ?FoldingSetNodeID ID;
> + ?ID.AddInteger(scUnknown);
> + ?ID.AddPointer(Old);
> + ?void *IP;
> + ?if (SCEVUnknown *S = cast_or_null<SCEVUnknown>(
> + ? ? ? ?SE->UniqueSCEVs.FindNodeOrInsertPos(ID, IP))) {
> + ? ?S->V = V;

gcc-4.0 seems unhappy with this:
--
ScalarEvolution.cpp: In member function 'virtual void
llvm::ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(llvm::Value*)':
/Users/buildslave/zorg/buildbot/smooshlab/slave/build.clang-i386-darwin9/llvm/include/llvm/Analysis/ScalarEvolutionExpressions.h:528:
error: 'llvm::Value* llvm::SCEVUnknown::V' is private
ScalarEvolution.cpp:5709: error: within this context
--

 - Daniel

> + ? ?SE->UniqueSCEVs.RemoveNode(S);
> + ? ?SE->ValuesAtScopes.erase(S);
> + ?}
> +
> ? // Forget all the expressions associated with users of the old value,
> ? // so that future queries will recompute the expressions using the new
> ? // value.
> ? SmallVector<User *, 16> Worklist;
> ? SmallPtrSet<User *, 8> Visited;
> - ?Value *Old = getValPtr();
> ? for (Value::use_iterator UI = Old->use_begin(), UE = Old->use_end();
> ? ? ? ?UI != UE; ++UI)
> ? ? Worklist.push_back(*UI);
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From renato.golin at arm.com  Wed Jul 28 03:36:36 2010
From: renato.golin at arm.com (Renato Golin)
Date: Wed, 28 Jul 2010 09:36:36 +0100
Subject: [llvm-commits] Patch proposal: DIDescriptor constructors
Message-ID: <AANLkTikaBk6MFvhz-2nc4xwJkO0POANM1USm9LwkoYTq@mail.gmail.com>

Debug info in LLVM is now done by passing small objects, rather than
pointers. To some degree it works, but it's difficult to pass derived
objects into functions that would expect base ones.

For that purpose, I have added a few constructors to DIDescriptor to
enable that functionality without changing the way the debug
infrastructure works, but that doesn't work in the opposite direction.

Patch extracted from yesterday's (Tue, 27 Jul, 2010) trunk.

cheers,
--renato
-------------- next part --------------
A non-text attachment was scrubbed...
Name: debug-info.patch
Type: text/x-patch
Size: 1786 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100728/b2ef5888/attachment.bin 

From renato.golin at arm.com  Wed Jul 28 03:40:03 2010
From: renato.golin at arm.com (Renato Golin)
Date: Wed, 28 Jul 2010 09:40:03 +0100
Subject: [llvm-commits] Patch proposal: DIDescriptor constructors (patch
	inline)
Message-ID: <AANLkTinD+0OXezaJBbONhrmGeHfMieEaYcOuDmz-Yk70@mail.gmail.com>

Debug info in LLVM is now done by passing small objects, rather than
pointers. To some degree it works, but it's difficult to pass derived
objects into functions that would expect base ones.
For that purpose, I have added a few constructors to DIDescriptor to
enable that functionality without changing the way the debug
infrastructure works, but that doesn't work in the opposite direction.

Patch extracted from yesterday's (Tue, 27 Jul, 2010) trunk, now inline.

Modifyed:
  include/llvm/Analysis/DebugInfo.h
  lib/Analysis/DebugInfo.cpp


Index: include/llvm/Analysis/DebugInfo.h
===================================================================
--- include/llvm/Analysis/DebugInfo.h   (revision 109509)
+++ include/llvm/Analysis/DebugInfo.h   (working copy)
@@ -36,6 +36,12 @@
   class LLVMContext;
   class raw_ostream;

+  class DIFile;
+  class DISubprogram;
+  class DILexicalBlock;
+  class DIVariable;
+  class DIType;
+
   /// DIDescriptor - A thin wraper around MDNode to access encoded debug info.
   /// This should not be stored in a container, because underly MDNode may
   /// change in certain situations.
@@ -61,6 +67,11 @@
   public:
     explicit DIDescriptor() : DbgNode(0) {}
     explicit DIDescriptor(const MDNode *N) : DbgNode(N) {}
+    explicit DIDescriptor(const DIFile F);
+    explicit DIDescriptor(const DISubprogram F);
+    explicit DIDescriptor(const DILexicalBlock F);
+    explicit DIDescriptor(const DIVariable F);
+    explicit DIDescriptor(const DIType F);

     bool Verify() const { return DbgNode != 0; }

Index: lib/Analysis/DebugInfo.cpp
===================================================================
--- lib/Analysis/DebugInfo.cpp  (revision 109509)
+++ lib/Analysis/DebugInfo.cpp  (working copy)
@@ -32,6 +32,21 @@
 // DIDescriptor
 //===----------------------------------------------------------------------===//

+DIDescriptor::DIDescriptor(const DIFile F) : DbgNode(F.DbgNode) {
+}
+
+DIDescriptor::DIDescriptor(const DISubprogram F) : DbgNode(F.DbgNode) {
+}
+
+DIDescriptor::DIDescriptor(const DILexicalBlock F) : DbgNode(F.DbgNode) {
+}
+
+DIDescriptor::DIDescriptor(const DIVariable F) : DbgNode(F.DbgNode) {
+}
+
+DIDescriptor::DIDescriptor(const DIType F) : DbgNode(F.DbgNode) {
+}
+
 StringRef
 DIDescriptor::getStringField(unsigned Elt) const {
   if (DbgNode == 0)


From baldrick at free.fr  Wed Jul 28 05:16:18 2010
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 28 Jul 2010 12:16:18 +0200
Subject: [llvm-commits] [llvm] r109169 - in
 /llvm/trunk/lib/ExecutionEngine/JIT: JITDwarfEmitter.cpp JITDwarfEmitter.h
In-Reply-To: <C2F703B2-4B16-4BBC-BA54-903C7C6B3899@apple.com>
References: <20100722212040.ABC6C2A6C12C@llvm.org> <4C49477B.7060200@free.fr>
	<C2F703B2-4B16-4BBC-BA54-903C7C6B3899@apple.com>
Message-ID: <4C500372.9070108@free.fr>

>> we should really arrange for svn to automatically add a diffstat section,
>> in order to better appreciate all the deleting you are doing :)
>
> We used to have that back in the CVS days, but it didn't make it over in the transition to SVN.

Maybe Owen can add it, since he seems to have been experimenting with the
subversion server recently...

Ciao,

Duncan.


From ggreif at gmail.com  Wed Jul 28 05:45:00 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 10:45:00 -0000
Subject: [llvm-commits] [llvm] r109576 -
	/llvm/trunk/include/llvm/Support/CallSite.h
Message-ID: <20100728104500.246A92A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 05:44:59 2010
New Revision: 109576

URL: http://llvm.org/viewvc/llvm-project?rev=109576&view=rev
Log:
we are supposed to only create proper CallSites from an instruction (esp. CallInst and InvokeInst)

Modified:
    llvm/trunk/include/llvm/Support/CallSite.h

Modified: llvm/trunk/include/llvm/Support/CallSite.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/CallSite.h?rev=109576&r1=109575&r2=109576&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/CallSite.h (original)
+++ llvm/trunk/include/llvm/Support/CallSite.h Wed Jul 28 05:44:59 2010
@@ -49,8 +49,8 @@
   PointerIntPair<InstrTy*, 1, bool> I;
 public:
   CallSiteBase() : I(0, false) {}
-  CallSiteBase(CallTy *CI) : I(CI, true) { /*assert(CI);*/ }
-  CallSiteBase(InvokeTy *II) : I(II, false) { /*assert(II);*/ }
+  CallSiteBase(CallTy *CI) : I(CI, true) { assert(CI); }
+  CallSiteBase(InvokeTy *II) : I(II, false) { assert(II); }
   CallSiteBase(ValTy *II) { *this = get(II); }
   CallSiteBase(InstrTy *II) {
     assert(II && "Null instruction given?");




From ggreif at gmail.com  Wed Jul 28 05:46:09 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 10:46:09 -0000
Subject: [llvm-commits] [llvm] r109577 -
	/llvm/trunk/lib/Analysis/AliasSetTracker.cpp
Message-ID: <20100728104609.D5BFB2A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 05:46:09 2010
New Revision: 109577

URL: http://llvm.org/viewvc/llvm-project?rev=109577&view=rev
Log:
simplify

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

Modified: llvm/trunk/lib/Analysis/AliasSetTracker.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/AliasSetTracker.cpp?rev=109577&r1=109576&r2=109577&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/AliasSetTracker.cpp (original)
+++ llvm/trunk/lib/Analysis/AliasSetTracker.cpp Wed Jul 28 05:46:09 2010
@@ -456,8 +456,7 @@
 
   // If this is a call instruction, remove the callsite from the appropriate
   // AliasSet.
-  CallSite CS = CallSite::get(PtrVal);
-  if (CS.getInstruction())
+  if (CallSite CS = PtrVal)
     if (!AA.doesNotAccessMemory(CS))
       if (AliasSet *AS = findAliasSetForCallSite(CS))
         AS->removeCallSite(CS);




From ggreif at gmail.com  Wed Jul 28 05:57:28 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 10:57:28 -0000
Subject: [llvm-commits] [llvm] r109578 -
	/llvm/trunk/lib/Analysis/CaptureTracking.cpp
Message-ID: <20100728105728.410592A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 05:57:28 2010
New Revision: 109578

URL: http://llvm.org/viewvc/llvm-project?rev=109578&view=rev
Log:
simplify

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

Modified: llvm/trunk/lib/Analysis/CaptureTracking.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/CaptureTracking.cpp?rev=109578&r1=109577&r2=109578&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/CaptureTracking.cpp (original)
+++ llvm/trunk/lib/Analysis/CaptureTracking.cpp Wed Jul 28 05:57:28 2010
@@ -69,7 +69,7 @@
     switch (I->getOpcode()) {
     case Instruction::Call:
     case Instruction::Invoke: {
-      CallSite CS = CallSite::get(I);
+      CallSite CS(I);
       // Not captured if the callee is readonly, doesn't return a copy through
       // its return value and doesn't unwind (a readonly function can leak bits
       // by throwing an exception or not depending on the input value).




From ggreif at gmail.com  Wed Jul 28 07:19:46 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 12:19:46 -0000
Subject: [llvm-commits] [llvm] r109579 -
	/llvm/trunk/lib/Analysis/IPA/CallGraph.cpp
Message-ID: <20100728121946.9EF972A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 07:19:46 2010
New Revision: 109579

URL: http://llvm.org/viewvc/llvm-project?rev=109579&view=rev
Log:
use Value* constructor of CallSite to create potentially improper site

Modified:
    llvm/trunk/lib/Analysis/IPA/CallGraph.cpp

Modified: llvm/trunk/lib/Analysis/IPA/CallGraph.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/IPA/CallGraph.cpp?rev=109579&r1=109578&r2=109579&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/IPA/CallGraph.cpp (original)
+++ llvm/trunk/lib/Analysis/IPA/CallGraph.cpp Wed Jul 28 07:19:46 2010
@@ -145,8 +145,8 @@
     for (Function::iterator BB = F->begin(), BBE = F->end(); BB != BBE; ++BB)
       for (BasicBlock::iterator II = BB->begin(), IE = BB->end();
            II != IE; ++II) {
-        CallSite CS = CallSite::get(II);
-        if (CS.getInstruction() && !isa<DbgInfoIntrinsic>(II)) {
+        CallSite CS(cast<Value>(II));
+        if (CS && !isa<DbgInfoIntrinsic>(II)) {
           const Function *Callee = CS.getCalledFunction();
           if (Callee)
             Node->addCalledFunction(CS, getOrInsertFunction(Callee));




From ggreif at gmail.com  Wed Jul 28 07:35:54 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 12:35:54 -0000
Subject: [llvm-commits] [llvm] r109580 -
	/llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp
Message-ID: <20100728123554.888D12A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 07:35:54 2010
New Revision: 109580

URL: http://llvm.org/viewvc/llvm-project?rev=109580&view=rev
Log:
use Value* constructor of CallSite to create potentially improper site, and test that

Modified:
    llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp

Modified: llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp?rev=109580&r1=109579&r2=109580&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp (original)
+++ llvm/trunk/lib/Analysis/IPA/CallGraphSCCPass.cpp Wed Jul 28 07:35:54 2010
@@ -209,7 +209,7 @@
           // If the call edge is not from a call or invoke, then the function
           // pass RAUW'd a call with another value.  This can happen when
           // constant folding happens of well known functions etc.
-          CallSite::get(I->first).getInstruction() == 0) {
+          !CallSite(I->first)) {
         assert(!CheckingMode &&
                "CallGraphSCCPass did not update the CallGraph correctly!");
         
@@ -245,8 +245,8 @@
     
     for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
       for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
-        CallSite CS = CallSite::get(I);
-        if (!CS.getInstruction() || isa<DbgInfoIntrinsic>(I)) continue;
+          CallSite CS(cast<Value>(I));
+        if (!CS || isa<DbgInfoIntrinsic>(I)) continue;
         
         // If this call site already existed in the callgraph, just verify it
         // matches up to expectations and remove it from CallSites.




From ggreif at gmail.com  Wed Jul 28 09:28:18 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 14:28:18 -0000
Subject: [llvm-commits] [llvm] r109581 -
	/llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp
Message-ID: <20100728142818.D14C12A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 09:28:18 2010
New Revision: 109581

URL: http://llvm.org/viewvc/llvm-project?rev=109581&view=rev
Log:
use Value* constructor of CallSite to create potentially improper site, and test that

Modified:
    llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp?rev=109581&r1=109580&r2=109581&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp Wed Jul 28 09:28:18 2010
@@ -401,10 +401,9 @@
       }
       
       continue;
-    } else if (CallSite::get(BBI).getInstruction() != 0) {
+    } else if (CallSite CS = cast<Value>(BBI)) {
       // If this call does not access memory, it can't
       // be undeadifying any of our pointers.
-      CallSite CS = CallSite::get(BBI);
       if (AA.doesNotAccessMemory(CS))
         continue;
       




From gohman at apple.com  Wed Jul 28 10:17:26 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 15:17:26 -0000
Subject: [llvm-commits] [llvm] r109584 - in /llvm/trunk/examples: BrainF/
 ExceptionDemo/ Fibonacci/ HowToUseJIT/ Kaleidoscope/Chapter2/
 Kaleidoscope/Chapter3/ Kaleidoscope/Chapter4/ Kaleidoscope/Chapter5/
 Kaleidoscope/Chapter6/ Kaleidoscope/Chapter7/ ModuleMaker/ ParallelJIT/
Message-ID: <20100728151727.035E42A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 10:17:26 2010
New Revision: 109584

URL: http://llvm.org/viewvc/llvm-project?rev=109584&view=rev
Log:
Update svn:ignore properties.

Modified:
    llvm/trunk/examples/BrainF/   (props changed)
    llvm/trunk/examples/ExceptionDemo/   (props changed)
    llvm/trunk/examples/Fibonacci/   (props changed)
    llvm/trunk/examples/HowToUseJIT/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter2/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter3/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter4/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter5/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter6/   (props changed)
    llvm/trunk/examples/Kaleidoscope/Chapter7/   (props changed)
    llvm/trunk/examples/ModuleMaker/   (props changed)
    llvm/trunk/examples/ParallelJIT/   (props changed)

Propchange: llvm/trunk/examples/BrainF/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -1,7 +1,10 @@
-Release
 Debug
+Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/ExceptionDemo/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -0,0 +1,10 @@
+Debug
+Release
+Release-Asserts
+Debug+Coverage-Asserts
+Release+Coverage-Asserts
+Debug+Coverage
+Release+Coverage
+Debug+Checks
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Fibonacci/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,6 +2,10 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
+Debug+Asserts
+Release+Asserts
+

Propchange: llvm/trunk/examples/HowToUseJIT/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,6 +2,10 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
+Debug+Asserts
+Release+Asserts
+

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter2/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter3/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter4/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter5/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter6/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/Kaleidoscope/Chapter7/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,7 +2,9 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
-
+Debug+Asserts
+Release+Asserts

Propchange: llvm/trunk/examples/ModuleMaker/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,6 +2,10 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
+Debug+Asserts
+Release+Asserts
+

Propchange: llvm/trunk/examples/ParallelJIT/
------------------------------------------------------------------------------
--- svn:ignore (original)
+++ svn:ignore Wed Jul 28 10:17:26 2010
@@ -2,6 +2,10 @@
 Release
 Release-Asserts
 Debug+Coverage-Asserts
+Release+Coverage-Asserts
 Debug+Coverage
 Release+Coverage
 Debug+Checks
+Debug+Asserts
+Release+Asserts
+




From ggreif at gmail.com  Wed Jul 28 10:31:37 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 15:31:37 -0000
Subject: [llvm-commits] [llvm] r109585 -
	/llvm/trunk/lib/Analysis/AliasAnalysisEvaluator.cpp
Message-ID: <20100728153137.EFF0F2A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 10:31:37 2010
New Revision: 109585

URL: http://llvm.org/viewvc/llvm-project?rev=109585&view=rev
Log:
simplify

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

Modified: llvm/trunk/lib/Analysis/AliasAnalysisEvaluator.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/AliasAnalysisEvaluator.cpp?rev=109585&r1=109584&r2=109585&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/AliasAnalysisEvaluator.cpp (original)
+++ llvm/trunk/lib/Analysis/AliasAnalysisEvaluator.cpp Wed Jul 28 10:31:37 2010
@@ -126,8 +126,7 @@
     if (I->getType()->isPointerTy()) // Add all pointer instructions.
       Pointers.insert(&*I);
     Instruction &Inst = *I;
-    CallSite CS = CallSite::get(&Inst);
-    if (CS) {
+    if (CallSite CS = cast<Value>(&Inst)) {
       Value *Callee = CS.getCalledValue();
       // Skip actual functions for direct function calls.
       if (!isa<Function>(Callee) && isInterestingPointer(Callee))
@@ -137,6 +136,7 @@
            AI != AE; ++AI)
         if (isInterestingPointer(*AI))
           Pointers.insert(*AI);
+      CallSites.insert(CS);
     } else {
       // Consider all operands.
       for (Instruction::op_iterator OI = Inst.op_begin(), OE = Inst.op_end();
@@ -144,8 +144,6 @@
         if (isInterestingPointer(*OI))
           Pointers.insert(*OI);
     }
-
-    if (CS.getInstruction()) CallSites.insert(CS);
   }
 
   if (PrintNoAlias || PrintMayAlias || PrintMustAlias ||




From daniel at zuster.org  Wed Jul 28 10:40:20 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 28 Jul 2010 15:40:20 -0000
Subject: [llvm-commits] [llvm] r109586 - in /llvm/trunk:
 include/llvm/Support/CrashRecoveryContext.h
 lib/Support/CrashRecoveryContext.cpp
Message-ID: <20100728154020.9AA472A6C12C@llvm.org>

Author: ddunbar
Date: Wed Jul 28 10:40:20 2010
New Revision: 109586

URL: http://llvm.org/viewvc/llvm-project?rev=109586&view=rev
Log:
Support: Add CrashRecoveryContext helper object.
 - Designed as a simple wrapper to allow clients to attempt to catch crashes
   (memory errors, assertion violations, etc.) and do some kind of recovery.

 - Currently doesn't actually attempt to catch crashes.

Added:
    llvm/trunk/include/llvm/Support/CrashRecoveryContext.h
    llvm/trunk/lib/Support/CrashRecoveryContext.cpp

Added: llvm/trunk/include/llvm/Support/CrashRecoveryContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/CrashRecoveryContext.h?rev=109586&view=auto
==============================================================================
--- llvm/trunk/include/llvm/Support/CrashRecoveryContext.h (added)
+++ llvm/trunk/include/llvm/Support/CrashRecoveryContext.h Wed Jul 28 10:40:20 2010
@@ -0,0 +1,82 @@
+//===--- CrashRecoveryContext.h - Crash Recovery ----------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_SUPPORT_CRASHRECOVERYCONTEXT_H
+#define LLVM_SUPPORT_CRASHRECOVERYCONTEXT_H
+
+#include <string>
+
+namespace llvm {
+class StringRef;
+
+/// \brief Crash recovery helper object.
+///
+/// This class implements support for running operations in a safe context so
+/// that crashes (memory errors, stack overflow, assertion violations) can be
+/// detected and control restored to the crashing thread. Crash detection is
+/// purely "best effort", the exact set of failures which can be recovered from
+/// is platform dependent.
+///
+/// Clients make use of this code by first calling
+/// CrashRecoveryContext::Enable(), and then executing unsafe operations via a
+/// CrashRecoveryContext object. For example:
+///
+///    void actual_work(void *);
+///
+///    void foo() {
+///      CrashRecoveryContext CRC;
+///
+///      if (!CRC.RunSafely(actual_work, 0)) {
+///         ... a crash was detected, report error to user ...
+///      }
+///
+///      ... no crash was detected ...
+///    }
+///
+/// Crash recovery contexts may not be nested.
+class CrashRecoveryContext {
+  void *Impl;
+
+public:
+  CrashRecoveryContext() : Impl(0) {}
+  ~CrashRecoveryContext();
+
+  /// \brief Enable crash recovery. This function is not thread safe, clients
+  /// should call it during startup or with a lock held.
+  static void Enable();
+
+  /// \brief Disable crash recovery. This function is not thread safe, clients
+  /// should call it during startup or with a lock held.
+  static void Disable();
+
+  /// \brief Execute the provide callback function (with the given arguments) in
+  /// a protected context.
+  ///
+  /// \return True if the function completed successfully, and false if the
+  /// function crashed (or HandleCrash was called explicitly). Clients should
+  /// make as little assumptions as possible about the program state when
+  /// RunSafely has returned false. Clients can use getBacktrace() to retrieve
+  /// the backtrace of the crash on failures.
+  bool RunSafely(void (*Fn)(void*), void *UserData);
+
+  /// \brief Explicitly trigger a crash recovery in the current process, and
+  /// return failure from RunSafely(). This function does not return.
+  void HandleCrash();
+
+  /// \brief Return a string containing the backtrace where the crash was
+  /// detected; or empty if the backtrace wasn't recovered.
+  ///
+  /// This function is only valid when a crash has been detected (i.e.,
+  /// RunSafely() has returned false.
+  const std::string &getBacktrace() const;
+};
+
+}
+
+#endif

Added: llvm/trunk/lib/Support/CrashRecoveryContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CrashRecoveryContext.cpp?rev=109586&view=auto
==============================================================================
--- llvm/trunk/lib/Support/CrashRecoveryContext.cpp (added)
+++ llvm/trunk/lib/Support/CrashRecoveryContext.cpp Wed Jul 28 10:40:20 2010
@@ -0,0 +1,88 @@
+//===--- CrashRecoveryContext.cpp - Crash Recovery ------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Support/CrashRecoveryContext.h"
+#include "llvm/ADT/SmallString.h"
+#include <setjmp.h>
+using namespace llvm;
+
+namespace {
+
+struct CrashRecoveryContextImpl;
+
+struct CrashRecoveryContextImpl {
+  std::string Backtrace;
+  ::jmp_buf JumpBuffer;
+  volatile unsigned Failed : 1;
+
+public:
+  CrashRecoveryContextImpl() : Failed(false) {}
+
+  void HandleCrash() {
+    assert(!Failed && "Crash recovery context already failed!");
+    Failed = true;
+
+    // FIXME: Stash the backtrace.
+
+    // Jump back to the RunSafely we were called under.
+    longjmp(JumpBuffer, 1);
+  }
+};
+
+}
+
+static bool gCrashRecoveryEnabled = false;
+
+CrashRecoveryContext::~CrashRecoveryContext() {
+  CrashRecoveryContextImpl *CRCI = (CrashRecoveryContextImpl *) Impl;
+  delete CRCI;
+}
+
+void CrashRecoveryContext::Enable() {
+  if (gCrashRecoveryEnabled)
+    return;
+
+  gCrashRecoveryEnabled = true;
+}
+
+void CrashRecoveryContext::Disable() {
+  if (!gCrashRecoveryEnabled)
+    return;
+
+  gCrashRecoveryEnabled = false;
+}
+
+bool CrashRecoveryContext::RunSafely(void (*Fn)(void*), void *UserData) {
+  // If crash recovery is disabled, do nothing.
+  if (gCrashRecoveryEnabled) {
+    assert(!Impl && "Crash recovery context already initialized!");
+    CrashRecoveryContextImpl *CRCI = new CrashRecoveryContextImpl;
+    Impl = CRCI;
+
+    if (setjmp(CRCI->JumpBuffer) != 0) {
+      return false;
+    }
+  }
+
+  Fn(UserData);
+  return true;
+}
+
+void CrashRecoveryContext::HandleCrash() {
+  CrashRecoveryContextImpl *CRCI = (CrashRecoveryContextImpl *) Impl;
+  assert(CRCI && "Crash recovery context never initialized!");
+  CRCI->HandleCrash();
+}
+
+const std::string &CrashRecoveryContext::getBacktrace() const {
+  CrashRecoveryContextImpl *CRC = (CrashRecoveryContextImpl *) Impl;
+  assert(CRC && "Crash recovery context never initialized!");
+  assert(CRC->Failed && "No crash was detected!");
+  return CRC->Backtrace;
+}




From ggreif at gmail.com  Wed Jul 28 10:52:43 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 15:52:43 -0000
Subject: [llvm-commits] [llvm] r109589 -
	/llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp
Message-ID: <20100728155243.34ED32A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 10:52:43 2010
New Revision: 109589

URL: http://llvm.org/viewvc/llvm-project?rev=109589&view=rev
Log:
simplify

Modified:
    llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp

Modified: llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp?rev=109589&r1=109588&r2=109589&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/BasicInliner.cpp Wed Jul 28 10:52:43 2010
@@ -82,8 +82,8 @@
     Function *F = *FI;
     for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
       for (BasicBlock::iterator I = BB->begin(); I != BB->end(); ++I) {
-        CallSite CS = CallSite::get(I);
-        if (CS.getInstruction() && CS.getCalledFunction()
+        CallSite CS(cast<Value>(I));
+        if (CS && CS.getCalledFunction()
             && !CS.getCalledFunction()->isDeclaration())
           CallSites.push_back(CS);
       }




From clattner at apple.com  Wed Jul 28 11:12:22 2010
From: clattner at apple.com (Chris Lattner)
Date: Wed, 28 Jul 2010 09:12:22 -0700
Subject: [llvm-commits] [llvm] r109586 - in /llvm/trunk:
	include/llvm/Support/CrashRecoveryContext.h
	lib/Support/CrashRecoveryContext.cpp
In-Reply-To: <20100728154020.9AA472A6C12C@llvm.org>
References: <20100728154020.9AA472A6C12C@llvm.org>
Message-ID: <C02B93B3-9E00-44C9-8851-E3444DCF1E58@apple.com>


On Jul 28, 2010, at 8:40 AM, Daniel Dunbar wrote:

> Author: ddunbar
> Date: Wed Jul 28 10:40:20 2010
> New Revision: 109586
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109586&view=rev
> Log:
> Support: Add CrashRecoveryContext helper object.
> - Designed as a simple wrapper to allow clients to attempt to catch crashes
>   (memory errors, assertion violations, etc.) and do some kind of recovery.
> 
> - Currently doesn't actually attempt to catch crashes.

Looks like a nice interface.

> +  /// \return True if the function completed successfully, and false if the
> +  /// function crashed (or HandleCrash was called explicitly). Clients should
> +  /// make as little assumptions as possible about the program state when
> +  /// RunSafely has returned false. Clients can use getBacktrace() to retrieve
> +  /// the backtrace of the crash on failures.
> +  bool RunSafely(void (*Fn)(void*), void *UserData);

It would be nice to add an #ifdef __BLOCKS__ version of this that took a block to execute.

-Chris




From geek4civic at gmail.com  Wed Jul 28 11:35:26 2010
From: geek4civic at gmail.com (NAKAMURA Takumi)
Date: Thu, 29 Jul 2010 01:35:26 +0900
Subject: [llvm-commits] Patch Proposal: tweak *.d on mingw-msys build
	environment
Message-ID: <AANLkTimugdBvFhk8XO7y-DUBMkr2xyzr1jCC3VB3KKYX@mail.gmail.com>

This patch substitutes DOSish driveletter(c:/xxx/) scheme to
msys-compatible(/c/xxx/) one
when build host is MingW.

I am using the msysgit (msys+mingw+git) package as development environment.
C:/msysgit/mingw/bin/g++.exe (TDM-1 mingw32) 4.4.0
C:/msysgit/mingw/bin/make.exe GNU Make 3.81

msys-make igonores dependency rules emitted by mingw-g++ with -MMD,
... without any warnings.

And I have confirmed this might not have any side effects on
fedora12-ppc and cygwin.


...Takumi
-------------- next part --------------
diff --git a/Makefile.rules b/Makefile.rules
index 0b0f02c..d8f8236 100644
--- a/Makefile.rules
+++ b/Makefile.rules
@@ -1457,8 +1457,15 @@ DEPEND_OPTIONS = -MMD -MP -MF "$(ObjDir)/$*.d.tmp" \
 
 # If the build succeeded, move the dependency file over, otherwise
 # remove it.
+ifeq ($(OS),MingW)
+DEPEND_MOVEFILE = then $(SED) 's/^\([ \t]*\)\([A-Za-z]\):/\1\/\2/g' \
+                        "$(ObjDir)/$*.d.tmp" > "$(ObjDir)/$*.d"; \
+                        $(RM) "$(ObjDir)/$*.d.tmp"; \
+                  else $(RM) "$(ObjDir)/$*.d.tmp"; exit 1; fi
+else
 DEPEND_MOVEFILE = then $(MV) -f "$(ObjDir)/$*.d.tmp" "$(ObjDir)/$*.d"; \
                   else $(RM) "$(ObjDir)/$*.d.tmp"; exit 1; fi
+endif
 
 $(ObjDir)/%.o: %.cpp $(ObjDir)/.dir $(BUILT_SOURCES) $(PROJ_SRC_DIR)/Makefile
 	$(Echo) "Compiling $*.cpp for $(BuildMode) build" $(PIC_FLAG)

From geek4civic at gmail.com  Wed Jul 28 11:46:28 2010
From: geek4civic at gmail.com (NAKAMURA Takumi)
Date: Thu, 29 Jul 2010 01:46:28 +0900
Subject: [llvm-commits] [llvm] r109385 - /llvm/trunk/Makefile.rules
In-Reply-To: <20100725223304.C4B6E2A6C12C@llvm.org>
References: <20100725223304.C4B6E2A6C12C@llvm.org>
Message-ID: <AANLkTimhjk8y207XRwz1awGqDRBYy7vCk0zx_FZ-+RSa@mail.gmail.com>

Nick,

I found this patch was incomplete. the improved one is attached.

with previous patch, the below is enabled on cygwin/mingw.
> ifneq ($(DARWIN_MAJVERS),4)
>   LD.Flags += $(RPATH) -Wl, at executable_path/../lib
> endif

I guess it would be less harmful but I did not forget former patch. XD


...Takumi



2010/7/26 Nick Lewycky <nicholas at mxc.ca>:
> Author: nicholas
> Date: Sun Jul 25 17:33:04 2010
> New Revision: 109385
>
> URL: http://llvm.org/viewvc/llvm-project?rev=109385&view=rev
> Log:
> Don't pass -export-dynamic to the linker on Cygwin and MinGW. These platforms
> accept the flag and do nothing but warn about it, cleverly bypassing our
> configure-based detection system. Patch by Takumi Nakamura!
>
> Modified:
> ? ?llvm/trunk/Makefile.rules
>
> Modified: llvm/trunk/Makefile.rules
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=109385&r1=109384&r2=109385&view=diff
> ==============================================================================
> --- llvm/trunk/Makefile.rules (original)
> +++ llvm/trunk/Makefile.rules Sun Jul 25 17:33:04 2010
> @@ -618,7 +618,7 @@
> ?endif
>
> ?# Adjust linker flags for building an executable
> -ifneq ($(HOST_OS),Darwin)
> +ifneq ($(HOST_OS), $(filter $(HOST_OS), Darwin Cygwin MingW))
> ?ifdef TOOLNAME
> ? LD.Flags += $(RPATH) -Wl,'$$ORIGIN/../lib'
> ? ifdef EXAMPLE_TOOL
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
-------------- next part --------------
diff --git a/Makefile.rules b/Makefile.rules
index 0b0f02c..85d9a75 100644
--- a/Makefile.rules
+++ b/Makefile.rules
@@ -618,7 +618,8 @@ ifndef KEEP_SYMBOLS
 endif
 
 # Adjust linker flags for building an executable
-ifneq ($(HOST_OS), $(filter $(HOST_OS), Darwin Cygwin MingW))
+ifneq ($(HOST_OS), $(filter $(HOST_OS), Cygwin MingW))
+ifneq ($(HOST_OS), Darwin)
 ifdef TOOLNAME
   LD.Flags += $(RPATH) -Wl,'$$ORIGIN/../lib'
   ifdef EXAMPLE_TOOL
@@ -632,6 +633,7 @@ ifneq ($(DARWIN_MAJVERS),4)
   LD.Flags += $(RPATH) -Wl, at executable_path/../lib
 endif
 endif
+endif
 
 
 #----------------------------------------------------------

From stoklund at 2pi.dk  Wed Jul 28 12:01:47 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Wed, 28 Jul 2010 17:01:47 -0000
Subject: [llvm-commits] [test-suite] r109592 - in
 /test-suite/trunk/MultiSource/Applications/lua: lua.reference_output
 tests.lua
Message-ID: <20100728170147.8CA8C2A6C12C@llvm.org>

Author: stoklund
Date: Wed Jul 28 12:01:47 2010
New Revision: 109592

URL: http://llvm.org/viewvc/llvm-project?rev=109592&view=rev
Log:
Disable the env.lua test. It was printing out $USER and $PATH from the
environment, making reference outputs futile.

Modified:
    test-suite/trunk/MultiSource/Applications/lua/lua.reference_output
    test-suite/trunk/MultiSource/Applications/lua/tests.lua

Modified: test-suite/trunk/MultiSource/Applications/lua/lua.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lua/lua.reference_output?rev=109592&r1=109591&r2=109592&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lua/lua.reference_output (original)
+++ test-suite/trunk/MultiSource/Applications/lua/lua.reference_output Wed Jul 28 12:01:47 2010
@@ -1 +1 @@
-80650b40ee6b6a44115e337b5089e026
+6badfbeedd84d040f153876007407785

Modified: test-suite/trunk/MultiSource/Applications/lua/tests.lua
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lua/tests.lua?rev=109592&r1=109591&r2=109592&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lua/tests.lua (original)
+++ test-suite/trunk/MultiSource/Applications/lua/tests.lua Wed Jul 28 12:01:47 2010
@@ -1,7 +1,7 @@
 dofile("test/bisect.lua");
 dofile("test/cf.lua");
 dofile("test/echo.lua");
-dofile("test/env.lua");
+-- dofile("test/env.lua");
 dofile("test/factorial.lua");
 dofile("test/fibfor.lua");
 dofile("test/fib.lua");




From gohman at apple.com  Wed Jul 28 12:09:24 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 17:09:24 -0000
Subject: [llvm-commits] [llvm] r109593 - in
 /llvm/trunk/include/llvm/Analysis: ScalarEvolution.h
 ScalarEvolutionExpressions.h
Message-ID: <20100728170924.4968F2A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 12:09:24 2010
New Revision: 109593

URL: http://llvm.org/viewvc/llvm-project?rev=109593&view=rev
Log:
Add some extra friend declarations to fix a gcc-4.0 compile error.
This is a temporary fix, until more elaborate changes are ready.

Modified:
    llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
    llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=109593&r1=109592&r2=109593&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Wed Jul 28 12:09:24 2010
@@ -44,6 +44,7 @@
   class Loop;
   class LoopInfo;
   class Operator;
+  class SCEVUnknown;
 
   /// SCEV - This class represents an analyzed expression in the program.  These
   /// are opaque objects that the client is not allowed to do much with
@@ -175,6 +176,7 @@
 
     friend class SCEVCallbackVH;
     friend class SCEVExpander;
+    friend class SCEVUnknown;
 
     /// F - The function we are analyzing.
     ///

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h?rev=109593&r1=109592&r2=109593&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h Wed Jul 28 12:09:24 2010
@@ -522,6 +522,7 @@
   ///
   class SCEVUnknown : public SCEV {
     friend class ScalarEvolution;
+    friend class ScalarEvolution::SCEVCallbackVH;
 
     // This should be an AssertingVH, however SCEVUnknowns are allocated in a
     // BumpPtrAllocator so their destructors are never called.




From gohman at apple.com  Wed Jul 28 12:11:36 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 17:11:36 -0000
Subject: [llvm-commits] [llvm] r109594 -
	/llvm/trunk/lib/Target/TargetData.cpp
Message-ID: <20100728171136.740D92A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 12:11:36 2010
New Revision: 109594

URL: http://llvm.org/viewvc/llvm-project?rev=109594&view=rev
Log:
Do GEP offset calculations with unsigned math rather than signed math
to avoid undefined behavior on overflow, noticed by John Regehr.

Modified:
    llvm/trunk/lib/Target/TargetData.cpp

Modified: llvm/trunk/lib/Target/TargetData.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/TargetData.cpp?rev=109594&r1=109593&r2=109594&view=diff
==============================================================================
--- llvm/trunk/lib/Target/TargetData.cpp (original)
+++ llvm/trunk/lib/Target/TargetData.cpp Wed Jul 28 12:11:36 2010
@@ -625,7 +625,7 @@
 
       // Get the array index and the size of each array element.
       if (int64_t arrayIdx = cast<ConstantInt>(Indices[CurIDX])->getSExtValue())
-        Result += arrayIdx * (int64_t)getTypeAllocSize(Ty);
+        Result += (uint64_t)arrayIdx * getTypeAllocSize(Ty);
     }
   }
 




From stoklund at 2pi.dk  Wed Jul 28 12:12:27 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Wed, 28 Jul 2010 17:12:27 -0000
Subject: [llvm-commits] [test-suite] r109595 -
 /test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile
Message-ID: <20100728171227.685AA2A6C12C@llvm.org>

Author: stoklund
Date: Wed Jul 28 12:12:27 2010
New Revision: 109595

URL: http://llvm.org/viewvc/llvm-project?rev=109595&view=rev
Log:
Allow MultiSource/Benchmarks/mafft to run outside srcdir.

Modified:
    test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile

Modified: test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile?rev=109595&r1=109594&r2=109595&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile (original)
+++ test-suite/trunk/MultiSource/Benchmarks/mafft/Makefile Wed Jul 28 12:12:27 2010
@@ -8,7 +8,7 @@
 PROG = pairlocalalign
 CPPFLAGS = -DLLVM
 RUN_OPTIONS = -b 62 -g 0.100 -f 2.00 -h 0.100 -L
-STDIN_FILENAME = pyruvate_decarboxylase.fasta
+STDIN_FILENAME = $(PROJ_SRC_DIR)/pyruvate_decarboxylase.fasta
 LDFLAGS = -lm
 FP_TOLERANCE = 0.00001
 HASH_PROGRAM_OUTPUT = 1




From gohman at apple.com  Wed Jul 28 12:14:23 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 17:14:23 -0000
Subject: [llvm-commits] [llvm] r109596 -
	/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
Message-ID: <20100728171423.E27C32A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 12:14:23 2010
New Revision: 109596

URL: http://llvm.org/viewvc/llvm-project?rev=109596&view=rev
Log:
When user code intentionally dereferences null, the alignment of the
dereference is theoretically infinite. Put a cap on the computed
alignment to avoid overflow, noticed by John Regehr.

Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=109596&r1=109595&r2=109596&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Wed Jul 28 12:14:23 2010
@@ -96,12 +96,17 @@
 /// increase the alignment of the ultimate object, making this check succeed.
 unsigned InstCombiner::GetOrEnforceKnownAlignment(Value *V,
                                                   unsigned PrefAlign) {
-  unsigned BitWidth = TD ? TD->getTypeSizeInBits(V->getType()) :
-                      sizeof(PrefAlign) * CHAR_BIT;
+  assert(V->getType()->isPointerTy() &&
+         "GetOrEnforceKnownAlignment expects a pointer!");
+  unsigned BitWidth = TD ? TD->getPointerSizeInBits() : 64;
   APInt Mask = APInt::getAllOnesValue(BitWidth);
   APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
   ComputeMaskedBits(V, Mask, KnownZero, KnownOne);
   unsigned TrailZ = KnownZero.countTrailingOnes();
+
+  // LLVM doesn't support alignments larger than this currently.
+  TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));
+
   unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
 
   if (PrefAlign > Align)




From gohman at apple.com  Wed Jul 28 12:15:36 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 17:15:36 -0000
Subject: [llvm-commits] [llvm] r109597 - /llvm/trunk/lib/Target/PIC16/PIC16.h
Message-ID: <20100728171536.4FA742A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 12:15:36 2010
New Revision: 109597

URL: http://llvm.org/viewvc/llvm-project?rev=109597&view=rev
Log:
Fix this code to avoid decrementing an iterator past the beginning
of a std::vector.

Modified:
    llvm/trunk/lib/Target/PIC16/PIC16.h

Modified: llvm/trunk/lib/Target/PIC16/PIC16.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PIC16/PIC16.h?rev=109597&r1=109596&r2=109597&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PIC16/PIC16.h (original)
+++ llvm/trunk/lib/Target/PIC16/PIC16.h Wed Jul 28 12:15:36 2010
@@ -58,13 +58,10 @@
     ESNames() {}
     public:
     ~ESNames() {
-      std::vector<char*>::iterator it = stk.end();
-      it--;
-      while(stk.end() != stk.begin())
+      while (!stk.empty())
         {
-        char* p = *it;
+        char* p = stk.back();
         delete [] p;
-        it--;
         stk.pop_back();
         }
     }




From dgregor at apple.com  Wed Jul 28 12:17:46 2010
From: dgregor at apple.com (Douglas Gregor)
Date: Wed, 28 Jul 2010 17:17:46 -0000
Subject: [llvm-commits] [llvm] r109598 -
	/llvm/trunk/lib/Support/CMakeLists.txt
Message-ID: <20100728171746.358E52A6C12C@llvm.org>

Author: dgregor
Date: Wed Jul 28 12:17:46 2010
New Revision: 109598

URL: http://llvm.org/viewvc/llvm-project?rev=109598&view=rev
Log:
Unbreak my CMake build, say you'll compile for me again...

Modified:
    llvm/trunk/lib/Support/CMakeLists.txt

Modified: llvm/trunk/lib/Support/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CMakeLists.txt?rev=109598&r1=109597&r2=109598&view=diff
==============================================================================
--- llvm/trunk/lib/Support/CMakeLists.txt (original)
+++ llvm/trunk/lib/Support/CMakeLists.txt Wed Jul 28 12:17:46 2010
@@ -6,6 +6,7 @@
   circular_raw_ostream.cpp
   CommandLine.cpp
   ConstantRange.cpp
+  CrashRecoveryContext.cpp
   Debug.cpp
   DeltaAlgorithm.cpp
   DAGDeltaAlgorithm.cpp




From rafael.espindola at gmail.com  Wed Jul 28 13:12:30 2010
From: rafael.espindola at gmail.com (Rafael Espindola)
Date: Wed, 28 Jul 2010 18:12:30 -0000
Subject: [llvm-commits] [llvm] r109602 - in /llvm/trunk/tools/bugpoint:
 BugDriver.h CrashDebugger.cpp ExtractFunction.cpp Miscompilation.cpp
 OptimizerDriver.cpp
Message-ID: <20100728181230.EDD942A6C12C@llvm.org>

Author: rafael
Date: Wed Jul 28 13:12:30 2010
New Revision: 109602

URL: http://llvm.org/viewvc/llvm-project?rev=109602&view=rev
Log:
Instead of abusing swapProgramIn, just add a Module argument to
EmitProgressBitcode.

Modified:
    llvm/trunk/tools/bugpoint/BugDriver.h
    llvm/trunk/tools/bugpoint/CrashDebugger.cpp
    llvm/trunk/tools/bugpoint/ExtractFunction.cpp
    llvm/trunk/tools/bugpoint/Miscompilation.cpp
    llvm/trunk/tools/bugpoint/OptimizerDriver.cpp

Modified: llvm/trunk/tools/bugpoint/BugDriver.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/BugDriver.h?rev=109602&r1=109601&r2=109602&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/BugDriver.h (original)
+++ llvm/trunk/tools/bugpoint/BugDriver.h Wed Jul 28 13:12:30 2010
@@ -205,10 +205,11 @@
                    bool RemoveBitcode = false,
                    std::string *Error = 0);
 
-  /// EmitProgressBitcode - This function is used to output the current Program
-  /// to a file named "bugpoint-ID.bc".
+  /// EmitProgressBitcode - This function is used to output M to a file named
+  /// "bugpoint-ID.bc".
   ///
-  void EmitProgressBitcode(const std::string &ID, bool NoFlyer = false);
+  void EmitProgressBitcode(const Module *M, const std::string &ID,
+                           bool NoFlyer = false);
 
   /// deleteInstructionFromProgram - This method clones the current Program and
   /// deletes the specified instruction from the cloned module.  It then runs a
@@ -274,7 +275,7 @@
   /// writeProgramToFile - This writes the current "Program" to the named
   /// bitcode file.  If an error occurs, true is returned.
   ///
-  bool writeProgramToFile(const std::string &Filename, Module *M = 0) const;
+  bool writeProgramToFile(const std::string &Filename, const Module *M) const;
 
 private:
   /// runPasses - Just like the method above, but this just returns true or

Modified: llvm/trunk/tools/bugpoint/CrashDebugger.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/CrashDebugger.cpp?rev=109602&r1=109601&r2=109602&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/CrashDebugger.cpp (original)
+++ llvm/trunk/tools/bugpoint/CrashDebugger.cpp Wed Jul 28 13:12:30 2010
@@ -471,7 +471,7 @@
             return true;
 
           if (GVs.size() < OldSize)
-            BD.EmitProgressBitcode("reduced-global-variables");
+            BD.EmitProgressBitcode(BD.getProgram(), "reduced-global-variables");
         }
       }
     }
@@ -492,7 +492,7 @@
     ReduceCrashingFunctions(BD, TestFn).reduceList(Functions, Error);
 
     if (Functions.size() < OldSize)
-      BD.EmitProgressBitcode("reduced-function");
+      BD.EmitProgressBitcode(BD.getProgram(), "reduced-function");
   }
 
   // Attempt to delete entire basic blocks at a time to speed up
@@ -509,7 +509,7 @@
     unsigned OldSize = Blocks.size();
     ReduceCrashingBlocks(BD, TestFn).reduceList(Blocks, Error);
     if (Blocks.size() < OldSize)
-      BD.EmitProgressBitcode("reduced-blocks");
+      BD.EmitProgressBitcode(BD.getProgram(), "reduced-blocks");
   }
 
   // Attempt to delete instructions using bisection. This should help out nasty
@@ -602,7 +602,7 @@
     }
   }
 
-  BD.EmitProgressBitcode("reduced-simplified");
+  BD.EmitProgressBitcode(BD.getProgram(), "reduced-simplified");
 
   return false;
 }
@@ -628,7 +628,7 @@
          << (PassesToRun.size() == 1 ? ": " : "es: ")
          << getPassesString(PassesToRun) << '\n';
 
-  EmitProgressBitcode(ID);
+  EmitProgressBitcode(Program, ID);
 
   bool Success = DebugACrash(*this, TestForOptimizerCrash, Error);
   assert(Error.empty());

Modified: llvm/trunk/tools/bugpoint/ExtractFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ExtractFunction.cpp?rev=109602&r1=109601&r2=109602&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ExtractFunction.cpp (original)
+++ llvm/trunk/tools/bugpoint/ExtractFunction.cpp Wed Jul 28 13:12:30 2010
@@ -143,11 +143,9 @@
 
   Module *NewM = runPassesOn(M, LoopExtractPasses);
   if (NewM == 0) {
-    Module *Old = swapProgramIn(M);
     outs() << "*** Loop extraction failed: ";
-    EmitProgressBitcode("loopextraction", true);
+    EmitProgressBitcode(M, "loopextraction", true);
     outs() << "*** Sorry. :(  Please report a bug!\n";
-    swapProgramIn(Old);
     return 0;
   }
 
@@ -327,9 +325,7 @@
   if (uniqueFilename.createTemporaryFileOnDisk(true, &ErrMsg)) {
     outs() << "*** Basic Block extraction failed!\n";
     errs() << "Error creating temporary file: " << ErrMsg << "\n";
-    M = swapProgramIn(M);
-    EmitProgressBitcode("basicblockextractfail", true);
-    swapProgramIn(M);
+    EmitProgressBitcode(M, "basicblockextractfail", true);
     return 0;
   }
   sys::RemoveFileOnSignal(uniqueFilename);
@@ -340,9 +336,7 @@
     outs() << "*** Basic Block extraction failed!\n";
     errs() << "Error writing list of blocks to not extract: " << ErrorInfo
            << "\n";
-    M = swapProgramIn(M);
-    EmitProgressBitcode("basicblockextractfail", true);
-    swapProgramIn(M);
+    EmitProgressBitcode(M, "basicblockextractfail", true);
     return 0;
   }
   for (std::vector<BasicBlock*>::const_iterator I = BBs.begin(), E = BBs.end();
@@ -368,9 +362,7 @@
 
   if (Ret == 0) {
     outs() << "*** Basic Block extraction failed, please report a bug!\n";
-    M = swapProgramIn(M);
-    EmitProgressBitcode("basicblockextractfail", true);
-    swapProgramIn(M);
+    EmitProgressBitcode(M, "basicblockextractfail", true);
   }
   return Ret;
 }

Modified: llvm/trunk/tools/bugpoint/Miscompilation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/Miscompilation.cpp?rev=109602&r1=109601&r2=109602&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/Miscompilation.cpp (original)
+++ llvm/trunk/tools/bugpoint/Miscompilation.cpp Wed Jul 28 13:12:30 2010
@@ -71,7 +71,7 @@
     errs() << " Error running this sequence of passes"
            << " on the input program!\n";
     BD.setPassesToRun(Suffix);
-    BD.EmitProgressBitcode("pass-error",  false);
+    BD.EmitProgressBitcode(BD.getProgram(), "pass-error",  false);
     exit(BD.debugOptimizerCrash());
   }
   
@@ -108,7 +108,7 @@
     errs() << " Error running this sequence of passes"
            << " on the input program!\n";
     BD.setPassesToRun(Prefix);
-    BD.EmitProgressBitcode("pass-error",  false);
+    BD.EmitProgressBitcode(BD.getProgram(), "pass-error",  false);
     exit(BD.debugOptimizerCrash());
   }
 
@@ -148,7 +148,7 @@
     errs() << " Error running this sequence of passes"
            << " on the input program!\n";
     BD.setPassesToRun(Suffix);
-    BD.EmitProgressBitcode("pass-error",  false);
+    BD.EmitProgressBitcode(BD.getProgram(), "pass-error",  false);
     exit(BD.debugOptimizerCrash());
   }
 
@@ -699,7 +699,7 @@
   outs() << "\n*** Found miscompiling pass"
          << (getPassesToRun().size() == 1 ? "" : "es") << ": "
          << getPassesString(getPassesToRun()) << '\n';
-  EmitProgressBitcode("passinput");
+  EmitProgressBitcode(Program, "passinput");
 
   std::vector<Function *> MiscompiledFunctions = 
     DebugAMiscompilation(*this, TestOptimizer, *Error);
@@ -715,14 +715,12 @@
                                                  VMap);
 
   outs() << "  Non-optimized portion: ";
-  ToNotOptimize = swapProgramIn(ToNotOptimize);
-  EmitProgressBitcode("tonotoptimize", true);
-  setNewProgram(ToNotOptimize);   // Delete hacked module.
+  EmitProgressBitcode(ToNotOptimize, "tonotoptimize", true);
+  delete ToNotOptimize;  // Delete hacked module.
 
   outs() << "  Portion that is input to optimizer: ";
-  ToOptimize = swapProgramIn(ToOptimize);
-  EmitProgressBitcode("tooptimize");
-  setNewProgram(ToOptimize);      // Delete hacked module.
+  EmitProgressBitcode(ToOptimize, "tooptimize");
+  delete ToOptimize;      // Delete hacked module.
 
   return;
 }

Modified: llvm/trunk/tools/bugpoint/OptimizerDriver.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/OptimizerDriver.cpp?rev=109602&r1=109601&r2=109602&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/OptimizerDriver.cpp (original)
+++ llvm/trunk/tools/bugpoint/OptimizerDriver.cpp Wed Jul 28 13:12:30 2010
@@ -51,13 +51,13 @@
 /// file.  If an error occurs, true is returned.
 ///
 bool BugDriver::writeProgramToFile(const std::string &Filename,
-                                   Module *M) const {
+                                   const Module *M) const {
   std::string ErrInfo;
   raw_fd_ostream Out(Filename.c_str(), ErrInfo,
                      raw_fd_ostream::F_Binary);
   if (!ErrInfo.empty()) return true;
   
-  WriteBitcodeToFile(M ? M : Program, Out);
+  WriteBitcodeToFile(M, Out);
   return false;
 }
 
@@ -65,12 +65,13 @@
 /// EmitProgressBitcode - This function is used to output the current Program
 /// to a file named "bugpoint-ID.bc".
 ///
-void BugDriver::EmitProgressBitcode(const std::string &ID, bool NoFlyer) {
+void BugDriver::EmitProgressBitcode(const Module *M,
+                                    const std::string &ID, bool NoFlyer) {
   // Output the input to the current pass to a bitcode file, emit a message
   // telling the user how to reproduce it: opt -foo blah.bc
   //
   std::string Filename = OutputPrefix + "-" + ID + ".bc";
-  if (writeProgramToFile(Filename)) {
+  if (writeProgramToFile(Filename, M)) {
     errs() <<  "Error opening file '" << Filename << "' for writing!\n";
     return;
   }
@@ -246,7 +247,7 @@
       errs() << " Error running this sequence of passes"
              << " on the input program!\n";
       delete OldProgram;
-      EmitProgressBitcode("pass-error",  false);
+      EmitProgressBitcode(Program, "pass-error",  false);
       exit(debugOptimizerCrash());
     }
     swapProgramIn(OldProgram);




From tonic at nondot.org  Wed Jul 28 13:18:45 2010
From: tonic at nondot.org (Tanya Lattner)
Date: Wed, 28 Jul 2010 18:18:45 -0000
Subject: [llvm-commits] [www] r109603 - /www/trunk/index.html
Message-ID: <20100728181845.3761A2A6C12C@llvm.org>

Author: tbrethou
Date: Wed Jul 28 13:18:45 2010
New Revision: 109603

URL: http://llvm.org/viewvc/llvm-project?rev=109603&view=rev
Log:
Test commit

Modified:
    www/trunk/index.html

Modified: www/trunk/index.html
URL: http://llvm.org/viewvc/llvm-project/www/trunk/index.html?rev=109603&r1=109602&r2=109603&view=diff
==============================================================================
--- www/trunk/index.html (original)
+++ www/trunk/index.html Wed Jul 28 13:18:45 2010
@@ -131,7 +131,7 @@
 
 <div class="www_text">
 <p>LLVM 2.8 release:</p>
-TBD
+Information coming soon.
 </div>
 
 <br>




From bob.wilson at apple.com  Wed Jul 28 13:21:10 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 28 Jul 2010 18:21:10 -0000
Subject: [llvm-commits] [llvm] r109605 -
	/llvm/trunk/utils/TableGen/NeonEmitter.cpp
Message-ID: <20100728182110.2CA682A6C12C@llvm.org>

Author: bwilson
Date: Wed Jul 28 13:21:10 2010
New Revision: 109605

URL: http://llvm.org/viewvc/llvm-project?rev=109605&view=rev
Log:
Fix a warning from gcc-4.0 (from the ppc buildbot).

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

Modified: llvm/trunk/utils/TableGen/NeonEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/NeonEmitter.cpp?rev=109605&r1=109604&r2=109605&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/NeonEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/NeonEmitter.cpp Wed Jul 28 13:21:10 2010
@@ -1016,6 +1016,8 @@
       throw "unhandled type!";
       break;
   }
+  assert(0 && "unreachable");
+  return 0;
 }
 
 /// runHeader - Emit a file with sections defining:




From stoklund at 2pi.dk  Wed Jul 28 13:35:47 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Wed, 28 Jul 2010 18:35:47 -0000
Subject: [llvm-commits] [llvm] r109608 -
	/llvm/trunk/lib/CodeGen/MachineInstr.cpp
Message-ID: <20100728183547.1E75C2A6C12C@llvm.org>

Author: stoklund
Date: Wed Jul 28 13:35:46 2010
New Revision: 109608

URL: http://llvm.org/viewvc/llvm-project?rev=109608&view=rev
Log:
Print out the regclass of any virtual registers used by a machine instruction.

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

Modified: llvm/trunk/lib/CodeGen/MachineInstr.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineInstr.cpp?rev=109608&r1=109607&r2=109608&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineInstr.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineInstr.cpp Wed Jul 28 13:35:46 2010
@@ -1236,12 +1236,18 @@
 void MachineInstr::print(raw_ostream &OS, const TargetMachine *TM) const {
   // We can be a bit tidier if we know the TargetMachine and/or MachineFunction.
   const MachineFunction *MF = 0;
+  const MachineRegisterInfo *MRI = 0;
   if (const MachineBasicBlock *MBB = getParent()) {
     MF = MBB->getParent();
     if (!TM && MF)
       TM = &MF->getTarget();
+    if (MF)
+      MRI = &MF->getRegInfo();
   }
 
+  // Save a list of virtual registers.
+  SmallVector<unsigned, 8> VirtRegs;
+
   // Print explicitly defined operands on the left of an assignment syntax.
   unsigned StartOp = 0, e = getNumOperands();
   for (; StartOp < e && getOperand(StartOp).isReg() &&
@@ -1250,6 +1256,9 @@
        ++StartOp) {
     if (StartOp != 0) OS << ", ";
     getOperand(StartOp).print(OS, TM);
+    unsigned Reg = getOperand(StartOp).getReg();
+    if (Reg && TargetRegisterInfo::isVirtualRegister(Reg))
+      VirtRegs.push_back(Reg);
   }
 
   if (StartOp != 0)
@@ -1264,6 +1273,10 @@
   for (unsigned i = StartOp, e = getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = getOperand(i);
 
+    if (MO.isReg() && MO.getReg() &&
+        TargetRegisterInfo::isVirtualRegister(MO.getReg()))
+      VirtRegs.push_back(MO.getReg());
+
     // Omit call-clobbered registers which aren't used anywhere. This makes
     // call instructions much less noisy on targets where calls clobber lots
     // of registers. Don't rely on MO.isDead() because we may be called before
@@ -1330,6 +1343,24 @@
     }
   }
 
+  // Print the regclass of any virtual registers encountered.
+  if (MRI && !VirtRegs.empty()) {
+    if (!HaveSemi) OS << ";"; HaveSemi = true;
+    for (unsigned i = 0; i != VirtRegs.size(); ++i) {
+      const TargetRegisterClass *RC = MRI->getRegClass(VirtRegs[i]);
+      OS << " " << RC->getName() << ":%reg" << VirtRegs[i];
+      for (unsigned j = i+1; j != VirtRegs.size();) {
+        if (MRI->getRegClass(VirtRegs[j]) != RC) {
+          ++j;
+          continue;
+        }
+        if (VirtRegs[i] != VirtRegs[j])
+          OS << "," << VirtRegs[j];
+        VirtRegs.erase(VirtRegs.begin()+j);
+      }
+    }
+  }
+
   if (!debugLoc.isUnknown() && MF) {
     if (!HaveSemi) OS << ";";
     OS << " dbg:";




From fjahanian at apple.com  Wed Jul 28 14:08:54 2010
From: fjahanian at apple.com (Fariborz Jahanian)
Date: Wed, 28 Jul 2010 19:08:54 -0000
Subject: [llvm-commits] [test-suite] r109621 -
	/test-suite/trunk/SingleSource/UnitTests/blockstret.c
Message-ID: <20100728190854.6E14D2A6C12C@llvm.org>

Author: fjahanian
Date: Wed Jul 28 14:08:54 2010
New Revision: 109621

URL: http://llvm.org/viewvc/llvm-project?rev=109621&view=rev
Log:
Test for clang's radar 8241648.


Added:
    test-suite/trunk/SingleSource/UnitTests/blockstret.c

Added: test-suite/trunk/SingleSource/UnitTests/blockstret.c
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/SingleSource/UnitTests/blockstret.c?rev=109621&view=auto
==============================================================================
--- test-suite/trunk/SingleSource/UnitTests/blockstret.c (added)
+++ test-suite/trunk/SingleSource/UnitTests/blockstret.c Wed Jul 28 14:08:54 2010
@@ -0,0 +1,101 @@
+#if defined(__BLOCKS__) && defined(__clang__)
+// rdar://8241648
+void *malloc(__SIZE_TYPE__ size);
+extern void abort (void);
+
+typedef struct bigbig {
+   int array[512];
+   char more[32];
+} BigStruct_t;
+
+BigStruct_t (^global)(void) = ^{ return *(BigStruct_t *)malloc(sizeof(struct bigbig)); };
+
+const char * getBlockSignature(void *);
+ 
+BigStruct_t foo(int param) {
+   BigStruct_t x;
+   BigStruct_t (^f)(int) = ^(int param) {
+     BigStruct_t *result = malloc(sizeof(BigStruct_t));
+     result->array[23] = param;
+     return *result;
+   };
+   getBlockSignature(f);
+   return x;
+}
+
+enum {
+    BLOCK_HAS_COPY_DISPOSE =  (1 << 25),
+    BLOCK_HAS_CXX_OBJ =       (1 << 26),
+    BLOCK_IS_GLOBAL =         (1 << 28),
+    BLOCK_USE_STRET =    (1 << 29),
+    BLOCK_HAS_OBJC_TYPE  =    (1 << 30)
+};
+
+struct block_descriptor_big {
+    unsigned long int reserved;
+    unsigned long int size;
+    void (*copy)(void *dst, void *src); // conditional on BLOCK_HAS_COPY_DISPOSE
+    void (*dispose)(void *);            // conditional on BLOCK_HAS_COPY_DISPOSE
+    const char *signature;                  // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+struct block_descriptor_small {
+    unsigned long int reserved;
+    unsigned long int size;
+    const char *signature;              // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+
+struct block_layout_abi { // can't change
+  void *isa;
+  int flags;
+  int reserved; 
+  void (*invoke)(void *, ...);
+  struct block_descriptor_big *descriptor;
+};
+
+const char *getBlockSignature(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   if ((layout->flags & BLOCK_HAS_OBJC_TYPE) != BLOCK_HAS_OBJC_TYPE) return 0;
+   if (layout->flags & BLOCK_HAS_COPY_DISPOSE) 
+      return layout->descriptor->signature;
+   else
+      return ((struct block_descriptor_small *)layout->descriptor)->signature;
+}
+
+int usesStruct(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   int want = BLOCK_HAS_OBJC_TYPE | BLOCK_USE_STRET;
+   return (layout->flags & want) == want;
+}
+#endif
+
+int main(int argc, char *argv[]) {
+#if defined(__BLOCKS__) && defined(__clang__)
+   if (!usesStruct(global))
+     abort();
+   BigStruct_t x;
+   BigStruct_t (^local)(int) = ^(int param) {
+     BigStruct_t *result = (BigStruct_t *)malloc(sizeof(BigStruct_t));
+     result->array[23] = argc;
+     return *result;
+   };
+   if (!usesStruct(global))
+     abort();
+   if (!usesStruct(local))
+     abort();
+   if (usesStruct(^void(int x){ }))
+     abort();
+#endif
+   return 0;
+}
+
+/*
+desired global flags: 1879048192
+desired stack flags: 1610612736
+should be non-zero: 1
+should be non-zero: 1
+should be non-zero: 1
+should be zero: 0
+
+*/




From gohman at apple.com  Wed Jul 28 15:12:04 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 20:12:04 -0000
Subject: [llvm-commits] [llvm] r109643 - in /llvm/trunk:
 include/llvm/Instructions.h lib/AsmParser/LLParser.cpp
 lib/Transforms/InstCombine/InstCombineCalls.cpp lib/VMCore/Instructions.cpp
 test/Assembler/align-inst-alloca.ll test/Assembler/align-inst-load.ll
 test/Assembler/align-inst-store.ll test/Assembler/align-inst.ll
Message-ID: <20100728201204.EB9B22A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 15:12:04 2010
New Revision: 109643

URL: http://llvm.org/viewvc/llvm-project?rev=109643&view=rev
Log:
Define a maximum supported alignment value for load, store, and
alloca instructions (constrained by their internal encoding),
and add error checking for it. Fix an instcombine bug which
generated huge alignment values (null is infinitely aligned).
This fixes undefined behavior noticed by John Regehr.

Added:
    llvm/trunk/test/Assembler/align-inst-alloca.ll
    llvm/trunk/test/Assembler/align-inst-load.ll
    llvm/trunk/test/Assembler/align-inst-store.ll
    llvm/trunk/test/Assembler/align-inst.ll
Modified:
    llvm/trunk/include/llvm/Instructions.h
    llvm/trunk/lib/AsmParser/LLParser.cpp
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
    llvm/trunk/lib/VMCore/Instructions.cpp

Modified: llvm/trunk/include/llvm/Instructions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Instructions.h?rev=109643&r1=109642&r2=109643&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Instructions.h (original)
+++ llvm/trunk/include/llvm/Instructions.h Wed Jul 28 15:12:04 2010
@@ -31,6 +31,10 @@
 class LLVMContext;
 class DominatorTree;
 
+/// MaximumAlignment - This is the greatest alignment value supported by
+/// load, store, and alloca instructions.
+static const unsigned MaximumAlignment = 1u << 29;
+
 //===----------------------------------------------------------------------===//
 //                                AllocaInst Class
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/lib/AsmParser/LLParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/AsmParser/LLParser.cpp?rev=109643&r1=109642&r2=109643&view=diff
==============================================================================
--- llvm/trunk/lib/AsmParser/LLParser.cpp (original)
+++ llvm/trunk/lib/AsmParser/LLParser.cpp Wed Jul 28 15:12:04 2010
@@ -1154,6 +1154,8 @@
   if (ParseUInt32(Alignment)) return true;
   if (!isPowerOf2_32(Alignment))
     return Error(AlignLoc, "alignment is not a power of two");
+  if (Alignment > MaximumAlignment)
+    return Error(AlignLoc, "huge alignments are not supported yet");
   return false;
 }
 
@@ -1176,6 +1178,7 @@
     if (Lex.getKind() != lltok::kw_align)
       return Error(Lex.getLoc(), "expected metadata or 'align'");
     
+    LocTy AlignLoc = Lex.getLoc();
     if (ParseOptionalAlignment(Alignment)) return true;
   }
 

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=109643&r1=109642&r2=109643&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Wed Jul 28 15:12:04 2010
@@ -104,11 +104,15 @@
   ComputeMaskedBits(V, Mask, KnownZero, KnownOne);
   unsigned TrailZ = KnownZero.countTrailingOnes();
 
-  // LLVM doesn't support alignments larger than this currently.
+  // Avoid trouble with rediculously large TrailZ values, such as
+  // those computed from a null pointer.
   TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));
 
   unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
 
+  // LLVM doesn't support alignments larger than this currently.
+  Align = std::min(Align, MaximumAlignment);
+
   if (PrefAlign > Align)
     Align = EnforceKnownAlignment(V, Align, PrefAlign);
   

Modified: llvm/trunk/lib/VMCore/Instructions.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Instructions.cpp?rev=109643&r1=109642&r2=109643&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Instructions.cpp (original)
+++ llvm/trunk/lib/VMCore/Instructions.cpp Wed Jul 28 15:12:04 2010
@@ -891,6 +891,8 @@
 
 void AllocaInst::setAlignment(unsigned Align) {
   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
+  assert(Align <= MaximumAlignment &&
+         "Alignment is greater than MaximumAlignment!");
   setInstructionSubclassData(Log2_32(Align) + 1);
   assert(getAlignment() == Align && "Alignment representation error!");
 }
@@ -1026,8 +1028,11 @@
 
 void LoadInst::setAlignment(unsigned Align) {
   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
+  assert(Align <= MaximumAlignment &&
+         "Alignment is greater than MaximumAlignment!");
   setInstructionSubclassData((getSubclassDataFromInstruction() & 1) |
                              ((Log2_32(Align)+1)<<1));
+  assert(getAlignment() == Align && "Alignment representation error!");
 }
 
 //===----------------------------------------------------------------------===//
@@ -1122,8 +1127,11 @@
 
 void StoreInst::setAlignment(unsigned Align) {
   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
+  assert(Align <= MaximumAlignment &&
+         "Alignment is greater than MaximumAlignment!");
   setInstructionSubclassData((getSubclassDataFromInstruction() & 1) |
                              ((Log2_32(Align)+1) << 1));
+  assert(getAlignment() == Align && "Alignment representation error!");
 }
 
 //===----------------------------------------------------------------------===//

Added: llvm/trunk/test/Assembler/align-inst-alloca.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-alloca.ll?rev=109643&view=auto
==============================================================================
--- llvm/trunk/test/Assembler/align-inst-alloca.ll (added)
+++ llvm/trunk/test/Assembler/align-inst-alloca.ll Wed Jul 28 15:12:04 2010
@@ -0,0 +1,6 @@
+; RUN: not llvm-as %s -o /dev/null 2>/dev/null
+
+define void @foo() {
+  %p = alloca i1, align 1073741824
+  ret void
+}

Added: llvm/trunk/test/Assembler/align-inst-load.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-load.ll?rev=109643&view=auto
==============================================================================
--- llvm/trunk/test/Assembler/align-inst-load.ll (added)
+++ llvm/trunk/test/Assembler/align-inst-load.ll Wed Jul 28 15:12:04 2010
@@ -0,0 +1,6 @@
+; RUN: not llvm-as %s -o /dev/null 2>/dev/null
+
+define void @foo() {
+  load i1* %p, align 1073741824
+  ret void
+}

Added: llvm/trunk/test/Assembler/align-inst-store.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-store.ll?rev=109643&view=auto
==============================================================================
--- llvm/trunk/test/Assembler/align-inst-store.ll (added)
+++ llvm/trunk/test/Assembler/align-inst-store.ll Wed Jul 28 15:12:04 2010
@@ -0,0 +1,6 @@
+; RUN: not llvm-as %s -o /dev/null 2>/dev/null
+
+define void @foo() {
+  store i1 false, i1* %p, align 1073741824
+  ret void
+}

Added: llvm/trunk/test/Assembler/align-inst.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst.ll?rev=109643&view=auto
==============================================================================
--- llvm/trunk/test/Assembler/align-inst.ll (added)
+++ llvm/trunk/test/Assembler/align-inst.ll Wed Jul 28 15:12:04 2010
@@ -0,0 +1,8 @@
+; RUN: llvm-as %s -o /dev/null
+
+define void @foo() {
+  %p = alloca i1, align 536870912
+  load i1* %p, align 536870912
+  store i1 false, i1* %p, align 536870912
+  ret void
+}




From criswell at uiuc.edu  Wed Jul 28 15:14:37 2010
From: criswell at uiuc.edu (John Criswell)
Date: Wed, 28 Jul 2010 20:14:37 -0000
Subject: [llvm-commits] [poolalloc] r109644 - in /poolalloc/trunk:
 include/poolalloc/PoolAllocate.h lib/DSA/TopDownClosure.cpp
 lib/PoolAllocate/PASimple.cpp
Message-ID: <20100728201437.445082A6C12C@llvm.org>

Author: criswell
Date: Wed Jul 28 15:14:37 2010
New Revision: 109644

URL: http://llvm.org/viewvc/llvm-project?rev=109644&view=rev
Log:
TopDownClosure.cpp:
Modified code so that it doesn't remove dead global nodes from the local graph.
This is important because it allows SAFECode to correctly find alignment of
the results of load instructions when the dereferenced pointer is a global.

include/poolalloc/PoolAllocate.h:
lib/PoolAllocate/PASimple.cpp:
Be more like DSA; use a separate globals graph.
Renamed function to be more descriptive.
Added comments.


Modified:
    poolalloc/trunk/include/poolalloc/PoolAllocate.h
    poolalloc/trunk/lib/DSA/TopDownClosure.cpp
    poolalloc/trunk/lib/PoolAllocate/PASimple.cpp

Modified: poolalloc/trunk/include/poolalloc/PoolAllocate.h
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/include/poolalloc/PoolAllocate.h?rev=109644&r1=109643&r2=109644&view=diff
==============================================================================
--- poolalloc/trunk/include/poolalloc/PoolAllocate.h (original)
+++ poolalloc/trunk/include/poolalloc/PoolAllocate.h Wed Jul 28 15:14:37 2010
@@ -30,6 +30,7 @@
 #include "llvm/Support/CommandLine.h"
 
 #include "dsa/DataStructure.h"
+#include "dsa/DSGraph.h"
 #include "poolalloc/Config/config.h"
 
 #include <utility>
@@ -526,7 +527,7 @@
   }
 
   virtual DSGraph* getGlobalsGraph () const {
-    return CombinedDSGraph;
+    return CombinedDSGraph->getGlobalsGraph();
   }
 
   virtual Value * getGlobalPool (const DSNode * Node) {

Modified: poolalloc/trunk/lib/DSA/TopDownClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/TopDownClosure.cpp?rev=109644&r1=109643&r2=109644&view=diff
==============================================================================
--- poolalloc/trunk/lib/DSA/TopDownClosure.cpp (original)
+++ poolalloc/trunk/lib/DSA/TopDownClosure.cpp Wed Jul 28 15:14:37 2010
@@ -276,8 +276,24 @@
   Flags |= DSGraph::IgnoreGlobals | DSGraph::MarkVAStart;
   DSG->markIncompleteNodes(Flags);
 
+  //
   // Delete dead nodes.  Treat globals that are unreachable as dead also.
+  //
+  // FIXME:
+  //  Do not delete unreachable globals as the comment describes.  For its
+  //  alignment checks on the results of load instructions, SAFECode must be
+  //  able to find the DSNode of both the result of the load as well as the
+  //  pointer dereferenced by the load.  If we remove unreachable globals, then
+  //  if the dereferenced pointer is a global, its DSNode will not reachable
+  //  from the local graph's scalar map, and chaos ensues.
+  //
+  //  So, for now, just remove dead nodes but leave the globals alone.
+  //
+#if 0
   DSG->removeDeadNodes(DSGraph::RemoveUnreachableGlobals);
+#else
+  DSG->removeDeadNodes(0);
+#endif
 
   // We are done with computing the current TD Graph!  Finally, before we can
   // finish processing this function, we figure out which functions it calls and

Modified: poolalloc/trunk/lib/PoolAllocate/PASimple.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PASimple.cpp?rev=109644&r1=109643&r2=109644&view=diff
==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PASimple.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PASimple.cpp Wed Jul 28 15:14:37 2010
@@ -84,10 +84,21 @@
   AU.setPreservesAll();
 }
 
+//
+// Function: FoldNodesInDSGraph()
+//
+// Description:
+//  This function will take the specified DSGraph and fold all DSNodes within
+//  it that are marked with the heap flag.
+//
 static void
-MergeNodesInDSGraph (DSGraph & Graph) {
+FoldNodesInDSGraph (DSGraph & Graph) {
+  // Worklist of heap nodes to process
   std::vector<DSNodeHandle> HeapNodes;
 
+  //
+  // Go find all of the heap nodes.
+  //
   DSGraph::node_iterator i;
   DSGraph::node_iterator e = Graph.node_end();
   for (i = Graph.node_begin(); i != e; ++i) {
@@ -96,6 +107,9 @@
       HeapNodes.push_back (DSNodeHandle(Node));
   }
 
+  //
+  // Fold all of the heap nodes; this makes them type-unknown.
+  //
   for (unsigned i = 0; i < HeapNodes.size(); ++i)
     HeapNodes[i].getNode()->foldNodeCompletely();
   return;
@@ -124,17 +138,27 @@
   AddPoolPrototypes(&M);
 
   //
-  // Merge all of the DSNodes in the DSGraphs.
+  // Create a single DSGraph which contains all of the information found in all
+  // the DSGraphs we got from DSA.  We do this because we're going to start
+  // making modifications to the points-to results.
   //
   GlobalECs = Graphs->getGlobalECs();
-  CombinedDSGraph = new DSGraph (GlobalECs, TD, Graphs->getTypeSS(), Graphs->getGlobalsGraph());
-  //CombinedDSGraph.cloneInto (getGlobalsGraph());
+  CombinedDSGraph = new DSGraph (GlobalECs,
+                                 TD,
+                                 Graphs->getTypeSS(),
+                                 Graphs->getGlobalsGraph());
   for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
     if (Graphs->hasDSGraph (*I))
       CombinedDSGraph->cloneInto (Graphs->getDSGraph(*I));
   }
-  CombinedDSGraph->cloneInto (Graphs->getGlobalsGraph());
-  MergeNodesInDSGraph (*CombinedDSGraph);
+
+  //
+  // Now fold all of the heap nodes in our DSGraph (i.e., make them
+  // type-unknown).  We do this because heap nodes may change type if we
+  // consider the effects of dangling pointers.
+  //
+  FoldNodesInDSGraph (*CombinedDSGraph);
+  FoldNodesInDSGraph (*(CombinedDSGraph->getGlobalsGraph()));
 
   //
   // Create the global pool.
@@ -177,10 +201,9 @@
   FunctionInfo.insert (std::make_pair(&F, FInfo));
 
   //
-  // Get the DSGraph for this function.
+  // Scan through all instructions in the function and modify call sites as
+  // necessary.
   //
-  DSGraph* ECG = Graphs->getDSGraph(F);
-
   for (Function::iterator i = F.begin(), e = F.end(); i != e; ++i) {
     for (BasicBlock::iterator ii = i->begin(), ee = i->end(); ii != ee; ++ii) {
       if (CallInst * CI = dyn_cast<CallInst>(ii)) {
@@ -201,7 +224,7 @@
         //
         if (CF && (CF->isDeclaration()) && (CF->getName() == "malloc")) {
           // Associate the global pool decriptor with the DSNode
-          DSNode * Node = ECG->getNodeForValue(CI).getNode();
+          DSNode * Node = CombinedDSGraph->getNodeForValue(CI).getNode();
           FInfo.PoolDescriptors.insert(std::make_pair(Node,TheGlobalPool));
 
           // Mark the call to malloc as an instruction to delete
@@ -248,7 +271,7 @@
           CI->replaceAllUsesWith(Casted);
         } else if (CF && (CF->isDeclaration()) && (CF->getName() == "realloc")) {
           // Associate the global pool decriptor with the DSNode
-          DSNode * Node = ECG->getNodeForValue(CI).getNode();
+          DSNode * Node = CombinedDSGraph->getNodeForValue(CI).getNode();
           FInfo.PoolDescriptors.insert(std::make_pair(Node,TheGlobalPool));
 
           // Mark the realloc as an instruction to delete
@@ -294,7 +317,7 @@
           CI->replaceAllUsesWith(Casted);
         } else if (CF && (CF->isDeclaration()) && (CF->getName() == "calloc")) {
           // Associate the global pool decriptor with the DSNode
-          DSNode * Node = ECG->getNodeForValue(CI).getNode();
+          DSNode * Node = CombinedDSGraph->getNodeForValue(CI).getNode();
           FInfo.PoolDescriptors.insert(std::make_pair(Node,TheGlobalPool));
 
           // Mark the realloc as an instruction to delete
@@ -341,7 +364,7 @@
           CI->replaceAllUsesWith(Casted);
         } else if (CF && (CF->isDeclaration()) && (CF->getName() == "strdup")) {
           // Associate the global pool decriptor with the DSNode
-          DSNode * Node = ECG->getNodeForValue(CI).getNode();
+          DSNode * Node = CombinedDSGraph->getNodeForValue(CI).getNode();
           FInfo.PoolDescriptors.insert(std::make_pair(Node,TheGlobalPool));
 
           // Mark the realloc as an instruction to delete




From daniel at zuster.org  Wed Jul 28 15:19:54 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 28 Jul 2010 13:19:54 -0700
Subject: [llvm-commits] [llvm] r109586 - in /llvm/trunk:
	include/llvm/Support/CrashRecoveryContext.h
	lib/Support/CrashRecoveryContext.cpp
In-Reply-To: <C02B93B3-9E00-44C9-8851-E3444DCF1E58@apple.com>
References: <20100728154020.9AA472A6C12C@llvm.org>
	<C02B93B3-9E00-44C9-8851-E3444DCF1E58@apple.com>
Message-ID: <AANLkTikwNYVyGnTbR_rC1Wt6ah4W3bkymVyLNsfY_LNd@mail.gmail.com>

On Wed, Jul 28, 2010 at 9:12 AM, Chris Lattner <clattner at apple.com> wrote:
>
> On Jul 28, 2010, at 8:40 AM, Daniel Dunbar wrote:
>
>> Author: ddunbar
>> Date: Wed Jul 28 10:40:20 2010
>> New Revision: 109586
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=109586&view=rev
>> Log:
>> Support: Add CrashRecoveryContext helper object.
>> - Designed as a simple wrapper to allow clients to attempt to catch crashes
>> ? (memory errors, assertion violations, etc.) and do some kind of recovery.
>>
>> - Currently doesn't actually attempt to catch crashes.
>
> Looks like a nice interface.

I hope so -- assuming it has more than 0 working implementations. :)

>> + ?/// \return True if the function completed successfully, and false if the
>> + ?/// function crashed (or HandleCrash was called explicitly). Clients should
>> + ?/// make as little assumptions as possible about the program state when
>> + ?/// RunSafely has returned false. Clients can use getBacktrace() to retrieve
>> + ?/// the backtrace of the crash on failures.
>> + ?bool RunSafely(void (*Fn)(void*), void *UserData);
>
> It would be nice to add an #ifdef __BLOCKS__ version of this that took a block to execute.

Yeah, this occurred to me as well. I'll try to get around to it.

 - Daniel

> -Chris
>
>



From echristo at apple.com  Wed Jul 28 15:26:34 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 20:26:34 -0000
Subject: [llvm-commits] [llvm] r109646 - /llvm/trunk/autoconf/configure.ac
Message-ID: <20100728202634.A8DE62A6C12C@llvm.org>

Author: echristo
Date: Wed Jul 28 15:26:34 2010
New Revision: 109646

URL: http://llvm.org/viewvc/llvm-project?rev=109646&view=rev
Log:
Use a C++ compiler for the atomic builtin check since we'll
be using a C++ compiler to build.

Patch by T?r?k Edwin!

Modified:
    llvm/trunk/autoconf/configure.ac

Modified: llvm/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/autoconf/configure.ac?rev=109646&r1=109645&r2=109646&view=diff
==============================================================================
--- llvm/trunk/autoconf/configure.ac (original)
+++ llvm/trunk/autoconf/configure.ac Wed Jul 28 15:26:34 2010
@@ -1355,6 +1355,9 @@
 
 dnl atomic builtins are required for threading support.
 AC_MSG_CHECKING(for GCC atomic builtins)
+dnl Since we'll be using these atomic builtins in C++ files we should test
+dnl the C++ compiler.
+AC_LANG_PUSH([C++])
 AC_LINK_IFELSE(
   AC_LANG_SOURCE(
     [[int main() {
@@ -1366,6 +1369,7 @@
         return 0;
       }
     ]]),
+  AC_LANG_POP([C++])
   AC_MSG_RESULT(yes)
   AC_DEFINE(LLVM_MULTITHREADED, 1, Build multithreading support into LLVM),
   AC_MSG_RESULT(no)




From echristo at apple.com  Wed Jul 28 15:26:44 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 20:26:44 -0000
Subject: [llvm-commits] [llvm] r109647 - /llvm/trunk/configure
Message-ID: <20100728202644.2A3342A6C12C@llvm.org>

Author: echristo
Date: Wed Jul 28 15:26:43 2010
New Revision: 109647

URL: http://llvm.org/viewvc/llvm-project?rev=109647&view=rev
Log:
Regenerate.

Modified:
    llvm/trunk/configure

Modified: llvm/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/configure?rev=109647&r1=109646&r2=109647&view=diff
==============================================================================
--- llvm/trunk/configure (original)
+++ llvm/trunk/configure Wed Jul 28 15:26:43 2010
@@ -20165,6 +20165,12 @@
 
 { echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
 echo $ECHO_N "checking for GCC atomic builtins... $ECHO_C" >&6; }
+ac_ext=cpp
+ac_cpp='$CXXCPP $CPPFLAGS'
+ac_compile='$CXX -c $CXXFLAGS $CPPFLAGS conftest.$ac_ext >&5'
+ac_link='$CXX -o conftest$ac_exeext $CXXFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'
+ac_compiler_gnu=$ac_cv_cxx_compiler_gnu
+
 cat >conftest.$ac_ext <<_ACEOF
 /* confdefs.h.  */
 _ACEOF
@@ -20215,6 +20221,12 @@
   ac_status=$?
   echo "$as_me:$LINENO: \$? = $ac_status" >&5
   (exit $ac_status); }; }; then
+  ac_ext=c
+ac_cpp='$CPP $CPPFLAGS'
+ac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'
+ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS >&5'
+ac_compiler_gnu=$ac_cv_c_compiler_gnu
+
   { echo "$as_me:$LINENO: result: yes" >&5
 echo "${ECHO_T}yes" >&6; }
 




From daniel at zuster.org  Wed Jul 28 15:27:09 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 28 Jul 2010 20:27:09 -0000
Subject: [llvm-commits] [test-suite] r109648 -
 /test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output
Message-ID: <20100728202709.2BB972A6C12D@llvm.org>

Author: ddunbar
Date: Wed Jul 28 15:27:09 2010
New Revision: 109648

URL: http://llvm.org/viewvc/llvm-project?rev=109648&view=rev
Log:
Add reference output.

Added:
    test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output

Added: test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output?rev=109648&view=auto
==============================================================================
--- test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output (added)
+++ test-suite/trunk/SingleSource/UnitTests/blockstret.reference_output Wed Jul 28 15:27:09 2010
@@ -0,0 +1 @@
+exit 0




From daniel at zuster.org  Wed Jul 28 15:28:45 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 28 Jul 2010 20:28:45 -0000
Subject: [llvm-commits] [llvm] r109649 - in /llvm/trunk:
 include/llvm/MC/MCAssembler.h lib/MC/MCAssembler.cpp
Message-ID: <20100728202845.6494B2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Jul 28 15:28:45 2010
New Revision: 109649

URL: http://llvm.org/viewvc/llvm-project?rev=109649&view=rev
Log:
MC: Put back the MCFragment vtable, so subclasses are destroyed properly (duh).

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

Modified: llvm/trunk/include/llvm/MC/MCAssembler.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCAssembler.h?rev=109649&r1=109648&r2=109649&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCAssembler.h (original)
+++ llvm/trunk/include/llvm/MC/MCAssembler.h Wed Jul 28 15:28:45 2010
@@ -87,6 +87,7 @@
 public:
   // Only for sentinel.
   MCFragment();
+  virtual ~MCFragment();
 
   FragmentType getKind() const { return Kind; }
 

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=109649&r1=109648&r2=109649&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Wed Jul 28 15:28:45 2010
@@ -178,6 +178,9 @@
 MCFragment::MCFragment() : Kind(FragmentType(~0)) {
 }
 
+MCFragment::~MCFragment() {
+}
+
 MCFragment::MCFragment(FragmentType _Kind, MCSectionData *_Parent)
   : Kind(_Kind), Parent(_Parent), Atom(0), EffectiveSize(~UINT64_C(0))
 {




From daniel at zuster.org  Wed Jul 28 15:28:50 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Wed, 28 Jul 2010 20:28:50 -0000
Subject: [llvm-commits] [llvm] r109650 -
	/llvm/trunk/lib/Analysis/RegionInfo.cpp
Message-ID: <20100728202851.03A312A6C12D@llvm.org>

Author: ddunbar
Date: Wed Jul 28 15:28:50 2010
New Revision: 109650

URL: http://llvm.org/viewvc/llvm-project?rev=109650&view=rev
Log:
RegionInfo: Make sure to free cached nodes; Tobias, please check!

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

Modified: llvm/trunk/lib/Analysis/RegionInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/RegionInfo.cpp?rev=109650&r1=109649&r2=109650&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/RegionInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/RegionInfo.cpp Wed Jul 28 15:28:50 2010
@@ -59,6 +59,11 @@
                : RegionNode(Parent, Entry, 1), RI(RInfo), DT(dt), exit(Exit) {}
 
 Region::~Region() {
+  // Free the cached nodes.
+  for (BBNodeMapT::iterator it = BBNodeMap.begin(),
+         ie = BBNodeMap.end(); it != ie; ++it)
+    delete it->second;
+
   // Only clean the cache for this Region. Caches of child Regions will be
   // cleaned when the child Regions are deleted.
   BBNodeMap.clear();




From enderby at apple.com  Wed Jul 28 15:55:35 2010
From: enderby at apple.com (Kevin Enderby)
Date: Wed, 28 Jul 2010 20:55:35 -0000
Subject: [llvm-commits] [llvm] r109651 - in /llvm/trunk:
 include/llvm/MC/MCContext.h include/llvm/MC/MCDwarf.h lib/MC/CMakeLists.txt
 lib/MC/MCContext.cpp lib/MC/MCDwarf.cpp lib/MC/MCParser/AsmParser.cpp
Message-ID: <20100728205535.D560F2A6C12C@llvm.org>

Author: enderby
Date: Wed Jul 28 15:55:35 2010
New Revision: 109651

URL: http://llvm.org/viewvc/llvm-project?rev=109651&view=rev
Log:
Added first bit of support for the dwarf .file directive.  This patch collects
the info from the .file directive and makes file and directory tables that
will eventually be put out as part of the dwarf info in the output file.

Added:
    llvm/trunk/include/llvm/MC/MCDwarf.h
    llvm/trunk/lib/MC/MCDwarf.cpp
Modified:
    llvm/trunk/include/llvm/MC/MCContext.h
    llvm/trunk/lib/MC/CMakeLists.txt
    llvm/trunk/lib/MC/MCContext.cpp
    llvm/trunk/lib/MC/MCParser/AsmParser.cpp

Modified: llvm/trunk/include/llvm/MC/MCContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCContext.h?rev=109651&r1=109650&r2=109651&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCContext.h (original)
+++ llvm/trunk/include/llvm/MC/MCContext.h Wed Jul 28 15:55:35 2010
@@ -15,6 +15,7 @@
 #include "llvm/ADT/StringMap.h"
 #include "llvm/Support/Allocator.h"
 #include "llvm/Support/raw_ostream.h"
+#include <vector> // FIXME: Shouldn't be needed.
 
 namespace llvm {
   class MCAsmInfo;
@@ -22,6 +23,7 @@
   class MCSection;
   class MCSymbol;
   class MCLabel;
+  class MCDwarfFile;
   class StringRef;
   class Twine;
   class MCSectionMachO;
@@ -66,6 +68,10 @@
     /// .secure_log_reset appearing between them.
     bool SecureLogUsed;
 
+    /// The dwarf file and directory tables from the dwarf .file directive.
+    std::vector<MCDwarfFile *> MCDwarfFiles;
+    std::vector<std::string *> MCDwarfDirs;
+
     /// Allocator - Allocator object used for creating machine code objects.
     ///
     /// We use a bump pointer allocator to avoid the need to track all allocated
@@ -139,6 +145,18 @@
     
     /// @}
 
+    /// @name Dwarf Managment
+    /// @{
+
+    /// GetDwarfFile - creates an entry in the dwarf file and directory tables.
+    unsigned GetDwarfFile(StringRef FileName, unsigned FileNumber);
+
+    const std::vector<MCDwarfFile *> &getMCDwarfFiles() {
+      return MCDwarfFiles;
+    }
+
+    /// @}
+
     char *getSecureLogFile() { return SecureLogFile; }
     raw_ostream *getSecureLog() { return SecureLog; }
     bool getSecureLogUsed() { return SecureLogUsed; }

Added: llvm/trunk/include/llvm/MC/MCDwarf.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCDwarf.h?rev=109651&view=auto
==============================================================================
--- llvm/trunk/include/llvm/MC/MCDwarf.h (added)
+++ llvm/trunk/include/llvm/MC/MCDwarf.h Wed Jul 28 15:55:35 2010
@@ -0,0 +1,61 @@
+//===- MCDwarf.h - Machine Code Dwarf support -------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains the declaration of the MCDwarfFile to support the dwarf
+// .file directive.
+// TODO: add the support needed for the .loc directive.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_MC_MCDWARF_H
+#define LLVM_MC_MCDWARF_H
+
+#include <string>
+
+namespace llvm {
+  class MCContext;
+  class raw_ostream;
+
+  /// MCDwarfFile - Instances of this class represent the name of the dwarf
+  /// .file directive and its associated dwarf file number in the MC file,
+  /// and MCDwarfFile's are created and unique'd by the MCContext class where
+  /// the file number for each is its index into the vector of DwarfFiles (note
+  /// index 0 is not used and not a valid dwarf file number).
+  class MCDwarfFile {
+    // Name - the base name of the file without its directory path.
+    std::string Name;
+
+    // DirIndex - the index into the list of directory names for this file name.
+    unsigned DirIndex;
+
+  private:  // MCContext creates and uniques these.
+    friend class MCContext;
+    MCDwarfFile(std::string name, unsigned dirIndex)
+      : Name(name), DirIndex(dirIndex) {}
+
+    MCDwarfFile(const MCDwarfFile&);       // DO NOT IMPLEMENT
+    void operator=(const MCDwarfFile&); // DO NOT IMPLEMENT
+  public:
+    /// getName - Get the base name of this MCDwarfFile.
+    std::string getName() const { return Name; }
+
+    /// print - Print the value to the stream \arg OS.
+    void print(raw_ostream &OS) const;
+
+    /// dump - Print the value to stderr.
+    void dump() const;
+  };
+
+  inline raw_ostream &operator<<(raw_ostream &OS, const MCDwarfFile &DwarfFile){
+    DwarfFile.print(OS);
+    return OS;
+  }
+} // end namespace llvm
+
+#endif

Modified: llvm/trunk/lib/MC/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/CMakeLists.txt?rev=109651&r1=109650&r2=109651&view=diff
==============================================================================
--- llvm/trunk/lib/MC/CMakeLists.txt (original)
+++ llvm/trunk/lib/MC/CMakeLists.txt Wed Jul 28 15:55:35 2010
@@ -11,6 +11,7 @@
   MCInst.cpp
   MCInstPrinter.cpp
   MCLabel.cpp
+  MCDwarf.cpp
   MCLoggingStreamer.cpp
   MCMachOStreamer.cpp
   MCNullStreamer.cpp

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=109651&r1=109650&r2=109651&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Wed Jul 28 15:55:35 2010
@@ -14,6 +14,7 @@
 #include "llvm/MC/MCSectionCOFF.h"
 #include "llvm/MC/MCSymbol.h"
 #include "llvm/MC/MCLabel.h"
+#include "llvm/MC/MCDwarf.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/Twine.h"
 using namespace llvm;
@@ -181,3 +182,63 @@
   Entry.setValue(Result);
   return Result;
 }
+
+//===----------------------------------------------------------------------===//
+// Dwarf Management
+//===----------------------------------------------------------------------===//
+
+/// GetDwarfFile - takes a file name an number to place in the dwarf file and
+/// directory tables.  If the file number has already been allocated it is an
+/// error and zero is returned and the client reports the error, else the
+/// allocated file number is returned.  The file numbers may be in any order.
+unsigned MCContext::GetDwarfFile(StringRef FileName, unsigned FileNumber) {
+  // TODO: a FileNumber of zero says to use the next available file number.
+  // Note: in GenericAsmParser::ParseDirectiveFile() FileNumber was checked
+  // to not be less than one.  This needs to be change to be not less than zero.
+
+  // Make space for this FileNumber in the MCDwarfFiles vector if needed.
+  if (FileNumber >= MCDwarfFiles.size()) {
+    MCDwarfFiles.resize(FileNumber + 1);
+  } else {
+    MCDwarfFile *&ExistingFile = MCDwarfFiles[FileNumber];
+    if (ExistingFile)
+      // It is an error to use see the same number more than once.
+      return 0;
+  }
+
+  // Get the new MCDwarfFile slot for this FileNumber.
+  MCDwarfFile *&File = MCDwarfFiles[FileNumber];
+
+  // Separate the directory part from the basename of the FileName.
+  std::pair<StringRef, StringRef> Slash = FileName.rsplit('/');
+
+  // Find or make a entry in the MCDwarfDirs vector for this Directory.
+  StringRef Directory;
+  StringRef Name;
+  unsigned DirIndex;
+  // Capture directory name.
+  if (Slash.second.empty()) {
+    Name = Slash.first;
+    DirIndex = 0; // For FileNames with no directories a DirIndex of 0 is used.
+  } else {
+    Directory = Slash.first;
+    Name = Slash.second;
+    for (DirIndex = 1; DirIndex < MCDwarfDirs.size(); DirIndex++) {
+      std::string *&Dir = MCDwarfDirs[DirIndex];
+      if (Directory == *Dir)
+	break;
+    }
+    if (DirIndex >= MCDwarfDirs.size()) {
+      MCDwarfDirs.resize(DirIndex + 1);
+      std::string *&NewDir = MCDwarfDirs[DirIndex];
+      NewDir = new (*this) std::string(Directory);
+    }
+  }
+  
+  // Now make the MCDwarfFile entry and place it in the slot in the MCDwarfFiles
+  // vector.
+  File = new (*this) MCDwarfFile(Name, DirIndex);
+
+  // return the allocated FileNumber.
+  return FileNumber;
+}

Added: llvm/trunk/lib/MC/MCDwarf.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCDwarf.cpp?rev=109651&view=auto
==============================================================================
--- llvm/trunk/lib/MC/MCDwarf.cpp (added)
+++ llvm/trunk/lib/MC/MCDwarf.cpp Wed Jul 28 15:55:35 2010
@@ -0,0 +1,21 @@
+//===- lib/MC/MCDwarf.cpp - MCDwarf implementation ------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/MC/MCDwarf.h"
+#include "llvm/Support/Debug.h"
+#include "llvm/Support/raw_ostream.h"
+using namespace llvm;
+
+void MCDwarfFile::print(raw_ostream &OS) const {
+  OS << '"' << getName() << '"';
+}
+
+void MCDwarfFile::dump() const {
+  print(dbgs());
+}

Modified: llvm/trunk/lib/MC/MCParser/AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCParser/AsmParser.cpp?rev=109651&r1=109650&r2=109651&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCParser/AsmParser.cpp (original)
+++ llvm/trunk/lib/MC/MCParser/AsmParser.cpp Wed Jul 28 15:55:35 2010
@@ -26,6 +26,7 @@
 #include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/MCStreamer.h"
 #include "llvm/MC/MCSymbol.h"
+#include "llvm/MC/MCDwarf.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/SourceMgr.h"
@@ -370,6 +371,16 @@
   if (TheCondState.TheCond != StartingCondState.TheCond ||
       TheCondState.Ignore != StartingCondState.Ignore)
     return TokError("unmatched .ifs or .elses");
+
+  // Check to see there are no empty DwarfFile slots.
+  const std::vector<MCDwarfFile *> &MCDwarfFiles =
+    getContext().getMCDwarfFiles();
+  for (unsigned i = 1; i < MCDwarfFiles.size(); i++) {
+    if (!MCDwarfFiles[i]){
+      TokError("unassigned file number: " + Twine(i) + " for .file directives");
+      HadError = true;
+    }
+  }
   
   // Finalize the output stream if there are no errors and if the client wants
   // us to.
@@ -1729,6 +1740,7 @@
 bool GenericAsmParser::ParseDirectiveFile(StringRef, SMLoc DirectiveLoc) {
   // FIXME: I'm not sure what this is.
   int64_t FileNumber = -1;
+  SMLoc FileNumberLoc = getLexer().getLoc();
   if (getLexer().is(AsmToken::Integer)) {
     FileNumber = getTok().getIntVal();
     Lex();
@@ -1749,8 +1761,11 @@
 
   if (FileNumber == -1)
     getStreamer().EmitFileDirective(Filename);
-  else
+  else {
+     if (getContext().GetDwarfFile(Filename, FileNumber) == 0)
+	Error(FileNumberLoc, "file number already allocated");
     getStreamer().EmitDwarfFileDirective(FileNumber, Filename);
+  }
 
   return false;
 }




From stoklund at 2pi.dk  Wed Jul 28 15:55:39 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Wed, 28 Jul 2010 20:55:39 -0000
Subject: [llvm-commits] [llvm] r109652 -
	/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
Message-ID: <20100728205539.24BDD2A6C12D@llvm.org>

Author: stoklund
Date: Wed Jul 28 15:55:38 2010
New Revision: 109652

URL: http://llvm.org/viewvc/llvm-project?rev=109652&view=rev
Log:
Create a fixed stack object for varargs that is as large as any register.

The size of this object isn't used for anything - technically it is of variable
size.

This avoids a false positive from the assert in
X86InstrInfo::loadRegFromStackSlot, and fixes PR7735.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109652&r1=109651&r2=109652&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Wed Jul 28 15:55:38 2010
@@ -1701,7 +1701,10 @@
   if (isVarArg) {
     if (Is64Bit || (CallConv != CallingConv::X86_FastCall &&
                     CallConv != CallingConv::X86_ThisCall)) {
-      FuncInfo->setVarArgsFrameIndex(MFI->CreateFixedObject(1, StackSize,true));
+      // The fixed varargs object is created large enough that we can load and
+      // store any register in it.
+      FuncInfo->setVarArgsFrameIndex(MFI->CreateFixedObject(32, StackSize,
+                                                            true));
     }
     if (Is64Bit) {
       unsigned TotalNumIntRegs = 0, TotalNumXMMRegs = 0;




From gohman at apple.com  Wed Jul 28 15:56:48 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 20:56:48 -0000
Subject: [llvm-commits] [llvm] r109653 - in /llvm/trunk:
 include/llvm/GlobalValue.h lib/VMCore/Globals.cpp
 test/Assembler/align-inst.ll
Message-ID: <20100728205648.A1B942A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 15:56:48 2010
New Revision: 109653

URL: http://llvm.org/viewvc/llvm-project?rev=109653&view=rev
Log:
Make GlobalValue alignment consistent with load, store, and alloca
alignment, fixing silent truncation of alignment values.

Modified:
    llvm/trunk/include/llvm/GlobalValue.h
    llvm/trunk/lib/VMCore/Globals.cpp
    llvm/trunk/test/Assembler/align-inst.ll

Modified: llvm/trunk/include/llvm/GlobalValue.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/GlobalValue.h?rev=109653&r1=109652&r2=109653&view=diff
==============================================================================
--- llvm/trunk/include/llvm/GlobalValue.h (original)
+++ llvm/trunk/include/llvm/GlobalValue.h Wed Jul 28 15:56:48 2010
@@ -74,11 +74,10 @@
     removeDeadConstantUsers();   // remove any dead constants using this.
   }
 
-  unsigned getAlignment() const { return Alignment; }
-  void setAlignment(unsigned Align) {
-    assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
-    Alignment = Align;
+  unsigned getAlignment() const {
+    return (1u << Alignment) >> 1;
   }
+  void setAlignment(unsigned Align);
 
   VisibilityTypes getVisibility() const { return VisibilityTypes(Visibility); }
   bool hasDefaultVisibility() const { return Visibility == DefaultVisibility; }

Modified: llvm/trunk/lib/VMCore/Globals.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Globals.cpp?rev=109653&r1=109652&r2=109653&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Globals.cpp (original)
+++ llvm/trunk/lib/VMCore/Globals.cpp Wed Jul 28 15:56:48 2010
@@ -102,7 +102,14 @@
   setVisibility(Src->getVisibility());
 }
 
-
+void GlobalValue::setAlignment(unsigned Align) {
+  assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
+  assert(Align <= MaximumAlignment &&
+         "Alignment is greater than MaximumAlignment!");
+  Alignment = Log2_32(Align) + 1;
+  assert(getAlignment() == Align && "Alignment representation error!");
+}
+  
 //===----------------------------------------------------------------------===//
 // GlobalVariable Implementation
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/test/Assembler/align-inst.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst.ll?rev=109653&r1=109652&r2=109653&view=diff
==============================================================================
--- llvm/trunk/test/Assembler/align-inst.ll (original)
+++ llvm/trunk/test/Assembler/align-inst.ll Wed Jul 28 15:56:48 2010
@@ -1,5 +1,7 @@
 ; RUN: llvm-as %s -o /dev/null
 
+ at A = global i1 0, align 536870912
+
 define void @foo() {
   %p = alloca i1, align 536870912
   load i1* %p, align 536870912




From gohman at apple.com  Wed Jul 28 16:02:38 2010
From: gohman at apple.com (Dan Gohman)
Date: Wed, 28 Jul 2010 21:02:38 -0000
Subject: [llvm-commits] [llvm] r109656 - in /llvm/trunk/include/llvm:
	Instructions.h Value.h
Message-ID: <20100728210238.B99902A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 16:02:38 2010
New Revision: 109656

URL: http://llvm.org/viewvc/llvm-project?rev=109656&view=rev
Log:
Move MaximumAlignment to Value.h, now that GlobalValue.h needs it.

Modified:
    llvm/trunk/include/llvm/Instructions.h
    llvm/trunk/include/llvm/Value.h

Modified: llvm/trunk/include/llvm/Instructions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Instructions.h?rev=109656&r1=109655&r2=109656&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Instructions.h (original)
+++ llvm/trunk/include/llvm/Instructions.h Wed Jul 28 16:02:38 2010
@@ -31,10 +31,6 @@
 class LLVMContext;
 class DominatorTree;
 
-/// MaximumAlignment - This is the greatest alignment value supported by
-/// load, store, and alloca instructions.
-static const unsigned MaximumAlignment = 1u << 29;
-
 //===----------------------------------------------------------------------===//
 //                                AllocaInst Class
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/include/llvm/Value.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Value.h?rev=109656&r1=109655&r2=109656&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Value.h (original)
+++ llvm/trunk/include/llvm/Value.h Wed Jul 28 16:02:38 2010
@@ -401,6 +401,10 @@
   enum { NumLowBitsAvailable = 2 };
 };
 
+/// MaximumAlignment - This is the greatest alignment value supported by
+/// load, store, and alloca instructions, and global values.
+static const unsigned MaximumAlignment = 1u << 29;
+
 } // End llvm namespace
 
 #endif




From echristo at apple.com  Wed Jul 28 16:05:43 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 21:05:43 -0000
Subject: [llvm-commits] [test-suite] r109658 - in /test-suite/trunk:
 External/Makefile External/skidmarks10/ External/skidmarks10/Makefile
 Makefile.config.in autoconf/configure.ac
Message-ID: <20100728210543.E29952A6C12C@llvm.org>

Author: echristo
Date: Wed Jul 28 16:05:43 2010
New Revision: 109658

URL: http://llvm.org/viewvc/llvm-project?rev=109658&view=rev
Log:
Add support for the External "skidmarks" test.

Added:
    test-suite/trunk/External/skidmarks10/
    test-suite/trunk/External/skidmarks10/Makefile
Modified:
    test-suite/trunk/External/Makefile
    test-suite/trunk/Makefile.config.in
    test-suite/trunk/autoconf/configure.ac

Modified: test-suite/trunk/External/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/Makefile?rev=109658&r1=109657&r2=109658&view=diff
==============================================================================
--- test-suite/trunk/External/Makefile (original)
+++ test-suite/trunk/External/Makefile Wed Jul 28 16:05:43 2010
@@ -8,7 +8,7 @@
 #
 # Create the list of directories to compile
 #
-PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim
+PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim skidmarks10
 
 ifneq (,$(findstring x86,$(ARCH)))
 # Nurbs uses SSE and only works on x86 and x86_64.

Added: test-suite/trunk/External/skidmarks10/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/skidmarks10/Makefile?rev=109658&view=auto
==============================================================================
--- test-suite/trunk/External/skidmarks10/Makefile (added)
+++ test-suite/trunk/External/skidmarks10/Makefile Wed Jul 28 16:05:43 2010
@@ -0,0 +1,21 @@
+LEVEL = ../..
+include $(LEVEL)/Makefile.config
+
+PROG = skidmarks
+
+DISABLE_JIT = 1
+DISABLE_CBE = 1
+
+SourceDir := $(SKIDMARKS10_ROOT)/source
+RUN_OPTIONS = 0 $(SKIDMARKS10_ROOT)/golden
+
+CPPFLAGS += -I$(SKIDMARKS10_ROOT)/source
+
+# skidmarks requires sse3 on x86
+ifneq (,$(findstring x86,$(ARCH)))
+CPPFLAGS += -msse3
+endif
+
+# TODO: We may want to set more flags than the default run here.
+
+include $(LEVEL)/MultiSource/Makefile.multisrc

Modified: test-suite/trunk/Makefile.config.in
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/Makefile.config.in?rev=109658&r1=109657&r2=109658&view=diff
==============================================================================
--- test-suite/trunk/Makefile.config.in (original)
+++ test-suite/trunk/Makefile.config.in Wed Jul 28 16:05:43 2010
@@ -121,6 +121,10 @@
 @USE_HMMER@
 HMMER_ROOT := @HMMER_ROOT@
 
+# Path to the Skidmarks source code
+ at USE_SKIDMARKS10@
+SKIDMARKS10_ROOT := @SKIDMARKS10_ROOT@
+
 # Disable LLC diffs for testing.
 @DISABLE_LLC_DIFFS@
 

Modified: test-suite/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/autoconf/configure.ac?rev=109658&r1=109657&r2=109658&view=diff
==============================================================================
--- test-suite/trunk/autoconf/configure.ac (original)
+++ test-suite/trunk/autoconf/configure.ac Wed Jul 28 16:05:43 2010
@@ -41,6 +41,7 @@
 AC_CONFIG_MAKEFILE(External/SPEC/CINT2000/Makefile)
 AC_CONFIG_MAKEFILE(External/SPEC/CFP95/Makefile)
 AC_CONFIG_MAKEFILE(External/SPEC/CINT95/Makefile)
+AC_CONFIG_MAKEFILE(External/skidmarks10/Makefile)
 AC_CONFIG_MAKEFILE(MultiSource/Makefile)
 AC_CONFIG_MAKEFILE(MultiSource/Makefile.multisrc)
 AC_CONFIG_MAKEFILE(SingleSource/Makefile)
@@ -79,6 +80,7 @@
 EXTERNAL_BENCHMARK(alp,${LLVM_EXTERNALS}/alp)
 EXTERNAL_BENCHMARK(nurbs,${LLVM_EXTERNALS}/nurbs)
 EXTERNAL_BENCHMARK(hmmer,${LLVM_EXTERNALS}/hmmer)
+EXTERNAL_BENCHMARK(skidmarks10,${LLVM_EXTERNALS}/skidmarks10)
 
 dnl LLC Diff Option
 AC_ARG_ENABLE(llc_diffs,




From echristo at apple.com  Wed Jul 28 16:05:55 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 21:05:55 -0000
Subject: [llvm-commits] [test-suite] r109659 - /test-suite/trunk/configure
Message-ID: <20100728210556.18A922A6C12C@llvm.org>

Author: echristo
Date: Wed Jul 28 16:05:55 2010
New Revision: 109659

URL: http://llvm.org/viewvc/llvm-project?rev=109659&view=rev
Log:
Regenerate.

Modified:
    test-suite/trunk/configure

Modified: test-suite/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/configure?rev=109659&r1=109658&r2=109659&view=diff
==============================================================================
--- test-suite/trunk/configure (original)
+++ test-suite/trunk/configure Wed Jul 28 16:05:55 2010
@@ -814,6 +814,8 @@
 USE_NURBS
 HMMER_ROOT
 USE_HMMER
+SKIDMARKS10_ROOT
+USE_SKIDMARKS10
 DISABLE_LLC_DIFFS
 CXX
 CXXFLAGS
@@ -2061,6 +2063,9 @@
 ac_config_commands="$ac_config_commands External/SPEC/CINT95/Makefile"
 
 
+ac_config_commands="$ac_config_commands External/skidmarks10/Makefile"
+
+
 ac_config_commands="$ac_config_commands MultiSource/Makefile"
 
 
@@ -2489,6 +2494,44 @@
 
 
 
+
+
+
+{ echo "$as_me:$LINENO: checking for skidmarks10 benchmark sources" >&5
+echo $ECHO_N "checking for skidmarks10 benchmark sources... $ECHO_C" >&6; }
+if test -d "${LLVM_EXTERNALS}/skidmarks10" ; then
+  if test -n "" ; then
+    if test -e "${LLVM_EXTERNALS}/skidmarks10/" ; then
+      SKIDMARKS10_ROOT=${LLVM_EXTERNALS}/skidmarks10
+
+      USE_SKIDMARKS10=USE_SKIDMARKS10=1
+
+      checkresult="yes, found in ${LLVM_EXTERNALS}/skidmarks10"
+    else
+      checkresult="no"
+    fi
+  else
+    SKIDMARKS10_ROOT=${LLVM_EXTERNALS}/skidmarks10
+
+    USE_SKIDMARKS10=USE_SKIDMARKS10=1
+
+    checkresult="yes, found in ${LLVM_EXTERNALS}/skidmarks10"
+  fi
+else
+  checkresult="no"
+fi
+if test "$checkresult" = "no" ; then
+
+
+  checkresult="no, not found in ${LLVM_EXTERNALS}/skidmarks10"
+fi
+{ echo "$as_me:$LINENO: result: $checkresult" >&5
+echo "${ECHO_T}$checkresult" >&6; }
+
+
+
+
+
 # Check whether --enable-llc_diffs was given.
 if test "${enable_llc_diffs+set}" = set; then
   enableval=$enable_llc_diffs;
@@ -4830,7 +4873,7 @@
   ;;
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 4833 "configure"' > conftest.$ac_ext
+  echo '#line 4876 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -6995,11 +7038,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:6998: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:7041: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:7002: \$? = $ac_status" >&5
+   echo "$as_me:7045: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -7263,11 +7306,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:7266: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:7309: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:7270: \$? = $ac_status" >&5
+   echo "$as_me:7313: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -7367,11 +7410,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:7370: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:7413: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:7374: \$? = $ac_status" >&5
+   echo "$as_me:7417: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -9819,7 +9862,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 9822 "configure"
+#line 9865 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -9919,7 +9962,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 9922 "configure"
+#line 9965 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12287,11 +12330,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:12290: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:12333: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:12294: \$? = $ac_status" >&5
+   echo "$as_me:12337: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -12391,11 +12434,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:12394: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:12437: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:12398: \$? = $ac_status" >&5
+   echo "$as_me:12441: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -13961,11 +14004,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:13964: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14007: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:13968: \$? = $ac_status" >&5
+   echo "$as_me:14011: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -14065,11 +14108,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:14068: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14111: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:14072: \$? = $ac_status" >&5
+   echo "$as_me:14115: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -16300,11 +16343,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:16303: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:16346: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:16307: \$? = $ac_status" >&5
+   echo "$as_me:16350: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -16568,11 +16611,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:16571: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:16614: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:16575: \$? = $ac_status" >&5
+   echo "$as_me:16618: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s "$ac_outfile"; then
      # The compiler can only warn and ignore the option if not recognized
      # So say no if there are warnings other than the usual output.
@@ -16672,11 +16715,11 @@
    -e 's:.*FLAGS}\{0,1\} :&$lt_compiler_flag :; t' \
    -e 's: [^ ]*conftest\.: $lt_compiler_flag&:; t' \
    -e 's:$: $lt_compiler_flag:'`
-   (eval echo "\"\$as_me:16675: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:16718: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:16679: \$? = $ac_status" >&5
+   echo "$as_me:16722: \$? = $ac_status" >&5
    if (exit $ac_status) && test -s out/conftest2.$ac_objext
    then
      # The compiler can only warn and ignore the option if not recognized
@@ -20233,7 +20276,7 @@
 int
 main ()
 {
-int main() { return 0; }
+
   ;
   return 0;
 }
@@ -21028,6 +21071,7 @@
     "External/SPEC/CINT2000/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS External/SPEC/CINT2000/Makefile" ;;
     "External/SPEC/CFP95/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS External/SPEC/CFP95/Makefile" ;;
     "External/SPEC/CINT95/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS External/SPEC/CINT95/Makefile" ;;
+    "External/skidmarks10/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS External/skidmarks10/Makefile" ;;
     "MultiSource/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS MultiSource/Makefile" ;;
     "MultiSource/Makefile.multisrc") CONFIG_COMMANDS="$CONFIG_COMMANDS MultiSource/Makefile.multisrc" ;;
     "SingleSource/Makefile") CONFIG_COMMANDS="$CONFIG_COMMANDS SingleSource/Makefile" ;;
@@ -21153,6 +21197,8 @@
 USE_NURBS!$USE_NURBS$ac_delim
 HMMER_ROOT!$HMMER_ROOT$ac_delim
 USE_HMMER!$USE_HMMER$ac_delim
+SKIDMARKS10_ROOT!$SKIDMARKS10_ROOT$ac_delim
+USE_SKIDMARKS10!$USE_SKIDMARKS10$ac_delim
 DISABLE_LLC_DIFFS!$DISABLE_LLC_DIFFS$ac_delim
 CXX!$CXX$ac_delim
 CXXFLAGS!$CXXFLAGS$ac_delim
@@ -21188,8 +21234,6 @@
 LIBTOOL!$LIBTOOL$ac_delim
 USE_F2C!$USE_F2C$ac_delim
 F2C!$F2C$ac_delim
-F2C_BIN!$F2C_BIN$ac_delim
-F2C_DIR!$F2C_DIR$ac_delim
 _ACEOF
 
   if test `sed -n "s/.*$ac_delim\$/X/p" conf$$subs.sed | grep -c X` = 97; then
@@ -21231,6 +21275,8 @@
 ac_delim='%!_!# '
 for ac_last_try in false false false false false :; do
   cat >conf$$subs.sed <<_ACEOF
+F2C_BIN!$F2C_BIN$ac_delim
+F2C_DIR!$F2C_DIR$ac_delim
 F2C_INC!$F2C_INC$ac_delim
 F2C_LIB!$F2C_LIB$ac_delim
 USE_F95!$USE_F95$ac_delim
@@ -21245,7 +21291,7 @@
 LTLIBOBJS!$LTLIBOBJS$ac_delim
 _ACEOF
 
-  if test `sed -n "s/.*$ac_delim\$/X/p" conf$$subs.sed | grep -c X` = 12; then
+  if test `sed -n "s/.*$ac_delim\$/X/p" conf$$subs.sed | grep -c X` = 14; then
     break
   elif $ac_last_try; then
     { { echo "$as_me:$LINENO: error: could not make $CONFIG_STATUS" >&5
@@ -21608,6 +21654,8 @@
    ${SHELL} ${llvm_src}/autoconf/install-sh -m 0644 -c ${srcdir}/External/SPEC/CFP95/Makefile External/SPEC/CFP95/Makefile ;;
     "External/SPEC/CINT95/Makefile":C) ${llvm_src}/autoconf/mkinstalldirs `dirname External/SPEC/CINT95/Makefile`
    ${SHELL} ${llvm_src}/autoconf/install-sh -m 0644 -c ${srcdir}/External/SPEC/CINT95/Makefile External/SPEC/CINT95/Makefile ;;
+    "External/skidmarks10/Makefile":C) ${llvm_src}/autoconf/mkinstalldirs `dirname External/skidmarks10/Makefile`
+   ${SHELL} ${llvm_src}/autoconf/install-sh -m 0644 -c ${srcdir}/External/skidmarks10/Makefile External/skidmarks10/Makefile ;;
     "MultiSource/Makefile":C) ${llvm_src}/autoconf/mkinstalldirs `dirname MultiSource/Makefile`
    ${SHELL} ${llvm_src}/autoconf/install-sh -m 0644 -c ${srcdir}/MultiSource/Makefile MultiSource/Makefile ;;
     "MultiSource/Makefile.multisrc":C) ${llvm_src}/autoconf/mkinstalldirs `dirname MultiSource/Makefile.multisrc`




From resistor at mac.com  Wed Jul 28 17:07:25 2010
From: resistor at mac.com (Owen Anderson)
Date: Wed, 28 Jul 2010 22:07:25 -0000
Subject: [llvm-commits] [llvm] r109679 -
	/llvm/trunk/lib/Analysis/LazyValueInfo.cpp
Message-ID: <20100728220725.5A97C2A6C12C@llvm.org>

Author: resistor
Date: Wed Jul 28 17:07:25 2010
New Revision: 109679

URL: http://llvm.org/viewvc/llvm-project?rev=109679&view=rev
Log:
Get rid of LVIQuery as a distinct data structure, so that we don't have to initialize a new set of maps on every query.

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

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109679&r1=109678&r2=109679&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Wed Jul 28 17:07:25 2010
@@ -204,7 +204,7 @@
   /// LazyValueInfoCache - This is the cache kept by LazyValueInfo which
   /// maintains information about queries across the clients' queries.
   class LazyValueInfoCache {
-  public:
+  private:
     /// BlockCacheEntryTy - This is a computed lattice value at the end of the
     /// specified basic block for a Value* that depends on context.
     typedef std::pair<BasicBlock*, LVILatticeVal> BlockCacheEntryTy;
@@ -214,7 +214,6 @@
     /// entries, allowing us to do a lookup with a binary search.
     typedef DenseMap<BasicBlock*, LVILatticeVal> ValueCacheEntryTy;
 
-  private:
     /// ValueCache - This is all of the cached information for all values,
     /// mapped from Value* to key information.
     DenseMap<Value*, ValueCacheEntryTy> ValueCache;
@@ -223,7 +222,54 @@
     /// values that are over-defined at the end of that block.  This is required
     /// for cache updating.
     DenseSet<std::pair<BasicBlock*, Value*> > OverDefinedCache;
+    
+    LVILatticeVal getBlockValue(ValueCacheEntryTy &Cache, BasicBlock *BB);
+    LVILatticeVal getEdgeValue(ValueCacheEntryTy &Cache,
+                               BasicBlock *Pred, BasicBlock *Succ);
+    LVILatticeVal &getCachedEntryForBlock(ValueCacheEntryTy &Cache,
+                                          BasicBlock *BB);
+    
+    /************* Begin Per-Query State *************/
+    /// This is the current value being queried for.
+    Value *Val;
+    
+    /// This is all of the cached information about this value.
+    //ValueCacheEntryTy *Cache; 
+    
+    ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
+    /// added to cache but that are not in sorted order.
+    DenseSet<BasicBlock*> NewBlockInfo;
+    
+    /// QuerySetup - An RAII helper to construct and tear-down per-query 
+    /// temporary state.
+    struct QuerySetup {
+      LazyValueInfoCache &Owner;
+      QuerySetup(LazyValueInfoCache &O, Value* Val) : Owner(O) {
+        assert(!Owner.Val && "Per-query info not cleared?");
+        Owner.Val = Val;
+        assert(Owner.NewBlockInfo.empty() && "Leaked block info!");
+      }
+      
+      ~QuerySetup() {
+        // When the query is done, insert the newly discovered facts into the
+        // cache in sorted order.
+        LazyValueInfoCache::ValueCacheEntryTy Cache = 
+          Owner.ValueCache[Owner.Val];
+        for (DenseSet<BasicBlock*>::iterator I = Owner.NewBlockInfo.begin(),
+             E = Owner.NewBlockInfo.end(); I != E; ++I) {
+          if (Cache[*I].isOverdefined())
+            Owner.OverDefinedCache.insert(std::make_pair(*I, Owner.Val));
+        }
+        
+        // Reset Per-Query State
+        Owner.Val = 0;
+        Owner.NewBlockInfo.clear();
+      }
+    };
+    /************* End Per-Query State *************/
+    
   public:
+    LazyValueInfoCache() : Val(0) { }
     
     /// getValueInBlock - This is the query interface to determine the lattice
     /// value for the specified Value* at the end of the specified block.
@@ -240,90 +286,20 @@
   };
 } // end anonymous namespace
 
-namespace {
-  struct BlockCacheEntryComparator {
-    static int Compare(const void *LHSv, const void *RHSv) {
-      const LazyValueInfoCache::BlockCacheEntryTy *LHS =
-        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(LHSv);
-      const LazyValueInfoCache::BlockCacheEntryTy *RHS =
-        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(RHSv);
-      if (LHS->first < RHS->first)
-        return -1;
-      if (LHS->first > RHS->first)
-        return 1;
-      return 0;
-    }
-    
-    bool operator()(const LazyValueInfoCache::BlockCacheEntryTy &LHS,
-                    const LazyValueInfoCache::BlockCacheEntryTy &RHS) const {
-      return LHS.first < RHS.first;
-    }
-  };
-}
-
-//===----------------------------------------------------------------------===//
-//                              LVIQuery Impl
-//===----------------------------------------------------------------------===//
-
-namespace {
-  /// LVIQuery - This is a transient object that exists while a query is
-  /// being performed.
-  ///
-  /// TODO: Reuse LVIQuery instead of recreating it for every query, this avoids
-  /// reallocation of the densemap on every query.
-  class LVIQuery {
-    typedef LazyValueInfoCache::BlockCacheEntryTy BlockCacheEntryTy;
-    typedef LazyValueInfoCache::ValueCacheEntryTy ValueCacheEntryTy;
-    
-    /// This is the current value being queried for.
-    Value *Val;
-    
-    /// This is all of the cached information about this value.
-    ValueCacheEntryTy &Cache;
-    
-    /// This tracks, for each block, what values are overdefined.
-    DenseSet<std::pair<BasicBlock*, Value*> > &OverDefinedCache;
-    
-    ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
-    /// added to cache but that are not in sorted order.
-    DenseSet<BasicBlock*> NewBlockInfo;
-  public:
-    
-    LVIQuery(Value *V, ValueCacheEntryTy &VC,
-             DenseSet<std::pair<BasicBlock*, Value*> > &ODC)
-      : Val(V), Cache(VC), OverDefinedCache(ODC) {
-    }
-
-    ~LVIQuery() {
-      // When the query is done, insert the newly discovered facts into the
-      // cache in sorted order.
-      if (NewBlockInfo.empty()) return;
-      
-      for (DenseSet<BasicBlock*>::iterator I = NewBlockInfo.begin(),
-           E = NewBlockInfo.end(); I != E; ++I) {
-        if (Cache[*I].isOverdefined())
-          OverDefinedCache.insert(std::make_pair(*I, Val));
-      }
-    }
-
-    LVILatticeVal getBlockValue(BasicBlock *BB);
-    LVILatticeVal getEdgeValue(BasicBlock *FromBB, BasicBlock *ToBB);
-
-  private:
-    LVILatticeVal &getCachedEntryForBlock(BasicBlock *BB);
-  };
-} // end anonymous namespace
 
 /// getCachedEntryForBlock - See if we already have a value for this block.  If
 /// so, return it, otherwise create a new entry in the Cache map to use.
-LVILatticeVal &LVIQuery::getCachedEntryForBlock(BasicBlock *BB) {
+LVILatticeVal& 
+LazyValueInfoCache::getCachedEntryForBlock(ValueCacheEntryTy &Cache,
+                                           BasicBlock *BB) {
   NewBlockInfo.insert(BB);
   return Cache[BB];
 }
 
-LVILatticeVal LVIQuery::getBlockValue(BasicBlock *BB) {
+LVILatticeVal
+LazyValueInfoCache::getBlockValue(ValueCacheEntryTy &Cache, BasicBlock *BB) {
   // See if we already have a value for this block.
-  LVILatticeVal &BBLV = getCachedEntryForBlock(BB);
+  LVILatticeVal &BBLV = getCachedEntryForBlock(Cache, BB);
   
   // If we've already computed this block's value, return it.
   if (!BBLV.isUndefined()) {
@@ -345,7 +321,7 @@
     // Loop over all of our predecessors, merging what we know from them into
     // result.
     for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
-      Result.mergeIn(getEdgeValue(*PI, BB));
+      Result.mergeIn(getEdgeValue(Cache, *PI, BB));
       
       // If we hit overdefined, exit early.  The BlockVals entry is already set
       // to overdefined.
@@ -367,7 +343,7 @@
     
     // Return the merged value, which is more precise than 'overdefined'.
     assert(!Result.isOverdefined());
-    return getCachedEntryForBlock(BB) = Result;
+    return getCachedEntryForBlock(Cache, BB) = Result;
   }
   
   // If this value is defined by an instruction in this block, we have to
@@ -384,12 +360,13 @@
 
   LVILatticeVal Result;
   Result.markOverdefined();
-  return getCachedEntryForBlock(BB) = Result;
+  return getCachedEntryForBlock(Cache, BB) = Result;
 }
 
 
 /// getEdgeValue - This method attempts to infer more complex 
-LVILatticeVal LVIQuery::getEdgeValue(BasicBlock *BBFrom, BasicBlock *BBTo) {
+LVILatticeVal LazyValueInfoCache::getEdgeValue(ValueCacheEntryTy &Cache, 
+                                         BasicBlock *BBFrom, BasicBlock *BBTo) {
   // TODO: Handle more complex conditionals.  If (v == 0 || v2 < 1) is false, we
   // know that v != 0.
   if (BranchInst *BI = dyn_cast<BranchInst>(BBFrom->getTerminator())) {
@@ -443,14 +420,9 @@
   }
   
   // Otherwise see if the value is known in the block.
-  return getBlockValue(BBFrom);
+  return getBlockValue(Cache, BBFrom);
 }
 
-
-//===----------------------------------------------------------------------===//
-//                         LazyValueInfoCache Impl
-//===----------------------------------------------------------------------===//
-
 LVILatticeVal LazyValueInfoCache::getValueInBlock(Value *V, BasicBlock *BB) {
   // If already a constant, there is nothing to compute.
   if (Constant *VC = dyn_cast<Constant>(V))
@@ -459,8 +431,9 @@
   DEBUG(dbgs() << "LVI Getting block end value " << *V << " at '"
         << BB->getName() << "'\n");
   
-  LVILatticeVal Result = LVIQuery(V, ValueCache[V], 
-                                  OverDefinedCache).getBlockValue(BB);
+  QuerySetup QS(*this, V);
+  ValueCacheEntryTy &Cache = ValueCache[V];
+  LVILatticeVal Result = getBlockValue(Cache, BB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   return Result;
@@ -475,9 +448,9 @@
   DEBUG(dbgs() << "LVI Getting edge value " << *V << " from '"
         << FromBB->getName() << "' to '" << ToBB->getName() << "'\n");
   
-  LVILatticeVal Result =
-    LVIQuery(V, ValueCache[V],
-             OverDefinedCache).getEdgeValue(FromBB, ToBB);
+  QuerySetup QS(*this, V);
+  ValueCacheEntryTy &Cache = ValueCache[V];
+  LVILatticeVal Result = getEdgeValue(Cache, FromBB, ToBB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   




From eli.friedman at gmail.com  Wed Jul 28 17:25:12 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Wed, 28 Jul 2010 15:25:12 -0700
Subject: [llvm-commits] [llvm] r109643 - in /llvm/trunk:
	include/llvm/Instructions.h lib/AsmParser/LLParser.cpp
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	lib/VMCore/Instructions.cpp test/Assembler/align-inst-alloca.ll
	test/Assembler/align-inst-load.
Message-ID: <AANLkTik08Eq95G373QgKs+Ytu-gNhVB+JXN1-ngVnxko@mail.gmail.com>

On Wed, Jul 28, 2010 at 1:12 PM, Dan Gohman <gohman at apple.com> wrote:
> @@ -104,11 +104,15 @@
> ? ComputeMaskedBits(V, Mask, KnownZero, KnownOne);
> ? unsigned TrailZ = KnownZero.countTrailingOnes();
>
> - ?// LLVM doesn't support alignments larger than this currently.
> + ?// Avoid trouble with rediculously large TrailZ values, such as
> + ?// those computed from a null pointer.
> ? TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));

Spelling.

-Eli



From resistor at mac.com  Wed Jul 28 17:30:53 2010
From: resistor at mac.com (Owen Anderson)
Date: Wed, 28 Jul 2010 22:30:53 -0000
Subject: [llvm-commits] [llvm] r109683 -
	/llvm/trunk/include/llvm/System/ThreadLocal.h
Message-ID: <20100728223053.DC7E22A6C12C@llvm.org>

Author: resistor
Date: Wed Jul 28 17:30:53 2010
New Revision: 109683

URL: http://llvm.org/viewvc/llvm-project?rev=109683&view=rev
Log:
Add more doxygen comments for llvm::ThreadLocal.

Modified:
    llvm/trunk/include/llvm/System/ThreadLocal.h

Modified: llvm/trunk/include/llvm/System/ThreadLocal.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/ThreadLocal.h?rev=109683&r1=109682&r2=109683&view=diff
==============================================================================
--- llvm/trunk/include/llvm/System/ThreadLocal.h (original)
+++ llvm/trunk/include/llvm/System/ThreadLocal.h Wed Jul 28 17:30:53 2010
@@ -19,6 +19,8 @@
 
 namespace llvm {
   namespace sys {
+    // ThreadLocalImpl - Common base class of all ThreadLocal instantiations.
+    // YOU SHOULD NEVER USE THIS DIRECTLY.
     class ThreadLocalImpl {
       void* data;
     public:
@@ -28,11 +30,18 @@
       const void* getInstance();
     };
     
+    /// ThreadLocal - A class used to abstract thread-local storage.  It holds,
+    /// for each thread, a pointer a single object of type T.
     template<class T>
     class ThreadLocal : public ThreadLocalImpl {
     public:
       ThreadLocal() : ThreadLocalImpl() { }
+      
+      /// get - Fetches a pointer to the object associated with the current
+      /// thread.  If no object has yet been associated, it returns NULL;
       T* get() { return static_cast<T*>(getInstance()); }
+      
+      // set - Associates a pointer to an object with the current thread.
       void set(T* d) { setInstance(d); }
     };
   }




From stuart at apple.com  Wed Jul 28 17:39:07 2010
From: stuart at apple.com (Stuart Hastings)
Date: Wed, 28 Jul 2010 22:39:07 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109684 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
Message-ID: <20100728223907.CB9D52A6C12C@llvm.org>

Author: stuart
Date: Wed Jul 28 17:39:07 2010
New Revision: 109684

URL: http://llvm.org/viewvc/llvm-project?rev=109684&view=rev
Log:
Revert r108986; it was breaking 2004-06-18-VariableLengthArrayOfStructures.c.  Radar 8239581.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp?rev=109684&r1=109683&r2=109684&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp Wed Jul 28 17:39:07 2010
@@ -820,8 +820,8 @@
   // every time we see a variable.
   if (SeenVLA &&
       GreatestAlignment > TheTarget->getFrameInfo()->getStackAlignment())
-      error ("alignment for %q+D conflicts with a dynamically realigned stack",
-             SeenVLA);
+      error ("alignment for %q+D conflicts with either a dynamically "
+             "realigned stack or the maximum stack alignment", SeenVLA);
 
   return Fn;
 }
@@ -1772,6 +1772,13 @@
       DECL_SIZE(decl) != 0 && TREE_CODE(DECL_SIZE_UNIT(decl)) != INTEGER_CST)
     SeenVLA = decl;
       
+  // If this is just the rotten husk of a variable that the gimplifier
+  // eliminated all uses of, but is preserving for debug info, ignore it.
+  // TODO: This affects the correctness of the warning we're attempting to
+  // watch above.
+  if (TREE_CODE(decl) == VAR_DECL && DECL_VALUE_EXPR(decl))
+    return;
+      
   // Gimple temporaries are handled specially: their DECL_LLVM is set when the
   // definition is encountered.
   if (isGimpleTemporary(decl))
@@ -1821,18 +1828,6 @@
       Alignment = DECL_ALIGN(decl) / 8;
   }
 
-  // Record the alignment if it's the largest we've seen and is explicitly
-  // requested.
-  if (DECL_USER_ALIGN(decl) && Alignment > GreatestAlignment)
-    GreatestAlignment = Alignment;
-
-  // If this is just the rotten husk of a variable that the gimplifier
-  // eliminated all uses of, but is preserving for debug info, ignore it.
-  // We're doing this late so that we can get the alignment information from
-  // the code.
-  if (TREE_CODE(decl) == VAR_DECL && DECL_VALUE_EXPR(decl))
-    return;
-      
   const char *Name;      // Name of variable
   if (DECL_NAME(decl))
     Name = IDENTIFIER_POINTER(DECL_NAME(decl));
@@ -1852,6 +1847,10 @@
 
   AI->setAlignment(Alignment);
   
+  // Record the alignment if it's the largest we've seen.
+  if (Alignment > GreatestAlignment)
+    GreatestAlignment = Alignment;
+
   SET_DECL_LLVM(decl, AI);
 
   // Handle annotate attributes




From echristo at apple.com  Wed Jul 28 17:42:01 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 15:42:01 -0700
Subject: [llvm-commits] [llvm-gcc-4.2] r109684 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
In-Reply-To: <20100728223907.CB9D52A6C12C@llvm.org>
References: <20100728223907.CB9D52A6C12C@llvm.org>
Message-ID: <E14C3E29-3BD9-4997-9775-BD1063BAE7CE@apple.com>


On Jul 28, 2010, at 3:39 PM, Stuart Hastings wrote:

> Author: stuart
> Date: Wed Jul 28 17:39:07 2010
> New Revision: 109684
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109684&view=rev
> Log:
> Revert r108986; it was breaking 2004-06-18-VariableLengthArrayOfStructures.c.  Radar 8239581.

Uh, how's it breaking it? My tests were clean.

-eric


From resistor at mac.com  Wed Jul 28 17:49:44 2010
From: resistor at mac.com (Owen Anderson)
Date: Wed, 28 Jul 2010 22:49:44 -0000
Subject: [llvm-commits] [llvm] r109686 - in /llvm/trunk:
 include/llvm/System/ThreadLocal.h lib/System/ThreadLocal.cpp
 lib/System/Unix/ThreadLocal.inc lib/System/Win32/ThreadLocal.inc
Message-ID: <20100728224944.209652A6C12C@llvm.org>

Author: resistor
Date: Wed Jul 28 17:49:43 2010
New Revision: 109686

URL: http://llvm.org/viewvc/llvm-project?rev=109686&view=rev
Log:
Add an erase() method to llvm::ThreadLocal.

Modified:
    llvm/trunk/include/llvm/System/ThreadLocal.h
    llvm/trunk/lib/System/ThreadLocal.cpp
    llvm/trunk/lib/System/Unix/ThreadLocal.inc
    llvm/trunk/lib/System/Win32/ThreadLocal.inc

Modified: llvm/trunk/include/llvm/System/ThreadLocal.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/ThreadLocal.h?rev=109686&r1=109685&r2=109686&view=diff
==============================================================================
--- llvm/trunk/include/llvm/System/ThreadLocal.h (original)
+++ llvm/trunk/include/llvm/System/ThreadLocal.h Wed Jul 28 17:49:43 2010
@@ -28,6 +28,7 @@
       virtual ~ThreadLocalImpl();
       void setInstance(const void* d);
       const void* getInstance();
+      void removeInstance();
     };
     
     /// ThreadLocal - A class used to abstract thread-local storage.  It holds,
@@ -43,6 +44,9 @@
       
       // set - Associates a pointer to an object with the current thread.
       void set(T* d) { setInstance(d); }
+      
+      // erase - Removes the pointer associated with the current thread.
+      void erase() { removeInstance(); }
     };
   }
 }

Modified: llvm/trunk/lib/System/ThreadLocal.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/ThreadLocal.cpp?rev=109686&r1=109685&r2=109686&view=diff
==============================================================================
--- llvm/trunk/lib/System/ThreadLocal.cpp (original)
+++ llvm/trunk/lib/System/ThreadLocal.cpp Wed Jul 28 17:49:43 2010
@@ -67,6 +67,10 @@
   return pthread_getspecific(*key);
 }
 
+void ThreadLocalImpl::removeInstance() {
+  setInstance(0);
+}
+
 }
 
 #elif defined(LLVM_ON_UNIX)

Modified: llvm/trunk/lib/System/Unix/ThreadLocal.inc
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Unix/ThreadLocal.inc?rev=109686&r1=109685&r2=109686&view=diff
==============================================================================
--- llvm/trunk/lib/System/Unix/ThreadLocal.inc (original)
+++ llvm/trunk/lib/System/Unix/ThreadLocal.inc Wed Jul 28 17:49:43 2010
@@ -22,4 +22,5 @@
 ThreadLocalImpl::~ThreadLocalImpl() { }
 void ThreadLocalImpl::setInstance(const void* d) { data = const_cast<void*>(d);}
 const void* ThreadLocalImpl::getInstance() { return data; }
+void ThreadLocalImpl::removeInstance() { setInstance(0); }
 }

Modified: llvm/trunk/lib/System/Win32/ThreadLocal.inc
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Win32/ThreadLocal.inc?rev=109686&r1=109685&r2=109686&view=diff
==============================================================================
--- llvm/trunk/lib/System/Win32/ThreadLocal.inc (original)
+++ llvm/trunk/lib/System/Win32/ThreadLocal.inc Wed Jul 28 17:49:43 2010
@@ -46,4 +46,8 @@
   assert(errorcode != 0);
 }
 
+void ThreadLocalImpl::removeInstance() {
+  setInstance(0);
+}
+
 }




From ggreif at gmail.com  Wed Jul 28 17:50:26 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Wed, 28 Jul 2010 22:50:26 -0000
Subject: [llvm-commits] [llvm] r109687 - in /llvm/trunk:
 lib/Transforms/IPO/ArgumentPromotion.cpp
 lib/Transforms/IPO/DeadArgumentElimination.cpp
 lib/Transforms/IPO/FunctionAttrs.cpp
 lib/Transforms/IPO/IPConstantPropagation.cpp lib/Transforms/IPO/Inliner.cpp
 lib/Transforms/IPO/StructRetPromotion.cpp
 lib/Transforms/Scalar/MemCpyOptimizer.cpp lib/Transforms/Scalar/SCCP.cpp
 tools/opt/AnalysisWrappers.cpp
Message-ID: <20100728225026.CFBD42A6C12C@llvm.org>

Author: ggreif
Date: Wed Jul 28 17:50:26 2010
New Revision: 109687

URL: http://llvm.org/viewvc/llvm-project?rev=109687&view=rev
Log:
simplify by using CallSite constructors; virtually eliminates CallSite::get from the tree

Modified:
    llvm/trunk/lib/Transforms/IPO/ArgumentPromotion.cpp
    llvm/trunk/lib/Transforms/IPO/DeadArgumentElimination.cpp
    llvm/trunk/lib/Transforms/IPO/FunctionAttrs.cpp
    llvm/trunk/lib/Transforms/IPO/IPConstantPropagation.cpp
    llvm/trunk/lib/Transforms/IPO/Inliner.cpp
    llvm/trunk/lib/Transforms/IPO/StructRetPromotion.cpp
    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
    llvm/trunk/lib/Transforms/Scalar/SCCP.cpp
    llvm/trunk/tools/opt/AnalysisWrappers.cpp

Modified: llvm/trunk/lib/Transforms/IPO/ArgumentPromotion.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/ArgumentPromotion.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/ArgumentPromotion.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/ArgumentPromotion.cpp Wed Jul 28 17:50:26 2010
@@ -208,8 +208,8 @@
   // have direct callees.
   for (Value::use_iterator UI = Callee->use_begin(), E = Callee->use_end();
        UI != E; ++UI) {
-    CallSite CS = CallSite::get(*UI);
-    assert(CS.getInstruction() && "Should only have direct calls!");
+    CallSite CS(*UI);
+    assert(CS && "Should only have direct calls!");
 
     if (!IsAlwaysValidPointer(CS.getArgument(ArgNo)))
       return false;
@@ -619,14 +619,13 @@
   
   // Get a new callgraph node for NF.
   CallGraphNode *NF_CGN = CG.getOrInsertFunction(NF);
-  
 
   // Loop over all of the callers of the function, transforming the call sites
   // to pass in the loaded pointers.
   //
   SmallVector<Value*, 16> Args;
   while (!F->use_empty()) {
-    CallSite CS = CallSite::get(F->use_back());
+    CallSite CS(F->use_back());
     assert(CS.getCalledFunction() == F);
     Instruction *Call = CS.getInstruction();
     const AttrListPtr &CallPAL = CS.getAttributes();

Modified: llvm/trunk/lib/Transforms/IPO/DeadArgumentElimination.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/DeadArgumentElimination.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/DeadArgumentElimination.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/DeadArgumentElimination.cpp Wed Jul 28 17:50:26 2010
@@ -220,11 +220,11 @@
   //
   std::vector<Value*> Args;
   while (!Fn.use_empty()) {
-    CallSite CS = CallSite::get(Fn.use_back());
+    CallSite CS(Fn.use_back());
     Instruction *Call = CS.getInstruction();
 
     // Pass all the same arguments.
-    Args.assign(CS.arg_begin(), CS.arg_begin()+NumArgs);
+    Args.assign(CS.arg_begin(), CS.arg_begin() + NumArgs);
 
     // Drop any attributes that were on the vararg arguments.
     AttrListPtr PAL = CS.getAttributes();
@@ -724,7 +724,7 @@
   //
   std::vector<Value*> Args;
   while (!F->use_empty()) {
-    CallSite CS = CallSite::get(F->use_back());
+    CallSite CS(F->use_back());
     Instruction *Call = CS.getInstruction();
 
     AttributesVec.clear();

Modified: llvm/trunk/lib/Transforms/IPO/FunctionAttrs.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/FunctionAttrs.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/FunctionAttrs.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/FunctionAttrs.cpp Wed Jul 28 17:50:26 2010
@@ -162,8 +162,8 @@
 
       // Some instructions can be ignored even if they read or write memory.
       // Detect these now, skipping to the next instruction if one is found.
-      CallSite CS = CallSite::get(I);
-      if (CS.getInstruction() && CS.getCalledFunction()) {
+      CallSite CS(cast<Value>(I));
+      if (CS && CS.getCalledFunction()) {
         // Ignore calls to functions in the same SCC.
         if (SCCNodes.count(CS.getCalledFunction()))
           continue;

Modified: llvm/trunk/lib/Transforms/IPO/IPConstantPropagation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/IPConstantPropagation.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/IPConstantPropagation.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/IPConstantPropagation.cpp Wed Jul 28 17:50:26 2010
@@ -94,7 +94,7 @@
     if (!isa<CallInst>(U) && !isa<InvokeInst>(U))
       return false;
     
-    CallSite CS = CallSite::get(cast<Instruction>(U));
+    CallSite CS(cast<Instruction>(U));
     if (!CS.isCallee(UI))
       return false;
 
@@ -219,7 +219,7 @@
   // constant.
   bool MadeChange = false;
   for (Value::use_iterator UI = F.use_begin(), E = F.use_end(); UI != E; ++UI) {
-    CallSite CS = CallSite::get(*UI);
+    CallSite CS(*UI);
     Instruction* Call = CS.getInstruction();
 
     // Not a call instruction or a call instruction that's not calling F

Modified: llvm/trunk/lib/Transforms/IPO/Inliner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/Inliner.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/Inliner.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/Inliner.cpp Wed Jul 28 17:50:26 2010
@@ -238,11 +238,11 @@
     bool someOuterCallWouldNotBeInlined = false;
     for (Value::use_iterator I = Caller->use_begin(), E =Caller->use_end(); 
          I != E; ++I) {
-      CallSite CS2 = CallSite::get(*I);
+      CallSite CS2(*I);
 
       // If this isn't a call to Caller (it could be some other sort
       // of reference) skip it.
-      if (CS2.getInstruction() == 0 || CS2.getCalledFunction() != Caller)
+      if (!CS2 || CS2.getCalledFunction() != Caller)
         continue;
 
       InlineCost IC2 = getInlineCost(CS2);
@@ -334,10 +334,10 @@
     
     for (Function::iterator BB = F->begin(), E = F->end(); BB != E; ++BB)
       for (BasicBlock::iterator I = BB->begin(), E = BB->end(); I != E; ++I) {
-        CallSite CS = CallSite::get(I);
+        CallSite CS(cast<Value>(I));
         // If this isn't a call, or it is a call to an intrinsic, it can
         // never be inlined.
-        if (CS.getInstruction() == 0 || isa<IntrinsicInst>(I))
+        if (!CS || isa<IntrinsicInst>(I))
           continue;
         
         // If this is a direct call to an external function, we can never inline

Modified: llvm/trunk/lib/Transforms/IPO/StructRetPromotion.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/StructRetPromotion.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/StructRetPromotion.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/StructRetPromotion.cpp Wed Jul 28 17:50:26 2010
@@ -156,7 +156,7 @@
        FnUseI != FnUseE; ++FnUseI) {
     // The function is passed in as an argument to (possibly) another function,
     // we can't change it!
-    CallSite CS = CallSite::get(*FnUseI);
+    CallSite CS(*FnUseI);
     Instruction *Call = CS.getInstruction();
     // The function is used by something else than a call or invoke instruction,
     // we can't change it!
@@ -271,7 +271,7 @@
   CallGraphNode *NF_CGN = CG.getOrInsertFunction(NF);
 
   while (!F->use_empty()) {
-    CallSite CS = CallSite::get(*F->use_begin());
+    CallSite CS(*F->use_begin());
     Instruction *Call = CS.getInstruction();
 
     const AttrListPtr &PAL = F->getAttributes();

Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Wed Jul 28 17:50:26 2010
@@ -508,7 +508,7 @@
   // because we'll need to do type comparisons based on the underlying type.
   Value *cpyDest = cpy->getDest();
   Value *cpySrc = cpy->getSource();
-  CallSite CS = CallSite::get(C);
+  CallSite CS(C);
 
   // We need to be able to reason about the size of the memcpy, so we require
   // that it be a constant.
@@ -636,10 +636,11 @@
   return true;
 }
 
-/// processMemCpy - perform simplication of memcpy's.  If we have memcpy A which
-/// copies X to Y, and memcpy B which copies Y to Z, then we can rewrite B to be
-/// a memcpy from X to Z (or potentially a memmove, depending on circumstances).
-///  This allows later passes to remove the first memcpy altogether.
+/// processMemCpy - perform simplification of memcpy's.  If we have memcpy A
+/// which copies X to Y, and memcpy B which copies Y to Z, then we can rewrite
+/// B to be a memcpy from X to Z (or potentially a memmove, depending on
+/// circumstances). This allows later passes to remove the first memcpy
+/// altogether.
 bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
   MemoryDependenceAnalysis &MD = getAnalysis<MemoryDependenceAnalysis>();
 

Modified: llvm/trunk/lib/Transforms/Scalar/SCCP.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/SCCP.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/SCCP.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/SCCP.cpp Wed Jul 28 17:50:26 2010
@@ -508,10 +508,10 @@
   void visitLoadInst      (LoadInst &I);
   void visitGetElementPtrInst(GetElementPtrInst &I);
   void visitCallInst      (CallInst &I) {
-    visitCallSite(CallSite::get(&I));
+    visitCallSite(&I);
   }
   void visitInvokeInst    (InvokeInst &II) {
-    visitCallSite(CallSite::get(&II));
+    visitCallSite(&II);
     visitTerminatorInst(II);
   }
   void visitCallSite      (CallSite CS);

Modified: llvm/trunk/tools/opt/AnalysisWrappers.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/opt/AnalysisWrappers.cpp?rev=109687&r1=109686&r2=109687&view=diff
==============================================================================
--- llvm/trunk/tools/opt/AnalysisWrappers.cpp (original)
+++ llvm/trunk/tools/opt/AnalysisWrappers.cpp Wed Jul 28 17:50:26 2010
@@ -42,8 +42,8 @@
           Instruction *User = dyn_cast<Instruction>(*UI);
           if (!User) continue;
           
-          CallSite CS = CallSite::get(User);
-          if (!CS.getInstruction()) continue;
+          CallSite CS(cast<Value>(User));
+          if (!CS) continue;
           
           for (CallSite::arg_iterator AI = CS.arg_begin(),
                E = CS.arg_end(); AI != E; ++AI) {




From echristo at apple.com  Wed Jul 28 17:55:35 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 15:55:35 -0700
Subject: [llvm-commits] [llvm-gcc-4.2] r109684 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
In-Reply-To: <E14C3E29-3BD9-4997-9775-BD1063BAE7CE@apple.com>
References: <20100728223907.CB9D52A6C12C@llvm.org>
	<E14C3E29-3BD9-4997-9775-BD1063BAE7CE@apple.com>
Message-ID: <7F927F05-469D-4694-86DE-54735E6DD4C5@apple.com>


On Jul 28, 2010, at 3:42 PM, Eric Christopher wrote:

> 
> On Jul 28, 2010, at 3:39 PM, Stuart Hastings wrote:
> 
>> Author: stuart
>> Date: Wed Jul 28 17:39:07 2010
>> New Revision: 109684
>> 
>> URL: http://llvm.org/viewvc/llvm-project?rev=109684&view=rev
>> Log:
>> Revert r108986; it was breaking 2004-06-18-VariableLengthArrayOfStructures.c.  Radar 8239581.
> 
> Uh, how's it breaking it? My tests were clean.

*looks at the radar*  Aah, it's probably the "we can't handle that" part of the converter.  Just needs another check, I'll add a check in there.

-eric


From grosbach at apple.com  Wed Jul 28 18:16:12 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 28 Jul 2010 23:16:12 -0000
Subject: [llvm-commits] [llvm] r109691 -
	/llvm/trunk/lib/Target/ARM/ARMISelLowering.h
Message-ID: <20100728231612.BFAA12A6C12C@llvm.org>

Author: grosbach
Date: Wed Jul 28 18:16:12 2010
New Revision: 109691

URL: http://llvm.org/viewvc/llvm-project?rev=109691&view=rev
Log:
Remove dead prototype

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

Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.h?rev=109691&r1=109690&r2=109691&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.h (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.h Wed Jul 28 18:16:12 2010
@@ -340,7 +340,6 @@
     SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerRETURNADDR(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const;
-    SDValue LowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerShiftRightParts(SDValue Op, SelectionDAG &DAG) const;
     SDValue LowerShiftLeftParts(SDValue Op, SelectionDAG &DAG) const;
 




From echristo at apple.com  Wed Jul 28 18:16:18 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 23:16:18 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109692 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
Message-ID: <20100728231619.02B9E2A6C12D@llvm.org>

Author: echristo
Date: Wed Jul 28 18:16:18 2010
New Revision: 109692

URL: http://llvm.org/viewvc/llvm-project?rev=109692&view=rev
Log:
Revert 109684 and add a patch to make sure we're only computing the type
for the alignment for var decls we're not going to use anyways.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp?rev=109692&r1=109691&r2=109692&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp Wed Jul 28 18:16:18 2010
@@ -820,8 +820,8 @@
   // every time we see a variable.
   if (SeenVLA &&
       GreatestAlignment > TheTarget->getFrameInfo()->getStackAlignment())
-      error ("alignment for %q+D conflicts with either a dynamically "
-             "realigned stack or the maximum stack alignment", SeenVLA);
+      error ("alignment for %q+D conflicts with a dynamically realigned stack",
+             SeenVLA);
 
   return Fn;
 }
@@ -1772,13 +1772,6 @@
       DECL_SIZE(decl) != 0 && TREE_CODE(DECL_SIZE_UNIT(decl)) != INTEGER_CST)
     SeenVLA = decl;
       
-  // If this is just the rotten husk of a variable that the gimplifier
-  // eliminated all uses of, but is preserving for debug info, ignore it.
-  // TODO: This affects the correctness of the warning we're attempting to
-  // watch above.
-  if (TREE_CODE(decl) == VAR_DECL && DECL_VALUE_EXPR(decl))
-    return;
-      
   // Gimple temporaries are handled specially: their DECL_LLVM is set when the
   // definition is encountered.
   if (isGimpleTemporary(decl))
@@ -1795,8 +1788,11 @@
       TODO(decl);
       abort();
     }
-  } else if (TREE_CODE(DECL_SIZE_UNIT(decl)) == INTEGER_CST) {
-    // Variable of fixed size that goes on the stack.
+  } else if (TREE_CODE(DECL_SIZE_UNIT(decl)) == INTEGER_CST
+             || (TREE_CODE(decl) == VAR_DECL && DECL_VALUE_EXPR(decl))) {
+    // Variable of fixed size that goes on the stack or the type for a
+    // variable we're not going to emit anyways, but need later for alignment
+    // calculations.
     Ty = ConvertType(type);
   } else {
     // Dynamic-size object: must push space on the stack.
@@ -1828,6 +1824,18 @@
       Alignment = DECL_ALIGN(decl) / 8;
   }
 
+  // Record the alignment if it's the largest we've seen and is explicitly
+  // requested.
+  if (DECL_USER_ALIGN(decl) && Alignment > GreatestAlignment)
+    GreatestAlignment = Alignment;
+
+  // If this is just the rotten husk of a variable that the gimplifier
+  // eliminated all uses of, but is preserving for debug info, ignore it.
+  // We're doing this late so that we can get the alignment information from
+  // the code.
+  if (TREE_CODE(decl) == VAR_DECL && DECL_VALUE_EXPR(decl))
+    return;
+      
   const char *Name;      // Name of variable
   if (DECL_NAME(decl))
     Name = IDENTIFIER_POINTER(DECL_NAME(decl));
@@ -1847,10 +1855,6 @@
 
   AI->setAlignment(Alignment);
   
-  // Record the alignment if it's the largest we've seen.
-  if (Alignment > GreatestAlignment)
-    GreatestAlignment = Alignment;
-
   SET_DECL_LLVM(decl, AI);
 
   // Handle annotate attributes




From grosbach at apple.com  Wed Jul 28 18:17:45 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 28 Jul 2010 23:17:45 -0000
Subject: [llvm-commits] [llvm] r109693 -
	/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
Message-ID: <20100728231745.9EDEA2A6C12C@llvm.org>

Author: grosbach
Date: Wed Jul 28 18:17:45 2010
New Revision: 109693

URL: http://llvm.org/viewvc/llvm-project?rev=109693&view=rev
Log:
Remove incorrect substitution pattern for UXTB16. It wrongly assumed the input shift was actually a rotate. rdar://8240138

Modified:
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=109693&r1=109692&r2=109693&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Wed Jul 28 18:17:45 2010
@@ -1352,8 +1352,12 @@
 defm t2UXTB16 : T2I_unary_rrot_uxtb16<0b011, "uxtb16",
                                UnOpFrag<(and node:$Src, 0x00FF00FF)>>;
 
-def : T2Pat<(and (shl GPR:$Src, (i32 8)), 0xFF00FF),
-            (t2UXTB16r_rot GPR:$Src, 24)>, Requires<[HasT2ExtractPack]>;
+// FIXME: This pattern incorrectly assumes the shl operator is a rotate.
+//        The transformation should probably be done as a combiner action
+//        instead so we can include a check for masking back in the upper
+//        eight bits of the source into the lower eight bits of the result.
+//def : T2Pat<(and (shl GPR:$Src, (i32 8)), 0xFF00FF),
+//            (t2UXTB16r_rot GPR:$Src, 24)>, Requires<[HasT2ExtractPack]>;
 def : T2Pat<(and (srl GPR:$Src, (i32 8)), 0xFF00FF),
             (t2UXTB16r_rot GPR:$Src, 8)>, Requires<[HasT2ExtractPack]>;
 




From echristo at apple.com  Wed Jul 28 18:18:11 2010
From: echristo at apple.com (Eric Christopher)
Date: Wed, 28 Jul 2010 16:18:11 -0700
Subject: [llvm-commits] [llvm-gcc-4.2] r109684 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
In-Reply-To: <7F927F05-469D-4694-86DE-54735E6DD4C5@apple.com>
References: <20100728223907.CB9D52A6C12C@llvm.org>
	<E14C3E29-3BD9-4997-9775-BD1063BAE7CE@apple.com>
	<7F927F05-469D-4694-86DE-54735E6DD4C5@apple.com>
Message-ID: <065ABD22-7334-47AD-9993-50616EA54546@apple.com>


On Jul 28, 2010, at 3:55 PM, Eric Christopher wrote:

> 
> On Jul 28, 2010, at 3:42 PM, Eric Christopher wrote:
> 
>> 
>> On Jul 28, 2010, at 3:39 PM, Stuart Hastings wrote:
>> 
>>> Author: stuart
>>> Date: Wed Jul 28 17:39:07 2010
>>> New Revision: 109684
>>> 
>>> URL: http://llvm.org/viewvc/llvm-project?rev=109684&view=rev
>>> Log:
>>> Revert r108986; it was breaking 2004-06-18-VariableLengthArrayOfStructures.c.  Radar 8239581.
>> 
>> Uh, how's it breaking it? My tests were clean.
> 
> *looks at the radar*  Aah, it's probably the "we can't handle that" part of the converter.  Just needs another check, I'll add a check in there.

Fixed in: 109692

Tested using an --enable-checking llvm-gcc. Clean run.

-eric


From grosbach at apple.com  Wed Jul 28 18:25:44 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Wed, 28 Jul 2010 23:25:44 -0000
Subject: [llvm-commits] [llvm] r109696 -
	/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
Message-ID: <20100728232544.68A082A6C12C@llvm.org>

Author: grosbach
Date: Wed Jul 28 18:25:44 2010
New Revision: 109696

URL: http://llvm.org/viewvc/llvm-project?rev=109696&view=rev
Log:
ARM mode version of r109693. Remove incorrect substitution pattern for UXTB16. It wrongly assumed the input shift was actually a rotate. rdar://8240138

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

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=109696&r1=109695&r2=109696&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Wed Jul 28 18:25:44 2010
@@ -1567,8 +1567,12 @@
 defm UXTB16 : AI_unary_rrot<0b01101100,
                             "uxtb16", UnOpFrag<(and node:$Src, 0x00FF00FF)>>;
 
-def : ARMV6Pat<(and (shl GPR:$Src, (i32 8)), 0xFF00FF),
-               (UXTB16r_rot GPR:$Src, 24)>;
+// FIXME: This pattern incorrectly assumes the shl operator is a rotate.
+//        The transformation should probably be done as a combiner action
+//        instead so we can include a check for masking back in the upper
+//        eight bits of the source into the lower eight bits of the result.
+//def : ARMV6Pat<(and (shl GPR:$Src, (i32 8)), 0xFF00FF),
+//               (UXTB16r_rot GPR:$Src, 24)>;
 def : ARMV6Pat<(and (srl GPR:$Src, (i32 8)), 0xFF00FF),
                (UXTB16r_rot GPR:$Src, 8)>;
 




From resistor at mac.com  Wed Jul 28 18:50:08 2010
From: resistor at mac.com (Owen Anderson)
Date: Wed, 28 Jul 2010 23:50:08 -0000
Subject: [llvm-commits] [llvm] r109701 -
	/llvm/trunk/lib/Analysis/LazyValueInfo.cpp
Message-ID: <20100728235008.C93B82A6C12C@llvm.org>

Author: resistor
Date: Wed Jul 28 18:50:08 2010
New Revision: 109701

URL: http://llvm.org/viewvc/llvm-project?rev=109701&view=rev
Log:
Pass the queried value by argument rather than in a member, in preparation for supporting PHI translation.

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

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109701&r1=109700&r2=109701&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Wed Jul 28 18:50:08 2010
@@ -223,53 +223,43 @@
     /// for cache updating.
     DenseSet<std::pair<BasicBlock*, Value*> > OverDefinedCache;
     
-    LVILatticeVal getBlockValue(ValueCacheEntryTy &Cache, BasicBlock *BB);
-    LVILatticeVal getEdgeValue(ValueCacheEntryTy &Cache,
-                               BasicBlock *Pred, BasicBlock *Succ);
-    LVILatticeVal &getCachedEntryForBlock(ValueCacheEntryTy &Cache,
+    LVILatticeVal getBlockValue(Value *Val, BasicBlock *BB);
+    LVILatticeVal getEdgeValue(Value *Val, BasicBlock *Pred, BasicBlock *Succ);
+    LVILatticeVal &getCachedEntryForBlock(Value *Val, ValueCacheEntryTy &Cache,
                                           BasicBlock *BB);
     
     /************* Begin Per-Query State *************/
-    /// This is the current value being queried for.
-    Value *Val;
-    
-    /// This is all of the cached information about this value.
-    //ValueCacheEntryTy *Cache; 
     
     ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
     /// added to cache but that are not in sorted order.
-    DenseSet<BasicBlock*> NewBlockInfo;
+    DenseSet<std::pair<BasicBlock*,Value*> > NewBlockInfo;
     
     /// QuerySetup - An RAII helper to construct and tear-down per-query 
     /// temporary state.
     struct QuerySetup {
       LazyValueInfoCache &Owner;
-      QuerySetup(LazyValueInfoCache &O, Value* Val) : Owner(O) {
-        assert(!Owner.Val && "Per-query info not cleared?");
-        Owner.Val = Val;
+      QuerySetup(LazyValueInfoCache &O) : Owner(O) {
         assert(Owner.NewBlockInfo.empty() && "Leaked block info!");
       }
       
       ~QuerySetup() {
         // When the query is done, insert the newly discovered facts into the
         // cache in sorted order.
-        LazyValueInfoCache::ValueCacheEntryTy Cache = 
-          Owner.ValueCache[Owner.Val];
-        for (DenseSet<BasicBlock*>::iterator I = Owner.NewBlockInfo.begin(),
-             E = Owner.NewBlockInfo.end(); I != E; ++I) {
-          if (Cache[*I].isOverdefined())
-            Owner.OverDefinedCache.insert(std::make_pair(*I, Owner.Val));
+        for (DenseSet<std::pair<BasicBlock*,Value*> >::iterator
+             I = Owner.NewBlockInfo.begin(), E = Owner.NewBlockInfo.end();
+             I != E; ++I) {
+          if (Owner.ValueCache[I->second][I->first].isOverdefined())
+            Owner.OverDefinedCache.insert(*I);
         }
         
         // Reset Per-Query State
-        Owner.Val = 0;
         Owner.NewBlockInfo.clear();
       }
     };
     /************* End Per-Query State *************/
     
   public:
-    LazyValueInfoCache() : Val(0) { }
+    LazyValueInfoCache() { }
     
     /// getValueInBlock - This is the query interface to determine the lattice
     /// value for the specified Value* at the end of the specified block.
@@ -290,16 +280,17 @@
 /// getCachedEntryForBlock - See if we already have a value for this block.  If
 /// so, return it, otherwise create a new entry in the Cache map to use.
 LVILatticeVal& 
-LazyValueInfoCache::getCachedEntryForBlock(ValueCacheEntryTy &Cache,
+LazyValueInfoCache::getCachedEntryForBlock(Value *Val, ValueCacheEntryTy &Cache,
                                            BasicBlock *BB) {
-  NewBlockInfo.insert(BB);
+  NewBlockInfo.insert(std::make_pair(BB, Val));
   return Cache[BB];
 }
 
 LVILatticeVal
-LazyValueInfoCache::getBlockValue(ValueCacheEntryTy &Cache, BasicBlock *BB) {
+LazyValueInfoCache::getBlockValue(Value *Val, BasicBlock *BB) {
   // See if we already have a value for this block.
-  LVILatticeVal &BBLV = getCachedEntryForBlock(Cache, BB);
+  ValueCacheEntryTy &Cache = ValueCache[Val];
+  LVILatticeVal &BBLV = getCachedEntryForBlock(Val, Cache, BB);
   
   // If we've already computed this block's value, return it.
   if (!BBLV.isUndefined()) {
@@ -321,7 +312,7 @@
     // Loop over all of our predecessors, merging what we know from them into
     // result.
     for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
-      Result.mergeIn(getEdgeValue(Cache, *PI, BB));
+      Result.mergeIn(getEdgeValue(Val, *PI, BB));
       
       // If we hit overdefined, exit early.  The BlockVals entry is already set
       // to overdefined.
@@ -343,7 +334,7 @@
     
     // Return the merged value, which is more precise than 'overdefined'.
     assert(!Result.isOverdefined());
-    return getCachedEntryForBlock(Cache, BB) = Result;
+    return getCachedEntryForBlock(Val, Cache, BB) = Result;
   }
   
   // If this value is defined by an instruction in this block, we have to
@@ -360,12 +351,12 @@
 
   LVILatticeVal Result;
   Result.markOverdefined();
-  return getCachedEntryForBlock(Cache, BB) = Result;
+  return getCachedEntryForBlock(Val, Cache, BB) = Result;
 }
 
 
 /// getEdgeValue - This method attempts to infer more complex 
-LVILatticeVal LazyValueInfoCache::getEdgeValue(ValueCacheEntryTy &Cache, 
+LVILatticeVal LazyValueInfoCache::getEdgeValue(Value *Val, 
                                          BasicBlock *BBFrom, BasicBlock *BBTo) {
   // TODO: Handle more complex conditionals.  If (v == 0 || v2 < 1) is false, we
   // know that v != 0.
@@ -420,7 +411,7 @@
   }
   
   // Otherwise see if the value is known in the block.
-  return getBlockValue(Cache, BBFrom);
+  return getBlockValue(Val, BBFrom);
 }
 
 LVILatticeVal LazyValueInfoCache::getValueInBlock(Value *V, BasicBlock *BB) {
@@ -431,9 +422,8 @@
   DEBUG(dbgs() << "LVI Getting block end value " << *V << " at '"
         << BB->getName() << "'\n");
   
-  QuerySetup QS(*this, V);
-  ValueCacheEntryTy &Cache = ValueCache[V];
-  LVILatticeVal Result = getBlockValue(Cache, BB);
+  QuerySetup QS(*this);
+  LVILatticeVal Result = getBlockValue(V, BB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   return Result;
@@ -448,9 +438,8 @@
   DEBUG(dbgs() << "LVI Getting edge value " << *V << " from '"
         << FromBB->getName() << "' to '" << ToBB->getName() << "'\n");
   
-  QuerySetup QS(*this, V);
-  ValueCacheEntryTy &Cache = ValueCache[V];
-  LVILatticeVal Result = getEdgeValue(Cache, FromBB, ToBB);
+  QuerySetup QS(*this);
+  LVILatticeVal Result = getEdgeValue(V, FromBB, ToBB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   




From isanbard at gmail.com  Wed Jul 28 18:54:15 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 28 Jul 2010 23:54:15 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109702 - in /llvm-gcc-4.2/trunk/gcc:
 c-common.h c-decl.c c-parser.c cp/parser.c testsuite/g++.apple/blockstret.C
 testsuite/gcc.apple/blockstret.c
Message-ID: <20100728235416.124B72A6C12C@llvm.org>

Author: void
Date: Wed Jul 28 18:54:15 2010
New Revision: 109702

URL: http://llvm.org/viewvc/llvm-project?rev=109702&view=rev
Log:
Title: BLOCKS: Need flag to indicate structure return semantics

This patch allows seamless integration of BLOCK_HAS_STRET in gcc/llvmgcc
while binaries produced by clang do the real work. libclosure will have the corresponding
changes to ignore the new fields.

Patch does:
1. Sets BLOCK_HAS_SIGNATURE always.
2. Include the `const char *signature` and `const char *layout` fields in the block descriptor.
3. Set signature=NULL and layout=NULL.

2010-07-28 Fariborz Jahanian <fjahanian at apple.com>

       Radar 8143947
       * c-decl.c (build_block_descriptor_type): Two new fields
       for 'struct __block_descriptor' type.
       * c-parser.c (build_descriptor_block_decl): Initialize new fields to 0.
       * c-common.h (BLOCK_HAS_SIGNATURE): New block flag.

       * parser.c (build_block_descriptor_type): Two new fields
       for 'struct __block_descriptor' type.
       (build_descriptor_block_decl): Initialize new fields to 0.

       gcc.apple/blockstret.c: New
       g++.apple/blockstret.C: New


Added:
    llvm-gcc-4.2/trunk/gcc/testsuite/g++.apple/blockstret.C
    llvm-gcc-4.2/trunk/gcc/testsuite/gcc.apple/blockstret.c
Modified:
    llvm-gcc-4.2/trunk/gcc/c-common.h
    llvm-gcc-4.2/trunk/gcc/c-decl.c
    llvm-gcc-4.2/trunk/gcc/c-parser.c
    llvm-gcc-4.2/trunk/gcc/cp/parser.c

Modified: llvm-gcc-4.2/trunk/gcc/c-common.h
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/c-common.h?rev=109702&r1=109701&r2=109702&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/c-common.h (original)
+++ llvm-gcc-4.2/trunk/gcc/c-common.h Wed Jul 28 18:54:15 2010
@@ -1157,7 +1157,9 @@
      /* APPLE LOCAL radar 5822844 */
      BLOCK_IS_GLOBAL = 	       (1 << 28),
      /* APPLE LOCAL radar 7735196 */
-     BLOCK_USE_STRET =    (1 << 29)
+     BLOCK_USE_STRET =    (1 << 29),
+     /* APPLE LOCAL radar 8143947 */
+     BLOCK_HAS_SIGNATURE  =    (1 << 30)
 };
 
 struct block_sema_info {

Modified: llvm-gcc-4.2/trunk/gcc/c-decl.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/c-decl.c?rev=109702&r1=109701&r2=109702&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/c-decl.c (original)
+++ llvm-gcc-4.2/trunk/gcc/c-decl.c Wed Jul 28 18:54:15 2010
@@ -8136,6 +8136,11 @@
  // optional helper functions
  void *CopyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE is set (withCopyDispose true)
  void *DestroyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE is set (withCopyDispose true)
+
+ // APPLE LOCAL begin radar 8143947
+ const char *signature;   // the block signature
+ const char *layout;      // reserved
+ // APPLE LOCAL end radar 8143947
 } *descriptor_ptr_type;
 
 Objects of this type will always be static. This is one main component of abi change.
@@ -8174,6 +8179,16 @@
     chainon (field_decl_chain, field_decl);
   }
   
+
+  /* APPLE LOCAL begin radar 8143947 */
+  /* char * signature */
+  field_decl = build_decl (FIELD_DECL, get_identifier ("signature"), build_pointer_type (char_type_node));
+  chainon (field_decl_chain, field_decl);
+  /* char * layout */
+  field_decl = build_decl (FIELD_DECL, get_identifier ("layout"), build_pointer_type (char_type_node));
+  chainon (field_decl_chain, field_decl);
+  /* APPLE LOCAL end radar 8143947 */
+
    /* Mark this struct as being a block struct rather than a 'normal'
       struct.  */
   TYPE_BLOCK_IMPL_STRUCT (main_type) = 1;

Modified: llvm-gcc-4.2/trunk/gcc/c-parser.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/c-parser.c?rev=109702&r1=109701&r2=109702&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/c-parser.c (original)
+++ llvm-gcc-4.2/trunk/gcc/c-parser.c Wed Jul 28 18:54:15 2010
@@ -9428,6 +9428,10 @@
         {
           unsigned long int reserved;
 	  unsigned long int Size;
+          // APPLE LOCAL begin radar 8143927
+          const char *signature;   // the block signature
+          const char *layout;      // reserved
+          // APPLE LOCAL end radar 8143927
 	} *__descriptor;
    };
 
@@ -9488,6 +9492,10 @@
     // optional helper functions
     void *CopyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE
     void *DestroyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE 
+    // APPLE LOCAL begin radar 8143927
+    const char *signature;   // the block signature
+    const char *layout;      // reserved
+    // APPLE LOCAL end radar 8143927
  } *__descriptor;
  
  // imported variables
@@ -9590,6 +9598,10 @@
   {0, sizeof(struct literal_block_n), 
    copy_helper_block_1, // only if block BLOCK_HAS_COPY_DISPOSE
    destroy_helper_block_1, // only if block BLOCK_HAS_COPY_DISPOSE
+   // APPLE LOCAL begin radar 8143947
+   // const char *signature;   // the block signature set to 0
+   // const char *layout;      // reserved set to 0
+   // APPLE LOCAL end radar 8143947 
   }
 */
 static tree
@@ -9638,6 +9650,19 @@
       helper_addr = convert (ptr_type_node, helper_addr);
       initlist = tree_cons (fields, helper_addr, initlist);
     }
+  /* APPLE LOCAL begin radar 8143947 */
+  /* signature field is set to 0 */
+  fields = TREE_CHAIN (fields);
+  initlist = tree_cons (fields,
+                        build_int_cst (build_pointer_type (char_type_node), 0),
+                        initlist);
+  /* layout field is set to 0 */
+  fields = TREE_CHAIN (fields);
+  initlist = tree_cons (fields,
+                        build_int_cst (build_pointer_type (char_type_node), 0),
+                        initlist);
+  /* APPLE LOCAL end radar 8143947 */
+  
   constructor = build_constructor_from_list (descriptor_type,
                                              nreverse (initlist));
   TREE_CONSTANT (constructor) = 1;
@@ -9658,6 +9683,7 @@
  build_block_struct_initlist - builds the initializer list:
  { &_NSConcreteStackBlock or &_NSConcreteGlobalBlock // __isa,
    BLOCK_USE_STRET | BLOCK_HAS_COPY_DISPOSE | BLOCK_IS_GLOBAL // __flags,
+   | BLOCK_HAS_SIGNATURE // __flags
    0, // __reserved
    &helper_1, // __FuncPtr,
    &static_descriptor_variable // __descriptor,
@@ -9673,7 +9699,7 @@
   tree initlist, helper_addr;
   tree chain, fields;
   /* APPLE LOCAL radar 7735196 */
-  unsigned int flags = 0;
+  unsigned int flags = BLOCK_HAS_SIGNATURE;
   static tree NSConcreteStackBlock_decl = NULL_TREE;
   static tree NSConcreteGlobalBlock_decl = NULL_TREE;
   tree descriptor_block_decl = build_descriptor_block_decl (block_struct_type, block_impl);
@@ -9819,6 +9845,10 @@
     // optional helper functions
     void *CopyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE
     void *DestroyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE
+    // APPLE LOCAL begin radar 8143927
+    const char *signature;   // the block signature
+    const char *layout;      // reserved
+    // APPLE LOCAL end radar 8143927
  } *__descriptor;
 
  // imported variables
@@ -9833,6 +9863,7 @@
  struct __block_literal_n I = {
    &_NSConcreteStackBlock or &_NSConcreteGlobalBlock // __isa,
    BLOCK_USE_STRET | BLOCK_HAS_COPY_DISPOSE | BLOCK_IS_GLOBAL // __flags,
+   | BLOCK_HAS_SIGNATURE // __flags
    0, // __reserved
    &helper_1, // __FuncPtr 
    &static_descriptor_variable // __descriptor,

Modified: llvm-gcc-4.2/trunk/gcc/cp/parser.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/cp/parser.c?rev=109702&r1=109701&r2=109702&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/cp/parser.c (original)
+++ llvm-gcc-4.2/trunk/gcc/cp/parser.c Wed Jul 28 18:54:15 2010
@@ -20599,6 +20599,10 @@
   {0, sizeof(struct literal_block_n), 
    copy_helper_block_1, // only if block BLOCK_HAS_COPY_DISPOSE
    destroy_helper_block_1, // only if block BLOCK_HAS_COPY_DISPOSE
+   // APPLE LOCAL begin radar 8143947
+   // const char *signature;   // the block signature set to 0
+   // const char *layout;      // reserved set to 0
+   // APPLE LOCAL end radar 8143947
   }
 */
 static tree
@@ -20641,6 +20645,15 @@
       helper_addr = convert (ptr_type_node, helper_addr);
       CONSTRUCTOR_APPEND_ELT(impl_v, NULL_TREE, helper_addr);
     }
+  /* APPLE LOCAL begin radar 8143947 */
+  /* signature field is set to 0 */
+  CONSTRUCTOR_APPEND_ELT(impl_v, NULL_TREE, 
+                         build_int_cst (build_pointer_type (char_type_node), 0));
+  /* layout field is set to 0 */
+  CONSTRUCTOR_APPEND_ELT(impl_v, NULL_TREE, 
+                         build_int_cst (build_pointer_type (char_type_node), 0));
+  /* APPLE LOCAL end radar 8143947 */
+
   /* Create a CONSTRUCTOR to represent the braced-initializer.  */
   constructor = make_node (CONSTRUCTOR);
   CONSTRUCTOR_ELTS (constructor) = impl_v;
@@ -20673,6 +20686,10 @@
         {
           unsigned long int reserved;
 	  unsigned long int Size;
+          // APPLE LOCAL begin radar 8143927
+          const char *signature;   // the block signature
+          const char *layout;      // reserved
+          // APPLE LOCAL end radar 8143927
 	} *__descriptor;
    };
 
@@ -20872,6 +20889,7 @@
  build_block_struct_initlist - builds the initializer list:
  { &_NSConcreteStackBlock or &_NSConcreteGlobalBlock // __isa,
    BLOCK_USE_STRET | BLOCK_HAS_COPY_DISPOSE | BLOCK_IS_GLOBAL // __flags,
+   | BLOCK_HAS_SIGNATURE // __flags
    0, // __reserved,
    &helper_1, // __FuncPtr,
    &static_descriptor_variable // __descriptor,
@@ -20889,7 +20907,7 @@
 {
   tree expr, chain, helper_addr;
   /* APPLE LOCAL radar 7735196 */
-  unsigned flags = 0;
+  unsigned flags = BLOCK_HAS_SIGNATURE;
   static tree NSConcreteStackBlock_decl = NULL_TREE;
   static tree NSConcreteGlobalBlock_decl = NULL_TREE;
   VEC(constructor_elt,gc) *impl_v = NULL;
@@ -21882,6 +21900,10 @@
  // optional helper functions
  void *CopyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE is set (withCopyDispose true)
  void *DestroyFuncPtr; // When BLOCK_HAS_COPY_DISPOSE is set (withCopyDispose true)
+ // APPLE LOCAL begin radar 8143947
+ const char *signature;   // the block signature
+ const char *layout;      // reserved
+ // APPLE LOCAL end radar 8143947
 } *descriptor_ptr_type;
 
 Objects of this type will always be static. This is one main component of abi change.
@@ -21922,6 +21944,17 @@
     field_decl_chain = field_decl;
   }
 
+  /* APPLE LOCAL begin radar 8143947 */
+  /* char * signature */
+  field_decl = build_decl (FIELD_DECL, get_identifier ("signature"), build_pointer_type (char_type_node));
+  TREE_CHAIN (field_decl) = field_decl_chain;
+  field_decl_chain = field_decl;
+  /* char * layout */
+  field_decl = build_decl (FIELD_DECL, get_identifier ("layout"), build_pointer_type (char_type_node));
+  TREE_CHAIN (field_decl) = field_decl_chain;
+  field_decl_chain = field_decl;
+  /* APPLE LOCAL end radar 8143947 */
+
    /* Mark this struct as being a block struct rather than a 'normal'
       struct.  */
   TYPE_BLOCK_IMPL_STRUCT (main_type) = 1;

Added: llvm-gcc-4.2/trunk/gcc/testsuite/g++.apple/blockstret.C
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/g%2B%2B.apple/blockstret.C?rev=109702&view=auto
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/testsuite/g++.apple/blockstret.C (added)
+++ llvm-gcc-4.2/trunk/gcc/testsuite/g++.apple/blockstret.C Wed Jul 28 18:54:15 2010
@@ -0,0 +1,102 @@
+/* APPLE LOCAL file radar 8241648 */
+/* { dg-do run { target *-*-darwin[1-2][0-9]* } } */
+/* { dg-options "-mmacosx-version-min=10.6 -fblocks" { target *-*-darwin* } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } } */
+
+// rdar://8241648
+extern "C" void *malloc(__SIZE_TYPE__ size);
+extern "C" void abort (void);
+
+typedef struct bigbig {
+   int array[512];
+   char more[32];
+} BigStruct_t;
+
+BigStruct_t (^global)(void) = ^{ return *(BigStruct_t *)malloc(sizeof(struct bigbig)); };
+
+const char * getBlockSignature(void *);
+ 
+BigStruct_t foo(int param) {
+   BigStruct_t x;
+   BigStruct_t (^f)(int) = ^(int param) {
+     BigStruct_t *result = (BigStruct_t *)malloc(sizeof(BigStruct_t));
+     result->array[23] = param;
+     return *result;
+   };
+   getBlockSignature(f);
+   return x;
+}
+
+enum {
+    BLOCK_HAS_COPY_DISPOSE =  (1 << 25),
+    BLOCK_HAS_CXX_OBJ =       (1 << 26),
+    BLOCK_IS_GLOBAL =         (1 << 28),
+    BLOCK_USE_STRET =    (1 << 29),
+    BLOCK_HAS_OBJC_TYPE  =    (1 << 30)
+};
+
+struct block_descriptor_big {
+    unsigned long int reserved;
+    unsigned long int size;
+    void (*copy)(void *dst, void *src); // conditional on BLOCK_HAS_COPY_DISPOSE
+    void (*dispose)(void *);            // conditional on BLOCK_HAS_COPY_DISPOSE
+    const char *signature;                  // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+struct block_descriptor_small {
+    unsigned long int reserved;
+    unsigned long int size;
+    const char *signature;              // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+
+struct block_layout_abi { // can't change
+  void *isa;
+  int flags;
+  int reserved; 
+  void (*invoke)(void *, ...);
+  struct block_descriptor_big *descriptor;
+};
+
+const char *getBlockSignature(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   if ((layout->flags & BLOCK_HAS_OBJC_TYPE) != BLOCK_HAS_OBJC_TYPE) return 0;
+   if (layout->flags & BLOCK_HAS_COPY_DISPOSE) 
+      return layout->descriptor->signature;
+   else
+      return ((struct block_descriptor_small *)layout->descriptor)->signature;
+}
+
+int usesStruct(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   int want = BLOCK_HAS_OBJC_TYPE | BLOCK_USE_STRET;
+   return (layout->flags & want) == want;
+}
+
+int main(int argc, char *argv[]) {
+   if (!usesStruct(global))
+     abort();
+   BigStruct_t x;
+   BigStruct_t (^local)(int) = ^(int param) {
+     BigStruct_t *result = (BigStruct_t *)malloc(sizeof(BigStruct_t));
+     result->array[23] = argc;
+     return *result;
+   };
+   if (!usesStruct(global))
+     abort();
+   if (!usesStruct(local))
+     abort();
+   if (usesStruct(^void(int x){ }))
+     abort();
+   return 0;
+}
+
+/*
+desired global flags: 1879048192
+desired stack flags: 1610612736
+should be non-zero: 1
+should be non-zero: 1
+should be non-zero: 1
+should be zero: 0
+
+*/

Added: llvm-gcc-4.2/trunk/gcc/testsuite/gcc.apple/blockstret.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.apple/blockstret.c?rev=109702&view=auto
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/testsuite/gcc.apple/blockstret.c (added)
+++ llvm-gcc-4.2/trunk/gcc/testsuite/gcc.apple/blockstret.c Wed Jul 28 18:54:15 2010
@@ -0,0 +1,102 @@
+/* APPLE LOCAL file radar 8241648 */
+/* { dg-do run { target *-*-darwin[1-2][0-9]* } } */
+/* { dg-options "-mmacosx-version-min=10.6 -fblocks" { target *-*-darwin* } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } } */
+
+// rdar://8241648
+void *malloc(__SIZE_TYPE__ size);
+extern void abort (void);
+
+typedef struct bigbig {
+   int array[512];
+   char more[32];
+} BigStruct_t;
+
+BigStruct_t (^global)(void) = ^{ return *(BigStruct_t *)malloc(sizeof(struct bigbig)); };
+
+const char * getBlockSignature(void *);
+ 
+BigStruct_t foo(int param) {
+   BigStruct_t x;
+   BigStruct_t (^f)(int) = ^(int param) {
+     BigStruct_t *result = (BigStruct_t *)malloc(sizeof(BigStruct_t));
+     result->array[23] = param;
+     return *result;
+   };
+   getBlockSignature(f);
+   return x;
+}
+
+enum {
+    BLOCK_HAS_COPY_DISPOSE =  (1 << 25),
+    BLOCK_HAS_CXX_OBJ =       (1 << 26),
+    BLOCK_IS_GLOBAL =         (1 << 28),
+    BLOCK_USE_STRET =    (1 << 29),
+    BLOCK_HAS_OBJC_TYPE  =    (1 << 30)
+};
+
+struct block_descriptor_big {
+    unsigned long int reserved;
+    unsigned long int size;
+    void (*copy)(void *dst, void *src); // conditional on BLOCK_HAS_COPY_DISPOSE
+    void (*dispose)(void *);            // conditional on BLOCK_HAS_COPY_DISPOSE
+    const char *signature;                  // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+struct block_descriptor_small {
+    unsigned long int reserved;
+    unsigned long int size;
+    const char *signature;              // conditional on BLOCK_HAS_OBJC
+    const char *layout;                 // conditional on BLOCK_HAS_OBJC
+};
+
+struct block_layout_abi { // can't change
+  void *isa;
+  int flags;
+  int reserved; 
+  void (*invoke)(void *, ...);
+  struct block_descriptor_big *descriptor;
+};
+
+const char *getBlockSignature(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   if ((layout->flags & BLOCK_HAS_OBJC_TYPE) != BLOCK_HAS_OBJC_TYPE) return 0;
+   if (layout->flags & BLOCK_HAS_COPY_DISPOSE) 
+      return layout->descriptor->signature;
+   else
+      return ((struct block_descriptor_small *)layout->descriptor)->signature;
+}
+
+int usesStruct(void *block) {
+   struct block_layout_abi *layout = (struct block_layout_abi *)block;
+   int want = BLOCK_HAS_OBJC_TYPE | BLOCK_USE_STRET;
+   return (layout->flags & want) == want;
+}
+
+int main(int argc, char *argv[]) {
+   if (!usesStruct(global))
+     abort();
+   BigStruct_t x;
+   BigStruct_t (^local)(int) = ^(int param) {
+     BigStruct_t *result = (BigStruct_t *)malloc(sizeof(BigStruct_t));
+     result->array[23] = argc;
+     return *result;
+   };
+   if (!usesStruct(global))
+     abort();
+   if (!usesStruct(local))
+     abort();
+   if (usesStruct(^void(int x){ }))
+     abort();
+   return 0;
+}
+
+/*
+desired global flags: 1879048192
+desired stack flags: 1610612736
+should be non-zero: 1
+should be non-zero: 1
+should be non-zero: 1
+should be zero: 0
+
+*/




From gohman at apple.com  Wed Jul 28 19:17:55 2010
From: gohman at apple.com (Dan Gohman)
Date: Thu, 29 Jul 2010 00:17:55 -0000
Subject: [llvm-commits] [llvm] r109705 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h
 include/llvm/Analysis/ScalarEvolutionExpressions.h
 lib/Analysis/ScalarEvolution.cpp unittests/Analysis/
 unittests/Analysis/Makefile unittests/Analysis/ScalarEvolutionTest.cpp
 unittests/Makefile
Message-ID: <20100729001755.A88E52A6C12C@llvm.org>

Author: djg
Date: Wed Jul 28 19:17:55 2010
New Revision: 109705

URL: http://llvm.org/viewvc/llvm-project?rev=109705&view=rev
Log:
Factor out some of the code for updating old SCEVUnknown values, and
extend it to handle the case where multiple RAUWs affect a single
SCEVUnknown.

Add a ScalarEvolution unittest to test for this situation.

Added:
    llvm/trunk/unittests/Analysis/   (with props)
    llvm/trunk/unittests/Analysis/Makefile
      - copied, changed from r109674, llvm/trunk/unittests/VMCore/Makefile
    llvm/trunk/unittests/Analysis/ScalarEvolutionTest.cpp
Modified:
    llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
    llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp
    llvm/trunk/unittests/Makefile

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=109705&r1=109704&r2=109705&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Wed Jul 28 19:17:55 2010
@@ -386,6 +386,11 @@
     bool isKnownPredicateWithRanges(ICmpInst::Predicate Pred,
                                     const SCEV *LHS, const SCEV *RHS);
 
+    /// forgetSCEVUnknown - V is being deleted or RAUW'd; remove the
+    /// SCEVUnknown for it from the uniquing map, and optionally
+    /// clear its contents to point to a replacement value.
+    void forgetSCEVUnknown(Value *V, Value *NewV);
+
   public:
     static char ID; // Pass identification, replacement for typeid
     ScalarEvolution();

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h?rev=109705&r1=109704&r2=109705&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h Wed Jul 28 19:17:55 2010
@@ -524,11 +524,26 @@
     friend class ScalarEvolution;
     friend class ScalarEvolution::SCEVCallbackVH;
 
-    // This should be an AssertingVH, however SCEVUnknowns are allocated in a
-    // BumpPtrAllocator so their destructors are never called.
+    /// V - The Value represented by this SCEVUnknown.
+    /// This should be an AssertingVH, however SCEVUnknowns are allocated in a
+    /// BumpPtrAllocator so their destructors are never called.
     Value *V;
+
+    /// UpdateList - When values are RAUW'd with new values, and the new
+    /// values already have their own SCEVUnknowns, they can end up with
+    /// muliple SCEVUnknowns. This pointer links them all together so that
+    /// they can all be updated when another RAUW happens.
+    SCEVUnknown *UpdateList;
+
+    /// getUpdateListBack - Return the last SCEVUnknown in te UpdateList.
+    SCEVUnknown *getUpdateListBack() {
+      SCEVUnknown *P = this;
+      while (SCEVUnknown *Q = P->UpdateList) P = Q;
+      return P;
+    }
+
     SCEVUnknown(const FoldingSetNodeIDRef ID, Value *v) :
-      SCEV(ID, scUnknown), V(v) {}
+      SCEV(ID, scUnknown), V(v), UpdateList(0) {}
 
   public:
     Value *getValue() const { return V; }

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=109705&r1=109704&r2=109705&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Wed Jul 28 19:17:55 2010
@@ -3660,29 +3660,59 @@
   }
 }
 
-/// forgetValue - This method should be called by the client when it has
-/// changed a value in a way that may effect its value, or which may
-/// disconnect it from a def-use chain linking it to a loop.
-void ScalarEvolution::forgetValue(Value *V) {
-  // If there's a SCEVUnknown tying this value into the SCEV
-  // space, remove it from the folding set map. The SCEVUnknown
-  // object and any other SCEV objects which reference it
-  // (transitively) remain allocated, effectively leaked until
-  // the underlying BumpPtrAllocator is freed.
-  //
+/// forgetSCEVUnknown - V is being deleted or RAUW'd; remove the
+/// SCEVUnknown for it from the uniquing map, and optionally
+/// clear its contents to point to a replacement value.
+void ScalarEvolution::forgetSCEVUnknown(Value *V, Value *NewV) {
   // This permits SCEV pointers to be used as keys in maps
   // such as the ValuesAtScopes map.
+
+  // Check for an exisitng SCEVUnknown for V in the uniquing map.
   FoldingSetNodeID ID;
   ID.AddInteger(scUnknown);
   ID.AddPointer(V);
   void *IP;
   if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
+    // Remove the old SCEVUnknown from the uniquing map, effectively
+    // leaking it from subsequent use by ScalarEvolution. There
+    // may be existing SCEVs live which still point to it though.
     UniqueSCEVs.RemoveNode(S);
 
-    // This isn't necessary, but we might as well remove the
-    // value from the ValuesAtScopes map too.
-    ValuesAtScopes.erase(S);
+    // If there's a replacement value, update the old SCEVUnknowns
+    // so that SCEVs which still point to it see the new value.
+    if (NewV) {
+      // Update the old SCEVUnknown and all SCEVUnknowns on its update list
+      // to point to NewV instead of V.
+      for (SCEVUnknown *U = cast<SCEVUnknown>(S); U; U = U->UpdateList) {
+        assert(U->V == V && "SCEVUnknown update list is inconsistent!");
+        U->V = NewV;
+      }
+
+      // Check for an existing SCEVUnknown for NewV in the uniquing map.
+      ID.clear();
+      ID.AddInteger(scUnknown);
+      ID.AddPointer(NewV);
+      if (SCEV *NewS = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
+        // The uniquing map already had an entry for the new
+        // value. Append the old SCEVUnknown (and its update list) to the
+        // new SCEVUnknown's update list.
+        cast<SCEVUnknown>(NewS)->getUpdateListBack()->UpdateList =
+          cast<SCEVUnknown>(S);
+      } else {
+        // The uniquing map did not have an entry for the new
+        // value yet. Set the old SCEVUnknown, which has now been
+        // updated, to be the SCEVUnknown in the uniquing map.
+        UniqueSCEVs.InsertNode(S, IP);
+      }
+    }
   }
+}
+
+/// forgetValue - This method should be called by the client when it has
+/// changed a value in a way that may effect its value, or which may
+/// disconnect it from a def-use chain linking it to a loop.
+void ScalarEvolution::forgetValue(Value *V) {
+  forgetSCEVUnknown(V, 0);
 
   Instruction *I = dyn_cast<Instruction>(V);
   if (!I) return;
@@ -5692,24 +5722,10 @@
 
 void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *V) {
   assert(SE && "SCEVCallbackVH called with a null ScalarEvolution!");
-
   Value *Old = getValPtr();
 
-  // If there's a SCEVUnknown tying this value into the SCEV
-  // space, replace the SCEVUnknown's value with the new value
-  // for the benefit of any SCEVs still referencing it, and
-  // and remove it from the folding set map so that new scevs
-  // don't reference it.
-  FoldingSetNodeID ID;
-  ID.AddInteger(scUnknown);
-  ID.AddPointer(Old);
-  void *IP;
-  if (SCEVUnknown *S = cast_or_null<SCEVUnknown>(
-        SE->UniqueSCEVs.FindNodeOrInsertPos(ID, IP))) {
-    S->V = V;
-    SE->UniqueSCEVs.RemoveNode(S);
-    SE->ValuesAtScopes.erase(S);
-  }
+  // First, update existing SCEVUnknowns.
+  SE->forgetSCEVUnknown(Old, V);
 
   // Forget all the expressions associated with users of the old value,
   // so that future queries will recompute the expressions using the new

Propchange: llvm/trunk/unittests/Analysis/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Wed Jul 28 19:17:55 2010
@@ -0,0 +1,9 @@
+Debug
+Release
+Release-Asserts
+Debug+Coverage-Asserts
+Debug+Coverage
+Release+Coverage
+Debug+Checks
+Debug+Asserts
+Release+Asserts

Copied: llvm/trunk/unittests/Analysis/Makefile (from r109674, llvm/trunk/unittests/VMCore/Makefile)
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Analysis/Makefile?p2=llvm/trunk/unittests/Analysis/Makefile&p1=llvm/trunk/unittests/VMCore/Makefile&r1=109674&r2=109705&rev=109705&view=diff
==============================================================================
--- llvm/trunk/unittests/VMCore/Makefile (original)
+++ llvm/trunk/unittests/Analysis/Makefile Wed Jul 28 19:17:55 2010
@@ -1,4 +1,4 @@
-##===- unittests/VMCore/Makefile ---------------------------*- Makefile -*-===##
+##===- unittests/Analysis/Makefile -------------------------*- Makefile -*-===##
 #
 #                     The LLVM Compiler Infrastructure
 #
@@ -8,8 +8,8 @@
 ##===----------------------------------------------------------------------===##
 
 LEVEL = ../..
-TESTNAME = VMCore
-LINK_COMPONENTS := core support target ipa
+TESTNAME = Analysis
+LINK_COMPONENTS := core support target analysis ipa
 
 include $(LEVEL)/Makefile.config
 include $(LLVM_SRC_ROOT)/unittests/Makefile.unittest

Added: llvm/trunk/unittests/Analysis/ScalarEvolutionTest.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Analysis/ScalarEvolutionTest.cpp?rev=109705&view=auto
==============================================================================
--- llvm/trunk/unittests/Analysis/ScalarEvolutionTest.cpp (added)
+++ llvm/trunk/unittests/Analysis/ScalarEvolutionTest.cpp Wed Jul 28 19:17:55 2010
@@ -0,0 +1,78 @@
+//===- ScalarEvolutionsTest.cpp - ScalarEvolution unit tests --------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include <llvm/Analysis/ScalarEvolutionExpressions.h>
+#include <llvm/GlobalVariable.h>
+#include <llvm/Constants.h>
+#include <llvm/LLVMContext.h>
+#include <llvm/Module.h>
+#include <llvm/PassManager.h>
+#include "gtest/gtest.h"
+
+namespace llvm {
+namespace {
+
+TEST(ScalarEvolutionsTest, ReturnInst) {
+  LLVMContext Context;
+  Module M("world", Context);
+
+  const FunctionType *FTy = FunctionType::get(Type::getVoidTy(Context),
+                                              std::vector<const Type *>(), false);
+  Function *F = cast<Function>(M.getOrInsertFunction("f", FTy));
+  BasicBlock *BB = BasicBlock::Create(Context, "entry", F);
+  ReturnInst::Create(Context, 0, BB);
+
+  const Type *Ty = Type::getInt1Ty(Context);
+  Constant *Init = Constant::getNullValue(Ty);
+  Value *V0 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V0");
+  Value *V1 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V1");
+  Value *V2 = new GlobalVariable(M, Ty, false, GlobalValue::ExternalLinkage, Init, "V2");
+
+  // Create a ScalarEvolution and "run" it so that it gets initialized.
+  PassManager PM;
+  ScalarEvolution &SE = *new ScalarEvolution();
+  PM.add(&SE);
+  PM.run(M);
+
+  const SCEV *S0 = SE.getSCEV(V0);
+  const SCEV *S1 = SE.getSCEV(V1);
+  const SCEV *S2 = SE.getSCEV(V2);
+
+  const SCEV *P0 = SE.getAddExpr(S0, S0);
+  const SCEV *P1 = SE.getAddExpr(S1, S1);
+  const SCEV *P2 = SE.getAddExpr(S2, S2);
+
+  const SCEVMulExpr *M0 = cast<SCEVMulExpr>(P0);
+  const SCEVMulExpr *M1 = cast<SCEVMulExpr>(P1);
+  const SCEVMulExpr *M2 = cast<SCEVMulExpr>(P2);
+
+  EXPECT_EQ(cast<SCEVConstant>(M0->getOperand(0))->getValue()->getZExtValue(),
+            2u);
+  EXPECT_EQ(cast<SCEVConstant>(M1->getOperand(0))->getValue()->getZExtValue(),
+            2u);
+  EXPECT_EQ(cast<SCEVConstant>(M2->getOperand(0))->getValue()->getZExtValue(),
+            2u);
+
+  // Before the RAUWs, these are all pointing to separate values.
+  EXPECT_EQ(cast<SCEVUnknown>(M0->getOperand(1))->getValue(), V0);
+  EXPECT_EQ(cast<SCEVUnknown>(M1->getOperand(1))->getValue(), V1);
+  EXPECT_EQ(cast<SCEVUnknown>(M2->getOperand(1))->getValue(), V2);
+
+  // Do some RAUWs.
+  V2->replaceAllUsesWith(V1);
+  V1->replaceAllUsesWith(V0);
+
+  // After the RAUWs, these should all be pointing to V0.
+  EXPECT_EQ(cast<SCEVUnknown>(M0->getOperand(1))->getValue(), V0);
+  EXPECT_EQ(cast<SCEVUnknown>(M1->getOperand(1))->getValue(), V0);
+  EXPECT_EQ(cast<SCEVUnknown>(M2->getOperand(1))->getValue(), V0);
+}
+
+}  // end anonymous namespace
+}  // end namespace llvm

Modified: llvm/trunk/unittests/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile?rev=109705&r1=109704&r2=109705&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile (original)
+++ llvm/trunk/unittests/Makefile Wed Jul 28 19:17:55 2010
@@ -9,7 +9,7 @@
 
 LEVEL = ..
 
-PARALLEL_DIRS = ADT ExecutionEngine Support Transforms VMCore
+PARALLEL_DIRS = ADT ExecutionEngine Support Transforms VMCore Analysis
 
 include $(LEVEL)/Makefile.common
 




From nlewycky at google.com  Wed Jul 28 19:39:45 2010
From: nlewycky at google.com (Nick Lewycky)
Date: Wed, 28 Jul 2010 17:39:45 -0700
Subject: [llvm-commits] [patch] Fix bugpoint -inline ....
In-Reply-To: <AANLkTin+z9EcYdUYQOWypc7j5fOd-BwcdF8vNiAisEmq@mail.gmail.com>
References: <AANLkTinSzYL39PhJs3RcOQFN4rUHZQHbUvxqamytmGJG@mail.gmail.com> 
	<AANLkTim6=XP_jz_YanRGMjTWqiuoaK2256-a5mq47OpS@mail.gmail.com> 
	<4C4C0173.50203@free.fr>
	<AANLkTi=F2C6_VtJiPX=U-kJtV+kiFSqQcGUgC5oBg0tM@mail.gmail.com> 
	<AANLkTinqRp8UmpcPD4p9goGgDGX_rfALp3tFj46tZCv6@mail.gmail.com> 
	<AANLkTin+z9EcYdUYQOWypc7j5fOd-BwcdF8vNiAisEmq@mail.gmail.com>
Message-ID: <AANLkTineRm6uF=Hmd85=FGkLh9W8qgv-vWSFCTjJHZ3a@mail.gmail.com>

+  for (unsigned i = 0, e = FunctionsBeingTested.size(); i !=e; ++i) {

Space after '!=' and before 'e' please.

+  for (unsigned i = 0, e = BBs.size(); i !=e; ++i) {

Here too!

Looks good!

On 25 July 2010 17:21, Rafael Espindola <espindola at google.com> wrote:

> > Do you want to use a SmallVector here?
>
> It is passed to a function and I wanted to avoid changing the
> interface. I implemented the other suggestions.
>
> The attached patch does a similar thing to the block reduction logic.
> With it bugpoint can run without crashing on the testcase I am trying
> to reduce.
>
> I did noticed another less serious problem with bugpoint that I might
> try to fix in another patch: When splitting the module into two parts,
> it will copy the functions it wants to optimize. When doing that it
> changes internal functions into regular ones and this sometimes masks
> the bug.
>
> For example, say a module has f1, f2 and f3. f2 is internal and the
> bug can be reproduced with only f1 and f2 but is masked if f2 is not
> internal. Bugpoint will fail to remove f3.
>

Generating testcases that are too large isn't my big worry with bugpoint.
I'm much more concerned about things like marking (or not marking) a
function fastcc as a result of it being inline.

Nick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100728/a656f491/attachment.html 

From stoklund at 2pi.dk  Wed Jul 28 19:52:17 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 29 Jul 2010 00:52:17 -0000
Subject: [llvm-commits] [llvm] r109714 -
	/llvm/trunk/tools/bugpoint/ToolRunner.cpp
Message-ID: <20100729005217.138AC2A6C12C@llvm.org>

Author: stoklund
Date: Wed Jul 28 19:52:16 2010
New Revision: 109714

URL: http://llvm.org/viewvc/llvm-project?rev=109714&view=rev
Log:
Use the right gcc tool args for IsARMArchitecture.

Modified:
    llvm/trunk/tools/bugpoint/ToolRunner.cpp

Modified: llvm/trunk/tools/bugpoint/ToolRunner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ToolRunner.cpp?rev=109714&r1=109713&r2=109714&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ToolRunner.cpp (original)
+++ llvm/trunk/tools/bugpoint/ToolRunner.cpp Wed Jul 28 19:52:16 2010
@@ -627,8 +627,8 @@
 // GCC abstraction
 //
 
-static bool IsARMArchitecture(std::vector<std::string> Args) {
-  for (std::vector<std::string>::const_iterator
+static bool IsARMArchitecture(std::vector<const char*> Args) {
+  for (std::vector<const char*>::const_iterator
          I = Args.begin(), E = Args.end(); I != E; ++I) {
     if (StringRef(*I).equals_lower("-arch")) {
       ++I;
@@ -673,7 +673,7 @@
       // explicitly told what architecture it is working on, so we get
       // it from gcc flags
       if ((TargetTriple.getOS() == Triple::Darwin) &&
-          !IsARMArchitecture(ArgsForGCC))
+          !IsARMArchitecture(GCCArgs))
         GCCArgs.push_back("-force_cpusubtype_ALL");
     }
   }




From stoklund at 2pi.dk  Wed Jul 28 19:52:19 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 29 Jul 2010 00:52:19 -0000
Subject: [llvm-commits] [llvm] r109715 -
	/llvm/trunk/lib/CodeGen/RegAllocFast.cpp
Message-ID: <20100729005219.7332C2A6C12D@llvm.org>

Author: stoklund
Date: Wed Jul 28 19:52:19 2010
New Revision: 109715

URL: http://llvm.org/viewvc/llvm-project?rev=109715&view=rev
Log:
Fix a bug in the -regalloc=fast handling of exotic two-address instruction with
multiple defs, like t2LDRSB_POST.

The first def could accidentally steal the physreg that the second, tied def was
required to be allocated to.

Now, the tied use-def is treated more like an early clobber, and the physreg is
reserved before allocating the other defs.

This would never be a problem when the tied def was the only def which is the
usual case.

This fixes MallocBench/gs for thumb2 -O0.

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

Modified: llvm/trunk/lib/CodeGen/RegAllocFast.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/RegAllocFast.cpp?rev=109715&r1=109714&r2=109715&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/RegAllocFast.cpp (original)
+++ llvm/trunk/lib/CodeGen/RegAllocFast.cpp Wed Jul 28 19:52:19 2010
@@ -847,13 +847,18 @@
     // operands. If there are also physical defs, these registers must avoid
     // both physical defs and uses, making them more constrained than normal
     // operands.
+    // Similarly, if there are multiple defs and tied operands, we must make sure
+    // the same register is allocated to uses and defs.
     // We didn't detect inline asm tied operands above, so just make this extra
     // pass for all inline asm.
     if (MI->isInlineAsm() || hasEarlyClobbers || hasPartialRedefs ||
-        (hasTiedOps && hasPhysDefs)) {
+        (hasTiedOps && (hasPhysDefs || TID.getNumDefs() > 1))) {
       handleThroughOperands(MI, VirtDead);
       // Don't attempt coalescing when we have funny stuff going on.
       CopyDst = 0;
+      // Pretend we have early clobbers so the use operands get marked below.
+      // This is not necessary for the common case of a single tied use.
+      hasEarlyClobbers = true;
     }
 
     // Second scan.
@@ -874,14 +879,17 @@
 
     MRI->addPhysRegsUsed(UsedInInstr);
 
-    // Track registers defined by instruction - early clobbers at this point.
+    // Track registers defined by instruction - early clobbers and tied uses at
+    // this point.
     UsedInInstr.reset();
     if (hasEarlyClobbers) {
       for (unsigned i = 0, e = MI->getNumOperands(); i != e; ++i) {
         MachineOperand &MO = MI->getOperand(i);
-        if (!MO.isReg() || !MO.isDef()) continue;
+        if (!MO.isReg()) continue;
         unsigned Reg = MO.getReg();
         if (!Reg || !TargetRegisterInfo::isPhysicalRegister(Reg)) continue;
+        // Look for physreg defs and tied uses.
+        if (!MO.isDef() && !MI->isRegTiedToDefOperand(i)) continue;
         UsedInInstr.set(Reg);
         for (const unsigned *AS = TRI->getAliasSet(Reg); *AS; ++AS)
           UsedInInstr.set(*AS);




From daniel at zuster.org  Wed Jul 28 20:21:47 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 01:21:47 -0000
Subject: [llvm-commits] [llvm] r109717 -
	/llvm/trunk/lib/Support/CrashRecoveryContext.cpp
Message-ID: <20100729012147.4CE4C2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Jul 28 20:21:47 2010
New Revision: 109717

URL: http://llvm.org/viewvc/llvm-project?rev=109717&view=rev
Log:
CrashRecoveryContext: Add a simple POSIX implementation.
 - This works, but won't handle crashes on stack overflow, or signals delivered
   to a thread other than the one that crashed. The latter is particular annoying
   on Darwin, because SIGABRT tends to go to the main thread.

Modified:
    llvm/trunk/lib/Support/CrashRecoveryContext.cpp

Modified: llvm/trunk/lib/Support/CrashRecoveryContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CrashRecoveryContext.cpp?rev=109717&r1=109716&r2=109717&view=diff
==============================================================================
--- llvm/trunk/lib/Support/CrashRecoveryContext.cpp (original)
+++ llvm/trunk/lib/Support/CrashRecoveryContext.cpp Wed Jul 28 20:21:47 2010
@@ -9,20 +9,29 @@
 
 #include "llvm/Support/CrashRecoveryContext.h"
 #include "llvm/ADT/SmallString.h"
+#include "llvm/System/ThreadLocal.h"
 #include <setjmp.h>
+#include <cstdio>
 using namespace llvm;
 
 namespace {
 
 struct CrashRecoveryContextImpl;
 
+static sys::ThreadLocal<const CrashRecoveryContextImpl> CurrentContext;
+
 struct CrashRecoveryContextImpl {
   std::string Backtrace;
   ::jmp_buf JumpBuffer;
   volatile unsigned Failed : 1;
 
 public:
-  CrashRecoveryContextImpl() : Failed(false) {}
+  CrashRecoveryContextImpl() : Failed(false) {
+    CurrentContext.set(this);
+  }
+  ~CrashRecoveryContextImpl() {
+    CurrentContext.set(0);
+  }
 
   void HandleCrash() {
     assert(!Failed && "Crash recovery context already failed!");
@@ -44,6 +53,10 @@
   delete CRCI;
 }
 
+#ifdef LLVM_ON_WIN32
+
+// FIXME: No real Win32 implementation currently.
+
 void CrashRecoveryContext::Enable() {
   if (gCrashRecoveryEnabled)
     return;
@@ -58,6 +71,94 @@
   gCrashRecoveryEnabled = false;
 }
 
+#else
+
+// Generic POSIX implementation.
+//
+// This implementation relies on synchronous signals being delivered to the
+// current thread. We use a thread local object to keep track of the active
+// crash recovery context, and install signal handlers to invoke HandleCrash on
+// the active object.
+//
+// This implementation does not to attempt to chain signal handlers in any
+// reliable fashion -- if we get a signal outside of a crash recovery context we
+// simply disable crash recovery and raise the signal again.
+
+#include <signal.h>
+
+static struct {
+  int Signal;
+  struct sigaction PrevAction;
+} SignalInfo[] = {
+  { SIGABRT, {} },
+  { SIGBUS,  {} },
+  { SIGFPE,  {} },
+  { SIGILL,  {} },
+  { SIGSEGV, {} },
+  { SIGTRAP, {} },
+};
+static const unsigned NumSignals = sizeof(SignalInfo) / sizeof(SignalInfo[0]);
+
+static void CrashRecoverySignalHandler(int Signal) {
+  // Lookup the current thread local recovery object.
+  const CrashRecoveryContextImpl *CRCI = CurrentContext.get();
+
+  if (!CRCI) {
+    // We didn't find a crash recovery context -- this means either we got a
+    // signal on a thread we didn't expect it on, the application got a signal
+    // outside of a crash recovery context, or something else went horribly
+    // wrong.
+    //
+    // Disable crash recovery and raise the signal again. The assumption here is
+    // that the enclosing application will terminate soon, and we won't want to
+    // attempt crash recovery again.
+    //
+    // This call of Disable isn't thread safe, but it doesn't actually matter.
+    CrashRecoveryContext::Disable();
+    raise(Signal);
+  }
+
+  // Unblock the signal we received.
+  sigset_t SigMask;
+  sigemptyset(&SigMask);
+  sigaddset(&SigMask, Signal);
+  sigprocmask(SIG_UNBLOCK, &SigMask, 0);
+
+  if (CRCI)
+    const_cast<CrashRecoveryContextImpl*>(CRCI)->HandleCrash();
+}
+
+void CrashRecoveryContext::Enable() {
+  if (gCrashRecoveryEnabled)
+    return;
+
+  gCrashRecoveryEnabled = true;
+
+  // Setup the signal handler.
+  struct sigaction Handler;
+  Handler.sa_handler = CrashRecoverySignalHandler;
+  Handler.sa_flags = 0;
+  sigemptyset(&Handler.sa_mask);
+
+  for (unsigned i = 0; i != NumSignals; ++i) {
+    sigaction(SignalInfo[i].Signal, &Handler,
+              &SignalInfo[i].PrevAction);
+  }
+}
+
+void CrashRecoveryContext::Disable() {
+  if (!gCrashRecoveryEnabled)
+    return;
+
+  gCrashRecoveryEnabled = false;
+
+  // Restore the previous signal handlers.
+  for (unsigned i = 0; i != NumSignals; ++i)
+    sigaction(SignalInfo[i].Signal, &SignalInfo[i].PrevAction, 0);
+}
+
+#endif
+
 bool CrashRecoveryContext::RunSafely(void (*Fn)(void*), void *UserData) {
   // If crash recovery is disabled, do nothing.
   if (gCrashRecoveryEnabled) {




From echristo at apple.com  Wed Jul 28 20:25:38 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 01:25:38 -0000
Subject: [llvm-commits] [llvm] r109718 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h
 include/llvm/Analysis/ScalarEvolutionExpressions.h
 lib/Analysis/ScalarEvolution.cpp unittests/Analysis/ unittests/Makefile
Message-ID: <20100729012538.B82032A6C12C@llvm.org>

Author: echristo
Date: Wed Jul 28 20:25:38 2010
New Revision: 109718

URL: http://llvm.org/viewvc/llvm-project?rev=109718&view=rev
Log:
Speculatively revert r109705 since it seems to be causing some build bot
angst.

Removed:
    llvm/trunk/unittests/Analysis/
Modified:
    llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
    llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp
    llvm/trunk/unittests/Makefile

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=109718&r1=109717&r2=109718&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Wed Jul 28 20:25:38 2010
@@ -386,11 +386,6 @@
     bool isKnownPredicateWithRanges(ICmpInst::Predicate Pred,
                                     const SCEV *LHS, const SCEV *RHS);
 
-    /// forgetSCEVUnknown - V is being deleted or RAUW'd; remove the
-    /// SCEVUnknown for it from the uniquing map, and optionally
-    /// clear its contents to point to a replacement value.
-    void forgetSCEVUnknown(Value *V, Value *NewV);
-
   public:
     static char ID; // Pass identification, replacement for typeid
     ScalarEvolution();

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h?rev=109718&r1=109717&r2=109718&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h Wed Jul 28 20:25:38 2010
@@ -524,26 +524,11 @@
     friend class ScalarEvolution;
     friend class ScalarEvolution::SCEVCallbackVH;
 
-    /// V - The Value represented by this SCEVUnknown.
-    /// This should be an AssertingVH, however SCEVUnknowns are allocated in a
-    /// BumpPtrAllocator so their destructors are never called.
+    // This should be an AssertingVH, however SCEVUnknowns are allocated in a
+    // BumpPtrAllocator so their destructors are never called.
     Value *V;
-
-    /// UpdateList - When values are RAUW'd with new values, and the new
-    /// values already have their own SCEVUnknowns, they can end up with
-    /// muliple SCEVUnknowns. This pointer links them all together so that
-    /// they can all be updated when another RAUW happens.
-    SCEVUnknown *UpdateList;
-
-    /// getUpdateListBack - Return the last SCEVUnknown in te UpdateList.
-    SCEVUnknown *getUpdateListBack() {
-      SCEVUnknown *P = this;
-      while (SCEVUnknown *Q = P->UpdateList) P = Q;
-      return P;
-    }
-
     SCEVUnknown(const FoldingSetNodeIDRef ID, Value *v) :
-      SCEV(ID, scUnknown), V(v), UpdateList(0) {}
+      SCEV(ID, scUnknown), V(v) {}
 
   public:
     Value *getValue() const { return V; }

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=109718&r1=109717&r2=109718&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Wed Jul 28 20:25:38 2010
@@ -3660,59 +3660,29 @@
   }
 }
 
-/// forgetSCEVUnknown - V is being deleted or RAUW'd; remove the
-/// SCEVUnknown for it from the uniquing map, and optionally
-/// clear its contents to point to a replacement value.
-void ScalarEvolution::forgetSCEVUnknown(Value *V, Value *NewV) {
+/// forgetValue - This method should be called by the client when it has
+/// changed a value in a way that may effect its value, or which may
+/// disconnect it from a def-use chain linking it to a loop.
+void ScalarEvolution::forgetValue(Value *V) {
+  // If there's a SCEVUnknown tying this value into the SCEV
+  // space, remove it from the folding set map. The SCEVUnknown
+  // object and any other SCEV objects which reference it
+  // (transitively) remain allocated, effectively leaked until
+  // the underlying BumpPtrAllocator is freed.
+  //
   // This permits SCEV pointers to be used as keys in maps
   // such as the ValuesAtScopes map.
-
-  // Check for an exisitng SCEVUnknown for V in the uniquing map.
   FoldingSetNodeID ID;
   ID.AddInteger(scUnknown);
   ID.AddPointer(V);
   void *IP;
   if (SCEV *S = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
-    // Remove the old SCEVUnknown from the uniquing map, effectively
-    // leaking it from subsequent use by ScalarEvolution. There
-    // may be existing SCEVs live which still point to it though.
     UniqueSCEVs.RemoveNode(S);
 
-    // If there's a replacement value, update the old SCEVUnknowns
-    // so that SCEVs which still point to it see the new value.
-    if (NewV) {
-      // Update the old SCEVUnknown and all SCEVUnknowns on its update list
-      // to point to NewV instead of V.
-      for (SCEVUnknown *U = cast<SCEVUnknown>(S); U; U = U->UpdateList) {
-        assert(U->V == V && "SCEVUnknown update list is inconsistent!");
-        U->V = NewV;
-      }
-
-      // Check for an existing SCEVUnknown for NewV in the uniquing map.
-      ID.clear();
-      ID.AddInteger(scUnknown);
-      ID.AddPointer(NewV);
-      if (SCEV *NewS = UniqueSCEVs.FindNodeOrInsertPos(ID, IP)) {
-        // The uniquing map already had an entry for the new
-        // value. Append the old SCEVUnknown (and its update list) to the
-        // new SCEVUnknown's update list.
-        cast<SCEVUnknown>(NewS)->getUpdateListBack()->UpdateList =
-          cast<SCEVUnknown>(S);
-      } else {
-        // The uniquing map did not have an entry for the new
-        // value yet. Set the old SCEVUnknown, which has now been
-        // updated, to be the SCEVUnknown in the uniquing map.
-        UniqueSCEVs.InsertNode(S, IP);
-      }
-    }
+    // This isn't necessary, but we might as well remove the
+    // value from the ValuesAtScopes map too.
+    ValuesAtScopes.erase(S);
   }
-}
-
-/// forgetValue - This method should be called by the client when it has
-/// changed a value in a way that may effect its value, or which may
-/// disconnect it from a def-use chain linking it to a loop.
-void ScalarEvolution::forgetValue(Value *V) {
-  forgetSCEVUnknown(V, 0);
 
   Instruction *I = dyn_cast<Instruction>(V);
   if (!I) return;
@@ -5722,10 +5692,24 @@
 
 void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *V) {
   assert(SE && "SCEVCallbackVH called with a null ScalarEvolution!");
+
   Value *Old = getValPtr();
 
-  // First, update existing SCEVUnknowns.
-  SE->forgetSCEVUnknown(Old, V);
+  // If there's a SCEVUnknown tying this value into the SCEV
+  // space, replace the SCEVUnknown's value with the new value
+  // for the benefit of any SCEVs still referencing it, and
+  // and remove it from the folding set map so that new scevs
+  // don't reference it.
+  FoldingSetNodeID ID;
+  ID.AddInteger(scUnknown);
+  ID.AddPointer(Old);
+  void *IP;
+  if (SCEVUnknown *S = cast_or_null<SCEVUnknown>(
+        SE->UniqueSCEVs.FindNodeOrInsertPos(ID, IP))) {
+    S->V = V;
+    SE->UniqueSCEVs.RemoveNode(S);
+    SE->ValuesAtScopes.erase(S);
+  }
 
   // Forget all the expressions associated with users of the old value,
   // so that future queries will recompute the expressions using the new

Modified: llvm/trunk/unittests/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile?rev=109718&r1=109717&r2=109718&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile (original)
+++ llvm/trunk/unittests/Makefile Wed Jul 28 20:25:38 2010
@@ -9,7 +9,7 @@
 
 LEVEL = ..
 
-PARALLEL_DIRS = ADT ExecutionEngine Support Transforms VMCore Analysis
+PARALLEL_DIRS = ADT ExecutionEngine Support Transforms VMCore
 
 include $(LEVEL)/Makefile.common
 




From daniel at zuster.org  Wed Jul 28 20:51:55 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 01:51:55 -0000
Subject: [llvm-commits] [llvm] r109720 -
	/llvm/trunk/lib/MC/MCParser/AsmParser.cpp
Message-ID: <20100729015155.841FA2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Jul 28 20:51:55 2010
New Revision: 109720

URL: http://llvm.org/viewvc/llvm-project?rev=109720&view=rev
Log:
MC: Destroy Macro instances.

Modified:
    llvm/trunk/lib/MC/MCParser/AsmParser.cpp

Modified: llvm/trunk/lib/MC/MCParser/AsmParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCParser/AsmParser.cpp?rev=109720&r1=109719&r2=109720&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCParser/AsmParser.cpp (original)
+++ llvm/trunk/lib/MC/MCParser/AsmParser.cpp Wed Jul 28 20:51:55 2010
@@ -279,6 +279,13 @@
 }
 
 AsmParser::~AsmParser() {
+  assert(ActiveMacros.empty() && "Unexpected active macro instantiation!");
+
+  // Destroy any macros.
+  for (StringMap<Macro*>::iterator it = MacroMap.begin(),
+         ie = MacroMap.end(); it != ie; ++it)
+    delete it->getValue();
+
   delete PlatformParser;
   delete GenericParser;
 }




From daniel at zuster.org  Wed Jul 28 20:52:04 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 01:52:04 -0000
Subject: [llvm-commits] [llvm] r109721 -
	/llvm/trunk/lib/Support/CrashRecoveryContext.cpp
Message-ID: <20100729015204.60C1E2A6C12C@llvm.org>

Author: ddunbar
Date: Wed Jul 28 20:52:04 2010
New Revision: 109721

URL: http://llvm.org/viewvc/llvm-project?rev=109721&view=rev
Log:
Make sure to include config.h, to pickup LLVM_ON_WIN32.

Modified:
    llvm/trunk/lib/Support/CrashRecoveryContext.cpp

Modified: llvm/trunk/lib/Support/CrashRecoveryContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CrashRecoveryContext.cpp?rev=109721&r1=109720&r2=109721&view=diff
==============================================================================
--- llvm/trunk/lib/Support/CrashRecoveryContext.cpp (original)
+++ llvm/trunk/lib/Support/CrashRecoveryContext.cpp Wed Jul 28 20:52:04 2010
@@ -9,6 +9,7 @@
 
 #include "llvm/Support/CrashRecoveryContext.h"
 #include "llvm/ADT/SmallString.h"
+#include "llvm/Config/config.h"
 #include "llvm/System/ThreadLocal.h"
 #include <setjmp.h>
 #include <cstdio>




From clattner at apple.com  Wed Jul 28 23:27:21 2010
From: clattner at apple.com (Chris Lattner)
Date: Wed, 28 Jul 2010 21:27:21 -0700
Subject: [llvm-commits] [llvm] r109643 - in /llvm/trunk:
	include/llvm/Instructions.h lib/AsmParser/LLParser.cpp
	lib/Transforms/InstCombine/InstCombineCalls.cpp
	lib/VMCore/Instructions.cpp test/Assembler/align-inst-alloca.ll
	test/Assembler/align-inst-load.ll test/Assembler/align-inst-store.ll
	test/Assembler/align-inst.ll
In-Reply-To: <20100728201204.EB9B22A6C12C@llvm.org>
References: <20100728201204.EB9B22A6C12C@llvm.org>
Message-ID: <C314A96D-8E9B-4EBD-AEC6-6322402F1D85@apple.com>


On Jul 28, 2010, at 1:12 PM, Dan Gohman wrote:

> Author: djg
> Date: Wed Jul 28 15:12:04 2010
> New Revision: 109643
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109643&view=rev
> Log:
> Define a maximum supported alignment value for load, store, and
> alloca instructions (constrained by their internal encoding),
> and add error checking for it. Fix an instcombine bug which
> generated huge alignment values (null is infinitely aligned).
> This fixes undefined behavior noticed by John Regehr.

Cool.

> 
> +/// MaximumAlignment - This is the greatest alignment value supported by
> +/// load, store, and alloca instructions.
> +static const unsigned MaximumAlignment = 1u << 29;

How about making this an enum in the Instruction class (so it is llvm::Instruction::MaximumAlignment), instead of being llvm::MaximumAlignment?

-Chris


> +
> //===----------------------------------------------------------------------===//
> //                                AllocaInst Class
> //===----------------------------------------------------------------------===//
> 
> Modified: llvm/trunk/lib/AsmParser/LLParser.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/AsmParser/LLParser.cpp?rev=109643&r1=109642&r2=109643&view=diff
> ==============================================================================
> --- llvm/trunk/lib/AsmParser/LLParser.cpp (original)
> +++ llvm/trunk/lib/AsmParser/LLParser.cpp Wed Jul 28 15:12:04 2010
> @@ -1154,6 +1154,8 @@
>   if (ParseUInt32(Alignment)) return true;
>   if (!isPowerOf2_32(Alignment))
>     return Error(AlignLoc, "alignment is not a power of two");
> +  if (Alignment > MaximumAlignment)
> +    return Error(AlignLoc, "huge alignments are not supported yet");
>   return false;
> }
> 
> @@ -1176,6 +1178,7 @@
>     if (Lex.getKind() != lltok::kw_align)
>       return Error(Lex.getLoc(), "expected metadata or 'align'");
> 
> +    LocTy AlignLoc = Lex.getLoc();
>     if (ParseOptionalAlignment(Alignment)) return true;
>   }
> 
> 
> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=109643&r1=109642&r2=109643&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Wed Jul 28 15:12:04 2010
> @@ -104,11 +104,15 @@
>   ComputeMaskedBits(V, Mask, KnownZero, KnownOne);
>   unsigned TrailZ = KnownZero.countTrailingOnes();
> 
> -  // LLVM doesn't support alignments larger than this currently.
> +  // Avoid trouble with rediculously large TrailZ values, such as
> +  // those computed from a null pointer.
>   TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));
> 
>   unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
> 
> +  // LLVM doesn't support alignments larger than this currently.
> +  Align = std::min(Align, MaximumAlignment);
> +
>   if (PrefAlign > Align)
>     Align = EnforceKnownAlignment(V, Align, PrefAlign);
> 
> 
> Modified: llvm/trunk/lib/VMCore/Instructions.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Instructions.cpp?rev=109643&r1=109642&r2=109643&view=diff
> ==============================================================================
> --- llvm/trunk/lib/VMCore/Instructions.cpp (original)
> +++ llvm/trunk/lib/VMCore/Instructions.cpp Wed Jul 28 15:12:04 2010
> @@ -891,6 +891,8 @@
> 
> void AllocaInst::setAlignment(unsigned Align) {
>   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
> +  assert(Align <= MaximumAlignment &&
> +         "Alignment is greater than MaximumAlignment!");
>   setInstructionSubclassData(Log2_32(Align) + 1);
>   assert(getAlignment() == Align && "Alignment representation error!");
> }
> @@ -1026,8 +1028,11 @@
> 
> void LoadInst::setAlignment(unsigned Align) {
>   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
> +  assert(Align <= MaximumAlignment &&
> +         "Alignment is greater than MaximumAlignment!");
>   setInstructionSubclassData((getSubclassDataFromInstruction() & 1) |
>                              ((Log2_32(Align)+1)<<1));
> +  assert(getAlignment() == Align && "Alignment representation error!");
> }
> 
> //===----------------------------------------------------------------------===//
> @@ -1122,8 +1127,11 @@
> 
> void StoreInst::setAlignment(unsigned Align) {
>   assert((Align & (Align-1)) == 0 && "Alignment is not a power of 2!");
> +  assert(Align <= MaximumAlignment &&
> +         "Alignment is greater than MaximumAlignment!");
>   setInstructionSubclassData((getSubclassDataFromInstruction() & 1) |
>                              ((Log2_32(Align)+1) << 1));
> +  assert(getAlignment() == Align && "Alignment representation error!");
> }
> 
> //===----------------------------------------------------------------------===//
> 
> Added: llvm/trunk/test/Assembler/align-inst-alloca.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-alloca.ll?rev=109643&view=auto
> ==============================================================================
> --- llvm/trunk/test/Assembler/align-inst-alloca.ll (added)
> +++ llvm/trunk/test/Assembler/align-inst-alloca.ll Wed Jul 28 15:12:04 2010
> @@ -0,0 +1,6 @@
> +; RUN: not llvm-as %s -o /dev/null 2>/dev/null
> +
> +define void @foo() {
> +  %p = alloca i1, align 1073741824
> +  ret void
> +}
> 
> Added: llvm/trunk/test/Assembler/align-inst-load.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-load.ll?rev=109643&view=auto
> ==============================================================================
> --- llvm/trunk/test/Assembler/align-inst-load.ll (added)
> +++ llvm/trunk/test/Assembler/align-inst-load.ll Wed Jul 28 15:12:04 2010
> @@ -0,0 +1,6 @@
> +; RUN: not llvm-as %s -o /dev/null 2>/dev/null
> +
> +define void @foo() {
> +  load i1* %p, align 1073741824
> +  ret void
> +}
> 
> Added: llvm/trunk/test/Assembler/align-inst-store.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst-store.ll?rev=109643&view=auto
> ==============================================================================
> --- llvm/trunk/test/Assembler/align-inst-store.ll (added)
> +++ llvm/trunk/test/Assembler/align-inst-store.ll Wed Jul 28 15:12:04 2010
> @@ -0,0 +1,6 @@
> +; RUN: not llvm-as %s -o /dev/null 2>/dev/null
> +
> +define void @foo() {
> +  store i1 false, i1* %p, align 1073741824
> +  ret void
> +}
> 
> Added: llvm/trunk/test/Assembler/align-inst.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Assembler/align-inst.ll?rev=109643&view=auto
> ==============================================================================
> --- llvm/trunk/test/Assembler/align-inst.ll (added)
> +++ llvm/trunk/test/Assembler/align-inst.ll Wed Jul 28 15:12:04 2010
> @@ -0,0 +1,8 @@
> +; RUN: llvm-as %s -o /dev/null
> +
> +define void @foo() {
> +  %p = alloca i1, align 536870912
> +  load i1* %p, align 536870912
> +  store i1 false, i1* %p, align 536870912
> +  ret void
> +}
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From clattner at apple.com  Wed Jul 28 23:30:28 2010
From: clattner at apple.com (Chris Lattner)
Date: Wed, 28 Jul 2010 21:30:28 -0700
Subject: [llvm-commits] [llvm] r109643 - in /llvm/trunk:
	include/llvm/Instructions.h lib/AsmParser/LLParser.cpp
	lib/Transforms/InstCombine/InstCombineCalls.cpp	lib/VMCore/Instructions.cpp
	test/Assembler/align-inst-alloca.ll	test/Assembler/align-inst-load.ll
	test/Assembler/align-inst-store.ll	test/Assembler/align-inst.ll
In-Reply-To: <C314A96D-8E9B-4EBD-AEC6-6322402F1D85@apple.com>
References: <20100728201204.EB9B22A6C12C@llvm.org>
	<C314A96D-8E9B-4EBD-AEC6-6322402F1D85@apple.com>
Message-ID: <C2A0EDFB-EB51-4752-90D1-11EB7803AC64@apple.com>


On Jul 28, 2010, at 9:27 PM, Chris Lattner wrote:

>> 
>> +/// MaximumAlignment - This is the greatest alignment value supported by
>> +/// load, store, and alloca instructions.
>> +static const unsigned MaximumAlignment = 1u << 29;
> 
> How about making this an enum in the Instruction class (so it is llvm::Instruction::MaximumAlignment), instead of being llvm::MaximumAlignment?

Ok, same comment, but how about Value::MaximumAlignment now?

-Chris


From rjmccall at apple.com  Thu Jul 29 02:53:27 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 07:53:27 -0000
Subject: [llvm-commits] [llvm] r109739 - in /llvm/trunk/tools/llvm-diff: ./
 CMakeLists.txt DifferenceEngine.cpp DifferenceEngine.h Makefile
 llvm-diff.cpp
Message-ID: <20100729075327.507082A6C12D@llvm.org>

Author: rjmccall
Date: Thu Jul 29 02:53:27 2010
New Revision: 109739

URL: http://llvm.org/viewvc/llvm-project?rev=109739&view=rev
Log:
Add the llvm-diff tool, which performs a relatively naive structural
diff of a function.  There's a lot of cruft in the current version, and
it's pretty far from perfect, but it's usable.

Currently only capable of comparing functions.  Currently ignores metadata.
Currently ignores most attributes of functions and instructions.

Patches welcome.


Added:
    llvm/trunk/tools/llvm-diff/   (with props)
    llvm/trunk/tools/llvm-diff/CMakeLists.txt
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
    llvm/trunk/tools/llvm-diff/DifferenceEngine.h
    llvm/trunk/tools/llvm-diff/Makefile
    llvm/trunk/tools/llvm-diff/llvm-diff.cpp

Propchange: llvm/trunk/tools/llvm-diff/
------------------------------------------------------------------------------
--- svn:ignore (added)
+++ svn:ignore Thu Jul 29 02:53:27 2010
@@ -0,0 +1,10 @@
+Debug
+Debug+Checks
+Debug+Coverage
+Debug+Coverage-Asserts
+Release
+Release-Asserts
+Release+Coverage
+Debug+Asserts
+Release+Asserts
+

Added: llvm/trunk/tools/llvm-diff/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/CMakeLists.txt?rev=109739&view=auto
==============================================================================
--- llvm/trunk/tools/llvm-diff/CMakeLists.txt (added)
+++ llvm/trunk/tools/llvm-diff/CMakeLists.txt Thu Jul 29 02:53:27 2010
@@ -0,0 +1,6 @@
+set(LLVM_LINK_COMPONENTS support asmparser bitreader)
+
+add_llvm_tool(llvm-diff
+  llvm-diff.cpp
+  DifferenceEngine.cpp
+  )

Added: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109739&view=auto
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (added)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 02:53:27 2010
@@ -0,0 +1,600 @@
+//===-- DifferenceEngine.cpp - Structural function/module comparison ------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This header defines the interface to the LLVM difference engine,
+// which structurally compares functions within a module.
+//
+//===----------------------------------------------------------------------===//
+
+#include <utility>
+
+#include <llvm/ADT/DenseMap.h>
+#include <llvm/ADT/DenseSet.h>
+#include <llvm/ADT/SmallVector.h>
+#include <llvm/ADT/StringRef.h>
+#include <llvm/ADT/StringSet.h>
+
+#include <llvm/Module.h>
+#include <llvm/Function.h>
+#include <llvm/Instructions.h>
+#include <llvm/Support/CFG.h>
+
+#include <llvm/Support/raw_ostream.h>
+#include <llvm/Support/type_traits.h>
+#include <llvm/Support/ErrorHandling.h>
+#include <llvm/Support/CallSite.h>
+
+#include "DifferenceEngine.h"
+
+using namespace llvm;
+
+namespace {
+
+/// A priority queue, implemented as a heap.
+template <class T, class Sorter, unsigned InlineCapacity>
+class PriorityQueue {
+  Sorter Precedes;
+  llvm::SmallVector<T, InlineCapacity> Storage;
+
+public:
+  PriorityQueue(const Sorter &Precedes) : Precedes(Precedes) {}
+
+  /// Checks whether the heap is empty.
+  bool empty() const { return Storage.empty(); }
+
+  /// Insert a new value on the heap.
+  void insert(const T &V) {
+    unsigned Index = Storage.size();
+    Storage.push_back(V);
+    if (Index == 0) return;
+
+    T *data = Storage.data();
+    while (true) {
+      unsigned Target = (Index + 1) / 2 - 1;
+      if (!Precedes(data[Index], data[Target])) return;
+      std::swap(data[Index], data[Target]);
+      if (Target == 0) return;
+      Index = Target;
+    }
+  }
+
+  /// Remove the minimum value in the heap.  Only valid on a non-empty heap.
+  T remove_min() {
+    assert(!empty());
+    T tmp = Storage[0];
+    
+    unsigned NewSize = Storage.size() - 1;
+    if (NewSize) {
+      // Move the slot at the end to the beginning.
+      if (isPodLike<T>::value)
+        Storage[0] = Storage[NewSize];
+      else
+        std::swap(Storage[0], Storage[NewSize]);
+
+      // Bubble the root up as necessary.
+      unsigned Index = 0;
+      while (true) {
+        // With a 1-based index, the children would be Index*2 and Index*2+1.
+        unsigned R = (Index + 1) * 2;
+        unsigned L = R - 1;
+
+        // If R is out of bounds, we're done after this in any case.
+        if (R >= NewSize) {
+          // If L is also out of bounds, we're done immediately.
+          if (L >= NewSize) break;
+
+          // Otherwise, test whether we should swap L and Index.
+          if (Precedes(Storage[L], Storage[Index]))
+            std::swap(Storage[L], Storage[Index]);
+          break;
+        }
+
+        // Otherwise, we need to compare with the smaller of L and R.
+        // Prefer R because it's closer to the end of the array.
+        unsigned IndexToTest = (Precedes(Storage[L], Storage[R]) ? L : R);
+
+        // If Index is >= the min of L and R, then heap ordering is restored.
+        if (!Precedes(Storage[IndexToTest], Storage[Index]))
+          break;
+
+        // Otherwise, keep bubbling up.
+        std::swap(Storage[IndexToTest], Storage[Index]);
+        Index = IndexToTest;
+      }
+    }
+    Storage.pop_back();
+
+    return tmp;
+  }
+};
+
+/// A function-scope difference engine.
+class FunctionDifferenceEngine {
+  DifferenceEngine &Engine;
+
+  /// The current mapping from old local values to new local values.
+  DenseMap<Value*, Value*> Values;
+
+  /// The current mapping from old blocks to new blocks.
+  DenseMap<BasicBlock*, BasicBlock*> Blocks;
+
+  DenseSet<std::pair<Value*, Value*> > TentativeValuePairs;
+
+  unsigned getUnprocPredCount(BasicBlock *Block) const {
+    unsigned Count = 0;
+    for (pred_iterator I = pred_begin(Block), E = pred_end(Block); I != E; ++I)
+      if (!Blocks.count(*I)) Count++;
+    return Count;
+  }
+
+  typedef std::pair<BasicBlock*, BasicBlock*> BlockPair;
+
+  /// A type which sorts a priority queue by the number of unprocessed
+  /// predecessor blocks it has remaining.
+  ///
+  /// This is actually really expensive to calculate.
+  struct QueueSorter {
+    const FunctionDifferenceEngine &fde;
+    explicit QueueSorter(const FunctionDifferenceEngine &fde) : fde(fde) {}
+
+    bool operator()(const BlockPair &Old, const BlockPair &New) {
+      return fde.getUnprocPredCount(Old.first)
+           < fde.getUnprocPredCount(New.first);
+    }
+  };
+
+  /// A queue of unified blocks to process.
+  PriorityQueue<BlockPair, QueueSorter, 20> Queue;
+
+  /// Try to unify the given two blocks.  Enqueues them for processing
+  /// if they haven't already been processed.
+  ///
+  /// Returns true if there was a problem unifying them.
+  bool tryUnify(BasicBlock *L, BasicBlock *R) {
+    BasicBlock *&Ref = Blocks[L];
+
+    if (Ref) {
+      if (Ref == R) return false;
+
+      Engine.logf("successor %s cannot be equivalent to %s; "
+                  "it's already equivalent to %s")
+        << L << R << Ref;
+      return true;
+    }
+
+    Ref = R;
+    Queue.insert(BlockPair(L, R));
+    return false;
+  }
+
+  void processQueue() {
+    while (!Queue.empty()) {
+      BlockPair Pair = Queue.remove_min();
+      diff(Pair.first, Pair.second);
+    }
+  }
+
+  void diff(BasicBlock *L, BasicBlock *R) {
+    DifferenceEngine::Context C(Engine, L, R);
+
+    BasicBlock::iterator LI = L->begin(), LE = L->end();
+    BasicBlock::iterator RI = R->begin(), RE = R->end();
+
+    do {
+      assert(LI != LE && RI != RE);
+      Instruction *LeftI = &*LI, *RightI = &*RI;
+
+      // If the instructions differ, start the more sophisticated diff
+      // algorithm here.
+      if (diff(LeftI, RightI, false, true))
+        return runBlockDiff(LI, RI);
+
+      // Otherwise, unify them.
+      if (!LeftI->use_empty())
+        Values[LeftI] = RightI;
+
+      ++LI, ++RI;
+    } while (LI != LE); // This is sufficient: we can't get equality of
+                        // terminators if there are residual instructions.
+  }
+
+  bool matchForBlockDiff(Instruction *L, Instruction *R);
+  void runBlockDiff(BasicBlock::iterator LI, BasicBlock::iterator RI);
+
+  bool diffCallSites(CallSite L, CallSite R, bool Complain) {
+    // FIXME: call attributes
+    if (!equivalentAsOperands(L.getCalledValue(), R.getCalledValue())) {
+      if (Complain) Engine.log("called functions differ");
+      return true;
+    }
+    if (L.arg_size() != R.arg_size()) {
+      if (Complain) Engine.log("argument counts differ");
+      return true;
+    }
+    for (unsigned I = 0, E = L.arg_size(); I != E; ++I)
+      if (!equivalentAsOperands(L.getArgument(I), R.getArgument(I))) {
+        if (Complain)
+          Engine.logf("arguments %s and %s differ")
+            << L.getArgument(I) << R.getArgument(I);
+        return true;
+      }
+    return false;
+  }
+
+  bool diff(Instruction *L, Instruction *R, bool Complain, bool TryUnify) {
+    // FIXME: metadata (if Complain is set)
+
+    // Different opcodes always imply different operations.
+    if (L->getOpcode() != R->getOpcode()) {
+      if (Complain) Engine.log("different instruction types");
+      return true;
+    }
+
+    if (isa<CmpInst>(L)) {
+      if (cast<CmpInst>(L)->getPredicate()
+            != cast<CmpInst>(R)->getPredicate()) {
+        if (Complain) Engine.log("different predicates");
+        return true;
+      }
+    } else if (isa<CallInst>(L)) {
+      return diffCallSites(CallSite(L), CallSite(R), Complain);
+    } else if (isa<PHINode>(L)) {
+      // FIXME: implement.
+
+      // This is really wierd;  type uniquing is broken?
+      if (L->getType() != R->getType()) {
+        if (!L->getType()->isPointerTy() || !R->getType()->isPointerTy()) {
+          if (Complain) Engine.log("different phi types");
+          return true;
+        }
+      }
+      return false;
+
+    // Terminators.
+    } else if (isa<InvokeInst>(L)) {
+      InvokeInst *LI = cast<InvokeInst>(L);
+      InvokeInst *RI = cast<InvokeInst>(R);
+      if (diffCallSites(CallSite(LI), CallSite(RI), Complain))
+        return true;
+
+      if (TryUnify) {
+        tryUnify(LI->getNormalDest(), RI->getNormalDest());
+        tryUnify(LI->getUnwindDest(), RI->getUnwindDest());
+      }
+      return false;
+
+    } else if (isa<BranchInst>(L)) {
+      BranchInst *LI = cast<BranchInst>(L);
+      BranchInst *RI = cast<BranchInst>(R);
+      if (LI->isConditional() != RI->isConditional()) {
+        if (Complain) Engine.log("branch conditionality differs");
+        return true;
+      }
+
+      if (LI->isConditional()) {
+        if (!equivalentAsOperands(LI->getCondition(), RI->getCondition())) {
+          if (Complain) Engine.log("branch conditions differ");
+          return true;
+        }
+        if (TryUnify) tryUnify(LI->getSuccessor(1), RI->getSuccessor(1));
+      }
+      if (TryUnify) tryUnify(LI->getSuccessor(0), RI->getSuccessor(0));
+      return false;
+
+    } else if (isa<SwitchInst>(L)) {
+      SwitchInst *LI = cast<SwitchInst>(L);
+      SwitchInst *RI = cast<SwitchInst>(R);
+      if (!equivalentAsOperands(LI->getCondition(), RI->getCondition())) {
+        if (Complain) Engine.log("switch conditions differ");
+        return true;
+      }
+      if (TryUnify) tryUnify(LI->getDefaultDest(), RI->getDefaultDest());
+
+      bool Difference = false;
+
+      DenseMap<ConstantInt*,BasicBlock*> LCases;
+      for (unsigned I = 1, E = LI->getNumCases(); I != E; ++I)
+        LCases[LI->getCaseValue(I)] = LI->getSuccessor(I);
+      for (unsigned I = 1, E = RI->getNumCases(); I != E; ++I) {
+        ConstantInt *CaseValue = RI->getCaseValue(I);
+        BasicBlock *LCase = LCases[CaseValue];
+        if (LCase) {
+          if (TryUnify) tryUnify(LCase, RI->getSuccessor(I));
+          LCases.erase(CaseValue);
+        } else if (!Difference) {
+          if (Complain)
+            Engine.logf("right switch has extra case %s") << CaseValue;
+          Difference = true;
+        }
+      }
+      if (!Difference)
+        for (DenseMap<ConstantInt*,BasicBlock*>::iterator
+               I = LCases.begin(), E = LCases.end(); I != E; ++I) {
+          if (Complain)
+            Engine.logf("left switch has extra case %s") << I->first;
+          Difference = true;
+        }
+      return Difference;
+    } else if (isa<UnreachableInst>(L)) {
+      return false;
+    }
+
+    if (L->getNumOperands() != R->getNumOperands()) {
+      if (Complain) Engine.log("instructions have different operand counts");
+      return true;
+    }
+
+    for (unsigned I = 0, E = L->getNumOperands(); I != E; ++I) {
+      Value *LO = L->getOperand(I), *RO = R->getOperand(I);
+      if (!equivalentAsOperands(LO, RO)) {
+        if (Complain) Engine.logf("operands %s and %s differ") << LO << RO;
+        return true;
+      }
+    }
+
+    return false;
+  }
+
+  bool equivalentAsOperands(Constant *L, Constant *R) {
+    // Use equality as a preliminary filter.
+    if (L == R)
+      return true;
+
+    if (L->getValueID() != R->getValueID())
+      return false;
+    
+    // Ask the engine about global values.
+    if (isa<GlobalValue>(L))
+      return Engine.equivalentAsOperands(cast<GlobalValue>(L),
+                                         cast<GlobalValue>(R));
+
+    // Compare constant expressions structurally.
+    if (isa<ConstantExpr>(L))
+      return equivalentAsOperands(cast<ConstantExpr>(L),
+                                  cast<ConstantExpr>(R));
+
+    // Nulls of the "same type" don't always actually have the same
+    // type; I don't know why.  Just white-list them.
+    if (isa<ConstantPointerNull>(L))
+      return true;
+
+    // Block addresses only match if we've already encountered the
+    // block.  FIXME: tentative matches?
+    if (isa<BlockAddress>(L))
+      return Blocks[cast<BlockAddress>(L)->getBasicBlock()]
+                 == cast<BlockAddress>(R)->getBasicBlock();
+
+    return false;
+  }
+
+  bool equivalentAsOperands(ConstantExpr *L, ConstantExpr *R) {
+    if (L == R)
+      return true;
+    if (L->getOpcode() != R->getOpcode())
+      return false;
+
+    switch (L->getOpcode()) {
+    case Instruction::ICmp:
+    case Instruction::FCmp:
+      if (L->getPredicate() != R->getPredicate())
+        return false;
+      break;
+
+    case Instruction::GetElementPtr:
+      // FIXME: inbounds?
+      break;
+
+    default:
+      break;
+    }
+
+    if (L->getNumOperands() != R->getNumOperands())
+      return false;
+
+    for (unsigned I = 0, E = L->getNumOperands(); I != E; ++I)
+      if (!equivalentAsOperands(L->getOperand(I), R->getOperand(I)))
+        return false;
+
+    return true;
+  }
+
+  bool equivalentAsOperands(Value *L, Value *R) {
+    // Fall out if the values have different kind.
+    // This possibly shouldn't take priority over oracles.
+    if (L->getValueID() != R->getValueID())
+      return false;
+
+    // Value subtypes:  Argument, Constant, Instruction, BasicBlock,
+    //                  InlineAsm, MDNode, MDString, PseudoSourceValue
+
+    if (isa<Constant>(L))
+      return equivalentAsOperands(cast<Constant>(L), cast<Constant>(R));
+
+    if (isa<Instruction>(L))
+      return Values[L] == R || TentativeValuePairs.count(std::make_pair(L, R));
+
+    if (isa<Argument>(L))
+      return Values[L] == R;
+
+    if (isa<BasicBlock>(L))
+      return Blocks[cast<BasicBlock>(L)] != R;
+
+    // Pretend everything else is identical.
+    return true;
+  }
+
+  // Avoid a gcc warning about accessing 'this' in an initializer.
+  FunctionDifferenceEngine *this_() { return this; }
+
+public:
+  FunctionDifferenceEngine(DifferenceEngine &Engine) :
+    Engine(Engine), Queue(QueueSorter(*this_())) {}
+
+  void diff(Function *L, Function *R) {
+    if (L->arg_size() != R->arg_size())
+      Engine.log("different argument counts");
+
+    // Map the arguments.
+    for (Function::arg_iterator
+           LI = L->arg_begin(), LE = L->arg_end(),
+           RI = R->arg_begin(), RE = R->arg_end();
+         LI != LE && RI != RE; ++LI, ++RI)
+      Values[&*LI] = &*RI;
+
+    tryUnify(&*L->begin(), &*R->begin());
+    processQueue();
+  }
+};
+
+struct DiffEntry {
+  DiffEntry() : Cost(0) {}
+
+  unsigned Cost;
+  llvm::SmallVector<char, 8> Path; // actually of DifferenceEngine::DiffChange
+};
+
+bool FunctionDifferenceEngine::matchForBlockDiff(Instruction *L,
+                                                 Instruction *R) {
+  return !diff(L, R, false, false);
+}
+
+void FunctionDifferenceEngine::runBlockDiff(BasicBlock::iterator LStart,
+                                            BasicBlock::iterator RStart) {
+  BasicBlock::iterator LE = LStart->getParent()->end();
+  BasicBlock::iterator RE = RStart->getParent()->end();
+
+  unsigned NL = std::distance(LStart, LE);
+
+  SmallVector<DiffEntry, 20> Paths1(NL+1);
+  SmallVector<DiffEntry, 20> Paths2(NL+1);
+
+  DiffEntry *Cur = Paths1.data();
+  DiffEntry *Next = Paths2.data();
+
+  assert(TentativeValuePairs.empty());
+
+  // Initialize the first column.
+  for (unsigned I = 0; I != NL+1; ++I) {
+    Cur[I].Cost = I;
+    for (unsigned J = 0; J != I; ++J)
+      Cur[I].Path.push_back(DifferenceEngine::DC_left);
+  }
+
+  for (BasicBlock::iterator RI = RStart; RI != RE; ++RI) {
+    // Initialize the first row.
+    Next[0] = Cur[0];
+    Next[0].Path.push_back(DifferenceEngine::DC_right);
+
+    unsigned Index = 1;
+    for (BasicBlock::iterator LI = LStart; LI != LE; ++LI, ++Index) {
+      if (matchForBlockDiff(&*LI, &*RI)) {
+        Next[Index] = Cur[Index-1];
+        Next[Index].Path.push_back(DifferenceEngine::DC_match);
+        TentativeValuePairs.insert(std::make_pair(&*LI, &*RI));
+      } else if (Next[Index-1].Cost <= Cur[Index].Cost) {
+        Next[Index] = Next[Index-1];
+        Next[Index].Path.push_back(DifferenceEngine::DC_left);
+      } else {
+        Next[Index] = Cur[Index];
+        Next[Index].Path.push_back(DifferenceEngine::DC_right);
+      }
+    }
+
+    std::swap(Cur, Next);
+  }
+
+  SmallVectorImpl<char> &Path = Cur[NL].Path;
+  BasicBlock::iterator LI = LStart, RI = RStart;
+
+  DifferenceEngine::DiffLogBuilder Diff(Engine);
+
+  // Drop trailing matches.
+  while (Path.back() == DifferenceEngine::DC_match)
+    Path.pop_back();
+
+  for (SmallVectorImpl<char>::iterator
+         PI = Path.begin(), PE = Path.end(); PI != PE; ++PI) {
+    switch (static_cast<DifferenceEngine::DiffChange>(*PI)) {
+    case DifferenceEngine::DC_match:
+      assert(LI != LE && RI != RE);
+      {
+        Instruction *L = &*LI, *R = &*RI;
+        DifferenceEngine::Context C(Engine, L, R);
+        diff(L, R, false, true);
+        Diff.addMatch(L, R);
+      }
+      ++LI; ++RI;
+      break;
+
+    case DifferenceEngine::DC_left:
+      assert(LI != LE);
+      Diff.addLeft(&*LI);
+      ++LI;
+      break;
+
+    case DifferenceEngine::DC_right:
+      assert(RI != RE);
+      Diff.addRight(&*RI);
+      ++RI;
+      break;
+    }
+  }
+
+  TentativeValuePairs.clear();
+}
+
+}
+
+void DifferenceEngine::diff(Function *L, Function *R) {
+  Context C(*this, L, R);
+
+  // FIXME: types
+  // FIXME: attributes and CC
+  // FIXME: parameter attributes
+  
+  // If both are declarations, we're done.
+  if (L->empty() && R->empty())
+    return;
+  else if (L->empty())
+    log("left function is declaration, right function is definition");
+  else if (R->empty())
+    log("right function is declaration, left function is definition");
+  else
+    FunctionDifferenceEngine(*this).diff(L, R);
+}
+
+void DifferenceEngine::diff(Module *L, Module *R) {
+  StringSet<> LNames;
+  SmallVector<std::pair<Function*,Function*>, 20> Queue;
+
+  for (Module::iterator I = L->begin(), E = L->end(); I != E; ++I) {
+    Function *LFn = &*I;
+    LNames.insert(LFn->getName());
+
+    if (Function *RFn = R->getFunction(LFn->getName()))
+      Queue.push_back(std::make_pair(LFn, RFn));
+    else
+      logf("function %s exists only in left module") << LFn;
+  }
+
+  for (Module::iterator I = R->begin(), E = R->end(); I != E; ++I) {
+    Function *RFn = &*I;
+    if (!LNames.count(RFn->getName()))
+      logf("function %s exists only in right module") << RFn;
+  }
+
+  for (SmallVectorImpl<std::pair<Function*,Function*> >::iterator
+         I = Queue.begin(), E = Queue.end(); I != E; ++I)
+    diff(I->first, I->second);
+}
+
+bool DifferenceEngine::equivalentAsOperands(GlobalValue *L, GlobalValue *R) {
+  if (globalValueOracle) return (*globalValueOracle)(L, R);
+  return L->getName() == R->getName();
+}

Added: llvm/trunk/tools/llvm-diff/DifferenceEngine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.h?rev=109739&view=auto
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.h (added)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.h Thu Jul 29 02:53:27 2010
@@ -0,0 +1,171 @@
+//===-- DifferenceEngine.h - Module comparator ------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This header defines the interface to the LLVM difference engine,
+// which structurally compares functions within a module.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LLVM_DIFFERENCE_ENGINE_H_
+#define _LLVM_DIFFERENCE_ENGINE_H_
+
+#include <utility>
+#include <llvm/ADT/SmallVector.h>
+#include <llvm/ADT/StringRef.h>
+
+namespace llvm {
+  class LLVMContext;
+  class Module;
+  class Function;
+  class Twine;
+  class Value;
+  class GlobalValue;
+
+  /// A class for performing structural comparisons of LLVM assembly.
+  class DifferenceEngine {
+  public:
+    /// A temporary-object class for building up log messages.
+    class LogBuilder {
+      DifferenceEngine &Engine;
+
+      /// The use of a stored StringRef here is okay because
+      /// LogBuilder should be used only as a temporary, and as a
+      /// temporary it will be destructed before whatever temporary
+      /// might be initializing this format.
+      StringRef Format;
+
+      SmallVector<Value*, 4> Arguments;
+
+    public:
+      LogBuilder(DifferenceEngine &Engine, StringRef Format)
+        : Engine(Engine), Format(Format) {}
+
+      LogBuilder &operator<<(Value *V) {
+        Arguments.push_back(V);
+        return *this;
+      }
+
+      ~LogBuilder() {
+        Engine.consumer.logf(*this);
+      }
+
+      StringRef getFormat() const { return Format; }
+
+      unsigned getNumArguments() const { return Arguments.size(); }
+      Value *getArgument(unsigned I) const { return Arguments[I]; }
+    };
+
+    enum DiffChange { DC_match, DC_left, DC_right };
+
+    /// A temporary-object class for building up diff messages.
+    class DiffLogBuilder {
+      typedef std::pair<Instruction*,Instruction*> DiffRecord;
+      SmallVector<DiffRecord, 20> Diff;
+
+      DifferenceEngine &Engine;
+
+    public:
+      DiffLogBuilder(DifferenceEngine &Engine) : Engine(Engine) {}
+      ~DiffLogBuilder() { Engine.consumer.logd(*this); }
+
+      void addMatch(Instruction *L, Instruction *R) {
+        Diff.push_back(DiffRecord(L, R));
+      }
+      void addLeft(Instruction *L) { Diff.push_back(DiffRecord(L, 0)); }
+      void addRight(Instruction *R) { Diff.push_back(DiffRecord(0, R)); }
+
+      unsigned getNumLines() const { return Diff.size(); }
+      DiffChange getLineKind(unsigned I) const {
+        return (Diff[I].first ? (Diff[I].second ? DC_match : DC_left)
+                              : DC_right);
+      }
+      Instruction *getLeft(unsigned I) const { return Diff[I].first; }
+      Instruction *getRight(unsigned I) const { return Diff[I].second; }
+    };
+
+    /// The interface for consumers of difference data.
+    struct Consumer {
+      /// Record that a local context has been entered.  Left and
+      /// Right are IR "containers" of some sort which are being
+      /// considered for structural equivalence: global variables,
+      /// functions, blocks, instructions, etc.
+      virtual void enterContext(Value *Left, Value *Right) = 0;
+
+      /// Record that a local context has been exited.
+      virtual void exitContext() = 0;
+
+      /// Record a difference within the current context.
+      virtual void log(StringRef Text) = 0;
+
+      /// Record a formatted difference within the current context.
+      virtual void logf(const LogBuilder &Log) = 0;
+
+      /// Record a line-by-line instruction diff.
+      virtual void logd(const DiffLogBuilder &Log) = 0;
+
+    protected:
+      ~Consumer() {}
+    };
+
+    /// A RAII object for recording the current context.
+    struct Context {
+      Context(DifferenceEngine &Engine, Value *L, Value *R) : Engine(Engine) {
+        Engine.consumer.enterContext(L, R);
+      }
+
+      ~Context() {
+        Engine.consumer.exitContext();
+      }
+
+    private:
+      DifferenceEngine &Engine;
+    };
+
+    /// An oracle for answering whether two values are equivalent as
+    /// operands.
+    struct Oracle {
+      virtual bool operator()(Value *L, Value *R) = 0;
+
+    protected:
+      ~Oracle() {}
+    };
+
+    DifferenceEngine(LLVMContext &context, Consumer &consumer)
+      : context(context), consumer(consumer), globalValueOracle(0) {}
+
+    void diff(Module *L, Module *R);
+    void diff(Function *L, Function *R);
+
+    void log(StringRef text) {
+      consumer.log(text);
+    }
+
+    LogBuilder logf(StringRef text) {
+      return LogBuilder(*this, text);
+    }
+
+    /// Installs an oracle to decide whether two global values are
+    /// equivalent as operands.  Without an oracle, global values are
+    /// considered equivalent as operands precisely when they have the
+    /// same name.
+    void setGlobalValueOracle(Oracle *oracle) {
+      globalValueOracle = oracle;
+    }
+
+    /// Determines whether two global values are equivalent.
+    bool equivalentAsOperands(GlobalValue *L, GlobalValue *R);
+
+  private:
+    LLVMContext &context;
+    Consumer &consumer;
+    Oracle *globalValueOracle;
+  };
+}
+
+#endif

Added: llvm/trunk/tools/llvm-diff/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/Makefile?rev=109739&view=auto
==============================================================================
--- llvm/trunk/tools/llvm-diff/Makefile (added)
+++ llvm/trunk/tools/llvm-diff/Makefile Thu Jul 29 02:53:27 2010
@@ -0,0 +1,17 @@
+##===- tools/llvm-diff/Makefile ----------------------------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL = ../..
+TOOLNAME = llvm-diff
+LINK_COMPONENTS := asmparser bitreader
+
+# This tool has no plugins, optimize startup time.
+TOOL_NO_EXPORTS = 1
+
+include $(LEVEL)/Makefile.common

Added: llvm/trunk/tools/llvm-diff/llvm-diff.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/llvm-diff.cpp?rev=109739&view=auto
==============================================================================
--- llvm/trunk/tools/llvm-diff/llvm-diff.cpp (added)
+++ llvm/trunk/tools/llvm-diff/llvm-diff.cpp Thu Jul 29 02:53:27 2010
@@ -0,0 +1,301 @@
+#include <string>
+#include <utility>
+
+#include <llvm/ADT/StringRef.h>
+#include <llvm/ADT/SmallVector.h>
+#include <llvm/ADT/DenseMap.h>
+
+// Required to parse .ll files.
+#include <llvm/Support/SourceMgr.h>
+#include <llvm/Assembly/Parser.h>
+
+// Required to parse .bc files.
+#include <llvm/Support/MemoryBuffer.h>
+#include <llvm/Bitcode/ReaderWriter.h>
+
+#include <llvm/Support/raw_ostream.h>
+#include <llvm/LLVMContext.h>
+#include <llvm/Module.h>
+#include <llvm/Type.h>
+#include <llvm/Instructions.h>
+
+#include "DifferenceEngine.h"
+
+using namespace llvm;
+
+/// Reads a module from a file.  If the filename ends in .ll, it is
+/// interpreted as an assembly file;  otherwise, it is interpreted as
+/// bitcode.  On error, messages are written to stderr and null is
+/// returned.
+static Module *ReadModule(LLVMContext &Context, StringRef Name) {
+  // LLVM assembly path.
+  if (Name.endswith(".ll")) {
+    SMDiagnostic Diag;
+    Module *M = ParseAssemblyFile(Name, Diag, Context);
+    if (M) return M;
+
+    Diag.Print("llvmdiff", errs());
+    return 0;
+  }
+
+  // Bitcode path.
+  MemoryBuffer *Buffer = MemoryBuffer::getFile(Name);
+
+  // ParseBitcodeFile takes ownership of the buffer if it succeeds.
+  std::string Error;
+  Module *M = ParseBitcodeFile(Buffer, Context, &Error);
+  if (M) return M;
+
+  errs() << "error parsing " << Name << ": " << Error;
+  delete Buffer;
+  return 0;
+}
+
+static int usage() {
+  errs() << "expected usage:\n";
+  errs() << "  llvm-diff oldmodule.ll newmodule.ll [function list]\n";
+  errs() << "Assembly or bitcode modules may be used interchangeably.\n";
+  errs() << "If no functions are provided, all functions will be compared.\n";
+  return 1;
+}
+
+namespace {
+struct DiffContext {
+  DiffContext(Value *L, Value *R)
+    : L(L), R(R), Differences(false), IsFunction(isa<Function>(L)) {}
+  Value *L;
+  Value *R;
+  bool Differences;
+  bool IsFunction;
+  DenseMap<Value*,unsigned> LNumbering;
+  DenseMap<Value*,unsigned> RNumbering;
+};
+
+void ComputeNumbering(Function *F, DenseMap<Value*,unsigned> &Numbering) {
+  unsigned BBN = 0;
+  unsigned IN = 0;
+
+  // Arguments get the first numbers.
+  for (Function::arg_iterator
+         AI = F->arg_begin(), AE = F->arg_end(); AI != AE; ++AI)
+    if (!AI->hasName())
+      Numbering[&*AI] = IN++;
+
+  // Walk the basic blocks in order.
+  for (Function::iterator FI = F->begin(), FE = F->end(); FI != FE; ++FI) {
+    // Basic blocks have their own 'namespace'.
+    if (!FI->hasName())
+      Numbering[&*FI] = BBN++;
+
+    // Walk the instructions in order.
+    for (BasicBlock::iterator BI = FI->begin(), BE = FI->end(); BI != BE; ++BI)
+      // void instructions don't get numbers.
+      if (!BI->hasName() && !BI->getType()->isVoidTy())
+        Numbering[&*BI] = IN++;
+  }
+
+  assert(!Numbering.empty() && "asked for numbering but numbering was no-op");
+}
+
+class DiffConsumer : public DifferenceEngine::Consumer {
+private:
+  Module *LModule;
+  Module *RModule;
+  SmallVector<DiffContext, 5> contexts;
+  bool Differences;
+  unsigned Indent;
+
+  void printValue(Value *V, bool isL) {
+    if (V->hasName()) {
+      errs() << (isa<GlobalValue>(V) ? '@' : '%') << V->getName();
+      return;
+    }
+    if (V->getType()->isVoidTy()) {
+      if (isa<StoreInst>(V)) {
+        errs() << "store to ";
+        printValue(cast<StoreInst>(V)->getPointerOperand(), isL);
+      } else if (isa<CallInst>(V)) {
+        errs() << "call to ";
+        printValue(cast<CallInst>(V)->getCalledValue(), isL);
+      } else if (isa<InvokeInst>(V)) {
+        errs() << "invoke to ";
+        printValue(cast<InvokeInst>(V)->getCalledValue(), isL);
+      } else {
+        errs() << *V;
+      }
+      return;
+    }
+
+    unsigned N = contexts.size();
+    while (N > 0) {
+      --N;
+      DiffContext &ctxt = contexts[N];
+      if (!ctxt.IsFunction) continue;
+      if (isL) {
+        if (ctxt.LNumbering.empty())
+          ComputeNumbering(cast<Function>(ctxt.L), ctxt.LNumbering);
+        errs() << '%' << ctxt.LNumbering[V];
+        return;
+      } else {
+        if (ctxt.RNumbering.empty())
+          ComputeNumbering(cast<Function>(ctxt.R), ctxt.RNumbering);
+        errs() << '%' << ctxt.RNumbering[V];
+        return;
+      }
+    }
+
+    errs() << "<anonymous>";
+  }
+
+  void header() {
+    if (contexts.empty()) return;
+    for (SmallVectorImpl<DiffContext>::iterator
+           I = contexts.begin(), E = contexts.end(); I != E; ++I) {
+      if (I->Differences) continue;
+      if (isa<Function>(I->L)) {
+        // Extra newline between functions.
+        if (Differences) errs() << "\n";
+
+        Function *L = cast<Function>(I->L);
+        Function *R = cast<Function>(I->R);
+        if (L->getName() != R->getName())
+          errs() << "in function " << L->getName() << " / " << R->getName() << ":\n";
+        else
+          errs() << "in function " << L->getName() << ":\n";
+      } else if (isa<BasicBlock>(I->L)) {
+        BasicBlock *L = cast<BasicBlock>(I->L);
+        BasicBlock *R = cast<BasicBlock>(I->R);
+        errs() << "  in block ";
+        printValue(L, true);
+        errs() << " / ";
+        printValue(R, false);
+        errs() << ":\n";
+      } else if (isa<Instruction>(I->L)) {
+        errs() << "    in instruction ";
+        printValue(I->L, true);
+        errs() << " / ";
+        printValue(I->R, false);
+        errs() << ":\n";
+      }
+
+      I->Differences = true;
+    }
+  }
+
+  void indent() {
+    unsigned N = Indent;
+    while (N--) errs() << ' ';
+  }
+
+public:
+  DiffConsumer(Module *L, Module *R)
+    : LModule(L), RModule(R), Differences(false), Indent(0) {}
+
+  bool hadDifferences() const { return Differences; }
+
+  void enterContext(Value *L, Value *R) {
+    contexts.push_back(DiffContext(L, R));
+    Indent += 2;
+  }
+  void exitContext() {
+    Differences |= contexts.back().Differences;
+    contexts.pop_back();
+    Indent -= 2;
+  }
+
+  void log(StringRef text) {
+    header();
+    indent();
+    errs() << text << "\n";
+  }
+
+  void logf(const DifferenceEngine::LogBuilder &Log) {
+    header();
+    indent();
+
+    // FIXME: we don't know whether these are l-values or r-values (ha!)
+    // Print them in some saner way!
+    errs() << Log.getFormat() << "\n";
+    for (unsigned I = 0, E = Log.getNumArguments(); I != E; ++I)
+      Log.getArgument(I)->dump();
+  }
+
+  void logd(const DifferenceEngine::DiffLogBuilder &Log) {
+    header();
+
+    for (unsigned I = 0, E = Log.getNumLines(); I != E; ++I) {
+      indent();
+      switch (Log.getLineKind(I)) {
+      case DifferenceEngine::DC_match:
+        errs() << "  ";
+        Log.getLeft(I)->dump();
+        //printValue(Log.getLeft(I), true);
+        break;
+      case DifferenceEngine::DC_left:
+        errs() << "< ";
+        Log.getLeft(I)->dump();
+        //printValue(Log.getLeft(I), true);
+        break;
+      case DifferenceEngine::DC_right:
+        errs() << "> ";
+        Log.getRight(I)->dump();
+        //printValue(Log.getRight(I), false);
+        break;
+      }
+      //errs() << "\n";
+    }
+  }
+  
+};
+}
+
+int main(int argc, const char **argv) {
+  if (argc < 3) return usage();
+
+  // Don't make StringRef locals like this at home.
+  StringRef LModuleFile = argv[1];
+  StringRef RModuleFile = argv[2];
+
+  LLVMContext Context;
+  
+  // Load both modules.  Die if that fails.
+  Module *LModule = ReadModule(Context, LModuleFile);
+  Module *RModule = ReadModule(Context, RModuleFile);
+  if (!LModule || !RModule) return 1;
+
+  DiffConsumer Consumer(LModule, RModule);
+  DifferenceEngine Engine(Context, Consumer);
+
+  // If any function names were given, just diff those.
+  const char **FnNames = argv + 3;
+  unsigned NumFnNames = argc - 3;
+  if (NumFnNames) {
+    for (unsigned I = 0; I != NumFnNames; ++I) {
+      StringRef FnName = FnNames[I];
+
+      // Drop leading sigils from the function name.
+      if (FnName.startswith("@")) FnName = FnName.substr(1);
+
+      Function *LFn = LModule->getFunction(FnName);
+      Function *RFn = RModule->getFunction(FnName);
+      if (LFn && RFn)
+        Engine.diff(LFn, RFn);
+      else {
+        if (!LFn && !RFn)
+          errs() << "No function named @" << FnName << " in either module\n";
+        else if (!LFn)
+          errs() << "No function named @" << FnName << " in left module\n";
+        else
+          errs() << "No function named @" << FnName << " in right module\n";
+      }
+    }
+  } else {
+    // Otherwise, diff all functions in the modules.
+    Engine.diff(LModule, RModule);
+  }
+
+  delete LModule;
+  delete RModule;
+
+  return Consumer.hadDifferences();
+}




From rjmccall at apple.com  Thu Jul 29 03:14:41 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 08:14:41 -0000
Subject: [llvm-commits] [llvm] r109740 - in /llvm/trunk/tools/llvm-diff:
 DifferenceEngine.cpp llvm-diff.cpp
Message-ID: <20100729081441.A46822A6C12D@llvm.org>

Author: rjmccall
Date: Thu Jul 29 03:14:41 2010
New Revision: 109740

URL: http://llvm.org/viewvc/llvm-project?rev=109740&view=rev
Log:
Cache the result of errs() and implement formatted logging.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
    llvm/trunk/tools/llvm-diff/llvm-diff.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109740&r1=109739&r2=109740&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 03:14:41 2010
@@ -162,8 +162,8 @@
     if (Ref) {
       if (Ref == R) return false;
 
-      Engine.logf("successor %s cannot be equivalent to %s; "
-                  "it's already equivalent to %s")
+      Engine.logf("successor %l cannot be equivalent to %r; "
+                  "it's already equivalent to %r")
         << L << R << Ref;
       return true;
     }
@@ -220,7 +220,7 @@
     for (unsigned I = 0, E = L.arg_size(); I != E; ++I)
       if (!equivalentAsOperands(L.getArgument(I), R.getArgument(I))) {
         if (Complain)
-          Engine.logf("arguments %s and %s differ")
+          Engine.logf("arguments %l and %r differ")
             << L.getArgument(I) << R.getArgument(I);
         return true;
       }
@@ -309,7 +309,7 @@
           LCases.erase(CaseValue);
         } else if (!Difference) {
           if (Complain)
-            Engine.logf("right switch has extra case %s") << CaseValue;
+            Engine.logf("right switch has extra case %r") << CaseValue;
           Difference = true;
         }
       }
@@ -317,7 +317,7 @@
         for (DenseMap<ConstantInt*,BasicBlock*>::iterator
                I = LCases.begin(), E = LCases.end(); I != E; ++I) {
           if (Complain)
-            Engine.logf("left switch has extra case %s") << I->first;
+            Engine.logf("left switch has extra case %l") << I->first;
           Difference = true;
         }
       return Difference;
@@ -333,7 +333,7 @@
     for (unsigned I = 0, E = L->getNumOperands(); I != E; ++I) {
       Value *LO = L->getOperand(I), *RO = R->getOperand(I);
       if (!equivalentAsOperands(LO, RO)) {
-        if (Complain) Engine.logf("operands %s and %s differ") << LO << RO;
+        if (Complain) Engine.logf("operands %l and %r differ") << LO << RO;
         return true;
       }
     }
@@ -580,13 +580,13 @@
     if (Function *RFn = R->getFunction(LFn->getName()))
       Queue.push_back(std::make_pair(LFn, RFn));
     else
-      logf("function %s exists only in left module") << LFn;
+      logf("function %l exists only in left module") << LFn;
   }
 
   for (Module::iterator I = R->begin(), E = R->end(); I != E; ++I) {
     Function *RFn = &*I;
     if (!LNames.count(RFn->getName()))
-      logf("function %s exists only in right module") << RFn;
+      logf("function %r exists only in right module") << RFn;
   }
 
   for (SmallVectorImpl<std::pair<Function*,Function*> >::iterator

Modified: llvm/trunk/tools/llvm-diff/llvm-diff.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/llvm-diff.cpp?rev=109740&r1=109739&r2=109740&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/llvm-diff.cpp (original)
+++ llvm/trunk/tools/llvm-diff/llvm-diff.cpp Thu Jul 29 03:14:41 2010
@@ -14,6 +14,7 @@
 #include <llvm/Bitcode/ReaderWriter.h>
 
 #include <llvm/Support/raw_ostream.h>
+#include <llvm/Support/ErrorHandling.h>
 #include <llvm/LLVMContext.h>
 #include <llvm/Module.h>
 #include <llvm/Type.h>
@@ -99,6 +100,7 @@
 
 class DiffConsumer : public DifferenceEngine::Consumer {
 private:
+  raw_ostream &out;
   Module *LModule;
   Module *RModule;
   SmallVector<DiffContext, 5> contexts;
@@ -107,21 +109,21 @@
 
   void printValue(Value *V, bool isL) {
     if (V->hasName()) {
-      errs() << (isa<GlobalValue>(V) ? '@' : '%') << V->getName();
+      out << (isa<GlobalValue>(V) ? '@' : '%') << V->getName();
       return;
     }
     if (V->getType()->isVoidTy()) {
       if (isa<StoreInst>(V)) {
-        errs() << "store to ";
+        out << "store to ";
         printValue(cast<StoreInst>(V)->getPointerOperand(), isL);
       } else if (isa<CallInst>(V)) {
-        errs() << "call to ";
+        out << "call to ";
         printValue(cast<CallInst>(V)->getCalledValue(), isL);
       } else if (isa<InvokeInst>(V)) {
-        errs() << "invoke to ";
+        out << "invoke to ";
         printValue(cast<InvokeInst>(V)->getCalledValue(), isL);
       } else {
-        errs() << *V;
+        out << *V;
       }
       return;
     }
@@ -134,17 +136,17 @@
       if (isL) {
         if (ctxt.LNumbering.empty())
           ComputeNumbering(cast<Function>(ctxt.L), ctxt.LNumbering);
-        errs() << '%' << ctxt.LNumbering[V];
+        out << '%' << ctxt.LNumbering[V];
         return;
       } else {
         if (ctxt.RNumbering.empty())
           ComputeNumbering(cast<Function>(ctxt.R), ctxt.RNumbering);
-        errs() << '%' << ctxt.RNumbering[V];
+        out << '%' << ctxt.RNumbering[V];
         return;
       }
     }
 
-    errs() << "<anonymous>";
+    out << "<anonymous>";
   }
 
   void header() {
@@ -154,28 +156,28 @@
       if (I->Differences) continue;
       if (isa<Function>(I->L)) {
         // Extra newline between functions.
-        if (Differences) errs() << "\n";
+        if (Differences) out << "\n";
 
         Function *L = cast<Function>(I->L);
         Function *R = cast<Function>(I->R);
         if (L->getName() != R->getName())
-          errs() << "in function " << L->getName() << " / " << R->getName() << ":\n";
+          out << "in function " << L->getName() << " / " << R->getName() << ":\n";
         else
-          errs() << "in function " << L->getName() << ":\n";
+          out << "in function " << L->getName() << ":\n";
       } else if (isa<BasicBlock>(I->L)) {
         BasicBlock *L = cast<BasicBlock>(I->L);
         BasicBlock *R = cast<BasicBlock>(I->R);
-        errs() << "  in block ";
+        out << "  in block ";
         printValue(L, true);
-        errs() << " / ";
+        out << " / ";
         printValue(R, false);
-        errs() << ":\n";
+        out << ":\n";
       } else if (isa<Instruction>(I->L)) {
-        errs() << "    in instruction ";
+        out << "    in instruction ";
         printValue(I->L, true);
-        errs() << " / ";
+        out << " / ";
         printValue(I->R, false);
-        errs() << ":\n";
+        out << ":\n";
       }
 
       I->Differences = true;
@@ -184,12 +186,12 @@
 
   void indent() {
     unsigned N = Indent;
-    while (N--) errs() << ' ';
+    while (N--) out << ' ';
   }
 
 public:
   DiffConsumer(Module *L, Module *R)
-    : LModule(L), RModule(R), Differences(false), Indent(0) {}
+    : out(errs()), LModule(L), RModule(R), Differences(false), Indent(0) {}
 
   bool hadDifferences() const { return Differences; }
 
@@ -206,18 +208,37 @@
   void log(StringRef text) {
     header();
     indent();
-    errs() << text << "\n";
+    out << text << '\n';
   }
 
   void logf(const DifferenceEngine::LogBuilder &Log) {
     header();
     indent();
 
-    // FIXME: we don't know whether these are l-values or r-values (ha!)
-    // Print them in some saner way!
-    errs() << Log.getFormat() << "\n";
-    for (unsigned I = 0, E = Log.getNumArguments(); I != E; ++I)
-      Log.getArgument(I)->dump();
+    unsigned arg = 0;
+
+    StringRef format = Log.getFormat();
+    while (true) {
+      size_t percent = format.find('%');
+      if (percent == StringRef::npos) {
+        out << format;
+        break;
+      }
+      assert(format[percent] == '%');
+
+      if (percent > 0) out << format.substr(0, percent);
+
+      switch (format[percent+1]) {
+      case '%': out << '%'; break;
+      case 'l': printValue(Log.getArgument(arg++), true); break;
+      case 'r': printValue(Log.getArgument(arg++), false); break;
+      default: llvm_unreachable("unknown format character");
+      }
+
+      format = format.substr(percent+2);
+    }
+
+    out << '\n';
   }
 
   void logd(const DifferenceEngine::DiffLogBuilder &Log) {
@@ -227,22 +248,22 @@
       indent();
       switch (Log.getLineKind(I)) {
       case DifferenceEngine::DC_match:
-        errs() << "  ";
+        out << "  ";
         Log.getLeft(I)->dump();
         //printValue(Log.getLeft(I), true);
         break;
       case DifferenceEngine::DC_left:
-        errs() << "< ";
+        out << "< ";
         Log.getLeft(I)->dump();
         //printValue(Log.getLeft(I), true);
         break;
       case DifferenceEngine::DC_right:
-        errs() << "> ";
+        out << "> ";
         Log.getRight(I)->dump();
         //printValue(Log.getRight(I), false);
         break;
       }
-      //errs() << "\n";
+      //out << "\n";
     }
   }
   




From rjmccall at apple.com  Thu Jul 29 03:53:59 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 08:53:59 -0000
Subject: [llvm-commits] [llvm] r109741 -
	/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
Message-ID: <20100729085359.AE8112A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 03:53:59 2010
New Revision: 109741

URL: http://llvm.org/viewvc/llvm-project?rev=109741&view=rev
Log:
Somehow I was getting reasonable results for the test cases I was interested
in despite not ever incrementing any path costs, so that the only nonzero costs
arose from the all-left path in the first column.  Anyway.  Perform the diff
starting from the beginning of the block to avoid capturing (say) loads of
allocas.

Vastly improves diff results on code that hasn't been mem2reg'ed.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109741&r1=109740&r2=109741&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 03:53:59 2010
@@ -124,7 +124,7 @@
   /// The current mapping from old blocks to new blocks.
   DenseMap<BasicBlock*, BasicBlock*> Blocks;
 
-  DenseSet<std::pair<Value*, Value*> > TentativeValuePairs;
+  DenseSet<std::pair<Value*, Value*> > TentativeValues;
 
   unsigned getUnprocPredCount(BasicBlock *Block) const {
     unsigned Count = 0;
@@ -186,22 +186,32 @@
     BasicBlock::iterator LI = L->begin(), LE = L->end();
     BasicBlock::iterator RI = R->begin(), RE = R->end();
 
+    llvm::SmallVector<std::pair<Instruction*,Instruction*>, 20> TentativePairs;
+
     do {
       assert(LI != LE && RI != RE);
       Instruction *LeftI = &*LI, *RightI = &*RI;
 
       // If the instructions differ, start the more sophisticated diff
-      // algorithm here.
-      if (diff(LeftI, RightI, false, true))
-        return runBlockDiff(LI, RI);
+      // algorithm at the start of the block.
+      if (diff(LeftI, RightI, false, true)) {
+        TentativeValues.clear();
+        return runBlockDiff(L->begin(), R->begin());
+      }
 
-      // Otherwise, unify them.
+      // Otherwise, tentatively unify them.
       if (!LeftI->use_empty())
-        Values[LeftI] = RightI;
+        TentativeValues.insert(std::make_pair(LeftI, RightI));
 
       ++LI, ++RI;
     } while (LI != LE); // This is sufficient: we can't get equality of
                         // terminators if there are residual instructions.
+
+    // Make all the tentative pairs solid.
+    for (llvm::DenseSet<std::pair<Value*,Value*> >::iterator
+           I = TentativeValues.begin(), E = TentativeValues.end(); I != E; ++I)
+      Values[I->first] = I->second;
+    TentativeValues.clear();
   }
 
   bool matchForBlockDiff(Instruction *L, Instruction *R);
@@ -417,7 +427,7 @@
       return equivalentAsOperands(cast<Constant>(L), cast<Constant>(R));
 
     if (isa<Instruction>(L))
-      return Values[L] == R || TentativeValuePairs.count(std::make_pair(L, R));
+      return Values[L] == R || TentativeValues.count(std::make_pair(L, R));
 
     if (isa<Argument>(L))
       return Values[L] == R;
@@ -477,11 +487,15 @@
   DiffEntry *Cur = Paths1.data();
   DiffEntry *Next = Paths2.data();
 
-  assert(TentativeValuePairs.empty());
+  const unsigned LeftCost = 2;
+  const unsigned RightCost = 2;
+  const unsigned MatchCost = 0;
+
+  assert(TentativeValues.empty());
 
   // Initialize the first column.
   for (unsigned I = 0; I != NL+1; ++I) {
-    Cur[I].Cost = I;
+    Cur[I].Cost = I * LeftCost;
     for (unsigned J = 0; J != I; ++J)
       Cur[I].Path.push_back(DifferenceEngine::DC_left);
   }
@@ -489,19 +503,23 @@
   for (BasicBlock::iterator RI = RStart; RI != RE; ++RI) {
     // Initialize the first row.
     Next[0] = Cur[0];
+    Next[0].Cost += RightCost;
     Next[0].Path.push_back(DifferenceEngine::DC_right);
 
     unsigned Index = 1;
     for (BasicBlock::iterator LI = LStart; LI != LE; ++LI, ++Index) {
       if (matchForBlockDiff(&*LI, &*RI)) {
         Next[Index] = Cur[Index-1];
+        Next[Index].Cost += MatchCost;
         Next[Index].Path.push_back(DifferenceEngine::DC_match);
-        TentativeValuePairs.insert(std::make_pair(&*LI, &*RI));
+        TentativeValues.insert(std::make_pair(&*LI, &*RI));
       } else if (Next[Index-1].Cost <= Cur[Index].Cost) {
         Next[Index] = Next[Index-1];
+        Next[Index].Cost += LeftCost;
         Next[Index].Path.push_back(DifferenceEngine::DC_left);
       } else {
         Next[Index] = Cur[Index];
+        Next[Index].Cost += RightCost;
         Next[Index].Path.push_back(DifferenceEngine::DC_right);
       }
     }
@@ -518,15 +536,21 @@
   while (Path.back() == DifferenceEngine::DC_match)
     Path.pop_back();
 
-  for (SmallVectorImpl<char>::iterator
-         PI = Path.begin(), PE = Path.end(); PI != PE; ++PI) {
+  // Skip leading matches.
+  SmallVectorImpl<char>::iterator
+    PI = Path.begin(), PE = Path.end();
+  while (PI != PE && *PI == DifferenceEngine::DC_match)
+    ++PI, ++LI, ++RI;
+
+  for (; PI != PE; ++PI) {
     switch (static_cast<DifferenceEngine::DiffChange>(*PI)) {
     case DifferenceEngine::DC_match:
       assert(LI != LE && RI != RE);
       {
         Instruction *L = &*LI, *R = &*RI;
         DifferenceEngine::Context C(Engine, L, R);
-        diff(L, R, false, true);
+        diff(L, R, false, true); // unify successors
+        Values[L] = R; // make non-tentative
         Diff.addMatch(L, R);
       }
       ++LI; ++RI;
@@ -546,7 +570,7 @@
     }
   }
 
-  TentativeValuePairs.clear();
+  TentativeValues.clear();
 }
 
 }




From rjmccall at apple.com  Thu Jul 29 03:59:28 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 08:59:28 -0000
Subject: [llvm-commits] [llvm] r109742 -
	/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
Message-ID: <20100729085928.25D3A2A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 03:59:27 2010
New Revision: 109742

URL: http://llvm.org/viewvc/llvm-project?rev=109742&view=rev
Log:
When unifying instructions during a block diff, actually complain about
any differences we see.  This should only happen if there are "non-structural"
differences between the instructions, i.e. differences which wouldn't cause
diff to return true.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109742&r1=109741&r2=109742&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 03:59:27 2010
@@ -549,7 +549,7 @@
       {
         Instruction *L = &*LI, *R = &*RI;
         DifferenceEngine::Context C(Engine, L, R);
-        diff(L, R, false, true); // unify successors
+        diff(L, R, true, true); // complain and unify successors
         Values[L] = R; // make non-tentative
         Diff.addMatch(L, R);
       }




From rjmccall at apple.com  Thu Jul 29 04:04:45 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 09:04:45 -0000
Subject: [llvm-commits] [llvm] r109743 -
	/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
Message-ID: <20100729090445.A4BEF2A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 04:04:45 2010
New Revision: 109743

URL: http://llvm.org/viewvc/llvm-project?rev=109743&view=rev
Log:
Diagnose non-structural differences in the case where blocks were
structurally identical.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109743&r1=109742&r2=109743&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 04:04:45 2010
@@ -194,7 +194,7 @@
 
       // If the instructions differ, start the more sophisticated diff
       // algorithm at the start of the block.
-      if (diff(LeftI, RightI, false, true)) {
+      if (diff(LeftI, RightI, false, false)) {
         TentativeValues.clear();
         return runBlockDiff(L->begin(), R->begin());
       }
@@ -207,11 +207,22 @@
     } while (LI != LE); // This is sufficient: we can't get equality of
                         // terminators if there are residual instructions.
 
-    // Make all the tentative pairs solid.
-    for (llvm::DenseSet<std::pair<Value*,Value*> >::iterator
-           I = TentativeValues.begin(), E = TentativeValues.end(); I != E; ++I)
-      Values[I->first] = I->second;
     TentativeValues.clear();
+
+    // Do another pass over the block, this time in complaints mode.
+    LI = L->begin(); RI = R->begin();
+    do {
+      assert(LI != LE && RI != RE);
+      bool Result = diff(&*LI, &*RI, true, true);
+      assert(!Result && "structural differences second time around?");
+      (void) Result;
+
+      // Make the mapping non-tentative this time.
+      if (!LI->use_empty())
+        Values[&*LI] = &*RI;
+      
+      ++LI, ++RI;
+    } while (LI != LE);
   }
 
   bool matchForBlockDiff(Instruction *L, Instruction *R);




From rjmccall at apple.com  Thu Jul 29 04:20:34 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 09:20:34 -0000
Subject: [llvm-commits] [llvm] r109744 -
	/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
Message-ID: <20100729092034.3E4562A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 04:20:34 2010
New Revision: 109744

URL: http://llvm.org/viewvc/llvm-project?rev=109744&view=rev
Log:
Centralize the logic to permanently unify two instructions and make sure
it establishes a context and does a complaining diff.  Also make sure we
unify the prelude and postlude of a diff after a block-diff call.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109744&r1=109743&r2=109744&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 04:20:34 2010
@@ -172,6 +172,18 @@
     Queue.insert(BlockPair(L, R));
     return false;
   }
+  
+  /// Unifies two instructions, given that they're known not to have
+  /// structural differences.
+  void unify(Instruction *L, Instruction *R) {
+    DifferenceEngine::Context C(Engine, L, R);
+
+    bool Result = diff(L, R, true, true);
+    assert(!Result && "structural differences second time around?");
+    (void) Result;
+    if (!L->use_empty())
+      Values[L] = R;
+  }
 
   void processQueue() {
     while (!Queue.empty()) {
@@ -207,22 +219,10 @@
     } while (LI != LE); // This is sufficient: we can't get equality of
                         // terminators if there are residual instructions.
 
+    // Unify everything in the block, non-tentatively this time.
     TentativeValues.clear();
-
-    // Do another pass over the block, this time in complaints mode.
-    LI = L->begin(); RI = R->begin();
-    do {
-      assert(LI != LE && RI != RE);
-      bool Result = diff(&*LI, &*RI, true, true);
-      assert(!Result && "structural differences second time around?");
-      (void) Result;
-
-      // Make the mapping non-tentative this time.
-      if (!LI->use_empty())
-        Values[&*LI] = &*RI;
-      
-      ++LI, ++RI;
-    } while (LI != LE);
+    for (LI = L->begin(), RI = R->begin(); LI != LE; ++LI, ++RI)
+      unify(&*LI, &*RI);
   }
 
   bool matchForBlockDiff(Instruction *L, Instruction *R);
@@ -538,6 +538,10 @@
     std::swap(Cur, Next);
   }
 
+  // We don't need the tentative values anymore; everything from here
+  // on out should be non-tentative.
+  TentativeValues.clear();
+
   SmallVectorImpl<char> &Path = Cur[NL].Path;
   BasicBlock::iterator LI = LStart, RI = RStart;
 
@@ -550,8 +554,10 @@
   // Skip leading matches.
   SmallVectorImpl<char>::iterator
     PI = Path.begin(), PE = Path.end();
-  while (PI != PE && *PI == DifferenceEngine::DC_match)
+  while (PI != PE && *PI == DifferenceEngine::DC_match) {
+    unify(&*LI, &*RI);
     ++PI, ++LI, ++RI;
+  }
 
   for (; PI != PE; ++PI) {
     switch (static_cast<DifferenceEngine::DiffChange>(*PI)) {
@@ -559,9 +565,7 @@
       assert(LI != LE && RI != RE);
       {
         Instruction *L = &*LI, *R = &*RI;
-        DifferenceEngine::Context C(Engine, L, R);
-        diff(L, R, true, true); // complain and unify successors
-        Values[L] = R; // make non-tentative
+        unify(L, R);
         Diff.addMatch(L, R);
       }
       ++LI; ++RI;
@@ -581,7 +585,13 @@
     }
   }
 
-  TentativeValues.clear();
+  // Finishing unifying and complaining about the tails of the block,
+  // which should be matches all the way through.
+  while (LI != LE) {
+    assert(RI != RE);
+    unify(&*LI, &*RI);
+    ++LI, ++RI;
+  }
 }
 
 }




From benny.kra at googlemail.com  Thu Jul 29 06:57:59 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 11:57:59 -0000
Subject: [llvm-commits] [llvm] r109745 -
	/llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp
Message-ID: <20100729115759.6F0182A6C12C@llvm.org>

Author: d0k
Date: Thu Jul 29 06:57:59 2010
New Revision: 109745

URL: http://llvm.org/viewvc/llvm-project?rev=109745&view=rev
Log:
COFFObjectWriter: Don't leak COFFSymbols and COFFSections.

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

Modified: llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp?rev=109745&r1=109744&r2=109745&view=diff
==============================================================================
--- llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp (original)
+++ llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp Thu Jul 29 06:57:59 2010
@@ -131,6 +131,7 @@
   symbol_map  SymbolMap;
 
   WinCOFFObjectWriter(raw_ostream &OS);
+  ~WinCOFFObjectWriter();
 
   COFFSymbol *createSymbol(llvm::StringRef Name);
   COFFSection *createSection(llvm::StringRef Name);
@@ -275,6 +276,13 @@
   Header.Machine = 0x14C; // x86
 }
 
+WinCOFFObjectWriter::~WinCOFFObjectWriter() {
+  for (symbols::iterator I = Symbols.begin(), E = Symbols.end(); I != E; ++I)
+    delete *I;
+  for (sections::iterator I = Sections.begin(), E = Sections.end(); I != E; ++I)
+    delete *I;
+}
+
 COFFSymbol *WinCOFFObjectWriter::createSymbol(llvm::StringRef Name) {
   return createCOFFEntity<COFFSymbol>(Name, Symbols);
 }




From benny.kra at googlemail.com  Thu Jul 29 08:53:19 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 13:53:19 -0000
Subject: [llvm-commits] [llvm] r109746 - in /llvm/trunk:
 include/llvm/MC/MCContext.h include/llvm/MC/MCDwarf.h lib/MC/MCContext.cpp
Message-ID: <20100729135320.010FD2A6C12C@llvm.org>

Author: d0k
Date: Thu Jul 29 08:53:19 2010
New Revision: 109746

URL: http://llvm.org/viewvc/llvm-project?rev=109746&view=rev
Log:
Stop leaking std::strings in GetDwarfFile.

Modified:
    llvm/trunk/include/llvm/MC/MCContext.h
    llvm/trunk/include/llvm/MC/MCDwarf.h
    llvm/trunk/lib/MC/MCContext.cpp

Modified: llvm/trunk/include/llvm/MC/MCContext.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCContext.h?rev=109746&r1=109745&r2=109746&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCContext.h (original)
+++ llvm/trunk/include/llvm/MC/MCContext.h Thu Jul 29 08:53:19 2010
@@ -70,7 +70,7 @@
 
     /// The dwarf file and directory tables from the dwarf .file directive.
     std::vector<MCDwarfFile *> MCDwarfFiles;
-    std::vector<std::string *> MCDwarfDirs;
+    std::vector<StringRef> MCDwarfDirs;
 
     /// Allocator - Allocator object used for creating machine code objects.
     ///

Modified: llvm/trunk/include/llvm/MC/MCDwarf.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCDwarf.h?rev=109746&r1=109745&r2=109746&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCDwarf.h (original)
+++ llvm/trunk/include/llvm/MC/MCDwarf.h Thu Jul 29 08:53:19 2010
@@ -16,7 +16,7 @@
 #ifndef LLVM_MC_MCDWARF_H
 #define LLVM_MC_MCDWARF_H
 
-#include <string>
+#include "llvm/ADT/StringRef.h"
 
 namespace llvm {
   class MCContext;
@@ -29,21 +29,22 @@
   /// index 0 is not used and not a valid dwarf file number).
   class MCDwarfFile {
     // Name - the base name of the file without its directory path.
-    std::string Name;
+    // The StringRef references memory allocated in the MCContext.
+    StringRef Name;
 
     // DirIndex - the index into the list of directory names for this file name.
     unsigned DirIndex;
 
   private:  // MCContext creates and uniques these.
     friend class MCContext;
-    MCDwarfFile(std::string name, unsigned dirIndex)
+    MCDwarfFile(StringRef name, unsigned dirIndex)
       : Name(name), DirIndex(dirIndex) {}
 
     MCDwarfFile(const MCDwarfFile&);       // DO NOT IMPLEMENT
     void operator=(const MCDwarfFile&); // DO NOT IMPLEMENT
   public:
     /// getName - Get the base name of this MCDwarfFile.
-    std::string getName() const { return Name; }
+    StringRef getName() const { return Name; }
 
     /// print - Print the value to the stream \arg OS.
     void print(raw_ostream &OS) const;

Modified: llvm/trunk/lib/MC/MCContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCContext.cpp?rev=109746&r1=109745&r2=109746&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCContext.cpp (original)
+++ llvm/trunk/lib/MC/MCContext.cpp Thu Jul 29 08:53:19 2010
@@ -213,7 +213,6 @@
   std::pair<StringRef, StringRef> Slash = FileName.rsplit('/');
 
   // Find or make a entry in the MCDwarfDirs vector for this Directory.
-  StringRef Directory;
   StringRef Name;
   unsigned DirIndex;
   // Capture directory name.
@@ -221,23 +220,24 @@
     Name = Slash.first;
     DirIndex = 0; // For FileNames with no directories a DirIndex of 0 is used.
   } else {
-    Directory = Slash.first;
+    StringRef Directory = Slash.first;
     Name = Slash.second;
     for (DirIndex = 1; DirIndex < MCDwarfDirs.size(); DirIndex++) {
-      std::string *&Dir = MCDwarfDirs[DirIndex];
-      if (Directory == *Dir)
+      if (Directory == MCDwarfDirs[DirIndex])
 	break;
     }
     if (DirIndex >= MCDwarfDirs.size()) {
-      MCDwarfDirs.resize(DirIndex + 1);
-      std::string *&NewDir = MCDwarfDirs[DirIndex];
-      NewDir = new (*this) std::string(Directory);
+      char *Buf = static_cast<char *>(Allocate(Directory.size()));
+      memcpy(Buf, Directory.data(), Directory.size());
+      MCDwarfDirs.push_back(StringRef(Buf, Directory.size()));
     }
   }
   
   // Now make the MCDwarfFile entry and place it in the slot in the MCDwarfFiles
   // vector.
-  File = new (*this) MCDwarfFile(Name, DirIndex);
+  char *Buf = static_cast<char *>(Allocate(Name.size()));
+  memcpy(Buf, Name.data(), Name.size());
+  File = new (*this) MCDwarfFile(StringRef(Buf, Name.size()), DirIndex);
 
   // return the allocated FileNumber.
   return FileNumber;




From rafael.espindola at gmail.com  Thu Jul 29 09:20:59 2010
From: rafael.espindola at gmail.com (Rafael Espindola)
Date: Thu, 29 Jul 2010 14:20:59 -0000
Subject: [llvm-commits] [llvm] r109747 -
	/llvm/trunk/tools/bugpoint/Miscompilation.cpp
Message-ID: <20100729142059.58A182A6C12C@llvm.org>

Author: rafael
Date: Thu Jul 29 09:20:59 2010
New Revision: 109747

URL: http://llvm.org/viewvc/llvm-project?rev=109747&view=rev
Log:
Make the test while reducing blocks functional. This avoids accessing freed
memory when one of the original BB is destroyed.

Modified:
    llvm/trunk/tools/bugpoint/Miscompilation.cpp

Modified: llvm/trunk/tools/bugpoint/Miscompilation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/Miscompilation.cpp?rev=109747&r1=109746&r2=109747&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/Miscompilation.cpp (original)
+++ llvm/trunk/tools/bugpoint/Miscompilation.cpp Thu Jul 29 09:20:59 2010
@@ -470,18 +470,36 @@
 
   // Split the module into the two halves of the program we want.
   ValueMap<const Value*, Value*> VMap;
+  Module *Clone = CloneModule(BD.getProgram(), VMap);
+  Module *Orig = BD.swapProgramIn(Clone);
+  std::vector<Function*> FuncsOnClone;
+  std::vector<BasicBlock*> BBsOnClone;
+  for (unsigned i = 0, e = FunctionsBeingTested.size(); i != e; ++i) {
+    Function *F = cast<Function>(VMap[FunctionsBeingTested[i]]);
+    FuncsOnClone.push_back(F);
+  }
+  for (unsigned i = 0, e = BBs.size(); i != e; ++i) {
+    BasicBlock *BB = cast<BasicBlock>(VMap[BBs[i]]);
+    BBsOnClone.push_back(BB);
+  }
+  VMap.clear();
+
   Module *ToNotOptimize = CloneModule(BD.getProgram(), VMap);
   Module *ToOptimize = SplitFunctionsOutOfModule(ToNotOptimize,
-                                                 FunctionsBeingTested,
+                                                 FuncsOnClone,
                                                  VMap);
 
   // Try the extraction.  If it doesn't work, then the block extractor crashed
   // or something, in which case bugpoint can't chase down this possibility.
-  if (Module *New = BD.ExtractMappedBlocksFromModule(BBs, ToOptimize)) {
+  if (Module *New = BD.ExtractMappedBlocksFromModule(BBsOnClone, ToOptimize)) {
     delete ToOptimize;
-    // Run the predicate, not that the predicate will delete both input modules.
-    return TestFn(BD, New, ToNotOptimize, Error);
+    // Run the predicate,
+    // note that the predicate will delete both input modules.
+    bool Ret = TestFn(BD, New, ToNotOptimize, Error);
+    delete BD.swapProgramIn(Orig);
+    return Ret;
   }
+  delete BD.swapProgramIn(Orig);
   delete ToOptimize;
   delete ToNotOptimize;
   return false;




From daniel at zuster.org  Thu Jul 29 10:24:21 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 15:24:21 -0000
Subject: [llvm-commits] [llvm] r109752 -
	/llvm/trunk/lib/Support/CrashRecoveryContext.cpp
Message-ID: <20100729152422.081BB2A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 10:24:21 2010
New Revision: 109752

URL: http://llvm.org/viewvc/llvm-project?rev=109752&view=rev
Log:
CrashRecovery: Use ThreadLocal::erase() instead of set(0).

Modified:
    llvm/trunk/lib/Support/CrashRecoveryContext.cpp

Modified: llvm/trunk/lib/Support/CrashRecoveryContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CrashRecoveryContext.cpp?rev=109752&r1=109751&r2=109752&view=diff
==============================================================================
--- llvm/trunk/lib/Support/CrashRecoveryContext.cpp (original)
+++ llvm/trunk/lib/Support/CrashRecoveryContext.cpp Thu Jul 29 10:24:21 2010
@@ -31,7 +31,7 @@
     CurrentContext.set(this);
   }
   ~CrashRecoveryContextImpl() {
-    CurrentContext.set(0);
+    CurrentContext.erase();
   }
 
   void HandleCrash() {




From benny.kra at googlemail.com  Thu Jul 29 11:51:07 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 18:51:07 +0200
Subject: [llvm-commits] [llvm] r107027 - in /llvm/trunk:
	lib/CodeGen/AsmPrinter/DwarfDebug.cpp
	lib/CodeGen/AsmPrinter/DwarfDebug.h
	test/FrontendC/2010-06-28-DbgLocalVar.c
In-Reply-To: <20100628182503.718AC2A6C12C@llvm.org>
References: <20100628182503.718AC2A6C12C@llvm.org>
Message-ID: <6339CA4A-5120-4E56-91BD-D5ECBCBF0B28@gmail.com>


On 28.06.2010, at 20:25, Devang Patel wrote:

> Author: dpatel
> Date: Mon Jun 28 13:25:03 2010
> New Revision: 107027
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=107027&view=rev
> Log:
> Preserve deleted function's local variables' debug info.
> Radar 8122864.
> 
> Added:
>    llvm/trunk/test/FrontendC/2010-06-28-DbgLocalVar.c
> Modified:
>    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
>    llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
> 
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp?rev=107027&r1=107026&r2=107027&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon Jun 28 13:25:03 2010
> @@ -1386,6 +1386,7 @@
> /// If there are global variables in this scope then create and insert
> /// DIEs for these variables.
> DIE *DwarfDebug::updateSubprogramScopeDIE(const MDNode *SPNode) {
> +  ProcessedSPNodes.insert(SPNode);
>   CompileUnit *SPCU = getCompileUnit(SPNode);
>   DIE *SPDie = SPCU->getDIE(SPNode);
>   assert(SPDie && "Unable to find subprogram DIE!");
> @@ -2005,6 +2006,40 @@
> ///
> void DwarfDebug::endModule() {
>   if (!FirstCU) return;
> +  const Module *M = MMI->getModule();
> +  if (NamedMDNode *AllSPs = M->getNamedMetadata("llvm.dbg.sp")) {
> +    for (unsigned SI = 0, SE = AllSPs->getNumOperands(); SI != SE; ++SI) {
> +      if (ProcessedSPNodes.count(AllSPs->getOperand(SI)) != 0) continue;
> +      DISubprogram SP(AllSPs->getOperand(SI));
> +      if (!SP.Verify()) continue;
> +
> +      // Collect info for variables that were optimized out.
> +      StringRef FName = SP.getLinkageName();
> +      if (FName.empty())
> +        FName = SP.getName();
> +      NamedMDNode *NMD = 
> +        M->getNamedMetadata(Twine("llvm.dbg.lv.", getRealLinkageName(FName)));
> +      if (!NMD) continue;
> +      unsigned E = NMD->getNumOperands();
> +      if (!E) continue;
> +      DbgScope *Scope = new DbgScope(NULL, DIDescriptor(SP), NULL);

Hi Devang,

valgrind says those DbgScopes are leaked:


******************** TEST 'LLVM :: DebugInfo/2010-07-19-Crash.ll' FAILED ********************Script:
--
llc -o /dev/null < /build/buildbot-llvm/llvm-i686-linux-vg_leak/llvm/test/DebugInfo/2010-07-19-Crash.ll
--
Exit Code: 1
Command Output (stdout):
--
Command has output on stderr!

--
Command Output (stderr):
--
==22197== 196 (184 direct, 12 indirect) bytes in 1 blocks are definitely lost in loss record 138 of 139
==22197==    at 0x7F336CA: operator new(unsigned int) (vg_replace_malloc.c:255)
==22197==    by 0x8A8551D: llvm::DwarfDebug::endModule() (DwarfDebug.cpp:2051)
==22197==    by 0x8A73326: llvm::AsmPrinter::doFinalization(llvm::Module&) (AsmPrinter.cpp:738)
==22197==    by 0x8E1A150: llvm::FPPassManager::doFinalization(llvm::Module&) (PassManager.cpp:1471)
==22197==    by 0x8E1A065: llvm::FPPassManager::runOnModule(llvm::Module&) (PassManager.cpp:1455)
==22197==    by 0x8E1A31D: llvm::MPPassManager::runOnModule(llvm::Module&) (PassManager.cpp:1507)
==22197==    by 0x8E1A7D3: llvm::PassManagerImpl::run(llvm::Module&) (PassManager.cpp:1588)
==22197==    by 0x8E1ACB0: llvm::PassManager::run(llvm::Module&) (PassManager.cpp:1631)
==22197==    by 0x85B738F: main (llc.cpp:343)
==22197== 
--

********************

> +      for (unsigned I = 0; I != E; ++I) {
> +        DIVariable DV(NMD->getOperand(I));
> +        if (!DV.Verify()) continue;
> +        Scope->addVariable(new DbgVariable(DV));
> +      }
> +      
> +      // Construct subprogram DIE and add variables DIEs.
> +      constructSubprogramDIE(SP);
> +      DIE *ScopeDIE = getCompileUnit(SP)->getDIE(SP);
> +      const SmallVector<DbgVariable *, 8> &Variables = Scope->getVariables();
> +      for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
> +        DIE *VariableDIE = constructVariableDIE(Variables[i], Scope);
> +        if (VariableDIE)
> +          ScopeDIE->addChild(VariableDIE);
> +      }
> +    }
> +  }
> 
>   // Attach DW_AT_inline attribute with inlined subprogram DIEs.
>   for (SmallPtrSet<DIE *, 4>::iterator AI = InlinedSubprogramDIEs.begin(),



From dpatel at apple.com  Thu Jul 29 12:02:03 2010
From: dpatel at apple.com (Devang Patel)
Date: Thu, 29 Jul 2010 10:02:03 -0700
Subject: [llvm-commits] [llvm] r109739 - in /llvm/trunk/tools/llvm-diff:
	./ CMakeLists.txt DifferenceEngine.cpp DifferenceEngine.h
	Makefile llvm-diff.cpp
In-Reply-To: <20100729075327.507082A6C12D@llvm.org>
References: <20100729075327.507082A6C12D@llvm.org>
Message-ID: <E24A36D6-2C37-4C13-A942-C4ACC46AF674@apple.com>

John,
On Jul 29, 2010, at 12:53 AM, John McCall wrote:

> +  bool diff(Instruction *L, Instruction *R, bool Complain, bool TryUnify) {

Any reason to not directly use one of the following Instruction methods here ?

  /// isIdenticalTo - Return true if the specified instruction is exactly                                     
  /// identical to the current one.  This means that all operands match and any                               
  /// extra information (e.g. load is volatile) agree.                                                        
  bool isIdenticalTo(const Instruction *I) const;

  /// isIdenticalToWhenDefined - This is like isIdenticalTo, except that it                                   
  /// ignores the SubclassOptionalData flags, which specify conditions                                        
  /// under which the instruction's result is undefined.                                                      
  bool isIdenticalToWhenDefined(const Instruction *I) const;

  /// This function determines if the specified instruction executes the same                                 
  /// operation as the current one. This means that the opcodes, type, operand                                
  /// types and any other factors affecting the operation must be the same. This                              
  /// is similar to isIdenticalTo except the operands themselves don't have to                                
  /// be identical.                                                                                           
  /// @returns true if the specified instruction is the same operation as                                     
  /// the current one.                                                                                        
  /// @brief Determine if one instruction is the same operation as another.                                   
  bool isSameOperationAs(const Instruction *I) const;

-
Devang


From rjmccall at apple.com  Thu Jul 29 12:07:32 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 10:07:32 -0700
Subject: [llvm-commits] [llvm] r109739 - in /llvm/trunk/tools/llvm-diff:
	./ CMakeLists.txt DifferenceEngine.cpp DifferenceEngine.h
	Makefile llvm-diff.cpp
In-Reply-To: <E24A36D6-2C37-4C13-A942-C4ACC46AF674@apple.com>
References: <20100729075327.507082A6C12D@llvm.org>
	<E24A36D6-2C37-4C13-A942-C4ACC46AF674@apple.com>
Message-ID: <AD65CCAA-76E2-4DC6-AE36-12E720457402@apple.com>

On Jul 29, 2010, at 10:02 AM, Devang Patel wrote:
>> +  bool diff(Instruction *L, Instruction *R, bool Complain, bool TryUnify) {
> 
> Any reason to not directly use one of the following Instruction methods here ?

I can't use the isIdenticalTo methods because I have to use my own rules for
operand equality.  I didn't know about isSameOperationAs, but I suspect I still
don't want to use it because I don't want to consider (say) an inbounds GEP to
be structurally different from an undecorated one:  I'd rather match them and then
point out the difference.

John.


From clattner at apple.com  Thu Jul 29 12:09:07 2010
From: clattner at apple.com (Chris Lattner)
Date: Thu, 29 Jul 2010 10:09:07 -0700
Subject: [llvm-commits] [llvm] r109739 - in /llvm/trunk/tools/llvm-diff:
	./ CMakeLists.txt DifferenceEngine.cpp DifferenceEngine.h
	Makefile llvm-diff.cpp
In-Reply-To: <20100729075327.507082A6C12D@llvm.org>
References: <20100729075327.507082A6C12D@llvm.org>
Message-ID: <3784B293-99B0-4E93-9EF4-D579F1109A01@apple.com>


On Jul 29, 2010, at 12:53 AM, John McCall wrote:

> Author: rjmccall
> Date: Thu Jul 29 02:53:27 2010
> New Revision: 109739
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109739&view=rev
> Log:
> Add the llvm-diff tool, which performs a relatively naive structural
> diff of a function.  There's a lot of cruft in the current version, and
> it's pretty far from perfect, but it's usable.
> 
> Currently only capable of comparing functions.  Currently ignores metadata.
> Currently ignores most attributes of functions and instructions.

Nifty!  I hope to have a chance to look at this more in detail later, but two comments.

1. Please use the CommandLine library instead of rolling your own option processing:
http://llvm.org/docs/CommandLine.html

2. Please follow the LLVM #include conventions:
http://llvm.org/docs/CodingStandards.html#scf_includes

Apparently not mentioned there, but please use "llvm/x" and not <llvm/x> when inside the llvm code base.

Thanks John!

-Chris




From stoklund at 2pi.dk  Thu Jul 29 12:18:15 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 29 Jul 2010 17:18:15 -0000
Subject: [llvm-commits] [test-suite] r109761 -
	/test-suite/trunk/External/Makefile
Message-ID: <20100729171815.8B9802A6C12C@llvm.org>

Author: stoklund
Date: Thu Jul 29 12:18:15 2010
New Revision: 109761

URL: http://llvm.org/viewvc/llvm-project?rev=109761&view=rev
Log:
Don't attempt to run skidmarks10 if the sources were not found by configure.

Modified:
    test-suite/trunk/External/Makefile

Modified: test-suite/trunk/External/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/Makefile?rev=109761&r1=109760&r2=109761&view=diff
==============================================================================
--- test-suite/trunk/External/Makefile (original)
+++ test-suite/trunk/External/Makefile Thu Jul 29 12:18:15 2010
@@ -39,6 +39,10 @@
 PARALLEL_DIRS := $(filter-out HMMER, $(PARALLEL_DIRS))
 endif
 
+ifndef USE_SKIDMARKS10
+PARALLEL_DIRS := $(filter-out skidmarks10, $(PARALLEL_DIRS))
+endif
+
 # Sparc can't handle Namd: infinite loop, cause unknown
 ifeq ($(ARCH),Sparc)
 PARALLEL_DIRS := $(filter-out Namd, $(PARALLEL_DIRS))




From echristo at apple.com  Thu Jul 29 12:27:07 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 10:27:07 -0700
Subject: [llvm-commits] [test-suite] r109761 -
	/test-suite/trunk/External/Makefile
In-Reply-To: <20100729171815.8B9802A6C12C@llvm.org>
References: <20100729171815.8B9802A6C12C@llvm.org>
Message-ID: <69DBA094-86CC-450F-9956-93EC03BB5B87@apple.com>


On Jul 29, 2010, at 10:18 AM, Jakob Stoklund Olesen wrote:

> Author: stoklund
> Date: Thu Jul 29 12:18:15 2010
> New Revision: 109761
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109761&view=rev
> Log:
> Don't attempt to run skidmarks10 if the sources were not found by configure.
> 
> Modified:
>    test-suite/trunk/External/Makefile
> 
> Modified: test-suite/trunk/External/Makefile
> URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/Makefile?rev=109761&r1=109760&r2=109761&view=diff
> ==============================================================================
> --- test-suite/trunk/External/Makefile (original)
> +++ test-suite/trunk/External/Makefile Thu Jul 29 12:18:15 2010
> @@ -39,6 +39,10 @@
> PARALLEL_DIRS := $(filter-out HMMER, $(PARALLEL_DIRS))
> endif
> 
> +ifndef USE_SKIDMARKS10
> +PARALLEL_DIRS := $(filter-out skidmarks10, $(PARALLEL_DIRS))
> +endif
> +
> # Sparc can't handle Namd: infinite loop, cause unknown
> ifeq ($(ARCH),Sparc)
> PARALLEL_DIRS := $(filter-out Namd, $(PARALLEL_DIRS))
> 

Oops, forgot to commit that.

Thanks!

-eric


From stoklund at 2pi.dk  Thu Jul 29 12:42:27 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 29 Jul 2010 17:42:27 -0000
Subject: [llvm-commits] [llvm] r109764 - in /llvm/trunk/lib/Target/X86:
 X86ISelLowering.cpp X86InstrInfo.cpp
Message-ID: <20100729174227.DFA022A6C12C@llvm.org>

Author: stoklund
Date: Thu Jul 29 12:42:27 2010
New Revision: 109764

URL: http://llvm.org/viewvc/llvm-project?rev=109764&view=rev
Log:
Revert r109652, and remove the offending assert in loadRegFromStackSlot instead.

We do sometimes load from a too small stack slot when dealing with x86 arguments
(varargs and smaller-than-32-bit args). It looks like we know what we are doing
in those cases, so I am going to remove the assert instead of artifically
enlarging stack slot sizes.

The assert in storeRegToStackSlot stays in. We don't want to write beyond the
bounds of a stack slot.

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86InstrInfo.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109764&r1=109763&r2=109764&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Thu Jul 29 12:42:27 2010
@@ -1701,10 +1701,7 @@
   if (isVarArg) {
     if (Is64Bit || (CallConv != CallingConv::X86_FastCall &&
                     CallConv != CallingConv::X86_ThisCall)) {
-      // The fixed varargs object is created large enough that we can load and
-      // store any register in it.
-      FuncInfo->setVarArgsFrameIndex(MFI->CreateFixedObject(32, StackSize,
-                                                            true));
+      FuncInfo->setVarArgsFrameIndex(MFI->CreateFixedObject(1, StackSize,true));
     }
     if (Is64Bit) {
       unsigned TotalNumIntRegs = 0, TotalNumXMMRegs = 0;

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=109764&r1=109763&r2=109764&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Thu Jul 29 12:42:27 2010
@@ -2033,8 +2033,6 @@
                                         const TargetRegisterClass *RC,
                                         const TargetRegisterInfo *TRI) const {
   const MachineFunction &MF = *MBB.getParent();
-  assert(MF.getFrameInfo()->getObjectSize(FrameIdx) >= RC->getSize() &&
-         "Stack slot too small for load");
   bool isAligned = (RI.getStackAlignment() >= 16) || RI.canRealignStack(MF);
   unsigned Opc = getLoadRegOpcode(DestReg, RC, isAligned, TM);
   DebugLoc DL = MBB.findDebugLoc(MI);




From dalej at apple.com  Thu Jul 29 12:45:24 2010
From: dalej at apple.com (Dale Johannesen)
Date: Thu, 29 Jul 2010 17:45:24 -0000
Subject: [llvm-commits] [llvm] r109765 -
	/llvm/trunk/lib/CodeGen/MachineLICM.cpp
Message-ID: <20100729174524.A824B2A6C12C@llvm.org>

Author: johannes
Date: Thu Jul 29 12:45:24 2010
New Revision: 109765

URL: http://llvm.org/viewvc/llvm-project?rev=109765&view=rev
Log:
Comment typo.


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

Modified: llvm/trunk/lib/CodeGen/MachineLICM.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineLICM.cpp?rev=109765&r1=109764&r2=109765&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/MachineLICM.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineLICM.cpp Thu Jul 29 12:45:24 2010
@@ -68,7 +68,7 @@
 
     BitVector AllocatableSet;
 
-    // For each opcode, keep a list of potentail CSE instructions.
+    // For each opcode, keep a list of potential CSE instructions.
     DenseMap<unsigned, std::vector<const MachineInstr*> > CSEMap;
 
   public:




From benny.kra at googlemail.com  Thu Jul 29 12:48:07 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 17:48:07 -0000
Subject: [llvm-commits] [llvm] r109767 - in /llvm/trunk:
 include/llvm/MC/MCStreamer.h include/llvm/Target/TargetRegistry.h
 lib/CodeGen/LLVMTargetMachine.cpp lib/MC/MCAsmStreamer.cpp
 lib/MC/MCObjectStreamer.cpp tools/llvm-mc/llvm-mc.cpp
Message-ID: <20100729174807.22ADA2A6C12D@llvm.org>

Author: d0k
Date: Thu Jul 29 12:48:06 2010
New Revision: 109767

URL: http://llvm.org/viewvc/llvm-project?rev=109767&view=rev
Log:
Plug the remaining MC leaks by giving MCObjectStreamer/MCAsmStreamer ownership of the TargetAsmBackend and the MCCodeEmitter.

Modified:
    llvm/trunk/include/llvm/MC/MCStreamer.h
    llvm/trunk/include/llvm/Target/TargetRegistry.h
    llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
    llvm/trunk/lib/MC/MCAsmStreamer.cpp
    llvm/trunk/lib/MC/MCObjectStreamer.cpp
    llvm/trunk/tools/llvm-mc/llvm-mc.cpp

Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Thu Jul 29 12:48:06 2010
@@ -331,7 +331,7 @@
   /// InstPrint.
   ///
   /// \param CE - If given, a code emitter to use to show the instruction
-  /// encoding inline with the assembly.
+  /// encoding inline with the assembly. This method takes ownership of \arg CE.
   ///
   /// \param ShowInst - Whether to show the MCInst representation inline with
   /// the assembly.
@@ -343,12 +343,16 @@
 
   /// createMachOStreamer - Create a machine code streamer which will generate
   /// Mach-O format object files.
+  ///
+  /// Takes ownership of \arg TAB and \arg CE.
   MCStreamer *createMachOStreamer(MCContext &Ctx, TargetAsmBackend &TAB,
                                   raw_ostream &OS, MCCodeEmitter *CE,
                                   bool RelaxAll = false);
 
   /// createWinCOFFStreamer - Create a machine code streamer which will
   /// generate Microsoft COFF format object files.
+  ///
+  /// Takes ownership of \arg TAB and \arg CE.
   MCStreamer *createWinCOFFStreamer(MCContext &Ctx,
                                     TargetAsmBackend &TAB,
                                     MCCodeEmitter &CE, raw_ostream &OS);

Modified: llvm/trunk/include/llvm/Target/TargetRegistry.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetRegistry.h?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetRegistry.h (original)
+++ llvm/trunk/include/llvm/Target/TargetRegistry.h Thu Jul 29 12:48:06 2010
@@ -278,9 +278,9 @@
     ///
     /// \arg TT - The target triple.
     /// \arg Ctx - The target context.
-    /// \arg TAB - The target assembler backend object.
+    /// \arg TAB - The target assembler backend object. Takes ownership.
     /// \arg _OS - The stream object.
-    /// \arg _Emitter - The target independent assembler object.
+    /// \arg _Emitter - The target independent assembler object.Takes ownership.
     /// \arg RelaxAll - Relax all fixups?
     MCStreamer *createObjectStreamer(const std::string &TT, MCContext &Ctx,
                                      TargetAsmBackend &TAB,

Modified: llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp (original)
+++ llvm/trunk/lib/CodeGen/LLVMTargetMachine.cpp Thu Jul 29 12:48:06 2010
@@ -139,8 +139,6 @@
       getTarget().createMCInstPrinter(MAI.getAssemblerDialect(), MAI);
 
     // Create a code emitter if asked to show the encoding.
-    //
-    // FIXME: These are currently leaked.
     MCCodeEmitter *MCE = 0;
     if (ShowMCEncoding)
       MCE = getTarget().createCodeEmitter(*this, *Context);
@@ -154,8 +152,6 @@
   case CGFT_ObjectFile: {
     // Create the code emitter for the target if it exists.  If not, .o file
     // emission fails.
-    //
-    // FIXME: These are currently leaked.
     MCCodeEmitter *MCE = getTarget().createCodeEmitter(*this, *Context);
     TargetAsmBackend *TAB = getTarget().createAsmBackend(TargetTriple);
     if (MCE == 0 || TAB == 0)

Modified: llvm/trunk/lib/MC/MCAsmStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAsmStreamer.cpp?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAsmStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCAsmStreamer.cpp Thu Jul 29 12:48:06 2010
@@ -31,7 +31,7 @@
   formatted_raw_ostream &OS;
   const MCAsmInfo &MAI;
   OwningPtr<MCInstPrinter> InstPrinter;
-  MCCodeEmitter *Emitter;
+  OwningPtr<MCCodeEmitter> Emitter;
   
   SmallString<128> CommentToEmit;
   raw_svector_ostream CommentStream;

Modified: llvm/trunk/lib/MC/MCObjectStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCObjectStreamer.cpp?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCObjectStreamer.cpp (original)
+++ llvm/trunk/lib/MC/MCObjectStreamer.cpp Thu Jul 29 12:48:06 2010
@@ -11,7 +11,9 @@
 
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/MC/MCAssembler.h"
+#include "llvm/MC/MCCodeEmitter.h"
 #include "llvm/MC/MCExpr.h"
+#include "llvm/Target/TargetAsmBackend.h"
 using namespace llvm;
 
 MCObjectStreamer::MCObjectStreamer(MCContext &Context, TargetAsmBackend &TAB,
@@ -23,6 +25,8 @@
 }
 
 MCObjectStreamer::~MCObjectStreamer() {
+  delete &Assembler->getBackend();
+  delete &Assembler->getEmitter();
   delete Assembler;
 }
 

Modified: llvm/trunk/tools/llvm-mc/llvm-mc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-mc/llvm-mc.cpp?rev=109767&r1=109766&r2=109767&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-mc/llvm-mc.cpp (original)
+++ llvm/trunk/tools/llvm-mc/llvm-mc.cpp Thu Jul 29 12:48:06 2010
@@ -287,25 +287,24 @@
     return 1;
   }
 
-  OwningPtr<MCCodeEmitter> CE;
   OwningPtr<MCStreamer> Str;
-  OwningPtr<TargetAsmBackend> TAB;
 
   if (FileType == OFT_AssemblyFile) {
     MCInstPrinter *IP =
       TheTarget->createMCInstPrinter(OutputAsmVariant, *MAI);
+    MCCodeEmitter *CE = 0;
     if (ShowEncoding)
-      CE.reset(TheTarget->createCodeEmitter(*TM, Ctx));
+      CE = TheTarget->createCodeEmitter(*TM, Ctx);
     Str.reset(createAsmStreamer(Ctx, *Out,TM->getTargetData()->isLittleEndian(),
-                                /*asmverbose*/true, IP, CE.get(), ShowInst));
+                                /*asmverbose*/true, IP, CE, ShowInst));
   } else if (FileType == OFT_Null) {
     Str.reset(createNullStreamer(Ctx));
   } else {
     assert(FileType == OFT_ObjectFile && "Invalid file type!");
-    CE.reset(TheTarget->createCodeEmitter(*TM, Ctx));
-    TAB.reset(TheTarget->createAsmBackend(TripleName));
+    MCCodeEmitter *CE = TheTarget->createCodeEmitter(*TM, Ctx);
+    TargetAsmBackend *TAB = TheTarget->createAsmBackend(TripleName);
     Str.reset(TheTarget->createObjectStreamer(TripleName, Ctx, *TAB,
-                                              *Out, CE.get(), RelaxAll));
+                                              *Out, CE, RelaxAll));
   }
 
   if (EnableLogging) {




From rjmccall at apple.com  Thu Jul 29 12:55:00 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 17:55:00 -0000
Subject: [llvm-commits] [llvm] r109769 -
	/llvm/trunk/tools/llvm-diff/llvm-diff.cpp
Message-ID: <20100729175500.712802A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 12:55:00 2010
New Revision: 109769

URL: http://llvm.org/viewvc/llvm-project?rev=109769&view=rev
Log:
Switch to using the LLVM CommandLine library so that our help
message is properly contaminated with nonsense about timing passes
that doesn't apply at all to this utility. :)


Modified:
    llvm/trunk/tools/llvm-diff/llvm-diff.cpp

Modified: llvm/trunk/tools/llvm-diff/llvm-diff.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/llvm-diff.cpp?rev=109769&r1=109768&r2=109769&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/llvm-diff.cpp (original)
+++ llvm/trunk/tools/llvm-diff/llvm-diff.cpp Thu Jul 29 12:55:00 2010
@@ -13,6 +13,7 @@
 #include <llvm/Support/MemoryBuffer.h>
 #include <llvm/Bitcode/ReaderWriter.h>
 
+#include <llvm/Support/CommandLine.h>
 #include <llvm/Support/raw_ostream.h>
 #include <llvm/Support/ErrorHandling.h>
 #include <llvm/LLVMContext.h>
@@ -52,14 +53,6 @@
   return 0;
 }
 
-static int usage() {
-  errs() << "expected usage:\n";
-  errs() << "  llvm-diff oldmodule.ll newmodule.ll [function list]\n";
-  errs() << "Assembly or bitcode modules may be used interchangeably.\n";
-  errs() << "If no functions are provided, all functions will be compared.\n";
-  return 1;
-}
-
 namespace {
 struct DiffContext {
   DiffContext(Value *L, Value *R)
@@ -270,48 +263,47 @@
 };
 }
 
-int main(int argc, const char **argv) {
-  if (argc < 3) return usage();
+static void diffGlobal(DifferenceEngine &Engine, Module *L, Module *R,
+                       StringRef Name) {
+  // Drop leading sigils from the global name.
+  if (Name.startswith("@")) Name = Name.substr(1);
+
+  Function *LFn = L->getFunction(Name);
+  Function *RFn = R->getFunction(Name);
+  if (LFn && RFn)
+    Engine.diff(LFn, RFn);
+  else if (!LFn && !RFn)
+    errs() << "No function named @" << Name << " in either module\n";
+  else if (!LFn)
+    errs() << "No function named @" << Name << " in left module\n";
+  else
+    errs() << "No function named @" << Name << " in right module\n";
+}
+
+cl::opt<std::string> LeftFilename(cl::Positional, cl::desc("<first file>"), cl::Required);
+cl::opt<std::string> RightFilename(cl::Positional, cl::desc("<second file>"), cl::Required);
+cl::list<std::string> GlobalsToCompare(cl::Positional, cl::desc("<globals to compare>"));
 
-  // Don't make StringRef locals like this at home.
-  StringRef LModuleFile = argv[1];
-  StringRef RModuleFile = argv[2];
+int main(int argc, char **argv) {
+  cl::ParseCommandLineOptions(argc, argv);
 
   LLVMContext Context;
   
   // Load both modules.  Die if that fails.
-  Module *LModule = ReadModule(Context, LModuleFile);
-  Module *RModule = ReadModule(Context, RModuleFile);
+  Module *LModule = ReadModule(Context, LeftFilename);
+  Module *RModule = ReadModule(Context, RightFilename);
   if (!LModule || !RModule) return 1;
 
   DiffConsumer Consumer(LModule, RModule);
   DifferenceEngine Engine(Context, Consumer);
 
-  // If any function names were given, just diff those.
-  const char **FnNames = argv + 3;
-  unsigned NumFnNames = argc - 3;
-  if (NumFnNames) {
-    for (unsigned I = 0; I != NumFnNames; ++I) {
-      StringRef FnName = FnNames[I];
-
-      // Drop leading sigils from the function name.
-      if (FnName.startswith("@")) FnName = FnName.substr(1);
-
-      Function *LFn = LModule->getFunction(FnName);
-      Function *RFn = RModule->getFunction(FnName);
-      if (LFn && RFn)
-        Engine.diff(LFn, RFn);
-      else {
-        if (!LFn && !RFn)
-          errs() << "No function named @" << FnName << " in either module\n";
-        else if (!LFn)
-          errs() << "No function named @" << FnName << " in left module\n";
-        else
-          errs() << "No function named @" << FnName << " in right module\n";
-      }
-    }
+  // If any global names were given, just diff those.
+  if (!GlobalsToCompare.empty()) {
+    for (unsigned I = 0, E = GlobalsToCompare.size(); I != E; ++I)
+      diffGlobal(Engine, LModule, RModule, GlobalsToCompare[I]);
+
+  // Otherwise, diff everything in the module.
   } else {
-    // Otherwise, diff all functions in the modules.
     Engine.diff(LModule, RModule);
   }
 




From natebegeman at mac.com  Thu Jul 29 12:56:55 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Thu, 29 Jul 2010 17:56:55 -0000
Subject: [llvm-commits] [llvm] r109770 - in /llvm/trunk:
 include/llvm/IntrinsicsARM.td lib/Target/ARM/ARMInstrInfo.td
 lib/Target/ARM/ARMInstrThumb2.td
Message-ID: <20100729175655.611902A6C12C@llvm.org>

Author: sampo
Date: Thu Jul 29 12:56:55 2010
New Revision: 109770

URL: http://llvm.org/viewvc/llvm-project?rev=109770&view=rev
Log:
Add intrinsics __builtin_arm_qadd & __builtin_arm_qsub to allow access to the QADD & QSUB instructions.
Behave identically to __qadd & __qsub RealView instruction intrinsics.

Modified:
    llvm/trunk/include/llvm/IntrinsicsARM.td
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/include/llvm/IntrinsicsARM.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/IntrinsicsARM.td?rev=109770&r1=109769&r2=109770&view=diff
==============================================================================
--- llvm/trunk/include/llvm/IntrinsicsARM.td (original)
+++ llvm/trunk/include/llvm/IntrinsicsARM.td Thu Jul 29 12:56:55 2010
@@ -21,6 +21,17 @@
 }
 
 //===----------------------------------------------------------------------===//
+// Saturating Arithmentic
+
+let TargetPrefix = "arm" in {  // All intrinsics start with "llvm.arm.".
+  def int_arm_qadd : GCCBuiltin<"__builtin_arm_qadd">,
+              Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty],
+                        [IntrNoMem, Commutative]>;
+  def int_arm_qsub : GCCBuiltin<"__builtin_arm_qsub">,
+              Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+}
+
+//===----------------------------------------------------------------------===//
 // Advanced SIMD (NEON)
 
 let TargetPrefix = "arm" in {  // All intrinsics start with "llvm.arm.".

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=109770&r1=109769&r2=109770&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Thu Jul 29 12:56:55 2010
@@ -1714,24 +1714,26 @@
 
 // ARM Arithmetic Instruction -- for disassembly only
 // GPR:$dst = GPR:$a op GPR:$b
-class AAI<bits<8> op27_20, bits<4> op7_4, string opc>
+class AAI<bits<8> op27_20, bits<4> op7_4, string opc,
+          list<dag> pattern = [/* For disassembly only; pattern left blank */]>
   : AI<(outs GPR:$dst), (ins GPR:$a, GPR:$b), DPFrm, IIC_iALUr,
-       opc, "\t$dst, $a, $b",
-       [/* For disassembly only; pattern left blank */]> {
+       opc, "\t$dst, $a, $b", pattern> {
   let Inst{27-20} = op27_20;
   let Inst{7-4} = op7_4;
 }
 
 // Saturating add/subtract -- for disassembly only
 
-def QADD    : AAI<0b00010000, 0b0101, "qadd">;
+def QADD    : AAI<0b00010000, 0b0101, "qadd",
+                  [(set GPR:$dst, (int_arm_qadd GPR:$a, GPR:$b))]>;
 def QADD16  : AAI<0b01100010, 0b0001, "qadd16">;
 def QADD8   : AAI<0b01100010, 0b1001, "qadd8">;
 def QASX    : AAI<0b01100010, 0b0011, "qasx">;
 def QDADD   : AAI<0b00010100, 0b0101, "qdadd">;
 def QDSUB   : AAI<0b00010110, 0b0101, "qdsub">;
 def QSAX    : AAI<0b01100010, 0b0101, "qsax">;
-def QSUB    : AAI<0b00010010, 0b0101, "qsub">;
+def QSUB    : AAI<0b00010010, 0b0101, "qsub",
+                  [(set GPR:$dst, (int_arm_qsub GPR:$a, GPR:$b))]>;
 def QSUB16  : AAI<0b01100010, 0b0111, "qsub16">;
 def QSUB8   : AAI<0b01100010, 0b1111, "qsub8">;
 def UQADD16 : AAI<0b01100110, 0b0001, "uqadd16">;

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=109770&r1=109769&r2=109770&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Thu Jul 29 12:56:55 2010
@@ -1441,9 +1441,10 @@
 
 // A6.3.13, A6.3.14, A6.3.15 Parallel addition and subtraction (signed/unsigned)
 // And Miscellaneous operations -- for disassembly only
-class T2I_pam<bits<3> op22_20, bits<4> op7_4, string opc>
+class T2I_pam<bits<3> op22_20, bits<4> op7_4, string opc,
+              list<dag> pat = [/* For disassembly only; pattern left blank */]>
   : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), NoItinerary, opc,
-        "\t$dst, $a, $b", [/* For disassembly only; pattern left blank */]> {
+        "\t$dst, $a, $b", pat> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0101;
   let Inst{22-20} = op22_20;
@@ -1453,14 +1454,16 @@
 
 // Saturating add/subtract -- for disassembly only
 
-def t2QADD    : T2I_pam<0b000, 0b1000, "qadd">;
+def t2QADD    : T2I_pam<0b000, 0b1000, "qadd",
+                        [(set GPR:$dst, (int_arm_qadd GPR:$a, GPR:$b))]>;
 def t2QADD16  : T2I_pam<0b001, 0b0001, "qadd16">;
 def t2QADD8   : T2I_pam<0b000, 0b0001, "qadd8">;
 def t2QASX    : T2I_pam<0b010, 0b0001, "qasx">;
 def t2QDADD   : T2I_pam<0b000, 0b1001, "qdadd">;
 def t2QDSUB   : T2I_pam<0b000, 0b1011, "qdsub">;
 def t2QSAX    : T2I_pam<0b110, 0b0001, "qsax">;
-def t2QSUB    : T2I_pam<0b000, 0b1010, "qsub">;
+def t2QSUB    : T2I_pam<0b000, 0b1010, "qsub",
+                        [(set GPR:$dst, (int_arm_qsub GPR:$a, GPR:$b))]>;
 def t2QSUB16  : T2I_pam<0b101, 0b0001, "qsub16">;
 def t2QSUB8   : T2I_pam<0b100, 0b0001, "qsub8">;
 def t2UQADD16 : T2I_pam<0b001, 0b0101, "uqadd16">;




From resistor at mac.com  Thu Jul 29 12:57:58 2010
From: resistor at mac.com (Owen Anderson)
Date: Thu, 29 Jul 2010 17:57:58 -0000
Subject: [llvm-commits] [llvm] r109772 - /llvm/trunk/utils/vim/llvm.vim
Message-ID: <20100729175758.218202A6C12C@llvm.org>

Author: resistor
Date: Thu Jul 29 12:57:57 2010
New Revision: 109772

URL: http://llvm.org/viewvc/llvm-project?rev=109772&view=rev
Log:
Make un-named values legible in certain vim configurations.

Modified:
    llvm/trunk/utils/vim/llvm.vim

Modified: llvm/trunk/utils/vim/llvm.vim
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/vim/llvm.vim?rev=109772&r1=109771&r2=109772&view=diff
==============================================================================
--- llvm/trunk/utils/vim/llvm.vim (original)
+++ llvm/trunk/utils/vim/llvm.vim Thu Jul 29 12:57:57 2010
@@ -64,7 +64,7 @@
 syn keyword llvmError  getresult big little endian begin end
 
 " Misc syntax.
-syn match   llvmIgnore /[%@]\d\+\>/
+syn match   llvmNoName /[%@]\d\+\>/
 syn match   llvmNumber /-\?\<\d\+\>/
 syn match   llvmFloat  /-\?\<\d\+\.\d*\(e[+-]\d\+\)\?\>/
 syn match   llvmFloat  /\<0x\x\+\>/
@@ -99,7 +99,7 @@
   HiLink llvmKeyword Keyword
   HiLink llvmBoolean Boolean
   HiLink llvmFloat Float
-  HiLink llvmIgnore Ignore
+  HiLink llvmNoName Identifier
   HiLink llvmConstant Constant
   HiLink llvmSpecialComment SpecialComment
   HiLink llvmError Error




From echristo at apple.com  Thu Jul 29 12:59:11 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 10:59:11 -0700
Subject: [llvm-commits] [llvm] r109769 -
	/llvm/trunk/tools/llvm-diff/llvm-diff.cpp
In-Reply-To: <20100729175500.712802A6C12C@llvm.org>
References: <20100729175500.712802A6C12C@llvm.org>
Message-ID: <312B2C86-7E5B-4123-876F-888A997CDDFA@apple.com>


On Jul 29, 2010, at 10:55 AM, John McCall wrote:

> Switch to using the LLVM CommandLine library so that our help
> message is properly contaminated with nonsense about timing passes
> that doesn't apply at all to this utility. :)

Totally need to remove the documentation for -print-before and -print-after.

-eric


From eli.friedman at gmail.com  Thu Jul 29 13:03:33 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Thu, 29 Jul 2010 18:03:33 -0000
Subject: [llvm-commits] [llvm] r109773 -
 /llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp
Message-ID: <20100729180333.2D17A2A6C12C@llvm.org>

Author: efriedma
Date: Thu Jul 29 13:03:33 2010
New Revision: 109773

URL: http://llvm.org/viewvc/llvm-project?rev=109773&view=rev
Log:
PR7750: !CExpr->isNullValue() only properly computes whether CExpr is nonnull
if CExpr is a ConstantInt.


Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp?rev=109773&r1=109772&r2=109773&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCompares.cpp Thu Jul 29 13:03:33 2010
@@ -1374,7 +1374,7 @@
       case Instruction::Or:
         // If bits are being or'd in that are not present in the constant we
         // are comparing against, then the comparison could never succeed!
-        if (Constant *BOC = dyn_cast<Constant>(BO->getOperand(1))) {
+        if (ConstantInt *BOC = dyn_cast<ConstantInt>(BO->getOperand(1))) {
           Constant *NotCI = ConstantExpr::getNot(RHS);
           if (!ConstantExpr::getAnd(BOC, NotCI)->isNullValue())
             return ReplaceInstUsesWith(ICI,




From rjmccall at apple.com  Thu Jul 29 13:08:23 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 18:08:23 -0000
Subject: [llvm-commits] [llvm] r109774 - in /llvm/trunk/tools/llvm-diff:
 DifferenceEngine.cpp DifferenceEngine.h llvm-diff.cpp
Message-ID: <20100729180823.5D7E92A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 13:08:23 2010
New Revision: 109774

URL: http://llvm.org/viewvc/llvm-project?rev=109774&view=rev
Log:
Make the header self-contained and follow #include guidelines.


Modified:
    llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
    llvm/trunk/tools/llvm-diff/DifferenceEngine.h
    llvm/trunk/tools/llvm-diff/llvm-diff.cpp

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp?rev=109774&r1=109773&r2=109774&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.cpp Thu Jul 29 13:08:23 2010
@@ -7,30 +7,28 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This header defines the interface to the LLVM difference engine,
-// which structurally compares functions within a module.
+// This header defines the implementation of the LLVM difference
+// engine, which structurally compares global values within a module.
 //
 //===----------------------------------------------------------------------===//
 
-#include <utility>
+#include "DifferenceEngine.h"
 
-#include <llvm/ADT/DenseMap.h>
-#include <llvm/ADT/DenseSet.h>
-#include <llvm/ADT/SmallVector.h>
-#include <llvm/ADT/StringRef.h>
-#include <llvm/ADT/StringSet.h>
-
-#include <llvm/Module.h>
-#include <llvm/Function.h>
-#include <llvm/Instructions.h>
-#include <llvm/Support/CFG.h>
-
-#include <llvm/Support/raw_ostream.h>
-#include <llvm/Support/type_traits.h>
-#include <llvm/Support/ErrorHandling.h>
-#include <llvm/Support/CallSite.h>
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/StringSet.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/type_traits.h"
 
-#include "DifferenceEngine.h"
+#include <utility>
 
 using namespace llvm;
 

Modified: llvm/trunk/tools/llvm-diff/DifferenceEngine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/DifferenceEngine.h?rev=109774&r1=109773&r2=109774&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/DifferenceEngine.h (original)
+++ llvm/trunk/tools/llvm-diff/DifferenceEngine.h Thu Jul 29 13:08:23 2010
@@ -15,17 +15,19 @@
 #ifndef _LLVM_DIFFERENCE_ENGINE_H_
 #define _LLVM_DIFFERENCE_ENGINE_H_
 
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+
 #include <utility>
-#include <llvm/ADT/SmallVector.h>
-#include <llvm/ADT/StringRef.h>
 
 namespace llvm {
+  class Function;
+  class GlobalValue;
+  class Instruction;
   class LLVMContext;
   class Module;
-  class Function;
   class Twine;
   class Value;
-  class GlobalValue;
 
   /// A class for performing structural comparisons of LLVM assembly.
   class DifferenceEngine {

Modified: llvm/trunk/tools/llvm-diff/llvm-diff.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/llvm-diff.cpp?rev=109774&r1=109773&r2=109774&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/llvm-diff.cpp (original)
+++ llvm/trunk/tools/llvm-diff/llvm-diff.cpp Thu Jul 29 13:08:23 2010
@@ -1,27 +1,36 @@
+//===-- llvm-diff.cpp - Module comparator command-line driver ---*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the command-line driver for the difference engine.
+//
+//===----------------------------------------------------------------------===//
+
+#include "DifferenceEngine.h"
+
+#include "llvm/Instructions.h"
+#include "llvm/LLVMContext.h"
+#include "llvm/Module.h"
+#include "llvm/Type.h"
+#include "llvm/Assembly/Parser.h"
+#include "llvm/Bitcode/ReaderWriter.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/StringRef.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/SourceMgr.h"
+
 #include <string>
 #include <utility>
 
-#include <llvm/ADT/StringRef.h>
-#include <llvm/ADT/SmallVector.h>
-#include <llvm/ADT/DenseMap.h>
-
-// Required to parse .ll files.
-#include <llvm/Support/SourceMgr.h>
-#include <llvm/Assembly/Parser.h>
-
-// Required to parse .bc files.
-#include <llvm/Support/MemoryBuffer.h>
-#include <llvm/Bitcode/ReaderWriter.h>
-
-#include <llvm/Support/CommandLine.h>
-#include <llvm/Support/raw_ostream.h>
-#include <llvm/Support/ErrorHandling.h>
-#include <llvm/LLVMContext.h>
-#include <llvm/Module.h>
-#include <llvm/Type.h>
-#include <llvm/Instructions.h>
-
-#include "DifferenceEngine.h"
 
 using namespace llvm;
 




From echristo at apple.com  Thu Jul 29 13:11:28 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 18:11:28 -0000
Subject: [llvm-commits] [llvm] r109775 -
	/llvm/trunk/include/llvm/Target/TargetOptions.h
Message-ID: <20100729181128.7D48D2A6C12C@llvm.org>

Author: echristo
Date: Thu Jul 29 13:11:28 2010
New Revision: 109775

URL: http://llvm.org/viewvc/llvm-project?rev=109775&view=rev
Log:
Grammar.

Modified:
    llvm/trunk/include/llvm/Target/TargetOptions.h

Modified: llvm/trunk/include/llvm/Target/TargetOptions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetOptions.h?rev=109775&r1=109774&r2=109775&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetOptions.h (original)
+++ llvm/trunk/include/llvm/Target/TargetOptions.h Thu Jul 29 13:11:28 2010
@@ -140,8 +140,8 @@
   /// StackAlignment - Override default stack alignment for target.
   extern unsigned StackAlignment;
 
-  /// RealignStack - This flag indicates, whether stack should be automatically
-  /// realigned, if needed.
+  /// RealignStack - This flag indicates whether the stack should be
+  /// automatically realigned, if needed.
   extern bool RealignStack;
 
   /// DisableJumpTables - This flag indicates jump tables should not be 




From daniel at zuster.org  Thu Jul 29 13:15:00 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:00 -0000
Subject: [llvm-commits] [zorg] r109777 -
	/zorg/trunk/lnt/lnt/util/NTEmailReport.py
Message-ID: <20100729181500.28BE22A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:14:59 2010
New Revision: 109777

URL: http://llvm.org/viewvc/llvm-project?rev=109777&view=rev
Log:
LNT/simple: Don't list unchanged tests in reports.

Modified:
    zorg/trunk/lnt/lnt/util/NTEmailReport.py

Modified: zorg/trunk/lnt/lnt/util/NTEmailReport.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/util/NTEmailReport.py?rev=109777&r1=109776&r2=109777&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/util/NTEmailReport.py (original)
+++ zorg/trunk/lnt/lnt/util/NTEmailReport.py Thu Jul 29 13:14:59 2010
@@ -271,7 +271,7 @@
 <p>
 <h3>Changes Detail</h3>"""
         for name,items,show_perf in items_info:
-            if not items:
+            if not items or name == 'Unchanged Tests':
                 continue
 
             show_pset = items.items()[0][0] or len(items) > 1




From daniel at zuster.org  Thu Jul 29 13:15:02 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:02 -0000
Subject: [llvm-commits] [zorg] r109778 -
	/zorg/trunk/lnt/lnt/db/perfdbsummary.py
Message-ID: <20100729181502.A8D9B2A6C12D@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:02 2010
New Revision: 109778

URL: http://llvm.org/viewvc/llvm-project?rev=109778&view=rev
Log:
LNT/perfdbsummary: Add a function for getting the status kind for a particular run.

Modified:
    zorg/trunk/lnt/lnt/db/perfdbsummary.py

Modified: zorg/trunk/lnt/lnt/db/perfdbsummary.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/db/perfdbsummary.py?rev=109778&r1=109777&r2=109778&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/db/perfdbsummary.py (original)
+++ zorg/trunk/lnt/lnt/db/perfdbsummary.py Thu Jul 29 13:15:02 2010
@@ -200,6 +200,7 @@
         self.runs_in_order = runs_in_order
         self.order_by_run = order_by_run
         self.machine_id_by_run = machine_id_by_run
+        self.run_status_kinds = {}
 
     def is_up_to_date(self, db):
         return (not db.modified_run and
@@ -231,3 +232,16 @@
             id = self.runs_in_order[i]
             if machine_id == self.machine_id_by_run[id]:
                 return id
+
+    def get_run_status_kind(self, db, run_id):
+        kind = self.run_status_kinds.get(run_id)
+        if kind is None:
+            # Compute the status kind by for .success tests in this run.
+            if db.session.query(Test.name).join(Sample)\
+                    .filter(Sample.run_id == run_id)\
+                    .filter(Test.name.endswith(".success")).first() is not None:
+                kind = False
+            else:
+                kind = True
+        self.run_status_kinds[run_id] = kind
+        return kind




From daniel at zuster.org  Thu Jul 29 13:15:05 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:05 -0000
Subject: [llvm-commits] [zorg] r109779 -
	/zorg/trunk/lnt/lnt/util/NTEmailReport.py
Message-ID: <20100729181505.790A82A6C12E@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:05 2010
New Revision: 109779

URL: http://llvm.org/viewvc/llvm-project?rev=109779&view=rev
Log:
LNT/simple: Stop listing existing failures twice.

Modified:
    zorg/trunk/lnt/lnt/util/NTEmailReport.py

Modified: zorg/trunk/lnt/lnt/util/NTEmailReport.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/util/NTEmailReport.py?rev=109779&r1=109778&r2=109779&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/util/NTEmailReport.py (original)
+++ zorg/trunk/lnt/lnt/util/NTEmailReport.py Thu Jul 29 13:15:05 2010
@@ -325,35 +325,6 @@
                 print >>html_report, """
 </table>"""
 
-    # Generate a list of the existing failures.
-    if existing_failures:
-        show_pset = existing_failures.items()[0][0] or len(items) > 1
-
-        print >>report
-        print >>report, """================="""
-        print >>report, """Existing Failures"""
-        print >>report, """================="""
-        print >>html_report, """
-<hr>
-<h3>Existing Failures</h3>"""
-        for pset,tests in existing_failures.items():
-            if show_pset:
-                print >>report
-                print >>report, "Parameter Set:", pset
-                table_name = "Existing Failures - %s" % (name, pset)
-            else:
-                table_name = "Existing Failures"
-
-            print >>html_report, """
-<p>
-<table class="sortable">
-<tr><th>%s</th></tr>""" % table_name
-            for name,cr in tests:
-                print >>report, '  %s' % (name,)
-                print >>html_report, """<tr><td>%s</td></tr>""" % (name,)
-            print >>html_report, """
-</table>"""
-
     # Finish up the HTML report.
     if graphs:
         # Get the test ids we want data for.




From daniel at zuster.org  Thu Jul 29 13:15:08 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:08 -0000
Subject: [llvm-commits] [zorg] r109780 - in /zorg/trunk/lnt/lnt:
 db/runinfo.py util/NTEmailReport.py viewer/simple.ptl
Message-ID: <20100729181508.CACE12A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:08 2010
New Revision: 109780

URL: http://llvm.org/viewvc/llvm-project?rev=109780&view=rev
Log:
LNT/simple: Compute status kinds on a per run basis instead of assuming the kind is consistent across the entire database, which was totally bogus.

Modified:
    zorg/trunk/lnt/lnt/db/runinfo.py
    zorg/trunk/lnt/lnt/util/NTEmailReport.py
    zorg/trunk/lnt/lnt/viewer/simple.ptl

Modified: zorg/trunk/lnt/lnt/db/runinfo.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/db/runinfo.py?rev=109780&r1=109779&r2=109780&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/db/runinfo.py (original)
+++ zorg/trunk/lnt/lnt/db/runinfo.py Thu Jul 29 13:15:08 2010
@@ -92,8 +92,31 @@
         self.sample_map = Util.multidict()
         self.loaded_samples = set()
 
-    def get_run_comparison_result(self, run, compare_to, test_name, pset,
-                                  comparison_window=[]):
+    def get_test_status_in_run(self, run, status_kind, test_name, pset):
+        if status_kind == False: # .success
+            status_name = test_name + '.success'
+            status_test_id = self.test_suite_summary.test_id_map.get(
+                (status_name, pset))
+            run_status = self.sample_map.get((run.id, status_test_id))
+            if run_status and int(run_status[0]) == 1:
+                return PASS
+            else:
+                return FAIL
+        else:
+            status_name = test_name + '.status'
+            status_test_id = self.test_suite_summary.test_id_map.get(
+                (status_name, pset))
+            run_status = self.sample_map.get((run.id, status_test_id))
+            if not run_status:
+                return PASS
+            else:
+                # FIXME: What to do about the multiple entries here. We could
+                # start by just treating non-matching samples as errors.
+                return int(run_status[0])
+
+    def get_run_comparison_result(self, run, run_status_kind,
+                                  compare_to, compare_to_status_kind,
+                                  test_name, pset, comparison_window=[]):
         # Get the test.
         test_id = self.test_suite_summary.test_id_map.get((test_name, pset))
         if test_id is None:
@@ -101,15 +124,6 @@
                                     pct_delta=None, stddev=None, MAD=None,
                                     cur_failed=None, prev_failed=None)
 
-        # Get the test status info.
-        status_info = self.test_suite_summary.test_status_map.get(test_name)
-        if status_info is not None:
-            status_name,status_kind = status_info
-            status_test_id = self.test_suite_summary.test_id_map.get(
-                (status_name, pset))
-        else:
-            status_test_id = status_kind = None
-
         # Load the sample data for the current and previous runs and the
         # comparison window.
         if compare_to is None:
@@ -129,25 +143,18 @@
         # Determine whether this (test,pset) passed or failed in the current and
         # previous runs.
         run_failed = prev_failed = False
-        if not status_test_id:
-            # Assume status kind is '.status' if missing, in which case no
-            # values indicates success.
-            run_failed = bool(run_values)
-            prev_failed = bool(prev_values)
-        else:
-            run_status = self.sample_map.get((run.id,status_test_id))
-            prev_status = self.sample_map.get((compare_id,status_test_id))
-
-            # FIXME: Support XFAILs better.
-            #
-            # FIXME: What to do about the multiple entries here. We could start
-            # by just treating non-matching samples as errors.
-            if status_kind == False: # .success style
-                run_failed = not run_status or not run_status[0]
-                prev_failed = not prev_status or not prev_status[0]
-            else:
-                run_failed = run_status and run_status[0] == FAIL
-                prev_failed = prev_status and prev_status[0] == FAIL
+        run_status = prev_status = None
+        run_status = self.get_test_status_in_run(
+            run, run_status_kind, test_name, pset)
+        if compare_to:
+            prev_status = self.get_test_status_in_run(
+                compare_to, compare_to_status_kind, test_name, pset)
+        else:
+            prev_status = None
+
+        # FIXME: Support XFAILs better.
+        run_failed = run_status == FAIL
+        prev_failed = prev_status == FAIL
 
         # Get the current and previous values.
         if run_values:

Modified: zorg/trunk/lnt/lnt/util/NTEmailReport.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/util/NTEmailReport.py?rev=109780&r1=109779&r2=109780&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/util/NTEmailReport.py (original)
+++ zorg/trunk/lnt/lnt/util/NTEmailReport.py Thu Jul 29 13:15:08 2010
@@ -114,6 +114,14 @@
         # FIXME: Look for run across machine.
         compare_to = None
 
+    # Get the test status style used in each run.
+    run_status_kind = run_summary.get_run_status_kind(db, run.id)
+    if compare_to:
+        compare_to_status_kind = run_summary.get_run_status_kind(
+            db, compare_to.id)
+    else:
+        compare_to_status_kind = None
+
     # Get the list of tests we are interested in.
     interesting_runs = [run.id]
     if compare_to:
@@ -132,8 +140,9 @@
     num_total_tests = len(test_names) * len(ts_summary.parameter_sets)
     for name in test_names:
         for pset in ts_summary.parameter_sets:
-            cr = sri.get_run_comparison_result(run, compare_to, name, pset,
-                                               comparison_window)
+            cr = sri.get_run_comparison_result(
+                run, run_status_kind, compare_to, compare_to_status_kind,
+                name, pset, comparison_window)
             test_status = cr.get_test_status()
             perf_status = cr.get_value_status()
             if test_status == runinfo.REGRESSED:

Modified: zorg/trunk/lnt/lnt/viewer/simple.ptl
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/viewer/simple.ptl?rev=109780&r1=109779&r2=109780&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/viewer/simple.ptl (original)
+++ zorg/trunk/lnt/lnt/viewer/simple.ptl Thu Jul 29 13:15:08 2010
@@ -435,6 +435,14 @@
         return text_report
 
     def _q_index_body [html] (self, db, run, run_summary, compare_to):
+        # Get the test status style used in each run.
+        run_status_kind = run_summary.get_run_status_kind(db, run.id)
+        if compare_to:
+            compare_to_status_kind = run_summary.get_run_status_kind(
+                db, compare_to.id)
+        else:
+            compare_to_status_kind = None
+
         # Load the test suite summary.
         ts_summary = perfdbsummary.get_simple_suite_summary(db, self.tag)
         sri = runinfo.SimpleRunInfo(db, ts_summary)
@@ -611,8 +619,9 @@
             <td><input type="checkbox" name="test.%s"></td>
             <td>%s</td>""" % (name, name)
             for pset in ts_summary.parameter_sets:
-                cr = sri.get_run_comparison_result(run, compare_to, name, pset,
-                                                   comparison_window)
+                cr = sri.get_run_comparison_result(
+                    run, run_status_kind, compare_to, compare_to_status_kind,
+                    name, pset, comparison_window)
                 get_cell_value(cr)
             """
           </tr>"""




From daniel at zuster.org  Thu Jul 29 13:15:11 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:11 -0000
Subject: [llvm-commits] [zorg] r109781 -
	/zorg/trunk/lnt/lnt/db/perfdbsummary.py
Message-ID: <20100729181511.3DABE2A6C12D@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:11 2010
New Revision: 109781

URL: http://llvm.org/viewvc/llvm-project?rev=109781&view=rev
Log:
LNT/simple: Eliminate the test status map, which was bogus and is now unused.

Modified:
    zorg/trunk/lnt/lnt/db/perfdbsummary.py

Modified: zorg/trunk/lnt/lnt/db/perfdbsummary.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/db/perfdbsummary.py?rev=109781&r1=109780&r2=109781&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/db/perfdbsummary.py (original)
+++ zorg/trunk/lnt/lnt/db/perfdbsummary.py Thu Jul 29 13:15:11 2010
@@ -50,8 +50,6 @@
         test_names = set()
         parameter_sets = set()
         test_id_map = {}
-        has_status_markers = False
-        has_success_markers = False
         for t in tests:
             name = t.name.split('.', 1)[1]
 
@@ -62,26 +60,13 @@
 
             if name.endswith('.success'):
                 test_name = name.rsplit('.', 1)[0]
-                has_success_markers = True
             elif name.endswith('.status'):
                 test_name = name.rsplit('.', 1)[0]
-                has_status_markers = True
             else:
                 test_name = name
 
             test_names.add(test_name)
 
-        # Compute the test status info.
-        test_status_map = {}
-        if has_status_markers:
-            for test_name in test_names:
-                marker_name = '%s.status' % test_name
-                test_status_map[test_name] = (marker_name, True)
-        elif has_success_markers:
-            for test_name in test_names:
-                marker_name = '%s.success' % test_name
-                test_status_map[test_name] = (marker_name, False)
-
         # Order the test names.
         test_names = list(test_names)
         test_names.sort()
@@ -96,17 +81,14 @@
         parameter_sets.sort()
 
         return SimpleSuiteSummary(revision, tag, test_names,
-                                  test_id_map, test_status_map,
-                                  parameter_keys, parameter_sets)
+                                  test_id_map, parameter_keys, parameter_sets)
 
     def __init__(self, revision, tag, test_names,
-                 test_id_map, test_status_map,
-                 parameter_keys, parameter_sets):
+                 test_id_map, parameter_keys, parameter_sets):
         self.revision = revision
         self.tag = tag
         self.test_names = test_names
         self.test_id_map = test_id_map
-        self.test_status_map = test_status_map
         self.parameter_keys = parameter_keys
         self.parameter_sets = parameter_sets
         self.test_info_map = dict([(v,k) for k,v in test_id_map.items()])




From daniel at zuster.org  Thu Jul 29 13:15:13 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:13 -0000
Subject: [llvm-commits] [zorg] r109782 - /zorg/trunk/lnt/lnt/db/runinfo.py
Message-ID: <20100729181513.A65E32A6C12E@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:13 2010
New Revision: 109782

URL: http://llvm.org/viewvc/llvm-project?rev=109782&view=rev
Log:
LNT/simple: When computing statistics on previous runs (for determining performance regressions), make sure to ignore failing runs.

Modified:
    zorg/trunk/lnt/lnt/db/runinfo.py

Modified: zorg/trunk/lnt/lnt/db/runinfo.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/db/runinfo.py?rev=109782&r1=109781&r2=109782&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/db/runinfo.py (original)
+++ zorg/trunk/lnt/lnt/db/runinfo.py Thu Jul 29 13:15:13 2010
@@ -92,12 +92,12 @@
         self.sample_map = Util.multidict()
         self.loaded_samples = set()
 
-    def get_test_status_in_run(self, run, status_kind, test_name, pset):
+    def get_test_status_in_run(self, run_id, status_kind, test_name, pset):
         if status_kind == False: # .success
             status_name = test_name + '.success'
             status_test_id = self.test_suite_summary.test_id_map.get(
                 (status_name, pset))
-            run_status = self.sample_map.get((run.id, status_test_id))
+            run_status = self.sample_map.get((run_id, status_test_id))
             if run_status and int(run_status[0]) == 1:
                 return PASS
             else:
@@ -106,7 +106,7 @@
             status_name = test_name + '.status'
             status_test_id = self.test_suite_summary.test_id_map.get(
                 (status_name, pset))
-            run_status = self.sample_map.get((run.id, status_test_id))
+            run_status = self.sample_map.get((run_id, status_test_id))
             if not run_status:
                 return PASS
             else:
@@ -145,10 +145,10 @@
         run_failed = prev_failed = False
         run_status = prev_status = None
         run_status = self.get_test_status_in_run(
-            run, run_status_kind, test_name, pset)
+            run.id, run_status_kind, test_name, pset)
         if compare_to:
             prev_status = self.get_test_status_in_run(
-                compare_to, compare_to_status_kind, test_name, pset)
+                compare_to.id, compare_to_status_kind, test_name, pset)
         else:
             prev_status = None
 
@@ -180,9 +180,13 @@
         else:
             pct_delta = 0.0
 
-        # Get all previous values in the comparison window.
+        # Get all previous values in the comparison window, for passing runs.
+        #
+        # FIXME: This is using the wrong status kind. :/
         prev_values = [v for run_id in comparison_window
-                       for v in self.sample_map.get((run_id, test_id), ())]
+                       for v in self.sample_map.get((run_id, test_id), ())
+                       if self.get_test_status_in_run(run_id, run_status_kind,
+                                                      test_name, pset) == PASS]
         if prev_values:
             stddev = stats.standard_deviation(prev_values)
             MAD = stats.median_absolute_deviation(prev_values)




From daniel at zuster.org  Thu Jul 29 13:15:16 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 18:15:16 -0000
Subject: [llvm-commits] [zorg] r109783 - /zorg/trunk/lnt/lnt/db/runinfo.py
Message-ID: <20100729181516.16E7B2A6C12F@llvm.org>

Author: ddunbar
Date: Thu Jul 29 13:15:15 2010
New Revision: 109783

URL: http://llvm.org/viewvc/llvm-project?rev=109783&view=rev
Log:
LNT/simple: Always ignore performance changes under 1%, even if they are considered significant -- we don't have enough time to investigate them currently.

Modified:
    zorg/trunk/lnt/lnt/db/runinfo.py

Modified: zorg/trunk/lnt/lnt/db/runinfo.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/db/runinfo.py?rev=109783&r1=109782&r2=109783&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/db/runinfo.py (original)
+++ zorg/trunk/lnt/lnt/db/runinfo.py Thu Jul 29 13:15:15 2010
@@ -53,10 +53,12 @@
         # sample at; otherwise quantization means that we can't measure the
         # standard deviation with enough accuracy.
         if abs(self.delta) <= 2 * value_precision * confidence_interval:
-            if self.failed:
-                return UNCHANGED_FAIL
-            else:
-                return UNCHANGED_PASS
+            return UNCHANGED_PASS
+
+        # Always ignore percentage changes below 1%, for now, we just don't have enough
+        # time to investigate that level of stuff.
+        if abs(self.pct_delta) < .01:
+            return UNCHANGED_PASS
 
         # If we have a comparison window, then measure using a symmetic
         # confidence interval.
@@ -67,10 +69,7 @@
                 else:
                     return REGRESSED
             else:
-                if self.failed:
-                    return UNCHANGED_FAIL
-                else:
-                    return UNCHANGED_PASS
+                return UNCHANGED_PASS
 
         # Otherwise, use the old "significant change" metric of > 5%.
         if abs(self.pct_delta) >= .05:
@@ -79,10 +78,7 @@
             else:
                 return REGRESSED
         else:
-            if self.failed:
-                return UNCHANGED_FAIL
-            else:
-                return UNCHANGED_PASS
+            return UNCHANGED_PASS
 
 class SimpleRunInfo:
     def __init__(self, db, test_suite_summary):




From rjmccall at apple.com  Thu Jul 29 13:20:14 2010
From: rjmccall at apple.com (John McCall)
Date: Thu, 29 Jul 2010 18:20:14 -0000
Subject: [llvm-commits] [llvm] r109787 -
	/llvm/trunk/tools/llvm-diff/llvm-diff.cpp
Message-ID: <20100729182014.27DB92A6C12C@llvm.org>

Author: rjmccall
Date: Thu Jul 29 13:20:13 2010
New Revision: 109787

URL: http://llvm.org/viewvc/llvm-project?rev=109787&view=rev
Log:
Transcribe IRC to svn.  Also don't print basic block names twice if they match.


Modified:
    llvm/trunk/tools/llvm-diff/llvm-diff.cpp

Modified: llvm/trunk/tools/llvm-diff/llvm-diff.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-diff/llvm-diff.cpp?rev=109787&r1=109786&r2=109787&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-diff/llvm-diff.cpp (original)
+++ llvm/trunk/tools/llvm-diff/llvm-diff.cpp Thu Jul 29 13:20:13 2010
@@ -163,17 +163,22 @@
         Function *L = cast<Function>(I->L);
         Function *R = cast<Function>(I->R);
         if (L->getName() != R->getName())
-          out << "in function " << L->getName() << " / " << R->getName() << ":\n";
+          out << "in function " << L->getName()
+              << " / " << R->getName() << ":\n";
         else
           out << "in function " << L->getName() << ":\n";
       } else if (isa<BasicBlock>(I->L)) {
         BasicBlock *L = cast<BasicBlock>(I->L);
         BasicBlock *R = cast<BasicBlock>(I->R);
-        out << "  in block ";
-        printValue(L, true);
-        out << " / ";
-        printValue(R, false);
-        out << ":\n";
+        if (L->hasName() && R->hasName() && L->getName() == R->getName())
+          out << "  in block %" << L->getName() << ":\n";
+        else {
+          out << "  in block ";
+          printValue(L, true);
+          out << " / ";
+          printValue(R, false);
+          out << ":\n";
+        }
       } else if (isa<Instruction>(I->L)) {
         out << "    in instruction ";
         printValue(I->L, true);
@@ -289,9 +294,14 @@
     errs() << "No function named @" << Name << " in right module\n";
 }
 
-cl::opt<std::string> LeftFilename(cl::Positional, cl::desc("<first file>"), cl::Required);
-cl::opt<std::string> RightFilename(cl::Positional, cl::desc("<second file>"), cl::Required);
-cl::list<std::string> GlobalsToCompare(cl::Positional, cl::desc("<globals to compare>"));
+cl::opt<std::string> LeftFilename(cl::Positional,
+                                  cl::desc("<first file>"),
+                                  cl::Required);
+cl::opt<std::string> RightFilename(cl::Positional,
+                                   cl::desc("<second file>"),
+                                   cl::Required);
+cl::list<std::string> GlobalsToCompare(cl::Positional,
+                                       cl::desc("<globals to compare>"));
 
 int main(int argc, char **argv) {
   cl::ParseCommandLineOptions(argc, argv);




From bob.wilson at apple.com  Thu Jul 29 13:29:28 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 29 Jul 2010 18:29:28 -0000
Subject: [llvm-commits] [llvm] r109788 -
 /llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
Message-ID: <20100729182928.9EAD22A6C12C@llvm.org>

Author: bwilson
Date: Thu Jul 29 13:29:28 2010
New Revision: 109788

URL: http://llvm.org/viewvc/llvm-project?rev=109788&view=rev
Log:
Don't assert on an unrecognized BrMiscFrm instruction.
PR7745.

Modified:
    llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp

Modified: llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp?rev=109788&r1=109787&r2=109788&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp Thu Jul 29 13:29:28 2010
@@ -887,7 +887,6 @@
     return true;
   }
 
-  assert(0 && "Unexpected BrMiscFrm Opcode");
   return false;
 }
 




From stuart at apple.com  Thu Jul 29 13:33:06 2010
From: stuart at apple.com (Stuart Hastings)
Date: Thu, 29 Jul 2010 18:33:06 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109789 -
 /llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c
Message-ID: <20100729183306.5C90E2A6C12C@llvm.org>

Author: stuart
Date: Thu Jul 29 13:33:06 2010
New Revision: 109789

URL: http://llvm.org/viewvc/llvm-project?rev=109789&view=rev
Log:
XFAIL a nested-function test.  Radar 8251322.

Modified:
    llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c

Modified: llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c?rev=109789&r1=109788&r2=109789&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c (original)
+++ llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c Thu Jul 29 13:33:06 2010
@@ -1,6 +1,8 @@
 /* APPLE LOCAL begin testsuite nested functions */
 /* { dg-options "-fnested-functions" } */
 /* APPLE LOCAL end testsuite nested functions */
+/* LLVM LOCAL 8251263 */
+/* { dg-xfail-if "" { *-*-darwin* } { "*" } { "" } } */
 
 extern void abort (void);
 extern void exit (int);




From echristo at apple.com  Thu Jul 29 13:38:24 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 11:38:24 -0700
Subject: [llvm-commits] [llvm-gcc-4.2] r109789 -
	/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c
In-Reply-To: <20100729183306.5C90E2A6C12C@llvm.org>
References: <20100729183306.5C90E2A6C12C@llvm.org>
Message-ID: <5F5FEABD-DD32-44F6-954C-E5ECFA46975E@apple.com>


On Jul 29, 2010, at 11:33 AM, Stuart Hastings wrote:

> Author: stuart
> Date: Thu Jul 29 13:33:06 2010
> New Revision: 109789
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109789&view=rev
> Log:
> XFAIL a nested-function test.  Radar 8251322.
> 
> Modified:
>    llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c
> 
> Modified: llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c
> URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c?rev=109789&r1=109788&r2=109789&view=diff
> ==============================================================================
> --- llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c (original)
> +++ llvm-gcc-4.2/trunk/gcc/testsuite/gcc.c-torture/compile/920415-1.c Thu Jul 29 13:33:06 2010
> @@ -1,6 +1,8 @@
> /* APPLE LOCAL begin testsuite nested functions */
> /* { dg-options "-fnested-functions" } */
> /* APPLE LOCAL end testsuite nested functions */
> +/* LLVM LOCAL 8251263 */
> +/* { dg-xfail-if "" { *-*-darwin* } { "*" } { "" } } */

Uh.... the -fnested-functions should make it pass, otherwise it's a bug.

-eric


From dalej at apple.com  Thu Jul 29 15:10:08 2010
From: dalej at apple.com (Dale Johannesen)
Date: Thu, 29 Jul 2010 20:10:08 -0000
Subject: [llvm-commits] [llvm] r109799 - in /llvm/trunk:
 lib/Target/ARM/ARMISelLowering.cpp test/CodeGen/Thumb2/machine-licm-vdup.ll
Message-ID: <20100729201009.07D012A6C12C@llvm.org>

Author: johannes
Date: Thu Jul 29 15:10:08 2010
New Revision: 109799

URL: http://llvm.org/viewvc/llvm-project?rev=109799&view=rev
Log:
Implement vector constants which are splat of
integers with mov + vdup.  8003375.  This is
currently disabled by default because LICM will
not hoist a VDUP, so it pessimizes the code if
the construct occurs inside a loop (8248029).


Added:
    llvm/trunk/test/CodeGen/Thumb2/machine-licm-vdup.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=109799&r1=109798&r2=109799&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Thu Jul 29 15:10:08 2010
@@ -57,6 +57,13 @@
   cl::desc("Generate tail calls (TEMPORARY OPTION)."),
   cl::init(true));
 
+// This option should go away when Machine LICM is smart enough to hoist a 
+// reg-to-reg VDUP.
+static cl::opt<bool>
+EnableARMVDUPsplat("arm-vdup-splat", cl::Hidden,
+  cl::desc("Generate VDUP for integer constant splats (TEMPORARY OPTION)."),
+  cl::init(false));
+
 static cl::opt<bool>
 EnableARMLongCalls("arm-long-calls", cl::Hidden,
   cl::desc("Generate calls via indirect call instructions"),
@@ -3257,9 +3264,30 @@
   return true;
 }
 
+// If N is an integer constant that can be moved into a register in one
+// instruction, return an SDValue of such a constant (will become a MOV
+// instruction).  Otherwise return null.
+static SDValue IsSingleInstrConstant(SDValue N, SelectionDAG &DAG,
+                                     const ARMSubtarget *ST, DebugLoc dl) {
+  uint64_t Val;
+  if (!isa<ConstantSDNode>(N))
+    return SDValue();
+  Val = cast<ConstantSDNode>(N)->getZExtValue();
+
+  if (ST->isThumb1Only()) {
+    if (Val <= 255 || ~Val <= 255)
+      return DAG.getConstant(Val, MVT::i32);
+  } else {
+    if (ARM_AM::getSOImmVal(Val) != -1 || ARM_AM::getSOImmVal(~Val) != -1)
+      return DAG.getConstant(Val, MVT::i32);
+  }
+  return SDValue();
+}
+
 // If this is a case we can't handle, return null and let the default
 // expansion code take care of it.
-static SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG) {
+static SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG, 
+                                 const ARMSubtarget *ST) {
   BuildVectorSDNode *BVN = cast<BuildVectorSDNode>(Op.getNode());
   DebugLoc dl = Op.getDebugLoc();
   EVT VT = Op.getValueType();
@@ -3319,15 +3347,41 @@
   if (isOnlyLowElement)
     return DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, VT, Value);
 
-  // If all elements are constants, fall back to the default expansion, which
-  // will generate a load from the constant pool.
+  unsigned EltSize = VT.getVectorElementType().getSizeInBits();
+
+  if (EnableARMVDUPsplat) {
+    // Use VDUP for non-constant splats.  For f32 constant splats, reduce to
+    // i32 and try again.
+    if (usesOnlyOneValue && EltSize <= 32) {
+      if (!isConstant)
+        return DAG.getNode(ARMISD::VDUP, dl, VT, Value);
+      if (VT.getVectorElementType().isFloatingPoint()) {
+        SmallVector<SDValue, 8> Ops;
+        for (unsigned i = 0; i < NumElts; ++i)
+          Ops.push_back(DAG.getNode(ISD::BIT_CONVERT, dl, MVT::i32, 
+                                    Op.getOperand(i)));
+        SDValue Val = DAG.getNode(ISD::BUILD_VECTOR, dl, MVT::v4i32, &Ops[0],
+                                  NumElts);
+        return DAG.getNode(ISD::BIT_CONVERT, dl, VT, 
+                           LowerBUILD_VECTOR(Val, DAG, ST));
+      }
+      SDValue Val = IsSingleInstrConstant(Value, DAG, ST, dl);
+      if (Val.getNode())
+        return DAG.getNode(ARMISD::VDUP, dl, VT, Val);
+    }
+  }
+
+  // If all elements are constants and the case above didn't get hit, fall back
+  // to the default expansion, which will generate a load from the constant
+  // pool.
   if (isConstant)
     return SDValue();
 
-  // Use VDUP for non-constant splats.
-  unsigned EltSize = VT.getVectorElementType().getSizeInBits();
-  if (usesOnlyOneValue && EltSize <= 32)
-    return DAG.getNode(ARMISD::VDUP, dl, VT, Value);
+  if (!EnableARMVDUPsplat) {
+    // Use VDUP for non-constant splats.
+    if (usesOnlyOneValue && EltSize <= 32)
+      return DAG.getNode(ARMISD::VDUP, dl, VT, Value);
+  }
 
   // Vectors with 32- or 64-bit elements can be built by directly assigning
   // the subregisters.  Lower it to an ARMISD::BUILD_VECTOR so the operands
@@ -3647,7 +3701,7 @@
   case ISD::SRA_PARTS:     return LowerShiftRightParts(Op, DAG);
   case ISD::CTTZ:          return LowerCTTZ(Op.getNode(), DAG, Subtarget);
   case ISD::VSETCC:        return LowerVSETCC(Op, DAG);
-  case ISD::BUILD_VECTOR:  return LowerBUILD_VECTOR(Op, DAG);
+  case ISD::BUILD_VECTOR:  return LowerBUILD_VECTOR(Op, DAG, Subtarget);
   case ISD::VECTOR_SHUFFLE: return LowerVECTOR_SHUFFLE(Op, DAG);
   case ISD::EXTRACT_VECTOR_ELT: return LowerEXTRACT_VECTOR_ELT(Op, DAG);
   case ISD::CONCAT_VECTORS: return LowerCONCAT_VECTORS(Op, DAG);

Added: llvm/trunk/test/CodeGen/Thumb2/machine-licm-vdup.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/machine-licm-vdup.ll?rev=109799&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb2/machine-licm-vdup.ll (added)
+++ llvm/trunk/test/CodeGen/Thumb2/machine-licm-vdup.ll Thu Jul 29 15:10:08 2010
@@ -0,0 +1,38 @@
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -disable-fp-elim                -arm-vdup-splat | FileCheck %s
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin -mcpu=cortex-a8 -relocation-model=pic -disable-fp-elim -arm-vdup-splat | FileCheck %s 
+; Modified version of machine-licm.ll with -arm-vdup-splat turned on, 8003375.
+; Eventually this should become the default and be moved into machine-licm.ll.
+; FIXME: the vdup should be hoisted out of the loop, 8248029.
+
+define void @t2(i8* %ptr1, i8* %ptr2) nounwind {
+entry:
+; CHECK: t2:
+; CHECK: mov.w r3, #1065353216
+  br i1 undef, label %bb1, label %bb2
+
+bb1:
+; CHECK-NEXT: %bb1
+; CHECK: vdup.32 q1, r3
+  %indvar = phi i32 [ %indvar.next, %bb1 ], [ 0, %entry ]
+  %tmp1 = shl i32 %indvar, 2
+  %gep1 = getelementptr i8* %ptr1, i32 %tmp1
+  %tmp2 = call <4 x float> @llvm.arm.neon.vld1.v4f32(i8* %gep1)
+  %tmp3 = call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> <float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, <4 x float> %tmp2)
+  %gep2 = getelementptr i8* %ptr2, i32 %tmp1
+  call void @llvm.arm.neon.vst1.v4f32(i8* %gep2, <4 x float> %tmp3)
+  %indvar.next = add i32 %indvar, 1
+  %cond = icmp eq i32 %indvar.next, 10
+  br i1 %cond, label %bb2, label %bb1
+
+bb2:
+  ret void
+}
+
+; CHECK-NOT: LCPI1_0:
+; CHECK: .subsections_via_symbols
+
+declare <4 x float> @llvm.arm.neon.vld1.v4f32(i8*) nounwind readonly
+
+declare void @llvm.arm.neon.vst1.v4f32(i8*, <4 x float>) nounwind
+
+declare <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float>, <4 x float>) nounwind readnone




From espindola at google.com  Thu Jul 29 15:21:20 2010
From: espindola at google.com (Rafael Espindola)
Date: Thu, 29 Jul 2010 16:21:20 -0400
Subject: [llvm-commits] [patch] Another small refactoring in bugpoint
Message-ID: <AANLkTinCc=4x2LZaHyGQ898aDb=YObPBreH-nhAevHHx@mail.gmail.com>

Attached is another small refactoring on the same lines of the
previous one. It adds a Program option to diffProgram to avoid another
silly swapProgramIn.

Cheers,
-- 
Rafael ?vila de Esp?ndola
-------------- next part --------------
A non-text attachment was scrubbed...
Name: bugpoint.patch
Type: text/x-patch
Size: 8916 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100729/9c33a00e/attachment.bin 

From bob.wilson at apple.com  Thu Jul 29 15:34:14 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 29 Jul 2010 20:34:14 -0000
Subject: [llvm-commits] [llvm] r109800 -
	/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
Message-ID: <20100729203414.A1BA42A6C12C@llvm.org>

Author: bwilson
Date: Thu Jul 29 15:34:14 2010
New Revision: 109800

URL: http://llvm.org/viewvc/llvm-project?rev=109800&view=rev
Log:
Refactor ARM-specific DAG combining in preparation for adding some more
transformations.

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

Modified: llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp?rev=109800&r1=109799&r2=109800&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Thu Jul 29 15:34:14 2010
@@ -4221,30 +4221,43 @@
   return SDValue();
 }
 
-/// PerformADDCombine - Target-specific dag combine xforms for ISD::ADD.
-static SDValue PerformADDCombine(SDNode *N,
-                                 TargetLowering::DAGCombinerInfo &DCI) {
-  // added by evan in r37685 with no testcase.
-  SDValue N0 = N->getOperand(0), N1 = N->getOperand(1);
-
+/// PerformADDCombineWithOperands - Try DAG combinations for an ADD with
+/// operands N0 and N1.  This is a helper for PerformADDCombine that is
+/// called with the default operands, and if that fails, with commuted
+/// operands.
+static SDValue PerformADDCombineWithOperands(SDNode *N, SDValue N0, SDValue N1,
+                                         TargetLowering::DAGCombinerInfo &DCI) {
   // fold (add (select cc, 0, c), x) -> (select cc, x, (add, x, c))
   if (N0.getOpcode() == ISD::SELECT && N0.getNode()->hasOneUse()) {
     SDValue Result = combineSelectAndUse(N, N0, N1, DCI);
     if (Result.getNode()) return Result;
   }
-  if (N1.getOpcode() == ISD::SELECT && N1.getNode()->hasOneUse()) {
-    SDValue Result = combineSelectAndUse(N, N1, N0, DCI);
-    if (Result.getNode()) return Result;
-  }
 
   return SDValue();
 }
 
+/// PerformADDCombine - Target-specific dag combine xforms for ISD::ADD.
+///
+static SDValue PerformADDCombine(SDNode *N,
+                                 TargetLowering::DAGCombinerInfo &DCI) {
+  SDValue N0 = N->getOperand(0);
+  SDValue N1 = N->getOperand(1);
+
+  // First try with the default operand order.
+  SDValue Result = PerformADDCombineWithOperands(N, N0, N1, DCI);
+  if (Result.getNode())
+    return Result;
+
+  // If that didn't work, try again with the operands commuted.
+  return PerformADDCombineWithOperands(N, N1, N0, DCI);
+}
+
 /// PerformSUBCombine - Target-specific dag combine xforms for ISD::SUB.
+///
 static SDValue PerformSUBCombine(SDNode *N,
                                  TargetLowering::DAGCombinerInfo &DCI) {
-  // added by evan in r37685 with no testcase.
-  SDValue N0 = N->getOperand(0), N1 = N->getOperand(1);
+  SDValue N0 = N->getOperand(0);
+  SDValue N1 = N->getOperand(1);
 
   // fold (sub x, (select cc, 0, c)) -> (select cc, x, (sub, x, c))
   if (N1.getOpcode() == ISD::SELECT && N1.getNode()->hasOneUse()) {




From benny.kra at googlemail.com  Thu Jul 29 15:36:36 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 20:36:36 -0000
Subject: [llvm-commits] [llvm] r109801 -
	/llvm/trunk/test/Other/inline-asm-newline-terminator.ll
Message-ID: <20100729203636.5702E2A6C12C@llvm.org>

Author: d0k
Date: Thu Jul 29 15:36:36 2010
New Revision: 109801

URL: http://llvm.org/viewvc/llvm-project?rev=109801&view=rev
Log:
Remove XFAIL, test doesn't leak anymore.

Modified:
    llvm/trunk/test/Other/inline-asm-newline-terminator.ll

Modified: llvm/trunk/test/Other/inline-asm-newline-terminator.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Other/inline-asm-newline-terminator.ll?rev=109801&r1=109800&r2=109801&view=diff
==============================================================================
--- llvm/trunk/test/Other/inline-asm-newline-terminator.ll (original)
+++ llvm/trunk/test/Other/inline-asm-newline-terminator.ll Thu Jul 29 15:36:36 2010
@@ -1,5 +1,4 @@
 ; RUN: llc -filetype=obj -o - < %s
-; XFAIL: vg_leak
 
 ; ModuleID = 't.c'
 target triple = "x86_64-apple-darwin10.0.0"




From daniel at zuster.org  Thu Jul 29 17:18:50 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 29 Jul 2010 22:18:50 -0000
Subject: [llvm-commits] [llvm] r109809 - /llvm/trunk/unittests/Makefile
Message-ID: <20100729221850.A16DC2A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 17:18:50 2010
New Revision: 109809

URL: http://llvm.org/viewvc/llvm-project?rev=109809&view=rev
Log:
Temporary hack to ensure stale Analysis dir is cleaned up on llvm-i686-linux.

Modified:
    llvm/trunk/unittests/Makefile

Modified: llvm/trunk/unittests/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile?rev=109809&r1=109808&r2=109809&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile (original)
+++ llvm/trunk/unittests/Makefile Thu Jul 29 17:18:50 2010
@@ -14,4 +14,4 @@
 include $(LEVEL)/Makefile.common
 
 clean::
-	$(Verb) $(RM) -f *Tests
+	$(Verb) $(RM) -f *Tests Analysis




From bob.wilson at apple.com  Thu Jul 29 17:33:35 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 29 Jul 2010 22:33:35 -0000
Subject: [llvm-commits] [llvm] r109810 - /llvm/trunk/unittests/Makefile
Message-ID: <20100729223335.3D43A2A6C130@llvm.org>

Author: bwilson
Date: Thu Jul 29 17:33:35 2010
New Revision: 109810

URL: http://llvm.org/viewvc/llvm-project?rev=109810&view=rev
Log:
Fix the temporary hack.

Modified:
    llvm/trunk/unittests/Makefile

Modified: llvm/trunk/unittests/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile?rev=109810&r1=109809&r2=109810&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile (original)
+++ llvm/trunk/unittests/Makefile Thu Jul 29 17:33:35 2010
@@ -14,4 +14,5 @@
 include $(LEVEL)/Makefile.common
 
 clean::
-	$(Verb) $(RM) -f *Tests Analysis
+	$(Verb) $(RM) -rf Analysis
+	$(Verb) $(RM) -f *Tests




From natebegeman at mac.com  Thu Jul 29 17:48:09 2010
From: natebegeman at mac.com (Nate Begeman)
Date: Thu, 29 Jul 2010 22:48:09 -0000
Subject: [llvm-commits] [llvm] r109813 - in /llvm/trunk:
 include/llvm/IntrinsicsARM.td lib/Target/ARM/ARMInstrInfo.td
 lib/Target/ARM/ARMInstrThumb2.td
Message-ID: <20100729224809.681A82A6C12C@llvm.org>

Author: sampo
Date: Thu Jul 29 17:48:09 2010
New Revision: 109813

URL: http://llvm.org/viewvc/llvm-project?rev=109813&view=rev
Log:
Add builtins for ssat/usat, similar to RealView's __ssat and __usat intrinsics.

Modified:
    llvm/trunk/include/llvm/IntrinsicsARM.td
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/include/llvm/IntrinsicsARM.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/IntrinsicsARM.td?rev=109813&r1=109812&r2=109813&view=diff
==============================================================================
--- llvm/trunk/include/llvm/IntrinsicsARM.td (original)
+++ llvm/trunk/include/llvm/IntrinsicsARM.td Thu Jul 29 17:48:09 2010
@@ -29,6 +29,10 @@
                         [IntrNoMem, Commutative]>;
   def int_arm_qsub : GCCBuiltin<"__builtin_arm_qsub">,
               Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_arm_ssat : GCCBuiltin<"__builtin_arm_ssat">,
+              Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
+  def int_arm_usat : GCCBuiltin<"__builtin_arm_usat">,
+              Intrinsic<[llvm_i32_ty], [llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
 }
 
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=109813&r1=109812&r2=109813&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Thu Jul 29 17:48:09 2010
@@ -1835,6 +1835,9 @@
   let Inst{7-4} = 0b0011;
 }
 
+def : ARMV6Pat<(int_arm_ssat GPR:$a, imm:$pos), (SSATlsl imm:$pos, GPR:$a, 0)>;
+def : ARMV6Pat<(int_arm_usat GPR:$a, imm:$pos), (USATlsl imm:$pos, GPR:$a, 0)>;
+
 //===----------------------------------------------------------------------===//
 //  Bitwise Instructions.
 //

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=109813&r1=109812&r2=109813&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Thu Jul 29 17:48:09 2010
@@ -1579,6 +1579,9 @@
   let Inst{7-6} = 0b00;    // imm2 = '00'
 }
 
+def : T2Pat<(int_arm_ssat GPR:$a, imm:$pos), (t2SSATlsl imm:$pos, GPR:$a, 0)>;
+def : T2Pat<(int_arm_usat GPR:$a, imm:$pos), (t2USATlsl imm:$pos, GPR:$a, 0)>;
+
 //===----------------------------------------------------------------------===//
 //  Shift and rotate Instructions.
 //




From bob.wilson at apple.com  Thu Jul 29 17:55:29 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Thu, 29 Jul 2010 22:55:29 -0000
Subject: [llvm-commits] [llvm] r109815 - /llvm/trunk/unittests/Makefile
Message-ID: <20100729225529.7A68D2A6C12C@llvm.org>

Author: bwilson
Date: Thu Jul 29 17:55:29 2010
New Revision: 109815

URL: http://llvm.org/viewvc/llvm-project?rev=109815&view=rev
Log:
Remove the temporary hack, now that the llvm-i686-linux buildbot has
completed a successful build.

Modified:
    llvm/trunk/unittests/Makefile

Modified: llvm/trunk/unittests/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/Makefile?rev=109815&r1=109814&r2=109815&view=diff
==============================================================================
--- llvm/trunk/unittests/Makefile (original)
+++ llvm/trunk/unittests/Makefile Thu Jul 29 17:55:29 2010
@@ -14,5 +14,4 @@
 include $(LEVEL)/Makefile.common
 
 clean::
-	$(Verb) $(RM) -rf Analysis
 	$(Verb) $(RM) -f *Tests




From benny.kra at googlemail.com  Thu Jul 29 18:15:04 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Thu, 29 Jul 2010 23:15:04 -0000
Subject: [llvm-commits] [llvm] r109818 - in /llvm/trunk/utils/valgrind:
 i386-pc-linux-gnu.supp x86_64-pc-linux-gnu.supp
Message-ID: <20100729231504.C48612A6C12C@llvm.org>

Author: d0k
Date: Thu Jul 29 18:15:04 2010
New Revision: 109818

URL: http://llvm.org/viewvc/llvm-project?rev=109818&view=rev
Log:
Supress valgrind errors from python.

Modified:
    llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp
    llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp

Modified: llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp?rev=109818&r1=109817&r2=109818&view=diff
==============================================================================
--- llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp (original)
+++ llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp Thu Jul 29 18:15:04 2010
@@ -5,3 +5,31 @@
    fun:_ZN83_GLOBAL_*PassRegistrar12RegisterPassERKN4llvm8PassInfoE
    fun:_ZN4llvm8PassInfo12registerPassEv
 }
+
+# Python false positives according to
+# http://svn.python.org/projects/python/trunk/Misc/README.valgrind
+
+{
+   ADDRESS_IN_RANGE/Invalid read of size 4
+   Memcheck:Addr4
+   obj:/usr/bin/python2.5
+}
+
+{
+   ADDRESS_IN_RANGE/Invalid read of size 4
+   Memcheck:Value4
+   obj:/usr/bin/python2.5
+}
+
+{
+   ADDRESS_IN_RANGE/Conditional jump or move depends on uninitialised value
+   Memcheck:Cond
+   obj:/usr/bin/python2.5
+}
+
+{
+   We don't care if python leaks
+   Memcheck:Leak
+   obj:/usr/bin/python2.5
+}
+

Modified: llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp?rev=109818&r1=109817&r2=109818&view=diff
==============================================================================
--- llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp (original)
+++ llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp Thu Jul 29 18:15:04 2010
@@ -5,3 +5,30 @@
    fun:_ZN83_GLOBAL_*PassRegistrar12RegisterPassERKN4llvm8PassInfoE
    fun:_ZN4llvm8PassInfo12registerPassEv
 }
+
+# Python false positives according to
+# http://svn.python.org/projects/python/trunk/Misc/README.valgrind
+
+{
+   ADDRESS_IN_RANGE/Invalid read of size 4
+   Memcheck:Addr4
+   obj:/usr/bin/python2.5
+}
+
+{
+   ADDRESS_IN_RANGE/Invalid read of size 4
+   Memcheck:Value8
+   obj:/usr/bin/python2.5
+}
+
+{
+   ADDRESS_IN_RANGE/Conditional jump or move depends on uninitialised value
+   Memcheck:Cond
+   obj:/usr/bin/python2.5
+}
+
+{
+   We don't care if python leaks
+   Memcheck:Leak
+   obj:/usr/bin/python2.5
+}




From echristo at apple.com  Thu Jul 29 18:20:10 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 23:20:10 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r109819 -
	/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
Message-ID: <20100729232010.A1EEE2A6C12C@llvm.org>

Author: echristo
Date: Thu Jul 29 18:20:10 2010
New Revision: 109819

URL: http://llvm.org/viewvc/llvm-project?rev=109819&view=rev
Log:
Make this a warning instead so it can be ignored.

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

Modified: llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp?rev=109819&r1=109818&r2=109819&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp (original)
+++ llvm-gcc-4.2/trunk/gcc/llvm-convert.cpp Thu Jul 29 18:20:10 2010
@@ -816,13 +816,12 @@
 
   // If we've seen a vla in this function and we'll possibly need to
   // either dynamically realign or this is greater than the maximum stack
-  // alignment, error out now.  This is here so we don't output an error
-  // every time we see a variable.
+  // alignment, output a warning.  This is here so we don't warn every time
+  // we see a variable.
   if (SeenVLA &&
       GreatestAlignment > TheTarget->getFrameInfo()->getStackAlignment())
-      error ("alignment for %q+D conflicts with a dynamically realigned stack",
-             SeenVLA);
-
+      warning (0, "alignment for %q+D conflicts with a dynamically realigned "
+                  "stack", SeenVLA);
   return Fn;
 }
 




From echristo at apple.com  Thu Jul 29 18:20:29 2010
From: echristo at apple.com (Eric Christopher)
Date: Thu, 29 Jul 2010 23:20:29 -0000
Subject: [llvm-commits] [llvm] r109820 - /llvm/trunk/test/FrontendC/vla-1.c
Message-ID: <20100729232029.CFFFC2A6C12C@llvm.org>

Author: echristo
Date: Thu Jul 29 18:20:29 2010
New Revision: 109820

URL: http://llvm.org/viewvc/llvm-project?rev=109820&view=rev
Log:
Fix this up per llvm-gcc r109819.

Modified:
    llvm/trunk/test/FrontendC/vla-1.c

Modified: llvm/trunk/test/FrontendC/vla-1.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/vla-1.c?rev=109820&r1=109819&r2=109820&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC/vla-1.c (original)
+++ llvm/trunk/test/FrontendC/vla-1.c Thu Jul 29 18:20:29 2010
@@ -1,4 +1,4 @@
-// RUN: not %llvmgcc -std=gnu99 %s -S |& grep "error: alignment for"
+// RUN: %llvmgcc_only -std=gnu99 %s -S |& grep "warning: alignment for"
 
 int foo(int a)
 {




From ggreif at gmail.com  Thu Jul 29 18:35:00 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Thu, 29 Jul 2010 23:35:00 -0000
Subject: [llvm-commits] [llvm] r109821 -
	/llvm/trunk/include/llvm/Support/CallSite.h
Message-ID: <20100729233500.995EF2A6C12C@llvm.org>

Author: ggreif
Date: Thu Jul 29 18:35:00 2010
New Revision: 109821

URL: http://llvm.org/viewvc/llvm-project?rev=109821&view=rev
Log:
remove a layer of cruft

Modified:
    llvm/trunk/include/llvm/Support/CallSite.h

Modified: llvm/trunk/include/llvm/Support/CallSite.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/CallSite.h?rev=109821&r1=109820&r2=109821&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Support/CallSite.h (original)
+++ llvm/trunk/include/llvm/Support/CallSite.h Thu Jul 29 18:35:00 2010
@@ -116,13 +116,13 @@
 
   ValTy *getArgument(unsigned ArgNo) const {
     assert(arg_begin() + ArgNo < arg_end() && "Argument # out of range!");
-    return *(arg_begin()+ArgNo);
+    return *(arg_begin() + ArgNo);
   }
 
   void setArgument(unsigned ArgNo, Value* newVal) {
     assert(getInstruction() && "Not a call or invoke instruction!");
     assert(arg_begin() + ArgNo < arg_end() && "Argument # out of range!");
-    getInstruction()->setOperand(getArgumentOffset() + ArgNo, newVal);
+    getInstruction()->setOperand(ArgNo, newVal);
   }
 
   /// Given a value use iterator, returns the argument that corresponds to it.
@@ -143,7 +143,7 @@
   IterTy arg_begin() const {
     assert(getInstruction() && "Not a call or invoke instruction!");
     // Skip non-arguments
-    return (*this)->op_begin() + getArgumentOffset();
+    return (*this)->op_begin();
   }
 
   IterTy arg_end() const { return (*this)->op_end() - getArgumentEndOffset(); }
@@ -253,12 +253,6 @@
   }
 
 private:
-  /// Returns the operand number of the first argument
-  /// FIXME: remove this func!
-  unsigned getArgumentOffset() const {
-    return 0; // Args are at the front
-  }
-
   unsigned getArgumentEndOffset() const {
     if (isCall())
       return 1; // Skip Callee




From nlewycky at google.com  Thu Jul 29 19:26:36 2010
From: nlewycky at google.com (Nick Lewycky)
Date: Thu, 29 Jul 2010 17:26:36 -0700
Subject: [llvm-commits] [patch] Another small refactoring in bugpoint
In-Reply-To: <AANLkTinCc=4x2LZaHyGQ898aDb=YObPBreH-nhAevHHx@mail.gmail.com>
References: <AANLkTinCc=4x2LZaHyGQ898aDb=YObPBreH-nhAevHHx@mail.gmail.com>
Message-ID: <AANLkTimnsXVO6NWgbk2mTqvP+i5b564uHuVbb1RUctG=@mail.gmail.com>

+  std::string executeProgram(const Module* Program,

Note that in the rest of the file, the space comes before the star. This
happens in several places, I won't mention them all.

Besides that it looks good. I'm quite surprised that even the comments on
diffProgram don't need to change. :)

Nick

On 29 July 2010 13:21, Rafael Espindola <espindola at google.com> wrote:

> Attached is another small refactoring on the same lines of the
> previous one. It adds a Program option to diffProgram to avoid another
> silly swapProgramIn.
>
> Cheers,
> --
> Rafael ?vila de Esp?ndola
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100729/92a9783f/attachment.html 

From echristo at apple.com  Thu Jul 29 19:34:38 2010
From: echristo at apple.com (Eric Christopher)
Date: Fri, 30 Jul 2010 00:34:38 -0000
Subject: [llvm-commits] [test-suite] r109824 -
	/test-suite/trunk/External/Makefile
Message-ID: <20100730003438.B6A322A6C12C@llvm.org>

Author: echristo
Date: Thu Jul 29 19:34:38 2010
New Revision: 109824

URL: http://llvm.org/viewvc/llvm-project?rev=109824&view=rev
Log:
Temporarily disable skidmarks - it's undergoing a bit of a rewrite.

Modified:
    test-suite/trunk/External/Makefile

Modified: test-suite/trunk/External/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/Makefile?rev=109824&r1=109823&r2=109824&view=diff
==============================================================================
--- test-suite/trunk/External/Makefile (original)
+++ test-suite/trunk/External/Makefile Thu Jul 29 19:34:38 2010
@@ -8,7 +8,9 @@
 #
 # Create the list of directories to compile
 #
-PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim skidmarks10
+PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim
+# Temporarily disable skidmarks
+# skidmarks10
 
 ifneq (,$(findstring x86,$(ARCH)))
 # Nurbs uses SSE and only works on x86 and x86_64.




From dirty at apple.com  Thu Jul 29 18:56:05 2010
From: dirty at apple.com (Cameron Esfahani)
Date: Thu, 29 Jul 2010 16:56:05 -0700
Subject: [llvm-commits] COFF patch
Message-ID: <B7F7863E-3D87-41DB-AFC0-B13206374BAE@apple.com>

My first patch, so please be gentle...

This fixes a problem I was seeing with the COFF back end.  All of the string data was getting put into one data fragment, and so none of the fixup references were correct.

I also made sure the TimeDateStamp in the COFF header had a correct value.

I also fixed a spot in MCFragment where Offset wasn't being properly initialized.

This patch was generated against revision 109821.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: COFFfragment.patch
Type: application/octet-stream
Size: 1983 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100729/6a4e1156/attachment.obj 
-------------- next part --------------


Cameron Esfahani
dirty at apple.com

"I cannot for the life of me understand why, while people without driver's licenses are not allowed on public roads, in bookstores one can find any number of books by persons without decency - let alone knowledge."

"His Master's Voice", Stanislaw Lem



From grosbach at apple.com  Thu Jul 29 21:41:01 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Fri, 30 Jul 2010 02:41:01 -0000
Subject: [llvm-commits] [llvm] r109842 - in /llvm/trunk:
 lib/Target/ARM/ARMBaseInstrInfo.cpp lib/Target/ARM/ARMInstrThumb2.td
 lib/Target/ARM/ARMRegisterInfo.td
 lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
 lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h
 lib/Target/ARM/Thumb2InstrInfo.cpp
 test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll
 test/CodeGen/Thumb2/thumb2-badreg-operands.ll utils/TableGen/EDEmitter.cpp
Message-ID: <20100730024101.B1A382A6C12C@llvm.org>

Author: grosbach
Date: Thu Jul 29 21:41:01 2010
New Revision: 109842

URL: http://llvm.org/viewvc/llvm-project?rev=109842&view=rev
Log:
Many Thumb2 instructions can reference the full ARM register set (i.e.,
have 4 bits per register in the operand encoding), but have undefined
behavior when the operand value is 13 or 15 (SP and PC, respectively).
The trivial coalescer in linear scan sometimes will merge a copy from
SP into a subsequent instruction which uses the copy, and if that
instruction cannot legally reference SP, we get bad code such as:
  mls r0,r9,r0,sp
instead of:
  mov r2, sp
  mls r0, r9, r0, r2

This patch adds a new register class for use by Thumb2 that excludes
the problematic registers (SP and PC) and is used instead of GPR
for those operands which cannot legally reference PC or SP. The
trivial coalescer explicitly requires that the register class
of the destination for the COPY instruction contain the source
register for the COPY to be considered for coalescing. This prevents
errant instructions like that above.

PR7499



Added:
    llvm/trunk/test/CodeGen/Thumb2/thumb2-badreg-operands.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
    llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td
    llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
    llvm/trunk/lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h
    llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp
    llvm/trunk/test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll
    llvm/trunk/utils/TableGen/EDEmitter.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp Thu Jul 29 21:41:01 2010
@@ -721,8 +721,9 @@
                             Align);
 
   // tGPR is used sometimes in ARM instructions that need to avoid using
-  // certain registers.  Just treat it as GPR here.
-  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass)
+  // certain registers.  Just treat it as GPR here. Likewise, rGPR.
+  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass
+      || RC == ARM::rGPRRegisterClass)
     RC = ARM::GPRRegisterClass;
 
   switch (RC->getID()) {
@@ -823,7 +824,8 @@
 
   // tGPR is used sometimes in ARM instructions that need to avoid using
   // certain registers.  Just treat it as GPR here.
-  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass)
+  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass
+      || RC == ARM::rGPRRegisterClass)
     RC = ARM::GPRRegisterClass;
 
   switch (RC->getID()) {

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Thu Jul 29 21:41:01 2010
@@ -32,7 +32,7 @@
                 ComplexPattern<i32, 2, "SelectT2ShifterOperandReg",
                                [shl,srl,sra,rotr]> {
   let PrintMethod = "printT2SOOperand";
-  let MIOperandInfo = (ops GPR, i32imm);
+  let MIOperandInfo = (ops rGPR, i32imm);
 }
 
 // t2_so_imm_not_XFORM - Return the complement of a t2_so_imm value
@@ -162,7 +162,7 @@
 def t2addrmode_so_reg : Operand<i32>,
                         ComplexPattern<i32, 3, "SelectT2AddrModeSoReg", []> {
   let PrintMethod = "printT2AddrModeSoRegOperand";
-  let MIOperandInfo = (ops GPR:$base, GPR:$offsreg, i32imm:$offsimm);
+  let MIOperandInfo = (ops GPR:$base, rGPR:$offsreg, i32imm:$offsimm);
 }
 
 
@@ -176,9 +176,9 @@
 multiclass T2I_un_irs<bits<4> opcod, string opc, PatFrag opnode,
                       bit Cheap = 0, bit ReMat = 0> {
    // shifted imm
-   def i : T2sI<(outs GPR:$dst), (ins t2_so_imm:$src), IIC_iMOVi,
+   def i : T2sI<(outs rGPR:$dst), (ins t2_so_imm:$src), IIC_iMOVi,
                 opc, "\t$dst, $src",
-                [(set GPR:$dst, (opnode t2_so_imm:$src))]> {
+                [(set rGPR:$dst, (opnode t2_so_imm:$src))]> {
      let isAsCheapAsAMove = Cheap;
      let isReMaterializable = ReMat;
      let Inst{31-27} = 0b11110;
@@ -189,9 +189,9 @@
      let Inst{15} = 0;
    }
    // register
-   def r : T2sI<(outs GPR:$dst), (ins GPR:$src), IIC_iMOVr,
+   def r : T2sI<(outs rGPR:$dst), (ins rGPR:$src), IIC_iMOVr,
                 opc, ".w\t$dst, $src",
-                [(set GPR:$dst, (opnode GPR:$src))]> {
+                [(set rGPR:$dst, (opnode rGPR:$src))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -202,9 +202,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def s : T2sI<(outs GPR:$dst), (ins t2_so_reg:$src), IIC_iMOVsi,
+   def s : T2sI<(outs rGPR:$dst), (ins t2_so_reg:$src), IIC_iMOVsi,
                 opc, ".w\t$dst, $src",
-                [(set GPR:$dst, (opnode t2_so_reg:$src))]> {
+                [(set rGPR:$dst, (opnode t2_so_reg:$src))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -219,9 +219,9 @@
 multiclass T2I_bin_irs<bits<4> opcod, string opc, PatFrag opnode,
                        bit Commutable = 0, string wide =""> {
    // shifted imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
+   def ri : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
                  opc, "\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_imm:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
      let Inst{24-21} = opcod;
@@ -229,9 +229,9 @@
      let Inst{15} = 0;
    }
    // register
-   def rr : T2sI<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
+   def rr : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, rGPR:$rhs), IIC_iALUr,
                  opc, !strconcat(wide, "\t$dst, $lhs, $rhs"),
-                 [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, rGPR:$rhs))]> {
      let isCommutable = Commutable;
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
@@ -242,9 +242,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def rs : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
+   def rs : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
                  opc, !strconcat(wide, "\t$dst, $lhs, $rhs"),
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_reg:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -263,9 +263,9 @@
 /// T2I_bin_irs counterpart.
 multiclass T2I_rbin_is<bits<4> opcod, string opc, PatFrag opnode> {
    // shifted imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$rhs, t2_so_imm:$lhs), IIC_iALUi,
+   def ri : T2sI<(outs rGPR:$dst), (ins rGPR:$rhs, t2_so_imm:$lhs), IIC_iALUi,
                  opc, ".w\t$dst, $rhs, $lhs",
-                 [(set GPR:$dst, (opnode t2_so_imm:$lhs, GPR:$rhs))]> {
+                 [(set rGPR:$dst, (opnode t2_so_imm:$lhs, rGPR:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
      let Inst{24-21} = opcod;
@@ -273,9 +273,9 @@
      let Inst{15} = 0;
    }
    // shifted register
-   def rs : T2sI<(outs GPR:$dst), (ins GPR:$rhs, t2_so_reg:$lhs), IIC_iALUsi,
+   def rs : T2sI<(outs rGPR:$dst), (ins rGPR:$rhs, t2_so_reg:$lhs), IIC_iALUsi,
                  opc, "\t$dst, $rhs, $lhs",
-                 [(set GPR:$dst, (opnode t2_so_reg:$lhs, GPR:$rhs))]> {
+                 [(set rGPR:$dst, (opnode t2_so_reg:$lhs, rGPR:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -289,9 +289,9 @@
 multiclass T2I_bin_s_irs<bits<4> opcod, string opc, PatFrag opnode,
                          bit Commutable = 0> {
    // shifted imm
-   def ri : T2I<(outs GPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
+   def ri : T2I<(outs rGPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
                 !strconcat(opc, "s"), ".w\t$dst, $lhs, $rhs",
-                [(set GPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]> {
+                [(set rGPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
      let Inst{24-21} = opcod;
@@ -299,9 +299,9 @@
      let Inst{15} = 0;
    }
    // register
-   def rr : T2I<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
+   def rr : T2I<(outs rGPR:$dst), (ins GPR:$lhs, rGPR:$rhs), IIC_iALUr,
                 !strconcat(opc, "s"), ".w\t$dst, $lhs, $rhs",
-                [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]> {
+                [(set rGPR:$dst, (opnode GPR:$lhs, rGPR:$rhs))]> {
      let isCommutable = Commutable;
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
@@ -312,9 +312,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def rs : T2I<(outs GPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
+   def rs : T2I<(outs rGPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
                 !strconcat(opc, "s"), ".w\t$dst, $lhs, $rhs",
-                [(set GPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]> {
+                [(set rGPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -328,9 +328,9 @@
 multiclass T2I_bin_ii12rs<bits<3> op23_21, string opc, PatFrag opnode,
                           bit Commutable = 0> {
    // shifted imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
+   def ri : T2sI<(outs rGPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]> {
+                 [(set rGPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
      let Inst{24} = 1;
@@ -339,9 +339,9 @@
      let Inst{15} = 0;
    }
    // 12-bit imm
-   def ri12 : T2I<(outs GPR:$dst), (ins GPR:$lhs, imm0_4095:$rhs), IIC_iALUi,
+   def ri12 : T2I<(outs rGPR:$dst), (ins GPR:$lhs, imm0_4095:$rhs), IIC_iALUi,
                   !strconcat(opc, "w"), "\t$dst, $lhs, $rhs",
-                  [(set GPR:$dst, (opnode GPR:$lhs, imm0_4095:$rhs))]> {
+                  [(set rGPR:$dst, (opnode GPR:$lhs, imm0_4095:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 1;
      let Inst{24} = 0;
@@ -350,9 +350,9 @@
      let Inst{15} = 0;
    }
    // register
-   def rr : T2sI<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
+   def rr : T2sI<(outs rGPR:$dst), (ins GPR:$lhs, rGPR:$rhs), IIC_iALUr,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]> {
+                 [(set rGPR:$dst, (opnode GPR:$lhs, rGPR:$rhs))]> {
      let isCommutable = Commutable;
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
@@ -364,9 +364,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def rs : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
+   def rs : T2sI<(outs rGPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]> {
+                 [(set rGPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24} = 1;
@@ -382,9 +382,9 @@
 multiclass T2I_adde_sube_irs<bits<4> opcod, string opc, PatFrag opnode,
                              bit Commutable = 0> {
    // shifted imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
+   def ri : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
                  opc, "\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_imm:$rhs))]>,
                  Requires<[IsThumb2]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
@@ -393,9 +393,9 @@
      let Inst{15} = 0;
    }
    // register
-   def rr : T2sI<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
+   def rr : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, rGPR:$rhs), IIC_iALUr,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, rGPR:$rhs))]>,
                  Requires<[IsThumb2]> {
      let isCommutable = Commutable;
      let Inst{31-27} = 0b11101;
@@ -407,9 +407,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def rs : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
+   def rs : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_reg:$rhs))]>,
                  Requires<[IsThumb2]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
@@ -423,9 +423,9 @@
 multiclass T2I_adde_sube_s_irs<bits<4> opcod, string opc, PatFrag opnode,
                                bit Commutable = 0> {
    // shifted imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
+   def ri : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_imm:$rhs), IIC_iALUi,
                  opc, "\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_imm:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_imm:$rhs))]>,
                  Requires<[IsThumb2]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
@@ -434,9 +434,9 @@
      let Inst{15} = 0;
    }
    // register
-   def rr : T2sI<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iALUr,
+   def rr : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, rGPR:$rhs), IIC_iALUr,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, rGPR:$rhs))]>,
                  Requires<[IsThumb2]> {
      let isCommutable = Commutable;
      let Inst{31-27} = 0b11101;
@@ -448,9 +448,9 @@
      let Inst{5-4} = 0b00; // type
    }
    // shifted register
-   def rs : T2sI<(outs GPR:$dst), (ins GPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
+   def rs : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, t2_so_reg:$rhs), IIC_iALUsi,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, t2_so_reg:$rhs))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, t2_so_reg:$rhs))]>,
                  Requires<[IsThumb2]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
@@ -465,9 +465,9 @@
 let Defs = [CPSR] in {
 multiclass T2I_rbin_s_is<bits<4> opcod, string opc, PatFrag opnode> {
    // shifted imm
-   def ri : T2I<(outs GPR:$dst), (ins GPR:$rhs, t2_so_imm:$lhs), IIC_iALUi,
+   def ri : T2I<(outs rGPR:$dst), (ins rGPR:$rhs, t2_so_imm:$lhs), IIC_iALUi,
                 !strconcat(opc, "s"), ".w\t$dst, $rhs, $lhs",
-                [(set GPR:$dst, (opnode t2_so_imm:$lhs, GPR:$rhs))]> {
+                [(set rGPR:$dst, (opnode t2_so_imm:$lhs, rGPR:$rhs))]> {
      let Inst{31-27} = 0b11110;
      let Inst{25} = 0;
      let Inst{24-21} = opcod;
@@ -475,9 +475,9 @@
      let Inst{15} = 0;
    }
    // shifted register
-   def rs : T2I<(outs GPR:$dst), (ins GPR:$rhs, t2_so_reg:$lhs), IIC_iALUsi,
+   def rs : T2I<(outs rGPR:$dst), (ins rGPR:$rhs, t2_so_reg:$lhs), IIC_iALUsi,
                 !strconcat(opc, "s"), "\t$dst, $rhs, $lhs",
-                [(set GPR:$dst, (opnode t2_so_reg:$lhs, GPR:$rhs))]> {
+                [(set rGPR:$dst, (opnode t2_so_reg:$lhs, rGPR:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -490,18 +490,18 @@
 //  rotate operation that produces a value.
 multiclass T2I_sh_ir<bits<2> opcod, string opc, PatFrag opnode> {
    // 5-bit imm
-   def ri : T2sI<(outs GPR:$dst), (ins GPR:$lhs, i32imm:$rhs), IIC_iMOVsi,
+   def ri : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, i32imm:$rhs), IIC_iMOVsi,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, imm1_31:$rhs))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, imm1_31:$rhs))]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-21} = 0b010010;
      let Inst{19-16} = 0b1111; // Rn
      let Inst{5-4} = opcod;
    }
    // register
-   def rr : T2sI<(outs GPR:$dst), (ins GPR:$lhs, GPR:$rhs), IIC_iMOVsr,
+   def rr : T2sI<(outs rGPR:$dst), (ins rGPR:$lhs, rGPR:$rhs), IIC_iMOVsr,
                  opc, ".w\t$dst, $lhs, $rhs",
-                 [(set GPR:$dst, (opnode GPR:$lhs, GPR:$rhs))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$lhs, rGPR:$rhs))]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
      let Inst{22-21} = opcod;
@@ -527,9 +527,9 @@
      let Inst{11-8} = 0b1111; // Rd
    }
    // register
-   def rr : T2I<(outs), (ins GPR:$lhs, GPR:$rhs), IIC_iCMPr,
+   def rr : T2I<(outs), (ins GPR:$lhs, rGPR:$rhs), IIC_iCMPr,
                 opc, ".w\t$lhs, $rhs",
-                [(opnode GPR:$lhs, GPR:$rhs)]> {
+                [(opnode GPR:$lhs, rGPR:$rhs)]> {
      let Inst{31-27} = 0b11101;
      let Inst{26-25} = 0b01;
      let Inst{24-21} = opcod;
@@ -639,9 +639,9 @@
 /// T2I_unary_rrot - A unary operation with two forms: one whose operand is a
 /// register and one whose operand is a register rotated by 8/16/24.
 multiclass T2I_unary_rrot<bits<3> opcod, string opc, PatFrag opnode> {
-  def r     : T2I<(outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+  def r     : T2I<(outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                   opc, ".w\t$dst, $src",
-                 [(set GPR:$dst, (opnode GPR:$src))]> {
+                 [(set rGPR:$dst, (opnode rGPR:$src))]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
      let Inst{22-20} = opcod;
@@ -650,9 +650,9 @@
      let Inst{7} = 1;
      let Inst{5-4} = 0b00; // rotate
    }
-  def r_rot : T2I<(outs GPR:$dst), (ins GPR:$src, i32imm:$rot), IIC_iUNAsi,
+  def r_rot : T2I<(outs rGPR:$dst), (ins rGPR:$src, i32imm:$rot), IIC_iUNAsi,
                   opc, ".w\t$dst, $src, ror $rot",
-                 [(set GPR:$dst, (opnode (rotr GPR:$src, rot_imm:$rot)))]> {
+                 [(set rGPR:$dst, (opnode (rotr rGPR:$src, rot_imm:$rot)))]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
      let Inst{22-20} = opcod;
@@ -665,9 +665,9 @@
 
 // UXTB16 - Requres T2ExtractPack, does not need the .w qualifier.
 multiclass T2I_unary_rrot_uxtb16<bits<3> opcod, string opc, PatFrag opnode> {
-  def r     : T2I<(outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+  def r     : T2I<(outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                   opc, "\t$dst, $src",
-                 [(set GPR:$dst, (opnode GPR:$src))]>,
+                 [(set rGPR:$dst, (opnode rGPR:$src))]>,
                  Requires<[HasT2ExtractPack]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -677,9 +677,9 @@
      let Inst{7} = 1;
      let Inst{5-4} = 0b00; // rotate
    }
-  def r_rot : T2I<(outs GPR:$dst), (ins GPR:$src, i32imm:$rot), IIC_iUNAsi,
+  def r_rot : T2I<(outs rGPR:$dst), (ins rGPR:$src, i32imm:$rot), IIC_iUNAsi,
                   opc, "\t$dst, $src, ror $rot",
-                 [(set GPR:$dst, (opnode (rotr GPR:$src, rot_imm:$rot)))]>,
+                 [(set rGPR:$dst, (opnode (rotr rGPR:$src, rot_imm:$rot)))]>,
                  Requires<[HasT2ExtractPack]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -694,7 +694,7 @@
 // SXTB16 - Requres T2ExtractPack, does not need the .w qualifier, no pattern
 // supported yet.
 multiclass T2I_unary_rrot_sxtb16<bits<3> opcod, string opc> {
-  def r     : T2I<(outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+  def r     : T2I<(outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                   opc, "\t$dst, $src", []> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -704,7 +704,7 @@
      let Inst{7} = 1;
      let Inst{5-4} = 0b00; // rotate
    }
-  def r_rot : T2I<(outs GPR:$dst), (ins GPR:$src, i32imm:$rot), IIC_iUNAsi,
+  def r_rot : T2I<(outs rGPR:$dst), (ins rGPR:$src, i32imm:$rot), IIC_iUNAsi,
                   opc, "\t$dst, $src, ror $rot", []> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -719,9 +719,9 @@
 /// T2I_bin_rrot - A binary operation with two forms: one whose operand is a
 /// register and one whose operand is a register rotated by 8/16/24.
 multiclass T2I_bin_rrot<bits<3> opcod, string opc, PatFrag opnode> {
-  def rr     : T2I<(outs GPR:$dst), (ins GPR:$LHS, GPR:$RHS), IIC_iALUr,
+  def rr     : T2I<(outs rGPR:$dst), (ins rGPR:$LHS, rGPR:$RHS), IIC_iALUr,
                   opc, "\t$dst, $LHS, $RHS",
-                  [(set GPR:$dst, (opnode GPR:$LHS, GPR:$RHS))]>,
+                  [(set rGPR:$dst, (opnode rGPR:$LHS, rGPR:$RHS))]>,
                   Requires<[HasT2ExtractPack]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -730,10 +730,10 @@
      let Inst{7} = 1;
      let Inst{5-4} = 0b00; // rotate
    }
-  def rr_rot : T2I<(outs GPR:$dst), (ins GPR:$LHS, GPR:$RHS, i32imm:$rot),
+  def rr_rot : T2I<(outs rGPR:$dst), (ins rGPR:$LHS, rGPR:$RHS, i32imm:$rot),
                   IIC_iALUsr, opc, "\t$dst, $LHS, $RHS, ror $rot",
-                  [(set GPR:$dst, (opnode GPR:$LHS,
-                                          (rotr GPR:$RHS, rot_imm:$rot)))]>,
+                  [(set rGPR:$dst, (opnode rGPR:$LHS,
+                                          (rotr rGPR:$RHS, rot_imm:$rot)))]>,
                   Requires<[HasT2ExtractPack]> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -747,7 +747,7 @@
 // DO variant - disassembly only, no pattern
 
 multiclass T2I_bin_rrot_DO<bits<3> opcod, string opc> {
-  def rr     : T2I<(outs GPR:$dst), (ins GPR:$LHS, GPR:$RHS), IIC_iALUr,
+  def rr     : T2I<(outs rGPR:$dst), (ins rGPR:$LHS, rGPR:$RHS), IIC_iALUr,
                   opc, "\t$dst, $LHS, $RHS", []> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -756,7 +756,7 @@
      let Inst{7} = 1;
      let Inst{5-4} = 0b00; // rotate
    }
-  def rr_rot : T2I<(outs GPR:$dst), (ins GPR:$LHS, GPR:$RHS, i32imm:$rot),
+  def rr_rot : T2I<(outs rGPR:$dst), (ins rGPR:$LHS, rGPR:$RHS, i32imm:$rot),
                   IIC_iALUsr, opc, "\t$dst, $LHS, $RHS, ror $rot", []> {
      let Inst{31-27} = 0b11111;
      let Inst{26-23} = 0b0100;
@@ -779,7 +779,7 @@
 // assembler.
 let neverHasSideEffects = 1 in {
 let isReMaterializable = 1 in
-def t2LEApcrel : T2XI<(outs GPR:$dst), (ins i32imm:$label, pred:$p), IIC_iALUi,
+def t2LEApcrel : T2XI<(outs rGPR:$dst), (ins i32imm:$label, pred:$p), IIC_iALUi,
                       "adr$p.w\t$dst, #$label", []> {
   let Inst{31-27} = 0b11110;
   let Inst{25-24} = 0b10;
@@ -790,7 +790,7 @@
   let Inst{15} = 0;
 }
 } // neverHasSideEffects
-def t2LEApcrelJT : T2XI<(outs GPR:$dst),
+def t2LEApcrelJT : T2XI<(outs rGPR:$dst),
                         (ins i32imm:$label, nohash_imm:$id, pred:$p), IIC_iALUi,
                         "adr$p.w\t$dst, #${label}_${id}", []> {
   let Inst{31-27} = 0b11110;
@@ -866,9 +866,9 @@
 }
 
 // Signed and unsigned division on v7-M
-def t2SDIV : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iALUi, 
+def t2SDIV : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iALUi, 
                  "sdiv", "\t$dst, $a, $b",
-                 [(set GPR:$dst, (sdiv GPR:$a, GPR:$b))]>,
+                 [(set rGPR:$dst, (sdiv rGPR:$a, rGPR:$b))]>,
                  Requires<[HasDivide]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-21} = 0b011100;
@@ -877,9 +877,9 @@
   let Inst{7-4} = 0b1111;
 }
 
-def t2UDIV : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iALUi, 
+def t2UDIV : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iALUi, 
                  "udiv", "\t$dst, $a, $b",
-                 [(set GPR:$dst, (udiv GPR:$a, GPR:$b))]>,
+                 [(set rGPR:$dst, (udiv rGPR:$a, rGPR:$b))]>,
                  Requires<[HasDivide]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-21} = 0b011101;
@@ -889,6 +889,9 @@
 }
 
 // Pseudo instruction that will expand into a t2SUBrSPi + a copy.
+// FIXME: Now that we have rGPR, do we need these pseudos? It seems
+//        that the coalescer will now properly know how to do the right
+//        thing without them.
 let usesCustomInserter = 1 in { // Expanded after instruction selection.
 def t2SUBrSPi_   : PseudoInst<(outs GPR:$dst), (ins GPR:$sp, t2_so_imm:$imm),
                    NoItinerary, "${:comment} sub.w\t$dst, $sp, $imm", []>;
@@ -917,10 +920,10 @@
 
 let mayLoad = 1, neverHasSideEffects = 1, hasExtraDefRegAllocReq = 1 in {
 // Load doubleword
-def t2LDRDi8  : T2Ii8s4<1, 0, 1, (outs GPR:$dst1, GPR:$dst2),
+def t2LDRDi8  : T2Ii8s4<1, 0, 1, (outs rGPR:$dst1, rGPR:$dst2),
                         (ins t2addrmode_imm8s4:$addr),
                         IIC_iLoadi, "ldrd", "\t$dst1, $addr", []>;
-def t2LDRDpci : T2Ii8s4<1, 0, 1, (outs GPR:$dst1, GPR:$dst2),
+def t2LDRDpci : T2Ii8s4<1, 0, 1, (outs rGPR:$dst1, rGPR:$dst2),
                         (ins i32imm:$addr), IIC_iLoadi,
                        "ldrd", "\t$dst1, $addr", []> {
   let Inst{19-16} = 0b1111; // Rn
@@ -967,6 +970,11 @@
 def : T2Pat<(extloadi16 (ARMWrapper tconstpool:$addr)),
             (t2LDRHpci  tconstpool:$addr)>;
 
+// FIXME: The destination register of the loads and stores can't be PC, but
+//        can be SP. We need another regclass (similar to rGPR) to represent
+//        that. Not a pressing issue since these are selected manually,
+//        not via pattern.
+
 // Indexed loads
 let mayLoad = 1, neverHasSideEffects = 1 in {
 def t2LDR_PRE  : T2Iidxldst<0, 0b10, 1, 1, (outs GPR:$dst, GPR:$base_wb),
@@ -1286,9 +1294,9 @@
 
 // AddedComplexity to ensure isel tries t2MOVi before t2MOVi16.
 let isReMaterializable = 1, isAsCheapAsAMove = 1, AddedComplexity = 1 in
-def t2MOVi : T2sI<(outs GPR:$dst), (ins t2_so_imm:$src), IIC_iMOVi,
+def t2MOVi : T2sI<(outs rGPR:$dst), (ins t2_so_imm:$src), IIC_iMOVi,
                    "mov", ".w\t$dst, $src",
-                   [(set GPR:$dst, t2_so_imm:$src)]> {
+                   [(set rGPR:$dst, t2_so_imm:$src)]> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 0;
   let Inst{24-21} = 0b0010;
@@ -1298,9 +1306,9 @@
 }
 
 let isReMaterializable = 1, isAsCheapAsAMove = 1 in
-def t2MOVi16 : T2I<(outs GPR:$dst), (ins i32imm:$src), IIC_iMOVi,
+def t2MOVi16 : T2I<(outs rGPR:$dst), (ins i32imm:$src), IIC_iMOVi,
                    "movw", "\t$dst, $src",
-                   [(set GPR:$dst, imm0_65535:$src)]> {
+                   [(set rGPR:$dst, imm0_65535:$src)]> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
   let Inst{24-21} = 0b0010;
@@ -1309,10 +1317,10 @@
 }
 
 let Constraints = "$src = $dst" in
-def t2MOVTi16 : T2I<(outs GPR:$dst), (ins GPR:$src, i32imm:$imm), IIC_iMOVi,
+def t2MOVTi16 : T2I<(outs rGPR:$dst), (ins rGPR:$src, i32imm:$imm), IIC_iMOVi,
                     "movt", "\t$dst, $imm",
-                    [(set GPR:$dst,
-                          (or (and GPR:$src, 0xffff), lo16AllZero:$imm))]> {
+                    [(set rGPR:$dst,
+                          (or (and rGPR:$src, 0xffff), lo16AllZero:$imm))]> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
   let Inst{24-21} = 0b0110;
@@ -1320,7 +1328,7 @@
   let Inst{15} = 0;
 }
 
-def : T2Pat<(or GPR:$src, 0xffff0000), (t2MOVTi16 GPR:$src, 0xffff)>;
+def : T2Pat<(or rGPR:$src, 0xffff0000), (t2MOVTi16 rGPR:$src, 0xffff)>;
 
 //===----------------------------------------------------------------------===//
 //  Extend Instructions.
@@ -1356,10 +1364,10 @@
 //        The transformation should probably be done as a combiner action
 //        instead so we can include a check for masking back in the upper
 //        eight bits of the source into the lower eight bits of the result.
-//def : T2Pat<(and (shl GPR:$Src, (i32 8)), 0xFF00FF),
-//            (t2UXTB16r_rot GPR:$Src, 24)>, Requires<[HasT2ExtractPack]>;
-def : T2Pat<(and (srl GPR:$Src, (i32 8)), 0xFF00FF),
-            (t2UXTB16r_rot GPR:$Src, 8)>, Requires<[HasT2ExtractPack]>;
+//def : T2Pat<(and (shl rGPR:$Src, (i32 8)), 0xFF00FF),
+//            (t2UXTB16r_rot rGPR:$Src, 24)>, Requires<[HasT2ExtractPack]>;
+def : T2Pat<(and (srl rGPR:$Src, (i32 8)), 0xFF00FF),
+            (t2UXTB16r_rot rGPR:$Src, 8)>, Requires<[HasT2ExtractPack]>;
 
 defm t2UXTAB : T2I_bin_rrot<0b101, "uxtab",
                            BinOpFrag<(add node:$LHS, (and node:$RHS, 0x00FF))>>;
@@ -1413,18 +1421,18 @@
 def : T2Pat<(add        GPR:$src, imm0_4095_neg:$imm),
             (t2SUBri12  GPR:$src, imm0_4095_neg:$imm)>;
 let AddedComplexity = 1 in
-def : T2Pat<(addc       GPR:$src, imm0_255_neg:$imm),
-            (t2SUBSri   GPR:$src, imm0_255_neg:$imm)>;
-def : T2Pat<(addc       GPR:$src, t2_so_imm_neg:$imm),
-            (t2SUBSri   GPR:$src, t2_so_imm_neg:$imm)>;
+def : T2Pat<(addc       rGPR:$src, imm0_255_neg:$imm),
+            (t2SUBSri   rGPR:$src, imm0_255_neg:$imm)>;
+def : T2Pat<(addc       rGPR:$src, t2_so_imm_neg:$imm),
+            (t2SUBSri   rGPR:$src, t2_so_imm_neg:$imm)>;
 // The with-carry-in form matches bitwise not instead of the negation.
 // Effectively, the inverse interpretation of the carry flag already accounts
 // for part of the negation.
 let AddedComplexity = 1 in
-def : T2Pat<(adde       GPR:$src, imm0_255_not:$imm),
-            (t2SBCSri   GPR:$src, imm0_255_not:$imm)>;
-def : T2Pat<(adde       GPR:$src, t2_so_imm_not:$imm),
-            (t2SBCSri   GPR:$src, t2_so_imm_not:$imm)>;
+def : T2Pat<(adde       rGPR:$src, imm0_255_not:$imm),
+            (t2SBCSri   rGPR:$src, imm0_255_not:$imm)>;
+def : T2Pat<(adde       rGPR:$src, t2_so_imm_not:$imm),
+            (t2SBCSri   rGPR:$src, t2_so_imm_not:$imm)>;
 
 // Select Bytes -- for disassembly only
 
@@ -1443,7 +1451,7 @@
 // And Miscellaneous operations -- for disassembly only
 class T2I_pam<bits<3> op22_20, bits<4> op7_4, string opc,
               list<dag> pat = [/* For disassembly only; pattern left blank */]>
-  : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), NoItinerary, opc,
+  : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), NoItinerary, opc,
         "\t$dst, $a, $b", pat> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0101;
@@ -1455,7 +1463,7 @@
 // Saturating add/subtract -- for disassembly only
 
 def t2QADD    : T2I_pam<0b000, 0b1000, "qadd",
-                        [(set GPR:$dst, (int_arm_qadd GPR:$a, GPR:$b))]>;
+                        [(set rGPR:$dst, (int_arm_qadd rGPR:$a, rGPR:$b))]>;
 def t2QADD16  : T2I_pam<0b001, 0b0001, "qadd16">;
 def t2QADD8   : T2I_pam<0b000, 0b0001, "qadd8">;
 def t2QASX    : T2I_pam<0b010, 0b0001, "qasx">;
@@ -1463,7 +1471,7 @@
 def t2QDSUB   : T2I_pam<0b000, 0b1011, "qdsub">;
 def t2QSAX    : T2I_pam<0b110, 0b0001, "qsax">;
 def t2QSUB    : T2I_pam<0b000, 0b1010, "qsub",
-                        [(set GPR:$dst, (int_arm_qsub GPR:$a, GPR:$b))]>;
+                        [(set rGPR:$dst, (int_arm_qsub rGPR:$a, rGPR:$b))]>;
 def t2QSUB16  : T2I_pam<0b101, 0b0001, "qsub16">;
 def t2QSUB8   : T2I_pam<0b100, 0b0001, "qsub8">;
 def t2UQADD16 : T2I_pam<0b001, 0b0101, "uqadd16">;
@@ -1505,17 +1513,18 @@
 
 // Unsigned Sum of Absolute Differences [and Accumulate] -- for disassembly only
 
-def t2USAD8   : T2I_mac<0, 0b111, 0b0000, (outs GPR:$dst), (ins GPR:$a, GPR:$b),
+def t2USAD8   : T2I_mac<0, 0b111, 0b0000, (outs rGPR:$dst),
+                                           (ins rGPR:$a, rGPR:$b),
                         NoItinerary, "usad8", "\t$dst, $a, $b", []> {
   let Inst{15-12} = 0b1111;
 }
-def t2USADA8  : T2I_mac<0, 0b111, 0b0000, (outs GPR:$dst),
-                        (ins GPR:$a, GPR:$b, GPR:$acc), NoItinerary, "usada8",
+def t2USADA8  : T2I_mac<0, 0b111, 0b0000, (outs rGPR:$dst),
+                       (ins rGPR:$a, rGPR:$b, rGPR:$acc), NoItinerary, "usada8",
                         "\t$dst, $a, $b, $acc", []>;
 
 // Signed/Unsigned saturate -- for disassembly only
 
-def t2SSATlsl : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos,GPR:$a,i32imm:$shamt),
+def t2SSATlsl:T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos,rGPR:$a,i32imm:$shamt),
                     NoItinerary, "ssat", "\t$dst, $bit_pos, $a, lsl $shamt",
                     [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1525,7 +1534,7 @@
   let Inst{21} = 0;        // sh = '0'
 }
 
-def t2SSATasr : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos,GPR:$a,i32imm:$shamt),
+def t2SSATasr:T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos,rGPR:$a,i32imm:$shamt),
                     NoItinerary, "ssat", "\t$dst, $bit_pos, $a, asr $shamt",
                     [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1535,7 +1544,7 @@
   let Inst{21} = 1;        // sh = '1'
 }
 
-def t2SSAT16 : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos, GPR:$a), NoItinerary,
+def t2SSAT16: T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos, rGPR:$a), NoItinerary,
                    "ssat16", "\t$dst, $bit_pos, $a",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1547,7 +1556,7 @@
   let Inst{7-6} = 0b00;    // imm2 = '00'
 }
 
-def t2USATlsl : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos,GPR:$a,i32imm:$shamt),
+def t2USATlsl:T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos,rGPR:$a,i32imm:$shamt),
                      NoItinerary, "usat", "\t$dst, $bit_pos, $a, lsl $shamt",
                      [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1557,7 +1566,7 @@
   let Inst{21} = 0;        // sh = '0'
 }
 
-def t2USATasr : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos,GPR:$a,i32imm:$shamt),
+def t2USATasr:T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos,rGPR:$a,i32imm:$shamt),
                      NoItinerary, "usat", "\t$dst, $bit_pos, $a, asr $shamt",
                      [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1567,7 +1576,7 @@
   let Inst{21} = 1;        // sh = '1'
 }
 
-def t2USAT16 : T2I<(outs GPR:$dst), (ins i32imm:$bit_pos, GPR:$a), NoItinerary,
+def t2USAT16: T2I<(outs rGPR:$dst), (ins i32imm:$bit_pos, rGPR:$a), NoItinerary,
                    "usat16", "\t$dst, $bit_pos, $a",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -1592,9 +1601,9 @@
 defm t2ROR  : T2I_sh_ir<0b11, "ror", BinOpFrag<(rotr node:$LHS, node:$RHS)>>;
 
 let Uses = [CPSR] in {
-def t2MOVrx : T2sI<(outs GPR:$dst), (ins GPR:$src), IIC_iMOVsi,
+def t2MOVrx : T2sI<(outs rGPR:$dst), (ins rGPR:$src), IIC_iMOVsi,
                    "rrx", "\t$dst, $src",
-                   [(set GPR:$dst, (ARMrrx GPR:$src))]> {
+                   [(set rGPR:$dst, (ARMrrx rGPR:$src))]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-25} = 0b01;
   let Inst{24-21} = 0b0010;
@@ -1606,9 +1615,9 @@
 }
 
 let Defs = [CPSR] in {
-def t2MOVsrl_flag : T2I<(outs GPR:$dst), (ins GPR:$src), IIC_iMOVsi,
+def t2MOVsrl_flag : T2I<(outs rGPR:$dst), (ins rGPR:$src), IIC_iMOVsi,
                         "lsrs", ".w\t$dst, $src, #1",
-                        [(set GPR:$dst, (ARMsrl_flag GPR:$src))]> {
+                        [(set rGPR:$dst, (ARMsrl_flag rGPR:$src))]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-25} = 0b01;
   let Inst{24-21} = 0b0010;
@@ -1619,9 +1628,9 @@
   let Inst{14-12} = 0b000;
   let Inst{7-6} = 0b01;
 }
-def t2MOVsra_flag : T2I<(outs GPR:$dst), (ins GPR:$src), IIC_iMOVsi,
+def t2MOVsra_flag : T2I<(outs rGPR:$dst), (ins rGPR:$src), IIC_iMOVsi,
                         "asrs", ".w\t$dst, $src, #1",
-                        [(set GPR:$dst, (ARMsra_flag GPR:$src))]> {
+                        [(set rGPR:$dst, (ARMsra_flag rGPR:$src))]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-25} = 0b01;
   let Inst{24-21} = 0b0010;
@@ -1649,9 +1658,9 @@
                             BinOpFrag<(and node:$LHS, (not node:$RHS))>>;
 
 let Constraints = "$src = $dst" in
-def t2BFC : T2I<(outs GPR:$dst), (ins GPR:$src, bf_inv_mask_imm:$imm),
+def t2BFC : T2I<(outs rGPR:$dst), (ins rGPR:$src, bf_inv_mask_imm:$imm),
                 IIC_iUNAsi, "bfc", "\t$dst, $imm",
-                [(set GPR:$dst, (and GPR:$src, bf_inv_mask_imm:$imm))]> {
+                [(set rGPR:$dst, (and rGPR:$src, bf_inv_mask_imm:$imm))]> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
   let Inst{24-20} = 0b10110;
@@ -1659,7 +1668,7 @@
   let Inst{15} = 0;
 }
 
-def t2SBFX : T2I<(outs GPR:$dst), (ins GPR:$src, imm0_31:$lsb, imm0_31:$width),
+def t2SBFX: T2I<(outs rGPR:$dst), (ins rGPR:$src, imm0_31:$lsb, imm0_31:$width),
                  IIC_iALUi, "sbfx", "\t$dst, $src, $lsb, $width", []> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
@@ -1667,7 +1676,7 @@
   let Inst{15} = 0;
 }
 
-def t2UBFX : T2I<(outs GPR:$dst), (ins GPR:$src, imm0_31:$lsb, imm0_31:$width),
+def t2UBFX: T2I<(outs rGPR:$dst), (ins rGPR:$src, imm0_31:$lsb, imm0_31:$width),
                  IIC_iALUi, "ubfx", "\t$dst, $src, $lsb, $width", []> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
@@ -1677,10 +1686,10 @@
 
 // A8.6.18  BFI - Bitfield insert (Encoding T1)
 let Constraints = "$src = $dst" in
-def t2BFI : T2I<(outs GPR:$dst),
-                (ins GPR:$src, GPR:$val, bf_inv_mask_imm:$imm),
+def t2BFI : T2I<(outs rGPR:$dst),
+                (ins rGPR:$src, rGPR:$val, bf_inv_mask_imm:$imm),
                 IIC_iALUi, "bfi", "\t$dst, $val, $imm",
-                [(set GPR:$dst, (ARMbfi GPR:$src, GPR:$val,
+                [(set rGPR:$dst, (ARMbfi rGPR:$src, rGPR:$val,
                                  bf_inv_mask_imm:$imm))]> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 1;
@@ -1697,12 +1706,12 @@
 
 
 let AddedComplexity = 1 in
-def : T2Pat<(and     GPR:$src, t2_so_imm_not:$imm),
-            (t2BICri GPR:$src, t2_so_imm_not:$imm)>;
+def : T2Pat<(and     rGPR:$src, t2_so_imm_not:$imm),
+            (t2BICri rGPR:$src, t2_so_imm_not:$imm)>;
 
 // FIXME: Disable this pattern on Darwin to workaround an assembler bug.
-def : T2Pat<(or      GPR:$src, t2_so_imm_not:$imm),
-            (t2ORNri GPR:$src, t2_so_imm_not:$imm)>,
+def : T2Pat<(or      rGPR:$src, t2_so_imm_not:$imm),
+            (t2ORNri rGPR:$src, t2_so_imm_not:$imm)>,
             Requires<[IsThumb2]>;
 
 def : T2Pat<(t2_so_imm_not:$src),
@@ -1712,9 +1721,9 @@
 //  Multiply Instructions.
 //
 let isCommutable = 1 in
-def t2MUL: T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+def t2MUL: T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
                 "mul", "\t$dst, $a, $b",
-                [(set GPR:$dst, (mul GPR:$a, GPR:$b))]> {
+                [(set rGPR:$dst, (mul rGPR:$a, rGPR:$b))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b000;
@@ -1722,9 +1731,9 @@
   let Inst{7-4} = 0b0000; // Multiply
 }
 
-def t2MLA: T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2MLA: T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
 		"mla", "\t$dst, $a, $b, $c",
-		[(set GPR:$dst, (add (mul GPR:$a, GPR:$b), GPR:$c))]> {
+		[(set rGPR:$dst, (add (mul rGPR:$a, rGPR:$b), rGPR:$c))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b000;
@@ -1732,9 +1741,9 @@
   let Inst{7-4} = 0b0000; // Multiply
 }
 
-def t2MLS: T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2MLS: T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
 		"mls", "\t$dst, $a, $b, $c",
-                [(set GPR:$dst, (sub GPR:$c, (mul GPR:$a, GPR:$b)))]> {
+                [(set rGPR:$dst, (sub rGPR:$c, (mul rGPR:$a, rGPR:$b)))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b000;
@@ -1745,7 +1754,7 @@
 // Extra precision multiplies with low / high results
 let neverHasSideEffects = 1 in {
 let isCommutable = 1 in {
-def t2SMULL : T2I<(outs GPR:$ldst, GPR:$hdst), (ins GPR:$a, GPR:$b), IIC_iMUL64,
+def t2SMULL : T2I<(outs rGPR:$ldst, rGPR:$hdst), (ins rGPR:$a, rGPR:$b), IIC_iMUL64,
                    "smull", "\t$ldst, $hdst, $a, $b", []> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0111;
@@ -1753,7 +1762,7 @@
   let Inst{7-4} = 0b0000;
 }
 
-def t2UMULL : T2I<(outs GPR:$ldst, GPR:$hdst), (ins GPR:$a, GPR:$b), IIC_iMUL64,
+def t2UMULL : T2I<(outs rGPR:$ldst, rGPR:$hdst), (ins rGPR:$a, rGPR:$b), IIC_iMUL64,
                    "umull", "\t$ldst, $hdst, $a, $b", []> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0111;
@@ -1763,7 +1772,7 @@
 } // isCommutable
 
 // Multiply + accumulate
-def t2SMLAL : T2I<(outs GPR:$ldst, GPR:$hdst), (ins GPR:$a, GPR:$b), IIC_iMAC64,
+def t2SMLAL : T2I<(outs rGPR:$ldst, rGPR:$hdst), (ins rGPR:$a, rGPR:$b), IIC_iMAC64,
                   "smlal", "\t$ldst, $hdst, $a, $b", []>{
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0111;
@@ -1771,7 +1780,7 @@
   let Inst{7-4} = 0b0000;
 }
 
-def t2UMLAL : T2I<(outs GPR:$ldst, GPR:$hdst), (ins GPR:$a, GPR:$b), IIC_iMAC64,
+def t2UMLAL : T2I<(outs rGPR:$ldst, rGPR:$hdst), (ins rGPR:$a, rGPR:$b), IIC_iMAC64,
                   "umlal", "\t$ldst, $hdst, $a, $b", []>{
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0111;
@@ -1779,7 +1788,7 @@
   let Inst{7-4} = 0b0000;
 }
 
-def t2UMAAL : T2I<(outs GPR:$ldst, GPR:$hdst), (ins GPR:$a, GPR:$b), IIC_iMAC64,
+def t2UMAAL : T2I<(outs rGPR:$ldst, rGPR:$hdst), (ins rGPR:$a, rGPR:$b), IIC_iMAC64,
                   "umaal", "\t$ldst, $hdst, $a, $b", []>{
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0111;
@@ -1791,9 +1800,9 @@
 // Rounding variants of the below included for disassembly only
 
 // Most significant word multiply
-def t2SMMUL : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+def t2SMMUL : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
                   "smmul", "\t$dst, $a, $b",
-                  [(set GPR:$dst, (mulhs GPR:$a, GPR:$b))]> {
+                  [(set rGPR:$dst, (mulhs rGPR:$a, rGPR:$b))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b101;
@@ -1801,7 +1810,7 @@
   let Inst{7-4} = 0b0000; // No Rounding (Inst{4} = 0)
 }
 
-def t2SMMULR : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+def t2SMMULR : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
                   "smmulr", "\t$dst, $a, $b", []> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
@@ -1810,9 +1819,9 @@
   let Inst{7-4} = 0b0001; // Rounding (Inst{4} = 1)
 }
 
-def t2SMMLA : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2SMMLA : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
                   "smmla", "\t$dst, $a, $b, $c",
-                  [(set GPR:$dst, (add (mulhs GPR:$a, GPR:$b), GPR:$c))]> {
+                  [(set rGPR:$dst, (add (mulhs rGPR:$a, rGPR:$b), rGPR:$c))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b101;
@@ -1820,7 +1829,7 @@
   let Inst{7-4} = 0b0000; // No Rounding (Inst{4} = 0)
 }
 
-def t2SMMLAR : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2SMMLAR : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
                   "smmlar", "\t$dst, $a, $b, $c", []> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
@@ -1829,9 +1838,9 @@
   let Inst{7-4} = 0b0001; // Rounding (Inst{4} = 1)
 }
 
-def t2SMMLS : T2I <(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2SMMLS : T2I <(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
                    "smmls", "\t$dst, $a, $b, $c",
-                   [(set GPR:$dst, (sub GPR:$c, (mulhs GPR:$a, GPR:$b)))]> {
+                   [(set rGPR:$dst, (sub rGPR:$c, (mulhs rGPR:$a, rGPR:$b)))]> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
   let Inst{22-20} = 0b110;
@@ -1839,7 +1848,7 @@
   let Inst{7-4} = 0b0000; // No Rounding (Inst{4} = 0)
 }
 
-def t2SMMLSR : T2I <(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$c), IIC_iMAC32,
+def t2SMMLSR : T2I <(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$c), IIC_iMAC32,
                    "smmlsr", "\t$dst, $a, $b, $c", []> {
   let Inst{31-27} = 0b11111;
   let Inst{26-23} = 0b0110;
@@ -1849,10 +1858,10 @@
 }
 
 multiclass T2I_smul<string opc, PatFrag opnode> {
-  def BB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+  def BB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
               !strconcat(opc, "bb"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (opnode (sext_inreg GPR:$a, i16),
-                                      (sext_inreg GPR:$b, i16)))]> {
+              [(set rGPR:$dst, (opnode (sext_inreg rGPR:$a, i16),
+                                      (sext_inreg rGPR:$b, i16)))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1861,10 +1870,10 @@
     let Inst{5-4} = 0b00;
   }
 
-  def BT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+  def BT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
               !strconcat(opc, "bt"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (opnode (sext_inreg GPR:$a, i16),
-                                      (sra GPR:$b, (i32 16))))]> {
+              [(set rGPR:$dst, (opnode (sext_inreg rGPR:$a, i16),
+                                      (sra rGPR:$b, (i32 16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1873,10 +1882,10 @@
     let Inst{5-4} = 0b01;
   }
 
-  def TB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+  def TB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
               !strconcat(opc, "tb"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (opnode (sra GPR:$a, (i32 16)),
-                                      (sext_inreg GPR:$b, i16)))]> {
+              [(set rGPR:$dst, (opnode (sra rGPR:$a, (i32 16)),
+                                      (sext_inreg rGPR:$b, i16)))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1885,10 +1894,10 @@
     let Inst{5-4} = 0b10;
   }
 
-  def TT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL32,
+  def TT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL32,
               !strconcat(opc, "tt"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (opnode (sra GPR:$a, (i32 16)),
-                                      (sra GPR:$b, (i32 16))))]> {
+              [(set rGPR:$dst, (opnode (sra rGPR:$a, (i32 16)),
+                                      (sra rGPR:$b, (i32 16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1897,10 +1906,10 @@
     let Inst{5-4} = 0b11;
   }
 
-  def WB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL16,
+  def WB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL16,
               !strconcat(opc, "wb"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (sra (opnode GPR:$a,
-                                    (sext_inreg GPR:$b, i16)), (i32 16)))]> {
+              [(set rGPR:$dst, (sra (opnode rGPR:$a,
+                                    (sext_inreg rGPR:$b, i16)), (i32 16)))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b011;
@@ -1909,10 +1918,10 @@
     let Inst{5-4} = 0b00;
   }
 
-  def WT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b), IIC_iMUL16,
+  def WT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b), IIC_iMUL16,
               !strconcat(opc, "wt"), "\t$dst, $a, $b",
-              [(set GPR:$dst, (sra (opnode GPR:$a,
-                                    (sra GPR:$b, (i32 16))), (i32 16)))]> {
+              [(set rGPR:$dst, (sra (opnode rGPR:$a,
+                                    (sra rGPR:$b, (i32 16))), (i32 16)))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b011;
@@ -1924,11 +1933,11 @@
 
 
 multiclass T2I_smla<string opc, PatFrag opnode> {
-  def BB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def BB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
               !strconcat(opc, "bb"), "\t$dst, $a, $b, $acc",
-              [(set GPR:$dst, (add GPR:$acc,
-                               (opnode (sext_inreg GPR:$a, i16),
-                                       (sext_inreg GPR:$b, i16))))]> {
+              [(set rGPR:$dst, (add rGPR:$acc,
+                               (opnode (sext_inreg rGPR:$a, i16),
+                                       (sext_inreg rGPR:$b, i16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1937,10 +1946,10 @@
     let Inst{5-4} = 0b00;
   }
 
-  def BT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def BT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
              !strconcat(opc, "bt"), "\t$dst, $a, $b, $acc",
-             [(set GPR:$dst, (add GPR:$acc, (opnode (sext_inreg GPR:$a, i16),
-                                                   (sra GPR:$b, (i32 16)))))]> {
+             [(set rGPR:$dst, (add rGPR:$acc, (opnode (sext_inreg rGPR:$a, i16),
+                                                   (sra rGPR:$b, (i32 16)))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1949,10 +1958,10 @@
     let Inst{5-4} = 0b01;
   }
 
-  def TB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def TB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
               !strconcat(opc, "tb"), "\t$dst, $a, $b, $acc",
-              [(set GPR:$dst, (add GPR:$acc, (opnode (sra GPR:$a, (i32 16)),
-                                                 (sext_inreg GPR:$b, i16))))]> {
+              [(set rGPR:$dst, (add rGPR:$acc, (opnode (sra rGPR:$a, (i32 16)),
+                                                 (sext_inreg rGPR:$b, i16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1961,10 +1970,10 @@
     let Inst{5-4} = 0b10;
   }
 
-  def TT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def TT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
               !strconcat(opc, "tt"), "\t$dst, $a, $b, $acc",
-             [(set GPR:$dst, (add GPR:$acc, (opnode (sra GPR:$a, (i32 16)),
-                                                   (sra GPR:$b, (i32 16)))))]> {
+             [(set rGPR:$dst, (add rGPR:$acc, (opnode (sra rGPR:$a, (i32 16)),
+                                                   (sra rGPR:$b, (i32 16)))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b001;
@@ -1973,10 +1982,10 @@
     let Inst{5-4} = 0b11;
   }
 
-  def WB : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def WB : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
               !strconcat(opc, "wb"), "\t$dst, $a, $b, $acc",
-              [(set GPR:$dst, (add GPR:$acc, (sra (opnode GPR:$a,
-                                      (sext_inreg GPR:$b, i16)), (i32 16))))]> {
+              [(set rGPR:$dst, (add rGPR:$acc, (sra (opnode rGPR:$a,
+                                      (sext_inreg rGPR:$b, i16)), (i32 16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b011;
@@ -1985,10 +1994,10 @@
     let Inst{5-4} = 0b00;
   }
 
-  def WT : T2I<(outs GPR:$dst), (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC16,
+  def WT : T2I<(outs rGPR:$dst), (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC16,
               !strconcat(opc, "wt"), "\t$dst, $a, $b, $acc",
-              [(set GPR:$dst, (add GPR:$acc, (sra (opnode GPR:$a,
-                                        (sra GPR:$b, (i32 16))), (i32 16))))]> {
+              [(set rGPR:$dst, (add rGPR:$acc, (sra (opnode rGPR:$a,
+                                        (sra rGPR:$b, (i32 16))), (i32 16))))]> {
     let Inst{31-27} = 0b11111;
     let Inst{26-23} = 0b0110;
     let Inst{22-20} = 0b011;
@@ -2002,61 +2011,61 @@
 defm t2SMLA : T2I_smla<"smla", BinOpFrag<(mul node:$LHS, node:$RHS)>>;
 
 // Halfword multiple accumulate long: SMLAL<x><y> -- for disassembly only
-def t2SMLALBB : T2I_mac<1, 0b100, 0b1000, (outs GPR:$ldst,GPR:$hdst),
-           (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlalbb", "\t$ldst, $hdst, $a, $b",
+def t2SMLALBB : T2I_mac<1, 0b100, 0b1000, (outs rGPR:$ldst,rGPR:$hdst),
+           (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlalbb", "\t$ldst, $hdst, $a, $b",
            [/* For disassembly only; pattern left blank */]>;
-def t2SMLALBT : T2I_mac<1, 0b100, 0b1001, (outs GPR:$ldst,GPR:$hdst),
-           (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlalbt", "\t$ldst, $hdst, $a, $b",
+def t2SMLALBT : T2I_mac<1, 0b100, 0b1001, (outs rGPR:$ldst,rGPR:$hdst),
+           (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlalbt", "\t$ldst, $hdst, $a, $b",
            [/* For disassembly only; pattern left blank */]>;
-def t2SMLALTB : T2I_mac<1, 0b100, 0b1010, (outs GPR:$ldst,GPR:$hdst),
-           (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlaltb", "\t$ldst, $hdst, $a, $b",
+def t2SMLALTB : T2I_mac<1, 0b100, 0b1010, (outs rGPR:$ldst,rGPR:$hdst),
+           (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlaltb", "\t$ldst, $hdst, $a, $b",
            [/* For disassembly only; pattern left blank */]>;
-def t2SMLALTT : T2I_mac<1, 0b100, 0b1011, (outs GPR:$ldst,GPR:$hdst),
-           (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlaltt", "\t$ldst, $hdst, $a, $b",
+def t2SMLALTT : T2I_mac<1, 0b100, 0b1011, (outs rGPR:$ldst,rGPR:$hdst),
+           (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlaltt", "\t$ldst, $hdst, $a, $b",
            [/* For disassembly only; pattern left blank */]>;
 
 // Dual halfword multiple: SMUAD, SMUSD, SMLAD, SMLSD, SMLALD, SMLSLD
 // These are for disassembly only.
 
-def t2SMUAD   : T2I_mac<0, 0b010, 0b0000, (outs GPR:$dst), (ins GPR:$a, GPR:$b),
+def t2SMUAD   : T2I_mac<0, 0b010, 0b0000, (outs rGPR:$dst), (ins rGPR:$a, rGPR:$b),
                         IIC_iMAC32, "smuad", "\t$dst, $a, $b", []> {
   let Inst{15-12} = 0b1111;
 }
-def t2SMUADX  : T2I_mac<0, 0b010, 0b0001, (outs GPR:$dst), (ins GPR:$a, GPR:$b),
+def t2SMUADX  : T2I_mac<0, 0b010, 0b0001, (outs rGPR:$dst), (ins rGPR:$a, rGPR:$b),
                         IIC_iMAC32, "smuadx", "\t$dst, $a, $b", []> {
   let Inst{15-12} = 0b1111;
 }
-def t2SMUSD   : T2I_mac<0, 0b100, 0b0000, (outs GPR:$dst), (ins GPR:$a, GPR:$b),
+def t2SMUSD   : T2I_mac<0, 0b100, 0b0000, (outs rGPR:$dst), (ins rGPR:$a, rGPR:$b),
                         IIC_iMAC32, "smusd", "\t$dst, $a, $b", []> {
   let Inst{15-12} = 0b1111;
 }
-def t2SMUSDX  : T2I_mac<0, 0b100, 0b0001, (outs GPR:$dst), (ins GPR:$a, GPR:$b),
+def t2SMUSDX  : T2I_mac<0, 0b100, 0b0001, (outs rGPR:$dst), (ins rGPR:$a, rGPR:$b),
                         IIC_iMAC32, "smusdx", "\t$dst, $a, $b", []> {
   let Inst{15-12} = 0b1111;
 }
-def t2SMLAD   : T2I_mac<0, 0b010, 0b0000, (outs GPR:$dst),
-                        (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC32, "smlad",
+def t2SMLAD   : T2I_mac<0, 0b010, 0b0000, (outs rGPR:$dst),
+                        (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC32, "smlad",
                         "\t$dst, $a, $b, $acc", []>;
-def t2SMLADX  : T2I_mac<0, 0b010, 0b0001, (outs GPR:$dst),
-                        (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC32, "smladx",
+def t2SMLADX  : T2I_mac<0, 0b010, 0b0001, (outs rGPR:$dst),
+                        (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC32, "smladx",
                         "\t$dst, $a, $b, $acc", []>;
-def t2SMLSD   : T2I_mac<0, 0b100, 0b0000, (outs GPR:$dst),
-                        (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC32, "smlsd",
+def t2SMLSD   : T2I_mac<0, 0b100, 0b0000, (outs rGPR:$dst),
+                        (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC32, "smlsd",
                         "\t$dst, $a, $b, $acc", []>;
-def t2SMLSDX  : T2I_mac<0, 0b100, 0b0001, (outs GPR:$dst),
-                        (ins GPR:$a, GPR:$b, GPR:$acc), IIC_iMAC32, "smlsdx",
+def t2SMLSDX  : T2I_mac<0, 0b100, 0b0001, (outs rGPR:$dst),
+                        (ins rGPR:$a, rGPR:$b, rGPR:$acc), IIC_iMAC32, "smlsdx",
                         "\t$dst, $a, $b, $acc", []>;
-def t2SMLALD  : T2I_mac<1, 0b100, 0b1100, (outs GPR:$ldst,GPR:$hdst),
-                        (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlald",
+def t2SMLALD  : T2I_mac<1, 0b100, 0b1100, (outs rGPR:$ldst,rGPR:$hdst),
+                        (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlald",
                         "\t$ldst, $hdst, $a, $b", []>;
-def t2SMLALDX : T2I_mac<1, 0b100, 0b1101, (outs GPR:$ldst,GPR:$hdst),
-                        (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlaldx",
+def t2SMLALDX : T2I_mac<1, 0b100, 0b1101, (outs rGPR:$ldst,rGPR:$hdst),
+                        (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlaldx",
                         "\t$ldst, $hdst, $a, $b", []>;
-def t2SMLSLD  : T2I_mac<1, 0b101, 0b1100, (outs GPR:$ldst,GPR:$hdst),
-                        (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlsld",
+def t2SMLSLD  : T2I_mac<1, 0b101, 0b1100, (outs rGPR:$ldst,rGPR:$hdst),
+                        (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlsld",
                         "\t$ldst, $hdst, $a, $b", []>;
-def t2SMLSLDX : T2I_mac<1, 0b101, 0b1101, (outs GPR:$ldst,GPR:$hdst),
-                        (ins GPR:$a,GPR:$b), IIC_iMAC64, "smlsldx",
+def t2SMLSLDX : T2I_mac<1, 0b101, 0b1101, (outs rGPR:$ldst,rGPR:$hdst),
+                        (ins rGPR:$a,rGPR:$b), IIC_iMAC64, "smlsldx",
                         "\t$ldst, $hdst, $a, $b", []>;
 
 //===----------------------------------------------------------------------===//
@@ -2074,35 +2083,35 @@
   let Inst{5-4} = op2;
 }
 
-def t2CLZ : T2I_misc<0b11, 0b00, (outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
-                    "clz", "\t$dst, $src", [(set GPR:$dst, (ctlz GPR:$src))]>;
+def t2CLZ : T2I_misc<0b11, 0b00, (outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
+                    "clz", "\t$dst, $src", [(set rGPR:$dst, (ctlz rGPR:$src))]>;
 
-def t2RBIT : T2I_misc<0b01, 0b10, (outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+def t2RBIT : T2I_misc<0b01, 0b10, (outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                       "rbit", "\t$dst, $src",
-                      [(set GPR:$dst, (ARMrbit GPR:$src))]>;
+                      [(set rGPR:$dst, (ARMrbit rGPR:$src))]>;
 
-def t2REV : T2I_misc<0b01, 0b00, (outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
-                   "rev", ".w\t$dst, $src", [(set GPR:$dst, (bswap GPR:$src))]>;
+def t2REV : T2I_misc<0b01, 0b00, (outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
+                   "rev", ".w\t$dst, $src", [(set rGPR:$dst, (bswap rGPR:$src))]>;
 
-def t2REV16 : T2I_misc<0b01, 0b01, (outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+def t2REV16 : T2I_misc<0b01, 0b01, (outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                        "rev16", ".w\t$dst, $src",
-                [(set GPR:$dst,
-                    (or (and (srl GPR:$src, (i32 8)), 0xFF),
-                        (or (and (shl GPR:$src, (i32 8)), 0xFF00),
-                            (or (and (srl GPR:$src, (i32 8)), 0xFF0000),
-                                (and (shl GPR:$src, (i32 8)), 0xFF000000)))))]>;
+                [(set rGPR:$dst,
+                    (or (and (srl rGPR:$src, (i32 8)), 0xFF),
+                        (or (and (shl rGPR:$src, (i32 8)), 0xFF00),
+                            (or (and (srl rGPR:$src, (i32 8)), 0xFF0000),
+                                (and (shl rGPR:$src, (i32 8)), 0xFF000000)))))]>;
 
-def t2REVSH : T2I_misc<0b01, 0b11, (outs GPR:$dst), (ins GPR:$src), IIC_iUNAr,
+def t2REVSH : T2I_misc<0b01, 0b11, (outs rGPR:$dst), (ins rGPR:$src), IIC_iUNAr,
                        "revsh", ".w\t$dst, $src",
-                 [(set GPR:$dst,
+                 [(set rGPR:$dst,
                     (sext_inreg
-                      (or (srl (and GPR:$src, 0xFF00), (i32 8)),
-                          (shl GPR:$src, (i32 8))), i16))]>;
+                      (or (srl (and rGPR:$src, 0xFF00), (i32 8)),
+                          (shl rGPR:$src, (i32 8))), i16))]>;
 
-def t2PKHBT : T2I<(outs GPR:$dst), (ins GPR:$src1, GPR:$src2, i32imm:$shamt),
+def t2PKHBT : T2I<(outs rGPR:$dst), (ins rGPR:$src1, rGPR:$src2, i32imm:$shamt),
                   IIC_iALUsi, "pkhbt", "\t$dst, $src1, $src2, lsl $shamt",
-                  [(set GPR:$dst, (or (and GPR:$src1, 0xFFFF),
-                                      (and (shl GPR:$src2, (i32 imm:$shamt)),
+                  [(set rGPR:$dst, (or (and rGPR:$src1, 0xFFFF),
+                                      (and (shl rGPR:$src2, (i32 imm:$shamt)),
                                            0xFFFF0000)))]>,
                   Requires<[HasT2ExtractPack]> {
   let Inst{31-27} = 0b11101;
@@ -2113,17 +2122,17 @@
 }
 
 // Alternate cases for PKHBT where identities eliminate some nodes.
-def : T2Pat<(or (and GPR:$src1, 0xFFFF), (and GPR:$src2, 0xFFFF0000)),
-            (t2PKHBT GPR:$src1, GPR:$src2, 0)>,
+def : T2Pat<(or (and rGPR:$src1, 0xFFFF), (and rGPR:$src2, 0xFFFF0000)),
+            (t2PKHBT rGPR:$src1, rGPR:$src2, 0)>,
             Requires<[HasT2ExtractPack]>;
-def : T2Pat<(or (and GPR:$src1, 0xFFFF), (shl GPR:$src2, imm16_31:$shamt)),
-            (t2PKHBT GPR:$src1, GPR:$src2, imm16_31:$shamt)>,
+def : T2Pat<(or (and rGPR:$src1, 0xFFFF), (shl rGPR:$src2, imm16_31:$shamt)),
+            (t2PKHBT rGPR:$src1, rGPR:$src2, imm16_31:$shamt)>,
             Requires<[HasT2ExtractPack]>;
 
-def t2PKHTB : T2I<(outs GPR:$dst), (ins GPR:$src1, GPR:$src2, i32imm:$shamt),
+def t2PKHTB : T2I<(outs rGPR:$dst), (ins rGPR:$src1, rGPR:$src2, i32imm:$shamt),
                   IIC_iALUsi, "pkhtb", "\t$dst, $src1, $src2, asr $shamt",
-                  [(set GPR:$dst, (or (and GPR:$src1, 0xFFFF0000),
-                                      (and (sra GPR:$src2, imm16_31:$shamt),
+                  [(set rGPR:$dst, (or (and rGPR:$src1, 0xFFFF0000),
+                                      (and (sra rGPR:$src2, imm16_31:$shamt),
                                            0xFFFF)))]>,
                   Requires<[HasT2ExtractPack]> {
   let Inst{31-27} = 0b11101;
@@ -2135,12 +2144,12 @@
 
 // Alternate cases for PKHTB where identities eliminate some nodes.  Note that
 // a shift amount of 0 is *not legal* here, it is PKHBT instead.
-def : T2Pat<(or (and GPR:$src1, 0xFFFF0000), (srl GPR:$src2, (i32 16))),
-            (t2PKHTB GPR:$src1, GPR:$src2, 16)>,
+def : T2Pat<(or (and rGPR:$src1, 0xFFFF0000), (srl rGPR:$src2, (i32 16))),
+            (t2PKHTB rGPR:$src1, rGPR:$src2, 16)>,
             Requires<[HasT2ExtractPack]>;
-def : T2Pat<(or (and GPR:$src1, 0xFFFF0000),
-                     (and (srl GPR:$src2, imm1_15:$shamt), 0xFFFF)),
-            (t2PKHTB GPR:$src1, GPR:$src2, imm1_15:$shamt)>,
+def : T2Pat<(or (and rGPR:$src1, 0xFFFF0000),
+                     (and (srl rGPR:$src2, imm1_15:$shamt), 0xFFFF)),
+            (t2PKHTB rGPR:$src1, rGPR:$src2, imm1_15:$shamt)>,
             Requires<[HasT2ExtractPack]>;
 
 //===----------------------------------------------------------------------===//
@@ -2179,9 +2188,9 @@
 // FIXME: should be able to write a pattern for ARMcmov, but can't use
 // a two-value operand where a dag node expects two operands. :(
 let neverHasSideEffects = 1 in {
-def t2MOVCCr : T2I<(outs GPR:$dst), (ins GPR:$false, GPR:$true), IIC_iCMOVr,
+def t2MOVCCr : T2I<(outs rGPR:$dst), (ins rGPR:$false, rGPR:$true), IIC_iCMOVr,
                    "mov", ".w\t$dst, $true",
-      [/*(set GPR:$dst, (ARMcmov GPR:$false, GPR:$true, imm:$cc, CCR:$ccr))*/]>,
+   [/*(set rGPR:$dst, (ARMcmov rGPR:$false, rGPR:$true, imm:$cc, CCR:$ccr))*/]>,
                 RegConstraint<"$false = $dst"> {
   let Inst{31-27} = 0b11101;
   let Inst{26-25} = 0b01;
@@ -2192,9 +2201,9 @@
   let Inst{7-4} = 0b0000;
 }
 
-def t2MOVCCi : T2I<(outs GPR:$dst), (ins GPR:$false, t2_so_imm:$true),
+def t2MOVCCi : T2I<(outs rGPR:$dst), (ins rGPR:$false, t2_so_imm:$true),
                    IIC_iCMOVi, "mov", ".w\t$dst, $true",
-[/*(set GPR:$dst, (ARMcmov GPR:$false, t2_so_imm:$true, imm:$cc, CCR:$ccr))*/]>,
+[/*(set rGPR:$dst,(ARMcmov rGPR:$false,t2_so_imm:$true, imm:$cc, CCR:$ccr))*/]>,
                    RegConstraint<"$false = $dst"> {
   let Inst{31-27} = 0b11110;
   let Inst{25} = 0;
@@ -2214,20 +2223,20 @@
   let Inst{19-16} = 0b1111; // Rn
   let Inst{5-4} = opcod; // Shift type.
 }
-def t2MOVCClsl : T2I_movcc_sh<0b00, (outs GPR:$dst),
-                             (ins GPR:$false, GPR:$true, i32imm:$rhs),
+def t2MOVCClsl : T2I_movcc_sh<0b00, (outs rGPR:$dst),
+                             (ins rGPR:$false, rGPR:$true, i32imm:$rhs),
                              IIC_iCMOVsi, "lsl", ".w\t$dst, $true, $rhs", []>,
                  RegConstraint<"$false = $dst">;
-def t2MOVCClsr : T2I_movcc_sh<0b01, (outs GPR:$dst),
-                             (ins GPR:$false, GPR:$true, i32imm:$rhs),
+def t2MOVCClsr : T2I_movcc_sh<0b01, (outs rGPR:$dst),
+                             (ins rGPR:$false, rGPR:$true, i32imm:$rhs),
                              IIC_iCMOVsi, "lsr", ".w\t$dst, $true, $rhs", []>,
                  RegConstraint<"$false = $dst">;
-def t2MOVCCasr : T2I_movcc_sh<0b10, (outs GPR:$dst),
-                             (ins GPR:$false, GPR:$true, i32imm:$rhs),
+def t2MOVCCasr : T2I_movcc_sh<0b10, (outs rGPR:$dst),
+                             (ins rGPR:$false, rGPR:$true, i32imm:$rhs),
                              IIC_iCMOVsi, "asr", ".w\t$dst, $true, $rhs", []>,
                  RegConstraint<"$false = $dst">;
-def t2MOVCCror : T2I_movcc_sh<0b11, (outs GPR:$dst),
-                             (ins GPR:$false, GPR:$true, i32imm:$rhs),
+def t2MOVCCror : T2I_movcc_sh<0b11, (outs rGPR:$dst),
+                             (ins rGPR:$false, rGPR:$true, i32imm:$rhs),
                              IIC_iCMOVsi, "ror", ".w\t$dst, $true, $rhs", []>,
                  RegConstraint<"$false = $dst">;
 } // neverHasSideEffects
@@ -2342,13 +2351,13 @@
 }
 
 let mayLoad = 1 in {
-def t2LDREXB : T2I_ldrex<0b00, (outs GPR:$dest), (ins GPR:$ptr), AddrModeNone,
+def t2LDREXB : T2I_ldrex<0b00, (outs rGPR:$dest), (ins rGPR:$ptr), AddrModeNone,
                          Size4Bytes, NoItinerary, "ldrexb", "\t$dest, [$ptr]",
                          "", []>;
-def t2LDREXH : T2I_ldrex<0b01, (outs GPR:$dest), (ins GPR:$ptr), AddrModeNone,
+def t2LDREXH : T2I_ldrex<0b01, (outs rGPR:$dest), (ins rGPR:$ptr), AddrModeNone,
                          Size4Bytes, NoItinerary, "ldrexh", "\t$dest, [$ptr]",
                          "", []>;
-def t2LDREX  : Thumb2I<(outs GPR:$dest), (ins GPR:$ptr), AddrModeNone,
+def t2LDREX  : Thumb2I<(outs rGPR:$dest), (ins rGPR:$ptr), AddrModeNone,
                        Size4Bytes, NoItinerary,
                        "ldrex", "\t$dest, [$ptr]", "",
                       []> {
@@ -2357,20 +2366,20 @@
   let Inst{11-8} = 0b1111;
   let Inst{7-0} = 0b00000000; // imm8 = 0
 }
-def t2LDREXD : T2I_ldrex<0b11, (outs GPR:$dest, GPR:$dest2), (ins GPR:$ptr),
+def t2LDREXD : T2I_ldrex<0b11, (outs rGPR:$dest, rGPR:$dest2), (ins rGPR:$ptr),
                          AddrModeNone, Size4Bytes, NoItinerary,
                          "ldrexd", "\t$dest, $dest2, [$ptr]", "",
                          [], {?, ?, ?, ?}>;
 }
 
 let mayStore = 1, Constraints = "@earlyclobber $success" in {
-def t2STREXB : T2I_strex<0b00, (outs GPR:$success), (ins GPR:$src, GPR:$ptr),
+def t2STREXB : T2I_strex<0b00, (outs rGPR:$success), (ins rGPR:$src, rGPR:$ptr),
                          AddrModeNone, Size4Bytes, NoItinerary,
                          "strexb", "\t$success, $src, [$ptr]", "", []>;
-def t2STREXH : T2I_strex<0b01, (outs GPR:$success), (ins GPR:$src, GPR:$ptr),
+def t2STREXH : T2I_strex<0b01, (outs rGPR:$success), (ins rGPR:$src, rGPR:$ptr),
                          AddrModeNone, Size4Bytes, NoItinerary,
                          "strexh", "\t$success, $src, [$ptr]", "", []>;
-def t2STREX  : Thumb2I<(outs GPR:$success), (ins GPR:$src, GPR:$ptr),
+def t2STREX  : Thumb2I<(outs rGPR:$success), (ins rGPR:$src, rGPR:$ptr),
                        AddrModeNone, Size4Bytes, NoItinerary,
                        "strex", "\t$success, $src, [$ptr]", "",
                       []> {
@@ -2378,8 +2387,8 @@
   let Inst{26-20} = 0b0000100;
   let Inst{7-0} = 0b00000000; // imm8 = 0
 }
-def t2STREXD : T2I_strex<0b11, (outs GPR:$success),
-                         (ins GPR:$src, GPR:$src2, GPR:$ptr),
+def t2STREXD : T2I_strex<0b11, (outs rGPR:$success),
+                         (ins rGPR:$src, rGPR:$src2, rGPR:$ptr),
                          AddrModeNone, Size4Bytes, NoItinerary,
                          "strexd", "\t$success, $src, $src2, [$ptr]", "", [],
                          {?, ?, ?, ?}>;
@@ -2573,7 +2582,7 @@
 
 // Branch and Exchange Jazelle -- for disassembly only
 // Rm = Inst{19-16}
-def t2BXJ : T2I<(outs), (ins GPR:$func), NoItinerary, "bxj", "\t$func",
+def t2BXJ : T2I<(outs), (ins rGPR:$func), NoItinerary, "bxj", "\t$func",
               [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2660,25 +2669,25 @@
 }
 
 // Return From Exception is a system instruction -- for disassembly only
-def t2RFEDBW : T2I<(outs), (ins GPR:$base), NoItinerary, "rfedb", "\t$base!",
+def t2RFEDBW : T2I<(outs), (ins rGPR:$base), NoItinerary, "rfedb", "\t$base!",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0000011; // W = 1
 }
 
-def t2RFEDB  : T2I<(outs), (ins GPR:$base), NoItinerary, "rfeab", "\t$base",
+def t2RFEDB  : T2I<(outs), (ins rGPR:$base), NoItinerary, "rfeab", "\t$base",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0000001; // W = 0
 }
 
-def t2RFEIAW : T2I<(outs), (ins GPR:$base), NoItinerary, "rfeia", "\t$base!",
+def t2RFEIAW : T2I<(outs), (ins rGPR:$base), NoItinerary, "rfeia", "\t$base!",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0011011; // W = 1
 }
 
-def t2RFEIA  : T2I<(outs), (ins GPR:$base), NoItinerary, "rfeia", "\t$base",
+def t2RFEIA  : T2I<(outs), (ins rGPR:$base), NoItinerary, "rfeia", "\t$base",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0011001; // W = 0
@@ -2689,26 +2698,26 @@
 //
 
 // Two piece so_imms.
-def : T2Pat<(or GPR:$LHS, t2_so_imm2part:$RHS),
-             (t2ORRri (t2ORRri GPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
+def : T2Pat<(or rGPR:$LHS, t2_so_imm2part:$RHS),
+             (t2ORRri (t2ORRri rGPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
                     (t2_so_imm2part_2 imm:$RHS))>;
-def : T2Pat<(xor GPR:$LHS, t2_so_imm2part:$RHS),
-             (t2EORri (t2EORri GPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
+def : T2Pat<(xor rGPR:$LHS, t2_so_imm2part:$RHS),
+             (t2EORri (t2EORri rGPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
                     (t2_so_imm2part_2 imm:$RHS))>;
-def : T2Pat<(add GPR:$LHS, t2_so_imm2part:$RHS),
-             (t2ADDri (t2ADDri GPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
+def : T2Pat<(add rGPR:$LHS, t2_so_imm2part:$RHS),
+             (t2ADDri (t2ADDri rGPR:$LHS, (t2_so_imm2part_1 imm:$RHS)),
                     (t2_so_imm2part_2 imm:$RHS))>;
-def : T2Pat<(add GPR:$LHS, t2_so_neg_imm2part:$RHS),
-             (t2SUBri (t2SUBri GPR:$LHS, (t2_so_neg_imm2part_1 imm:$RHS)),
+def : T2Pat<(add rGPR:$LHS, t2_so_neg_imm2part:$RHS),
+             (t2SUBri (t2SUBri rGPR:$LHS, (t2_so_neg_imm2part_1 imm:$RHS)),
                     (t2_so_neg_imm2part_2 imm:$RHS))>;
 
 // 32-bit immediate using movw + movt.
 // This is a single pseudo instruction to make it re-materializable. Remove
 // when we can do generalized remat.
 let isReMaterializable = 1 in
-def t2MOVi32imm : T2Ix2<(outs GPR:$dst), (ins i32imm:$src), IIC_iMOVi,
+def t2MOVi32imm : T2Ix2<(outs rGPR:$dst), (ins i32imm:$src), IIC_iMOVi,
                    "movw", "\t$dst, ${src:lo16}\n\tmovt${p}\t$dst, ${src:hi16}",
-                     [(set GPR:$dst, (i32 imm:$src))]>;
+                     [(set rGPR:$dst, (i32 imm:$src))]>;
 
 // ConstantPool, GlobalAddress, and JumpTable
 def : T2Pat<(ARMWrapper  tglobaladdr :$dst), (t2LEApcrel tglobaladdr :$dst)>,
@@ -2736,7 +2745,7 @@
 //
 
 // Rd = Instr{11-8}
-def t2MRS : T2I<(outs GPR:$dst), (ins), NoItinerary, "mrs", "\t$dst, cpsr",
+def t2MRS : T2I<(outs rGPR:$dst), (ins), NoItinerary, "mrs", "\t$dst, cpsr",
                 [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2747,7 +2756,7 @@
 }
 
 // Rd = Instr{11-8}
-def t2MRSsys : T2I<(outs GPR:$dst), (ins), NoItinerary, "mrs", "\t$dst, spsr",
+def t2MRSsys : T2I<(outs rGPR:$dst), (ins), NoItinerary, "mrs", "\t$dst, spsr",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
   let Inst{26} = 0;
@@ -2758,7 +2767,7 @@
 }
 
 // Rn = Inst{19-16}
-def t2MSR : T2I<(outs), (ins GPR:$src, msr_mask:$mask), NoItinerary, "msr",
+def t2MSR : T2I<(outs), (ins rGPR:$src, msr_mask:$mask), NoItinerary, "msr",
                 "\tcpsr$mask, $src",
                 [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;
@@ -2770,7 +2779,7 @@
 }
 
 // Rn = Inst{19-16}
-def t2MSRsys : T2I<(outs), (ins GPR:$src, msr_mask:$mask), NoItinerary, "msr",
+def t2MSRsys : T2I<(outs), (ins rGPR:$src, msr_mask:$mask), NoItinerary, "msr",
                    "\tspsr$mask, $src",
                    [/* For disassembly only; pattern left blank */]> {
   let Inst{31-27} = 0b11110;

Modified: llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMRegisterInfo.td Thu Jul 29 21:41:01 2010
@@ -318,6 +318,115 @@
   }];
 }
 
+// restricted GPR register class. Many Thumb2 instructions allow the full
+// register range for operands, but have undefined behaviours when PC
+// or SP (R13 or R15) are used. The ARM ARM refers to these operands
+// via the BadReg() pseudo-code description.
+def rGPR : RegisterClass<"ARM", [i32], 32, [R0, R1, R2, R3, R4, R5, R6,
+                                            R7, R8, R9, R10, R11, R12, LR]> {
+  let MethodProtos = [{
+    iterator allocation_order_begin(const MachineFunction &MF) const;
+    iterator allocation_order_end(const MachineFunction &MF) const;
+  }];
+  let MethodBodies = [{
+    // FP is R11, R9 is available.
+    static const unsigned ARM_rGPRAO_1[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R12,ARM::LR,
+      ARM::R4, ARM::R5, ARM::R6, ARM::R7,
+      ARM::R8, ARM::R9, ARM::R10,
+      ARM::R11 };
+    // FP is R11, R9 is not available.
+    static const unsigned ARM_rGPRAO_2[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R12,ARM::LR,
+      ARM::R4, ARM::R5, ARM::R6, ARM::R7,
+      ARM::R8, ARM::R10,
+      ARM::R11 };
+    // FP is R7, R9 is available as non-callee-saved register.
+    // This is used by Darwin.
+    static const unsigned ARM_rGPRAO_3[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R9, ARM::R12,ARM::LR,
+      ARM::R4, ARM::R5, ARM::R6,
+      ARM::R8, ARM::R10,ARM::R11,ARM::R7 };
+    // FP is R7, R9 is not available.
+    static const unsigned ARM_rGPRAO_4[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R12,ARM::LR,
+      ARM::R4, ARM::R5, ARM::R6,
+      ARM::R8, ARM::R10,ARM::R11,
+      ARM::R7 };
+    // FP is R7, R9 is available as callee-saved register.
+    // This is used by non-Darwin platform in Thumb mode.
+    static const unsigned ARM_rGPRAO_5[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R12,ARM::LR,
+      ARM::R4, ARM::R5, ARM::R6,
+      ARM::R8, ARM::R9, ARM::R10,ARM::R11,ARM::R7 };
+
+    // For Thumb1 mode, we don't want to allocate hi regs at all, as we
+    // don't know how to spill them. If we make our prologue/epilogue code
+    // smarter at some point, we can go back to using the above allocation
+    // orders for the Thumb1 instructions that know how to use hi regs.
+    static const unsigned THUMB_rGPRAO[] = {
+      ARM::R0, ARM::R1, ARM::R2, ARM::R3,
+      ARM::R4, ARM::R5, ARM::R6, ARM::R7 };
+
+    rGPRClass::iterator
+    rGPRClass::allocation_order_begin(const MachineFunction &MF) const {
+      const TargetMachine &TM = MF.getTarget();
+      const ARMSubtarget &Subtarget = TM.getSubtarget<ARMSubtarget>();
+      if (Subtarget.isThumb1Only())
+        return THUMB_rGPRAO;
+      if (Subtarget.isTargetDarwin()) {
+        if (Subtarget.isR9Reserved())
+          return ARM_rGPRAO_4;
+        else
+          return ARM_rGPRAO_3;
+      } else {
+        if (Subtarget.isR9Reserved())
+          return ARM_rGPRAO_2;
+        else if (Subtarget.isThumb())
+          return ARM_rGPRAO_5;
+        else
+          return ARM_rGPRAO_1;
+      }
+    }
+
+    rGPRClass::iterator
+    rGPRClass::allocation_order_end(const MachineFunction &MF) const {
+      const TargetMachine &TM = MF.getTarget();
+      const TargetRegisterInfo *RI = TM.getRegisterInfo();
+      const ARMSubtarget &Subtarget = TM.getSubtarget<ARMSubtarget>();
+      GPRClass::iterator I;
+
+      if (Subtarget.isThumb1Only()) {
+        I = THUMB_rGPRAO + (sizeof(THUMB_rGPRAO)/sizeof(unsigned));
+        // Mac OS X requires FP not to be clobbered for backtracing purpose.
+        return (Subtarget.isTargetDarwin() || RI->hasFP(MF)) ? I-1 : I;
+      }
+
+      if (Subtarget.isTargetDarwin()) {
+        if (Subtarget.isR9Reserved())
+          I = ARM_rGPRAO_4 + (sizeof(ARM_rGPRAO_4)/sizeof(unsigned));
+        else
+          I = ARM_rGPRAO_3 + (sizeof(ARM_rGPRAO_3)/sizeof(unsigned));
+      } else {
+        if (Subtarget.isR9Reserved())
+          I = ARM_rGPRAO_2 + (sizeof(ARM_rGPRAO_2)/sizeof(unsigned));
+        else if (Subtarget.isThumb())
+          I = ARM_rGPRAO_5 + (sizeof(ARM_rGPRAO_5)/sizeof(unsigned));
+        else
+          I = ARM_rGPRAO_1 + (sizeof(ARM_rGPRAO_1)/sizeof(unsigned));
+      }
+
+      // Mac OS X requires FP not to be clobbered for backtracing purpose.
+      return (Subtarget.isTargetDarwin() || RI->hasFP(MF)) ? I-1 : I;
+    }
+  }];
+}
+
 // Thumb registers are R0-R7 normally. Some instructions can still use
 // the general GPR register class above (MOV, e.g.)
 def tGPR : RegisterClass<"ARM", [i32], 32, [R0, R1, R2, R3, R4, R5, R6, R7]> {

Modified: llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp Thu Jul 29 21:41:01 2010
@@ -93,6 +93,9 @@
     RegClassID = ARM::DPRRegClassID;
   }
 
+  // For this purpose, we can treat rGPR as if it were GPR.
+  if (RegClassID == ARM::rGPRRegClassID) RegClassID = ARM::GPRRegClassID;
+
   // See also decodeNEONRd(), decodeNEONRn(), decodeNEONRm().
   unsigned RegNum =
     RegClassID == ARM::QPRRegClassID ? RawRegister >> 1 : RawRegister;

Modified: llvm/trunk/lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h (original)
+++ llvm/trunk/lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h Thu Jul 29 21:41:01 2010
@@ -103,7 +103,7 @@
 }
 
 static inline bool IsGPR(unsigned RegClass) {
-  return RegClass == ARM::GPRRegClassID;
+  return RegClass == ARM::GPRRegClassID || RegClass == ARM::rGPRRegClassID;
 }
 
 // Utilities for 32-bit Thumb instructions.
@@ -1324,7 +1324,7 @@
            && OpInfo[1].RegClass == ARM::GPRRegClassID
            && OpInfo[2].RegClass < 0
            && OpInfo[3].RegClass < 0
-           && "Exactlt 4 operands expect and first two as reg operands");
+           && "Exactly 4 operands expect and first two as reg operands");
     // Only need to populate the src reg operand.
     MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
                                                        decodeRm(insn))));
@@ -1338,17 +1338,20 @@
   OpIdx = 0;
 
   assert(NumOps >= 2
-         && OpInfo[0].RegClass == ARM::GPRRegClassID
-         && OpInfo[1].RegClass == ARM::GPRRegClassID
+         && (OpInfo[0].RegClass == ARM::GPRRegClassID ||
+             OpInfo[0].RegClass == ARM::rGPRRegClassID)
+         && (OpInfo[1].RegClass == ARM::GPRRegClassID ||
+             OpInfo[1].RegClass == ARM::rGPRRegClassID)
          && "Expect >= 2 operands and first two as reg operands");
 
-  bool ThreeReg = (NumOps > 2 && OpInfo[2].RegClass == ARM::GPRRegClassID);
+  bool ThreeReg = (NumOps > 2 && (OpInfo[2].RegClass == ARM::GPRRegClassID ||
+                                  OpInfo[2].RegClass == ARM::rGPRRegClassID));
   bool NoDstReg = (decodeRs(insn) == 0xF);
 
   // Build the register operands, followed by the constant shift specifier.
 
   MI.addOperand(MCOperand::CreateReg(
-                  getRegisterEnum(B, ARM::GPRRegClassID,
+                  getRegisterEnum(B, OpInfo[0].RegClass,
                                   NoDstReg ? decodeRn(insn) : decodeRs(insn))));
   ++OpIdx;
 
@@ -1359,7 +1362,7 @@
       MI.addOperand(MI.getOperand(Idx));
       ++OpIdx;
     } else if (!NoDstReg) {
-      MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+      MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, OpInfo[1].RegClass,
                                                          decodeRn(insn))));
       ++OpIdx;
     } else {
@@ -1368,7 +1371,7 @@
     }
   }
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, OpInfo[OpIdx].RegClass,
                                                      decodeRm(insn))));
   ++OpIdx;
 
@@ -1416,16 +1419,20 @@
 
   OpIdx = 0;
 
-  assert(NumOps >= 2 && OpInfo[0].RegClass == ARM::GPRRegClassID
+  unsigned RdRegClassID = OpInfo[0].RegClass;
+  assert(NumOps >= 2 && (RdRegClassID == ARM::GPRRegClassID ||
+                         RdRegClassID == ARM::rGPRRegClassID)
          && "Expect >= 2 operands and first one as reg operand");
 
-  bool TwoReg = (OpInfo[1].RegClass == ARM::GPRRegClassID);
+  unsigned RnRegClassID = OpInfo[1].RegClass;
+  bool TwoReg = (RnRegClassID == ARM::GPRRegClassID
+                 || RnRegClassID == ARM::rGPRRegClassID);
   bool NoDstReg = (decodeRs(insn) == 0xF);
 
   // Build the register operands, followed by the modified immediate.
 
   MI.addOperand(MCOperand::CreateReg(
-                  getRegisterEnum(B, ARM::GPRRegClassID,
+                  getRegisterEnum(B, RdRegClassID,
                                   NoDstReg ? decodeRn(insn) : decodeRs(insn))));
   ++OpIdx;
 
@@ -1434,7 +1441,7 @@
       DEBUG(errs()<<"Thumb2 encoding error: d==15 for DPModImm 2-reg instr.\n");
       return false;
     }
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, RnRegClassID,
                                                        decodeRn(insn))));
     ++OpIdx;
   }
@@ -1506,14 +1513,18 @@
 
   OpIdx = 0;
 
-  assert(NumOps >= 2 && OpInfo[0].RegClass == ARM::GPRRegClassID
+  unsigned RdRegClassID = OpInfo[0].RegClass;
+  assert(NumOps >= 2 && (RdRegClassID == ARM::GPRRegClassID ||
+                         RdRegClassID == ARM::rGPRRegClassID)
          && "Expect >= 2 operands and first one as reg operand");
 
-  bool TwoReg = (OpInfo[1].RegClass == ARM::GPRRegClassID);
+  unsigned RnRegClassID = OpInfo[1].RegClass;
+  bool TwoReg = (RnRegClassID == ARM::GPRRegClassID
+                 || RnRegClassID == ARM::rGPRRegClassID);
 
   // Build the register operand(s), followed by the immediate(s).
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, RdRegClassID,
                                                      decodeRs(insn))));
   ++OpIdx;
 
@@ -1521,7 +1532,7 @@
   if (Thumb2SaturateOpcode(Opcode)) {
     MI.addOperand(MCOperand::CreateImm(decodeThumb2SaturatePos(Opcode, insn)));
 
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, RnRegClassID,
                                                        decodeRn(insn))));
 
     if (Opcode == ARM::t2SSAT16 || Opcode == ARM::t2USAT16) {
@@ -1549,7 +1560,7 @@
       MI.addOperand(MI.getOperand(Idx));
     } else {
       // Add src reg operand.
-      MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+      MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, RnRegClassID,
                                                          decodeRn(insn))));
     }
     ++OpIdx;
@@ -1557,7 +1568,7 @@
 
   if (Opcode == ARM::t2BFI) {
     // Add val reg operand.
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, RnRegClassID,
                                                        decodeRn(insn))));
     ++OpIdx;
   }
@@ -1959,25 +1970,25 @@
   OpIdx = 0;
 
   assert(NumOps >= 2 &&
-         OpInfo[0].RegClass == ARM::GPRRegClassID &&
-         OpInfo[1].RegClass == ARM::GPRRegClassID &&
+         OpInfo[0].RegClass == ARM::rGPRRegClassID &&
+         OpInfo[1].RegClass == ARM::rGPRRegClassID &&
          "Expect >= 2 operands and first two as reg operands");
 
   // Build the register operands, followed by the optional rotation amount.
 
-  bool ThreeReg = NumOps > 2 && OpInfo[2].RegClass == ARM::GPRRegClassID;
+  bool ThreeReg = NumOps > 2 && OpInfo[2].RegClass == ARM::rGPRRegClassID;
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRs(insn))));
   ++OpIdx;
 
   if (ThreeReg) {
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                        decodeRn(insn))));
     ++OpIdx;
   }
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRm(insn))));
   ++OpIdx;
 
@@ -2009,26 +2020,26 @@
   const TargetOperandInfo *OpInfo = ARMInsts[Opcode].OpInfo;
 
   assert(NumOps >= 3 &&
-         OpInfo[0].RegClass == ARM::GPRRegClassID &&
-         OpInfo[1].RegClass == ARM::GPRRegClassID &&
-         OpInfo[2].RegClass == ARM::GPRRegClassID &&
+         OpInfo[0].RegClass == ARM::rGPRRegClassID &&
+         OpInfo[1].RegClass == ARM::rGPRRegClassID &&
+         OpInfo[2].RegClass == ARM::rGPRRegClassID &&
          "Expect >= 3 operands and first three as reg operands");
 
   // Build the register operands.
 
-  bool FourReg = NumOps > 3 && OpInfo[3].RegClass == ARM::GPRRegClassID;
+  bool FourReg = NumOps > 3 && OpInfo[3].RegClass == ARM::rGPRRegClassID;
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRs(insn))));
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRn(insn))));
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRm(insn))));
 
   if (FourReg)
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                        decodeRd(insn))));
 
   NumOpsAdded = FourReg ? 4 : 3;
@@ -2054,26 +2065,26 @@
   const TargetOperandInfo *OpInfo = ARMInsts[Opcode].OpInfo;
 
   assert(NumOps >= 3 &&
-         OpInfo[0].RegClass == ARM::GPRRegClassID &&
-         OpInfo[1].RegClass == ARM::GPRRegClassID &&
-         OpInfo[2].RegClass == ARM::GPRRegClassID &&
+         OpInfo[0].RegClass == ARM::rGPRRegClassID &&
+         OpInfo[1].RegClass == ARM::rGPRRegClassID &&
+         OpInfo[2].RegClass == ARM::rGPRRegClassID &&
          "Expect >= 3 operands and first three as reg operands");
 
-  bool FourReg = NumOps > 3 && OpInfo[3].RegClass == ARM::GPRRegClassID;
+  bool FourReg = NumOps > 3 && OpInfo[3].RegClass == ARM::rGPRRegClassID;
 
   // Build the register operands.
 
   if (FourReg)
-    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+    MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                        decodeRd(insn))));
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRs(insn))));
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRn(insn))));
 
-  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::GPRRegClassID,
+  MI.addOperand(MCOperand::CreateReg(getRegisterEnum(B, ARM::rGPRRegClassID,
                                                      decodeRm(insn))));
 
   if (FourReg)

Modified: llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Thumb2InstrInfo.cpp Thu Jul 29 21:41:01 2010
@@ -147,8 +147,8 @@
                     unsigned SrcReg, bool isKill, int FI,
                     const TargetRegisterClass *RC,
                     const TargetRegisterInfo *TRI) const {
-  if (RC == ARM::GPRRegisterClass || RC == ARM::tGPRRegisterClass ||
-      RC == ARM::tcGPRRegisterClass) {
+  if (RC == ARM::GPRRegisterClass   || RC == ARM::tGPRRegisterClass ||
+      RC == ARM::tcGPRRegisterClass || RC == ARM::rGPRRegisterClass) {
     DebugLoc DL;
     if (I != MBB.end()) DL = I->getDebugLoc();
 
@@ -173,8 +173,8 @@
                      unsigned DestReg, int FI,
                      const TargetRegisterClass *RC,
                      const TargetRegisterInfo *TRI) const {
-  if (RC == ARM::GPRRegisterClass || RC == ARM::tGPRRegisterClass ||
-      RC == ARM::tcGPRRegisterClass) {
+  if (RC == ARM::GPRRegisterClass   || RC == ARM::tGPRRegisterClass ||
+      RC == ARM::tcGPRRegisterClass || RC == ARM::rGPRRegisterClass) {
     DebugLoc DL;
     if (I != MBB.end()) DL = I->getDebugLoc();
 

Modified: llvm/trunk/test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll (original)
+++ llvm/trunk/test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll Thu Jul 29 21:41:01 2010
@@ -7,13 +7,17 @@
 define void @t() nounwind ssp {
 entry:
 ; CHECK: t:
-; CHECK: bic r0, sp, #7
-; CHECK: subs r0, #16
-; CHECK: mov sp, r0
-; Yes, this is stupid codegen, but it's correct.
-; CHECK: bic r0, sp, #7
-; CHECK: subs r0, #16
-; CHECK: mov sp, r0
+; CHECK:  push  {r4, r7}
+; CHECK:  mov r0, sp
+; CHECK:  add r7, sp, #4
+; CHECK:  bic r0, r0, #7
+; CHECK:  subs  r0, #16
+; CHECK:  mov sp, r0
+; CHECK:  mov r0, sp
+; CHECK:  bic r0, r0, #7
+; CHECK:  subs  r0, #16
+; CHECK:  mov sp, r0
+
   %size = mul i32 8, 2
   %vla_a = alloca i8, i32 %size, align 8
   %vla_b = alloca i8, i32 %size, align 8

Added: llvm/trunk/test/CodeGen/Thumb2/thumb2-badreg-operands.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Thumb2/thumb2-badreg-operands.ll?rev=109842&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Thumb2/thumb2-badreg-operands.ll (added)
+++ llvm/trunk/test/CodeGen/Thumb2/thumb2-badreg-operands.ll Thu Jul 29 21:41:01 2010
@@ -0,0 +1,15 @@
+; RUN: llc < %s -mtriple=thumbv7-apple-darwin10 | FileCheck %s
+
+define void @b(i32 %x) nounwind optsize {
+entry:
+; CHECK: b
+; CHECK: mov r2, sp
+; CHECK: mls r0, r0, r1, r2
+; CHECK: mov sp, r0
+  %0 = mul i32 %x, 24                             ; <i32> [#uses=1]
+  %vla = alloca i8, i32 %0, align 1               ; <i8*> [#uses=1]
+  call arm_aapcscc  void @a(i8* %vla) nounwind optsize
+  ret void
+}
+
+declare void @a(i8*) optsize

Modified: llvm/trunk/utils/TableGen/EDEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/EDEmitter.cpp?rev=109842&r1=109841&r2=109842&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/EDEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/EDEmitter.cpp Thu Jul 29 21:41:01 2010
@@ -578,6 +578,7 @@
 static int ARMFlagFromOpName(LiteralConstantEmitter *type,
                              const std::string &name) {
   REG("GPR");
+  REG("rGPR");
   REG("tcGPR");
   REG("cc_out");
   REG("s_cc_out");




From daniel at zuster.org  Thu Jul 29 21:54:33 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 02:54:33 -0000
Subject: [llvm-commits] [test-suite] r109844 - in /test-suite/trunk:
 MultiSource/Examples/ MultiSource/Examples/Makefile
 MultiSource/Examples/MultipleResults/
 MultiSource/Examples/MultipleResults/Makefile
 MultiSource/Examples/MultipleResults/MultipleResults.c
 MultiSource/Examples/MultipleResults/MultipleResults.reference_output
 ParseMultipleResults TEST.simple.Makefile
Message-ID: <20100730025433.B12532A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 21:54:33 2010
New Revision: 109844

URL: http://llvm.org/viewvc/llvm-project?rev=109844&view=rev
Log:
TEST=simple: Add support for tests which can generate multiple results.

Test is expected to write Output/TESTNAME.extra-results.txt, in the format:
--
Result-FOO: NNN.NNNN
Result-BAR: NNN.NNNN
--
which we will then report as fake tests (with no compile info). Good enough for
now, perhaps.

Added:
    test-suite/trunk/MultiSource/Examples/
    test-suite/trunk/MultiSource/Examples/Makefile
    test-suite/trunk/MultiSource/Examples/MultipleResults/
    test-suite/trunk/MultiSource/Examples/MultipleResults/Makefile
    test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.c
    test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.reference_output
    test-suite/trunk/ParseMultipleResults   (with props)
Modified:
    test-suite/trunk/TEST.simple.Makefile

Added: test-suite/trunk/MultiSource/Examples/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Examples/Makefile?rev=109844&view=auto
==============================================================================
--- test-suite/trunk/MultiSource/Examples/Makefile (added)
+++ test-suite/trunk/MultiSource/Examples/Makefile Thu Jul 29 21:54:33 2010
@@ -0,0 +1,8 @@
+# MultiSource/Examples Makefile:  Build all subdirectories automatically
+
+LEVEL = ../..
+
+PARALLEL_DIRS = MultipleResults
+
+include $(LEVEL)/Makefile.config
+include $(LEVEL)/Makefile.programs

Added: test-suite/trunk/MultiSource/Examples/MultipleResults/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Examples/MultipleResults/Makefile?rev=109844&view=auto
==============================================================================
--- test-suite/trunk/MultiSource/Examples/MultipleResults/Makefile (added)
+++ test-suite/trunk/MultiSource/Examples/MultipleResults/Makefile Thu Jul 29 21:54:33 2010
@@ -0,0 +1,10 @@
+LEVEL = ../../../
+PROG = MultipleResults
+LDFLAGS =
+
+include $(LEVEL)/Makefile.config
+
+EXTRA_TEST_RESULTS_FILE := $(PROJ_OBJ_DIR)/Output/MultipleResults.extra-results.txt
+RUN_OPTIONS := $(EXTRA_TEST_RESULTS_FILE)
+
+include ../../Makefile.multisrc

Added: test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.c
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.c?rev=109844&view=auto
==============================================================================
--- test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.c (added)
+++ test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.c Thu Jul 29 21:54:33 2010
@@ -0,0 +1,11 @@
+#include <stdio.h>
+#include <assert.h>
+
+int main(int argc, char **argv) {
+  assert(argc == 1);
+  FILE *f = fopen(argv[1], "w");
+  fprintf(f, "Result-A: %.2f\n", 0.1);
+  fprintf(f, "Result-B: %.2f\n", 0.3);
+  fclose(f);
+  return 0;
+}

Added: test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.reference_output?rev=109844&view=auto
==============================================================================
--- test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.reference_output (added)
+++ test-suite/trunk/MultiSource/Examples/MultipleResults/MultipleResults.reference_output Thu Jul 29 21:54:33 2010
@@ -0,0 +1 @@
+exit 0

Added: test-suite/trunk/ParseMultipleResults
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/ParseMultipleResults?rev=109844&view=auto
==============================================================================
--- test-suite/trunk/ParseMultipleResults (added)
+++ test-suite/trunk/ParseMultipleResults Thu Jul 29 21:54:33 2010
@@ -0,0 +1,22 @@
+#!/usr/bin/env python
+
+import re
+import sys
+
+def main():
+    _,testname,path = sys.argv
+    data = open(path).read()
+    for result in re.finditer('Result-([^:]*): (.*)', data):
+        name, value = result.groups()
+        value = float(value)
+        print
+        print '-' * 62
+        print ">>> ========= '%s.Subtest.%s' Program " % (testname, name)
+        print '-' * 62
+        print
+        print 'TEST-PASS: exec %s.Subtest.%s' % (testname, name)
+        print 'TEST-RESULT-exec-success: pass'
+        print 'TEST-RESULT-exec-time: program %f' % value
+
+if __name__ == '__main__':
+    main()

Propchange: test-suite/trunk/ParseMultipleResults
------------------------------------------------------------------------------
    svn:executable = *

Modified: test-suite/trunk/TEST.simple.Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/TEST.simple.Makefile?rev=109844&r1=109843&r2=109844&view=diff
==============================================================================
--- test-suite/trunk/TEST.simple.Makefile (original)
+++ test-suite/trunk/TEST.simple.Makefile Thu Jul 29 21:54:33 2010
@@ -50,6 +50,9 @@
 	fi
 	@-printf "TEST-RESULT-exec-time: " >> $@
 	@-grep "^program" Output/$*.out-simple.time >> $@
+	if test -f Output/$*.extra-results.txt; then \
+	  $(PROGDIR)/ParseMultipleResults $(RELDIR)/$* Output/$*.extra-results.txt >> $@; \
+	fi
 
 # Overall tests: just run subordinate tests
 $(PROGRAMS_TO_TEST:%=Output/%.$(TEST).report.txt): \




From daniel at zuster.org  Thu Jul 29 21:56:56 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 02:56:56 -0000
Subject: [llvm-commits] [zorg] r109845 - /zorg/trunk/lnt/lnt/tests/nt.py
Message-ID: <20100730025656.B492E2A6C12C@llvm.org>

Author: ddunbar
Date: Thu Jul 29 21:56:56 2010
New Revision: 109845

URL: http://llvm.org/viewvc/llvm-project?rev=109845&view=rev
Log:
LNT/simple: Make --only-test work for multilevel paths (i.e., MultiSource/Applications), by initializing the directories and copy Makefiles to the obj root.

Modified:
    zorg/trunk/lnt/lnt/tests/nt.py

Modified: zorg/trunk/lnt/lnt/tests/nt.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/tests/nt.py?rev=109845&r1=109844&r2=109845&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/tests/nt.py (original)
+++ zorg/trunk/lnt/lnt/tests/nt.py Thu Jul 29 21:56:56 2010
@@ -1,6 +1,7 @@
 import csv
 import os
 import re
+import shutil
 import subprocess
 import sys
 import time
@@ -255,6 +256,20 @@
         if res != 0:
             fatal('configure failed, log is here: %r' % configure_log_path)
 
+    # If running with --only-test, creating any dirs which might be missing and
+    # copy Makefiles.
+    if opts.only_test is not None:
+        suffix = ''
+        for component in opts.only_test.split('/'):
+            suffix = os.path.join(suffix, component)
+            obj_path = os.path.join(basedir, suffix)
+            src_path = os.path.join(opts.test_suite_root, suffix)
+            if not os.path.exists(obj_path):
+                print '%s: initializing test dir %s' % (timestamp(), suffix)
+                os.mkdir(obj_path)
+                shutil.copyfile(os.path.join(src_path, 'Makefile'),
+                                os.path.join(obj_path, 'Makefile'))
+
     # Always blow away any existing report.
     report_path = os.path.join(basedir)
     if opts.only_test is not None:




From daniel at zuster.org  Thu Jul 29 21:56:59 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 02:56:59 -0000
Subject: [llvm-commits] [zorg] r109846 - /zorg/trunk/lnt/lnt/tests/nt.py
Message-ID: <20100730025659.B122D2A6C12D@llvm.org>

Author: ddunbar
Date: Thu Jul 29 21:56:59 2010
New Revision: 109846

URL: http://llvm.org/viewvc/llvm-project?rev=109846&view=rev
Log:
LNT/simple: Add explicit support for subtest results, so we don't get spurious compile failure test results for them.

Modified:
    zorg/trunk/lnt/lnt/tests/nt.py

Modified: zorg/trunk/lnt/lnt/tests/nt.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/lnt/lnt/tests/nt.py?rev=109846&r1=109845&r2=109846&view=diff
==============================================================================
--- zorg/trunk/lnt/lnt/tests/nt.py (original)
+++ zorg/trunk/lnt/lnt/tests/nt.py Thu Jul 29 21:56:59 2010
@@ -362,6 +362,15 @@
         if opts.only_test is not None:
             program = os.path.join(opts.only_test, program)
         test_base_name = '%s.%s' % (test_namespace, program.replace('.','_'))
+
+        # Check if this is a subtest result, in which case we ignore missing
+        # values.
+        if '_Subtest_' in test_base_name:
+            is_subtest = True
+            test_base_name = test_base_name.replace('_Subtest_', '.')
+        else:
+            is_subtest = False
+
         for info in sample_keys:
             if len(info) == 3:
                 name,key,tname = info
@@ -378,6 +387,8 @@
 
             # FIXME: Move to simpler and more succinct format, using .failed.
             if success_value == '*':
+                if is_subtest:
+                    continue
                 status_value = lnt.testing.FAIL
             elif success_value == 'xfail':
                 status_value = lnt.testing.XFAIL




From stoklund at 2pi.dk  Thu Jul 29 22:37:57 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Thu, 29 Jul 2010 20:37:57 -0700
Subject: [llvm-commits] [llvm] r109842 - in /llvm/trunk:
	lib/Target/ARM/ARMBaseInstrInfo.cpp lib/Target/ARM/ARMInstrThumb2.td
	lib/Target/ARM/ARMRegisterInfo.td
	lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
	lib/Target/ARM/Disassembler/ThumbDisassemblerCore.h
	lib/Target/ARM/Thumb2InstrInfo.cpp
	test/CodeGen/Thumb2/2010-04-15-DynAllocBug.ll
	test/CodeGen/Thumb2/thumb2-badreg-operands.ll
	utils/TableGen/EDEmitter.cpp
In-Reply-To: <20100730024101.B1A382A6C12C@llvm.org>
References: <20100730024101.B1A382A6C12C@llvm.org>
Message-ID: <B1CD6343-D25A-4C7F-868D-023F4F41E45E@2pi.dk>


On Jul 29, 2010, at 7:41 PM, Jim Grosbach wrote:
> This patch adds a new register class for use by Thumb2 that excludes
> the problematic registers (SP and PC) and is used instead of GPR
> for those operands which cannot legally reference PC or SP. The
> trivial coalescer explicitly requires that the register class
> of the destination for the COPY instruction contain the source
> register for the COPY to be considered for coalescing. This prevents
> errant instructions like that above.

Thanks, Jim. This is awesome.

It is really important that target instruction descriptors accurately model the instruction constraints. It lets us do more target independent optimizations.

> Modified: llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp?rev=109842&r1=109841&r2=109842&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp (original)
> +++ llvm/trunk/lib/Target/ARM/ARMBaseInstrInfo.cpp Thu Jul 29 21:41:01 2010
> @@ -721,8 +721,9 @@
>                             Align);
> 
>   // tGPR is used sometimes in ARM instructions that need to avoid using
> -  // certain registers.  Just treat it as GPR here.
> -  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass)
> +  // certain registers.  Just treat it as GPR here. Likewise, rGPR.
> +  if (RC == ARM::tGPRRegisterClass || RC == ARM::tcGPRRegisterClass
> +      || RC == ARM::rGPRRegisterClass)
>     RC = ARM::GPRRegisterClass;

This should be folded into the switch below:

>   switch (RC->getID()) {

/jakob

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1929 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100729/efd7f8fa/attachment.bin 

From baldrick at free.fr  Fri Jul 30 00:49:32 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 30 Jul 2010 05:49:32 -0000
Subject: [llvm-commits] [llvm] r109851 - in /llvm/trunk:
 include/llvm/ADT/ValueMap.h unittests/ADT/ValueMapTest.cpp
Message-ID: <20100730054932.CE2EF2A6C12C@llvm.org>

Author: baldrick
Date: Fri Jul 30 00:49:32 2010
New Revision: 109851

URL: http://llvm.org/viewvc/llvm-project?rev=109851&view=rev
Log:
Fix the ValueMap copy constructor.  The issue is that the map keys are value
handles with a pointer to the containing map.  When a map is copied, these
pointers need to be corrected to point to the new map.  If not, then consider
the case of a map M1 which maps a value V to something.  Create a copy M2 of
M1.  At this point there are two value handles on V, one representing V as a
key in M1, the other representing V as a key in M2.  But both value handles
point to M1 as the containing map.  Now delete V.  The value handles remove
themselves from their containing map (which destroys them), but only the first
value handle is successful: the second one cannot remove itself from M1 as
(once the first one has removed itself) there is nothing there to remove; it
is therefore not destroyed.  This causes an assertion failure "All references
to V were not removed?".

Modified:
    llvm/trunk/include/llvm/ADT/ValueMap.h
    llvm/trunk/unittests/ADT/ValueMapTest.cpp

Modified: llvm/trunk/include/llvm/ADT/ValueMap.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ADT/ValueMap.h?rev=109851&r1=109850&r2=109851&view=diff
==============================================================================
--- llvm/trunk/include/llvm/ADT/ValueMap.h (original)
+++ llvm/trunk/include/llvm/ADT/ValueMap.h Fri Jul 30 00:49:32 2010
@@ -87,7 +87,12 @@
   typedef ValueT mapped_type;
   typedef std::pair<KeyT, ValueT> value_type;
 
-  ValueMap(const ValueMap& Other) : Map(Other.Map), Data(Other.Data) {}
+  ValueMap(const ValueMap& Other) : Map(Other.Map), Data(Other.Data) {
+    // Each ValueMapCVH key contains a pointer to the containing ValueMap.
+    // The keys in the new map need to point to the new map, not Other.
+    for (typename MapT::iterator I = Map.begin(), E = Map.end(); I != E; ++I)
+      I->first.Map = this;
+  }
 
   explicit ValueMap(unsigned NumInitBuckets = 64)
     : Map(NumInitBuckets), Data() {}

Modified: llvm/trunk/unittests/ADT/ValueMapTest.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/unittests/ADT/ValueMapTest.cpp?rev=109851&r1=109850&r2=109851&view=diff
==============================================================================
--- llvm/trunk/unittests/ADT/ValueMapTest.cpp (original)
+++ llvm/trunk/unittests/ADT/ValueMapTest.cpp Fri Jul 30 00:49:32 2010
@@ -39,6 +39,15 @@
 typedef ::testing::Types<Value, Instruction, const Instruction> KeyTypes;
 TYPED_TEST_CASE(ValueMapTest, KeyTypes);
 
+TYPED_TEST(ValueMapTest, CopyConstructor) {
+  ValueMap<TypeParam*, int> VM1;
+  VM1[this->AddV.get()] = 7;
+  ValueMap<TypeParam*, int> VM2(VM1);
+  this->AddV.reset();
+  EXPECT_TRUE(VM1.empty());
+  EXPECT_TRUE(VM2.empty());
+}
+
 TYPED_TEST(ValueMapTest, Null) {
   ValueMap<TypeParam*, int> VM1;
   VM1[NULL] = 7;




From baldrick at free.fr  Fri Jul 30 00:50:45 2010
From: baldrick at free.fr (Duncan Sands)
Date: Fri, 30 Jul 2010 05:50:45 -0000
Subject: [llvm-commits] [llvm] r109852 -
	/llvm/trunk/tools/bugpoint/ExtractFunction.cpp
Message-ID: <20100730055045.A3CA12A6C12C@llvm.org>

Author: baldrick
Date: Fri Jul 30 00:50:45 2010
New Revision: 109852

URL: http://llvm.org/viewvc/llvm-project?rev=109852&view=rev
Log:
Do not pass a copy of the value map, pass a reference to it.

Modified:
    llvm/trunk/tools/bugpoint/ExtractFunction.cpp

Modified: llvm/trunk/tools/bugpoint/ExtractFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ExtractFunction.cpp?rev=109852&r1=109851&r2=109852&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ExtractFunction.cpp (original)
+++ llvm/trunk/tools/bugpoint/ExtractFunction.cpp Fri Jul 30 00:50:45 2010
@@ -199,7 +199,7 @@
 /// static ctors/dtors, we need to add an llvm.global_[cd]tors global to M2, and
 /// prune appropriate entries out of M1s list.
 static void SplitStaticCtorDtor(const char *GlobalName, Module *M1, Module *M2,
-                                ValueMap<const Value*, Value*> VMap) {
+                                ValueMap<const Value*, Value*> &VMap) {
   GlobalVariable *GV = M1->getNamedGlobal(GlobalName);
   if (!GV || GV->isDeclaration() || GV->hasLocalLinkage() ||
       !GV->use_empty()) return;




From evan.cheng at apple.com  Fri Jul 30 00:52:41 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Thu, 29 Jul 2010 22:52:41 -0700
Subject: [llvm-commits] [patch] - Bugfix for
	TargetLowering::SimplifySetCC
In-Reply-To: <000001caf81e$c614ce00$523e6a00$%muller@arm.com>
References: <000001caf81e$c614ce00$523e6a00$%muller@arm.com>
Message-ID: <CAA7B1D6-E805-42FD-9123-40D10380C513@apple.com>

Sorry I completely missed this patch. I agree there is a bug here but I don't know this code well enough to determine if this is the right fix. Eli, you wrote the optimization. Do you mind take a look at this?

Evan

On May 20, 2010, at 6:17 AM, Evzen Muller wrote:

> Hi all,
> 
> When I compile this code (clang generates something similar for bitfield
> access):
> 
> target datalayout = "e-p:32:32-f64:32:32-i64:32:32-n32"
> target triple = "arm-none-eabi"
> 
> define i32 @extend2bit_v2(i32 %val) {
> entry:
>  %0 = trunc i32 %val to i2                       ; <i2> [#uses=1]
>  %1 = sext i2 %0 to i32                          ; <i32> [#uses=1]
>  %2 = icmp eq i32 %1, 3                          ; <i1> [#uses=1]
>  %3 = zext i1 %2 to i32                          ; <i32> [#uses=1]
>  ret i32 %3
> }
> 
> ARM codegen produces:
> @ BB#0:                                 @ %entry
>        and     r1, r0, #3
>        mov     r0, #0
>        cmp     r1, #3
>        moveq   r0, #1
>        bx      lr
> 
> This is not correct, because when %val == 3 this will return 1, but expected
> result is 0 because i2 constant (3) should be sign-extended before icmp.
> 
> There is an optimization in TargetLowering::SimplifySetCC (around line 1790)
> which works fine if C1 has more or less active bits than ExtSrcTyBits, but I
> think that if C1.getActiveBits() == ExtSrcTyBits we should do normal
> truncate/sign-extend.
> 
> Attached patch should fix this problem.
> 
> Thanks
> Evzen
> 
> <simplify_setcc_bugfix.patch>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From eli.friedman at gmail.com  Fri Jul 30 01:44:31 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Fri, 30 Jul 2010 06:44:31 -0000
Subject: [llvm-commits] [llvm] r109854 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/TargetLowering.cpp
 test/CodeGen/X86/2010-07-29-SetccSimplify.ll
Message-ID: <20100730064431.D55032A6C12C@llvm.org>

Author: efriedma
Date: Fri Jul 30 01:44:31 2010
New Revision: 109854

URL: http://llvm.org/viewvc/llvm-project?rev=109854&view=rev
Log:
Fix for bug reported by Evzen Muller on llvm-commits: make sure to correctly
check the range of the constant when optimizing a comparison between a
constant and a sign_extend_inreg node.


Added:
    llvm/trunk/test/CodeGen/X86/2010-07-29-SetccSimplify.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=109854&r1=109853&r2=109854&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Fri Jul 30 01:44:31 2010
@@ -1968,12 +1968,9 @@
       EVT ExtDstTy = N0.getValueType();
       unsigned ExtDstTyBits = ExtDstTy.getSizeInBits();
 
-      // If the extended part has any inconsistent bits, it cannot ever
-      // compare equal.  In other words, they have to be all ones or all
-      // zeros.
-      APInt ExtBits =
-        APInt::getHighBitsSet(ExtDstTyBits, ExtDstTyBits - ExtSrcTyBits);
-      if ((C1 & ExtBits) != 0 && (C1 & ExtBits) != ExtBits)
+      // If the constant doesn't fit into the number of bits for the source of
+      // the sign extension, it is impossible for both sides to be equal.
+      if (C1.getMinSignedBits() > ExtSrcTyBits)
         return DAG.getConstant(Cond == ISD::SETNE, VT);
       
       SDValue ZextOp;

Added: llvm/trunk/test/CodeGen/X86/2010-07-29-SetccSimplify.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2010-07-29-SetccSimplify.ll?rev=109854&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/2010-07-29-SetccSimplify.ll (added)
+++ llvm/trunk/test/CodeGen/X86/2010-07-29-SetccSimplify.ll Fri Jul 30 01:44:31 2010
@@ -0,0 +1,14 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s
+
+define i32 @extend2bit_v2(i32 %val) {
+entry:
+  %0 = trunc i32 %val to i2                       ; <i2> [#uses=1]
+  %1 = sext i2 %0 to i32                          ; <i32> [#uses=1]
+  %2 = icmp eq i32 %1, 3                          ; <i1> [#uses=1]
+  %3 = zext i1 %2 to i32                          ; <i32> [#uses=1]
+  ret i32 %3
+}
+
+; CHECK: extend2bit_v2:
+; CHECK: xorl	%eax, %eax
+; CHECK-NEXT: ret




From rjmccall at apple.com  Fri Jul 30 01:47:29 2010
From: rjmccall at apple.com (John McCall)
Date: Fri, 30 Jul 2010 06:47:29 -0000
Subject: [llvm-commits] [test-suite] r109855 -
 /test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m
Message-ID: <20100730064729.8D85B2A6C12C@llvm.org>

Author: rjmccall
Date: Fri Jul 30 01:47:29 2010
New Revision: 109855

URL: http://llvm.org/viewvc/llvm-project?rev=109855&view=rev
Log:
Fix a race condition in this test causing it to occasionally fail to
terminate (and therefore have halting thrust upon it by RunToolSafely).


Modified:
    test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m

Modified: test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m?rev=109855&r1=109854&r2=109855&view=diff
==============================================================================
--- test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m (original)
+++ test-suite/trunk/SingleSource/UnitTests/ObjC/synchronized.m Fri Jul 30 01:47:29 2010
@@ -29,8 +29,7 @@
 
 int main() {
   unsigned i;
-  numThreads = 10;
-  for (i = 0; i < numThreads; ++i) {
+  for (i = 0; i < kNumThreads; ++i) {
     [[[A alloc] init] start];
   }
   while (numThreads) ;




From eli.friedman at gmail.com  Fri Jul 30 01:50:22 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Thu, 29 Jul 2010 23:50:22 -0700
Subject: [llvm-commits] [patch] - Bugfix for
	TargetLowering::SimplifySetCC
In-Reply-To: <CAA7B1D6-E805-42FD-9123-40D10380C513@apple.com>
References: <000001caf81e$c614ce00$523e6a00$%muller@arm.com>
	<CAA7B1D6-E805-42FD-9123-40D10380C513@apple.com>
Message-ID: <AANLkTimxoYmVPq5QYe4=ddXAa6eX0RAZOGwWv4QiA=+V@mail.gmail.com>

Not actually my code (I only reindented it); in any case, fixed in r109854.

-Eli

On Thu, Jul 29, 2010 at 10:52 PM, Evan Cheng <evan.cheng at apple.com> wrote:
> Sorry I completely missed this patch. I agree there is a bug here but I don't know this code well enough to determine if this is the right fix. Eli, you wrote the optimization. Do you mind take a look at this?
>
> Evan
>
> On May 20, 2010, at 6:17 AM, Evzen Muller wrote:
>
>> Hi all,
>>
>> When I compile this code (clang generates something similar for bitfield
>> access):
>>
>> target datalayout = "e-p:32:32-f64:32:32-i64:32:32-n32"
>> target triple = "arm-none-eabi"
>>
>> define i32 @extend2bit_v2(i32 %val) {
>> entry:
>> ?%0 = trunc i32 %val to i2 ? ? ? ? ? ? ? ? ? ? ? ; <i2> [#uses=1]
>> ?%1 = sext i2 %0 to i32 ? ? ? ? ? ? ? ? ? ? ? ? ?; <i32> [#uses=1]
>> ?%2 = icmp eq i32 %1, 3 ? ? ? ? ? ? ? ? ? ? ? ? ?; <i1> [#uses=1]
>> ?%3 = zext i1 %2 to i32 ? ? ? ? ? ? ? ? ? ? ? ? ?; <i32> [#uses=1]
>> ?ret i32 %3
>> }
>>
>> ARM codegen produces:
>> @ BB#0: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? @ %entry
>> ? ? ? ?and ? ? r1, r0, #3
>> ? ? ? ?mov ? ? r0, #0
>> ? ? ? ?cmp ? ? r1, #3
>> ? ? ? ?moveq ? r0, #1
>> ? ? ? ?bx ? ? ?lr
>>
>> This is not correct, because when %val == 3 this will return 1, but expected
>> result is 0 because i2 constant (3) should be sign-extended before icmp.
>>
>> There is an optimization in TargetLowering::SimplifySetCC (around line 1790)
>> which works fine if C1 has more or less active bits than ExtSrcTyBits, but I
>> think that if C1.getActiveBits() == ExtSrcTyBits we should do normal
>> truncate/sign-extend.
>>
>> Attached patch should fix this problem.
>>
>> Thanks
>> Evzen
>>
>> <simplify_setcc_bugfix.patch>_______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



From benny.kra at googlemail.com  Fri Jul 30 07:18:36 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Fri, 30 Jul 2010 12:18:36 -0000
Subject: [llvm-commits] [llvm] r109858 - in /llvm/trunk/utils/valgrind:
 i386-pc-linux-gnu.supp x86_64-pc-linux-gnu.supp
Message-ID: <20100730121836.85B512A6C12C@llvm.org>

Author: d0k
Date: Fri Jul 30 07:18:36 2010
New Revision: 109858

URL: http://llvm.org/viewvc/llvm-project?rev=109858&view=rev
Log:
Tweak suppressions.

Modified:
    llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp
    llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp

Modified: llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp?rev=109858&r1=109857&r2=109858&view=diff
==============================================================================
--- llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp (original)
+++ llvm/trunk/utils/valgrind/i386-pc-linux-gnu.supp Fri Jul 30 07:18:36 2010
@@ -30,6 +30,6 @@
 {
    We don't care if python leaks
    Memcheck:Leak
+   fun:malloc
    obj:/usr/bin/python2.5
 }
-

Modified: llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp?rev=109858&r1=109857&r2=109858&view=diff
==============================================================================
--- llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp (original)
+++ llvm/trunk/utils/valgrind/x86_64-pc-linux-gnu.supp Fri Jul 30 07:18:36 2010
@@ -30,5 +30,6 @@
 {
    We don't care if python leaks
    Memcheck:Leak
+   fun:malloc
    obj:/usr/bin/python2.5
 }




From rafael.espindola at gmail.com  Fri Jul 30 09:19:00 2010
From: rafael.espindola at gmail.com (Rafael Espindola)
Date: Fri, 30 Jul 2010 14:19:00 -0000
Subject: [llvm-commits] [llvm] r109859 - in /llvm/trunk/tools/bugpoint:
 BugDriver.cpp BugDriver.h ExecutionDriver.cpp FindBugs.cpp
 Miscompilation.cpp
Message-ID: <20100730141900.5263C2A6C12C@llvm.org>

Author: rafael
Date: Fri Jul 30 09:19:00 2010
New Revision: 109859

URL: http://llvm.org/viewvc/llvm-project?rev=109859&view=rev
Log:
Add a Program argument to diffProgram to avoid a use of swapProgramIn.

Modified:
    llvm/trunk/tools/bugpoint/BugDriver.cpp
    llvm/trunk/tools/bugpoint/BugDriver.h
    llvm/trunk/tools/bugpoint/ExecutionDriver.cpp
    llvm/trunk/tools/bugpoint/FindBugs.cpp
    llvm/trunk/tools/bugpoint/Miscompilation.cpp

Modified: llvm/trunk/tools/bugpoint/BugDriver.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/BugDriver.cpp?rev=109859&r1=109858&r2=109859&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/BugDriver.cpp (original)
+++ llvm/trunk/tools/bugpoint/BugDriver.cpp Fri Jul 30 09:19:00 2010
@@ -211,7 +211,7 @@
   // matches, then we assume there is a miscompilation bug and try to 
   // diagnose it.
   outs() << "*** Checking the code generator...\n";
-  bool Diff = diffProgram("", "", false, &Error);
+  bool Diff = diffProgram(Program, "", "", false, &Error);
   if (!Error.empty()) {
     errs() << Error;
     return debugCodeGeneratorCrash(ErrMsg);

Modified: llvm/trunk/tools/bugpoint/BugDriver.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/BugDriver.h?rev=109859&r1=109858&r2=109859&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/BugDriver.h (original)
+++ llvm/trunk/tools/bugpoint/BugDriver.h Fri Jul 30 09:19:00 2010
@@ -174,7 +174,8 @@
   /// executeProgram - This method runs "Program", capturing the output of the
   /// program to a file.  A recommended filename may be optionally specified.
   ///
-  std::string executeProgram(std::string OutputFilename,
+  std::string executeProgram(const Module *Program,
+                             std::string OutputFilename,
                              std::string Bitcode,
                              const std::string &SharedObjects,
                              AbstractInterpreter *AI,
@@ -185,7 +186,8 @@
   /// the code generator (e.g., llc crashes), this will return false and set
   /// Error.
   ///
-  std::string executeProgramSafely(std::string OutputFile, std::string *Error);
+  std::string executeProgramSafely(const Module *Program,
+                                   std::string OutputFile, std::string *Error);
 
   /// createReferenceFile - calls compileProgram and then records the output
   /// into ReferenceOutputFile. Returns true if reference file created, false 
@@ -200,7 +202,8 @@
   /// is different, 1 is returned.  If there is a problem with the code
   /// generator (e.g., llc crashes), this will return -1 and set Error.
   ///
-  bool diffProgram(const std::string &BitcodeFile = "",
+  bool diffProgram(const Module *Program,
+                   const std::string &BitcodeFile = "",
                    const std::string &SharedObj = "",
                    bool RemoveBitcode = false,
                    std::string *Error = 0);

Modified: llvm/trunk/tools/bugpoint/ExecutionDriver.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ExecutionDriver.cpp?rev=109859&r1=109858&r2=109859&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ExecutionDriver.cpp (original)
+++ llvm/trunk/tools/bugpoint/ExecutionDriver.cpp Fri Jul 30 09:19:00 2010
@@ -320,7 +320,8 @@
 /// program to a file, returning the filename of the file.  A recommended
 /// filename may be optionally specified.
 ///
-std::string BugDriver::executeProgram(std::string OutputFile,
+std::string BugDriver::executeProgram(const Module *Program,
+                                      std::string OutputFile,
                                       std::string BitcodeFile,
                                       const std::string &SharedObj,
                                       AbstractInterpreter *AI,
@@ -399,9 +400,10 @@
 /// executeProgramSafely - Used to create reference output with the "safe"
 /// backend, if reference output is not provided.
 ///
-std::string BugDriver::executeProgramSafely(std::string OutputFile,
+std::string BugDriver::executeProgramSafely(const Module *Program,
+                                            std::string OutputFile,
                                             std::string *Error) {
-  return executeProgram(OutputFile, "", "", SafeInterpreter, Error);
+  return executeProgram(Program, OutputFile, "", "", SafeInterpreter, Error);
 }
 
 std::string BugDriver::compileSharedObject(const std::string &BitcodeFile,
@@ -440,7 +442,7 @@
   if (!Error.empty())
     return false;
 
-  ReferenceOutputFile = executeProgramSafely(Filename, &Error);
+  ReferenceOutputFile = executeProgramSafely(Program, Filename, &Error);
   if (!Error.empty()) {
     errs() << Error;
     if (Interpreter != SafeInterpreter) {
@@ -460,12 +462,14 @@
 /// is different, 1 is returned.  If there is a problem with the code
 /// generator (e.g., llc crashes), this will return -1 and set Error.
 ///
-bool BugDriver::diffProgram(const std::string &BitcodeFile,
+bool BugDriver::diffProgram(const Module *Program,
+                            const std::string &BitcodeFile,
                             const std::string &SharedObject,
                             bool RemoveBitcode,
                             std::string *ErrMsg) {
   // Execute the program, generating an output file...
-  sys::Path Output(executeProgram("", BitcodeFile, SharedObject, 0, ErrMsg));
+  sys::Path Output(executeProgram(Program, "", BitcodeFile, SharedObject, 0,
+                                  ErrMsg));
   if (!ErrMsg->empty())
     return false;
 

Modified: llvm/trunk/tools/bugpoint/FindBugs.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/FindBugs.cpp?rev=109859&r1=109858&r2=109859&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/FindBugs.cpp (original)
+++ llvm/trunk/tools/bugpoint/FindBugs.cpp Fri Jul 30 09:19:00 2010
@@ -89,7 +89,7 @@
     // output (created above).
     //
     outs() << "*** Checking if passes caused miscompliation:\n";
-    bool Diff = diffProgram(Filename, "", false, &Error);
+    bool Diff = diffProgram(Program, Filename, "", false, &Error);
     if (Error.empty() && Diff) {
       outs() << "\n*** diffProgram returned true!\n";
       debugMiscompilation(&Error);

Modified: llvm/trunk/tools/bugpoint/Miscompilation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/Miscompilation.cpp?rev=109859&r1=109858&r2=109859&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/Miscompilation.cpp (original)
+++ llvm/trunk/tools/bugpoint/Miscompilation.cpp Fri Jul 30 09:19:00 2010
@@ -76,8 +76,8 @@
   }
   
   // Check to see if the finished program matches the reference output...
-  bool Diff = BD.diffProgram(BitcodeResult, "", true /*delete bitcode*/,
-                             &Error);
+  bool Diff = BD.diffProgram(BD.getProgram(), BitcodeResult, "",
+                             true /*delete bitcode*/, &Error);
   if (!Error.empty())
     return InternalError;
   if (Diff) {
@@ -113,7 +113,7 @@
   }
 
   // If the prefix maintains the predicate by itself, only keep the prefix!
-  Diff = BD.diffProgram(BitcodeResult, "", false, &Error);
+  Diff = BD.diffProgram(BD.getProgram(), BitcodeResult, "", false, &Error);
   if (!Error.empty())
     return InternalError;
   if (Diff) {
@@ -153,7 +153,8 @@
   }
 
   // Run the result...
-  Diff = BD.diffProgram(BitcodeResult, "", true /*delete bitcode*/, &Error);
+  Diff = BD.diffProgram(BD.getProgram(), BitcodeResult, "",
+                        true /*delete bitcode*/, &Error);
   if (!Error.empty())
     return InternalError;
   if (Diff) {
@@ -223,15 +224,15 @@
   }
   delete M2;   // We are done with this module.
 
-  OwningPtr<Module> OldProgram(BD.swapProgramIn(M1));
-
   // Execute the program.  If it does not match the expected output, we must
   // return true.
-  bool Broken = BD.diffProgram("", "", false, &Error);
+  bool Broken = BD.diffProgram(M1, "", "", false, &Error);
   if (!Error.empty()) {
-    // Delete the linked module & restore the original
-    delete BD.swapProgramIn(OldProgram.take());
+    // Delete the linked module
+    delete M1;
   }
+  // Delete the original and set the new program.
+  delete BD.swapProgramIn(M1);
   return Broken;
 }
 
@@ -958,7 +959,8 @@
 
   // Run the code generator on the `Test' code, loading the shared library.
   // The function returns whether or not the new output differs from reference.
-  bool Result = BD.diffProgram(TestModuleBC.str(), SharedObject, false, &Error);
+  bool Result = BD.diffProgram(BD.getProgram(), TestModuleBC.str(),
+                               SharedObject, false, &Error);
   if (!Error.empty())
     return false;
 
@@ -975,7 +977,8 @@
 ///
 bool BugDriver::debugCodeGenerator(std::string *Error) {
   if ((void*)SafeInterpreter == (void*)Interpreter) {
-    std::string Result = executeProgramSafely("bugpoint.safe.out", Error);
+    std::string Result = executeProgramSafely(Program, "bugpoint.safe.out",
+                                              Error);
     if (Error->empty()) {
       outs() << "\n*** The \"safe\" i.e. 'known good' backend cannot match "
              << "the reference diff.  This may be due to a\n    front-end "




From espindola at google.com  Fri Jul 30 10:38:11 2010
From: espindola at google.com (Rafael Espindola)
Date: Fri, 30 Jul 2010 11:38:11 -0400
Subject: [llvm-commits] [patch] Another small refactoring in bugpoint
In-Reply-To: <AANLkTimnsXVO6NWgbk2mTqvP+i5b564uHuVbb1RUctG=@mail.gmail.com>
References: <AANLkTinCc=4x2LZaHyGQ898aDb=YObPBreH-nhAevHHx@mail.gmail.com>
	<AANLkTimnsXVO6NWgbk2mTqvP+i5b564uHuVbb1RUctG=@mail.gmail.com>
Message-ID: <AANLkTi=7MbAA2HeAvaB28Ow+wVCCUtQuaVyEUb4tBk=t@mail.gmail.com>

New patch attached. This one adds const to some methods and changes
TestMergedProgram to return the merged module and take a const
BugDriver.

> Nick

Cheers,
-- 
Rafael ?vila de Esp?ndola
-------------- next part --------------
A non-text attachment was scrubbed...
Name: bugpoint.patch
Type: text/x-patch
Size: 6547 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100730/1277bdbd/attachment.bin 

From grosbach at apple.com  Fri Jul 30 10:50:57 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Fri, 30 Jul 2010 08:50:57 -0700
Subject: [llvm-commits] [llvm] r109434 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
In-Reply-To: <20100726210118.BFB9B2A6C12C@llvm.org>
References: <20100726210118.BFB9B2A6C12C@llvm.org>
Message-ID: <716F6A74-8A65-4512-9848-DDECE6D02179@apple.com>

Minor thing. Should work as-is, but the predicates should be using "&&" rather than "&".

-Jim

On Jul 26, 2010, at 2:01 PM, Bruno Cardoso Lopes wrote:

> Author: bruno
> Date: Mon Jul 26 16:01:18 2010
> New Revision: 109434
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109434&view=rev
> Log:
> Temporary hack to let codegen assert or generate poor code in case
> we are using AVX and no AVX version of the desired intruction is present,
> this is better for incremental dev (without fallbacks it's easier to spot
> what's missing). Not sure this is the best hack thought (we can also disable
> all HasSSE* predicates by dinamically marking them 'false' if AVX is present)
> 
> 
> Modified:
>    llvm/trunk/lib/Target/X86/X86InstrInfo.td
> 
> Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=109434&r1=109433&r2=109434&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Jul 26 16:01:18 2010
> @@ -348,14 +348,19 @@
> // X86 Instruction Predicate Definitions.
> def HasCMov      : Predicate<"Subtarget->hasCMov()">;
> def NoCMov       : Predicate<"!Subtarget->hasCMov()">;
> -def HasMMX       : Predicate<"Subtarget->hasMMX()">;
> -def HasSSE1      : Predicate<"Subtarget->hasSSE1()">;
> -def HasSSE2      : Predicate<"Subtarget->hasSSE2()">;
> -def HasSSE3      : Predicate<"Subtarget->hasSSE3()">;
> -def HasSSSE3     : Predicate<"Subtarget->hasSSSE3()">;
> -def HasSSE41     : Predicate<"Subtarget->hasSSE41()">;
> -def HasSSE42     : Predicate<"Subtarget->hasSSE42()">;
> -def HasSSE4A     : Predicate<"Subtarget->hasSSE4A()">;
> +
> +// FIXME: temporary hack to let codegen assert or generate poor code in case
> +// no AVX version of the desired intructions is present, this is better for
> +// incremental dev (without fallbacks it's easier to spot what's missing)
> +def HasMMX       : Predicate<"Subtarget->hasMMX() & !Subtarget->hasAVX()">;
> +def HasSSE1      : Predicate<"Subtarget->hasSSE1() & !Subtarget->hasAVX()">;
> +def HasSSE2      : Predicate<"Subtarget->hasSSE2() & !Subtarget->hasAVX()">;
> +def HasSSE3      : Predicate<"Subtarget->hasSSE3() & !Subtarget->hasAVX()">;
> +def HasSSSE3     : Predicate<"Subtarget->hasSSSE3() & !Subtarget->hasAVX()">;
> +def HasSSE41     : Predicate<"Subtarget->hasSSE41() & !Subtarget->hasAVX()">;
> +def HasSSE42     : Predicate<"Subtarget->hasSSE42() & !Subtarget->hasAVX()">;
> +def HasSSE4A     : Predicate<"Subtarget->hasSSE4A() & !Subtarget->hasAVX()">;
> +
> def HasAVX       : Predicate<"Subtarget->hasAVX()">;
> def HasCLMUL     : Predicate<"Subtarget->hasCLMUL()">;
> def HasFMA3      : Predicate<"Subtarget->hasFMA3()">;
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From grosbach at apple.com  Fri Jul 30 11:01:40 2010
From: grosbach at apple.com (Jim Grosbach)
Date: Fri, 30 Jul 2010 09:01:40 -0700
Subject: [llvm-commits] [llvm] r109608 -
	/llvm/trunk/lib/CodeGen/MachineInstr.cpp
In-Reply-To: <20100728183547.1E75C2A6C12C@llvm.org>
References: <20100728183547.1E75C2A6C12C@llvm.org>
Message-ID: <FB6272CB-0402-44C6-9A09-1DAFB25FC9E1@apple.com>

This is going to be very useful, and not just for those crazy enough to hack on register allocators. ;) Thanks, Jakob.

-Jim

On Jul 28, 2010, at 11:35 AM, Jakob Stoklund Olesen wrote:

> Author: stoklund
> Date: Wed Jul 28 13:35:46 2010
> New Revision: 109608
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109608&view=rev
> Log:
> Print out the regclass of any virtual registers used by a machine instruction.
> 
> Modified:
>    llvm/trunk/lib/CodeGen/MachineInstr.cpp
> 
> Modified: llvm/trunk/lib/CodeGen/MachineInstr.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineInstr.cpp?rev=109608&r1=109607&r2=109608&view=diff
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/MachineInstr.cpp (original)
> +++ llvm/trunk/lib/CodeGen/MachineInstr.cpp Wed Jul 28 13:35:46 2010
> @@ -1236,12 +1236,18 @@
> void MachineInstr::print(raw_ostream &OS, const TargetMachine *TM) const {
>   // We can be a bit tidier if we know the TargetMachine and/or MachineFunction.
>   const MachineFunction *MF = 0;
> +  const MachineRegisterInfo *MRI = 0;
>   if (const MachineBasicBlock *MBB = getParent()) {
>     MF = MBB->getParent();
>     if (!TM && MF)
>       TM = &MF->getTarget();
> +    if (MF)
> +      MRI = &MF->getRegInfo();
>   }
> 
> +  // Save a list of virtual registers.
> +  SmallVector<unsigned, 8> VirtRegs;
> +
>   // Print explicitly defined operands on the left of an assignment syntax.
>   unsigned StartOp = 0, e = getNumOperands();
>   for (; StartOp < e && getOperand(StartOp).isReg() &&
> @@ -1250,6 +1256,9 @@
>        ++StartOp) {
>     if (StartOp != 0) OS << ", ";
>     getOperand(StartOp).print(OS, TM);
> +    unsigned Reg = getOperand(StartOp).getReg();
> +    if (Reg && TargetRegisterInfo::isVirtualRegister(Reg))
> +      VirtRegs.push_back(Reg);
>   }
> 
>   if (StartOp != 0)
> @@ -1264,6 +1273,10 @@
>   for (unsigned i = StartOp, e = getNumOperands(); i != e; ++i) {
>     const MachineOperand &MO = getOperand(i);
> 
> +    if (MO.isReg() && MO.getReg() &&
> +        TargetRegisterInfo::isVirtualRegister(MO.getReg()))
> +      VirtRegs.push_back(MO.getReg());
> +
>     // Omit call-clobbered registers which aren't used anywhere. This makes
>     // call instructions much less noisy on targets where calls clobber lots
>     // of registers. Don't rely on MO.isDead() because we may be called before
> @@ -1330,6 +1343,24 @@
>     }
>   }
> 
> +  // Print the regclass of any virtual registers encountered.
> +  if (MRI && !VirtRegs.empty()) {
> +    if (!HaveSemi) OS << ";"; HaveSemi = true;
> +    for (unsigned i = 0; i != VirtRegs.size(); ++i) {
> +      const TargetRegisterClass *RC = MRI->getRegClass(VirtRegs[i]);
> +      OS << " " << RC->getName() << ":%reg" << VirtRegs[i];
> +      for (unsigned j = i+1; j != VirtRegs.size();) {
> +        if (MRI->getRegClass(VirtRegs[j]) != RC) {
> +          ++j;
> +          continue;
> +        }
> +        if (VirtRegs[i] != VirtRegs[j])
> +          OS << "," << VirtRegs[j];
> +        VirtRegs.erase(VirtRegs.begin()+j);
> +      }
> +    }
> +  }
> +
>   if (!debugLoc.isUnknown() && MF) {
>     if (!HaveSemi) OS << ";";
>     OS << " dbg:";
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits




From bruno.cardoso at gmail.com  Fri Jul 30 11:21:59 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 09:21:59 -0700
Subject: [llvm-commits] [llvm] r109434 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
In-Reply-To: <716F6A74-8A65-4512-9848-DDECE6D02179@apple.com>
References: <20100726210118.BFB9B2A6C12C@llvm.org>
	<716F6A74-8A65-4512-9848-DDECE6D02179@apple.com>
Message-ID: <AANLkTi=sBsKrGqP5Of1AErE7-cF3yCUoqn+dpvOh+a4g@mail.gmail.com>

This is actually a typo, thanks Jim

On Friday, July 30, 2010, Jim Grosbach <grosbach at apple.com> wrote:
> Minor thing. Should work as-is, but the predicates should be using "&&" rather than "&".
>
> -Jim
>
> On Jul 26, 2010, at 2:01 PM, Bruno Cardoso Lopes wrote:
>
>> Author: bruno
>> Date: Mon Jul 26 16:01:18 2010
>> New Revision: 109434
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=109434&view=rev
>> Log:
>> Temporary hack to let codegen assert or generate poor code in case
>> we are using AVX and no AVX version of the desired intruction is present,
>> this is better for incremental dev (without fallbacks it's easier to spot
>> what's missing). Not sure this is the best hack thought (we can also disable
>> all HasSSE* predicates by dinamically marking them 'false' if AVX is present)
>>
>>
>> Modified:
>> ? ?llvm/trunk/lib/Target/X86/X86InstrInfo.td
>>
>> Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=109434&r1=109433&r2=109434&view=diff
>> ==============================================================================
>> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
>> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon Jul 26 16:01:18 2010
>> @@ -348,14 +348,19 @@
>> // X86 Instruction Predicate Definitions.
>> def HasCMov ? ? ?: Predicate<"Subtarget->hasCMov()">;
>> def NoCMov ? ? ? : Predicate<"!Subtarget->hasCMov()">;
>> -def HasMMX ? ? ? : Predicate<"Subtarget->hasMMX()">;
>> -def HasSSE1 ? ? ?: Predicate<"Subtarget->hasSSE1()">;
>> -def HasSSE2 ? ? ?: Predicate<"Subtarget->hasSSE2()">;
>> -def HasSSE3 ? ? ?: Predicate<"Subtarget->hasSSE3()">;
>> -def HasSSSE3 ? ? : Predicate<"Subtarget->hasSSSE3()">;
>> -def HasSSE41 ? ? : Predicate<"Subtarget->hasSSE41()">;
>> -def HasSSE42 ? ? : Predicate<"Subtarget->hasSSE42()">;
>> -def HasSSE4A ? ? : Predicate<"Subtarget->hasSSE4A()">;
>> +
>> +// FIXME: temporary hack to let codegen assert or generate poor code in case
>> +// no AVX version of the desired intructions is present, this is better for
>> +// incremental dev (without fallbacks it's easier to spot what's missing)
>> +def HasMMX ? ? ? : Predicate<"Subtarget->hasMMX() & !Subtarget->hasAVX()">;
>> +def HasSSE1 ? ? ?: Predicate<"Subtarget->hasSSE1() & !Subtarget->hasAVX()">;
>> +def HasSSE2 ? ? ?: Predicate<"Subtarget->hasSSE2() & !Subtarget->hasAVX()">;
>> +def HasSSE3 ? ? ?: Predicate<"Subtarget->hasSSE3() & !Subtarget->hasAVX()">;
>> +def HasSSSE3 ? ? : Predicate<"Subtarget->hasSSSE3() & !Subtarget->hasAVX()">;
>> +def HasSSE41 ? ? : Predicate<"Subtarget->hasSSE41() & !Subtarget->hasAVX()">;
>> +def HasSSE42 ? ? : Predicate<"Subtarget->hasSSE42() & !Subtarget->hasAVX()">;
>> +def HasSSE4A ? ? : Predicate<"Subtarget->hasSSE4A() & !Subtarget->hasAVX()">;
>> +
>> def HasAVX ? ? ? : Predicate<"Subtarget->hasAVX()">;
>> def HasCLMUL ? ? : Predicate<"Subtarget->hasCLMUL()">;
>> def HasFMA3 ? ? ?: Predicate<"Subtarget->hasFMA3()">;
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>

-- 
Bruno Cardoso Lopes
http://www.brunocardoso.cc



From criswell at uiuc.edu  Fri Jul 30 11:30:28 2010
From: criswell at uiuc.edu (John Criswell)
Date: Fri, 30 Jul 2010 16:30:28 -0000
Subject: [llvm-commits] [llvm] r109863 - /llvm/trunk/docs/LangRef.html
Message-ID: <20100730163029.23A7D2A6C12C@llvm.org>

Author: criswell
Date: Fri Jul 30 11:30:28 2010
New Revision: 109863

URL: http://llvm.org/viewvc/llvm-project?rev=109863&view=rev
Log:
Grammar cop pullover: Corrected and improved some grammar in the description of
the llvm.memset() intrinsic family.
No content changes.

Modified:
    llvm/trunk/docs/LangRef.html

Modified: llvm/trunk/docs/LangRef.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/LangRef.html?rev=109863&r1=109862&r2=109863&view=diff
==============================================================================
--- llvm/trunk/docs/LangRef.html (original)
+++ llvm/trunk/docs/LangRef.html Fri Jul 30 11:30:28 2010
@@ -6119,8 +6119,8 @@
 
 <h5>Syntax:</h5>
 <p>This is an overloaded intrinsic. You can use llvm.memset on any integer bit
-   width and for different address spaces. Not all targets support all bit
-   widths however.</p>
+   width and for different address spaces. However, not all targets support all
+   bit widths.</p>
 
 <pre>
   declare void @llvm.memset.p0i8.i32(i8* &lt;dest&gt;, i8 &lt;val&gt;,
@@ -6134,14 +6134,14 @@
    particular byte value.</p>
 
 <p>Note that, unlike the standard libc function, the <tt>llvm.memset</tt>
-   intrinsic does not return a value, takes extra alignment/volatile arguments,
-   and the destination can be in an arbitrary address space.</p>
+   intrinsic does not return a value and takes extra alignment/volatile
+   arguments.  Also, the destination can be in an arbitrary address space.</p>
 
 <h5>Arguments:</h5>
 <p>The first argument is a pointer to the destination to fill, the second is the
-   byte value to fill it with, the third argument is an integer argument
+   byte value with which to fill it, the third argument is an integer argument
    specifying the number of bytes to fill, and the fourth argument is the known
-   alignment of destination location.</p>
+   alignment of the destination location.</p>
 
 <p>If the call to this intrinsic has an alignment value that is not 0 or 1,
    then the caller guarantees that the destination pointer is aligned to that




From evan.cheng at apple.com  Fri Jul 30 11:57:10 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Fri, 30 Jul 2010 09:57:10 -0700
Subject: [llvm-commits] [patch] - Bugfix for
	TargetLowering::SimplifySetCC
In-Reply-To: <AANLkTimxoYmVPq5QYe4=ddXAa6eX0RAZOGwWv4QiA=+V@mail.gmail.com>
References: <000001caf81e$c614ce00$523e6a00$%muller@arm.com>
	<CAA7B1D6-E805-42FD-9123-40D10380C513@apple.com>
	<AANLkTimxoYmVPq5QYe4=ddXAa6eX0RAZOGwWv4QiA=+V@mail.gmail.com>
Message-ID: <E3168495-AB39-47E6-8164-F243FDD7AC7F@apple.com>

Thanks Eli.

Evan
On Jul 29, 2010, at 11:50 PM, Eli Friedman wrote:

> Not actually my code (I only reindented it); in any case, fixed in r109854.
> 
> -Eli
> 
> On Thu, Jul 29, 2010 at 10:52 PM, Evan Cheng <evan.cheng at apple.com> wrote:
>> Sorry I completely missed this patch. I agree there is a bug here but I don't know this code well enough to determine if this is the right fix. Eli, you wrote the optimization. Do you mind take a look at this?
>> 
>> Evan
>> 
>> On May 20, 2010, at 6:17 AM, Evzen Muller wrote:
>> 
>>> Hi all,
>>> 
>>> When I compile this code (clang generates something similar for bitfield
>>> access):
>>> 
>>> target datalayout = "e-p:32:32-f64:32:32-i64:32:32-n32"
>>> target triple = "arm-none-eabi"
>>> 
>>> define i32 @extend2bit_v2(i32 %val) {
>>> entry:
>>>  %0 = trunc i32 %val to i2                       ; <i2> [#uses=1]
>>>  %1 = sext i2 %0 to i32                          ; <i32> [#uses=1]
>>>  %2 = icmp eq i32 %1, 3                          ; <i1> [#uses=1]
>>>  %3 = zext i1 %2 to i32                          ; <i32> [#uses=1]
>>>  ret i32 %3
>>> }
>>> 
>>> ARM codegen produces:
>>> @ BB#0:                                 @ %entry
>>>        and     r1, r0, #3
>>>        mov     r0, #0
>>>        cmp     r1, #3
>>>        moveq   r0, #1
>>>        bx      lr
>>> 
>>> This is not correct, because when %val == 3 this will return 1, but expected
>>> result is 0 because i2 constant (3) should be sign-extended before icmp.
>>> 
>>> There is an optimization in TargetLowering::SimplifySetCC (around line 1790)
>>> which works fine if C1 has more or less active bits than ExtSrcTyBits, but I
>>> think that if C1.getActiveBits() == ExtSrcTyBits we should do normal
>>> truncate/sign-extend.
>>> 
>>> Attached patch should fix this problem.
>>> 
>>> Thanks
>>> Evzen
>>> 
>>> <simplify_setcc_bugfix.patch>_______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>> 
>> 




From krister.walfridsson at gmail.com  Fri Jul 30 12:32:46 2010
From: krister.walfridsson at gmail.com (Krister Walfridsson)
Date: Fri, 30 Jul 2010 19:32:46 +0200 (CEST)
Subject: [llvm-commits] PATCH: Fix 2004-08-12-InlinerAndAllocas.c for NetBSD
Message-ID: <Pine.NEB.4.64.1007301931510.17143@pc20.kwa>

NetBSD does not have alloca.h, so it need to be handled in the same way as
FreeBSD/OpenBSD.

    /Krister
-------------- next part --------------
Index: SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c
===================================================================
--- SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c	(revision 109871)
+++ SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c	(working copy)
@@ -1,7 +1,7 @@
 // A compiler cannot inline Callee into main unless it is prepared to reclaim
 // the stack memory allocated in it.
 
-#if defined(__FreeBSD__) || defined(__OpenBSD__)
+#if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__)
 #include <stdlib.h>
 #else
 #include <alloca.h>

From daniel at zuster.org  Fri Jul 30 12:49:05 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 17:49:05 -0000
Subject: [llvm-commits] [llvm] r109872 -
	/llvm/trunk/lib/Support/CrashRecoveryContext.cpp
Message-ID: <20100730174905.0FF002A6C12C@llvm.org>

Author: ddunbar
Date: Fri Jul 30 12:49:04 2010
New Revision: 109872

URL: http://llvm.org/viewvc/llvm-project?rev=109872&view=rev
Log:
Fix -Wmissing-field-initializers warnings.

Modified:
    llvm/trunk/lib/Support/CrashRecoveryContext.cpp

Modified: llvm/trunk/lib/Support/CrashRecoveryContext.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/CrashRecoveryContext.cpp?rev=109872&r1=109871&r2=109872&view=diff
==============================================================================
--- llvm/trunk/lib/Support/CrashRecoveryContext.cpp (original)
+++ llvm/trunk/lib/Support/CrashRecoveryContext.cpp Fri Jul 30 12:49:04 2010
@@ -87,18 +87,9 @@
 
 #include <signal.h>
 
-static struct {
-  int Signal;
-  struct sigaction PrevAction;
-} SignalInfo[] = {
-  { SIGABRT, {} },
-  { SIGBUS,  {} },
-  { SIGFPE,  {} },
-  { SIGILL,  {} },
-  { SIGSEGV, {} },
-  { SIGTRAP, {} },
-};
-static const unsigned NumSignals = sizeof(SignalInfo) / sizeof(SignalInfo[0]);
+static int Signals[] = { SIGABRT, SIGBUS, SIGFPE, SIGILL, SIGSEGV, SIGTRAP };
+static const unsigned NumSignals = sizeof(Signals) / sizeof(Signals[0]);
+static struct sigaction PrevActions[NumSignals];
 
 static void CrashRecoverySignalHandler(int Signal) {
   // Lookup the current thread local recovery object.
@@ -142,8 +133,7 @@
   sigemptyset(&Handler.sa_mask);
 
   for (unsigned i = 0; i != NumSignals; ++i) {
-    sigaction(SignalInfo[i].Signal, &Handler,
-              &SignalInfo[i].PrevAction);
+    sigaction(Signals[i], &Handler, &PrevActions[i]);
   }
 }
 
@@ -155,7 +145,7 @@
 
   // Restore the previous signal handlers.
   for (unsigned i = 0; i != NumSignals; ++i)
-    sigaction(SignalInfo[i].Signal, &SignalInfo[i].PrevAction, 0);
+    sigaction(Signals[i], &PrevActions[i], 0);
 }
 
 #endif




From bruno.cardoso at gmail.com  Fri Jul 30 13:15:34 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 18:15:34 -0000
Subject: [llvm-commits] [test-suite] r109873 -
 /test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c
Message-ID: <20100730181534.26C402A6C12C@llvm.org>

Author: bruno
Date: Fri Jul 30 13:15:33 2010
New Revision: 109873

URL: http://llvm.org/viewvc/llvm-project?rev=109873&view=rev
Log:
Handle NetBSD in the same way as FreeBSD/OpenBSD. Patch by Krister Walfridsson

Modified:
    test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c

Modified: test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c?rev=109873&r1=109872&r2=109873&view=diff
==============================================================================
--- test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c (original)
+++ test-suite/trunk/SingleSource/Regression/C/2004-08-12-InlinerAndAllocas.c Fri Jul 30 13:15:33 2010
@@ -1,7 +1,7 @@
 // A compiler cannot inline Callee into main unless it is prepared to reclaim
 // the stack memory allocated in it.
 
-#if defined(__FreeBSD__) || defined(__OpenBSD__)
+#if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__)
 #include <stdlib.h>
 #else
 #include <alloca.h>




From bruno.cardoso at gmail.com  Fri Jul 30 13:16:07 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 11:16:07 -0700
Subject: [llvm-commits] PATCH: Fix 2004-08-12-InlinerAndAllocas.c for
	NetBSD
In-Reply-To: <Pine.NEB.4.64.1007301931510.17143@pc20.kwa>
References: <Pine.NEB.4.64.1007301931510.17143@pc20.kwa>
Message-ID: <AANLkTi=2kFF9fj6-ppHaGjkAcao6mt=ujQTEujb-M-Yk@mail.gmail.com>

Commited in r109873. Thanks!

On Fri, Jul 30, 2010 at 10:32 AM, Krister Walfridsson
<krister.walfridsson at gmail.com> wrote:
> NetBSD does not have alloca.h, so it need to be handled in the same way as
> FreeBSD/OpenBSD.
>
> ? /Krister
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



-- 
Bruno Cardoso Lopes
http://www.brunocardoso.cc



From daniel at zuster.org  Fri Jul 30 13:30:31 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 11:30:31 -0700
Subject: [llvm-commits] COFF patch
In-Reply-To: <B7F7863E-3D87-41DB-AFC0-B13206374BAE@apple.com>
References: <B7F7863E-3D87-41DB-AFC0-B13206374BAE@apple.com>
Message-ID: <AANLkTikU8kS3ot3HNPwbj3Kvn-EQLNCs_1cCqqvFnneL@mail.gmail.com>

On Thu, Jul 29, 2010 at 4:56 PM, Cameron Esfahani <dirty at apple.com> wrote:
> My first patch, so please be gentle...
>
> This fixes a problem I was seeing with the COFF back end. ?All of the string data was getting put into one data fragment, and so none of the fixup references were correct.

I'm not sure that this is the right fix. The fixups should be able to
deal with symbols being embedded inside a fragment via the Offset
field. In general, we don't want to create any more fragments than
necessary because they have a large impact on the assembly time. I
haven't looked at the COFF streamer or writer in detail, some I can't
speak to where the real bug is.

> I also made sure the TimeDateStamp in the COFF header had a correct value.

Ok, I assume this is fine but I don't know anything about COFF.

> I also fixed a spot in MCFragment where Offset wasn't being properly initialized.

I'm going to cherry pick this hunk.

 - Daniel

> This patch was generated against revision 109821.
>
>
>
>
>
> Cameron Esfahani
> dirty at apple.com
>
> "I cannot for the life of me understand why, while people without driver's licenses are not allowed on public roads, in bookstores one can find any number of books by persons without decency - let alone knowledge."
>
> "His Master's Voice", Stanislaw Lem
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



From daniel at zuster.org  Fri Jul 30 13:32:09 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 18:32:09 -0000
Subject: [llvm-commits] [llvm] r109875 - /llvm/trunk/lib/MC/MCAssembler.cpp
Message-ID: <20100730183209.7D5AB2A6C12C@llvm.org>

Author: ddunbar
Date: Fri Jul 30 13:32:09 2010
New Revision: 109875

URL: http://llvm.org/viewvc/llvm-project?rev=109875&view=rev
Log:
MC: Initialize MCFragment::Offset, noticed by Cameron Esfahani.

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

Modified: llvm/trunk/lib/MC/MCAssembler.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/MCAssembler.cpp?rev=109875&r1=109874&r2=109875&view=diff
==============================================================================
--- llvm/trunk/lib/MC/MCAssembler.cpp (original)
+++ llvm/trunk/lib/MC/MCAssembler.cpp Fri Jul 30 13:32:09 2010
@@ -182,7 +182,8 @@
 }
 
 MCFragment::MCFragment(FragmentType _Kind, MCSectionData *_Parent)
-  : Kind(_Kind), Parent(_Parent), Atom(0), EffectiveSize(~UINT64_C(0))
+  : Kind(_Kind), Parent(_Parent), Atom(0), Offset(~UINT64_C(0)),
+    EffectiveSize(~UINT64_C(0))
 {
   if (Parent)
     Parent->getFragmentList().push_back(this);




From bob.wilson at apple.com  Fri Jul 30 13:52:48 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Fri, 30 Jul 2010 18:52:48 -0000
Subject: [llvm-commits] [llvm] r109876 - /llvm/trunk/test/FrontendC/vla-1.c
Message-ID: <20100730185248.0F1ED2A6C12C@llvm.org>

Author: bwilson
Date: Fri Jul 30 13:52:47 2010
New Revision: 109876

URL: http://llvm.org/viewvc/llvm-project?rev=109876&view=rev
Log:
Attempt to fix the llvm-gcc-powerpc-darwin9 buildbot.

Modified:
    llvm/trunk/test/FrontendC/vla-1.c

Modified: llvm/trunk/test/FrontendC/vla-1.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/vla-1.c?rev=109876&r1=109875&r2=109876&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC/vla-1.c (original)
+++ llvm/trunk/test/FrontendC/vla-1.c Fri Jul 30 13:52:47 2010
@@ -1,4 +1,4 @@
-// RUN: %llvmgcc_only -std=gnu99 %s -S |& grep "warning: alignment for"
+// RUN: %llvmgcc_only -std=gnu99 %s -S |& grep {warning: alignment for}
 
 int foo(int a)
 {




From bruno.cardoso at gmail.com  Fri Jul 30 14:41:24 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 19:41:24 -0000
Subject: [llvm-commits] [llvm] r109877 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
Message-ID: <20100730194124.E91F32A6C12C@llvm.org>

Author: bruno
Date: Fri Jul 30 14:41:24 2010
New Revision: 109877

URL: http://llvm.org/viewvc/llvm-project?rev=109877&view=rev
Log:
Fix typo!

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

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=109877&r1=109876&r2=109877&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Fri Jul 30 14:41:24 2010
@@ -352,14 +352,14 @@
 // FIXME: temporary hack to let codegen assert or generate poor code in case
 // no AVX version of the desired intructions is present, this is better for
 // incremental dev (without fallbacks it's easier to spot what's missing)
-def HasMMX       : Predicate<"Subtarget->hasMMX() & !Subtarget->hasAVX()">;
-def HasSSE1      : Predicate<"Subtarget->hasSSE1() & !Subtarget->hasAVX()">;
-def HasSSE2      : Predicate<"Subtarget->hasSSE2() & !Subtarget->hasAVX()">;
-def HasSSE3      : Predicate<"Subtarget->hasSSE3() & !Subtarget->hasAVX()">;
-def HasSSSE3     : Predicate<"Subtarget->hasSSSE3() & !Subtarget->hasAVX()">;
-def HasSSE41     : Predicate<"Subtarget->hasSSE41() & !Subtarget->hasAVX()">;
-def HasSSE42     : Predicate<"Subtarget->hasSSE42() & !Subtarget->hasAVX()">;
-def HasSSE4A     : Predicate<"Subtarget->hasSSE4A() & !Subtarget->hasAVX()">;
+def HasMMX       : Predicate<"Subtarget->hasMMX() && !Subtarget->hasAVX()">;
+def HasSSE1      : Predicate<"Subtarget->hasSSE1() && !Subtarget->hasAVX()">;
+def HasSSE2      : Predicate<"Subtarget->hasSSE2() && !Subtarget->hasAVX()">;
+def HasSSE3      : Predicate<"Subtarget->hasSSE3() && !Subtarget->hasAVX()">;
+def HasSSSE3     : Predicate<"Subtarget->hasSSSE3() && !Subtarget->hasAVX()">;
+def HasSSE41     : Predicate<"Subtarget->hasSSE41() && !Subtarget->hasAVX()">;
+def HasSSE42     : Predicate<"Subtarget->hasSSE42() && !Subtarget->hasAVX()">;
+def HasSSE4A     : Predicate<"Subtarget->hasSSE4A() && !Subtarget->hasAVX()">;
 
 def HasAVX       : Predicate<"Subtarget->hasAVX()">;
 def HasCLMUL     : Predicate<"Subtarget->hasCLMUL()">;




From bruno.cardoso at gmail.com  Fri Jul 30 14:54:34 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 19:54:34 -0000
Subject: [llvm-commits] [llvm] r109878 - in /llvm/trunk/lib/Target/X86:
 X86ISelLowering.cpp X86Instr64bit.td X86InstrSSE.td
Message-ID: <20100730195434.51A122A6C12C@llvm.org>

Author: bruno
Date: Fri Jul 30 14:54:33 2010
New Revision: 109878

URL: http://llvm.org/viewvc/llvm-project?rev=109878&view=rev
Log:
Support all 128-bit AVX vector intrinsics. Most part of them I already
declared during the addition of the assembler support, the additional
changes are:
- Add missing intrinsics
- Move all SSE conversion instructions in X86InstInfo64.td to the SSE.td file.
- Duplicate some patterns to AVX mode.
- Step into PCMPEST/PCMPIST custom inserter and add AVX versions.


Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/lib/Target/X86/X86Instr64bit.td
    llvm/trunk/lib/Target/X86/X86InstrSSE.td

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=109878&r1=109877&r2=109878&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Jul 30 14:54:33 2010
@@ -8543,19 +8543,31 @@
 }
 
 // FIXME: When we get size specific XMM0 registers, i.e. XMM0_V16I8
-// all of this code can be replaced with that in the .td file.
+// or XMM0_V32I8 in AVX all of this code can be replaced with that
+// in the .td file.
 MachineBasicBlock *
 X86TargetLowering::EmitPCMP(MachineInstr *MI, MachineBasicBlock *BB,
                             unsigned numArgs, bool memArg) const {
 
+  assert((Subtarget->hasSSE42() || Subtarget->hasAVX()) &&
+         "Target must have SSE4.2 or AVX features enabled");
+
   DebugLoc dl = MI->getDebugLoc();
   const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
 
   unsigned Opc;
-  if (memArg)
-    Opc = numArgs == 3 ? X86::PCMPISTRM128rm : X86::PCMPESTRM128rm;
-  else
-    Opc = numArgs == 3 ? X86::PCMPISTRM128rr : X86::PCMPESTRM128rr;
+
+  if (!Subtarget->hasAVX()) {
+    if (memArg)
+      Opc = numArgs == 3 ? X86::PCMPISTRM128rm : X86::PCMPESTRM128rm;
+    else
+      Opc = numArgs == 3 ? X86::PCMPISTRM128rr : X86::PCMPESTRM128rr;
+  } else {
+    if (memArg)
+      Opc = numArgs == 3 ? X86::VPCMPISTRM128rm : X86::VPCMPESTRM128rm;
+    else
+      Opc = numArgs == 3 ? X86::VPCMPISTRM128rr : X86::VPCMPESTRM128rr;
+  }
 
   MachineInstrBuilder MIB = BuildMI(BB, dl, TII->get(Opc));
 
@@ -8902,12 +8914,16 @@
   }
     // String/text processing lowering.
   case X86::PCMPISTRM128REG:
+  case X86::VPCMPISTRM128REG:
     return EmitPCMP(MI, BB, 3, false /* in-mem */);
   case X86::PCMPISTRM128MEM:
+  case X86::VPCMPISTRM128MEM:
     return EmitPCMP(MI, BB, 3, true /* in-mem */);
   case X86::PCMPESTRM128REG:
+  case X86::VPCMPESTRM128REG:
     return EmitPCMP(MI, BB, 5, false /* in mem */);
   case X86::PCMPESTRM128MEM:
+  case X86::VPCMPESTRM128MEM:
     return EmitPCMP(MI, BB, 5, true /* in mem */);
 
     // Atomic Lowering.

Modified: llvm/trunk/lib/Target/X86/X86Instr64bit.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Instr64bit.td?rev=109878&r1=109877&r2=109878&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86Instr64bit.td (original)
+++ llvm/trunk/lib/Target/X86/X86Instr64bit.td Fri Jul 30 14:54:33 2010
@@ -1540,116 +1540,6 @@
           (SETB_C64r)>;
 
 //===----------------------------------------------------------------------===//
-//  Conversion Instructions...
-//
-
-// f64 -> signed i64
-def CVTSD2SI64rr: RSDI<0x2D, MRMSrcReg, (outs GR64:$dst), (ins FR64:$src),
-                       "cvtsd2si{q}\t{$src, $dst|$dst, $src}", []>;
-def CVTSD2SI64rm: RSDI<0x2D, MRMSrcMem, (outs GR64:$dst), (ins f64mem:$src),
-                       "cvtsd2si{q}\t{$src, $dst|$dst, $src}", []>;
-def Int_CVTSD2SI64rr: RSDI<0x2D, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
-                           "cvtsd2si{q}\t{$src, $dst|$dst, $src}",
-                           [(set GR64:$dst,
-                             (int_x86_sse2_cvtsd2si64 VR128:$src))]>;
-def Int_CVTSD2SI64rm: RSDI<0x2D, MRMSrcMem, (outs GR64:$dst), 
-                           (ins f128mem:$src),
-                           "cvtsd2si{q}\t{$src, $dst|$dst, $src}",
-                           [(set GR64:$dst, (int_x86_sse2_cvtsd2si64
-                                             (load addr:$src)))]>;
-def CVTTSD2SI64rr: RSDI<0x2C, MRMSrcReg, (outs GR64:$dst), (ins FR64:$src),
-                        "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
-                        [(set GR64:$dst, (fp_to_sint FR64:$src))]>;
-def CVTTSD2SI64rm: RSDI<0x2C, MRMSrcMem, (outs GR64:$dst), (ins f64mem:$src),
-                        "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
-                        [(set GR64:$dst, (fp_to_sint (loadf64 addr:$src)))]>;
-def Int_CVTTSD2SI64rr: RSDI<0x2C, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
-                            "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
-                            [(set GR64:$dst,
-                              (int_x86_sse2_cvttsd2si64 VR128:$src))]>;
-def Int_CVTTSD2SI64rm: RSDI<0x2C, MRMSrcMem, (outs GR64:$dst), 
-                            (ins f128mem:$src),
-                            "cvttsd2si{q}\t{$src, $dst|$dst, $src}",
-                            [(set GR64:$dst,
-                              (int_x86_sse2_cvttsd2si64
-                               (load addr:$src)))]>;
-
-// Signed i64 -> f64
-def CVTSI2SD64rr: RSDI<0x2A, MRMSrcReg, (outs FR64:$dst), (ins GR64:$src),
-                       "cvtsi2sd{q}\t{$src, $dst|$dst, $src}",
-                       [(set FR64:$dst, (sint_to_fp GR64:$src))]>;
-def CVTSI2SD64rm: RSDI<0x2A, MRMSrcMem, (outs FR64:$dst), (ins i64mem:$src),
-                       "cvtsi2sd{q}\t{$src, $dst|$dst, $src}",
-                       [(set FR64:$dst, (sint_to_fp (loadi64 addr:$src)))]>;
-
-let Constraints = "$src1 = $dst" in {
-def Int_CVTSI2SD64rr: RSDI<0x2A, MRMSrcReg,
-                           (outs VR128:$dst), (ins VR128:$src1, GR64:$src2),
-                           "cvtsi2sd{q}\t{$src2, $dst|$dst, $src2}",
-                           [(set VR128:$dst,
-                             (int_x86_sse2_cvtsi642sd VR128:$src1,
-                              GR64:$src2))]>;
-def Int_CVTSI2SD64rm: RSDI<0x2A, MRMSrcMem,
-                           (outs VR128:$dst), (ins VR128:$src1, i64mem:$src2),
-                           "cvtsi2sd{q}\t{$src2, $dst|$dst, $src2}",
-                           [(set VR128:$dst,
-                             (int_x86_sse2_cvtsi642sd VR128:$src1,
-                              (loadi64 addr:$src2)))]>;
-} // Constraints = "$src1 = $dst"
-
-// Signed i64 -> f32
-def CVTSI2SS64rr: RSSI<0x2A, MRMSrcReg, (outs FR32:$dst), (ins GR64:$src),
-                       "cvtsi2ss{q}\t{$src, $dst|$dst, $src}",
-                       [(set FR32:$dst, (sint_to_fp GR64:$src))]>;
-def CVTSI2SS64rm: RSSI<0x2A, MRMSrcMem, (outs FR32:$dst), (ins i64mem:$src),
-                       "cvtsi2ss{q}\t{$src, $dst|$dst, $src}",
-                       [(set FR32:$dst, (sint_to_fp (loadi64 addr:$src)))]>;
-
-let Constraints = "$src1 = $dst" in {
-  def Int_CVTSI2SS64rr : RSSI<0x2A, MRMSrcReg,
-                              (outs VR128:$dst), (ins VR128:$src1, GR64:$src2),
-                              "cvtsi2ss{q}\t{$src2, $dst|$dst, $src2}",
-                              [(set VR128:$dst,
-                                (int_x86_sse_cvtsi642ss VR128:$src1,
-                                 GR64:$src2))]>;
-  def Int_CVTSI2SS64rm : RSSI<0x2A, MRMSrcMem,
-                              (outs VR128:$dst), 
-                              (ins VR128:$src1, i64mem:$src2),
-                              "cvtsi2ss{q}\t{$src2, $dst|$dst, $src2}",
-                              [(set VR128:$dst,
-                                (int_x86_sse_cvtsi642ss VR128:$src1,
-                                 (loadi64 addr:$src2)))]>;
-} // Constraints = "$src1 = $dst"
-
-// f32 -> signed i64
-def CVTSS2SI64rr: RSSI<0x2D, MRMSrcReg, (outs GR64:$dst), (ins FR32:$src),
-                       "cvtss2si{q}\t{$src, $dst|$dst, $src}", []>;
-def CVTSS2SI64rm: RSSI<0x2D, MRMSrcMem, (outs GR64:$dst), (ins f32mem:$src),
-                       "cvtss2si{q}\t{$src, $dst|$dst, $src}", []>;
-def Int_CVTSS2SI64rr: RSSI<0x2D, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
-                           "cvtss2si{q}\t{$src, $dst|$dst, $src}",
-                           [(set GR64:$dst,
-                             (int_x86_sse_cvtss2si64 VR128:$src))]>;
-def Int_CVTSS2SI64rm: RSSI<0x2D, MRMSrcMem, (outs GR64:$dst), (ins f32mem:$src),
-                           "cvtss2si{q}\t{$src, $dst|$dst, $src}",
-                           [(set GR64:$dst, (int_x86_sse_cvtss2si64
-                                             (load addr:$src)))]>;
-def CVTTSS2SI64rr: RSSI<0x2C, MRMSrcReg, (outs GR64:$dst), (ins FR32:$src),
-                        "cvttss2si{q}\t{$src, $dst|$dst, $src}",
-                        [(set GR64:$dst, (fp_to_sint FR32:$src))]>;
-def CVTTSS2SI64rm: RSSI<0x2C, MRMSrcMem, (outs GR64:$dst), (ins f32mem:$src),
-                        "cvttss2si{q}\t{$src, $dst|$dst, $src}",
-                        [(set GR64:$dst, (fp_to_sint (loadf32 addr:$src)))]>;
-def Int_CVTTSS2SI64rr: RSSI<0x2C, MRMSrcReg, (outs GR64:$dst), (ins VR128:$src),
-                            "cvttss2si{q}\t{$src, $dst|$dst, $src}",
-                            [(set GR64:$dst,
-                              (int_x86_sse_cvttss2si64 VR128:$src))]>;
-def Int_CVTTSS2SI64rm: RSSI<0x2C, MRMSrcMem, (outs GR64:$dst),
-                            (ins f32mem:$src),
-                            "cvttss2si{q}\t{$src, $dst|$dst, $src}",
-                            [(set GR64:$dst,
-                              (int_x86_sse_cvttss2si64 (load addr:$src)))]>;
-                              
 // Descriptor-table support instructions
 
 // LLDT is not interpreted specially in 64-bit mode because there is no sign

Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=109878&r1=109877&r2=109878&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Fri Jul 30 14:54:33 2010
@@ -543,41 +543,49 @@
 }
 
 let isAsmParserOnly = 1 in {
-defm VCVTTSS2SI    : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
-                                 "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX;
-defm VCVTTSS2SIr64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
-                                 "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX,
-                                 VEX_W;
-defm VCVTTSD2SI    : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
-                                 "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, VEX;
-defm VCVTTSD2SIr64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
-                                 "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, VEX,
-                                 VEX_W;
+defm VCVTTSS2SI   : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
+                                "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX;
+defm VCVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
+                                "cvttss2si\t{$src, $dst|$dst, $src}">, XS, VEX,
+                                VEX_W;
+defm VCVTTSD2SI   : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
+                                "cvttsd2si\t{$src, $dst|$dst, $src}">, XD, VEX;
+defm VCVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
+                                "cvttsd2si\t{$src, $dst|$dst, $src}">, XD,
+                                VEX, VEX_W;
 
 // The assembler can recognize rr 64-bit instructions by seeing a rxx
 // register, but the same isn't true when only using memory operands,
 // provide other assembly "l" and "q" forms to address this explicitly
 // where appropriate to do so.
-defm VCVTSI2SS  : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">, XS,
-                                 VEX_4V;
-defm VCVTSI2SSQ : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ssq">, XS,
-                                 VEX_4V, VEX_W;
-defm VCVTSI2SD  : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">, XD,
-                                 VEX_4V;
-defm VCVTSI2SDL : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sdl">, XD,
-                                 VEX_4V;
-defm VCVTSI2SDQ : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sdq">, XD,
-                                 VEX_4V, VEX_W;
+defm VCVTSI2SS   : sse12_vcvt_avx<0x2A, GR32, FR32, i32mem, "cvtsi2ss">, XS,
+                                  VEX_4V;
+defm VCVTSI2SS64 : sse12_vcvt_avx<0x2A, GR64, FR32, i64mem, "cvtsi2ss{q}">, XS,
+                                  VEX_4V, VEX_W;
+defm VCVTSI2SD   : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd">, XD,
+                                  VEX_4V;
+defm VCVTSI2SDL  : sse12_vcvt_avx<0x2A, GR32, FR64, i32mem, "cvtsi2sd{l}">, XD,
+                                  VEX_4V;
+defm VCVTSI2SD64 : sse12_vcvt_avx<0x2A, GR64, FR64, i64mem, "cvtsi2sd{q}">, XD,
+                                  VEX_4V, VEX_W;
 }
 
 defm CVTTSS2SI : sse12_cvt_s<0x2C, FR32, GR32, fp_to_sint, f32mem, loadf32,
                       "cvttss2si\t{$src, $dst|$dst, $src}">, XS;
+defm CVTTSS2SI64 : sse12_cvt_s<0x2C, FR32, GR64, fp_to_sint, f32mem, loadf32,
+                      "cvttss2si{q}\t{$src, $dst|$dst, $src}">, XS, REX_W;
 defm CVTTSD2SI : sse12_cvt_s<0x2C, FR64, GR32, fp_to_sint, f64mem, loadf64,
                       "cvttsd2si\t{$src, $dst|$dst, $src}">, XD;
+defm CVTTSD2SI64 : sse12_cvt_s<0x2C, FR64, GR64, fp_to_sint, f64mem, loadf64,
+                      "cvttsd2si{q}\t{$src, $dst|$dst, $src}">, XD, REX_W;
 defm CVTSI2SS  : sse12_cvt_s<0x2A, GR32, FR32, sint_to_fp, i32mem, loadi32,
                       "cvtsi2ss\t{$src, $dst|$dst, $src}">, XS;
+defm CVTSI2SS64 : sse12_cvt_s<0x2A, GR64, FR32, sint_to_fp, i64mem, loadi64,
+                      "cvtsi2ss{q}\t{$src, $dst|$dst, $src}">, XS, REX_W;
 defm CVTSI2SD  : sse12_cvt_s<0x2A, GR32, FR64, sint_to_fp, i32mem, loadi32,
                       "cvtsi2sd\t{$src, $dst|$dst, $src}">, XD;
+defm CVTSI2SD64 : sse12_cvt_s<0x2A, GR64, FR64, sint_to_fp, i64mem, loadi64,
+                      "cvtsi2sd{q}\t{$src, $dst|$dst, $src}">, XD, REX_W;
 
 // Conversion Instructions Intrinsics - Match intrinsics which expect MM
 // and/or XMM operand(s).
@@ -593,10 +601,12 @@
 multiclass sse12_cvt_sint<bits<8> opc, RegisterClass SrcRC, RegisterClass DstRC,
                          Intrinsic Int, X86MemOperand x86memop, PatFrag ld_frag,
                          string asm> {
-  def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src), asm,
-                        [(set DstRC:$dst, (Int SrcRC:$src))]>;
-  def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src), asm,
-                        [(set DstRC:$dst, (Int (ld_frag addr:$src)))]>;
+  def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins SrcRC:$src),
+              !strconcat(asm, "\t{$src, $dst|$dst, $src}"),
+              [(set DstRC:$dst, (Int SrcRC:$src))]>;
+  def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst), (ins x86memop:$src),
+              !strconcat(asm, "\t{$src, $dst|$dst, $src}"),
+              [(set DstRC:$dst, (Int (ld_frag addr:$src)))]>;
 }
 
 multiclass sse12_cvt_pint_3addr<bits<8> opc, RegisterClass SrcRC,
@@ -611,21 +621,32 @@
 
 multiclass sse12_cvt_sint_3addr<bits<8> opc, RegisterClass SrcRC,
                     RegisterClass DstRC, Intrinsic Int, X86MemOperand x86memop,
-                    PatFrag ld_frag, string asm> {
+                    PatFrag ld_frag, string asm, bit Is2Addr = 1> {
   def rr : SI<opc, MRMSrcReg, (outs DstRC:$dst), (ins DstRC:$src1, SrcRC:$src2),
-              asm, [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))]>;
+              !if(Is2Addr,
+                  !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
+                  !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
+              [(set DstRC:$dst, (Int DstRC:$src1, SrcRC:$src2))]>;
   def rm : SI<opc, MRMSrcMem, (outs DstRC:$dst),
-                   (ins DstRC:$src1, x86memop:$src2), asm,
+              (ins DstRC:$src1, x86memop:$src2),
+              !if(Is2Addr,
+                  !strconcat(asm, "\t{$src2, $dst|$dst, $src2}"),
+                  !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}")),
               [(set DstRC:$dst, (Int DstRC:$src1, (ld_frag addr:$src2)))]>;
 }
 
 let isAsmParserOnly = 1 in {
   defm Int_VCVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
-                        f32mem, load, "cvtss2si\t{$src, $dst|$dst, $src}">, XS,
-                        VEX;
+                        f32mem, load, "cvtss2si">, XS, VEX;
+  defm Int_VCVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64,
+                          int_x86_sse_cvtss2si64, f32mem, load, "cvtss2si">,
+                          XS, VEX, VEX_W;
   defm Int_VCVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
-                        f128mem, load, "cvtsd2si\t{$src, $dst|$dst, $src}">, XD,
-                        VEX;
+                        f128mem, load, "cvtsd2si">, XD, VEX;
+  defm Int_VCVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64,
+                        int_x86_sse2_cvtsd2si64, f128mem, load, "cvtsd2si">,
+                        XD, VEX, VEX_W;
+
   // FIXME: The asm matcher has a hack to ignore instructions with _Int and Int_
   // Get rid of this hack or rename the intrinsics, there are several
   // intructions that only match with the intrinsic form, why create duplicates
@@ -636,18 +657,43 @@
                         "cvtsd2si\t{$src, $dst|$dst, $src}">, XD, VEX, VEX_W;
 }
 defm Int_CVTSS2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse_cvtss2si,
-                      f32mem, load, "cvtss2si\t{$src, $dst|$dst, $src}">, XS;
+                      f32mem, load, "cvtss2si">, XS;
+defm Int_CVTSS2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse_cvtss2si64,
+                      f32mem, load, "cvtss2si{q}">, XS, REX_W;
 defm Int_CVTSD2SI : sse12_cvt_sint<0x2D, VR128, GR32, int_x86_sse2_cvtsd2si,
-                      f128mem, load, "cvtsd2si\t{$src, $dst|$dst, $src}">, XD;
+                      f128mem, load, "cvtsd2si">, XD;
+defm Int_CVTSD2SI64 : sse12_cvt_sint<0x2D, VR128, GR64, int_x86_sse2_cvtsd2si64,
+                        f128mem, load, "cvtsd2si">, XD, REX_W;
 
+defm CVTSD2SI64 : sse12_cvt_s_np<0x2D, VR128, GR64, f64mem, "cvtsd2si{q}">, XD,
+                        REX_W;
+
+let isAsmParserOnly = 1 in {
+  defm Int_VCVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
+            int_x86_sse_cvtsi2ss, i32mem, loadi32, "cvtsi2ss", 0>, XS, VEX_4V;
+  defm Int_VCVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
+            int_x86_sse_cvtsi642ss, i64mem, loadi64, "cvtsi2ss", 0>, XS, VEX_4V,
+            VEX_W;
+  defm Int_VCVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
+            int_x86_sse2_cvtsi2sd, i32mem, loadi32, "cvtsi2sd", 0>, XD, VEX_4V;
+  defm Int_VCVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
+            int_x86_sse2_cvtsi642sd, i64mem, loadi64, "cvtsi2sd", 0>, XD,
+            VEX_4V, VEX_W;
+}
 
 let Constraints = "$src1 = $dst" in {
   defm Int_CVTSI2SS : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
                         int_x86_sse_cvtsi2ss, i32mem, loadi32,
-                        "cvtsi2ss\t{$src2, $dst|$dst, $src2}">, XS;
+                        "cvtsi2ss">, XS;
+  defm Int_CVTSI2SS64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
+                        int_x86_sse_cvtsi642ss, i64mem, loadi64,
+                        "cvtsi2ss{q}">, XS, REX_W;
   defm Int_CVTSI2SD : sse12_cvt_sint_3addr<0x2A, GR32, VR128,
                         int_x86_sse2_cvtsi2sd, i32mem, loadi32,
-                        "cvtsi2ss\t{$src2, $dst|$dst, $src2}">, XD;
+                        "cvtsi2sd">, XD;
+  defm Int_CVTSI2SD64 : sse12_cvt_sint_3addr<0x2A, GR64, VR128,
+                        int_x86_sse2_cvtsi642sd, i64mem, loadi64,
+                        "cvtsi2sd">, XD, REX_W;
 }
 
 // Instructions below don't have an AVX form.
@@ -676,20 +722,28 @@
 /// SSE 1 Only
 
 // Aliases for intrinsics
-let isAsmParserOnly = 1, Pattern = []<dag> in {
-defm Int_VCVTTSS2SI : sse12_cvt_sint_3addr<0x2C, VR128, GR32,
-                int_x86_sse_cvttss2si, f32mem, load,
-                "cvttss2si\t{$src2, $src1, $dst|$dst, $src1, $src2}">, XS;
-defm Int_VCVTTSD2SI : sse12_cvt_sint_3addr<0x2C, VR128, GR32,
-                int_x86_sse2_cvttsd2si, f128mem, load,
-                "cvttss2si\t{$src2, $src1, $dst|$dst, $src1, $src2}">, XD;
+let isAsmParserOnly = 1 in {
+defm Int_VCVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
+                                    f32mem, load, "cvttss2si">, XS, VEX;
+defm Int_VCVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
+                                    int_x86_sse_cvttss2si64, f32mem, load,
+                                    "cvttss2si">, XS, VEX, VEX_W;
+defm Int_VCVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
+                                    f128mem, load, "cvttss2si">, XD, VEX;
+defm Int_VCVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
+                                    int_x86_sse2_cvttsd2si64, f128mem, load,
+                                    "cvttss2si">, XD, VEX, VEX_W;
 }
 defm Int_CVTTSS2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse_cvttss2si,
-                          f32mem, load, "cvttss2si\t{$src, $dst|$dst, $src}">,
-                          XS;
+                                    f32mem, load, "cvttss2si">, XS;
+defm Int_CVTTSS2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
+                                    int_x86_sse_cvttss2si64, f32mem, load,
+                                    "cvttss2si{q}">, XS, REX_W;
 defm Int_CVTTSD2SI : sse12_cvt_sint<0x2C, VR128, GR32, int_x86_sse2_cvttsd2si,
-                          f128mem, load, "cvttss2si\t{$src, $dst|$dst, $src}">,
-                          XD;
+                                    f128mem, load, "cvttss2si">, XD;
+defm Int_CVTTSD2SI64 : sse12_cvt_sint<0x2C, VR128, GR64,
+                                    int_x86_sse2_cvttsd2si64, f128mem, load,
+                                    "cvttss2si{q}">, XD, REX_W;
 
 let isAsmParserOnly = 1, Pattern = []<dag> in {
 defm VCVTSS2SI   : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load,
@@ -707,6 +761,8 @@
 let Pattern = []<dag> in {
 defm CVTSS2SI : sse12_cvt_s<0x2D, FR32, GR32, undef, f32mem, load /*dummy*/,
                           "cvtss2si{l}\t{$src, $dst|$dst, $src}">, XS;
+defm CVTSS2SI64 : sse12_cvt_s<0x2D, FR32, GR64, undef, f32mem, load /*dummy*/,
+                          "cvtss2si{q}\t{$src, $dst|$dst, $src}">, XS, REX_W;
 defm CVTDQ2PS : sse12_cvt_p<0x5B, VR128, VR128, undef, f128mem, load /*dummy*/,
                             "cvtdq2ps\t{$src, $dst|$dst, $src}",
                             SSEPackedSingle>, TB; /* PD SSE3 form is avaiable */
@@ -735,13 +791,11 @@
 
 let isAsmParserOnly = 1 in
 defm Int_VCVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
-                    int_x86_sse2_cvtsd2ss, f64mem, load,
-                    "cvtsd2ss\t{$src2, $src1, $dst|$dst, $src1, $src2}">,
-                    XS, VEX_4V;
+                      int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss", 0>,
+                      XS, VEX_4V;
 let Constraints = "$src1 = $dst" in
 defm Int_CVTSD2SS: sse12_cvt_sint_3addr<0x5A, VR128, VR128,
-             int_x86_sse2_cvtsd2ss, f64mem, load,
-             "cvtsd2ss\t{$src2, $dst|$dst, $src2}">, XS;
+                      int_x86_sse2_cvtsd2ss, f64mem, load, "cvtsd2ss">, XS;
 
 // Convert scalar single to scalar double
 let isAsmParserOnly = 1 in { // SSE2 instructions with XS prefix
@@ -998,11 +1052,11 @@
 
 let isAsmParserOnly = 1 in {
 def Int_VCVTPS2PDrr : I<0x5A, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
-                       "cvtps2pd\t{$src, $dst|$dst, $src}",
+                       "vcvtps2pd\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (int_x86_sse2_cvtps2pd VR128:$src))]>,
                      VEX, Requires<[HasAVX]>;
 def Int_VCVTPS2PDrm : I<0x5A, MRMSrcMem, (outs VR128:$dst), (ins f64mem:$src),
-                       "cvtps2pd\t{$src, $dst|$dst, $src}",
+                       "vcvtps2pd\t{$src, $dst|$dst, $src}",
                        [(set VR128:$dst, (int_x86_sse2_cvtps2pd
                                           (load addr:$src)))]>,
                      VEX, Requires<[HasAVX]>;
@@ -1618,7 +1672,7 @@
 }
 
 multiclass basic_sse12_fp_binop_s_int<bits<8> opc, string OpcodeStr,
-                                   bit Is2Addr = 1> {
+                                      bit Is2Addr = 1> {
   defm SS : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
      !strconcat(OpcodeStr, "ss"), "", "_ss", ssmem, sse_load_f32, Is2Addr>, XS;
   defm SD : sse12_fp_scalar_int<opc, OpcodeStr, VR128,
@@ -1626,7 +1680,7 @@
 }
 
 multiclass basic_sse12_fp_binop_p_int<bits<8> opc, string OpcodeStr,
-                                   bit Is2Addr = 1> {
+                                      bit Is2Addr = 1> {
   defm PS : sse12_fp_packed_int<opc, OpcodeStr, VR128,
      !strconcat(OpcodeStr, "ps"), "", "_ps", f128mem, memopv4f32,
                                               SSEPackedSingle, Is2Addr>, TB;
@@ -1639,24 +1693,32 @@
 // Binary Arithmetic instructions
 let isAsmParserOnly = 1 in {
   defm VADD : basic_sse12_fp_binop_s<0x58, "add", fadd, 0>,
+              basic_sse12_fp_binop_s_int<0x58, "add", 0>,
               basic_sse12_fp_binop_p<0x58, "add", fadd, 0>,
               basic_sse12_fp_binop_p_y<0x58, "add", fadd>, VEX_4V;
   defm VMUL : basic_sse12_fp_binop_s<0x59, "mul", fmul, 0>,
+              basic_sse12_fp_binop_s_int<0x59, "mul", 0>,
               basic_sse12_fp_binop_p<0x59, "mul", fmul, 0>,
               basic_sse12_fp_binop_p_y<0x59, "mul", fmul>, VEX_4V;
 
   let isCommutable = 0 in {
     defm VSUB : basic_sse12_fp_binop_s<0x5C, "sub", fsub, 0>,
+                basic_sse12_fp_binop_s_int<0x5C, "sub", 0>,
                 basic_sse12_fp_binop_p<0x5C, "sub", fsub, 0>,
                 basic_sse12_fp_binop_p_y<0x5C, "sub", fsub>, VEX_4V;
     defm VDIV : basic_sse12_fp_binop_s<0x5E, "div", fdiv, 0>,
+                basic_sse12_fp_binop_s_int<0x5E, "div", 0>,
                 basic_sse12_fp_binop_p<0x5E, "div", fdiv, 0>,
                 basic_sse12_fp_binop_p_y<0x5E, "div", fdiv>, VEX_4V;
     defm VMAX : basic_sse12_fp_binop_s<0x5F, "max", X86fmax, 0>,
+                basic_sse12_fp_binop_s_int<0x5F, "max", 0>,
                 basic_sse12_fp_binop_p<0x5F, "max", X86fmax, 0>,
+                basic_sse12_fp_binop_p_int<0x5F, "max", 0>,
                 basic_sse12_fp_binop_p_y<0x5F, "max", X86fmax>, VEX_4V;
     defm VMIN : basic_sse12_fp_binop_s<0x5D, "min", X86fmin, 0>,
+                basic_sse12_fp_binop_s_int<0x5D, "min", 0>,
                 basic_sse12_fp_binop_p<0x5D, "min", X86fmin, 0>,
+                basic_sse12_fp_binop_p_int<0x5D, "min", 0>,
                 basic_sse12_fp_binop_p_y<0x5D, "min", X86fmin>, VEX_4V;
   }
 }
@@ -1721,20 +1783,20 @@
 multiclass sse1_fp_unop_s_avx<bits<8> opc, string OpcodeStr,
                               SDNode OpNode, Intrinsic F32Int> {
   def SSr : SSI<opc, MRMSrcReg, (outs FR32:$dst), (ins FR32:$src1, FR32:$src2),
-                !strconcat(!strconcat("v", OpcodeStr),
+                !strconcat(OpcodeStr,
                            "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
   def SSm : I<opc, MRMSrcMem, (outs FR32:$dst), (ins FR32:$src1, f32mem:$src2),
-                !strconcat(!strconcat("v", OpcodeStr),
+                !strconcat(OpcodeStr,
                            "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
                 []>, XS, Requires<[HasAVX, OptForSize]>;
-  def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst),
-                (ins VR128:$src1, VR128:$src2),
-                !strconcat(!strconcat("v", OpcodeStr),
-                           "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
-  def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst),
-                (ins VR128:$src1, ssmem:$src2),
-                !strconcat(!strconcat("v", OpcodeStr),
-                           "ss\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
+  def SSr_Int : SSI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+                !strconcat(OpcodeStr,
+                           "ss\t{$src, $dst, $dst|$dst, $dst, $src}"),
+                [(set VR128:$dst, (F32Int VR128:$src))]>;
+  def SSm_Int : SSI<opc, MRMSrcMem, (outs VR128:$dst), (ins ssmem:$src),
+                !strconcat(OpcodeStr,
+                           "ss\t{$src, $dst, $dst|$dst, $dst, $src}"),
+                [(set VR128:$dst, (F32Int sse_load_f32:$src))]>;
 }
 
 /// sse1_fp_unop_p - SSE1 unops in packed form.
@@ -1791,21 +1853,19 @@
 /// sse2_fp_unop_s_avx - AVX SSE2 unops in scalar form.
 multiclass sse2_fp_unop_s_avx<bits<8> opc, string OpcodeStr,
                               SDNode OpNode, Intrinsic F64Int> {
-  def SDr : VSDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
-                !strconcat(OpcodeStr,
-                           "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
-  def SDm : VSDI<opc, MRMSrcMem, (outs FR64:$dst),
-                (ins FR64:$src1, f64mem:$src2),
-                !strconcat(OpcodeStr,
-                           "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
-  def SDr_Int : VSDI<opc, MRMSrcReg, (outs VR128:$dst),
-           (ins VR128:$src1, VR128:$src2),
-           !strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                    []>;
-  def SDm_Int : VSDI<opc, MRMSrcMem, (outs VR128:$dst),
-           (ins VR128:$src1, sdmem:$src2),
-           !strconcat(OpcodeStr, "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
-                    []>;
+  def SDr : SDI<opc, MRMSrcReg, (outs FR64:$dst), (ins FR64:$src1, FR64:$src2),
+               !strconcat(OpcodeStr,
+                          "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
+  def SDm : SDI<opc, MRMSrcMem, (outs FR64:$dst),
+               (ins FR64:$src1, f64mem:$src2),
+               !strconcat(OpcodeStr,
+                          "sd\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>;
+  def SDr_Int : SDI<opc, MRMSrcReg, (outs VR128:$dst), (ins VR128:$src),
+           !strconcat(OpcodeStr, "sd\t{$src, $dst, $dst|$dst, $dst, $src}"),
+           [(set VR128:$dst, (F64Int VR128:$src))]>;
+  def SDm_Int : SDI<opc, MRMSrcMem, (outs VR128:$dst), (ins sdmem:$src),
+           !strconcat(OpcodeStr, "sd\t{$src, $dst, $dst|$dst, $dst, $src}"),
+           [(set VR128:$dst, (F64Int sse_load_f64:$src))]>;
 }
 
 /// sse2_fp_unop_p - SSE2 unops in vector forms.
@@ -1842,27 +1902,31 @@
 
 let isAsmParserOnly = 1, Predicates = [HasAVX] in {
   // Square root.
-  defm VSQRT  : sse1_fp_unop_s_avx<0x51, "sqrt", fsqrt, int_x86_sse_sqrt_ss>,
-                sse2_fp_unop_s_avx<0x51, "sqrt", fsqrt, int_x86_sse2_sqrt_sd>,
+  defm VSQRT  : sse1_fp_unop_s_avx<0x51, "vsqrt", fsqrt, int_x86_sse_sqrt_ss>,
+                sse2_fp_unop_s_avx<0x51, "vsqrt", fsqrt, int_x86_sse2_sqrt_sd>,
                 VEX_4V;
 
   defm VSQRT  : sse1_fp_unop_p<0x51, "vsqrt", fsqrt>,
                 sse2_fp_unop_p<0x51, "vsqrt", fsqrt>,
                 sse1_fp_unop_p_y<0x51, "vsqrt", fsqrt>,
                 sse2_fp_unop_p_y<0x51, "vsqrt", fsqrt>,
+                sse1_fp_unop_p_int<0x51, "vsqrt",  int_x86_sse_sqrt_ps>,
+                sse2_fp_unop_p_int<0x51, "vsqrt", int_x86_sse2_sqrt_pd>,
                 VEX;
 
   // Reciprocal approximations. Note that these typically require refinement
   // in order to obtain suitable precision.
-  defm VRSQRT : sse1_fp_unop_s_avx<0x52, "rsqrt", X86frsqrt,
+  defm VRSQRT : sse1_fp_unop_s_avx<0x52, "vrsqrt", X86frsqrt,
                                    int_x86_sse_rsqrt_ss>, VEX_4V;
   defm VRSQRT : sse1_fp_unop_p<0x52, "vrsqrt", X86frsqrt>,
-                sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt>, VEX;
+                sse1_fp_unop_p_y<0x52, "vrsqrt", X86frsqrt>,
+                sse1_fp_unop_p_int<0x52, "vrsqrt", int_x86_sse_rsqrt_ps>, VEX;
 
-  defm VRCP   : sse1_fp_unop_s_avx<0x53, "rcp", X86frcp, int_x86_sse_rcp_ss>,
+  defm VRCP   : sse1_fp_unop_s_avx<0x53, "vrcp", X86frcp, int_x86_sse_rcp_ss>,
                                    VEX_4V;
   defm VRCP   : sse1_fp_unop_p<0x53, "vrcp", X86frcp>,
-                sse1_fp_unop_p_y<0x53, "vrcp", X86frcp>, VEX;
+                sse1_fp_unop_p_y<0x53, "vrcp", X86frcp>,
+                sse1_fp_unop_p_int<0x53, "vrcp", int_x86_sse_rcp_ps>, VEX;
 }
 
 // Square root.
@@ -2442,6 +2506,25 @@
 }
 } // Constraints = "$src1 = $dst"
 
+let Predicates = [HasAVX] in {
+  def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2),
+            (v2i64 (VPSLLDQri VR128:$src1, (BYTE_imm imm:$src2)))>;
+  def : Pat<(int_x86_sse2_psrl_dq VR128:$src1, imm:$src2),
+            (v2i64 (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>;
+  def : Pat<(int_x86_sse2_psll_dq_bs VR128:$src1, imm:$src2),
+            (v2i64 (VPSLLDQri VR128:$src1, imm:$src2))>;
+  def : Pat<(int_x86_sse2_psrl_dq_bs VR128:$src1, imm:$src2),
+            (v2i64 (VPSRLDQri VR128:$src1, imm:$src2))>;
+  def : Pat<(v2f64 (X86fsrl VR128:$src1, i32immSExt8:$src2)),
+            (v2f64 (VPSRLDQri VR128:$src1, (BYTE_imm imm:$src2)))>;
+
+  // Shift up / down and insert zero's.
+  def : Pat<(v2i64 (X86vshl  VR128:$src, (i8 imm:$amt))),
+            (v2i64 (VPSLLDQri VR128:$src, (BYTE_imm imm:$amt)))>;
+  def : Pat<(v2i64 (X86vshr  VR128:$src, (i8 imm:$amt))),
+            (v2i64 (VPSRLDQri VR128:$src, (BYTE_imm imm:$amt)))>;
+}
+
 let Predicates = [HasSSE2] in {
   def : Pat<(int_x86_sse2_psll_dq VR128:$src1, imm:$src2),
             (v2i64 (PSLLDQri VR128:$src1, (BYTE_imm imm:$src2)))>;
@@ -4257,7 +4340,11 @@
   defm VINSERTPS : SS41I_insertf32<0x21, "vinsertps", 0>, VEX_4V;
 
 def : Pat<(int_x86_sse41_insertps VR128:$src1, VR128:$src2, imm:$src3),
-          (INSERTPSrr VR128:$src1, VR128:$src2, imm:$src3)>;
+          (VINSERTPSrr VR128:$src1, VR128:$src2, imm:$src3)>,
+          Requires<[HasAVX]>;
+def : Pat<(int_x86_sse41_insertps VR128:$src1, VR128:$src2, imm:$src3),
+          (INSERTPSrr VR128:$src1, VR128:$src2, imm:$src3)>,
+          Requires<[HasSSE41]>;
 
 //===----------------------------------------------------------------------===//
 // SSE4.1 - Round Instructions
@@ -4682,27 +4769,39 @@
 /// SS41I_quaternary_int_avx - AVX SSE 4.1 with 4 operators
 let isAsmParserOnly = 1, Predicates = [HasAVX] in {
 multiclass SS41I_quaternary_int_avx<bits<8> opc, string OpcodeStr,
-                                    RegisterClass RC, X86MemOperand x86memop> {
+                                    RegisterClass RC, X86MemOperand x86memop,
+                                    PatFrag mem_frag, Intrinsic IntId> {
   def rr : I<opc, MRMSrcReg, (outs RC:$dst),
                   (ins RC:$src1, RC:$src2, RC:$src3),
                   !strconcat(OpcodeStr,
                     "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
-                  [], SSEPackedInt>, OpSize, TA, VEX_4V, VEX_I8IMM;
+                  [(set RC:$dst, (IntId RC:$src1, RC:$src2, RC:$src3))],
+                  SSEPackedInt>, OpSize, TA, VEX_4V, VEX_I8IMM;
 
   def rm : I<opc, MRMSrcMem, (outs RC:$dst),
                   (ins RC:$src1, x86memop:$src2, RC:$src3),
                   !strconcat(OpcodeStr,
                     "\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}"),
-                  [], SSEPackedInt>, OpSize, TA, VEX_4V, VEX_I8IMM;
+                  [(set RC:$dst,
+                        (IntId RC:$src1, (bitconvert (mem_frag addr:$src2)),
+                               RC:$src3))],
+                  SSEPackedInt>, OpSize, TA, VEX_4V, VEX_I8IMM;
 }
 }
 
-defm VBLENDVPD  : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, i128mem>;
-defm VBLENDVPS  : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, i128mem>;
-defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, i256mem>;
-defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, i256mem>;
-
-defm VPBLENDVB : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem>;
+defm VBLENDVPD  : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR128, i128mem,
+                                           memopv16i8, int_x86_sse41_blendvpd>;
+defm VBLENDVPS  : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR128, i128mem,
+                                           memopv16i8, int_x86_sse41_blendvps>;
+defm VPBLENDVB  : SS41I_quaternary_int_avx<0x4C, "vpblendvb", VR128, i128mem,
+                                           memopv16i8, int_x86_sse41_pblendvb>;
+
+let Pattern = []<dag> in { // FIXME: implement 256 intrinsics here.
+defm VBLENDVPDY : SS41I_quaternary_int_avx<0x4B, "vblendvpd", VR256, i256mem,
+                                           memopv32i8, int_x86_sse41_blendvpd>;
+defm VBLENDVPSY : SS41I_quaternary_int_avx<0x4A, "vblendvps", VR256, i256mem,
+                                           memopv32i8, int_x86_sse41_blendvps>;
+}
 
 /// SS41I_ternary_int - SSE 4.1 ternary operator
 let Uses = [XMM0], Constraints = "$src1 = $dst" in {
@@ -4778,17 +4877,20 @@
 //===----------------------------------------------------------------------===//
 
 // Packed Compare Implicit Length Strings, Return Mask
-let Defs = [EFLAGS], usesCustomInserter = 1 in {
-  def PCMPISTRM128REG : SS42AI<0, Pseudo, (outs VR128:$dst),
-    (ins VR128:$src1, VR128:$src2, i8imm:$src3),
-    "#PCMPISTRM128rr PSEUDO!",
+multiclass pseudo_pcmpistrm<string asm> {
+  def REG : Ii8<0, Pseudo, (outs VR128:$dst),
+    (ins VR128:$src1, VR128:$src2, i8imm:$src3), !strconcat(asm, "rr PSEUDO"),
     [(set VR128:$dst, (int_x86_sse42_pcmpistrm128 VR128:$src1, VR128:$src2,
-                                                  imm:$src3))]>, OpSize;
-  def PCMPISTRM128MEM : SS42AI<0, Pseudo, (outs VR128:$dst),
-    (ins VR128:$src1, i128mem:$src2, i8imm:$src3),
-    "#PCMPISTRM128rm PSEUDO!",
+                                                  imm:$src3))]>;
+  def MEM : Ii8<0, Pseudo, (outs VR128:$dst),
+    (ins VR128:$src1, i128mem:$src2, i8imm:$src3), !strconcat(asm, "rm PSEUDO"),
     [(set VR128:$dst, (int_x86_sse42_pcmpistrm128
-                          VR128:$src1, (load addr:$src2), imm:$src3))]>, OpSize;
+                       VR128:$src1, (load addr:$src2), imm:$src3))]>;
+}
+
+let Defs = [EFLAGS], usesCustomInserter = 1 in {
+  defm PCMPISTRM128 : pseudo_pcmpistrm<"#PCMPISTRM128">, Requires<[HasSSE42]>;
+  defm VPCMPISTRM128 : pseudo_pcmpistrm<"#VPCMPISTRM128">, Requires<[HasAVX]>;
 }
 
 let Defs = [XMM0, EFLAGS], isAsmParserOnly = 1,
@@ -4811,20 +4913,20 @@
 }
 
 // Packed Compare Explicit Length Strings, Return Mask
-let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
-  def PCMPESTRM128REG : SS42AI<0, Pseudo, (outs VR128:$dst),
-    (ins VR128:$src1, VR128:$src3, i8imm:$src5),
-    "#PCMPESTRM128rr PSEUDO!",
-    [(set VR128:$dst,
-          (int_x86_sse42_pcmpestrm128
-           VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5))]>, OpSize;
-
-  def PCMPESTRM128MEM : SS42AI<0, Pseudo, (outs VR128:$dst),
-    (ins VR128:$src1, i128mem:$src3, i8imm:$src5),
-    "#PCMPESTRM128rm PSEUDO!",
+multiclass pseudo_pcmpestrm<string asm> {
+  def REG : Ii8<0, Pseudo, (outs VR128:$dst),
+    (ins VR128:$src1, VR128:$src3, i8imm:$src5), !strconcat(asm, "rr PSEUDO"),
+    [(set VR128:$dst, (int_x86_sse42_pcmpestrm128
+                       VR128:$src1, EAX, VR128:$src3, EDX, imm:$src5))]>;
+  def MEM : Ii8<0, Pseudo, (outs VR128:$dst),
+    (ins VR128:$src1, i128mem:$src3, i8imm:$src5), !strconcat(asm, "rm PSEUDO"),
     [(set VR128:$dst, (int_x86_sse42_pcmpestrm128
-                       VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5))]>,
-    OpSize;
+                       VR128:$src1, EAX, (load addr:$src3), EDX, imm:$src5))]>;
+}
+
+let Defs = [EFLAGS], Uses = [EAX, EDX], usesCustomInserter = 1 in {
+  defm PCMPESTRM128 : pseudo_pcmpestrm<"#PCMPESTRM128">, Requires<[HasSSE42]>;
+  defm VPCMPESTRM128 : pseudo_pcmpestrm<"#VPCMPESTRM128">, Requires<[HasAVX]>;
 }
 
 let isAsmParserOnly = 1, Predicates = [HasAVX],




From bob.wilson at apple.com  Fri Jul 30 14:55:47 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Fri, 30 Jul 2010 19:55:47 -0000
Subject: [llvm-commits] [llvm] r109879 -
	/llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp
Message-ID: <20100730195547.4F6112A6C12C@llvm.org>

Author: bwilson
Date: Fri Jul 30 14:55:47 2010
New Revision: 109879

URL: http://llvm.org/viewvc/llvm-project?rev=109879&view=rev
Log:
Add the __TEXT,__StaticInit section to the list of sections emitted at the
beginning on ARM Darwin assembly files so that it won't be placed after
debug sections.  Radar 8252813.

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

Modified: llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp?rev=109879&r1=109878&r2=109879&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp Fri Jul 30 14:55:47 2010
@@ -1165,6 +1165,12 @@
                                      16, SectionKind::getText());
         OutStreamer.SwitchSection(sect);
       }
+      const MCSection *StaticInitSect =
+        OutContext.getMachOSection("__TEXT", "__StaticInit",
+                                   MCSectionMachO::S_REGULAR |
+                                   MCSectionMachO::S_ATTR_PURE_INSTRUCTIONS,
+                                   SectionKind::getText());
+      OutStreamer.SwitchSection(StaticInitSect);
     }
   }
 




From bruno.cardoso at gmail.com  Fri Jul 30 14:57:56 2010
From: bruno.cardoso at gmail.com (Bruno Cardoso Lopes)
Date: Fri, 30 Jul 2010 19:57:56 -0000
Subject: [llvm-commits] [llvm] r109881 - in /llvm/trunk/test/CodeGen/X86:
 avx-intrinsics-x86.ll avx-intrinsics-x86_64.ll
Message-ID: <20100730195756.CD3DA2A6C12C@llvm.org>

Author: bruno
Date: Fri Jul 30 14:57:56 2010
New Revision: 109881

URL: http://llvm.org/viewvc/llvm-project?rev=109881&view=rev
Log:
A *bunch* of tests for AVX intrinsics

Added:
    llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
    llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll

Added: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll?rev=109881&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll (added)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll Fri Jul 30 14:57:56 2010
@@ -0,0 +1,1956 @@
+; RUN: llc < %s -march=x86 -mattr=avx,aes | FileCheck %s
+
+define <2 x i64> @test_x86_aesni_aesdec(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vaesdec
+  %res = call <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <2 x i64> @test_x86_aesni_aesdeclast(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vaesdeclast
+  %res = call <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <2 x i64> @test_x86_aesni_aesenc(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vaesenc
+  %res = call <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <2 x i64> @test_x86_aesni_aesenclast(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vaesenclast
+  %res = call <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <2 x i64> @test_x86_aesni_aesimc(<2 x i64> %a0) {
+  ; CHECK: vaesimc
+  %res = call <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64>) nounwind readnone
+
+
+define <2 x i64> @test_x86_aesni_aeskeygenassist(<2 x i64> %a0) {
+  ; CHECK: vaeskeygenassist
+  %res = call <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64> %a0, i8 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64>, i8) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_add_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vaddsd
+  %res = call <2 x double> @llvm.x86.sse2.add.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.add.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cmp_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcmpordpd
+  %res = call <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double>, <2 x double>, i8) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cmp_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcmpordsd
+  %res = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double>, <2 x double>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse2_comieq_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: sete
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comieq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comieq.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_comige_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: setae
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comige.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comige.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_comigt_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: seta
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comigt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comigt.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_comile_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: setbe
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comile.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comile.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_comilt_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: setb
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comilt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comilt.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_comineq_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vcomisd
+  ; CHECK: setne
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.comineq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.comineq.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cvtdq2pd(<4 x i32> %a0) {
+  ; CHECK: vcvtdq2pd
+  %res = call <2 x double> @llvm.x86.sse2.cvtdq2pd(<4 x i32> %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cvtdq2pd(<4 x i32>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse2_cvtdq2ps(<4 x i32> %a0) {
+  ; CHECK: vcvtdq2ps
+  %res = call <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_cvtpd2dq(<2 x double> %a0) {
+  ; CHECK: vcvtpd2dq
+  %res = call <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse2_cvtpd2ps(<2 x double> %a0) {
+  ; CHECK: vcvtpd2ps
+  %res = call <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_cvtps2dq(<4 x float> %a0) {
+  ; CHECK: vcvtps2dq
+  %res = call <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cvtps2pd(<4 x float> %a0) {
+  ; CHECK: vcvtps2pd
+  %res = call <2 x double> @llvm.x86.sse2.cvtps2pd(<4 x float> %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cvtps2pd(<4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse2_cvtsd2si(<2 x double> %a0) {
+  ; CHECK: vcvtsd2si
+  %res = call i32 @llvm.x86.sse2.cvtsd2si(<2 x double> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.cvtsd2si(<2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse2_cvtsd2ss(<4 x float> %a0, <2 x double> %a1) {
+  ; CHECK: vcvtsd2ss
+  %res = call <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float> %a0, <2 x double> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cvtsi2sd(<2 x double> %a0) {
+  ; CHECK: movl
+  ; CHECK: vcvtsi2sd
+  %res = call <2 x double> @llvm.x86.sse2.cvtsi2sd(<2 x double> %a0, i32 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cvtsi2sd(<2 x double>, i32) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cvtss2sd(<2 x double> %a0, <4 x float> %a1) {
+  ; CHECK: vcvtss2sd
+  %res = call <2 x double> @llvm.x86.sse2.cvtss2sd(<2 x double> %a0, <4 x float> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cvtss2sd(<2 x double>, <4 x float>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_cvttpd2dq(<2 x double> %a0) {
+  ; CHECK: vcvttpd2dq
+  %res = call <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_cvttps2dq(<4 x float> %a0) {
+  ; CHECK: vcvttps2dq
+  %res = call <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse2_cvttsd2si(<2 x double> %a0) {
+  ; CHECK: vcvttss2si
+  %res = call i32 @llvm.x86.sse2.cvttsd2si(<2 x double> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.cvttsd2si(<2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_div_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vdivsd
+  %res = call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.div.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_loadu_dq(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vmovdqu
+  %res = call <16 x i8> @llvm.x86.sse2.loadu.dq(i8* %a0) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.loadu.dq(i8*) nounwind readonly
+
+
+define <2 x double> @test_x86_sse2_loadu_pd(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vmovupd
+  %res = call <2 x double> @llvm.x86.sse2.loadu.pd(i8* %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.loadu.pd(i8*) nounwind readonly
+
+
+define void @test_x86_sse2_maskmov_dqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2) {
+  ; CHECK: pushl
+  ; CHECK: movl
+  ; CHECK: vmaskmovdqu
+  ; CHECK: popl
+  call void @llvm.x86.sse2.maskmov.dqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2)
+  ret void
+}
+declare void @llvm.x86.sse2.maskmov.dqu(<16 x i8>, <16 x i8>, i8*) nounwind
+
+
+define <2 x double> @test_x86_sse2_max_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vmaxpd
+  %res = call <2 x double> @llvm.x86.sse2.max.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.max.pd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_max_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vmaxsd
+  %res = call <2 x double> @llvm.x86.sse2.max.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.max.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_min_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vminpd
+  %res = call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_min_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vminsd
+  %res = call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.min.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_movmsk_pd(<2 x double> %a0) {
+  ; CHECK: vmovmskpd
+  %res = call i32 @llvm.x86.sse2.movmsk.pd(<2 x double> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.movmsk.pd(<2 x double>) nounwind readnone
+
+
+define void @test_x86_sse2_movnt_dq(i8* %a0, <2 x i64> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovntdq
+  call void @llvm.x86.sse2.movnt.dq(i8* %a0, <2 x i64> %a1)
+  ret void
+}
+declare void @llvm.x86.sse2.movnt.dq(i8*, <2 x i64>) nounwind
+
+
+define void @test_x86_sse2_movnt_pd(i8* %a0, <2 x double> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovntpd
+  call void @llvm.x86.sse2.movnt.pd(i8* %a0, <2 x double> %a1)
+  ret void
+}
+declare void @llvm.x86.sse2.movnt.pd(i8*, <2 x double>) nounwind
+
+
+define <2 x double> @test_x86_sse2_mul_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vmulsd
+  %res = call <2 x double> @llvm.x86.sse2.mul.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.mul.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_packssdw_128(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpackssdw
+  %res = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a0, <4 x i32> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_packsswb_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpacksswb
+  %res = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a0, <8 x i16> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_packuswb_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpackuswb
+  %res = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a0, <8 x i16> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_padds_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpaddsb
+  %res = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_padds_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpaddsw
+  %res = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_paddus_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpaddusb
+  %res = call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_paddus_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpaddusw
+  %res = call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_pavg_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpavgb
+  %res = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pavg_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpavgw
+  %res = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_pcmpeq_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpeqb
+  %res = call <16 x i8> @llvm.x86.sse2.pcmpeq.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.pcmpeq.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_pcmpeq_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpcmpeqd
+  %res = call <4 x i32> @llvm.x86.sse2.pcmpeq.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.pcmpeq.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pcmpeq_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpcmpeqw
+  %res = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_pcmpgt_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpgtb
+  %res = call <16 x i8> @llvm.x86.sse2.pcmpgt.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.pcmpgt.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_pcmpgt_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpcmpgtd
+  %res = call <4 x i32> @llvm.x86.sse2.pcmpgt.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.pcmpgt.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pcmpgt_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpcmpgtw
+  %res = call <8 x i16> @llvm.x86.sse2.pcmpgt.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pcmpgt.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_pmadd_wd(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmaddwd
+  %res = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %a0, <8 x i16> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pmaxs_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmaxsw
+  %res = call <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_pmaxu_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpmaxub
+  %res = call <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pmins_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpminsw
+  %res = call <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_pminu_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpminub
+  %res = call <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define i32 @test_x86_sse2_pmovmskb_128(<16 x i8> %a0) {
+  ; CHECK: vpmovmskb
+  %res = call i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pmulh_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmulhw
+  %res = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pmulhu_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmulhuw
+  %res = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_pmulu_dq(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpmuludq
+  %res = call <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32> %a0, <4 x i32> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psad_bw(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpsadbw
+  %res = call <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8> %a0, <16 x i8> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_psll_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpslld
+  %res = call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psll_dq(<2 x i64> %a0) {
+  ; CHECK: vpslldq
+  %res = call <2 x i64> @llvm.x86.sse2.psll.dq(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psll.dq(<2 x i64>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psll_dq_bs(<2 x i64> %a0) {
+  ; CHECK: vpslldq
+  %res = call <2 x i64> @llvm.x86.sse2.psll.dq.bs(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psll.dq.bs(<2 x i64>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psll_q(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpsllq
+  %res = call <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psll_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsllw
+  %res = call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_pslli_d(<4 x i32> %a0) {
+  ; CHECK: vpslld
+  %res = call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_pslli_q(<2 x i64> %a0) {
+  ; CHECK: vpsllq
+  %res = call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64>, i32) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_pslli_w(<8 x i16> %a0) {
+  ; CHECK: vpsllw
+  %res = call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16>, i32) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_psra_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsrad
+  %res = call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psra_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsraw
+  %res = call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_psrai_d(<4 x i32> %a0) {
+  ; CHECK: vpsrad
+  %res = call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32>, i32) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psrai_w(<8 x i16> %a0) {
+  ; CHECK: vpsraw
+  %res = call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16>, i32) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_psrl_d(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsrld
+  %res = call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psrl_dq(<2 x i64> %a0) {
+  ; CHECK: vpsrldq
+  %res = call <2 x i64> @llvm.x86.sse2.psrl.dq(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psrl.dq(<2 x i64>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psrl_dq_bs(<2 x i64> %a0) {
+  ; CHECK: vpsrldq
+  %res = call <2 x i64> @llvm.x86.sse2.psrl.dq.bs(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psrl.dq.bs(<2 x i64>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psrl_q(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpsrlq
+  %res = call <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psrl_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsrlw
+  %res = call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse2_psrli_d(<4 x i32> %a0) {
+  ; CHECK: vpsrld
+  %res = call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse2_psrli_q(<2 x i64> %a0) {
+  ; CHECK: vpsrlq
+  %res = call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64>, i32) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psrli_w(<8 x i16> %a0) {
+  ; CHECK: vpsrlw
+  %res = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16>, i32) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_psubs_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpsubsb
+  %res = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psubs_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsubsw
+  %res = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse2_psubus_b(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpsubusb
+  %res = call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse2_psubus_w(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsubusw
+  %res = call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_sqrt_pd(<2 x double> %a0) {
+  ; CHECK: vsqrtpd
+  %res = call <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double> %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_sqrt_sd(<2 x double> %a0) {
+  ; CHECK: vsqrtsd
+  %res = call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %a0) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone
+
+
+define void @test_x86_sse2_storel_dq(i8* %a0, <4 x i32> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovq
+  call void @llvm.x86.sse2.storel.dq(i8* %a0, <4 x i32> %a1)
+  ret void
+}
+declare void @llvm.x86.sse2.storel.dq(i8*, <4 x i32>) nounwind
+
+
+define void @test_x86_sse2_storeu_dq(i8* %a0, <16 x i8> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovdqu
+  call void @llvm.x86.sse2.storeu.dq(i8* %a0, <16 x i8> %a1)
+  ret void
+}
+declare void @llvm.x86.sse2.storeu.dq(i8*, <16 x i8>) nounwind
+
+
+define void @test_x86_sse2_storeu_pd(i8* %a0, <2 x double> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovupd
+  call void @llvm.x86.sse2.storeu.pd(i8* %a0, <2 x double> %a1)
+  ret void
+}
+declare void @llvm.x86.sse2.storeu.pd(i8*, <2 x double>) nounwind
+
+
+define <2 x double> @test_x86_sse2_sub_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vsubsd
+  %res = call <2 x double> @llvm.x86.sse2.sub.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.sub.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomieq_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: sete
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomieq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomieq.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomige_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: setae
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomige.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomige.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomigt_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: seta
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomigt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomigt.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomile_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: setbe
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomile.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomile.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomilt_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: setb
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomilt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomilt.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define i32 @test_x86_sse2_ucomineq_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vucomisd
+  ; CHECK: setne
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse2.ucomineq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse2.ucomineq.sd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse3_addsub_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vaddsubpd
+  %res = call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse3_addsub_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vaddsubps
+  %res = call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse3_hadd_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vhaddpd
+  %res = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse3_hadd_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vhaddps
+  %res = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse3_hsub_pd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vhsubpd
+  %res = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double>, <2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse3_hsub_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vhsubps
+  %res = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse3_ldu_dq(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vlddqu
+  %res = call <16 x i8> @llvm.x86.sse3.ldu.dq(i8* %a0) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse3.ldu.dq(i8*) nounwind readonly
+
+
+define <2 x double> @test_x86_sse41_blendpd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vblendpd
+  %res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_blendps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vblendps
+  %res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32) nounwind readnone
+
+
+define <2 x double> @test_x86_sse41_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
+  ; CHECK: vblendvpd
+  %res = call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse41.blendvpd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
+  ; CHECK: vblendvps
+  %res = call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse41_dppd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vdppd
+  %res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i32) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_dpps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vdpps
+  %res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i32) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_insertps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vinsertps
+  %res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_movntdqa(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vmovntdqa
+  %res = call <2 x i64> @llvm.x86.sse41.movntdqa(i8* %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.movntdqa(i8*) nounwind readonly
+
+
+define <16 x i8> @test_x86_sse41_mpsadbw(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vmpsadbw
+  %res = call <16 x i8> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i32 7) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i32) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_packusdw(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpackusdw
+  %res = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a0, <4 x i32> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse41_pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2) {
+  ; CHECK: vpblendvb
+  %res = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_pblendw(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpblendw
+  %res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 7) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pcmpeqq(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpcmpeqq
+  %res = call <2 x i64> @llvm.x86.sse41.pcmpeqq(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pcmpeqq(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_phminposuw(<8 x i16> %a0) {
+  ; CHECK: vphminposuw
+  %res = call <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16> %a0) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse41_pmaxsb(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpmaxsb
+  %res = call <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmaxsd(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpmaxsd
+  %res = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmaxud(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpmaxud
+  %res = call <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_pmaxuw(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmaxuw
+  %res = call <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse41_pminsb(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpminsb
+  %res = call <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pminsd(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpminsd
+  %res = call <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pminud(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpminud
+  %res = call <4 x i32> @llvm.x86.sse41.pminud(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pminud(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_pminuw(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpminuw
+  %res = call <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmovsxbd(<16 x i8> %a0) {
+  ; CHECK: vpmovsxbd
+  %res = call <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovsxbq(<16 x i8> %a0) {
+  ; CHECK: vpmovsxbq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovsxbq(<16 x i8> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovsxbq(<16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_pmovsxbw(<16 x i8> %a0) {
+  ; CHECK: vpmovsxbw
+  %res = call <8 x i16> @llvm.x86.sse41.pmovsxbw(<16 x i8> %a0) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.pmovsxbw(<16 x i8>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovsxdq(<4 x i32> %a0) {
+  ; CHECK: vpmovsxdq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovsxdq(<4 x i32> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovsxdq(<4 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmovsxwd(<8 x i16> %a0) {
+  ; CHECK: vpmovsxwd
+  %res = call <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovsxwq(<8 x i16> %a0) {
+  ; CHECK: vpmovsxwq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovsxwq(<8 x i16> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovsxwq(<8 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmovzxbd(<16 x i8> %a0) {
+  ; CHECK: vpmovzxbd
+  %res = call <4 x i32> @llvm.x86.sse41.pmovzxbd(<16 x i8> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmovzxbd(<16 x i8>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovzxbq(<16 x i8> %a0) {
+  ; CHECK: vpmovzxbq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8>) nounwind readnone
+
+
+define <8 x i16> @test_x86_sse41_pmovzxbw(<16 x i8> %a0) {
+  ; CHECK: vpmovzxbw
+  %res = call <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8> %a0) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovzxdq(<4 x i32> %a0) {
+  ; CHECK: vpmovzxdq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovzxdq(<4 x i32> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovzxdq(<4 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_sse41_pmovzxwd(<8 x i16> %a0) {
+  ; CHECK: vpmovzxwd
+  %res = call <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmovzxwq(<8 x i16> %a0) {
+  ; CHECK: vpmovzxwq
+  %res = call <2 x i64> @llvm.x86.sse41.pmovzxwq(<8 x i16> %a0) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmovzxwq(<8 x i16>) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse41_pmuldq(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpmuldq
+  %res = call <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32> %a0, <4 x i32> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define i32 @test_x86_sse41_ptestc(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vptest 
+  ; CHECK: setb
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse41.ptestc(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse41.ptestc(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse41_ptestnzc(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vptest 
+  ; CHECK: seta
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse41.ptestnzc(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse41.ptestnzc(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse41_ptestz(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vptest 
+  ; CHECK: sete
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse41.ptestz(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse41_round_pd(<2 x double> %a0) {
+  ; CHECK: vroundpd
+  %res = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse41.round.pd(<2 x double>, i32) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_round_ps(<4 x float> %a0) {
+  ; CHECK: vroundps
+  %res = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.round.ps(<4 x float>, i32) nounwind readnone
+
+
+define <2 x double> @test_x86_sse41_round_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vroundsd
+  %res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse41.round.sd(<2 x double>, <2 x double>, i32) nounwind readnone
+
+
+define <4 x float> @test_x86_sse41_round_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vroundss
+  %res = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestri128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestri128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestri128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestria128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestria128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestria128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestric128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestric128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestric128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestrio128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestrio128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestrio128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestris128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestris128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestris128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpestriz128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpestriz128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpestriz128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse42_pcmpestrm128(<16 x i8> %a0, <16 x i8> %a2) {
+  ; CHECK: movl
+  ; CHECK: movl
+  ; CHECK: vpcmpestrm
+  %res = call <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
+
+
+define <2 x i64> @test_x86_sse42_pcmpgtq(<2 x i64> %a0, <2 x i64> %a1) {
+  ; CHECK: vpcmpgtq
+  %res = call <2 x i64> @llvm.x86.sse42.pcmpgtq(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
+  ret <2 x i64> %res
+}
+declare <2 x i64> @llvm.x86.sse42.pcmpgtq(<2 x i64>, <2 x i64>) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistri128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistri128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistri128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistria128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistria128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistria128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistric128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistric128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistric128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistrio128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistrio128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistrio128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistris128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistris128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistris128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse42_pcmpistriz128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistri
+  ; CHECK: movl
+  %res = call i32 @llvm.x86.sse42.pcmpistriz128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse42.pcmpistriz128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define <16 x i8> @test_x86_sse42_pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpcmpistrm
+  %res = call <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8>, <16 x i8>, i8) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_add_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vaddss
+  %res = call <4 x float> @llvm.x86.sse.add.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.add.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_cmp_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcmpordps
+  %res = call <4 x float> @llvm.x86.sse.cmp.ps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.cmp.ps(<4 x float>, <4 x float>, i8) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_cmp_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcmpordss
+  %res = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.cmp.ss(<4 x float>, <4 x float>, i8) nounwind readnone
+
+
+define i32 @test_x86_sse_comieq_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: sete
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comieq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comieq.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_comige_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: setae
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comige.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comige.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_comigt_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: seta
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comigt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comigt.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_comile_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: setbe
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comile.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comile.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_comilt_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: setb
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comilt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comilt.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_comineq_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vcomiss
+  ; CHECK: setne
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.comineq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.comineq.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_cvtsi2ss(<4 x float> %a0) {
+  ; CHECK: movl
+  ; CHECK: vcvtsi2ss
+  %res = call <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float> %a0, i32 7) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float>, i32) nounwind readnone
+
+
+define i32 @test_x86_sse_cvtss2si(<4 x float> %a0) {
+  ; CHECK: vcvtss2si
+  %res = call i32 @llvm.x86.sse.cvtss2si(<4 x float> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.cvtss2si(<4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_cvttss2si(<4 x float> %a0) {
+  ; CHECK: vcvttss2si
+  %res = call i32 @llvm.x86.sse.cvttss2si(<4 x float> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.cvttss2si(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_div_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vdivss
+  %res = call <4 x float> @llvm.x86.sse.div.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.div.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_ldmxcsr(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vldmxcsr
+  call void @llvm.x86.sse.ldmxcsr(i8* %a0)
+  ret void
+}
+declare void @llvm.x86.sse.ldmxcsr(i8*) nounwind
+
+
+define <4 x float> @test_x86_sse_loadu_ps(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vmovups
+  %res = call <4 x float> @llvm.x86.sse.loadu.ps(i8* %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.loadu.ps(i8*) nounwind readonly
+
+
+define <4 x float> @test_x86_sse_max_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vmaxps
+  %res = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_max_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vmaxss
+  %res = call <4 x float> @llvm.x86.sse.max.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_min_ps(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vminps
+  %res = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_min_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vminss
+  %res = call <4 x float> @llvm.x86.sse.min.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_movmsk_ps(<4 x float> %a0) {
+  ; CHECK: vmovmskps
+  %res = call i32 @llvm.x86.sse.movmsk.ps(<4 x float> %a0) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.movmsk.ps(<4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_movnt_ps(i8* %a0, <4 x float> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovntps
+  call void @llvm.x86.sse.movnt.ps(i8* %a0, <4 x float> %a1)
+  ret void
+}
+declare void @llvm.x86.sse.movnt.ps(i8*, <4 x float>) nounwind
+
+
+define <4 x float> @test_x86_sse_mul_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vmulss
+  %res = call <4 x float> @llvm.x86.sse.mul.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.mul.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_rcp_ps(<4 x float> %a0) {
+  ; CHECK: vrcpps
+  %res = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.rcp.ps(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_rcp_ss(<4 x float> %a0) {
+  ; CHECK: vrcpss
+  %res = call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.rcp.ss(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_rsqrt_ps(<4 x float> %a0) {
+  ; CHECK: vrsqrtps
+  %res = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_rsqrt_ss(<4 x float> %a0) {
+  ; CHECK: vrsqrtss
+  %res = call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_sqrt_ps(<4 x float> %a0) {
+  ; CHECK: vsqrtps
+  %res = call <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_sqrt_ss(<4 x float> %a0) {
+  ; CHECK: vsqrtss
+  %res = call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float>) nounwind readnone
+
+
+define void @test_x86_sse_stmxcsr(i8* %a0) {
+  ; CHECK: movl
+  ; CHECK: vstmxcsr
+  call void @llvm.x86.sse.stmxcsr(i8* %a0)
+  ret void
+}
+declare void @llvm.x86.sse.stmxcsr(i8*) nounwind
+
+
+define void @test_x86_sse_storeu_ps(i8* %a0, <4 x float> %a1) {
+  ; CHECK: movl
+  ; CHECK: vmovups
+  call void @llvm.x86.sse.storeu.ps(i8* %a0, <4 x float> %a1)
+  ret void
+}
+declare void @llvm.x86.sse.storeu.ps(i8*, <4 x float>) nounwind
+
+
+define <4 x float> @test_x86_sse_sub_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vsubss
+  %res = call <4 x float> @llvm.x86.sse.sub.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.sub.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomieq_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: sete
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomieq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomieq.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomige_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: setae
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomige.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomige.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomigt_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: seta
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomigt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomigt.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomile_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: setbe
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomile.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomile.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomilt_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: setb
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomilt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomilt.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define i32 @test_x86_sse_ucomineq_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vucomiss
+  ; CHECK: setne
+  ; CHECK: movzbl
+  %res = call i32 @llvm.x86.sse.ucomineq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
+  ret i32 %res
+}
+declare i32 @llvm.x86.sse.ucomineq.ss(<4 x float>, <4 x float>) nounwind readnone
+
+
+define <8 x i8> @test_x86_ssse3_pabs_b(<8 x i8> %a0) {
+  ; CHECK: vpabsb
+  %res = call <8 x i8> @llvm.x86.ssse3.pabs.b(<8 x i8> %a0) ; <<8 x i8>> [#uses=1]
+  ret <8 x i8> %res
+}
+declare <8 x i8> @llvm.x86.ssse3.pabs.b(<8 x i8>) nounwind readnone
+
+
+define <16 x i8> @test_x86_ssse3_pabs_b_128(<16 x i8> %a0) {
+  ; CHECK: vpabsb
+  %res = call <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8> %a0) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8>) nounwind readnone
+
+
+define <2 x i32> @test_x86_ssse3_pabs_d(<2 x i32> %a0) {
+  ; CHECK: vpabsd
+  %res = call <2 x i32> @llvm.x86.ssse3.pabs.d(<2 x i32> %a0) ; <<2 x i32>> [#uses=1]
+  ret <2 x i32> %res
+}
+declare <2 x i32> @llvm.x86.ssse3.pabs.d(<2 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_ssse3_pabs_d_128(<4 x i32> %a0) {
+  ; CHECK: vpabsd
+  %res = call <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32> %a0) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_pabs_w(<4 x i16> %a0) {
+  ; CHECK: vpabsw
+  %res = call <4 x i16> @llvm.x86.ssse3.pabs.w(<4 x i16> %a0) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.pabs.w(<4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_pabs_w_128(<8 x i16> %a0) {
+  ; CHECK: vpabsw
+  %res = call <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16> %a0) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16>) nounwind readnone
+
+
+define <2 x i32> @test_x86_ssse3_phadd_d(<2 x i32> %a0, <2 x i32> %a1) {
+  ; CHECK: vphaddd
+  %res = call <2 x i32> @llvm.x86.ssse3.phadd.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
+  ret <2 x i32> %res
+}
+declare <2 x i32> @llvm.x86.ssse3.phadd.d(<2 x i32>, <2 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_ssse3_phadd_d_128(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vphaddd
+  %res = call <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_phadd_sw(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vphaddsw
+  %res = call <4 x i16> @llvm.x86.ssse3.phadd.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.phadd.sw(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <4 x i32> @test_x86_ssse3_phadd_sw_128(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vphaddsw
+  %res = call <4 x i32> @llvm.x86.ssse3.phadd.sw.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.ssse3.phadd.sw.128(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_phadd_w(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vphaddw
+  %res = call <4 x i16> @llvm.x86.ssse3.phadd.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.phadd.w(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_phadd_w_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vphaddw
+  %res = call <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <2 x i32> @test_x86_ssse3_phsub_d(<2 x i32> %a0, <2 x i32> %a1) {
+  ; CHECK: vphsubd
+  %res = call <2 x i32> @llvm.x86.ssse3.phsub.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
+  ret <2 x i32> %res
+}
+declare <2 x i32> @llvm.x86.ssse3.phsub.d(<2 x i32>, <2 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_ssse3_phsub_d_128(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vphsubd
+  %res = call <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_phsub_sw(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vphsubsw
+  %res = call <4 x i16> @llvm.x86.ssse3.phsub.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.phsub.sw(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_phsub_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vphsubsw
+  %res = call <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_phsub_w(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vphsubw
+  %res = call <4 x i16> @llvm.x86.ssse3.phsub.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.phsub.w(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_phsub_w_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vphsubw
+  %res = call <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_pmadd_ub_sw(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vpmaddubsw
+  %res = call <4 x i16> @llvm.x86.ssse3.pmadd.ub.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.pmadd.ub.sw(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_pmadd_ub_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmaddubsw
+  %res = call <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_pmul_hr_sw(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vpmulhrsw
+  %res = call <4 x i16> @llvm.x86.ssse3.pmul.hr.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.pmul.hr.sw(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_pmul_hr_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpmulhrsw
+  %res = call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
+
+
+define <8 x i8> @test_x86_ssse3_pshuf_b(<8 x i8> %a0, <8 x i8> %a1) {
+  ; CHECK: vpshufb
+  %res = call <8 x i8> @llvm.x86.ssse3.pshuf.b(<8 x i8> %a0, <8 x i8> %a1) ; <<8 x i8>> [#uses=1]
+  ret <8 x i8> %res
+}
+declare <8 x i8> @llvm.x86.ssse3.pshuf.b(<8 x i8>, <8 x i8>) nounwind readnone
+
+
+define <16 x i8> @test_x86_ssse3_pshuf_b_128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpshufb
+  %res = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <8 x i8> @test_x86_ssse3_psign_b(<8 x i8> %a0, <8 x i8> %a1) {
+  ; CHECK: vpsignb
+  %res = call <8 x i8> @llvm.x86.ssse3.psign.b(<8 x i8> %a0, <8 x i8> %a1) ; <<8 x i8>> [#uses=1]
+  ret <8 x i8> %res
+}
+declare <8 x i8> @llvm.x86.ssse3.psign.b(<8 x i8>, <8 x i8>) nounwind readnone
+
+
+define <16 x i8> @test_x86_ssse3_psign_b_128(<16 x i8> %a0, <16 x i8> %a1) {
+  ; CHECK: vpsignb
+  %res = call <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
+  ret <16 x i8> %res
+}
+declare <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8>, <16 x i8>) nounwind readnone
+
+
+define <2 x i32> @test_x86_ssse3_psign_d(<2 x i32> %a0, <2 x i32> %a1) {
+  ; CHECK: vpsignd
+  %res = call <2 x i32> @llvm.x86.ssse3.psign.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
+  ret <2 x i32> %res
+}
+declare <2 x i32> @llvm.x86.ssse3.psign.d(<2 x i32>, <2 x i32>) nounwind readnone
+
+
+define <4 x i32> @test_x86_ssse3_psign_d_128(<4 x i32> %a0, <4 x i32> %a1) {
+  ; CHECK: vpsignd
+  %res = call <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
+  ret <4 x i32> %res
+}
+declare <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32>, <4 x i32>) nounwind readnone
+
+
+define <4 x i16> @test_x86_ssse3_psign_w(<4 x i16> %a0, <4 x i16> %a1) {
+  ; CHECK: vpsignw
+  %res = call <4 x i16> @llvm.x86.ssse3.psign.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
+  ret <4 x i16> %res
+}
+declare <4 x i16> @llvm.x86.ssse3.psign.w(<4 x i16>, <4 x i16>) nounwind readnone
+
+
+define <8 x i16> @test_x86_ssse3_psign_w_128(<8 x i16> %a0, <8 x i16> %a1) {
+  ; CHECK: vpsignw
+  %res = call <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
+  ret <8 x i16> %res
+}
+declare <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16>, <8 x i16>) nounwind readnone
+

Added: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll?rev=109881&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll (added)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll Fri Jul 30 14:57:56 2010
@@ -0,0 +1,50 @@
+; RUN: llc < %s -march=x86-64 -mattr=avx,aes | FileCheck %s
+
+define i64 @test_x86_sse2_cvtsd2si64(<2 x double> %a0) {
+  ; CHECK: vcvtsd2si
+  %res = call i64 @llvm.x86.sse2.cvtsd2si64(<2 x double> %a0) ; <i64> [#uses=1]
+  ret i64 %res
+}
+declare i64 @llvm.x86.sse2.cvtsd2si64(<2 x double>) nounwind readnone
+
+
+define <2 x double> @test_x86_sse2_cvtsi642sd(<2 x double> %a0, i64 %a1) {
+  ; CHECK: vcvtsi2sd
+  %res = call <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double> %a0, i64 %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double>, i64) nounwind readnone
+
+
+define i64 @test_x86_sse2_cvttsd2si64(<2 x double> %a0) {
+  ; CHECK: vcvttss2si
+  %res = call i64 @llvm.x86.sse2.cvttsd2si64(<2 x double> %a0) ; <i64> [#uses=1]
+  ret i64 %res
+}
+declare i64 @llvm.x86.sse2.cvttsd2si64(<2 x double>) nounwind readnone
+
+
+define i64 @test_x86_sse_cvtss2si64(<4 x float> %a0) {
+  ; CHECK: vcvtss2si
+  %res = call i64 @llvm.x86.sse.cvtss2si64(<4 x float> %a0) ; <i64> [#uses=1]
+  ret i64 %res
+}
+declare i64 @llvm.x86.sse.cvtss2si64(<4 x float>) nounwind readnone
+
+
+define <4 x float> @test_x86_sse_cvtsi642ss(<4 x float> %a0, i64 %a1) {
+  ; CHECK: vcvtsi2ss
+  %res = call <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float> %a0, i64 %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float>, i64) nounwind readnone
+
+
+define i64 @test_x86_sse_cvttss2si64(<4 x float> %a0) {
+  ; CHECK: vcvttss2si
+  %res = call i64 @llvm.x86.sse.cvttss2si64(<4 x float> %a0) ; <i64> [#uses=1]
+  ret i64 %res
+}
+declare i64 @llvm.x86.sse.cvttss2si64(<4 x float>) nounwind readnone
+
+




From peter at pcc.me.uk  Fri Jul 30 15:15:30 2010
From: peter at pcc.me.uk (Peter Collingbourne)
Date: Fri, 30 Jul 2010 21:15:30 +0100
Subject: [llvm-commits] [PATCH] Phi node folding improvements
Message-ID: <20100730201530.GA1396@pcc.me.uk>

Hi,

The attached patches improve the flexibility of the phi node folding
mechanism by assigning a cost to each non-trapping instruction and
allowing the user to specify a maximum cost threshold.  By increasing
the threshold the user can apply phi node folding to more complex
expressions.

I developed this patch as part of my work with the Klee symbolic
virtual machine, which can handle symbolic expressions containing
selects more efficiently than branches conditional on symbolic
expressions.  However I imagine it could also provide an advantage on
architectures with relatively expensive branch operations (I haven't
done any experiments to determine if this is the case though).

The patches should not contain any changes to default functionality,
except for adding more instructions to the list of "cheap" (folded
by default) instructions.

OK to commit?

Thanks,
-- 
Peter
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Add-Trunc-ZExt-and-SExt-to-the-list-of-cheap-instruc.patch
Type: text/x-diff
Size: 927 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100730/1af696fd/attachment.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0002-Add-CostRemaining-parameter-to-DominatesMergePoint.patch
Type: text/x-diff
Size: 4957 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100730/1af696fd/attachment-0001.bin 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0003-Expose-phi-node-folding-cost-threshold-as-command-li.patch
Type: text/x-diff
Size: 1618 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100730/1af696fd/attachment-0002.bin 

From nicholas at mxc.ca  Fri Jul 30 15:19:09 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Fri, 30 Jul 2010 20:19:09 -0000
Subject: [llvm-commits] [llvm] r109883 -
	/llvm/trunk/include/llvm/Analysis/LibCallAliasAnalysis.h
Message-ID: <20100730201909.51F602A6C12C@llvm.org>

Author: nicholas
Date: Fri Jul 30 15:19:09 2010
New Revision: 109883

URL: http://llvm.org/viewvc/llvm-project?rev=109883&view=rev
Log:
LibCallAliasAnalysis uses multiple inheritance, so it needs to implement
getAdjustedAnalysisPointer. Part of a fix to PR7760.

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

Modified: llvm/trunk/include/llvm/Analysis/LibCallAliasAnalysis.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/LibCallAliasAnalysis.h?rev=109883&r1=109882&r2=109883&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Analysis/LibCallAliasAnalysis.h (original)
+++ llvm/trunk/include/llvm/Analysis/LibCallAliasAnalysis.h Fri Jul 30 15:19:09 2010
@@ -49,6 +49,16 @@
       return false;
     }
     
+    /// getAdjustedAnalysisPointer - This method is used when a pass implements
+    /// an analysis interface through multiple inheritance.  If needed, it
+    /// should override this to adjust the this pointer as needed for the
+    /// specified pass info.
+    virtual void *getAdjustedAnalysisPointer(const PassInfo *PI) {
+      if (PI->isPassID(&AliasAnalysis::ID))
+        return (AliasAnalysis*)this;
+      return this;
+    }
+    
   private:
     ModRefResult AnalyzeLibCallDetails(const LibCallFunctionInfo *FI,
                                        CallSite CS, Value *P, unsigned Size);




From stoklund at 2pi.dk  Fri Jul 30 15:23:46 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Fri, 30 Jul 2010 20:23:46 -0000
Subject: [llvm-commits] [test-suite] r109884 - in
 /test-suite/trunk/External/SPEC/CFP2006/444.namd: 444.namd.reference_output
 Makefile
Message-ID: <20100730202346.281762A6C12C@llvm.org>

Author: stoklund
Date: Fri Jul 30 15:23:45 2010
New Revision: 109884

URL: http://llvm.org/viewvc/llvm-project?rev=109884&view=rev
Log:
Let 444.namd write its results to stdout instead of namd.out.

RunSafely.sh is already redirecting stdout to namd.out, and we want to avoid the
race opening the output file.

The reference output is updated to contain the stdout messages that were
overwritten before.

Modified:
    test-suite/trunk/External/SPEC/CFP2006/444.namd/444.namd.reference_output
    test-suite/trunk/External/SPEC/CFP2006/444.namd/Makefile

Modified: test-suite/trunk/External/SPEC/CFP2006/444.namd/444.namd.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/SPEC/CFP2006/444.namd/444.namd.reference_output?rev=109884&r1=109883&r2=109884&view=diff
==============================================================================
--- test-suite/trunk/External/SPEC/CFP2006/444.namd/444.namd.reference_output (original)
+++ test-suite/trunk/External/SPEC/CFP2006/444.namd/444.namd.reference_output Fri Jul 30 15:23:45 2010
@@ -1,3 +1,6 @@
+*** TEST RUN - 1 ITERATIONS ***
+sending output to stdout
+iteration 0: 1 0 0
 RESULTSET_BEGIN
 92224 1 0 0
 -355359.705797 0.000000 24229.348329
@@ -11,6 +14,7 @@
 0.000000 0.000000 0.000000 0.000000
 0.000000 -0.000000 0.000000 24.034240
 RESULTSET_END
+iteration 0: 1 1 0
 RESULTSET_BEGIN
 92224 1 1 0
 -355359.705797 1175488.426964 24229.348329
@@ -24,6 +28,7 @@
 -0.000000 -0.000000 -0.000000 0.654402
 0.000000 -0.000000 0.000000 24.079992
 RESULTSET_END
+iteration 0: 1 1 1
 RESULTSET_BEGIN
 92224 1 1 1
 0.000000 820128.721167 24229.348329
@@ -37,6 +42,7 @@
 0.000000 -0.000000 0.000000 24.079992
 0.000000 -0.000000 0.000000 24.079992
 RESULTSET_END
+iteration 0: 0 0 0
 RESULTSET_BEGIN
 92224 0 0 0
 0.000000 0.000000 0.000000
@@ -50,6 +56,7 @@
 0.000000 0.000000 0.000000 0.000000
 0.000000 -0.000000 0.000000 24.034240
 RESULTSET_END
+iteration 0: 0 1 0
 RESULTSET_BEGIN
 92224 0 1 0
 0.000000 0.000000 0.000000
@@ -63,6 +70,7 @@
 -0.000000 -0.000000 -0.000000 0.654402
 0.000000 -0.000000 0.000000 24.079992
 RESULTSET_END
+iteration 0: 0 1 1
 RESULTSET_BEGIN
 92224 0 1 1
 0.000000 0.000000 0.000000
@@ -76,4 +84,6 @@
 0.000000 -0.000000 0.000000 24.079992
 0.000000 -0.000000 0.000000 24.079992
 RESULTSET_END
+*** TEST RUN - 1 ITERATIONS ***
+SUCCESSFUL COMPLETION
 exit 0

Modified: test-suite/trunk/External/SPEC/CFP2006/444.namd/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/SPEC/CFP2006/444.namd/Makefile?rev=109884&r1=109883&r2=109884&view=diff
==============================================================================
--- test-suite/trunk/External/SPEC/CFP2006/444.namd/Makefile (original)
+++ test-suite/trunk/External/SPEC/CFP2006/444.namd/Makefile Fri Jul 30 15:23:45 2010
@@ -11,7 +11,7 @@
 # This test is over the 500s default limit on ARM. Thumb1 takes a *long* time.
 RUNTIMELIMIT = 3000
 
-RUN_OPTIONS      = --input namd.input --iterations 1 --output namd.out
+RUN_OPTIONS      = --input namd.input --iterations 1 --output -
 STDOUT_FILENAME := namd.out
 
 LDFLAGS = -lstdc++ -lm




From stoklund at 2pi.dk  Fri Jul 30 15:23:48 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Fri, 30 Jul 2010 20:23:48 -0000
Subject: [llvm-commits] [test-suite] r109885 - in
 /test-suite/trunk/External/SPEC: CINT95/147.vortex/Makefile Makefile.spec
Message-ID: <20100730202348.CB6DE2A6C12D@llvm.org>

Author: stoklund
Date: Fri Jul 30 15:23:48 2010
New Revision: 109885

URL: http://llvm.org/viewvc/llvm-project?rev=109885&view=rev
Log:
Fix stdout race for 147.vortex.

This test insists on writing its output to a file, so we need to concatenate it
with the exit code that RunSafely.sh appended to STDOUT_FILENAME.

Modified:
    test-suite/trunk/External/SPEC/CINT95/147.vortex/Makefile
    test-suite/trunk/External/SPEC/Makefile.spec

Modified: test-suite/trunk/External/SPEC/CINT95/147.vortex/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/SPEC/CINT95/147.vortex/Makefile?rev=109885&r1=109884&r2=109885&view=diff
==============================================================================
--- test-suite/trunk/External/SPEC/CINT95/147.vortex/Makefile (original)
+++ test-suite/trunk/External/SPEC/CINT95/147.vortex/Makefile Fri Jul 30 15:23:48 2010
@@ -2,12 +2,15 @@
 
 #include $(LEVEL)/Makefile.config
 
-STDOUT_FILENAME := vortex.out
 LDFLAGS += -lm
 LIBS    += -lm
 
 include ../../Makefile.spec95
 
+# The executable writes virtex.out, and RunSafely.sh appends the exit code to
+# $(STDOUT_FILENAME). We need both.
+LOCAL_OUTPUTS := vortex.out $(STDOUT_FILENAME)
+
 ifeq ($(ARCH),Sparc)
   ## SPEC portability note for vortex says to use this flag on 64-bit machines
   CPPFLAGS += -D__RISC_64__

Modified: test-suite/trunk/External/SPEC/Makefile.spec
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/SPEC/Makefile.spec?rev=109885&r1=109884&r2=109885&view=diff
==============================================================================
--- test-suite/trunk/External/SPEC/Makefile.spec (original)
+++ test-suite/trunk/External/SPEC/Makefile.spec Fri Jul 30 15:23:48 2010
@@ -13,7 +13,7 @@
 # Information about testing the program...
 REF_IN_DIR  := $(SPEC_BENCH_DIR)/data/$(RUN_TYPE)/input/
 REF_OUT_DIR := $(SPEC_BENCH_DIR)/data/$(RUN_TYPE)/output/
-LOCAL_OUTPUTS := $(notdir $(wildcard $(REF_OUT_DIR)/*))
+LOCAL_OUTPUTS = $(notdir $(wildcard $(REF_OUT_DIR)/*))
 
 # Sometimes a test will output its program name or some other gunk that
 # we don't want to diff against. Tests can override this definition to




From nicholas at mxc.ca  Fri Jul 30 15:27:01 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Fri, 30 Jul 2010 20:27:01 -0000
Subject: [llvm-commits] [llvm] r109886 -
	/llvm/trunk/lib/Transforms/Scalar/LICM.cpp
Message-ID: <20100730202701.CB1522A6C12C@llvm.org>

Author: nicholas
Date: Fri Jul 30 15:27:01 2010
New Revision: 109886

URL: http://llvm.org/viewvc/llvm-project?rev=109886&view=rev
Log:
Add missing newline to debug statement.

Modified:
    llvm/trunk/lib/Transforms/Scalar/LICM.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/LICM.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LICM.cpp?rev=109886&r1=109885&r2=109886&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LICM.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LICM.cpp Fri Jul 30 15:27:01 2010
@@ -457,7 +457,7 @@
 /// position, and may either delete it or move it to outside of the loop.
 ///
 void LICM::sink(Instruction &I) {
-  DEBUG(dbgs() << "LICM sinking instruction: " << I);
+  DEBUG(dbgs() << "LICM sinking instruction: " << I << "\n");
 
   SmallVector<BasicBlock*, 8> ExitBlocks;
   CurLoop->getExitBlocks(ExitBlocks);




From stoklund at 2pi.dk  Fri Jul 30 15:49:24 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Fri, 30 Jul 2010 20:49:24 -0000
Subject: [llvm-commits] [test-suite] r109888 - in
 /test-suite/trunk/MultiSource/Applications/lemon: lemon.c
 lemon.reference_output
Message-ID: <20100730204924.E1EEE2A6C12C@llvm.org>

Author: stoklund
Date: Fri Jul 30 15:49:24 2010
New Revision: 109888

URL: http://llvm.org/viewvc/llvm-project?rev=109888&view=rev
Log:
Don't print the number of parsing conflicts to stderr. The text gets merged with
stdout in a way that depends on the stdout buffer size, foiling reference
outputs.

Note that lem.nconflict is printed to stdout as well, so no test information is
lost.

Modified:
    test-suite/trunk/MultiSource/Applications/lemon/lemon.c
    test-suite/trunk/MultiSource/Applications/lemon/lemon.reference_output

Modified: test-suite/trunk/MultiSource/Applications/lemon/lemon.c
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lemon/lemon.c?rev=109888&r1=109887&r2=109888&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lemon/lemon.c (original)
+++ test-suite/trunk/MultiSource/Applications/lemon/lemon.c Fri Jul 30 15:49:24 2010
@@ -1515,9 +1515,6 @@
     printf("                   %d states, %d parser table entries, %d conflicts\n",
       lem.nstate, lem.tablesize, lem.nconflict);
   }
-  if( lem.nconflict ){
-    fprintf(stderr,"%d parsing conflicts.\n",lem.nconflict);
-  }
   exit(lem.errorcnt + lem.nconflict);
   return (lem.errorcnt + lem.nconflict);
 }

Modified: test-suite/trunk/MultiSource/Applications/lemon/lemon.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/lemon/lemon.reference_output?rev=109888&r1=109887&r2=109888&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Applications/lemon/lemon.reference_output (original)
+++ test-suite/trunk/MultiSource/Applications/lemon/lemon.reference_output Fri Jul 30 15:49:24 2010
@@ -1 +1 @@
-1ec0c020ebeb276de5d975275c657078
+b0e45716410d6ea87f60851c46267162




From resistor at mac.com  Fri Jul 30 15:56:07 2010
From: resistor at mac.com (Owen Anderson)
Date: Fri, 30 Jul 2010 20:56:07 -0000
Subject: [llvm-commits] [llvm] r109889 -
	/llvm/trunk/lib/Analysis/LazyValueInfo.cpp
Message-ID: <20100730205607.B3A3F2A6C12C@llvm.org>

Author: resistor
Date: Fri Jul 30 15:56:07 2010
New Revision: 109889

URL: http://llvm.org/viewvc/llvm-project?rev=109889&view=rev
Log:
Revert my last two patches to LVI, which recent changes have exposed a miscompilation in.

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

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109889&r1=109888&r2=109889&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Fri Jul 30 15:56:07 2010
@@ -204,7 +204,7 @@
   /// LazyValueInfoCache - This is the cache kept by LazyValueInfo which
   /// maintains information about queries across the clients' queries.
   class LazyValueInfoCache {
-  private:
+  public:
     /// BlockCacheEntryTy - This is a computed lattice value at the end of the
     /// specified basic block for a Value* that depends on context.
     typedef std::pair<BasicBlock*, LVILatticeVal> BlockCacheEntryTy;
@@ -214,6 +214,7 @@
     /// entries, allowing us to do a lookup with a binary search.
     typedef DenseMap<BasicBlock*, LVILatticeVal> ValueCacheEntryTy;
 
+  private:
     /// ValueCache - This is all of the cached information for all values,
     /// mapped from Value* to key information.
     DenseMap<Value*, ValueCacheEntryTy> ValueCache;
@@ -222,44 +223,7 @@
     /// values that are over-defined at the end of that block.  This is required
     /// for cache updating.
     DenseSet<std::pair<BasicBlock*, Value*> > OverDefinedCache;
-    
-    LVILatticeVal getBlockValue(Value *Val, BasicBlock *BB);
-    LVILatticeVal getEdgeValue(Value *Val, BasicBlock *Pred, BasicBlock *Succ);
-    LVILatticeVal &getCachedEntryForBlock(Value *Val, ValueCacheEntryTy &Cache,
-                                          BasicBlock *BB);
-    
-    /************* Begin Per-Query State *************/
-    
-    ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
-    /// added to cache but that are not in sorted order.
-    DenseSet<std::pair<BasicBlock*,Value*> > NewBlockInfo;
-    
-    /// QuerySetup - An RAII helper to construct and tear-down per-query 
-    /// temporary state.
-    struct QuerySetup {
-      LazyValueInfoCache &Owner;
-      QuerySetup(LazyValueInfoCache &O) : Owner(O) {
-        assert(Owner.NewBlockInfo.empty() && "Leaked block info!");
-      }
-      
-      ~QuerySetup() {
-        // When the query is done, insert the newly discovered facts into the
-        // cache in sorted order.
-        for (DenseSet<std::pair<BasicBlock*,Value*> >::iterator
-             I = Owner.NewBlockInfo.begin(), E = Owner.NewBlockInfo.end();
-             I != E; ++I) {
-          if (Owner.ValueCache[I->second][I->first].isOverdefined())
-            Owner.OverDefinedCache.insert(*I);
-        }
-        
-        // Reset Per-Query State
-        Owner.NewBlockInfo.clear();
-      }
-    };
-    /************* End Per-Query State *************/
-    
   public:
-    LazyValueInfoCache() { }
     
     /// getValueInBlock - This is the query interface to determine the lattice
     /// value for the specified Value* at the end of the specified block.
@@ -276,21 +240,90 @@
   };
 } // end anonymous namespace
 
+namespace {
+  struct BlockCacheEntryComparator {
+    static int Compare(const void *LHSv, const void *RHSv) {
+      const LazyValueInfoCache::BlockCacheEntryTy *LHS =
+        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(LHSv);
+      const LazyValueInfoCache::BlockCacheEntryTy *RHS =
+        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(RHSv);
+      if (LHS->first < RHS->first)
+        return -1;
+      if (LHS->first > RHS->first)
+        return 1;
+      return 0;
+    }
+    
+    bool operator()(const LazyValueInfoCache::BlockCacheEntryTy &LHS,
+                    const LazyValueInfoCache::BlockCacheEntryTy &RHS) const {
+      return LHS.first < RHS.first;
+    }
+  };
+}
+
+//===----------------------------------------------------------------------===//
+//                              LVIQuery Impl
+//===----------------------------------------------------------------------===//
+
+namespace {
+  /// LVIQuery - This is a transient object that exists while a query is
+  /// being performed.
+  ///
+  /// TODO: Reuse LVIQuery instead of recreating it for every query, this avoids
+  /// reallocation of the densemap on every query.
+  class LVIQuery {
+    typedef LazyValueInfoCache::BlockCacheEntryTy BlockCacheEntryTy;
+    typedef LazyValueInfoCache::ValueCacheEntryTy ValueCacheEntryTy;
+    
+    /// This is the current value being queried for.
+    Value *Val;
+    
+    /// This is all of the cached information about this value.
+    ValueCacheEntryTy &Cache;
+    
+    /// This tracks, for each block, what values are overdefined.
+    DenseSet<std::pair<BasicBlock*, Value*> > &OverDefinedCache;
+    
+    ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
+    /// added to cache but that are not in sorted order.
+    DenseSet<BasicBlock*> NewBlockInfo;
+  public:
+    
+    LVIQuery(Value *V, ValueCacheEntryTy &VC,
+             DenseSet<std::pair<BasicBlock*, Value*> > &ODC)
+      : Val(V), Cache(VC), OverDefinedCache(ODC) {
+    }
+
+    ~LVIQuery() {
+      // When the query is done, insert the newly discovered facts into the
+      // cache in sorted order.
+      if (NewBlockInfo.empty()) return;
+      
+      for (DenseSet<BasicBlock*>::iterator I = NewBlockInfo.begin(),
+           E = NewBlockInfo.end(); I != E; ++I) {
+        if (Cache[*I].isOverdefined())
+          OverDefinedCache.insert(std::make_pair(*I, Val));
+      }
+    }
+
+    LVILatticeVal getBlockValue(BasicBlock *BB);
+    LVILatticeVal getEdgeValue(BasicBlock *FromBB, BasicBlock *ToBB);
+
+  private:
+    LVILatticeVal &getCachedEntryForBlock(BasicBlock *BB);
+  };
+} // end anonymous namespace
 
 /// getCachedEntryForBlock - See if we already have a value for this block.  If
 /// so, return it, otherwise create a new entry in the Cache map to use.
-LVILatticeVal& 
-LazyValueInfoCache::getCachedEntryForBlock(Value *Val, ValueCacheEntryTy &Cache,
-                                           BasicBlock *BB) {
-  NewBlockInfo.insert(std::make_pair(BB, Val));
+LVILatticeVal &LVIQuery::getCachedEntryForBlock(BasicBlock *BB) {
+  NewBlockInfo.insert(BB);
   return Cache[BB];
 }
 
-LVILatticeVal
-LazyValueInfoCache::getBlockValue(Value *Val, BasicBlock *BB) {
+LVILatticeVal LVIQuery::getBlockValue(BasicBlock *BB) {
   // See if we already have a value for this block.
-  ValueCacheEntryTy &Cache = ValueCache[Val];
-  LVILatticeVal &BBLV = getCachedEntryForBlock(Val, Cache, BB);
+  LVILatticeVal &BBLV = getCachedEntryForBlock(BB);
   
   // If we've already computed this block's value, return it.
   if (!BBLV.isUndefined()) {
@@ -312,7 +345,7 @@
     // Loop over all of our predecessors, merging what we know from them into
     // result.
     for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
-      Result.mergeIn(getEdgeValue(Val, *PI, BB));
+      Result.mergeIn(getEdgeValue(*PI, BB));
       
       // If we hit overdefined, exit early.  The BlockVals entry is already set
       // to overdefined.
@@ -334,7 +367,7 @@
     
     // Return the merged value, which is more precise than 'overdefined'.
     assert(!Result.isOverdefined());
-    return getCachedEntryForBlock(Val, Cache, BB) = Result;
+    return getCachedEntryForBlock(BB) = Result;
   }
   
   // If this value is defined by an instruction in this block, we have to
@@ -351,13 +384,12 @@
 
   LVILatticeVal Result;
   Result.markOverdefined();
-  return getCachedEntryForBlock(Val, Cache, BB) = Result;
+  return getCachedEntryForBlock(BB) = Result;
 }
 
 
 /// getEdgeValue - This method attempts to infer more complex 
-LVILatticeVal LazyValueInfoCache::getEdgeValue(Value *Val, 
-                                         BasicBlock *BBFrom, BasicBlock *BBTo) {
+LVILatticeVal LVIQuery::getEdgeValue(BasicBlock *BBFrom, BasicBlock *BBTo) {
   // TODO: Handle more complex conditionals.  If (v == 0 || v2 < 1) is false, we
   // know that v != 0.
   if (BranchInst *BI = dyn_cast<BranchInst>(BBFrom->getTerminator())) {
@@ -411,9 +443,14 @@
   }
   
   // Otherwise see if the value is known in the block.
-  return getBlockValue(Val, BBFrom);
+  return getBlockValue(BBFrom);
 }
 
+
+//===----------------------------------------------------------------------===//
+//                         LazyValueInfoCache Impl
+//===----------------------------------------------------------------------===//
+
 LVILatticeVal LazyValueInfoCache::getValueInBlock(Value *V, BasicBlock *BB) {
   // If already a constant, there is nothing to compute.
   if (Constant *VC = dyn_cast<Constant>(V))
@@ -422,8 +459,8 @@
   DEBUG(dbgs() << "LVI Getting block end value " << *V << " at '"
         << BB->getName() << "'\n");
   
-  QuerySetup QS(*this);
-  LVILatticeVal Result = getBlockValue(V, BB);
+  LVILatticeVal Result = LVIQuery(V, ValueCache[V], 
+                                  OverDefinedCache).getBlockValue(BB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   return Result;
@@ -438,8 +475,9 @@
   DEBUG(dbgs() << "LVI Getting edge value " << *V << " from '"
         << FromBB->getName() << "' to '" << ToBB->getName() << "'\n");
   
-  QuerySetup QS(*this);
-  LVILatticeVal Result = getEdgeValue(V, FromBB, ToBB);
+  LVILatticeVal Result =
+    LVIQuery(V, ValueCache[V],
+             OverDefinedCache).getEdgeValue(FromBB, ToBB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
   




From gohman at apple.com  Fri Jul 30 16:07:05 2010
From: gohman at apple.com (Dan Gohman)
Date: Fri, 30 Jul 2010 21:07:05 -0000
Subject: [llvm-commits] [llvm] r109891 - in /llvm/trunk:
 include/llvm/Value.h lib/AsmParser/LLParser.cpp
 lib/Transforms/InstCombine/InstCombineCalls.cpp
Message-ID: <20100730210705.53B6D2A6C12C@llvm.org>

Author: djg
Date: Fri Jul 30 16:07:05 2010
New Revision: 109891

URL: http://llvm.org/viewvc/llvm-project?rev=109891&view=rev
Log:
Move MaximumAlignment to be a member of the Value class.

Modified:
    llvm/trunk/include/llvm/Value.h
    llvm/trunk/lib/AsmParser/LLParser.cpp
    llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp

Modified: llvm/trunk/include/llvm/Value.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Value.h?rev=109891&r1=109890&r2=109891&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Value.h (original)
+++ llvm/trunk/include/llvm/Value.h Fri Jul 30 16:07:05 2010
@@ -307,6 +307,10 @@
     return const_cast<Value*>(this)->DoPHITranslation(CurBB, PredBB);
   }
   
+  /// MaximumAlignment - This is the greatest alignment value supported by
+  /// load, store, and alloca instructions, and global values.
+  static const unsigned MaximumAlignment = 1u << 29;
+  
 protected:
   unsigned short getSubclassDataFromValue() const { return SubclassData; }
   void setValueSubclassData(unsigned short D) { SubclassData = D; }
@@ -401,10 +405,6 @@
   enum { NumLowBitsAvailable = 2 };
 };
 
-/// MaximumAlignment - This is the greatest alignment value supported by
-/// load, store, and alloca instructions, and global values.
-static const unsigned MaximumAlignment = 1u << 29;
-
 } // End llvm namespace
 
 #endif

Modified: llvm/trunk/lib/AsmParser/LLParser.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/AsmParser/LLParser.cpp?rev=109891&r1=109890&r2=109891&view=diff
==============================================================================
--- llvm/trunk/lib/AsmParser/LLParser.cpp (original)
+++ llvm/trunk/lib/AsmParser/LLParser.cpp Fri Jul 30 16:07:05 2010
@@ -1154,7 +1154,7 @@
   if (ParseUInt32(Alignment)) return true;
   if (!isPowerOf2_32(Alignment))
     return Error(AlignLoc, "alignment is not a power of two");
-  if (Alignment > MaximumAlignment)
+  if (Alignment > Value::MaximumAlignment)
     return Error(AlignLoc, "huge alignments are not supported yet");
   return false;
 }

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=109891&r1=109890&r2=109891&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Fri Jul 30 16:07:05 2010
@@ -109,9 +109,10 @@
   TrailZ = std::min(TrailZ, unsigned(sizeof(unsigned) * CHAR_BIT - 1));
 
   unsigned Align = 1u << std::min(BitWidth - 1, TrailZ);
+  unsigned MaxAlign = Value::MaximumAlignment;
 
   // LLVM doesn't support alignments larger than this currently.
-  Align = std::min(Align, MaximumAlignment);
+  Align = std::min(Align, MaxAlign);
 
   if (PrefAlign > Align)
     Align = EnforceKnownAlignment(V, Align, PrefAlign);




From gohman at apple.com  Fri Jul 30 16:07:30 2010
From: gohman at apple.com (Dan Gohman)
Date: Fri, 30 Jul 2010 14:07:30 -0700
Subject: [llvm-commits] [llvm] r109643 - in /llvm/trunk:
	include/llvm/Instructions.h lib/AsmParser/LLParser.cpp
	lib/Transforms/InstCombine/InstCombineCalls.cpp	lib/VMCore/Instructions.cpp
	test/Assembler/align-inst-alloca.ll	test/Assembler/align-inst-load.ll
	test/Assembler/align-inst-store.ll	test/Assembler/align-inst.ll
In-Reply-To: <C2A0EDFB-EB51-4752-90D1-11EB7803AC64@apple.com>
References: <20100728201204.EB9B22A6C12C@llvm.org>
	<C314A96D-8E9B-4EBD-AEC6-6322402F1D85@apple.com>
	<C2A0EDFB-EB51-4752-90D1-11EB7803AC64@apple.com>
Message-ID: <F1412566-DE02-47C2-9FE9-83CD84785193@apple.com>


On Jul 28, 2010, at 9:30 PM, Chris Lattner wrote:

> 
> On Jul 28, 2010, at 9:27 PM, Chris Lattner wrote:
> 
>>> 
>>> +/// MaximumAlignment - This is the greatest alignment value supported by
>>> +/// load, store, and alloca instructions.
>>> +static const unsigned MaximumAlignment = 1u << 29;
>> 
>> How about making this an enum in the Instruction class (so it is llvm::Instruction::MaximumAlignment), instead of being llvm::MaximumAlignment?
> 
> Ok, same comment, but how about Value::MaximumAlignment now?

Ok, done.

Dan




From dalej at apple.com  Fri Jul 30 16:09:48 2010
From: dalej at apple.com (Dale Johannesen)
Date: Fri, 30 Jul 2010 21:09:48 -0000
Subject: [llvm-commits] [llvm] r109892 - in /llvm/trunk:
 lib/Target/PowerPC/PPCRegisterInfo.cpp test/FrontendC/vla-1.c
Message-ID: <20100730210948.71CBF2A6C12C@llvm.org>

Author: johannes
Date: Fri Jul 30 16:09:48 2010
New Revision: 109892

URL: http://llvm.org/viewvc/llvm-project?rev=109892&view=rev
Log:
PPC doesn't supported VLA with large alignment.  This was
formerly rejected by the FE, so asserted in the BE; now the FE only
warns, so we treat it as a legitimate fatal error in PPC BE.
This means the test for the feature won't pass, so it's xfail'd.


Modified:
    llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp
    llvm/trunk/test/FrontendC/vla-1.c

Modified: llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp?rev=109892&r1=109891&r2=109892&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCRegisterInfo.cpp Fri Jul 30 16:09:48 2010
@@ -449,8 +449,8 @@
   // Get stack alignments.
   unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
   unsigned MaxAlign = MFI->getMaxAlignment();
-  assert(MaxAlign <= TargetAlign &&
-         "Dynamic alloca with large aligns not supported");
+  if (MaxAlign > TargetAlign)
+    report_fatal_error("Dynamic alloca with large aligns not supported");
 
   // Determine the previous frame's address.  If FrameSize can't be
   // represented as 16 bits or we need special alignment, then we load the

Modified: llvm/trunk/test/FrontendC/vla-1.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/vla-1.c?rev=109892&r1=109891&r2=109892&view=diff
==============================================================================
--- llvm/trunk/test/FrontendC/vla-1.c (original)
+++ llvm/trunk/test/FrontendC/vla-1.c Fri Jul 30 16:09:48 2010
@@ -1,4 +1,6 @@
 // RUN: %llvmgcc_only -std=gnu99 %s -S |& grep {warning: alignment for}
+// ppc does not support this feature, and gets a fatal error at runtime.
+// XFAIL: powerpc
 
 int foo(int a)
 {




From isanbard at gmail.com  Fri Jul 30 17:48:39 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Fri, 30 Jul 2010 22:48:39 -0000
Subject: [llvm-commits] [llvm] r109901 - in /llvm/trunk:
 include/llvm/Target/Target.td include/llvm/Target/TargetInstrDesc.h
 utils/TableGen/CodeGenInstruction.cpp utils/TableGen/CodeGenInstruction.h
 utils/TableGen/InstrInfoEmitter.cpp
Message-ID: <20100730224839.532A32A6C12C@llvm.org>

Author: void
Date: Fri Jul 30 17:48:39 2010
New Revision: 109901

URL: http://llvm.org/viewvc/llvm-project?rev=109901&view=rev
Log:
Add a "Compare" flag to the target instruction descriptor. This will be used
later to identify and possibly remove superfluous compare instructions -- those
that are testing for and setting a status flag that should already be set.

Modified:
    llvm/trunk/include/llvm/Target/Target.td
    llvm/trunk/include/llvm/Target/TargetInstrDesc.h
    llvm/trunk/utils/TableGen/CodeGenInstruction.cpp
    llvm/trunk/utils/TableGen/CodeGenInstruction.h
    llvm/trunk/utils/TableGen/InstrInfoEmitter.cpp

Modified: llvm/trunk/include/llvm/Target/Target.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Target.td?rev=109901&r1=109900&r2=109901&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Target.td (original)
+++ llvm/trunk/include/llvm/Target/Target.td Fri Jul 30 17:48:39 2010
@@ -198,6 +198,7 @@
   bit isReturn     = 0;     // Is this instruction a return instruction?
   bit isBranch     = 0;     // Is this instruction a branch instruction?
   bit isIndirectBranch = 0; // Is this instruction an indirect branch?
+  bit isCompare    = 0;     // Is this instruction a comparison instruction?
   bit isBarrier    = 0;     // Can control flow fall through this instruction?
   bit isCall       = 0;     // Is this instruction a call instruction?
   bit canFoldAsLoad = 0;    // Can this be folded as a simple memory operand?

Modified: llvm/trunk/include/llvm/Target/TargetInstrDesc.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetInstrDesc.h?rev=109901&r1=109900&r2=109901&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetInstrDesc.h (original)
+++ llvm/trunk/include/llvm/Target/TargetInstrDesc.h Fri Jul 30 17:48:39 2010
@@ -105,6 +105,7 @@
     IndirectBranch,
     Predicable,
     NotDuplicable,
+    Compare,
     DelaySlot,
     FoldableAsLoad,
     MayLoad,
@@ -315,7 +316,7 @@
   bool isIndirectBranch() const {
     return Flags & (1 << TID::IndirectBranch);
   }
-  
+
   /// isConditionalBranch - Return true if this is a branch which may fall
   /// through to the next instruction or may transfer control flow to some other
   /// block.  The TargetInstrInfo::AnalyzeBranch method can be used to get more
@@ -340,6 +341,11 @@
     return Flags & (1 << TID::Predicable);
   }
   
+  /// isCompare - Return true if this instruction is a comparison.
+  bool isCompare() const {
+    return Flags & (1 << TID::Compare);
+  }
+  
   /// isNotDuplicable - Return true if this instruction cannot be safely
   /// duplicated.  For example, if the instruction has a unique labels attached
   /// to it, duplicating it would cause multiple definition errors.

Modified: llvm/trunk/utils/TableGen/CodeGenInstruction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/CodeGenInstruction.cpp?rev=109901&r1=109900&r2=109901&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/CodeGenInstruction.cpp (original)
+++ llvm/trunk/utils/TableGen/CodeGenInstruction.cpp Fri Jul 30 17:48:39 2010
@@ -102,6 +102,7 @@
   isReturn     = R->getValueAsBit("isReturn");
   isBranch     = R->getValueAsBit("isBranch");
   isIndirectBranch = R->getValueAsBit("isIndirectBranch");
+  isCompare    = R->getValueAsBit("isCompare");
   isBarrier    = R->getValueAsBit("isBarrier");
   isCall       = R->getValueAsBit("isCall");
   canFoldAsLoad = R->getValueAsBit("canFoldAsLoad");

Modified: llvm/trunk/utils/TableGen/CodeGenInstruction.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/CodeGenInstruction.h?rev=109901&r1=109900&r2=109901&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/CodeGenInstruction.h (original)
+++ llvm/trunk/utils/TableGen/CodeGenInstruction.h Fri Jul 30 17:48:39 2010
@@ -123,6 +123,7 @@
     bool isReturn;
     bool isBranch;
     bool isIndirectBranch;
+    bool isCompare;
     bool isBarrier;
     bool isCall;
     bool canFoldAsLoad;

Modified: llvm/trunk/utils/TableGen/InstrInfoEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/InstrInfoEmitter.cpp?rev=109901&r1=109900&r2=109901&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/InstrInfoEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/InstrInfoEmitter.cpp Fri Jul 30 17:48:39 2010
@@ -270,6 +270,7 @@
   if (Inst.isReturn)           OS << "|(1<<TID::Return)";
   if (Inst.isBranch)           OS << "|(1<<TID::Branch)";
   if (Inst.isIndirectBranch)   OS << "|(1<<TID::IndirectBranch)";
+  if (Inst.isCompare)          OS << "|(1<<TID::Compare)";
   if (Inst.isBarrier)          OS << "|(1<<TID::Barrier)";
   if (Inst.hasDelaySlot)       OS << "|(1<<TID::DelaySlot)";
   if (Inst.isCall)             OS << "|(1<<TID::Call)";




From stoklund at 2pi.dk  Fri Jul 30 17:53:47 2010
From: stoklund at 2pi.dk (Jakob Stoklund Olesen)
Date: Fri, 30 Jul 2010 22:53:47 -0000
Subject: [llvm-commits] [test-suite] r109902 - in
 /test-suite/trunk/MultiSource/Benchmarks/Olden/health: health.c health.h
Message-ID: <20100730225347.9350B2A6C12C@llvm.org>

Author: stoklund
Date: Fri Jul 30 17:53:47 2010
New Revision: 109902

URL: http://llvm.org/viewvc/llvm-project?rev=109902&view=rev
Log:
Remove a few more dependencies on sizeof(long).

Modified:
    test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.c
    test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.h

Modified: test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.c
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.c?rev=109902&r1=109901&r2=109902&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.c (original)
+++ test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.c Fri Jul 30 17:53:47 2010
@@ -11,7 +11,8 @@
 #include <assert.h>
 
 int  max_level;
-long max_time, seed;
+long max_time;
+long long seed;
 
 struct Village *alloc_tree(int level, int label, struct Village *back) {
   if (level == 0)
@@ -129,7 +130,7 @@
     if (p->time_left == 0) { 
       s = village->seed;
       rand = my_rand(s);
-      village->seed = (long)(rand * IM);
+      village->seed = (long long)(rand * IM);
       label = village->label;
       if (rand > 0.1 || label == 0) {
         removeList(&village->hosp.assess, p);
@@ -200,7 +201,7 @@
   
   s = village->seed;
   rand = my_rand(s);
-  village->seed = (long)(rand * IM);
+  village->seed = (long long)(rand * IM);
   newseed = village->seed;
   label = village->label;
   if (rand > 0.666) {

Modified: test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.h
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.h?rev=109902&r1=109901&r2=109902&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.h (original)
+++ test-suite/trunk/MultiSource/Benchmarks/Olden/health/health.h Fri Jul 30 17:53:47 2010
@@ -21,7 +21,8 @@
 #define MASK 123459876
 
 extern int  max_level;
-extern long max_time, seed;
+extern long max_time;
+extern long long seed;
 
 struct Results {
   float                   total_patients;




From daniel at zuster.org  Fri Jul 30 18:21:49 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Fri, 30 Jul 2010 23:21:49 -0000
Subject: [llvm-commits] [test-suite] r109932 -
 /test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile
Message-ID: <20100730232149.850472A6C12D@llvm.org>

Author: ddunbar
Date: Fri Jul 30 18:21:49 2010
New Revision: 109932

URL: http://llvm.org/viewvc/llvm-project?rev=109932&view=rev
Log:
Unxfail MultiSource/Benchmarks/Olden/health for x86-32, Jakob fixed it!

Modified:
    test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile

Modified: test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile?rev=109932&r1=109931&r2=109932&view=diff
==============================================================================
--- test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile (original)
+++ test-suite/trunk/MultiSource/Benchmarks/Olden/health/Makefile Fri Jul 30 18:21:49 2010
@@ -16,10 +16,4 @@
 endif
 endif
 
-ifeq ($(ARCH),x86)
-ifdef USE_REFERENCE_OUTPUT
-EXEC_XFAILS = health
-endif
-endif
-
 include $(LEVEL)/MultiSource/Makefile.multisrc




From bob.wilson at apple.com  Fri Jul 30 18:27:59 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Fri, 30 Jul 2010 23:27:59 -0000
Subject: [llvm-commits] [llvm] r109933 -
 /llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
Message-ID: <20100730232800.09FA42A6C12C@llvm.org>

Author: bwilson
Date: Fri Jul 30 18:27:59 2010
New Revision: 109933

URL: http://llvm.org/viewvc/llvm-project?rev=109933&view=rev
Log:
Add a check in the ARM disassembler for NEON instructions that would
reference registers past the end of the NEON register file, and report them
as invalid instead of asserting when trying to print them.  PR7746.

Modified:
    llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp

Modified: llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp?rev=109933&r1=109932&r2=109933&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp Fri Jul 30 18:27:59 2010
@@ -2248,9 +2248,10 @@
 
   // We have homogeneous NEON registers for Load/Store.
   unsigned RegClass = 0;
+  bool DRegPair = UseDRegPair(Opcode);
 
   // Double-spaced registers have increments of 2.
-  unsigned Inc = DblSpaced ? 2 : 1;
+  unsigned Inc = (DblSpaced || DRegPair) ? 2 : 1;
 
   unsigned Rn = decodeRn(insn);
   unsigned Rm = decodeRm(insn);
@@ -2296,8 +2297,7 @@
     RegClass = OpInfo[OpIdx].RegClass;
     while (OpIdx < NumOps && (unsigned)OpInfo[OpIdx].RegClass == RegClass) {
       MI.addOperand(MCOperand::CreateReg(
-                      getRegisterEnum(B, RegClass, Rd,
-                                      UseDRegPair(Opcode))));
+                      getRegisterEnum(B, RegClass, Rd, DRegPair)));
       Rd += Inc;
       ++OpIdx;
     }
@@ -2316,8 +2316,7 @@
 
     while (OpIdx < NumOps && (unsigned)OpInfo[OpIdx].RegClass == RegClass) {
       MI.addOperand(MCOperand::CreateReg(
-                      getRegisterEnum(B, RegClass, Rd,
-                                      UseDRegPair(Opcode))));
+                      getRegisterEnum(B, RegClass, Rd, DRegPair)));
       Rd += Inc;
       ++OpIdx;
     }
@@ -2355,6 +2354,11 @@
     }
   }
 
+  // Accessing registers past the end of the NEON register file is not
+  // defined.
+  if (Rd > 32)
+    return false;
+
   return true;
 }
 




From evan.cheng at apple.com  Fri Jul 30 18:33:55 2010
From: evan.cheng at apple.com (Evan Cheng)
Date: Fri, 30 Jul 2010 23:33:55 -0000
Subject: [llvm-commits] [llvm] r109934 -
	/llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp
Message-ID: <20100730233355.236A82A6C12C@llvm.org>

Author: evancheng
Date: Fri Jul 30 18:33:54 2010
New Revision: 109934

URL: http://llvm.org/viewvc/llvm-project?rev=109934&view=rev
Log:
Add -disable-shifter-op to disable isel of shifter ops. On Cortex-a9 the shifts cost extra instructions so it might be better to emit them separately to take advantage of dual-issues.

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

Modified: llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp?rev=109934&r1=109933&r2=109934&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp Fri Jul 30 18:33:54 2010
@@ -36,6 +36,11 @@
 
 using namespace llvm;
 
+static cl::opt<bool>
+DisableShifterOp("disable-shifter-op", cl::Hidden,
+  cl::desc("Disable isel of shifter-op"),
+  cl::init(false));
+
 //===--------------------------------------------------------------------===//
 /// ARMDAGToDAGISel - ARM specific code to select ARM machine
 /// instructions for SelectionDAG operations.
@@ -220,6 +225,9 @@
                                               SDValue &BaseReg,
                                               SDValue &ShReg,
                                               SDValue &Opc) {
+  if (DisableShifterOp)
+    return false;
+
   ARM_AM::ShiftOpc ShOpcVal = ARM_AM::getShiftOpcForNode(N);
 
   // Don't match base register only case. That is matched to a separate
@@ -666,6 +674,9 @@
 bool ARMDAGToDAGISel::SelectT2ShifterOperandReg(SDNode *Op, SDValue N,
                                                 SDValue &BaseReg,
                                                 SDValue &Opc) {
+  if (DisableShifterOp)
+    return false;
+
   ARM_AM::ShiftOpc ShOpcVal = ARM_AM::getShiftOpcForNode(N);
 
   // Don't match base register only case. That is matched to a separate




From resistor at mac.com  Fri Jul 30 18:59:40 2010
From: resistor at mac.com (Owen Anderson)
Date: Fri, 30 Jul 2010 23:59:40 -0000
Subject: [llvm-commits] [llvm] r109935 -
	/llvm/trunk/lib/Analysis/LazyValueInfo.cpp
Message-ID: <20100730235940.8F1E12A6C12C@llvm.org>

Author: resistor
Date: Fri Jul 30 18:59:40 2010
New Revision: 109935

URL: http://llvm.org/viewvc/llvm-project?rev=109935&view=rev
Log:
Add an initial implementation of PHI translation for LazyValueInfo.  This involves rolling back some
of my earlier data structure improvements until I can ensure that there are no iterator invalidation problems.

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

Modified: llvm/trunk/lib/Analysis/LazyValueInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/LazyValueInfo.cpp?rev=109935&r1=109934&r2=109935&view=diff
==============================================================================
--- llvm/trunk/lib/Analysis/LazyValueInfo.cpp (original)
+++ llvm/trunk/lib/Analysis/LazyValueInfo.cpp Fri Jul 30 18:59:40 2010
@@ -21,6 +21,7 @@
 #include "llvm/Support/CFG.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
+#include "llvm/Support/ValueHandle.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/DenseSet.h"
 #include "llvm/ADT/PointerIntPair.h"
@@ -212,17 +213,36 @@
     /// ValueCacheEntryTy - This is all of the cached block information for
     /// exactly one Value*.  The entries are sorted by the BasicBlock* of the
     /// entries, allowing us to do a lookup with a binary search.
-    typedef DenseMap<BasicBlock*, LVILatticeVal> ValueCacheEntryTy;
+    typedef std::map<BasicBlock*, LVILatticeVal> ValueCacheEntryTy;
 
   private:
+     /// LVIValueHandle - A callback value handle update the cache when
+     /// values are erased.
+    struct LVIValueHandle : public CallbackVH {
+      LazyValueInfoCache *Parent;
+      
+      LVIValueHandle(Value *V, LazyValueInfoCache *P)
+        : CallbackVH(V), Parent(P) { }
+      
+      void deleted();
+      void allUsesReplacedWith(Value* V) {
+        deleted();
+      }
+
+      LVIValueHandle &operator=(Value *V) {
+        return *this = LVIValueHandle(V, Parent);
+      }
+    };
+
     /// ValueCache - This is all of the cached information for all values,
     /// mapped from Value* to key information.
-    DenseMap<Value*, ValueCacheEntryTy> ValueCache;
+    std::map<LVIValueHandle, ValueCacheEntryTy> ValueCache;
     
     /// OverDefinedCache - This tracks, on a per-block basis, the set of 
     /// values that are over-defined at the end of that block.  This is required
     /// for cache updating.
-    DenseSet<std::pair<BasicBlock*, Value*> > OverDefinedCache;
+    std::set<std::pair<BasicBlock*, Value*> > OverDefinedCache;
+
   public:
     
     /// getValueInBlock - This is the query interface to determine the lattice
@@ -240,27 +260,6 @@
   };
 } // end anonymous namespace
 
-namespace {
-  struct BlockCacheEntryComparator {
-    static int Compare(const void *LHSv, const void *RHSv) {
-      const LazyValueInfoCache::BlockCacheEntryTy *LHS =
-        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(LHSv);
-      const LazyValueInfoCache::BlockCacheEntryTy *RHS =
-        static_cast<const LazyValueInfoCache::BlockCacheEntryTy *>(RHSv);
-      if (LHS->first < RHS->first)
-        return -1;
-      if (LHS->first > RHS->first)
-        return 1;
-      return 0;
-    }
-    
-    bool operator()(const LazyValueInfoCache::BlockCacheEntryTy &LHS,
-                    const LazyValueInfoCache::BlockCacheEntryTy &RHS) const {
-      return LHS.first < RHS.first;
-    }
-  };
-}
-
 //===----------------------------------------------------------------------===//
 //                              LVIQuery Impl
 //===----------------------------------------------------------------------===//
@@ -278,20 +277,24 @@
     /// This is the current value being queried for.
     Value *Val;
     
+    /// This is a pointer to the owning cache, for recursive queries.
+    LazyValueInfoCache &Parent;
+
     /// This is all of the cached information about this value.
     ValueCacheEntryTy &Cache;
     
     /// This tracks, for each block, what values are overdefined.
-    DenseSet<std::pair<BasicBlock*, Value*> > &OverDefinedCache;
+    std::set<std::pair<BasicBlock*, Value*> > &OverDefinedCache;
     
     ///  NewBlocks - This is a mapping of the new BasicBlocks which have been
     /// added to cache but that are not in sorted order.
     DenseSet<BasicBlock*> NewBlockInfo;
   public:
     
-    LVIQuery(Value *V, ValueCacheEntryTy &VC,
-             DenseSet<std::pair<BasicBlock*, Value*> > &ODC)
-      : Val(V), Cache(VC), OverDefinedCache(ODC) {
+    LVIQuery(Value *V, LazyValueInfoCache &P,
+             ValueCacheEntryTy &VC,
+             std::set<std::pair<BasicBlock*, Value*> > &ODC)
+      : Val(V), Parent(P), Cache(VC), OverDefinedCache(ODC) {
     }
 
     ~LVIQuery() {
@@ -314,6 +317,20 @@
   };
 } // end anonymous namespace
 
+void LazyValueInfoCache::LVIValueHandle::deleted() {
+  Parent->ValueCache.erase(*this);
+  for (std::set<std::pair<BasicBlock*, Value*> >::iterator
+       I = Parent->OverDefinedCache.begin(),
+       E = Parent->OverDefinedCache.end();
+       I != E; ) {
+    std::set<std::pair<BasicBlock*, Value*> >::iterator tmp = I;
+    ++I;
+    if (tmp->second == getValPtr())
+      Parent->OverDefinedCache.erase(tmp);
+  }
+}
+
+
 /// getCachedEntryForBlock - See if we already have a value for this block.  If
 /// so, return it, otherwise create a new entry in the Cache map to use.
 LVILatticeVal &LVIQuery::getCachedEntryForBlock(BasicBlock *BB) {
@@ -373,8 +390,27 @@
   // If this value is defined by an instruction in this block, we have to
   // process it here somehow or return overdefined.
   if (PHINode *PN = dyn_cast<PHINode>(BBI)) {
-    (void)PN;
-    // TODO: PHI Translation in preds.
+    LVILatticeVal Result;  // Start Undefined.
+    
+    // Loop over all of our predecessors, merging what we know from them into
+    // result.
+    for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
+      Value* PhiVal = PN->getIncomingValueForBlock(*PI);
+      Result.mergeIn(Parent.getValueOnEdge(PhiVal, *PI, BB));
+      
+      // If we hit overdefined, exit early.  The BlockVals entry is already set
+      // to overdefined.
+      if (Result.isOverdefined()) {
+        DEBUG(dbgs() << " compute BB '" << BB->getName()
+                     << "' - overdefined because of pred.\n");
+        return Result;
+      }
+    }
+    
+    // Return the merged value, which is more precise than 'overdefined'.
+    assert(!Result.isOverdefined());
+    return getCachedEntryForBlock(BB) = Result;
+
   } else {
     
   }
@@ -459,7 +495,8 @@
   DEBUG(dbgs() << "LVI Getting block end value " << *V << " at '"
         << BB->getName() << "'\n");
   
-  LVILatticeVal Result = LVIQuery(V, ValueCache[V], 
+  LVILatticeVal Result = LVIQuery(V, *this,
+                                  ValueCache[LVIValueHandle(V, this)], 
                                   OverDefinedCache).getBlockValue(BB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
@@ -476,7 +513,7 @@
         << FromBB->getName() << "' to '" << ToBB->getName() << "'\n");
   
   LVILatticeVal Result =
-    LVIQuery(V, ValueCache[V],
+    LVIQuery(V, *this, ValueCache[LVIValueHandle(V, this)],
              OverDefinedCache).getEdgeValue(FromBB, ToBB);
   
   DEBUG(dbgs() << "  Result = " << Result << "\n");
@@ -500,7 +537,7 @@
   worklist.push_back(OldSucc);
   
   DenseSet<Value*> ClearSet;
-  for (DenseSet<std::pair<BasicBlock*, Value*> >::iterator
+  for (std::set<std::pair<BasicBlock*, Value*> >::iterator
        I = OverDefinedCache.begin(), E = OverDefinedCache.end(); I != E; ++I) {
     if (I->first == OldSucc)
       ClearSet.insert(I->second);
@@ -521,12 +558,12 @@
     for (DenseSet<Value*>::iterator I = ClearSet.begin(),E = ClearSet.end();
          I != E; ++I) {
       // If a value was marked overdefined in OldSucc, and is here too...
-      DenseSet<std::pair<BasicBlock*, Value*> >::iterator OI =
+      std::set<std::pair<BasicBlock*, Value*> >::iterator OI =
         OverDefinedCache.find(std::make_pair(ToUpdate, *I));
       if (OI == OverDefinedCache.end()) continue;
 
       // Remove it from the caches.
-      ValueCacheEntryTy &Entry = ValueCache[*I];
+      ValueCacheEntryTy &Entry = ValueCache[LVIValueHandle(*I, this)];
       ValueCacheEntryTy::iterator CI = Entry.find(ToUpdate);
         
       assert(CI != Entry.end() && "Couldn't find entry to update?");




From rafael.espindola at gmail.com  Fri Jul 30 19:32:17 2010
From: rafael.espindola at gmail.com (Rafael Espindola)
Date: Sat, 31 Jul 2010 00:32:17 -0000
Subject: [llvm-commits] [llvm] r109936 - in /llvm/trunk:
 include/llvm/Transforms/IPO.h lib/Transforms/IPO/LoopExtractor.cpp
 tools/bugpoint/ExtractFunction.cpp
Message-ID: <20100731003218.089E72A6C12C@llvm.org>

Author: rafael
Date: Fri Jul 30 19:32:17 2010
New Revision: 109936

URL: http://llvm.org/viewvc/llvm-project?rev=109936&view=rev
Log:
The BlockExtractorPass() constructor was not reading the BlockFile and that was
exactly what bugpoint expected it to do.

There was also only one user of
BlockExtractorPass(const std::vector<BasicBlock*> &B), so just remove it and
make BlockExtractorPass read BlockFile.

This fixes bugpoint's block extraction.

Nick, please review.

Modified:
    llvm/trunk/include/llvm/Transforms/IPO.h
    llvm/trunk/lib/Transforms/IPO/LoopExtractor.cpp
    llvm/trunk/tools/bugpoint/ExtractFunction.cpp

Modified: llvm/trunk/include/llvm/Transforms/IPO.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Transforms/IPO.h?rev=109936&r1=109935&r2=109936&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Transforms/IPO.h (original)
+++ llvm/trunk/include/llvm/Transforms/IPO.h Fri Jul 30 19:32:17 2010
@@ -181,7 +181,7 @@
 /// createBlockExtractorPass - This pass extracts all blocks (except those
 /// specified in the argument list) from the functions in the module.
 ///
-ModulePass *createBlockExtractorPass(const std::vector<BasicBlock*> &BTNE);
+ModulePass *createBlockExtractorPass();
 
 /// createStripDeadPrototypesPass - This pass removes any function declarations
 /// (prototypes) that are not used.

Modified: llvm/trunk/lib/Transforms/IPO/LoopExtractor.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/IPO/LoopExtractor.cpp?rev=109936&r1=109935&r2=109936&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/IPO/LoopExtractor.cpp (original)
+++ llvm/trunk/lib/Transforms/IPO/LoopExtractor.cpp Fri Jul 30 19:32:17 2010
@@ -147,12 +147,10 @@
     std::vector<std::pair<std::string, std::string> > BlocksToNotExtractByName;
   public:
     static char ID; // Pass identification, replacement for typeid
-    explicit BlockExtractorPass(const std::vector<BasicBlock*> &B) 
-      : ModulePass(&ID), BlocksToNotExtract(B) {
+    BlockExtractorPass() : ModulePass(&ID) {
       if (!BlockFile.empty())
         LoadFile(BlockFile.c_str());
     }
-    BlockExtractorPass() : ModulePass(&ID) {}
 
     bool runOnModule(Module &M);
   };
@@ -166,9 +164,9 @@
 // createBlockExtractorPass - This pass extracts all blocks (except those
 // specified in the argument list) from the functions in the module.
 //
-ModulePass *llvm::createBlockExtractorPass(const std::vector<BasicBlock*> &BTNE)
+ModulePass *llvm::createBlockExtractorPass()
 {
-  return new BlockExtractorPass(BTNE);
+  return new BlockExtractorPass();
 }
 
 void BlockExtractorPass::LoadFile(const char *Filename) {

Modified: llvm/trunk/tools/bugpoint/ExtractFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ExtractFunction.cpp?rev=109936&r1=109935&r2=109936&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ExtractFunction.cpp (original)
+++ llvm/trunk/tools/bugpoint/ExtractFunction.cpp Fri Jul 30 19:32:17 2010
@@ -354,8 +354,7 @@
   const char *ExtraArg = uniqueFN.c_str();
 
   std::vector<const PassInfo*> PI;
-  std::vector<BasicBlock *> EmptyBBs; // This parameter is ignored.
-  PI.push_back(getPI(createBlockExtractorPass(EmptyBBs)));
+  PI.push_back(getPI(createBlockExtractorPass()));
   Module *Ret = runPassesOn(M, PI, false, 1, &ExtraArg);
 
   uniqueFilename.eraseFromDisk(); // Free disk space




From echristo at apple.com  Fri Jul 30 21:35:10 2010
From: echristo at apple.com (Eric Christopher)
Date: Sat, 31 Jul 2010 02:35:10 -0000
Subject: [llvm-commits] [test-suite] r109940 -
	/test-suite/trunk/External/Makefile
Message-ID: <20100731023510.9D7E02A6C12C@llvm.org>

Author: echristo
Date: Fri Jul 30 21:35:10 2010
New Revision: 109940

URL: http://llvm.org/viewvc/llvm-project?rev=109940&view=rev
Log:
Turn these back on after Daniel's changes for multiple results.

Modified:
    test-suite/trunk/External/Makefile

Modified: test-suite/trunk/External/Makefile
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/Makefile?rev=109940&r1=109939&r2=109940&view=diff
==============================================================================
--- test-suite/trunk/External/Makefile (original)
+++ test-suite/trunk/External/Makefile Fri Jul 30 21:35:10 2010
@@ -8,9 +8,7 @@
 #
 # Create the list of directories to compile
 #
-PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim
-# Temporarily disable skidmarks
-# skidmarks10
+PARALLEL_DIRS := SPEC Povray Namd FPGrowth BoxedSim skidmarks10
 
 ifneq (,$(findstring x86,$(ARCH)))
 # Nurbs uses SSE and only works on x86 and x86_64.




From daniel at zuster.org  Fri Jul 30 23:29:23 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 04:29:23 -0000
Subject: [llvm-commits] [zorg] r109942 -
	/zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py
Message-ID: <20100731042923.1E7F42A6C12C@llvm.org>

Author: ddunbar
Date: Fri Jul 30 23:29:22 2010
New Revision: 109942

URL: http://llvm.org/viewvc/llvm-project?rev=109942&view=rev
Log:
buildbot/LLVM: Add config name, default branch, and configured targets options.

Modified:
    zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py

Modified: zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py?rev=109942&r1=109941&r2=109942&view=diff
==============================================================================
--- zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py (original)
+++ zorg/trunk/zorg/buildbot/builders/LLVMBuilder.py Fri Jul 30 23:29:22 2010
@@ -9,11 +9,14 @@
 
 from zorg.buildbot.commands.ClangTestCommand import ClangTestCommand
 
+from Util import getConfigArgs
+
 def getLLVMBuildFactory(triple=None, clean=True, test=True,
                         expensive_checks=False, examples=False, valgrind=False,
                         valgrindLeakCheck=False, valgrindSuppressions=None,
                         jobs='%(jobs)s', timeout=20, make='make',
-                        enable_shared=False):
+                        enable_shared=False, enable_targets=None, defaultBranch='trunk',
+                        config_name='Debug+Asserts'):
     f = buildbot.process.factory.BuildFactory()
 
     # Determine the build directory.
@@ -26,15 +29,14 @@
     # Checkout sources.
     f.addStep(SVN(name='svn-llvm',
                   mode='update', baseURL='http://llvm.org/svn/llvm-project/llvm/',
-                  defaultBranch='trunk',
+                  defaultBranch=defaultBranch,
                   workdir='llvm'))
 
     # Force without llvm-gcc so we don't run afoul of Frontend test failures.
     configure_args = ["./configure", "--without-llvmgcc", "--without-llvmgxx"]
-    config_name = 'Debug+Asserts'
-    if expensive_checks:
-        configure_args.append('--enable-expensive-checks')
-        config_name += '+Checks'
+    configure_args += getConfigArgs(config_name)
+    if enable_targets is not None:
+        configure_args.append('--enable-targets %s' % enable_targets)
     if triple:
         configure_args += ['--build=%s' % triple,
                            '--host=%s' % triple,




From daniel at zuster.org  Fri Jul 30 23:30:16 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 04:30:16 -0000
Subject: [llvm-commits] [zorg] r109943 -
	/zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py
Message-ID: <20100731043016.2E5FE2A6C12C@llvm.org>

Author: ddunbar
Date: Fri Jul 30 23:30:16 2010
New Revision: 109943

URL: http://llvm.org/viewvc/llvm-project?rev=109943&view=rev
Log:
buildbot: Start sketching a KLEE build factory.

Added:
    zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py

Added: zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py?rev=109943&view=auto
==============================================================================
--- zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py (added)
+++ zorg/trunk/zorg/buildbot/builders/KLEEBuilder.py Fri Jul 30 23:30:16 2010
@@ -0,0 +1,75 @@
+import os
+
+import buildbot
+import buildbot.process.factory
+from buildbot.steps.source import SVN
+from buildbot.steps.shell import Configure, ShellCommand
+from buildbot.steps.shell import WarningCountingShellCommand
+from buildbot.process.properties import WithProperties
+
+import ClangBuilder
+import LLVMBuilder
+reload(LLVMBuilder)
+import LLVMBuilder
+from Util import getConfigArgs
+
+def getKLEEBuildFactory(triple, jobs='%(jobs)d', llvm_branch='trunk',
+                        config_name='Release+Asserts', clean=True, llvmgccdir=None,
+                        *args, **kwargs):
+    if False:
+        f = buildbot.process.factory.BuildFactory()
+
+        # Determine the build directory.
+        f.addStep(buildbot.steps.shell.SetProperty(name="get_builddir",
+                                                   command=["pwd"],
+                                                   property="builddir",
+                                                   description="set build dir",
+                                                   workdir="."))
+    else:
+        # If we are building from trunk, we need to build Clang as
+        # well so we have access to an LLVM capable compiler.
+        if llvm_branch == 'trunk':
+            f = ClangBuilder.getClangBuildFactory(triple, jobs=jobs,
+                                                  stage1_config=config_name, extra_configure_args=['--with-built-clang',
+                                                                                                   '--enable-targets=host'],
+                                                  clean=clean, test=False, *args, **kwargs)
+        else:
+            f = LLVMBuilder.getLLVMBuildFactory(triple, jobs=jobs, defaultBranch=llvm_branch,
+                                                config_name=config_name, enable_targets='host',
+                                                clean=clean, test=False, *args, **kwargs)
+
+    # Checkout sources.
+    f.addStep(SVN(name='svn-klee',
+                  mode='update', baseURL='http://llvm.org/svn/llvm-project/klee/',
+                  defaultBranch='trunk', workdir='klee'))
+
+    # Configure.
+    configure_args = ["./configure", WithProperties("--with-llvm=%(builddir)s/llvm")]
+    configure_args += getConfigArgs(config_name)
+    if triple:
+        configure_args += ['--build=%s' % triple,
+                           '--host=%s' % triple,
+                           '--target=%s' % triple]
+    if llvmgccdir:
+        configure_args += ['--with-llvmgccdir=%s' % llvmgccdir]
+    f.addStep(Configure(command=configure_args, workdir='klee',
+                        description=['configure','klee',config_name]))
+
+    # Clean, if requested.
+    if clean:
+        f.addStep(WarningCountingShellCommand(name="clean-klee",
+                                              command=['make', 'clean'],
+                                              haltOnFailure=True,
+                                              description="clean klee",
+                                              workdir='klee'))
+
+    # Compile.
+    f.addStep(WarningCountingShellCommand(name="compile",
+                                          command=['nice', '-n', '10',
+                                                   'make', WithProperties("-j%s" % jobs)],
+                                          haltOnFailure=True, description="compile klee",
+                                          workdir='klee'))
+
+    return f
+
+




From nicholas at mxc.ca  Fri Jul 30 23:46:24 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Fri, 30 Jul 2010 21:46:24 -0700
Subject: [llvm-commits] [llvm] r109385 - /llvm/trunk/Makefile.rules
In-Reply-To: <AANLkTimhjk8y207XRwz1awGqDRBYy7vCk0zx_FZ-+RSa@mail.gmail.com>
References: <20100725223304.C4B6E2A6C12C@llvm.org>
	<AANLkTimhjk8y207XRwz1awGqDRBYy7vCk0zx_FZ-+RSa@mail.gmail.com>
Message-ID: <4C53AAA0.9050506@mxc.ca>

NAKAMURA Takumi wrote:
> Nick,
>
> I found this patch was incomplete. the improved one is attached.
>
> with previous patch, the below is enabled on cygwin/mingw.
>> ifneq ($(DARWIN_MAJVERS),4)
>>    LD.Flags += $(RPATH) -Wl, at executable_path/../lib
>> endif
>
> I guess it would be less harmful but I did not forget former patch. XD

Does:

   ifeq (,$(findstring $(HOST_OS),Darwin Cygwin MingW))
     LD.Flags += ...
   endif

work? And do you think that reads more clearly? Elsewhere in the 
makefile, we also have this:

   ifeq ($(HOST_OS), $(filter $(HOST_OS), Linux NetBSD FreeBSD))
   ifneq ($(ARCH), Mips)
     LD.Flags += 
-Wl,--version-script=$(LLVM_SRC_ROOT)/autoconf/ExportMap.map
   endif
   endif

which is why I thought the construct was correct. Should we update this 
one too? Or is there a subtlety I'm missing.

Nick

>
> ...Takumi
>
>
>
> 2010/7/26 Nick Lewycky<nicholas at mxc.ca>:
>> Author: nicholas
>> Date: Sun Jul 25 17:33:04 2010
>> New Revision: 109385
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=109385&view=rev
>> Log:
>> Don't pass -export-dynamic to the linker on Cygwin and MinGW. These platforms
>> accept the flag and do nothing but warn about it, cleverly bypassing our
>> configure-based detection system. Patch by Takumi Nakamura!
>>
>> Modified:
>>     llvm/trunk/Makefile.rules
>>
>> Modified: llvm/trunk/Makefile.rules
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=109385&r1=109384&r2=109385&view=diff
>> ==============================================================================
>> --- llvm/trunk/Makefile.rules (original)
>> +++ llvm/trunk/Makefile.rules Sun Jul 25 17:33:04 2010
>> @@ -618,7 +618,7 @@
>>   endif
>>
>>   # Adjust linker flags for building an executable
>> -ifneq ($(HOST_OS),Darwin)
>> +ifneq ($(HOST_OS), $(filter $(HOST_OS), Darwin Cygwin MingW))
>>   ifdef TOOLNAME
>>    LD.Flags += $(RPATH) -Wl,'$$ORIGIN/../lib'
>>    ifdef EXAMPLE_TOOL
>>
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>



From daniel at zuster.org  Sat Jul 31 00:26:20 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 05:26:20 -0000
Subject: [llvm-commits] [zorg] r109944 -
	/zorg/trunk/zorg/buildbot/commands/StandardizedTest.py
Message-ID: <20100731052620.BD2342A6C12C@llvm.org>

Author: ddunbar
Date: Sat Jul 31 00:26:20 2010
New Revision: 109944

URL: http://llvm.org/viewvc/llvm-project?rev=109944&view=rev
Log:
buildbot/StandardizedTest: Strip test names, and don't warn about
ignored tests which pass, if there is also an ignored test with the
same name that fails. Nice naming!

Modified:
    zorg/trunk/zorg/buildbot/commands/StandardizedTest.py

Modified: zorg/trunk/zorg/buildbot/commands/StandardizedTest.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/commands/StandardizedTest.py?rev=109944&r1=109943&r2=109944&view=diff
==============================================================================
--- zorg/trunk/zorg/buildbot/commands/StandardizedTest.py (original)
+++ zorg/trunk/zorg/buildbot/commands/StandardizedTest.py Sat Jul 31 00:26:20 2010
@@ -39,6 +39,7 @@
         lines = self.getLog(self.testLogName).readlines()
         hasIgnored = False
         for result,test,log in self.parseLog(lines):
+            test = test.strip()
             if result not in self.knownCodes:
                 raise ValueError,'test command return invalid result code!'
 
@@ -60,6 +61,19 @@
                 if log is not None and log.strip():
                     logs.append((test, log))
 
+        # Explicitly remove any ignored warnings for tests which are
+        # also in the an ignored failing set (some tests may appear
+        # twice).
+        ignored_failures = set()
+        for code in self.failingCodes:
+            results = results_by_code.get('IGNORE ' + code)
+            if results:
+                ignored_failures |= set(results)
+        for code in self.warningCodes:
+            results = results_by_code.get(code)
+            if results:
+                results_by_code[code] -= ignored_failures
+
         # Summarize result counts.
         total = failed = passed = warnings = 0
         for code in self.allKnownCodes:




From daniel at zuster.org  Sat Jul 31 00:29:16 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 05:29:16 -0000
Subject: [llvm-commits] [zorg] r109945 -
	/zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
Message-ID: <20100731052916.E0FF12A6C12C@llvm.org>

Author: ddunbar
Date: Sat Jul 31 00:29:16 2010
New Revision: 109945

URL: http://llvm.org/viewvc/llvm-project?rev=109945&view=rev
Log:
buildbot: Add a builder for running the GCC testsuite in the clang-tests repo. Also, change two-stage build to install to 'llvm.install', for consistency with single stage builds.

Modified:
    zorg/trunk/zorg/buildbot/builders/ClangBuilder.py

Modified: zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/builders/ClangBuilder.py?rev=109945&r1=109944&r2=109945&view=diff
==============================================================================
--- zorg/trunk/zorg/buildbot/builders/ClangBuilder.py (original)
+++ zorg/trunk/zorg/buildbot/builders/ClangBuilder.py Sat Jul 31 00:29:16 2010
@@ -10,6 +10,7 @@
 
 from zorg.buildbot.commands.ClangTestCommand import ClangTestCommand
 from zorg.buildbot.commands.BatchFileDownload import BatchFileDownload
+from zorg.buildbot.commands import DejaGNUCommand
 
 from Util import getConfigArgs
 
@@ -35,7 +36,7 @@
         llvm_1_objdir = "llvm.obj"
         llvm_1_installdir = "llvm.install.1"
         llvm_2_objdir = "llvm.obj.2"
-        llvm_2_installdir = "llvm.install.2"
+        llvm_2_installdir = "llvm.install"
 
     f = buildbot.process.factory.BuildFactory()
 
@@ -312,3 +313,21 @@
                                workdir="llvm\\build\\tools\\clang\\test"))
 
     return f
+
+def addClangTests(f, ignores={}):
+    make_vars = [WithProperties(
+            'CC_UNDER_TEST=%(builddir)s/llvm.install/bin/clang'),
+                 WithProperties(
+            'CXX_UNDER_TEST=%(builddir)s/llvm.install/bin/clang++')]
+    f.addStep(SVN(name='svn-clang-tests', mode='update',
+                  baseURL='http://llvm.org/svn/llvm-project/clang-tests/',
+                  defaultBranch='trunk', workdir='clang-tests'))
+    gcc_dg_ignores = ignores.get('gcc-4_2-testsuite', {})
+    for lang in ('gcc', 'g++', 'objc', 'obj-c++'):
+        f.addStep(DejaGNUCommand.DejaGNUCommand(
+            name='test-gcc-4_2-testsuite-%s' % lang,
+            command=["make", "-k", "check-%s" % lang] + make_vars,
+            description="gcc-4_2-testsuite (%s)" % lang,
+            workdir='clang-tests/gcc-4_2-testsuite',
+            logfiles={ 'dg.sum' : 'obj/%s/%s.sum' % (lang, lang) },
+            ignore=gcc_dg_ignores.get(lang, [])))




From bob.wilson at apple.com  Sat Jul 31 00:57:44 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Sat, 31 Jul 2010 05:57:44 -0000
Subject: [llvm-commits] [llvm] r109946 - in /llvm/trunk:
 lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
 test/MC/Disassembler/neon-tests.txt
Message-ID: <20100731055744.4EB8A2A6C12C@llvm.org>

Author: bwilson
Date: Sat Jul 31 00:57:44 2010
New Revision: 109946

URL: http://llvm.org/viewvc/llvm-project?rev=109946&view=rev
Log:
Add support for disassembling VMVN (immediate) instructions.  PR7747.

Modified:
    llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
    llvm/trunk/test/MC/Disassembler/neon-tests.txt

Modified: llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp?rev=109946&r1=109945&r2=109946&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp (original)
+++ llvm/trunk/lib/Target/ARM/Disassembler/ARMDisassemblerCore.cpp Sat Jul 31 00:57:44 2010
@@ -2431,10 +2431,14 @@
     break;
   case ARM::VMOVv4i16:
   case ARM::VMOVv8i16:
+  case ARM::VMVNv4i16:
+  case ARM::VMVNv8i16:
     esize = ESize16;
     break;
   case ARM::VMOVv2i32:
   case ARM::VMOVv4i32:
+  case ARM::VMVNv2i32:
+  case ARM::VMVNv4i32:
     esize = ESize32;
     break;
   case ARM::VMOVv1i64:

Modified: llvm/trunk/test/MC/Disassembler/neon-tests.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/MC/Disassembler/neon-tests.txt?rev=109946&r1=109945&r2=109946&view=diff
==============================================================================
--- llvm/trunk/test/MC/Disassembler/neon-tests.txt (original)
+++ llvm/trunk/test/MC/Disassembler/neon-tests.txt Sat Jul 31 00:57:44 2010
@@ -25,6 +25,9 @@
 # CHECK:	vmov.i64	q6, #0xFF00FF00FF
 0x75 0xce 0x81 0xf2
 
+# CHECK:	vmvn.i32	d0, #0x0
+0x30 0x00 0x80 0xf2
+
 # CHECK:	vmul.f32	d0, d0, d6
 0x16 0x0d 0x00 0xf3
 




From bigcheesegs at gmail.com  Sat Jul 31 01:22:29 2010
From: bigcheesegs at gmail.com (Michael J. Spencer)
Date: Sat, 31 Jul 2010 06:22:29 -0000
Subject: [llvm-commits] [llvm] r109947 - in /llvm/trunk:
 include/llvm/MC/MCStreamer.h lib/MC/WinCOFFStreamer.cpp
 lib/Target/X86/X86TargetMachine.cpp
Message-ID: <20100731062229.B82BC2A6C12C@llvm.org>

Author: mspencer
Date: Sat Jul 31 01:22:29 2010
New Revision: 109947

URL: http://llvm.org/viewvc/llvm-project?rev=109947&view=rev
Log:
Add relax all support to the COFF object streamer.

Modified:
    llvm/trunk/include/llvm/MC/MCStreamer.h
    llvm/trunk/lib/MC/WinCOFFStreamer.cpp
    llvm/trunk/lib/Target/X86/X86TargetMachine.cpp

Modified: llvm/trunk/include/llvm/MC/MCStreamer.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/MC/MCStreamer.h?rev=109947&r1=109946&r2=109947&view=diff
==============================================================================
--- llvm/trunk/include/llvm/MC/MCStreamer.h (original)
+++ llvm/trunk/include/llvm/MC/MCStreamer.h Sat Jul 31 01:22:29 2010
@@ -355,7 +355,8 @@
   /// Takes ownership of \arg TAB and \arg CE.
   MCStreamer *createWinCOFFStreamer(MCContext &Ctx,
                                     TargetAsmBackend &TAB,
-                                    MCCodeEmitter &CE, raw_ostream &OS);
+                                    MCCodeEmitter &CE, raw_ostream &OS,
+                                    bool RelaxAll = false);
 
   /// createLoggingStreamer - Create a machine code streamer which just logs the
   /// API calls and then dispatches to another streamer.

Modified: llvm/trunk/lib/MC/WinCOFFStreamer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/WinCOFFStreamer.cpp?rev=109947&r1=109946&r2=109947&view=diff
==============================================================================
--- llvm/trunk/lib/MC/WinCOFFStreamer.cpp (original)
+++ llvm/trunk/lib/MC/WinCOFFStreamer.cpp Sat Jul 31 01:22:29 2010
@@ -339,7 +339,10 @@
   MCStreamer *createWinCOFFStreamer(MCContext &Context,
                                     TargetAsmBackend &TAB,
                                     MCCodeEmitter &CE,
-                                    raw_ostream &OS) {
-    return new WinCOFFStreamer(Context, TAB, CE, OS);
+                                    raw_ostream &OS,
+                                    bool RelaxAll) {
+    WinCOFFStreamer *S = new WinCOFFStreamer(Context, TAB, CE, OS);
+    S->getAssembler().setRelaxAll(RelaxAll);
+    return S;
   }
 }

Modified: llvm/trunk/lib/Target/X86/X86TargetMachine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetMachine.cpp?rev=109947&r1=109946&r2=109947&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetMachine.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetMachine.cpp Sat Jul 31 01:22:29 2010
@@ -47,7 +47,7 @@
   Triple TheTriple(TT);
   switch (TheTriple.getOS()) {
   case Triple::Win32:
-    return createWinCOFFStreamer(Ctx, TAB, *_Emitter, _OS);
+    return createWinCOFFStreamer(Ctx, TAB, *_Emitter, _OS, RelaxAll);
   default:
     return createMachOStreamer(Ctx, TAB, _OS, _Emitter, RelaxAll);
   }




From bob.wilson at apple.com  Sat Jul 31 01:28:10 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Sat, 31 Jul 2010 06:28:10 -0000
Subject: [llvm-commits] [llvm] r109948 - in /llvm/trunk/lib/Target/ARM:
 ARMAsmPrinter.cpp ARMInstrInfo.td ARMInstrThumb.td ARMInstrThumb2.td
Message-ID: <20100731062810.80B942A6C12C@llvm.org>

Author: bwilson
Date: Sat Jul 31 01:28:10 2010
New Revision: 109948

URL: http://llvm.org/viewvc/llvm-project?rev=109948&view=rev
Log:
Move newlines before inline jumptables from the asm strings in .td files to
the jtblock_operand print methods.  This avoids extra newlines in the
disassembler's output.  PR7757.

Modified:
    llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp
    llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb.td
    llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

Modified: llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp?rev=109948&r1=109947&r2=109948&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMAsmPrinter.cpp Sat Jul 31 01:28:10 2010
@@ -938,7 +938,7 @@
   MCSymbol *JTISymbol = GetARMJTIPICJumpTableLabel2(JTI, MO2.getImm());
   // Can't use EmitLabel until instprinter happens, label comes out in the wrong
   // order.
-  O << *JTISymbol << ":\n";
+  O << "\n" << *JTISymbol << ":\n";
 
   const char *JTEntryDirective = MAI->getData32bitsDirective();
 
@@ -980,7 +980,7 @@
   
   // Can't use EmitLabel until instprinter happens, label comes out in the wrong
   // order.
-  O << *JTISymbol << ":\n";
+  O << "\n" << *JTISymbol << ":\n";
 
   const MachineJumpTableInfo *MJTI = MF->getJumpTableInfo();
   const std::vector<MachineJumpTableEntry> &JT = MJTI->getJumpTables();

Modified: llvm/trunk/lib/Target/ARM/ARMInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrInfo.td?rev=109948&r1=109947&r2=109948&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrInfo.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrInfo.td Sat Jul 31 01:28:10 2010
@@ -1111,7 +1111,7 @@
 
   let isNotDuplicable = 1, isIndirectBranch = 1 in {
   def BR_JTr : JTI<(outs), (ins GPR:$target, jtblock_operand:$jt, i32imm:$id),
-                    IIC_Br, "mov\tpc, $target \n$jt",
+                    IIC_Br, "mov\tpc, $target$jt",
                     [(ARMbrjt GPR:$target, tjumptable:$jt, imm:$id)]> {
     let Inst{11-4}  = 0b00000000;
     let Inst{15-12} = 0b1111;
@@ -1121,7 +1121,7 @@
   }
   def BR_JTm : JTI<(outs),
                    (ins addrmode2:$target, jtblock_operand:$jt, i32imm:$id),
-                   IIC_Br, "ldr\tpc, $target \n$jt",
+                   IIC_Br, "ldr\tpc, $target$jt",
                    [(ARMbrjt (i32 (load addrmode2:$target)), tjumptable:$jt,
                      imm:$id)]> {
     let Inst{15-12} = 0b1111;
@@ -1133,7 +1133,7 @@
   }
   def BR_JTadd : JTI<(outs),
                    (ins GPR:$target, GPR:$idx, jtblock_operand:$jt, i32imm:$id),
-                    IIC_Br, "add\tpc, $target, $idx \n$jt",
+                    IIC_Br, "add\tpc, $target, $idx$jt",
                     [(ARMbrjt (add GPR:$target, GPR:$idx), tjumptable:$jt,
                       imm:$id)]> {
     let Inst{15-12} = 0b1111;

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb.td?rev=109948&r1=109947&r2=109948&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb.td Sat Jul 31 01:28:10 2010
@@ -378,7 +378,7 @@
 
   def tBR_JTr : T1JTI<(outs),
                       (ins tGPR:$target, jtblock_operand:$jt, i32imm:$id),
-                      IIC_Br, "mov\tpc, $target\n\t.align\t2\n$jt",
+                      IIC_Br, "mov\tpc, $target\n\t.align\t2$jt",
                       [(ARMbrjt tGPR:$target, tjumptable:$jt, imm:$id)]>,
                 Encoding16 {
     let Inst{15-7} = 0b010001101;

Modified: llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td?rev=109948&r1=109947&r2=109948&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td Sat Jul 31 01:28:10 2010
@@ -2504,7 +2504,7 @@
 def t2BR_JT :
     T2JTI<(outs),
           (ins GPR:$target, GPR:$index, jt2block_operand:$jt, i32imm:$id),
-           IIC_Br, "mov\tpc, $target\n$jt",
+           IIC_Br, "mov\tpc, $target$jt",
           [(ARMbr2jt GPR:$target, GPR:$index, tjumptable:$jt, imm:$id)]> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0100100;
@@ -2518,7 +2518,7 @@
 def t2TBB :
     T2JTI<(outs),
         (ins tb_addrmode:$index, jt2block_operand:$jt, i32imm:$id),
-         IIC_Br, "tbb\t$index\n$jt", []> {
+         IIC_Br, "tbb\t$index$jt", []> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0001101;
   let Inst{19-16} = 0b1111; // Rn = pc (table follows this instruction)
@@ -2529,7 +2529,7 @@
 def t2TBH :
     T2JTI<(outs),
         (ins tb_addrmode:$index, jt2block_operand:$jt, i32imm:$id),
-         IIC_Br, "tbh\t$index\n$jt", []> {
+         IIC_Br, "tbh\t$index$jt", []> {
   let Inst{31-27} = 0b11101;
   let Inst{26-20} = 0b0001101;
   let Inst{19-16} = 0b1111; // Rn = pc (table follows this instruction)




From bigcheesegs at gmail.com  Sat Jul 31 02:21:44 2010
From: bigcheesegs at gmail.com (Michael J. Spencer)
Date: Sat, 31 Jul 2010 07:21:44 -0000
Subject: [llvm-commits] [llvm] r109949 -
	/llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
Message-ID: <20100731072144.A9A732A6C12C@llvm.org>

Author: mspencer
Date: Sat Jul 31 02:21:44 2010
New Revision: 109949

URL: http://llvm.org/viewvc/llvm-project?rev=109949&view=rev
Log:
MC: Remove HasAbsolutizedSet from WindowsX86AsmBackend.

Modified:
    llvm/trunk/lib/Target/X86/X86AsmBackend.cpp

Modified: llvm/trunk/lib/Target/X86/X86AsmBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86AsmBackend.cpp?rev=109949&r1=109948&r2=109949&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86AsmBackend.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86AsmBackend.cpp Sat Jul 31 02:21:44 2010
@@ -217,7 +217,6 @@
 public:
   WindowsX86AsmBackend(const Target &T)
     : X86AsmBackend(T) {
-    HasAbsolutizedSet = true;
     HasScatteredSymbols = true;
   }
 




From bigcheesegs at gmail.com  Sat Jul 31 02:25:44 2010
From: bigcheesegs at gmail.com (Michael Spencer)
Date: Sat, 31 Jul 2010 03:25:44 -0400
Subject: [llvm-commits] [llc:patch] Add -mc-relax-all to llc.
Message-ID: <AANLkTi=8Jqg4LYNwtoyBcX=0mD4OSgk0OndAOn9Rg8Sj@mail.gmail.com>

This is a very simple patch that adds -mc-relax-all to llc. If used
without filetype=obj it warns that the argument is ignored.

- Michael Spencer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: llc-mc-relax-all.patch
Type: application/octet-stream
Size: 1086 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100731/e9702aae/attachment.obj 

From bigcheesegs at gmail.com  Sat Jul 31 03:02:46 2010
From: bigcheesegs at gmail.com (Michael Spencer)
Date: Sat, 31 Jul 2010 04:02:46 -0400
Subject: [llvm-commits] COFF patch
In-Reply-To: <B7F7863E-3D87-41DB-AFC0-B13206374BAE@apple.com>
References: <B7F7863E-3D87-41DB-AFC0-B13206374BAE@apple.com>
Message-ID: <AANLkTikQH7ngpeey8Qnk0YVRj+6NQ7KDJx5C9hLEPZf=@mail.gmail.com>

On Thu, Jul 29, 2010 at 7:56 PM, Cameron Esfahani <dirty at apple.com> wrote:
> My first patch, so please be gentle...
>
> This fixes a problem I was seeing with the COFF back end. ?All of the string data was getting put into one data fragment, and so none of the fixup references were correct.

Can you post a test case so I can take a look? I'm not quite sure what
the problem is. But I know this isn't the best way to fix the problem.

> I also made sure the TimeDateStamp in the COFF header had a correct value.

This is right.

> I also fixed a spot in MCFragment where Offset wasn't being properly initialized.
>
> This patch was generated against revision 109821.

Also, was this change required to get it to work? The SymbolTableIndex
should be set down on line 674.

@@ -578,12 +580,13 @@ void WinCOFFObjectWriter::RecordRelocation(const
MCAssembler &Asm,
   FixedValue = Target.getConstant();

   COFFRelocation Reloc;

   Reloc.Data.VirtualAddress = Layout.getFragmentOffset(Fragment);
   Reloc.Symb = coff_symbol;
+  Reloc.Data.SymbolTableIndex = 0;

   Reloc.Data.VirtualAddress += Fixup.getOffset();

   switch (Fixup.getKind()) {
   case FirstTargetFixupKind: // reloc_pcrel_4byte
     Reloc.Data.Type = COFF::IMAGE_REL_I386_REL32;

- Michael Spencer



From ggreif at gmail.com  Sat Jul 31 03:35:21 2010
From: ggreif at gmail.com (Gabor Greif)
Date: Sat, 31 Jul 2010 08:35:21 -0000
Subject: [llvm-commits] [llvm] r109950 -
	/llvm/trunk/include/llvm/Instructions.h
Message-ID: <20100731083521.ABF8D2A6C12C@llvm.org>

Author: ggreif
Date: Sat Jul 31 03:35:21 2010
New Revision: 109950

URL: http://llvm.org/viewvc/llvm-project?rev=109950&view=rev
Log:
doxygenize argument accessors

Modified:
    llvm/trunk/include/llvm/Instructions.h

Modified: llvm/trunk/include/llvm/Instructions.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Instructions.h?rev=109950&r1=109949&r2=109950&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Instructions.h (original)
+++ llvm/trunk/include/llvm/Instructions.h Sat Jul 31 03:35:21 2010
@@ -964,7 +964,12 @@
 # undef protected
 public:
 
+  /// getNumArgOperands - Return the number of call arguments.
+  ///
   unsigned getNumArgOperands() const { return getNumOperands() - 1; }
+
+  /// getArgOperand/setArgOperand - Return/set the i-th call argument.
+  ///
   Value *getArgOperand(unsigned i) const { return getOperand(i); }
   void setArgOperand(unsigned i, Value *v) { setOperand(i, v); }
 
@@ -2460,7 +2465,12 @@
   /// Provide fast operand accessors
   DECLARE_TRANSPARENT_OPERAND_ACCESSORS(Value);
 
+  /// getNumArgOperands - Return the number of invoke arguments.
+  ///
   unsigned getNumArgOperands() const { return getNumOperands() - 3; }
+
+  /// getArgOperand/setArgOperand - Return/set the i-th invoke argument.
+  ///
   Value *getArgOperand(unsigned i) const { return getOperand(i); }
   void setArgOperand(unsigned i, Value *v) { setOperand(i, v); }
 




From echristo at apple.com  Sat Jul 31 04:49:42 2010
From: echristo at apple.com (Eric Christopher)
Date: Sat, 31 Jul 2010 02:49:42 -0700
Subject: [llvm-commits] [llvm] r109892 - in /llvm/trunk:
	lib/Target/PowerPC/PPCRegisterInfo.cpp test/FrontendC/vla-1.c
In-Reply-To: <20100730210948.71CBF2A6C12C@llvm.org>
References: <20100730210948.71CBF2A6C12C@llvm.org>
Message-ID: <2EB1C48D-A306-4061-8C49-96D59FA869F4@apple.com>


On Jul 30, 2010, at 2:09 PM, Dale Johannesen wrote:

> Author: johannes
> Date: Fri Jul 30 16:09:48 2010
> New Revision: 109892
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=109892&view=rev
> Log:
> PPC doesn't supported VLA with large alignment.  This was
> formerly rejected by the FE, so asserted in the BE; now the FE only
> warns, so we treat it as a legitimate fatal error in PPC BE.
> This means the test for the feature won't pass, so it's xfail'd.
> 

Thanks Dale!

-eric


From rafael.espindola at gmail.com  Sat Jul 31 09:34:49 2010
From: rafael.espindola at gmail.com (Rafael Espindola)
Date: Sat, 31 Jul 2010 14:34:49 -0000
Subject: [llvm-commits] [llvm] r109951 - in /llvm/trunk/tools/bugpoint:
 BugDriver.h ExecutionDriver.cpp Miscompilation.cpp
Message-ID: <20100731143449.53BC42A6C12C@llvm.org>

Author: rafael
Date: Sat Jul 31 09:34:49 2010
New Revision: 109951

URL: http://llvm.org/viewvc/llvm-project?rev=109951&view=rev
Log:
Add const to some methods and change TestMergedProgram to return the merged
module and take a const BugDriver.

Modified:
    llvm/trunk/tools/bugpoint/BugDriver.h
    llvm/trunk/tools/bugpoint/ExecutionDriver.cpp
    llvm/trunk/tools/bugpoint/Miscompilation.cpp

Modified: llvm/trunk/tools/bugpoint/BugDriver.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/BugDriver.h?rev=109951&r1=109950&r2=109951&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/BugDriver.h (original)
+++ llvm/trunk/tools/bugpoint/BugDriver.h Sat Jul 31 09:34:49 2010
@@ -179,7 +179,7 @@
                              std::string Bitcode,
                              const std::string &SharedObjects,
                              AbstractInterpreter *AI,
-                             std::string *Error);
+                             std::string *Error) const;
 
   /// executeProgramSafely - Used to create reference output with the "safe"
   /// backend, if reference output is not provided.  If there is a problem with
@@ -187,7 +187,8 @@
   /// Error.
   ///
   std::string executeProgramSafely(const Module *Program,
-                                   std::string OutputFile, std::string *Error);
+                                   std::string OutputFile,
+                                   std::string *Error) const;
 
   /// createReferenceFile - calls compileProgram and then records the output
   /// into ReferenceOutputFile. Returns true if reference file created, false 
@@ -206,7 +207,7 @@
                    const std::string &BitcodeFile = "",
                    const std::string &SharedObj = "",
                    bool RemoveBitcode = false,
-                   std::string *Error = 0);
+                   std::string *Error = 0) const;
 
   /// EmitProgressBitcode - This function is used to output M to a file named
   /// "bugpoint-ID.bc".

Modified: llvm/trunk/tools/bugpoint/ExecutionDriver.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/ExecutionDriver.cpp?rev=109951&r1=109950&r2=109951&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/ExecutionDriver.cpp (original)
+++ llvm/trunk/tools/bugpoint/ExecutionDriver.cpp Sat Jul 31 09:34:49 2010
@@ -325,7 +325,7 @@
                                       std::string BitcodeFile,
                                       const std::string &SharedObj,
                                       AbstractInterpreter *AI,
-                                      std::string *Error) {
+                                      std::string *Error) const {
   if (AI == 0) AI = Interpreter;
   assert(AI && "Interpreter should have been created already!");
   bool CreatedBitcode = false;
@@ -402,7 +402,7 @@
 ///
 std::string BugDriver::executeProgramSafely(const Module *Program,
                                             std::string OutputFile,
-                                            std::string *Error) {
+                                            std::string *Error) const {
   return executeProgram(Program, OutputFile, "", "", SafeInterpreter, Error);
 }
 
@@ -466,7 +466,7 @@
                             const std::string &BitcodeFile,
                             const std::string &SharedObject,
                             bool RemoveBitcode,
-                            std::string *ErrMsg) {
+                            std::string *ErrMsg) const {
   // Execute the program, generating an output file...
   sys::Path Output(executeProgram(Program, "", BitcodeFile, SharedObject, 0,
                                   ErrMsg));

Modified: llvm/trunk/tools/bugpoint/Miscompilation.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/bugpoint/Miscompilation.cpp?rev=109951&r1=109950&r2=109951&view=diff
==============================================================================
--- llvm/trunk/tools/bugpoint/Miscompilation.cpp (original)
+++ llvm/trunk/tools/bugpoint/Miscompilation.cpp Sat Jul 31 09:34:49 2010
@@ -204,13 +204,15 @@
 }
 
 /// TestMergedProgram - Given two modules, link them together and run the
-/// program, checking to see if the program matches the diff.  If the diff
-/// matches, return false, otherwise return true.  If the DeleteInputs argument
-/// is set to true then this function deletes both input modules before it
-/// returns.
-///
-static bool TestMergedProgram(BugDriver &BD, Module *M1, Module *M2,
-                              bool DeleteInputs, std::string &Error) {
+/// program, checking to see if the program matches the diff. If there is
+/// an error, return NULL. If not, return the merged module. The Broken argument
+/// will be set to true if the output is different. If the DeleteInputs
+/// argument is set to true then this function deletes both input
+/// modules before it returns.
+///
+static Module *TestMergedProgram(const BugDriver &BD, Module *M1, Module *M2,
+                                 bool DeleteInputs, std::string &Error,
+                                 bool &Broken) {
   // Link the two portions of the program back to together.
   std::string ErrorMsg;
   if (!DeleteInputs) {
@@ -224,16 +226,14 @@
   }
   delete M2;   // We are done with this module.
 
-  // Execute the program.  If it does not match the expected output, we must
-  // return true.
-  bool Broken = BD.diffProgram(M1, "", "", false, &Error);
+  // Execute the program.
+  Broken = BD.diffProgram(M1, "", "", false, &Error);
   if (!Error.empty()) {
     // Delete the linked module
     delete M1;
+    return NULL;
   }
-  // Delete the original and set the new program.
-  delete BD.swapProgramIn(M1);
-  return Broken;
+  return M1;
 }
 
 /// TestFuncs - split functions in a Module into two groups: those that are
@@ -329,10 +329,13 @@
     // has broken.  If something broke, then we'll inform the user and stop
     // extraction.
     AbstractInterpreter *AI = BD.switchToSafeInterpreter();
-    bool Failure = TestMergedProgram(BD, ToOptimizeLoopExtracted, ToNotOptimize,
-                                     false, Error);
-    if (!Error.empty())
+    bool Failure;
+    Module *New = TestMergedProgram(BD, ToOptimizeLoopExtracted, ToNotOptimize,
+                                    false, Error, Failure);
+    if (!New)
       return false;
+    // Delete the original and set the new program.
+    delete BD.swapProgramIn(New);
     if (Failure) {
       BD.switchToInterpreter(AI);
 
@@ -695,8 +698,13 @@
   delete Test;
 
   outs() << "  Checking to see if the merged program executes correctly: ";
-  bool Broken = TestMergedProgram(BD, Optimized, Safe, true, Error);
-  if (Error.empty()) outs() << (Broken ? " nope.\n" : " yup.\n");
+  bool Broken;
+  Module *New = TestMergedProgram(BD, Optimized, Safe, true, Error, Broken);
+  if (New) {
+    outs() << (Broken ? " nope.\n" : " yup.\n");
+    // Delete the original and set the new program.
+    delete BD.swapProgramIn(New);
+  }
   return Broken;
 }
 




From clattner at apple.com  Sat Jul 31 14:09:22 2010
From: clattner at apple.com (Chris Lattner)
Date: Sat, 31 Jul 2010 12:09:22 -0700
Subject: [llvm-commits] [llc:patch] Add -mc-relax-all to llc.
In-Reply-To: <AANLkTi=8Jqg4LYNwtoyBcX=0mD4OSgk0OndAOn9Rg8Sj@mail.gmail.com>
References: <AANLkTi=8Jqg4LYNwtoyBcX=0mD4OSgk0OndAOn9Rg8Sj@mail.gmail.com>
Message-ID: <9D627F5D-F10B-420A-AB3D-5E16215335D7@apple.com>

LGTM, please apply!

-Chris

On Jul 31, 2010, at 12:25 AM, Michael Spencer wrote:

> This is a very simple patch that adds -mc-relax-all to llc. If used
> without filetype=obj it warns that the argument is ignored.
> 
> - Michael Spencer
> <llc-mc-relax-all.patch>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits



From daniel at zuster.org  Sat Jul 31 14:51:11 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 19:51:11 -0000
Subject: [llvm-commits] [llvm] r109953 - in /llvm/trunk:
 lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
 test/Transforms/InstCombine/bit-checks.ll
Message-ID: <20100731195111.62FD92A6C12C@llvm.org>

Author: ddunbar
Date: Sat Jul 31 14:51:11 2010
New Revision: 109953

URL: http://llvm.org/viewvc/llvm-project?rev=109953&view=rev
Log:
Speculatively revert r108614, "Another attempt at getting the clang self-host to
like my instcombine patch.", in an attempt to fix Clang i386 bootstrap.
 - Also PR7719.

Removed:
    llvm/trunk/test/Transforms/InstCombine/bit-checks.ll
Modified:
    llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp

Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp?rev=109953&r1=109952&r2=109953&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp Sat Jul 31 14:51:11 2010
@@ -472,22 +472,6 @@
       Value *NewOr = Builder->CreateOr(Val, Val2);
       return Builder->CreateICmp(LHSCC, NewOr, LHSCst);
     }
-    
-    // (icmp ne (A & C1), 0) & (icmp ne (A & C2), 0) -->
-    // (icmp eq (A & (C1|C2)), (C1|C2)) where C1 and C2 are non-zero POT
-    if (LHSCC == ICmpInst::ICMP_NE && LHSCst->isZero()) {
-      Value *Op1 = 0, *Op2 = 0;
-      ConstantInt *CI1 = 0, *CI2 = 0;
-      if (match(LHS->getOperand(0), m_And(m_Value(Op1), m_ConstantInt(CI1))) &&
-          match(RHS->getOperand(0), m_And(m_Value(Op2), m_ConstantInt(CI2)))) {
-        if (Op1 == Op2 && !CI1->isZero() && !CI2->isZero() &&
-            CI1->getValue().isPowerOf2() && CI2->getValue().isPowerOf2()) {
-          Constant *ConstOr = ConstantExpr::getOr(CI1, CI2);
-          Value *NewAnd = Builder->CreateAnd(Op1, ConstOr);
-          return Builder->CreateICmp(ICmpInst::ICMP_EQ, NewAnd, ConstOr);
-        }
-      }
-    }
   }
   
   // From here on, we only handle:
@@ -1174,22 +1158,6 @@
     return Builder->CreateICmp(LHSCC, NewOr, LHSCst);
   }
   
-  // (icmp eq (A & C1), 0) | (icmp eq (A & C2), 0) -->
-  // (icmp ne (A & (C1|C2)), (C1|C2)) where C1 and C2 are non-zero POT
-  if (LHSCC == ICmpInst::ICMP_EQ && LHSCst->isZero()) {
-    Value *Op1 = 0, *Op2 = 0;
-    ConstantInt *CI1 = 0, *CI2 = 0;
-    if (match(LHS->getOperand(0), m_And(m_Value(Op1), m_ConstantInt(CI1))) &&
-        match(RHS->getOperand(0), m_And(m_Value(Op2), m_ConstantInt(CI2)))) {
-      if (Op1 == Op2 && !CI1->isZero() && !CI2->isZero() &&
-          CI1->getValue().isPowerOf2() && CI2->getValue().isPowerOf2()) {
-        Constant *ConstOr = ConstantExpr::getOr(CI1, CI2);
-        Value *NewAnd = Builder->CreateAnd(Op1, ConstOr);
-        return Builder->CreateICmp(ICmpInst::ICMP_NE, NewAnd, ConstOr);
-      }
-    }
-  }
-  
   // From here on, we only handle:
   //    (icmp1 A, C1) | (icmp2 A, C2) --> something simpler.
   if (Val != Val2) return 0;

Removed: llvm/trunk/test/Transforms/InstCombine/bit-checks.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/bit-checks.ll?rev=109952&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/bit-checks.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/bit-checks.ll (removed)
@@ -1,26 +0,0 @@
-; This test makes sure that these instructions are properly eliminated.
-;
-; RUN: opt < %s -instcombine -S | \
-; RUN:    not grep {tobool}
-; END.
-define i32 @main(i32 %argc, i8** %argv) nounwind ssp {
-entry:
-  %and = and i32 %argc, 1                         ; <i32> [#uses=1]
-  %tobool = icmp ne i32 %and, 0                   ; <i1> [#uses=1]
-  %and2 = and i32 %argc, 2                        ; <i32> [#uses=1]
-  %tobool3 = icmp ne i32 %and2, 0                 ; <i1> [#uses=1]
-  %or.cond = and i1 %tobool, %tobool3             ; <i1> [#uses=1]
-  %retval.0 = select i1 %or.cond, i32 2, i32 1    ; <i32> [#uses=1]
-  ret i32 %retval.0
-}
-
-define i32 @main2(i32 %argc, i8** nocapture %argv) nounwind readnone ssp {
-entry:
-  %and = and i32 %argc, 1                         ; <i32> [#uses=1]
-  %tobool = icmp eq i32 %and, 0                   ; <i1> [#uses=1]
-  %and2 = and i32 %argc, 2                        ; <i32> [#uses=1]
-  %tobool3 = icmp eq i32 %and2, 0                 ; <i1> [#uses=1]
-  %or.cond = or i1 %tobool, %tobool3              ; <i1> [#uses=1]
-  %storemerge = select i1 %or.cond, i32 0, i32 1  ; <i32> [#uses=1]
-  ret i32 %storemerge
-}
\ No newline at end of file




From bigcheesegs at gmail.com  Sat Jul 31 14:57:02 2010
From: bigcheesegs at gmail.com (Michael J. Spencer)
Date: Sat, 31 Jul 2010 19:57:02 -0000
Subject: [llvm-commits] [llvm] r109954 - /llvm/trunk/tools/llc/llc.cpp
Message-ID: <20100731195703.088932A6C12C@llvm.org>

Author: mspencer
Date: Sat Jul 31 14:57:02 2010
New Revision: 109954

URL: http://llvm.org/viewvc/llvm-project?rev=109954&view=rev
Log:
llc: Add -mc-relax-all.

Modified:
    llvm/trunk/tools/llc/llc.cpp

Modified: llvm/trunk/tools/llc/llc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llc/llc.cpp?rev=109954&r1=109953&r2=109954&view=diff
==============================================================================
--- llvm/trunk/tools/llc/llc.cpp (original)
+++ llvm/trunk/tools/llc/llc.cpp Sat Jul 31 14:57:02 2010
@@ -76,6 +76,9 @@
   cl::desc("Target specific attributes (-mattr=help for details)"),
   cl::value_desc("a1,+a2,-a3,..."));
 
+static cl::opt<bool>
+RelaxAll("mc-relax-all", cl::desc("Relax all fixups"));
+
 cl::opt<TargetMachine::CodeGenFileType>
 FileType("filetype", cl::init(TargetMachine::CGFT_AssemblyFile),
   cl::desc("Choose a file type (not all types are supported by all targets):"),
@@ -329,6 +332,14 @@
   // Override default to generate verbose assembly.
   Target.setAsmVerbosityDefault(true);
 
+  if (RelaxAll) {
+    if (FileType != TargetMachine::CGFT_ObjectFile)
+      errs() << argv[0]
+             << ": warning: ignoring -mc-relax-all because filetype != obj";
+    else
+      Target.setMCRelaxAll(true);
+  }
+
   // Ask the target to add backend passes as necessary.
   if (Target.addPassesToEmitFile(PM, *Out, FileType, OLvl,
                                  DisableVerify)) {




From daniel at zuster.org  Sat Jul 31 15:27:27 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 13:27:27 -0700
Subject: [llvm-commits] [llvm] r108614 - in /llvm/trunk:
	lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
	test/Transforms/InstCombine/bit-checks.ll
In-Reply-To: <20100717065635.EBD872A6C12C@llvm.org>
References: <20100717065635.EBD872A6C12C@llvm.org>
Message-ID: <AANLkTimuta+Ceej9=43BVy-VgOmNTNPNPWiTtSB-xN_C@mail.gmail.com>

Hi Owen,

Clang self host still didn't like this -- I ended up reverting it in 108614. See
  http://llvm.org/bugs/show_bug.cgi?id=7719
for more information.

 - Daniel

On Fri, Jul 16, 2010 at 11:56 PM, Owen Anderson <resistor at mac.com> wrote:
> Author: resistor
> Date: Sat Jul 17 01:56:35 2010
> New Revision: 108614
>
> URL: http://llvm.org/viewvc/llvm-project?rev=108614&view=rev
> Log:
> Another attempt at getting the clang self-host to like my instcombine patch.
>
> Added:
> ? ?llvm/trunk/test/Transforms/InstCombine/bit-checks.ll
> Modified:
> ? ?llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
>
> Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp?rev=108614&r1=108613&r2=108614&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp (original)
> +++ llvm/trunk/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp Sat Jul 17 01:56:35 2010
> @@ -472,6 +472,22 @@
> ? ? ? Value *NewOr = Builder->CreateOr(Val, Val2);
> ? ? ? return Builder->CreateICmp(LHSCC, NewOr, LHSCst);
> ? ? }
> +
> + ? ?// (icmp ne (A & C1), 0) & (icmp ne (A & C2), 0) -->
> + ? ?// (icmp eq (A & (C1|C2)), (C1|C2)) where C1 and C2 are non-zero POT
> + ? ?if (LHSCC == ICmpInst::ICMP_NE && LHSCst->isZero()) {
> + ? ? ?Value *Op1 = 0, *Op2 = 0;
> + ? ? ?ConstantInt *CI1 = 0, *CI2 = 0;
> + ? ? ?if (match(LHS->getOperand(0), m_And(m_Value(Op1), m_ConstantInt(CI1))) &&
> + ? ? ? ? ?match(RHS->getOperand(0), m_And(m_Value(Op2), m_ConstantInt(CI2)))) {
> + ? ? ? ?if (Op1 == Op2 && !CI1->isZero() && !CI2->isZero() &&
> + ? ? ? ? ? ?CI1->getValue().isPowerOf2() && CI2->getValue().isPowerOf2()) {
> + ? ? ? ? ?Constant *ConstOr = ConstantExpr::getOr(CI1, CI2);
> + ? ? ? ? ?Value *NewAnd = Builder->CreateAnd(Op1, ConstOr);
> + ? ? ? ? ?return Builder->CreateICmp(ICmpInst::ICMP_EQ, NewAnd, ConstOr);
> + ? ? ? ?}
> + ? ? ?}
> + ? ?}
> ? }
>
> ? // From here on, we only handle:
> @@ -1158,6 +1174,22 @@
> ? ? return Builder->CreateICmp(LHSCC, NewOr, LHSCst);
> ? }
>
> + ?// (icmp eq (A & C1), 0) | (icmp eq (A & C2), 0) -->
> + ?// (icmp ne (A & (C1|C2)), (C1|C2)) where C1 and C2 are non-zero POT
> + ?if (LHSCC == ICmpInst::ICMP_EQ && LHSCst->isZero()) {
> + ? ?Value *Op1 = 0, *Op2 = 0;
> + ? ?ConstantInt *CI1 = 0, *CI2 = 0;
> + ? ?if (match(LHS->getOperand(0), m_And(m_Value(Op1), m_ConstantInt(CI1))) &&
> + ? ? ? ?match(RHS->getOperand(0), m_And(m_Value(Op2), m_ConstantInt(CI2)))) {
> + ? ? ?if (Op1 == Op2 && !CI1->isZero() && !CI2->isZero() &&
> + ? ? ? ? ?CI1->getValue().isPowerOf2() && CI2->getValue().isPowerOf2()) {
> + ? ? ? ?Constant *ConstOr = ConstantExpr::getOr(CI1, CI2);
> + ? ? ? ?Value *NewAnd = Builder->CreateAnd(Op1, ConstOr);
> + ? ? ? ?return Builder->CreateICmp(ICmpInst::ICMP_NE, NewAnd, ConstOr);
> + ? ? ?}
> + ? ?}
> + ?}
> +
> ? // From here on, we only handle:
> ? // ? ?(icmp1 A, C1) | (icmp2 A, C2) --> something simpler.
> ? if (Val != Val2) return 0;
>
> Added: llvm/trunk/test/Transforms/InstCombine/bit-checks.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/bit-checks.ll?rev=108614&view=auto
> ==============================================================================
> --- llvm/trunk/test/Transforms/InstCombine/bit-checks.ll (added)
> +++ llvm/trunk/test/Transforms/InstCombine/bit-checks.ll Sat Jul 17 01:56:35 2010
> @@ -0,0 +1,26 @@
> +; This test makes sure that these instructions are properly eliminated.
> +;
> +; RUN: opt < %s -instcombine -S | \
> +; RUN: ? ?not grep {tobool}
> +; END.
> +define i32 @main(i32 %argc, i8** %argv) nounwind ssp {
> +entry:
> + ?%and = and i32 %argc, 1 ? ? ? ? ? ? ? ? ? ? ? ? ; <i32> [#uses=1]
> + ?%tobool = icmp ne i32 %and, 0 ? ? ? ? ? ? ? ? ? ; <i1> [#uses=1]
> + ?%and2 = and i32 %argc, 2 ? ? ? ? ? ? ? ? ? ? ? ?; <i32> [#uses=1]
> + ?%tobool3 = icmp ne i32 %and2, 0 ? ? ? ? ? ? ? ? ; <i1> [#uses=1]
> + ?%or.cond = and i1 %tobool, %tobool3 ? ? ? ? ? ? ; <i1> [#uses=1]
> + ?%retval.0 = select i1 %or.cond, i32 2, i32 1 ? ?; <i32> [#uses=1]
> + ?ret i32 %retval.0
> +}
> +
> +define i32 @main2(i32 %argc, i8** nocapture %argv) nounwind readnone ssp {
> +entry:
> + ?%and = and i32 %argc, 1 ? ? ? ? ? ? ? ? ? ? ? ? ; <i32> [#uses=1]
> + ?%tobool = icmp eq i32 %and, 0 ? ? ? ? ? ? ? ? ? ; <i1> [#uses=1]
> + ?%and2 = and i32 %argc, 2 ? ? ? ? ? ? ? ? ? ? ? ?; <i32> [#uses=1]
> + ?%tobool3 = icmp eq i32 %and2, 0 ? ? ? ? ? ? ? ? ; <i1> [#uses=1]
> + ?%or.cond = or i1 %tobool, %tobool3 ? ? ? ? ? ? ?; <i1> [#uses=1]
> + ?%storemerge = select i1 %or.cond, i32 0, i32 1 ?; <i32> [#uses=1]
> + ?ret i32 %storemerge
> +}
> \ No newline at end of file
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From echristo at apple.com  Sat Jul 31 15:56:08 2010
From: echristo at apple.com (Eric Christopher)
Date: Sat, 31 Jul 2010 20:56:08 -0000
Subject: [llvm-commits] [test-suite] r109955 -
 /test-suite/trunk/External/skidmarks10/skidmarks.reference_output
Message-ID: <20100731205608.A52E62A6C12D@llvm.org>

Author: echristo
Date: Sat Jul 31 15:56:08 2010
New Revision: 109955

URL: http://llvm.org/viewvc/llvm-project?rev=109955&view=rev
Log:
Add "reference output".

Added:
    test-suite/trunk/External/skidmarks10/skidmarks.reference_output

Added: test-suite/trunk/External/skidmarks10/skidmarks.reference_output
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/External/skidmarks10/skidmarks.reference_output?rev=109955&view=auto
==============================================================================
--- test-suite/trunk/External/skidmarks10/skidmarks.reference_output (added)
+++ test-suite/trunk/External/skidmarks10/skidmarks.reference_output Sat Jul 31 15:56:08 2010
@@ -0,0 +1 @@
+exit 0




From daniel at zuster.org  Sat Jul 31 16:08:54 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 21:08:54 -0000
Subject: [llvm-commits] [llvm] r109956 - in /llvm/trunk/lib:
 MC/WinCOFFObjectWriter.cpp Target/ARM/ARMInstrNEON.td
Message-ID: <20100731210854.6797D2A6C12C@llvm.org>

Author: ddunbar
Date: Sat Jul 31 16:08:54 2010
New Revision: 109956

URL: http://llvm.org/viewvc/llvm-project?rev=109956&view=rev
Log:
Silence some -Asserts uninitialized variable warnings.

Modified:
    llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp
    llvm/trunk/lib/Target/ARM/ARMInstrNEON.td

Modified: llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp?rev=109956&r1=109955&r2=109956&view=diff
==============================================================================
--- llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp (original)
+++ llvm/trunk/lib/MC/WinCOFFObjectWriter.cpp Sat Jul 31 16:08:54 2010
@@ -579,6 +579,7 @@
 
   COFFRelocation Reloc;
 
+  Reloc.Data.SymbolTableIndex = 0;
   Reloc.Data.VirtualAddress = Layout.getFragmentOffset(Fragment);
   Reloc.Symb = coff_symbol;
 

Modified: llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrNEON.td?rev=109956&r1=109955&r2=109956&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrNEON.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrNEON.td Sat Jul 31 16:08:54 2010
@@ -100,14 +100,14 @@
 
 def NEONimmAllZerosV: PatLeaf<(NEONvmovImm (i32 timm)), [{
   ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
-  unsigned EltBits;
+  unsigned EltBits = 0;
   uint64_t EltVal = ARM_AM::decodeNEONModImm(ConstVal->getZExtValue(), EltBits);
   return (EltBits == 32 && EltVal == 0);
 }]>;
 
 def NEONimmAllOnesV: PatLeaf<(NEONvmovImm (i32 timm)), [{
   ConstantSDNode *ConstVal = cast<ConstantSDNode>(N->getOperand(0));
-  unsigned EltBits;
+  unsigned EltBits = 0;
   uint64_t EltVal = ARM_AM::decodeNEONModImm(ConstVal->getZExtValue(), EltBits);
   return (EltBits == 8 && EltVal == 0xff);
 }]>;




From daniel at zuster.org  Sat Jul 31 16:32:57 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sat, 31 Jul 2010 21:32:57 -0000
Subject: [llvm-commits] [llvm] r109957 - in /llvm/trunk/tools: edis/Makefile
 llvm-shlib/Makefile lto/Makefile
Message-ID: <20100731213257.20C802A6C12C@llvm.org>

Author: ddunbar
Date: Sat Jul 31 16:32:56 2010
New Revision: 109957

URL: http://llvm.org/viewvc/llvm-project?rev=109957&view=rev
Log:
build: Don't pass -avoid-version or -no-undefined on Darwin, they don't do
anything.

Modified:
    llvm/trunk/tools/edis/Makefile
    llvm/trunk/tools/llvm-shlib/Makefile
    llvm/trunk/tools/lto/Makefile

Modified: llvm/trunk/tools/edis/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/edis/Makefile?rev=109957&r1=109956&r2=109957&view=diff
==============================================================================
--- llvm/trunk/tools/edis/Makefile (original)
+++ llvm/trunk/tools/edis/Makefile Sat Jul 31 16:32:56 2010
@@ -36,7 +36,6 @@
 ifeq ($(HOST_OS),Darwin)
     # extra options to override libtool defaults 
     LLVMLibsOptions    := $(LLVMLibsOptions)  \
-                         -avoid-version \
                          -Wl,-dead_strip
 
     ifdef EDIS_VERSION
@@ -48,7 +47,7 @@
     DARWIN_VERS := $(shell echo $(TARGET_TRIPLE) | sed 's/.*darwin\([0-9]*\).*/\1/')
     ifneq ($(DARWIN_VERS),8)
        LLVMLibsOptions    := $(LLVMLibsOptions)  \
-                            -no-undefined -Wl,-install_name \
+                            -Wl,-install_name \
                             -Wl,"@rpath/lib$(LIBRARYNAME)$(SHLIBEXT)"
     endif
 endif

Modified: llvm/trunk/tools/llvm-shlib/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/llvm-shlib/Makefile?rev=109957&r1=109956&r2=109957&view=diff
==============================================================================
--- llvm/trunk/tools/llvm-shlib/Makefile (original)
+++ llvm/trunk/tools/llvm-shlib/Makefile Sat Jul 31 16:32:56 2010
@@ -38,7 +38,6 @@
     LLVMLibsOptions    := $(LLVMLibsOptions) -all_load
     # extra options to override libtool defaults 
     LLVMLibsOptions    := $(LLVMLibsOptions)  \
-                         -avoid-version \
                          -Wl,-dead_strip \
                          -Wl,-seg1addr -Wl,0xE0000000 
 

Modified: llvm/trunk/tools/lto/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/lto/Makefile?rev=109957&r1=109956&r2=109957&view=diff
==============================================================================
--- llvm/trunk/tools/lto/Makefile (original)
+++ llvm/trunk/tools/lto/Makefile Sat Jul 31 16:32:56 2010
@@ -41,7 +41,6 @@
     endif
     # extra options to override libtool defaults 
     LLVMLibsOptions    := $(LLVMLibsOptions)  \
-                         -avoid-version \
                          -Wl,-dead_strip \
                          -Wl,-seg1addr -Wl,0xE0000000 
 
@@ -49,7 +48,7 @@
     DARWIN_VERS := $(shell echo $(TARGET_TRIPLE) | sed 's/.*darwin\([0-9]*\).*/\1/')
     ifneq ($(DARWIN_VERS),8)
        LLVMLibsOptions    := $(LLVMLibsOptions)  \
-                            -no-undefined -Wl,-install_name \
+                            -Wl,-install_name \
                             -Wl,"@executable_path/../lib/lib$(LIBRARYNAME)$(SHLIBEXT)"
     endif
 endif




From bob.wilson at apple.com  Sat Jul 31 17:36:03 2010
From: bob.wilson at apple.com (Bob Wilson)
Date: Sat, 31 Jul 2010 22:36:03 -0000
Subject: [llvm-commits] [llvm] r109959 - in /llvm/trunk/test/CodeGen/X86:
 avx-intrinsics-x86.ll avx-intrinsics-x86_64.ll
Message-ID: <20100731223603.E5D452A6C12E@llvm.org>

Author: bwilson
Date: Sat Jul 31 17:36:03 2010
New Revision: 109959

URL: http://llvm.org/viewvc/llvm-project?rev=109959&view=rev
Log:
Revert new AVX intrinsic tests.  They are breaking buildbots and Bruno is
away from a computer now.
--- Reverse-merging r109881 into '.':
D    test/CodeGen/X86/avx-intrinsics-x86.ll
D    test/CodeGen/X86/avx-intrinsics-x86_64.ll

Removed:
    llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
    llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll

Removed: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll?rev=109958&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll (removed)
@@ -1,1956 +0,0 @@
-; RUN: llc < %s -march=x86 -mattr=avx,aes | FileCheck %s
-
-define <2 x i64> @test_x86_aesni_aesdec(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vaesdec
-  %res = call <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aesdec(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <2 x i64> @test_x86_aesni_aesdeclast(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vaesdeclast
-  %res = call <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aesdeclast(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <2 x i64> @test_x86_aesni_aesenc(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vaesenc
-  %res = call <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aesenc(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <2 x i64> @test_x86_aesni_aesenclast(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vaesenclast
-  %res = call <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aesenclast(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <2 x i64> @test_x86_aesni_aesimc(<2 x i64> %a0) {
-  ; CHECK: vaesimc
-  %res = call <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aesimc(<2 x i64>) nounwind readnone
-
-
-define <2 x i64> @test_x86_aesni_aeskeygenassist(<2 x i64> %a0) {
-  ; CHECK: vaeskeygenassist
-  %res = call <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64> %a0, i8 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.aesni.aeskeygenassist(<2 x i64>, i8) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_add_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vaddsd
-  %res = call <2 x double> @llvm.x86.sse2.add.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.add.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cmp_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcmpordpd
-  %res = call <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cmp.pd(<2 x double>, <2 x double>, i8) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cmp_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcmpordsd
-  %res = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> %a0, <2 x double> %a1, i8 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double>, <2 x double>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse2_comieq_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: sete
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comieq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comieq.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_comige_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: setae
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comige.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comige.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_comigt_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: seta
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comigt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comigt.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_comile_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: setbe
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comile.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comile.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_comilt_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: setb
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comilt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comilt.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_comineq_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vcomisd
-  ; CHECK: setne
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.comineq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.comineq.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cvtdq2pd(<4 x i32> %a0) {
-  ; CHECK: vcvtdq2pd
-  %res = call <2 x double> @llvm.x86.sse2.cvtdq2pd(<4 x i32> %a0) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cvtdq2pd(<4 x i32>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse2_cvtdq2ps(<4 x i32> %a0) {
-  ; CHECK: vcvtdq2ps
-  %res = call <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse2.cvtdq2ps(<4 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_cvtpd2dq(<2 x double> %a0) {
-  ; CHECK: vcvtpd2dq
-  %res = call <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.cvtpd2dq(<2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse2_cvtpd2ps(<2 x double> %a0) {
-  ; CHECK: vcvtpd2ps
-  %res = call <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse2.cvtpd2ps(<2 x double>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_cvtps2dq(<4 x float> %a0) {
-  ; CHECK: vcvtps2dq
-  %res = call <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.cvtps2dq(<4 x float>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cvtps2pd(<4 x float> %a0) {
-  ; CHECK: vcvtps2pd
-  %res = call <2 x double> @llvm.x86.sse2.cvtps2pd(<4 x float> %a0) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cvtps2pd(<4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse2_cvtsd2si(<2 x double> %a0) {
-  ; CHECK: vcvtsd2si
-  %res = call i32 @llvm.x86.sse2.cvtsd2si(<2 x double> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.cvtsd2si(<2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse2_cvtsd2ss(<4 x float> %a0, <2 x double> %a1) {
-  ; CHECK: vcvtsd2ss
-  %res = call <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float> %a0, <2 x double> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse2.cvtsd2ss(<4 x float>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cvtsi2sd(<2 x double> %a0) {
-  ; CHECK: movl
-  ; CHECK: vcvtsi2sd
-  %res = call <2 x double> @llvm.x86.sse2.cvtsi2sd(<2 x double> %a0, i32 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cvtsi2sd(<2 x double>, i32) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cvtss2sd(<2 x double> %a0, <4 x float> %a1) {
-  ; CHECK: vcvtss2sd
-  %res = call <2 x double> @llvm.x86.sse2.cvtss2sd(<2 x double> %a0, <4 x float> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cvtss2sd(<2 x double>, <4 x float>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_cvttpd2dq(<2 x double> %a0) {
-  ; CHECK: vcvttpd2dq
-  %res = call <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.cvttpd2dq(<2 x double>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_cvttps2dq(<4 x float> %a0) {
-  ; CHECK: vcvttps2dq
-  %res = call <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.cvttps2dq(<4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse2_cvttsd2si(<2 x double> %a0) {
-  ; CHECK: vcvttss2si
-  %res = call i32 @llvm.x86.sse2.cvttsd2si(<2 x double> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.cvttsd2si(<2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_div_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vdivsd
-  %res = call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.div.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_loadu_dq(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vmovdqu
-  %res = call <16 x i8> @llvm.x86.sse2.loadu.dq(i8* %a0) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.loadu.dq(i8*) nounwind readonly
-
-
-define <2 x double> @test_x86_sse2_loadu_pd(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vmovupd
-  %res = call <2 x double> @llvm.x86.sse2.loadu.pd(i8* %a0) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.loadu.pd(i8*) nounwind readonly
-
-
-define void @test_x86_sse2_maskmov_dqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2) {
-  ; CHECK: pushl
-  ; CHECK: movl
-  ; CHECK: vmaskmovdqu
-  ; CHECK: popl
-  call void @llvm.x86.sse2.maskmov.dqu(<16 x i8> %a0, <16 x i8> %a1, i8* %a2)
-  ret void
-}
-declare void @llvm.x86.sse2.maskmov.dqu(<16 x i8>, <16 x i8>, i8*) nounwind
-
-
-define <2 x double> @test_x86_sse2_max_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vmaxpd
-  %res = call <2 x double> @llvm.x86.sse2.max.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.max.pd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_max_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vmaxsd
-  %res = call <2 x double> @llvm.x86.sse2.max.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.max.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_min_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vminpd
-  %res = call <2 x double> @llvm.x86.sse2.min.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.min.pd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_min_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vminsd
-  %res = call <2 x double> @llvm.x86.sse2.min.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.min.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_movmsk_pd(<2 x double> %a0) {
-  ; CHECK: vmovmskpd
-  %res = call i32 @llvm.x86.sse2.movmsk.pd(<2 x double> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.movmsk.pd(<2 x double>) nounwind readnone
-
-
-define void @test_x86_sse2_movnt_dq(i8* %a0, <2 x i64> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovntdq
-  call void @llvm.x86.sse2.movnt.dq(i8* %a0, <2 x i64> %a1)
-  ret void
-}
-declare void @llvm.x86.sse2.movnt.dq(i8*, <2 x i64>) nounwind
-
-
-define void @test_x86_sse2_movnt_pd(i8* %a0, <2 x double> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovntpd
-  call void @llvm.x86.sse2.movnt.pd(i8* %a0, <2 x double> %a1)
-  ret void
-}
-declare void @llvm.x86.sse2.movnt.pd(i8*, <2 x double>) nounwind
-
-
-define <2 x double> @test_x86_sse2_mul_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vmulsd
-  %res = call <2 x double> @llvm.x86.sse2.mul.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.mul.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_packssdw_128(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpackssdw
-  %res = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a0, <4 x i32> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_packsswb_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpacksswb
-  %res = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a0, <8 x i16> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_packuswb_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpackuswb
-  %res = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a0, <8 x i16> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_padds_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpaddsb
-  %res = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_padds_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpaddsw
-  %res = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_paddus_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpaddusb
-  %res = call <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.paddus.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_paddus_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpaddusw
-  %res = call <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.paddus.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_pavg_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpavgb
-  %res = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pavg_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpavgw
-  %res = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_pcmpeq_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpeqb
-  %res = call <16 x i8> @llvm.x86.sse2.pcmpeq.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.pcmpeq.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_pcmpeq_d(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpcmpeqd
-  %res = call <4 x i32> @llvm.x86.sse2.pcmpeq.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.pcmpeq.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pcmpeq_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpcmpeqw
-  %res = call <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pcmpeq.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_pcmpgt_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpgtb
-  %res = call <16 x i8> @llvm.x86.sse2.pcmpgt.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.pcmpgt.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_pcmpgt_d(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpcmpgtd
-  %res = call <4 x i32> @llvm.x86.sse2.pcmpgt.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.pcmpgt.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pcmpgt_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpcmpgtw
-  %res = call <8 x i16> @llvm.x86.sse2.pcmpgt.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pcmpgt.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_pmadd_wd(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmaddwd
-  %res = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %a0, <8 x i16> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pmaxs_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmaxsw
-  %res = call <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pmaxs.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_pmaxu_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpmaxub
-  %res = call <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.pmaxu.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pmins_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpminsw
-  %res = call <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pmins.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_pminu_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpminub
-  %res = call <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.pminu.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define i32 @test_x86_sse2_pmovmskb_128(<16 x i8> %a0) {
-  ; CHECK: vpmovmskb
-  %res = call i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.pmovmskb.128(<16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pmulh_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmulhw
-  %res = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pmulhu_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmulhuw
-  %res = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_pmulu_dq(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpmuludq
-  %res = call <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32> %a0, <4 x i32> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.pmulu.dq(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psad_bw(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpsadbw
-  %res = call <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8> %a0, <16 x i8> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psad.bw(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_psll_d(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpslld
-  %res = call <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.psll.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psll_dq(<2 x i64> %a0) {
-  ; CHECK: vpslldq
-  %res = call <2 x i64> @llvm.x86.sse2.psll.dq(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psll.dq(<2 x i64>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psll_dq_bs(<2 x i64> %a0) {
-  ; CHECK: vpslldq
-  %res = call <2 x i64> @llvm.x86.sse2.psll.dq.bs(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psll.dq.bs(<2 x i64>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psll_q(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vpsllq
-  %res = call <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psll.q(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psll_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsllw
-  %res = call <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psll.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_pslli_d(<4 x i32> %a0) {
-  ; CHECK: vpslld
-  %res = call <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.pslli.d(<4 x i32>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_pslli_q(<2 x i64> %a0) {
-  ; CHECK: vpsllq
-  %res = call <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.pslli.q(<2 x i64>, i32) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_pslli_w(<8 x i16> %a0) {
-  ; CHECK: vpsllw
-  %res = call <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.pslli.w(<8 x i16>, i32) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_psra_d(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpsrad
-  %res = call <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.psra.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psra_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsraw
-  %res = call <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psra.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_psrai_d(<4 x i32> %a0) {
-  ; CHECK: vpsrad
-  %res = call <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.psrai.d(<4 x i32>, i32) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psrai_w(<8 x i16> %a0) {
-  ; CHECK: vpsraw
-  %res = call <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psrai.w(<8 x i16>, i32) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_psrl_d(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpsrld
-  %res = call <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.psrl.d(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psrl_dq(<2 x i64> %a0) {
-  ; CHECK: vpsrldq
-  %res = call <2 x i64> @llvm.x86.sse2.psrl.dq(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psrl.dq(<2 x i64>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psrl_dq_bs(<2 x i64> %a0) {
-  ; CHECK: vpsrldq
-  %res = call <2 x i64> @llvm.x86.sse2.psrl.dq.bs(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psrl.dq.bs(<2 x i64>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psrl_q(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vpsrlq
-  %res = call <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psrl.q(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psrl_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsrlw
-  %res = call <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psrl.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse2_psrli_d(<4 x i32> %a0) {
-  ; CHECK: vpsrld
-  %res = call <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32> %a0, i32 7) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse2.psrli.d(<4 x i32>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse2_psrli_q(<2 x i64> %a0) {
-  ; CHECK: vpsrlq
-  %res = call <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64> %a0, i32 7) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse2.psrli.q(<2 x i64>, i32) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psrli_w(<8 x i16> %a0) {
-  ; CHECK: vpsrlw
-  %res = call <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16> %a0, i32 7) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psrli.w(<8 x i16>, i32) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_psubs_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpsubsb
-  %res = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psubs_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsubsw
-  %res = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse2_psubus_b(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpsubusb
-  %res = call <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse2.psubus.b(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse2_psubus_w(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsubusw
-  %res = call <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse2.psubus.w(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_sqrt_pd(<2 x double> %a0) {
-  ; CHECK: vsqrtpd
-  %res = call <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double> %a0) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.sqrt.pd(<2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_sqrt_sd(<2 x double> %a0) {
-  ; CHECK: vsqrtsd
-  %res = call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %a0) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone
-
-
-define void @test_x86_sse2_storel_dq(i8* %a0, <4 x i32> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovq
-  call void @llvm.x86.sse2.storel.dq(i8* %a0, <4 x i32> %a1)
-  ret void
-}
-declare void @llvm.x86.sse2.storel.dq(i8*, <4 x i32>) nounwind
-
-
-define void @test_x86_sse2_storeu_dq(i8* %a0, <16 x i8> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovdqu
-  call void @llvm.x86.sse2.storeu.dq(i8* %a0, <16 x i8> %a1)
-  ret void
-}
-declare void @llvm.x86.sse2.storeu.dq(i8*, <16 x i8>) nounwind
-
-
-define void @test_x86_sse2_storeu_pd(i8* %a0, <2 x double> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovupd
-  call void @llvm.x86.sse2.storeu.pd(i8* %a0, <2 x double> %a1)
-  ret void
-}
-declare void @llvm.x86.sse2.storeu.pd(i8*, <2 x double>) nounwind
-
-
-define <2 x double> @test_x86_sse2_sub_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vsubsd
-  %res = call <2 x double> @llvm.x86.sse2.sub.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.sub.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomieq_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: sete
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomieq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomieq.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomige_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: setae
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomige.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomige.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomigt_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: seta
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomigt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomigt.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomile_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: setbe
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomile.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomile.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomilt_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: setb
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomilt.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomilt.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define i32 @test_x86_sse2_ucomineq_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vucomisd
-  ; CHECK: setne
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse2.ucomineq.sd(<2 x double> %a0, <2 x double> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse2.ucomineq.sd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse3_addsub_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vaddsubpd
-  %res = call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse3_addsub_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vaddsubps
-  %res = call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse3_hadd_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vhaddpd
-  %res = call <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse3.hadd.pd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse3_hadd_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vhaddps
-  %res = call <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse3.hadd.ps(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse3_hsub_pd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vhsubpd
-  %res = call <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse3.hsub.pd(<2 x double>, <2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse3_hsub_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vhsubps
-  %res = call <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse3.hsub.ps(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse3_ldu_dq(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vlddqu
-  %res = call <16 x i8> @llvm.x86.sse3.ldu.dq(i8* %a0) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse3.ldu.dq(i8*) nounwind readonly
-
-
-define <2 x double> @test_x86_sse41_blendpd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vblendpd
-  %res = call <2 x double> @llvm.x86.sse41.blendpd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse41.blendpd(<2 x double>, <2 x double>, i32) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_blendps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vblendps
-  %res = call <4 x float> @llvm.x86.sse41.blendps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.blendps(<4 x float>, <4 x float>, i32) nounwind readnone
-
-
-define <2 x double> @test_x86_sse41_blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) {
-  ; CHECK: vblendvpd
-  %res = call <2 x double> @llvm.x86.sse41.blendvpd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse41.blendvpd(<2 x double>, <2 x double>, <2 x double>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) {
-  ; CHECK: vblendvps
-  %res = call <4 x float> @llvm.x86.sse41.blendvps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.blendvps(<4 x float>, <4 x float>, <4 x float>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse41_dppd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vdppd
-  %res = call <2 x double> @llvm.x86.sse41.dppd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse41.dppd(<2 x double>, <2 x double>, i32) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_dpps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vdpps
-  %res = call <4 x float> @llvm.x86.sse41.dpps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.dpps(<4 x float>, <4 x float>, i32) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_insertps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vinsertps
-  %res = call <4 x float> @llvm.x86.sse41.insertps(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.insertps(<4 x float>, <4 x float>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_movntdqa(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vmovntdqa
-  %res = call <2 x i64> @llvm.x86.sse41.movntdqa(i8* %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.movntdqa(i8*) nounwind readonly
-
-
-define <16 x i8> @test_x86_sse41_mpsadbw(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vmpsadbw
-  %res = call <16 x i8> @llvm.x86.sse41.mpsadbw(<16 x i8> %a0, <16 x i8> %a1, i32 7) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse41.mpsadbw(<16 x i8>, <16 x i8>, i32) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_packusdw(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpackusdw
-  %res = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a0, <4 x i32> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse41_pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2) {
-  ; CHECK: vpblendvb
-  %res = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %a2) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_pblendw(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpblendw
-  %res = call <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16> %a0, <8 x i16> %a1, i32 7) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.pblendw(<8 x i16>, <8 x i16>, i32) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pcmpeqq(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vpcmpeqq
-  %res = call <2 x i64> @llvm.x86.sse41.pcmpeqq(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pcmpeqq(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_phminposuw(<8 x i16> %a0) {
-  ; CHECK: vphminposuw
-  %res = call <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16> %a0) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.phminposuw(<8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse41_pmaxsb(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpmaxsb
-  %res = call <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse41.pmaxsb(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmaxsd(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpmaxsd
-  %res = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmaxud(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpmaxud
-  %res = call <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmaxud(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_pmaxuw(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmaxuw
-  %res = call <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.pmaxuw(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse41_pminsb(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpminsb
-  %res = call <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse41.pminsb(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pminsd(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpminsd
-  %res = call <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pminsd(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pminud(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpminud
-  %res = call <4 x i32> @llvm.x86.sse41.pminud(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pminud(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_pminuw(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpminuw
-  %res = call <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.pminuw(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmovsxbd(<16 x i8> %a0) {
-  ; CHECK: vpmovsxbd
-  %res = call <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmovsxbd(<16 x i8>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovsxbq(<16 x i8> %a0) {
-  ; CHECK: vpmovsxbq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovsxbq(<16 x i8> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovsxbq(<16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_pmovsxbw(<16 x i8> %a0) {
-  ; CHECK: vpmovsxbw
-  %res = call <8 x i16> @llvm.x86.sse41.pmovsxbw(<16 x i8> %a0) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.pmovsxbw(<16 x i8>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovsxdq(<4 x i32> %a0) {
-  ; CHECK: vpmovsxdq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovsxdq(<4 x i32> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovsxdq(<4 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmovsxwd(<8 x i16> %a0) {
-  ; CHECK: vpmovsxwd
-  %res = call <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmovsxwd(<8 x i16>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovsxwq(<8 x i16> %a0) {
-  ; CHECK: vpmovsxwq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovsxwq(<8 x i16> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovsxwq(<8 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmovzxbd(<16 x i8> %a0) {
-  ; CHECK: vpmovzxbd
-  %res = call <4 x i32> @llvm.x86.sse41.pmovzxbd(<16 x i8> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmovzxbd(<16 x i8>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovzxbq(<16 x i8> %a0) {
-  ; CHECK: vpmovzxbq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovzxbq(<16 x i8>) nounwind readnone
-
-
-define <8 x i16> @test_x86_sse41_pmovzxbw(<16 x i8> %a0) {
-  ; CHECK: vpmovzxbw
-  %res = call <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8> %a0) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.sse41.pmovzxbw(<16 x i8>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovzxdq(<4 x i32> %a0) {
-  ; CHECK: vpmovzxdq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovzxdq(<4 x i32> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovzxdq(<4 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_sse41_pmovzxwd(<8 x i16> %a0) {
-  ; CHECK: vpmovzxwd
-  %res = call <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmovzxwq(<8 x i16> %a0) {
-  ; CHECK: vpmovzxwq
-  %res = call <2 x i64> @llvm.x86.sse41.pmovzxwq(<8 x i16> %a0) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmovzxwq(<8 x i16>) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse41_pmuldq(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpmuldq
-  %res = call <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32> %a0, <4 x i32> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse41.pmuldq(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define i32 @test_x86_sse41_ptestc(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vptest 
-  ; CHECK: setb
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse41.ptestc(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse41.ptestc(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse41_ptestnzc(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vptest 
-  ; CHECK: seta
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse41.ptestnzc(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse41.ptestnzc(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse41_ptestz(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vptest 
-  ; CHECK: sete
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse41.ptestz(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse41.ptestz(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse41_round_pd(<2 x double> %a0) {
-  ; CHECK: vroundpd
-  %res = call <2 x double> @llvm.x86.sse41.round.pd(<2 x double> %a0, i32 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse41.round.pd(<2 x double>, i32) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_round_ps(<4 x float> %a0) {
-  ; CHECK: vroundps
-  %res = call <4 x float> @llvm.x86.sse41.round.ps(<4 x float> %a0, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.round.ps(<4 x float>, i32) nounwind readnone
-
-
-define <2 x double> @test_x86_sse41_round_sd(<2 x double> %a0, <2 x double> %a1) {
-  ; CHECK: vroundsd
-  %res = call <2 x double> @llvm.x86.sse41.round.sd(<2 x double> %a0, <2 x double> %a1, i32 7) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse41.round.sd(<2 x double>, <2 x double>, i32) nounwind readnone
-
-
-define <4 x float> @test_x86_sse41_round_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vroundss
-  %res = call <4 x float> @llvm.x86.sse41.round.ss(<4 x float> %a0, <4 x float> %a1, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse41.round.ss(<4 x float>, <4 x float>, i32) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestri128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestri128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestri128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestria128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestria128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestria128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestric128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestric128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestric128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestrio128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestrio128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestrio128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestris128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestris128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestris128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpestriz128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpestriz128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpestriz128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse42_pcmpestrm128(<16 x i8> %a0, <16 x i8> %a2) {
-  ; CHECK: movl
-  ; CHECK: movl
-  ; CHECK: vpcmpestrm
-  %res = call <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8> %a0, i32 7, <16 x i8> %a2, i32 7, i8 7) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse42.pcmpestrm128(<16 x i8>, i32, <16 x i8>, i32, i8) nounwind readnone
-
-
-define <2 x i64> @test_x86_sse42_pcmpgtq(<2 x i64> %a0, <2 x i64> %a1) {
-  ; CHECK: vpcmpgtq
-  %res = call <2 x i64> @llvm.x86.sse42.pcmpgtq(<2 x i64> %a0, <2 x i64> %a1) ; <<2 x i64>> [#uses=1]
-  ret <2 x i64> %res
-}
-declare <2 x i64> @llvm.x86.sse42.pcmpgtq(<2 x i64>, <2 x i64>) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistri128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistri128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistri128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistria128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistria128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistria128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistric128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistric128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistric128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistrio128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistrio128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistrio128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistris128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistris128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistris128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse42_pcmpistriz128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistri
-  ; CHECK: movl
-  %res = call i32 @llvm.x86.sse42.pcmpistriz128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse42.pcmpistriz128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define <16 x i8> @test_x86_sse42_pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpcmpistrm
-  %res = call <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8> %a0, <16 x i8> %a1, i8 7) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.sse42.pcmpistrm128(<16 x i8>, <16 x i8>, i8) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_add_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vaddss
-  %res = call <4 x float> @llvm.x86.sse.add.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.add.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_cmp_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcmpordps
-  %res = call <4 x float> @llvm.x86.sse.cmp.ps(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.cmp.ps(<4 x float>, <4 x float>, i8) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_cmp_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcmpordss
-  %res = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> %a0, <4 x float> %a1, i8 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.cmp.ss(<4 x float>, <4 x float>, i8) nounwind readnone
-
-
-define i32 @test_x86_sse_comieq_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: sete
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comieq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comieq.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_comige_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: setae
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comige.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comige.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_comigt_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: seta
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comigt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comigt.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_comile_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: setbe
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comile.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comile.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_comilt_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: setb
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comilt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comilt.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_comineq_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vcomiss
-  ; CHECK: setne
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.comineq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.comineq.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_cvtsi2ss(<4 x float> %a0) {
-  ; CHECK: movl
-  ; CHECK: vcvtsi2ss
-  %res = call <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float> %a0, i32 7) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.cvtsi2ss(<4 x float>, i32) nounwind readnone
-
-
-define i32 @test_x86_sse_cvtss2si(<4 x float> %a0) {
-  ; CHECK: vcvtss2si
-  %res = call i32 @llvm.x86.sse.cvtss2si(<4 x float> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.cvtss2si(<4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_cvttss2si(<4 x float> %a0) {
-  ; CHECK: vcvttss2si
-  %res = call i32 @llvm.x86.sse.cvttss2si(<4 x float> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.cvttss2si(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_div_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vdivss
-  %res = call <4 x float> @llvm.x86.sse.div.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.div.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define void @test_x86_sse_ldmxcsr(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vldmxcsr
-  call void @llvm.x86.sse.ldmxcsr(i8* %a0)
-  ret void
-}
-declare void @llvm.x86.sse.ldmxcsr(i8*) nounwind
-
-
-define <4 x float> @test_x86_sse_loadu_ps(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vmovups
-  %res = call <4 x float> @llvm.x86.sse.loadu.ps(i8* %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.loadu.ps(i8*) nounwind readonly
-
-
-define <4 x float> @test_x86_sse_max_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vmaxps
-  %res = call <4 x float> @llvm.x86.sse.max.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.max.ps(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_max_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vmaxss
-  %res = call <4 x float> @llvm.x86.sse.max.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.max.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_min_ps(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vminps
-  %res = call <4 x float> @llvm.x86.sse.min.ps(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.min.ps(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_min_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vminss
-  %res = call <4 x float> @llvm.x86.sse.min.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.min.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_movmsk_ps(<4 x float> %a0) {
-  ; CHECK: vmovmskps
-  %res = call i32 @llvm.x86.sse.movmsk.ps(<4 x float> %a0) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.movmsk.ps(<4 x float>) nounwind readnone
-
-
-define void @test_x86_sse_movnt_ps(i8* %a0, <4 x float> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovntps
-  call void @llvm.x86.sse.movnt.ps(i8* %a0, <4 x float> %a1)
-  ret void
-}
-declare void @llvm.x86.sse.movnt.ps(i8*, <4 x float>) nounwind
-
-
-define <4 x float> @test_x86_sse_mul_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vmulss
-  %res = call <4 x float> @llvm.x86.sse.mul.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.mul.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_rcp_ps(<4 x float> %a0) {
-  ; CHECK: vrcpps
-  %res = call <4 x float> @llvm.x86.sse.rcp.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.rcp.ps(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_rcp_ss(<4 x float> %a0) {
-  ; CHECK: vrcpss
-  %res = call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.rcp.ss(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_rsqrt_ps(<4 x float> %a0) {
-  ; CHECK: vrsqrtps
-  %res = call <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.rsqrt.ps(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_rsqrt_ss(<4 x float> %a0) {
-  ; CHECK: vrsqrtss
-  %res = call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_sqrt_ps(<4 x float> %a0) {
-  ; CHECK: vsqrtps
-  %res = call <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.sqrt.ps(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_sqrt_ss(<4 x float> %a0) {
-  ; CHECK: vsqrtss
-  %res = call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %a0) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float>) nounwind readnone
-
-
-define void @test_x86_sse_stmxcsr(i8* %a0) {
-  ; CHECK: movl
-  ; CHECK: vstmxcsr
-  call void @llvm.x86.sse.stmxcsr(i8* %a0)
-  ret void
-}
-declare void @llvm.x86.sse.stmxcsr(i8*) nounwind
-
-
-define void @test_x86_sse_storeu_ps(i8* %a0, <4 x float> %a1) {
-  ; CHECK: movl
-  ; CHECK: vmovups
-  call void @llvm.x86.sse.storeu.ps(i8* %a0, <4 x float> %a1)
-  ret void
-}
-declare void @llvm.x86.sse.storeu.ps(i8*, <4 x float>) nounwind
-
-
-define <4 x float> @test_x86_sse_sub_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vsubss
-  %res = call <4 x float> @llvm.x86.sse.sub.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.sub.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomieq_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: sete
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomieq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomieq.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomige_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: setae
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomige.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomige.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomigt_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: seta
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomigt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomigt.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomile_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: setbe
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomile.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomile.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomilt_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: setb
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomilt.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomilt.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define i32 @test_x86_sse_ucomineq_ss(<4 x float> %a0, <4 x float> %a1) {
-  ; CHECK: vucomiss
-  ; CHECK: setne
-  ; CHECK: movzbl
-  %res = call i32 @llvm.x86.sse.ucomineq.ss(<4 x float> %a0, <4 x float> %a1) ; <i32> [#uses=1]
-  ret i32 %res
-}
-declare i32 @llvm.x86.sse.ucomineq.ss(<4 x float>, <4 x float>) nounwind readnone
-
-
-define <8 x i8> @test_x86_ssse3_pabs_b(<8 x i8> %a0) {
-  ; CHECK: vpabsb
-  %res = call <8 x i8> @llvm.x86.ssse3.pabs.b(<8 x i8> %a0) ; <<8 x i8>> [#uses=1]
-  ret <8 x i8> %res
-}
-declare <8 x i8> @llvm.x86.ssse3.pabs.b(<8 x i8>) nounwind readnone
-
-
-define <16 x i8> @test_x86_ssse3_pabs_b_128(<16 x i8> %a0) {
-  ; CHECK: vpabsb
-  %res = call <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8> %a0) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.ssse3.pabs.b.128(<16 x i8>) nounwind readnone
-
-
-define <2 x i32> @test_x86_ssse3_pabs_d(<2 x i32> %a0) {
-  ; CHECK: vpabsd
-  %res = call <2 x i32> @llvm.x86.ssse3.pabs.d(<2 x i32> %a0) ; <<2 x i32>> [#uses=1]
-  ret <2 x i32> %res
-}
-declare <2 x i32> @llvm.x86.ssse3.pabs.d(<2 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_ssse3_pabs_d_128(<4 x i32> %a0) {
-  ; CHECK: vpabsd
-  %res = call <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32> %a0) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.ssse3.pabs.d.128(<4 x i32>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_pabs_w(<4 x i16> %a0) {
-  ; CHECK: vpabsw
-  %res = call <4 x i16> @llvm.x86.ssse3.pabs.w(<4 x i16> %a0) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.pabs.w(<4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_pabs_w_128(<8 x i16> %a0) {
-  ; CHECK: vpabsw
-  %res = call <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16> %a0) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.pabs.w.128(<8 x i16>) nounwind readnone
-
-
-define <2 x i32> @test_x86_ssse3_phadd_d(<2 x i32> %a0, <2 x i32> %a1) {
-  ; CHECK: vphaddd
-  %res = call <2 x i32> @llvm.x86.ssse3.phadd.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
-  ret <2 x i32> %res
-}
-declare <2 x i32> @llvm.x86.ssse3.phadd.d(<2 x i32>, <2 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_ssse3_phadd_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vphaddd
-  %res = call <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.ssse3.phadd.d.128(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_phadd_sw(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vphaddsw
-  %res = call <4 x i16> @llvm.x86.ssse3.phadd.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.phadd.sw(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <4 x i32> @test_x86_ssse3_phadd_sw_128(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vphaddsw
-  %res = call <4 x i32> @llvm.x86.ssse3.phadd.sw.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.ssse3.phadd.sw.128(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_phadd_w(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vphaddw
-  %res = call <4 x i16> @llvm.x86.ssse3.phadd.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.phadd.w(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_phadd_w_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vphaddw
-  %res = call <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.phadd.w.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <2 x i32> @test_x86_ssse3_phsub_d(<2 x i32> %a0, <2 x i32> %a1) {
-  ; CHECK: vphsubd
-  %res = call <2 x i32> @llvm.x86.ssse3.phsub.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
-  ret <2 x i32> %res
-}
-declare <2 x i32> @llvm.x86.ssse3.phsub.d(<2 x i32>, <2 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_ssse3_phsub_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vphsubd
-  %res = call <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.ssse3.phsub.d.128(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_phsub_sw(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vphsubsw
-  %res = call <4 x i16> @llvm.x86.ssse3.phsub.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.phsub.sw(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_phsub_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vphsubsw
-  %res = call <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.phsub.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_phsub_w(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vphsubw
-  %res = call <4 x i16> @llvm.x86.ssse3.phsub.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.phsub.w(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_phsub_w_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vphsubw
-  %res = call <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.phsub.w.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_pmadd_ub_sw(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vpmaddubsw
-  %res = call <4 x i16> @llvm.x86.ssse3.pmadd.ub.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.pmadd.ub.sw(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_pmadd_ub_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmaddubsw
-  %res = call <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_pmul_hr_sw(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vpmulhrsw
-  %res = call <4 x i16> @llvm.x86.ssse3.pmul.hr.sw(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.pmul.hr.sw(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_pmul_hr_sw_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpmulhrsw
-  %res = call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16>, <8 x i16>) nounwind readnone
-
-
-define <8 x i8> @test_x86_ssse3_pshuf_b(<8 x i8> %a0, <8 x i8> %a1) {
-  ; CHECK: vpshufb
-  %res = call <8 x i8> @llvm.x86.ssse3.pshuf.b(<8 x i8> %a0, <8 x i8> %a1) ; <<8 x i8>> [#uses=1]
-  ret <8 x i8> %res
-}
-declare <8 x i8> @llvm.x86.ssse3.pshuf.b(<8 x i8>, <8 x i8>) nounwind readnone
-
-
-define <16 x i8> @test_x86_ssse3_pshuf_b_128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpshufb
-  %res = call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <8 x i8> @test_x86_ssse3_psign_b(<8 x i8> %a0, <8 x i8> %a1) {
-  ; CHECK: vpsignb
-  %res = call <8 x i8> @llvm.x86.ssse3.psign.b(<8 x i8> %a0, <8 x i8> %a1) ; <<8 x i8>> [#uses=1]
-  ret <8 x i8> %res
-}
-declare <8 x i8> @llvm.x86.ssse3.psign.b(<8 x i8>, <8 x i8>) nounwind readnone
-
-
-define <16 x i8> @test_x86_ssse3_psign_b_128(<16 x i8> %a0, <16 x i8> %a1) {
-  ; CHECK: vpsignb
-  %res = call <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8> %a0, <16 x i8> %a1) ; <<16 x i8>> [#uses=1]
-  ret <16 x i8> %res
-}
-declare <16 x i8> @llvm.x86.ssse3.psign.b.128(<16 x i8>, <16 x i8>) nounwind readnone
-
-
-define <2 x i32> @test_x86_ssse3_psign_d(<2 x i32> %a0, <2 x i32> %a1) {
-  ; CHECK: vpsignd
-  %res = call <2 x i32> @llvm.x86.ssse3.psign.d(<2 x i32> %a0, <2 x i32> %a1) ; <<2 x i32>> [#uses=1]
-  ret <2 x i32> %res
-}
-declare <2 x i32> @llvm.x86.ssse3.psign.d(<2 x i32>, <2 x i32>) nounwind readnone
-
-
-define <4 x i32> @test_x86_ssse3_psign_d_128(<4 x i32> %a0, <4 x i32> %a1) {
-  ; CHECK: vpsignd
-  %res = call <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32> %a0, <4 x i32> %a1) ; <<4 x i32>> [#uses=1]
-  ret <4 x i32> %res
-}
-declare <4 x i32> @llvm.x86.ssse3.psign.d.128(<4 x i32>, <4 x i32>) nounwind readnone
-
-
-define <4 x i16> @test_x86_ssse3_psign_w(<4 x i16> %a0, <4 x i16> %a1) {
-  ; CHECK: vpsignw
-  %res = call <4 x i16> @llvm.x86.ssse3.psign.w(<4 x i16> %a0, <4 x i16> %a1) ; <<4 x i16>> [#uses=1]
-  ret <4 x i16> %res
-}
-declare <4 x i16> @llvm.x86.ssse3.psign.w(<4 x i16>, <4 x i16>) nounwind readnone
-
-
-define <8 x i16> @test_x86_ssse3_psign_w_128(<8 x i16> %a0, <8 x i16> %a1) {
-  ; CHECK: vpsignw
-  %res = call <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16> %a0, <8 x i16> %a1) ; <<8 x i16>> [#uses=1]
-  ret <8 x i16> %res
-}
-declare <8 x i16> @llvm.x86.ssse3.psign.w.128(<8 x i16>, <8 x i16>) nounwind readnone
-

Removed: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll?rev=109958&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86_64.ll (removed)
@@ -1,50 +0,0 @@
-; RUN: llc < %s -march=x86-64 -mattr=avx,aes | FileCheck %s
-
-define i64 @test_x86_sse2_cvtsd2si64(<2 x double> %a0) {
-  ; CHECK: vcvtsd2si
-  %res = call i64 @llvm.x86.sse2.cvtsd2si64(<2 x double> %a0) ; <i64> [#uses=1]
-  ret i64 %res
-}
-declare i64 @llvm.x86.sse2.cvtsd2si64(<2 x double>) nounwind readnone
-
-
-define <2 x double> @test_x86_sse2_cvtsi642sd(<2 x double> %a0, i64 %a1) {
-  ; CHECK: vcvtsi2sd
-  %res = call <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double> %a0, i64 %a1) ; <<2 x double>> [#uses=1]
-  ret <2 x double> %res
-}
-declare <2 x double> @llvm.x86.sse2.cvtsi642sd(<2 x double>, i64) nounwind readnone
-
-
-define i64 @test_x86_sse2_cvttsd2si64(<2 x double> %a0) {
-  ; CHECK: vcvttss2si
-  %res = call i64 @llvm.x86.sse2.cvttsd2si64(<2 x double> %a0) ; <i64> [#uses=1]
-  ret i64 %res
-}
-declare i64 @llvm.x86.sse2.cvttsd2si64(<2 x double>) nounwind readnone
-
-
-define i64 @test_x86_sse_cvtss2si64(<4 x float> %a0) {
-  ; CHECK: vcvtss2si
-  %res = call i64 @llvm.x86.sse.cvtss2si64(<4 x float> %a0) ; <i64> [#uses=1]
-  ret i64 %res
-}
-declare i64 @llvm.x86.sse.cvtss2si64(<4 x float>) nounwind readnone
-
-
-define <4 x float> @test_x86_sse_cvtsi642ss(<4 x float> %a0, i64 %a1) {
-  ; CHECK: vcvtsi2ss
-  %res = call <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float> %a0, i64 %a1) ; <<4 x float>> [#uses=1]
-  ret <4 x float> %res
-}
-declare <4 x float> @llvm.x86.sse.cvtsi642ss(<4 x float>, i64) nounwind readnone
-
-
-define i64 @test_x86_sse_cvttss2si64(<4 x float> %a0) {
-  ; CHECK: vcvttss2si
-  %res = call i64 @llvm.x86.sse.cvttss2si64(<4 x float> %a0) ; <i64> [#uses=1]
-  ret i64 %res
-}
-declare i64 @llvm.x86.sse.cvttss2si64(<4 x float>) nounwind readnone
-
-




From isanbard at gmail.com  Sat Jul 31 20:34:21 2010
From: isanbard at gmail.com (Bill Wendling)
Date: Sun, 01 Aug 2010 01:34:21 -0000
Subject: [llvm-commits] [llvm] r109966 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
Message-ID: <20100801013421.7F4182A6C12C@llvm.org>

Author: void
Date: Sat Jul 31 20:34:21 2010
New Revision: 109966

URL: http://llvm.org/viewvc/llvm-project?rev=109966&view=rev
Log:
Reference the personalities. Don't copy them into a new vector.

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

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp?rev=109966&r1=109965&r2=109966&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfException.cpp Sat Jul 31 20:34:21 2010
@@ -894,7 +894,7 @@
   if (!shouldEmitMovesModule && !shouldEmitTableModule)
     return;
 
-  const std::vector<const Function *> Personalities = MMI->getPersonalities();
+  const std::vector<const Function*> &Personalities = MMI->getPersonalities();
 
   for (unsigned I = 0, E = Personalities.size(); I < E; ++I)
     EmitCIE(Personalities[I], I);




From benny.kra at googlemail.com  Sun Aug  1 06:43:26 2010
From: benny.kra at googlemail.com (Benjamin Kramer)
Date: Sun, 01 Aug 2010 11:43:26 -0000
Subject: [llvm-commits] [llvm] r109971 - /llvm/trunk/lib/VMCore/Constants.cpp
Message-ID: <20100801114326.3E5912A6C12C@llvm.org>

Author: d0k
Date: Sun Aug  1 06:43:26 2010
New Revision: 109971

URL: http://llvm.org/viewvc/llvm-project?rev=109971&view=rev
Log:
Preallocate vector, avoid unnecessary vector growth.

Modified:
    llvm/trunk/lib/VMCore/Constants.cpp

Modified: llvm/trunk/lib/VMCore/Constants.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Constants.cpp?rev=109971&r1=109970&r2=109971&view=diff
==============================================================================
--- llvm/trunk/lib/VMCore/Constants.cpp (original)
+++ llvm/trunk/lib/VMCore/Constants.cpp Sun Aug  1 06:43:26 2010
@@ -526,6 +526,7 @@
 Constant* ConstantArray::get(LLVMContext &Context, StringRef Str,
                              bool AddNull) {
   std::vector<Constant*> ElementVals;
+  ElementVals.reserve(Str.size() + size_t(AddNull));
   for (unsigned i = 0; i < Str.size(); ++i)
     ElementVals.push_back(ConstantInt::get(Type::getInt8Ty(Context), Str[i]));
 




From steinke-dirk at web.de  Sun Aug  1 08:46:02 2010
From: steinke-dirk at web.de (Dirk Steinke)
Date: Sun, 01 Aug 2010 15:46:02 +0200
Subject: [llvm-commits] About: /llvm/trunk:
 lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
 test/Transforms/InstCombine/bit-checks.ll
Message-ID: <4C557A9A.2050203@web.de>

Hi everyone,

just my two bits about Owens patch to InstCombineAndOrXor.cpp.
The culprit should be the first line of the second transformation:

// (icmp eq (A & C1), 0) | (icmp eq (A & C2), 0) -->
// (icmp ne (A & (C1|C2)), (C1|C2)) where C1 and C2 are non-zero POT
if (LHSCC == ICmpInst::ICMP_EQ && LHSCst->isZero()) {
   Value *Op1 = 0, *Op2 = 0;

It is missing the "LHSCst == RHSCst && LHSCC == RHSCC" part.

The transformations themselves can't be wrong. They are actually too
specific.
The transformations should be
    (icmp eq (A & C1), C1) & (icmp eq (A & C2), C2)
     --> (icmp eq (A & (C1|C2)), (C1|C2))

    (icmp eq (A & C1), 0) & (icmp eq (A & C2), 0)
     --> (icmp eq (A & (C1|C2)), 0)

    (icmp ne (A & C1), C1) | (icmp ne (A & C2), C2)
     --> (icmp ne (A & (C1|C2)), (C1|C2))

    (icmp ne (A & C1), 0) | (icmp ne (A & C2), 0)
     --> (icmp ne (A & (C1|C2)), 0)

The current state only handles the special case, where C1 and
C2 only have a single bit set, so that
    (icmp eq (A & C1), C1)  iff  (icmp ne (A & C1), 0)
    (icmp ne (A & C1), C1)  iff  (icmp eq (A & C1), 0)

And: C1 and C2 don't have to be constant for those transformations to
work.

Another transformation not handled yet
    (icmp eq (A & C1), A) & (icmp eq (A & C2), A)
     --> (icmp eq (A & (C1&C2)), A)

I'll try to provide a patch, that handles the generic cases
    (icmp eq (A & B), C) & (icmp eq (A & D), E)
and
    (icmp ne (A & B), C) | (icmp ne (A & D), E)
when possible, in a couple of days. At least for the special
case (C == 0 || C == B) && (E == 0 || E == D), or when
B, C, D, and E are all constant.

bye
Dirk



From eli.friedman at gmail.com  Sun Aug  1 16:13:28 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Sun, 01 Aug 2010 21:13:28 -0000
Subject: [llvm-commits] [llvm] r109985 - in /llvm/trunk:
 lib/Target/Alpha/AlphaISelDAGToDAG.cpp
 test/CodeGen/Alpha/2010-08-01-mulreduce64.ll
Message-ID: <20100801211328.B5AD12A6C12C@llvm.org>

Author: efriedma
Date: Sun Aug  1 16:13:28 2010
New Revision: 109985

URL: http://llvm.org/viewvc/llvm-project?rev=109985&view=rev
Log:
PR7774: Fix undefined shifts in Alpha backend.  As a bonus, this actually
improves the generated code in some cases.


Added:
    llvm/trunk/test/CodeGen/Alpha/2010-08-01-mulreduce64.ll
Modified:
    llvm/trunk/lib/Target/Alpha/AlphaISelDAGToDAG.cpp

Modified: llvm/trunk/lib/Target/Alpha/AlphaISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/Alpha/AlphaISelDAGToDAG.cpp?rev=109985&r1=109984&r2=109985&view=diff
==============================================================================
--- llvm/trunk/lib/Target/Alpha/AlphaISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/Alpha/AlphaISelDAGToDAG.cpp Sun Aug  1 16:13:28 2010
@@ -113,8 +113,8 @@
     static uint64_t getNearPower2(uint64_t x) {
       if (!x) return 0;
       unsigned at = CountLeadingZeros_64(x);
-      uint64_t complow = 1 << (63 - at);
-      uint64_t comphigh = 1 << (64 - at);
+      uint64_t complow = 1ULL << (63 - at);
+      uint64_t comphigh = 1ULL << (64 - at);
       //cerr << x << ":" << complow << ":" << comphigh << "\n";
       if (abs64(complow - x) <= abs64(comphigh - x))
         return complow;

Added: llvm/trunk/test/CodeGen/Alpha/2010-08-01-mulreduce64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/Alpha/2010-08-01-mulreduce64.ll?rev=109985&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/Alpha/2010-08-01-mulreduce64.ll (added)
+++ llvm/trunk/test/CodeGen/Alpha/2010-08-01-mulreduce64.ll Sun Aug  1 16:13:28 2010
@@ -0,0 +1,11 @@
+; RUN: llc < %s -march=alpha | FileCheck %s
+
+define fastcc i64 @getcount(i64 %s) {
+	%tmp431 = mul i64 %s, 12884901888
+	ret i64 %tmp431
+}
+
+; CHECK: sll $16,33,$0
+; CHECK-NEXT: sll $16,32,$1
+; CHECK-NEXT: addq $0,$1,$0
+




From matt at console-pimps.org  Sun Aug  1 17:56:59 2010
From: matt at console-pimps.org (Matt Fleming)
Date: Sun, 01 Aug 2010 23:56:59 +0100
Subject: [llvm-commits] [PATCH 4/5] Add ELF ObjectWriter and Streamer
	support.
In-Reply-To: <AANLkTikpCshvf6iioHhQcZtRfsno7+GKPxzFuYjc8OtC@mail.gmail.com>
References: <cover.1280173465.git.matt@console-pimps.org>
	<cac33388f858fe37905e5eb017a333827b3ae1ba.1280173465.git.matt@console-pimps.org>
	<AANLkTikpCshvf6iioHhQcZtRfsno7+GKPxzFuYjc8OtC@mail.gmail.com>
Message-ID: <87ocdmx7pg.fsf@linux-g6p1.site>

On Mon, 26 Jul 2010 14:36:35 -0700, Eli Friedman <eli.friedman at gmail.com> wrote:

Hi Eli,

thanks for this detailed review. It's much appreciated. Sorry it took so
long to get back to you.

> On Mon, Jul 26, 2010 at 1:00 PM, Matt Fleming <matt at console-pimps.org> wrote:
> +// Because all the symbol flags need to be stored in the MCSymbolData
> +// 'flags' variable we need to provide shift constants per flag type.
> +enum {
> +  STT_SHIFT = 0, // Shift value for STT_* flags.
> +  STB_SHIFT = 4, // Shift value for STB_* flags.
> +  STV_SHIFT = 8  // Shift value ofr STV_* flags.
> +};
> 
> I don't think this belongs in ELF.h; it seems specific to the MC implementation.

Agreed. I've moved it into include/llvm/MC/MCELFSymbolFlags.h and added
some more enums to stop requiring all the shifting.

> > +  static bool isFixupKindPCRel(unsigned Kind) {
> > +    switch (Kind) {
> > +    default:
> > +      return false;
> > +    case X86::reloc_pcrel_1byte:
> > +    case X86::reloc_pcrel_4byte:
> > +    case X86::reloc_riprel_4byte:
> > +    case X86::reloc_riprel_4byte_movq_load:
> > +      return true;
> > +    }
> > +  }
> 
> Change the name to indicate that this x86-specific?

Done.

> > +  static bool isFixupKindRIPRel(unsigned Kind) {
> > +    return Kind == X86::reloc_riprel_4byte ||
> > +      Kind == X86::reloc_riprel_4byte_movq_load;
> > +  }
> 
> Same.

Done.

> > +  // Emit the ELF header.
> > +  void WriteHeader(uint64_t SectionDataSize, unsigned NumberOfSections) {
> 
> (General comment) Could you move the large method implementations out of the
> class definition?

Yeah, that seems fair. Done.

> > +    Write8(ELF::EV_CURRENT);        // e_ident[EI_VERSION]
> > +    Write8(ELF::ELFOSABI_LINUX);    // e_ident[EI_OSABI]
> > +    Write8(0);                 // e_ident[EI_ABIVERSION]
> 
> Factor this out so that it's obvious something needs to change here to make
> it crossplatform.

Hmm.. I'm not quite sure how to do that. The infrastructure for this
type of thing doesn't exist. Daniel, do you have any advice on this or
how it should be done?

> > +    // FIXME
> > +    Write16(ELF::EM_X86_64); // e_machine = target
> 
> Same.
> 
> > +    Write32(ELF::EV_CURRENT);         // e_version
> > +    WriteWord(0);                    // e_entry, no entry point in .o file
> > +    WriteWord(0);                    // e_phoff, no program header for .o
> > +    WriteWord(SectionDataSize + 64);  // e_shoff = sec hdr table off in bytes
> 
> Odd-looking indentation.

I've sorted out the indentation for all this code.

> > +    // FIXME
> > +    Write32(0);   // e_flags = whatever the target wants
> 
> Factor out.
> 
> > +    Write16(Is64Bit ? 64 : 52);  // e_ehsize = ELF header size
> 
> sizeof(Elf32_Ehdr)?  Or at least name these values somehow...

Yeah, sizeof(Elf32_Ehdr), etc seems sane. Done.

> > +    // e_shentsize = Section header entry size
> > +    Write16(Is64Bit ? 64 : 40);
> 
> See above.

Done.

> > +    // e_shnum     = # of section header ents
> > +    Write16(NumberOfSections);
> > +
> > +    // e_shstrndx  = Section # of '.shstrtab'
> > +    Write16(ShstrtabIndex);
> > +  }
> > +
> > +  void WriteSymbolEntry(MCDataFragment *F, uint64_t name, uint8_t info,
> > +                        uint64_t value, uint64_t size,
> > +                        uint8_t other, uint16_t shndx) {
> > +    if (Is64Bit) {
> > +      F->getContents() += StringRef((const char *)&name, 4);  // st_name
> > +      F->getContents() += StringRef((const char *)&info, 1);  // st_info
> > +      F->getContents() += StringRef((const char *)&other, 1); // st_other
> > +      F->getContents() += StringRef((const char *)&shndx, 2); // st_shndx
> > +      F->getContents() += StringRef((const char *)&value, 8); // st_value
> > +      // FIXME
> > +      F->getContents() += StringRef((const char *)&size, 8);  // st_size
> 
> FIXME what?  Same for all the unlableled FIXMEs.

These particular FIXMEs are no longer needed. I'll add descriptions to
the other unlabeled FIXMEs.

> > +      Value = Layout.getSymbolAddress(&Data);
> > +      MCValue Res;
> > +      if (Data.getSizeSymbol()->EvaluateAsRelocatable(Res, &Layout)) {
> > +        MCSymbolData &A =
> > +            Layout.getAssembler().getSymbolData(Res.getSymA()->getSymbol());
> > +        MCSymbolData &B =
> > +            Layout.getAssembler().getSymbolData(Res.getSymB()->getSymbol());
> > +
> > +       Size = Layout.getSymbolAddress(&A) - Layout.getSymbolAddress(&B);
> 
> How is this different from EvaluateAsAbsolute?

I'm not sure if EvalulateAsAbsolute would work here. I think the
expression may not be constant. Roman, what do you think?

> > +  // XXX: this is currently X86_64 only
> 
> Don't use XXX; use FIXME instead.

Done.

> > +      if (Symbol.isUndefined()) {
> > +        MSD.SectionIndex = ELF::SHN_UNDEF;
> > +       // XXX: for some reason we dont Emit* this
> > +       it->setFlags(it->getFlags() | (ELF::STB_GLOBAL << ELF::STB_SHIFT));
> > +        UndefinedSymbolData.push_back(MSD);
> 
> Indentation?
>
> > +
> > +  void WriteRelocations(MCAssembler &Asm, MCAsmLayout &Layout) {
> > +    for (MCAssembler::const_iterator it = Asm.begin(),
> > +           ie = Asm.end(); it != ie; ++it) {
> > +      const MCSectionData &SD = *it;
> > +
> > +      WriteRelocation(Asm, Layout, SD);
> 
> WriteRelocation(Asm, Layout, *it);

Fixed.

> > +  void WriteSecHdrEntry(uint32_t Name, uint32_t Type, uint64_t Flags,
> > +                       uint64_t Address, uint64_t Offset,
> > +                       uint64_t Size, uint32_t Link, uint32_t Info,
> > +                       uint64_t Alignment, uint64_t EntrySize) {
> > +    // Suppose that all non-initialized EntrySizes are actually a zero
> > +    if (EntrySize == ~0U)
> > +      EntrySize = 0;
> 
> Messy... would it be possible to just set it correctly in the first place?

Yeah, I've reworked this because of your comments on the EntrySize
patch.

> > +         F->getContents() +=  StringRef((const char *)&entry.r_offset, 8);
> > +         F->getContents() +=  StringRef((const char *)&entry.r_info, 8);
> 
> Is this endian-safe?

Nope. This is little endian specific. I've fixed this up now.

> > +    // Write out the ELF header ...
> > +    WriteHeader(SectionDataSize, NumSections);
> > +    FileOff = Is64Bit ? 64 : 52;
> 
> Again, names...

Done.

> > +    // ... and then the section header table.
> > +    // Should we align the section header table?
> > +    //
> > +    // Null secton first.
> 
> *section

Fixed.

> > +      switch(Section.getType()) {
> > +      case ELF::SHT_DYNAMIC:
> > +        sh_link = SectionStringTableIndex[&it->getSection()];
> > +        sh_info = 0;
> > +        break;
> > +
> > +      case ELF::SHT_REL:
> > +      case ELF::SHT_RELA:
> > +        const MCSection *SymtabSection;
> > +        const MCSection *InfoSection;
> > +        const StringRef *SectionName;
> > +        const MCSectionData *SymtabSD;
> > +        const MCSectionData *InfoSD;
> > +
> > +        SymtabSection = Asm.getContext().getELFSection(".symtab", ELF::SHT_SYMTAB, 0,
> > +                        SectionKind::getReadOnly(), false);
> > +        SymtabSD = &Asm.getSectionData(*SymtabSection);
> > +        // we have to count the empty section in too
> > +        sh_link = SymtabSD->getLayoutOrder() + 1;
> > +
> > +        SectionName = &Section.getSectionName();
> > +        SectionName = &SectionName->slice(5, SectionName->size());
> > +        InfoSection = Asm.getContext().getELFSection(*SectionName,
> > +          ELF::SHT_PROGBITS, 0, SectionKind::getReadOnly(), false);
> > +        InfoSD = &Asm.getSectionData(*InfoSection);
> > +        sh_info = InfoSD->getLayoutOrder() + 1;
> > +        break;
> > +
> > +      case ELF::SHT_SYMTAB:
> > +      case ELF::SHT_DYNSYM:
> > +        sh_link = StringTableIndex;
> > +        sh_info = LastLocalSymbolIndex;
> > +        break;
> > +
> > +      case ELF::SHT_HASH:
> > +      case ELF::SHT_GROUP:
> > +      case ELF::SHT_SYMTAB_SHNDX:
> 
> "default:" ?

Sure, that seems sensible. Fixed.

> > +        assert(0 && "FIXME: sh_type value not supported!");
> > +        break;
> > +      }
> > +
> > +      WriteSecHdrEntry(SectionStringTableIndex[&it->getSection()],
> > +                      Section.getType(), Section.getFlags(),
> > +                      Layout.getSectionAddress(&SD),
> > +                      SectionOffsetMap.lookup(&SD.getSection()),
> > +                      Layout.getSectionSize(&SD), sh_link,
> > +                       sh_info, SD.getAlignment(),
> > +                       SD.getEntrySize());
> 
> Indentation.
> 
> +namespace {
> +
> +class MCELFStreamer : public MCStreamer {
> +private:
> 
> Should inherit from MCObjectStreamer.

Fixed. 

> > +  MCFragment *getCurrentFragment() const {
> > +    assert(CurSectionData && "No current section!");
> > +
> > +    if (!CurSectionData->empty())
> > +      return &CurSectionData->getFragmentList().back();
> > +
> > +    return 0;
> > +  }
> 
> Shadows MCObjectStreamer::getCurrentFragment().
> 

Deleted.

> > +  /// fragment is not a data fragment.
> > +  MCDataFragment *getOrCreateDataFragment() const {
> > +    MCDataFragment *F = dyn_cast_or_null<MCDataFragment>(getCurrentFragment());
> > +    if (!F)
> > +      F = new MCDataFragment(CurSectionData);
> > +    return F;
> > +  }
> 
> Shadows MCObjectStreamer::getOrCreateDataFragment().

Deleted.

> > +  MCAssembler &getAssembler() { return Assembler; }
> 
> Shadows MCObjectStreamer::getAssembler().

Deleted.

> > +  const MCExpr *AddValueSymbols(const MCExpr *Value) {
> > +    switch (Value->getKind()) {
> > +    case MCExpr::Target: assert(0 && "Can't handle target exprs yet!");
> > +    case MCExpr::Constant:
> > +      break;
> > +
> > +    case MCExpr::Binary: {
> > +      const MCBinaryExpr *BE = cast<MCBinaryExpr>(Value);
> > +      AddValueSymbols(BE->getLHS());
> > +      AddValueSymbols(BE->getRHS());
> > +      break;
> > +    }
> > +
> > +    case MCExpr::SymbolRef:
> > +      Assembler.getOrCreateSymbolData(
> > +        cast<MCSymbolRefExpr>(Value)->getSymbol());
> > +      break;
> > +
> > +    case MCExpr::Unary:
> > +      AddValueSymbols(cast<MCUnaryExpr>(Value)->getSubExpr());
> > +      break;
> > +    }
> > +
> > +    return Value;
> > +  }
> 
> Shadows MCObjectStreamer::AddValueSymbols.

Deleted.

> > +
> > +  virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) {
> 
> Push definition outside of class defn.

I fixed a bug in this method and now it's only 2 lines long. Does the
definition still need to be pushed outside of the class definition?

> > +  virtual void EmitDwarfFileDirective(unsigned FileNo, StringRef Filename) {
> > +    errs() << "FIXME: MCELFStreamer:EmitDwarfFileDirective not implemented\n";
> > +  }
> 
> DEBUG(dbgs() << "")?

OK. Fixed.

> > +  virtual void EmitInstruction(const MCInst &Inst);
> > +  virtual void Finish();
> > +
> > +  /// @}
> > +};
> > +
> > +} // end anonymous namespace.
> > +
> > +void MCELFStreamer::SwitchSection(const MCSection *Section) {
> > +  assert(Section && "Cannot switch to a null section!");
> > +
> > +  // If already in this section, then this is a noop.
> > +  if (Section == CurSection) return;
> > +
> > +  CurSection = Section;
> > +  CurSectionData = &Assembler.getOrCreateSectionData(*Section);
> > +}
> 
> Identical to MCObjectStreamer::SwitchSection.

Deleted.

> > +void MCELFStreamer::EmitAssignment(MCSymbol *Symbol, const MCExpr *Value) {
> > +  // Only absolute symbols can be redefined.
> > +  assert((Symbol->isUndefined() || Symbol->isAbsolute()) &&
> > +         "Cannot define a symbol twice!");
> > +
> > +  // FIXME: Lift context changes into super class.
> > +  // FIXME: Set associated section.
> > +  //  Symbol->setValue(AddValueSymbols(Value));
> > +}
> 
> Should this be different from MCMachOStreamer::EmitAssignment?

I don't think so. I've fixed this now.

> > +void MCELFStreamer::EmitSymbolDesc(MCSymbol *Symbol, unsigned DescValue) {
> > +}
> 
> assert(0)?  Or what is this supposed to be useful for?

I actually don't think the .desc directive makes sense for ELF. I've
deleted it.

> > +void MCELFStreamer::EmitCommonSymbol(MCSymbol *Symbol, uint64_t Size,
> > +                                       unsigned ByteAlignment) {
> > +  MCSymbolData &SD = Assembler.getOrCreateSymbolData(*Symbol);
> > +
> > +  if ((SD.getFlags() & (0xff << ELF::STB_SHIFT)) == ELF::STB_LOCAL << ELF::STB_SHIFT) {
> 
> STB_* is only 4 bits, so 0xf, no?

Correct. Fixed.

> > +void MCELFStreamer::EmitZerofill(const MCSection *Section, MCSymbol *Symbol,
> > +                                   unsigned Size, unsigned ByteAlignment) {
> 
> AFAIK, .zerofill is Darwin-only.

True. Deleted.

> > +void MCELFStreamer::EmitTBSSSymbol(const MCSection *Section, MCSymbol *Symbol,
> > +                                   uint64_t Size, unsigned ByteAlignment) {
> 
> Also Darwin only, AFAIK.

True. Deleted.

> > +void MCELFStreamer::EmitBytes(StringRef Data, unsigned AddrSpace) {
> > +  getOrCreateDataFragment()->getContents().append(Data.begin(), Data.end());
> > +}
> 
> Same as MCMachOStreamer::EmitBytes; toss in a TODO?

Done.

> > +void MCELFStreamer::EmitValue(const MCExpr *Value, unsigned Size,
> > +                                unsigned AddrSpace) {
> > +  MCDataFragment *DF = getOrCreateDataFragment();
> > +
> > +  // Avoid fixups when possible.
> > +  int64_t AbsValue;
> > +  if (AddValueSymbols(Value)->EvaluateAsAbsolute(AbsValue)) {
> > +    // FIXME: Endianness assumption.
> > +    for (unsigned i = 0; i != Size; ++i)
> > +      DF->getContents().push_back(uint8_t(AbsValue >> (i * 8)));
> > +  } else {
> > +    DF->addFixup(MCFixup::Create(DF->getContents().size(), AddValueSymbols(Value),
> > +                                 MCFixup::getKindForSize(Size)));
> > +    DF->getContents().resize(DF->getContents().size() + Size, 0);
> > +  }
> > +}
> 
> Same as MCMachOStreamer::EmitValue; toss in a TODO?

Done.

> > +void MCELFStreamer::EmitValueToAlignment(unsigned ByteAlignment,
> > +                                           int64_t Value, unsigned ValueSize,
> > +                                           unsigned MaxBytesToEmit) {
> > +  if (MaxBytesToEmit == 0)
> > +    MaxBytesToEmit = ByteAlignment;
> > +  new MCAlignFragment(ByteAlignment, Value, ValueSize, MaxBytesToEmit,
> > +                      CurSectionData);
> > +
> > +  // Update the maximum alignment on the current section if necessary.
> > +  if (ByteAlignment > CurSectionData->getAlignment())
> > +    CurSectionData->setAlignment(ByteAlignment);
> > +}
> 
> Same as MCMachOStreamer::EmitValueToAlignment; toss in a TODO?

Done.

> > +void MCELFStreamer::EmitCodeAlignment(unsigned ByteAlignment,
> > +                                        unsigned MaxBytesToEmit) {
> > +  if (MaxBytesToEmit == 0)
> > +    MaxBytesToEmit = ByteAlignment;
> > +  new MCAlignFragment(ByteAlignment, 0, 1, MaxBytesToEmit,
> > +                      CurSectionData);
> > +
> > +  // Update the maximum alignment on the current section if necessary.
> > +  if (ByteAlignment > CurSectionData->getAlignment())
> > +    CurSectionData->setAlignment(ByteAlignment);
> > +}
> 
> Same as MCMachOStreamer::EmitCodeAlignment; toss in a TODO?

Done.

> > +void MCELFStreamer::EmitValueToOffset(const MCExpr *Offset,
> > +                                        unsigned char Value) {
> > +  new MCOrgFragment(*Offset, Value, CurSectionData);
> > +}
> 
> Same as MCMachOStreamer::EmitValueToOffset; toss in a TODO?

Done. I'll post the updated series sometime this week.

Thanks!


From daniel at zuster.org  Sun Aug  1 18:12:31 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sun, 01 Aug 2010 23:12:31 -0000
Subject: [llvm-commits] [zorg] r109993 -
	/zorg/trunk/buildbot/osuosl/master/config/builders.py
Message-ID: <20100801231232.013DD2A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 18:12:31 2010
New Revision: 109993

URL: http://llvm.org/viewvc/llvm-project?rev=109993&view=rev
Log:
buildbot: Switch FreeBSD builder back to using incremental builds.

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

Modified: zorg/trunk/buildbot/osuosl/master/config/builders.py
URL: http://llvm.org/viewvc/llvm-project/zorg/trunk/buildbot/osuosl/master/config/builders.py?rev=109993&r1=109992&r2=109993&view=diff
==============================================================================
--- zorg/trunk/buildbot/osuosl/master/config/builders.py (original)
+++ zorg/trunk/buildbot/osuosl/master/config/builders.py Sun Aug  1 18:12:31 2010
@@ -39,7 +39,7 @@
         {'name': "llvm-arm-linux",
          'slavenames':["ranby1"],
          'builddir':"llvm-arm-linux",
-         'factory': LLVMBuilder.getLLVMBuildFactory("arm-pc-linux-gnu", jobs=1, clean=False,
+         'factory': LLVMBuilder.getLLVMBuildFactory("arm-pc-linux-gnu", jobs=1, clean=True,
                                                     timeout=40)},
          {'name': "llvm-i686-linux-vg_leak",
           'slavenames':["osu8"],




From nicholas at mxc.ca  Sun Aug  1 18:18:45 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Sun, 01 Aug 2010 23:18:45 -0000
Subject: [llvm-commits] [llvm] r109995 -
	/llvm/trunk/docs/ProgrammersManual.html
Message-ID: <20100801231846.1D6F52A6C12C@llvm.org>

Author: nicholas
Date: Sun Aug  1 18:18:45 2010
New Revision: 109995

URL: http://llvm.org/viewvc/llvm-project?rev=109995&view=rev
Log:
Fix word choice.

Modified:
    llvm/trunk/docs/ProgrammersManual.html

Modified: llvm/trunk/docs/ProgrammersManual.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/ProgrammersManual.html?rev=109995&r1=109994&r2=109995&view=diff
==============================================================================
--- llvm/trunk/docs/ProgrammersManual.html (original)
+++ llvm/trunk/docs/ProgrammersManual.html Sun Aug  1 18:18:45 2010
@@ -1436,7 +1436,7 @@
 
 <p>The StringMap is very fast for several reasons: quadratic probing is very
 cache efficient for lookups, the hash value of strings in buckets is not
-recomputed when lookup up an element, StringMap rarely has to touch the
+recomputed when looking up an element, StringMap rarely has to touch the
 memory for unrelated objects when looking up a value (even when hash collisions
 happen), hash table growth does not recompute the hash values for strings
 already in the table, and each pair in the map is store in a single allocation




From daniel at zuster.org  Sun Aug  1 18:36:18 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Sun, 01 Aug 2010 23:36:18 -0000
Subject: [llvm-commits] [llvm] r109996 -
	/llvm/trunk/include/llvm/Target/TargetSelect.h
Message-ID: <20100801233618.436802A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 18:36:18 2010
New Revision: 109996

URL: http://llvm.org/viewvc/llvm-project?rev=109996&view=rev
Log:
Targets: Add InitializeNativeTargetAsmPrinter(), patch by Jan Sjodin, although
rewritten by me to not require updating all the target initialization routine
names.

Modified:
    llvm/trunk/include/llvm/Target/TargetSelect.h

Modified: llvm/trunk/include/llvm/Target/TargetSelect.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/TargetSelect.h?rev=109996&r1=109995&r2=109996&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/TargetSelect.h (original)
+++ llvm/trunk/include/llvm/Target/TargetSelect.h Sun Aug  1 18:36:18 2010
@@ -30,6 +30,13 @@
 #define LLVM_ASM_PRINTER(TargetName) void LLVMInitialize##TargetName##AsmPrinter();
 #include "llvm/Config/AsmPrinters.def"
 
+  // FIXME: Workaround for unfortunate definition of LLVM_NATIVE_ARCH.
+#define LLVM_ASM_PRINTER(TargetName) \
+  static inline void LLVMInitialize##TargetName##TargetAsmPrinter() { \
+    LLVMInitialize##TargetName##AsmPrinter(); \
+  }
+#include "llvm/Config/AsmPrinters.def"
+
   // Declare all of the available assembly parser initialization functions.
 #define LLVM_ASM_PARSER(TargetName) void LLVMInitialize##TargetName##AsmParser();
 #include "llvm/Config/AsmParsers.def"
@@ -113,6 +120,23 @@
     return true;
 #endif
   }  
+
+  /// InitializeNativeTargetAsmPrinter - The main program should call
+  /// this function to initialize the native target asm printer.
+  inline bool InitializeNativeTargetAsmPrinter() {
+  // If we have a native target, initialize the corresponding asm printer.
+#ifdef LLVM_NATIVE_ARCH
+#define DoInit2(TARG) \
+    LLVMInitialize ## TARG ## AsmPrinter ();
+#define DoInit(T) DoInit2(T)
+    DoInit(LLVM_NATIVE_ARCH);
+    return false;
+#undef DoInit
+#undef DoInit2
+#else
+    return true;
+#endif
+  }  
 }
 
 #endif




From eli.friedman at gmail.com  Sun Aug  1 18:43:26 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Sun, 1 Aug 2010 16:43:26 -0700
Subject: [llvm-commits] [PATCH 4/5] Add ELF ObjectWriter and Streamer
	support.
In-Reply-To: <87ocdmx7pg.fsf@linux-g6p1.site>
References: <cover.1280173465.git.matt@console-pimps.org>
	<cac33388f858fe37905e5eb017a333827b3ae1ba.1280173465.git.matt@console-pimps.org>
	<AANLkTikpCshvf6iioHhQcZtRfsno7+GKPxzFuYjc8OtC@mail.gmail.com>
	<87ocdmx7pg.fsf@linux-g6p1.site>
Message-ID: <AANLkTi=i7nCM2mFux5W6xS_Awn2nkYzEnHDdMam=MzHm@mail.gmail.com>

On Sun, Aug 1, 2010 at 3:56 PM, Matt Fleming <matt at console-pimps.org> wrote:
>> > +
>> > + ?virtual void EmitELFSize(MCSymbol *Symbol, const MCExpr *Value) {
>>
>> Push definition outside of class defn.
>
> I fixed a bug in this method and now it's only 2 lines long. Does the
> definition still need to be pushed outside of the class definition?

No, that should be fine.

-Eli



From daniel at zuster.org  Sun Aug  1 19:05:18 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:05:18 -0000
Subject: [llvm-commits] [llvm] r109997 - in /llvm/trunk: Makefile.rules
	test/Makefile
Message-ID: <20100802000518.2E8302A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:05:18 2010
New Revision: 109997

URL: http://llvm.org/viewvc/llvm-project?rev=109997&view=rev
Log:
tests: Make 'lit' the default test tool. You can still use 'make check-dg' to
run the tests using DejaGNU, but not for much longer. This is a last call for
DejaGNU supporters, if no one complains soon the DejaGNU support is going to
die.

Modified:
    llvm/trunk/Makefile.rules
    llvm/trunk/test/Makefile

Modified: llvm/trunk/Makefile.rules
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=109997&r1=109996&r2=109997&view=diff
==============================================================================
--- llvm/trunk/Makefile.rules (original)
+++ llvm/trunk/Makefile.rules Sun Aug  1 19:05:18 2010
@@ -1858,11 +1858,13 @@
 	  $(EchoCmd) No test directory ; \
 	fi
 
-check-lit::
+check-lit:: check
+
+check-dg::
 	$(Verb) if test -d "$(PROJ_OBJ_ROOT)/test" ; then \
 	  if test -f "$(PROJ_OBJ_ROOT)/test/Makefile" ; then \
 	    $(EchoCmd) Running test suite ; \
-	    $(MAKE) -C $(PROJ_OBJ_ROOT)/test check-local-lit ; \
+	    $(MAKE) -C $(PROJ_OBJ_ROOT)/test check-local-dg ; \
 	  else \
 	    $(EchoCmd) No Makefile in test directory ; \
 	  fi ; \

Modified: llvm/trunk/test/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Makefile?rev=109997&r1=109996&r2=109997&view=diff
==============================================================================
--- llvm/trunk/test/Makefile (original)
+++ llvm/trunk/test/Makefile Sun Aug  1 19:05:18 2010
@@ -10,11 +10,11 @@
 LEVEL = ..
 DIRS  =
 
-#
-# Make Dejagnu the default for testing
-#
 all:: check-local
 
+# 'lit' is the default test runner.Make Dejagnu the default for testing
+check-local:: check-local-lit
+
 # Include other test rules
 include Makefile.tests
 
@@ -90,12 +90,12 @@
 endif # SunOS
 
 ifneq ($(RUNTEST),)
-check-local:: site.exp
+check-local-dg:: site.exp
 	( $(ULIMIT) \
 	  PATH="$(LLVMToolDir):$(LLVM_SRC_ROOT)/test/Scripts:$(LLVMGCCDIR)/bin:$(PATH)" \
 	  $(RUNTEST) $(RUNTESTFLAGS) )
 else
-check-local:: site.exp
+check-local-dg:: site.exp
 	@echo "*** dejagnu not found.  Make sure 'runtest' is in your PATH, then reconfigure LLVM."
 endif
 




From echristo at apple.com  Sun Aug  1 19:07:06 2010
From: echristo at apple.com (Eric Christopher)
Date: Sun, 1 Aug 2010 17:07:06 -0700
Subject: [llvm-commits] [llvm] r109997 - in /llvm/trunk: Makefile.rules
	test/Makefile
In-Reply-To: <20100802000518.2E8302A6C12C@llvm.org>
References: <20100802000518.2E8302A6C12C@llvm.org>
Message-ID: <0E9FFB67-D4BB-4E23-9FBD-0AA67C5057AA@apple.com>


On Aug 1, 2010, at 5:05 PM, Daniel Dunbar wrote:

> tests: Make 'lit' the default test tool.

Yay!

Thanks for all the work Daniel!  I'm loving lit.

-eric


From eli.friedman at gmail.com  Sun Aug  1 19:18:19 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 02 Aug 2010 00:18:19 -0000
Subject: [llvm-commits] [llvm] r109998 - in /llvm/trunk:
 lib/Target/PowerPC/PPCISelLowering.cpp
 test/CodeGen/PowerPC/vec_constants.ll
Message-ID: <20100802001819.89C662A6C12C@llvm.org>

Author: efriedma
Date: Sun Aug  1 19:18:19 2010
New Revision: 109998

URL: http://llvm.org/viewvc/llvm-project?rev=109998&view=rev
Log:
PR7781: Fix incorrect shifting in PPCTargetLowering::LowerBUILD_VECTOR.


Modified:
    llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp
    llvm/trunk/test/CodeGen/PowerPC/vec_constants.ll

Modified: llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp?rev=109998&r1=109997&r2=109998&view=diff
==============================================================================
--- llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/PowerPC/PPCISelLowering.cpp Sun Aug  1 19:18:19 2010
@@ -3942,17 +3942,17 @@
     }
 
     // t = vsplti c, result = vsldoi t, t, 1
-    if (SextVal == ((i << 8) | (i >> (TypeShiftAmt-8)))) {
+    if (SextVal == ((i << 8) | (i < 0 ? 0xFF : 0))) {
       SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
       return BuildVSLDOI(T, T, 1, Op.getValueType(), DAG, dl);
     }
     // t = vsplti c, result = vsldoi t, t, 2
-    if (SextVal == ((i << 16) | (i >> (TypeShiftAmt-16)))) {
+    if (SextVal == ((i << 16) | (i < 0 ? 0xFFFF : 0))) {
       SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
       return BuildVSLDOI(T, T, 2, Op.getValueType(), DAG, dl);
     }
     // t = vsplti c, result = vsldoi t, t, 3
-    if (SextVal == ((i << 24) | (i >> (TypeShiftAmt-24)))) {
+    if (SextVal == ((i << 24) | (i < 0 ? 0xFFFFFF : 0))) {
       SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
       return BuildVSLDOI(T, T, 3, Op.getValueType(), DAG, dl);
     }

Modified: llvm/trunk/test/CodeGen/PowerPC/vec_constants.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/vec_constants.ll?rev=109998&r1=109997&r2=109998&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/vec_constants.ll (original)
+++ llvm/trunk/test/CodeGen/PowerPC/vec_constants.ll Sun Aug  1 19:18:19 2010
@@ -1,6 +1,6 @@
 ; RUN: llc < %s -march=ppc32 -mcpu=g5 | not grep CPI
 
-define void @test1(<4 x i32>* %P1, <4 x i32>* %P2, <4 x float>* %P3) {
+define void @test1(<4 x i32>* %P1, <4 x i32>* %P2, <4 x float>* %P3) nounwind {
 	%tmp = load <4 x i32>* %P1		; <<4 x i32>> [#uses=1]
 	%tmp4 = and <4 x i32> %tmp, < i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648 >		; <<4 x i32>> [#uses=1]
 	store <4 x i32> %tmp4, <4 x i32>* %P1
@@ -15,26 +15,30 @@
 	ret void
 }
 
-define <4 x i32> @test_30() {
+define <4 x i32> @test_30() nounwind {
 	ret <4 x i32> < i32 30, i32 30, i32 30, i32 30 >
 }
 
-define <4 x i32> @test_29() {
+define <4 x i32> @test_29() nounwind {
 	ret <4 x i32> < i32 29, i32 29, i32 29, i32 29 >
 }
 
-define <8 x i16> @test_n30() {
+define <8 x i16> @test_n30() nounwind {
 	ret <8 x i16> < i16 -30, i16 -30, i16 -30, i16 -30, i16 -30, i16 -30, i16 -30, i16 -30 >
 }
 
-define <16 x i8> @test_n104() {
+define <16 x i8> @test_n104() nounwind {
 	ret <16 x i8> < i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104, i8 -104 >
 }
 
-define <4 x i32> @test_vsldoi() {
+define <4 x i32> @test_vsldoi() nounwind {
 	ret <4 x i32> < i32 512, i32 512, i32 512, i32 512 >
 }
 
-define <4 x i32> @test_rol() {
+define <8 x i16> @test_vsldoi_65023() nounwind {
+	ret <8 x i16> < i16 65023, i16 65023,i16 65023,i16 65023,i16 65023,i16 65023,i16 65023,i16 65023 >
+}
+
+define <4 x i32> @test_rol() nounwind {
 	ret <4 x i32> < i32 -11534337, i32 -11534337, i32 -11534337, i32 -11534337 >
 }




From daniel at zuster.org  Sun Aug  1 19:39:39 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:39:39 -0000
Subject: [llvm-commits] [llvm] r109999 - /llvm/trunk/utils/lit/lit/lit.py
Message-ID: <20100802003939.12EF12A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:39:38 2010
New Revision: 109999

URL: http://llvm.org/viewvc/llvm-project?rev=109999&view=rev
Log:
lit: Allow clients to define predefined parameters.

Modified:
    llvm/trunk/utils/lit/lit/lit.py

Modified: llvm/trunk/utils/lit/lit/lit.py
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/lit/lit/lit.py?rev=109999&r1=109998&r2=109999&view=diff
==============================================================================
--- llvm/trunk/utils/lit/lit/lit.py (original)
+++ llvm/trunk/utils/lit/lit/lit.py Sun Aug  1 19:39:38 2010
@@ -358,8 +358,7 @@
     from LitTestCase import LitTestCase
     return unittest.TestSuite([LitTestCase(test, litConfig) for test in tests])
 
-def main():
-    # Bump the GIL check interval, its more important to get any one thread to a
+def main(builtinParameters = {}):    # Bump the GIL check interval, its more important to get any one thread to a
     # blocking operation (hopefully exec) than to try and unblock other threads.
     #
     # FIXME: This is a hack.
@@ -469,7 +468,7 @@
     inputs = args
 
     # Create the user defined parameters.
-    userParams = {}
+    userParams = dict(builtinParameters)
     for entry in opts.userParameters:
         if '=' not in entry:
             name,val = entry,''




From daniel at zuster.org  Sun Aug  1 19:39:42 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:39:42 -0000
Subject: [llvm-commits] [llvm] r110000 - in /llvm/trunk/utils: Makefile
 llvm-lit/ llvm-lit/Makefile llvm-lit/llvm-lit.in
Message-ID: <20100802003942.6E9342A6C12D@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:39:42 2010
New Revision: 110000

URL: http://llvm.org/viewvc/llvm-project?rev=110000&view=rev
Log:
tests: Add an 'llvm-lit' tool (script), which gets generated as part of the
build and has the object build directory baked into it. This allows 'llvm-lit'
to properly find the information needed to run the test suite in all cases,
without requiring the user to have LLVM or 'lit' available in their PATH, for
example.

Added:
    llvm/trunk/utils/llvm-lit/
    llvm/trunk/utils/llvm-lit/Makefile
    llvm/trunk/utils/llvm-lit/llvm-lit.in
Modified:
    llvm/trunk/utils/Makefile

Modified: llvm/trunk/utils/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/Makefile?rev=110000&r1=109999&r2=110000&view=diff
==============================================================================
--- llvm/trunk/utils/Makefile (original)
+++ llvm/trunk/utils/Makefile Sun Aug  1 19:39:42 2010
@@ -8,7 +8,8 @@
 ##===----------------------------------------------------------------------===##
 
 LEVEL = ..
-PARALLEL_DIRS := TableGen fpcmp PerfectShuffle FileCheck FileUpdate count not unittest
+PARALLEL_DIRS := FileCheck FileUpdate TableGen PerfectShuffle \
+	      count fpcmp llvm-lit not unittest
 
 EXTRA_DIST := cgiplotNLT.pl check-each-file codegen-diff countloc.sh cvsupdate \
               DSAclean.py DSAextract.py emacs findsym.pl GenLibDeps.pl \

Added: llvm/trunk/utils/llvm-lit/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/llvm-lit/Makefile?rev=110000&view=auto
==============================================================================
--- llvm/trunk/utils/llvm-lit/Makefile (added)
+++ llvm/trunk/utils/llvm-lit/Makefile Sun Aug  1 19:39:42 2010
@@ -0,0 +1,21 @@
+##===- utils/llvm-lit/Makefile -----------------------------*- Makefile -*-===##
+# 
+#                     The LLVM Compiler Infrastructure
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+# 
+##===----------------------------------------------------------------------===##
+
+LEVEL = ../..
+
+include $(LEVEL)/Makefile.common
+
+all:: $(ToolDir)/llvm-lit
+
+$(ToolDir)/llvm-lit: llvm-lit.in $(ObjDir)/.dir
+	$(Echo) "Creating 'llvm-lit' script..."
+	$(Verb)sed -e "s#@LLVM_SOURCE_DIR@#$(LLVM_SRC_ROOT)#g" \
+	     -e "s#@LLVM_BINARY_DIR@#$(LLVM_OBJ_ROOT)#g" \
+	     $< > $@
+	$(Verb)chmod +x $@

Added: llvm/trunk/utils/llvm-lit/llvm-lit.in
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/llvm-lit/llvm-lit.in?rev=110000&view=auto
==============================================================================
--- llvm/trunk/utils/llvm-lit/llvm-lit.in (added)
+++ llvm/trunk/utils/llvm-lit/llvm-lit.in Sun Aug  1 19:39:42 2010
@@ -0,0 +1,21 @@
+#!/usr/bin/env python
+
+import os
+import sys
+
+# Variables configured at build time.
+llvm_source_root = "@LLVM_SOURCE_DIR@"
+llvm_obj_root = "@LLVM_BINARY_DIR@"
+
+# Make sure we can find the lit package.
+sys.path.append(os.path.join(llvm_source_root, 'utils', 'lit'))
+
+# Set up some builtin parameters, so that by default the LLVM test suite
+# configuration file knows how to find the object tree.
+builtin_parameters = {
+    'llvm_site_config' : os.path.join(llvm_obj_root, 'test', 'lit.site.cfg')
+    }
+
+if __name__=='__main__':
+    import lit
+    lit.main(builtin_parameters)




From daniel at zuster.org  Sun Aug  1 19:45:41 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:45:41 -0000
Subject: [llvm-commits] [llvm] r110001 - /llvm/trunk/utils/llvm-lit/Makefile
Message-ID: <20100802004541.283BA2A6C12E@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:45:41 2010
New Revision: 110001

URL: http://llvm.org/viewvc/llvm-project?rev=110001&view=rev
Log:
build: Fix a dependency.

Modified:
    llvm/trunk/utils/llvm-lit/Makefile

Modified: llvm/trunk/utils/llvm-lit/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/llvm-lit/Makefile?rev=110001&r1=110000&r2=110001&view=diff
==============================================================================
--- llvm/trunk/utils/llvm-lit/Makefile (original)
+++ llvm/trunk/utils/llvm-lit/Makefile Sun Aug  1 19:45:41 2010
@@ -13,7 +13,7 @@
 
 all:: $(ToolDir)/llvm-lit
 
-$(ToolDir)/llvm-lit: llvm-lit.in $(ObjDir)/.dir
+$(ToolDir)/llvm-lit: llvm-lit.in $(ToolDir)/.dir
 	$(Echo) "Creating 'llvm-lit' script..."
 	$(Verb)sed -e "s#@LLVM_SOURCE_DIR@#$(LLVM_SRC_ROOT)#g" \
 	     -e "s#@LLVM_BINARY_DIR@#$(LLVM_OBJ_ROOT)#g" \




From daniel at zuster.org  Sun Aug  1 19:52:42 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:52:42 -0000
Subject: [llvm-commits] [llvm] r110002 - /llvm/trunk/test/TestRunner.sh
Message-ID: <20100802005242.10A602A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:52:41 2010
New Revision: 110002

URL: http://llvm.org/viewvc/llvm-project?rev=110002&view=rev
Log:
tests: Deprecate TestRunner.sh, and have it just invoke 'llvm-lit' (which will
need to be in your path). Please move to using 'llvm-lit' if you are still using
TestRunner.sh.

Modified:
    llvm/trunk/test/TestRunner.sh

Modified: llvm/trunk/test/TestRunner.sh
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/TestRunner.sh?rev=110002&r1=110001&r2=110002&view=diff
==============================================================================
--- llvm/trunk/test/TestRunner.sh (original)
+++ llvm/trunk/test/TestRunner.sh Sun Aug  1 19:52:41 2010
@@ -1,36 +1,5 @@
 #!/bin/sh
-#
-#  TestRunner.sh - This script is used to run the deja-gnu tests exactly like
-#  deja-gnu does, by executing the Tcl script specified in the test case's 
-#  RUN: lines. This is made possible by a simple make target supported by the
-#  test/Makefile. All this script does is invoke that make target. 
-#
-#  Usage:
-#     TestRunner.sh {script_names}
-#
-#     This script is typically used by cd'ing to a test directory and then
-#     running TestRunner.sh with a list of test file names you want to run.
-#
-TESTPATH=`pwd`
-SUBDIR=""
-if test `dirname $1` = "." ; then
-  while test `basename $TESTPATH` != "test" -a ! -z "$TESTPATH" ; do
-    tmp=`basename $TESTPATH`
-    SUBDIR="$tmp/$SUBDIR"
-    TESTPATH=`dirname $TESTPATH`
-  done
-fi
+# Deprecated, use 'llvm-lit'.
 
-for TESTFILE in "$@" ; do 
-  if test `dirname $TESTFILE` = . ; then
-    if test -d "$TESTPATH" ; then
-      cd $TESTPATH
-      make check-one TESTONE="$SUBDIR$TESTFILE"
-      cd $PWD
-    else
-      echo "Can't find llvm/test directory in " `pwd`
-    fi
-  else
-    make check-one TESTONE=$TESTFILE
-  fi
-done
+echo "warning: '$0' is deprecated, use 'llvm-lit' instead."
+llvm-lit "$@"




From daniel at zuster.org  Sun Aug  1 19:52:44 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 00:52:44 -0000
Subject: [llvm-commits] [llvm] r110003 - in /llvm/trunk: Makefile
	test/Makefile
Message-ID: <20100802005244.D4B5E2A6C12E@llvm.org>

Author: ddunbar
Date: Sun Aug  1 19:52:44 2010
New Revision: 110003

URL: http://llvm.org/viewvc/llvm-project?rev=110003&view=rev
Log:
tests: Kill off custom targets which were just there for TestRunner.sh.

Modified:
    llvm/trunk/Makefile
    llvm/trunk/test/Makefile

Modified: llvm/trunk/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile?rev=110003&r1=110002&r2=110003&view=diff
==============================================================================
--- llvm/trunk/Makefile (original)
+++ llvm/trunk/Makefile Sun Aug  1 19:52:44 2010
@@ -195,9 +195,6 @@
 check-llvm2cpp:
 	$(Verb)$(MAKE) check TESTSUITE=Feature RUNLLVM2CPP=1
 
-check-one:
-	$(Verb)$(MAKE) -C test check-one TESTONE=$(TESTONE)
-
 srpm: $(LLVM_OBJ_ROOT)/llvm.spec
 	rpmbuild -bs $(LLVM_OBJ_ROOT)/llvm.spec
 

Modified: llvm/trunk/test/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Makefile?rev=110003&r1=110002&r2=110003&view=diff
==============================================================================
--- llvm/trunk/test/Makefile (original)
+++ llvm/trunk/test/Makefile Sun Aug  1 19:52:44 2010
@@ -107,26 +107,6 @@
 	( $(ULIMIT) \
 	  $(LLVM_SRC_ROOT)/utils/lit/lit.py $(LIT_ARGS) $(LIT_ALL_TESTSUITES) )
 
-ifdef TESTONE
-CLEANED_TESTONE := $(patsubst %/,%,$(TESTONE))
-CLEANED_TESTONE := $(patsubst test/%,%,$(CLEANED_TESTONE))
-SUBDIR := $(shell dirname $(CLEANED_TESTONE))
-TESTPATH := $(LLVM_SRC_ROOT)/test/$(CLEANED_TESTONE)
-check-one: site.exp $(TCLSH)
-	$(Verb)( echo "source $(LLVM_OBJ_ROOT)/test/site.exp" ; \
-	  echo "set subdir $(SUBDIR)" ; \
-	  echo "proc pass  { msg } { puts \"PASS: \$$msg\" } "; \
-	  echo "proc fail  { msg } { puts \"FAIL: \$$msg\" }" ; \
-	  echo "proc xfail { msg } { puts \"XFAIL: \$$msg\" }" ; \
-	  echo "proc xpass { msg } { puts \"XPASS: \$$msg\" }" ; \
-	  echo "proc verbose args { }" ; \
-	  echo "source $(LLVM_SRC_ROOT)/test/lib/llvm.exp" ; \
-	  echo "RunLLVMTests $(TESTPATH)" ) | \
-	( $(ULIMIT) \
-	  PATH="$(LLVMToolDir):$(LLVM_SRC_ROOT)/test/Scripts:$(PATH)" \
-	  $(TCLSH) )
-endif
-
 clean::
 	$(RM) -rf `find $(LLVM_OBJ_ROOT)/test -name Output -type d -print`
 




From daniel at zuster.org  Sun Aug  1 20:20:21 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 01:20:21 -0000
Subject: [llvm-commits] [llvm] r110004 - /llvm/trunk/docs/TestingGuide.html
Message-ID: <20100802012021.21C0F2A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 20:20:20 2010
New Revision: 110004

URL: http://llvm.org/viewvc/llvm-project?rev=110004&view=rev
Log:
docs/TestingGuide: Rip out description of old nightly tester infrastructure. The
official recommendation is to use LNT instead. I will write docs on setting up
an LNT installation for submission to llvm.org, one day.

Modified:
    llvm/trunk/docs/TestingGuide.html

Modified: llvm/trunk/docs/TestingGuide.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/TestingGuide.html?rev=110004&r1=110003&r2=110004&view=diff
==============================================================================
--- llvm/trunk/docs/TestingGuide.html (original)
+++ llvm/trunk/docs/TestingGuide.html Sun Aug  1 20:20:20 2010
@@ -43,7 +43,6 @@
       <li><a href="#testsuitecustom">Writing custom tests for llvm-test</a></li>
    </ul>
   </li>
-  <li><a href="#nightly">Running the nightly tester</a></li>
 </ol>
 
 <div class="doc_author">
@@ -1135,66 +1134,6 @@
 
 </div>
 
-<!--=========================================================================-->
-<div class="doc_section"><a name="nightly">Running the nightly tester</a></div>
-<!--=========================================================================-->
-
-<div class="doc_text">
-
-<p>
-The <a href="http://llvm.org/nightlytest/">LLVM Nightly Testers</a>
-automatically check out an LLVM tree, build it, run the "nightly" 
-program test (described above), run all of the DejaGNU tests, 
-delete the checked out tree, and then submit the results to 
-<a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>. 
-After test results are submitted to 
-<a href="http://llvm.org/nightlytest/">http://llvm.org/nightlytest/</a>,
-they are processed and displayed on the tests page. An email to 
-<a href="http://lists.cs.uiuc.edu/pipermail/llvm-testresults/">
-llvm-testresults at cs.uiuc.edu</a> summarizing the results is also generated. 
-This testing scheme is designed to ensure that programs don't break as well 
-as keep track of LLVM's progress over time.</p>
-
-<p>If you'd like to set up an instance of the nightly tester to run on your 
-machine, take a look at the comments at the top of the 
-<tt>utils/NewNightlyTest.pl</tt> file. If you decide to set up a nightly tester 
-please choose a unique nickname and invoke <tt>utils/NewNightlyTest.pl</tt> 
-with the "-nickname [yournickname]" command line option. 
-
-<p>You can create a shell script to encapsulate the running of the script.
-The optimized x86 Linux nightly test is run from just such a script:</p>
-
-<div class="doc_code">
-<pre>
-#!/bin/bash
-BASE=/proj/work/llvm/nightlytest
-export BUILDDIR=$BASE/build 
-export WEBDIR=$BASE/testresults 
-export LLVMGCCDIR=/proj/work/llvm/cfrontend/install
-export PATH=/proj/install/bin:$LLVMGCCDIR/bin:$PATH
-export LD_LIBRARY_PATH=/proj/install/lib
-cd $BASE
-cp /proj/work/llvm/llvm/utils/NewNightlyTest.pl .
-nice ./NewNightlyTest.pl -nice -release -verbose -parallel -enable-linscan \
-   -nickname NightlyTester -noexternals &gt; output.log 2&gt;&amp;1 
-</pre>
-</div>
-
-<p>It is also possible to specify the the location your nightly test results
-are submitted. You can do this by passing the command line option
-"-submit-server [server_address]" and "-submit-script [script_on_server]" to
-<tt>utils/NewNightlyTest.pl</tt>. For example, to submit to the llvm.org 
-nightly test results page, you would invoke the nightly test script with 
-"-submit-server llvm.org -submit-script /nightlytest/NightlyTestAccept.cgi". 
-If these options are not specified, the nightly test script sends the results 
-to the llvm.org nightly test results page.</p>
-
-<p>Take a look at the <tt>NewNightlyTest.pl</tt> file to see what all of the
-flags and strings do.  If you start running the nightly tests, please let us
-know. Thanks!</p>
-
-</div>
-
 <!-- *********************************************************************** -->
 
 <hr>




From daniel at zuster.org  Sun Aug  1 20:20:23 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 01:20:23 -0000
Subject: [llvm-commits] [llvm] r110005 - /llvm/trunk/docs/TestingGuide.html
Message-ID: <20100802012023.E35522A6C12D@llvm.org>

Author: ddunbar
Date: Sun Aug  1 20:20:23 2010
New Revision: 110005

URL: http://llvm.org/viewvc/llvm-project?rev=110005&view=rev
Log:
docs/TestingGuide: Minimal update to describe 'lit' based regression testing
instead of DejaGNU. Still a bit kooky, since the current test format still has
some strong Tcl roots. Oh well!

Modified:
    llvm/trunk/docs/TestingGuide.html

Modified: llvm/trunk/docs/TestingGuide.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/TestingGuide.html?rev=110005&r1=110004&r2=110005&view=diff
==============================================================================
--- llvm/trunk/docs/TestingGuide.html (original)
+++ llvm/trunk/docs/TestingGuide.html Sun Aug  1 20:20:23 2010
@@ -16,22 +16,22 @@
   <li><a href="#requirements">Requirements</a></li>
   <li><a href="#org">LLVM testing infrastructure organization</a>
     <ul>
-      <li><a href="#dejagnu">DejaGNU tests</a></li>
+      <li><a href="#regressiontests">Regression tests</a></li>
       <li><a href="#testsuite">Test suite</a></li>
     </ul>
   </li>
   <li><a href="#quick">Quick start</a>
     <ul>
-      <li><a href="#quickdejagnu">DejaGNU tests</a></li>
+      <li><a href="#quickregressiontests">Regression tests</a></li>
       <li><a href="#quicktestsuite">Test suite</a></li>
    </ul>
   </li>
-  <li><a href="#dgstructure">DejaGNU structure</a>
+  <li><a href="#rtstructure">Regression test structure</a>
     <ul>
-      <li><a href="#dgcustom">Writing new DejaGNU tests</a></li>
+      <li><a href="#rtcustom">Writing new regression tests</a></li>
       <li><a href="#FileCheck">The FileCheck utility</a></li>
-      <li><a href="#dgvars">Variables and substitutions</a></li>
-      <li><a href="#dgfeatures">Other features</a></li>
+      <li><a href="#rtvars">Variables and substitutions</a></li>
+      <li><a href="#rtfeatures">Other features</a></li>
    </ul>
   </li>
   <li><a href="#testsuitestructure">Test suite structure</a></li>
@@ -46,8 +46,7 @@
 </ol>
 
 <div class="doc_author">
-  <p>Written by John T. Criswell, <a
-  href="http://llvm.x10sys.com/rspencer">Reid Spencer</a>, and Tanya Lattner</p>
+  <p>Written by John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner</p>
 </div>
 
 <!--=========================================================================-->
@@ -56,9 +55,9 @@
 
 <div class="doc_text">
 
-<p>This document is the reference manual for the LLVM testing infrastructure. It documents
-the structure of the LLVM testing infrastructure, the tools needed to use it,
-and how to add and run tests.</p>
+<p>This document is the reference manual for the LLVM testing infrastructure. It
+documents the structure of the LLVM testing infrastructure, the tools needed to
+use it, and how to add and run tests.</p>
 
 </div>
 
@@ -68,17 +67,9 @@
 
 <div class="doc_text">
 
-<p>In order to use the LLVM testing infrastructure, you will need all of the software
-required to build LLVM, plus the following:</p>
-
-<dl>
-<dt><a href="http://www.gnu.org/software/dejagnu/">DejaGNU</a></dt>
-<dd>The Feature and Regressions tests are organized and run by DejaGNU.</dd>
-<dt><a href="http://expect.nist.gov/">Expect</a></dt>
-<dd>Expect is required by DejaGNU.</dd>
-<dt><a href="http://www.tcl.tk/software/tcltk/">tcl</a></dt>
-<dd>Tcl is required by DejaGNU. </dd>
-</dl>
+<p>In order to use the LLVM testing infrastructure, you will need all of the
+software required to build LLVM, as well
+as <a href="http://python.org">Python</a> 2.4 or later.</p>
 
 </div>
 
@@ -88,29 +79,28 @@
 
 <div class="doc_text">
 
-<p>The LLVM testing infrastructure contains two major categories of tests: code
-fragments and whole programs. Code fragments are referred to as the "DejaGNU
-tests" and are in the <tt>llvm</tt> module in subversion under the
-<tt>llvm/test</tt> directory. The whole programs tests are referred to as the
-"Test suite" and are in the <tt>test-suite</tt> module in subversion.
+<p>The LLVM testing infrastructure contains two major categories of tests:
+regression tests and whole programs. The regression tests are contained inside
+the LLVM repository itself under <tt>llvm/test</tt> and are expected to always
+pass -- they should be run before every commit. The whole programs tests are
+referred to as the "LLVM test suite" and are in the <tt>test-suite</tt> module
+in subversion.
 </p>
 
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="dejagnu">DejaGNU tests</a></div>
+<div class="doc_subsection"><a name="regressiontests">Regression tests</a></div>
 <!-- _______________________________________________________________________ -->
 
 <div class="doc_text">
 
-<p>Code fragments are small pieces of code that test a specific
-feature of LLVM or trigger a specific bug in LLVM.  They are usually
-written in LLVM assembly language, but can be written in other
-languages if the test targets a particular language front end (and the
-appropriate <tt>--with-llvmgcc</tt> options were used
-at <tt>configure</tt> time of the <tt>llvm</tt> module). These tests
-are driven by the DejaGNU testing framework, which is hidden behind a
-few simple makefiles.</p>
+<p>The regression tests are small pieces of code that test a specific feature of
+LLVM or trigger a specific bug in LLVM.  They are usually written in LLVM
+assembly language, but can be written in other languages if the test targets a
+particular language front end (and the appropriate <tt>--with-llvmgcc</tt>
+options were used at <tt>configure</tt> time of the <tt>llvm</tt> module). These
+tests are driven by the 'lit' testing tool, which is part of LLVM.</p>
 
 <p>These code fragments are not complete programs. The code generated
 from them is never executed to determine correct behavior.</p>
@@ -157,8 +147,8 @@
 
 <div class="doc_text">
 
-  <p>The tests are located in two separate Subversion modules. The
-  DejaGNU tests are in the main "llvm" module under the directory 
+  <p>The tests are located in two separate Subversion modules. The regressions
+  tests are in the main "llvm" module under the directory
   <tt>llvm/test</tt> (so you get these tests for free with the main llvm tree).
   The more comprehensive test suite that includes whole 
 programs in C and C++ is in the <tt>test-suite</tt> module. This module should
@@ -170,10 +160,10 @@
 Alternatively, you can configure the <tt>test-suite</tt> module manually.</p>
 
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="quickdejagnu">DejaGNU tests</a></div>
+<div class="doc_subsection"><a name="quickregressiontests">Regression tests</a></div>
 <!-- _______________________________________________________________________ -->
-<p>To run all of the simple tests in LLVM using DejaGNU, use the master Makefile
- in the <tt>llvm/test</tt> directory:</p>
+<p>To run all of the LLVM regression tests, use master Makefile in
+ the <tt>llvm/test</tt> directory:</p>
 
 <div class="doc_code">
 <pre>
@@ -189,38 +179,47 @@
 </pre>
 </div>
 
-<p>To run only a subdirectory of tests in <tt>llvm/test</tt> using DejaGNU (ie.
-Transforms), just set the TESTSUITE variable to the path of the
-subdirectory (relative to <tt>llvm/test</tt>):</p>
+<p>If you have <a href="http://clang.llvm.org">Clang</a> checked out and built,
+you can run the LLVM and Clang tests simultaneously using:</p>
+
+<p>or</p>
 
 <div class="doc_code">
 <pre>
-% gmake TESTSUITE=Transforms check
+% gmake check-all
 </pre>
 </div>
 
-<p><b>Note: If you are running the tests with <tt>objdir != subdir</tt>, you
-must have run the complete testsuite before you can specify a
-subdirectory.</b></p>
+<p>To run the tests with Valgrind (Memcheck by default), just append
+<tt>VG=1</tt> to the commands above, e.g.:</p>
+
+<div class="doc_code">
+<pre>
+% gmake check VG=1
+</pre>
+</div>
 
-<p>To run only a single test, set <tt>TESTONE</tt> to its path (relative to
-<tt>llvm/test</tt>) and make the <tt>check-one</tt> target:</p>
+<p>To run individual tests or subsets of tests, you can use the 'llvm-lit'
+script which is built as part of LLVM. For example, to run the
+'Integer/BitCast.ll' test by itself you can run:</p>
 
 <div class="doc_code">
 <pre>
-% gmake TESTONE=Feature/basictest.ll check-one
+% llvm-lit ~/llvm/test/Integer/BitCast.ll 
 </pre>
 </div>
 
-<p>To run the tests with Valgrind (Memcheck by default), just append
-<tt>VG=1</tt> to the commands above, e.g.:</p>
+<p>or to run all of the ARM CodeGen tests:</p>
 
 <div class="doc_code">
 <pre>
-% gmake check VG=1
+% llvm-lit ~/llvm/test/CodeGen/ARM
 </pre>
 </div>
 
+<p>For more information on using the 'lit' tool, see 'llvm-lit --help' or the
+'lit' man page.</p>
+
 <!-- _______________________________________________________________________ -->
 <div class="doc_subsection"><a name="quicktestsuite">Test suite</a></div>
 <!-- _______________________________________________________________________ -->
@@ -274,11 +273,11 @@
 </div>
 
 <!--=========================================================================-->
-<div class="doc_section"><a name="dgstructure">DejaGNU structure</a></div>
+<div class="doc_section"><a name="rtstructure">Regression test structure</a></div>
 <!--=========================================================================-->
 <div class="doc_text">
-  <p>The LLVM DejaGNU tests are driven by DejaGNU together with GNU Make and are
-  located in the <tt>llvm/test</tt> directory.
+  <p>The LLVM regression tests are driven by 'lit' and are located in
+  the <tt>llvm/test</tt> directory.
 
   <p>This directory contains a large array of small tests
   that exercise various features of LLVM and to ensure that regressions do not
@@ -301,23 +300,24 @@
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="dgcustom">Writing new DejaGNU tests</a></div>
+<div class="doc_subsection"><a name="rtcustom">Writing new regression tests</a></div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_text">
-  <p>The DejaGNU structure is very simple, but does require some information to 
-  be set. This information is gathered via <tt>configure</tt> and is written 
-  to a file, <tt>site.exp</tt> in <tt>llvm/test</tt>. The <tt>llvm/test</tt> 
-  Makefile does this work for you.</p>
-
-  <p>In order for DejaGNU to work, each directory of tests must have a 
-  <tt>dg.exp</tt> file. DejaGNU looks for this file to determine how to run the
-  tests. This file is just a Tcl script and it can do anything you want, but 
-  we've standardized it for the LLVM regression tests. If you're adding a
+  <p>The regression test structure is very simple, but does require some
+  information to be set. This information is gathered via <tt>configure</tt> and
+  is written to a file, <tt>lit.site.cfg</tt>
+  in <tt>llvm/test</tt>. The <tt>llvm/test</tt> Makefile does this work for
+  you.</p>
+
+  <p>In order for the regression tests to work, each directory of tests must
+  have a <tt>dg.exp</tt> file. Lit looks for this file to determine how to
+  run the tests. This file is just a Tcl script and it can do anything you want,
+  but we've standardized it for the LLVM regression tests. If you're adding a
   directory of tests, just copy <tt>dg.exp</tt> from another directory to get
-  running. The standard <tt>dg.exp</tt> simply loads a Tcl 
-  library (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt> 
-  function defined in that library with a list of file names to run. The names 
-  are obtained by using Tcl's glob command.  Any directory that contains only
+  running. The standard <tt>dg.exp</tt> simply loads a Tcl library
+  (<tt>test/lib/llvm.exp</tt>) and calls the <tt>llvm_runtests</tt> function
+  defined in that library with a list of file names to run. The names are
+  obtained by using Tcl's glob command.  Any directory that contains only
   directories does not need the <tt>dg.exp</tt> file.</p>
 
   <p>The <tt>llvm-runtests</tt> function lookas at each file that is passed to
@@ -378,7 +378,8 @@
 
   <p>There are some quoting rules that you must pay attention to when writing
   your RUN lines. In general nothing needs to be quoted. Tcl won't strip off any
-  ' or " so they will get passed to the invoked program. For example:</p>
+  quote characters so they will get passed to the invoked program. For
+  example:</p>
 
 <div class="doc_code">
 <pre>
@@ -695,7 +696,7 @@
 </div>
 
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="dgvars">Variables and
+<div class="doc_subsection"><a name="rtvars">Variables and
 substitutions</a></div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_text">
@@ -791,7 +792,7 @@
 </div>
   
 <!-- _______________________________________________________________________ -->
-<div class="doc_subsection"><a name="dgfeatures">Other Features</a></div>
+<div class="doc_subsection"><a name="rtfeatures">Other Features</a></div>
 <!-- _______________________________________________________________________ -->
 <div class="doc_text">
   <p>To make RUN line writing easier, there are several shell scripts located
@@ -817,7 +818,7 @@
   <p>Sometimes it is necessary to mark a test case as "expected fail" or XFAIL.
   You can easily mark a test as XFAIL just by including <tt>XFAIL: </tt> on a
   line near the top of the file. This signals that the test case should succeed
-  if the test fails. Such test cases are counted separately by DejaGnu. To
+  if the test fails. Such test cases are counted separately by the testing tool. To
   specify an expected fail, use the XFAIL keyword in the comments of the test
   program followed by a colon and one or more regular expressions (separated by
   a comma). The regular expressions allow you to XFAIL the test conditionally by
@@ -905,9 +906,10 @@
 organizations should be relatively self explanatory.</p>
 
 <p>Some tests are known to fail.  Some are bugs that we have not fixed yet;
-others are features that we haven't added yet (or may never add).  In DejaGNU,
-the result for such tests will be XFAIL (eXpected FAILure).  In this way, you
-can tell the difference between an expected and unexpected failure.</p>
+others are features that we haven't added yet (or may never add).  In the
+regression tests, the result for such tests will be XFAIL (eXpected FAILure).
+In this way, you can tell the difference between an expected and unexpected
+failure.</p>
 
 <p>The tests in the test suite have no such feature at this time. If the
 test passes, only warnings and other miscellaneous output will be generated.  If
@@ -1143,7 +1145,7 @@
   <a href="http://validator.w3.org/check/referer"><img
   src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
 
-  John T. Criswell, Reid Spencer, and Tanya Lattner<br>
+  John T. Criswell, Daniel Dunbar, Reid Spencer, and Tanya Lattner<br>
   <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
   Last modified: $Date$
 </address>




From daniel at zuster.org  Sun Aug  1 20:25:20 2010
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 02 Aug 2010 01:25:20 -0000
Subject: [llvm-commits] [llvm] r110006 - /llvm/trunk/test/Makefile
Message-ID: <20100802012520.951852A6C12C@llvm.org>

Author: ddunbar
Date: Sun Aug  1 20:25:20 2010
New Revision: 110006

URL: http://llvm.org/viewvc/llvm-project?rev=110006&view=rev
Log:
Fix comment.

Modified:
    llvm/trunk/test/Makefile

Modified: llvm/trunk/test/Makefile
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Makefile?rev=110006&r1=110005&r2=110006&view=diff
==============================================================================
--- llvm/trunk/test/Makefile (original)
+++ llvm/trunk/test/Makefile Sun Aug  1 20:25:20 2010
@@ -12,7 +12,7 @@
 
 all:: check-local
 
-# 'lit' is the default test runner.Make Dejagnu the default for testing
+# 'lit' is the default test runner.
 check-local:: check-local-lit
 
 # Include other test rules




From nicholas at mxc.ca  Sun Aug  1 22:16:19 2010
From: nicholas at mxc.ca (Nick Lewycky)
Date: Mon, 02 Aug 2010 03:16:19 -0000
Subject: [llvm-commits] [llvm] r110016 - /llvm/trunk/Makefile.rules
Message-ID: <20100802031620.097F52A6C12C@llvm.org>

Author: nicholas
Date: Sun Aug  1 22:16:19 2010
New Revision: 110016

URL: http://llvm.org/viewvc/llvm-project?rev=110016&view=rev
Log:
Fix this condition; it has an else clause attached for Darwin only. Patch
by Takumi Nakamura.

Modified:
    llvm/trunk/Makefile.rules

Modified: llvm/trunk/Makefile.rules
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/Makefile.rules?rev=110016&r1=110015&r2=110016&view=diff
==============================================================================
--- llvm/trunk/Makefile.rules (original)
+++ llvm/trunk/Makefile.rules Sun Aug  1 22:16:19 2010
@@ -618,7 +618,8 @@
 endif
 
 # Adjust linker flags for building an executable
-ifneq ($(HOST_OS), $(filter $(HOST_OS), Darwin Cygwin MingW))
+ifneq ($(HOST_OS), $(filter $(HOST_OS), Cygwin MingW))
+ifneq ($(HOST_OS), Darwin)
 ifdef TOOLNAME
   LD.Flags += $(RPATH) -Wl,'$$ORIGIN/../lib'
   ifdef EXAMPLE_TOOL
@@ -632,6 +633,7 @@
   LD.Flags += $(RPATH) -Wl, at executable_path/../lib
 endif
 endif
+endif
 
 
 #----------------------------------------------------------




From echristo at apple.com  Sun Aug  1 22:26:49 2010
From: echristo at apple.com (Eric Christopher)
Date: Sun, 1 Aug 2010 20:26:49 -0700
Subject: [llvm-commits] Patch Proposal: tweak *.d on mingw-msys build
	environment
In-Reply-To: <AANLkTimugdBvFhk8XO7y-DUBMkr2xyzr1jCC3VB3KKYX@mail.gmail.com>
References: <AANLkTimugdBvFhk8XO7y-DUBMkr2xyzr1jCC3VB3KKYX@mail.gmail.com>
Message-ID: <8AB39C99-F624-4EE6-BB1F-305D9260404D@apple.com>


On Jul 28, 2010, at 9:35 AM, NAKAMURA Takumi wrote:

> This patch substitutes DOSish driveletter(c:/xxx/) scheme to
> msys-compatible(/c/xxx/) one
> when build host is MingW.
> 
> I am using the msysgit (msys+mingw+git) package as development environment.
> C:/msysgit/mingw/bin/g++.exe (TDM-1 mingw32) 4.4.0
> C:/msysgit/mingw/bin/make.exe GNU Make 3.81
> 
> msys-make igonores dependency rules emitted by mingw-g++ with -MMD,
> ... without any warnings.
> 
> And I have confirmed this might not have any side effects on
> fedora12-ppc and cygwin.
> 


Anton: Can you give this a run on your system? I think you use mingw yes?

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: r109430-mingw-msys.diff.txt
Url: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20100801/65b46ca2/attachment.txt 
-------------- next part --------------


-eric

From echristo at apple.com  Sun Aug  1 23:06:45 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 02 Aug 2010 04:06:45 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r110017 -
 /llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s
Message-ID: <20100802040645.162B82A6C12C@llvm.org>

Author: echristo
Date: Sun Aug  1 23:06:44 2010
New Revision: 110017

URL: http://llvm.org/viewvc/llvm-project?rev=110017&view=rev
Log:
Make sure the stack is aligned on entry.

Modified:
    llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s

Modified: llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s?rev=110017&r1=110016&r2=110017&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s (original)
+++ llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support-darwin.s Sun Aug  1 23:06:44 2010
@@ -42,7 +42,9 @@
 .globl _snapshot_ret
 _snapshot_ret:
 	movq	%rdi, _rdi(%rip)
+	subq  $8, %rsp
 	call	*_callthis(%rip)
+	addq  $8, %rsp
 	movq	%rax, _rax(%rip)
 	movq	%rdx, _rdx(%rip)
 	movdqu	%xmm0, _xmm_regs+0(%rip)




From echristo at apple.com  Sun Aug  1 23:15:06 2010
From: echristo at apple.com (Eric Christopher)
Date: Mon, 02 Aug 2010 04:15:06 -0000
Subject: [llvm-commits] [llvm-gcc-4.2] r110018 -
 /llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s
Message-ID: <20100802041506.73AC72A6C12C@llvm.org>

Author: echristo
Date: Sun Aug  1 23:15:06 2010
New Revision: 110018

URL: http://llvm.org/viewvc/llvm-project?rev=110018&view=rev
Log:
Here too.

Modified:
    llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s

Modified: llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s
URL: http://llvm.org/viewvc/llvm-project/llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s?rev=110018&r1=110017&r2=110018&view=diff
==============================================================================
--- llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s (original)
+++ llvm-gcc-4.2/trunk/gcc/testsuite/gcc.target/x86_64/abi/asm-support.s Sun Aug  1 23:15:06 2010
@@ -46,7 +46,9 @@
 	.type	snapshot_ret, @function
 snapshot_ret:
 	movq	%rdi, rdi(%rip)
+	subq  $8, %rsp
 	call	*callthis(%rip)
+	addq  $8, %rsp
 	movq	%rax, rax(%rip)
 	movq	%rdx, rdx(%rip)
 	movdqu	%xmm0, xmm_regs+0(%rip)




From eli.friedman at gmail.com  Sun Aug  1 23:42:25 2010
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 02 Aug 2010 04:42:25 -0000
Subject: [llvm-commits] [llvm] r110019 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
Message-ID: <20100802044225.2D39B2A6C12C@llvm.org>

Author: efriedma
Date: Sun Aug  1 23:42:25 2010
New Revision: 110019

URL: http://llvm.org/viewvc/llvm-project?rev=110019&view=rev
Log:
PR7586: Make sure we don't claim that unknown bits are actually known in the
ISD::AND case of TargetLowering::SimplifyDemandedBits.


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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp?rev=110019&r1=110018&r2=110019&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/TargetLowering.cpp Sun Aug  1 23:42:25 2010
@@ -1497,11 +1497,10 @@
     // present in the input.
     APInt NewBits =
       APInt::getHighBitsSet(BitWidth,
-                            BitWidth - EVT.getScalarType().getSizeInBits()) &
-      NewMask;
+                            BitWidth - EVT.getScalarType().getSizeInBits());
     
     // If none of the extended bits are demanded, eliminate the sextinreg.
-    if (NewBits == 0)
+    if ((NewBits & NewMask) == 0)
       return TLO.CombineTo(Op, Op.getOperand(0));
 
     APInt InSignBit = APInt::getSignBit(EVT.getScalarType().getSizeInBits());