From echristo at apple.com  Mon May 18 00:33:05 2009
From: echristo at apple.com (Eric Christopher)
Date: Sun, 17 May 2009 22:33:05 -0700
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
Message-ID: <E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>


On May 17, 2009, at 7:47 PM, Owen Anderson wrote:

> <config.patch>

Couple of comments:

a) the idea of testing host configuration (which may not work) for  
something target specific seems a bit problematic.

b) the indenting is a bit weird too, ok, now real stuff

c)

  AC_RUN_IFELSE(AC_LANG_SOURCE([[int main() {
+                 volatile unsigned long val = 1;
+                 __sync_synchronize();
+                 if (__sync_val_compare_and_swap(&val, 1, 0) != 1 ||
+                     val != 0)
+                   return 1;
+                 return 0; }]]),

Do you care if it runs or if it links? I'd have thought the latter,  
but it's up to you.

d)

+                AC_DEFINE(ENABLE_THREADS, [0])

You can do this using something like:

AC_DEFINE(ENABLE_THREADS, 0,
         Define if gcc atomic builtins work on your system.)

e) ENABLE_THREADS seems like a really bad name :)

f) It seems like you want to do a configure time flag? I'm not  
positive though. If you did then something like this may work better:

AC_ARG_ENABLE(enable-gcc-atomics,
[  --enable-gcc-atomics   gcc atomic functions],,
enable_gcc_atomics=no)

(or yes, depending on what you want it to default to) and then  
something like:

if test x$enable_gcc_atomics = xyes ; then
   AC_DEFINE(ENABLE_THREADS, 1,
         Define if gcc atomic builtins work on your system.)
fi

g)

+                AC_SUBST(ENABLE_THREADS, [0])

You probably don't mean to SUBST and DEFINE this. Is it for the  
environment or the compiled C/C++ files?

Sorry to seemingly pick it apart so much, but you did ask :)

-eric


From eli.friedman at gmail.com  Mon May 18 03:44:07 2009
From: eli.friedman at gmail.com (Eli Friedman)
Date: Mon, 18 May 2009 08:44:07 -0000
Subject: [llvm-commits] [llvm] r71998 -
	/llvm/trunk/lib/Support/MemoryBuffer.cpp
Message-ID: <200905180844.n4I8i7sn030779@zion.cs.uiuc.edu>

Author: efriedma
Date: Mon May 18 03:44:04 2009
New Revision: 71998

URL: http://llvm.org/viewvc/llvm-project?rev=71998&view=rev
Log:
Tweak MemoryBuffer::getSTDIN so that it returns after the first EOF.  
It doesn't matter for piped input, but it's annoying when typing at the 
console.


Modified:
    llvm/trunk/lib/Support/MemoryBuffer.cpp

Modified: llvm/trunk/lib/Support/MemoryBuffer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/MemoryBuffer.cpp?rev=71998&r1=71997&r2=71998&view=diff

==============================================================================
--- llvm/trunk/lib/Support/MemoryBuffer.cpp (original)
+++ llvm/trunk/lib/Support/MemoryBuffer.cpp Mon May 18 03:44:04 2009
@@ -258,13 +258,16 @@
 
 MemoryBuffer *MemoryBuffer::getSTDIN() {
   char Buffer[4096*4];
-  
+
   std::vector<char> FileData;
-  
+
   // Read in all of the data from stdin, we cannot mmap stdin.
   sys::Program::ChangeStdinToBinary();
-  while (size_t ReadBytes = fread(Buffer, sizeof(char), 4096*4, stdin))
+  size_t ReadBytes;
+  do {
+    ReadBytes = fread(Buffer, sizeof(char), sizeof(Buffer), stdin);
     FileData.insert(FileData.end(), Buffer, Buffer+ReadBytes);
+  } while (ReadBytes == sizeof(Buffer));
 
   FileData.push_back(0); // &FileData[Size] is invalid. So is &*FileData.end().
   size_t Size = FileData.size();




From resistor at mac.com  Mon May 18 03:50:48 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 01:50:48 -0700
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
Message-ID: <73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>


On May 17, 2009, at 10:33 PM, Eric Christopher wrote:

> a) the idea of testing host configuration (which may not work) for  
> something target specific seems a bit problematic.

I don't really have a better idea.

> c)
>
> AC_RUN_IFELSE(AC_LANG_SOURCE([[int main() {
> +                 volatile unsigned long val = 1;
> +                 __sync_synchronize();
> +                 if (__sync_val_compare_and_swap(&val, 1, 0) != 1 ||
> +                     val != 0)
> +                   return 1;
> +                 return 0; }]]),
>
> Do you care if it runs or if it links? I'd have thought the latter,  
> but it's up to you.

Probably just links, I guess.

> d)
>
> +                AC_DEFINE(ENABLE_THREADS, [0])
>
> You can do this using something like:
>
> AC_DEFINE(ENABLE_THREADS, 0,
>        Define if gcc atomic builtins work on your system.)

What does this achieve?

> e) ENABLE_THREADS seems like a really bad name :)

I didn't choose it.  It's already part of LLVM's configure.ac.

> f) It seems like you want to do a configure time flag? I'm not  
> positive though. If you did then something like this may work better:
>
> AC_ARG_ENABLE(enable-gcc-atomics,
> [  --enable-gcc-atomics   gcc atomic functions],,
> enable_gcc_atomics=no)
>
> (or yes, depending on what you want it to default to) and then  
> something like:
>
> if test x$enable_gcc_atomics = xyes ; then
>  AC_DEFINE(ENABLE_THREADS, 1,
>        Define if gcc atomic builtins work on your system.)
> fi

There's already a --enable-threads configure thread, that defaults to  
true.

> g)
>
> +                AC_SUBST(ENABLE_THREADS, [0])
>
> You probably don't mean to SUBST and DEFINE this. Is it for the  
> environment or the compiled C/C++ files?

I'm somewhat confused on this one myself.  Early parts of the  
configure.ac use AC_SUBST, but there is a #define for it in config.h,  
which doesn't get changed if I just use AC_SUBST, hence whey I added  
the AC_DEFINE.  Perhaps it's a bug in the rest of configure.ac?

> Sorry to seemingly pick it apart so much, but you did ask :)

Much appreciated!

--Owen

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/447527d4/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2620 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/447527d4/attachment.bin 

From baldrick at free.fr  Mon May 18 09:05:11 2009
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 18 May 2009 16:05:11 +0200
Subject: [llvm-commits] [llvm] r71986
	-	/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
In-Reply-To: <533C8EB1-0A3D-4C41-AFEA-95C91DBF7C8F@gmail.com>
References: <200905171937.n4HJb3NQ025904@zion.cs.uiuc.edu>
	<533C8EB1-0A3D-4C41-AFEA-95C91DBF7C8F@gmail.com>
Message-ID: <4A116B17.2030806@free.fr>

Hi Bill,

> It's very strange, but on a self-hosted release build, I'm getting  
> this error:
> 
> Running /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/ 
> FrontendC/dg.exp ... FAIL:
> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/FrontendC/ 
> 2009-05-17-AlwaysInline.c Failed with exit(1) at line 1 while running:
> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/ 
> llvmgcc42~dst/Developer/usr/bin/llvm-gcc-4.2 -m32 -fstrict-aliasing - 
> Wstrict-aliasing -I/usr/include/c++/4.0.0/i686-apple-darwin9 -I/usr/ 
> include/c++/4.0.0 -emit-llvm -w -S
> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/FrontendC/ 
> 2009-05-17-AlwaysInline.c -O0 -o - -mllvm --disable-llvm-optzns | /usr/ 
> bin/grep bar cc1: Unknown command line argument '-fdisable-llvm- 
> optzns'. Try: 'cc1 --help'
> I'm not sure where the '-fdisable-llvm-optzns' command is coming from...

this is similar to what you get when you are compiling with the system
gcc and not llvm-gcc, and you pass it -mllvm --disable-llvm-optzns.  So
is it possible that it's not llvm-gcc being used here, but standard gcc?

Ciao,

Duncan.

> 
> -bw
> 
> On May 17, 2009, at 12:37 PM, Duncan Sands wrote:
> 
>> Author: baldrick
>> Date: Sun May 17 14:37:02 2009
>> New Revision: 71986
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=71986&view=rev
>> Log:
>> Check that the gcc front-end is not doing inlining
>> when not doing unit-at-a-time.
>>
>> Added:
>>    llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
>>
>> Added: llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
>> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c?rev=71986&view=auto
>>
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> ======================================================================
>> --- llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c (added)
>> +++ llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c Sun May 17  
>> 14:37:02 2009
>> @@ -0,0 +1,17 @@
>> +// RUN: %llvmgcc -S %s -O0 -o - -mllvm --disable-llvm-optzns | grep  
>> bar
>> +// Check that the gcc inliner is turned off.
>> +
>> +#include <stdio.h>
>> +static __inline__ __attribute__ ((always_inline))
>> +     int bar (int x)
>> +{
>> +  return 4;
>> +}
>> +
>> +void
>> +foo ()
>> +{
>> +  long long b = 1;
>> +  int Y = bar (4);
>> +  printf ("%d\n", Y);
>> +}
>>
>>
>> _______________________________________________
>> 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 criswell at uiuc.edu  Mon May 18 10:08:29 2009
From: criswell at uiuc.edu (John Criswell)
Date: Mon, 18 May 2009 15:08:29 -0000
Subject: [llvm-commits] [poolalloc] r72001 - in /poolalloc/trunk/lib:
 DSA/DataStructure.cpp DSA/Local.cpp PoolAllocate/Heuristic.cpp
 PoolAllocate/PASimple.cpp PoolAllocate/PointerCompress.cpp
 PoolAllocate/TransformFunctionBody.cpp
Message-ID: <200905181508.n4IF8TPb017038@zion.cs.uiuc.edu>

Author: criswell
Date: Mon May 18 10:08:28 2009
New Revision: 72001

URL: http://llvm.org/viewvc/llvm-project?rev=72001&view=rev
Log:
Updated to new LLVM API for TargetData.
Patch contributed by Haohui Mai.

Modified:
    poolalloc/trunk/lib/DSA/DataStructure.cpp
    poolalloc/trunk/lib/DSA/Local.cpp
    poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp
    poolalloc/trunk/lib/PoolAllocate/PASimple.cpp
    poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp
    poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp

Modified: poolalloc/trunk/lib/DSA/DataStructure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/DataStructure.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/DSA/DataStructure.cpp (original)
+++ poolalloc/trunk/lib/DSA/DataStructure.cpp Mon May 18 10:08:28 2009
@@ -400,7 +400,7 @@
           const ArrayType *AT = cast<ArrayType>(SS.Ty);
           ++SS.Idx;
           if (SS.Idx != AT->getNumElements()) {
-            SS.Offset += unsigned(TD.getTypePaddedSize(AT->getElementType()));
+            SS.Offset += unsigned(TD.getTypeAllocSize(AT->getElementType()));
             return;
           }
           Stack.pop_back();  // At the end of the array
@@ -435,7 +435,7 @@
             assert(SS.Idx < AT->getNumElements());
             Stack.push_back(StackState(AT->getElementType(),
                                        SS.Offset+SS.Idx*
-                             unsigned(TD.getTypePaddedSize(AT->getElementType()))));
+                             unsigned(TD.getTypeAllocSize(AT->getElementType()))));
           }
         }
       }
@@ -492,7 +492,7 @@
           (Size == 0 && !Ty->isSized() && !isArray()) ||
           (Size == 1 && Ty == Type::VoidTy && isArray()) ||
           (Size == 0 && !Ty->isSized() && !isArray()) ||
-          (TD.getTypePaddedSize(Ty) == Size)) &&
+          (TD.getTypeAllocSize(Ty) == Size)) &&
          "Size member of DSNode doesn't match the type structure!");
   assert(NewTy != Type::VoidTy && "Cannot merge void type into DSNode!");
 
@@ -515,7 +515,7 @@
   }
 
   // Figure out how big the new type we're merging in is...
-  unsigned NewTySize = NewTy->isSized() ? (unsigned)TD.getTypePaddedSize(NewTy) : 0;
+  unsigned NewTySize = NewTy->isSized() ? (unsigned)TD.getTypeAllocSize(NewTy) : 0;
 
   // Otherwise check to see if we can fold this type into the current node.  If
   // we can't, we fold the node completely, if we can, we potentially update our
@@ -655,7 +655,7 @@
   unsigned O = 0;
   const Type *SubType = Ty;
   while (O < Offset) {
-    assert(Offset-O < TD.getTypePaddedSize(SubType) && "Offset out of range!");
+    assert(Offset-O < TD.getTypeAllocSize(SubType) && "Offset out of range!");
 
     switch (SubType->getTypeID()) {
     case Type::StructTyID: {
@@ -670,7 +670,7 @@
     }
     case Type::ArrayTyID: {
       SubType = cast<ArrayType>(SubType)->getElementType();
-      unsigned ElSize = (unsigned)TD.getTypePaddedSize(SubType);
+      unsigned ElSize = (unsigned)TD.getTypeAllocSize(SubType);
       unsigned Remainder = (Offset-O) % ElSize;
       O = Offset-Remainder;
       break;
@@ -692,7 +692,7 @@
       isa<FunctionType>(NewTy)) return false;
 
   unsigned SubTypeSize = SubType->isSized() ?
-       (unsigned)TD.getTypePaddedSize(SubType) : 0;
+       (unsigned)TD.getTypeAllocSize(SubType) : 0;
 
   // Ok, we are getting desperate now.  Check for physical subtyping, where we
   // just require each element in the node to be compatible.
@@ -720,12 +720,12 @@
       else
         NextPadSize = SubTypeSize;
       NextSubType = STy->getElementType(0);
-      NextSubTypeSize = (unsigned)TD.getTypePaddedSize(NextSubType);
+      NextSubTypeSize = (unsigned)TD.getTypeAllocSize(NextSubType);
       break;
     }
     case Type::ArrayTyID:
       NextSubType = cast<ArrayType>(SubType)->getElementType();
-      NextSubTypeSize = (unsigned)TD.getTypePaddedSize(NextSubType);
+      NextSubTypeSize = (unsigned)TD.getTypeAllocSize(NextSubType);
       NextPadSize = NextSubTypeSize;
       break;
     default: ;

Modified: poolalloc/trunk/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/DSA/Local.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/DSA/Local.cpp (original)
+++ poolalloc/trunk/lib/DSA/Local.cpp Mon May 18 10:08:28 2009
@@ -447,7 +447,7 @@
       if (ConstantInt *CS = dyn_cast<ConstantInt>(GEP.getOperand(i))) {
         Offset += 
           (CS->getType()->isSigned() ? CS->getSExtValue() : CS->getZExtValue())
-          * TD.getTypePaddedSize(CurTy);
+          * TD.getTypeAllocSize(CurTy);
       } else {
         // Variable index into a node.  We must merge all of the elements of the
         // sequential type here.
@@ -455,7 +455,7 @@
           cerr << "Pointer indexing not handled yet!\n";
         else {
           const ArrayType *ATy = cast<ArrayType>(STy);
-          unsigned ElSize = TD.getTypePaddedSize(CurTy);
+          unsigned ElSize = TD.getTypeAllocSize(CurTy);
           DSNode *N = Value.getNode();
           assert(N && "Value must have a node!");
           unsigned RawOffset = Offset+Value.getOffset();

Modified: poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/Heuristic.cpp Mon May 18 10:08:28 2009
@@ -57,7 +57,7 @@
 unsigned Heuristic::getRecommendedSize(const DSNode *N) {
   unsigned PoolSize = 0;
   if (!N->isArray() && N->getType()->isSized()) {
-    PoolSize = N->getParentGraph()->getTargetData().getTypePaddedSize(N->getType());
+    PoolSize = N->getParentGraph()->getTargetData().getTypeAllocSize(N->getType());
   }
   if (PoolSize == 1) PoolSize = 0;
   return PoolSize;

Modified: poolalloc/trunk/lib/PoolAllocate/PASimple.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PASimple.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PASimple.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PASimple.cpp Mon May 18 10:08:28 2009
@@ -178,7 +178,7 @@
         if (MI->isArrayAllocation()) {
           Value * NumElements = MI->getArraySize();
           Value * ElementSize = ConstantInt::get (Type::Int32Ty,
-                                                  TD.getTypePaddedSize(MI->getAllocatedType()));
+                                                  TD.getTypeAllocSize(MI->getAllocatedType()));
           AllocSize = BinaryOperator::Create (Instruction::Mul,
                                               ElementSize,
                                               NumElements,
@@ -186,7 +186,7 @@
                                               MI);
         } else {
           AllocSize = ConstantInt::get (Type::Int32Ty,
-                                        TD.getTypePaddedSize(MI->getAllocatedType()));
+                                        TD.getTypeAllocSize(MI->getAllocatedType()));
         }
 
         Value* args[] = {TheGlobalPool, AllocSize};

Modified: poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/PointerCompress.cpp Mon May 18 10:08:28 2009
@@ -226,7 +226,7 @@
   NewTy = ComputeCompressedType(Pool->getType(), 0, Nodes);
 
   // Get the compressed type size.
-  NewSize = NewTy->isSized() ? TD.getTypePaddedSize(NewTy) : 0;
+  NewSize = NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0;
 }
 
 
@@ -318,9 +318,9 @@
   const TargetData &TD = getNode()->getParentGraph()->getTargetData();
   std::cerr << "  From size: "
             << (getNode()->getType()->isSized() ? 
-                        TD.getTypePaddedSize(getNode()->getType()) : 0)
+                        TD.getTypeAllocSize(getNode()->getType()) : 0)
             << "  To size: "
-            << (NewTy->isSized() ? TD.getTypePaddedSize(NewTy) : 0) << "\n";
+            << (NewTy->isSized() ? TD.getTypeAllocSize(NewTy) : 0) << "\n";
   std::cerr << "Node: "; getNode()->dump();
   std::cerr << "New Type: " << *NewTy << "\n";
 }
@@ -733,7 +733,7 @@
           Idx = CastInst::CreateSExtOrBitCast(Idx, SCALARUINTTYPE, Idx->getName(), &GEPI);
 
         Constant *Scale = ConstantInt::get(SCALARUINTTYPE,
-                                            TD.getTypePaddedSize(ElTy));
+                                            TD.getTypeAllocSize(ElTy));
         Idx = BinaryOperator::CreateMul(Idx, Scale, "fieldidx", &GEPI);
         Val = BinaryOperator::CreateAdd(Val, Idx, GEPI.getName(), &GEPI);
       }

Modified: poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp?rev=72001&r1=72000&r2=72001&view=diff

==============================================================================
--- poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp (original)
+++ poolalloc/trunk/lib/PoolAllocate/TransformFunctionBody.cpp Mon May 18 10:08:28 2009
@@ -216,7 +216,7 @@
 
   TargetData &TD = PAInfo.getAnalysis<TargetData>();
   Value *AllocSize =
-    ConstantInt::get(Type::Int32Ty, TD.getTypePaddedSize(MI.getAllocatedType()));
+    ConstantInt::get(Type::Int32Ty, TD.getTypeAllocSize(MI.getAllocatedType()));
 
   if (MI.isArrayAllocation())
     AllocSize = BinaryOperator::Create(Instruction::Mul, AllocSize,
@@ -274,7 +274,7 @@
     if (PH == 0 || isa<ConstantPointerNull>(PH)) return;
     TargetData &TD = PAInfo.getAnalysis<TargetData>();
     Value *AllocSize =
-      ConstantInt::get(Type::Int32Ty, TD.getTypePaddedSize(MI.getAllocatedType()));
+      ConstantInt::get(Type::Int32Ty, TD.getTypeAllocSize(MI.getAllocatedType()));
     
     if (MI.isArrayAllocation())
       AllocSize = BinaryOperator::Create(Instruction::Mul, AllocSize,
@@ -335,7 +335,7 @@
 
 void FuncTransform::visitCallocCall(CallSite CS) {
   TargetData& TD = PAInfo.getAnalysis<TargetData>();
-  bool useLong = TD.getTypePaddedSize(PointerType::getUnqual(Type::Int8Ty)) != 4;
+  bool useLong = TD.getTypeAllocSize(PointerType::getUnqual(Type::Int8Ty)) != 4;
   
   Module *M = CS.getInstruction()->getParent()->getParent()->getParent();
   assert(CS.arg_end()-CS.arg_begin() == 2 && "calloc takes two arguments!");




From gohman at apple.com  Mon May 18 10:16:49 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 15:16:49 -0000
Subject: [llvm-commits] [llvm] r72002 - in /llvm/trunk/test:
 Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll
 CodeGen/ARM/arm-negative-stride.ll CodeGen/ARM/lsr-scale-addr-mode.ll
 CodeGen/X86/2007-11-06-InstrSched.ll CodeGen/X86/loop-hoist.ll
 Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll
Message-ID: <200905181516.n4IFGo8U017308@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 10:16:49 2009
New Revision: 72002

URL: http://llvm.org/viewvc/llvm-project?rev=72002&view=rev
Log:
Add nounwind to a few tests.

Modified:
    llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll
    llvm/trunk/test/CodeGen/ARM/arm-negative-stride.ll
    llvm/trunk/test/CodeGen/ARM/lsr-scale-addr-mode.ll
    llvm/trunk/test/CodeGen/X86/2007-11-06-InstrSched.ll
    llvm/trunk/test/CodeGen/X86/loop-hoist.ll
    llvm/trunk/test/Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll

Modified: llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll (original)
+++ llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll Mon May 18 10:16:49 2009
@@ -3,7 +3,7 @@
 ; RUN: grep -F "backedge-taken count is (-2147483632 + (2147483632 smax (-1 + (-1 * %x)) smax (-1 + (-1 * %y))))"
 ; PR2607
 
-define i32 @b(i32 %x, i32 %y) {
+define i32 @b(i32 %x, i32 %y) nounwind {
 entry:
 	%cmp2 = icmp slt i32 %y, %x
 	%cond3 = select i1 %cmp2, i32 %y, i32 %x

Modified: llvm/trunk/test/CodeGen/ARM/arm-negative-stride.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/arm-negative-stride.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/ARM/arm-negative-stride.ll (original)
+++ llvm/trunk/test/CodeGen/ARM/arm-negative-stride.ll Mon May 18 10:16:49 2009
@@ -1,6 +1,6 @@
 ; RUN: llvm-as < %s | llc -march=arm | grep {str r1, \\\[r.*, -r.*, lsl #2\}
 
-define void @test(i32* %P, i32 %A, i32 %i) {
+define void @test(i32* %P, i32 %A, i32 %i) nounwind {
 entry:
         icmp eq i32 %i, 0               ; <i1>:0 [#uses=1]
         br i1 %0, label %return, label %bb

Modified: llvm/trunk/test/CodeGen/ARM/lsr-scale-addr-mode.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/lsr-scale-addr-mode.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/ARM/lsr-scale-addr-mode.ll (original)
+++ llvm/trunk/test/CodeGen/ARM/lsr-scale-addr-mode.ll Mon May 18 10:16:49 2009
@@ -1,7 +1,7 @@
 ; RUN: llvm-as < %s | llc -march=arm | grep -F {str r2, \[r0, +r3, lsl #2\]}
 ; Should use scaled addressing mode.
 
-define void @sintzero(i32* %a) {
+define void @sintzero(i32* %a) nounwind {
 entry:
 	store i32 0, i32* %a
 	br label %cond_next

Modified: llvm/trunk/test/CodeGen/X86/2007-11-06-InstrSched.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2007-11-06-InstrSched.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/X86/2007-11-06-InstrSched.ll (original)
+++ llvm/trunk/test/CodeGen/X86/2007-11-06-InstrSched.ll Mon May 18 10:16:49 2009
@@ -1,6 +1,6 @@
 ; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse2 | not grep lea
 
-define float @foo(i32* %x, float* %y, i32 %c) {
+define float @foo(i32* %x, float* %y, i32 %c) nounwind {
 entry:
 	%tmp2132 = icmp eq i32 %c, 0		; <i1> [#uses=1]
 	br i1 %tmp2132, label %bb23, label %bb18

Modified: llvm/trunk/test/CodeGen/X86/loop-hoist.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/loop-hoist.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/X86/loop-hoist.ll (original)
+++ llvm/trunk/test/CodeGen/X86/loop-hoist.ll Mon May 18 10:16:49 2009
@@ -8,7 +8,7 @@
 
 @Arr = external global [0 x i32]		; <[0 x i32]*> [#uses=1]
 
-define void @foo(i32 %N.in, i32 %x) {
+define void @foo(i32 %N.in, i32 %x) nounwind {
 entry:
 	%N = bitcast i32 %N.in to i32		; <i32> [#uses=1]
 	br label %cond_true

Modified: llvm/trunk/test/Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll?rev=72002&r1=72001&r2=72002&view=diff

==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll (original)
+++ llvm/trunk/test/Transforms/IndVarSimplify/2007-01-08-X86-64-Pointer.ll Mon May 18 10:16:49 2009
@@ -1,7 +1,7 @@
 ; RUN: llvm-as < %s | llc -march=x86-64 | grep {(%rdi,%rax,8)}
 ; RUN: llvm-as < %s | llc -march=x86-64 | not grep {addq.*8}
 
-define void @foo(double* %y) {
+define void @foo(double* %y) nounwind {
 entry:
         br label %bb
 




From gohman at apple.com  Mon May 18 10:22:40 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 15:22:40 -0000
Subject: [llvm-commits] [llvm] r72003 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905181522.n4IFMeAq017531@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 10:22:39 2009
New Revision: 72003

URL: http://llvm.org/viewvc/llvm-project?rev=72003&view=rev
Log:
Add an isOne() utility function to ScalarEvolution, similar to isZero()
and similar to ConstantInt's isOne().

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

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=72003&r1=72002&r2=72003&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Mon May 18 10:22:39 2009
@@ -77,6 +77,10 @@
     ///
     bool isZero() const;
 
+    /// isOne - Return true if the expression is a constant one.
+    ///
+    bool isOne() const;
+
     /// replaceSymbolicValuesWithConcrete - If this SCEV internally references
     /// the symbolic value "Sym", construct and return a new SCEV that produces
     /// the same value, but which uses the concrete value Conc instead of the

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72003&r1=72002&r2=72003&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 10:22:39 2009
@@ -127,6 +127,11 @@
   return false;
 }
 
+bool SCEV::isOne() const {
+  if (const SCEVConstant *SC = dyn_cast<SCEVConstant>(this))
+    return SC->getValue()->isOne();
+  return false;
+}
 
 SCEVCouldNotCompute::SCEVCouldNotCompute() : SCEV(scCouldNotCompute) {}
 SCEVCouldNotCompute::~SCEVCouldNotCompute() {}
@@ -3392,7 +3397,7 @@
     const SCEVConstant *CStep = dyn_cast<SCEVConstant>(Step);
     if (!CStep || CStep->isZero())
       return UnknownValue;
-    if (CStep->getValue()->getValue() == 1) {
+    if (CStep->isOne()) {
       // With unit stride, the iteration never steps past the limit value.
     } else if (CStep->getValue()->getValue().isStrictlyPositive()) {
       if (const SCEVConstant *CLimit = dyn_cast<SCEVConstant>(RHS)) {




From gohman at apple.com  Mon May 18 10:36:10 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 15:36:10 -0000
Subject: [llvm-commits] [llvm] r72004 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h lib/Analysis/ScalarEvolution.cpp
 test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll
Message-ID: <200905181536.n4IFaARp018136@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 10:36:09 2009
New Revision: 72004

URL: http://llvm.org/viewvc/llvm-project?rev=72004&view=rev
Log:
Make ScalarEvolution::isLoopGuardedByCond work even when the edge
entering a loop is a non-split critical edge.

Modified:
    llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp
    llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=72004&r1=72003&r2=72004&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Mon May 18 10:36:09 2009
@@ -341,6 +341,10 @@
     BackedgeTakenInfo HowManyLessThans(const SCEV *LHS, const SCEV *RHS,
                                        const Loop *L, bool isSigned);
 
+    /// getLoopPredecessor - If the given loop's header has exactly one unique
+    /// predecessor outside the loop, return it. Otherwise return null.
+    BasicBlock *getLoopPredecessor(const Loop *L);
+
     /// getPredecessorWithUniqueSuccessorForBB - Return a predecessor of BB
     /// (which may not be an immediate predecessor) which has exactly one
     /// successor from which BB is reachable, or null if no such block is

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72004&r1=72003&r2=72004&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 10:36:09 2009
@@ -3246,6 +3246,21 @@
   return UnknownValue;
 }
 
+/// getLoopPredecessor - If the given loop's header has exactly one unique
+/// predecessor outside the loop, return it. Otherwise return null.
+///
+BasicBlock *ScalarEvolution::getLoopPredecessor(const Loop *L) {
+  BasicBlock *Header = L->getHeader();
+  BasicBlock *Pred = 0;
+  for (pred_iterator PI = pred_begin(Header), E = pred_end(Header);
+       PI != E; ++PI)
+    if (!L->contains(*PI)) {
+      if (Pred && Pred != *PI) return 0; // Multiple predecessors.
+      Pred = *PI;
+    }
+  return Pred;
+}
+
 /// getPredecessorWithUniqueSuccessorForBB - Return a predecessor of BB
 /// (which may not be an immediate predecessor) which has exactly one
 /// successor from which BB is reachable, or null if no such block is
@@ -3260,11 +3275,10 @@
     return Pred;
 
   // A loop's header is defined to be a block that dominates the loop.
-  // If the loop has a preheader, it must be a block that has exactly
-  // one successor that can reach BB. This is slightly more strict
-  // than necessary, but works if critical edges are split.
+  // If the header has a unique predecessor outside the loop, it must be
+  // a block that has exactly one successor that can reach the loop.
   if (Loop *L = LI->getLoopFor(BB))
-    return L->getLoopPreheader();
+    return getLoopPredecessor(L);
 
   return 0;
 }
@@ -3275,18 +3289,18 @@
 bool ScalarEvolution::isLoopGuardedByCond(const Loop *L,
                                           ICmpInst::Predicate Pred,
                                           const SCEV *LHS, const SCEV *RHS) {
-  BasicBlock *Preheader = L->getLoopPreheader();
-  BasicBlock *PreheaderDest = L->getHeader();
+  BasicBlock *Predecessor = getLoopPredecessor(L);
+  BasicBlock *PredecessorDest = L->getHeader();
 
-  // Starting at the preheader, climb up the predecessor chain, as long as
-  // there are predecessors that can be found that have unique successors
+  // Starting at the loop predecessor, climb up the predecessor chain, as long
+  // as there are predecessors that can be found that have unique successors
   // leading to the original header.
-  for (; Preheader;
-       PreheaderDest = Preheader,
-       Preheader = getPredecessorWithUniqueSuccessorForBB(Preheader)) {
+  for (; Predecessor;
+       PredecessorDest = Predecessor,
+       Predecessor = getPredecessorWithUniqueSuccessorForBB(Predecessor)) {
 
     BranchInst *LoopEntryPredicate =
-      dyn_cast<BranchInst>(Preheader->getTerminator());
+      dyn_cast<BranchInst>(Predecessor->getTerminator());
     if (!LoopEntryPredicate ||
         LoopEntryPredicate->isUnconditional())
       continue;
@@ -3299,7 +3313,7 @@
     Value *PreCondLHS = ICI->getOperand(0);
     Value *PreCondRHS = ICI->getOperand(1);
     ICmpInst::Predicate Cond;
-    if (LoopEntryPredicate->getSuccessor(0) == PreheaderDest)
+    if (LoopEntryPredicate->getSuccessor(0) == PredecessorDest)
       Cond = ICI->getPredicate();
     else
       Cond = ICI->getInversePredicate();

Modified: llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll?rev=72004&r1=72003&r2=72004&view=diff

==============================================================================
--- llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll (original)
+++ llvm/trunk/test/Analysis/ScalarEvolution/2008-07-29-SMinExpr.ll Mon May 18 10:36:09 2009
@@ -1,6 +1,6 @@
 ; RUN: llvm-as < %s | opt -analyze -scalar-evolution -disable-output \
 ; RUN:   -scalar-evolution-max-iterations=0 | \
-; RUN: grep -F "backedge-taken count is (-2147483632 + (2147483632 smax (-1 + (-1 * %x)) smax (-1 + (-1 * %y))))"
+; RUN: grep -F "backedge-taken count is (-2147483632 + ((-1 + (-1 * %x)) smax (-1 + (-1 * %y))))"
 ; PR2607
 
 define i32 @b(i32 %x, i32 %y) nounwind {




From baldrick at free.fr  Mon May 18 10:39:14 2009
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 18 May 2009 17:39:14 +0200
Subject: [llvm-commits] [llvm] r71410
	-	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfWriter.cpp
In-Reply-To: <48386810-85B9-46D5-8CE6-13FC50D21FFE@apple.com>
References: <200905102314.n4ANEdP9010260@zion.cs.uiuc.edu>	<99CEAE6F-A380-404D-9081-1BA1D8CB74F3@apple.com>
	<48386810-85B9-46D5-8CE6-13FC50D21FFE@apple.com>
Message-ID: <4A118122.9040303@free.fr>

Hi Dale,

>>> This doesn't take care of all of the problems with debug info for
>>> inlined
>>> functions, but it's a step in the right direction. For one thing,
>>> llvm-gcc
>>> generates wrong IR (it's missing some llvm.dbg intrinsics at the
>>> point where the
>>> function's inlined) for this example:
>>>
>>> #include <stdio.h>
>>> static __inline__ __attribute__((always_inline))  int bar(int x)
>>> { return 4; }
>>> void foo() {
>>> long long b = 1;
>>> int Y = bar(4);
>>> printf("%d\n", Y);
>>> }
>> The body for "bar" is gone by the time llvm_asm_file_end is called: I
>> think the tree inliner is still being used???
> 
> Yeah, looks like it's not disabled fully at -O0 (no-unit-at-a-time  
> seems to be the relevant setting).  A significantly different high- 
> level code path, implemented by having  if(flag_unit_at_a_time)   all  
> over the place in common code....

yes, the cgraph stuff is icky.  I've fixed the problem by having all
functions be classified as uninlineable.

Ciao,

Duncan.


From gohman at apple.com  Mon May 18 10:44:59 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 15:44:59 -0000
Subject: [llvm-commits] [llvm] r72006 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905181544.n4IFixNb018461@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 10:44:58 2009
New Revision: 72006

URL: http://llvm.org/viewvc/llvm-project?rev=72006&view=rev
Log:
Add assertion checks to the SCEV operator creation methods to catch
type mismatches.

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=72006&r1=72005&r2=72006&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 10:44:58 2009
@@ -921,6 +921,12 @@
 SCEVHandle ScalarEvolution::getAddExpr(std::vector<SCEVHandle> &Ops) {
   assert(!Ops.empty() && "Cannot get empty add!");
   if (Ops.size() == 1) return Ops[0];
+#ifndef NDEBUG
+  for (unsigned i = 1, e = Ops.size(); i != e; ++i)
+    assert(getEffectiveSCEVType(Ops[i]->getType()) ==
+           getEffectiveSCEVType(Ops[0]->getType()) &&
+           "SCEVAddExpr operand types don't match!");
+#endif
 
   // Sort by complexity, this groups all similar expression types together.
   GroupByComplexity(Ops, LI);
@@ -1205,6 +1211,12 @@
 
 SCEVHandle ScalarEvolution::getMulExpr(std::vector<SCEVHandle> &Ops) {
   assert(!Ops.empty() && "Cannot get empty mul!");
+#ifndef NDEBUG
+  for (unsigned i = 1, e = Ops.size(); i != e; ++i)
+    assert(getEffectiveSCEVType(Ops[i]->getType()) ==
+           getEffectiveSCEVType(Ops[0]->getType()) &&
+           "SCEVMulExpr operand types don't match!");
+#endif
 
   // Sort by complexity, this groups all similar expression types together.
   GroupByComplexity(Ops, LI);
@@ -1362,6 +1374,10 @@
 
 SCEVHandle ScalarEvolution::getUDivExpr(const SCEVHandle &LHS,
                                         const SCEVHandle &RHS) {
+  assert(getEffectiveSCEVType(LHS->getType()) ==
+         getEffectiveSCEVType(RHS->getType()) &&
+         "SCEVUDivExpr operand types don't match!");
+
   if (const SCEVConstant *RHSC = dyn_cast<SCEVConstant>(RHS)) {
     if (RHSC->getValue()->equalsInt(1))
       return LHS;                            // X udiv 1 --> x
@@ -1466,6 +1482,12 @@
 SCEVHandle ScalarEvolution::getAddRecExpr(std::vector<SCEVHandle> &Operands,
                                           const Loop *L) {
   if (Operands.size() == 1) return Operands[0];
+#ifndef NDEBUG
+  for (unsigned i = 1, e = Operands.size(); i != e; ++i)
+    assert(getEffectiveSCEVType(Operands[i]->getType()) ==
+           getEffectiveSCEVType(Operands[0]->getType()) &&
+           "SCEVAddRecExpr operand types don't match!");
+#endif
 
   if (Operands.back()->isZero()) {
     Operands.pop_back();
@@ -1502,6 +1524,12 @@
 SCEVHandle ScalarEvolution::getSMaxExpr(std::vector<SCEVHandle> Ops) {
   assert(!Ops.empty() && "Cannot get empty smax!");
   if (Ops.size() == 1) return Ops[0];
+#ifndef NDEBUG
+  for (unsigned i = 1, e = Ops.size(); i != e; ++i)
+    assert(getEffectiveSCEVType(Ops[i]->getType()) ==
+           getEffectiveSCEVType(Ops[0]->getType()) &&
+           "SCEVSMaxExpr operand types don't match!");
+#endif
 
   // Sort by complexity, this groups all similar expression types together.
   GroupByComplexity(Ops, LI);
@@ -1582,6 +1610,12 @@
 SCEVHandle ScalarEvolution::getUMaxExpr(std::vector<SCEVHandle> Ops) {
   assert(!Ops.empty() && "Cannot get empty umax!");
   if (Ops.size() == 1) return Ops[0];
+#ifndef NDEBUG
+  for (unsigned i = 1, e = Ops.size(); i != e; ++i)
+    assert(getEffectiveSCEVType(Ops[i]->getType()) ==
+           getEffectiveSCEVType(Ops[0]->getType()) &&
+           "SCEVUMaxExpr operand types don't match!");
+#endif
 
   // Sort by complexity, this groups all similar expression types together.
   GroupByComplexity(Ops, LI);




From gohman at apple.com  Mon May 18 10:58:40 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 15:58:40 -0000
Subject: [llvm-commits] [llvm] r72007 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905181558.n4IFwelE018892@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 10:58:39 2009
New Revision: 72007

URL: http://llvm.org/viewvc/llvm-project?rev=72007&view=rev
Log:
Minor code cleanups. Do more of the work before the if statements
instead of within their controlling expressions.

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=72007&r1=72006&r2=72007&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 10:58:39 2009
@@ -806,8 +806,9 @@
         // the addrec's type. The count is always unsigned.
         SCEVHandle CastedMaxBECount =
           getTruncateOrZeroExtend(MaxBECount, Start->getType());
-        if (MaxBECount ==
-            getTruncateOrZeroExtend(CastedMaxBECount, MaxBECount->getType())) {
+        SCEVHandle RecastedMaxBECount =
+          getTruncateOrZeroExtend(CastedMaxBECount, MaxBECount->getType());
+        if (MaxBECount == RecastedMaxBECount) {
           const Type *WideTy =
             IntegerType::get(getTypeSizeInBits(Start->getType()) * 2);
           // Check whether Start+Step*MaxBECount has no unsigned overflow.
@@ -815,10 +816,11 @@
             getMulExpr(CastedMaxBECount,
                        getTruncateOrZeroExtend(Step, Start->getType()));
           SCEVHandle Add = getAddExpr(Start, ZMul);
-          if (getZeroExtendExpr(Add, WideTy) ==
-              getAddExpr(getZeroExtendExpr(Start, WideTy),
-                         getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
-                                    getZeroExtendExpr(Step, WideTy))))
+          SCEVHandle OperandExtendedAdd =
+            getAddExpr(getZeroExtendExpr(Start, WideTy),
+                       getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
+                                  getZeroExtendExpr(Step, WideTy)));
+          if (getZeroExtendExpr(Add, WideTy) == OperandExtendedAdd)
             // Return the expression with the addrec on the outside.
             return getAddRecExpr(getZeroExtendExpr(Start, Ty),
                                  getZeroExtendExpr(Step, Ty),
@@ -830,10 +832,11 @@
             getMulExpr(CastedMaxBECount,
                        getTruncateOrSignExtend(Step, Start->getType()));
           Add = getAddExpr(Start, SMul);
-          if (getZeroExtendExpr(Add, WideTy) ==
-              getAddExpr(getZeroExtendExpr(Start, WideTy),
-                         getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
-                                    getSignExtendExpr(Step, WideTy))))
+          OperandExtendedAdd =
+            getAddExpr(getZeroExtendExpr(Start, WideTy),
+                       getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
+                                  getSignExtendExpr(Step, WideTy)));
+          if (getZeroExtendExpr(Add, WideTy) == OperandExtendedAdd)
             // Return the expression with the addrec on the outside.
             return getAddRecExpr(getZeroExtendExpr(Start, Ty),
                                  getSignExtendExpr(Step, Ty),
@@ -891,8 +894,9 @@
         // the addrec's type. The count is always unsigned.
         SCEVHandle CastedMaxBECount =
           getTruncateOrZeroExtend(MaxBECount, Start->getType());
-        if (MaxBECount ==
-            getTruncateOrZeroExtend(CastedMaxBECount, MaxBECount->getType())) {
+        SCEVHandle RecastedMaxBECount =
+          getTruncateOrZeroExtend(CastedMaxBECount, MaxBECount->getType());
+        if (MaxBECount == RecastedMaxBECount) {
           const Type *WideTy =
             IntegerType::get(getTypeSizeInBits(Start->getType()) * 2);
           // Check whether Start+Step*MaxBECount has no signed overflow.
@@ -900,10 +904,11 @@
             getMulExpr(CastedMaxBECount,
                        getTruncateOrSignExtend(Step, Start->getType()));
           SCEVHandle Add = getAddExpr(Start, SMul);
-          if (getSignExtendExpr(Add, WideTy) ==
-              getAddExpr(getSignExtendExpr(Start, WideTy),
-                         getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
-                                    getSignExtendExpr(Step, WideTy))))
+          SCEVHandle OperandExtendedAdd =
+            getAddExpr(getSignExtendExpr(Start, WideTy),
+                       getMulExpr(getZeroExtendExpr(CastedMaxBECount, WideTy),
+                                  getSignExtendExpr(Step, WideTy)));
+          if (getSignExtendExpr(Add, WideTy) == OperandExtendedAdd)
             // Return the expression with the addrec on the outside.
             return getAddRecExpr(getSignExtendExpr(Start, Ty),
                                  getSignExtendExpr(Step, Ty),




From gohman at apple.com  Mon May 18 11:03:58 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 16:03:58 -0000
Subject: [llvm-commits] [llvm] r72008 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905181603.n4IG3wVh019068@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 11:03:58 2009
New Revision: 72008

URL: http://llvm.org/viewvc/llvm-project?rev=72008&view=rev
Log:
Fix ScalarEvolution::isLoopGuardedByCond to accept a null Loop*, for
consistency with other routines that use a null Loop* to mean code
not contained by any loop.

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=72008&r1=72007&r2=72008&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 11:03:58 2009
@@ -3328,6 +3328,10 @@
 bool ScalarEvolution::isLoopGuardedByCond(const Loop *L,
                                           ICmpInst::Predicate Pred,
                                           const SCEV *LHS, const SCEV *RHS) {
+  // Interpret a null as meaning no loop, where there is obviously no guard
+  // (interprocedural conditions notwithstanding).
+  if (!L) return false;
+
   BasicBlock *Predecessor = getLoopPredecessor(L);
   BasicBlock *PredecessorDest = L->getHeader();
 




From gohman at apple.com  Mon May 18 11:17:45 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 16:17:45 -0000
Subject: [llvm-commits] [llvm] r72009 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905181617.n4IGHjHP019521@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 11:17:44 2009
New Revision: 72009

URL: http://llvm.org/viewvc/llvm-project?rev=72009&view=rev
Log:
Delete a redundant 'else'.

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=72009&r1=72008&r2=72009&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 11:17:44 2009
@@ -2201,7 +2201,7 @@
                           getSCEV(U->getOperand(1)));
 
       // If the RHS of xor is -1, then this is a not operation.
-      else if (CI->isAllOnesValue())
+      if (CI->isAllOnesValue())
         return getNotSCEV(getSCEV(U->getOperand(0)));
     }
     break;




From gohman at apple.com  Mon May 18 11:29:05 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 16:29:05 -0000
Subject: [llvm-commits] [llvm] r72010 - in /llvm/trunk:
 lib/Analysis/ScalarEvolution.cpp test/Analysis/ScalarEvolution/and-xor.ll
Message-ID: <200905181629.n4IGT5dh019885@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 11:29:04 2009
New Revision: 72010

URL: http://llvm.org/viewvc/llvm-project?rev=72010&view=rev
Log:
Teach ScalarEvolution to recognize x^-1 in the case where non-demanded
bits have been stripped out by instcombine.

Added:
    llvm/trunk/test/Analysis/ScalarEvolution/and-xor.ll
Modified:
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72010&r1=72009&r2=72010&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 11:29:04 2009
@@ -2203,6 +2203,19 @@
       // If the RHS of xor is -1, then this is a not operation.
       if (CI->isAllOnesValue())
         return getNotSCEV(getSCEV(U->getOperand(0)));
+
+      // Model xor(and(x, C), C) as and(~x, C), if C is a low-bits mask.
+      // This is a variant of the check for xor with -1, and it handles
+      // the case where instcombine has trimmed non-demanded bits out
+      // of an xor with -1.
+      if (BinaryOperator *BO = dyn_cast<BinaryOperator>(U->getOperand(0)))
+        if (ConstantInt *LCI = dyn_cast<ConstantInt>(BO->getOperand(1)))
+          if (BO->getOpcode() == Instruction::And &&
+              LCI->getValue() == CI->getValue())
+            if (const SCEVZeroExtendExpr *Z =
+                  dyn_cast<SCEVZeroExtendExpr>(getSCEV(U->getOperand(0))))
+              return getZeroExtendExpr(getNotSCEV(Z->getOperand()),
+                                       U->getType());
     }
     break;
 

Added: llvm/trunk/test/Analysis/ScalarEvolution/and-xor.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Analysis/ScalarEvolution/and-xor.ll?rev=72010&view=auto

==============================================================================
--- llvm/trunk/test/Analysis/ScalarEvolution/and-xor.ll (added)
+++ llvm/trunk/test/Analysis/ScalarEvolution/and-xor.ll Mon May 18 11:29:04 2009
@@ -0,0 +1,8 @@
+; RUN: llvm-as < %s | opt -scalar-evolution -analyze -disable-output \
+; RUN:   | grep {\\-->  (zext} | count 2
+
+define i32 @foo(i32 %x) {
+  %n = and i32 %x, 255
+  %y = xor i32 %n, 255
+  ret i32 %y
+}




From daniel at zuster.org  Mon May 18 11:43:04 2009
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 18 May 2009 16:43:04 -0000
Subject: [llvm-commits] [llvm] r72011 - in /llvm/trunk/lib:
 Analysis/ScalarEvolution.cpp CodeGen/SelectionDAG/DAGCombiner.cpp
 CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp
Message-ID: <200905181643.n4IGh4ii020305@zion.cs.uiuc.edu>

Author: ddunbar
Date: Mon May 18 11:43:04 2009
New Revision: 72011

URL: http://llvm.org/viewvc/llvm-project?rev=72011&view=rev
Log:
Silence Release-Asserts warnings.

Modified:
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72011&r1=72010&r2=72011&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 11:43:04 2009
@@ -3081,6 +3081,7 @@
   }
 
   assert(0 && "Unknown SCEV type!");
+  return 0;
 }
 
 /// getSCEVAtScope - This is a convenience function which does

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=72011&r1=72010&r2=72011&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Mon May 18 11:43:04 2009
@@ -6000,9 +6000,9 @@
 
   // Get alias information for node.
   SDValue Ptr;
-  int64_t Size;
-  const Value *SrcValue;
-  int SrcValueOffset;
+  int64_t Size = 0;
+  const Value *SrcValue = 0;
+  int SrcValueOffset = 0;
   bool IsLoad = FindAliasInfo(N, Ptr, Size, SrcValue, SrcValueOffset);
 
   // Starting off.
@@ -6028,9 +6028,9 @@
     case ISD::STORE: {
       // Get alias information for Chain.
       SDValue OpPtr;
-      int64_t OpSize;
-      const Value *OpSrcValue;
-      int OpSrcValueOffset;
+      int64_t OpSize = 0;
+      const Value *OpSrcValue = 0;
+      int OpSrcValueOffset = 0;
       bool IsOpLoad = FindAliasInfo(Chain.getNode(), OpPtr, OpSize,
                                     OpSrcValue, OpSrcValueOffset);
 

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp?rev=72011&r1=72010&r2=72011&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodesEmit.cpp Mon May 18 11:43:04 2009
@@ -119,6 +119,7 @@
                                      DstRC, SrcRC);
 
     assert(Emitted && "Unable to issue a copy instruction!\n");
+    (void) Emitted;
   }
 
   SDValue Op(Node, ResNo);
@@ -254,6 +255,7 @@
       bool Emitted = TII->copyRegToReg(*BB, InsertPos, NewVReg, VReg,
                                        DstRC, SrcRC);
       assert(Emitted && "Unable to issue a copy instruction!\n");
+      (void) Emitted;
       VReg = NewVReg;
     }
   }
@@ -445,6 +447,7 @@
                                    DstRC, SrcRC);
   assert(Emitted &&
          "Unable to issue a copy instruction for a COPY_TO_REGCLASS node!\n");
+  (void) Emitted;
 
   SDValue Op(Node, 0);
   bool isNew = VRBaseMap.insert(std::make_pair(Op, NewVReg)).second;
@@ -568,6 +571,7 @@
     bool Emitted = TII->copyRegToReg(*BB, InsertPos, DestReg, SrcReg,
                                      DstTRC, SrcTRC);
     assert(Emitted && "Unable to issue a copy instruction!\n");
+    (void) Emitted;
     break;
   }
   case ISD::CopyFromReg: {




From gohman at apple.com  Mon May 18 11:45:29 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 16:45:29 -0000
Subject: [llvm-commits] [llvm] r72012 -
	/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
Message-ID: <200905181645.n4IGjT2o020397@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 11:45:28 2009
New Revision: 72012

URL: http://llvm.org/viewvc/llvm-project?rev=72012&view=rev
Log:
Rename UseTy to AccessTy, for consistency with getAccessType, and to
avoid ambiguity with the word "use" in IVStrideUse.

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

Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=72012&r1=72011&r2=72012&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Mon May 18 11:45:28 2009
@@ -291,9 +291,9 @@
 
 /// getAccessType - Return the type of the memory being accessed.
 static const Type *getAccessType(const Instruction *Inst) {
-  const Type *UseTy = Inst->getType();
+  const Type *AccessTy = Inst->getType();
   if (const StoreInst *SI = dyn_cast<StoreInst>(Inst))
-    UseTy = SI->getOperand(0)->getType();
+    AccessTy = SI->getOperand(0)->getType();
   else if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
     // Addressing modes can also be folded into prefetches and a variety
     // of intrinsics.
@@ -303,11 +303,11 @@
     case Intrinsic::x86_sse2_storeu_pd:
     case Intrinsic::x86_sse2_storeu_dq:
     case Intrinsic::x86_sse2_storel_dq:
-      UseTy = II->getOperand(1)->getType();
+      AccessTy = II->getOperand(1)->getType();
       break;
     }
   }
-  return UseTy;
+  return AccessTy;
 }
 
 namespace {
@@ -547,7 +547,7 @@
 
 /// fitsInAddressMode - Return true if V can be subsumed within an addressing
 /// mode, and does not need to be put in a register first.
-static bool fitsInAddressMode(const SCEVHandle &V, const Type *UseTy,
+static bool fitsInAddressMode(const SCEVHandle &V, const Type *AccessTy,
                              const TargetLowering *TLI, bool HasBaseReg) {
   if (const SCEVConstant *SC = dyn_cast<SCEVConstant>(V)) {
     int64_t VC = SC->getValue()->getSExtValue();
@@ -555,7 +555,7 @@
       TargetLowering::AddrMode AM;
       AM.BaseOffs = VC;
       AM.HasBaseReg = HasBaseReg;
-      return TLI->isLegalAddressingMode(AM, UseTy);
+      return TLI->isLegalAddressingMode(AM, AccessTy);
     } else {
       // Defaults to PPC. PPC allows a sign-extended 16-bit immediate field.
       return (VC > -(1 << 16) && VC < (1 << 16)-1);
@@ -568,7 +568,7 @@
         TargetLowering::AddrMode AM;
         AM.BaseGV = GV;
         AM.HasBaseReg = HasBaseReg;
-        return TLI->isLegalAddressingMode(AM, UseTy);
+        return TLI->isLegalAddressingMode(AM, AccessTy);
       } else {
         // Default: assume global addresses are not legal.
       }
@@ -620,7 +620,7 @@
 /// that can fit into the immediate field of instructions in the target.
 /// Accumulate these immediate values into the Imm value.
 static void MoveImmediateValues(const TargetLowering *TLI,
-                                const Type *UseTy,
+                                const Type *AccessTy,
                                 SCEVHandle &Val, SCEVHandle &Imm,
                                 bool isAddress, Loop *L,
                                 ScalarEvolution *SE) {
@@ -630,7 +630,7 @@
     
     for (unsigned i = 0; i != SAE->getNumOperands(); ++i) {
       SCEVHandle NewOp = SAE->getOperand(i);
-      MoveImmediateValues(TLI, UseTy, NewOp, Imm, isAddress, L, SE);
+      MoveImmediateValues(TLI, AccessTy, NewOp, Imm, isAddress, L, SE);
       
       if (!NewOp->isLoopInvariant(L)) {
         // If this is a loop-variant expression, it must stay in the immediate
@@ -649,7 +649,7 @@
   } else if (const SCEVAddRecExpr *SARE = dyn_cast<SCEVAddRecExpr>(Val)) {
     // Try to pull immediates out of the start value of nested addrec's.
     SCEVHandle Start = SARE->getStart();
-    MoveImmediateValues(TLI, UseTy, Start, Imm, isAddress, L, SE);
+    MoveImmediateValues(TLI, AccessTy, Start, Imm, isAddress, L, SE);
     
     if (Start != SARE->getStart()) {
       std::vector<SCEVHandle> Ops(SARE->op_begin(), SARE->op_end());
@@ -659,12 +659,13 @@
     return;
   } else if (const SCEVMulExpr *SME = dyn_cast<SCEVMulExpr>(Val)) {
     // Transform "8 * (4 + v)" -> "32 + 8*V" if "32" fits in the immed field.
-    if (isAddress && fitsInAddressMode(SME->getOperand(0), UseTy, TLI, false) &&
+    if (isAddress &&
+        fitsInAddressMode(SME->getOperand(0), AccessTy, TLI, false) &&
         SME->getNumOperands() == 2 && SME->isLoopInvariant(L)) {
 
       SCEVHandle SubImm = SE->getIntegerSCEV(0, Val->getType());
       SCEVHandle NewOp = SME->getOperand(1);
-      MoveImmediateValues(TLI, UseTy, NewOp, SubImm, isAddress, L, SE);
+      MoveImmediateValues(TLI, AccessTy, NewOp, SubImm, isAddress, L, SE);
       
       // If we extracted something out of the subexpressions, see if we can 
       // simplify this!
@@ -672,7 +673,7 @@
         // Scale SubImm up by "8".  If the result is a target constant, we are
         // good.
         SubImm = SE->getMulExpr(SubImm, SME->getOperand(0));
-        if (fitsInAddressMode(SubImm, UseTy, TLI, false)) {
+        if (fitsInAddressMode(SubImm, AccessTy, TLI, false)) {
           // Accumulate the immediate.
           Imm = SE->getAddExpr(Imm, SubImm);
           
@@ -686,7 +687,7 @@
 
   // Loop-variant expressions must stay in the immediate field of the
   // expression.
-  if ((isAddress && fitsInAddressMode(Val, UseTy, TLI, false)) ||
+  if ((isAddress && fitsInAddressMode(Val, AccessTy, TLI, false)) ||
       !Val->isLoopInvariant(L)) {
     Imm = SE->getAddExpr(Imm, Val);
     Val = SE->getIntegerSCEV(0, Val->getType());
@@ -701,8 +702,8 @@
                                 SCEVHandle &Val, SCEVHandle &Imm,
                                 bool isAddress, Loop *L,
                                 ScalarEvolution *SE) {
-  const Type *UseTy = getAccessType(User);
-  MoveImmediateValues(TLI, UseTy, Val, Imm, isAddress, L, SE);
+  const Type *AccessTy = getAccessType(User);
+  MoveImmediateValues(TLI, AccessTy, Val, Imm, isAddress, L, SE);
 }
 
 /// SeparateSubExprs - Decompose Expr into all of the subexpressions that are
@@ -796,11 +797,11 @@
     // If this use is as an address we may be able to put CSEs in the addressing
     // mode rather than hoisting them.
     bool isAddrUse = isAddressUse(Uses[i].Inst, Uses[i].OperandValToReplace);
-    // We may need the UseTy below, but only when isAddrUse, so compute it
+    // We may need the AccessTy below, but only when isAddrUse, so compute it
     // only in that case.
-    const Type *UseTy = 0;
+    const Type *AccessTy = 0;
     if (isAddrUse)
-      UseTy = getAccessType(Uses[i].Inst);
+      AccessTy = getAccessType(Uses[i].Inst);
 
     // Split the expression into subexprs.
     SeparateSubExprs(SubExprs, Uses[i].Base, SE);
@@ -811,7 +812,7 @@
     for (unsigned j = 0, e = SubExprs.size(); j != e; ++j) {
       if (++SubExpressionUseData[SubExprs[j]].Count == 1)
         UniqueSubExprs.push_back(SubExprs[j]);
-      if (!isAddrUse || !fitsInAddressMode(SubExprs[j], UseTy, TLI, false))
+      if (!isAddrUse || !fitsInAddressMode(SubExprs[j], AccessTy, TLI, false))
         SubExpressionUseData[SubExprs[j]].notAllUsesAreFree = true;
     }
     SubExprs.clear();
@@ -845,8 +846,8 @@
         continue;
       // We know this is an addressing mode use; if there are any uses that
       // are not, FreeResult would be Zero.
-      const Type *UseTy = getAccessType(Uses[i].Inst);
-      if (!fitsInAddressMode(FreeResult, UseTy, TLI, Result!=Zero)) {
+      const Type *AccessTy = getAccessType(Uses[i].Inst);
+      if (!fitsInAddressMode(FreeResult, AccessTy, TLI, Result!=Zero)) {
         // FIXME:  could split up FreeResult into pieces here, some hoisted
         // and some not.  There is no obvious advantage to this.
         Result = SE->getAddExpr(Result, FreeResult);
@@ -1258,11 +1259,11 @@
         if (!CurImm) CurImm = SE->getIntegerSCEV(0, Stride->getType());
         if (!Imm)       Imm = SE->getIntegerSCEV(0, Stride->getType());
         const Instruction *Inst = UsersToProcess[i].Inst;
-        const Type *UseTy = getAccessType(Inst);
+        const Type *AccessTy = getAccessType(Inst);
         SCEVHandle Diff = SE->getMinusSCEV(UsersToProcess[i].Imm, Imm);
         if (!Diff->isZero() &&
             (!AllUsesAreAddresses ||
-             !fitsInAddressMode(Diff, UseTy, TLI, /*HasBaseReg=*/true)))
+             !fitsInAddressMode(Diff, AccessTy, TLI, /*HasBaseReg=*/true)))
           return false;
       }
     } while (++i != e && Base == UsersToProcess[i].Base);




From dgregor at apple.com  Mon May 18 12:21:34 2009
From: dgregor at apple.com (Douglas Gregor)
Date: Mon, 18 May 2009 17:21:34 -0000
Subject: [llvm-commits] [llvm] r72018 - in /llvm/trunk:
 autoconf/configure.ac cmake/config-ix.cmake configure
 include/llvm/Config/config.h.cmake include/llvm/Config/config.h.in
 lib/System/Unix/Process.inc
Message-ID: <200905181721.n4IHLZbj021853@zion.cs.uiuc.edu>

Author: dgregor
Date: Mon May 18 12:21:34 2009
New Revision: 72018

URL: http://llvm.org/viewvc/llvm-project?rev=72018&view=rev
Log:
termios.h contains the winsize structure we need to determine the
width of a terminal. Don't try to get the width of a terminal if we
don't have this header.

Modified:
    llvm/trunk/autoconf/configure.ac
    llvm/trunk/cmake/config-ix.cmake
    llvm/trunk/configure
    llvm/trunk/include/llvm/Config/config.h.cmake
    llvm/trunk/include/llvm/Config/config.h.in
    llvm/trunk/lib/System/Unix/Process.inc

Modified: llvm/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/autoconf/configure.ac?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/autoconf/configure.ac (original)
+++ llvm/trunk/autoconf/configure.ac Mon May 18 12:21:34 2009
@@ -823,8 +823,8 @@
 AC_HEADER_TIME
 
 AC_CHECK_HEADERS([dlfcn.h execinfo.h fcntl.h inttypes.h limits.h link.h])
-AC_CHECK_HEADERS([malloc.h setjmp.h signal.h stdint.h unistd.h utime.h])
-AC_CHECK_HEADERS([windows.h])
+AC_CHECK_HEADERS([malloc.h setjmp.h signal.h stdint.h termios.h unistd.h])
+AC_CHECK_HEADERS([utime.h windows.h])
 AC_CHECK_HEADERS([sys/mman.h sys/param.h sys/resource.h sys/time.h])
 AC_CHECK_HEADERS([sys/types.h sys/ioctl.h malloc/malloc.h mach/mach.h])
 if test "$ENABLE_THREADS" -eq 1 ; then

Modified: llvm/trunk/cmake/config-ix.cmake
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/cmake/config-ix.cmake?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/cmake/config-ix.cmake (original)
+++ llvm/trunk/cmake/config-ix.cmake Mon May 18 12:21:34 2009
@@ -34,6 +34,7 @@
 check_include_file(sys/stat.h HAVE_SYS_STAT_H)
 check_include_file(sys/time.h HAVE_SYS_TIME_H)
 check_include_file(sys/types.h HAVE_SYS_TYPES_H)
+check_include_file(termios.h HAVE_TERMIOS_H)
 check_include_file(unistd.h HAVE_UNISTD_H)
 check_include_file(utime.h HAVE_UTIME_H)
 check_include_file(windows.h HAVE_WINDOWS_H)

Modified: llvm/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/configure?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/configure (original)
+++ llvm/trunk/configure Mon May 18 12:21:34 2009
@@ -28963,7 +28963,7 @@
 
 
 
-for ac_header in malloc.h setjmp.h signal.h stdint.h unistd.h utime.h
+for ac_header in malloc.h setjmp.h signal.h stdint.h termios.h unistd.h
 do
 as_ac_Header=`echo "ac_cv_header_$ac_header" | $as_tr_sh`
 if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
@@ -29132,7 +29132,8 @@
 done
 
 
-for ac_header in windows.h
+
+for ac_header in utime.h windows.h
 do
 as_ac_Header=`echo "ac_cv_header_$ac_header" | $as_tr_sh`
 if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then

Modified: llvm/trunk/include/llvm/Config/config.h.cmake
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Config/config.h.cmake?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Config/config.h.cmake (original)
+++ llvm/trunk/include/llvm/Config/config.h.cmake Mon May 18 12:21:34 2009
@@ -419,6 +419,9 @@
 /* Define to 1 if the system has the type `uint64_t'. */
 #undef HAVE_UINT64_T
 
+/* Define to 1 if you have the <termios.h> header file. */
+#cmakedefine HAVE_TERMIOS_H ${HAVE_TERMIOS_H}
+
 /* Define to 1 if you have the <unistd.h> header file. */
 #cmakedefine HAVE_UNISTD_H ${HAVE_UNISTD_H}
 

Modified: llvm/trunk/include/llvm/Config/config.h.in
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Config/config.h.in?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Config/config.h.in (original)
+++ llvm/trunk/include/llvm/Config/config.h.in Mon May 18 12:21:34 2009
@@ -419,6 +419,9 @@
 /* Define to 1 if you have <sys/wait.h> that is POSIX.1 compatible. */
 #undef HAVE_SYS_WAIT_H
 
+/* Define to 1 if you have the <termios.h> header file. */
+#undef HAVE_TERMIOS_H
+
 /* Define to 1 if the system has the type `uint64_t'. */
 #undef HAVE_UINT64_T
 

Modified: llvm/trunk/lib/System/Unix/Process.inc
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Unix/Process.inc?rev=72018&r1=72017&r2=72018&view=diff

==============================================================================
--- llvm/trunk/lib/System/Unix/Process.inc (original)
+++ llvm/trunk/lib/System/Unix/Process.inc Mon May 18 12:21:34 2009
@@ -27,6 +27,9 @@
 #ifdef HAVE_SYS_IOCTL_H
 #  include <sys/ioctl.h>
 #endif
+#ifdef HAVE_TERMIOS_H
+#  include <termios.h>
+#endif
 
 //===----------------------------------------------------------------------===//
 //=== WARNING: Implementation here must contain only generic UNIX code that
@@ -204,7 +207,7 @@
 
   unsigned Columns = 0;
 
-#ifdef HAVE_SYS_IOCTL_H
+#if defined(HAVE_SYS_IOCTL_H) && defined(HAVE_TERMIOS_H)
   // Try to determine the width of the terminal.
   struct winsize ws;
   if (ioctl(FileID, TIOCGWINSZ, &ws) == 0)




From evan at fallingsnow.net  Mon May 18 12:31:25 2009
From: evan at fallingsnow.net (Evan Phoenix)
Date: Mon, 18 May 2009 10:31:25 -0700
Subject: [llvm-commits] [PATCH] Generate info about the JIT process
In-Reply-To: <0A9506FE-111E-4D3C-BC04-D962645AA1DE@apple.com>
References: <978B1AEF-4B6B-4BA1-A73F-6FEF5C95A12D@fallingsnow.net>
	<4A0A6C85.1030909@gmail.com>
	<E84E4E33-642E-4A2F-A645-D6CFF6569E3E@fallingsnow.net>
	<4A0A90FB.3000907@gmail.com>
	<0A9506FE-111E-4D3C-BC04-D962645AA1DE@apple.com>
Message-ID: <0F735E86-C6C7-4352-8262-B394C70A5D26@fallingsnow.net>

I haven't heard anything from Argiris, or any other thoughts on what  
the JITListener would look like. I hesitate to write such a thing  
without an idea of what people want, so I'm hoping we can commit my  
patch as is.

When there is more discussion about the JITListener, it will be easy  
to make MachineCodeInfo sit on top of it without any external API  
changes.

I've attached an updated version of the patch against SVN HEAD.

  - Evan

-------------- next part --------------
A non-text attachment was scrubbed...
Name: jit-info.diff
Type: application/octet-stream
Size: 6510 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/36f9b50d/attachment.obj 
-------------- next part --------------


On May 13, 2009, at 10:21 AM, Chris Lattner wrote:

> On May 13, 2009, at 2:20 AM, Argiris Kirtzidis wrote:
>>> Thats basically what MachineCodeInfo enables. ATM, there are only 2
>>> things it publishes directly to the MCI, but future APIs can  
>>> easily be
>>> added using virtual methods to allow for subclasses of  
>>> MachineCodeInfo
>>> to do new things.
>>>
>>> This is far from a final set of things MCI will do, it's primarily a
>>> way of introduce an API that we can add future things to.
>>
>> Ok, that sounds great. It's just that MachineCodeInfo is restricted  
>> to
>> the "function level", how about "module level" info, specifically
>> getting notifications during code execution like:
>> -"I lazy JITed this function"
>> -"I allocated this memory chunk for this global variable"
>>
>> Of course, MachineCodeInfo could be the one for the "function  
>> level" and
>> something like "JITListener" at the module level. The JITListener  
>> could
>> have a method that passes a MachineCodeInfo before function JITing.  
>> IMO
>> this is flexible and clean.
>
> I agree with Argiris: there should be a simple abstract interface  
> with methods like the above.  The JIT should have an optional  
> pointer to it.  If non-null, it would invoke the methods on the  
> JITListener.  This would allow lots of detailed instrumentation, but  
> would introduce no overhead if it's not being used.
>
> -Chris


From dalej at apple.com  Mon May 18 12:44:15 2009
From: dalej at apple.com (Dale Johannesen)
Date: Mon, 18 May 2009 17:44:15 -0000
Subject: [llvm-commits] [llvm] r72022 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
Message-ID: <200905181744.n4IHiFPc022776@zion.cs.uiuc.edu>

Author: johannes
Date: Mon May 18 12:44:15 2009
New Revision: 72022

URL: http://llvm.org/viewvc/llvm-project?rev=72022&view=rev
Log:
Fill in the missing patterns for ADC and SBB.
Some comment cleanup.


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=72022&r1=72021&r2=72022&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon May 18 12:44:15 2009
@@ -34,7 +34,6 @@
                                             [SDTCisSameAs<0, 1>,
                                              SDTCisSameAs<0, 2>,
                                              SDTCisInt<0>]>;
-
 def SDTX86BrCond  : SDTypeProfile<0, 3,
                                   [SDTCisVT<0, OtherVT>,
                                    SDTCisVT<1, i8>, SDTCisVT<2, i32>]>;
@@ -2275,24 +2274,69 @@
 
 let Uses = [EFLAGS] in {
 let isCommutable = 1 in {  // X = ADC Y, Z --> X = ADC Z, Y
-def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst), (ins GR32:$src1, GR32:$src2),
+def ADC8rr   : I<0x10, MRMDestReg, (outs GR8:$dst), (ins GR8:$src1, GR8:$src2),
+                 "adc{b}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR8:$dst, (adde GR8:$src1, GR8:$src2))]>;
+def ADC16rr  : I<0x11, MRMDestReg, (outs GR16:$dst),
+                                   (ins GR16:$src1, GR16:$src2),
+                 "adc{w}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>;
+def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst),
+                                   (ins GR32:$src1, GR32:$src2),
                  "adc{l}\t{$src2, $dst|$dst, $src2}",
                  [(set GR32:$dst, (adde GR32:$src1, GR32:$src2))]>;
 }
-def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst), (ins GR32:$src1, i32mem:$src2),
+def ADC8rm   : I<0x12, MRMSrcMem , (outs GR8:$dst), 
+                                   (ins GR8:$src1, i8mem:$src2),
+                 "adc{b}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR8:$dst, (adde GR8:$src1, (load addr:$src2)))]>;
+def ADC16rm  : I<0x13, MRMSrcMem , (outs GR16:$dst),
+                                   (ins GR16:$src1, i16mem:$src2),
+                 "adc{w}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR16:$dst, (adde GR16:$src1, (load addr:$src2)))]>;
+def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst),
+                                   (ins GR32:$src1, i32mem:$src2),
                  "adc{l}\t{$src2, $dst|$dst, $src2}",
                  [(set GR32:$dst, (adde GR32:$src1, (load addr:$src2)))]>;
-def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst), (ins GR32:$src1, i32imm:$src2),
+def ADC8ri   : Ii8<0x80, MRM2r, (outs GR8:$dst), (ins GR8:$src1, i8imm:$src2),
+                    "adc{b}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR8:$dst, (adde GR8:$src1, imm:$src2))]>;
+def ADC16ri  : Ii16<0x81, MRM2r, (outs GR16:$dst),
+                                 (ins GR16:$src1, i16imm:$src2),
+                    "adc{w}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>;
+def ADC16ri8 : Ii8<0x83, MRM2r, (outs GR16:$dst),
+                                (ins GR16:$src1, i16i8imm:$src2),
+                   "adc{w}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR16:$dst, (adde GR16:$src1, i16immSExt8:$src2))]>;
+def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst),
+                                 (ins GR32:$src1, i32imm:$src2),
                     "adc{l}\t{$src2, $dst|$dst, $src2}",
                  [(set GR32:$dst, (adde GR32:$src1, imm:$src2))]>;
-def ADC32ri8 : Ii8<0x83, MRM2r, (outs GR32:$dst), (ins GR32:$src1, i32i8imm:$src2),
+def ADC32ri8 : Ii8<0x83, MRM2r, (outs GR32:$dst),
+                                (ins GR32:$src1, i32i8imm:$src2),
                    "adc{l}\t{$src2, $dst|$dst, $src2}",
                  [(set GR32:$dst, (adde GR32:$src1, i32immSExt8:$src2))]>;
 
 let isTwoAddress = 0 in {
+  def ADC8mr   : I<0x10, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src2),
+                   "adc{b}\t{$src2, $dst|$dst, $src2}",
+                   [(store (adde (load addr:$dst), GR8:$src2), addr:$dst)]>;
+  def ADC16mr  : I<0x11, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2),
+                   "adc{w}\t{$src2, $dst|$dst, $src2}",
+                   [(store (adde (load addr:$dst), GR16:$src2), addr:$dst)]>;
   def ADC32mr  : I<0x11, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2),
                    "adc{l}\t{$src2, $dst|$dst, $src2}",
                    [(store (adde (load addr:$dst), GR32:$src2), addr:$dst)]>;
+  def ADC8mi   : Ii8<0x80, MRM2m, (outs), (ins i8mem:$dst, i8imm:$src2),
+                      "adc{b}\t{$src2, $dst|$dst, $src2}",
+                  [(store (adde (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
+  def ADC16mi  : Ii16<0x81, MRM2m, (outs), (ins i16mem:$dst, i16imm:$src2),
+                      "adc{w}\t{$src2, $dst|$dst, $src2}",
+                  [(store (adde (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
+  def ADC16mi8 : Ii8<0x83, MRM2m, (outs), (ins i16mem:$dst, i16i8imm :$src2),
+                     "adc{w}\t{$src2, $dst|$dst, $src2}",
+               [(store (adde (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
   def ADC32mi  : Ii32<0x81, MRM2m, (outs), (ins i32mem:$dst, i32imm:$src2),
                       "adc{l}\t{$src2, $dst|$dst, $src2}",
                   [(store (adde (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
@@ -2401,17 +2445,38 @@
 }
 
 let Uses = [EFLAGS] in {
-def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst), (ins GR32:$src1, GR32:$src2),
+def SBB8rr     : I<0x18, MRMDestReg, (outs GR8:$dst),
+                                     (ins GR8:$src1, GR8:$src2),
+                  "sbb{b}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR8:$dst, (sube GR8:$src1, GR8:$src2))]>;
+def SBB16rr    : I<0x19, MRMDestReg, (outs GR16:$dst),
+                                     (ins GR16:$src1, GR16:$src2),
+                  "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>;
+def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst),
+                                      (ins GR32:$src1, GR32:$src2),
                   "sbb{l}\t{$src2, $dst|$dst, $src2}",
                  [(set GR32:$dst, (sube GR32:$src1, GR32:$src2))]>;
 
 let isTwoAddress = 0 in {
+  def SBB8mr   : I<0x18, MRMDestMem, (outs), (ins i8mem:$dst, GR8:$src2), 
+                   "sbb{b}\t{$src2, $dst|$dst, $src2}",
+                   [(store (sube (load addr:$dst), GR8:$src2), addr:$dst)]>;
+  def SBB16mr  : I<0x19, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), 
+                   "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                   [(store (sube (load addr:$dst), GR16:$src2), addr:$dst)]>;
   def SBB32mr  : I<0x19, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), 
                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
                    [(store (sube (load addr:$dst), GR32:$src2), addr:$dst)]>;
   def SBB8mi  : Ii32<0x80, MRM3m, (outs), (ins i8mem:$dst, i8imm:$src2), 
                       "sbb{b}\t{$src2, $dst|$dst, $src2}",
                    [(store (sube (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
+  def SBB16mi  : Ii16<0x81, MRM3m, (outs), (ins i16mem:$dst, i16imm:$src2), 
+                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                  [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
+  def SBB16mi8 : Ii8<0x83, MRM3m, (outs), (ins i16mem:$dst, i16i8imm :$src2), 
+                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
+               [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
   def SBB32mi  : Ii32<0x81, MRM3m, (outs), (ins i32mem:$dst, i32imm:$src2), 
                       "sbb{l}\t{$src2, $dst|$dst, $src2}",
                   [(store (sube (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
@@ -2419,13 +2484,34 @@
                      "sbb{l}\t{$src2, $dst|$dst, $src2}",
                [(store (sube (load addr:$dst), i32immSExt8:$src2), addr:$dst)]>;
 }
-def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst), (ins GR32:$src1, i32mem:$src2),
+def SBB8rm   : I<0x1A, MRMSrcMem, (outs GR8:$dst), (ins GR8:$src1, i8mem:$src2),
+                    "sbb{b}\t{$src2, $dst|$dst, $src2}",
+                    [(set GR8:$dst, (sube GR8:$src1, (load addr:$src2)))]>;
+def SBB16rm  : I<0x1B, MRMSrcMem, (outs GR16:$dst),
+                                  (ins GR16:$src1, i16mem:$src2),
+                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                    [(set GR16:$dst, (sube GR16:$src1, (load addr:$src2)))]>;
+def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst),
+                                  (ins GR32:$src1, i32mem:$src2),
                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
                     [(set GR32:$dst, (sube GR32:$src1, (load addr:$src2)))]>;
-def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst), (ins GR32:$src1, i32imm:$src2),
+def SBB8ri   : Ii8<0x80, MRM3r, (outs GR8:$dst), (ins GR8:$src1, i8imm:$src2),
+                    "sbb{b}\t{$src2, $dst|$dst, $src2}",
+                    [(set GR8:$dst, (sube GR8:$src1, imm:$src2))]>;
+def SBB16ri  : Ii16<0x81, MRM3r, (outs GR16:$dst),
+                                 (ins GR16:$src1, i16imm:$src2),
+                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                    [(set GR16:$dst, (sube GR16:$src1, imm:$src2))]>;
+def SBB16ri8 : Ii8<0x83, MRM3r, (outs GR16:$dst),
+                                (ins GR16:$src1, i16i8imm:$src2),
+                   "sbb{w}\t{$src2, $dst|$dst, $src2}",
+                   [(set GR16:$dst, (sube GR16:$src1, i16immSExt8:$src2))]>;
+def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst), 
+                                 (ins GR32:$src1, i32imm:$src2),
                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
                     [(set GR32:$dst, (sube GR32:$src1, imm:$src2))]>;
-def SBB32ri8 : Ii8<0x83, MRM3r, (outs GR32:$dst), (ins GR32:$src1, i32i8imm:$src2),
+def SBB32ri8 : Ii8<0x83, MRM3r, (outs GR32:$dst),
+                                (ins GR32:$src1, i32i8imm:$src2),
                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
                    [(set GR32:$dst, (sube GR32:$src1, i32immSExt8:$src2))]>;
 } // Uses = [EFLAGS]
@@ -3598,8 +3684,6 @@
 def : Pat<(parallel (X86add_flag GR8:$src1, GR8:$src2),
                     (implicit EFLAGS)),
           (ADD8rr GR8:$src1, GR8:$src2)>;
-
-// Register-Register Addition with EFLAGS result
 def : Pat<(parallel (X86add_flag GR16:$src1, GR16:$src2),
                     (implicit EFLAGS)),
           (ADD16rr GR16:$src1, GR16:$src2)>;
@@ -3622,8 +3706,6 @@
 def : Pat<(parallel (X86add_flag GR8:$src1, imm:$src2),
                     (implicit EFLAGS)),
           (ADD8ri GR8:$src1, imm:$src2)>;
-
-// Register-Integer Addition with EFLAGS result
 def : Pat<(parallel (X86add_flag GR16:$src1, imm:$src2),
                     (implicit EFLAGS)),
           (ADD16ri GR16:$src1, imm:$src2)>;
@@ -3650,6 +3732,8 @@
                            addr:$dst),
                     (implicit EFLAGS)),
           (ADD32mr addr:$dst, GR32:$src2)>;
+
+// Memory-Integer Addition with EFLAGS result
 def : Pat<(parallel (store (X86add_flag (loadi8 addr:$dst), imm:$src2),
                            addr:$dst),
                     (implicit EFLAGS)),
@@ -3836,7 +3920,6 @@
                     (implicit EFLAGS)),
           (DEC32m addr:$dst)>, Requires<[In32BitMode]>;
 
-
 //===----------------------------------------------------------------------===//
 // Floating Point Stack Support
 //===----------------------------------------------------------------------===//




From isanbard at gmail.com  Mon May 18 13:04:58 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 11:04:58 -0700
Subject: [llvm-commits] [llvm] r71986 -
	/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
In-Reply-To: <4A116B17.2030806@free.fr>
References: <200905171937.n4HJb3NQ025904@zion.cs.uiuc.edu>
	<533C8EB1-0A3D-4C41-AFEA-95C91DBF7C8F@gmail.com>
	<4A116B17.2030806@free.fr>
Message-ID: <16e5fdf90905181104q1cbfcd4cyae2bab1e902e1fbd@mail.gmail.com>

On Mon, May 18, 2009 at 7:05 AM, Duncan Sands <baldrick at free.fr> wrote:
> Hi Bill,
>
>> It's very strange, but on a self-hosted release build, I'm getting
>> this error:
>>
>> Running /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/
>> FrontendC/dg.exp ... FAIL:
>> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/FrontendC/
>> 2009-05-17-AlwaysInline.c Failed with exit(1) at line 1 while running:
>> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmgcc42.roots/
>> llvmgcc42~dst/Developer/usr/bin/llvm-gcc-4.2 -m32 -fstrict-aliasing -
>> Wstrict-aliasing -I/usr/include/c++/4.0.0/i686-apple-darwin9 -I/usr/
>> include/c++/4.0.0 -emit-llvm -w -S
>> /Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvmCore/test/FrontendC/
>> 2009-05-17-AlwaysInline.c -O0 -o - -mllvm --disable-llvm-optzns | /usr/
>> bin/grep bar cc1: Unknown command line argument '-fdisable-llvm-
>> optzns'. Try: 'cc1 --help'
>> I'm not sure where the '-fdisable-llvm-optzns' command is coming from...
>
> this is similar to what you get when you are compiling with the system
> gcc and not llvm-gcc, and you pass it -mllvm --disable-llvm-optzns. ?So
> is it possible that it's not llvm-gcc being used here, but standard gcc?
>
I doubt it. The log file shows that it's using "llvm-gcc". This must
be a release-mode deal, as it's failing in release mode for me (and
not just self-hosting). Here's an option. Use "-disable-llvm-optzns"
instead of "--disable-llvm-optzns". It seems that '--' is converted to
'-f' in gcc. I have no clue why:

$ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
cc1: for the -relocation-model=pic option: : may only occur zero or one times!

$ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
cc1: Unknown command line argument '-frelocation-model=pic'.  Try: 'cc1 --help'

-bw



From isanbard at gmail.com  Mon May 18 13:09:37 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 18:09:37 -0000
Subject: [llvm-commits] [llvm] r72023 -
	/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
Message-ID: <200905181809.n4II9bTv023638@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 13:09:36 2009
New Revision: 72023

URL: http://llvm.org/viewvc/llvm-project?rev=72023&view=rev
Log:
Commands beginning with '--' are converted to '-f' by gcc. Blech!

Modified:
    llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c

Modified: llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c?rev=72023&r1=72022&r2=72023&view=diff

==============================================================================
--- llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c (original)
+++ llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c Mon May 18 13:09:36 2009
@@ -1,4 +1,4 @@
-// RUN: %llvmgcc -S %s -O0 -o - -mllvm --disable-llvm-optzns | grep bar
+// RUN: %llvmgcc -S %s -O0 -o - -mllvm -disable-llvm-optzns | grep bar
 // Check that the gcc inliner is turned off.
 
 #include <stdio.h>




From isanbard at gmail.com  Mon May 18 13:17:23 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 18:17:23 -0000
Subject: [llvm-commits] [llvm] r72024 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
Message-ID: <200905181817.n4IIHNvL023916@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 13:17:22 2009
New Revision: 72024

URL: http://llvm.org/viewvc/llvm-project?rev=72024&view=rev
Log:
Don't call RegionInlinedFnEnd if our optimization level isn't -O0.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72024&r1=72023&r2=72024&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Mon May 18 13:17:22 2009
@@ -3929,7 +3929,8 @@
       MachineFunction &MF = DAG.getMachineFunction();
       DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
 
-      if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
+      if (OptLevel == CodeGenOpt::None ||
+          Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
         unsigned LabelID =
           DW->RecordRegionEnd(cast<GlobalVariable>(REI.getContext()));
         DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),




From gohman at apple.com  Mon May 18 13:18:57 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 18:18:57 -0000
Subject: [llvm-commits] [llvm] r72025 -
	/llvm/trunk/lib/VMCore/Instructions.cpp
Message-ID: <200905181818.n4IIIwkf023971@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 13:18:57 2009
New Revision: 72025

URL: http://llvm.org/viewvc/llvm-project?rev=72025&view=rev
Log:
Add assertions to CastInst::getCastOpcode to catch attempted conversions
between integers and pointers when the source type is marked signed,
since inttoptr and ptrtoint always use zero-extension when the destination
is larger than the source.

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

Modified: llvm/trunk/lib/VMCore/Instructions.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Instructions.cpp?rev=72025&r1=72024&r2=72025&view=diff

==============================================================================
--- llvm/trunk/lib/VMCore/Instructions.cpp (original)
+++ llvm/trunk/lib/VMCore/Instructions.cpp Mon May 18 13:18:57 2009
@@ -2220,6 +2220,8 @@
     } else {
       assert(isa<PointerType>(SrcTy) &&
              "Casting from a value that is not first-class type");
+      assert(!SrcIsSigned &&
+             "Pointer types cannot be considered signed for conversions!");
       return PtrToInt;                              // ptr -> int
     }
   } else if (DestTy->isFloatingPoint()) {           // Casting to floating pt
@@ -2259,6 +2261,7 @@
     if (isa<PointerType>(SrcTy)) {
       return BitCast;                               // ptr -> ptr
     } else if (SrcTy->isInteger()) {
+      assert(!SrcIsSigned && "Cannot cast signed integer to pointer!");
       return IntToPtr;                              // int -> ptr
     } else {
       assert(!"Casting pointer to other than pointer or int");




From isanbard at gmail.com  Mon May 18 13:21:03 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 18:21:03 -0000
Subject: [llvm-commits] [llvm] r72026 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
Message-ID: <200905181821.n4IIL3sW024083@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 13:21:03 2009
New Revision: 72026

URL: http://llvm.org/viewvc/llvm-project?rev=72026&view=rev
Log:
Revert last commit. It was wrong.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72026&r1=72025&r2=72026&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Mon May 18 13:21:03 2009
@@ -3929,8 +3929,7 @@
       MachineFunction &MF = DAG.getMachineFunction();
       DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
 
-      if (OptLevel == CodeGenOpt::None ||
-          Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
+      if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
         unsigned LabelID =
           DW->RecordRegionEnd(cast<GlobalVariable>(REI.getContext()));
         DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),




From echristo at apple.com  Mon May 18 13:46:57 2009
From: echristo at apple.com (Eric Christopher)
Date: Mon, 18 May 2009 18:46:57 -0000
Subject: [llvm-commits] [llvm] r72028 -
	/llvm/trunk/include/llvm/Instructions.h
Message-ID: <200905181846.n4IIkvhx025117@zion.cs.uiuc.edu>

Author: echristo
Date: Mon May 18 13:46:57 2009
New Revision: 72028

URL: http://llvm.org/viewvc/llvm-project?rev=72028&view=rev
Log:
Remove getType() overrides for ExtractValueInst and InsertValueInst.

Patch by John McCall.


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=72028&r1=72027&r2=72028&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Instructions.h (original)
+++ llvm/trunk/include/llvm/Instructions.h Mon May 18 13:46:57 2009
@@ -1616,11 +1616,6 @@
 
   virtual ExtractValueInst *clone() const;
 
-  // getType - Overload to return most specific pointer type...
-  const PointerType *getType() const {
-    return reinterpret_cast<const PointerType*>(Instruction::getType());
-  }
-
   /// getIndexedType - Returns the type of the element that would be extracted
   /// with an extractvalue instruction with the specified parameters.
   ///
@@ -1794,11 +1789,6 @@
   /// Transparently provide more efficient getOperand methods.
   DECLARE_TRANSPARENT_OPERAND_ACCESSORS(Value);
 
-  // getType - Overload to return most specific pointer type...
-  const PointerType *getType() const {
-    return reinterpret_cast<const PointerType*>(Instruction::getType());
-  }
-
   typedef const unsigned* idx_iterator;
   inline idx_iterator idx_begin() const { return Indices.begin(); }
   inline idx_iterator idx_end()   const { return Indices.end(); }




From gohman at apple.com  Mon May 18 13:55:40 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 18:55:40 -0000
Subject: [llvm-commits] [llvm] r72029 -
	/llvm/trunk/lib/VMCore/Instructions.cpp
Message-ID: <200905181855.n4IIte3R025392@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 13:55:39 2009
New Revision: 72029

URL: http://llvm.org/viewvc/llvm-project?rev=72029&view=rev
Log:
Revert r72025. It is possible for clients to convert between signed types
and pointer types safely if they only do so when the sizes are the same.
llvm-gcc is such a client.

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

Modified: llvm/trunk/lib/VMCore/Instructions.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/VMCore/Instructions.cpp?rev=72029&r1=72028&r2=72029&view=diff

==============================================================================
--- llvm/trunk/lib/VMCore/Instructions.cpp (original)
+++ llvm/trunk/lib/VMCore/Instructions.cpp Mon May 18 13:55:39 2009
@@ -2220,8 +2220,6 @@
     } else {
       assert(isa<PointerType>(SrcTy) &&
              "Casting from a value that is not first-class type");
-      assert(!SrcIsSigned &&
-             "Pointer types cannot be considered signed for conversions!");
       return PtrToInt;                              // ptr -> int
     }
   } else if (DestTy->isFloatingPoint()) {           // Casting to floating pt
@@ -2261,7 +2259,6 @@
     if (isa<PointerType>(SrcTy)) {
       return BitCast;                               // ptr -> ptr
     } else if (SrcTy->isInteger()) {
-      assert(!SrcIsSigned && "Cannot cast signed integer to pointer!");
       return IntToPtr;                              // int -> ptr
     } else {
       assert(!"Casting pointer to other than pointer or int");




From gohman at apple.com  Mon May 18 14:02:08 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 12:02:08 -0700
Subject: [llvm-commits] [llvm] r71757 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
In-Reply-To: <95C426D7-58A9-4A1A-B276-40C8FEB9C03D@apple.com>
References: <200905140030.n4E0UJ3H020876@zion.cs.uiuc.edu>
	<95C426D7-58A9-4A1A-B276-40C8FEB9C03D@apple.com>
Message-ID: <B20E2F34-5632-4604-86AF-399A1941FEEF@apple.com>


On May 13, 2009, at 6:38 PM, Chris Lattner wrote:

>
> On May 13, 2009, at 5:30 PM, Dan Gohman wrote:
>
>> Author: djg
>> Date: Wed May 13 19:30:16 2009
>> New Revision: 71757
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=71757&view=rev
>> Log:
>> Add an assert to turn a segfault on an unsupported inline
>> asm construct into an assertion failure.
>
> Should this use printf+exit(1)?


I don't know. I was just following similar code elsewhere in
the same file. Is there a conscious decision to use printf+exit
for certain cases now?

Dan



From lhames at gmail.com  Mon May 18 14:03:16 2009
From: lhames at gmail.com (Lang Hames)
Date: Mon, 18 May 2009 19:03:16 -0000
Subject: [llvm-commits] [llvm] r72030 - in /llvm/trunk/lib/CodeGen:
 RegAllocLinearScan.cpp Spiller.cpp Spiller.h
Message-ID: <200905181903.n4IJ3G6o025658@zion.cs.uiuc.edu>

Author: lhames
Date: Mon May 18 14:03:16 2009
New Revision: 72030

URL: http://llvm.org/viewvc/llvm-project?rev=72030&view=rev
Log:
New Spiller interface and trivial implementation.

Added:
    llvm/trunk/lib/CodeGen/Spiller.cpp
    llvm/trunk/lib/CodeGen/Spiller.h
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=72030&r1=72029&r2=72030&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp (original)
+++ llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp Mon May 18 14:03:16 2009
@@ -14,6 +14,7 @@
 #define DEBUG_TYPE "regalloc"
 #include "VirtRegMap.h"
 #include "VirtRegRewriter.h"
+#include "Spiller.h"
 #include "llvm/Function.h"
 #include "llvm/CodeGen/LiveIntervalAnalysis.h"
 #include "llvm/CodeGen/LiveStackAnalysis.h"
@@ -56,6 +57,11 @@
                   cl::desc("Pre-register allocation live interval splitting"),
                   cl::init(false), cl::Hidden);
 
+static cl::opt<bool>
+NewSpillFramework("new-spill-framework",
+                  cl::desc("New spilling framework"),
+                  cl::init(false), cl::Hidden);
+
 static RegisterRegAlloc
 linearscanRegAlloc("linearscan", "linear scan register allocator",
                    createLinearScanRegisterAllocator);
@@ -127,6 +133,8 @@
 
     std::auto_ptr<VirtRegRewriter> rewriter_;
 
+    std::auto_ptr<Spiller> spiller_;
+
   public:
     virtual const char* getPassName() const {
       return "Linear Scan Register Allocator";
@@ -420,6 +428,13 @@
 
   vrm_ = &getAnalysis<VirtRegMap>();
   if (!rewriter_.get()) rewriter_.reset(createVirtRegRewriter());
+  
+  if (NewSpillFramework) {
+    spiller_.reset(createSpiller(mf_, li_, vrm_));
+  }
+  else {
+    spiller_.reset(0);
+  }
 
   initIntervalSets();
 
@@ -1108,8 +1123,15 @@
   if (cur->weight != HUGE_VALF && cur->weight <= minWeight) {
     DOUT << "\t\t\tspilling(c): " << *cur << '\n';
     SmallVector<LiveInterval*, 8> spillIs;
-    std::vector<LiveInterval*> added =
-      li_->addIntervalsForSpills(*cur, spillIs, loopInfo, *vrm_);
+    std::vector<LiveInterval*> added;
+    
+    if (!NewSpillFramework) {
+      added = li_->addIntervalsForSpills(*cur, spillIs, loopInfo, *vrm_);
+    }
+    else {
+      added = spiller_->spill(cur); 
+    }
+
     std::sort(added.begin(), added.end(), LISorter());
     addStackInterval(cur, ls_, li_, mri_, *vrm_);
     if (added.empty())

Added: llvm/trunk/lib/CodeGen/Spiller.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.cpp?rev=72030&view=auto

==============================================================================
--- llvm/trunk/lib/CodeGen/Spiller.cpp (added)
+++ llvm/trunk/lib/CodeGen/Spiller.cpp Mon May 18 14:03:16 2009
@@ -0,0 +1,206 @@
+//===-- llvm/CodeGen/Spiller.cpp -  Spiller -------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#define DEBUG_TYPE "spiller"
+
+#include "Spiller.h"
+#include "VirtRegMap.h"
+#include "llvm/CodeGen/LiveIntervalAnalysis.h"
+#include "llvm/CodeGen/MachineFunction.h"
+#include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/MachineFrameInfo.h"
+#include "llvm/Target/TargetMachine.h"
+#include "llvm/Target/TargetInstrInfo.h"
+#include "llvm/Support/Debug.h"
+
+#include <algorithm>
+#include <map>
+
+using namespace llvm;
+
+Spiller::~Spiller() {}
+
+namespace {
+
+class TrivialSpiller : public Spiller {
+public:
+  TrivialSpiller(MachineFunction *mf, LiveIntervals *lis, VirtRegMap *vrm) :
+    mf(mf), lis(lis), vrm(vrm)
+  {
+    mfi = mf->getFrameInfo();
+    mri = &mf->getRegInfo();
+    tii = mf->getTarget().getInstrInfo();
+  }
+
+  std::vector<LiveInterval*> spill(LiveInterval *li) {
+
+    DOUT << "Trivial spiller spilling " << *li << "\n";
+
+    assert(li->weight != HUGE_VALF &&
+           "Attempting to spill already spilled value.");
+
+    assert(!li->isStackSlot() &&
+           "Trying to spill a stack slot.");
+
+    std::vector<LiveInterval*> added;
+    
+    const TargetRegisterClass *trc = mri->getRegClass(li->reg);
+    /*unsigned ss = mfi->CreateStackObject(trc->getSize(),
+                                         trc->getAlignment());*/
+    unsigned ss = vrm->assignVirt2StackSlot(li->reg);
+
+    MachineRegisterInfo::reg_iterator regItr = mri->reg_begin(li->reg);
+    
+    while (regItr != mri->reg_end()) {
+
+      MachineInstr *mi = &*regItr;
+
+      SmallVector<unsigned, 2> indices;
+      bool hasUse = false;
+      bool hasDef = false;
+    
+      for (unsigned i = 0; i != mi->getNumOperands(); ++i) {
+        MachineOperand &op = mi->getOperand(i);
+
+        if (!op.isReg() || op.getReg() != li->reg)
+          continue;
+      
+        hasUse |= mi->getOperand(i).isUse();
+        hasDef |= mi->getOperand(i).isDef();
+      
+        indices.push_back(i);
+      }
+
+      unsigned newVReg = mri->createVirtualRegister(trc);
+      LiveInterval *newLI = &lis->getOrCreateInterval(newVReg);
+      newLI->weight = HUGE_VALF;
+
+      vrm->grow();
+      vrm->assignVirt2StackSlot(newVReg, ss);
+
+      for (unsigned i = 0; i < indices.size(); ++i) {
+        mi->getOperand(indices[i]).setReg(newVReg);
+
+        if (mi->getOperand(indices[i]).isUse()) {
+          mi->getOperand(indices[i]).setIsKill(true);
+        }
+      }
+
+      if (hasUse) {
+        unsigned loadInstIdx = insertLoadFor(mi, ss, newVReg, trc);
+        unsigned start = lis->getDefIndex(loadInstIdx),
+                 end = lis->getUseIndex(lis->getInstructionIndex(mi));
+
+        VNInfo *vni =
+          newLI->getNextValue(loadInstIdx, 0, lis->getVNInfoAllocator());
+        vni->kills.push_back(lis->getInstructionIndex(mi));
+        LiveRange lr(start, end, vni);
+
+        newLI->addRange(lr);
+        added.push_back(newLI);
+      }
+
+      if (hasDef) {
+        unsigned storeInstIdx = insertStoreFor(mi, ss, newVReg, trc);
+        unsigned start = lis->getDefIndex(lis->getInstructionIndex(mi)),
+                 end = lis->getUseIndex(storeInstIdx);
+
+        VNInfo *vni =
+          newLI->getNextValue(storeInstIdx, 0, lis->getVNInfoAllocator());
+        vni->kills.push_back(storeInstIdx);
+        LiveRange lr(start, end, vni);
+      
+        newLI->addRange(lr);
+        added.push_back(newLI);
+      }
+
+      regItr = mri->reg_begin(li->reg);
+    }
+
+
+    return added;
+  }
+
+
+private:
+
+  MachineFunction *mf;
+  LiveIntervals *lis;
+  MachineFrameInfo *mfi;
+  MachineRegisterInfo *mri;
+  const TargetInstrInfo *tii;
+  VirtRegMap *vrm;
+
+
+
+  void makeRoomForInsertBefore(MachineInstr *mi) {
+    if (!lis->hasGapBeforeInstr(lis->getInstructionIndex(mi))) {
+      lis->computeNumbering();
+    }
+
+    assert(lis->hasGapBeforeInstr(lis->getInstructionIndex(mi)));
+  }
+
+  unsigned insertStoreFor(MachineInstr *mi, unsigned ss,
+                          unsigned newVReg,
+                          const TargetRegisterClass *trc) {
+    MachineBasicBlock::iterator nextInstItr(mi); 
+    ++nextInstItr;
+
+    makeRoomForInsertBefore(&*nextInstItr);
+
+    unsigned miIdx = lis->getInstructionIndex(mi);
+
+    tii->storeRegToStackSlot(*mi->getParent(), nextInstItr, newVReg,
+                             true, ss, trc);
+    MachineBasicBlock::iterator storeInstItr(mi);
+    ++storeInstItr;
+    MachineInstr *storeInst = &*storeInstItr;
+    unsigned storeInstIdx = miIdx + LiveIntervals::InstrSlots::NUM;
+
+    assert(lis->getInstructionFromIndex(storeInstIdx) == 0 &&
+           "Store inst index already in use.");
+    
+    lis->InsertMachineInstrInMaps(storeInst, storeInstIdx);
+
+    return storeInstIdx;
+  }
+
+  unsigned insertLoadFor(MachineInstr *mi, unsigned ss,
+                         unsigned newVReg,
+                         const TargetRegisterClass *trc) {
+    MachineBasicBlock::iterator useInstItr(mi);
+
+    makeRoomForInsertBefore(mi);
+ 
+    unsigned miIdx = lis->getInstructionIndex(mi);
+    
+    tii->loadRegFromStackSlot(*mi->getParent(), useInstItr, newVReg, ss, trc);
+    MachineBasicBlock::iterator loadInstItr(mi);
+    --loadInstItr;
+    MachineInstr *loadInst = &*loadInstItr;
+    unsigned loadInstIdx = miIdx - LiveIntervals::InstrSlots::NUM;
+
+    assert(lis->getInstructionFromIndex(loadInstIdx) == 0 &&
+           "Load inst index already in use.");
+
+    lis->InsertMachineInstrInMaps(loadInst, loadInstIdx);
+
+    return loadInstIdx;
+  }
+
+};
+
+}
+
+
+llvm::Spiller* llvm::createSpiller(MachineFunction *mf, LiveIntervals *lis,
+                                   VirtRegMap *vrm) {
+  return new TrivialSpiller(mf, lis, vrm);
+}

Added: llvm/trunk/lib/CodeGen/Spiller.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.h?rev=72030&view=auto

==============================================================================
--- llvm/trunk/lib/CodeGen/Spiller.h (added)
+++ llvm/trunk/lib/CodeGen/Spiller.h Mon May 18 14:03:16 2009
@@ -0,0 +1,32 @@
+//===-- llvm/CodeGen/Spiller.h - Spiller -*- C++ -*------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CODEGEN_SPILLER_H
+#define LLVM_CODEGEN_SPILLER_H
+
+#include <vector>
+
+namespace llvm {
+
+  /// Spiller interface.
+  ///
+  /// Implementations are utility classes which insert spill or remat code on
+  /// demand.
+  class Spiller {
+  public:
+    virtual ~Spiller() = 0;
+    virtual std::vector<class LiveInterval*> spill(class LiveInterval *li) = 0;
+  };
+
+  /// Create and return a spiller object, as specified on the command line.
+  Spiller* createSpiller(class MachineFunction *mf, class LiveIntervals *li,
+                         class VirtRegMap *vrm);
+}
+
+#endif




From akyrtzi at gmail.com  Mon May 18 15:17:23 2009
From: akyrtzi at gmail.com (Argiris Kirtzidis)
Date: Mon, 18 May 2009 23:17:23 +0300
Subject: [llvm-commits] Add scope info in DebugLoc
In-Reply-To: <16e5fdf90905121328r50ee3767n290bc2ff7000c6a6@mail.gmail.com>
References: <4A037F8D.90001@gmail.com>
	<16e5fdf90905121328r50ee3767n290bc2ff7000c6a6@mail.gmail.com>
Message-ID: <4A11C253.8050401@gmail.com>

Hi Bill,

Bill Wendling wrote:
> -  /// DebugLocTuple - Debug location tuple of filename id, line and column.
> +  /// DebugScope - Debug scope id.
>
> Please flesh out this comment. Add more about what a debug scope is
> and how it will be used.
>
> +  class DebugScope {
> +    unsigned Idx;
> +
> +  public:
> +    DebugScope() : Idx(~0U) {}  // Defaults to null.
>
> Well...it's not really Null. :-) Maybe it would be better to name it
> "invalid" (with appropriate changes to the other methods' names)?
>
> +  /// DebugScopeInfo - Contains info about the scope global variable and the
> +  /// parent debug scope.
>
> Could you explain why this is different from DebugScope.
>   

Ok to all.

> +  /// DebugLocTuple - Debug location tuple of compile unit, scope,
> +  /// line and column.
>
> The term "compile unit" is overloaded. There's the idea of a CU in the
> debug info, then the actual CompileUnit class, then a DICompileUnit
> class. Could you name it something else? Just for my sanity? :-) That
> is, unless CompileUnit is really absolutely correct here...
>   

I modified it to "DICompileUnit global variable", is this more clear ?

> +  void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
> +  void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
>
> FastISel has MF available as an ivar. Do you need to pass this in as a
> parameter? Also, please comment these methods.
>   

Moved them into a new DebugScopeTracker class.

> +DebugScope MachineFunction::CreateDebugScope(GlobalVariable *ScopeGV,
> +                                             DebugScope Parent) {
> +  DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
> +  return DebugScope::get(DbgScopeInfos.size() - 1);
>
> Should you verify that DbgScopeInfos.size() > 0 here?

Is it really needed ? A new item is pushed into the DbgScopeInfos in the 
immediate line above it.

>  Actually, why
> subtract one here at all? In the assert, you can verify that Idx <=
> DbgScopeInfos.size().
>   

Well, there's not much difference but I'd prefer to have 
DebugScope::getIndex return the "actual index" into the vector.
This is similar to the way DebugLoc's index is created.

> +  // FIXME: Proper debug scope.
> +  DebugScope DbgScope;
>
> What's stopping you from using a proper debug scope here?
>   

I added comments about this. Let me know what you think about this part.

> @@ -3817,6 +3821,25 @@
>    setValue(&I, result);
>  }
>
> +void SelectionDAGLowering::EnterDebugScope(GlobalVariable *ScopeGV,
> +                                           MachineFunction &MF) {
> +  assert(ScopeGV && "GlobalVariable for scope is null!");
> +  setCurDebugScope(MF.CreateDebugScope(ScopeGV, getCurDebugScope()));
> +}
> +
> +void SelectionDAGLowering::ExitDebugScope(GlobalVariable *ScopeGV,
> +                                          MachineFunction &MF) {
> +  assert(ScopeGV && "GlobalVariable for scope is null!");
> +  assert(!getCurDebugScope().isNull() && "Mismatched region.end ?");
> +  // We may have skipped a region.end because it was in an unreachable block.
> +  // Go up the scope chain until we reach the scope that ScopeGV points to.
> +  DebugScopeInfo DSI;
> +  do {
> +    DSI = MF.getDebugScopeInfo(getCurDebugScope());
> +    setCurDebugScope(DSI.Parent);
> +  } while (!DSI.Parent.isNull() && DSI.GV != ScopeGV);
> +}
>
> Because these methods are used in both SelectionDAGBuild.cpp and
> FastISel.cpp, is there a way to factor them out into another module?
> Maybe a CodeGen/DebugLoc.cpp module?
>   

Done.

I attached a new patch, thanks for reviewing!


-Argiris
-------------- next part --------------
A non-text attachment was scrubbed...
Name: debugscope.patch
Type: text/x-diff
Size: 21354 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/8f5c062a/attachment.bin 

From evan.cheng at apple.com  Mon May 18 15:29:32 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 18 May 2009 13:29:32 -0700
Subject: [llvm-commits] [llvm] r72030 - in /llvm/trunk/lib/CodeGen:
	RegAllocLinearScan.cpp Spiller.cpp Spiller.h
In-Reply-To: <200905181903.n4IJ3G6o025658@zion.cs.uiuc.edu>
References: <200905181903.n4IJ3G6o025658@zion.cs.uiuc.edu>
Message-ID: <DC25BC54-95B7-488C-85D1-C6EC144DB4A3@apple.com>

Hi Lang,

Thanks. Some comments:

+  if (NewSpillFramework) {
+    spiller_.reset(createSpiller(mf_, li_, vrm_));
+  }
+  else {
+    spiller_.reset(0);
+  }

+  std::vector<LiveInterval*> spill(LiveInterval *li) {
+
+    DOUT << "Trivial spiller spilling " << *li << "\n";

spill() is the top level interface used by the register allocators,  
right? I am assuming you are going to grow this spiller to do  
splitting etc. If so, please move the part that does the trivial  
spilling into a second route.

Is the else part necessary? Also, the coding convention used  
throughout llvm is
if () {
} else {
}

+    /*unsigned ss = mfi->CreateStackObject(trc->getSize(),
+                                         trc->getAlignment());*/

Please delete the commented lines.

+    MachineRegisterInfo::reg_iterator regItr = mri->reg_begin(li->reg);
+
+    while (regItr != mri->reg_end()) {
+
+      MachineInstr *mi = &*regItr;
...
+      regItr = mri->reg_begin(li->reg);
+    }

Why not just use a for loop iterating from begin() to end()?

   for (MachineRegisterInfo::reg_iterator regItr = mri_->reg_begin(li- 
 >reg),
          E = mri_->reg_end(); regItr != E; ) {
     MachineInstr *mi = &*regItr;
     ++regItr;

Please add some comments. :-)

Thanks,

Evan


On May 18, 2009, at 12:03 PM, Lang Hames wrote:

> Author: lhames
> Date: Mon May 18 14:03:16 2009
> New Revision: 72030
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72030&view=rev
> Log:
> New Spiller interface and trivial implementation.
>
> Added:
>    llvm/trunk/lib/CodeGen/Spiller.cpp
>    llvm/trunk/lib/CodeGen/Spiller.h
> 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=72030&r1=72029&r2=72030&view=diff
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp (original)
> +++ llvm/trunk/lib/CodeGen/RegAllocLinearScan.cpp Mon May 18  
> 14:03:16 2009
> @@ -14,6 +14,7 @@
> #define DEBUG_TYPE "regalloc"
> #include "VirtRegMap.h"
> #include "VirtRegRewriter.h"
> +#include "Spiller.h"
> #include "llvm/Function.h"
> #include "llvm/CodeGen/LiveIntervalAnalysis.h"
> #include "llvm/CodeGen/LiveStackAnalysis.h"
> @@ -56,6 +57,11 @@
>                   cl::desc("Pre-register allocation live interval  
> splitting"),
>                   cl::init(false), cl::Hidden);
>
> +static cl::opt<bool>
> +NewSpillFramework("new-spill-framework",
> +                  cl::desc("New spilling framework"),
> +                  cl::init(false), cl::Hidden);
> +
> static RegisterRegAlloc
> linearscanRegAlloc("linearscan", "linear scan register allocator",
>                    createLinearScanRegisterAllocator);
> @@ -127,6 +133,8 @@
>
>     std::auto_ptr<VirtRegRewriter> rewriter_;
>
> +    std::auto_ptr<Spiller> spiller_;
> +
>   public:
>     virtual const char* getPassName() const {
>       return "Linear Scan Register Allocator";
> @@ -420,6 +428,13 @@
>
>   vrm_ = &getAnalysis<VirtRegMap>();
>   if (!rewriter_.get()) rewriter_.reset(createVirtRegRewriter());
> +
> +  if (NewSpillFramework) {
> +    spiller_.reset(createSpiller(mf_, li_, vrm_));
> +  }
> +  else {
> +    spiller_.reset(0);
> +  }
>
>   initIntervalSets();
>
> @@ -1108,8 +1123,15 @@
>   if (cur->weight != HUGE_VALF && cur->weight <= minWeight) {
>     DOUT << "\t\t\tspilling(c): " << *cur << '\n';
>     SmallVector<LiveInterval*, 8> spillIs;
> -    std::vector<LiveInterval*> added =
> -      li_->addIntervalsForSpills(*cur, spillIs, loopInfo, *vrm_);
> +    std::vector<LiveInterval*> added;
> +
> +    if (!NewSpillFramework) {
> +      added = li_->addIntervalsForSpills(*cur, spillIs, loopInfo,  
> *vrm_);
> +    }
> +    else {
> +      added = spiller_->spill(cur);
> +    }
> +
>     std::sort(added.begin(), added.end(), LISorter());
>     addStackInterval(cur, ls_, li_, mri_, *vrm_);
>     if (added.empty())
>
> Added: llvm/trunk/lib/CodeGen/Spiller.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.cpp?rev=72030&view=auto
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/lib/CodeGen/Spiller.cpp (added)
> +++ llvm/trunk/lib/CodeGen/Spiller.cpp Mon May 18 14:03:16 2009
> @@ -0,0 +1,206 @@
> +//===-- llvm/CodeGen/Spiller.cpp -  Spiller  
> -------------------------------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open  
> Source
> +// License. See LICENSE.TXT for details.
> +//
> +// 
> = 
> = 
> = 
> ----------------------------------------------------------------------= 
> ==//
> +
> +#define DEBUG_TYPE "spiller"
> +
> +#include "Spiller.h"
> +#include "VirtRegMap.h"
> +#include "llvm/CodeGen/LiveIntervalAnalysis.h"
> +#include "llvm/CodeGen/MachineFunction.h"
> +#include "llvm/CodeGen/MachineRegisterInfo.h"
> +#include "llvm/CodeGen/MachineFrameInfo.h"
> +#include "llvm/Target/TargetMachine.h"
> +#include "llvm/Target/TargetInstrInfo.h"
> +#include "llvm/Support/Debug.h"
> +
> +#include <algorithm>
> +#include <map>
> +
> +using namespace llvm;
> +
> +Spiller::~Spiller() {}
> +
> +namespace {
> +
> +class TrivialSpiller : public Spiller {
> +public:
> +  TrivialSpiller(MachineFunction *mf, LiveIntervals *lis,  
> VirtRegMap *vrm) :
> +    mf(mf), lis(lis), vrm(vrm)
> +  {
> +    mfi = mf->getFrameInfo();
> +    mri = &mf->getRegInfo();
> +    tii = mf->getTarget().getInstrInfo();
> +  }
> +
> +  std::vector<LiveInterval*> spill(LiveInterval *li) {
> +
> +    DOUT << "Trivial spiller spilling " << *li << "\n";
> +
> +    assert(li->weight != HUGE_VALF &&
> +           "Attempting to spill already spilled value.");
> +
> +    assert(!li->isStackSlot() &&
> +           "Trying to spill a stack slot.");
> +
> +    std::vector<LiveInterval*> added;
> +
> +    const TargetRegisterClass *trc = mri->getRegClass(li->reg);
> +    /*unsigned ss = mfi->CreateStackObject(trc->getSize(),
> +                                         trc->getAlignment());*/
> +    unsigned ss = vrm->assignVirt2StackSlot(li->reg);
> +
> +    MachineRegisterInfo::reg_iterator regItr = mri->reg_begin(li- 
> >reg);
> +
> +    while (regItr != mri->reg_end()) {
> +
> +      MachineInstr *mi = &*regItr;
> +
> +      SmallVector<unsigned, 2> indices;
> +      bool hasUse = false;
> +      bool hasDef = false;
> +
> +      for (unsigned i = 0; i != mi->getNumOperands(); ++i) {
> +        MachineOperand &op = mi->getOperand(i);
> +
> +        if (!op.isReg() || op.getReg() != li->reg)
> +          continue;
> +
> +        hasUse |= mi->getOperand(i).isUse();
> +        hasDef |= mi->getOperand(i).isDef();
> +
> +        indices.push_back(i);
> +      }
> +
> +      unsigned newVReg = mri->createVirtualRegister(trc);
> +      LiveInterval *newLI = &lis->getOrCreateInterval(newVReg);
> +      newLI->weight = HUGE_VALF;
> +
> +      vrm->grow();
> +      vrm->assignVirt2StackSlot(newVReg, ss);
> +
> +      for (unsigned i = 0; i < indices.size(); ++i) {
> +        mi->getOperand(indices[i]).setReg(newVReg);
> +
> +        if (mi->getOperand(indices[i]).isUse()) {
> +          mi->getOperand(indices[i]).setIsKill(true);
> +        }
> +      }
> +
> +      if (hasUse) {
> +        unsigned loadInstIdx = insertLoadFor(mi, ss, newVReg, trc);
> +        unsigned start = lis->getDefIndex(loadInstIdx),
> +                 end = lis->getUseIndex(lis->getInstructionIndex 
> (mi));
> +
> +        VNInfo *vni =
> +          newLI->getNextValue(loadInstIdx, 0, lis- 
> >getVNInfoAllocator());
> +        vni->kills.push_back(lis->getInstructionIndex(mi));
> +        LiveRange lr(start, end, vni);
> +
> +        newLI->addRange(lr);
> +        added.push_back(newLI);
> +      }
> +
> +      if (hasDef) {
> +        unsigned storeInstIdx = insertStoreFor(mi, ss, newVReg, trc);
> +        unsigned start = lis->getDefIndex(lis->getInstructionIndex 
> (mi)),
> +                 end = lis->getUseIndex(storeInstIdx);
> +
> +        VNInfo *vni =
> +          newLI->getNextValue(storeInstIdx, 0, lis- 
> >getVNInfoAllocator());
> +        vni->kills.push_back(storeInstIdx);
> +        LiveRange lr(start, end, vni);
> +
> +        newLI->addRange(lr);
> +        added.push_back(newLI);
> +      }
> +
> +      regItr = mri->reg_begin(li->reg);
> +    }
> +
> +
> +    return added;
> +  }
> +
> +
> +private:
> +
> +  MachineFunction *mf;
> +  LiveIntervals *lis;
> +  MachineFrameInfo *mfi;
> +  MachineRegisterInfo *mri;
> +  const TargetInstrInfo *tii;
> +  VirtRegMap *vrm;
> +
> +
> +
> +  void makeRoomForInsertBefore(MachineInstr *mi) {
> +    if (!lis->hasGapBeforeInstr(lis->getInstructionIndex(mi))) {
> +      lis->computeNumbering();
> +    }
> +
> +    assert(lis->hasGapBeforeInstr(lis->getInstructionIndex(mi)));
> +  }
> +
> +  unsigned insertStoreFor(MachineInstr *mi, unsigned ss,
> +                          unsigned newVReg,
> +                          const TargetRegisterClass *trc) {
> +    MachineBasicBlock::iterator nextInstItr(mi);
> +    ++nextInstItr;
> +
> +    makeRoomForInsertBefore(&*nextInstItr);
> +
> +    unsigned miIdx = lis->getInstructionIndex(mi);
> +
> +    tii->storeRegToStackSlot(*mi->getParent(), nextInstItr, newVReg,
> +                             true, ss, trc);
> +    MachineBasicBlock::iterator storeInstItr(mi);
> +    ++storeInstItr;
> +    MachineInstr *storeInst = &*storeInstItr;
> +    unsigned storeInstIdx = miIdx + LiveIntervals::InstrSlots::NUM;
> +
> +    assert(lis->getInstructionFromIndex(storeInstIdx) == 0 &&
> +           "Store inst index already in use.");
> +
> +    lis->InsertMachineInstrInMaps(storeInst, storeInstIdx);
> +
> +    return storeInstIdx;
> +  }
> +
> +  unsigned insertLoadFor(MachineInstr *mi, unsigned ss,
> +                         unsigned newVReg,
> +                         const TargetRegisterClass *trc) {
> +    MachineBasicBlock::iterator useInstItr(mi);
> +
> +    makeRoomForInsertBefore(mi);
> +
> +    unsigned miIdx = lis->getInstructionIndex(mi);
> +
> +    tii->loadRegFromStackSlot(*mi->getParent(), useInstItr,  
> newVReg, ss, trc);
> +    MachineBasicBlock::iterator loadInstItr(mi);
> +    --loadInstItr;
> +    MachineInstr *loadInst = &*loadInstItr;
> +    unsigned loadInstIdx = miIdx - LiveIntervals::InstrSlots::NUM;
> +
> +    assert(lis->getInstructionFromIndex(loadInstIdx) == 0 &&
> +           "Load inst index already in use.");
> +
> +    lis->InsertMachineInstrInMaps(loadInst, loadInstIdx);
> +
> +    return loadInstIdx;
> +  }
> +
> +};
> +
> +}
> +
> +
> +llvm::Spiller* llvm::createSpiller(MachineFunction *mf,  
> LiveIntervals *lis,
> +                                   VirtRegMap *vrm) {
> +  return new TrivialSpiller(mf, lis, vrm);
> +}
>
> Added: llvm/trunk/lib/CodeGen/Spiller.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.h?rev=72030&view=auto
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/lib/CodeGen/Spiller.h (added)
> +++ llvm/trunk/lib/CodeGen/Spiller.h Mon May 18 14:03:16 2009
> @@ -0,0 +1,32 @@
> +//===-- llvm/CodeGen/Spiller.h - Spiller -*- C++ - 
> *------------------------===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open  
> Source
> +// License. See LICENSE.TXT for details.
> +//
> +// 
> = 
> = 
> = 
> ----------------------------------------------------------------------= 
> ==//
> +
> +#ifndef LLVM_CODEGEN_SPILLER_H
> +#define LLVM_CODEGEN_SPILLER_H
> +
> +#include <vector>
> +
> +namespace llvm {
> +
> +  /// Spiller interface.
> +  ///
> +  /// Implementations are utility classes which insert spill or  
> remat code on
> +  /// demand.
> +  class Spiller {
> +  public:
> +    virtual ~Spiller() = 0;
> +    virtual std::vector<class LiveInterval*> spill(class  
> LiveInterval *li) = 0;
> +  };
> +
> +  /// Create and return a spiller object, as specified on the  
> command line.
> +  Spiller* createSpiller(class MachineFunction *mf, class  
> LiveIntervals *li,
> +                         class VirtRegMap *vrm);
> +}
> +
> +#endif
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits



From baldrick at free.fr  Mon May 18 15:30:30 2009
From: baldrick at free.fr (Duncan Sands)
Date: Mon, 18 May 2009 22:30:30 +0200
Subject: [llvm-commits] [llvm] r71986
	-	/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
In-Reply-To: <16e5fdf90905181104q1cbfcd4cyae2bab1e902e1fbd@mail.gmail.com>
References: <200905171937.n4HJb3NQ025904@zion.cs.uiuc.edu>	<533C8EB1-0A3D-4C41-AFEA-95C91DBF7C8F@gmail.com>	<4A116B17.2030806@free.fr>
	<16e5fdf90905181104q1cbfcd4cyae2bab1e902e1fbd@mail.gmail.com>
Message-ID: <4A11C566.5080808@free.fr>

Hi Bill,

> I doubt it. The log file shows that it's using "llvm-gcc". This must
> be a release-mode deal, as it's failing in release mode for me (and
> not just self-hosting). Here's an option. Use "-disable-llvm-optzns"
> instead of "--disable-llvm-optzns". It seems that '--' is converted to
> '-f' in gcc. I have no clue why:
> 
> $ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
> cc1: for the -relocation-model=pic option: : may only occur zero or one times!
> 
> $ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
> cc1: Unknown command line argument '-frelocation-model=pic'.  Try: 'cc1 --help'

that's... very strange!  I'm not planning to look into it though.

Ciao,

Duncan.


From isanbard at gmail.com  Mon May 18 15:39:18 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 13:39:18 -0700
Subject: [llvm-commits] [llvm] r71986 -
	/llvm/trunk/test/FrontendC/2009-05-17-AlwaysInline.c
In-Reply-To: <4A11C566.5080808@free.fr>
References: <200905171937.n4HJb3NQ025904@zion.cs.uiuc.edu>
	<533C8EB1-0A3D-4C41-AFEA-95C91DBF7C8F@gmail.com>
	<4A116B17.2030806@free.fr>
	<16e5fdf90905181104q1cbfcd4cyae2bab1e902e1fbd@mail.gmail.com>
	<4A11C566.5080808@free.fr>
Message-ID: <16e5fdf90905181339y8710f86h3c2e16c244ea11a8@mail.gmail.com>

On Mon, May 18, 2009 at 1:30 PM, Duncan Sands <baldrick at free.fr> wrote:
> Hi Bill,
>
>> I doubt it. The log file shows that it's using "llvm-gcc". This must
>> be a release-mode deal, as it's failing in release mode for me (and
>> not just self-hosting). Here's an option. Use "-disable-llvm-optzns"
>> instead of "--disable-llvm-optzns". It seems that '--' is converted to
>> '-f' in gcc. I have no clue why:
>>
>> $ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
>> cc1: for the -relocation-model=pic option: : may only occur zero or one times!
>>
>> $ /Developer/usr/bin/llvm-gcc -c t.c -o /dev/null -mllvm --relocation-model=pic
>> cc1: Unknown command line argument '-frelocation-model=pic'. ?Try: 'cc1 --help'
>
> that's... very strange!

I agree!

>?I'm not planning to look into it though.
>
I wouldn't suggest it either. :-) This way lies madness.

-bw



From bob.wilson at apple.com  Mon May 18 15:55:41 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Mon, 18 May 2009 20:55:41 -0000
Subject: [llvm-commits] [llvm] r72037 -
	/llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp
Message-ID: <200905182055.n4IKtmPi029794@zion.cs.uiuc.edu>

Author: bwilson
Date: Mon May 18 15:55:32 2009
New Revision: 72037

URL: http://llvm.org/viewvc/llvm-project?rev=72037&view=rev
Log:
Fix pr4202: Disable CodePlacementOpt for ARM.  The ARMConstantIslandPass has
to run last because it needs to know the exact size and position of every
basic block.  Currently CodePlacementOpt is set up to run last.  It might be
worthwhile to investigate reordering these passes, but for now, let's just
make it work.

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=72037&r1=72036&r2=72037&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.cpp Mon May 18 15:55:32 2009
@@ -288,7 +288,10 @@
   setIfCvtDupBlockSizeLimit(Subtarget->isThumb() ? 0 : 2);
 
   maxStoresPerMemcpy = 1;   //// temporary - rewrite interface to use type
-  benefitFromCodePlacementOpt = true;
+  // Do not enable CodePlacementOpt for now: it currently runs after the
+  // ARMConstantIslandPass and messes up branch relaxation and placement
+  // of constant islands.
+  // benefitFromCodePlacementOpt = true;
 }
 
 const char *ARMTargetLowering::getTargetNodeName(unsigned Opcode) const {




From bob.wilson at apple.com  Mon May 18 16:03:03 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Mon, 18 May 2009 21:03:03 -0000
Subject: [llvm-commits] [llvm] r72038 -
	/llvm/trunk/lib/CodeGen/CodePlacementOpt.cpp
Message-ID: <200905182103.n4IL37c5030058@zion.cs.uiuc.edu>

Author: bwilson
Date: Mon May 18 16:02:18 2009
New Revision: 72038

URL: http://llvm.org/viewvc/llvm-project?rev=72038&view=rev
Log:
Fix CodePlacementOpt::OptimizeIntraLoopEdges so that its return value
correctly indicates whether it changed the code.

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

Modified: llvm/trunk/lib/CodeGen/CodePlacementOpt.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/CodePlacementOpt.cpp?rev=72038&r1=72037&r2=72038&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/CodePlacementOpt.cpp (original)
+++ llvm/trunk/lib/CodeGen/CodePlacementOpt.cpp Mon May 18 16:02:18 2009
@@ -134,6 +134,7 @@
       TII->RemoveBranch(*MBB);
       ChangedMBBs.insert(MBB);
       ++NumIntraElim;
+      Changed = true;
       continue;
     }
 
@@ -231,6 +232,7 @@
       TII->InsertBranch(*FtMBB, FtTBB, FtFBB, FtCond);
       ChangedMBBs.insert(FtMBB);
     }
+    Changed = true;
 
     // If BB is the loop latch, we may have a new loop headr.
     if (MBB == L->getLoopLatch()) {




From akyrtzi at gmail.com  Mon May 18 16:06:48 2009
From: akyrtzi at gmail.com (Argiris Kirtzidis)
Date: Mon, 18 May 2009 21:06:48 -0000
Subject: [llvm-commits] [llvm] r72040 - in /llvm/trunk:
 include/llvm/ExecutionEngine/ExecutionEngine.h
 lib/ExecutionEngine/JIT/JIT.cpp lib/ExecutionEngine/JIT/JIT.h
 lib/ExecutionEngine/JIT/JITEmitter.cpp
Message-ID: <200905182106.n4IL6xl1030221@zion.cs.uiuc.edu>

Author: akirtzidis
Date: Mon May 18 16:06:40 2009
New Revision: 72040

URL: http://llvm.org/viewvc/llvm-project?rev=72040&view=rev
Log:
Allow the JIT ExecutionEngine to report details about the generated machine code.
Introduce a new class (MachineCodeInfo) that the JIT can fill in with details. Right now, just the address and the size of the machine code are reported.

Patch by Evan Phoenix!

Modified:
    llvm/trunk/include/llvm/ExecutionEngine/ExecutionEngine.h
    llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp
    llvm/trunk/lib/ExecutionEngine/JIT/JIT.h
    llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp

Modified: llvm/trunk/include/llvm/ExecutionEngine/ExecutionEngine.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/ExecutionEngine/ExecutionEngine.h?rev=72040&r1=72039&r2=72040&view=diff

==============================================================================
--- llvm/trunk/include/llvm/ExecutionEngine/ExecutionEngine.h (original)
+++ llvm/trunk/include/llvm/ExecutionEngine/ExecutionEngine.h Mon May 18 16:06:40 2009
@@ -35,6 +35,7 @@
 class Type;
 class MutexGuard;
 class JITMemoryManager;
+class MachineCodeInfo;
 
 class ExecutionEngineState {
 private:
@@ -241,6 +242,9 @@
     return getPointerToFunction(F);
   }
 
+  // The JIT overrides a version that actually does this.
+  virtual void runJITOnFunction(Function *F, MachineCodeInfo *MCI = 0) { }
+
   /// getGlobalValueAtAddress - Return the LLVM global value object that starts
   /// at the specified address.
   ///

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp?rev=72040&r1=72039&r2=72040&view=diff

==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JIT.cpp Mon May 18 16:06:40 2009
@@ -21,6 +21,7 @@
 #include "llvm/ModuleProvider.h"
 #include "llvm/CodeGen/MachineCodeEmitter.h"
 #include "llvm/ExecutionEngine/GenericValue.h"
+#include "llvm/CodeGen/MachineCodeInfo.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetMachine.h"
 #include "llvm/Target/TargetJITInfo.h"
@@ -512,9 +513,14 @@
 /// just-in-time compilation passes on F, hopefully filling in
 /// GlobalAddress[F] with the address of F's machine code.
 ///
-void JIT::runJITOnFunction(Function *F) {
+void JIT::runJITOnFunction(Function *F, MachineCodeInfo *MCI) {
   MutexGuard locked(lock);
+
+  registerMachineCodeInfo(MCI);
+
   runJITOnFunctionUnlocked(F, locked);
+
+  registerMachineCodeInfo(0);
 }
 
 void JIT::runJITOnFunctionUnlocked(Function *F, const MutexGuard &locked) {

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JIT.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JIT.h?rev=72040&r1=72039&r2=72040&view=diff

==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JIT.h (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JIT.h Mon May 18 16:06:40 2009
@@ -23,6 +23,7 @@
 class TargetMachine;
 class TargetJITInfo;
 class MachineCodeEmitter;
+class MachineCodeInfo;
 
 class JITState {
 private:
@@ -151,14 +152,18 @@
   static ExecutionEngine *createJIT(ModuleProvider *MP, std::string *Err,
                                     JITMemoryManager *JMM,
                                     CodeGenOpt::Level OptLevel);
-  
+
+
+  // Run the JIT on F and return information about the generated code
+  void runJITOnFunction(Function *F, MachineCodeInfo *MCI = 0);
+
 private:
   static MachineCodeEmitter *createEmitter(JIT &J, JITMemoryManager *JMM);
-  void runJITOnFunction(Function *F);
+  void registerMachineCodeInfo(MachineCodeInfo *MCI);
   void runJITOnFunctionUnlocked(Function *F, const MutexGuard &locked);
   void updateFunctionStub(Function *F);
   void updateDlsymStubTable();
-  
+
 protected:
 
   /// getMemoryforGV - Allocate memory for a global variable.

Modified: llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp?rev=72040&r1=72039&r2=72040&view=diff

==============================================================================
--- llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp (original)
+++ llvm/trunk/lib/ExecutionEngine/JIT/JITEmitter.cpp Mon May 18 16:06:40 2009
@@ -26,6 +26,7 @@
 #include "llvm/CodeGen/MachineRelocation.h"
 #include "llvm/ExecutionEngine/JITMemoryManager.h"
 #include "llvm/ExecutionEngine/GenericValue.h"
+#include "llvm/CodeGen/MachineCodeInfo.h"
 #include "llvm/Target/TargetData.h"
 #include "llvm/Target/TargetJITInfo.h"
 #include "llvm/Target/TargetMachine.h"
@@ -614,9 +615,12 @@
     // ExtFnStubs - A map of external function names to stubs which have entries
     // in the JITResolver's ExternalFnToStubMap.
     StringMap<void *> ExtFnStubs;
-    
+
+    // MCI - A pointer to a MachineCodeInfo object to update with information.
+    MachineCodeInfo *MCI;
+
   public:
-    JITEmitter(JIT &jit, JITMemoryManager *JMM) : Resolver(jit), CurFn(0) {
+    JITEmitter(JIT &jit, JITMemoryManager *JMM) : Resolver(jit), CurFn(0), MCI(0) {
       MemMgr = JMM ? JMM : JITMemoryManager::CreateDefaultMemManager();
       if (jit.getJITInfo().needsGOT()) {
         MemMgr->AllocateGOT();
@@ -712,6 +716,10 @@
     
     JITMemoryManager *getMemMgr(void) const { return MemMgr; }
 
+    void setMachineCodeInfo(MachineCodeInfo *mci) {
+      MCI = mci;
+    }
+
   private:
     void *getPointerToGlobal(GlobalValue *GV, void *Reference, bool NoNeedStub);
     void *getPointerToGVIndirectSym(GlobalValue *V, void *Reference,
@@ -1155,6 +1163,12 @@
        << "] Function: " << F.getFunction()->getName()
        << ": " << (FnEnd-FnStart) << " bytes of text, "
        << Relocations.size() << " relocations\n";
+
+  if (MCI) {
+    MCI->setAddress(FnStart);
+    MCI->setSize(FnEnd-FnStart);
+  }
+
   Relocations.clear();
   ConstPoolAddresses.clear();
 
@@ -1478,6 +1492,13 @@
   return JE->getJITResolver().getFunctionStub(F);
 }
 
+void JIT::registerMachineCodeInfo(MachineCodeInfo *mc) {
+  assert(isa<JITEmitter>(MCE) && "Unexpected MCE?");
+  JITEmitter *JE = cast<JITEmitter>(getCodeEmitter());
+
+  JE->setMachineCodeInfo(mc);
+}
+
 void JIT::updateFunctionStub(Function *F) {
   // Get the empty stub we generated earlier.
   assert(isa<JITEmitter>(MCE) && "Unexpected MCE?");




From akyrtzi at gmail.com  Mon May 18 16:08:46 2009
From: akyrtzi at gmail.com (Argiris Kirtzidis)
Date: Mon, 18 May 2009 21:08:46 -0000
Subject: [llvm-commits] [llvm] r72042 -
	/llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h
Message-ID: <200905182108.n4IL8lOs030292@zion.cs.uiuc.edu>

Author: akirtzidis
Date: Mon May 18 16:08:45 2009
New Revision: 72042

URL: http://llvm.org/viewvc/llvm-project?rev=72042&view=rev
Log:
Add missing file.

Added:
    llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h

Added: llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h?rev=72042&view=auto

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h (added)
+++ llvm/trunk/include/llvm/CodeGen/MachineCodeInfo.h Mon May 18 16:08:45 2009
@@ -0,0 +1,51 @@
+//===-- MachineCodeInfo.h - Class used to report JIT info -------*- 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 MachineCodeInfo, a class used by the JIT ExecutionEngine
+// to report information about the generated machine code.
+//
+// See JIT::runJITOnFunction for usage.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef EE_MACHINE_CODE_INFO_H
+#define EE_MACHINE_CODE_INFO_H
+
+namespace llvm {
+
+class MachineCodeInfo {
+private:
+  size_t Size;   // Number of bytes in memory used
+  void *Address; // The address of the function in memory
+
+public:
+  MachineCodeInfo() : Size(0), Address(0) {}
+
+  void setSize(size_t s) {
+    Size = s;
+  }
+
+  void setAddress(void *a) {
+    Address = a;
+  }
+
+  size_t size() const {
+    return Size;
+  }
+
+  void *address() const {
+    return Address;
+  }
+
+};
+
+}
+
+#endif
+




From akyrtzi at gmail.com  Mon May 18 16:13:28 2009
From: akyrtzi at gmail.com (Argiris Kirtzidis)
Date: Tue, 19 May 2009 00:13:28 +0300
Subject: [llvm-commits] [PATCH] Generate info about the JIT process
In-Reply-To: <0F735E86-C6C7-4352-8262-B394C70A5D26@fallingsnow.net>
References: <978B1AEF-4B6B-4BA1-A73F-6FEF5C95A12D@fallingsnow.net>	<4A0A6C85.1030909@gmail.com>	<E84E4E33-642E-4A2F-A645-D6CFF6569E3E@fallingsnow.net>	<4A0A90FB.3000907@gmail.com>	<0A9506FE-111E-4D3C-BC04-D962645AA1DE@apple.com>
	<0F735E86-C6C7-4352-8262-B394C70A5D26@fallingsnow.net>
Message-ID: <4A11CF78.8020301@gmail.com>

Hi Evan,

I commited the patch, thanks!

-Argiris


Evan Phoenix wrote:
> I haven't heard anything from Argiris, or any other thoughts on what 
> the JITListener would look like. I hesitate to write such a thing 
> without an idea of what people want, so I'm hoping we can commit my 
> patch as is.
>
> When there is more discussion about the JITListener, it will be easy 
> to make MachineCodeInfo sit on top of it without any external API 
> changes.
>
> I've attached an updated version of the patch against SVN HEAD.
>
>  - Evan
>
>
>
> On May 13, 2009, at 10:21 AM, Chris Lattner wrote:
>
>> On May 13, 2009, at 2:20 AM, Argiris Kirtzidis wrote:
>>>> Thats basically what MachineCodeInfo enables. ATM, there are only 2
>>>> things it publishes directly to the MCI, but future APIs can easily be
>>>> added using virtual methods to allow for subclasses of MachineCodeInfo
>>>> to do new things.
>>>>
>>>> This is far from a final set of things MCI will do, it's primarily a
>>>> way of introduce an API that we can add future things to.
>>>
>>> Ok, that sounds great. It's just that MachineCodeInfo is restricted to
>>> the "function level", how about "module level" info, specifically
>>> getting notifications during code execution like:
>>> -"I lazy JITed this function"
>>> -"I allocated this memory chunk for this global variable"
>>>
>>> Of course, MachineCodeInfo could be the one for the "function level" 
>>> and
>>> something like "JITListener" at the module level. The JITListener could
>>> have a method that passes a MachineCodeInfo before function JITing. IMO
>>> this is flexible and clean.
>>
>> I agree with Argiris: there should be a simple abstract interface 
>> with methods like the above.  The JIT should have an optional pointer 
>> to it.  If non-null, it would invoke the methods on the JITListener.  
>> This would allow lots of detailed instrumentation, but would 
>> introduce no overhead if it's not being used.
>>
>> -Chris
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>   


From bob.wilson at apple.com  Mon May 18 16:17:21 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Mon, 18 May 2009 21:17:21 -0000
Subject: [llvm-commits] [nightly-test-server] r72043 -
 /nightly-test-server/trunk/NightlyTestAccept.php
Message-ID: <200905182117.n4ILHS2b030542@zion.cs.uiuc.edu>

Author: bwilson
Date: Mon May 18 16:17:03 2009
New Revision: 72043

URL: http://llvm.org/viewvc/llvm-project?rev=72043&view=rev
Log:
Fix pr4030: Include the $machine_id in the URL for the buildlog.

Modified:
    nightly-test-server/trunk/NightlyTestAccept.php

Modified: nightly-test-server/trunk/NightlyTestAccept.php
URL: http://llvm.org/viewvc/llvm-project/nightly-test-server/trunk/NightlyTestAccept.php?rev=72043&r1=72042&r2=72043&view=diff

==============================================================================
--- nightly-test-server/trunk/NightlyTestAccept.php (original)
+++ nightly-test-server/trunk/NightlyTestAccept.php Mon May 18 16:17:03 2009
@@ -939,7 +939,7 @@
   }
   else{
     $email .= "\nBuildlog available at http://llvm.org/nightlytest/".
-              "machines/$db_date-Build-Log.txt\n";
+              "machines/$machine_id/$db_date-Build-Log.txt\n";
   }
 
   $email_addr = "llvm-testresults\@cs.uiuc.edu";




From evan at fallingsnow.net  Mon May 18 16:38:59 2009
From: evan at fallingsnow.net (Evan Phoenix)
Date: Mon, 18 May 2009 14:38:59 -0700
Subject: [llvm-commits] [PATCH] Generate info about the JIT process
In-Reply-To: <4A11CF78.8020301@gmail.com>
References: <978B1AEF-4B6B-4BA1-A73F-6FEF5C95A12D@fallingsnow.net>	<4A0A6C85.1030909@gmail.com>	<E84E4E33-642E-4A2F-A645-D6CFF6569E3E@fallingsnow.net>	<4A0A90FB.3000907@gmail.com>	<0A9506FE-111E-4D3C-BC04-D962645AA1DE@apple.com>
	<0F735E86-C6C7-4352-8262-B394C70A5D26@fallingsnow.net>
	<4A11CF78.8020301@gmail.com>
Message-ID: <B00B3EF9-437A-4C2A-83FD-644F1EB586ED@fallingsnow.net>


On May 18, 2009, at 2:13 PM, Argiris Kirtzidis wrote:

> Hi Evan,
>
> I commited the patch, thanks!

Wonderful! Thanks so much!

  - Evan

>
>
> -Argiris
>
>
> Evan Phoenix wrote:
>> I haven't heard anything from Argiris, or any other thoughts on  
>> what the JITListener would look like. I hesitate to write such a  
>> thing without an idea of what people want, so I'm hoping we can  
>> commit my patch as is.
>>
>> When there is more discussion about the JITListener, it will be  
>> easy to make MachineCodeInfo sit on top of it without any external  
>> API changes.
>>
>> I've attached an updated version of the patch against SVN HEAD.
>>
>> - Evan
>>
>>
>>
>> On May 13, 2009, at 10:21 AM, Chris Lattner wrote:
>>
>>> On May 13, 2009, at 2:20 AM, Argiris Kirtzidis wrote:
>>>>> Thats basically what MachineCodeInfo enables. ATM, there are  
>>>>> only 2
>>>>> things it publishes directly to the MCI, but future APIs can  
>>>>> easily be
>>>>> added using virtual methods to allow for subclasses of  
>>>>> MachineCodeInfo
>>>>> to do new things.
>>>>>
>>>>> This is far from a final set of things MCI will do, it's  
>>>>> primarily a
>>>>> way of introduce an API that we can add future things to.
>>>>
>>>> Ok, that sounds great. It's just that MachineCodeInfo is  
>>>> restricted to
>>>> the "function level", how about "module level" info, specifically
>>>> getting notifications during code execution like:
>>>> -"I lazy JITed this function"
>>>> -"I allocated this memory chunk for this global variable"
>>>>
>>>> Of course, MachineCodeInfo could be the one for the "function  
>>>> level" and
>>>> something like "JITListener" at the module level. The JITListener  
>>>> could
>>>> have a method that passes a MachineCodeInfo before function  
>>>> JITing. IMO
>>>> this is flexible and clean.
>>>
>>> I agree with Argiris: there should be a simple abstract interface  
>>> with methods like the above.  The JIT should have an optional  
>>> pointer to it.  If non-null, it would invoke the methods on the  
>>> JITListener.  This would allow lots of detailed instrumentation,  
>>> but would introduce no overhead if it's not being used.
>>>
>>> -Chris
>>
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> llvm-commits mailing list
>> llvm-commits at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>



From dalej at apple.com  Mon May 18 16:42:26 2009
From: dalej at apple.com (Dale Johannesen)
Date: Mon, 18 May 2009 21:42:26 -0000
Subject: [llvm-commits] [llvm] r72045 -
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
Message-ID: <200905182142.n4ILgWgP031346@zion.cs.uiuc.edu>

Author: johannes
Date: Mon May 18 16:41:59 2009
New Revision: 72045

URL: http://llvm.org/viewvc/llvm-project?rev=72045&view=rev
Log:
Add OpSize to 16-bit ADC and SBB.


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=72045&r1=72044&r2=72045&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon May 18 16:41:59 2009
@@ -2280,7 +2280,7 @@
 def ADC16rr  : I<0x11, MRMDestReg, (outs GR16:$dst),
                                    (ins GR16:$src1, GR16:$src2),
                  "adc{w}\t{$src2, $dst|$dst, $src2}",
-                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>;
+                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>, OpSize;
 def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst),
                                    (ins GR32:$src1, GR32:$src2),
                  "adc{l}\t{$src2, $dst|$dst, $src2}",
@@ -2293,7 +2293,8 @@
 def ADC16rm  : I<0x13, MRMSrcMem , (outs GR16:$dst),
                                    (ins GR16:$src1, i16mem:$src2),
                  "adc{w}\t{$src2, $dst|$dst, $src2}",
-                 [(set GR16:$dst, (adde GR16:$src1, (load addr:$src2)))]>;
+                 [(set GR16:$dst, (adde GR16:$src1, (load addr:$src2)))]>,
+                 OpSize;
 def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst),
                                    (ins GR32:$src1, i32mem:$src2),
                  "adc{l}\t{$src2, $dst|$dst, $src2}",
@@ -2304,11 +2305,12 @@
 def ADC16ri  : Ii16<0x81, MRM2r, (outs GR16:$dst),
                                  (ins GR16:$src1, i16imm:$src2),
                     "adc{w}\t{$src2, $dst|$dst, $src2}",
-                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>;
+                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>, OpSize;
 def ADC16ri8 : Ii8<0x83, MRM2r, (outs GR16:$dst),
                                 (ins GR16:$src1, i16i8imm:$src2),
                    "adc{w}\t{$src2, $dst|$dst, $src2}",
-                 [(set GR16:$dst, (adde GR16:$src1, i16immSExt8:$src2))]>;
+                 [(set GR16:$dst, (adde GR16:$src1, i16immSExt8:$src2))]>,
+                 OpSize;
 def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst),
                                  (ins GR32:$src1, i32imm:$src2),
                     "adc{l}\t{$src2, $dst|$dst, $src2}",
@@ -2324,7 +2326,8 @@
                    [(store (adde (load addr:$dst), GR8:$src2), addr:$dst)]>;
   def ADC16mr  : I<0x11, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2),
                    "adc{w}\t{$src2, $dst|$dst, $src2}",
-                   [(store (adde (load addr:$dst), GR16:$src2), addr:$dst)]>;
+                   [(store (adde (load addr:$dst), GR16:$src2), addr:$dst)]>,
+                   OpSize;
   def ADC32mr  : I<0x11, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2),
                    "adc{l}\t{$src2, $dst|$dst, $src2}",
                    [(store (adde (load addr:$dst), GR32:$src2), addr:$dst)]>;
@@ -2333,10 +2336,12 @@
                   [(store (adde (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
   def ADC16mi  : Ii16<0x81, MRM2m, (outs), (ins i16mem:$dst, i16imm:$src2),
                       "adc{w}\t{$src2, $dst|$dst, $src2}",
-                  [(store (adde (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
+                  [(store (adde (loadi16 addr:$dst), imm:$src2), addr:$dst)]>,
+                  OpSize;
   def ADC16mi8 : Ii8<0x83, MRM2m, (outs), (ins i16mem:$dst, i16i8imm :$src2),
                      "adc{w}\t{$src2, $dst|$dst, $src2}",
-               [(store (adde (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
+               [(store (adde (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>,
+               OpSize;
   def ADC32mi  : Ii32<0x81, MRM2m, (outs), (ins i32mem:$dst, i32imm:$src2),
                       "adc{l}\t{$src2, $dst|$dst, $src2}",
                   [(store (adde (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
@@ -2452,7 +2457,7 @@
 def SBB16rr    : I<0x19, MRMDestReg, (outs GR16:$dst),
                                      (ins GR16:$src1, GR16:$src2),
                   "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>;
+                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>, OpSize;
 def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst),
                                       (ins GR32:$src1, GR32:$src2),
                   "sbb{l}\t{$src2, $dst|$dst, $src2}",
@@ -2464,7 +2469,8 @@
                    [(store (sube (load addr:$dst), GR8:$src2), addr:$dst)]>;
   def SBB16mr  : I<0x19, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), 
                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                   [(store (sube (load addr:$dst), GR16:$src2), addr:$dst)]>;
+                   [(store (sube (load addr:$dst), GR16:$src2), addr:$dst)]>,
+                   OpSize;
   def SBB32mr  : I<0x19, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), 
                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
                    [(store (sube (load addr:$dst), GR32:$src2), addr:$dst)]>;
@@ -2473,10 +2479,12 @@
                    [(store (sube (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
   def SBB16mi  : Ii16<0x81, MRM3m, (outs), (ins i16mem:$dst, i16imm:$src2), 
                       "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                  [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
+                  [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>,
+                  OpSize;
   def SBB16mi8 : Ii8<0x83, MRM3m, (outs), (ins i16mem:$dst, i16i8imm :$src2), 
                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
-               [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
+               [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>,
+               OpSize;
   def SBB32mi  : Ii32<0x81, MRM3m, (outs), (ins i32mem:$dst, i32imm:$src2), 
                       "sbb{l}\t{$src2, $dst|$dst, $src2}",
                   [(store (sube (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
@@ -2490,7 +2498,8 @@
 def SBB16rm  : I<0x1B, MRMSrcMem, (outs GR16:$dst),
                                   (ins GR16:$src1, i16mem:$src2),
                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                    [(set GR16:$dst, (sube GR16:$src1, (load addr:$src2)))]>;
+                    [(set GR16:$dst, (sube GR16:$src1, (load addr:$src2)))]>,
+                    OpSize;
 def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst),
                                   (ins GR32:$src1, i32mem:$src2),
                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
@@ -2501,11 +2510,12 @@
 def SBB16ri  : Ii16<0x81, MRM3r, (outs GR16:$dst),
                                  (ins GR16:$src1, i16imm:$src2),
                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                    [(set GR16:$dst, (sube GR16:$src1, imm:$src2))]>;
+                    [(set GR16:$dst, (sube GR16:$src1, imm:$src2))]>, OpSize;
 def SBB16ri8 : Ii8<0x83, MRM3r, (outs GR16:$dst),
                                 (ins GR16:$src1, i16i8imm:$src2),
                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
-                   [(set GR16:$dst, (sube GR16:$src1, i16immSExt8:$src2))]>;
+                   [(set GR16:$dst, (sube GR16:$src1, i16immSExt8:$src2))]>,
+                   OpSize;
 def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst), 
                                  (ins GR32:$src1, i32imm:$src2),
                     "sbb{l}\t{$src2, $dst|$dst, $src2}",




From isanbard at gmail.com  Mon May 18 17:02:44 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 22:02:44 -0000
Subject: [llvm-commits] [llvm] r72047 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <200905182202.n4IM2jVA032127@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 17:02:36 2009
New Revision: 72047

URL: http://llvm.org/viewvc/llvm-project?rev=72047&view=rev
Log:
RecordVariable is called each time a DECLARE node is encountered. For an inlined
function, this could be many, many times. We don't want to re-add variables to
that DIE for each time. We just want to add them once. Check to make sure that
we haven't added them already.

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=72047&r1=72046&r2=72047&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon May 18 17:02:36 2009
@@ -855,7 +855,8 @@
 /// CreateSubprogramDIE - Create new DIE using SP.
 DIE *DwarfDebug::CreateSubprogramDIE(CompileUnit *DW_Unit,
                                      const DISubprogram &SP,
-                                     bool IsConstructor) {
+                                     bool IsConstructor,
+                                     bool IsInlined) {
   DIE *SPDie = new DIE(dwarf::DW_TAG_subprogram);
 
   std::string Name;
@@ -903,7 +904,7 @@
       }
   }
 
-  if (!SP.isLocalToUnit())
+  if (!SP.isLocalToUnit() && !IsInlined)
     AddUInt(SPDie, dwarf::DW_AT_external, dwarf::DW_FORM_flag, 1);
 
   // DW_TAG_inlined_subroutine may refer to this DIE.
@@ -2425,7 +2426,7 @@
   return ID;
 }
 
-/// RecordVariable - Indicate the declaration of  a local variable.
+/// RecordVariable - Indicate the declaration of a local variable.
 void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
                                 const MachineInstr *MI) {
   if (TimePassesIsEnabled)
@@ -2453,12 +2454,33 @@
       DenseMap<const GlobalVariable *, DbgScope *>::iterator
         AI = AbstractInstanceRootMap.find(V);
 
-      if (AI != AbstractInstanceRootMap.end())
+      if (AI != AbstractInstanceRootMap.end()) {
+        // This method is called each time a DECLARE node is encountered. For an
+        // inlined function, this could be many, many times. We don't want to
+        // re-add variables to that DIE for each time. We just want to add them
+        // once. Check to make sure that we haven't added them already.
+        DenseMap<const GlobalVariable *,
+          SmallSet<const GlobalVariable *, 32> >::iterator
+          IP = InlinedParamMap.find(V);
+
+        if (IP != InlinedParamMap.end()) {
+          SmallSet<const GlobalVariable*, 32> &S = IP->second;
+
+          if (S.count(GV) > 0) {
+            if (TimePassesIsEnabled)
+              DebugTimer->stopTimer();
+            return;
+          }
+
+        }
+
         // or GV is an inlined local variable.
         Scope = AI->second;
-      else
+        InlinedParamMap[V].insert(GV);
+      } else {
         // or GV is a local variable.
         Scope = getOrCreateScope(V);
+      }
     }
   }
 
@@ -2494,13 +2516,13 @@
     CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
     DIE *SPDie = Unit->getDieMapSlotFor(GV);
     if (!SPDie)
-      SPDie = CreateSubprogramDIE(Unit, SP);
+      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
 
     // Mark as being inlined. This makes this subprogram entry an abstract
     // instance root.
     // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
-    // that it's defined. It probably won't change in the future, but this
-    // could be more elegant.
+    // that it's defined. That probably won't change in the future. However,
+    // this could be more elegant.
     AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
 
     // Keep track of the abstract scope for this function.
@@ -2566,6 +2588,7 @@
     I = DbgConcreteScopeMap.find(GV);
 
   if (I == DbgConcreteScopeMap.end()) {
+    // FIXME: Can this situation actually happen? And if so, should it?
     if (TimePassesIsEnabled)
       DebugTimer->stopTimer();
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=72047&r1=72046&r2=72047&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Mon May 18 17:02:36 2009
@@ -22,6 +22,7 @@
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/FoldingSet.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/UniqueVector.h"
 #include <string>
@@ -163,6 +164,11 @@
   /// attribute.
   DenseMap<const GlobalVariable *, DbgScope *> AbstractInstanceRootMap;
 
+  /// InlinedParamMap - A map keeping track of which parameters are assigned to
+  /// which abstract instance.
+  DenseMap<const GlobalVariable *,
+    SmallSet<const GlobalVariable *, 32> > InlinedParamMap;
+
   /// AbstractInstanceRootList - List of abstract instance roots of inlined
   /// functions. These are subroutine entries that contain a DW_AT_inline
   /// attribute.
@@ -333,7 +339,8 @@
   /// CreateSubprogramDIE - Create new DIE using SP.
   DIE *CreateSubprogramDIE(CompileUnit *DW_Unit,
                            const DISubprogram &SP,
-                           bool IsConstructor = false);
+                           bool IsConstructor = false,
+                           bool IsInlined = false);
 
   /// FindCompileUnit - Get the compile unit for the given descriptor. 
   ///




From evan.cheng at apple.com  Mon May 18 17:14:46 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Mon, 18 May 2009 22:14:46 -0000
Subject: [llvm-commits] [llvm] r72050 -
	/llvm/trunk/docs/TableGenFundamentals.html
Message-ID: <200905182214.n4IMEkcm032623@zion.cs.uiuc.edu>

Author: evancheng
Date: Mon May 18 17:14:45 2009
New Revision: 72050

URL: http://llvm.org/viewvc/llvm-project?rev=72050&view=rev
Log:
Add short descriptions of 'implicit' and 'parallel'.

Modified:
    llvm/trunk/docs/TableGenFundamentals.html

Modified: llvm/trunk/docs/TableGenFundamentals.html
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/TableGenFundamentals.html?rev=72050&r1=72049&r2=72050&view=diff

==============================================================================
--- llvm/trunk/docs/TableGenFundamentals.html (original)
+++ llvm/trunk/docs/TableGenFundamentals.html Mon May 18 17:14:45 2009
@@ -395,6 +395,13 @@
   <dd>a dag value.  The first element is required to be a record definition, the
   remaining elements in the list may be arbitrary other values, including nested
   `<tt>dag</tt>' values.</dd>
+<dt><tt>(implicit a)</tt></dt>
+  <dd>an implicitly defined physical register.  This tells the dag instruction
+  selection emitter the input pattern's extra definitions matches implicit
+  physical register definitions.</dd>
+<dt><tt>(parallel (a), (b))</tt></dt>
+  <dd>a list of dags specifying parallel operations which map to the same
+  instruction.</dd>
 <dt><tt>!strconcat(a, b)</tt></dt>
   <dd>A string value that is the result of concatenating the 'a' and 'b'
   strings.</dd>




From gohman at apple.com  Mon May 18 17:15:25 2009
From: gohman at apple.com (Dan Gohman)
Date: Mon, 18 May 2009 22:15:25 -0000
Subject: [llvm-commits] [llvm] r72051 -
 /llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h
Message-ID: <200905182215.n4IMFQJA032653@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 17:15:25 2009
New Revision: 72051

URL: http://llvm.org/viewvc/llvm-project?rev=72051&view=rev
Log:
Update a comment to reflect changes in the surrounding code.

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=72051&r1=72050&r2=72051&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpressions.h Mon May 18 17:15:25 2009
@@ -220,7 +220,7 @@
       return true;
     }
 
-    // hasComputableLoopEvolution - Commutative expressions have computable loop
+    // hasComputableLoopEvolution - N-ary expressions have computable loop
     // evolutions iff they have at least one operand that varies with the loop,
     // but that all varying operands are computable.
     virtual bool hasComputableLoopEvolution(const Loop *L) const {




From isanbard at gmail.com  Mon May 18 17:33:01 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 22:33:01 -0000
Subject: [llvm-commits] [llvm] r72057 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <200905182233.n4IMX1fi000848@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 17:33:01 2009
New Revision: 72057

URL: http://llvm.org/viewvc/llvm-project?rev=72057&view=rev
Log:
Small code cleanup.

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=72057&r1=72056&r2=72057&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon May 18 17:33:01 2009
@@ -2463,15 +2463,10 @@
           SmallSet<const GlobalVariable *, 32> >::iterator
           IP = InlinedParamMap.find(V);
 
-        if (IP != InlinedParamMap.end()) {
-          SmallSet<const GlobalVariable*, 32> &S = IP->second;
-
-          if (S.count(GV) > 0) {
-            if (TimePassesIsEnabled)
-              DebugTimer->stopTimer();
-            return;
-          }
-
+        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
+          if (TimePassesIsEnabled)
+            DebugTimer->stopTimer();
+          return;
         }
 
         // or GV is an inlined local variable.




From resistor at mac.com  Mon May 18 16:44:42 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 16:44:42 -0500
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
Message-ID: <7296517323013093458502609274541346967-Webmail@me.com>

And here's Take 3, in which I incorporate feedback from Eric, and figure out the difference between AC_SUBST and AC_DEFINE!

--Owen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: config.patch
Type: application/octet-stream
Size: 1314 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/b190eb81/attachment.obj 

From echristo at apple.com  Mon May 18 18:06:14 2009
From: echristo at apple.com (Eric Christopher)
Date: Mon, 18 May 2009 16:06:14 -0700
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <7296517323013093458502609274541346967-Webmail@me.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
	<7296517323013093458502609274541346967-Webmail@me.com>
Message-ID: <26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>


On May 18, 2009, at 2:44 PM, Owen Anderson wrote:


> And here's Take 3, in which I incorporate feedback from Eric, and  
> figure out the difference between AC_SUBST and AC_DEFINE!

Quite a better, couple of questions/comments:

a)  If you're doing this then why the configure flag? Shouldn't  
something just fail to link ultimately if this is wrong?

b) I'm pretty sure that you don't need the [] around the values - at  
least not for AC_SUBST and AC_DEFINE.

c) The bits above for enable_thread should probably look like what I  
sent last night.

d) This should also probably be right under that in the configure.ac  
file. Confusing otherwise :)

-eric


From isanbard at gmail.com  Mon May 18 18:08:55 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Mon, 18 May 2009 23:08:55 -0000
Subject: [llvm-commits] [llvm] r72064 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <200905182308.n4IN8tEn002301@zion.cs.uiuc.edu>

Author: void
Date: Mon May 18 18:08:55 2009
New Revision: 72064

URL: http://llvm.org/viewvc/llvm-project?rev=72064&view=rev
Log:
Don't set the "location" information for inlined functions' variables.

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=72064&r1=72063&r2=72064&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Mon May 18 18:08:55 2009
@@ -129,12 +129,15 @@
 class VISIBILITY_HIDDEN DbgVariable {
   DIVariable Var;                    // Variable Descriptor.
   unsigned FrameIndex;               // Variable frame index.
+  bool InlinedFnVar;                 // Variable for an inlined function.
 public:
-  DbgVariable(DIVariable V, unsigned I) : Var(V), FrameIndex(I)  {}
+  DbgVariable(DIVariable V, unsigned I, bool IFV)
+    : Var(V), FrameIndex(I), InlinedFnVar(IFV)  {}
 
   // Accessors.
-  DIVariable getVariable()  const { return Var; }
+  DIVariable getVariable() const { return Var; }
   unsigned getFrameIndex() const { return FrameIndex; }
+  bool isInlinedFnVar() const { return InlinedFnVar; }
 };
 
 //===----------------------------------------------------------------------===//
@@ -956,10 +959,14 @@
   AddType(Unit, VariableDie, VD.getType());
 
   // Add variable address.
-  MachineLocation Location;
-  Location.set(RI->getFrameRegister(*MF),
-               RI->getFrameIndexOffset(*MF, DV->getFrameIndex()));
-  AddAddress(VariableDie, dwarf::DW_AT_location, Location);
+  if (!DV->isInlinedFnVar()) {
+    // Variables for abstract instances of inlined functions don't get a
+    // location.
+    MachineLocation Location;
+    Location.set(RI->getFrameRegister(*MF),
+                 RI->getFrameIndexOffset(*MF, DV->getFrameIndex()));
+    AddAddress(VariableDie, dwarf::DW_AT_location, Location);
+  }
 
   return VariableDie;
 }
@@ -2434,6 +2441,7 @@
 
   DIDescriptor Desc(GV);
   DbgScope *Scope = NULL;
+  bool InlinedFnVar = false;
 
   if (Desc.getTag() == dwarf::DW_TAG_variable) {
     // GV is a global variable.
@@ -2472,6 +2480,7 @@
         // or GV is an inlined local variable.
         Scope = AI->second;
         InlinedParamMap[V].insert(GV);
+        InlinedFnVar = true;
       } else {
         // or GV is a local variable.
         Scope = getOrCreateScope(V);
@@ -2480,7 +2489,7 @@
   }
 
   assert(Scope && "Unable to find the variable's scope");
-  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex);
+  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
   Scope->AddVariable(DV);
 
   if (TimePassesIsEnabled)




From kremenek at apple.com  Mon May 18 18:20:16 2009
From: kremenek at apple.com (Ted Kremenek)
Date: Mon, 18 May 2009 23:20:16 -0000
Subject: [llvm-commits] [llvm] r72068 - /llvm/tags/checker/checker-0.209/
Message-ID: <200905182320.n4INKGUX002728@zion.cs.uiuc.edu>

Author: kremenek
Date: Mon May 18 18:20:15 2009
New Revision: 72068

URL: http://llvm.org/viewvc/llvm-project?rev=72068&view=rev
Log:
Tagging checker-0.209.

Added:
    llvm/tags/checker/checker-0.209/
      - copied from r72067, llvm/trunk/



From daniel at zuster.org  Mon May 18 18:24:27 2009
From: daniel at zuster.org (Daniel Dunbar)
Date: Mon, 18 May 2009 23:24:27 -0000
Subject: [llvm-commits] [llvm] r72070 - /llvm/trunk/utils/NewNightlyTest.pl
Message-ID: <200905182324.n4INORhv002912@zion.cs.uiuc.edu>

Author: ddunbar
Date: Mon May 18 18:24:26 2009
New Revision: 72070

URL: http://llvm.org/viewvc/llvm-project?rev=72070&view=rev
Log:
Add -submit-aux option to NewNightlyTest.pl
 - If given, the argument will be run using system with the path to the sent
   data. Useful for testing nightlytest server replacements.

Modified:
    llvm/trunk/utils/NewNightlyTest.pl

Modified: llvm/trunk/utils/NewNightlyTest.pl
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/NewNightlyTest.pl?rev=72070&r1=72069&r2=72070&view=diff

==============================================================================
--- llvm/trunk/utils/NewNightlyTest.pl (original)
+++ llvm/trunk/utils/NewNightlyTest.pl Mon May 18 18:24:26 2009
@@ -75,6 +75,9 @@
 #                   this option is not specified it defaults to
 #                   /nightlytest/NightlyTestAccept.php. This is basically
 #                   everything after the www.yourserver.org.
+#  -submit-aux      If specified, an auxiliary script to run in addition to the
+#                   normal submit script. The script will be passed the path to
+#                   the "sentdata.txt" file as its sole argument.
 #  -nosubmit        Do not report the test results back to a submit server.
 #
 # CVSROOT is the CVS repository from which the tree will be checked out,
@@ -128,6 +131,7 @@
 $MAKECMD="make";
 $SUBMITSERVER = "llvm.org";
 $SUBMITSCRIPT = "/nightlytest/NightlyTestAccept.php";
+$SUBMITAUX="";
 $SUBMIT = 1;
 
 while (scalar(@ARGV) and ($_ = $ARGV[0], /^[-+]/)) {
@@ -168,6 +172,7 @@
                              shift; next; }
   if (/^-submit-server/)   { $SUBMITSERVER = "$ARGV[0]"; shift; next; }
   if (/^-submit-script/)   { $SUBMITSCRIPT = "$ARGV[0]"; shift; next; }
+  if (/^-submit-aux/)      { $SUBMITAUX = "$ARGV[0]"; shift; next; }
   if (/^-nosubmit$/)       { $SUBMIT = 0; next; }
   if (/^-nickname$/)       { $nickname = "$ARGV[0]"; shift; next; }
   if (/^-gccpath/)         { $CONFIGUREARGS .=
@@ -477,6 +482,9 @@
     }
     WriteFile "$Prefix-sentdata.txt", $sentdata;
 
+    if (!($SUBMITAUX eq "")) {
+      system "$SUBMITAUX \"$Prefix-sentdata.txt\"";
+    }
 
     return $result;
 }




From resistor at mac.com  Mon May 18 18:29:07 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 18:29:07 -0500
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
	<7296517323013093458502609274541346967-Webmail@me.com>
	<26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>
Message-ID: <148449344759468636145448259517377278506-Webmail@me.com>

 On Monday, May 18, 2009, at 06:06PM, "Eric Christopher" <echristo at apple.com> wrote:
>Quite a better, couple of questions/comments:
>
>a)  If you're doing this then why the configure flag? Shouldn't  
>something just fail to link ultimately if this is wrong?

LLVM can be built in threads-disabled mode, which turns mutexes into no-ops, etc.
The sensible policy seems (to me) to be to try to honor the user's request for
thread-safety if possible.  If not, we print a warning and continue to build
in thread-unsafe mode.

Thinking about it a bit more, maybe we should make it a hard error if the
user explicitly asked for thread-safety on an unsupported target, and
a warning if it's just thread-safety-as-default.

>b) I'm pretty sure that you don't need the [] around the values - at  
>least not for AC_SUBST and AC_DEFINE.

I have no idea.  I just copied and pasted from elsewhere.  :-)

--Owen


From echristo at apple.com  Mon May 18 18:35:33 2009
From: echristo at apple.com (Eric Christopher)
Date: Mon, 18 May 2009 16:35:33 -0700
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <148449344759468636145448259517377278506-Webmail@me.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
	<7296517323013093458502609274541346967-Webmail@me.com>
	<26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>
	<148449344759468636145448259517377278506-Webmail@me.com>
Message-ID: <2DEEB59B-0AAB-485A-96FB-D2E75033D8EC@apple.com>


On May 18, 2009, at 4:29 PM, Owen Anderson wrote:


> On Monday, May 18, 2009, at 06:06PM, "Eric Christopher" <echristo at apple.com 
> > wrote:
>
>> Quite a better, couple of questions/comments:
>>
>>
>>
>> a)  If you're doing this then why the configure flag? Shouldn't
>>
>> something just fail to link ultimately if this is wrong?
>>
>
> LLVM can be built in threads-disabled mode, which turns mutexes into  
> no-ops, etc.
> The sensible policy seems (to me) to be to try to honor the user's  
> request for
> thread-safety if possible.  If not, we print a warning and continue  
> to build
> in thread-unsafe mode.
>
> Thinking about it a bit more, maybe we should make it a hard error  
> if the
> user explicitly asked for thread-safety on an unsupported target, and
> a warning if it's just thread-safety-as-default.
>

Don't know how you'd guess target though since llvm is multi-arch at  
build time.

Honestly I think you get what you ask for it might be much more simple  
to
just mark which targets work and then have a preprocessor flag that sees
if it's set and the native platform.

*shrug* :)

>
>> b) I'm pretty sure that you don't need the [] around the values - at
>>
>> least not for AC_SUBST and AC_DEFINE.
>>
>
> I have no idea.  I just copied and pasted from elsewhere.  :-)
>

Aaah. I don't believe you need it.

-eric



From resistor at mac.com  Mon May 18 18:46:37 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 18:46:37 -0500
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <2DEEB59B-0AAB-485A-96FB-D2E75033D8EC@apple.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
	<7296517323013093458502609274541346967-Webmail@me.com>
	<26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>
	<148449344759468636145448259517377278506-Webmail@me.com>
	<2DEEB59B-0AAB-485A-96FB-D2E75033D8EC@apple.com>
Message-ID: <79026032486662333965431321961251389077-Webmail@me.com>

 
On Monday, May 18, 2009, at 06:35PM, "Eric Christopher" <echristo at apple.com> wrote:
>Don't know how you'd guess target though since llvm is multi-arch at  
>build time.
>
>Honestly I think you get what you ask for it might be much more simple  
>to
>just mark which targets work and then have a preprocessor flag that sees
>if it's set and the native platform.
>
>*shrug* :)
>

I'm adding support for threading LLVM itself, nothing to do with the code that it generates.  For LLVM itself to be thread-safe, we need to have access to certain atomic intrinsics on the target for which LLVM is being built.

I think the current solution handles that, and should work just fine on a cross-compiled system now that I've switched to link-testing rather than run-testing.  Right?

--Owen


From echristo at apple.com  Mon May 18 18:52:26 2009
From: echristo at apple.com (Eric Christopher)
Date: Mon, 18 May 2009 16:52:26 -0700
Subject: [llvm-commits] [PATCH] Detect GCC atomic builtins in configure
In-Reply-To: <79026032486662333965431321961251389077-Webmail@me.com>
References: <EAB97055-04A1-46F2-9EE1-8E67FB92FAD0@mac.com>
	<7FDB65CF-9461-4DA6-B206-FD2F5EEFA615@mac.com>
	<39E5492E-0BA1-465F-8A42-BAA459433D4B@mac.com>
	<E095B5F5-6BF4-4E00-BA70-E5B1FAE4A38B@apple.com>
	<73637BEF-61A8-4BA9-88FB-985D955108FF@mac.com>
	<7296517323013093458502609274541346967-Webmail@me.com>
	<26BF1D56-5396-4B37-9D07-C51DE955D79A@apple.com>
	<148449344759468636145448259517377278506-Webmail@me.com>
	<2DEEB59B-0AAB-485A-96FB-D2E75033D8EC@apple.com>
	<79026032486662333965431321961251389077-Webmail@me.com>
Message-ID: <C1A2A2B7-5552-433E-B969-80744F55AD6D@apple.com>

>
> I think the current solution handles that, and should work just fine  
> on a cross-compiled system now that I've switched to link-testing  
> rather than run-testing.  Right?

link should be fine. I'd probably do this as a test on the target and  
a switch statement, but that's just me :)

patch looks fine now.

-eric


From resistor at mac.com  Mon May 18 18:58:52 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 23:58:52 -0000
Subject: [llvm-commits] [llvm] r72075 - in /llvm/trunk:
	autoconf/configure.ac configure
Message-ID: <200905182358.n4INwrhS004191@zion.cs.uiuc.edu>

Author: resistor
Date: Mon May 18 18:58:51 2009
New Revision: 72075

URL: http://llvm.org/viewvc/llvm-project?rev=72075&view=rev
Log:
Test for the presence of GCC atomic builtins at configure time.  If not found,
disable building LLVM in thread-safe mode and print a nice warning.

Regenerate configure for these changes.

Modified:
    llvm/trunk/autoconf/configure.ac
    llvm/trunk/configure

Modified: llvm/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/autoconf/configure.ac?rev=72075&r1=72074&r2=72075&view=diff

==============================================================================
--- llvm/trunk/autoconf/configure.ac (original)
+++ llvm/trunk/autoconf/configure.ac Mon May 18 18:58:51 2009
@@ -372,7 +372,6 @@
   default) AC_SUBST(ENABLE_THREADS,[1]) ;;
   *) AC_MSG_ERROR([Invalid setting for --enable-threads. Use "yes" or "no"]) ;;
 esac
-AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
 
 dnl Allow building without position independent code
 AC_ARG_ENABLE(pic,
@@ -897,6 +896,26 @@
   fi
 fi
 
+dnl atomic builtins are required for threading support.
+if test "$ENABLE_THREADS" -eq 1 ; then
+  AC_MSG_CHECKING(for GCC atomic builtins)
+  AC_LINK_IFELSE(
+	AC_LANG_SOURCE(
+          [[int main() { 
+              volatile unsigned long val = 1;
+              __sync_synchronize();
+              __sync_val_compare_and_swap(&val, 1, 0);
+              return 0;
+            }
+          ]]),
+        AC_MSG_RESULT(yes),
+        AC_MSG_RESULT(no)
+        AC_SUBST(ENABLE_THREADS, [0])
+        AC_MSG_WARN([Threading support disabled because atomic builtins are missing]))
+fi
+
+AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
+
 dnl===-----------------------------------------------------------------------===
 dnl===
 dnl=== SECTION 9: Additional checks, variables, etc.

Modified: llvm/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/configure?rev=72075&r1=72074&r2=72075&view=diff

==============================================================================
--- llvm/trunk/configure (original)
+++ llvm/trunk/configure Mon May 18 18:58:51 2009
@@ -4870,11 +4870,6 @@
    { (exit 1); exit 1; }; } ;;
 esac
 
-cat >>confdefs.h <<_ACEOF
-#define ENABLE_THREADS $ENABLE_THREADS
-_ACEOF
-
-
 # Check whether --enable-pic was given.
 if test "${enable_pic+set}" = set; then
   enableval=$enable_pic;
@@ -10575,7 +10570,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 10578 "configure"
+#line 10573 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12719,7 +12714,7 @@
   ;;
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 12722 "configure"' > conftest.$ac_ext
+  echo '#line 12717 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -14437,11 +14432,11 @@
    -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:14440: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14435: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:14444: \$? = $ac_status" >&5
+   echo "$as_me:14439: \$? = $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.
@@ -14705,11 +14700,11 @@
    -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:14708: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14703: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:14712: \$? = $ac_status" >&5
+   echo "$as_me:14707: \$? = $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.
@@ -14809,11 +14804,11 @@
    -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:14812: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14807: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:14816: \$? = $ac_status" >&5
+   echo "$as_me:14811: \$? = $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
@@ -17261,7 +17256,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 17264 "configure"
+#line 17259 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -17361,7 +17356,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 17364 "configure"
+#line 17359 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -19729,11 +19724,11 @@
    -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:19732: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:19727: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:19736: \$? = $ac_status" >&5
+   echo "$as_me:19731: \$? = $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.
@@ -19833,11 +19828,11 @@
    -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:19836: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:19831: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:19840: \$? = $ac_status" >&5
+   echo "$as_me:19835: \$? = $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
@@ -21403,11 +21398,11 @@
    -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:21406: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:21401: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:21410: \$? = $ac_status" >&5
+   echo "$as_me:21405: \$? = $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.
@@ -21507,11 +21502,11 @@
    -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:21510: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:21505: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:21514: \$? = $ac_status" >&5
+   echo "$as_me:21509: \$? = $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
@@ -23742,11 +23737,11 @@
    -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:23745: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:23740: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:23749: \$? = $ac_status" >&5
+   echo "$as_me:23744: \$? = $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.
@@ -24010,11 +24005,11 @@
    -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:24013: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:24008: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:24017: \$? = $ac_status" >&5
+   echo "$as_me:24012: \$? = $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.
@@ -24114,11 +24109,11 @@
    -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:24117: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:24112: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:24121: \$? = $ac_status" >&5
+   echo "$as_me:24116: \$? = $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
@@ -33607,6 +33602,81 @@
   fi
 fi
 
+if test "$ENABLE_THREADS" -eq 1 ; then
+  { echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
+echo $ECHO_N "checking for GCC atomic builtins... $ECHO_C" >&6; }
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+int main() {
+              volatile unsigned long val = 1;
+              __sync_synchronize();
+              __sync_val_compare_and_swap(&val, 1, 0);
+              return 0;
+            }
+
+_ACEOF
+rm -f conftest.$ac_objext conftest$ac_exeext
+if { (ac_try="$ac_link"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_link") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest$ac_exeext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  { echo "$as_me:$LINENO: result: yes" >&5
+echo "${ECHO_T}yes" >&6; }
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	{ echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+        ENABLE_THREADS=0
+
+        { echo "$as_me:$LINENO: WARNING: Threading support disabled because atomic builtins are missing" >&5
+echo "$as_me: WARNING: Threading support disabled because atomic builtins are missing" >&2;}
+fi
+
+rm -f core conftest.err conftest.$ac_objext \
+      conftest$ac_exeext conftest.$ac_ext
+fi
+
+
+cat >>confdefs.h <<_ACEOF
+#define ENABLE_THREADS $ENABLE_THREADS
+_ACEOF
+
+
 
 
 for ac_func in __dso_handle




From resistor at mac.com  Mon May 18 19:08:29 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 00:08:29 -0000
Subject: [llvm-commits] [llvm] r72077 - in /llvm/trunk: LICENSE.TXT
	include/llvm/System/Atomic.h
Message-ID: <200905190008.n4J08TuW004524@zion.cs.uiuc.edu>

Author: resistor
Date: Mon May 18 19:08:29 2009
New Revision: 72077

URL: http://llvm.org/viewvc/llvm-project?rev=72077&view=rev
Log:
Now that we have atomics support properly detected by configure,
use it to implement Atomic.h.

This expunges the code previously imported from libatomic_ops.

Modified:
    llvm/trunk/LICENSE.TXT
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/LICENSE.TXT
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/LICENSE.TXT?rev=72077&r1=72076&r2=72077&view=diff

==============================================================================
--- llvm/trunk/LICENSE.TXT (original)
+++ llvm/trunk/LICENSE.TXT Mon May 18 19:08:29 2009
@@ -67,4 +67,3 @@
                     llvm/projects/sample/autoconf
 CellSPU backend     llvm/lib/Target/CellSPU/README.txt
 Google Test         llvm/utils/unittest/googletest
-Atomics Library     llvm/include/llvm/System/Atomic.h
\ No newline at end of file

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72077&r1=72076&r2=72077&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Mon May 18 19:08:29 2009
@@ -9,197 +9,64 @@
 //
 // This file declares the llvm::sys atomic operations.
 //
-// Portions of this file use code from libatomic_ops, for which the following 
-// license applies:
-//
-// Copyright (c) 2003 by Hewlett-Packard Company.  All rights reserved.
-// 
-// Permission is hereby granted, free of charge, to any person obtaining a copy
-// of this software and associated documentation files (the "Software"), to deal
-// in the Software without restriction, including without limitation the rights
-// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-// copies of the Software, and to permit persons to whom the Software is
-// furnished to do so, subject to the following conditions:
-// 
-// The above copyright notice and this permission notice shall be included in
-// all copies or substantial portions of the Software.
-// 
-// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-// SOFTWARE.
-//
 //===----------------------------------------------------------------------===//
 
 #ifndef LLVM_SYSTEM_ATOMIC_H
 #define LLVM_SYSTEM_ATOMIC_H
 
-#include <stdint.h>
-
-#if defined(_HPUX_SOURCE) && defined(__ia64)
-#include <machine/sys/inline.h>
-#elif defined(_MSC_VER)
+#if defined(_MSC_VER)
 #include <windows.h>
-#endif // defined(_HPUX_SOURCE) && defined(__ia64)
+#endif
 
 
 namespace llvm {
   namespace sys {
     
-    inline void CompilerFence() {
-#if defined(__GNUC__) && !defined(__INTEL_COMPILER)
+    inline void MemoryFence() {
+#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
+#  if defined(__GNUC__)
       __asm__ __volatile__("" : : : "memory");
-#elif defined(_MSC_VER)
+#  elif defined(_MSC_VER)
       __asm { };
-#elif defined(__INTEL_COMPILER)
-      __memory_barrier(); /* Too strong? IA64-only? */
+#  else
+#    error No memory fence implementation for your platform!
+#  endif
 #else
-    /* We conjecture that the following usually gives us the right 	*/
-    /* semantics or an error.						*/
-      asm("");
-#endif // defined(__GNUC__) && !defined(__INTEL_COMPILER)
+#  if defined(__GNUC__)
+      __sync_synchronize();
+#  elif defined(_MSC_VER)
+      MemoryBarrier();
+#  else
+#    error No memory fence implementation for your platform!
+#  endif
+#endif
 }
 
 #if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
-    inline void MemoryFence() {
-      CompilerFence();
-    }
-
-    typedef uint32_t cas_flag;
+    typedef unsigned long cas_flag;
     inline cas_flag CompareAndSwap(cas_flag* dest, cas_flag exc, cas_flag c) {
       cas_flag result = *dest;
       if (result == c)
         *dest = exc;
       return result;
     }
-
 #elif defined(__GNUC__)
-
-    inline void MemoryFence() {
-#  if defined(__i386__) || defined(__x86_64__)
-#    if defined(__SSE2__)
-      __asm__ __volatile__("mfence" : : : "memory");
-#    else
-      unsigned char dummy = 0;
-      volatile unsigned char* addr = &dummy;
-      unsigned char oldval;
-      __asm __ __volatile__("xchgb %0, %1" : "=r"(oldval),
-        "=m"(*addr), "0"(0xff), "m"(*addr) : "memory");
-#    endif // defined(__SSE2__)
-#  elif defined(__ia64__)
-      __asm__ __volatile__("mf" : : : "memory");
-# elif defined(__alpha__)
-      __asm__ __volatile__("mb" : : : "memory");
-# elif defined(__sparc__)
-      __asm__ __volatile__("membar	#StoreStore | #LoadStore | #LoadLoad | #StoreLoad");
-# elif defined(__powerpc__) || defined(__ppc__)
-      __asm__ __volatile__("sync" : : : "memory");
-# elif defined(__arm__)
-      __asm__ __volatile__ ("mcr	p15, 0, r0, c7, c10, 5	@ dmb");
-# endif
-    } // defined(__i386__) || defined(__x86_64__)
-    
     typedef unsigned long cas_flag;
     inline cas_flag CompareAndSwap(cas_flag* ptr,
                                    cas_flag new_value,
                                    cas_flag old_value) {
-      cas_flag prev;
-#  if defined(__i386__) || defined(__x86_64__)
-      __asm__ __volatile__("lock; cmpxchgl %1,%2"
-                           : "=a" (prev)
-                           : "q" (new_value), "m" (*ptr), "0" (old_value)
-                           : "memory");
-#  elif defined(__ia64__)
-      MemoryFence();
-#  if defined(_ILP32)
-      __asm__("zxt4 %1=%1": "=r"(prev) : "0"(prev));
-      __asm__ __volatile__("addp4 %1=0,%1;;\n"
-        "mov ar.ccv=%[old] ;; cmpxchg 4"
-        ".acq %0=[%1],%[new_val],ar.ccv"
-        : "=r"(prev) "1"(addr),
-        : "=r"(addr), [new_value]"r"(new_value), [old_value]"r"(old_value)
-        : "memory");
-#  else
-      __asm__ __volatile__(
-        "mov ar.ccv=%[old] ;; cmpxchg 8"
-        ".acq %0=[%1],%[new_val],ar.ccv"
-        : "=r"(prev)
-        : "r"(ptr), [new_value]"r"(new_value), 
-        [old_value]"r"(old_value)
-        : "memory");
-#  endif // defined(_ILP32)
-#  elif defined(__alpha__)
-      cas_flag was_equal;
-      __asm__ __volatile__(
-        "1:     ldq_l %0,%1\n"
-        "       cmpeq %0,%4,%2\n"
-        "	     mov %3,%0\n"
-        "       beq %2,2f\n"
-        "       stq_c %0,%1\n"
-        "       beq %0,1b\n"
-        "2:\n"
-        :"=&r" (prev), "=m" (*ptr), "=&r" (was_equal)
-        : "r" (new_value), "Ir" (old_value)
-        :"memory");
-#elif defined(__sparc__)
-#error No CAS implementation for SPARC yet.
-#elif defined(__powerpc__) || defined(__ppc__)
-      int result = 0;
-      __asm__ __volatile__(
-        "1:lwarx %0,0,%2\n"   /* load and reserve              */
-        "cmpw %0, %4\n"      /* if load is not equal to 	*/
-        "bne 2f\n"            /*   old, fail			*/
-        "stwcx. %3,0,%2\n"    /* else store conditional         */
-        "bne- 1b\n"           /* retry if lost reservation      */
-        "li %1,1\n"	     /* result = 1;			*/
-        "2:\n"
-        : "=&r"(prev), "=&r"(result)
-        : "r"(ptr), "r"(new_value), "r"(old_value), "1"(result)
-        : "memory", "cc");
-#elif defined(__arm__)
-      int result;
-      __asm__ __volatile__ (
-        "\n"
-        "0:\t"
-        "ldr     %1,[%2] \n\t"
-        "mov     %0,#0 \n\t"
-        "cmp     %1,%4 \n\t"
-        "bne     1f \n\t"
-        "swp     %0,%3,[%2] \n\t"
-        "cmp     %1,%0 \n\t"
-        "swpne   %1,%0,[%2] \n\t"
-        "bne     0b \n\t"
-        "mov     %0,#1 \n"
-        "1:\n\t"
-        ""
-        : "=&r"(result), "=&r"(prev) 
-        : "r" ptr), "r" (new_value), "r" (old_value) 
-        : "cc", "memory");
-#endif // defined(__i386__)
-      return prev;
+      return __sync_val_compare_and_swap(ptr, old_value, new_value);
     }
-
 #elif defined(_MSC_VER) && _M_IX86 > 400
-    inline void MemoryFence() {
-      LONG dummy = 0;
-      InterlockedExchanged((LONG volatile *)&dummy, (LONG)0);
-    }
-    
-    typedef DWORD cas_flag;
+    typedef LONG cas_flag;
     inline cas_flag CompareAndSwap(cas_flag* ptr,
                                    cas_flag new_value,
                                    cas_flag old_value) {
-      /* FIXME - This is nearly useless on win64.			*/
-      /* Use InterlockedCompareExchange64 for win64?	*/
-      return InterlockedCompareExchange((DWORD volatile *)addr,
-                                        (DWORD)new_value, (DWORD) old_value)
+      return InterlockedCompareExchange(addr, new_value, old_value);
     }
 #else
-#error No atomics implementation found for your platform.
-#endif // !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
+#  error No compare-and-swap implementation for your platform!
+#endif
 
   }
 }




From echristo at apple.com  Mon May 18 19:16:07 2009
From: echristo at apple.com (Eric Christopher)
Date: Mon, 18 May 2009 17:16:07 -0700
Subject: [llvm-commits] [llvm] r72077 - in /llvm/trunk: LICENSE.TXT
	include/llvm/System/Atomic.h
In-Reply-To: <200905190008.n4J08TuW004524@zion.cs.uiuc.edu>
References: <200905190008.n4J08TuW004524@zion.cs.uiuc.edu>
Message-ID: <70ABE7D9-1463-4CF9-B08D-183B3F601C40@apple.com>


On May 18, 2009, at 5:08 PM, Owen Anderson wrote:


> +#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
>

Since it's going to be defined either way how about

#if ENABLE_THREADS == 0

or perhaps even better let DCE handle it:

if (!ENABLE_THREADS)

:)

-eric


From mrs at apple.com  Mon May 18 19:18:14 2009
From: mrs at apple.com (Mike Stump)
Date: Tue, 19 May 2009 00:18:14 -0000
Subject: [llvm-commits] [llvm] r72078 - in /llvm/trunk/lib/CodeGen:
 AsmPrinter/CMakeLists.txt CMakeLists.txt
Message-ID: <200905190018.n4J0IGIo004888@zion.cs.uiuc.edu>

Author: mrs
Date: Mon May 18 19:18:14 2009
New Revision: 72078

URL: http://llvm.org/viewvc/llvm-project?rev=72078&view=rev
Log:
Fix cmake builds.

Modified:
    llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt
    llvm/trunk/lib/CodeGen/CMakeLists.txt

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt?rev=72078&r1=72077&r2=72078&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/CMakeLists.txt Mon May 18 19:18:14 2009
@@ -1,5 +1,10 @@
 add_llvm_library(LLVMAsmPrinter
   AsmPrinter.cpp
+  DIE.cpp
+  DwarfDebug.cpp
+  DwarfException.cpp
+  DwarfLabel.cpp
+  DwarfPrinter.cpp
   DwarfWriter.cpp
   OcamlGCPrinter.cpp
   )

Modified: llvm/trunk/lib/CodeGen/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/CMakeLists.txt?rev=72078&r1=72077&r2=72078&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/CMakeLists.txt (original)
+++ llvm/trunk/lib/CodeGen/CMakeLists.txt Mon May 18 19:18:14 2009
@@ -49,6 +49,7 @@
   ShadowStackGC.cpp
   ShrinkWrapping.cpp
   SimpleRegisterCoalescing.cpp
+  Spiller.cpp
   StackProtector.cpp
   StackSlotColoring.cpp
   StrongPHIElimination.cpp




From dalej at apple.com  Mon May 18 19:46:45 2009
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 19 May 2009 00:46:45 -0000
Subject: [llvm-commits] [llvm] r72083 -
	/llvm/trunk/lib/Target/CBackend/CBackend.cpp
Message-ID: <200905190046.n4J0kkBb005953@zion.cs.uiuc.edu>

Author: johannes
Date: Mon May 18 19:46:42 2009
New Revision: 72083

URL: http://llvm.org/viewvc/llvm-project?rev=72083&view=rev
Log:
Spacing fix.


Modified:
    llvm/trunk/lib/Target/CBackend/CBackend.cpp

Modified: llvm/trunk/lib/Target/CBackend/CBackend.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/CBackend/CBackend.cpp?rev=72083&r1=72082&r2=72083&view=diff

==============================================================================
--- llvm/trunk/lib/Target/CBackend/CBackend.cpp (original)
+++ llvm/trunk/lib/Target/CBackend/CBackend.cpp Mon May 18 19:46:42 2009
@@ -1127,7 +1127,7 @@
         Out << CI->getZExtValue() << 'u';
       else
         Out << CI->getSExtValue();
-       Out << ')';
+      Out << ')';
     }
     return;
   } 




From resistor at mac.com  Mon May 18 20:02:27 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 01:02:27 -0000
Subject: [llvm-commits] [llvm] r72086 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905190102.n4J12RRh006490@zion.cs.uiuc.edu>

Author: resistor
Date: Mon May 18 20:02:27 2009
New Revision: 72086

URL: http://llvm.org/viewvc/llvm-project?rev=72086&view=rev
Log:
Template CompareAndSwap function.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72086&r1=72085&r2=72086&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Mon May 18 20:02:27 2009
@@ -44,24 +44,27 @@
 
 #if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
     typedef unsigned long cas_flag;
-    inline cas_flag CompareAndSwap(cas_flag* dest, cas_flag exc, cas_flag c) {
-      cas_flag result = *dest;
+    template<typename T>
+    inline T CompareAndSwap(volatile T* dest,
+			    T exc, T c) {
+      T result = *dest;
       if (result == c)
         *dest = exc;
       return result;
     }
 #elif defined(__GNUC__)
     typedef unsigned long cas_flag;
-    inline cas_flag CompareAndSwap(cas_flag* ptr,
-                                   cas_flag new_value,
-                                   cas_flag old_value) {
+    template<typename T>
+    inline T CompareAndSwap(volatile T* ptr,
+			    T new_value,
+			    T old_value) {
       return __sync_val_compare_and_swap(ptr, old_value, new_value);
     }
-#elif defined(_MSC_VER) && _M_IX86 > 400
+#elif defined(_MSC_VER)
     typedef LONG cas_flag;
-    inline cas_flag CompareAndSwap(cas_flag* ptr,
-                                   cas_flag new_value,
-                                   cas_flag old_value) {
+    inline T CompareAndSwap(volatile T* ptr,
+			    T new_value,
+			    T old_value) {
       return InterlockedCompareExchange(addr, new_value, old_value);
     }
 #else




From resistor at mac.com  Mon May 18 20:07:40 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 01:07:40 -0000
Subject: [llvm-commits] [llvm] r72087 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>

Author: resistor
Date: Mon May 18 20:07:40 2009
New Revision: 72087

URL: http://llvm.org/viewvc/llvm-project?rev=72087&view=rev
Log:
Fix up the Windows portion of Atomic.h.  This is untested, but it is my best understanding of what should work.
I'd be much obliged if someone on MSVC++ could try this out and let me know if it works.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72087&r1=72086&r2=72087&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Mon May 18 20:07:40 2009
@@ -62,11 +62,24 @@
     }
 #elif defined(_MSC_VER)
     typedef LONG cas_flag;
+    template<typename T>
     inline T CompareAndSwap(volatile T* ptr,
 			    T new_value,
 			    T old_value) {
-      return InterlockedCompareExchange(addr, new_value, old_value);
+      if (sizeof(T) == 4)
+	return InterlockedCompareExchange(ptr, new_value, old_value);
+      else
+	return InterlockedCompareExchange64(ptr, new_value, old_value);
     }
+    
+    template<typename T>
+    inline T* CompareAndSwap<T*>(volatile T** ptr,
+				 T* new_value,
+				 T* old_value) {
+      return InterlockedCompareExchangePtr(ptr, new_value, old_value);
+    }
+
+
 #else
 #  error No compare-and-swap implementation for your platform!
 #endif




From resistor at mac.com  Mon May 18 20:08:36 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 01:08:36 -0000
Subject: [llvm-commits] [llvm] r72088 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905190108.n4J18afK006720@zion.cs.uiuc.edu>

Author: resistor
Date: Mon May 18 20:08:36 2009
New Revision: 72088

URL: http://llvm.org/viewvc/llvm-project?rev=72088&view=rev
Log:
Make using an unsupported CAS size a hard error on Windows.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72088&r1=72087&r2=72088&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Mon May 18 20:08:36 2009
@@ -68,8 +68,10 @@
 			    T old_value) {
       if (sizeof(T) == 4)
 	return InterlockedCompareExchange(ptr, new_value, old_value);
-      else
+      else if (sizeof(T) == 8)
 	return InterlockedCompareExchange64(ptr, new_value, old_value);
+      else
+	assert(0 && "Unsupported compare-and-swap size!");
     }
     
     template<typename T>




From resistor at mac.com  Mon May 18 20:25:38 2009
From: resistor at mac.com (Owen Anderson)
Date: Mon, 18 May 2009 20:25:38 -0500
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
Message-ID: <5464957333051772950217202212009693918-Webmail@me.com>

So, now that we have the basic infrastructure more or less in place, here's another attempt at making a thread-safe version of ManagedStatic based on double-checked locking with explicit fences.  It also proved necessary to provide atomic update facilities for the global StaticList as well.

Please see the included block comment for more details on exactly what safety guarantees this code provides.

--Owen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ManagedStatic.diff
Type: application/octet-stream
Size: 4719 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090518/0f336639/attachment.obj 

From gohman at apple.com  Mon May 18 20:52:46 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 01:52:46 -0000
Subject: [llvm-commits] [llvm] r72091 -
	/llvm/trunk/include/llvm/Analysis/LoopInfo.h
Message-ID: <200905190152.n4J1qk2x008083@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 20:52:46 2009
New Revision: 72091

URL: http://llvm.org/viewvc/llvm-project?rev=72091&view=rev
Log:
Add some comments noting how IndVarSimplify facilitates
LoopInfo functionality.

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

Modified: llvm/trunk/include/llvm/Analysis/LoopInfo.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/LoopInfo.h?rev=72091&r1=72090&r2=72091&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/LoopInfo.h (original)
+++ llvm/trunk/include/llvm/Analysis/LoopInfo.h Mon May 18 20:52:46 2009
@@ -351,6 +351,9 @@
   /// by one each time through the loop.  If so, return the phi node that
   /// corresponds to it.
   ///
+  /// The IndVarSimplify pass transforms loops to have a canonical induction
+  /// variable.
+  ///
   inline PHINode *getCanonicalInductionVariable() const {
     BlockT *H = getHeader();
 
@@ -406,6 +409,9 @@
   /// of the loop executes N-1 times.  If the trip-count cannot be determined,
   /// this returns null.
   ///
+  /// The IndVarSimplify pass transforms loops to have a form that this
+  /// function easily understands.
+  ///
   inline Value *getTripCount() const {
     // Canonical loops will end with a 'cmp ne I, V', where I is the incremented
     // canonical induction variable and V is the trip count of the loop.




From gohman at apple.com  Mon May 18 21:15:55 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 02:15:55 -0000
Subject: [llvm-commits] [llvm] r72093 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h
 include/llvm/Analysis/ScalarEvolutionExpander.h
 lib/Analysis/ScalarEvolution.cpp lib/Analysis/ScalarEvolutionExpander.cpp
 lib/Transforms/Scalar/IndVarSimplify.cpp
 lib/Transforms/Scalar/LoopStrengthReduce.cpp
 lib/Transforms/Utils/AddrModeMatcher.cpp
 test/Transforms/IndVarSimplify/preserve-gep.ll
 test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll
Message-ID: <200905190215.n4J2FuCD009024@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 21:15:55 2009
New Revision: 72093

URL: http://llvm.org/viewvc/llvm-project?rev=72093&view=rev
Log:
Teach SCEVExpander to expand arithmetic involving pointers into GEP
instructions. It attempts to create high-level multi-operand GEPs,
though in cases where this isn't possible it falls back to casting
the pointer to i8* and emitting a GEP with that. Using GEP instructions
instead of ptrtoint+arithmetic+inttoptr helps pointer analyses that
don't use ScalarEvolution, such as BasicAliasAnalysis.

Also, make the AddrModeMatcher more aggressive in handling GEPs.
Previously it assumed that operand 0 of a GEP would require a register
in almost all cases. It now does extra checking and can do more
matching if operand 0 of the GEP is foldable. This fixes a problem
that was exposed by SCEVExpander using GEPs.

Added:
    llvm/trunk/test/Transforms/IndVarSimplify/preserve-gep.ll
Modified:
    llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
    llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h
    llvm/trunk/lib/Analysis/ScalarEvolution.cpp
    llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
    llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
    llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
    llvm/trunk/lib/Transforms/Utils/AddrModeMatcher.cpp
    llvm/trunk/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Mon May 18 21:15:55 2009
@@ -213,6 +213,7 @@
   ///
   class ScalarEvolution : public FunctionPass {
     friend class SCEVCallbackVH;
+    friend class SCEVExpander;
 
     /// F - The function we are analyzing.
     ///

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolutionExpander.h Mon May 18 21:15:55 2009
@@ -28,7 +28,6 @@
   /// memory.
   struct SCEVExpander : public SCEVVisitor<SCEVExpander, Value*> {
     ScalarEvolution &SE;
-    LoopInfo &LI;
     std::map<SCEVHandle, Value*> InsertedExpressions;
     std::set<Value*> InsertedValues;
 
@@ -36,10 +35,8 @@
 
     friend struct SCEVVisitor<SCEVExpander, Value*>;
   public:
-    SCEVExpander(ScalarEvolution &se, LoopInfo &li)
-      : SE(se), LI(li) {}
-
-    LoopInfo &getLoopInfo() const { return LI; }
+    explicit SCEVExpander(ScalarEvolution &se)
+      : SE(se) {}
 
     /// clear - Erase the contents of the InsertedExpressions map so that users
     /// trying to expand the same expression into multiple BasicBlocks or
@@ -83,8 +80,9 @@
 
     /// expandCodeFor - Insert code to directly compute the specified SCEV
     /// expression into the program.  The inserted code is inserted into the
-    /// SCEVExpander's current insertion point.
-    Value *expandCodeFor(SCEVHandle SH, const Type *Ty);
+    /// SCEVExpander's current insertion point. If a type is specified, the
+    /// result will be expanded to have that type, with a cast if necessary.
+    Value *expandCodeFor(SCEVHandle SH, const Type *Ty = 0);
 
     /// expandCodeFor - Insert code to directly compute the specified SCEV
     /// expression into the program.  The inserted code is inserted into the
@@ -110,6 +108,11 @@
                        Value *RHS, BasicBlock::iterator InsertPt);
 
   private:
+    /// expandAddToGEP - Expand a SCEVAddExpr with a pointer type into a GEP
+    /// instead of using ptrtoint+arithmetic+inttoptr.
+    Value *expandAddToGEP(const SCEVAddExpr *S, const PointerType *PTy,
+                          const Type *Ty, Value *V);
+
     Value *expand(const SCEV *S);
 
     Value *visitConstant(const SCEVConstant *S) {

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Mon May 18 21:15:55 2009
@@ -456,6 +456,13 @@
       if (const SCEVUnknown *LU = dyn_cast<SCEVUnknown>(LHS)) {
         const SCEVUnknown *RU = cast<SCEVUnknown>(RHS);
 
+        // Order pointer values after integer values. This helps SCEVExpander
+        // form GEPs.
+        if (isa<PointerType>(LU->getType()) && !isa<PointerType>(RU->getType()))
+          return false;
+        if (isa<PointerType>(RU->getType()) && !isa<PointerType>(LU->getType()))
+          return true;
+
         // Compare getValueID values.
         if (LU->getValue()->getValueID() != RU->getValue()->getValueID())
           return LU->getValue()->getValueID() < RU->getValue()->getValueID();

Modified: llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp Mon May 18 21:15:55 2009
@@ -15,6 +15,7 @@
 
 #include "llvm/Analysis/ScalarEvolutionExpander.h"
 #include "llvm/Analysis/LoopInfo.h"
+#include "llvm/Target/TargetData.h"
 using namespace llvm;
 
 /// InsertCastOfTo - Insert a cast of V to the specified type, doing what
@@ -130,10 +131,9 @@
     BasicBlock::iterator IP = InsertPt;
     --IP;
     for (; ScanLimit; --IP, --ScanLimit) {
-      if (BinaryOperator *BinOp = dyn_cast<BinaryOperator>(IP))
-        if (BinOp->getOpcode() == Opcode && BinOp->getOperand(0) == LHS &&
-            BinOp->getOperand(1) == RHS)
-          return BinOp;
+      if (IP->getOpcode() == (unsigned)Opcode && IP->getOperand(0) == LHS &&
+          IP->getOperand(1) == RHS)
+        return IP;
       if (IP == BlockBegin) break;
     }
   }
@@ -144,9 +144,156 @@
   return BO;
 }
 
+/// expandAddToGEP - Expand a SCEVAddExpr with a pointer type into a GEP
+/// instead of using ptrtoint+arithmetic+inttoptr.
+Value *SCEVExpander::expandAddToGEP(const SCEVAddExpr *S,
+                                    const PointerType *PTy,
+                                    const Type *Ty,
+                                    Value *V) {
+  const Type *ElTy = PTy->getElementType();
+  SmallVector<Value *, 4> GepIndices;
+  std::vector<SCEVHandle> Ops = S->getOperands();
+  bool AnyNonZeroIndices = false;
+  Ops.pop_back();
+
+  // Decend down the pointer's type and attempt to convert the other
+  // operands into GEP indices, at each level. The first index in a GEP
+  // indexes into the array implied by the pointer operand; the rest of
+  // the indices index into the element or field type selected by the
+  // preceding index.
+  for (;;) {
+    APInt ElSize = APInt(SE.getTypeSizeInBits(Ty),
+                         ElTy->isSized() ?  SE.TD->getTypeAllocSize(ElTy) : 0);
+    std::vector<SCEVHandle> NewOps;
+    std::vector<SCEVHandle> ScaledOps;
+    for (unsigned i = 0, e = Ops.size(); i != e; ++i) {
+      if (ElSize != 0) {
+        if (const SCEVConstant *C = dyn_cast<SCEVConstant>(Ops[i]))
+          if (!C->getValue()->getValue().srem(ElSize)) {
+            ConstantInt *CI =
+              ConstantInt::get(C->getValue()->getValue().sdiv(ElSize));
+            SCEVHandle Div = SE.getConstant(CI);
+            ScaledOps.push_back(Div);
+            continue;
+          }
+        if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(Ops[i]))
+          if (const SCEVConstant *C = dyn_cast<SCEVConstant>(M->getOperand(0)))
+            if (C->getValue()->getValue() == ElSize) {
+              for (unsigned j = 1, f = M->getNumOperands(); j != f; ++j)
+                ScaledOps.push_back(M->getOperand(j));
+              continue;
+            }
+        if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(Ops[i]))
+          if (BinaryOperator *BO = dyn_cast<BinaryOperator>(U->getValue()))
+            if (BO->getOpcode() == Instruction::Mul)
+              if (ConstantInt *CI = dyn_cast<ConstantInt>(BO->getOperand(1)))
+                if (CI->getValue() == ElSize) {
+                  ScaledOps.push_back(SE.getUnknown(BO->getOperand(0)));
+                  continue;
+                }
+        if (ElSize == 1) {
+          ScaledOps.push_back(Ops[i]);
+          continue;
+        }
+      }
+      NewOps.push_back(Ops[i]);
+    }
+    Ops = NewOps;
+    AnyNonZeroIndices |= !ScaledOps.empty();
+    Value *Scaled = ScaledOps.empty() ?
+                    Constant::getNullValue(Ty) :
+                    expandCodeFor(SE.getAddExpr(ScaledOps), Ty);
+    GepIndices.push_back(Scaled);
+
+    // Collect struct field index operands.
+    if (!Ops.empty())
+      while (const StructType *STy = dyn_cast<StructType>(ElTy)) {
+        if (const SCEVConstant *C = dyn_cast<SCEVConstant>(Ops[0]))
+          if (SE.getTypeSizeInBits(C->getType()) <= 64) {
+            const StructLayout &SL = *SE.TD->getStructLayout(STy);
+            uint64_t FullOffset = C->getValue()->getZExtValue();
+            if (FullOffset < SL.getSizeInBytes()) {
+              unsigned ElIdx = SL.getElementContainingOffset(FullOffset);
+              GepIndices.push_back(ConstantInt::get(Type::Int32Ty, ElIdx));
+              ElTy = STy->getTypeAtIndex(ElIdx);
+              Ops[0] =
+                SE.getConstant(ConstantInt::get(Ty,
+                                                FullOffset -
+                                                  SL.getElementOffset(ElIdx)));
+              AnyNonZeroIndices = true;
+              continue;
+            }
+          }
+        break;
+      }
+
+    if (const ArrayType *ATy = dyn_cast<ArrayType>(ElTy)) {
+      ElTy = ATy->getElementType();
+      continue;
+    }
+    break;
+  }
+
+  // If none of the operands were convertable to proper GEP indices, cast
+  // the base to i8* and do an ugly getelementptr with that. It's still
+  // better than ptrtoint+arithmetic+inttoptr at least.
+  if (!AnyNonZeroIndices) {
+    V = InsertNoopCastOfTo(V,
+                           Type::Int8Ty->getPointerTo(PTy->getAddressSpace()));
+    Value *Idx = expand(SE.getAddExpr(Ops));
+    Idx = InsertNoopCastOfTo(Idx, Ty);
+
+    // Fold a GEP with constant operands.
+    if (Constant *CLHS = dyn_cast<Constant>(V))
+      if (Constant *CRHS = dyn_cast<Constant>(Idx))
+        return ConstantExpr::get(Instruction::GetElementPtr, CLHS, CRHS);
+
+    // Do a quick scan to see if we have this GEP nearby.  If so, reuse it.
+    unsigned ScanLimit = 6;
+    BasicBlock::iterator BlockBegin = InsertPt->getParent()->begin();
+    if (InsertPt != BlockBegin) {
+      // Scanning starts from the last instruction before InsertPt.
+      BasicBlock::iterator IP = InsertPt;
+      --IP;
+      for (; ScanLimit; --IP, --ScanLimit) {
+        if (IP->getOpcode() == Instruction::GetElementPtr &&
+            IP->getOperand(0) == V && IP->getOperand(1) == Idx)
+          return IP;
+        if (IP == BlockBegin) break;
+      }
+    }
+
+    Value *GEP = GetElementPtrInst::Create(V, Idx, "scevgep", InsertPt);
+    InsertedValues.insert(GEP);
+    return GEP;
+  }
+
+  // Insert a pretty getelementptr.
+  Value *GEP = GetElementPtrInst::Create(V,
+                                         GepIndices.begin(),
+                                         GepIndices.end(),
+                                         "scevgep", InsertPt);
+  Ops.push_back(SE.getUnknown(GEP));
+  InsertedValues.insert(GEP);
+  return expand(SE.getAddExpr(Ops));
+}
+
 Value *SCEVExpander::visitAddExpr(const SCEVAddExpr *S) {
   const Type *Ty = SE.getEffectiveSCEVType(S->getType());
   Value *V = expand(S->getOperand(S->getNumOperands()-1));
+
+  // Turn things like ptrtoint+arithmetic+inttoptr into GEP. This helps
+  // BasicAliasAnalysis analyze the result. However, it suffers from the
+  // underlying bug described in PR2831. Addition in LLVM currently always
+  // has two's complement wrapping guaranteed. However, the semantics for
+  // getelementptr overflow are ambiguous. In the common case though, this
+  // expansion gets used when a GEP in the original code has been converted
+  // into integer arithmetic, in which case the resulting code will be no
+  // more undefined than it was originally.
+  if (SE.TD)
+    if (const PointerType *PTy = dyn_cast<PointerType>(V->getType()))
+      return expandAddToGEP(S, PTy, Ty, V);
+
   V = InsertNoopCastOfTo(V, Ty);
 
   // Emit a bunch of add instructions
@@ -157,7 +304,7 @@
   }
   return V;
 }
-    
+
 Value *SCEVExpander::visitMulExpr(const SCEVMulExpr *S) {
   const Type *Ty = SE.getEffectiveSCEVType(S->getType());
   int FirstOp = 0;  // Set if we should emit a subtract.
@@ -206,15 +353,10 @@
 
   // {X,+,F} --> X + {0,+,F}
   if (!S->getStart()->isZero()) {
-    Value *Start = expand(S->getStart());
-    Start = InsertNoopCastOfTo(Start, Ty);
-    std::vector<SCEVHandle> NewOps(S->op_begin(), S->op_end());
+    std::vector<SCEVHandle> NewOps(S->getOperands());
     NewOps[0] = SE.getIntegerSCEV(0, Ty);
     Value *Rest = expand(SE.getAddRecExpr(NewOps, L));
-    Rest = InsertNoopCastOfTo(Rest, Ty);
-
-    // FIXME: look for an existing add to use.
-    return InsertBinop(Instruction::Add, Rest, Start, InsertPt);
+    return expand(SE.getAddExpr(S->getStart(), SE.getUnknown(Rest)));
   }
 
   // {0,+,1} --> Insert a canonical induction variable into the loop!
@@ -265,7 +407,7 @@
     // point loop.  If we can, move the multiply to the outer most loop that it
     // is safe to be in.
     BasicBlock::iterator MulInsertPt = getInsertionPoint();
-    Loop *InsertPtLoop = LI.getLoopFor(MulInsertPt->getParent());
+    Loop *InsertPtLoop = SE.LI->getLoopFor(MulInsertPt->getParent());
     if (InsertPtLoop != L && InsertPtLoop &&
         L->contains(InsertPtLoop->getHeader())) {
       do {
@@ -363,10 +505,13 @@
 
 Value *SCEVExpander::expandCodeFor(SCEVHandle SH, const Type *Ty) {
   // Expand the code for this SCEV.
-  assert(SE.getTypeSizeInBits(Ty) == SE.getTypeSizeInBits(SH->getType()) &&
-         "non-trivial casts should be done with the SCEVs directly!");
   Value *V = expand(SH);
-  return InsertNoopCastOfTo(V, Ty);
+  if (Ty) {
+    assert(SE.getTypeSizeInBits(Ty) == SE.getTypeSizeInBits(SH->getType()) &&
+           "non-trivial casts should be done with the SCEVs directly!");
+    V = InsertNoopCastOfTo(V, Ty);
+  }
+  return V;
 }
 
 Value *SCEVExpander::expand(const SCEV *S) {

Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Mon May 18 21:15:55 2009
@@ -172,7 +172,7 @@
 
   // Expand the code for the iteration count into the preheader of the loop.
   BasicBlock *Preheader = L->getLoopPreheader();
-  Value *ExitCnt = Rewriter.expandCodeFor(RHS, IndVar->getType(),
+  Value *ExitCnt = Rewriter.expandCodeFor(RHS, CmpIndVar->getType(),
                                           Preheader->getTerminator());
 
   // Insert a new icmp_ne or icmp_eq instruction before the branch.
@@ -218,7 +218,7 @@
 
   // Scan all of the instructions in the loop, looking at those that have
   // extra-loop users and which are recurrences.
-  SCEVExpander Rewriter(*SE, *LI);
+  SCEVExpander Rewriter(*SE);
 
   // We insert the code into the preheader of the loop if the loop contains
   // multiple exit blocks, or in the exit block if there is exactly one.
@@ -386,7 +386,7 @@
   }
 
   // Create a rewriter object which we'll use to transform the code with.
-  SCEVExpander Rewriter(*SE, *LI);
+  SCEVExpander Rewriter(*SE);
 
   // Now that we know the largest of of the induction variable expressions
   // in this loop, insert a canonical induction variable of the largest size.
@@ -478,7 +478,7 @@
         BasicBlock::iterator I = Rewriter.getInsertionPoint();
         // Expand loop-invariant values in the loop preheader. They will
         // be sunk to the exit block later, if possible.
-          NewVal =
+        NewVal =
           Rewriter.expandCodeFor(AR, LargestType,
                                  L->getLoopPreheader()->getTerminator());
         Rewriter.setInsertionPoint(I);
@@ -523,7 +523,7 @@
         NewAR = SE->getAddExpr(NewAR, PromotedOffset);
 
         // Expand the addrec into instructions.
-        Value *V = Rewriter.expandCodeFor(NewAR, LargestType);
+        Value *V = Rewriter.expandCodeFor(NewAR);
 
         // Insert an explicit cast if necessary to truncate the value
         // down to the original stride type. This is done outside of
@@ -533,7 +533,7 @@
           if (SE->getTypeSizeInBits(IVTy) != SE->getTypeSizeInBits(LargestType))
             NewAR = SE->getTruncateExpr(NewAR, IVTy);
           if (Rewriter.isInsertedExpression(NewAR))
-            V = Rewriter.expandCodeFor(NewAR, IVTy);
+            V = Rewriter.expandCodeFor(NewAR);
           else {
             V = Rewriter.InsertCastOfTo(CastInst::getCastOpcode(V, false,
                                                                 IVTy, false),

Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Mon May 18 21:15:55 2009
@@ -367,12 +367,14 @@
     void RewriteInstructionToUseNewBase(const SCEVHandle &NewBase,
                                         Instruction *InsertPt,
                                        SCEVExpander &Rewriter, Loop *L, Pass *P,
+                                        LoopInfo &LI,
                                         SmallVectorImpl<WeakVH> &DeadInsts);
     
     Value *InsertCodeForBaseAtPosition(const SCEVHandle &NewBase, 
                                        const Type *Ty,
                                        SCEVExpander &Rewriter,
-                                       Instruction *IP, Loop *L);
+                                       Instruction *IP, Loop *L,
+                                       LoopInfo &LI);
     void dump() const;
   };
 }
@@ -386,12 +388,12 @@
 Value *BasedUser::InsertCodeForBaseAtPosition(const SCEVHandle &NewBase, 
                                               const Type *Ty,
                                               SCEVExpander &Rewriter,
-                                              Instruction *IP, Loop *L) {
+                                              Instruction *IP, Loop *L,
+                                              LoopInfo &LI) {
   // Figure out where we *really* want to insert this code.  In particular, if
   // the user is inside of a loop that is nested inside of L, we really don't
   // want to insert this expression before the user, we'd rather pull it out as
   // many loops as possible.
-  LoopInfo &LI = Rewriter.getLoopInfo();
   Instruction *BaseInsertPt = IP;
   
   // Figure out the most-nested loop that IP is in.
@@ -405,8 +407,7 @@
       InsertLoop = InsertLoop->getParentLoop();
     }
   
-  Value *Base = Rewriter.expandCodeFor(NewBase, NewBase->getType(),
-                                       BaseInsertPt);
+  Value *Base = Rewriter.expandCodeFor(NewBase, 0, BaseInsertPt);
 
   SCEVHandle NewValSCEV = SE->getUnknown(Base);
 
@@ -439,6 +440,7 @@
 void BasedUser::RewriteInstructionToUseNewBase(const SCEVHandle &NewBase,
                                                Instruction *NewBasePt,
                                       SCEVExpander &Rewriter, Loop *L, Pass *P,
+                                      LoopInfo &LI,
                                       SmallVectorImpl<WeakVH> &DeadInsts) {
   if (!isa<PHINode>(Inst)) {
     // By default, insert code at the user instruction.
@@ -468,7 +470,7 @@
     }
     Value *NewVal = InsertCodeForBaseAtPosition(NewBase,
                                                 OperandValToReplace->getType(),
-                                                Rewriter, InsertPt, L);
+                                                Rewriter, InsertPt, L, LI);
     // Replace the use of the operand Value with the new Phi we just created.
     Inst->replaceUsesOfWith(OperandValToReplace, NewVal);
 
@@ -527,7 +529,7 @@
                                 PN->getIncomingBlock(i)->getTerminator() :
                                 OldLoc->getParent()->getTerminator();
         Code = InsertCodeForBaseAtPosition(NewBase, PN->getType(),
-                                           Rewriter, InsertPt, L);
+                                           Rewriter, InsertPt, L, LI);
 
         DOUT << "      Changing PHI use to ";
         DEBUG(WriteAsOperand(*DOUT, Code, /*PrintType=*/false));
@@ -1580,8 +1582,8 @@
        << *Stride << ":\n"
        << "  Common base: " << *CommonExprs << "\n";
 
-  SCEVExpander Rewriter(*SE, *LI);
-  SCEVExpander PreheaderRewriter(*SE, *LI);
+  SCEVExpander Rewriter(*SE);
+  SCEVExpander PreheaderRewriter(*SE);
 
   BasicBlock  *Preheader = L->getLoopPreheader();
   Instruction *PreInsertPt = Preheader->getTerminator();
@@ -1636,8 +1638,7 @@
     // Emit the code for Base into the preheader.
     Value *BaseV = 0;
     if (!Base->isZero()) {
-      BaseV = PreheaderRewriter.expandCodeFor(Base, Base->getType(),
-                                              PreInsertPt);
+      BaseV = PreheaderRewriter.expandCodeFor(Base, 0, PreInsertPt);
 
       DOUT << "  INSERTING code for BASE = " << *Base << ":";
       if (BaseV->hasName())
@@ -1758,7 +1759,7 @@
         RewriteExpr = SE->getAddExpr(RewriteExpr, SE->getUnknown(BaseV));
 
       User.RewriteInstructionToUseNewBase(RewriteExpr, NewBasePt,
-                                          Rewriter, L, this,
+                                          Rewriter, L, this, *LI,
                                           DeadInsts);
 
       // Mark old value we replaced as possibly dead, so that it is eliminated

Modified: llvm/trunk/lib/Transforms/Utils/AddrModeMatcher.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Utils/AddrModeMatcher.cpp?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Utils/AddrModeMatcher.cpp (original)
+++ llvm/trunk/lib/Transforms/Utils/AddrModeMatcher.cpp Mon May 18 21:15:55 2009
@@ -255,43 +255,44 @@
 
     // Save the valid addressing mode in case we can't match.
     ExtAddrMode BackupAddrMode = AddrMode;
-    
-    // Check that this has no base reg yet.  If so, we won't have a place to
-    // put the base of the GEP (assuming it is not a null ptr).
-    bool SetBaseReg = true;
-    if (isa<ConstantPointerNull>(AddrInst->getOperand(0)))
-      SetBaseReg = false;   // null pointer base doesn't need representation.
-    else if (AddrMode.HasBaseReg)
-      return false;  // Base register already specified, can't match GEP.
-    else {
-      // Otherwise, we'll use the GEP base as the BaseReg.
+    unsigned OldSize = AddrModeInsts.size();
+
+    // See if the scale and offset amount is valid for this target.
+    AddrMode.BaseOffs += ConstantOffset;
+
+    // Match the base operand of the GEP.
+    if (!MatchAddr(AddrInst->getOperand(0), Depth+1)) {
+      // If it couldn't be matched, just stuff the value in a register.
+      if (AddrMode.HasBaseReg) {
+        AddrMode = BackupAddrMode;
+        AddrModeInsts.resize(OldSize);
+        return false;
+      }
       AddrMode.HasBaseReg = true;
       AddrMode.BaseReg = AddrInst->getOperand(0);
     }
-    
-    // See if the scale and offset amount is valid for this target.
-    AddrMode.BaseOffs += ConstantOffset;
-    
+
+    // Match the remaining variable portion of the GEP.
     if (!MatchScaledValue(AddrInst->getOperand(VariableOperand), VariableScale,
                           Depth)) {
+      // If it couldn't be matched, try stuffing the base into a register
+      // instead of matching it, and retrying the match of the scale.
       AddrMode = BackupAddrMode;
-      return false;
+      AddrModeInsts.resize(OldSize);
+      if (AddrMode.HasBaseReg)
+        return false;
+      AddrMode.HasBaseReg = true;
+      AddrMode.BaseReg = AddrInst->getOperand(0);
+      AddrMode.BaseOffs += ConstantOffset;
+      if (!MatchScaledValue(AddrInst->getOperand(VariableOperand),
+                            VariableScale, Depth)) {
+        // If even that didn't work, bail.
+        AddrMode = BackupAddrMode;
+        AddrModeInsts.resize(OldSize);
+        return false;
+      }
     }
-    
-    // If we have a null as the base of the GEP, folding in the constant offset
-    // plus variable scale is all we can do.
-    if (!SetBaseReg) return true;
-      
-    // If this match succeeded, we know that we can form an address with the
-    // GepBase as the basereg.  Match the base pointer of the GEP more
-    // aggressively by zeroing out BaseReg and rematching.  If the base is
-    // (for example) another GEP, this allows merging in that other GEP into
-    // the addressing mode we're forming.
-    AddrMode.HasBaseReg = false;
-    AddrMode.BaseReg = 0;
-    bool Success = MatchAddr(AddrInst->getOperand(0), Depth+1);
-    assert(Success && "MatchAddr should be able to fill in BaseReg!");
-    Success=Success;
+
     return true;
   }
   }

Added: llvm/trunk/test/Transforms/IndVarSimplify/preserve-gep.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/IndVarSimplify/preserve-gep.ll?rev=72093&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/IndVarSimplify/preserve-gep.ll (added)
+++ llvm/trunk/test/Transforms/IndVarSimplify/preserve-gep.ll Mon May 18 21:15:55 2009
@@ -0,0 +1,39 @@
+; RUN: llvm-as < %s | opt -indvars | llvm-dis > %t
+; RUN: not grep ptrtoint %t
+; RUN: not grep inttoptr %t
+; RUN: grep getelementptr %t | count 1
+
+; Indvars shouldn't leave getelementptrs expanded out as
+; inttoptr+ptrtoint in its output in common cases.
+
+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"
+target triple = "x86_64-unknown-linux-gnu"
+	%struct.Foo = type { i32, i32, [10 x i32], i32 }
+
+define void @me(%struct.Foo* nocapture %Bar) nounwind {
+entry:
+	br i1 false, label %return, label %bb.nph
+
+bb.nph:		; preds = %entry
+	br label %bb
+
+bb:		; preds = %bb1, %bb.nph
+	%i.01 = phi i64 [ %4, %bb1 ], [ 0, %bb.nph ]		; <i64> [#uses=3]
+	%0 = getelementptr %struct.Foo* %Bar, i64 %i.01, i32 2, i64 3		; <i32*> [#uses=1]
+	%1 = load i32* %0, align 4		; <i32> [#uses=1]
+	%2 = mul i32 %1, 113		; <i32> [#uses=1]
+	%3 = getelementptr %struct.Foo* %Bar, i64 %i.01, i32 2, i64 3		; <i32*> [#uses=1]
+	store i32 %2, i32* %3, align 4
+	%4 = add i64 %i.01, 1		; <i64> [#uses=2]
+	br label %bb1
+
+bb1:		; preds = %bb
+	%phitmp = icmp sgt i64 %4, 19999		; <i1> [#uses=1]
+	br i1 %phitmp, label %bb1.return_crit_edge, label %bb
+
+bb1.return_crit_edge:		; preds = %bb1
+	br label %return
+
+return:		; preds = %bb1.return_crit_edge, %entry
+	ret void
+}

Modified: llvm/trunk/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll?rev=72093&r1=72092&r2=72093&view=diff

==============================================================================
--- llvm/trunk/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll (original)
+++ llvm/trunk/test/Transforms/LoopStrengthReduce/2009-04-28-no-reduce-mul.ll Mon May 18 21:15:55 2009
@@ -1,4 +1,5 @@
-; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis | grep {mul.*%lsr.iv} | count 2
+; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis \
+; RUN:   | grep {getelementptr.*%lsr.iv.*%lsr.iv.*<i32\\*>}
 ; The multiply in bb2 must not be reduced to an add, as the sext causes the
 ; %1 argument to become negative after a while.
 ; ModuleID = '<stdin>'




From gohman at apple.com  Mon May 18 21:19:57 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 02:19:57 -0000
Subject: [llvm-commits] [llvm] r72094 -
	/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
Message-ID: <200905190219.n4J2JwED009203@zion.cs.uiuc.edu>

Author: djg
Date: Mon May 18 21:19:57 2009
New Revision: 72094

URL: http://llvm.org/viewvc/llvm-project?rev=72094&view=rev
Log:
Remove the #ifndef NDEBUG from the FastISel debugging options. This
fixes dejagnu tests that use these options.

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

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp?rev=72094&r1=72093&r2=72094&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp Mon May 18 21:19:57 2009
@@ -54,7 +54,6 @@
 
 static cl::opt<bool>
 DisableLegalizeTypes("disable-legalize-types", cl::Hidden);
-#ifndef NDEBUG
 static cl::opt<bool>
 EnableFastISelVerbose("fast-isel-verbose", cl::Hidden,
           cl::desc("Enable verbose messages in the \"fast\" "
@@ -62,10 +61,6 @@
 static cl::opt<bool>
 EnableFastISelAbort("fast-isel-abort", cl::Hidden,
           cl::desc("Enable abort calls when \"fast\" instruction fails"));
-#else
-static const bool EnableFastISelVerbose = false,
-                  EnableFastISelAbort = false;
-#endif
 static cl::opt<bool>
 SchedLiveInCopies("schedule-livein-copies",
                   cl::desc("Schedule copies of livein registers"),




From bob.wilson at apple.com  Tue May 19 00:53:43 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Tue, 19 May 2009 05:53:43 -0000
Subject: [llvm-commits] [llvm] r72105 - in /llvm/trunk:
 lib/Target/ARM/ARMISelDAGToDAG.cpp
 lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
 test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll
Message-ID: <200905190553.n4J5rhoI017012@zion.cs.uiuc.edu>

Author: bwilson
Date: Tue May 19 00:53:42 2009
New Revision: 72105

URL: http://llvm.org/viewvc/llvm-project?rev=72105&view=rev
Log:
Fix pr4091: Add support for "m" constraint in ARM inline assembly.

Added:
    llvm/trunk/test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp
    llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp

Modified: llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp?rev=72105&r1=72104&r2=72105&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelDAGToDAG.cpp Tue May 19 00:53:42 2009
@@ -89,6 +89,13 @@
   
   // Include the pieces autogenerated from the target description.
 #include "ARMGenDAGISel.inc"
+
+private:
+    /// SelectInlineAsmMemoryOperand - Implement addressing mode selection for
+    /// inline asm expressions.
+    virtual bool SelectInlineAsmMemoryOperand(const SDValue &Op,
+                                              char ConstraintCode,
+                                              std::vector<SDValue> &OutOps);
 };
 }
 
@@ -881,6 +888,21 @@
   return SelectCode(Op);
 }
 
+bool ARMDAGToDAGISel::
+SelectInlineAsmMemoryOperand(const SDValue &Op, char ConstraintCode,
+                             std::vector<SDValue> &OutOps) {
+  assert(ConstraintCode == 'm' && "unexpected asm memory constraint");
+
+  SDValue Base, Offset, Opc;
+  if (!SelectAddrMode2(Op, Op, Base, Offset, Opc))
+    return true;
+  
+  OutOps.push_back(Base);
+  OutOps.push_back(Offset);
+  OutOps.push_back(Opc);
+  return false;
+}
+
 /// createARMISelDag - This pass converts a legalized DAG into a
 /// ARM-specific DAG, ready for instruction scheduling.
 ///

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=72105&r1=72104&r2=72105&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Tue May 19 00:53:42 2009
@@ -124,6 +124,9 @@
 
     virtual bool PrintAsmOperand(const MachineInstr *MI, unsigned OpNo,
                                  unsigned AsmVariant, const char *ExtraCode);
+    virtual bool PrintAsmMemoryOperand(const MachineInstr *MI, unsigned OpNo,
+                                       unsigned AsmVariant,
+                                       const char *ExtraCode);
 
     void printModuleLevelGV(const GlobalVariable* GVar);
     bool printInstruction(const MachineInstr *MI);  // autogenerated.
@@ -769,6 +772,15 @@
   return false;
 }
 
+bool ARMAsmPrinter::PrintAsmMemoryOperand(const MachineInstr *MI,
+                                          unsigned OpNo, unsigned AsmVariant,
+                                          const char *ExtraCode) {
+  if (ExtraCode && ExtraCode[0])
+    return true; // Unknown modifier.
+  printAddrMode2Operand(MI, OpNo);
+  return false;
+}
+
 void ARMAsmPrinter::printMachineInstruction(const MachineInstr *MI) {
   ++EmittedInsts;
 

Added: llvm/trunk/test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll?rev=72105&view=auto

==============================================================================
--- llvm/trunk/test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll (added)
+++ llvm/trunk/test/CodeGen/ARM/2009-05-18-InlineAsmMem.ll Tue May 19 00:53:42 2009
@@ -0,0 +1,7 @@
+; RUN: llvm-as < %s | llc -march=arm | grep swp
+; PR4091
+
+define void @foo(i32 %i, i32* %p) nounwind {
+	%asmtmp = call i32 asm sideeffect "swp $0, $2, $3", "=&r,=*m,r,*m,~{memory}"(i32* %p, i32 %i, i32* %p) nounwind
+	ret void
+}




From nicolas.geoffray at lip6.fr  Tue May 19 04:25:00 2009
From: nicolas.geoffray at lip6.fr (Nicolas Geoffray)
Date: Tue, 19 May 2009 11:25:00 +0200
Subject: [llvm-commits] [llvm] r72045
	-	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
In-Reply-To: <200905182142.n4ILgWgP031346@zion.cs.uiuc.edu>
References: <200905182142.n4ILgWgP031346@zion.cs.uiuc.edu>
Message-ID: <4A127AEC.1070608@lip6.fr>

Hi Dale,

Dale Johannesen wrote:
> Author: johannes
> Date: Mon May 18 16:41:59 2009
> New Revision: 72045
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72045&view=rev
> Log:
> Add OpSize to 16-bit ADC and SBB.
>   

What is OpSize? This patch is causing lots of memory errors in VMKit, 
I'm not sure why. Does the JIT in LLVM understand OpSize?

Thanks,
Nicolas

>
> 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=72045&r1=72044&r2=72045&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon May 18 16:41:59 2009
> @@ -2280,7 +2280,7 @@
>  def ADC16rr  : I<0x11, MRMDestReg, (outs GR16:$dst),
>                                     (ins GR16:$src1, GR16:$src2),
>                   "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>;
> +                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>, OpSize;
>  def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst),
>                                     (ins GR32:$src1, GR32:$src2),
>                   "adc{l}\t{$src2, $dst|$dst, $src2}",
> @@ -2293,7 +2293,8 @@
>  def ADC16rm  : I<0x13, MRMSrcMem , (outs GR16:$dst),
>                                     (ins GR16:$src1, i16mem:$src2),
>                   "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                 [(set GR16:$dst, (adde GR16:$src1, (load addr:$src2)))]>;
> +                 [(set GR16:$dst, (adde GR16:$src1, (load addr:$src2)))]>,
> +                 OpSize;
>  def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst),
>                                     (ins GR32:$src1, i32mem:$src2),
>                   "adc{l}\t{$src2, $dst|$dst, $src2}",
> @@ -2304,11 +2305,12 @@
>  def ADC16ri  : Ii16<0x81, MRM2r, (outs GR16:$dst),
>                                   (ins GR16:$src1, i16imm:$src2),
>                      "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>;
> +                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>, OpSize;
>  def ADC16ri8 : Ii8<0x83, MRM2r, (outs GR16:$dst),
>                                  (ins GR16:$src1, i16i8imm:$src2),
>                     "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                 [(set GR16:$dst, (adde GR16:$src1, i16immSExt8:$src2))]>;
> +                 [(set GR16:$dst, (adde GR16:$src1, i16immSExt8:$src2))]>,
> +                 OpSize;
>  def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst),
>                                   (ins GR32:$src1, i32imm:$src2),
>                      "adc{l}\t{$src2, $dst|$dst, $src2}",
> @@ -2324,7 +2326,8 @@
>                     [(store (adde (load addr:$dst), GR8:$src2), addr:$dst)]>;
>    def ADC16mr  : I<0x11, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2),
>                     "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                   [(store (adde (load addr:$dst), GR16:$src2), addr:$dst)]>;
> +                   [(store (adde (load addr:$dst), GR16:$src2), addr:$dst)]>,
> +                   OpSize;
>    def ADC32mr  : I<0x11, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2),
>                     "adc{l}\t{$src2, $dst|$dst, $src2}",
>                     [(store (adde (load addr:$dst), GR32:$src2), addr:$dst)]>;
> @@ -2333,10 +2336,12 @@
>                    [(store (adde (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
>    def ADC16mi  : Ii16<0x81, MRM2m, (outs), (ins i16mem:$dst, i16imm:$src2),
>                        "adc{w}\t{$src2, $dst|$dst, $src2}",
> -                  [(store (adde (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
> +                  [(store (adde (loadi16 addr:$dst), imm:$src2), addr:$dst)]>,
> +                  OpSize;
>    def ADC16mi8 : Ii8<0x83, MRM2m, (outs), (ins i16mem:$dst, i16i8imm :$src2),
>                       "adc{w}\t{$src2, $dst|$dst, $src2}",
> -               [(store (adde (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
> +               [(store (adde (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>,
> +               OpSize;
>    def ADC32mi  : Ii32<0x81, MRM2m, (outs), (ins i32mem:$dst, i32imm:$src2),
>                        "adc{l}\t{$src2, $dst|$dst, $src2}",
>                    [(store (adde (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
> @@ -2452,7 +2457,7 @@
>  def SBB16rr    : I<0x19, MRMDestReg, (outs GR16:$dst),
>                                       (ins GR16:$src1, GR16:$src2),
>                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>;
> +                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>, OpSize;
>  def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst),
>                                        (ins GR32:$src1, GR32:$src2),
>                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
> @@ -2464,7 +2469,8 @@
>                     [(store (sube (load addr:$dst), GR8:$src2), addr:$dst)]>;
>    def SBB16mr  : I<0x19, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2), 
>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                   [(store (sube (load addr:$dst), GR16:$src2), addr:$dst)]>;
> +                   [(store (sube (load addr:$dst), GR16:$src2), addr:$dst)]>,
> +                   OpSize;
>    def SBB32mr  : I<0x19, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2), 
>                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
>                     [(store (sube (load addr:$dst), GR32:$src2), addr:$dst)]>;
> @@ -2473,10 +2479,12 @@
>                     [(store (sube (loadi8 addr:$dst), imm:$src2), addr:$dst)]>;
>    def SBB16mi  : Ii16<0x81, MRM3m, (outs), (ins i16mem:$dst, i16imm:$src2), 
>                        "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                  [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>;
> +                  [(store (sube (loadi16 addr:$dst), imm:$src2), addr:$dst)]>,
> +                  OpSize;
>    def SBB16mi8 : Ii8<0x83, MRM3m, (outs), (ins i16mem:$dst, i16i8imm :$src2), 
>                       "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -               [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>;
> +               [(store (sube (load addr:$dst), i16immSExt8:$src2), addr:$dst)]>,
> +               OpSize;
>    def SBB32mi  : Ii32<0x81, MRM3m, (outs), (ins i32mem:$dst, i32imm:$src2), 
>                        "sbb{l}\t{$src2, $dst|$dst, $src2}",
>                    [(store (sube (loadi32 addr:$dst), imm:$src2), addr:$dst)]>;
> @@ -2490,7 +2498,8 @@
>  def SBB16rm  : I<0x1B, MRMSrcMem, (outs GR16:$dst),
>                                    (ins GR16:$src1, i16mem:$src2),
>                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                    [(set GR16:$dst, (sube GR16:$src1, (load addr:$src2)))]>;
> +                    [(set GR16:$dst, (sube GR16:$src1, (load addr:$src2)))]>,
> +                    OpSize;
>  def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst),
>                                    (ins GR32:$src1, i32mem:$src2),
>                      "sbb{l}\t{$src2, $dst|$dst, $src2}",
> @@ -2501,11 +2510,12 @@
>  def SBB16ri  : Ii16<0x81, MRM3r, (outs GR16:$dst),
>                                   (ins GR16:$src1, i16imm:$src2),
>                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                    [(set GR16:$dst, (sube GR16:$src1, imm:$src2))]>;
> +                    [(set GR16:$dst, (sube GR16:$src1, imm:$src2))]>, OpSize;
>  def SBB16ri8 : Ii8<0x83, MRM3r, (outs GR16:$dst),
>                                  (ins GR16:$src1, i16i8imm:$src2),
>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
> -                   [(set GR16:$dst, (sube GR16:$src1, i16immSExt8:$src2))]>;
> +                   [(set GR16:$dst, (sube GR16:$src1, i16immSExt8:$src2))]>,
> +                   OpSize;
>  def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst), 
>                                   (ins GR32:$src1, i32imm:$src2),
>                      "sbb{l}\t{$src2, $dst|$dst, $src2}",
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>   



From cristic at cs.stanford.edu  Mon May 18 20:56:09 2009
From: cristic at cs.stanford.edu (Cristian Cadar)
Date: Tue, 19 May 2009 01:56:09 -0000
Subject: [llvm-commits] [llvm] r72092 - /llvm/trunk/docs/doxygen.cfg.in
Message-ID: <200905190156.n4J1u9h2008184@zion.cs.uiuc.edu>

Author: cristic
Date: Mon May 18 20:56:08 2009
New Revision: 72092

URL: http://llvm.org/viewvc/llvm-project?rev=72092&view=rev
Log:
Test commit.

Modified:
    llvm/trunk/docs/doxygen.cfg.in

Modified: llvm/trunk/docs/doxygen.cfg.in
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/docs/doxygen.cfg.in?rev=72092&r1=72091&r2=72092&view=diff

==============================================================================
--- llvm/trunk/docs/doxygen.cfg.in (original)
+++ llvm/trunk/docs/doxygen.cfg.in Mon May 18 20:56:08 2009
@@ -11,7 +11,7 @@
 # Values that contain spaces should be placed between quotes (" ")
 
 #---------------------------------------------------------------------------
-# Project related configuration options
+# Project related configuration options 
 #---------------------------------------------------------------------------
 
 # This tag specifies the encoding used for all characters in the config file 




From howard0su at gmail.com  Tue May 19 04:44:18 2009
From: howard0su at gmail.com (Howard Su)
Date: Tue, 19 May 2009 17:44:18 +0800
Subject: [llvm-commits] [PATCH] Fix C2888 compile error in VS2008
Message-ID: <f126fae00905190244u48ac8319t32e8777ce420af5b@mail.gmail.com>

The attached patch to fix two compiler error C2888 when compiling under
VS2008.


-- 
-Howard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090519/36ea9100/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: build_under_vs8.patch
Type: application/octet-stream
Size: 675 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090519/36ea9100/attachment.obj 

From bob.wilson at apple.com  Tue May 19 05:02:41 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Tue, 19 May 2009 10:02:41 -0000
Subject: [llvm-commits] [llvm] r72106 - in /llvm/trunk:
 lib/Target/ARM/ARMCallingConv.td
 test/CodeGen/ARM/arguments-nosplit-double.ll
 test/CodeGen/ARM/arguments-nosplit-i64.ll
Message-ID: <200905191002.n4JA2gVE005722@zion.cs.uiuc.edu>

Author: bwilson
Date: Tue May 19 05:02:36 2009
New Revision: 72106

URL: http://llvm.org/viewvc/llvm-project?rev=72106&view=rev
Log:
Fix pr4058 and pr4059.  Do not split i64 or double arguments between r3 and
the stack.  Patch by Sandeep Patel.

Added:
    llvm/trunk/test/CodeGen/ARM/arguments-nosplit-double.ll
    llvm/trunk/test/CodeGen/ARM/arguments-nosplit-i64.ll
Modified:
    llvm/trunk/lib/Target/ARM/ARMCallingConv.td

Modified: llvm/trunk/lib/Target/ARM/ARMCallingConv.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMCallingConv.td?rev=72106&r1=72105&r2=72106&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMCallingConv.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMCallingConv.td Tue May 19 05:02:36 2009
@@ -51,11 +51,14 @@
 
   // i64/f64 is passed in even pairs of GPRs
   // i64 is 8-aligned i32 here, so we may need to eat R1 as a pad register
+  // (and the same is true for f64 if VFP is not enabled)
   CCIfType<[i32], CCIfAlign<"8", CCAssignToRegWithShadow<[R0, R2], [R0, R1]>>>,
   CCIfType<[f64], CCCustom<"CC_ARM_AAPCS_Custom_f64">>,
 
   CCIfType<[f32], CCBitConvertToType<i32>>,
-  CCIfType<[i32], CCAssignToReg<[R0, R1, R2, R3]>>,
+  CCIfType<[i32], CCIf<"State.getNextStackOffset() == 0 &&"
+                       "ArgFlags.getOrigAlign() != 8",
+                       CCAssignToReg<[R0, R1, R2, R3]>>>,
 
   CCIfType<[i32], CCAssignToStack<4, 4>>,
   CCIfType<[f64], CCAssignToStack<8, 8>>

Added: llvm/trunk/test/CodeGen/ARM/arguments-nosplit-double.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/arguments-nosplit-double.ll?rev=72106&view=auto

==============================================================================
--- llvm/trunk/test/CodeGen/ARM/arguments-nosplit-double.ll (added)
+++ llvm/trunk/test/CodeGen/ARM/arguments-nosplit-double.ll Tue May 19 05:02:36 2009
@@ -0,0 +1,9 @@
+; RUN: llvm-as < %s | llc -mtriple=arm-linux-gnueabi | not grep r3
+; PR4059
+
+define i32 @f(i64 %z, i32 %a, double %b) {
+	%tmp = call i32 @g(double %b)
+	ret i32 %tmp
+}
+
+declare i32 @g(double)

Added: llvm/trunk/test/CodeGen/ARM/arguments-nosplit-i64.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/ARM/arguments-nosplit-i64.ll?rev=72106&view=auto

==============================================================================
--- llvm/trunk/test/CodeGen/ARM/arguments-nosplit-i64.ll (added)
+++ llvm/trunk/test/CodeGen/ARM/arguments-nosplit-i64.ll Tue May 19 05:02:36 2009
@@ -0,0 +1,9 @@
+; RUN: llvm-as < %s | llc -mtriple=arm-linux-gnueabi | not grep r3
+; PR4058
+
+define i32 @f(i64 %z, i32 %a, i64 %b) {
+	%tmp = call i32 @g(i64 %b)
+	ret i32 %tmp
+}
+
+declare i32 @g(i64)




From clattner at apple.com  Tue May 19 05:08:30 2009
From: clattner at apple.com (Chris Lattner)
Date: Tue, 19 May 2009 03:08:30 -0700
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <5464957333051772950217202212009693918-Webmail@me.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
Message-ID: <A091075B-9531-4FA9-BAE4-A39049026353@apple.com>


On May 18, 2009, at 6:25 PM, Owen Anderson wrote:

> So, now that we have the basic infrastructure more or less in place,  
> here's another attempt at making a thread-safe version of  
> ManagedStatic based on double-checked locking with explicit fences.   
> It also proved necessary to provide atomic update facilities for the  
> global StaticList as well.
>
> Please see the included block comment for more details on exactly  
> what safety guarantees this code provides.

The basic algorithm looks fine, however:

1. Please move the meat of the code for LazyInit out of line somehow.   
Only the case when tmp != 0 should be inline ideally.
2. Again, we only need a single global lock, we do not need a mutex  
per ManagedStatic.
3. Why do you create your own mutex out of atomic operations instead  
of just using a pthreads lock?  spinning is not always an efficient  
solution to contention, particularly if your machine really only has  
one cpu! (in which case you end up spinning away the rest of your  
timeslice)
4. Instead of going through the trouble to make the locking more  
efficient, I think it would be good to special case the scenario when  
there is only one LLVM thread.  Why not make clients explicitly opt-in  
to multithreaded llvm, by making an explicit llvm_multithread call?   
If multithreading is disabled, this call should return an error.  If  
enabled, it would set a global.  All the various "locking" clients  
could just check the global before taking potential heavy-weight locks.

-Chris




From resistor at mac.com  Tue May 19 06:26:12 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 04:26:12 -0700
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
Message-ID: <358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>


On May 19, 2009, at 3:08 AM, Chris Lattner wrote:
> The basic algorithm looks fine, however:
>
> 1. Please move the meat of the code for LazyInit out of line somehow.
> Only the case when tmp != 0 should be inline ideally.

Done.

> 2. Again, we only need a single global lock, we do not need a mutex
> per ManagedStatic.

Conversely, why have only one?  Also, see my answer to #3.

> 3. Why do you create your own mutex out of atomic operations instead
> of just using a pthreads lock?  spinning is not always an efficient
> solution to contention, particularly if your machine really only has
> one cpu! (in which case you end up spinning away the rest of your
> timeslice)

Portability.  llvm::sys::Mutex _cannot_ be made to work, because on  
Windows
it has to have a non-trivial constructor.

Also, this is pretty much the ideal case for spinlocks.   
Initialization only happens once and
we're using a separate mutex for each object, so we don't expect  
significant contention over
the life of the program.   Plus, the initialization itself should be  
fairly fast.  This means that
  the spinning thread can continue immediately after the lock is  
released, rather than having
  to wait for the OS to wake it up.

> 4. Instead of going through the trouble to make the locking more
> efficient, I think it would be good to special case the scenario when
> there is only one LLVM thread.  Why not make clients explicitly opt-in
> to multithreaded llvm, by making an explicit llvm_multithread call?
> If multithreading is disabled, this call should return an error.  If
> enabled, it would set a global.  All the various "locking" clients
> could just check the global before taking potential heavy-weight  
> locks.

We already handle this in a somewhat different way.  When threading is  
disabled,
llvm::sys::Mutex operations are turned into no-ops, and MemoryFence()  
and
CompareAndSwap() are given non-atomic implementations.

--Owen
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2620 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090519/f543b925/attachment.bin 

From resistor at mac.com  Tue May 19 06:29:05 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 04:29:05 -0700
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
Message-ID: <5841C435-F5D5-4C62-A17B-1E406A9C8374@mac.com>


On May 19, 2009, at 4:26 AM, Owen Anderson wrote:
>> 3. Why do you create your own mutex out of atomic operations instead
>> of just using a pthreads lock?  spinning is not always an efficient
>> solution to contention, particularly if your machine really only has
>> one cpu! (in which case you end up spinning away the rest of your
>> timeslice)
>
> Portability.  llvm::sys::Mutex _cannot_ be made to work, because on  
> Windows
> it has to have a non-trivial constructor.

To build on my own comment, this is also the reason behind having one  
lock per
object.  Since we have to write our own mutex implementation for this  
to work on Windows,
we are free to write one that uses zero as the unlocked state, and can  
thus be put within
each ManagedStatic, reducing lock contention.

Basically, given that we need to implement our own mutex, making one- 
per-object
seems like the natural next step.

--Owen
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2620 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090519/3998a26d/attachment.bin 

From sebastian.redl at getdesigned.at  Tue May 19 08:59:56 2009
From: sebastian.redl at getdesigned.at (Sebastian Redl)
Date: Tue, 19 May 2009 15:59:56 +0200
Subject: [llvm-commits] [llvm] r72087
	-	/llvm/trunk/include/llvm/System/Atomic.h
In-Reply-To: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
Message-ID: <4A12BB5C.9040200@getdesigned.at>

Owen Anderson wrote:
> Author: resistor
> Date: Mon May 18 20:07:40 2009
> New Revision: 72087
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72087&view=rev
> Log:
> Fix up the Windows portion of Atomic.h.  This is untested, but it is my best understanding of what should work.
> I'd be much obliged if someone on MSVC++ could try this out and let me know if it works.
>
> Modified:
>     llvm/trunk/include/llvm/System/Atomic.h
>
> Modified: llvm/trunk/include/llvm/System/Atomic.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72087&r1=72086&r2=72087&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/System/Atomic.h (original)
> +++ llvm/trunk/include/llvm/System/Atomic.h Mon May 18 20:07:40 2009
> @@ -62,11 +62,24 @@
>      }
>  #elif defined(_MSC_VER)
>      typedef LONG cas_flag;
> +    template<typename T>
>      inline T CompareAndSwap(volatile T* ptr,
>  			    T new_value,
>  			    T old_value) {
> -      return InterlockedCompareExchange(addr, new_value, old_value);
> +      if (sizeof(T) == 4)
> +	return InterlockedCompareExchange(ptr, new_value, old_value);
> +      else
> +	return InterlockedCompareExchange64(ptr, new_value, old_value);
>      }
>   
I'm not booted into Windows, but this cannot possibly compile. An if
must have all branches valid even if the condition is a compile-time
constant. You could force some casts, or you could specialize the
function on 4-byte types. Eh, you'll need the casts anyway, because the
WinAPI sucks.

> +      else
> +	assert(0 && "Unsupported compare-and-swap size!");
>   

However, it would be really nice to catch this at compile time.
And another thing, the function is called InterlockedCompareExchangePointer.

The code below is untested, but should do the Right Thing.

template <typename T, size_t N>
struct CompareAndSwapImpl;

template <typename T>
struct CompareAndSwapImpl<T, 4> {
  static T DoCAS(volatile T* ptr, T n, T o) {
    return InterlockedCompareExchange(reinterpret_cast<LONG
volatile*>(ptr), n, o);
  }
}

template <typename T>
struct CompareAndSwapImpl<T, 8> {
  static T DoCAS(volatile T* ptr, T n, T o) {
    return InterlockedCompareExchange64(reinterpret_cast<LONGLONG
volatile*>(ptr), n, o);
  }
}

template <typename T>
struct CompareAndSwapImpl<T*, sizeof(T*)> {
  static T DoCAS(T* volatile* ptr, T* n, T* o) {
    return InterlockedCompareExchangePointer(reinterpret_cast<void*
volatile*>(ptr), n, o);
  }
}

template <typename T>
inline T CompareAndSwap(T volatile *ptr, T new_value, T old_value) {
  return CompareAndSwapImpl<T, sizeof(T)>::DoCAS(ptr, new_value, old_value);
}

Sebastian


From dalej at apple.com  Tue May 19 11:51:05 2009
From: dalej at apple.com (Dale Johannesen)
Date: Tue, 19 May 2009 09:51:05 -0700
Subject: [llvm-commits] [llvm] r72045	-
	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
In-Reply-To: <4A127AEC.1070608@lip6.fr>
References: <200905182142.n4ILgWgP031346@zion.cs.uiuc.edu>
	<4A127AEC.1070608@lip6.fr>
Message-ID: <8AD27C27-C211-4F58-A924-05FAF49952F0@apple.com>


On May 19, 2009, at 2:25 AMPDT, Nicolas Geoffray wrote:

> Hi Dale,
>
> Dale Johannesen wrote:
>> Author: johannes
>> Date: Mon May 18 16:41:59 2009
>> New Revision: 72045
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=72045&view=rev
>> Log:
>> Add OpSize to 16-bit ADC and SBB.
>>
>
> What is OpSize? This patch is causing lots of memory errors in VMKit,
> I'm not sure why.

It indicates the operand size prefix, 0x66, should be emitted.  It is  
normal on 16-bit operations.

> Does the JIT in LLVM understand OpSize?

Yes, in fact it's entirely for use by the JIT.  See XCodeEmitter.

Are you sure it's this patch?  I'd be surprised if these new patterns  
were getting matched at all right now.

> Thanks,
> Nicolas
>
>>
>> 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=72045&r1=72044&r2=72045&view=diff
>>
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> = 
>> =====================================================================
>> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
>> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon May 18 16:41:59  
>> 2009
>> @@ -2280,7 +2280,7 @@
>> def ADC16rr  : I<0x11, MRMDestReg, (outs GR16:$dst),
>>                                    (ins GR16:$src1, GR16:$src2),
>>                  "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>;
>> +                 [(set GR16:$dst, (adde GR16:$src1,  
>> GR16:$src2))]>, OpSize;
>> def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst),
>>                                    (ins GR32:$src1, GR32:$src2),
>>                  "adc{l}\t{$src2, $dst|$dst, $src2}",
>> @@ -2293,7 +2293,8 @@
>> def ADC16rm  : I<0x13, MRMSrcMem , (outs GR16:$dst),
>>                                    (ins GR16:$src1, i16mem:$src2),
>>                  "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                 [(set GR16:$dst, (adde GR16:$src1, (load addr: 
>> $src2)))]>;
>> +                 [(set GR16:$dst, (adde GR16:$src1, (load addr: 
>> $src2)))]>,
>> +                 OpSize;
>> def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst),
>>                                    (ins GR32:$src1, i32mem:$src2),
>>                  "adc{l}\t{$src2, $dst|$dst, $src2}",
>> @@ -2304,11 +2305,12 @@
>> def ADC16ri  : Ii16<0x81, MRM2r, (outs GR16:$dst),
>>                                  (ins GR16:$src1, i16imm:$src2),
>>                     "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>;
>> +                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>,  
>> OpSize;
>> def ADC16ri8 : Ii8<0x83, MRM2r, (outs GR16:$dst),
>>                                 (ins GR16:$src1, i16i8imm:$src2),
>>                    "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                 [(set GR16:$dst, (adde GR16:$src1,  
>> i16immSExt8:$src2))]>;
>> +                 [(set GR16:$dst, (adde GR16:$src1,  
>> i16immSExt8:$src2))]>,
>> +                 OpSize;
>> def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst),
>>                                  (ins GR32:$src1, i32imm:$src2),
>>                     "adc{l}\t{$src2, $dst|$dst, $src2}",
>> @@ -2324,7 +2326,8 @@
>>                    [(store (adde (load addr:$dst), GR8:$src2), addr: 
>> $dst)]>;
>>   def ADC16mr  : I<0x11, MRMDestMem, (outs), (ins i16mem:$dst,  
>> GR16:$src2),
>>                    "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                   [(store (adde (load addr:$dst), GR16:$src2),  
>> addr:$dst)]>;
>> +                   [(store (adde (load addr:$dst), GR16:$src2),  
>> addr:$dst)]>,
>> +                   OpSize;
>>   def ADC32mr  : I<0x11, MRMDestMem, (outs), (ins i32mem:$dst,  
>> GR32:$src2),
>>                    "adc{l}\t{$src2, $dst|$dst, $src2}",
>>                    [(store (adde (load addr:$dst), GR32:$src2),  
>> addr:$dst)]>;
>> @@ -2333,10 +2336,12 @@
>>                   [(store (adde (loadi8 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>>   def ADC16mi  : Ii16<0x81, MRM2m, (outs), (ins i16mem:$dst, i16imm: 
>> $src2),
>>                       "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -                  [(store (adde (loadi16 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>> +                  [(store (adde (loadi16 addr:$dst), imm:$src2),  
>> addr:$dst)]>,
>> +                  OpSize;
>>   def ADC16mi8 : Ii8<0x83, MRM2m, (outs), (ins i16mem:$dst,  
>> i16i8imm :$src2),
>>                      "adc{w}\t{$src2, $dst|$dst, $src2}",
>> -               [(store (adde (load addr:$dst), i16immSExt8:$src2),  
>> addr:$dst)]>;
>> +               [(store (adde (load addr:$dst), i16immSExt8:$src2),  
>> addr:$dst)]>,
>> +               OpSize;
>>   def ADC32mi  : Ii32<0x81, MRM2m, (outs), (ins i32mem:$dst, i32imm: 
>> $src2),
>>                       "adc{l}\t{$src2, $dst|$dst, $src2}",
>>                   [(store (adde (loadi32 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>> @@ -2452,7 +2457,7 @@
>> def SBB16rr    : I<0x19, MRMDestReg, (outs GR16:$dst),
>>                                      (ins GR16:$src1, GR16:$src2),
>>                   "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>;
>> +                 [(set GR16:$dst, (sube GR16:$src1,  
>> GR16:$src2))]>, OpSize;
>> def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst),
>>                                       (ins GR32:$src1, GR32:$src2),
>>                   "sbb{l}\t{$src2, $dst|$dst, $src2}",
>> @@ -2464,7 +2469,8 @@
>>                    [(store (sube (load addr:$dst), GR8:$src2), addr: 
>> $dst)]>;
>>   def SBB16mr  : I<0x19, MRMDestMem, (outs), (ins i16mem:$dst,  
>> GR16:$src2),
>>                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                   [(store (sube (load addr:$dst), GR16:$src2),  
>> addr:$dst)]>;
>> +                   [(store (sube (load addr:$dst), GR16:$src2),  
>> addr:$dst)]>,
>> +                   OpSize;
>>   def SBB32mr  : I<0x19, MRMDestMem, (outs), (ins i32mem:$dst,  
>> GR32:$src2),
>>                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>                    [(store (sube (load addr:$dst), GR32:$src2),  
>> addr:$dst)]>;
>> @@ -2473,10 +2479,12 @@
>>                    [(store (sube (loadi8 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>>   def SBB16mi  : Ii16<0x81, MRM3m, (outs), (ins i16mem:$dst, i16imm: 
>> $src2),
>>                       "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                  [(store (sube (loadi16 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>> +                  [(store (sube (loadi16 addr:$dst), imm:$src2),  
>> addr:$dst)]>,
>> +                  OpSize;
>>   def SBB16mi8 : Ii8<0x83, MRM3m, (outs), (ins i16mem:$dst,  
>> i16i8imm :$src2),
>>                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -               [(store (sube (load addr:$dst), i16immSExt8:$src2),  
>> addr:$dst)]>;
>> +               [(store (sube (load addr:$dst), i16immSExt8:$src2),  
>> addr:$dst)]>,
>> +               OpSize;
>>   def SBB32mi  : Ii32<0x81, MRM3m, (outs), (ins i32mem:$dst, i32imm: 
>> $src2),
>>                       "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>                   [(store (sube (loadi32 addr:$dst), imm:$src2),  
>> addr:$dst)]>;
>> @@ -2490,7 +2498,8 @@
>> def SBB16rm  : I<0x1B, MRMSrcMem, (outs GR16:$dst),
>>                                   (ins GR16:$src1, i16mem:$src2),
>>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                    [(set GR16:$dst, (sube GR16:$src1, (load addr: 
>> $src2)))]>;
>> +                    [(set GR16:$dst, (sube GR16:$src1, (load addr: 
>> $src2)))]>,
>> +                    OpSize;
>> def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst),
>>                                   (ins GR32:$src1, i32mem:$src2),
>>                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
>> @@ -2501,11 +2510,12 @@
>> def SBB16ri  : Ii16<0x81, MRM3r, (outs GR16:$dst),
>>                                  (ins GR16:$src1, i16imm:$src2),
>>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                    [(set GR16:$dst, (sube GR16:$src1, imm: 
>> $src2))]>;
>> +                    [(set GR16:$dst, (sube GR16:$src1, imm: 
>> $src2))]>, OpSize;
>> def SBB16ri8 : Ii8<0x83, MRM3r, (outs GR16:$dst),
>>                                 (ins GR16:$src1, i16i8imm:$src2),
>>                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
>> -                   [(set GR16:$dst, (sube GR16:$src1,  
>> i16immSExt8:$src2))]>;
>> +                   [(set GR16:$dst, (sube GR16:$src1,  
>> i16immSExt8:$src2))]>,
>> +                   OpSize;
>> def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst),
>>                                  (ins GR32:$src1, i32imm:$src2),
>>                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>
>>
>> _______________________________________________
>> 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 isanbard at gmail.com  Tue May 19 12:47:47 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 19 May 2009 10:47:47 -0700
Subject: [llvm-commits] [PATCH] Fix C2888 compile error in VS2008
In-Reply-To: <f126fae00905190244u48ac8319t32e8777ce420af5b@mail.gmail.com>
References: <f126fae00905190244u48ac8319t32e8777ce420af5b@mail.gmail.com>
Message-ID: <16e5fdf90905191047u270571fcrf695e4427beb73ca@mail.gmail.com>

"struct LiveInterva"? :-)

-bw

On Tue, May 19, 2009 at 2:44 AM, Howard Su <howard0su at gmail.com> wrote:
> The attached patch to fix two compiler error C2888?when compiling under
> VS2008.
>
> --
> -Howard
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



From isanbard at gmail.com  Tue May 19 12:50:19 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 19 May 2009 17:50:19 -0000
Subject: [llvm-commits] [llvm] r72112 - /llvm/trunk/lib/CodeGen/Spiller.h
Message-ID: <200905191750.n4JHoJZP022112@zion.cs.uiuc.edu>

Author: void
Date: Tue May 19 12:50:18 2009
New Revision: 72112

URL: http://llvm.org/viewvc/llvm-project?rev=72112&view=rev
Log:
Fix to compile on VS2008.

Modified:
    llvm/trunk/lib/CodeGen/Spiller.h

Modified: llvm/trunk/lib/CodeGen/Spiller.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.h?rev=72112&r1=72111&r2=72112&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/Spiller.h (original)
+++ llvm/trunk/lib/CodeGen/Spiller.h Tue May 19 12:50:18 2009
@@ -13,6 +13,7 @@
 #include <vector>
 
 namespace llvm {
+  struct LiveInterval;
 
   /// Spiller interface.
   ///
@@ -21,7 +22,7 @@
   class Spiller {
   public:
     virtual ~Spiller() = 0;
-    virtual std::vector<class LiveInterval*> spill(class LiveInterval *li) = 0;
+    virtual std::vector<LiveInterval*> spill(class LiveInterval *li) = 0;
   };
 
   /// Create and return a spiller object, as specified on the command line.




From isanbard at gmail.com  Tue May 19 12:50:32 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 19 May 2009 10:50:32 -0700
Subject: [llvm-commits] [PATCH] Fix C2888 compile error in VS2008
In-Reply-To: <f126fae00905190244u48ac8319t32e8777ce420af5b@mail.gmail.com>
References: <f126fae00905190244u48ac8319t32e8777ce420af5b@mail.gmail.com>
Message-ID: <16e5fdf90905191050i642307e6j2084e1c02aedb989@mail.gmail.com>

 I applied a modified version of your patch. Thanks!

-bw

On Tue, May 19, 2009 at 2:44 AM, Howard Su <howard0su at gmail.com> wrote:
> The attached patch to fix two compiler error C2888?when compiling under
> VS2008.
>
> --
> -Howard
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>



From isanbard at gmail.com  Tue May 19 12:51:27 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 19 May 2009 17:51:27 -0000
Subject: [llvm-commits] [llvm] r72113 - /llvm/trunk/lib/CodeGen/Spiller.h
Message-ID: <200905191751.n4JHpRl7022152@zion.cs.uiuc.edu>

Author: void
Date: Tue May 19 12:51:26 2009
New Revision: 72113

URL: http://llvm.org/viewvc/llvm-project?rev=72113&view=rev
Log:
Remove 'class' specifier.

Modified:
    llvm/trunk/lib/CodeGen/Spiller.h

Modified: llvm/trunk/lib/CodeGen/Spiller.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.h?rev=72113&r1=72112&r2=72113&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/Spiller.h (original)
+++ llvm/trunk/lib/CodeGen/Spiller.h Tue May 19 12:51:26 2009
@@ -22,12 +22,12 @@
   class Spiller {
   public:
     virtual ~Spiller() = 0;
-    virtual std::vector<LiveInterval*> spill(class LiveInterval *li) = 0;
+    virtual std::vector<LiveInterval*> spill(LiveInterval *li) = 0;
   };
 
   /// Create and return a spiller object, as specified on the command line.
-  Spiller* createSpiller(class MachineFunction *mf, class LiveIntervals *li,
-                         class VirtRegMap *vrm);
+  Spiller* createSpiller(MachineFunction *mf, LiveIntervals *li,
+                         VirtRegMap *vrm);
 }
 
 #endif




From isanbard at gmail.com  Tue May 19 12:52:32 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Tue, 19 May 2009 17:52:32 -0000
Subject: [llvm-commits] [llvm] r72114 - /llvm/trunk/lib/CodeGen/Spiller.h
Message-ID: <200905191752.n4JHqWOS022197@zion.cs.uiuc.edu>

Author: void
Date: Tue May 19 12:52:31 2009
New Revision: 72114

URL: http://llvm.org/viewvc/llvm-project?rev=72114&view=rev
Log:
Forward decls.

Modified:
    llvm/trunk/lib/CodeGen/Spiller.h

Modified: llvm/trunk/lib/CodeGen/Spiller.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/Spiller.h?rev=72114&r1=72113&r2=72114&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/Spiller.h (original)
+++ llvm/trunk/lib/CodeGen/Spiller.h Tue May 19 12:52:31 2009
@@ -14,6 +14,9 @@
 
 namespace llvm {
   struct LiveInterval;
+  class LiveIntervals;
+  class MachineFunction;
+  class VirtRegMap;
 
   /// Spiller interface.
   ///




From bob.wilson at apple.com  Tue May 19 13:33:35 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Tue, 19 May 2009 18:33:35 -0000
Subject: [llvm-commits] [llvm] r72115 -
	/llvm/trunk/lib/Target/ARM/ARMRegisterInfo.cpp
Message-ID: <200905191833.n4JIXb9F023834@zion.cs.uiuc.edu>

Author: bwilson
Date: Tue May 19 13:33:02 2009
New Revision: 72115

URL: http://llvm.org/viewvc/llvm-project?rev=72115&view=rev
Log:
Follow up on new support for memory operands in ARM inline assembly.
This fixes pr4233.

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

Modified: llvm/trunk/lib/Target/ARM/ARMRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMRegisterInfo.cpp?rev=72115&r1=72114&r2=72115&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/ARM/ARMRegisterInfo.cpp Tue May 19 13:33:02 2009
@@ -688,6 +688,10 @@
   unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask);
   bool isSub = false;
 
+  // Memory operands in inline assembly always use AddrMode2.
+  if (Opcode == ARM::INLINEASM)
+    AddrMode = ARMII::AddrMode2;
+
   if (Opcode == ARM::ADDri) {
     Offset += MI.getOperand(i+1).getImm();
     if (Offset == 0) {




From gohman at apple.com  Tue May 19 14:18:45 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 19:18:45 -0000
Subject: [llvm-commits] [llvm] r72120 -
	/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
Message-ID: <200905191918.n4JJItIS025462@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 14:18:01 2009
New Revision: 72120

URL: http://llvm.org/viewvc/llvm-project?rev=72120&view=rev
Log:
Create ConstantExpr GEPs the correct way. This fixes
MultiSource/Benchmarks/Prolangs-C/football and a variety of other
failures.

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

Modified: llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp?rev=72120&r1=72119&r2=72120&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolutionExpander.cpp Tue May 19 14:18:01 2009
@@ -246,7 +246,7 @@
     // Fold a GEP with constant operands.
     if (Constant *CLHS = dyn_cast<Constant>(V))
       if (Constant *CRHS = dyn_cast<Constant>(Idx))
-        return ConstantExpr::get(Instruction::GetElementPtr, CLHS, CRHS);
+        return ConstantExpr::getGetElementPtr(CLHS, &CRHS, 1);
 
     // Do a quick scan to see if we have this GEP nearby.  If so, reuse it.
     unsigned ScanLimit = 6;




From criswell at uiuc.edu  Tue May 19 14:22:02 2009
From: criswell at uiuc.edu (John Criswell)
Date: Tue, 19 May 2009 19:22:02 -0000
Subject: [llvm-commits] [poolalloc] r72121 -
	/poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp
Message-ID: <200905191922.n4JJM7A3025609@zion.cs.uiuc.edu>

Author: criswell
Date: Tue May 19 14:21:46 2009
New Revision: 72121

URL: http://llvm.org/viewvc/llvm-project?rev=72121&view=rev
Log:
Added code to merge metapools together for objects allocated out of the
same kmem_cache_t.
Re-ordered header files per LLVM coding standards.

Modified:
    poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp

Modified: poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp?rev=72121&r1=72120&r2=72121&view=diff

==============================================================================
--- poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp (original)
+++ poolalloc/branches/SVA/lib/DSA/TopDownClosure.cpp Tue May 19 14:21:46 2009
@@ -14,14 +14,18 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "dsa/DataStructure.h"
-#include "llvm/Module.h"
+#include "llvm/ADT/Statistic.h"
+#include "llvm/Instructions.h"
 #include "llvm/DerivedTypes.h"
-#include "dsa/DSGraph.h"
+#include "llvm/Module.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/Timer.h"
-#include "llvm/ADT/Statistic.h"
+
+#include "dsa/DataStructure.h"
+#include "dsa/DSGraph.h"
+
 #include <iostream>
+
 using namespace llvm;
 
 #if 0
@@ -149,7 +153,57 @@
 	   ii != ee; ++ii)
 	ii->getMP()->addFlags(ii->getNodeFlags());
     }
-      
+
+#ifdef LLVA_KERNEL
+  //
+  // Ugly hack:
+  //
+  // Memory objects returned from kmem_cache_alloc() alias if they are
+  // allocated from the same kernel pool.  This code forces them to have the
+  // same MetaPool.
+  //
+  Function* KMA = M.getNamedFunction("kmem_cache_alloc");
+  if (KMA) {
+    // Map from kmem_cache_t's metapool to the metapool of its return value
+    std::map<MetaPool*, MetaPool*> locs;
+
+    for (Value::use_iterator ii = KMA->use_begin(), ee = KMA->use_end();
+         ii != ee; ++ii) {
+      CallInst* CI = dyn_cast<CallInst>(*ii);
+      if ((CI) && (CI->getCalledFunction() == KMA)) {
+        // Function in which the call statement resides
+        Function * F = CI->getParent()->getParent();
+
+        // The pointer to the kmem_cache_t
+        Value* CacheT = CI->getOperand(1);
+
+        //
+        // Get the metapool for the kmem_cache_t
+        //
+        DSNodeHandle DSCacheT = DSInfo[F]->getNodeForValue(CacheT);
+        MetaPoolHandle MPCacheT (DSCacheT.getNode()->getMP());
+
+        //
+        // Get the DSNode handle of the object being allocated.
+        //
+        DSNodeHandle DSH = DSInfo[F]->getNodeForValue(CI);
+        MetaPoolHandle MPNode (DSH.getNode()->getMP());
+
+        //
+        // If the allocated object does not belong to the same metapool as
+        // other objects allocated from this kmem_cache_t, merge their
+        // metapools so that they do.
+        //
+        if (locs[MPCacheT.getPool()] != MPNode.getPool()) {
+          std::cerr << "kmem_cache_alloc recovered merge\n";
+          DSH.getNode()->getMP()->merge(locs[MPCacheT.getPool()]);
+        }
+        locs[MPCacheT.getPool()] = DSH.getNode()->getMP();
+      }
+    }
+  }
+#endif
+     
   return false;
 }
 




From gohman at apple.com  Tue May 19 14:22:49 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 19:22:49 -0000
Subject: [llvm-commits] [llvm] r72122 - in /llvm/trunk:
 include/llvm/Analysis/ScalarEvolution.h lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905191922.n4JJMoKw025685@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 14:22:47 2009
New Revision: 72122

URL: http://llvm.org/viewvc/llvm-project?rev=72122&view=rev
Log:
Make SCEVCallbackVH a private nested class inside ScalarEvolution, as
it's an implementation detail.

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

Modified: llvm/trunk/include/llvm/Analysis/ScalarEvolution.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/ScalarEvolution.h?rev=72122&r1=72121&r2=72122&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/ScalarEvolution.h (original)
+++ llvm/trunk/include/llvm/Analysis/ScalarEvolution.h Tue May 19 14:22:47 2009
@@ -145,16 +145,6 @@
     static bool classof(const SCEV *S);
   };
 
-  /// SCEVCallbackVH - A CallbackVH to arrange for ScalarEvolution to be
-  /// notified whenever a Value is deleted.
-  class SCEVCallbackVH : public CallbackVH {
-    ScalarEvolution *SE;
-    virtual void deleted();
-    virtual void allUsesReplacedWith(Value *New);
-  public:
-    SCEVCallbackVH(Value *V, ScalarEvolution *SE = 0);
-  };
-
   /// SCEVHandle - This class is used to maintain the SCEV object's refcounts,
   /// freeing the objects when the last reference is dropped.
   class SCEVHandle {
@@ -212,6 +202,16 @@
   /// they must ask this class for services.
   ///
   class ScalarEvolution : public FunctionPass {
+    /// SCEVCallbackVH - A CallbackVH to arrange for ScalarEvolution to be
+    /// notified whenever a Value is deleted.
+    class SCEVCallbackVH : public CallbackVH {
+      ScalarEvolution *SE;
+      virtual void deleted();
+      virtual void allUsesReplacedWith(Value *New);
+    public:
+      SCEVCallbackVH(Value *V, ScalarEvolution *SE = 0);
+    };
+
     friend class SCEVCallbackVH;
     friend class SCEVExpander;
 

Modified: llvm/trunk/lib/Analysis/ScalarEvolution.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ScalarEvolution.cpp?rev=72122&r1=72121&r2=72122&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue May 19 14:22:47 2009
@@ -3674,7 +3674,7 @@
 //                   SCEVCallbackVH Class Implementation
 //===----------------------------------------------------------------------===//
 
-void SCEVCallbackVH::deleted() {
+void ScalarEvolution::SCEVCallbackVH::deleted() {
   assert(SE && "SCEVCallbackVH called with a non-null ScalarEvolution!");
   if (PHINode *PN = dyn_cast<PHINode>(getValPtr()))
     SE->ConstantEvolutionLoopExitValue.erase(PN);
@@ -3684,7 +3684,7 @@
   // this now dangles!
 }
 
-void SCEVCallbackVH::allUsesReplacedWith(Value *) {
+void ScalarEvolution::SCEVCallbackVH::allUsesReplacedWith(Value *) {
   assert(SE && "SCEVCallbackVH called with a non-null ScalarEvolution!");
 
   // Forget all the expressions associated with users of the old value,
@@ -3724,7 +3724,7 @@
   // this may dangle!
 }
 
-SCEVCallbackVH::SCEVCallbackVH(Value *V, ScalarEvolution *se)
+ScalarEvolution::SCEVCallbackVH::SCEVCallbackVH(Value *V, ScalarEvolution *se)
   : CallbackVH(V), SE(se) {}
 
 //===----------------------------------------------------------------------===//




From luked at cs.rochester.edu  Tue May 19 09:55:10 2009
From: luked at cs.rochester.edu (Luke Dalessandro)
Date: Tue, 19 May 2009 10:55:10 -0400
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>	<5464957333051772950217202212009693918-Webmail@me.com>	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
Message-ID: <4A12C84E.3010203@cs.rochester.edu>

Owen Anderson wrote:
> 
> On May 19, 2009, at 3:08 AM, Chris Lattner wrote:
>> The basic algorithm looks fine, however:
>>
>> 3. Why do you create your own mutex out of atomic operations instead
>> of just using a pthreads lock?  spinning is not always an efficient
>> solution to contention, particularly if your machine really only has
>> one cpu! (in which case you end up spinning away the rest of your
>> timeslice)
> 
> Portability.  llvm::sys::Mutex _cannot_ be made to work, because on Windows
> it has to have a non-trivial constructor.

If there was just one static lock for the class, could you stick it 
inside the LazyInit() routine and get the non-trivial constructor to be 
threadsafe on Windows?

   void LazyInit() const {
     static sys::Mutex lock(false);
     etc...
   }

I'm actually curious about the answer. What does windows do here? Does 
it default to the equivalent of gcc's -fno-threadsafe-statics?

Luke

> 
> Also, this is pretty much the ideal case for spinlocks.  Initialization 
> only happens once and
> we're using a separate mutex for each object, so we don't expect 
> significant contention over
> the life of the program.   Plus, the initialization itself should be 
> fairly fast.  This means that
>  the spinning thread can continue immediately after the lock is 
> released, rather than having
>  to wait for the OS to wake it up.
> 
>> 4. Instead of going through the trouble to make the locking more
>> efficient, I think it would be good to special case the scenario when
>> there is only one LLVM thread.  Why not make clients explicitly opt-in
>> to multithreaded llvm, by making an explicit llvm_multithread call?
>> If multithreading is disabled, this call should return an error.  If
>> enabled, it would set a global.  All the various "locking" clients
>> could just check the global before taking potential heavy-weight locks.
> 
> We already handle this in a somewhat different way.  When threading is 
> disabled,
> llvm::sys::Mutex operations are turned into no-ops, and MemoryFence() and
> CompareAndSwap() are given non-atomic implementations.
> 
> --Owen
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits



From cristic at cs.stanford.edu  Tue May 19 14:29:19 2009
From: cristic at cs.stanford.edu (Cristian Cadar)
Date: Tue, 19 May 2009 19:29:19 -0000
Subject: [llvm-commits] [klee] r72123 - /klee/trunk/www/tutorials.html
Message-ID: <200905191929.n4JJTKmV025902@zion.cs.uiuc.edu>

Author: cristic
Date: Tue May 19 14:29:16 2009
New Revision: 72123

URL: http://llvm.org/viewvc/llvm-project?rev=72123&view=rev
Log:
Added empty www/tutorials.html page (test commit to klee repository).

Added:
    klee/trunk/www/tutorials.html

Added: klee/trunk/www/tutorials.html
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/tutorials.html?rev=72123&view=auto

==============================================================================
--- klee/trunk/www/tutorials.html (added)
+++ klee/trunk/www/tutorials.html Tue May 19 14:29:16 2009
@@ -0,0 +1,20 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 
+          "http://www.w3.org/TR/html4/strict.dtd">
+<!-- Material used from: HTML 4.01 specs: http://www.w3.org/TR/html401/ -->
+<html>
+<head>
+  <META http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
+  <title>The klee Symbolic Virtual Machine</title>
+  <link type="text/css" rel="stylesheet" href="menu.css">
+  <link type="text/css" rel="stylesheet" href="content.css">
+</head>
+<body>
+<include virtual="menu.html.incl">
+<div id="content">
+  <!--*********************************************************************-->
+  <h1>KLEE Tutorials</h1>
+  <!--*********************************************************************-->
+
+</div>
+</body>
+</html>




From criswell at uiuc.edu  Tue May 19 14:36:08 2009
From: criswell at uiuc.edu (John Criswell)
Date: Tue, 19 May 2009 19:36:08 -0000
Subject: [llvm-commits] [poolalloc] r72124 -
	/poolalloc/branches/SVA/lib/DSA/Local.cpp
Message-ID: <200905191936.n4JJaAel026115@zion.cs.uiuc.edu>

Author: criswell
Date: Tue May 19 14:35:52 2009
New Revision: 72124

URL: http://llvm.org/viewvc/llvm-project?rev=72124&view=rev
Log:
Moved the kmem_cache_alloc() hack from the local DSA pass to the Top-Down DSA
pass (where more aliasing information is available).

Modified:
    poolalloc/branches/SVA/lib/DSA/Local.cpp

Modified: poolalloc/branches/SVA/lib/DSA/Local.cpp
URL: http://llvm.org/viewvc/llvm-project/poolalloc/branches/SVA/lib/DSA/Local.cpp?rev=72124&r1=72123&r2=72124&view=diff

==============================================================================
--- poolalloc/branches/SVA/lib/DSA/Local.cpp (original)
+++ poolalloc/branches/SVA/lib/DSA/Local.cpp Tue May 19 14:35:52 2009
@@ -1911,33 +1911,6 @@
       EliminateUsesOfECGlobals(*I->second, ECGlobals);
   }
 
-#ifdef LLVA_KERNEL
-
-  // Ugly hack.  kmem_cache_allocs are in the same pool also if the kmem_cache_t's are the same
-  // this only works on global kmem_cache_ts
-  Function* KMA = M.getNamedFunction("kmem_cache_alloc");
-  if (KMA) {
-    for (Value::use_iterator ii = KMA->use_begin(), ee = KMA->use_end();
-         ii != ee; ++ii) {
-      std::map<Value*, MetaPool*> locs;
-      if (CallInst* CI = dyn_cast<CallInst>(*ii)) {
-        if (CI->getCalledFunction() == KMA && isa<GlobalValue>(CI->getOperand(1))) {
-          Value* V = CI->getOperand(1); //the kmem_cache_alloc
-          DSNodeHandle DSH = DSInfo[CI->getParent()->getParent()]->getNodeForValue(CI);
-          MetaPoolHandle L(locs[V]), N(DSH.getNode()->getMP());
-          if (L.getPool() != N.getPool()) {
-            std::cerr << "kmem_cache_alloc recovered merge\n";
-            MetaPoolHandle L(locs[V]), N(DSH.getNode()->getMP());
-            DSH.getNode()->getMP()->merge(locs[V]);
-          }
-          locs[V] = DSH.getNode()->getMP();
-        }
-      }
-    }
-  }
-
-#endif
-
   return true;
 }
 




From clattner at apple.com  Tue May 19 14:41:25 2009
From: clattner at apple.com (Chris Lattner)
Date: Tue, 19 May 2009 12:41:25 -0700
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
Message-ID: <8DD69FC6-D713-4451-899D-C8FE8D508E75@apple.com>

>> 2. Again, we only need a single global lock, we do not need a mutex
>> per ManagedStatic.
>
> Conversely, why have only one?  Also, see my answer to #3.

Because it takes less space.

>> 3. Why do you create your own mutex out of atomic operations instead
>> of just using a pthreads lock?  spinning is not always an efficient
>> solution to contention, particularly if your machine really only has
>> one cpu! (in which case you end up spinning away the rest of your
>> timeslice)
>
> Portability.  llvm::sys::Mutex _cannot_ be made to work, because on  
> Windows
> it has to have a non-trivial constructor.

This is irrelevant.  We are going to have a bunch of "global" state  
for all the stuff that uniques stuff.  The managed static is just the  
first piece of it.  When the app calls llvm_start_multithread_mode(),  
this function can set up the global mutex and do whatever else it  
wants to do, while the app is still single threaded.  This means that  
it is perfectly safe to have:

Mutex *TheGlobalMutex;

and use:

TheGlobalMutex->lock();

places, because you know that TheGlobalMutex will be allocated and  
initialized when the app is single threaded.


> Also, this is pretty much the ideal case for spinlocks.   
> Initialization only happens once and
> we're using a separate mutex for each object, so we don't expect  
> significant contention over
> the life of the program.   Plus, the initialization itself should be  
> fairly fast.  This means that
> the spinning thread can continue immediately after the lock is  
> released, rather than having
> to wait for the OS to wake it up.

ManagedStatic is a generic class, meaning that it can take an  
arbitrary amount of time to initialize.  One initialization can  
potentially also cause initialization of other objects: make sure the  
global mutex is recursive.

>>
>> efficient, I think it would be good to special case the scenario when
>> there is only one LLVM thread.  Why not make clients explicitly opt- 
>> in
>> to multithreaded llvm, by making an explicit llvm_multithread call?
>> If multithreading is disabled, this call should return an error.  If
>> enabled, it would set a global.  All the various "locking" clients
>> could just check the global before taking potential heavy-weight  
>> locks.
>
> We already handle this in a somewhat different way.  When threading  
> is disabled,
> llvm::sys::Mutex operations are turned into no-ops, and  
> MemoryFence() and
> CompareAndSwap() are given non-atomic implementations.

You are confusing the case when llvm is compiled without  
multithreading support with the case where it is compiled with  
threading support but that app is not using it.

-Chris


From resistor at mac.com  Tue May 19 15:18:07 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 15:18:07 -0500
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <4A12C84E.3010203@cs.rochester.edu>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
	<4A12C84E.3010203@cs.rochester.edu>
Message-ID: <87525340677602673349810574238295973857-Webmail@me.com>

 
On Tuesday, May 19, 2009, at 09:55AM, "Luke Dalessandro" <luked at cs.rochester.edu> wrote:

>
>If there was just one static lock for the class, could you stick it 
>inside the LazyInit() routine and get the non-trivial constructor to be 
>threadsafe on Windows?
>
>   void LazyInit() const {
>     static sys::Mutex lock(false);
>     etc...
>   }

Unfortunately, we're not implementing Singletons here.  There can well be more than one static instance of the class.

--Owen


From resistor at mac.com  Tue May 19 15:20:48 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 15:20:48 -0500
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <8DD69FC6-D713-4451-899D-C8FE8D508E75@apple.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
	<8DD69FC6-D713-4451-899D-C8FE8D508E75@apple.com>
Message-ID: <16551545510839478801965352385268027480-Webmail@me.com>

Alright, is this something more like what you had in mind?

--Owen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: ManagedStatic2.diff
Type: application/octet-stream
Size: 6466 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090519/d23ae220/attachment.obj 

From gohman at apple.com  Tue May 19 15:35:26 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 20:35:26 -0000
Subject: [llvm-commits] [llvm] r72130 - in
 /llvm/trunk/lib/Transforms/Scalar: IndVarSimplify.cpp
 LoopStrengthReduce.cpp
Message-ID: <200905192035.n4JKZQ8S028161@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 15:35:26 2009
New Revision: 72130

URL: http://llvm.org/viewvc/llvm-project?rev=72130&view=rev
Log:
Trim unneeded #includes.

Modified:
    llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
    llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp?rev=72130&r1=72129&r2=72130&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Tue May 19 15:35:26 2009
@@ -51,12 +51,10 @@
 #include "llvm/Support/CFG.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/Debug.h"
-#include "llvm/Support/GetElementPtrTypeIterator.h"
 #include "llvm/Transforms/Utils/Local.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Support/CommandLine.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/ADT/SetVector.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/STLExtras.h"
 using namespace llvm;

Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=72130&r1=72129&r2=72130&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Tue May 19 15:35:26 2009
@@ -27,7 +27,6 @@
 #include "llvm/Transforms/Utils/AddrModeMatcher.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
-#include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Support/CFG.h"
 #include "llvm/Support/Debug.h"




From gohman at apple.com  Tue May 19 15:37:36 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 20:37:36 -0000
Subject: [llvm-commits] [llvm] r72131 -
	/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
Message-ID: <200905192037.n4JKbaL7028343@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 15:37:36 2009
New Revision: 72131

URL: http://llvm.org/viewvc/llvm-project?rev=72131&view=rev
Log:
Add some more comments to the top of this file.

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

Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=72131&r1=72130&r2=72131&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Tue May 19 15:37:36 2009
@@ -7,8 +7,15 @@
 //
 //===----------------------------------------------------------------------===//
 //
+// This transformation analyzes and transforms the induction variables (and
+// computations derived from them) into forms suitable for efficient execution
+// on the target.
+//
 // This pass performs a strength reduction on array references inside loops that
-// have as one or more of their components the loop induction variable.
+// have as one or more of their components the loop induction variable, it
+// rewrites expressions to take advantage of scaled-index addressing modes
+// available on the target, and it performs a variety of other optimizations
+// related to loop induction variables.
 //
 //===----------------------------------------------------------------------===//
 




From gohman at apple.com  Tue May 19 15:38:47 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 20:38:47 -0000
Subject: [llvm-commits] [llvm] r72132 -
	/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
Message-ID: <200905192038.n4JKclu7028417@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 15:38:47 2009
New Revision: 72132

URL: http://llvm.org/viewvc/llvm-project?rev=72132&view=rev
Log:
Remove an irrelevant comment.

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

Modified: llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp?rev=72132&r1=72131&r2=72132&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/IndVarSimplify.cpp Tue May 19 15:38:47 2009
@@ -31,9 +31,7 @@
 //      expression, this transformation will make the loop dead.
 //
 // This transformation should be followed by strength reduction after all of the
-// desired loop transformations have been performed.  Additionally, on targets
-// where it is profitable, the loop could be transformed to count down to zero
-// (the "do loop" optimization).
+// desired loop transformations have been performed.
 //
 //===----------------------------------------------------------------------===//
 




From nicolas.geoffray at lip6.fr  Tue May 19 15:37:19 2009
From: nicolas.geoffray at lip6.fr (Nicolas Geoffray)
Date: Tue, 19 May 2009 22:37:19 +0200
Subject: [llvm-commits] [llvm]
	r72045	-	/llvm/trunk/lib/Target/X86/X86InstrInfo.td
In-Reply-To: <8AD27C27-C211-4F58-A924-05FAF49952F0@apple.com>
References: <200905182142.n4ILgWgP031346@zion.cs.uiuc.edu>	<4A127AEC.1070608@lip6.fr>
	<8AD27C27-C211-4F58-A924-05FAF49952F0@apple.com>
Message-ID: <4A13187F.4040306@lip6.fr>

Apologies Dale, your patch didn't cause my errors. I realized it's the 
new atomics checks that cause them.

Sorry for the noise.

Nicolas

Dale Johannesen wrote:
> On May 19, 2009, at 2:25 AMPDT, Nicolas Geoffray wrote:
>
>   
>> Hi Dale,
>>
>> Dale Johannesen wrote:
>>     
>>> Author: johannes
>>> Date: Mon May 18 16:41:59 2009
>>> New Revision: 72045
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=72045&view=rev
>>> Log:
>>> Add OpSize to 16-bit ADC and SBB.
>>>
>>>       
>> What is OpSize? This patch is causing lots of memory errors in VMKit,
>> I'm not sure why.
>>     
>
> It indicates the operand size prefix, 0x66, should be emitted.  It is  
> normal on 16-bit operations.
>
>   
>> Does the JIT in LLVM understand OpSize?
>>     
>
> Yes, in fact it's entirely for use by the JIT.  See XCodeEmitter.
>
> Are you sure it's this patch?  I'd be surprised if these new patterns  
> were getting matched at all right now.
>
>   
>> Thanks,
>> Nicolas
>>
>>     
>>> 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=72045&r1=72044&r2=72045&view=diff
>>>
>>> = 
>>> = 
>>> = 
>>> = 
>>> = 
>>> = 
>>> = 
>>> = 
>>> = 
>>> =====================================================================
>>> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
>>> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Mon May 18 16:41:59  
>>> 2009
>>> @@ -2280,7 +2280,7 @@
>>> def ADC16rr  : I<0x11, MRMDestReg, (outs GR16:$dst),
>>>                                    (ins GR16:$src1, GR16:$src2),
>>>                  "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                 [(set GR16:$dst, (adde GR16:$src1, GR16:$src2))]>;
>>> +                 [(set GR16:$dst, (adde GR16:$src1,  
>>> GR16:$src2))]>, OpSize;
>>> def ADC32rr  : I<0x11, MRMDestReg, (outs GR32:$dst),
>>>                                    (ins GR32:$src1, GR32:$src2),
>>>                  "adc{l}\t{$src2, $dst|$dst, $src2}",
>>> @@ -2293,7 +2293,8 @@
>>> def ADC16rm  : I<0x13, MRMSrcMem , (outs GR16:$dst),
>>>                                    (ins GR16:$src1, i16mem:$src2),
>>>                  "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                 [(set GR16:$dst, (adde GR16:$src1, (load addr: 
>>> $src2)))]>;
>>> +                 [(set GR16:$dst, (adde GR16:$src1, (load addr: 
>>> $src2)))]>,
>>> +                 OpSize;
>>> def ADC32rm  : I<0x13, MRMSrcMem , (outs GR32:$dst),
>>>                                    (ins GR32:$src1, i32mem:$src2),
>>>                  "adc{l}\t{$src2, $dst|$dst, $src2}",
>>> @@ -2304,11 +2305,12 @@
>>> def ADC16ri  : Ii16<0x81, MRM2r, (outs GR16:$dst),
>>>                                  (ins GR16:$src1, i16imm:$src2),
>>>                     "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>;
>>> +                 [(set GR16:$dst, (adde GR16:$src1, imm:$src2))]>,  
>>> OpSize;
>>> def ADC16ri8 : Ii8<0x83, MRM2r, (outs GR16:$dst),
>>>                                 (ins GR16:$src1, i16i8imm:$src2),
>>>                    "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                 [(set GR16:$dst, (adde GR16:$src1,  
>>> i16immSExt8:$src2))]>;
>>> +                 [(set GR16:$dst, (adde GR16:$src1,  
>>> i16immSExt8:$src2))]>,
>>> +                 OpSize;
>>> def ADC32ri  : Ii32<0x81, MRM2r, (outs GR32:$dst),
>>>                                  (ins GR32:$src1, i32imm:$src2),
>>>                     "adc{l}\t{$src2, $dst|$dst, $src2}",
>>> @@ -2324,7 +2326,8 @@
>>>                    [(store (adde (load addr:$dst), GR8:$src2), addr: 
>>> $dst)]>;
>>>   def ADC16mr  : I<0x11, MRMDestMem, (outs), (ins i16mem:$dst,  
>>> GR16:$src2),
>>>                    "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                   [(store (adde (load addr:$dst), GR16:$src2),  
>>> addr:$dst)]>;
>>> +                   [(store (adde (load addr:$dst), GR16:$src2),  
>>> addr:$dst)]>,
>>> +                   OpSize;
>>>   def ADC32mr  : I<0x11, MRMDestMem, (outs), (ins i32mem:$dst,  
>>> GR32:$src2),
>>>                    "adc{l}\t{$src2, $dst|$dst, $src2}",
>>>                    [(store (adde (load addr:$dst), GR32:$src2),  
>>> addr:$dst)]>;
>>> @@ -2333,10 +2336,12 @@
>>>                   [(store (adde (loadi8 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>>   def ADC16mi  : Ii16<0x81, MRM2m, (outs), (ins i16mem:$dst, i16imm: 
>>> $src2),
>>>                       "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -                  [(store (adde (loadi16 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>> +                  [(store (adde (loadi16 addr:$dst), imm:$src2),  
>>> addr:$dst)]>,
>>> +                  OpSize;
>>>   def ADC16mi8 : Ii8<0x83, MRM2m, (outs), (ins i16mem:$dst,  
>>> i16i8imm :$src2),
>>>                      "adc{w}\t{$src2, $dst|$dst, $src2}",
>>> -               [(store (adde (load addr:$dst), i16immSExt8:$src2),  
>>> addr:$dst)]>;
>>> +               [(store (adde (load addr:$dst), i16immSExt8:$src2),  
>>> addr:$dst)]>,
>>> +               OpSize;
>>>   def ADC32mi  : Ii32<0x81, MRM2m, (outs), (ins i32mem:$dst, i32imm: 
>>> $src2),
>>>                       "adc{l}\t{$src2, $dst|$dst, $src2}",
>>>                   [(store (adde (loadi32 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>> @@ -2452,7 +2457,7 @@
>>> def SBB16rr    : I<0x19, MRMDestReg, (outs GR16:$dst),
>>>                                      (ins GR16:$src1, GR16:$src2),
>>>                   "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                 [(set GR16:$dst, (sube GR16:$src1, GR16:$src2))]>;
>>> +                 [(set GR16:$dst, (sube GR16:$src1,  
>>> GR16:$src2))]>, OpSize;
>>> def SBB32rr    : I<0x19, MRMDestReg, (outs GR32:$dst),
>>>                                       (ins GR32:$src1, GR32:$src2),
>>>                   "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>> @@ -2464,7 +2469,8 @@
>>>                    [(store (sube (load addr:$dst), GR8:$src2), addr: 
>>> $dst)]>;
>>>   def SBB16mr  : I<0x19, MRMDestMem, (outs), (ins i16mem:$dst,  
>>> GR16:$src2),
>>>                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                   [(store (sube (load addr:$dst), GR16:$src2),  
>>> addr:$dst)]>;
>>> +                   [(store (sube (load addr:$dst), GR16:$src2),  
>>> addr:$dst)]>,
>>> +                   OpSize;
>>>   def SBB32mr  : I<0x19, MRMDestMem, (outs), (ins i32mem:$dst,  
>>> GR32:$src2),
>>>                    "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>>                    [(store (sube (load addr:$dst), GR32:$src2),  
>>> addr:$dst)]>;
>>> @@ -2473,10 +2479,12 @@
>>>                    [(store (sube (loadi8 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>>   def SBB16mi  : Ii16<0x81, MRM3m, (outs), (ins i16mem:$dst, i16imm: 
>>> $src2),
>>>                       "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                  [(store (sube (loadi16 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>> +                  [(store (sube (loadi16 addr:$dst), imm:$src2),  
>>> addr:$dst)]>,
>>> +                  OpSize;
>>>   def SBB16mi8 : Ii8<0x83, MRM3m, (outs), (ins i16mem:$dst,  
>>> i16i8imm :$src2),
>>>                      "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -               [(store (sube (load addr:$dst), i16immSExt8:$src2),  
>>> addr:$dst)]>;
>>> +               [(store (sube (load addr:$dst), i16immSExt8:$src2),  
>>> addr:$dst)]>,
>>> +               OpSize;
>>>   def SBB32mi  : Ii32<0x81, MRM3m, (outs), (ins i32mem:$dst, i32imm: 
>>> $src2),
>>>                       "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>>                   [(store (sube (loadi32 addr:$dst), imm:$src2),  
>>> addr:$dst)]>;
>>> @@ -2490,7 +2498,8 @@
>>> def SBB16rm  : I<0x1B, MRMSrcMem, (outs GR16:$dst),
>>>                                   (ins GR16:$src1, i16mem:$src2),
>>>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                    [(set GR16:$dst, (sube GR16:$src1, (load addr: 
>>> $src2)))]>;
>>> +                    [(set GR16:$dst, (sube GR16:$src1, (load addr: 
>>> $src2)))]>,
>>> +                    OpSize;
>>> def SBB32rm  : I<0x1B, MRMSrcMem, (outs GR32:$dst),
>>>                                   (ins GR32:$src1, i32mem:$src2),
>>>                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>> @@ -2501,11 +2510,12 @@
>>> def SBB16ri  : Ii16<0x81, MRM3r, (outs GR16:$dst),
>>>                                  (ins GR16:$src1, i16imm:$src2),
>>>                     "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                    [(set GR16:$dst, (sube GR16:$src1, imm: 
>>> $src2))]>;
>>> +                    [(set GR16:$dst, (sube GR16:$src1, imm: 
>>> $src2))]>, OpSize;
>>> def SBB16ri8 : Ii8<0x83, MRM3r, (outs GR16:$dst),
>>>                                 (ins GR16:$src1, i16i8imm:$src2),
>>>                    "sbb{w}\t{$src2, $dst|$dst, $src2}",
>>> -                   [(set GR16:$dst, (sube GR16:$src1,  
>>> i16immSExt8:$src2))]>;
>>> +                   [(set GR16:$dst, (sube GR16:$src1,  
>>> i16immSExt8:$src2))]>,
>>> +                   OpSize;
>>> def SBB32ri  : Ii32<0x81, MRM3r, (outs GR32:$dst),
>>>                                  (ins GR32:$src1, i32imm:$src2),
>>>                     "sbb{l}\t{$src2, $dst|$dst, $src2}",
>>>
>>>
>>> _______________________________________________
>>> 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
>>     
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>   



From bob.wilson at apple.com  Tue May 19 16:27:58 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Tue, 19 May 2009 21:27:58 -0000
Subject: [llvm-commits] [llvm] r72138 -
	/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
Message-ID: <200905192127.n4JLRwG2030089@zion.cs.uiuc.edu>

Author: bwilson
Date: Tue May 19 16:27:57 2009
New Revision: 72138

URL: http://llvm.org/viewvc/llvm-project?rev=72138&view=rev
Log:
Fix pr4227: Handle large immediate values in inline assembly.

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

Modified: llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp?rev=72138&r1=72137&r2=72138&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp (original)
+++ llvm/trunk/lib/Target/ARM/AsmPrinter/ARMAsmPrinter.cpp Tue May 19 16:27:57 2009
@@ -294,7 +294,7 @@
     if (!Modifier || strcmp(Modifier, "no_hash") != 0)
       O << "#";
 
-    O << (int)MO.getImm();
+    O << MO.getImm();
     break;
   }
   case MachineOperand::MO_MachineBasicBlock:




From resistor at mac.com  Tue May 19 17:18:56 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 22:18:56 -0000
Subject: [llvm-commits] [llvm] r72140 - in /llvm/trunk:
 autoconf/configure.ac configure include/llvm/System/Atomic.h
Message-ID: <200905192218.n4JMIv6n031660@zion.cs.uiuc.edu>

Author: resistor
Date: Tue May 19 17:18:56 2009
New Revision: 72140

URL: http://llvm.org/viewvc/llvm-project?rev=72140&view=rev
Log:
Split the ENABLE_THREADS #define in two.  Now ENABLE_THREADS indicates the ability to run multiple threads at once in the JIT, and requires only
mutex support.  LLVM_MULTITHREADED indicates (or will indicate) the ability to run LLVM itself across multiple threads, and requires atomics support.

Modified:
    llvm/trunk/autoconf/configure.ac
    llvm/trunk/configure
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/autoconf/configure.ac?rev=72140&r1=72139&r2=72140&view=diff

==============================================================================
--- llvm/trunk/autoconf/configure.ac (original)
+++ llvm/trunk/autoconf/configure.ac Tue May 19 17:18:56 2009
@@ -372,6 +372,7 @@
   default) AC_SUBST(ENABLE_THREADS,[1]) ;;
   *) AC_MSG_ERROR([Invalid setting for --enable-threads. Use "yes" or "no"]) ;;
 esac
+AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
 
 dnl Allow building without position independent code
 AC_ARG_ENABLE(pic,
@@ -897,24 +898,22 @@
 fi
 
 dnl atomic builtins are required for threading support.
-if test "$ENABLE_THREADS" -eq 1 ; then
-  AC_MSG_CHECKING(for GCC atomic builtins)
-  AC_LINK_IFELSE(
-	AC_LANG_SOURCE(
-          [[int main() { 
-              volatile unsigned long val = 1;
-              __sync_synchronize();
-              __sync_val_compare_and_swap(&val, 1, 0);
-              return 0;
-            }
-          ]]),
-        AC_MSG_RESULT(yes),
-        AC_MSG_RESULT(no)
-        AC_SUBST(ENABLE_THREADS, [0])
-        AC_MSG_WARN([Threading support disabled because atomic builtins are missing]))
-fi
+AC_MSG_CHECKING(for GCC atomic builtins)
+AC_LINK_IFELSE(
+  AC_LANG_SOURCE(
+    [[int main() { 
+        volatile unsigned long val = 1;
+        __sync_synchronize();
+        __sync_val_compare_and_swap(&val, 1, 0);
+        return 0;
+      }
+    ]]),
+  AC_MSG_RESULT(yes)
+  AC_DEFINE(LLVM_MULTITHREADED, 1, Build multithreading support into LLVM),
+  AC_MSG_RESULT(no)
+  AC_DEFINE(LLVM_MULTITHREADED, 0, Build multithreading support into LLVM)
+  AC_MSG_WARN([LLVM will be built thread-unsafe because atomic builtins are missing]))
 
-AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
 
 dnl===-----------------------------------------------------------------------===
 dnl===

Modified: llvm/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/configure?rev=72140&r1=72139&r2=72140&view=diff

==============================================================================
--- llvm/trunk/configure (original)
+++ llvm/trunk/configure Tue May 19 17:18:56 2009
@@ -4870,6 +4870,11 @@
    { (exit 1); exit 1; }; } ;;
 esac
 
+cat >>confdefs.h <<_ACEOF
+#define ENABLE_THREADS $ENABLE_THREADS
+_ACEOF
+
+
 # Check whether --enable-pic was given.
 if test "${enable_pic+set}" = set; then
   enableval=$enable_pic;
@@ -10570,7 +10575,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 10573 "configure"
+#line 10578 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -12714,7 +12719,7 @@
   ;;
 *-*-irix6*)
   # Find out which ABI we are using.
-  echo '#line 12717 "configure"' > conftest.$ac_ext
+  echo '#line 12722 "configure"' > conftest.$ac_ext
   if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
   (eval $ac_compile) 2>&5
   ac_status=$?
@@ -14432,11 +14437,11 @@
    -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:14435: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14440: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:14439: \$? = $ac_status" >&5
+   echo "$as_me:14444: \$? = $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.
@@ -14700,11 +14705,11 @@
    -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:14703: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14708: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:14707: \$? = $ac_status" >&5
+   echo "$as_me:14712: \$? = $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.
@@ -14804,11 +14809,11 @@
    -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:14807: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:14812: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:14811: \$? = $ac_status" >&5
+   echo "$as_me:14816: \$? = $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
@@ -17256,7 +17261,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 17259 "configure"
+#line 17264 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -17356,7 +17361,7 @@
   lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
   lt_status=$lt_dlunknown
   cat > conftest.$ac_ext <<EOF
-#line 17359 "configure"
+#line 17364 "configure"
 #include "confdefs.h"
 
 #if HAVE_DLFCN_H
@@ -19724,11 +19729,11 @@
    -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:19727: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:19732: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:19731: \$? = $ac_status" >&5
+   echo "$as_me:19736: \$? = $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.
@@ -19828,11 +19833,11 @@
    -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:19831: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:19836: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:19835: \$? = $ac_status" >&5
+   echo "$as_me:19840: \$? = $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
@@ -21398,11 +21403,11 @@
    -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:21401: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:21406: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:21405: \$? = $ac_status" >&5
+   echo "$as_me:21410: \$? = $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.
@@ -21502,11 +21507,11 @@
    -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:21505: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:21510: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:21509: \$? = $ac_status" >&5
+   echo "$as_me:21514: \$? = $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
@@ -23737,11 +23742,11 @@
    -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:23740: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:23745: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:23744: \$? = $ac_status" >&5
+   echo "$as_me:23749: \$? = $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.
@@ -24005,11 +24010,11 @@
    -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:24008: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:24013: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>conftest.err)
    ac_status=$?
    cat conftest.err >&5
-   echo "$as_me:24012: \$? = $ac_status" >&5
+   echo "$as_me:24017: \$? = $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.
@@ -24109,11 +24114,11 @@
    -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:24112: $lt_compile\"" >&5)
+   (eval echo "\"\$as_me:24117: $lt_compile\"" >&5)
    (eval "$lt_compile" 2>out/conftest.err)
    ac_status=$?
    cat out/conftest.err >&5
-   echo "$as_me:24116: \$? = $ac_status" >&5
+   echo "$as_me:24121: \$? = $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
@@ -33602,21 +33607,20 @@
   fi
 fi
 
-if test "$ENABLE_THREADS" -eq 1 ; then
-  { echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
+{ echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
 echo $ECHO_N "checking for GCC atomic builtins... $ECHO_C" >&6; }
-  cat >conftest.$ac_ext <<_ACEOF
+cat >conftest.$ac_ext <<_ACEOF
 /* confdefs.h.  */
 _ACEOF
 cat confdefs.h >>conftest.$ac_ext
 cat >>conftest.$ac_ext <<_ACEOF
 /* end confdefs.h.  */
 int main() {
-              volatile unsigned long val = 1;
-              __sync_synchronize();
-              __sync_val_compare_and_swap(&val, 1, 0);
-              return 0;
-            }
+        volatile unsigned long val = 1;
+        __sync_synchronize();
+        __sync_val_compare_and_swap(&val, 1, 0);
+        return 0;
+      }
 
 _ACEOF
 rm -f conftest.$ac_objext conftest$ac_exeext
@@ -33655,26 +33659,28 @@
   (exit $ac_status); }; }; then
   { echo "$as_me:$LINENO: result: yes" >&5
 echo "${ECHO_T}yes" >&6; }
+
+cat >>confdefs.h <<\_ACEOF
+#define LLVM_MULTITHREADED 1
+_ACEOF
+
 else
   echo "$as_me: failed program was:" >&5
 sed 's/^/| /' conftest.$ac_ext >&5
 
 	{ echo "$as_me:$LINENO: result: no" >&5
 echo "${ECHO_T}no" >&6; }
-        ENABLE_THREADS=0
 
-        { echo "$as_me:$LINENO: WARNING: Threading support disabled because atomic builtins are missing" >&5
-echo "$as_me: WARNING: Threading support disabled because atomic builtins are missing" >&2;}
+cat >>confdefs.h <<\_ACEOF
+#define LLVM_MULTITHREADED 0
+_ACEOF
+
+  { echo "$as_me:$LINENO: WARNING: LLVM will be built thread-unsafe because atomic builtins are missing" >&5
+echo "$as_me: WARNING: LLVM will be built thread-unsafe because atomic builtins are missing" >&2;}
 fi
 
 rm -f core conftest.err conftest.$ac_objext \
       conftest$ac_exeext conftest.$ac_ext
-fi
-
-
-cat >>confdefs.h <<_ACEOF
-#define ENABLE_THREADS $ENABLE_THREADS
-_ACEOF
 
 
 

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72140&r1=72139&r2=72140&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Tue May 19 17:18:56 2009
@@ -23,14 +23,8 @@
   namespace sys {
     
     inline void MemoryFence() {
-#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
-#  if defined(__GNUC__)
-      __asm__ __volatile__("" : : : "memory");
-#  elif defined(_MSC_VER)
-      __asm { };
-#  else
-#    error No memory fence implementation for your platform!
-#  endif
+#if LLVM_MULTITHREADED==0
+      return;
 #else
 #  if defined(__GNUC__)
       __sync_synchronize();
@@ -42,7 +36,7 @@
 #endif
 }
 
-#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
+#if LLVM_MULTITHREADED==0
     typedef unsigned long cas_flag;
     template<typename T>
     inline T CompareAndSwap(volatile T* dest,




From resistor at mac.com  Tue May 19 17:25:56 2009
From: resistor at mac.com (Owen Anderson)
Date: Tue, 19 May 2009 22:25:56 -0000
Subject: [llvm-commits] [llvm] r72141 -
	/llvm/trunk/include/llvm/Config/config.h.in
Message-ID: <200905192225.n4JMPu2s032032@zion.cs.uiuc.edu>

Author: resistor
Date: Tue May 19 17:25:55 2009
New Revision: 72141

URL: http://llvm.org/viewvc/llvm-project?rev=72141&view=rev
Log:
Forgot this file in my last commit.

Modified:
    llvm/trunk/include/llvm/Config/config.h.in

Modified: llvm/trunk/include/llvm/Config/config.h.in
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Config/config.h.in?rev=72141&r1=72140&r2=72141&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Config/config.h.in (original)
+++ llvm/trunk/include/llvm/Config/config.h.in Tue May 19 17:25:55 2009
@@ -470,6 +470,9 @@
 /* Installation directory for man pages */
 #undef LLVM_MANDIR
 
+/* Build multithreading support into LLVM */
+#undef LLVM_MULTITHREADED
+
 /* Define if this is Unixish platform */
 #undef LLVM_ON_UNIX
 




From resistor at mac.com  Tue May 19 19:31:29 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 00:31:29 -0000
Subject: [llvm-commits] [llvm] r72148 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905200031.n4K0VTjb003694@zion.cs.uiuc.edu>

Author: resistor
Date: Tue May 19 19:31:29 2009
New Revision: 72148

URL: http://llvm.org/viewvc/llvm-project?rev=72148&view=rev
Log:
We need to include config.h here so that the #defines are set properly.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72148&r1=72147&r2=72148&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Tue May 19 19:31:29 2009
@@ -14,6 +14,8 @@
 #ifndef LLVM_SYSTEM_ATOMIC_H
 #define LLVM_SYSTEM_ATOMIC_H
 
+#include "llvm/Config/config.h"
+
 #if defined(_MSC_VER)
 #include <windows.h>
 #endif




From gohman at apple.com  Tue May 19 19:34:08 2009
From: gohman at apple.com (Dan Gohman)
Date: Wed, 20 May 2009 00:34:08 -0000
Subject: [llvm-commits] [llvm] r72149 - in /llvm/trunk:
 lib/Transforms/Scalar/LoopStrengthReduce.cpp
 test/Transforms/LoopStrengthReduce/dont_reverse.ll
Message-ID: <200905200034.n4K0Y8KL003792@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 19:34:08 2009
New Revision: 72149

URL: http://llvm.org/viewvc/llvm-project?rev=72149&view=rev
Log:
Suppress the IV reversal transformation in the case that the RHS
of the comparison is defined inside the loop. This fixes a
use-before-def problem, because the transformation puts a use
of the RHS outside the loop.

Added:
    llvm/trunk/test/Transforms/LoopStrengthReduce/dont_reverse.ll
Modified:
    llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp?rev=72149&r1=72148&r2=72149&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/LoopStrengthReduce.cpp Tue May 19 19:34:08 2009
@@ -2456,6 +2456,11 @@
   SCEVHandle One = SE->getIntegerSCEV(1, BackedgeTakenCount->getType());
   if (!AR || !AR->isAffine() || AR->getStepRecurrence(*SE) != One)
     return;
+  // If the RHS of the comparison is defined inside the loop, the rewrite
+  // cannot be done.
+  if (Instruction *CR = dyn_cast<Instruction>(Cond->getOperand(1)))
+    if (L->contains(CR->getParent()))
+      return;
 
   // Make sure the IV is only used for counting.  Value may be preinc or
   // postinc; 2 uses in either case.

Added: llvm/trunk/test/Transforms/LoopStrengthReduce/dont_reverse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopStrengthReduce/dont_reverse.ll?rev=72149&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/LoopStrengthReduce/dont_reverse.ll (added)
+++ llvm/trunk/test/Transforms/LoopStrengthReduce/dont_reverse.ll Tue May 19 19:34:08 2009
@@ -0,0 +1,21 @@
+; RUN: llvm-as < %s | opt -loop-reduce | llvm-dis \
+; RUN:    | grep {icmp eq i2 %lsr.iv.next, %xmp4344}
+
+; Don't reverse the iteration if the rhs of the compare is defined
+; inside the loop.
+
+define void @Fill_Buffer() nounwind {
+entry:
+	br label %bb8
+
+bb8:
+	%indvar34 = phi i32 [ 0, %entry ], [ %indvar.next35, %bb8 ]
+	%indvar3451 = trunc i32 %indvar34 to i2
+	%xmp4344 = xor i2 0, -1
+	%xmp104 = icmp eq i2 %indvar3451, %xmp4344
+	%indvar.next35 = add i32 %indvar34, 1
+	br i1 %xmp104, label %bb10, label %bb8
+
+bb10:
+	unreachable
+}




From resistor at mac.com  Tue May 19 19:39:26 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 00:39:26 -0000
Subject: [llvm-commits] [llvm] r72151 - in /llvm/trunk:
 include/llvm/Support/ManagedStatic.h lib/Support/ManagedStatic.cpp
Message-ID: <200905200039.n4K0dRtm003968@zion.cs.uiuc.edu>

Author: resistor
Date: Tue May 19 19:39:20 2009
New Revision: 72151

URL: http://llvm.org/viewvc/llvm-project?rev=72151&view=rev
Log:
Add llvm_start_multithreaded(), which starts up the LLVM internals in thread-safe mode.  Provide double-check locking
initialization of ManagedStatic's when running in thread-safe mode.

Modified:
    llvm/trunk/include/llvm/Support/ManagedStatic.h
    llvm/trunk/lib/Support/ManagedStatic.cpp

Modified: llvm/trunk/include/llvm/Support/ManagedStatic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/ManagedStatic.h?rev=72151&r1=72150&r2=72151&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Support/ManagedStatic.h (original)
+++ llvm/trunk/include/llvm/Support/ManagedStatic.h Tue May 19 19:39:20 2009
@@ -14,8 +14,16 @@
 #ifndef LLVM_SUPPORT_MANAGED_STATIC_H
 #define LLVM_SUPPORT_MANAGED_STATIC_H
 
+#include "llvm/System/Atomic.h"
+
 namespace llvm {
 
+/// object_creator - Helper method for ManagedStatic.
+template<class C>
+void* object_creator() {
+  return new C();
+}
+
 /// object_deleter - Helper method for ManagedStatic.
 ///
 template<class C>
@@ -32,7 +40,7 @@
   mutable void (*DeleterFn)(void*);
   mutable const ManagedStaticBase *Next;
 
-  void RegisterManagedStatic(void *ObjPtr, void (*deleter)(void*)) const;
+  void RegisterManagedStatic(void *(*creator)(), void (*deleter)(void*)) const;
 public:
   /// isConstructed - Return true if this object has not been created yet.
   bool isConstructed() const { return Ptr != 0; }
@@ -51,25 +59,32 @@
 
   // Accessors.
   C &operator*() {
-    if (!Ptr) LazyInit();
+    void* tmp = Ptr;
+    sys::MemoryFence();
+    if (!tmp) RegisterManagedStatic(object_creator<C>, object_deleter<C>);
+
     return *static_cast<C*>(Ptr);
   }
   C *operator->() {
-    if (!Ptr) LazyInit();
+    void* tmp = Ptr;
+    sys::MemoryFence();
+    if (!tmp) RegisterManagedStatic(object_creator<C>, object_deleter<C>);
+
     return static_cast<C*>(Ptr);
   }
   const C &operator*() const {
-    if (!Ptr) LazyInit();
+    void* tmp = Ptr;
+    sys::MemoryFence();
+    if (!tmp) RegisterManagedStatic(object_creator<C>, object_deleter<C>);
+
     return *static_cast<C*>(Ptr);
   }
   const C *operator->() const {
-    if (!Ptr) LazyInit();
-    return static_cast<C*>(Ptr);
-  }
+    void* tmp = Ptr;
+    sys::MemoryFence();
+    if (!tmp) RegisterManagedStatic(object_creator<C>, object_deleter<C>);
 
-public:
-  void LazyInit() const {
-    RegisterManagedStatic(new C(), object_deleter<C>);
+    return static_cast<C*>(Ptr);
   }
 };
 
@@ -80,6 +95,10 @@
 };
 
 
+/// llvm_start_multithreaded - Allocate and initialize structures needed to
+/// make LLVM safe for multithreading.
+void llvm_start_multithreaded();
+
 /// llvm_shutdown - Deallocate and destroy all ManagedStatic variables.
 void llvm_shutdown();
 
@@ -87,7 +106,10 @@
 /// llvm_shutdown_obj - This is a simple helper class that calls
 /// llvm_shutdown() when it is destroyed.
 struct llvm_shutdown_obj {
-  llvm_shutdown_obj() {}
+  llvm_shutdown_obj() { }
+  explicit llvm_shutdown_obj(bool multithreaded) {
+    if (multithreaded) llvm_start_multithreaded();
+  }
   ~llvm_shutdown_obj() { llvm_shutdown(); }
 };
 

Modified: llvm/trunk/lib/Support/ManagedStatic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/ManagedStatic.cpp?rev=72151&r1=72150&r2=72151&view=diff

==============================================================================
--- llvm/trunk/lib/Support/ManagedStatic.cpp (original)
+++ llvm/trunk/lib/Support/ManagedStatic.cpp Tue May 19 19:39:20 2009
@@ -12,21 +12,44 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/Support/ManagedStatic.h"
+#include "llvm/Config/config.h"
+#include "llvm/System/Atomic.h"
+#include "llvm/System/Mutex.h"
 #include <cassert>
 using namespace llvm;
 
 static const ManagedStaticBase *StaticList = 0;
 
-void ManagedStaticBase::RegisterManagedStatic(void *ObjPtr,
+static sys::Mutex* ManagedStaticMutex = 0;
+
+void ManagedStaticBase::RegisterManagedStatic(void *(*Creator)(),
                                               void (*Deleter)(void*)) const {
-  assert(Ptr == 0 && DeleterFn == 0 && Next == 0 &&
-         "Partially init static?");
-  Ptr = ObjPtr;
-  DeleterFn = Deleter;
+  if (ManagedStaticMutex) {
+    ManagedStaticMutex->acquire();
+
+    if (Ptr == 0) {
+      void* tmp = Creator ? Creator() : 0;
+
+      sys::MemoryFence();
+      Ptr = tmp;
+      DeleterFn = Deleter;
+      
+      // Add to list of managed statics.
+      Next = StaticList;
+      StaticList = this;
+    }
+
+    ManagedStaticMutex->release();
+  } else {
+    assert(Ptr == 0 && DeleterFn == 0 && Next == 0 &&
+	   "Partially initialized ManagedStatic!?");
+    Ptr = Creator ? Creator() : 0;
+    DeleterFn = Deleter;
   
-  // Add to list of managed statics.
-  Next = StaticList;
-  StaticList = this;
+    // Add to list of managed statics.
+    Next = StaticList;
+    StaticList = this;
+  }
 }
 
 void ManagedStaticBase::destroy() const {
@@ -45,9 +68,23 @@
   DeleterFn = 0;
 }
 
+void llvm::llvm_start_multithreaded() {
+#if LLVM_MULTITHREADED
+  assert(ManagedStaticMutex == 0 && "Multithreaded LLVM already initialized!");
+  ManagedStaticMutex = new sys::Mutex(true);
+#else
+  assert(0 && "LLVM built without multithreading support!");
+#endif
+}
+
 /// llvm_shutdown - Deallocate and destroy all ManagedStatic variables.
 void llvm::llvm_shutdown() {
   while (StaticList)
     StaticList->destroy();
+
+  if (ManagedStaticMutex) {
+    delete ManagedStaticMutex;
+    ManagedStaticMutex = 0;
+  }
 }
 




From gohman at apple.com  Tue May 19 20:01:26 2009
From: gohman at apple.com (Dan Gohman)
Date: Wed, 20 May 2009 01:01:26 -0000
Subject: [llvm-commits] [llvm] r72152 -
	/llvm/trunk/lib/Analysis/ScalarEvolution.cpp
Message-ID: <200905200101.n4K11QHS004778@zion.cs.uiuc.edu>

Author: djg
Date: Tue May 19 20:01:24 2009
New Revision: 72152

URL: http://llvm.org/viewvc/llvm-project?rev=72152&view=rev
Log:
Teach SCEV::isLoopInvariant and SCEV::hasComputableLoopEvolution
about the convention from LoopInfo that a null Loop* means the entire
function body.

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=72152&r1=72151&r2=72152&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ScalarEvolution.cpp (original)
+++ llvm/trunk/lib/Analysis/ScalarEvolution.cpp Tue May 19 20:01:24 2009
@@ -388,7 +388,9 @@
 bool SCEVAddRecExpr::isLoopInvariant(const Loop *QueryLoop) const {
   // This recurrence is invariant w.r.t to QueryLoop iff QueryLoop doesn't
   // contain L and if the start is invariant.
-  return !QueryLoop->contains(L->getHeader()) &&
+  // Add recurrences are never invariant in the function-body (null loop).
+  return QueryLoop &&
+         !QueryLoop->contains(L->getHeader()) &&
          getOperand(0)->isLoopInvariant(QueryLoop);
 }
 
@@ -410,8 +412,10 @@
 bool SCEVUnknown::isLoopInvariant(const Loop *L) const {
   // All non-instruction values are loop invariant.  All instructions are loop
   // invariant if they are not contained in the specified loop.
+  // Instructions are never considered invariant in the function body
+  // (null loop) because they are defined within the "loop".
   if (Instruction *I = dyn_cast<Instruction>(V))
-    return !L->contains(I->getParent());
+    return L && !L->contains(I->getParent());
   return true;
 }
 




From evan.cheng at apple.com  Tue May 19 20:11:00 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 20 May 2009 01:11:00 -0000
Subject: [llvm-commits] [llvm] r72154 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/call-imm.ll
Message-ID: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>

Author: evancheng
Date: Tue May 19 20:11:00 2009
New Revision: 72154

URL: http://llvm.org/viewvc/llvm-project?rev=72154&view=rev
Log:
Cannot use immediate as call absolute target in PIC mode.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/test/CodeGen/X86/call-imm.ll

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=72154&r1=72153&r2=72154&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Tue May 19 20:11:00 2009
@@ -234,6 +234,7 @@
 def SmallCode    : Predicate<"TM.getCodeModel() == CodeModel::Small">;
 def NotSmallCode : Predicate<"TM.getCodeModel() != CodeModel::Small">;
 def IsStatic     : Predicate<"TM.getRelocationModel() == Reloc::Static">;
+def IsNotPIC     : Predicate<"TM.getRelocationModel() != Reloc::PIC_">;
 def OptForSpeed  : Predicate<"!OptForSize">;
 def FastBTMem    : Predicate<"!Subtarget->isBTMemSlow()">;
 
@@ -562,7 +563,7 @@
       Uses = [ESP] in {
     def CALLpcrel32 : Ii32<0xE8, RawFrm, (outs), (ins i32imm:$dst,variable_ops),
                            "call\t${dst:call}", [(X86call imm:$dst)]>,
-                      Requires<[In32BitMode]>;
+                      Requires<[In32BitMode, IsNotPIC]>;
     def CALL32r     : I<0xFF, MRM2r, (outs), (ins GR32:$dst, variable_ops),
                         "call\t{*}$dst", [(X86call GR32:$dst)]>;
     def CALL32m     : I<0xFF, MRM2m, (outs), (ins i32mem:$dst, variable_ops),

Modified: llvm/trunk/test/CodeGen/X86/call-imm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/call-imm.ll?rev=72154&r1=72153&r2=72154&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/X86/call-imm.ll (original)
+++ llvm/trunk/test/CodeGen/X86/call-imm.ll Tue May 19 20:11:00 2009
@@ -1,4 +1,5 @@
 ; RUN: llvm-as < %s | llc -march=x86    | grep {call.*12345678}
+; RUN: llvm-as < %s | llc -march=x86 -relocation-model=pic | not grep {call.*12345678}
 
 ; Call to immediate is not safe on x86-64 unless we *know* that the
 ; call will be within 32-bits pcrel from the dest immediate.
@@ -7,6 +8,7 @@
 
 ; PR3666
 ; PR3773
+; rdar://6904453
 
 define i32 @main() nounwind {
 entry:




From gohman at apple.com  Tue May 19 20:41:43 2009
From: gohman at apple.com (Dan Gohman)
Date: Tue, 19 May 2009 18:41:43 -0700
Subject: [llvm-commits] [llvm] r72154 - in /llvm/trunk:
	lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/call-imm.ll
In-Reply-To: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>
References: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>
Message-ID: <33592CC8-4FA5-486B-88A4-C218B929C699@apple.com>


On May 19, 2009, at 6:11 PM, Evan Cheng wrote:

> Author: evancheng
> Date: Tue May 19 20:11:00 2009
> New Revision: 72154
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72154&view=rev
> Log:
> Cannot use immediate as call absolute target in PIC mode.
>
> Modified:
>    llvm/trunk/lib/Target/X86/X86InstrInfo.td
>    llvm/trunk/test/CodeGen/X86/call-imm.ll
>
> Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=72154&r1=72153&r2=72154&view=diff
>
> =
> =
> =
> =
> =
> =
> =
> =
> ======================================================================
> --- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
> +++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Tue May 19 20:11:00 2009
> @@ -234,6 +234,7 @@
> def SmallCode    : Predicate<"TM.getCodeModel() == CodeModel::Small">;
> def NotSmallCode : Predicate<"TM.getCodeModel() != CodeModel::Small">;
> def IsStatic     : Predicate<"TM.getRelocationModel() ==  
> Reloc::Static">;
> +def IsNotPIC     : Predicate<"TM.getRelocationModel() !=  
> Reloc::PIC_">;
> def OptForSpeed  : Predicate<"!OptForSize">;
> def FastBTMem    : Predicate<"!Subtarget->isBTMemSlow()">;
>
> @@ -562,7 +563,7 @@
>       Uses = [ESP] in {
>     def CALLpcrel32 : Ii32<0xE8, RawFrm, (outs), (ins i32imm: 
> $dst,variable_ops),
>                            "call\t${dst:call}", [(X86call imm:$dst)]>,
> -                      Requires<[In32BitMode]>;
> +                      Requires<[In32BitMode, IsNotPIC]>;

This looks like it says that CALLpcrel32 is valid only for PIC,
which isn't true. It looks like the (X86call imm:$dst) pattern
should be moved out into a "def : Pat" like the other patterns
for CALLpcrel32, so that the Requires<[IsNotPIC]> can apply to
just that pattern, not the instruction itself.

Also, it looks like the Darwin linker rejects "call 12345678" in
dynamic-no-pic mode as well as PIC mode.

Dan



From evan.cheng at apple.com  Tue May 19 22:10:50 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 19 May 2009 20:10:50 -0700
Subject: [llvm-commits] [llvm] r72154 - in /llvm/trunk:
	lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/call-imm.ll
In-Reply-To: <33592CC8-4FA5-486B-88A4-C218B929C699@apple.com>
References: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>
	<33592CC8-4FA5-486B-88A4-C218B929C699@apple.com>
Message-ID: <D3940BC3-1B6D-4CF7-9AB9-5F9369859F9D@apple.com>


On May 19, 2009, at 6:41 PM, Dan Gohman wrote:

>>
>>
>> @@ -562,7 +563,7 @@
>>      Uses = [ESP] in {
>>    def CALLpcrel32 : Ii32<0xE8, RawFrm, (outs), (ins i32imm:
>> $dst,variable_ops),
>>                           "call\t${dst:call}", [(X86call imm:$dst)]>,
>> -                      Requires<[In32BitMode]>;
>> +                      Requires<[In32BitMode, IsNotPIC]>;
>
> This looks like it says that CALLpcrel32 is valid only for PIC,
> which isn't true. It looks like the (X86call imm:$dst) pattern
> should be moved out into a "def : Pat" like the other patterns
> for CALLpcrel32, so that the Requires<[IsNotPIC]> can apply to
> just that pattern, not the instruction itself.

The requirement does only apply for the pattern. I agree it reads a  
bit funny. I can move it to a def : Pat.

>
> Also, it looks like the Darwin linker rejects "call 12345678" in
> dynamic-no-pic mode as well as PIC mode.

Ah right. I'll fix.

Evan

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



From evan.cheng at apple.com  Tue May 19 22:12:37 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Tue, 19 May 2009 20:12:37 -0700
Subject: [llvm-commits] [llvm] r72154 - in /llvm/trunk:
	lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/call-imm.ll
In-Reply-To: <D3940BC3-1B6D-4CF7-9AB9-5F9369859F9D@apple.com>
References: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>
	<33592CC8-4FA5-486B-88A4-C218B929C699@apple.com>
	<D3940BC3-1B6D-4CF7-9AB9-5F9369859F9D@apple.com>
Message-ID: <F53680BE-B04E-48EF-9BB7-727D7EF96605@apple.com>


On May 19, 2009, at 8:10 PM, Evan Cheng wrote:

>
> On May 19, 2009, at 6:41 PM, Dan Gohman wrote:
>
>>>
>>>
>>> @@ -562,7 +563,7 @@
>>>     Uses = [ESP] in {
>>>   def CALLpcrel32 : Ii32<0xE8, RawFrm, (outs), (ins i32imm:
>>> $dst,variable_ops),
>>>                          "call\t${dst:call}", [(X86call imm:$dst)]>,
>>> -                      Requires<[In32BitMode]>;
>>> +                      Requires<[In32BitMode, IsNotPIC]>;
>>
>> This looks like it says that CALLpcrel32 is valid only for PIC,
>> which isn't true. It looks like the (X86call imm:$dst) pattern
>> should be moved out into a "def : Pat" like the other patterns
>> for CALLpcrel32, so that the Requires<[IsNotPIC]> can apply to
>> just that pattern, not the instruction itself.
>
> The requirement does only apply for the pattern. I agree it reads a
> bit funny. I can move it to a def : Pat.
>
>>
>> Also, it looks like the Darwin linker rejects "call 12345678" in
>> dynamic-no-pic mode as well as PIC mode.

Is it only Darwin linker that rejects this?

Evan

>
> Ah right. I'll fix.
>
> Evan
>
>>
>> Dan
>>
>> _______________________________________________
>> 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 eli.friedman at gmail.com  Tue May 19 22:54:53 2009
From: eli.friedman at gmail.com (Eli Friedman)
Date: Tue, 19 May 2009 20:54:53 -0700
Subject: [llvm-commits] [llvm] r72154 - in /llvm/trunk:
	lib/Target/X86/X86InstrInfo.td test/CodeGen/X86/call-imm.ll
In-Reply-To: <F53680BE-B04E-48EF-9BB7-727D7EF96605@apple.com>
References: <200905200111.n4K1B0QC005085@zion.cs.uiuc.edu>
	<33592CC8-4FA5-486B-88A4-C218B929C699@apple.com>
	<D3940BC3-1B6D-4CF7-9AB9-5F9369859F9D@apple.com>
	<F53680BE-B04E-48EF-9BB7-727D7EF96605@apple.com>
Message-ID: <f5aa3e9b0905192054q77cdcb6dy7474f8d939057ed1@mail.gmail.com>

On Tue, May 19, 2009 at 8:12 PM, Evan Cheng <evan.cheng at apple.com> wrote:
>>> Also, it looks like the Darwin linker rejects "call 12345678" in
>>> dynamic-no-pic mode as well as PIC mode.
>
> Is it only Darwin linker that rejects this?

I'm not sure if it's supported reliably outside of Linux.  I just
tried with mingw binutils, and it crashes with an assertion failure
there.

-Eli


From evan.cheng at apple.com  Tue May 19 23:53:59 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 20 May 2009 04:53:59 -0000
Subject: [llvm-commits] [llvm] r72160 - in /llvm/trunk:
 lib/Target/X86/X86InstrInfo.td lib/Target/X86/X86Subtarget.cpp
 lib/Target/X86/X86Subtarget.h test/CodeGen/X86/call-imm.ll
Message-ID: <200905200453.n4K4rxIL012191@zion.cs.uiuc.edu>

Author: evancheng
Date: Tue May 19 23:53:57 2009
New Revision: 72160

URL: http://llvm.org/viewvc/llvm-project?rev=72160&view=rev
Log:
Try again. Allow call to immediate address for ELF or when in static relocation mode.

Modified:
    llvm/trunk/lib/Target/X86/X86InstrInfo.td
    llvm/trunk/lib/Target/X86/X86Subtarget.cpp
    llvm/trunk/lib/Target/X86/X86Subtarget.h
    llvm/trunk/test/CodeGen/X86/call-imm.ll

Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.td?rev=72160&r1=72159&r2=72160&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.td Tue May 19 23:53:57 2009
@@ -234,9 +234,9 @@
 def SmallCode    : Predicate<"TM.getCodeModel() == CodeModel::Small">;
 def NotSmallCode : Predicate<"TM.getCodeModel() != CodeModel::Small">;
 def IsStatic     : Predicate<"TM.getRelocationModel() == Reloc::Static">;
-def IsNotPIC     : Predicate<"TM.getRelocationModel() != Reloc::PIC_">;
 def OptForSpeed  : Predicate<"!OptForSize">;
 def FastBTMem    : Predicate<"!Subtarget->isBTMemSlow()">;
+def CallImmAddr  : Predicate<"Subtarget->IsLegalToCallImmediateAddr(TM)">;
 
 //===----------------------------------------------------------------------===//
 // X86 Instruction Format Definitions.
@@ -562,8 +562,7 @@
               XMM8, XMM9, XMM10, XMM11, XMM12, XMM13, XMM14, XMM15, EFLAGS],
       Uses = [ESP] in {
     def CALLpcrel32 : Ii32<0xE8, RawFrm, (outs), (ins i32imm:$dst,variable_ops),
-                           "call\t${dst:call}", [(X86call imm:$dst)]>,
-                      Requires<[In32BitMode, IsNotPIC]>;
+                           "call\t${dst:call}", []>;
     def CALL32r     : I<0xFF, MRM2r, (outs), (ins GR32:$dst, variable_ops),
                         "call\t{*}$dst", [(X86call GR32:$dst)]>;
     def CALL32m     : I<0xFF, MRM2m, (outs), (ins i32mem:$dst, variable_ops),
@@ -3342,6 +3341,8 @@
           (CALLpcrel32 tglobaladdr:$dst)>;
 def : Pat<(X86call (i32 texternalsym:$dst)),
           (CALLpcrel32 texternalsym:$dst)>;
+def : Pat<(X86call (i32 imm:$dst)),
+          (CALLpcrel32 imm:$dst)>, Requires<[CallImmAddr]>;
 
 // X86 specific add which produces a flag.
 def : Pat<(addc GR32:$src1, GR32:$src2),

Modified: llvm/trunk/lib/Target/X86/X86Subtarget.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Subtarget.cpp?rev=72160&r1=72159&r2=72160&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86Subtarget.cpp Tue May 19 23:53:57 2009
@@ -74,8 +74,8 @@
 /// cases where GVRequiresExtraLoad is true.  Some variations of PIC require
 /// a register, but not an extra load.
 bool X86Subtarget::GVRequiresRegister(const GlobalValue *GV,
-                                       const TargetMachine& TM,
-                                       bool isDirectCall) const
+                                      const TargetMachine& TM,
+                                      bool isDirectCall) const
 {
   if (GVRequiresExtraLoad(GV, TM, isDirectCall))
     return true;
@@ -99,6 +99,14 @@
   return 0;
 }
 
+/// IsLegalToCallImmediateAddr - Return true if the subtarget allows calls
+/// to immediate address.
+bool X86Subtarget::IsLegalToCallImmediateAddr(const TargetMachine &TM) const {
+  if (Is64Bit)
+    return false;
+  return isTargetELF() || TM.getRelocationModel() == Reloc::Static;
+}
+
 /// getSpecialAddressLatency - For targets where it is beneficial to
 /// backschedule instructions that compute addresses, return a value
 /// indicating the number of scheduling cycles of backscheduling that

Modified: llvm/trunk/lib/Target/X86/X86Subtarget.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86Subtarget.h?rev=72160&r1=72159&r2=72160&view=diff

==============================================================================
--- llvm/trunk/lib/Target/X86/X86Subtarget.h (original)
+++ llvm/trunk/lib/Target/X86/X86Subtarget.h Tue May 19 23:53:57 2009
@@ -189,6 +189,10 @@
   bool GVRequiresRegister(const GlobalValue* GV, const TargetMachine& TM,
                            bool isDirectCall) const;
 
+  /// IsLegalToCallImmediateAddr - Return true if the subtarget allows calls
+  /// to immediate address.
+  bool IsLegalToCallImmediateAddr(const TargetMachine &TM) const;
+
   /// This function returns the name of a function which has an interface
   /// like the non-standard bzero function, if such a function exists on
   /// the current subtarget and it is considered prefereable over

Modified: llvm/trunk/test/CodeGen/X86/call-imm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/call-imm.ll?rev=72160&r1=72159&r2=72160&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/X86/call-imm.ll (original)
+++ llvm/trunk/test/CodeGen/X86/call-imm.ll Tue May 19 23:53:57 2009
@@ -1,4 +1,5 @@
-; RUN: llvm-as < %s | llc -march=x86    | grep {call.*12345678}
+; RUN: llvm-as < %s | llc -mtriple=i386-darwin-apple -relocation-model=static | grep {call.*12345678}
+; RUN: llvm-as < %s | llc -mtriple=i386-pc-linux -relocation-model=dynamic-no-pic | grep {call.*12345678}
 ; RUN: llvm-as < %s | llc -march=x86 -relocation-model=pic | not grep {call.*12345678}
 
 ; Call to immediate is not safe on x86-64 unless we *know* that the




From evan.cheng at apple.com  Wed May 20 00:45:37 2009
From: evan.cheng at apple.com (Evan Cheng)
Date: Wed, 20 May 2009 05:45:37 -0000
Subject: [llvm-commits] [llvm] r72161 -
	/llvm/trunk/test/CodeGen/X86/call-imm.ll
Message-ID: <200905200545.n4K5jbnr013806@zion.cs.uiuc.edu>

Author: evancheng
Date: Wed May 20 00:45:36 2009
New Revision: 72161

URL: http://llvm.org/viewvc/llvm-project?rev=72161&view=rev
Log:
Fix test on non-darwin hosts.

Modified:
    llvm/trunk/test/CodeGen/X86/call-imm.ll

Modified: llvm/trunk/test/CodeGen/X86/call-imm.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/call-imm.ll?rev=72161&r1=72160&r2=72161&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/X86/call-imm.ll (original)
+++ llvm/trunk/test/CodeGen/X86/call-imm.ll Wed May 20 00:45:36 2009
@@ -1,6 +1,6 @@
 ; RUN: llvm-as < %s | llc -mtriple=i386-darwin-apple -relocation-model=static | grep {call.*12345678}
+; RUN: llvm-as < %s | llc -mtriple=i386-darwin-apple -relocation-model=pic | not grep {call.*12345678}
 ; RUN: llvm-as < %s | llc -mtriple=i386-pc-linux -relocation-model=dynamic-no-pic | grep {call.*12345678}
-; RUN: llvm-as < %s | llc -march=x86 -relocation-model=pic | not grep {call.*12345678}
 
 ; Call to immediate is not safe on x86-64 unless we *know* that the
 ; call will be within 32-bits pcrel from the dest immediate.




From eli.friedman at gmail.com  Wed May 20 01:02:09 2009
From: eli.friedman at gmail.com (Eli Friedman)
Date: Wed, 20 May 2009 06:02:09 -0000
Subject: [llvm-commits] [llvm] r72162 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
 test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll test/CodeGen/XCore/fneg.ll
Message-ID: <200905200602.n4K62Ab6014388@zion.cs.uiuc.edu>

Author: efriedma
Date: Wed May 20 01:02:09 2009
New Revision: 72162

URL: http://llvm.org/viewvc/llvm-project?rev=72162&view=rev
Log:
Fix for PR4235: to build a floating-point value from integer parts, 
build an integer and cast that to a float.  This fixes a crash 
caused by trying to split an f32 into two f16's.

This changes the behavior in test/CodeGen/XCore/fneg.ll because that 
testcase now triggers a DAGCombine which converts the fneg into an integer
operation.  If someone is interested, it's probably possible to tweak 
the test to generate an actual fneg.


Added:
    llvm/trunk/test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
    llvm/trunk/test/CodeGen/XCore/fneg.ll

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72162&r1=72161&r2=72162&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Wed May 20 01:02:09 2009
@@ -426,7 +426,7 @@
 
   if (NumParts > 1) {
     // Assemble the value from multiple parts.
-    if (!ValueVT.isVector()) {
+    if (!ValueVT.isVector() && ValueVT.isInteger()) {
       unsigned PartBits = PartVT.getSizeInBits();
       unsigned ValueBits = ValueVT.getSizeInBits();
 
@@ -438,9 +438,7 @@
         ValueVT : MVT::getIntegerVT(RoundBits);
       SDValue Lo, Hi;
 
-      MVT HalfVT = ValueVT.isInteger() ?
-        MVT::getIntegerVT(RoundBits/2) :
-        MVT::getFloatingPointVT(RoundBits/2);
+      MVT HalfVT = MVT::getIntegerVT(RoundBits/2);
 
       if (RoundParts > 2) {
         Lo = getCopyFromParts(DAG, dl, Parts, RoundParts/2, PartVT, HalfVT);
@@ -473,7 +471,7 @@
         Lo = DAG.getNode(ISD::ZERO_EXTEND, dl, TotalVT, Lo);
         Val = DAG.getNode(ISD::OR, dl, TotalVT, Lo, Hi);
       }
-    } else {
+    } else if (ValueVT.isVector()) {
       // Handle a multi-element vector.
       MVT IntermediateVT, RegisterVT;
       unsigned NumIntermediates;
@@ -510,6 +508,22 @@
       Val = DAG.getNode(IntermediateVT.isVector() ?
                         ISD::CONCAT_VECTORS : ISD::BUILD_VECTOR, dl,
                         ValueVT, &Ops[0], NumIntermediates);
+    } else if (PartVT.isFloatingPoint()) {
+      // FP split into multiple FP parts (for ppcf128)
+      assert(ValueVT == MVT(MVT::ppcf128) && PartVT == MVT(MVT::f64) &&
+             "Unexpected split");
+      SDValue Lo, Hi;
+      Lo = DAG.getNode(ISD::BIT_CONVERT, dl, MVT(MVT::f64), Parts[0]);
+      Hi = DAG.getNode(ISD::BIT_CONVERT, dl, MVT(MVT::f64), Parts[1]);
+      if (TLI.isBigEndian())
+        std::swap(Lo, Hi);
+      Val = DAG.getNode(ISD::BUILD_PAIR, dl, ValueVT, Lo, Hi);
+    } else {
+      // FP split into integer parts (soft fp)
+      assert(ValueVT.isFloatingPoint() && PartVT.isInteger() &&
+             !PartVT.isVector() && "Unexpected split");
+      MVT IntVT = MVT::getIntegerVT(ValueVT.getSizeInBits());
+      Val = getCopyFromParts(DAG, dl, Parts, NumParts, PartVT, IntVT);
     }
   }
 

Added: llvm/trunk/test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll?rev=72162&view=auto

==============================================================================
--- llvm/trunk/test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll (added)
+++ llvm/trunk/test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll Wed May 20 01:02:09 2009
@@ -0,0 +1,11 @@
+; RUN: llvm-as < %s | llc -march=msp430
+
+define i16 @test(double %d) nounwind {
+entry:
+        %add = add double %d, 1.000000e+00
+        %call = tail call i16 @funct(double %add) nounwind
+        ret i16 %call
+}
+
+declare i16 @funct(double)
+

Modified: llvm/trunk/test/CodeGen/XCore/fneg.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/XCore/fneg.ll?rev=72162&r1=72161&r2=72162&view=diff

==============================================================================
--- llvm/trunk/test/CodeGen/XCore/fneg.ll (original)
+++ llvm/trunk/test/CodeGen/XCore/fneg.ll Wed May 20 01:02:09 2009
@@ -1,5 +1,5 @@
 ; RUN: llvm-as < %s | llc -march=xcore > %t1.s
-; RUN: grep "bl __subdf3" %t1.s | count 1
+; RUN: grep "xor" %t1.s | count 1
 define i1 @test(double %F) nounwind {
 entry:
 	%0 = sub double -0.000000e+00, %F




From cristic at cs.stanford.edu  Wed May 20 01:29:36 2009
From: cristic at cs.stanford.edu (Cristian Cadar)
Date: Wed, 20 May 2009 06:29:36 -0000
Subject: [llvm-commits] [klee] r72163 - in /klee/trunk/www: bugs.html
 content.css index.html install.html menu.html.incl
Message-ID: <200905200629.n4K6Tbis015243@zion.cs.uiuc.edu>

Author: cristic
Date: Wed May 20 01:29:36 2009
New Revision: 72163

URL: http://llvm.org/viewvc/llvm-project?rev=72163&view=rev
Log:
Web page changes. Added a nice way to format code (div.code) in our CSS file.
Added download and install instructions, and a bug report page.


Added:
    klee/trunk/www/bugs.html
    klee/trunk/www/install.html
Modified:
    klee/trunk/www/content.css
    klee/trunk/www/index.html
    klee/trunk/www/menu.html.incl

Added: klee/trunk/www/bugs.html
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/bugs.html?rev=72163&view=auto

==============================================================================
--- klee/trunk/www/bugs.html (added)
+++ klee/trunk/www/bugs.html Wed May 20 01:29:36 2009
@@ -0,0 +1,25 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 
+          "http://www.w3.org/TR/html4/strict.dtd">
+<!-- Material used from: HTML 4.01 specs: http://www.w3.org/TR/html401/ -->
+<html>
+<head>
+  <META http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
+  <title>The KLEE Symbolic Virtual Machine</title>
+  <link type="text/css" rel="stylesheet" href="menu.css">
+  <link type="text/css" rel="stylesheet" href="content.css">
+</head>
+<body>
+<include virtual="menu.html.incl">
+<div id="content">
+  <!--*********************************************************************-->
+  <h1>KLEE Bug Reports</h1>
+  <!--*********************************************************************-->
+
+  Please report any bugs in KLEE to
+  the <a href="http://keeda.stanford.edu/mailman/listinfo/klee-dev">klee-dev
+  mailing list</a>.  You need to be subscribe to the list in order to
+  send your report.
+
+</div>
+</body>
+</html>

Modified: klee/trunk/www/content.css
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/content.css?rev=72163&r1=72162&r2=72163&view=diff

==============================================================================
--- klee/trunk/www/content.css (original)
+++ klee/trunk/www/content.css Wed May 20 01:29:36 2009
@@ -23,3 +23,12 @@
 
 /* Tables */
 tr { vertical-align:top }
+
+/* Code */
+div.code{
+    border: 1px solid LightSteelBlue ;
+    font-family: Courier New;
+    background-color: #E3E3E3;
+    padding: 7px;
+    margin: 7px;
+}

Modified: klee/trunk/www/index.html
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/index.html?rev=72163&r1=72162&r2=72163&view=diff

==============================================================================
--- klee/trunk/www/index.html (original)
+++ klee/trunk/www/index.html Wed May 20 01:29:36 2009
@@ -4,7 +4,7 @@
 <html>
 <head>
   <META http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
-  <title>The klee Symbolic Virtual Machine</title>
+  <title>The KLEE Symbolic Virtual Machine</title>
   <link type="text/css" rel="stylesheet" href="menu.css">
   <link type="text/css" rel="stylesheet" href="content.css">
 </head>
@@ -12,10 +12,10 @@
 <!--#include virtual="menu.html.incl"-->
 <div id="content">
   <!--*********************************************************************-->
-  <h1>The klee Symbolic Virtual Machine</h1>
+  <h1>The KLEE Symbolic Virtual Machine</h1>
   <!--*********************************************************************-->
 
-  <p>klee is a symbolic virtual machine built on top of
+  <p>KLEE is a symbolic virtual machine built on top of
   the <a href="http://llvm.org">LLVM</a> compiler infrastructure.</p>
 
   <p>Stay tuned for more information on public availability. For now,
@@ -28,5 +28,6 @@
   2008</a> paper.</p>
 
 </div>
+
 </body>
 </html>

Added: klee/trunk/www/install.html
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/install.html?rev=72163&view=auto

==============================================================================
--- klee/trunk/www/install.html (added)
+++ klee/trunk/www/install.html Wed May 20 01:29:36 2009
@@ -0,0 +1,72 @@
+<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 
+          "http://www.w3.org/TR/html4/strict.dtd">
+<!-- Material used from: HTML 4.01 specs: http://www.w3.org/TR/html401/ -->
+<html>
+<head>
+  <META http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
+  <title>The KLEE Symbolic Virtual Machine</title>
+  <link type="text/css" rel="stylesheet" href="menu.css">
+  <link type="text/css" rel="stylesheet" href="content.css">
+</head>
+<body>
+<include virtual="menu.html.incl">
+<div id="content">
+  <!--*********************************************************************-->
+  <h1>KLEE Installation Guide</h1>
+  <!--*********************************************************************-->
+
+  <h2> Minimal Installation </h2>
+  <ol>
+  <li> Obtain the current version
+  of <a href="http://llvm.org/releases/download.html#svn">LLVM via
+  SVN</a>.  Build the release version of LLVM and install the GCC
+  front end as well.  Follow the instructions
+  at <a href="http://llvm.org/docs/GettingStarted.html">Getting
+  Started with the LLVM System</a>.</li>
+
+  <li> Obtain the current version of KLEE via SVN (readonly access):
+       <div class="code">svn co http://llvm.org/svn/llvm-project/klee/trunk klee</div>
+  </li>
+
+  <li> Configure KLEE: 
+       <div class="code">
+	 ./configure --with-llvm=<i>path/to/llvm ENABLE_OPTIMIZED=1</i>
+       </div>
+
+    <p>This assumes that you compiled LLVM in-place.  If you used a
+      different directory for the object files then use:
+      <div class="code">
+	./configure --with-llvmsrc=<i>path/to/llvm/src</i>--with-llvmobj=<i>path/to/llvm/obj</i>
+      </div>
+  </li>
+
+  <li> Build the release version of KLEE:
+       <div class="code">
+         make
+       </div>
+  </li>
+
+  <li> Run the test suite.  You should have no unexpected failures.but
+  please <a href="bugs.html">report</a> any such failures.
+       <div class="code">
+         make test
+       </div>
+  </li>
+
+  <li> You're ready to go!  Go to the <a href="tutorials.html">Tutorials</a> page to try KLEE.</li>
+  </ol>
+
+<!--   <h2> Full Installation </h2> -->
+  
+<!--   If you need uCLibc and/or POSIX support add <i>-with-uclibc</i> -->
+<!--   and <i>-enable-posix-runtime</i> to configure.  Thus, to enable -->
+<!--   both, replace step 3 above with: -->
+  
+<!--   <div class="code"> -->
+<!--     ./configure -with-llvm=<i>path/to/llvm</i> -with-uclibc -enable-posix-runtime ENABLE_OPTIMIZED=1 -->
+<!--   </div> -->
+<!--   However, note that... -->
+
+</div>
+</body>
+</html>

Modified: klee/trunk/www/menu.html.incl
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/www/menu.html.incl?rev=72163&r1=72162&r2=72163&view=diff

==============================================================================
--- klee/trunk/www/menu.html.incl (original)
+++ klee/trunk/www/menu.html.incl Wed May 20 01:29:36 2009
@@ -6,6 +6,7 @@
   <div class="submenu">
     <label>klee Info</label>
     <a href="index.html">About</a>
+    <a href="install.html">Download &amp; Install</a>
   </div>
 
   <div class="submenu">




From nicolas.geoffray at lip6.fr  Wed May 20 01:58:29 2009
From: nicolas.geoffray at lip6.fr (Nicolas Geoffray)
Date: Wed, 20 May 2009 08:58:29 +0200
Subject: [llvm-commits] [llvm] r72140 - in /llvm/trunk:
 autoconf/configure.ac configure include/llvm/System/Atomic.h
In-Reply-To: <200905192218.n4JMIv6n031660@zion.cs.uiuc.edu>
References: <200905192218.n4JMIv6n031660@zion.cs.uiuc.edu>
Message-ID: <4A13AA15.8090404@lip6.fr>

Great, it's working now on my machine. Thanks Owen for fixing the 
problem that quickly!

Nicolas

Owen Anderson wrote:
> Author: resistor
> Date: Tue May 19 17:18:56 2009
> New Revision: 72140
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72140&view=rev
> Log:
> Split the ENABLE_THREADS #define in two.  Now ENABLE_THREADS indicates the ability to run multiple threads at once in the JIT, and requires only
> mutex support.  LLVM_MULTITHREADED indicates (or will indicate) the ability to run LLVM itself across multiple threads, and requires atomics support.
>
> Modified:
>     llvm/trunk/autoconf/configure.ac
>     llvm/trunk/configure
>     llvm/trunk/include/llvm/System/Atomic.h
>
> Modified: llvm/trunk/autoconf/configure.ac
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/autoconf/configure.ac?rev=72140&r1=72139&r2=72140&view=diff
>
> ==============================================================================
> --- llvm/trunk/autoconf/configure.ac (original)
> +++ llvm/trunk/autoconf/configure.ac Tue May 19 17:18:56 2009
> @@ -372,6 +372,7 @@
>    default) AC_SUBST(ENABLE_THREADS,[1]) ;;
>    *) AC_MSG_ERROR([Invalid setting for --enable-threads. Use "yes" or "no"]) ;;
>  esac
> +AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
>  
>  dnl Allow building without position independent code
>  AC_ARG_ENABLE(pic,
> @@ -897,24 +898,22 @@
>  fi
>  
>  dnl atomic builtins are required for threading support.
> -if test "$ENABLE_THREADS" -eq 1 ; then
> -  AC_MSG_CHECKING(for GCC atomic builtins)
> -  AC_LINK_IFELSE(
> -	AC_LANG_SOURCE(
> -          [[int main() { 
> -              volatile unsigned long val = 1;
> -              __sync_synchronize();
> -              __sync_val_compare_and_swap(&val, 1, 0);
> -              return 0;
> -            }
> -          ]]),
> -        AC_MSG_RESULT(yes),
> -        AC_MSG_RESULT(no)
> -        AC_SUBST(ENABLE_THREADS, [0])
> -        AC_MSG_WARN([Threading support disabled because atomic builtins are missing]))
> -fi
> +AC_MSG_CHECKING(for GCC atomic builtins)
> +AC_LINK_IFELSE(
> +  AC_LANG_SOURCE(
> +    [[int main() { 
> +        volatile unsigned long val = 1;
> +        __sync_synchronize();
> +        __sync_val_compare_and_swap(&val, 1, 0);
> +        return 0;
> +      }
> +    ]]),
> +  AC_MSG_RESULT(yes)
> +  AC_DEFINE(LLVM_MULTITHREADED, 1, Build multithreading support into LLVM),
> +  AC_MSG_RESULT(no)
> +  AC_DEFINE(LLVM_MULTITHREADED, 0, Build multithreading support into LLVM)
> +  AC_MSG_WARN([LLVM will be built thread-unsafe because atomic builtins are missing]))
>  
> -AC_DEFINE_UNQUOTED([ENABLE_THREADS],$ENABLE_THREADS,[Define if threads enabled])
>  
>  dnl===-----------------------------------------------------------------------===
>  dnl===
>
> Modified: llvm/trunk/configure
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/configure?rev=72140&r1=72139&r2=72140&view=diff
>
> ==============================================================================
> --- llvm/trunk/configure (original)
> +++ llvm/trunk/configure Tue May 19 17:18:56 2009
> @@ -4870,6 +4870,11 @@
>     { (exit 1); exit 1; }; } ;;
>  esac
>  
> +cat >>confdefs.h <<_ACEOF
> +#define ENABLE_THREADS $ENABLE_THREADS
> +_ACEOF
> +
> +
>  # Check whether --enable-pic was given.
>  if test "${enable_pic+set}" = set; then
>    enableval=$enable_pic;
> @@ -10570,7 +10575,7 @@
>    lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
>    lt_status=$lt_dlunknown
>    cat > conftest.$ac_ext <<EOF
> -#line 10573 "configure"
> +#line 10578 "configure"
>  #include "confdefs.h"
>  
>  #if HAVE_DLFCN_H
> @@ -12714,7 +12719,7 @@
>    ;;
>  *-*-irix6*)
>    # Find out which ABI we are using.
> -  echo '#line 12717 "configure"' > conftest.$ac_ext
> +  echo '#line 12722 "configure"' > conftest.$ac_ext
>    if { (eval echo "$as_me:$LINENO: \"$ac_compile\"") >&5
>    (eval $ac_compile) 2>&5
>    ac_status=$?
> @@ -14432,11 +14437,11 @@
>     -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:14435: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:14440: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:14439: \$? = $ac_status" >&5
> +   echo "$as_me:14444: \$? = $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.
> @@ -14700,11 +14705,11 @@
>     -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:14703: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:14708: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:14707: \$? = $ac_status" >&5
> +   echo "$as_me:14712: \$? = $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.
> @@ -14804,11 +14809,11 @@
>     -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:14807: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:14812: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>out/conftest.err)
>     ac_status=$?
>     cat out/conftest.err >&5
> -   echo "$as_me:14811: \$? = $ac_status" >&5
> +   echo "$as_me:14816: \$? = $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
> @@ -17256,7 +17261,7 @@
>    lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
>    lt_status=$lt_dlunknown
>    cat > conftest.$ac_ext <<EOF
> -#line 17259 "configure"
> +#line 17264 "configure"
>  #include "confdefs.h"
>  
>  #if HAVE_DLFCN_H
> @@ -17356,7 +17361,7 @@
>    lt_dlunknown=0; lt_dlno_uscore=1; lt_dlneed_uscore=2
>    lt_status=$lt_dlunknown
>    cat > conftest.$ac_ext <<EOF
> -#line 17359 "configure"
> +#line 17364 "configure"
>  #include "confdefs.h"
>  
>  #if HAVE_DLFCN_H
> @@ -19724,11 +19729,11 @@
>     -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:19727: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:19732: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:19731: \$? = $ac_status" >&5
> +   echo "$as_me:19736: \$? = $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.
> @@ -19828,11 +19833,11 @@
>     -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:19831: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:19836: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>out/conftest.err)
>     ac_status=$?
>     cat out/conftest.err >&5
> -   echo "$as_me:19835: \$? = $ac_status" >&5
> +   echo "$as_me:19840: \$? = $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
> @@ -21398,11 +21403,11 @@
>     -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:21401: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:21406: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:21405: \$? = $ac_status" >&5
> +   echo "$as_me:21410: \$? = $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.
> @@ -21502,11 +21507,11 @@
>     -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:21505: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:21510: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>out/conftest.err)
>     ac_status=$?
>     cat out/conftest.err >&5
> -   echo "$as_me:21509: \$? = $ac_status" >&5
> +   echo "$as_me:21514: \$? = $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
> @@ -23737,11 +23742,11 @@
>     -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:23740: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:23745: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:23744: \$? = $ac_status" >&5
> +   echo "$as_me:23749: \$? = $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.
> @@ -24005,11 +24010,11 @@
>     -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:24008: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:24013: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>conftest.err)
>     ac_status=$?
>     cat conftest.err >&5
> -   echo "$as_me:24012: \$? = $ac_status" >&5
> +   echo "$as_me:24017: \$? = $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.
> @@ -24109,11 +24114,11 @@
>     -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:24112: $lt_compile\"" >&5)
> +   (eval echo "\"\$as_me:24117: $lt_compile\"" >&5)
>     (eval "$lt_compile" 2>out/conftest.err)
>     ac_status=$?
>     cat out/conftest.err >&5
> -   echo "$as_me:24116: \$? = $ac_status" >&5
> +   echo "$as_me:24121: \$? = $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
> @@ -33602,21 +33607,20 @@
>    fi
>  fi
>  
> -if test "$ENABLE_THREADS" -eq 1 ; then
> -  { echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
> +{ echo "$as_me:$LINENO: checking for GCC atomic builtins" >&5
>  echo $ECHO_N "checking for GCC atomic builtins... $ECHO_C" >&6; }
> -  cat >conftest.$ac_ext <<_ACEOF
> +cat >conftest.$ac_ext <<_ACEOF
>  /* confdefs.h.  */
>  _ACEOF
>  cat confdefs.h >>conftest.$ac_ext
>  cat >>conftest.$ac_ext <<_ACEOF
>  /* end confdefs.h.  */
>  int main() {
> -              volatile unsigned long val = 1;
> -              __sync_synchronize();
> -              __sync_val_compare_and_swap(&val, 1, 0);
> -              return 0;
> -            }
> +        volatile unsigned long val = 1;
> +        __sync_synchronize();
> +        __sync_val_compare_and_swap(&val, 1, 0);
> +        return 0;
> +      }
>  
>  _ACEOF
>  rm -f conftest.$ac_objext conftest$ac_exeext
> @@ -33655,26 +33659,28 @@
>    (exit $ac_status); }; }; then
>    { echo "$as_me:$LINENO: result: yes" >&5
>  echo "${ECHO_T}yes" >&6; }
> +
> +cat >>confdefs.h <<\_ACEOF
> +#define LLVM_MULTITHREADED 1
> +_ACEOF
> +
>  else
>    echo "$as_me: failed program was:" >&5
>  sed 's/^/| /' conftest.$ac_ext >&5
>  
>  	{ echo "$as_me:$LINENO: result: no" >&5
>  echo "${ECHO_T}no" >&6; }
> -        ENABLE_THREADS=0
>  
> -        { echo "$as_me:$LINENO: WARNING: Threading support disabled because atomic builtins are missing" >&5
> -echo "$as_me: WARNING: Threading support disabled because atomic builtins are missing" >&2;}
> +cat >>confdefs.h <<\_ACEOF
> +#define LLVM_MULTITHREADED 0
> +_ACEOF
> +
> +  { echo "$as_me:$LINENO: WARNING: LLVM will be built thread-unsafe because atomic builtins are missing" >&5
> +echo "$as_me: WARNING: LLVM will be built thread-unsafe because atomic builtins are missing" >&2;}
>  fi
>  
>  rm -f core conftest.err conftest.$ac_objext \
>        conftest$ac_exeext conftest.$ac_ext
> -fi
> -
> -
> -cat >>confdefs.h <<_ACEOF
> -#define ENABLE_THREADS $ENABLE_THREADS
> -_ACEOF
>  
>  
>  
>
> Modified: llvm/trunk/include/llvm/System/Atomic.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72140&r1=72139&r2=72140&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/System/Atomic.h (original)
> +++ llvm/trunk/include/llvm/System/Atomic.h Tue May 19 17:18:56 2009
> @@ -23,14 +23,8 @@
>    namespace sys {
>      
>      inline void MemoryFence() {
> -#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
> -#  if defined(__GNUC__)
> -      __asm__ __volatile__("" : : : "memory");
> -#  elif defined(_MSC_VER)
> -      __asm { };
> -#  else
> -#    error No memory fence implementation for your platform!
> -#  endif
> +#if LLVM_MULTITHREADED==0
> +      return;
>  #else
>  #  if defined(__GNUC__)
>        __sync_synchronize();
> @@ -42,7 +36,7 @@
>  #endif
>  }
>  
> -#if !defined(ENABLE_THREADS) || ENABLE_THREADS == 0
> +#if LLVM_MULTITHREADED==0
>      typedef unsigned long cas_flag;
>      template<typename T>
>      inline T CompareAndSwap(volatile T* dest,
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>   



From baldrick at free.fr  Wed May 20 02:10:14 2009
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 20 May 2009 09:10:14 +0200
Subject: [llvm-commits] [llvm] r72162 - in /llvm/trunk:
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
 test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll test/CodeGen/XCore/fneg.ll
In-Reply-To: <200905200602.n4K62Ab6014388@zion.cs.uiuc.edu>
References: <200905200602.n4K62Ab6014388@zion.cs.uiuc.edu>
Message-ID: <4A13ACD6.9050008@free.fr>

Hi Eli,

> Fix for PR4235: to build a floating-point value from integer parts, 
> build an integer and cast that to a float.  This fixes a crash 
> caused by trying to split an f32 into two f16's.

in general if getCopyFromParts has a problem then so does
getCopyToParts.  Can you please eyeball it to see if it is ok.

Thanks,

Duncan.


From isanbard at gmail.com  Wed May 20 04:24:31 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 02:24:31 -0700
Subject: [llvm-commits] Add scope info in DebugLoc
In-Reply-To: <4A11C253.8050401@gmail.com>
References: <4A037F8D.90001@gmail.com>
	<16e5fdf90905121328r50ee3767n290bc2ff7000c6a6@mail.gmail.com>
	<4A11C253.8050401@gmail.com>
Message-ID: <F01471C6-37FC-49AF-A504-FC53E054C0C7@gmail.com>

Hi Argiris,

On May 18, 2009, at 1:17 PM, Argiris Kirtzidis wrote:

> Hi Bill,
>
> Bill Wendling wrote:
>> -  /// DebugLocTuple - Debug location tuple of filename id, line  
>> and column.
>> +  /// DebugScope - Debug scope id.
>>
>> Please flesh out this comment. Add more about what a debug scope is
>> and how it will be used.
>>
>> +  class DebugScope {
>> +    unsigned Idx;
>> +
>> +  public:
>> +    DebugScope() : Idx(~0U) {}  // Defaults to null.
>>
>> Well...it's not really Null. :-) Maybe it would be better to name it
>> "invalid" (with appropriate changes to the other methods' names)?
>>
>> +  /// DebugScopeInfo - Contains info about the scope global  
>> variable and the
>> +  /// parent debug scope.
>>
>> Could you explain why this is different from DebugScope.
>>
>
> Ok to all.
>
Thanks! :)

>> +  /// DebugLocTuple - Debug location tuple of compile unit, scope,
>> +  /// line and column.
>>
>> The term "compile unit" is overloaded. There's the idea of a CU in  
>> the
>> debug info, then the actual CompileUnit class, then a DICompileUnit
>> class. Could you name it something else? Just for my sanity? :-) That
>> is, unless CompileUnit is really absolutely correct here...
>>
>
> I modified it to "DICompileUnit global variable", is this more clear ?
>
It is.

>> +  void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction  
>> &MF);
>> +  void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
>>
>> FastISel has MF available as an ivar. Do you need to pass this in  
>> as a
>> parameter? Also, please comment these methods.
>>
>
> Moved them into a new DebugScopeTracker class.
>
Thank you!

>> +DebugScope MachineFunction::CreateDebugScope(GlobalVariable  
>> *ScopeGV,
>> +                                             DebugScope Parent) {
>> +  DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
>> +  return DebugScope::get(DbgScopeInfos.size() - 1);
>>
>> Should you verify that DbgScopeInfos.size() > 0 here?
>
> Is it really needed ? A new item is pushed into the DbgScopeInfos in  
> the immediate line above it.
>
Ah! You're right. I was blind to it. :-)

>> Actually, why
>> subtract one here at all? In the assert, you can verify that Idx <=
>> DbgScopeInfos.size().
>>
>
> Well, there's not much difference but I'd prefer to have  
> DebugScope::getIndex return the "actual index" into the vector.
> This is similar to the way DebugLoc's index is created.
>
Now that I see that the original code wasn't buggy, this is fine.

>> +  // FIXME: Proper debug scope.
>> +  DebugScope DbgScope;
>>
>> What's stopping you from using a proper debug scope here?
>>
>
> I added comments about this. Let me know what you think about this  
> part.

Looks good!

>> @@ -3817,6 +3821,25 @@
>>   setValue(&I, result);
>> }
>>
>> +void SelectionDAGLowering::EnterDebugScope(GlobalVariable *ScopeGV,
>> +                                           MachineFunction &MF) {
>> +  assert(ScopeGV && "GlobalVariable for scope is null!");
>> +  setCurDebugScope(MF.CreateDebugScope(ScopeGV,  
>> getCurDebugScope()));
>> +}
>> +
>> +void SelectionDAGLowering::ExitDebugScope(GlobalVariable *ScopeGV,
>> +                                          MachineFunction &MF) {
>> +  assert(ScopeGV && "GlobalVariable for scope is null!");
>> +  assert(!getCurDebugScope().isNull() && "Mismatched region.end ?");
>> +  // We may have skipped a region.end because it was in an  
>> unreachable block.
>> +  // Go up the scope chain until we reach the scope that ScopeGV  
>> points to.
>> +  DebugScopeInfo DSI;
>> +  do {
>> +    DSI = MF.getDebugScopeInfo(getCurDebugScope());
>> +    setCurDebugScope(DSI.Parent);
>> +  } while (!DSI.Parent.isNull() && DSI.GV != ScopeGV);
>> +}
>>
>> Because these methods are used in both SelectionDAGBuild.cpp and
>> FastISel.cpp, is there a way to factor them out into another module?
>> Maybe a CodeGen/DebugLoc.cpp module?
>>
>
> Done.
>
> I attached a new patch, thanks for reviewing!
>
The patch looks good. :-) Be careful about 80-column violations (I  
think I saw one potential one?). Okay to commit.

Thank you for working on this!
-bw


From resistor at mac.com  Wed May 20 04:34:25 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 09:34:25 -0000
Subject: [llvm-commits] [llvm] r72167 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905200934.n4K9YTBR000917@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 04:34:13 2009
New Revision: 72167

URL: http://llvm.org/viewvc/llvm-project?rev=72167&view=rev
Log:
Compile fix for MSVC.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72167&r1=72166&r2=72167&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Wed May 20 04:34:13 2009
@@ -17,6 +17,7 @@
 #include "llvm/Config/config.h"
 
 #if defined(_MSC_VER)
+#define NOMINMAX
 #include <windows.h>
 #endif
 




From howard0su at gmail.com  Wed May 20 00:18:26 2009
From: howard0su at gmail.com (Howard Su)
Date: Wed, 20 May 2009 13:18:26 +0800
Subject: [llvm-commits] [llvm] r72148 -
	/llvm/trunk/include/llvm/System/Atomic.h
In-Reply-To: <f126fae00905192018o1b8eb416m18994741c74ae10c@mail.gmail.com>
References: <200905200031.n4K0VTjb003694@zion.cs.uiuc.edu>
	<f126fae00905192018o1b8eb416m18994741c74ae10c@mail.gmail.com>
Message-ID: <f126fae00905192218w40e031f1ub52b09362812abac@mail.gmail.com>

after run through the build, I found the suggested fix can not solve other
problem like:     1>..\..\..\llvm-ng\tools\llvm-ld\llvm-ld.cpp(414): error
C2039: 'CopyFileA' : is not a member of 'llvm::sys'
     1>..\..\..\llvm-ng\tools\llvm-ld\llvm-ld.cpp(414): error C2664:
'CopyFileA' : cannot convert parameter 1 from 'llvm::sys::Path' to 'LPCSTR'
In general, windows header will redefine CopyFile, FindExecutable, etc to a
unicode version and a non-unicode version. So the fix can be add a inline
method which implement MemoryBarrier instead of including windows.h.


On Wed, May 20, 2009 at 11:18 AM, Howard Su <howard0su at gmail.com> wrote:

> I got the following errors when compiling under vs2008:..\..\..\llvm-ng\lib\Support\Timer.cpp(178):
> error C2589: '(' : illegal token on right side of '::'
> ..\..\..\llvm-ng\lib\Support\Timer.cpp(178): error C2059: syntax error :
> '::'
>
> I think this is broken after rev:72087.
>
> Under VS, windows.h also defined a min,max macro which cause the conflict.
> I will suggest to fix as:
> #if defined(_MSC_VER)
> #define NOMINMAX
> #include <windows.h>
> #endif
>
> --
> -Howard
>



-- 
-Howard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/17d76356/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: build_under_windows.patch
Type: application/octet-stream
Size: 500 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/17d76356/attachment.obj 

From eli.friedman at gmail.com  Wed May 20 05:07:44 2009
From: eli.friedman at gmail.com (Eli Friedman)
Date: Wed, 20 May 2009 03:07:44 -0700
Subject: [llvm-commits] [llvm] r72162 - in /llvm/trunk:
	lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
	test/CodeGen/MSP430/2009-05-19-DoubleSplit.ll
	test/CodeGen/XCore/fneg.ll
In-Reply-To: <4A13ACD6.9050008@free.fr>
References: <200905200602.n4K62Ab6014388@zion.cs.uiuc.edu>
	<4A13ACD6.9050008@free.fr>
Message-ID: <f5aa3e9b0905200307j3d60f232r3552e97f5f1343e9@mail.gmail.com>

On Wed, May 20, 2009 at 12:10 AM, Duncan Sands <baldrick at free.fr> wrote:
> Hi Eli,
>
>> Fix for PR4235: to build a floating-point value from integer parts,
>> build an integer and cast that to a float. ?This fixes a crash
>> caused by trying to split an f32 into two f16's.
>
> in general if getCopyFromParts has a problem then so does
> getCopyToParts. ?Can you please eyeball it to see if it is ok.

It looks like getCopyToParts is already doing the right thing,
although in a slightly ugly way.

-Eli



From jay.foad at gmail.com  Wed May 20 05:34:07 2009
From: jay.foad at gmail.com (Jay Foad)
Date: Wed, 20 May 2009 11:34:07 +0100
Subject: [llvm-commits] implement and use SmallVector::data()
Message-ID: <ee2e06e90905200334j546ec53dq65b050ad11a04748@mail.gmail.com>

SmallVector::operator[] has this comment:

  /* These asserts could be "Begin + idx < End", but there are lots of places
     in llvm where we use &v[v.size()] instead of v.end(). */
  reference operator[](unsigned idx) {
    assert (Begin + idx <= End);
    return Begin[idx];
  }
  const_reference operator[](unsigned idx) const {
    assert (Begin + idx <= End);
    return Begin[idx];
  }

I think this is a bit ugly, and rather non-standard, because it means
SmallVector behaves differently from std::vector.

The attached patch adds a data() method to SmallVector, following DR 464:

http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#464

... and tightens up the asserts in operator[], and converts code like
"foo(&vec[0], vec.size())" to "foo(vec.data(), vec.size())" when it
was triggering the assert. It passes "make check" and cfe's "make
test", and llvm-gcc still bootstraps.

Thoughts?

Thanks,
Jay.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch.data
Type: application/octet-stream
Size: 46397 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/edf377dd/attachment.obj 

From howard0su at gmail.com  Wed May 20 07:27:51 2009
From: howard0su at gmail.com (Howard Su)
Date: Wed, 20 May 2009 20:27:51 +0800
Subject: [llvm-commits] [llvm] r72167 -
	/llvm/trunk/include/llvm/System/Atomic.h
In-Reply-To: <200905200934.n4K9YTBR000917@zion.cs.uiuc.edu>
References: <200905200934.n4K9YTBR000917@zion.cs.uiuc.edu>
Message-ID: <f126fae00905200527m24641e69g5befca994505316e@mail.gmail.com>

This fix doesn't fully make llvm compile under MSVC. I still have the link
error when compile llvm-ld. Since windows.h will define the macro for
CopyFile. This cause the following line link to the CopyFileA.

llvm-ld.cpp
Line414:  if (0 != sys::CopyFile(sys::Path(OutputFilename), llvmstub,
&ErrMsg))


On Wed, May 20, 2009 at 5:34 PM, Owen Anderson <resistor at mac.com> wrote:

> Author: resistor
> Date: Wed May 20 04:34:13 2009
> New Revision: 72167
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72167&view=rev
> Log:
> Compile fix for MSVC.
>
> Modified:
>    llvm/trunk/include/llvm/System/Atomic.h
>
> Modified: llvm/trunk/include/llvm/System/Atomic.h
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72167&r1=72166&r2=72167&view=diff
>
>
> ==============================================================================
> --- llvm/trunk/include/llvm/System/Atomic.h (original)
> +++ llvm/trunk/include/llvm/System/Atomic.h Wed May 20 04:34:13 2009
> @@ -17,6 +17,7 @@
>  #include "llvm/Config/config.h"
>
>  #if defined(_MSC_VER)
> +#define NOMINMAX
>  #include <windows.h>
>  #endif
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



-- 
-Howard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/5696ef31/attachment.html 

From baldrick at free.fr  Wed May 20 10:12:08 2009
From: baldrick at free.fr (Duncan Sands)
Date: Wed, 20 May 2009 15:12:08 -0000
Subject: [llvm-commits] [llvm] r72168 -
	/llvm/trunk/include/llvm/Analysis/Dominators.h
Message-ID: <200905201512.n4KFC9CI012166@zion.cs.uiuc.edu>

Author: baldrick
Date: Wed May 20 10:12:01 2009
New Revision: 72168

URL: http://llvm.org/viewvc/llvm-project?rev=72168&view=rev
Log:
When comparing DominanceFrontier's, advance iterators
before erasing nodes, not after.  Otherwise dom frontier
checking reads from freed memory.

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

Modified: llvm/trunk/include/llvm/Analysis/Dominators.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/Dominators.h?rev=72168&r1=72167&r2=72168&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Analysis/Dominators.h (original)
+++ llvm/trunk/include/llvm/Analysis/Dominators.h Wed May 20 10:12:01 2009
@@ -968,8 +968,8 @@
       tmpSet.insert(*I);
 
     for (DomSetType::const_iterator I = DS1.begin(),
-           E = DS1.end(); I != E; ++I) {
-      BasicBlock *Node = *I;
+           E = DS1.end(); I != E; ) {
+      BasicBlock *Node = *I++;
 
       if (tmpSet.erase(Node) == 0)
         // Node is in DS1 but not in DS2.
@@ -993,7 +993,7 @@
       tmpFrontiers.insert(std::make_pair(I->first, I->second));
 
     for (DomSetMapType::iterator I = tmpFrontiers.begin(),
-           E = tmpFrontiers.end(); I != E; ++I) {
+           E = tmpFrontiers.end(); I != E; ) {
       BasicBlock *Node = I->first;
       const_iterator DFI = find(Node);
       if (DFI == end()) 
@@ -1002,6 +1002,7 @@
       if (compareDomSet(I->second, DFI->second))
         return true;
 
+      ++I;
       tmpFrontiers.erase(Node);
     }
 




From bob.wilson at apple.com  Wed May 20 11:30:25 2009
From: bob.wilson at apple.com (Bob Wilson)
Date: Wed, 20 May 2009 16:30:25 -0000
Subject: [llvm-commits] [llvm] r72172 -
	/llvm/trunk/lib/Target/ARM/ARMISelLowering.h
Message-ID: <200905201630.n4KGUPPM015003@zion.cs.uiuc.edu>

Author: bwilson
Date: Wed May 20 11:30:25 2009
New Revision: 72172

URL: http://llvm.org/viewvc/llvm-project?rev=72172&view=rev
Log:
Minor formatting fixes.

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=72172&r1=72171&r2=72172&view=diff

==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMISelLowering.h (original)
+++ llvm/trunk/lib/Target/ARM/ARMISelLowering.h Wed May 20 11:30:25 2009
@@ -62,7 +62,7 @@
       RRX,          // V = RRX X, Flag     -> srl X, 1 + shift in carry flag.
 
       FMRRD,        // double to two gprs.
-      FMDRR,         // Two gprs to double.
+      FMDRR,        // Two gprs to double.
 
       EH_SJLJ_SETJMP,    // SjLj exception handling setjmp
       EH_SJLJ_LONGJMP,   // SjLj exception handling longjmp
@@ -71,7 +71,7 @@
     };
   }
 
-  //===----------------------------------------------------------------------===//
+  //===--------------------------------------------------------------------===//
   //  ARMTargetLowering - ARM Implementation of the TargetLowering interface
 
   class ARMTargetLowering : public TargetLowering {




From gohman at apple.com  Wed May 20 12:03:58 2009
From: gohman at apple.com (Dan Gohman)
Date: Wed, 20 May 2009 10:03:58 -0700
Subject: [llvm-commits] [llvm] r71805 - in /llvm/trunk:
	docs/TableGenFundamentals.html
	test/TableGen/lisp.td	utils/TableGen/Record.cpp
	utils/TableGen/Record.h	utils/TableGen/TGLexer.cpp
	utils/TableGen/TGLexer.h utils/TableGen/TGParser.cpp
In-Reply-To: <200905151033.34159.dag@cray.com>
References: <200905142238.n4EMcWwf015754@zion.cs.uiuc.edu>
	<d486b780-3c59-455a-9a80-a7c9c3ac94b6@s20g2000vbp.googlegroups.com>
	<200905151033.34159.dag@cray.com>
Message-ID: <00448720-2FC1-4925-A92D-249249D80982@apple.com>


On May 15, 2009, at 8:33 AM, David Greene wrote:

> On Friday 15 May 2009 05:27, Gabor Greif wrote:
>> David,
>>
>> did you consider adding the (more modern) Scheme
>> equivalents of these primitives instead:
>>
>> car --> head
>> cdr --> tail
>> (and possibly: null --> empty)
>>
>> Car and Cdr have some very specific historical baggage added,
>> an I do not think TG should willingly ingherit this legacy.
>
> I considered different names (not the Scheme ones specifically) but  
> thought
> that being consistent with history would be a benefit for those  
> already
> familiar with LISP.  I'm certainly open to changing this.  I don't
> particularly care what these things are named.

I agree with Gabor. car and cdr symbolize LISP in a way that would
be misleading for TG to suggest.

Also, please don't implement eval ;-).

Dan



From resistor at mac.com  Wed May 20 13:26:15 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 18:26:15 -0000
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
 include/llvm/System/Atomic.h lib/System/Atomic.cpp
Message-ID: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 13:26:15 2009
New Revision: 72174

URL: http://llvm.org/viewvc/llvm-project?rev=72174&view=rev
Log:
Move atomic operations' definitions out of line.  While this seems kind of silly,
all kinds of problems caused by including windows.h and/or config.h in an LLVM header.

Added:
    llvm/trunk/lib/System/Atomic.cpp
Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72174&r1=72173&r2=72174&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Wed May 20 13:26:15 2009
@@ -14,75 +14,16 @@
 #ifndef LLVM_SYSTEM_ATOMIC_H
 #define LLVM_SYSTEM_ATOMIC_H
 
-#include "llvm/Config/config.h"
-
-#if defined(_MSC_VER)
-#define NOMINMAX
-#include <windows.h>
-#endif
-
+#include <stdint.h>
 
 namespace llvm {
   namespace sys {
-    
-    inline void MemoryFence() {
-#if LLVM_MULTITHREADED==0
-      return;
-#else
-#  if defined(__GNUC__)
-      __sync_synchronize();
-#  elif defined(_MSC_VER)
-      MemoryBarrier();
-#  else
-#    error No memory fence implementation for your platform!
-#  endif
-#endif
-}
-
-#if LLVM_MULTITHREADED==0
-    typedef unsigned long cas_flag;
-    template<typename T>
-    inline T CompareAndSwap(volatile T* dest,
-			    T exc, T c) {
-      T result = *dest;
-      if (result == c)
-        *dest = exc;
-      return result;
-    }
-#elif defined(__GNUC__)
-    typedef unsigned long cas_flag;
-    template<typename T>
-    inline T CompareAndSwap(volatile T* ptr,
-			    T new_value,
-			    T old_value) {
-      return __sync_val_compare_and_swap(ptr, old_value, new_value);
-    }
-#elif defined(_MSC_VER)
-    typedef LONG cas_flag;
-    template<typename T>
-    inline T CompareAndSwap(volatile T* ptr,
-			    T new_value,
-			    T old_value) {
-      if (sizeof(T) == 4)
-	return InterlockedCompareExchange(ptr, new_value, old_value);
-      else if (sizeof(T) == 8)
-	return InterlockedCompareExchange64(ptr, new_value, old_value);
-      else
-	assert(0 && "Unsupported compare-and-swap size!");
-    }
-    
-    template<typename T>
-    inline T* CompareAndSwap<T*>(volatile T** ptr,
-				 T* new_value,
-				 T* old_value) {
-      return InterlockedCompareExchangePtr(ptr, new_value, old_value);
-    }
-
-
-#else
-#  error No compare-and-swap implementation for your platform!
-#endif
+    void MemoryFence();
 
+    typedef uint32_t cas_flag;
+    cas_flag CompareAndSwap(volatile cas_flag* ptr,
+    	                      cas_flag new_value,
+                            cas_flag old_value);
   }
 }
 

Added: llvm/trunk/lib/System/Atomic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72174&view=auto

==============================================================================
--- llvm/trunk/lib/System/Atomic.cpp (added)
+++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 13:26:15 2009
@@ -0,0 +1,52 @@
+//===-- Atomic.cpp - Atomic Operations --------------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+//  This header file implements atomic operations.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/System/Atomic.h"
+#include "llvm/Config/config.h"
+
+using namespace llvm;
+
+#if defined(_MSC_VER)
+#include <windows.h>
+#endif
+
+void sys::MemoryFence() {
+#if LLVM_MULTITHREADED==0
+  return;
+#else
+#  if defined(__GNUC__)
+  __sync_synchronize();
+#  elif defined(_MSC_VER)
+  MemoryBarrier();
+#  else
+# error No memory fence implementation for your platform!
+#  endif
+#endif
+}
+
+sys::cas_flag sys::CompareAndSwap(volatile sys::cas_flag* ptr,
+	                                sys::cas_flag new_value,
+	                                sys::cas_flag old_value) {
+#if LLVM_MULTITHREADED==0
+  T result = *dest;
+  if (result == c)
+    *dest = exc;
+  return result;
+#elif defined(__GNUC__)
+  return __sync_val_compare_and_swap(ptr, old_value, new_value);
+#elif defined(_MSC_VER)
+	return InterlockedCompareExchange(ptr, new_value, old_value);
+#else
+#  error No compare-and-swap implementation for your platform!
+#endif
+}
\ No newline at end of file




From mrs at apple.com  Wed May 20 13:34:22 2009
From: mrs at apple.com (Mike Stump)
Date: Wed, 20 May 2009 11:34:22 -0700
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
	include/llvm/System/Atomic.h lib/System/Atomic.cpp
In-Reply-To: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
References: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
Message-ID: <03F86A6F-6EBE-46A5-BB7C-DECDF3301A8C@apple.com>

On May 20, 2009, at 11:26 AM, Owen Anderson wrote:
> Author: resistor
> Date: Wed May 20 13:26:15 2009
> New Revision: 72174
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72174&view=rev
> Log:
> Move atomic operations' definitions out of line.  While this seems  
> kind of silly,
> all kinds of problems caused by including windows.h and/or config.h  
> in an LLVM header.
>
> Added:
>    llvm/trunk/lib/System/Atomic.cpp
> Modified:
>    llvm/trunk/include/llvm/System/Atomic.h

Atomic.cpp: In function 'llvm::sys::cas_flag llvm::sys::CompareAndSwap 
(volatile llvm::sys::cas_flag*, llvm::sys::cas_flag,  
llvm::sys::cas_flag)':
Atomic.cpp:41: error: 'T' was not declared in this scope

?


From resistor at mac.com  Wed May 20 13:47:53 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 18:47:53 -0000
Subject: [llvm-commits] [llvm] r72177 - /llvm/trunk/lib/System/Atomic.cpp
Message-ID: <200905201847.n4KIlrpC019737@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 13:47:53 2009
New Revision: 72177

URL: http://llvm.org/viewvc/llvm-project?rev=72177&view=rev
Log:
Copy-and-paste-o.

Modified:
    llvm/trunk/lib/System/Atomic.cpp

Modified: llvm/trunk/lib/System/Atomic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72177&r1=72176&r2=72177&view=diff

==============================================================================
--- llvm/trunk/lib/System/Atomic.cpp (original)
+++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 13:47:53 2009
@@ -38,7 +38,7 @@
 	                                sys::cas_flag new_value,
 	                                sys::cas_flag old_value) {
 #if LLVM_MULTITHREADED==0
-  T result = *dest;
+  sys::cas_flag result = *dest;
   if (result == c)
     *dest = exc;
   return result;




From resistor at mac.com  Wed May 20 14:01:51 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 19:01:51 -0000
Subject: [llvm-commits] [llvm] r72179 - /llvm/trunk/lib/System/Atomic.cpp
Message-ID: <200905201901.n4KJ1pgm020204@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 14:01:50 2009
New Revision: 72179

URL: http://llvm.org/viewvc/llvm-project?rev=72179&view=rev
Log:
I just fail today.

Hopefully this fixes the last build errors on systems with GCC < 4.1.

Modified:
    llvm/trunk/lib/System/Atomic.cpp

Modified: llvm/trunk/lib/System/Atomic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72179&r1=72178&r2=72179&view=diff

==============================================================================
--- llvm/trunk/lib/System/Atomic.cpp (original)
+++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 14:01:50 2009
@@ -38,9 +38,9 @@
 	                                sys::cas_flag new_value,
 	                                sys::cas_flag old_value) {
 #if LLVM_MULTITHREADED==0
-  sys::cas_flag result = *dest;
-  if (result == c)
-    *dest = exc;
+  sys::cas_flag result = *ptr;
+  if (result == old_value)
+    *ptr = new_value;
   return result;
 #elif defined(__GNUC__)
   return __sync_val_compare_and_swap(ptr, old_value, new_value);




From resistor at mac.com  Wed May 20 14:06:49 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 19:06:49 -0000
Subject: [llvm-commits] [llvm] r72180 - /llvm/trunk/lib/System/Atomic.cpp
Message-ID: <200905201906.n4KJ6oxJ020500@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 14:06:49 2009
New Revision: 72180

URL: http://llvm.org/viewvc/llvm-project?rev=72180&view=rev
Log:
Tabs, be gone!

Modified:
    llvm/trunk/lib/System/Atomic.cpp

Modified: llvm/trunk/lib/System/Atomic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72180&r1=72179&r2=72180&view=diff

==============================================================================
--- llvm/trunk/lib/System/Atomic.cpp (original)
+++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 14:06:49 2009
@@ -35,8 +35,8 @@
 }
 
 sys::cas_flag sys::CompareAndSwap(volatile sys::cas_flag* ptr,
-	                                sys::cas_flag new_value,
-	                                sys::cas_flag old_value) {
+                                  sys::cas_flag new_value,
+                                  sys::cas_flag old_value) {
 #if LLVM_MULTITHREADED==0
   sys::cas_flag result = *ptr;
   if (result == old_value)
@@ -45,7 +45,7 @@
 #elif defined(__GNUC__)
   return __sync_val_compare_and_swap(ptr, old_value, new_value);
 #elif defined(_MSC_VER)
-	return InterlockedCompareExchange(ptr, new_value, old_value);
+  return InterlockedCompareExchange(ptr, new_value, old_value);
 #else
 #  error No compare-and-swap implementation for your platform!
 #endif




From isanbard at gmail.com  Wed May 20 15:27:01 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 13:27:01 -0700
Subject: [llvm-commits] [llvm] r72180 - /llvm/trunk/lib/System/Atomic.cpp
In-Reply-To: <200905201906.n4KJ6oxJ020500@zion.cs.uiuc.edu>
References: <200905201906.n4KJ6oxJ020500@zion.cs.uiuc.edu>
Message-ID: <16e5fdf90905201327n709bb4afwbcf79bbfb7f99e42@mail.gmail.com>

This is now working. Thanks, Owen!

-bw

On Wed, May 20, 2009 at 12:06 PM, Owen Anderson <resistor at mac.com> wrote:
> Author: resistor
> Date: Wed May 20 14:06:49 2009
> New Revision: 72180
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72180&view=rev
> Log:
> Tabs, be gone!
>
> Modified:
> ? ?llvm/trunk/lib/System/Atomic.cpp
>
> Modified: llvm/trunk/lib/System/Atomic.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72180&r1=72179&r2=72180&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/System/Atomic.cpp (original)
> +++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 14:06:49 2009
> @@ -35,8 +35,8 @@
> ?}
>
> ?sys::cas_flag sys::CompareAndSwap(volatile sys::cas_flag* ptr,
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sys::cas_flag new_value,
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? sys::cas_flag old_value) {
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?sys::cas_flag new_value,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?sys::cas_flag old_value) {
> ?#if LLVM_MULTITHREADED==0
> ? sys::cas_flag result = *ptr;
> ? if (result == old_value)
> @@ -45,7 +45,7 @@
> ?#elif defined(__GNUC__)
> ? return __sync_val_compare_and_swap(ptr, old_value, new_value);
> ?#elif defined(_MSC_VER)
> - ? ? ? return InterlockedCompareExchange(ptr, new_value, old_value);
> + ?return InterlockedCompareExchange(ptr, new_value, old_value);
> ?#else
> ?# ?error No compare-and-swap implementation for your platform!
> ?#endif
>
>
> _______________________________________________
> 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 May 20 15:35:08 2009
From: clattner at apple.com (Chris Lattner)
Date: Wed, 20 May 2009 13:35:08 -0700
Subject: [llvm-commits] implement and use SmallVector::data()
In-Reply-To: <ee2e06e90905200334j546ec53dq65b050ad11a04748@mail.gmail.com>
References: <ee2e06e90905200334j546ec53dq65b050ad11a04748@mail.gmail.com>
Message-ID: <2D3D224C-A384-416F-B7A3-98E9CB6AB2C0@apple.com>


On May 20, 2009, at 3:34 AM, Jay Foad wrote:

> SmallVector::operator[] has this comment:
>
>  /* These asserts could be "Begin + idx < End", but there are lots  
> of places
>     in llvm where we use &v[v.size()] instead of v.end(). */
>  reference operator[](unsigned idx) {
>    assert (Begin + idx <= End);
>    return Begin[idx];
>  }
>  const_reference operator[](unsigned idx) const {
>    assert (Begin + idx <= End);
>    return Begin[idx];
>  }
>
> I think this is a bit ugly, and rather non-standard, because it means
> SmallVector behaves differently from std::vector.
>
> The attached patch adds a data() method to SmallVector, following DR  
> 464:
>
> http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-defects.html#464
>
> ... and tightens up the asserts in operator[], and converts code like
> "foo(&vec[0], vec.size())" to "foo(vec.data(), vec.size())" when it
> was triggering the assert. It passes "make check" and cfe's "make
> test", and llvm-gcc still bootstraps.

Looks nice to me, please apply!

-Chris


From clattner at apple.com  Wed May 20 15:36:29 2009
From: clattner at apple.com (Chris Lattner)
Date: Wed, 20 May 2009 13:36:29 -0700
Subject: [llvm-commits] [llvm] r72148
	-	/llvm/trunk/include/llvm/System/Atomic.h
In-Reply-To: <200905200031.n4K0VTjb003694@zion.cs.uiuc.edu>
References: <200905200031.n4K0VTjb003694@zion.cs.uiuc.edu>
Message-ID: <F1CDCB4B-F930-47D3-B263-54FCCFF49F4D@apple.com>

On May 19, 2009, at 5:31 PM, Owen Anderson wrote:
> Author: resistor
> Date: Tue May 19 19:31:29 2009
> New Revision: 72148
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72148&view=rev
> Log:
> We need to include config.h here so that the #defines are set  
> properly.

This isn't going to work: LLVM headers can't include config.h, because  
it conflicts with other projects that use their own config.h.  Why  
does "gcc atomics support" need to be autoconf'd?  Can' you just check  
against the gcc version macros?

-Chris

>
>
> Modified:
>    llvm/trunk/include/llvm/System/Atomic.h
>
> Modified: llvm/trunk/include/llvm/System/Atomic.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72148&r1=72147&r2=72148&view=diff
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/include/llvm/System/Atomic.h (original)
> +++ llvm/trunk/include/llvm/System/Atomic.h Tue May 19 19:31:29 2009
> @@ -14,6 +14,8 @@
> #ifndef LLVM_SYSTEM_ATOMIC_H
> #define LLVM_SYSTEM_ATOMIC_H
>
> +#include "llvm/Config/config.h"
> +
> #if defined(_MSC_VER)
> #include <windows.h>
> #endif
>
>
> _______________________________________________
> 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 May 20 15:38:00 2009
From: clattner at apple.com (Chris Lattner)
Date: Wed, 20 May 2009 13:38:00 -0700
Subject: [llvm-commits] [PATCH] Thread-safe ManagedStatic
In-Reply-To: <16551545510839478801965352385268027480-Webmail@me.com>
References: <200905190107.n4J17f3A006679@zion.cs.uiuc.edu>
	<5464957333051772950217202212009693918-Webmail@me.com>
	<A091075B-9531-4FA9-BAE4-A39049026353@apple.com>
	<358F1AAC-7DF4-4ABD-87DC-4B3B581B1EA7@mac.com>
	<8DD69FC6-D713-4451-899D-C8FE8D508E75@apple.com>
	<16551545510839478801965352385268027480-Webmail@me.com>
Message-ID: <D4CB013C-99D7-490B-98DB-1B3EBE5D6480@apple.com>


On May 19, 2009, at 1:20 PM, Owen Anderson wrote:

> Alright, is this something more like what you had in mind?

No, this still has a per-object mutex that you created yourself.  The  
idea is to have a single global "pthreads" mutex out of line in a .cpp  
file.

-Chris


From clattner at apple.com  Wed May 20 15:43:44 2009
From: clattner at apple.com (Chris Lattner)
Date: Wed, 20 May 2009 13:43:44 -0700
Subject: [llvm-commits] [llvm] r72151 - in /llvm/trunk:
	include/llvm/Support/ManagedStatic.h lib/Support/ManagedStatic.cpp
In-Reply-To: <200905200039.n4K0dRtm003968@zion.cs.uiuc.edu>
References: <200905200039.n4K0dRtm003968@zion.cs.uiuc.edu>
Message-ID: <6DA25E85-E172-450B-9E5B-A6BF017A44C2@apple.com>


On May 19, 2009, at 5:39 PM, Owen Anderson wrote:

> Author: resistor
> Date: Tue May 19 19:39:20 2009
> New Revision: 72151
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72151&view=rev
> Log:
> Add llvm_start_multithreaded(), which starts up the LLVM internals  
> in thread-safe mode.  Provide double-check locking
> initialization of ManagedStatic's when running in thread-safe mode.

This looks great!
>
> +void llvm::llvm_start_multithreaded() {
> +#if LLVM_MULTITHREADED
> +  assert(ManagedStaticMutex == 0 && "Multithreaded LLVM already  
> initialized!");
> +  ManagedStaticMutex = new sys::Mutex(true);
> +#else
> +  assert(0 && "LLVM built without multithreading support!");
> +#endif
> +}

How about making llvm_start_multithreaded return a bool indicating  
whether or not it was successful?  This way clients could detect and  
nicely report the error if multithreading wasn't enabled properly.

Thanks Owen,

-Chris


From resistor at mac.com  Wed May 20 16:03:08 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 21:03:08 -0000
Subject: [llvm-commits] [llvm] r72182 - in /llvm/trunk:
 include/llvm/Support/ManagedStatic.h lib/Support/ManagedStatic.cpp
Message-ID: <200905202103.n4KL38lM024284@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 16:03:06 2009
New Revision: 72182

URL: http://llvm.org/viewvc/llvm-project?rev=72182&view=rev
Log:
Have llvm_start_multithreaded return a bool indicating whether multithreaded
initialization succeeded or not, rather than just asserting.

Modified:
    llvm/trunk/include/llvm/Support/ManagedStatic.h
    llvm/trunk/lib/Support/ManagedStatic.cpp

Modified: llvm/trunk/include/llvm/Support/ManagedStatic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/ManagedStatic.h?rev=72182&r1=72181&r2=72182&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Support/ManagedStatic.h (original)
+++ llvm/trunk/include/llvm/Support/ManagedStatic.h Wed May 20 16:03:06 2009
@@ -96,8 +96,10 @@
 
 
 /// llvm_start_multithreaded - Allocate and initialize structures needed to
-/// make LLVM safe for multithreading.
-void llvm_start_multithreaded();
+/// make LLVM safe for multithreading.  The return value indicates whether
+/// multithreaded initialization succeeded.  LLVM will still be operational
+/// on "failed" return, but will not be safe to run multithreaded.
+bool llvm_start_multithreaded();
 
 /// llvm_shutdown - Deallocate and destroy all ManagedStatic variables.
 void llvm_shutdown();

Modified: llvm/trunk/lib/Support/ManagedStatic.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Support/ManagedStatic.cpp?rev=72182&r1=72181&r2=72182&view=diff

==============================================================================
--- llvm/trunk/lib/Support/ManagedStatic.cpp (original)
+++ llvm/trunk/lib/Support/ManagedStatic.cpp Wed May 20 16:03:06 2009
@@ -68,12 +68,13 @@
   DeleterFn = 0;
 }
 
-void llvm::llvm_start_multithreaded() {
+bool llvm::llvm_start_multithreaded() {
 #if LLVM_MULTITHREADED
   assert(ManagedStaticMutex == 0 && "Multithreaded LLVM already initialized!");
   ManagedStaticMutex = new sys::Mutex(true);
+  return true;
 #else
-  assert(0 && "LLVM built without multithreading support!");
+  return false;
 #endif
 }
 




From gohman at apple.com  Wed May 20 16:45:42 2009
From: gohman at apple.com (Dan Gohman)
Date: Wed, 20 May 2009 21:45:42 -0000
Subject: [llvm-commits] [llvm] r72184 -
	/llvm/trunk/include/llvm/Support/IRBuilder.h
Message-ID: <200905202145.n4KLjgSZ025574@zion.cs.uiuc.edu>

Author: djg
Date: Wed May 20 16:45:41 2009
New Revision: 72184

URL: http://llvm.org/viewvc/llvm-project?rev=72184&view=rev
Log:
Add an accessor method to return the insertion point.

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

Modified: llvm/trunk/include/llvm/Support/IRBuilder.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Support/IRBuilder.h?rev=72184&r1=72183&r2=72184&view=diff

==============================================================================
--- llvm/trunk/include/llvm/Support/IRBuilder.h (original)
+++ llvm/trunk/include/llvm/Support/IRBuilder.h Wed May 20 16:45:41 2009
@@ -67,6 +67,8 @@
 
   BasicBlock *GetInsertBlock() const { return BB; }
 
+  BasicBlock::iterator GetInsertPoint() const { return InsertPt; }
+
   /// SetInsertPoint - This specifies that created instructions should be
   /// appended to the end of the specified block.
   void SetInsertPoint(BasicBlock *TheBB) {




From kremenek at apple.com  Wed May 20 17:41:48 2009
From: kremenek at apple.com (Ted Kremenek)
Date: Wed, 20 May 2009 22:41:48 -0000
Subject: [llvm-commits] [llvm] r72188 - /llvm/tags/checker/checker-0.210/
Message-ID: <200905202241.n4KMfmni027628@zion.cs.uiuc.edu>

Author: kremenek
Date: Wed May 20 17:41:47 2009
New Revision: 72188

URL: http://llvm.org/viewvc/llvm-project?rev=72188&view=rev
Log:
Tagging checker-0.210.

Added:
    llvm/tags/checker/checker-0.210/
      - copied from r72187, llvm/trunk/



From akyrtzi at gmail.com  Wed May 20 17:57:17 2009
From: akyrtzi at gmail.com (Argiris Kirtzidis)
Date: Wed, 20 May 2009 22:57:17 -0000
Subject: [llvm-commits] [llvm] r72191 - in /llvm/trunk:
 include/llvm/CodeGen/DebugLoc.h include/llvm/CodeGen/FastISel.h
 include/llvm/CodeGen/MachineFunction.h
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/DebugLoc.cpp
 lib/CodeGen/MachineFunction.cpp lib/CodeGen/SelectionDAG/FastISel.cpp
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
Message-ID: <200905202257.n4KMvHIK028254@zion.cs.uiuc.edu>

Author: akirtzidis
Date: Wed May 20 17:57:17 2009
New Revision: 72191

URL: http://llvm.org/viewvc/llvm-project?rev=72191&view=rev
Log:
Introduce DebugScope which gets embedded into the machine instructions' DebugLoc.
DebugScope refers to a debug region, function or block.

Added:
    llvm/trunk/lib/CodeGen/DebugLoc.cpp
Modified:
    llvm/trunk/include/llvm/CodeGen/DebugLoc.h
    llvm/trunk/include/llvm/CodeGen/FastISel.h
    llvm/trunk/include/llvm/CodeGen/MachineFunction.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/MachineFunction.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h

Modified: llvm/trunk/include/llvm/CodeGen/DebugLoc.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DebugLoc.h?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/DebugLoc.h (original)
+++ llvm/trunk/include/llvm/CodeGen/DebugLoc.h Wed May 20 17:57:17 2009
@@ -20,18 +20,77 @@
 
 namespace llvm {
   class GlobalVariable;
+  class MachineFunction;
 
-  /// DebugLocTuple - Debug location tuple of filename id, line and column.
+  /// DebugScope - Debug scope id. This is carried into DebugLocTuple to index
+  /// into a vector of DebugScopeInfos. Provides the ability to associate a
+  /// SDNode/MachineInstr with the debug scope that it belongs to.
+  ///
+  class DebugScope {
+    unsigned Idx;
+
+  public:
+    DebugScope() : Idx(~0U) {}  // Defaults to null.
+
+    static DebugScope getInvalid()   { DebugScope S; S.Idx = ~0U; return S; }
+    static DebugScope get(unsigned idx) { DebugScope S; S.Idx = idx; return S; }
+
+    unsigned getIndex() const { return Idx; }
+
+    /// isInvalid - Return true if it doesn't refer to any scope.
+    bool isInvalid() const { return Idx == ~0U; }
+
+    bool operator==(const DebugScope &DS) const { return Idx == DS.Idx; }
+    bool operator!=(const DebugScope &DS) const { return !(*this == DS); }
+  };
+
+  /// DebugScopeInfo - Contains info about the scope global variable and the
+  /// parent debug scope. DebugScope is only a "cookie" that can point to a
+  /// specific DebugScopeInfo.
+  ///
+  struct DebugScopeInfo {
+    GlobalVariable *GV;
+    DebugScope Parent;
+
+    DebugScopeInfo(GlobalVariable *gv, DebugScope parent)
+      : GV(gv), Parent(parent) {}
+    DebugScopeInfo()
+      : GV(0), Parent(DebugScope::getInvalid()) {}
+  };
+
+  /// DebugScopeTracker - Create and keep track of the debug scope while
+  /// entering/exiting subprograms and blocks. Used by the instruction
+  /// selectors.
+  ///
+  class DebugScopeTracker {
+    DebugScope CurScope;
+
+  public:
+    /// getCurScope - The current debug scope that we "entered" through
+    /// EnterDebugScope.
+    DebugScope getCurScope() const { return CurScope; }
+
+    /// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
+    /// or a DIBlock.
+    void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
+
+    /// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
+    void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
+  };
+
+  /// DebugLocTuple - Debug location tuple of a DICompileUnit global variable,
+  /// debug scope, line and column.
   ///
   struct DebugLocTuple {
     GlobalVariable *CompileUnit;
+    DebugScope Scope;
     unsigned Line, Col;
 
-    DebugLocTuple(GlobalVariable *v, unsigned l, unsigned c)
-      : CompileUnit(v), Line(l), Col(c) {};
+    DebugLocTuple(GlobalVariable *v, DebugScope s, unsigned l, unsigned c)
+      : CompileUnit(v), Scope(s), Line(l), Col(c) {};
 
     bool operator==(const DebugLocTuple &DLT) const {
-      return CompileUnit == DLT.CompileUnit &&
+      return CompileUnit == DLT.CompileUnit && Scope == DLT.Scope &&
              Line == DLT.Line && Col == DLT.Col;
     }
     bool operator!=(const DebugLocTuple &DLT) const {
@@ -63,18 +122,20 @@
   // Partially specialize DenseMapInfo for DebugLocTyple.
   template<>  struct DenseMapInfo<DebugLocTuple> {
     static inline DebugLocTuple getEmptyKey() {
-      return DebugLocTuple(0, ~0U, ~0U);
+      return DebugLocTuple(0, DebugScope::getInvalid(), ~0U, ~0U);
     }
     static inline DebugLocTuple getTombstoneKey() {
-      return DebugLocTuple((GlobalVariable*)~1U, ~1U, ~1U);
+      return DebugLocTuple((GlobalVariable*)~1U,DebugScope::get(~1U), ~1U, ~1U);
     }
     static unsigned getHashValue(const DebugLocTuple &Val) {
       return DenseMapInfo<GlobalVariable*>::getHashValue(Val.CompileUnit) ^
+             DenseMapInfo<unsigned>::getHashValue(Val.Scope.getIndex()) ^
              DenseMapInfo<unsigned>::getHashValue(Val.Line) ^
              DenseMapInfo<unsigned>::getHashValue(Val.Col);
     }
     static bool isEqual(const DebugLocTuple &LHS, const DebugLocTuple &RHS) {
       return LHS.CompileUnit == RHS.CompileUnit &&
+             LHS.Scope       == RHS.Scope &&
              LHS.Line        == RHS.Line &&
              LHS.Col         == RHS.Col;
     }

Modified: llvm/trunk/include/llvm/CodeGen/FastISel.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/FastISel.h?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/FastISel.h (original)
+++ llvm/trunk/include/llvm/CodeGen/FastISel.h Wed May 20 17:57:17 2009
@@ -57,6 +57,7 @@
   MachineFrameInfo &MFI;
   MachineConstantPool &MCP;
   DebugLoc DL;
+  DebugScopeTracker DbgScopeTrack;
   const TargetMachine &TM;
   const TargetData &TD;
   const TargetInstrInfo &TII;

Modified: llvm/trunk/include/llvm/CodeGen/MachineFunction.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineFunction.h?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineFunction.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineFunction.h Wed May 20 17:57:17 2009
@@ -111,6 +111,8 @@
   // Tracks debug locations.
   DebugLocTracker DebugLocInfo;
 
+  std::vector<DebugScopeInfo> DbgScopeInfos;
+
 public:
   MachineFunction(const Function *Fn, const TargetMachine &TM);
   ~MachineFunction();
@@ -332,7 +334,7 @@
   /// getOrCreateDebugLocID - Look up the DebugLocTuple index with the given
   /// source file, line, and column. If none currently exists, create a new
   /// DebugLocTuple, and insert it into the DebugIdMap.
-  unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit,
+  unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit, DebugScope Scope,
                                  unsigned Line, unsigned Col);
 
   /// getDebugLocTuple - Get the DebugLocTuple for a given DebugLoc object.
@@ -345,6 +347,12 @@
   /// setDefaultDebugLoc - Get the default debug location for the machine
   /// function.
   void setDefaultDebugLoc(DebugLoc DL) { DefaultDebugLoc = DL; }
+
+  /// CreateDebugScope - Create a new debug scope.
+  DebugScope CreateDebugScope(GlobalVariable *ScopeGV, DebugScope Parent);
+
+  /// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
+  const DebugScopeInfo &getDebugScopeInfo(DebugScope DS) const;
 };
 
 //===--------------------------------------------------------------------===//

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Wed May 20 17:57:17 2009
@@ -1311,7 +1311,7 @@
 void AsmPrinter::processDebugLoc(DebugLoc DL) {
   if (TAI->doesSupportDebugInformation() && DW->ShouldEmitDwarfDebug()) {
     if (!DL.isUnknown()) {
-      static DebugLocTuple PrevDLT(0, ~0U, ~0U);
+      static DebugLocTuple PrevDLT(0, DebugScope::getInvalid(), ~0U, ~0U);
       DebugLocTuple CurDLT = MF->getDebugLocTuple(DL);
 
       if (CurDLT.CompileUnit != 0 && PrevDLT != CurDLT)

Added: llvm/trunk/lib/CodeGen/DebugLoc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DebugLoc.cpp?rev=72191&view=auto

==============================================================================
--- llvm/trunk/lib/CodeGen/DebugLoc.cpp (added)
+++ llvm/trunk/lib/CodeGen/DebugLoc.cpp Wed May 20 17:57:17 2009
@@ -0,0 +1,38 @@
+//===-- DebugLoc.cpp ------------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Implementation for DebugScopeTracker.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/CodeGen/DebugLoc.h"
+#include "llvm/CodeGen/MachineFunction.h"
+using namespace llvm;
+
+/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
+/// or a DIBlock.
+void DebugScopeTracker::EnterDebugScope(GlobalVariable *ScopeGV,
+                                        MachineFunction &MF) {
+  assert(ScopeGV && "GlobalVariable for scope is null!");
+  CurScope = MF.CreateDebugScope(ScopeGV, CurScope);
+}
+
+/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
+void DebugScopeTracker::ExitDebugScope(GlobalVariable *ScopeGV,
+                                       MachineFunction &MF) {
+  assert(ScopeGV && "GlobalVariable for scope is null!");
+  assert(!CurScope.isInvalid() && "Mismatched region.end ?");
+  // We may have skipped a region.end because it was in an unreachable block.
+  // Go up the scope chain until we reach the scope that ScopeGV points to.
+  DebugScopeInfo DSI;
+  do {
+    DSI =  MF.getDebugScopeInfo(CurScope);
+    CurScope = DSI.Parent;
+  } while (!DSI.Parent.isInvalid() && DSI.GV != ScopeGV);
+}

Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Wed May 20 17:57:17 2009
@@ -403,8 +403,9 @@
 /// source file, line, and column. If none currently exists, create a new
 /// DebugLocTuple, and insert it into the DebugIdMap.
 unsigned MachineFunction::getOrCreateDebugLocID(GlobalVariable *CompileUnit,
+                                                DebugScope Scope,
                                                 unsigned Line, unsigned Col) {
-  DebugLocTuple Tuple(CompileUnit, Line, Col);
+  DebugLocTuple Tuple(CompileUnit, Scope, Line, Col);
   DenseMap<DebugLocTuple, unsigned>::iterator II
     = DebugLocInfo.DebugIdMap.find(Tuple);
   if (II != DebugLocInfo.DebugIdMap.end())
@@ -424,6 +425,21 @@
   return DebugLocInfo.DebugLocations[Idx];
 }
 
+/// CreateDebugScope - Create a new debug scope.
+DebugScope MachineFunction::CreateDebugScope(GlobalVariable *ScopeGV,
+                                             DebugScope Parent) {
+  DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
+  return DebugScope::get(DbgScopeInfos.size() - 1);
+}
+
+/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
+const DebugScopeInfo &MachineFunction::getDebugScopeInfo(DebugScope DS) const {
+  unsigned Idx = DS.getIndex();
+  assert(Idx < DbgScopeInfos.size() && "Invalid index into debug scopes!");
+  return DbgScopeInfos[Idx];
+}
+
+
 //===----------------------------------------------------------------------===//
 //  MachineFrameInfo implementation
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Wed May 20 17:57:17 2009
@@ -331,17 +331,22 @@
       DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
       unsigned Line = SPI->getLine();
       unsigned Col = SPI->getColumn();
-      unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(), Line, Col);
+      unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(),
+                                              DbgScopeTrack.getCurScope(),
+                                              Line, Col);
       setCurDebugLoc(DebugLoc::get(Idx));
     }
     return true;
   }
   case Intrinsic::dbg_region_start: {
     DbgRegionStartInst *RSI = cast<DbgRegionStartInst>(I);
-    if (DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None) &&
-        DW && DW->ShouldEmitDwarfDebug()) {
-      unsigned ID = 
-        DW->RecordRegionStart(cast<GlobalVariable>(RSI->getContext()));
+    if (!DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None))
+      return true;
+
+    GlobalVariable *Rgn = cast<GlobalVariable>(RSI->getContext());
+    DbgScopeTrack.EnterDebugScope(Rgn, MF);
+    if (DW && DW->ShouldEmitDwarfDebug()) {
+      unsigned ID = DW->RecordRegionStart(Rgn);
       const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
       BuildMI(MBB, DL, II).addImm(ID);
     }
@@ -349,10 +354,14 @@
   }
   case Intrinsic::dbg_region_end: {
     DbgRegionEndInst *REI = cast<DbgRegionEndInst>(I);
-    if (DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None) &&
-        DW && DW->ShouldEmitDwarfDebug()) {
+    if (!DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None))
+      return true;
+
+    GlobalVariable *Rgn = cast<GlobalVariable>(REI->getContext());
+    DbgScopeTrack.ExitDebugScope(Rgn, MF);
+    if (DW && DW->ShouldEmitDwarfDebug()) {
      unsigned ID = 0;
-     DISubprogram Subprogram(cast<GlobalVariable>(REI->getContext()));
+     DISubprogram Subprogram(Rgn);
      if (!Subprogram.isNull() && !Subprogram.describes(MF.getFunction())) {
         // This is end of an inlined function.
         const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
@@ -382,6 +391,7 @@
     DebugLoc PrevLoc = DL;
     DISubprogram Subprogram(cast<GlobalVariable>(SP));
     DICompileUnit CompileUnit = Subprogram.getCompileUnit();
+    DbgScopeTrack.EnterDebugScope(Subprogram.getGV(), MF);
 
     if (!Subprogram.describes(MF.getFunction())) {
       // This is a beginning of an inlined function.
@@ -393,8 +403,10 @@
         return true;
       // Record the source line.
       unsigned Line = Subprogram.getLineNumber();
-      setCurDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
-                                              CompileUnit.getGV(), Line, 0)));
+      setCurDebugLoc(
+        DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
+                                               DbgScopeTrack.getCurScope(),
+                                               Line, 0)));
 
       if (DW && DW->ShouldEmitDwarfDebug()) {
         DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
@@ -408,8 +420,10 @@
     } else {
       // Record the source line.
       unsigned Line = Subprogram.getLineNumber();
-      MF.setDefaultDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
-                                              CompileUnit.getGV(), Line, 0)));
+      MF.setDefaultDebugLoc(
+        DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
+                                               DbgScopeTrack.getCurScope(),
+                                               Line, 0)));
       if (DW && DW->ShouldEmitDwarfDebug()) {
         // llvm.dbg.func_start also defines beginning of function scope.
         DW->RecordRegionStart(cast<GlobalVariable>(FSI->getSubprogram()));

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Wed May 20 17:57:17 2009
@@ -312,6 +312,17 @@
             !StaticAllocaMap.count(cast<AllocaInst>(I)))
           InitializeRegForValue(I);
 
+  // FIXME: PHI instructions currently get an invalid scope. This is because
+  // tracking of debug scopes is done through a simple mechanism where
+  // "entering" a scope implies that the scope is entered for the first time.
+  // If we keep track of debug scopes for the following loop, the PHI
+  // instructions would get scopes that will not be used again later by the
+  // instruction selectors.
+  // Either provide a mechanism to re-enter a previously created scope or wait
+  // for when the DebugLoc is retrieved from Instruction, in which case the
+  // current debug scope tracking mechanism will be obsolete.
+  DebugScope DbgScope;
+
   // Create an initial MachineBasicBlock for each LLVM BasicBlock in F.  This
   // also creates the initial PHI MachineInstrs, though none of the input
   // operands are populated.
@@ -337,6 +348,7 @@
                                              CodeGenOpt::Default)) {
               DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
               unsigned idx = MF->getOrCreateDebugLocID(CU.getGV(),
+                                                       DbgScope,
                                                        SPI->getLine(),
                                                        SPI->getColumn());
               DL = DebugLoc::get(idx);
@@ -352,7 +364,7 @@
               DISubprogram Subprogram(cast<GlobalVariable>(SP));
               DICompileUnit CU(Subprogram.getCompileUnit());
               unsigned Line = Subprogram.getLineNumber();
-              DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(),
+              DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(), DbgScope,
                                                            Line, 0));
             }
 
@@ -3909,7 +3921,9 @@
       MachineFunction &MF = DAG.getMachineFunction();
       DICompileUnit CU(cast<GlobalVariable>(SPI.getContext()));
       DebugLoc Loc = DebugLoc::get(MF.getOrCreateDebugLocID(CU.getGV(),
-                                              SPI.getLine(), SPI.getColumn()));
+                                                 DbgScopeTrack.getCurScope(),
+                                                           SPI.getLine(),
+                                                           SPI.getColumn()));
       setCurDebugLoc(Loc);
       
       if (OptLevel == CodeGenOpt::None)
@@ -3923,11 +3937,14 @@
   case Intrinsic::dbg_region_start: {
     DwarfWriter *DW = DAG.getDwarfWriter();
     DbgRegionStartInst &RSI = cast<DbgRegionStartInst>(I);
+    if (!DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel))
+      return 0;
 
-    if (DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel) &&
-        DW && DW->ShouldEmitDwarfDebug()) {
-      unsigned LabelID =
-        DW->RecordRegionStart(cast<GlobalVariable>(RSI.getContext()));
+    MachineFunction &MF = DAG.getMachineFunction();
+    GlobalVariable *Rgn = cast<GlobalVariable>(RSI.getContext());
+    DbgScopeTrack.EnterDebugScope(Rgn, MF);
+    if (DW && DW->ShouldEmitDwarfDebug()) {
+      unsigned LabelID = DW->RecordRegionStart(Rgn);
       DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),
                                getRoot(), LabelID));
     }
@@ -3937,11 +3954,14 @@
   case Intrinsic::dbg_region_end: {
     DwarfWriter *DW = DAG.getDwarfWriter();
     DbgRegionEndInst &REI = cast<DbgRegionEndInst>(I);
+    if (!DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel))
+      return 0;
 
-    if (DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel) &&
-        DW && DW->ShouldEmitDwarfDebug()) {
-      MachineFunction &MF = DAG.getMachineFunction();
-      DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
+    MachineFunction &MF = DAG.getMachineFunction();
+    GlobalVariable *Rgn = cast<GlobalVariable>(REI.getContext());
+    DbgScopeTrack.ExitDebugScope(Rgn, MF);
+    if (DW && DW->ShouldEmitDwarfDebug()) {
+      DISubprogram Subprogram(Rgn);
 
       if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
         unsigned LabelID =
@@ -3974,6 +3994,7 @@
       return 0;
 
     MachineFunction &MF = DAG.getMachineFunction();
+    DbgScopeTrack.EnterDebugScope(cast<GlobalVariable>(SP), MF);
     if (OptLevel == CodeGenOpt::None) {
       // llvm.dbg.func.start implicitly defines a dbg_stoppoint which is what
       // (most?) gdb expects.
@@ -3993,7 +4014,9 @@
         // Record the source line.
         unsigned Line = Subprogram.getLineNumber();
         setCurDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
+                                              DbgScopeTrack.getCurScope(),
+                                              Line, 0)));
 
         if (DW && DW->ShouldEmitDwarfDebug()) {
           DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
@@ -4008,7 +4031,9 @@
         // Record the source line.
         unsigned Line = Subprogram.getLineNumber();
         MF.setDefaultDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
+                                              DbgScopeTrack.getCurScope(),
+                                              Line, 0)));
         if (DW && DW->ShouldEmitDwarfDebug()) {
           // llvm.dbg.func_start also defines beginning of function scope.
           DW->RecordRegionStart(cast<GlobalVariable>(FSI.getSubprogram()));
@@ -4035,7 +4060,9 @@
       // create a label if this is a beginning of inlined function.
       unsigned Line = Subprogram.getLineNumber();
       setCurDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
+                                              DbgScopeTrack.getCurScope(),
+                                              Line, 0)));
       // FIXME -  Start new region because llvm.dbg.func_start also defines
       // beginning of function scope.
     }

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h?rev=72191&r1=72190&r2=72191&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h Wed May 20 17:57:17 2009
@@ -169,6 +169,8 @@
   /// CurDebugLoc - current file + line number.  Changes as we build the DAG.
   DebugLoc CurDebugLoc;
 
+  DebugScopeTracker DbgScopeTrack;
+
   DenseMap<const Value*, SDValue> NodeMap;
 
   /// PendingLoads - Loads are not emitted to the program immediately.  We bunch




From isanbard at gmail.com  Wed May 20 18:04:56 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:04:56 -0000
Subject: [llvm-commits] [llvm] r72192 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <200905202304.n4KN4u7n028487@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:04:56 2009
New Revision: 72192

URL: http://llvm.org/viewvc/llvm-project?rev=72192&view=rev
Log:
Do some mechanical changes. Combine the 'construct abastract dbg thingy' in with
the 'constract function dbg thingy'. Rename some methods to make them consistent
with the rest of the methods. Move the 'Emit' methods to the end of the file.

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=72192&r1=72191&r2=72192&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:04:56 2009
@@ -275,9 +275,9 @@
   }
 }
 
-/// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
-/// entry.
-DIEEntry *DwarfDebug::NewDIEEntry(DIE *Entry) {
+/// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
+/// information entry.
+DIEEntry *DwarfDebug::CreateDIEEntry(DIE *Entry) {
   DIEEntry *Value;
 
   if (Entry) {
@@ -545,7 +545,7 @@
   }
 
   // Set up proxy.
-  Slot = NewDIEEntry();
+  Slot = CreateDIEEntry();
 
   // Construct type.
   DIE Buffer(dwarf::DW_TAG_base_type);
@@ -925,9 +925,9 @@
   return *I->second;
 }
 
-/// NewDbgScopeVariable - Create a new scope variable.
+/// CreateDbgScopeVariable - Create a new scope variable.
 ///
-DIE *DwarfDebug::NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
+DIE *DwarfDebug::CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
   // Get the descriptor.
   const DIVariable &VD = DV->getVariable();
 
@@ -1012,7 +1012,7 @@
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = ParentScope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
-    DIE *VariableDie = NewDbgScopeVariable(Variables[i], Unit);
+    DIE *VariableDie = CreateDbgScopeVariable(Variables[i], Unit);
     if (VariableDie) ParentDie->AddChild(VariableDie);
   }
 
@@ -1092,7 +1092,8 @@
 
 /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
 ///
-void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope) {
+void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope,
+                                           bool AbstractScope) {
   // Exit if there is no root scope.
   if (!RootScope) return;
   DIDescriptor Desc = RootScope->getDesc();
@@ -1111,43 +1112,19 @@
   DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
   assert(SPDie && "Missing subprogram descriptor");
 
-  // Add the function bounds.
-  AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           DWLabel("func_begin", SubprogramCount));
-  AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           DWLabel("func_end", SubprogramCount));
-  MachineLocation Location(RI->getFrameRegister(*MF));
-  AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
+  if (!AbstractScope) {
+    // Add the function bounds.
+    AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
+             DWLabel("func_begin", SubprogramCount));
+    AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
+             DWLabel("func_end", SubprogramCount));
+    MachineLocation Location(RI->getFrameRegister(*MF));
+    AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
+  }
 
   ConstructDbgScope(RootScope, 0, 0, SPDie, Unit);
 }
 
-/// ConstructFunctionDbgScope - Construct the scope for the abstract debug
-/// scope.
-///
-void DwarfDebug::ConstructAbstractDbgScope(DbgScope *AbsScope) {
-  // Exit if there is no root scope.
-  if (!AbsScope) return;
-
-  DIDescriptor Desc = AbsScope->getDesc();
-  if (Desc.isNull())
-    return;
-
-  // Get the subprogram debug information entry.
-  DISubprogram SPD(Desc.getGV());
-
-  // Get the compile unit context.
-  CompileUnit *Unit = MainCU;
-  if (!Unit)
-    Unit = &FindCompileUnit(SPD.getCompileUnit());
-
-  // Get the subprogram die.
-  DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
-  assert(SPDie && "Missing subprogram descriptor");
-
-  ConstructDbgScope(AbsScope, 0, 0, SPDie, Unit);
-}
-
 /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
 ///
 void DwarfDebug::ConstructDefaultDbgScope(MachineFunction *MF) {
@@ -1196,1445 +1173,1451 @@
 #endif
 }
 
-/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
-/// tools to recognize the object file contains Dwarf information.
-void DwarfDebug::EmitInitial() {
-  // Check to see if we already emitted intial headers.
-  if (didInitial) return;
-  didInitial = true;
-
-  // Dwarf sections base addresses.
-  if (TAI->doesDwarfRequireFrameSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
-    EmitLabel("section_debug_frame", 0);
+/// GetOrCreateSourceID - Look up the source id with the given directory and
+/// source file names. If none currently exists, create a new id and insert it
+/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
+/// maps as well.
+unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  unsigned DId;
+  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
+  if (DI != DirectoryIdMap.end()) {
+    DId = DI->getValue();
+  } else {
+    DId = DirectoryNames.size() + 1;
+    DirectoryIdMap[DirName] = DId;
+    DirectoryNames.push_back(DirName);
   }
 
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
-  EmitLabel("section_info", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
-  EmitLabel("section_abbrev", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
-  EmitLabel("section_aranges", 0);
-
-  if (TAI->doesSupportMacInfoSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    EmitLabel("section_macinfo", 0);
+  unsigned FId;
+  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
+  if (FI != SourceFileIdMap.end()) {
+    FId = FI->getValue();
+  } else {
+    FId = SourceFileNames.size() + 1;
+    SourceFileIdMap[FileName] = FId;
+    SourceFileNames.push_back(FileName);
   }
 
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
-  EmitLabel("section_line", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  EmitLabel("section_loc", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
-  EmitLabel("section_pubnames", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
-  EmitLabel("section_str", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  EmitLabel("section_ranges", 0);
-
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_begin", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_begin", 0);
-}
+  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
+    SourceIdMap.find(std::make_pair(DId, FId));
+  if (SI != SourceIdMap.end())
+    return SI->second;
 
-/// EmitDIE - Recusively Emits a debug information entry.
-///
-void DwarfDebug::EmitDIE(DIE *Die) {
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
+  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
+  SourceIds.push_back(std::make_pair(DId, FId));
 
-  Asm->EOL();
+  return SrcId;
+}
 
-  // Emit the code (index) for the abbreviation.
-  Asm->EmitULEB128Bytes(AbbrevNumber);
+void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
+  DICompileUnit DIUnit(GV);
+  std::string Dir, FN, Prod;
+  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
+                                    DIUnit.getFilename(FN));
 
-  if (Asm->isVerbose())
-    Asm->EOL(std::string("Abbrev [" +
-                         utostr(AbbrevNumber) +
-                         "] 0x" + utohexstr(Die->getOffset()) +
-                         ":0x" + utohexstr(Die->getSize()) + " " +
-                         dwarf::TagString(Abbrev->getTag())));
-  else
-    Asm->EOL();
+  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
+  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
+                   DWLabel("section_line", 0), DWLabel("section_line", 0),
+                   false);
+  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
+            DIUnit.getProducer(Prod));
+  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
+          DIUnit.getLanguage());
+  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
 
-  SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+  if (!Dir.empty())
+    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
+  if (DIUnit.isOptimized())
+    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
 
-  // Emit the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
-    unsigned Attr = AbbrevData[i].getAttribute();
-    unsigned Form = AbbrevData[i].getForm();
-    assert(Form && "Too many attributes for DIE (check abbreviation)");
+  std::string Flags;
+  DIUnit.getFlags(Flags);
+  if (!Flags.empty())
+    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
 
-    switch (Attr) {
-    case dwarf::DW_AT_sibling:
-      Asm->EmitInt32(Die->SiblingOffset());
-      break;
-    case dwarf::DW_AT_abstract_origin: {
-      DIEEntry *E = cast<DIEEntry>(Values[i]);
-      DIE *Origin = E->getEntry();
-      unsigned Addr =
-        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
-        Origin->getOffset();
+  unsigned RVer = DIUnit.getRunTimeVersion();
+  if (RVer)
+    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
+            dwarf::DW_FORM_data1, RVer);
 
-      Asm->EmitInt32(Addr);
-      break;
-    }
-    default:
-      // Emit an attribute using the defined form.
-      Values[i]->EmitValue(this, Form);
-      break;
+  CompileUnit *Unit = new CompileUnit(ID, Die);
+  if (DIUnit.isMain()) {
+    assert(!MainCU && "Multiple main compile units are found!");
+    MainCU = Unit;
     }
 
-    Asm->EOL(dwarf::AttributeString(Attr));
-  }
-
-  // Emit the DIE children if any.
-  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
-    const std::vector<DIE *> &Children = Die->getChildren();
+  CompileUnitMap[DIUnit.getGV()] = Unit;
+  CompileUnits.push_back(Unit);
+}
 
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      EmitDIE(Children[j]);
+/// ConstructCompileUnits - Create a compile unit DIEs.
+void DwarfDebug::ConstructCompileUnits() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
+  if (!Root)
+    return;
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed compile unit descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed compile unit descriptor anchor type");
 
-    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
-  }
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI) {
+      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
+      ConstructCompileUnit(GV);
+    }
 }
 
-/// SizeAndOffsetDie - Compute the size and offset of a DIE.
-///
-unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
-  // Get the children.
-  const std::vector<DIE *> &Children = Die->getChildren();
+bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
+  DIGlobalVariable DI_GV(GV);
+  CompileUnit *DW_Unit = MainCU;
+  if (!DW_Unit)
+    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
 
-  // If not last sibling and has children then add sibling offset attribute.
-  if (!Last && !Children.empty()) Die->AddSiblingOffset();
+  // Check for pre-existence.
+  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
+  if (Slot)
+    return false;
 
-  // Record the abbreviation.
-  AssignAbbrevNumber(Die->getAbbrev());
+  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
 
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+  // Add address.
+  DIEBlock *Block = new DIEBlock();
+  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
+  std::string GLN;
+  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
+                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
+  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
 
-  // Set DIE offset
-  Die->setOffset(Offset);
+  // Add to map.
+  Slot = VariableDie;
 
-  // Start the size with the size of abbreviation code.
-  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
+  // Add to context owner.
+  DW_Unit->getDie()->AddChild(VariableDie);
 
-  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+  // Expose as global. FIXME - need to check external flag.
+  std::string Name;
+  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
+  return true;
+}
 
-  // Size the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i)
-    // Size attribute value.
-    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
+/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
+/// global variables. Return true if at least one global DIE is created.
+bool DwarfDebug::ConstructGlobalVariableDIEs() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
+  if (!Root)
+    return false;
 
-  // Size the DIE children if any.
-  if (!Children.empty()) {
-    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
-           "Children flag not set");
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed global variable descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed global variable descriptor anchor type");
 
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
-
-    // End of children marker.
-    Offset += sizeof(int8_t);
-  }
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
 
-  Die->setSize(Offset - Die->getOffset());
-  return Offset;
+  return Result;
 }
 
-/// SizeAndOffsets - Compute the size and offset of all the DIEs.
-///
-void DwarfDebug::SizeAndOffsets() {
-  // Compute size of compile unit header.
-  static unsigned Offset =
-    sizeof(int32_t) + // Length of Compilation Unit Info
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t);   // Pointer Size (in bytes)
+bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
+  DISubprogram SP(GV);
+  CompileUnit *Unit = MainCU;
+  if (!Unit)
+    Unit = &FindCompileUnit(SP.getCompileUnit());
 
-  // Process base compile unit.
-  if (MainCU) {
-    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
-    CompileUnitOffsets[MainCU] = 0;
-    return;
-  }
+  // Check for pre-existence.
+  DIE *&Slot = Unit->getDieMapSlotFor(GV);
+  if (Slot)
+    return false;
 
-  // Process all compile units.
-  unsigned PrevOffset = 0;
+  if (!SP.isDefinition())
+    // This is a method declaration which will be handled while constructing
+    // class type.
+    return false;
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
-    CompileUnit *Unit = CompileUnits[i];
-    CompileUnitOffsets[Unit] = PrevOffset;
-    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
-      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
-  }
-}
+  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
 
-/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
-///
-void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
-  DIE *Die = Unit->getDie();
+  // Add to map.
+  Slot = SubprogramDie;
 
-  // Emit the compile units header.
-  EmitLabel("info_begin", Unit->getID());
+  // Add to context owner.
+  Unit->getDie()->AddChild(SubprogramDie);
 
-  // Emit size of content not including length itself
-  unsigned ContentSize = Die->getSize() +
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t) +  // Pointer Size (in bytes)
-    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+  // Expose as global.
+  std::string Name;
+  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
+  return true;
+}
 
-  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
-  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
-  Asm->EOL("Offset Into Abbrev. Section");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+/// ConstructSubprograms - Create DIEs for each of the externally visible
+/// subprograms. Return true if at least one subprogram DIE is created.
+bool DwarfDebug::ConstructSubprograms() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
+  if (!Root)
+    return false;
 
-  EmitDIE(Die);
-  // FIXME - extra padding for gdb bug.
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  EmitLabel("info_end", Unit->getID());
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed subprogram descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed subprogram descriptor anchor type");
 
-  Asm->EOL();
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
+
+  return Result;
 }
 
-void DwarfDebug::EmitDebugInfo() {
-  // Start debug info section.
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
+/// This is inovked by the target AsmPrinter.
+void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
+
+  // Create all the compile unit DIEs.
+  ConstructCompileUnits();
+
+  if (CompileUnits.empty()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  if (MainCU) {
-    EmitDebugInfoPerCU(MainCU);
     return;
   }
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugInfoPerCU(CompileUnits[i]);
-}
+  // Create DIEs for each of the externally visible global variables.
+  bool globalDIEs = ConstructGlobalVariableDIEs();
 
-/// EmitAbbreviations - Emit the abbreviation section.
-///
-void DwarfDebug::EmitAbbreviations() const {
-  // Check to see if it is worth the effort.
-  if (!Abbreviations.empty()) {
-    // Start the debug abbrev section.
-    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
+  // Create DIEs for each of the externally visible subprograms.
+  bool subprogramDIEs = ConstructSubprograms();
 
-    EmitLabel("abbrev_begin", 0);
+  // If there is not any debug info available for any global variables and any
+  // subprograms then there is not any debug info to emit.
+  if (!globalDIEs && !subprogramDIEs) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-    // For each abbrevation.
-    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
-      // Get abbreviation data
-      const DIEAbbrev *Abbrev = Abbreviations[i];
+    return;
+  }
 
-      // Emit the abbrevations code (base 1 index.)
-      Asm->EmitULEB128Bytes(Abbrev->getNumber());
-      Asm->EOL("Abbreviation Code");
+  MMI = mmi;
+  shouldEmit = true;
+  MMI->setDebugInfoAvailability(true);
 
-      // Emit the abbreviations data.
-      Abbrev->Emit(Asm);
+  // Prime section data.
+  SectionMap.insert(TAI->getTextSection());
 
+  // Print out .file directives to specify files for .loc directives. These are
+  // printed out early so that they precede any .loc directives.
+  if (TAI->hasDotLocAndDotFile()) {
+    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
+      // Remember source id starts at 1.
+      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
+      sys::Path FullPath(getSourceDirectoryName(Id.first));
+      bool AppendOk =
+        FullPath.appendComponent(getSourceFileName(Id.second));
+      assert(AppendOk && "Could not append filename to directory!");
+      AppendOk = false;
+      Asm->EmitFile(i, FullPath.toString());
       Asm->EOL();
     }
-
-    // Mark end of abbreviations.
-    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
-
-    EmitLabel("abbrev_end", 0);
-    Asm->EOL();
   }
-}
 
-/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
-/// the line matrix.
-///
-void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
-  // Define last address of section.
-  Asm->EmitInt8(0); Asm->EOL("Extended Op");
-  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
+  // Emit initial sections
+  EmitInitial();
 
-  // Mark end of matrix.
-  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
-  Asm->EmitULEB128Bytes(1); Asm->EOL();
-  Asm->EmitInt8(1); Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitDebugLines - Emit source line information.
+/// EndModule - Emit all Dwarf sections that should come after the content.
 ///
-void DwarfDebug::EmitDebugLines() {
-  // If the target is using .loc/.file, the assembler will be emitting the
-  // .debug_line table automatically.
-  if (TAI->hasDotLocAndDotFile())
+void DwarfDebug::EndModule() {
+  if (!ShouldEmitDwarfDebug())
     return;
 
-  // Minimum line delta, thus ranging from -10..(255-10).
-  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
-  // Maximum line delta, thus ranging from -10..(255-10).
-  const int MaxLineDelta = 255 + MinLineDelta;
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  // Start the dwarf line section.
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
+  // Standard sections final addresses.
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_end", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_end", 0);
 
-  // Construct the section header.
-  EmitDifference("line_end", 0, "line_begin", 0, true);
-  Asm->EOL("Length of Source Line Info");
-  EmitLabel("line_begin", 0);
+  // End text sections.
+  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
+    Asm->SwitchToSection(SectionMap[i]);
+    EmitLabel("section_end", i);
+  }
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
+  // Emit common frame information.
+  EmitCommonDebugFrame();
 
-  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
-  Asm->EOL("Prolog Length");
-  EmitLabel("line_prolog_begin", 0);
+  // Emit function debug frame information
+  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
+         E = DebugFrames.end(); I != E; ++I)
+    EmitFunctionDebugFrame(*I);
 
-  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
+  // Compute DIE offsets and sizes.
+  SizeAndOffsets();
 
-  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
+  // Emit all the DIEs into a debug info section
+  EmitDebugInfo();
 
-  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
+  // Corresponding abbreviations into a abbrev section.
+  EmitAbbreviations();
 
-  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
+  // Emit source line correspondence into a debug line section.
+  EmitDebugLines();
 
-  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
+  // Emit info into a debug pubnames section.
+  EmitDebugPubNames();
 
-  // Line number standard opcode encodings argument count
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
-
-  // Emit directories.
-  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
-    Asm->EmitString(getSourceDirectoryName(DI));
-    Asm->EOL("Directory");
-  }
-
-  Asm->EmitInt8(0); Asm->EOL("End of directories");
-
-  // Emit files.
-  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
-    // Remember source id starts at 1.
-    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
-    Asm->EmitString(getSourceFileName(Id.second));
-    Asm->EOL("Source");
-    Asm->EmitULEB128Bytes(Id.first);
-    Asm->EOL("Directory #");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("Mod date");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("File size");
-  }
-
-  Asm->EmitInt8(0); Asm->EOL("End of files");
-
-  EmitLabel("line_prolog_end", 0);
+  // Emit info into a debug str section.
+  EmitDebugStr();
 
-  // A sequence for each text section.
-  unsigned SecSrcLinesSize = SectionSourceLines.size();
+  // Emit info into a debug loc section.
+  EmitDebugLoc();
 
-  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
-    // Isolate current sections line info.
-    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
+  // Emit info into a debug aranges section.
+  EmitDebugARanges();
 
-    if (Asm->isVerbose()) {
-      const Section* S = SectionMap[j + 1];
-      O << '\t' << TAI->getCommentString() << " Section"
-        << S->getName() << '\n';
-    } else {
-      Asm->EOL();
-    }
+  // Emit info into a debug ranges section.
+  EmitDebugRanges();
 
-    // Dwarf assumes we start with first line of first source file.
-    unsigned Source = 1;
-    unsigned Line = 1;
+  // Emit info into a debug macinfo section.
+  EmitDebugMacInfo();
 
-    // Construct rows of the address, source, line, column matrix.
-    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
-      const SrcLineInfo &LineInfo = LineInfos[i];
-      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
-      if (!LabelID) continue;
+  // Emit inline info.
+  EmitDebugInlineInfo();
 
-      if (!Asm->isVerbose())
-        Asm->EOL();
-      else {
-        std::pair<unsigned, unsigned> SourceID =
-          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
-        O << '\t' << TAI->getCommentString() << ' '
-          << getSourceDirectoryName(SourceID.first) << ' '
-          << getSourceFileName(SourceID.second)
-          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
-      }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+}
 
-      // Define the line address.
-      Asm->EmitInt8(0); Asm->EOL("Extended Op");
-      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-      EmitReference("label",  LabelID); Asm->EOL("Location label");
+/// BeginFunction - Gather pre-function debug information.  Assumes being
+/// emitted immediately after the function entry point.
+void DwarfDebug::BeginFunction(MachineFunction *MF) {
+  this->MF = MF;
 
-      // If change of source, then switch to the new source.
-      if (Source != LineInfo.getSourceID()) {
-        Source = LineInfo.getSourceID();
-        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
-        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
-      }
+  if (!ShouldEmitDwarfDebug()) return;
 
-      // If change of line.
-      if (Line != LineInfo.getLine()) {
-        // Determine offset.
-        int Offset = LineInfo.getLine() - Line;
-        int Delta = Offset - MinLineDelta;
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-        // Update line.
-        Line = LineInfo.getLine();
+  // Begin accumulating function debug information.
+  MMI->BeginFunction(MF);
 
-        // If delta is small enough and in range...
-        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
-          // ... then use fast opcode.
-          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
-        } else {
-          // ... otherwise use long hand.
-          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
-          Asm->EOL("DW_LNS_advance_line");
-          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
-          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-        }
-      } else {
-        // Copy the previous row (different address or source)
-        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-      }
-    }
+  // Assumes in correct section after the entry point.
+  EmitLabel("func_begin", ++SubprogramCount);
 
-    EmitEndOfLineMatrix(j + 1);
+  // Emit label for the implicitly defined dbg.stoppoint at the start of the
+  // function.
+  DebugLoc FDL = MF->getDefaultDebugLoc();
+  if (!FDL.isUnknown()) {
+    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
+    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
+                                        DICompileUnit(DLT.CompileUnit));
+    Asm->printLabel(LabelID);
   }
 
-  if (SecSrcLinesSize == 0)
-    // Because we're emitting a debug_line section, we still need a line
-    // table. The linker and friends expect it to exist. If there's nothing to
-    // put into it, emit an empty table.
-    EmitEndOfLineMatrix(1);
-
-  EmitLabel("line_end", 0);
-  Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
+/// EndFunction - Gather and emit post-function debug information.
 ///
-void DwarfDebug::EmitCommonDebugFrame() {
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
-
-  int stackGrowth =
-    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
-      TargetFrameInfo::StackGrowsUp ?
-    TD->getPointerSize() : -TD->getPointerSize();
-
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
-
-  EmitLabel("debug_frame_common", 0);
-  EmitDifference("debug_frame_common_end", 0,
-                 "debug_frame_common_begin", 0, true);
-  Asm->EOL("Length of Common Information Entry");
-
-  EmitLabel("debug_frame_common_begin", 0);
-  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
-  Asm->EOL("CIE Identifier Tag");
-  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
-  Asm->EOL("CIE Version");
-  Asm->EmitString("");
-  Asm->EOL("CIE Augmentation");
-  Asm->EmitULEB128Bytes(1);
-  Asm->EOL("CIE Code Alignment Factor");
-  Asm->EmitSLEB128Bytes(stackGrowth);
-  Asm->EOL("CIE Data Alignment Factor");
-  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
-  Asm->EOL("CIE RA Column");
-
-  std::vector<MachineMove> Moves;
-  RI->getInitialFrameState(Moves);
-
-  EmitFrameMoves(NULL, 0, Moves, false);
-
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_common_end", 0);
-
-  Asm->EOL();
-}
+void DwarfDebug::EndFunction(MachineFunction *MF) {
+  if (!ShouldEmitDwarfDebug()) return;
 
-/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
-/// section.
-void
-DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+  // Define end label for subprogram.
+  EmitLabel("func_end", SubprogramCount);
 
-  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
-                 "debug_frame_begin", DebugFrameInfo.Number, true);
-  Asm->EOL("Length of Frame Information Entry");
+  // Get function line info.
+  if (!Lines.empty()) {
+    // Get section line info.
+    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
+    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
+    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
+    // Append the function info to section info.
+    SectionLineInfos.insert(SectionLineInfos.end(),
+                            Lines.begin(), Lines.end());
+  }
 
-  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
+  // Construct the DbgScope for abstract instances.
+  for (SmallVector<DbgScope *, 32>::iterator
+         I = AbstractInstanceRootList.begin(),
+         E = AbstractInstanceRootList.end(); I != E; ++I)
+    ConstructFunctionDbgScope(*I, true);
 
-  EmitSectionOffset("debug_frame_common", "section_debug_frame",
-                    0, 0, true, false);
-  Asm->EOL("FDE CIE offset");
+  // Construct scopes for subprogram.
+  if (FunctionDbgScope)
+    ConstructFunctionDbgScope(FunctionDbgScope);
+  else
+    // FIXME: This is wrong. We are essentially getting past a problem with
+    // debug information not being able to handle unreachable blocks that have
+    // debug information in them. In particular, those unreachable blocks that
+    // have "region end" info in them. That situation results in the "root
+    // scope" not being created. If that's the case, then emit a "default"
+    // scope, i.e., one that encompasses the whole function. This isn't
+    // desirable. And a better way of handling this (and all of the debugging
+    // information) needs to be explored.
+    ConstructDefaultDbgScope(MF);
 
-  EmitReference("func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE initial location");
-  EmitDifference("func_end", DebugFrameInfo.Number,
-                 "func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE address range");
+  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
+                                               MMI->getFrameMoves()));
 
-  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
-                 false);
+  // Clear debug info
+  if (FunctionDbgScope) {
+    delete FunctionDbgScope;
+    DbgScopeMap.clear();
+    DbgAbstractScopeMap.clear();
+    DbgConcreteScopeMap.clear();
+    InlinedVariableScopes.clear();
+    FunctionDbgScope = NULL;
+    LexicalScopeStack.clear();
+    AbstractInstanceRootList.clear();
+  }
 
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
+  Lines.clear();
 
-  Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
-  EmitDifference("pubnames_end", Unit->getID(),
-                 "pubnames_begin", Unit->getID(), true);
-  Asm->EOL("Length of Public Names Info");
-
-  EmitLabel("pubnames_begin", Unit->getID());
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
+  CompileUnit *Unit = CompileUnitMap[V];
+  assert(Unit && "Unable to find CompileUnit");
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
 
-  EmitSectionOffset("info_begin", "section_info",
-                    Unit->getID(), 0, true, false);
-  Asm->EOL("Offset of Compilation Unit Info");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
-                 true);
-  Asm->EOL("Compilation Unit Length");
+  return ID;
+}
 
-  StringMap<DIE*> &Globals = Unit->getGlobals();
-  for (StringMap<DIE*>::const_iterator
-         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
-    const char *Name = GI->getKeyData();
-    DIE * Entity = GI->second;
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
+                                      DICompileUnit CU) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
-    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
-  }
+  std::string Dir, Fn;
+  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
+                                     CU.getFilename(Fn));
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
 
-  Asm->EmitInt32(0); Asm->EOL("End Mark");
-  EmitLabel("pubnames_end", Unit->getID());
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  Asm->EOL();
+  return ID;
 }
 
-/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
-///
-void DwarfDebug::EmitDebugPubNames() {
-  // Start the dwarf pubnames section.
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
+/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
+/// timed. Look up the source id with the given directory and source file
+/// names. If none currently exists, create a new id and insert it in the
+/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
+/// well.
+unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  if (MainCU) {
-    EmitDebugPubNamesPerCU(MainCU);
-    return;
-  }
+  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugPubNamesPerCU(CompileUnits[i]);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+
+  return SrcId;
 }
 
-/// EmitDebugStr - Emit visible names into a debug str section.
-///
-void DwarfDebug::EmitDebugStr() {
-  // Check to see if it is worth the effort.
-  if (!StringPool.empty()) {
-    // Start the dwarf str section.
-    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+/// RecordRegionStart - Indicate the start of a region.
+unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    // For each of strings in the string pool.
-    for (unsigned StringID = 1, N = StringPool.size();
-         StringID <= N; ++StringID) {
-      // Emit a label for reference from debug information entries.
-      EmitLabel("string", StringID);
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
+  LexicalScopeStack.push_back(Scope);
 
-      // Emit the string itself.
-      const std::string &String = StringPool[StringID];
-      Asm->EmitString(String); Asm->EOL();
-    }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-    Asm->EOL();
-  }
+  return ID;
 }
 
-/// EmitDebugLoc - Emit visible names into a debug loc section.
-///
-void DwarfDebug::EmitDebugLoc() {
-  // Start the dwarf loc section.
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  Asm->EOL();
-}
+/// RecordRegionEnd - Indicate the end of a region.
+unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-/// EmitDebugARanges - Emit visible names into a debug aranges section.
-///
-void DwarfDebug::EmitDebugARanges() {
-  // Start the dwarf aranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  Scope->setEndLabelID(ID);
+  if (LexicalScopeStack.size() != 0)
+    LexicalScopeStack.pop_back();
 
-  // FIXME - Mock up
-#if 0
-  CompileUnit *Unit = GetBaseCompileUnit();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  // Don't include size of length
-  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
+  return ID;
+}
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+/// RecordVariable - Indicate the declaration of a local variable.
+void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
+                                const MachineInstr *MI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitReference("info_begin", Unit->getID());
-  Asm->EOL("Offset of Compilation Unit Info");
+  DIDescriptor Desc(GV);
+  DbgScope *Scope = NULL;
+  bool InlinedFnVar = false;
 
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
+  if (Desc.getTag() == dwarf::DW_TAG_variable) {
+    // GV is a global variable.
+    DIGlobalVariable DG(GV);
+    Scope = getOrCreateScope(DG.getContext().getGV());
+  } else {
+    DenseMap<const MachineInstr *, DbgScope *>::iterator
+      SI = InlinedVariableScopes.find(MI);
 
-  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
+    if (SI != InlinedVariableScopes.end()) {
+      // or GV is an inlined local variable.
+      Scope = SI->second;
+    } else {
+      DIVariable DV(GV);
+      GlobalVariable *V = DV.getContext().getGV();
 
-  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
-  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
+      // FIXME: The code that checks for the inlined local variable is a hack!
+      DenseMap<const GlobalVariable *, DbgScope *>::iterator
+        AI = AbstractInstanceRootMap.find(V);
 
-  // Range 1
-  EmitReference("text_begin", 0); Asm->EOL("Address");
-  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
+      if (AI != AbstractInstanceRootMap.end()) {
+        // This method is called each time a DECLARE node is encountered. For an
+        // inlined function, this could be many, many times. We don't want to
+        // re-add variables to that DIE for each time. We just want to add them
+        // once. Check to make sure that we haven't added them already.
+        DenseMap<const GlobalVariable *,
+          SmallSet<const GlobalVariable *, 32> >::iterator
+          IP = InlinedParamMap.find(V);
 
-  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
-  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
-#endif
+        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
+          if (TimePassesIsEnabled)
+            DebugTimer->stopTimer();
+          return;
+        }
 
-  Asm->EOL();
-}
+        // or GV is an inlined local variable.
+        Scope = AI->second;
+        InlinedParamMap[V].insert(GV);
+        InlinedFnVar = true;
+      } else {
+        // or GV is a local variable.
+        Scope = getOrCreateScope(V);
+      }
+    }
+  }
 
-/// EmitDebugRanges - Emit visible names into a debug ranges section.
-///
-void DwarfDebug::EmitDebugRanges() {
-  // Start the dwarf ranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  Asm->EOL();
-}
+  assert(Scope && "Unable to find the variable's scope");
+  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
+  Scope->AddVariable(DV);
 
-/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
-///
-void DwarfDebug::EmitDebugMacInfo() {
-  if (TAI->doesSupportMacInfoSection()) {
-    // Start the dwarf macinfo section.
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    Asm->EOL();
-  }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitDebugInlineInfo - Emit inline info using following format.
-/// Section Header:
-/// 1. length of section
-/// 2. Dwarf version number
-/// 3. address size.
-///
-/// Entries (one "entry" for each function that was inlined):
-///
-/// 1. offset into __debug_str section for MIPS linkage name, if exists;
-///   otherwise offset into __debug_str for regular function name.
-/// 2. offset into __debug_str section for regular function name.
-/// 3. an unsigned LEB128 number indicating the number of distinct inlining
-/// instances for the function.
-///
-/// The rest of the entry consists of a {die_offset, low_pc} pair for each
-/// inlined instance; the die_offset points to the inlined_subroutine die in the
-/// __debug_info section, and the low_pc is the starting address for the
-/// inlining instance.
-void DwarfDebug::EmitDebugInlineInfo() {
-  if (!TAI->doesDwarfUsesInlineInfoSection())
-    return;
+/// RecordInlinedFnStart - Indicate the start of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
+                                          unsigned Line, unsigned Col) {
+  unsigned LabelID = MMI->NextLabelID();
 
-  if (!MainCU)
-    return;
+  if (!TAI->doesDwarfUsesInlineInfoSection())
+    return LabelID;
 
-  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
-  Asm->EOL();
-  EmitDifference("debug_inlined_end", 1,
-                 "debug_inlined_begin", 1, true);
-  Asm->EOL("Length of Debug Inlined Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitLabel("debug_inlined_begin", 1);
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<const GlobalVariable *, DbgScope *>::iterator
+    II = AbstractInstanceRootMap.find(GV);
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+  if (II == AbstractInstanceRootMap.end()) {
+    // Create an abstract instance entry for this inlined function if it doesn't
+    // already exist.
+    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
 
-  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
-    GlobalVariable *GV = I->first;
-    SmallVector<unsigned, 4> &Labels = I->second;
-    DISubprogram SP(GV);
-    std::string Name;
-    std::string LName;
+    // Get the compile unit context.
+    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+    DIE *SPDie = Unit->getDieMapSlotFor(GV);
+    if (!SPDie)
+      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
 
-    SP.getLinkageName(LName);
-    SP.getName(Name);
+    // Mark as being inlined. This makes this subprogram entry an abstract
+    // instance root.
+    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
+    // that it's defined. That probably won't change in the future. However,
+    // this could be more elegant.
+    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
 
-    Asm->EmitString(LName.empty() ? Name : LName);
-    Asm->EOL("MIPS linkage name");
+    // Keep track of the abstract scope for this function.
+    DbgAbstractScopeMap[GV] = Scope;
 
-    Asm->EmitString(Name); Asm->EOL("Function name");
+    AbstractInstanceRootMap[GV] = Scope;
+    AbstractInstanceRootList.push_back(Scope);
+  }
 
-    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
+  // Create a concrete inlined instance for this inlined function.
+  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
+  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
+  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+  ScopeDie->setAbstractCompileUnit(Unit);
 
-    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
-           LE = Labels.end(); LI != LE; ++LI) {
-      DIE *SP = MainCU->getDieMapSlotFor(GV);
-      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
+  DIE *Origin = Unit->getDieMapSlotFor(GV);
+  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
+              dwarf::DW_FORM_ref4, Origin);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
+  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
 
-      if (TD->getPointerSize() == sizeof(int32_t))
-        O << TAI->getData32bitsDirective();
-      else
-        O << TAI->getData64bitsDirective();
+  ConcreteScope->setDie(ScopeDie);
+  ConcreteScope->setStartLabelID(LabelID);
+  MMI->RecordUsedDbgLabel(LabelID);
 
-      PrintLabelName("label", *LI); Asm->EOL("low_pc");
-    }
-  }
+  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
+  LexicalScopeStack.push_back(ConcreteScope);
 
-  EmitLabel("debug_inlined_end", 1);
-  Asm->EOL();
-}
+  // Keep track of the concrete scope that's inlined into this function.
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    SI = DbgConcreteScopeMap.find(GV);
 
-/// GetOrCreateSourceID - Look up the source id with the given directory and
-/// source file names. If none currently exists, create a new id and insert it
-/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
-/// maps as well.
-unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  unsigned DId;
-  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
-  if (DI != DirectoryIdMap.end()) {
-    DId = DI->getValue();
-  } else {
-    DId = DirectoryNames.size() + 1;
-    DirectoryIdMap[DirName] = DId;
-    DirectoryNames.push_back(DirName);
-  }
+  if (SI == DbgConcreteScopeMap.end())
+    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
+  else
+    SI->second.push_back(ConcreteScope);
 
-  unsigned FId;
-  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
-  if (FI != SourceFileIdMap.end()) {
-    FId = FI->getValue();
-  } else {
-    FId = SourceFileNames.size() + 1;
-    SourceFileIdMap[FileName] = FId;
-    SourceFileNames.push_back(FileName);
-  }
+  // Track the start label for this inlined function.
+  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+    I = InlineInfo.find(GV);
 
-  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
-    SourceIdMap.find(std::make_pair(DId, FId));
-  if (SI != SourceIdMap.end())
-    return SI->second;
+  if (I == InlineInfo.end())
+    InlineInfo[GV].push_back(LabelID);
+  else
+    I->second.push_back(LabelID);
 
-  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
-  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
-  SourceIds.push_back(std::make_pair(DId, FId));
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  return SrcId;
+  return LabelID;
 }
 
-void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
-  DICompileUnit DIUnit(GV);
-  std::string Dir, FN, Prod;
-  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
-                                    DIUnit.getFilename(FN));
+/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
+  if (!TAI->doesDwarfUsesInlineInfoSection())
+    return 0;
 
-  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
-  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
-                   DWLabel("section_line", 0), DWLabel("section_line", 0),
-                   false);
-  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
-            DIUnit.getProducer(Prod));
-  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
-          DIUnit.getLanguage());
-  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  if (!Dir.empty())
-    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
-  if (DIUnit.isOptimized())
-    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    I = DbgConcreteScopeMap.find(GV);
 
-  std::string Flags;
-  DIUnit.getFlags(Flags);
-  if (!Flags.empty())
-    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
+  if (I == DbgConcreteScopeMap.end()) {
+    // FIXME: Can this situation actually happen? And if so, should it?
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  unsigned RVer = DIUnit.getRunTimeVersion();
-  if (RVer)
-    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
-            dwarf::DW_FORM_data1, RVer);
+    return 0;
+  }
 
-  CompileUnit *Unit = new CompileUnit(ID, Die);
-  if (DIUnit.isMain()) {
-    assert(!MainCU && "Multiple main compile units are found!");
-    MainCU = Unit;
-    }
+  SmallVector<DbgScope *, 8> &Scopes = I->second;
+  assert(!Scopes.empty() && "We should have at least one debug scope!");
+  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
+  unsigned ID = MMI->NextLabelID();
+  MMI->RecordUsedDbgLabel(ID);
+  Scope->setEndLabelID(ID);
+  LexicalScopeStack.pop_back();
 
-  CompileUnitMap[DIUnit.getGV()] = Unit;
-  CompileUnits.push_back(Unit);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+
+  return ID;
 }
 
-/// ConstructCompileUnits - Create a compile unit DIEs.
-void DwarfDebug::ConstructCompileUnits() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
-  if (!Root)
+/// RecordVariableScope - Record scope for the variable declared by
+/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
+/// for only inlined subroutine variables. Other variables's scopes are
+/// determined during RecordVariable().
+void DwarfDebug::RecordVariableScope(DIVariable &DV,
+                                     const MachineInstr *DeclareMI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
+
+  DISubprogram SP(DV.getContext().getGV());
+
+  if (SP.isNull()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
+
     return;
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed compile unit descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed compile unit descriptor anchor type");
+  }
 
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI) {
-      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
-      ConstructCompileUnit(GV);
-    }
+  DenseMap<GlobalVariable *, DbgScope *>::iterator
+    I = DbgAbstractScopeMap.find(SP.getGV());
+  if (I != DbgAbstractScopeMap.end())
+    InlinedVariableScopes[DeclareMI] = I->second;
+
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
-  DIGlobalVariable DI_GV(GV);
-  CompileUnit *DW_Unit = MainCU;
-  if (!DW_Unit)
-    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
+//===----------------------------------------------------------------------===//
+// Emit Methods
+//===----------------------------------------------------------------------===//
 
-  // Check for pre-existence.
-  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
-  if (Slot)
-    return false;
+/// SizeAndOffsetDie - Compute the size and offset of a DIE.
+///
+unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
+  // Get the children.
+  const std::vector<DIE *> &Children = Die->getChildren();
+
+  // If not last sibling and has children then add sibling offset attribute.
+  if (!Last && !Children.empty()) Die->AddSiblingOffset();
+
+  // Record the abbreviation.
+  AssignAbbrevNumber(Die->getAbbrev());
+
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+
+  // Set DIE offset
+  Die->setOffset(Offset);
+
+  // Start the size with the size of abbreviation code.
+  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
 
-  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
+  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-  // Add address.
-  DIEBlock *Block = new DIEBlock();
-  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
-  std::string GLN;
-  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
-                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
-  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
+  // Size the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i)
+    // Size attribute value.
+    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
 
-  // Add to map.
-  Slot = VariableDie;
+  // Size the DIE children if any.
+  if (!Children.empty()) {
+    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
+           "Children flag not set");
 
-  // Add to context owner.
-  DW_Unit->getDie()->AddChild(VariableDie);
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
 
-  // Expose as global. FIXME - need to check external flag.
-  std::string Name;
-  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
-  return true;
+    // End of children marker.
+    Offset += sizeof(int8_t);
+  }
+
+  Die->setSize(Offset - Die->getOffset());
+  return Offset;
 }
 
-/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
-/// global variables. Return true if at least one global DIE is created.
-bool DwarfDebug::ConstructGlobalVariableDIEs() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
-  if (!Root)
-    return false;
+/// SizeAndOffsets - Compute the size and offset of all the DIEs.
+///
+void DwarfDebug::SizeAndOffsets() {
+  // Compute size of compile unit header.
+  static unsigned Offset =
+    sizeof(int32_t) + // Length of Compilation Unit Info
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t);   // Pointer Size (in bytes)
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed global variable descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed global variable descriptor anchor type");
+  // Process base compile unit.
+  if (MainCU) {
+    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
+    CompileUnitOffsets[MainCU] = 0;
+    return;
+  }
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
+  // Process all compile units.
+  unsigned PrevOffset = 0;
 
-  return Result;
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
+    CompileUnit *Unit = CompileUnits[i];
+    CompileUnitOffsets[Unit] = PrevOffset;
+    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
+      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+  }
 }
 
-bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
-  DISubprogram SP(GV);
-  CompileUnit *Unit = MainCU;
-  if (!Unit)
-    Unit = &FindCompileUnit(SP.getCompileUnit());
-
-  // Check for pre-existence.
-  DIE *&Slot = Unit->getDieMapSlotFor(GV);
-  if (Slot)
-    return false;
+/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
+/// tools to recognize the object file contains Dwarf information.
+void DwarfDebug::EmitInitial() {
+  // Check to see if we already emitted intial headers.
+  if (didInitial) return;
+  didInitial = true;
 
-  if (!SP.isDefinition())
-    // This is a method declaration which will be handled while constructing
-    // class type.
-    return false;
+  // Dwarf sections base addresses.
+  if (TAI->doesDwarfRequireFrameSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+    EmitLabel("section_debug_frame", 0);
+  }
 
-  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+  EmitLabel("section_info", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
+  EmitLabel("section_abbrev", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+  EmitLabel("section_aranges", 0);
 
-  // Add to map.
-  Slot = SubprogramDie;
+  if (TAI->doesSupportMacInfoSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    EmitLabel("section_macinfo", 0);
+  }
 
-  // Add to context owner.
-  Unit->getDie()->AddChild(SubprogramDie);
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
+  EmitLabel("section_line", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  EmitLabel("section_loc", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
+  EmitLabel("section_pubnames", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+  EmitLabel("section_str", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  EmitLabel("section_ranges", 0);
 
-  // Expose as global.
-  std::string Name;
-  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
-  return true;
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_begin", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_begin", 0);
 }
 
-/// ConstructSubprograms - Create DIEs for each of the externally visible
-/// subprograms. Return true if at least one subprogram DIE is created.
-bool DwarfDebug::ConstructSubprograms() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
-  if (!Root)
-    return false;
+/// EmitDIE - Recusively Emits a debug information entry.
+///
+void DwarfDebug::EmitDIE(DIE *Die) {
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed subprogram descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed subprogram descriptor anchor type");
+  Asm->EOL();
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
+  // Emit the code (index) for the abbreviation.
+  Asm->EmitULEB128Bytes(AbbrevNumber);
 
-  return Result;
-}
+  if (Asm->isVerbose())
+    Asm->EOL(std::string("Abbrev [" +
+                         utostr(AbbrevNumber) +
+                         "] 0x" + utohexstr(Die->getOffset()) +
+                         ":0x" + utohexstr(Die->getSize()) + " " +
+                         dwarf::TagString(Abbrev->getTag())));
+  else
+    Asm->EOL();
 
-/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
-/// This is inovked by the target AsmPrinter.
-void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-  // Create all the compile unit DIEs.
-  ConstructCompileUnits();
+  // Emit the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
+    unsigned Attr = AbbrevData[i].getAttribute();
+    unsigned Form = AbbrevData[i].getForm();
+    assert(Form && "Too many attributes for DIE (check abbreviation)");
 
-  if (CompileUnits.empty()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    switch (Attr) {
+    case dwarf::DW_AT_sibling:
+      Asm->EmitInt32(Die->SiblingOffset());
+      break;
+    case dwarf::DW_AT_abstract_origin: {
+      DIEEntry *E = cast<DIEEntry>(Values[i]);
+      DIE *Origin = E->getEntry();
+      unsigned Addr =
+        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
+        Origin->getOffset();
 
-    return;
-  }
+      Asm->EmitInt32(Addr);
+      break;
+    }
+    default:
+      // Emit an attribute using the defined form.
+      Values[i]->EmitValue(this, Form);
+      break;
+    }
 
-  // Create DIEs for each of the externally visible global variables.
-  bool globalDIEs = ConstructGlobalVariableDIEs();
+    Asm->EOL(dwarf::AttributeString(Attr));
+  }
 
-  // Create DIEs for each of the externally visible subprograms.
-  bool subprogramDIEs = ConstructSubprograms();
+  // Emit the DIE children if any.
+  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
+    const std::vector<DIE *> &Children = Die->getChildren();
 
-  // If there is not any debug info available for any global variables and any
-  // subprograms then there is not any debug info to emit.
-  if (!globalDIEs && !subprogramDIEs) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      EmitDIE(Children[j]);
 
-    return;
+    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
   }
+}
 
-  MMI = mmi;
-  shouldEmit = true;
-  MMI->setDebugInfoAvailability(true);
+/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
+///
+void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
+  DIE *Die = Unit->getDie();
+
+  // Emit the compile units header.
+  EmitLabel("info_begin", Unit->getID());
+
+  // Emit size of content not including length itself
+  unsigned ContentSize = Die->getSize() +
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t) +  // Pointer Size (in bytes)
+    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+
+  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
+  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
+  Asm->EOL("Offset Into Abbrev. Section");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+
+  EmitDIE(Die);
+  // FIXME - extra padding for gdb bug.
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  EmitLabel("info_end", Unit->getID());
 
-  // Prime section data.
-  SectionMap.insert(TAI->getTextSection());
+  Asm->EOL();
+}
 
-  // Print out .file directives to specify files for .loc directives. These are
-  // printed out early so that they precede any .loc directives.
-  if (TAI->hasDotLocAndDotFile()) {
-    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
-      // Remember source id starts at 1.
-      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
-      sys::Path FullPath(getSourceDirectoryName(Id.first));
-      bool AppendOk =
-        FullPath.appendComponent(getSourceFileName(Id.second));
-      assert(AppendOk && "Could not append filename to directory!");
-      AppendOk = false;
-      Asm->EmitFile(i, FullPath.toString());
-      Asm->EOL();
-    }
-  }
+void DwarfDebug::EmitDebugInfo() {
+  // Start debug info section.
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
 
-  // Emit initial sections
-  EmitInitial();
+  if (MainCU) {
+    EmitDebugInfoPerCU(MainCU);
+    return;
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugInfoPerCU(CompileUnits[i]);
 }
 
-/// EndModule - Emit all Dwarf sections that should come after the content.
+/// EmitAbbreviations - Emit the abbreviation section.
 ///
-void DwarfDebug::EndModule() {
-  if (!ShouldEmitDwarfDebug())
-    return;
+void DwarfDebug::EmitAbbreviations() const {
+  // Check to see if it is worth the effort.
+  if (!Abbreviations.empty()) {
+    // Start the debug abbrev section.
+    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    EmitLabel("abbrev_begin", 0);
 
-  // Standard sections final addresses.
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_end", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_end", 0);
+    // For each abbrevation.
+    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
+      // Get abbreviation data
+      const DIEAbbrev *Abbrev = Abbreviations[i];
 
-  // End text sections.
-  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
-    Asm->SwitchToSection(SectionMap[i]);
-    EmitLabel("section_end", i);
-  }
+      // Emit the abbrevations code (base 1 index.)
+      Asm->EmitULEB128Bytes(Abbrev->getNumber());
+      Asm->EOL("Abbreviation Code");
 
-  // Emit common frame information.
-  EmitCommonDebugFrame();
+      // Emit the abbreviations data.
+      Abbrev->Emit(Asm);
 
-  // Emit function debug frame information
-  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
-         E = DebugFrames.end(); I != E; ++I)
-    EmitFunctionDebugFrame(*I);
+      Asm->EOL();
+    }
 
-  // Compute DIE offsets and sizes.
-  SizeAndOffsets();
+    // Mark end of abbreviations.
+    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
 
-  // Emit all the DIEs into a debug info section
-  EmitDebugInfo();
+    EmitLabel("abbrev_end", 0);
+    Asm->EOL();
+  }
+}
 
-  // Corresponding abbreviations into a abbrev section.
-  EmitAbbreviations();
+/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
+/// the line matrix.
+///
+void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
+  // Define last address of section.
+  Asm->EmitInt8(0); Asm->EOL("Extended Op");
+  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
 
-  // Emit source line correspondence into a debug line section.
-  EmitDebugLines();
+  // Mark end of matrix.
+  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
+  Asm->EmitULEB128Bytes(1); Asm->EOL();
+  Asm->EmitInt8(1); Asm->EOL();
+}
 
-  // Emit info into a debug pubnames section.
-  EmitDebugPubNames();
+/// EmitDebugLines - Emit source line information.
+///
+void DwarfDebug::EmitDebugLines() {
+  // If the target is using .loc/.file, the assembler will be emitting the
+  // .debug_line table automatically.
+  if (TAI->hasDotLocAndDotFile())
+    return;
 
-  // Emit info into a debug str section.
-  EmitDebugStr();
+  // Minimum line delta, thus ranging from -10..(255-10).
+  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
+  // Maximum line delta, thus ranging from -10..(255-10).
+  const int MaxLineDelta = 255 + MinLineDelta;
 
-  // Emit info into a debug loc section.
-  EmitDebugLoc();
+  // Start the dwarf line section.
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
 
-  // Emit info into a debug aranges section.
-  EmitDebugARanges();
+  // Construct the section header.
+  EmitDifference("line_end", 0, "line_begin", 0, true);
+  Asm->EOL("Length of Source Line Info");
+  EmitLabel("line_begin", 0);
 
-  // Emit info into a debug ranges section.
-  EmitDebugRanges();
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
 
-  // Emit info into a debug macinfo section.
-  EmitDebugMacInfo();
+  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
+  Asm->EOL("Prolog Length");
+  EmitLabel("line_prolog_begin", 0);
 
-  // Emit inline info.
-  EmitDebugInlineInfo();
+  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
 
-/// BeginFunction - Gather pre-function debug information.  Assumes being
-/// emitted immediately after the function entry point.
-void DwarfDebug::BeginFunction(MachineFunction *MF) {
-  this->MF = MF;
+  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
 
-  if (!ShouldEmitDwarfDebug()) return;
+  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
 
-  // Begin accumulating function debug information.
-  MMI->BeginFunction(MF);
+  // Line number standard opcode encodings argument count
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
 
-  // Assumes in correct section after the entry point.
-  EmitLabel("func_begin", ++SubprogramCount);
+  // Emit directories.
+  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
+    Asm->EmitString(getSourceDirectoryName(DI));
+    Asm->EOL("Directory");
+  }
 
-  // Emit label for the implicitly defined dbg.stoppoint at the start of the
-  // function.
-  DebugLoc FDL = MF->getDefaultDebugLoc();
-  if (!FDL.isUnknown()) {
-    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
-    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
-                                        DICompileUnit(DLT.CompileUnit));
-    Asm->printLabel(LabelID);
+  Asm->EmitInt8(0); Asm->EOL("End of directories");
+
+  // Emit files.
+  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
+    // Remember source id starts at 1.
+    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
+    Asm->EmitString(getSourceFileName(Id.second));
+    Asm->EOL("Source");
+    Asm->EmitULEB128Bytes(Id.first);
+    Asm->EOL("Directory #");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("Mod date");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("File size");
   }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+  Asm->EmitInt8(0); Asm->EOL("End of files");
 
-/// EndFunction - Gather and emit post-function debug information.
-///
-void DwarfDebug::EndFunction(MachineFunction *MF) {
-  if (!ShouldEmitDwarfDebug()) return;
+  EmitLabel("line_prolog_end", 0);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  // A sequence for each text section.
+  unsigned SecSrcLinesSize = SectionSourceLines.size();
 
-  // Define end label for subprogram.
-  EmitLabel("func_end", SubprogramCount);
+  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
+    // Isolate current sections line info.
+    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
 
-  // Get function line info.
-  if (!Lines.empty()) {
-    // Get section line info.
-    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
-    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
-    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
-    // Append the function info to section info.
-    SectionLineInfos.insert(SectionLineInfos.end(),
-                            Lines.begin(), Lines.end());
-  }
+    if (Asm->isVerbose()) {
+      const Section* S = SectionMap[j + 1];
+      O << '\t' << TAI->getCommentString() << " Section"
+        << S->getName() << '\n';
+    } else {
+      Asm->EOL();
+    }
 
-  // Construct the DbgScope for abstract instances.
-  for (SmallVector<DbgScope *, 32>::iterator
-         I = AbstractInstanceRootList.begin(),
-         E = AbstractInstanceRootList.end(); I != E; ++I)
-    ConstructAbstractDbgScope(*I);
+    // Dwarf assumes we start with first line of first source file.
+    unsigned Source = 1;
+    unsigned Line = 1;
 
-  // Construct scopes for subprogram.
-  if (FunctionDbgScope)
-    ConstructFunctionDbgScope(FunctionDbgScope);
-  else
-    // FIXME: This is wrong. We are essentially getting past a problem with
-    // debug information not being able to handle unreachable blocks that have
-    // debug information in them. In particular, those unreachable blocks that
-    // have "region end" info in them. That situation results in the "root
-    // scope" not being created. If that's the case, then emit a "default"
-    // scope, i.e., one that encompasses the whole function. This isn't
-    // desirable. And a better way of handling this (and all of the debugging
-    // information) needs to be explored.
-    ConstructDefaultDbgScope(MF);
+    // Construct rows of the address, source, line, column matrix.
+    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
+      const SrcLineInfo &LineInfo = LineInfos[i];
+      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
+      if (!LabelID) continue;
+
+      if (!Asm->isVerbose())
+        Asm->EOL();
+      else {
+        std::pair<unsigned, unsigned> SourceID =
+          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
+        O << '\t' << TAI->getCommentString() << ' '
+          << getSourceDirectoryName(SourceID.first) << ' '
+          << getSourceFileName(SourceID.second)
+          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
+      }
 
-  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
-                                               MMI->getFrameMoves()));
+      // Define the line address.
+      Asm->EmitInt8(0); Asm->EOL("Extended Op");
+      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+      EmitReference("label",  LabelID); Asm->EOL("Location label");
 
-  // Clear debug info
-  if (FunctionDbgScope) {
-    delete FunctionDbgScope;
-    DbgScopeMap.clear();
-    DbgAbstractScopeMap.clear();
-    DbgConcreteScopeMap.clear();
-    InlinedVariableScopes.clear();
-    FunctionDbgScope = NULL;
-    LexicalScopeStack.clear();
-    AbstractInstanceRootList.clear();
-  }
+      // If change of source, then switch to the new source.
+      if (Source != LineInfo.getSourceID()) {
+        Source = LineInfo.getSourceID();
+        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
+        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
+      }
 
-  Lines.clear();
+      // If change of line.
+      if (Line != LineInfo.getLine()) {
+        // Determine offset.
+        int Offset = LineInfo.getLine() - Line;
+        int Delta = Offset - MinLineDelta;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+        // Update line.
+        Line = LineInfo.getLine();
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+        // If delta is small enough and in range...
+        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
+          // ... then use fast opcode.
+          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
+        } else {
+          // ... otherwise use long hand.
+          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
+          Asm->EOL("DW_LNS_advance_line");
+          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
+          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+        }
+      } else {
+        // Copy the previous row (different address or source)
+        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+      }
+    }
 
-  CompileUnit *Unit = CompileUnitMap[V];
-  assert(Unit && "Unable to find CompileUnit");
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
+    EmitEndOfLineMatrix(j + 1);
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  if (SecSrcLinesSize == 0)
+    // Because we're emitting a debug_line section, we still need a line
+    // table. The linker and friends expect it to exist. If there's nothing to
+    // put into it, emit an empty table.
+    EmitEndOfLineMatrix(1);
 
-  return ID;
+  EmitLabel("line_end", 0);
+  Asm->EOL();
 }
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
-                                      DICompileUnit CU) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
+///
+void DwarfDebug::EmitCommonDebugFrame() {
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  std::string Dir, Fn;
-  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
-                                     CU.getFilename(Fn));
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
+  int stackGrowth =
+    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
+      TargetFrameInfo::StackGrowsUp ?
+    TD->getPointerSize() : -TD->getPointerSize();
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  return ID;
-}
+  EmitLabel("debug_frame_common", 0);
+  EmitDifference("debug_frame_common_end", 0,
+                 "debug_frame_common_begin", 0, true);
+  Asm->EOL("Length of Common Information Entry");
 
-/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
-/// timed. Look up the source id with the given directory and source file
-/// names. If none currently exists, create a new id and insert it in the
-/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
-/// well.
-unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("debug_frame_common_begin", 0);
+  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
+  Asm->EOL("CIE Identifier Tag");
+  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
+  Asm->EOL("CIE Version");
+  Asm->EmitString("");
+  Asm->EOL("CIE Augmentation");
+  Asm->EmitULEB128Bytes(1);
+  Asm->EOL("CIE Code Alignment Factor");
+  Asm->EmitSLEB128Bytes(stackGrowth);
+  Asm->EOL("CIE Data Alignment Factor");
+  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
+  Asm->EOL("CIE RA Column");
 
-  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
+  std::vector<MachineMove> Moves;
+  RI->getInitialFrameState(Moves);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitFrameMoves(NULL, 0, Moves, false);
 
-  return SrcId;
-}
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_common_end", 0);
 
-/// RecordRegionStart - Indicate the start of a region.
-unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EOL();
+}
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
-  LexicalScopeStack.push_back(Scope);
+/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
+/// section.
+void
+DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  return ID;
-}
+  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
+                 "debug_frame_begin", DebugFrameInfo.Number, true);
+  Asm->EOL("Length of Frame Information Entry");
 
-/// RecordRegionEnd - Indicate the end of a region.
-unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  Scope->setEndLabelID(ID);
-  if (LexicalScopeStack.size() != 0)
-    LexicalScopeStack.pop_back();
+  EmitSectionOffset("debug_frame_common", "section_debug_frame",
+                    0, 0, true, false);
+  Asm->EOL("FDE CIE offset");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitReference("func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE initial location");
+  EmitDifference("func_end", DebugFrameInfo.Number,
+                 "func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE address range");
 
-  return ID;
-}
+  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
+                 false);
 
-/// RecordVariable - Indicate the declaration of a local variable.
-void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
-                                const MachineInstr *MI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
 
-  DIDescriptor Desc(GV);
-  DbgScope *Scope = NULL;
-  bool InlinedFnVar = false;
+  Asm->EOL();
+}
 
-  if (Desc.getTag() == dwarf::DW_TAG_variable) {
-    // GV is a global variable.
-    DIGlobalVariable DG(GV);
-    Scope = getOrCreateScope(DG.getContext().getGV());
-  } else {
-    DenseMap<const MachineInstr *, DbgScope *>::iterator
-      SI = InlinedVariableScopes.find(MI);
+void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
+  EmitDifference("pubnames_end", Unit->getID(),
+                 "pubnames_begin", Unit->getID(), true);
+  Asm->EOL("Length of Public Names Info");
 
-    if (SI != InlinedVariableScopes.end()) {
-      // or GV is an inlined local variable.
-      Scope = SI->second;
-    } else {
-      DIVariable DV(GV);
-      GlobalVariable *V = DV.getContext().getGV();
+  EmitLabel("pubnames_begin", Unit->getID());
 
-      // FIXME: The code that checks for the inlined local variable is a hack!
-      DenseMap<const GlobalVariable *, DbgScope *>::iterator
-        AI = AbstractInstanceRootMap.find(V);
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
 
-      if (AI != AbstractInstanceRootMap.end()) {
-        // This method is called each time a DECLARE node is encountered. For an
-        // inlined function, this could be many, many times. We don't want to
-        // re-add variables to that DIE for each time. We just want to add them
-        // once. Check to make sure that we haven't added them already.
-        DenseMap<const GlobalVariable *,
-          SmallSet<const GlobalVariable *, 32> >::iterator
-          IP = InlinedParamMap.find(V);
+  EmitSectionOffset("info_begin", "section_info",
+                    Unit->getID(), 0, true, false);
+  Asm->EOL("Offset of Compilation Unit Info");
 
-        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
-          if (TimePassesIsEnabled)
-            DebugTimer->stopTimer();
-          return;
-        }
+  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
+                 true);
+  Asm->EOL("Compilation Unit Length");
 
-        // or GV is an inlined local variable.
-        Scope = AI->second;
-        InlinedParamMap[V].insert(GV);
-        InlinedFnVar = true;
-      } else {
-        // or GV is a local variable.
-        Scope = getOrCreateScope(V);
-      }
-    }
+  StringMap<DIE*> &Globals = Unit->getGlobals();
+  for (StringMap<DIE*>::const_iterator
+         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
+    const char *Name = GI->getKeyData();
+    DIE * Entity = GI->second;
+
+    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
+    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
   }
 
-  assert(Scope && "Unable to find the variable's scope");
-  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
-  Scope->AddVariable(DV);
+  Asm->EmitInt32(0); Asm->EOL("End Mark");
+  EmitLabel("pubnames_end", Unit->getID());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  Asm->EOL();
 }
 
-//// RecordInlinedFnStart - Indicate the start of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
-                                          unsigned Line, unsigned Col) {
-  unsigned LabelID = MMI->NextLabelID();
+/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
+///
+void DwarfDebug::EmitDebugPubNames() {
+  // Start the dwarf pubnames section.
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
 
-  if (!TAI->doesDwarfUsesInlineInfoSection())
-    return LabelID;
+  if (MainCU) {
+    EmitDebugPubNamesPerCU(MainCU);
+    return;
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugPubNamesPerCU(CompileUnits[i]);
+}
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<const GlobalVariable *, DbgScope *>::iterator
-    II = AbstractInstanceRootMap.find(GV);
+/// EmitDebugStr - Emit visible names into a debug str section.
+///
+void DwarfDebug::EmitDebugStr() {
+  // Check to see if it is worth the effort.
+  if (!StringPool.empty()) {
+    // Start the dwarf str section.
+    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
 
-  if (II == AbstractInstanceRootMap.end()) {
-    // Create an abstract instance entry for this inlined function if it doesn't
-    // already exist.
-    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
+    // For each of strings in the string pool.
+    for (unsigned StringID = 1, N = StringPool.size();
+         StringID <= N; ++StringID) {
+      // Emit a label for reference from debug information entries.
+      EmitLabel("string", StringID);
 
-    // Get the compile unit context.
-    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-    DIE *SPDie = Unit->getDieMapSlotFor(GV);
-    if (!SPDie)
-      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
+      // Emit the string itself.
+      const std::string &String = StringPool[StringID];
+      Asm->EmitString(String); Asm->EOL();
+    }
 
-    // Mark as being inlined. This makes this subprogram entry an abstract
-    // instance root.
-    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
-    // that it's defined. That probably won't change in the future. However,
-    // this could be more elegant.
-    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
+    Asm->EOL();
+  }
+}
 
-    // Keep track of the abstract scope for this function.
-    DbgAbstractScopeMap[GV] = Scope;
+/// EmitDebugLoc - Emit visible names into a debug loc section.
+///
+void DwarfDebug::EmitDebugLoc() {
+  // Start the dwarf loc section.
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  Asm->EOL();
+}
 
-    AbstractInstanceRootMap[GV] = Scope;
-    AbstractInstanceRootList.push_back(Scope);
-  }
+/// EmitDebugARanges - Emit visible names into a debug aranges section.
+///
+void DwarfDebug::EmitDebugARanges() {
+  // Start the dwarf aranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
 
-  // Create a concrete inlined instance for this inlined function.
-  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
-  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
-  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-  ScopeDie->setAbstractCompileUnit(Unit);
+  // FIXME - Mock up
+#if 0
+  CompileUnit *Unit = GetBaseCompileUnit();
 
-  DIE *Origin = Unit->getDieMapSlotFor(GV);
-  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
-              dwarf::DW_FORM_ref4, Origin);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
-  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
+  // Don't include size of length
+  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
 
-  ConcreteScope->setDie(ScopeDie);
-  ConcreteScope->setStartLabelID(LabelID);
-  MMI->RecordUsedDbgLabel(LabelID);
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
 
-  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
+  EmitReference("info_begin", Unit->getID());
+  Asm->EOL("Offset of Compilation Unit Info");
 
-  // Keep track of the concrete scope that's inlined into this function.
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    SI = DbgConcreteScopeMap.find(GV);
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
 
-  if (SI == DbgConcreteScopeMap.end())
-    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
-  else
-    SI->second.push_back(ConcreteScope);
+  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
 
-  // Track the start label for this inlined function.
-  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-    I = InlineInfo.find(GV);
+  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
+  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
 
-  if (I == InlineInfo.end())
-    InlineInfo[GV].push_back(LabelID);
-  else
-    I->second.push_back(LabelID);
+  // Range 1
+  EmitReference("text_begin", 0); Asm->EOL("Address");
+  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
+  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
+#endif
 
-  return LabelID;
+  Asm->EOL();
 }
 
-/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
+/// EmitDebugRanges - Emit visible names into a debug ranges section.
+///
+void DwarfDebug::EmitDebugRanges() {
+  // Start the dwarf ranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  Asm->EOL();
+}
+
+/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
+///
+void DwarfDebug::EmitDebugMacInfo() {
+  if (TAI->doesSupportMacInfoSection()) {
+    // Start the dwarf macinfo section.
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    Asm->EOL();
+  }
+}
+
+/// EmitDebugInlineInfo - Emit inline info using following format.
+/// Section Header:
+/// 1. length of section
+/// 2. Dwarf version number
+/// 3. address size.
+///
+/// Entries (one "entry" for each function that was inlined):
+///
+/// 1. offset into __debug_str section for MIPS linkage name, if exists;
+///   otherwise offset into __debug_str for regular function name.
+/// 2. offset into __debug_str section for regular function name.
+/// 3. an unsigned LEB128 number indicating the number of distinct inlining
+/// instances for the function.
+///
+/// The rest of the entry consists of a {die_offset, low_pc} pair for each
+/// inlined instance; the die_offset points to the inlined_subroutine die in the
+/// __debug_info section, and the low_pc is the starting address for the
+/// inlining instance.
+void DwarfDebug::EmitDebugInlineInfo() {
   if (!TAI->doesDwarfUsesInlineInfoSection())
-    return 0;
+    return;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  if (!MainCU)
+    return;
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    I = DbgConcreteScopeMap.find(GV);
+  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
+  Asm->EOL();
+  EmitDifference("debug_inlined_end", 1,
+                 "debug_inlined_begin", 1, true);
+  Asm->EOL("Length of Debug Inlined Information Entry");
 
-  if (I == DbgConcreteScopeMap.end()) {
-    // FIXME: Can this situation actually happen? And if so, should it?
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_begin", 1);
 
-    return 0;
-  }
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
 
-  SmallVector<DbgScope *, 8> &Scopes = I->second;
-  assert(!Scopes.empty() && "We should have at least one debug scope!");
-  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
-  unsigned ID = MMI->NextLabelID();
-  MMI->RecordUsedDbgLabel(ID);
-  Scope->setEndLabelID(ID);
+  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
+    GlobalVariable *GV = I->first;
+    SmallVector<unsigned, 4> &Labels = I->second;
+    DISubprogram SP(GV);
+    std::string Name;
+    std::string LName;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+    SP.getLinkageName(LName);
+    SP.getName(Name);
 
-  return ID;
-}
+    Asm->EmitString(LName.empty() ? Name : LName);
+    Asm->EOL("MIPS linkage name");
 
-/// RecordVariableScope - Record scope for the variable declared by
-/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
-/// for only inlined subroutine variables. Other variables's scopes are
-/// determined during RecordVariable().
-void DwarfDebug::RecordVariableScope(DIVariable &DV,
-                                     const MachineInstr *DeclareMI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    Asm->EmitString(Name); Asm->EOL("Function name");
 
-  DISubprogram SP(DV.getContext().getGV());
+    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
 
-  if (SP.isNull()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
+           LE = Labels.end(); LI != LE; ++LI) {
+      DIE *SP = MainCU->getDieMapSlotFor(GV);
+      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
 
-    return;
-  }
+      if (TD->getPointerSize() == sizeof(int32_t))
+        O << TAI->getData32bitsDirective();
+      else
+        O << TAI->getData64bitsDirective();
 
-  DenseMap<GlobalVariable *, DbgScope *>::iterator
-    I = DbgAbstractScopeMap.find(SP.getGV());
-  if (I != DbgAbstractScopeMap.end())
-    InlinedVariableScopes[DeclareMI] = I->second;
+      PrintLabelName("label", *LI); Asm->EOL("low_pc");
+    }
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_end", 1);
+  Asm->EOL();
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=72192&r1=72191&r2=72192&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Wed May 20 18:04:56 2009
@@ -230,16 +230,9 @@
   ///
   void AssignAbbrevNumber(DIEAbbrev &Abbrev);
 
-  /// NewString - Add a string to the constant pool and returns a label.
-  ///
-  DWLabel NewString(const std::string &String) {
-    unsigned StringID = StringPool.insert(String);
-    return DWLabel("string", StringID);
-  }
-
-  /// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
-  /// entry.
-  DIEEntry *NewDIEEntry(DIE *Entry = NULL);
+  /// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
+  /// information entry.
+  DIEEntry *CreateDIEEntry(DIE *Entry = NULL);
 
   /// SetDIEEntry - Set a DIEEntry once the debug information entry is defined.
   ///
@@ -282,7 +275,7 @@
   /// AddDIEEntry - Add a DIE attribute data and value.
   ///
   void AddDIEEntry(DIE *Die, unsigned Attribute, unsigned Form, DIE *Entry) {
-    Die->AddValue(Attribute, Form, NewDIEEntry(Entry));
+    Die->AddValue(Attribute, Form, CreateDIEEntry(Entry));
   }
 
   /// AddBlock - Add block data.
@@ -346,9 +339,9 @@
   ///
   CompileUnit &FindCompileUnit(DICompileUnit Unit) const;
 
-  /// NewDbgScopeVariable - Create a new scope variable.
+  /// CreateDbgScopeVariable - Create a new debug scope variable.
   ///
-  DIE *NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
+  DIE *CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
 
   /// getOrCreateScope - Returns the scope associated with the given descriptor.
   ///
@@ -362,17 +355,38 @@
 
   /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
   ///
-  void ConstructFunctionDbgScope(DbgScope *RootScope);
-
-  /// ConstructFunctionDbgScope - Construct the scope for the abstract debug
-  /// scope.
-  ///
-  void ConstructAbstractDbgScope(DbgScope *AbsScope);
+  void ConstructFunctionDbgScope(DbgScope *RootScope,
+                                 bool AbstractScope = false);
 
   /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
   ///
   void ConstructDefaultDbgScope(MachineFunction *MF);
 
+  /// GetOrCreateSourceID - Look up the source id with the given directory and
+  /// source file names. If none currently exists, create a new id and insert it
+  /// in the SourceIds map. This can update DirectoryNames and SourceFileNames maps
+  /// as well.
+  unsigned GetOrCreateSourceID(const std::string &DirName,
+                               const std::string &FileName);
+
+  void ConstructCompileUnit(GlobalVariable *GV);
+
+  /// ConstructCompileUnits - Create a compile unit DIEs.
+  void ConstructCompileUnits();
+
+  bool ConstructGlobalVariableDIE(GlobalVariable *GV);
+
+  /// ConstructGlobalVariableDIEs - Create DIEs for each of the externally 
+  /// visible global variables. Return true if at least one global DIE is
+  /// created.
+  bool ConstructGlobalVariableDIEs();
+
+  bool ConstructSubprogram(GlobalVariable *GV);
+
+  /// ConstructSubprograms - Create DIEs for each of the externally visible
+  /// subprograms. Return true if at least one subprogram DIE is created.
+  bool ConstructSubprograms();
+
   /// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
   /// tools to recognize the object file contains Dwarf information.
   void EmitInitial();
@@ -461,31 +475,6 @@
   /// the __debug_info section, and the low_pc is the starting address  for the
   ///  inlining instance.
   void EmitDebugInlineInfo();
-
-  /// GetOrCreateSourceID - Look up the source id with the given directory and
-  /// source file names. If none currently exists, create a new id and insert it
-  /// in the SourceIds map. This can update DirectoryNames and SourceFileNames maps
-  /// as well.
-  unsigned GetOrCreateSourceID(const std::string &DirName,
-                               const std::string &FileName);
-
-  void ConstructCompileUnit(GlobalVariable *GV);
-
-  /// ConstructCompileUnits - Create a compile unit DIEs.
-  void ConstructCompileUnits();
-
-  bool ConstructGlobalVariableDIE(GlobalVariable *GV);
-
-  /// ConstructGlobalVariableDIEs - Create DIEs for each of the externally 
-  /// visible global variables. Return true if at least one global DIE is
-  /// created.
-  bool ConstructGlobalVariableDIEs();
-
-  bool ConstructSubprogram(GlobalVariable *GV);
-
-  /// ConstructSubprograms - Create DIEs for each of the externally visible
-  /// subprograms. Return true if at least one subprogram DIE is created.
-  bool ConstructSubprograms();
 public:
   //===--------------------------------------------------------------------===//
   // Main entry points.




From isanbard at gmail.com  Wed May 20 18:19:14 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:19:14 -0000
Subject: [llvm-commits] [llvm] r72193 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <200905202319.n4KNJG0t028988@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:19:06 2009
New Revision: 72193

URL: http://llvm.org/viewvc/llvm-project?rev=72193&view=rev
Log:
Revert r72192. It was causing a build failure.

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=72193&r1=72192&r2=72193&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:19:06 2009
@@ -275,9 +275,9 @@
   }
 }
 
-/// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
-/// information entry.
-DIEEntry *DwarfDebug::CreateDIEEntry(DIE *Entry) {
+/// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
+/// entry.
+DIEEntry *DwarfDebug::NewDIEEntry(DIE *Entry) {
   DIEEntry *Value;
 
   if (Entry) {
@@ -545,7 +545,7 @@
   }
 
   // Set up proxy.
-  Slot = CreateDIEEntry();
+  Slot = NewDIEEntry();
 
   // Construct type.
   DIE Buffer(dwarf::DW_TAG_base_type);
@@ -925,9 +925,9 @@
   return *I->second;
 }
 
-/// CreateDbgScopeVariable - Create a new scope variable.
+/// NewDbgScopeVariable - Create a new scope variable.
 ///
-DIE *DwarfDebug::CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
+DIE *DwarfDebug::NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
   // Get the descriptor.
   const DIVariable &VD = DV->getVariable();
 
@@ -1012,7 +1012,7 @@
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = ParentScope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
-    DIE *VariableDie = CreateDbgScopeVariable(Variables[i], Unit);
+    DIE *VariableDie = NewDbgScopeVariable(Variables[i], Unit);
     if (VariableDie) ParentDie->AddChild(VariableDie);
   }
 
@@ -1092,8 +1092,7 @@
 
 /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
 ///
-void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope,
-                                           bool AbstractScope) {
+void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope) {
   // Exit if there is no root scope.
   if (!RootScope) return;
   DIDescriptor Desc = RootScope->getDesc();
@@ -1112,19 +1111,43 @@
   DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
   assert(SPDie && "Missing subprogram descriptor");
 
-  if (!AbstractScope) {
-    // Add the function bounds.
-    AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-             DWLabel("func_begin", SubprogramCount));
-    AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-             DWLabel("func_end", SubprogramCount));
-    MachineLocation Location(RI->getFrameRegister(*MF));
-    AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
-  }
+  // Add the function bounds.
+  AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
+           DWLabel("func_begin", SubprogramCount));
+  AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
+           DWLabel("func_end", SubprogramCount));
+  MachineLocation Location(RI->getFrameRegister(*MF));
+  AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
 
   ConstructDbgScope(RootScope, 0, 0, SPDie, Unit);
 }
 
+/// ConstructFunctionDbgScope - Construct the scope for the abstract debug
+/// scope.
+///
+void DwarfDebug::ConstructAbstractDbgScope(DbgScope *AbsScope) {
+  // Exit if there is no root scope.
+  if (!AbsScope) return;
+
+  DIDescriptor Desc = AbsScope->getDesc();
+  if (Desc.isNull())
+    return;
+
+  // Get the subprogram debug information entry.
+  DISubprogram SPD(Desc.getGV());
+
+  // Get the compile unit context.
+  CompileUnit *Unit = MainCU;
+  if (!Unit)
+    Unit = &FindCompileUnit(SPD.getCompileUnit());
+
+  // Get the subprogram die.
+  DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
+  assert(SPDie && "Missing subprogram descriptor");
+
+  ConstructDbgScope(AbsScope, 0, 0, SPDie, Unit);
+}
+
 /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
 ///
 void DwarfDebug::ConstructDefaultDbgScope(MachineFunction *MF) {
@@ -1173,1451 +1196,1445 @@
 #endif
 }
 
-/// GetOrCreateSourceID - Look up the source id with the given directory and
-/// source file names. If none currently exists, create a new id and insert it
-/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
-/// maps as well.
-unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  unsigned DId;
-  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
-  if (DI != DirectoryIdMap.end()) {
-    DId = DI->getValue();
-  } else {
-    DId = DirectoryNames.size() + 1;
-    DirectoryIdMap[DirName] = DId;
-    DirectoryNames.push_back(DirName);
-  }
+/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
+/// tools to recognize the object file contains Dwarf information.
+void DwarfDebug::EmitInitial() {
+  // Check to see if we already emitted intial headers.
+  if (didInitial) return;
+  didInitial = true;
 
-  unsigned FId;
-  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
-  if (FI != SourceFileIdMap.end()) {
-    FId = FI->getValue();
-  } else {
-    FId = SourceFileNames.size() + 1;
-    SourceFileIdMap[FileName] = FId;
-    SourceFileNames.push_back(FileName);
+  // Dwarf sections base addresses.
+  if (TAI->doesDwarfRequireFrameSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+    EmitLabel("section_debug_frame", 0);
   }
 
-  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
-    SourceIdMap.find(std::make_pair(DId, FId));
-  if (SI != SourceIdMap.end())
-    return SI->second;
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+  EmitLabel("section_info", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
+  EmitLabel("section_abbrev", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+  EmitLabel("section_aranges", 0);
 
-  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
-  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
-  SourceIds.push_back(std::make_pair(DId, FId));
+  if (TAI->doesSupportMacInfoSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    EmitLabel("section_macinfo", 0);
+  }
 
-  return SrcId;
-}
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
+  EmitLabel("section_line", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  EmitLabel("section_loc", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
+  EmitLabel("section_pubnames", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+  EmitLabel("section_str", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  EmitLabel("section_ranges", 0);
 
-void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
-  DICompileUnit DIUnit(GV);
-  std::string Dir, FN, Prod;
-  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
-                                    DIUnit.getFilename(FN));
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_begin", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_begin", 0);
+}
 
-  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
-  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
-                   DWLabel("section_line", 0), DWLabel("section_line", 0),
-                   false);
-  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
-            DIUnit.getProducer(Prod));
-  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
-          DIUnit.getLanguage());
-  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
+/// EmitDIE - Recusively Emits a debug information entry.
+///
+void DwarfDebug::EmitDIE(DIE *Die) {
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
 
-  if (!Dir.empty())
-    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
-  if (DIUnit.isOptimized())
-    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
+  Asm->EOL();
 
-  std::string Flags;
-  DIUnit.getFlags(Flags);
-  if (!Flags.empty())
-    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
+  // Emit the code (index) for the abbreviation.
+  Asm->EmitULEB128Bytes(AbbrevNumber);
 
-  unsigned RVer = DIUnit.getRunTimeVersion();
-  if (RVer)
-    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
-            dwarf::DW_FORM_data1, RVer);
+  if (Asm->isVerbose())
+    Asm->EOL(std::string("Abbrev [" +
+                         utostr(AbbrevNumber) +
+                         "] 0x" + utohexstr(Die->getOffset()) +
+                         ":0x" + utohexstr(Die->getSize()) + " " +
+                         dwarf::TagString(Abbrev->getTag())));
+  else
+    Asm->EOL();
 
-  CompileUnit *Unit = new CompileUnit(ID, Die);
-  if (DIUnit.isMain()) {
-    assert(!MainCU && "Multiple main compile units are found!");
-    MainCU = Unit;
-    }
+  SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-  CompileUnitMap[DIUnit.getGV()] = Unit;
-  CompileUnits.push_back(Unit);
-}
+  // Emit the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
+    unsigned Attr = AbbrevData[i].getAttribute();
+    unsigned Form = AbbrevData[i].getForm();
+    assert(Form && "Too many attributes for DIE (check abbreviation)");
 
-/// ConstructCompileUnits - Create a compile unit DIEs.
-void DwarfDebug::ConstructCompileUnits() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
-  if (!Root)
-    return;
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed compile unit descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed compile unit descriptor anchor type");
+    switch (Attr) {
+    case dwarf::DW_AT_sibling:
+      Asm->EmitInt32(Die->SiblingOffset());
+      break;
+    case dwarf::DW_AT_abstract_origin: {
+      DIEEntry *E = cast<DIEEntry>(Values[i]);
+      DIE *Origin = E->getEntry();
+      unsigned Addr =
+        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
+        Origin->getOffset();
 
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI) {
-      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
-      ConstructCompileUnit(GV);
+      Asm->EmitInt32(Addr);
+      break;
+    }
+    default:
+      // Emit an attribute using the defined form.
+      Values[i]->EmitValue(this, Form);
+      break;
     }
-}
 
-bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
-  DIGlobalVariable DI_GV(GV);
-  CompileUnit *DW_Unit = MainCU;
-  if (!DW_Unit)
-    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
+    Asm->EOL(dwarf::AttributeString(Attr));
+  }
 
-  // Check for pre-existence.
-  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
-  if (Slot)
-    return false;
+  // Emit the DIE children if any.
+  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
+    const std::vector<DIE *> &Children = Die->getChildren();
 
-  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      EmitDIE(Children[j]);
 
-  // Add address.
-  DIEBlock *Block = new DIEBlock();
-  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
-  std::string GLN;
-  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
-                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
-  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
+    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
+  }
+}
 
-  // Add to map.
-  Slot = VariableDie;
+/// SizeAndOffsetDie - Compute the size and offset of a DIE.
+///
+unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
+  // Get the children.
+  const std::vector<DIE *> &Children = Die->getChildren();
 
-  // Add to context owner.
-  DW_Unit->getDie()->AddChild(VariableDie);
+  // If not last sibling and has children then add sibling offset attribute.
+  if (!Last && !Children.empty()) Die->AddSiblingOffset();
 
-  // Expose as global. FIXME - need to check external flag.
-  std::string Name;
-  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
-  return true;
-}
+  // Record the abbreviation.
+  AssignAbbrevNumber(Die->getAbbrev());
 
-/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
-/// global variables. Return true if at least one global DIE is created.
-bool DwarfDebug::ConstructGlobalVariableDIEs() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
-  if (!Root)
-    return false;
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed global variable descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed global variable descriptor anchor type");
+  // Set DIE offset
+  Die->setOffset(Offset);
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
+  // Start the size with the size of abbreviation code.
+  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
 
-  return Result;
-}
+  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
-  DISubprogram SP(GV);
-  CompileUnit *Unit = MainCU;
-  if (!Unit)
-    Unit = &FindCompileUnit(SP.getCompileUnit());
+  // Size the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i)
+    // Size attribute value.
+    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
 
-  // Check for pre-existence.
-  DIE *&Slot = Unit->getDieMapSlotFor(GV);
-  if (Slot)
-    return false;
+  // Size the DIE children if any.
+  if (!Children.empty()) {
+    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
+           "Children flag not set");
 
-  if (!SP.isDefinition())
-    // This is a method declaration which will be handled while constructing
-    // class type.
-    return false;
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
 
-  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
+    // End of children marker.
+    Offset += sizeof(int8_t);
+  }
 
-  // Add to map.
-  Slot = SubprogramDie;
+  Die->setSize(Offset - Die->getOffset());
+  return Offset;
+}
 
-  // Add to context owner.
-  Unit->getDie()->AddChild(SubprogramDie);
+/// SizeAndOffsets - Compute the size and offset of all the DIEs.
+///
+void DwarfDebug::SizeAndOffsets() {
+  // Compute size of compile unit header.
+  static unsigned Offset =
+    sizeof(int32_t) + // Length of Compilation Unit Info
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t);   // Pointer Size (in bytes)
 
-  // Expose as global.
-  std::string Name;
-  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
-  return true;
+  // Process base compile unit.
+  if (MainCU) {
+    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
+    CompileUnitOffsets[MainCU] = 0;
+    return;
+  }
+
+  // Process all compile units.
+  unsigned PrevOffset = 0;
+
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
+    CompileUnit *Unit = CompileUnits[i];
+    CompileUnitOffsets[Unit] = PrevOffset;
+    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
+      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+  }
 }
 
-/// ConstructSubprograms - Create DIEs for each of the externally visible
-/// subprograms. Return true if at least one subprogram DIE is created.
-bool DwarfDebug::ConstructSubprograms() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
-  if (!Root)
-    return false;
+/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
+///
+void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
+  DIE *Die = Unit->getDie();
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed subprogram descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed subprogram descriptor anchor type");
+  // Emit the compile units header.
+  EmitLabel("info_begin", Unit->getID());
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
+  // Emit size of content not including length itself
+  unsigned ContentSize = Die->getSize() +
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t) +  // Pointer Size (in bytes)
+    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
 
-  return Result;
-}
+  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
+  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
+  Asm->EOL("Offset Into Abbrev. Section");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
 
-/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
-/// This is inovked by the target AsmPrinter.
-void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitDIE(Die);
+  // FIXME - extra padding for gdb bug.
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  EmitLabel("info_end", Unit->getID());
 
-  // Create all the compile unit DIEs.
-  ConstructCompileUnits();
+  Asm->EOL();
+}
 
-  if (CompileUnits.empty()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+void DwarfDebug::EmitDebugInfo() {
+  // Start debug info section.
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
 
+  if (MainCU) {
+    EmitDebugInfoPerCU(MainCU);
     return;
   }
 
-  // Create DIEs for each of the externally visible global variables.
-  bool globalDIEs = ConstructGlobalVariableDIEs();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugInfoPerCU(CompileUnits[i]);
+}
 
-  // Create DIEs for each of the externally visible subprograms.
-  bool subprogramDIEs = ConstructSubprograms();
+/// EmitAbbreviations - Emit the abbreviation section.
+///
+void DwarfDebug::EmitAbbreviations() const {
+  // Check to see if it is worth the effort.
+  if (!Abbreviations.empty()) {
+    // Start the debug abbrev section.
+    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
 
-  // If there is not any debug info available for any global variables and any
-  // subprograms then there is not any debug info to emit.
-  if (!globalDIEs && !subprogramDIEs) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    EmitLabel("abbrev_begin", 0);
 
-    return;
-  }
+    // For each abbrevation.
+    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
+      // Get abbreviation data
+      const DIEAbbrev *Abbrev = Abbreviations[i];
 
-  MMI = mmi;
-  shouldEmit = true;
-  MMI->setDebugInfoAvailability(true);
+      // Emit the abbrevations code (base 1 index.)
+      Asm->EmitULEB128Bytes(Abbrev->getNumber());
+      Asm->EOL("Abbreviation Code");
 
-  // Prime section data.
-  SectionMap.insert(TAI->getTextSection());
+      // Emit the abbreviations data.
+      Abbrev->Emit(Asm);
 
-  // Print out .file directives to specify files for .loc directives. These are
-  // printed out early so that they precede any .loc directives.
-  if (TAI->hasDotLocAndDotFile()) {
-    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
-      // Remember source id starts at 1.
-      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
-      sys::Path FullPath(getSourceDirectoryName(Id.first));
-      bool AppendOk =
-        FullPath.appendComponent(getSourceFileName(Id.second));
-      assert(AppendOk && "Could not append filename to directory!");
-      AppendOk = false;
-      Asm->EmitFile(i, FullPath.toString());
       Asm->EOL();
     }
-  }
 
-  // Emit initial sections
-  EmitInitial();
+    // Mark end of abbreviations.
+    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+    EmitLabel("abbrev_end", 0);
+    Asm->EOL();
+  }
 }
 
-/// EndModule - Emit all Dwarf sections that should come after the content.
+/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
+/// the line matrix.
 ///
-void DwarfDebug::EndModule() {
-  if (!ShouldEmitDwarfDebug())
-    return;
+void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
+  // Define last address of section.
+  Asm->EmitInt8(0); Asm->EOL("Extended Op");
+  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  // Mark end of matrix.
+  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
+  Asm->EmitULEB128Bytes(1); Asm->EOL();
+  Asm->EmitInt8(1); Asm->EOL();
+}
 
-  // Standard sections final addresses.
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_end", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_end", 0);
+/// EmitDebugLines - Emit source line information.
+///
+void DwarfDebug::EmitDebugLines() {
+  // If the target is using .loc/.file, the assembler will be emitting the
+  // .debug_line table automatically.
+  if (TAI->hasDotLocAndDotFile())
+    return;
 
-  // End text sections.
-  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
-    Asm->SwitchToSection(SectionMap[i]);
-    EmitLabel("section_end", i);
-  }
+  // Minimum line delta, thus ranging from -10..(255-10).
+  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
+  // Maximum line delta, thus ranging from -10..(255-10).
+  const int MaxLineDelta = 255 + MinLineDelta;
 
-  // Emit common frame information.
-  EmitCommonDebugFrame();
+  // Start the dwarf line section.
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
 
-  // Emit function debug frame information
-  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
-         E = DebugFrames.end(); I != E; ++I)
-    EmitFunctionDebugFrame(*I);
+  // Construct the section header.
+  EmitDifference("line_end", 0, "line_begin", 0, true);
+  Asm->EOL("Length of Source Line Info");
+  EmitLabel("line_begin", 0);
 
-  // Compute DIE offsets and sizes.
-  SizeAndOffsets();
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
 
-  // Emit all the DIEs into a debug info section
-  EmitDebugInfo();
+  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
+  Asm->EOL("Prolog Length");
+  EmitLabel("line_prolog_begin", 0);
 
-  // Corresponding abbreviations into a abbrev section.
-  EmitAbbreviations();
+  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
 
-  // Emit source line correspondence into a debug line section.
-  EmitDebugLines();
+  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
 
-  // Emit info into a debug pubnames section.
-  EmitDebugPubNames();
+  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
 
-  // Emit info into a debug str section.
-  EmitDebugStr();
+  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
 
-  // Emit info into a debug loc section.
-  EmitDebugLoc();
+  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
 
-  // Emit info into a debug aranges section.
-  EmitDebugARanges();
-
-  // Emit info into a debug ranges section.
-  EmitDebugRanges();
-
-  // Emit info into a debug macinfo section.
-  EmitDebugMacInfo();
+  // Line number standard opcode encodings argument count
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
 
-  // Emit inline info.
-  EmitDebugInlineInfo();
+  // Emit directories.
+  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
+    Asm->EmitString(getSourceDirectoryName(DI));
+    Asm->EOL("Directory");
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+  Asm->EmitInt8(0); Asm->EOL("End of directories");
 
-/// BeginFunction - Gather pre-function debug information.  Assumes being
-/// emitted immediately after the function entry point.
-void DwarfDebug::BeginFunction(MachineFunction *MF) {
-  this->MF = MF;
+  // Emit files.
+  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
+    // Remember source id starts at 1.
+    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
+    Asm->EmitString(getSourceFileName(Id.second));
+    Asm->EOL("Source");
+    Asm->EmitULEB128Bytes(Id.first);
+    Asm->EOL("Directory #");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("Mod date");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("File size");
+  }
 
-  if (!ShouldEmitDwarfDebug()) return;
+  Asm->EmitInt8(0); Asm->EOL("End of files");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("line_prolog_end", 0);
 
-  // Begin accumulating function debug information.
-  MMI->BeginFunction(MF);
+  // A sequence for each text section.
+  unsigned SecSrcLinesSize = SectionSourceLines.size();
 
-  // Assumes in correct section after the entry point.
-  EmitLabel("func_begin", ++SubprogramCount);
+  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
+    // Isolate current sections line info.
+    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
 
-  // Emit label for the implicitly defined dbg.stoppoint at the start of the
-  // function.
-  DebugLoc FDL = MF->getDefaultDebugLoc();
-  if (!FDL.isUnknown()) {
-    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
-    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
-                                        DICompileUnit(DLT.CompileUnit));
-    Asm->printLabel(LabelID);
-  }
+    if (Asm->isVerbose()) {
+      const Section* S = SectionMap[j + 1];
+      O << '\t' << TAI->getCommentString() << " Section"
+        << S->getName() << '\n';
+    } else {
+      Asm->EOL();
+    }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+    // Dwarf assumes we start with first line of first source file.
+    unsigned Source = 1;
+    unsigned Line = 1;
 
-/// EndFunction - Gather and emit post-function debug information.
-///
-void DwarfDebug::EndFunction(MachineFunction *MF) {
-  if (!ShouldEmitDwarfDebug()) return;
+    // Construct rows of the address, source, line, column matrix.
+    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
+      const SrcLineInfo &LineInfo = LineInfos[i];
+      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
+      if (!LabelID) continue;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+      if (!Asm->isVerbose())
+        Asm->EOL();
+      else {
+        std::pair<unsigned, unsigned> SourceID =
+          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
+        O << '\t' << TAI->getCommentString() << ' '
+          << getSourceDirectoryName(SourceID.first) << ' '
+          << getSourceFileName(SourceID.second)
+          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
+      }
 
-  // Define end label for subprogram.
-  EmitLabel("func_end", SubprogramCount);
+      // Define the line address.
+      Asm->EmitInt8(0); Asm->EOL("Extended Op");
+      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+      EmitReference("label",  LabelID); Asm->EOL("Location label");
 
-  // Get function line info.
-  if (!Lines.empty()) {
-    // Get section line info.
-    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
-    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
-    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
-    // Append the function info to section info.
-    SectionLineInfos.insert(SectionLineInfos.end(),
-                            Lines.begin(), Lines.end());
-  }
+      // If change of source, then switch to the new source.
+      if (Source != LineInfo.getSourceID()) {
+        Source = LineInfo.getSourceID();
+        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
+        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
+      }
 
-  // Construct the DbgScope for abstract instances.
-  for (SmallVector<DbgScope *, 32>::iterator
-         I = AbstractInstanceRootList.begin(),
-         E = AbstractInstanceRootList.end(); I != E; ++I)
-    ConstructFunctionDbgScope(*I, true);
+      // If change of line.
+      if (Line != LineInfo.getLine()) {
+        // Determine offset.
+        int Offset = LineInfo.getLine() - Line;
+        int Delta = Offset - MinLineDelta;
 
-  // Construct scopes for subprogram.
-  if (FunctionDbgScope)
-    ConstructFunctionDbgScope(FunctionDbgScope);
-  else
-    // FIXME: This is wrong. We are essentially getting past a problem with
-    // debug information not being able to handle unreachable blocks that have
-    // debug information in them. In particular, those unreachable blocks that
-    // have "region end" info in them. That situation results in the "root
-    // scope" not being created. If that's the case, then emit a "default"
-    // scope, i.e., one that encompasses the whole function. This isn't
-    // desirable. And a better way of handling this (and all of the debugging
-    // information) needs to be explored.
-    ConstructDefaultDbgScope(MF);
+        // Update line.
+        Line = LineInfo.getLine();
 
-  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
-                                               MMI->getFrameMoves()));
+        // If delta is small enough and in range...
+        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
+          // ... then use fast opcode.
+          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
+        } else {
+          // ... otherwise use long hand.
+          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
+          Asm->EOL("DW_LNS_advance_line");
+          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
+          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+        }
+      } else {
+        // Copy the previous row (different address or source)
+        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+      }
+    }
 
-  // Clear debug info
-  if (FunctionDbgScope) {
-    delete FunctionDbgScope;
-    DbgScopeMap.clear();
-    DbgAbstractScopeMap.clear();
-    DbgConcreteScopeMap.clear();
-    InlinedVariableScopes.clear();
-    FunctionDbgScope = NULL;
-    LexicalScopeStack.clear();
-    AbstractInstanceRootList.clear();
+    EmitEndOfLineMatrix(j + 1);
   }
 
-  Lines.clear();
+  if (SecSrcLinesSize == 0)
+    // Because we're emitting a debug_line section, we still need a line
+    // table. The linker and friends expect it to exist. If there's nothing to
+    // put into it, emit an empty table.
+    EmitEndOfLineMatrix(1);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitLabel("line_end", 0);
+  Asm->EOL();
 }
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
+///
+void DwarfDebug::EmitCommonDebugFrame() {
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  CompileUnit *Unit = CompileUnitMap[V];
-  assert(Unit && "Unable to find CompileUnit");
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
+  int stackGrowth =
+    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
+      TargetFrameInfo::StackGrowsUp ?
+    TD->getPointerSize() : -TD->getPointerSize();
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  return ID;
-}
+  EmitLabel("debug_frame_common", 0);
+  EmitDifference("debug_frame_common_end", 0,
+                 "debug_frame_common_begin", 0, true);
+  Asm->EOL("Length of Common Information Entry");
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
-                                      DICompileUnit CU) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("debug_frame_common_begin", 0);
+  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
+  Asm->EOL("CIE Identifier Tag");
+  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
+  Asm->EOL("CIE Version");
+  Asm->EmitString("");
+  Asm->EOL("CIE Augmentation");
+  Asm->EmitULEB128Bytes(1);
+  Asm->EOL("CIE Code Alignment Factor");
+  Asm->EmitSLEB128Bytes(stackGrowth);
+  Asm->EOL("CIE Data Alignment Factor");
+  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
+  Asm->EOL("CIE RA Column");
 
-  std::string Dir, Fn;
-  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
-                                     CU.getFilename(Fn));
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
+  std::vector<MachineMove> Moves;
+  RI->getInitialFrameState(Moves);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitFrameMoves(NULL, 0, Moves, false);
 
-  return ID;
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_common_end", 0);
+
+  Asm->EOL();
 }
 
-/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
-/// timed. Look up the source id with the given directory and source file
-/// names. If none currently exists, create a new id and insert it in the
-/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
-/// well.
-unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
+/// section.
+void
+DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
+                 "debug_frame_begin", DebugFrameInfo.Number, true);
+  Asm->EOL("Length of Frame Information Entry");
 
-  return SrcId;
-}
+  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
 
-/// RecordRegionStart - Indicate the start of a region.
-unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitSectionOffset("debug_frame_common", "section_debug_frame",
+                    0, 0, true, false);
+  Asm->EOL("FDE CIE offset");
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
-  LexicalScopeStack.push_back(Scope);
+  EmitReference("func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE initial location");
+  EmitDifference("func_end", DebugFrameInfo.Number,
+                 "func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE address range");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
+                 false);
 
-  return ID;
-}
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
 
-/// RecordRegionEnd - Indicate the end of a region.
-unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EOL();
+}
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  Scope->setEndLabelID(ID);
-  if (LexicalScopeStack.size() != 0)
-    LexicalScopeStack.pop_back();
+void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
+  EmitDifference("pubnames_end", Unit->getID(),
+                 "pubnames_begin", Unit->getID(), true);
+  Asm->EOL("Length of Public Names Info");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitLabel("pubnames_begin", Unit->getID());
 
-  return ID;
-}
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
 
-/// RecordVariable - Indicate the declaration of a local variable.
-void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
-                                const MachineInstr *MI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitSectionOffset("info_begin", "section_info",
+                    Unit->getID(), 0, true, false);
+  Asm->EOL("Offset of Compilation Unit Info");
 
-  DIDescriptor Desc(GV);
-  DbgScope *Scope = NULL;
-  bool InlinedFnVar = false;
+  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
+                 true);
+  Asm->EOL("Compilation Unit Length");
 
-  if (Desc.getTag() == dwarf::DW_TAG_variable) {
-    // GV is a global variable.
-    DIGlobalVariable DG(GV);
-    Scope = getOrCreateScope(DG.getContext().getGV());
-  } else {
-    DenseMap<const MachineInstr *, DbgScope *>::iterator
-      SI = InlinedVariableScopes.find(MI);
+  StringMap<DIE*> &Globals = Unit->getGlobals();
+  for (StringMap<DIE*>::const_iterator
+         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
+    const char *Name = GI->getKeyData();
+    DIE * Entity = GI->second;
 
-    if (SI != InlinedVariableScopes.end()) {
-      // or GV is an inlined local variable.
-      Scope = SI->second;
-    } else {
-      DIVariable DV(GV);
-      GlobalVariable *V = DV.getContext().getGV();
+    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
+    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
+  }
 
-      // FIXME: The code that checks for the inlined local variable is a hack!
-      DenseMap<const GlobalVariable *, DbgScope *>::iterator
-        AI = AbstractInstanceRootMap.find(V);
+  Asm->EmitInt32(0); Asm->EOL("End Mark");
+  EmitLabel("pubnames_end", Unit->getID());
 
-      if (AI != AbstractInstanceRootMap.end()) {
-        // This method is called each time a DECLARE node is encountered. For an
-        // inlined function, this could be many, many times. We don't want to
-        // re-add variables to that DIE for each time. We just want to add them
-        // once. Check to make sure that we haven't added them already.
-        DenseMap<const GlobalVariable *,
-          SmallSet<const GlobalVariable *, 32> >::iterator
-          IP = InlinedParamMap.find(V);
+  Asm->EOL();
+}
 
-        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
-          if (TimePassesIsEnabled)
-            DebugTimer->stopTimer();
-          return;
-        }
+/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
+///
+void DwarfDebug::EmitDebugPubNames() {
+  // Start the dwarf pubnames section.
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
 
-        // or GV is an inlined local variable.
-        Scope = AI->second;
-        InlinedParamMap[V].insert(GV);
-        InlinedFnVar = true;
-      } else {
-        // or GV is a local variable.
-        Scope = getOrCreateScope(V);
-      }
-    }
+  if (MainCU) {
+    EmitDebugPubNamesPerCU(MainCU);
+    return;
   }
 
-  assert(Scope && "Unable to find the variable's scope");
-  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
-  Scope->AddVariable(DV);
-
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugPubNamesPerCU(CompileUnits[i]);
 }
 
-/// RecordInlinedFnStart - Indicate the start of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
-                                          unsigned Line, unsigned Col) {
-  unsigned LabelID = MMI->NextLabelID();
-
-  if (!TAI->doesDwarfUsesInlineInfoSection())
-    return LabelID;
+/// EmitDebugStr - Emit visible names into a debug str section.
+///
+void DwarfDebug::EmitDebugStr() {
+  // Check to see if it is worth the effort.
+  if (!StringPool.empty()) {
+    // Start the dwarf str section.
+    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    // For each of strings in the string pool.
+    for (unsigned StringID = 1, N = StringPool.size();
+         StringID <= N; ++StringID) {
+      // Emit a label for reference from debug information entries.
+      EmitLabel("string", StringID);
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<const GlobalVariable *, DbgScope *>::iterator
-    II = AbstractInstanceRootMap.find(GV);
+      // Emit the string itself.
+      const std::string &String = StringPool[StringID];
+      Asm->EmitString(String); Asm->EOL();
+    }
 
-  if (II == AbstractInstanceRootMap.end()) {
-    // Create an abstract instance entry for this inlined function if it doesn't
-    // already exist.
-    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
+    Asm->EOL();
+  }
+}
 
-    // Get the compile unit context.
-    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-    DIE *SPDie = Unit->getDieMapSlotFor(GV);
-    if (!SPDie)
-      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
+/// EmitDebugLoc - Emit visible names into a debug loc section.
+///
+void DwarfDebug::EmitDebugLoc() {
+  // Start the dwarf loc section.
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  Asm->EOL();
+}
 
-    // Mark as being inlined. This makes this subprogram entry an abstract
-    // instance root.
-    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
-    // that it's defined. That probably won't change in the future. However,
-    // this could be more elegant.
-    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
+/// EmitDebugARanges - Emit visible names into a debug aranges section.
+///
+void DwarfDebug::EmitDebugARanges() {
+  // Start the dwarf aranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
 
-    // Keep track of the abstract scope for this function.
-    DbgAbstractScopeMap[GV] = Scope;
+  // FIXME - Mock up
+#if 0
+  CompileUnit *Unit = GetBaseCompileUnit();
 
-    AbstractInstanceRootMap[GV] = Scope;
-    AbstractInstanceRootList.push_back(Scope);
-  }
+  // Don't include size of length
+  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
 
-  // Create a concrete inlined instance for this inlined function.
-  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
-  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
-  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-  ScopeDie->setAbstractCompileUnit(Unit);
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
 
-  DIE *Origin = Unit->getDieMapSlotFor(GV);
-  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
-              dwarf::DW_FORM_ref4, Origin);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
-  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
+  EmitReference("info_begin", Unit->getID());
+  Asm->EOL("Offset of Compilation Unit Info");
 
-  ConcreteScope->setDie(ScopeDie);
-  ConcreteScope->setStartLabelID(LabelID);
-  MMI->RecordUsedDbgLabel(LabelID);
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
 
-  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
-  LexicalScopeStack.push_back(ConcreteScope);
+  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
 
-  // Keep track of the concrete scope that's inlined into this function.
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    SI = DbgConcreteScopeMap.find(GV);
+  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
+  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
 
-  if (SI == DbgConcreteScopeMap.end())
-    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
-  else
-    SI->second.push_back(ConcreteScope);
+  // Range 1
+  EmitReference("text_begin", 0); Asm->EOL("Address");
+  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
 
-  // Track the start label for this inlined function.
-  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-    I = InlineInfo.find(GV);
+  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
+  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
+#endif
 
-  if (I == InlineInfo.end())
-    InlineInfo[GV].push_back(LabelID);
-  else
-    I->second.push_back(LabelID);
+  Asm->EOL();
+}
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+/// EmitDebugRanges - Emit visible names into a debug ranges section.
+///
+void DwarfDebug::EmitDebugRanges() {
+  // Start the dwarf ranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  Asm->EOL();
+}
 
-  return LabelID;
+/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
+///
+void DwarfDebug::EmitDebugMacInfo() {
+  if (TAI->doesSupportMacInfoSection()) {
+    // Start the dwarf macinfo section.
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    Asm->EOL();
+  }
 }
 
-/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
+/// EmitDebugInlineInfo - Emit inline info using following format.
+/// Section Header:
+/// 1. length of section
+/// 2. Dwarf version number
+/// 3. address size.
+///
+/// Entries (one "entry" for each function that was inlined):
+///
+/// 1. offset into __debug_str section for MIPS linkage name, if exists;
+///   otherwise offset into __debug_str for regular function name.
+/// 2. offset into __debug_str section for regular function name.
+/// 3. an unsigned LEB128 number indicating the number of distinct inlining
+/// instances for the function.
+///
+/// The rest of the entry consists of a {die_offset, low_pc} pair for each
+/// inlined instance; the die_offset points to the inlined_subroutine die in the
+/// __debug_info section, and the low_pc is the starting address for the
+/// inlining instance.
+void DwarfDebug::EmitDebugInlineInfo() {
   if (!TAI->doesDwarfUsesInlineInfoSection())
-    return 0;
+    return;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  if (!MainCU)
+    return;
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    I = DbgConcreteScopeMap.find(GV);
+  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
+  Asm->EOL();
+  EmitDifference("debug_inlined_end", 1,
+                 "debug_inlined_begin", 1, true);
+  Asm->EOL("Length of Debug Inlined Information Entry");
 
-  if (I == DbgConcreteScopeMap.end()) {
-    // FIXME: Can this situation actually happen? And if so, should it?
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_begin", 1);
 
-    return 0;
-  }
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
 
-  SmallVector<DbgScope *, 8> &Scopes = I->second;
-  assert(!Scopes.empty() && "We should have at least one debug scope!");
-  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
-  unsigned ID = MMI->NextLabelID();
-  MMI->RecordUsedDbgLabel(ID);
-  Scope->setEndLabelID(ID);
-  LexicalScopeStack.pop_back();
+  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
+    GlobalVariable *GV = I->first;
+    SmallVector<unsigned, 4> &Labels = I->second;
+    DISubprogram SP(GV);
+    std::string Name;
+    std::string LName;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+    SP.getLinkageName(LName);
+    SP.getName(Name);
 
-  return ID;
-}
+    Asm->EmitString(LName.empty() ? Name : LName);
+    Asm->EOL("MIPS linkage name");
 
-/// RecordVariableScope - Record scope for the variable declared by
-/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
-/// for only inlined subroutine variables. Other variables's scopes are
-/// determined during RecordVariable().
-void DwarfDebug::RecordVariableScope(DIVariable &DV,
-                                     const MachineInstr *DeclareMI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    Asm->EmitString(Name); Asm->EOL("Function name");
 
-  DISubprogram SP(DV.getContext().getGV());
+    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
 
-  if (SP.isNull()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
+           LE = Labels.end(); LI != LE; ++LI) {
+      DIE *SP = MainCU->getDieMapSlotFor(GV);
+      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
 
-    return;
-  }
+      if (TD->getPointerSize() == sizeof(int32_t))
+        O << TAI->getData32bitsDirective();
+      else
+        O << TAI->getData64bitsDirective();
 
-  DenseMap<GlobalVariable *, DbgScope *>::iterator
-    I = DbgAbstractScopeMap.find(SP.getGV());
-  if (I != DbgAbstractScopeMap.end())
-    InlinedVariableScopes[DeclareMI] = I->second;
+      PrintLabelName("label", *LI); Asm->EOL("low_pc");
+    }
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_end", 1);
+  Asm->EOL();
 }
 
-//===----------------------------------------------------------------------===//
-// Emit Methods
-//===----------------------------------------------------------------------===//
-
-/// SizeAndOffsetDie - Compute the size and offset of a DIE.
-///
-unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
-  // Get the children.
-  const std::vector<DIE *> &Children = Die->getChildren();
-
-  // If not last sibling and has children then add sibling offset attribute.
-  if (!Last && !Children.empty()) Die->AddSiblingOffset();
+/// GetOrCreateSourceID - Look up the source id with the given directory and
+/// source file names. If none currently exists, create a new id and insert it
+/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
+/// maps as well.
+unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  unsigned DId;
+  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
+  if (DI != DirectoryIdMap.end()) {
+    DId = DI->getValue();
+  } else {
+    DId = DirectoryNames.size() + 1;
+    DirectoryIdMap[DirName] = DId;
+    DirectoryNames.push_back(DirName);
+  }
 
-  // Record the abbreviation.
-  AssignAbbrevNumber(Die->getAbbrev());
+  unsigned FId;
+  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
+  if (FI != SourceFileIdMap.end()) {
+    FId = FI->getValue();
+  } else {
+    FId = SourceFileNames.size() + 1;
+    SourceFileIdMap[FileName] = FId;
+    SourceFileNames.push_back(FileName);
+  }
 
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
+    SourceIdMap.find(std::make_pair(DId, FId));
+  if (SI != SourceIdMap.end())
+    return SI->second;
 
-  // Set DIE offset
-  Die->setOffset(Offset);
+  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
+  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
+  SourceIds.push_back(std::make_pair(DId, FId));
 
-  // Start the size with the size of abbreviation code.
-  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
+  return SrcId;
+}
 
-  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
+  DICompileUnit DIUnit(GV);
+  std::string Dir, FN, Prod;
+  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
+                                    DIUnit.getFilename(FN));
 
-  // Size the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i)
-    // Size attribute value.
-    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
+  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
+  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
+                   DWLabel("section_line", 0), DWLabel("section_line", 0),
+                   false);
+  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
+            DIUnit.getProducer(Prod));
+  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
+          DIUnit.getLanguage());
+  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
 
-  // Size the DIE children if any.
-  if (!Children.empty()) {
-    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
-           "Children flag not set");
+  if (!Dir.empty())
+    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
+  if (DIUnit.isOptimized())
+    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
 
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
+  std::string Flags;
+  DIUnit.getFlags(Flags);
+  if (!Flags.empty())
+    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
 
-    // End of children marker.
-    Offset += sizeof(int8_t);
-  }
+  unsigned RVer = DIUnit.getRunTimeVersion();
+  if (RVer)
+    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
+            dwarf::DW_FORM_data1, RVer);
 
-  Die->setSize(Offset - Die->getOffset());
-  return Offset;
-}
+  CompileUnit *Unit = new CompileUnit(ID, Die);
+  if (DIUnit.isMain()) {
+    assert(!MainCU && "Multiple main compile units are found!");
+    MainCU = Unit;
+    }
 
-/// SizeAndOffsets - Compute the size and offset of all the DIEs.
-///
-void DwarfDebug::SizeAndOffsets() {
-  // Compute size of compile unit header.
-  static unsigned Offset =
-    sizeof(int32_t) + // Length of Compilation Unit Info
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t);   // Pointer Size (in bytes)
+  CompileUnitMap[DIUnit.getGV()] = Unit;
+  CompileUnits.push_back(Unit);
+}
 
-  // Process base compile unit.
-  if (MainCU) {
-    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
-    CompileUnitOffsets[MainCU] = 0;
+/// ConstructCompileUnits - Create a compile unit DIEs.
+void DwarfDebug::ConstructCompileUnits() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
+  if (!Root)
     return;
-  }
-
-  // Process all compile units.
-  unsigned PrevOffset = 0;
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed compile unit descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed compile unit descriptor anchor type");
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
-    CompileUnit *Unit = CompileUnits[i];
-    CompileUnitOffsets[Unit] = PrevOffset;
-    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
-      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
-  }
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI) {
+      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
+      ConstructCompileUnit(GV);
+    }
 }
 
-/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
-/// tools to recognize the object file contains Dwarf information.
-void DwarfDebug::EmitInitial() {
-  // Check to see if we already emitted intial headers.
-  if (didInitial) return;
-  didInitial = true;
+bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
+  DIGlobalVariable DI_GV(GV);
+  CompileUnit *DW_Unit = MainCU;
+  if (!DW_Unit)
+    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
 
-  // Dwarf sections base addresses.
-  if (TAI->doesDwarfRequireFrameSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
-    EmitLabel("section_debug_frame", 0);
-  }
+  // Check for pre-existence.
+  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
+  if (Slot)
+    return false;
 
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
-  EmitLabel("section_info", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
-  EmitLabel("section_abbrev", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
-  EmitLabel("section_aranges", 0);
+  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
 
-  if (TAI->doesSupportMacInfoSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    EmitLabel("section_macinfo", 0);
-  }
+  // Add address.
+  DIEBlock *Block = new DIEBlock();
+  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
+  std::string GLN;
+  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
+                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
+  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
 
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
-  EmitLabel("section_line", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  EmitLabel("section_loc", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
-  EmitLabel("section_pubnames", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
-  EmitLabel("section_str", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  EmitLabel("section_ranges", 0);
+  // Add to map.
+  Slot = VariableDie;
 
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_begin", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_begin", 0);
-}
+  // Add to context owner.
+  DW_Unit->getDie()->AddChild(VariableDie);
 
-/// EmitDIE - Recusively Emits a debug information entry.
-///
-void DwarfDebug::EmitDIE(DIE *Die) {
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+  // Expose as global. FIXME - need to check external flag.
+  std::string Name;
+  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
+  return true;
+}
 
-  Asm->EOL();
+/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
+/// global variables. Return true if at least one global DIE is created.
+bool DwarfDebug::ConstructGlobalVariableDIEs() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
+  if (!Root)
+    return false;
 
-  // Emit the code (index) for the abbreviation.
-  Asm->EmitULEB128Bytes(AbbrevNumber);
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed global variable descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed global variable descriptor anchor type");
 
-  if (Asm->isVerbose())
-    Asm->EOL(std::string("Abbrev [" +
-                         utostr(AbbrevNumber) +
-                         "] 0x" + utohexstr(Die->getOffset()) +
-                         ":0x" + utohexstr(Die->getSize()) + " " +
-                         dwarf::TagString(Abbrev->getTag())));
-  else
-    Asm->EOL();
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
 
-  SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+  return Result;
+}
 
-  // Emit the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
-    unsigned Attr = AbbrevData[i].getAttribute();
-    unsigned Form = AbbrevData[i].getForm();
-    assert(Form && "Too many attributes for DIE (check abbreviation)");
+bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
+  DISubprogram SP(GV);
+  CompileUnit *Unit = MainCU;
+  if (!Unit)
+    Unit = &FindCompileUnit(SP.getCompileUnit());
 
-    switch (Attr) {
-    case dwarf::DW_AT_sibling:
-      Asm->EmitInt32(Die->SiblingOffset());
-      break;
-    case dwarf::DW_AT_abstract_origin: {
-      DIEEntry *E = cast<DIEEntry>(Values[i]);
-      DIE *Origin = E->getEntry();
-      unsigned Addr =
-        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
-        Origin->getOffset();
+  // Check for pre-existence.
+  DIE *&Slot = Unit->getDieMapSlotFor(GV);
+  if (Slot)
+    return false;
 
-      Asm->EmitInt32(Addr);
-      break;
-    }
-    default:
-      // Emit an attribute using the defined form.
-      Values[i]->EmitValue(this, Form);
-      break;
-    }
+  if (!SP.isDefinition())
+    // This is a method declaration which will be handled while constructing
+    // class type.
+    return false;
 
-    Asm->EOL(dwarf::AttributeString(Attr));
-  }
+  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
 
-  // Emit the DIE children if any.
-  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
-    const std::vector<DIE *> &Children = Die->getChildren();
+  // Add to map.
+  Slot = SubprogramDie;
 
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      EmitDIE(Children[j]);
+  // Add to context owner.
+  Unit->getDie()->AddChild(SubprogramDie);
 
-    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
-  }
+  // Expose as global.
+  std::string Name;
+  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
+  return true;
 }
 
-/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
-///
-void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
-  DIE *Die = Unit->getDie();
+/// ConstructSubprograms - Create DIEs for each of the externally visible
+/// subprograms. Return true if at least one subprogram DIE is created.
+bool DwarfDebug::ConstructSubprograms() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
+  if (!Root)
+    return false;
 
-  // Emit the compile units header.
-  EmitLabel("info_begin", Unit->getID());
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed subprogram descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed subprogram descriptor anchor type");
 
-  // Emit size of content not including length itself
-  unsigned ContentSize = Die->getSize() +
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t) +  // Pointer Size (in bytes)
-    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
 
-  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
-  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
-  Asm->EOL("Offset Into Abbrev. Section");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+  return Result;
+}
 
-  EmitDIE(Die);
-  // FIXME - extra padding for gdb bug.
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  EmitLabel("info_end", Unit->getID());
+/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
+/// This is inovked by the target AsmPrinter.
+void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  Asm->EOL();
-}
+  // Create all the compile unit DIEs.
+  ConstructCompileUnits();
 
-void DwarfDebug::EmitDebugInfo() {
-  // Start debug info section.
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+  if (CompileUnits.empty()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  if (MainCU) {
-    EmitDebugInfoPerCU(MainCU);
     return;
   }
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugInfoPerCU(CompileUnits[i]);
-}
+  // Create DIEs for each of the externally visible global variables.
+  bool globalDIEs = ConstructGlobalVariableDIEs();
 
-/// EmitAbbreviations - Emit the abbreviation section.
-///
-void DwarfDebug::EmitAbbreviations() const {
-  // Check to see if it is worth the effort.
-  if (!Abbreviations.empty()) {
-    // Start the debug abbrev section.
-    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
+  // Create DIEs for each of the externally visible subprograms.
+  bool subprogramDIEs = ConstructSubprograms();
 
-    EmitLabel("abbrev_begin", 0);
+  // If there is not any debug info available for any global variables and any
+  // subprograms then there is not any debug info to emit.
+  if (!globalDIEs && !subprogramDIEs) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-    // For each abbrevation.
-    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
-      // Get abbreviation data
-      const DIEAbbrev *Abbrev = Abbreviations[i];
+    return;
+  }
 
-      // Emit the abbrevations code (base 1 index.)
-      Asm->EmitULEB128Bytes(Abbrev->getNumber());
-      Asm->EOL("Abbreviation Code");
+  MMI = mmi;
+  shouldEmit = true;
+  MMI->setDebugInfoAvailability(true);
 
-      // Emit the abbreviations data.
-      Abbrev->Emit(Asm);
+  // Prime section data.
+  SectionMap.insert(TAI->getTextSection());
 
+  // Print out .file directives to specify files for .loc directives. These are
+  // printed out early so that they precede any .loc directives.
+  if (TAI->hasDotLocAndDotFile()) {
+    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
+      // Remember source id starts at 1.
+      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
+      sys::Path FullPath(getSourceDirectoryName(Id.first));
+      bool AppendOk =
+        FullPath.appendComponent(getSourceFileName(Id.second));
+      assert(AppendOk && "Could not append filename to directory!");
+      AppendOk = false;
+      Asm->EmitFile(i, FullPath.toString());
       Asm->EOL();
     }
-
-    // Mark end of abbreviations.
-    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
-
-    EmitLabel("abbrev_end", 0);
-    Asm->EOL();
   }
-}
 
-/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
-/// the line matrix.
-///
-void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
-  // Define last address of section.
-  Asm->EmitInt8(0); Asm->EOL("Extended Op");
-  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
+  // Emit initial sections
+  EmitInitial();
 
-  // Mark end of matrix.
-  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
-  Asm->EmitULEB128Bytes(1); Asm->EOL();
-  Asm->EmitInt8(1); Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitDebugLines - Emit source line information.
+/// EndModule - Emit all Dwarf sections that should come after the content.
 ///
-void DwarfDebug::EmitDebugLines() {
-  // If the target is using .loc/.file, the assembler will be emitting the
-  // .debug_line table automatically.
-  if (TAI->hasDotLocAndDotFile())
+void DwarfDebug::EndModule() {
+  if (!ShouldEmitDwarfDebug())
     return;
 
-  // Minimum line delta, thus ranging from -10..(255-10).
-  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
-  // Maximum line delta, thus ranging from -10..(255-10).
-  const int MaxLineDelta = 255 + MinLineDelta;
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  // Start the dwarf line section.
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
+  // Standard sections final addresses.
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_end", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_end", 0);
 
-  // Construct the section header.
-  EmitDifference("line_end", 0, "line_begin", 0, true);
-  Asm->EOL("Length of Source Line Info");
-  EmitLabel("line_begin", 0);
+  // End text sections.
+  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
+    Asm->SwitchToSection(SectionMap[i]);
+    EmitLabel("section_end", i);
+  }
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
+  // Emit common frame information.
+  EmitCommonDebugFrame();
 
-  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
-  Asm->EOL("Prolog Length");
-  EmitLabel("line_prolog_begin", 0);
+  // Emit function debug frame information
+  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
+         E = DebugFrames.end(); I != E; ++I)
+    EmitFunctionDebugFrame(*I);
 
-  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
+  // Compute DIE offsets and sizes.
+  SizeAndOffsets();
 
-  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
+  // Emit all the DIEs into a debug info section
+  EmitDebugInfo();
 
-  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
+  // Corresponding abbreviations into a abbrev section.
+  EmitAbbreviations();
 
-  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
+  // Emit source line correspondence into a debug line section.
+  EmitDebugLines();
 
-  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
+  // Emit info into a debug pubnames section.
+  EmitDebugPubNames();
 
-  // Line number standard opcode encodings argument count
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
+  // Emit info into a debug str section.
+  EmitDebugStr();
 
-  // Emit directories.
-  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
-    Asm->EmitString(getSourceDirectoryName(DI));
-    Asm->EOL("Directory");
-  }
+  // Emit info into a debug loc section.
+  EmitDebugLoc();
 
-  Asm->EmitInt8(0); Asm->EOL("End of directories");
+  // Emit info into a debug aranges section.
+  EmitDebugARanges();
 
-  // Emit files.
-  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
-    // Remember source id starts at 1.
-    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
-    Asm->EmitString(getSourceFileName(Id.second));
-    Asm->EOL("Source");
-    Asm->EmitULEB128Bytes(Id.first);
-    Asm->EOL("Directory #");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("Mod date");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("File size");
-  }
+  // Emit info into a debug ranges section.
+  EmitDebugRanges();
 
-  Asm->EmitInt8(0); Asm->EOL("End of files");
+  // Emit info into a debug macinfo section.
+  EmitDebugMacInfo();
 
-  EmitLabel("line_prolog_end", 0);
+  // Emit inline info.
+  EmitDebugInlineInfo();
 
-  // A sequence for each text section.
-  unsigned SecSrcLinesSize = SectionSourceLines.size();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+}
 
-  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
-    // Isolate current sections line info.
-    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
+/// BeginFunction - Gather pre-function debug information.  Assumes being
+/// emitted immediately after the function entry point.
+void DwarfDebug::BeginFunction(MachineFunction *MF) {
+  this->MF = MF;
 
-    if (Asm->isVerbose()) {
-      const Section* S = SectionMap[j + 1];
-      O << '\t' << TAI->getCommentString() << " Section"
-        << S->getName() << '\n';
-    } else {
-      Asm->EOL();
-    }
+  if (!ShouldEmitDwarfDebug()) return;
 
-    // Dwarf assumes we start with first line of first source file.
-    unsigned Source = 1;
-    unsigned Line = 1;
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    // Construct rows of the address, source, line, column matrix.
-    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
-      const SrcLineInfo &LineInfo = LineInfos[i];
-      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
-      if (!LabelID) continue;
+  // Begin accumulating function debug information.
+  MMI->BeginFunction(MF);
 
-      if (!Asm->isVerbose())
-        Asm->EOL();
-      else {
-        std::pair<unsigned, unsigned> SourceID =
-          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
-        O << '\t' << TAI->getCommentString() << ' '
-          << getSourceDirectoryName(SourceID.first) << ' '
-          << getSourceFileName(SourceID.second)
-          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
-      }
+  // Assumes in correct section after the entry point.
+  EmitLabel("func_begin", ++SubprogramCount);
 
-      // Define the line address.
-      Asm->EmitInt8(0); Asm->EOL("Extended Op");
-      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-      EmitReference("label",  LabelID); Asm->EOL("Location label");
+  // Emit label for the implicitly defined dbg.stoppoint at the start of the
+  // function.
+  DebugLoc FDL = MF->getDefaultDebugLoc();
+  if (!FDL.isUnknown()) {
+    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
+    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
+                                        DICompileUnit(DLT.CompileUnit));
+    Asm->printLabel(LabelID);
+  }
 
-      // If change of source, then switch to the new source.
-      if (Source != LineInfo.getSourceID()) {
-        Source = LineInfo.getSourceID();
-        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
-        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
-      }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+}
+
+/// EndFunction - Gather and emit post-function debug information.
+///
+void DwarfDebug::EndFunction(MachineFunction *MF) {
+  if (!ShouldEmitDwarfDebug()) return;
+
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
+
+  // Define end label for subprogram.
+  EmitLabel("func_end", SubprogramCount);
+
+  // Get function line info.
+  if (!Lines.empty()) {
+    // Get section line info.
+    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
+    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
+    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
+    // Append the function info to section info.
+    SectionLineInfos.insert(SectionLineInfos.end(),
+                            Lines.begin(), Lines.end());
+  }
 
-      // If change of line.
-      if (Line != LineInfo.getLine()) {
-        // Determine offset.
-        int Offset = LineInfo.getLine() - Line;
-        int Delta = Offset - MinLineDelta;
+  // Construct the DbgScope for abstract instances.
+  for (SmallVector<DbgScope *, 32>::iterator
+         I = AbstractInstanceRootList.begin(),
+         E = AbstractInstanceRootList.end(); I != E; ++I)
+    ConstructAbstractDbgScope(*I);
 
-        // Update line.
-        Line = LineInfo.getLine();
+  // Construct scopes for subprogram.
+  if (FunctionDbgScope)
+    ConstructFunctionDbgScope(FunctionDbgScope);
+  else
+    // FIXME: This is wrong. We are essentially getting past a problem with
+    // debug information not being able to handle unreachable blocks that have
+    // debug information in them. In particular, those unreachable blocks that
+    // have "region end" info in them. That situation results in the "root
+    // scope" not being created. If that's the case, then emit a "default"
+    // scope, i.e., one that encompasses the whole function. This isn't
+    // desirable. And a better way of handling this (and all of the debugging
+    // information) needs to be explored.
+    ConstructDefaultDbgScope(MF);
 
-        // If delta is small enough and in range...
-        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
-          // ... then use fast opcode.
-          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
-        } else {
-          // ... otherwise use long hand.
-          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
-          Asm->EOL("DW_LNS_advance_line");
-          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
-          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-        }
-      } else {
-        // Copy the previous row (different address or source)
-        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-      }
-    }
+  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
+                                               MMI->getFrameMoves()));
 
-    EmitEndOfLineMatrix(j + 1);
+  // Clear debug info
+  if (FunctionDbgScope) {
+    delete FunctionDbgScope;
+    DbgScopeMap.clear();
+    DbgAbstractScopeMap.clear();
+    DbgConcreteScopeMap.clear();
+    InlinedVariableScopes.clear();
+    FunctionDbgScope = NULL;
+    LexicalScopeStack.clear();
+    AbstractInstanceRootList.clear();
   }
 
-  if (SecSrcLinesSize == 0)
-    // Because we're emitting a debug_line section, we still need a line
-    // table. The linker and friends expect it to exist. If there's nothing to
-    // put into it, emit an empty table.
-    EmitEndOfLineMatrix(1);
+  Lines.clear();
 
-  EmitLabel("line_end", 0);
-  Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
-///
-void DwarfDebug::EmitCommonDebugFrame() {
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
-
-  int stackGrowth =
-    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
-      TargetFrameInfo::StackGrowsUp ?
-    TD->getPointerSize() : -TD->getPointerSize();
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+  CompileUnit *Unit = CompileUnitMap[V];
+  assert(Unit && "Unable to find CompileUnit");
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
 
-  EmitLabel("debug_frame_common", 0);
-  EmitDifference("debug_frame_common_end", 0,
-                 "debug_frame_common_begin", 0, true);
-  Asm->EOL("Length of Common Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  EmitLabel("debug_frame_common_begin", 0);
-  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
-  Asm->EOL("CIE Identifier Tag");
-  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
-  Asm->EOL("CIE Version");
-  Asm->EmitString("");
-  Asm->EOL("CIE Augmentation");
-  Asm->EmitULEB128Bytes(1);
-  Asm->EOL("CIE Code Alignment Factor");
-  Asm->EmitSLEB128Bytes(stackGrowth);
-  Asm->EOL("CIE Data Alignment Factor");
-  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
-  Asm->EOL("CIE RA Column");
+  return ID;
+}
 
-  std::vector<MachineMove> Moves;
-  RI->getInitialFrameState(Moves);
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
+                                      DICompileUnit CU) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitFrameMoves(NULL, 0, Moves, false);
+  std::string Dir, Fn;
+  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
+                                     CU.getFilename(Fn));
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
 
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_common_end", 0);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  Asm->EOL();
+  return ID;
 }
 
-/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
-/// section.
-void
-DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
+/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
+/// timed. Look up the source id with the given directory and source file
+/// names. If none currently exists, create a new id and insert it in the
+/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
+/// well.
+unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
 
-  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
-                 "debug_frame_begin", DebugFrameInfo.Number, true);
-  Asm->EOL("Length of Frame Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
+  return SrcId;
+}
 
-  EmitSectionOffset("debug_frame_common", "section_debug_frame",
-                    0, 0, true, false);
-  Asm->EOL("FDE CIE offset");
+/// RecordRegionStart - Indicate the start of a region.
+unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitReference("func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE initial location");
-  EmitDifference("func_end", DebugFrameInfo.Number,
-                 "func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE address range");
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
+  LexicalScopeStack.push_back(Scope);
 
-  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
-                 false);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
+  return ID;
+}
 
-  Asm->EOL();
+/// RecordRegionEnd - Indicate the end of a region.
+unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
+
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  Scope->setEndLabelID(ID);
+  if (LexicalScopeStack.size() != 0)
+    LexicalScopeStack.pop_back();
+
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+
+  return ID;
 }
 
-void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
-  EmitDifference("pubnames_end", Unit->getID(),
-                 "pubnames_begin", Unit->getID(), true);
-  Asm->EOL("Length of Public Names Info");
+/// RecordVariable - Indicate the declaration of a local variable.
+void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
+                                const MachineInstr *MI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitLabel("pubnames_begin", Unit->getID());
+  DIDescriptor Desc(GV);
+  DbgScope *Scope = NULL;
+  bool InlinedFnVar = false;
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
+  if (Desc.getTag() == dwarf::DW_TAG_variable) {
+    // GV is a global variable.
+    DIGlobalVariable DG(GV);
+    Scope = getOrCreateScope(DG.getContext().getGV());
+  } else {
+    DenseMap<const MachineInstr *, DbgScope *>::iterator
+      SI = InlinedVariableScopes.find(MI);
 
-  EmitSectionOffset("info_begin", "section_info",
-                    Unit->getID(), 0, true, false);
-  Asm->EOL("Offset of Compilation Unit Info");
+    if (SI != InlinedVariableScopes.end()) {
+      // or GV is an inlined local variable.
+      Scope = SI->second;
+    } else {
+      DIVariable DV(GV);
+      GlobalVariable *V = DV.getContext().getGV();
 
-  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
-                 true);
-  Asm->EOL("Compilation Unit Length");
+      // FIXME: The code that checks for the inlined local variable is a hack!
+      DenseMap<const GlobalVariable *, DbgScope *>::iterator
+        AI = AbstractInstanceRootMap.find(V);
 
-  StringMap<DIE*> &Globals = Unit->getGlobals();
-  for (StringMap<DIE*>::const_iterator
-         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
-    const char *Name = GI->getKeyData();
-    DIE * Entity = GI->second;
+      if (AI != AbstractInstanceRootMap.end()) {
+        // This method is called each time a DECLARE node is encountered. For an
+        // inlined function, this could be many, many times. We don't want to
+        // re-add variables to that DIE for each time. We just want to add them
+        // once. Check to make sure that we haven't added them already.
+        DenseMap<const GlobalVariable *,
+          SmallSet<const GlobalVariable *, 32> >::iterator
+          IP = InlinedParamMap.find(V);
 
-    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
-    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
+        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
+          if (TimePassesIsEnabled)
+            DebugTimer->stopTimer();
+          return;
+        }
+
+        // or GV is an inlined local variable.
+        Scope = AI->second;
+        InlinedParamMap[V].insert(GV);
+        InlinedFnVar = true;
+      } else {
+        // or GV is a local variable.
+        Scope = getOrCreateScope(V);
+      }
+    }
   }
 
-  Asm->EmitInt32(0); Asm->EOL("End Mark");
-  EmitLabel("pubnames_end", Unit->getID());
+  assert(Scope && "Unable to find the variable's scope");
+  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
+  Scope->AddVariable(DV);
 
-  Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
-///
-void DwarfDebug::EmitDebugPubNames() {
-  // Start the dwarf pubnames section.
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
-
-  if (MainCU) {
-    EmitDebugPubNamesPerCU(MainCU);
-    return;
-  }
-
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugPubNamesPerCU(CompileUnits[i]);
-}
+//// RecordInlinedFnStart - Indicate the start of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
+                                          unsigned Line, unsigned Col) {
+  unsigned LabelID = MMI->NextLabelID();
 
-/// EmitDebugStr - Emit visible names into a debug str section.
-///
-void DwarfDebug::EmitDebugStr() {
-  // Check to see if it is worth the effort.
-  if (!StringPool.empty()) {
-    // Start the dwarf str section.
-    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+  if (!TAI->doesDwarfUsesInlineInfoSection())
+    return LabelID;
 
-    // For each of strings in the string pool.
-    for (unsigned StringID = 1, N = StringPool.size();
-         StringID <= N; ++StringID) {
-      // Emit a label for reference from debug information entries.
-      EmitLabel("string", StringID);
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-      // Emit the string itself.
-      const std::string &String = StringPool[StringID];
-      Asm->EmitString(String); Asm->EOL();
-    }
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<const GlobalVariable *, DbgScope *>::iterator
+    II = AbstractInstanceRootMap.find(GV);
 
-    Asm->EOL();
-  }
-}
+  if (II == AbstractInstanceRootMap.end()) {
+    // Create an abstract instance entry for this inlined function if it doesn't
+    // already exist.
+    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
 
-/// EmitDebugLoc - Emit visible names into a debug loc section.
-///
-void DwarfDebug::EmitDebugLoc() {
-  // Start the dwarf loc section.
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  Asm->EOL();
-}
+    // Get the compile unit context.
+    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+    DIE *SPDie = Unit->getDieMapSlotFor(GV);
+    if (!SPDie)
+      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
 
-/// EmitDebugARanges - Emit visible names into a debug aranges section.
-///
-void DwarfDebug::EmitDebugARanges() {
-  // Start the dwarf aranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+    // Mark as being inlined. This makes this subprogram entry an abstract
+    // instance root.
+    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
+    // that it's defined. That probably won't change in the future. However,
+    // this could be more elegant.
+    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
 
-  // FIXME - Mock up
-#if 0
-  CompileUnit *Unit = GetBaseCompileUnit();
+    // Keep track of the abstract scope for this function.
+    DbgAbstractScopeMap[GV] = Scope;
 
-  // Don't include size of length
-  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
+    AbstractInstanceRootMap[GV] = Scope;
+    AbstractInstanceRootList.push_back(Scope);
+  }
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+  // Create a concrete inlined instance for this inlined function.
+  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
+  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
+  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+  ScopeDie->setAbstractCompileUnit(Unit);
 
-  EmitReference("info_begin", Unit->getID());
-  Asm->EOL("Offset of Compilation Unit Info");
+  DIE *Origin = Unit->getDieMapSlotFor(GV);
+  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
+              dwarf::DW_FORM_ref4, Origin);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
+  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
 
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
+  ConcreteScope->setDie(ScopeDie);
+  ConcreteScope->setStartLabelID(LabelID);
+  MMI->RecordUsedDbgLabel(LabelID);
 
-  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
+  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
 
-  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
-  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
+  // Keep track of the concrete scope that's inlined into this function.
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    SI = DbgConcreteScopeMap.find(GV);
 
-  // Range 1
-  EmitReference("text_begin", 0); Asm->EOL("Address");
-  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
+  if (SI == DbgConcreteScopeMap.end())
+    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
+  else
+    SI->second.push_back(ConcreteScope);
 
-  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
-  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
-#endif
+  // Track the start label for this inlined function.
+  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+    I = InlineInfo.find(GV);
 
-  Asm->EOL();
-}
+  if (I == InlineInfo.end())
+    InlineInfo[GV].push_back(LabelID);
+  else
+    I->second.push_back(LabelID);
 
-/// EmitDebugRanges - Emit visible names into a debug ranges section.
-///
-void DwarfDebug::EmitDebugRanges() {
-  // Start the dwarf ranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  Asm->EOL();
-}
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
-///
-void DwarfDebug::EmitDebugMacInfo() {
-  if (TAI->doesSupportMacInfoSection()) {
-    // Start the dwarf macinfo section.
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    Asm->EOL();
-  }
+  return LabelID;
 }
 
-/// EmitDebugInlineInfo - Emit inline info using following format.
-/// Section Header:
-/// 1. length of section
-/// 2. Dwarf version number
-/// 3. address size.
-///
-/// Entries (one "entry" for each function that was inlined):
-///
-/// 1. offset into __debug_str section for MIPS linkage name, if exists;
-///   otherwise offset into __debug_str for regular function name.
-/// 2. offset into __debug_str section for regular function name.
-/// 3. an unsigned LEB128 number indicating the number of distinct inlining
-/// instances for the function.
-///
-/// The rest of the entry consists of a {die_offset, low_pc} pair for each
-/// inlined instance; the die_offset points to the inlined_subroutine die in the
-/// __debug_info section, and the low_pc is the starting address for the
-/// inlining instance.
-void DwarfDebug::EmitDebugInlineInfo() {
+/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
   if (!TAI->doesDwarfUsesInlineInfoSection())
-    return;
-
-  if (!MainCU)
-    return;
+    return 0;
 
-  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
-  Asm->EOL();
-  EmitDifference("debug_inlined_end", 1,
-                 "debug_inlined_begin", 1, true);
-  Asm->EOL("Length of Debug Inlined Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitLabel("debug_inlined_begin", 1);
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    I = DbgConcreteScopeMap.find(GV);
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+  if (I == DbgConcreteScopeMap.end()) {
+    // FIXME: Can this situation actually happen? And if so, should it?
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
-    GlobalVariable *GV = I->first;
-    SmallVector<unsigned, 4> &Labels = I->second;
-    DISubprogram SP(GV);
-    std::string Name;
-    std::string LName;
+    return 0;
+  }
 
-    SP.getLinkageName(LName);
-    SP.getName(Name);
+  SmallVector<DbgScope *, 8> &Scopes = I->second;
+  assert(!Scopes.empty() && "We should have at least one debug scope!");
+  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
+  unsigned ID = MMI->NextLabelID();
+  MMI->RecordUsedDbgLabel(ID);
+  Scope->setEndLabelID(ID);
 
-    Asm->EmitString(LName.empty() ? Name : LName);
-    Asm->EOL("MIPS linkage name");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-    Asm->EmitString(Name); Asm->EOL("Function name");
+  return ID;
+}
 
-    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
+/// RecordVariableScope - Record scope for the variable declared by
+/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
+/// for only inlined subroutine variables. Other variables's scopes are
+/// determined during RecordVariable().
+void DwarfDebug::RecordVariableScope(DIVariable &DV,
+                                     const MachineInstr *DeclareMI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
-           LE = Labels.end(); LI != LE; ++LI) {
-      DIE *SP = MainCU->getDieMapSlotFor(GV);
-      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
+  DISubprogram SP(DV.getContext().getGV());
 
-      if (TD->getPointerSize() == sizeof(int32_t))
-        O << TAI->getData32bitsDirective();
-      else
-        O << TAI->getData64bitsDirective();
+  if (SP.isNull()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-      PrintLabelName("label", *LI); Asm->EOL("low_pc");
-    }
+    return;
   }
 
-  EmitLabel("debug_inlined_end", 1);
-  Asm->EOL();
+  DenseMap<GlobalVariable *, DbgScope *>::iterator
+    I = DbgAbstractScopeMap.find(SP.getGV());
+  if (I != DbgAbstractScopeMap.end())
+    InlinedVariableScopes[DeclareMI] = I->second;
+
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=72193&r1=72192&r2=72193&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Wed May 20 18:19:06 2009
@@ -230,9 +230,16 @@
   ///
   void AssignAbbrevNumber(DIEAbbrev &Abbrev);
 
-  /// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
-  /// information entry.
-  DIEEntry *CreateDIEEntry(DIE *Entry = NULL);
+  /// NewString - Add a string to the constant pool and returns a label.
+  ///
+  DWLabel NewString(const std::string &String) {
+    unsigned StringID = StringPool.insert(String);
+    return DWLabel("string", StringID);
+  }
+
+  /// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
+  /// entry.
+  DIEEntry *NewDIEEntry(DIE *Entry = NULL);
 
   /// SetDIEEntry - Set a DIEEntry once the debug information entry is defined.
   ///
@@ -275,7 +282,7 @@
   /// AddDIEEntry - Add a DIE attribute data and value.
   ///
   void AddDIEEntry(DIE *Die, unsigned Attribute, unsigned Form, DIE *Entry) {
-    Die->AddValue(Attribute, Form, CreateDIEEntry(Entry));
+    Die->AddValue(Attribute, Form, NewDIEEntry(Entry));
   }
 
   /// AddBlock - Add block data.
@@ -339,9 +346,9 @@
   ///
   CompileUnit &FindCompileUnit(DICompileUnit Unit) const;
 
-  /// CreateDbgScopeVariable - Create a new debug scope variable.
+  /// NewDbgScopeVariable - Create a new scope variable.
   ///
-  DIE *CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
+  DIE *NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
 
   /// getOrCreateScope - Returns the scope associated with the given descriptor.
   ///
@@ -355,38 +362,17 @@
 
   /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
   ///
-  void ConstructFunctionDbgScope(DbgScope *RootScope,
-                                 bool AbstractScope = false);
+  void ConstructFunctionDbgScope(DbgScope *RootScope);
+
+  /// ConstructFunctionDbgScope - Construct the scope for the abstract debug
+  /// scope.
+  ///
+  void ConstructAbstractDbgScope(DbgScope *AbsScope);
 
   /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
   ///
   void ConstructDefaultDbgScope(MachineFunction *MF);
 
-  /// GetOrCreateSourceID - Look up the source id with the given directory and
-  /// source file names. If none currently exists, create a new id and insert it
-  /// in the SourceIds map. This can update DirectoryNames and SourceFileNames maps
-  /// as well.
-  unsigned GetOrCreateSourceID(const std::string &DirName,
-                               const std::string &FileName);
-
-  void ConstructCompileUnit(GlobalVariable *GV);
-
-  /// ConstructCompileUnits - Create a compile unit DIEs.
-  void ConstructCompileUnits();
-
-  bool ConstructGlobalVariableDIE(GlobalVariable *GV);
-
-  /// ConstructGlobalVariableDIEs - Create DIEs for each of the externally 
-  /// visible global variables. Return true if at least one global DIE is
-  /// created.
-  bool ConstructGlobalVariableDIEs();
-
-  bool ConstructSubprogram(GlobalVariable *GV);
-
-  /// ConstructSubprograms - Create DIEs for each of the externally visible
-  /// subprograms. Return true if at least one subprogram DIE is created.
-  bool ConstructSubprograms();
-
   /// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
   /// tools to recognize the object file contains Dwarf information.
   void EmitInitial();
@@ -475,6 +461,31 @@
   /// the __debug_info section, and the low_pc is the starting address  for the
   ///  inlining instance.
   void EmitDebugInlineInfo();
+
+  /// GetOrCreateSourceID - Look up the source id with the given directory and
+  /// source file names. If none currently exists, create a new id and insert it
+  /// in the SourceIds map. This can update DirectoryNames and SourceFileNames maps
+  /// as well.
+  unsigned GetOrCreateSourceID(const std::string &DirName,
+                               const std::string &FileName);
+
+  void ConstructCompileUnit(GlobalVariable *GV);
+
+  /// ConstructCompileUnits - Create a compile unit DIEs.
+  void ConstructCompileUnits();
+
+  bool ConstructGlobalVariableDIE(GlobalVariable *GV);
+
+  /// ConstructGlobalVariableDIEs - Create DIEs for each of the externally 
+  /// visible global variables. Return true if at least one global DIE is
+  /// created.
+  bool ConstructGlobalVariableDIEs();
+
+  bool ConstructSubprogram(GlobalVariable *GV);
+
+  /// ConstructSubprograms - Create DIEs for each of the externally visible
+  /// subprograms. Return true if at least one subprogram DIE is created.
+  bool ConstructSubprograms();
 public:
   //===--------------------------------------------------------------------===//
   // Main entry points.




From isanbard at gmail.com  Wed May 20 18:21:42 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:21:42 -0000
Subject: [llvm-commits] [llvm] r72194 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <200905202321.n4KNLgLK029095@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:21:38 2009
New Revision: 72194

URL: http://llvm.org/viewvc/llvm-project?rev=72194&view=rev
Log:
Move 'Emit' methods down to their own place.

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=72194&r1=72193&r2=72194&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:21:38 2009
@@ -1196,1445 +1196,1445 @@
 #endif
 }
 
-/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
-/// tools to recognize the object file contains Dwarf information.
-void DwarfDebug::EmitInitial() {
-  // Check to see if we already emitted intial headers.
-  if (didInitial) return;
-  didInitial = true;
-
-  // Dwarf sections base addresses.
-  if (TAI->doesDwarfRequireFrameSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
-    EmitLabel("section_debug_frame", 0);
+/// GetOrCreateSourceID - Look up the source id with the given directory and
+/// source file names. If none currently exists, create a new id and insert it
+/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
+/// maps as well.
+unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  unsigned DId;
+  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
+  if (DI != DirectoryIdMap.end()) {
+    DId = DI->getValue();
+  } else {
+    DId = DirectoryNames.size() + 1;
+    DirectoryIdMap[DirName] = DId;
+    DirectoryNames.push_back(DirName);
   }
 
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
-  EmitLabel("section_info", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
-  EmitLabel("section_abbrev", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
-  EmitLabel("section_aranges", 0);
-
-  if (TAI->doesSupportMacInfoSection()) {
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    EmitLabel("section_macinfo", 0);
+  unsigned FId;
+  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
+  if (FI != SourceFileIdMap.end()) {
+    FId = FI->getValue();
+  } else {
+    FId = SourceFileNames.size() + 1;
+    SourceFileIdMap[FileName] = FId;
+    SourceFileNames.push_back(FileName);
   }
 
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
-  EmitLabel("section_line", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  EmitLabel("section_loc", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
-  EmitLabel("section_pubnames", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
-  EmitLabel("section_str", 0);
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  EmitLabel("section_ranges", 0);
-
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_begin", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_begin", 0);
-}
+  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
+    SourceIdMap.find(std::make_pair(DId, FId));
+  if (SI != SourceIdMap.end())
+    return SI->second;
 
-/// EmitDIE - Recusively Emits a debug information entry.
-///
-void DwarfDebug::EmitDIE(DIE *Die) {
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
+  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
+  SourceIds.push_back(std::make_pair(DId, FId));
 
-  Asm->EOL();
+  return SrcId;
+}
 
-  // Emit the code (index) for the abbreviation.
-  Asm->EmitULEB128Bytes(AbbrevNumber);
+void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
+  DICompileUnit DIUnit(GV);
+  std::string Dir, FN, Prod;
+  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
+                                    DIUnit.getFilename(FN));
 
-  if (Asm->isVerbose())
-    Asm->EOL(std::string("Abbrev [" +
-                         utostr(AbbrevNumber) +
-                         "] 0x" + utohexstr(Die->getOffset()) +
-                         ":0x" + utohexstr(Die->getSize()) + " " +
-                         dwarf::TagString(Abbrev->getTag())));
-  else
-    Asm->EOL();
+  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
+  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
+                   DWLabel("section_line", 0), DWLabel("section_line", 0),
+                   false);
+  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
+            DIUnit.getProducer(Prod));
+  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
+          DIUnit.getLanguage());
+  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
 
-  SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+  if (!Dir.empty())
+    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
+  if (DIUnit.isOptimized())
+    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
 
-  // Emit the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
-    unsigned Attr = AbbrevData[i].getAttribute();
-    unsigned Form = AbbrevData[i].getForm();
-    assert(Form && "Too many attributes for DIE (check abbreviation)");
+  std::string Flags;
+  DIUnit.getFlags(Flags);
+  if (!Flags.empty())
+    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
 
-    switch (Attr) {
-    case dwarf::DW_AT_sibling:
-      Asm->EmitInt32(Die->SiblingOffset());
-      break;
-    case dwarf::DW_AT_abstract_origin: {
-      DIEEntry *E = cast<DIEEntry>(Values[i]);
-      DIE *Origin = E->getEntry();
-      unsigned Addr =
-        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
-        Origin->getOffset();
+  unsigned RVer = DIUnit.getRunTimeVersion();
+  if (RVer)
+    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
+            dwarf::DW_FORM_data1, RVer);
 
-      Asm->EmitInt32(Addr);
-      break;
-    }
-    default:
-      // Emit an attribute using the defined form.
-      Values[i]->EmitValue(this, Form);
-      break;
+  CompileUnit *Unit = new CompileUnit(ID, Die);
+  if (DIUnit.isMain()) {
+    assert(!MainCU && "Multiple main compile units are found!");
+    MainCU = Unit;
     }
 
-    Asm->EOL(dwarf::AttributeString(Attr));
-  }
-
-  // Emit the DIE children if any.
-  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
-    const std::vector<DIE *> &Children = Die->getChildren();
-
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      EmitDIE(Children[j]);
-
-    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
-  }
+  CompileUnitMap[DIUnit.getGV()] = Unit;
+  CompileUnits.push_back(Unit);
 }
 
-/// SizeAndOffsetDie - Compute the size and offset of a DIE.
-///
-unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
-  // Get the children.
-  const std::vector<DIE *> &Children = Die->getChildren();
-
-  // If not last sibling and has children then add sibling offset attribute.
-  if (!Last && !Children.empty()) Die->AddSiblingOffset();
-
-  // Record the abbreviation.
-  AssignAbbrevNumber(Die->getAbbrev());
-
-  // Get the abbreviation for this DIE.
-  unsigned AbbrevNumber = Die->getAbbrevNumber();
-  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+/// ConstructCompileUnits - Create a compile unit DIEs.
+void DwarfDebug::ConstructCompileUnits() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
+  if (!Root)
+    return;
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed compile unit descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed compile unit descriptor anchor type");
 
-  // Set DIE offset
-  Die->setOffset(Offset);
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI) {
+      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
+      ConstructCompileUnit(GV);
+    }
+}
 
-  // Start the size with the size of abbreviation code.
-  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
+bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
+  DIGlobalVariable DI_GV(GV);
+  CompileUnit *DW_Unit = MainCU;
+  if (!DW_Unit)
+    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
 
-  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
-  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
+  // Check for pre-existence.
+  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
+  if (Slot)
+    return false;
 
-  // Size the DIE attribute values.
-  for (unsigned i = 0, N = Values.size(); i < N; ++i)
-    // Size attribute value.
-    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
+  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
 
-  // Size the DIE children if any.
-  if (!Children.empty()) {
-    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
-           "Children flag not set");
+  // Add address.
+  DIEBlock *Block = new DIEBlock();
+  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
+  std::string GLN;
+  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
+                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
+  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
 
-    for (unsigned j = 0, M = Children.size(); j < M; ++j)
-      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
+  // Add to map.
+  Slot = VariableDie;
 
-    // End of children marker.
-    Offset += sizeof(int8_t);
-  }
+  // Add to context owner.
+  DW_Unit->getDie()->AddChild(VariableDie);
 
-  Die->setSize(Offset - Die->getOffset());
-  return Offset;
+  // Expose as global. FIXME - need to check external flag.
+  std::string Name;
+  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
+  return true;
 }
 
-/// SizeAndOffsets - Compute the size and offset of all the DIEs.
-///
-void DwarfDebug::SizeAndOffsets() {
-  // Compute size of compile unit header.
-  static unsigned Offset =
-    sizeof(int32_t) + // Length of Compilation Unit Info
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t);   // Pointer Size (in bytes)
+/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
+/// global variables. Return true if at least one global DIE is created.
+bool DwarfDebug::ConstructGlobalVariableDIEs() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
+  if (!Root)
+    return false;
 
-  // Process base compile unit.
-  if (MainCU) {
-    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
-    CompileUnitOffsets[MainCU] = 0;
-    return;
-  }
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed global variable descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed global variable descriptor anchor type");
 
-  // Process all compile units.
-  unsigned PrevOffset = 0;
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
-    CompileUnit *Unit = CompileUnits[i];
-    CompileUnitOffsets[Unit] = PrevOffset;
-    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
-      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
-  }
+  return Result;
 }
 
-/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
-///
-void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
-  DIE *Die = Unit->getDie();
-
-  // Emit the compile units header.
-  EmitLabel("info_begin", Unit->getID());
-
-  // Emit size of content not including length itself
-  unsigned ContentSize = Die->getSize() +
-    sizeof(int16_t) + // DWARF version number
-    sizeof(int32_t) + // Offset Into Abbrev. Section
-    sizeof(int8_t) +  // Pointer Size (in bytes)
-    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
+  DISubprogram SP(GV);
+  CompileUnit *Unit = MainCU;
+  if (!Unit)
+    Unit = &FindCompileUnit(SP.getCompileUnit());
 
-  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
-  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
-  Asm->EOL("Offset Into Abbrev. Section");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+  // Check for pre-existence.
+  DIE *&Slot = Unit->getDieMapSlotFor(GV);
+  if (Slot)
+    return false;
 
-  EmitDIE(Die);
-  // FIXME - extra padding for gdb bug.
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
-  EmitLabel("info_end", Unit->getID());
+  if (!SP.isDefinition())
+    // This is a method declaration which will be handled while constructing
+    // class type.
+    return false;
 
-  Asm->EOL();
-}
+  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
 
-void DwarfDebug::EmitDebugInfo() {
-  // Start debug info section.
-  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+  // Add to map.
+  Slot = SubprogramDie;
 
-  if (MainCU) {
-    EmitDebugInfoPerCU(MainCU);
-    return;
-  }
+  // Add to context owner.
+  Unit->getDie()->AddChild(SubprogramDie);
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugInfoPerCU(CompileUnits[i]);
+  // Expose as global.
+  std::string Name;
+  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
+  return true;
 }
 
-/// EmitAbbreviations - Emit the abbreviation section.
-///
-void DwarfDebug::EmitAbbreviations() const {
-  // Check to see if it is worth the effort.
-  if (!Abbreviations.empty()) {
-    // Start the debug abbrev section.
-    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
-
-    EmitLabel("abbrev_begin", 0);
-
-    // For each abbrevation.
-    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
-      // Get abbreviation data
-      const DIEAbbrev *Abbrev = Abbreviations[i];
-
-      // Emit the abbrevations code (base 1 index.)
-      Asm->EmitULEB128Bytes(Abbrev->getNumber());
-      Asm->EOL("Abbreviation Code");
-
-      // Emit the abbreviations data.
-      Abbrev->Emit(Asm);
-
-      Asm->EOL();
-    }
-
-    // Mark end of abbreviations.
-    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
+/// ConstructSubprograms - Create DIEs for each of the externally visible
+/// subprograms. Return true if at least one subprogram DIE is created.
+bool DwarfDebug::ConstructSubprograms() {
+  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
+  if (!Root)
+    return false;
 
-    EmitLabel("abbrev_end", 0);
-    Asm->EOL();
-  }
-}
+  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
+         "Malformed subprogram descriptor anchor type");
+  Constant *RootC = cast<Constant>(*Root->use_begin());
+  assert(RootC->hasNUsesOrMore(1) &&
+         "Malformed subprogram descriptor anchor type");
 
-/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
-/// the line matrix.
-///
-void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
-  // Define last address of section.
-  Asm->EmitInt8(0); Asm->EOL("Extended Op");
-  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
+  bool Result = false;
+  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
+       UI != UE; ++UI)
+    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
+         UUI != UUE; ++UUI)
+      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
 
-  // Mark end of matrix.
-  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
-  Asm->EmitULEB128Bytes(1); Asm->EOL();
-  Asm->EmitInt8(1); Asm->EOL();
+  return Result;
 }
 
-/// EmitDebugLines - Emit source line information.
-///
-void DwarfDebug::EmitDebugLines() {
-  // If the target is using .loc/.file, the assembler will be emitting the
-  // .debug_line table automatically.
-  if (TAI->hasDotLocAndDotFile())
-    return;
-
-  // Minimum line delta, thus ranging from -10..(255-10).
-  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
-  // Maximum line delta, thus ranging from -10..(255-10).
-  const int MaxLineDelta = 255 + MinLineDelta;
-
-  // Start the dwarf line section.
-  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
-
-  // Construct the section header.
-  EmitDifference("line_end", 0, "line_begin", 0, true);
-  Asm->EOL("Length of Source Line Info");
-  EmitLabel("line_begin", 0);
-
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
-
-  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
-  Asm->EOL("Prolog Length");
-  EmitLabel("line_prolog_begin", 0);
+/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
+/// This is inovked by the target AsmPrinter.
+void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
+  // Create all the compile unit DIEs.
+  ConstructCompileUnits();
 
-  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
+  if (CompileUnits.empty()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
+    return;
+  }
 
-  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
+  // Create DIEs for each of the externally visible global variables.
+  bool globalDIEs = ConstructGlobalVariableDIEs();
 
-  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
+  // Create DIEs for each of the externally visible subprograms.
+  bool subprogramDIEs = ConstructSubprograms();
 
-  // Line number standard opcode encodings argument count
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
-  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
-  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
+  // If there is not any debug info available for any global variables and any
+  // subprograms then there is not any debug info to emit.
+  if (!globalDIEs && !subprogramDIEs) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  // Emit directories.
-  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
-    Asm->EmitString(getSourceDirectoryName(DI));
-    Asm->EOL("Directory");
+    return;
   }
 
-  Asm->EmitInt8(0); Asm->EOL("End of directories");
+  MMI = mmi;
+  shouldEmit = true;
+  MMI->setDebugInfoAvailability(true);
 
-  // Emit files.
-  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
-    // Remember source id starts at 1.
-    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
-    Asm->EmitString(getSourceFileName(Id.second));
-    Asm->EOL("Source");
-    Asm->EmitULEB128Bytes(Id.first);
-    Asm->EOL("Directory #");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("Mod date");
-    Asm->EmitULEB128Bytes(0);
-    Asm->EOL("File size");
+  // Prime section data.
+  SectionMap.insert(TAI->getTextSection());
+
+  // Print out .file directives to specify files for .loc directives. These are
+  // printed out early so that they precede any .loc directives.
+  if (TAI->hasDotLocAndDotFile()) {
+    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
+      // Remember source id starts at 1.
+      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
+      sys::Path FullPath(getSourceDirectoryName(Id.first));
+      bool AppendOk =
+        FullPath.appendComponent(getSourceFileName(Id.second));
+      assert(AppendOk && "Could not append filename to directory!");
+      AppendOk = false;
+      Asm->EmitFile(i, FullPath.toString());
+      Asm->EOL();
+    }
   }
 
-  Asm->EmitInt8(0); Asm->EOL("End of files");
+  // Emit initial sections
+  EmitInitial();
 
-  EmitLabel("line_prolog_end", 0);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+}
 
-  // A sequence for each text section.
-  unsigned SecSrcLinesSize = SectionSourceLines.size();
+/// EndModule - Emit all Dwarf sections that should come after the content.
+///
+void DwarfDebug::EndModule() {
+  if (!ShouldEmitDwarfDebug())
+    return;
 
-  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
-    // Isolate current sections line info.
-    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    if (Asm->isVerbose()) {
-      const Section* S = SectionMap[j + 1];
-      O << '\t' << TAI->getCommentString() << " Section"
-        << S->getName() << '\n';
-    } else {
-      Asm->EOL();
-    }
+  // Standard sections final addresses.
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_end", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_end", 0);
 
-    // Dwarf assumes we start with first line of first source file.
-    unsigned Source = 1;
-    unsigned Line = 1;
+  // End text sections.
+  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
+    Asm->SwitchToSection(SectionMap[i]);
+    EmitLabel("section_end", i);
+  }
 
-    // Construct rows of the address, source, line, column matrix.
-    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
-      const SrcLineInfo &LineInfo = LineInfos[i];
-      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
-      if (!LabelID) continue;
+  // Emit common frame information.
+  EmitCommonDebugFrame();
 
-      if (!Asm->isVerbose())
-        Asm->EOL();
-      else {
-        std::pair<unsigned, unsigned> SourceID =
-          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
-        O << '\t' << TAI->getCommentString() << ' '
-          << getSourceDirectoryName(SourceID.first) << ' '
-          << getSourceFileName(SourceID.second)
-          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
-      }
+  // Emit function debug frame information
+  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
+         E = DebugFrames.end(); I != E; ++I)
+    EmitFunctionDebugFrame(*I);
 
-      // Define the line address.
-      Asm->EmitInt8(0); Asm->EOL("Extended Op");
-      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
-      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
-      EmitReference("label",  LabelID); Asm->EOL("Location label");
+  // Compute DIE offsets and sizes.
+  SizeAndOffsets();
 
-      // If change of source, then switch to the new source.
-      if (Source != LineInfo.getSourceID()) {
-        Source = LineInfo.getSourceID();
-        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
-        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
-      }
+  // Emit all the DIEs into a debug info section
+  EmitDebugInfo();
 
-      // If change of line.
-      if (Line != LineInfo.getLine()) {
-        // Determine offset.
-        int Offset = LineInfo.getLine() - Line;
-        int Delta = Offset - MinLineDelta;
+  // Corresponding abbreviations into a abbrev section.
+  EmitAbbreviations();
 
-        // Update line.
-        Line = LineInfo.getLine();
+  // Emit source line correspondence into a debug line section.
+  EmitDebugLines();
 
-        // If delta is small enough and in range...
-        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
-          // ... then use fast opcode.
-          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
-        } else {
-          // ... otherwise use long hand.
-          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
-          Asm->EOL("DW_LNS_advance_line");
-          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
-          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-        }
-      } else {
-        // Copy the previous row (different address or source)
-        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
-      }
-    }
+  // Emit info into a debug pubnames section.
+  EmitDebugPubNames();
 
-    EmitEndOfLineMatrix(j + 1);
-  }
+  // Emit info into a debug str section.
+  EmitDebugStr();
 
-  if (SecSrcLinesSize == 0)
-    // Because we're emitting a debug_line section, we still need a line
-    // table. The linker and friends expect it to exist. If there's nothing to
-    // put into it, emit an empty table.
-    EmitEndOfLineMatrix(1);
+  // Emit info into a debug loc section.
+  EmitDebugLoc();
 
-  EmitLabel("line_end", 0);
-  Asm->EOL();
-}
+  // Emit info into a debug aranges section.
+  EmitDebugARanges();
 
-/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
-///
-void DwarfDebug::EmitCommonDebugFrame() {
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
+  // Emit info into a debug ranges section.
+  EmitDebugRanges();
 
-  int stackGrowth =
-    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
-      TargetFrameInfo::StackGrowsUp ?
-    TD->getPointerSize() : -TD->getPointerSize();
+  // Emit info into a debug macinfo section.
+  EmitDebugMacInfo();
 
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+  // Emit inline info.
+  EmitDebugInlineInfo();
 
-  EmitLabel("debug_frame_common", 0);
-  EmitDifference("debug_frame_common_end", 0,
-                 "debug_frame_common_begin", 0, true);
-  Asm->EOL("Length of Common Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+}
 
-  EmitLabel("debug_frame_common_begin", 0);
-  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
-  Asm->EOL("CIE Identifier Tag");
-  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
-  Asm->EOL("CIE Version");
-  Asm->EmitString("");
-  Asm->EOL("CIE Augmentation");
-  Asm->EmitULEB128Bytes(1);
-  Asm->EOL("CIE Code Alignment Factor");
-  Asm->EmitSLEB128Bytes(stackGrowth);
-  Asm->EOL("CIE Data Alignment Factor");
-  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
-  Asm->EOL("CIE RA Column");
+/// BeginFunction - Gather pre-function debug information.  Assumes being
+/// emitted immediately after the function entry point.
+void DwarfDebug::BeginFunction(MachineFunction *MF) {
+  this->MF = MF;
 
-  std::vector<MachineMove> Moves;
-  RI->getInitialFrameState(Moves);
+  if (!ShouldEmitDwarfDebug()) return;
 
-  EmitFrameMoves(NULL, 0, Moves, false);
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_common_end", 0);
+  // Begin accumulating function debug information.
+  MMI->BeginFunction(MF);
 
-  Asm->EOL();
+  // Assumes in correct section after the entry point.
+  EmitLabel("func_begin", ++SubprogramCount);
+
+  // Emit label for the implicitly defined dbg.stoppoint at the start of the
+  // function.
+  DebugLoc FDL = MF->getDefaultDebugLoc();
+  if (!FDL.isUnknown()) {
+    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
+    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
+                                        DICompileUnit(DLT.CompileUnit));
+    Asm->printLabel(LabelID);
+  }
+
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
-/// section.
-void
-DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
-  if (!TAI->doesDwarfRequireFrameSection())
-    return;
+/// EndFunction - Gather and emit post-function debug information.
+///
+void DwarfDebug::EndFunction(MachineFunction *MF) {
+  if (!ShouldEmitDwarfDebug()) return;
 
-  // Start the dwarf frame section.
-  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
-                 "debug_frame_begin", DebugFrameInfo.Number, true);
-  Asm->EOL("Length of Frame Information Entry");
+  // Define end label for subprogram.
+  EmitLabel("func_end", SubprogramCount);
 
-  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
+  // Get function line info.
+  if (!Lines.empty()) {
+    // Get section line info.
+    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
+    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
+    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
+    // Append the function info to section info.
+    SectionLineInfos.insert(SectionLineInfos.end(),
+                            Lines.begin(), Lines.end());
+  }
 
-  EmitSectionOffset("debug_frame_common", "section_debug_frame",
-                    0, 0, true, false);
-  Asm->EOL("FDE CIE offset");
+  // Construct the DbgScope for abstract instances.
+  for (SmallVector<DbgScope *, 32>::iterator
+         I = AbstractInstanceRootList.begin(),
+         E = AbstractInstanceRootList.end(); I != E; ++I)
+    ConstructAbstractDbgScope(*I);
 
-  EmitReference("func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE initial location");
-  EmitDifference("func_end", DebugFrameInfo.Number,
-                 "func_begin", DebugFrameInfo.Number);
-  Asm->EOL("FDE address range");
+  // Construct scopes for subprogram.
+  if (FunctionDbgScope)
+    ConstructFunctionDbgScope(FunctionDbgScope);
+  else
+    // FIXME: This is wrong. We are essentially getting past a problem with
+    // debug information not being able to handle unreachable blocks that have
+    // debug information in them. In particular, those unreachable blocks that
+    // have "region end" info in them. That situation results in the "root
+    // scope" not being created. If that's the case, then emit a "default"
+    // scope, i.e., one that encompasses the whole function. This isn't
+    // desirable. And a better way of handling this (and all of the debugging
+    // information) needs to be explored.
+    ConstructDefaultDbgScope(MF);
 
-  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
-                 false);
+  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
+                                               MMI->getFrameMoves()));
+
+  // Clear debug info
+  if (FunctionDbgScope) {
+    delete FunctionDbgScope;
+    DbgScopeMap.clear();
+    DbgAbstractScopeMap.clear();
+    DbgConcreteScopeMap.clear();
+    InlinedVariableScopes.clear();
+    FunctionDbgScope = NULL;
+    LexicalScopeStack.clear();
+    AbstractInstanceRootList.clear();
+  }
 
-  Asm->EmitAlignment(2, 0, 0, false);
-  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
+  Lines.clear();
 
-  Asm->EOL();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
-  EmitDifference("pubnames_end", Unit->getID(),
-                 "pubnames_begin", Unit->getID(), true);
-  Asm->EOL("Length of Public Names Info");
-
-  EmitLabel("pubnames_begin", Unit->getID());
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
+  CompileUnit *Unit = CompileUnitMap[V];
+  assert(Unit && "Unable to find CompileUnit");
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
 
-  EmitSectionOffset("info_begin", "section_info",
-                    Unit->getID(), 0, true, false);
-  Asm->EOL("Offset of Compilation Unit Info");
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
-                 true);
-  Asm->EOL("Compilation Unit Length");
+  return ID;
+}
 
-  StringMap<DIE*> &Globals = Unit->getGlobals();
-  for (StringMap<DIE*>::const_iterator
-         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
-    const char *Name = GI->getKeyData();
-    DIE * Entity = GI->second;
+/// RecordSourceLine - Records location information and associates it with a
+/// label. Returns a unique label ID used to generate a label and provide
+/// correspondence to the source line list.
+unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
+                                      DICompileUnit CU) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
-    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
-  }
+  std::string Dir, Fn;
+  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
+                                     CU.getFilename(Fn));
+  unsigned ID = MMI->NextLabelID();
+  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
 
-  Asm->EmitInt32(0); Asm->EOL("End Mark");
-  EmitLabel("pubnames_end", Unit->getID());
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  Asm->EOL();
+  return ID;
 }
 
-/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
-///
-void DwarfDebug::EmitDebugPubNames() {
-  // Start the dwarf pubnames section.
-  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
+/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
+/// timed. Look up the source id with the given directory and source file
+/// names. If none currently exists, create a new id and insert it in the
+/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
+/// well.
+unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
+                                         const std::string &FileName) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  if (MainCU) {
-    EmitDebugPubNamesPerCU(MainCU);
-    return;
-  }
+  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
 
-  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
-    EmitDebugPubNamesPerCU(CompileUnits[i]);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+
+  return SrcId;
 }
 
-/// EmitDebugStr - Emit visible names into a debug str section.
-///
-void DwarfDebug::EmitDebugStr() {
-  // Check to see if it is worth the effort.
-  if (!StringPool.empty()) {
-    // Start the dwarf str section.
-    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+/// RecordRegionStart - Indicate the start of a region.
+unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-    // For each of strings in the string pool.
-    for (unsigned StringID = 1, N = StringPool.size();
-         StringID <= N; ++StringID) {
-      // Emit a label for reference from debug information entries.
-      EmitLabel("string", StringID);
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
+  LexicalScopeStack.push_back(Scope);
 
-      // Emit the string itself.
-      const std::string &String = StringPool[StringID];
-      Asm->EmitString(String); Asm->EOL();
-    }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-    Asm->EOL();
-  }
+  return ID;
 }
 
-/// EmitDebugLoc - Emit visible names into a debug loc section.
-///
-void DwarfDebug::EmitDebugLoc() {
-  // Start the dwarf loc section.
-  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
-  Asm->EOL();
-}
+/// RecordRegionEnd - Indicate the end of a region.
+unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-/// EmitDebugARanges - Emit visible names into a debug aranges section.
-///
-void DwarfDebug::EmitDebugARanges() {
-  // Start the dwarf aranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+  DbgScope *Scope = getOrCreateScope(V);
+  unsigned ID = MMI->NextLabelID();
+  Scope->setEndLabelID(ID);
+  if (LexicalScopeStack.size() != 0)
+    LexicalScopeStack.pop_back();
 
-  // FIXME - Mock up
-#if 0
-  CompileUnit *Unit = GetBaseCompileUnit();
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  // Don't include size of length
-  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
+  return ID;
+}
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+/// RecordVariable - Indicate the declaration of a local variable.
+void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
+                                const MachineInstr *MI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitReference("info_begin", Unit->getID());
-  Asm->EOL("Offset of Compilation Unit Info");
+  DIDescriptor Desc(GV);
+  DbgScope *Scope = NULL;
+  bool InlinedFnVar = false;
 
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
+  if (Desc.getTag() == dwarf::DW_TAG_variable) {
+    // GV is a global variable.
+    DIGlobalVariable DG(GV);
+    Scope = getOrCreateScope(DG.getContext().getGV());
+  } else {
+    DenseMap<const MachineInstr *, DbgScope *>::iterator
+      SI = InlinedVariableScopes.find(MI);
 
-  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
+    if (SI != InlinedVariableScopes.end()) {
+      // or GV is an inlined local variable.
+      Scope = SI->second;
+    } else {
+      DIVariable DV(GV);
+      GlobalVariable *V = DV.getContext().getGV();
 
-  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
-  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
+      // FIXME: The code that checks for the inlined local variable is a hack!
+      DenseMap<const GlobalVariable *, DbgScope *>::iterator
+        AI = AbstractInstanceRootMap.find(V);
 
-  // Range 1
-  EmitReference("text_begin", 0); Asm->EOL("Address");
-  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
+      if (AI != AbstractInstanceRootMap.end()) {
+        // This method is called each time a DECLARE node is encountered. For an
+        // inlined function, this could be many, many times. We don't want to
+        // re-add variables to that DIE for each time. We just want to add them
+        // once. Check to make sure that we haven't added them already.
+        DenseMap<const GlobalVariable *,
+          SmallSet<const GlobalVariable *, 32> >::iterator
+          IP = InlinedParamMap.find(V);
 
-  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
-  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
-#endif
+        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
+          if (TimePassesIsEnabled)
+            DebugTimer->stopTimer();
+          return;
+        }
 
-  Asm->EOL();
-}
+        // or GV is an inlined local variable.
+        Scope = AI->second;
+        InlinedParamMap[V].insert(GV);
+        InlinedFnVar = true;
+      } else {
+        // or GV is a local variable.
+        Scope = getOrCreateScope(V);
+      }
+    }
+  }
 
-/// EmitDebugRanges - Emit visible names into a debug ranges section.
-///
-void DwarfDebug::EmitDebugRanges() {
-  // Start the dwarf ranges section.
-  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
-  Asm->EOL();
-}
+  assert(Scope && "Unable to find the variable's scope");
+  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
+  Scope->AddVariable(DV);
 
-/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
-///
-void DwarfDebug::EmitDebugMacInfo() {
-  if (TAI->doesSupportMacInfoSection()) {
-    // Start the dwarf macinfo section.
-    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
-    Asm->EOL();
-  }
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-/// EmitDebugInlineInfo - Emit inline info using following format.
-/// Section Header:
-/// 1. length of section
-/// 2. Dwarf version number
-/// 3. address size.
-///
-/// Entries (one "entry" for each function that was inlined):
-///
-/// 1. offset into __debug_str section for MIPS linkage name, if exists;
-///   otherwise offset into __debug_str for regular function name.
-/// 2. offset into __debug_str section for regular function name.
-/// 3. an unsigned LEB128 number indicating the number of distinct inlining
-/// instances for the function.
-///
-/// The rest of the entry consists of a {die_offset, low_pc} pair for each
-/// inlined instance; the die_offset points to the inlined_subroutine die in the
-/// __debug_info section, and the low_pc is the starting address for the
-/// inlining instance.
-void DwarfDebug::EmitDebugInlineInfo() {
-  if (!TAI->doesDwarfUsesInlineInfoSection())
-    return;
+//// RecordInlinedFnStart - Indicate the start of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
+                                          unsigned Line, unsigned Col) {
+  unsigned LabelID = MMI->NextLabelID();
 
-  if (!MainCU)
-    return;
+  if (!TAI->doesDwarfUsesInlineInfoSection())
+    return LabelID;
 
-  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
-  Asm->EOL();
-  EmitDifference("debug_inlined_end", 1,
-                 "debug_inlined_begin", 1, true);
-  Asm->EOL("Length of Debug Inlined Information Entry");
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  EmitLabel("debug_inlined_begin", 1);
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<const GlobalVariable *, DbgScope *>::iterator
+    II = AbstractInstanceRootMap.find(GV);
 
-  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
-  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+  if (II == AbstractInstanceRootMap.end()) {
+    // Create an abstract instance entry for this inlined function if it doesn't
+    // already exist.
+    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
 
-  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
-    GlobalVariable *GV = I->first;
-    SmallVector<unsigned, 4> &Labels = I->second;
-    DISubprogram SP(GV);
-    std::string Name;
-    std::string LName;
+    // Get the compile unit context.
+    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+    DIE *SPDie = Unit->getDieMapSlotFor(GV);
+    if (!SPDie)
+      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
 
-    SP.getLinkageName(LName);
-    SP.getName(Name);
+    // Mark as being inlined. This makes this subprogram entry an abstract
+    // instance root.
+    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
+    // that it's defined. That probably won't change in the future. However,
+    // this could be more elegant.
+    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
 
-    Asm->EmitString(LName.empty() ? Name : LName);
-    Asm->EOL("MIPS linkage name");
+    // Keep track of the abstract scope for this function.
+    DbgAbstractScopeMap[GV] = Scope;
 
-    Asm->EmitString(Name); Asm->EOL("Function name");
+    AbstractInstanceRootMap[GV] = Scope;
+    AbstractInstanceRootList.push_back(Scope);
+  }
 
-    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
+  // Create a concrete inlined instance for this inlined function.
+  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
+  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
+  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
+  ScopeDie->setAbstractCompileUnit(Unit);
 
-    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
-           LE = Labels.end(); LI != LE; ++LI) {
-      DIE *SP = MainCU->getDieMapSlotFor(GV);
-      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
+  DIE *Origin = Unit->getDieMapSlotFor(GV);
+  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
+              dwarf::DW_FORM_ref4, Origin);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
+  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
+  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
 
-      if (TD->getPointerSize() == sizeof(int32_t))
-        O << TAI->getData32bitsDirective();
-      else
-        O << TAI->getData64bitsDirective();
+  ConcreteScope->setDie(ScopeDie);
+  ConcreteScope->setStartLabelID(LabelID);
+  MMI->RecordUsedDbgLabel(LabelID);
 
-      PrintLabelName("label", *LI); Asm->EOL("low_pc");
-    }
-  }
+  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
 
-  EmitLabel("debug_inlined_end", 1);
-  Asm->EOL();
-}
+  // Keep track of the concrete scope that's inlined into this function.
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    SI = DbgConcreteScopeMap.find(GV);
 
-/// GetOrCreateSourceID - Look up the source id with the given directory and
-/// source file names. If none currently exists, create a new id and insert it
-/// in the SourceIds map. This can update DirectoryNames and SourceFileNames
-/// maps as well.
-unsigned DwarfDebug::GetOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  unsigned DId;
-  StringMap<unsigned>::iterator DI = DirectoryIdMap.find(DirName);
-  if (DI != DirectoryIdMap.end()) {
-    DId = DI->getValue();
-  } else {
-    DId = DirectoryNames.size() + 1;
-    DirectoryIdMap[DirName] = DId;
-    DirectoryNames.push_back(DirName);
-  }
+  if (SI == DbgConcreteScopeMap.end())
+    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
+  else
+    SI->second.push_back(ConcreteScope);
 
-  unsigned FId;
-  StringMap<unsigned>::iterator FI = SourceFileIdMap.find(FileName);
-  if (FI != SourceFileIdMap.end()) {
-    FId = FI->getValue();
-  } else {
-    FId = SourceFileNames.size() + 1;
-    SourceFileIdMap[FileName] = FId;
-    SourceFileNames.push_back(FileName);
-  }
+  // Track the start label for this inlined function.
+  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+    I = InlineInfo.find(GV);
 
-  DenseMap<std::pair<unsigned, unsigned>, unsigned>::iterator SI =
-    SourceIdMap.find(std::make_pair(DId, FId));
-  if (SI != SourceIdMap.end())
-    return SI->second;
+  if (I == InlineInfo.end())
+    InlineInfo[GV].push_back(LabelID);
+  else
+    I->second.push_back(LabelID);
 
-  unsigned SrcId = SourceIds.size() + 1;  // DW_AT_decl_file cannot be 0.
-  SourceIdMap[std::make_pair(DId, FId)] = SrcId;
-  SourceIds.push_back(std::make_pair(DId, FId));
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 
-  return SrcId;
+  return LabelID;
 }
 
-void DwarfDebug::ConstructCompileUnit(GlobalVariable *GV) {
-  DICompileUnit DIUnit(GV);
-  std::string Dir, FN, Prod;
-  unsigned ID = GetOrCreateSourceID(DIUnit.getDirectory(Dir),
-                                    DIUnit.getFilename(FN));
+/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
+unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
+  if (!TAI->doesDwarfUsesInlineInfoSection())
+    return 0;
 
-  DIE *Die = new DIE(dwarf::DW_TAG_compile_unit);
-  AddSectionOffset(Die, dwarf::DW_AT_stmt_list, dwarf::DW_FORM_data4,
-                   DWLabel("section_line", 0), DWLabel("section_line", 0),
-                   false);
-  AddString(Die, dwarf::DW_AT_producer, dwarf::DW_FORM_string,
-            DIUnit.getProducer(Prod));
-  AddUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data1,
-          DIUnit.getLanguage());
-  AddString(Die, dwarf::DW_AT_name, dwarf::DW_FORM_string, FN);
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
 
-  if (!Dir.empty())
-    AddString(Die, dwarf::DW_AT_comp_dir, dwarf::DW_FORM_string, Dir);
-  if (DIUnit.isOptimized())
-    AddUInt(Die, dwarf::DW_AT_APPLE_optimized, dwarf::DW_FORM_flag, 1);
+  GlobalVariable *GV = SP.getGV();
+  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
+    I = DbgConcreteScopeMap.find(GV);
 
-  std::string Flags;
-  DIUnit.getFlags(Flags);
-  if (!Flags.empty())
-    AddString(Die, dwarf::DW_AT_APPLE_flags, dwarf::DW_FORM_string, Flags);
+  if (I == DbgConcreteScopeMap.end()) {
+    // FIXME: Can this situation actually happen? And if so, should it?
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
 
-  unsigned RVer = DIUnit.getRunTimeVersion();
-  if (RVer)
-    AddUInt(Die, dwarf::DW_AT_APPLE_major_runtime_vers,
-            dwarf::DW_FORM_data1, RVer);
+    return 0;
+  }
 
-  CompileUnit *Unit = new CompileUnit(ID, Die);
-  if (DIUnit.isMain()) {
-    assert(!MainCU && "Multiple main compile units are found!");
-    MainCU = Unit;
-    }
+  SmallVector<DbgScope *, 8> &Scopes = I->second;
+  assert(!Scopes.empty() && "We should have at least one debug scope!");
+  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
+  unsigned ID = MMI->NextLabelID();
+  MMI->RecordUsedDbgLabel(ID);
+  Scope->setEndLabelID(ID);
 
-  CompileUnitMap[DIUnit.getGV()] = Unit;
-  CompileUnits.push_back(Unit);
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
+
+  return ID;
 }
 
-/// ConstructCompileUnits - Create a compile unit DIEs.
-void DwarfDebug::ConstructCompileUnits() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.compile_units");
-  if (!Root)
+/// RecordVariableScope - Record scope for the variable declared by
+/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
+/// for only inlined subroutine variables. Other variables's scopes are
+/// determined during RecordVariable().
+void DwarfDebug::RecordVariableScope(DIVariable &DV,
+                                     const MachineInstr *DeclareMI) {
+  if (TimePassesIsEnabled)
+    DebugTimer->startTimer();
+
+  DISubprogram SP(DV.getContext().getGV());
+
+  if (SP.isNull()) {
+    if (TimePassesIsEnabled)
+      DebugTimer->stopTimer();
+
     return;
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed compile unit descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed compile unit descriptor anchor type");
+  }
 
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI) {
-      GlobalVariable *GV = cast<GlobalVariable>(*UUI);
-      ConstructCompileUnit(GV);
-    }
+  DenseMap<GlobalVariable *, DbgScope *>::iterator
+    I = DbgAbstractScopeMap.find(SP.getGV());
+  if (I != DbgAbstractScopeMap.end())
+    InlinedVariableScopes[DeclareMI] = I->second;
+
+  if (TimePassesIsEnabled)
+    DebugTimer->stopTimer();
 }
 
-bool DwarfDebug::ConstructGlobalVariableDIE(GlobalVariable *GV) {
-  DIGlobalVariable DI_GV(GV);
-  CompileUnit *DW_Unit = MainCU;
-  if (!DW_Unit)
-    DW_Unit = &FindCompileUnit(DI_GV.getCompileUnit());
+/// SizeAndOffsetDie - Compute the size and offset of a DIE.
+///
+unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {
+  // Get the children.
+  const std::vector<DIE *> &Children = Die->getChildren();
 
-  // Check for pre-existence.
-  DIE *&Slot = DW_Unit->getDieMapSlotFor(DI_GV.getGV());
-  if (Slot)
-    return false;
+  // If not last sibling and has children then add sibling offset attribute.
+  if (!Last && !Children.empty()) Die->AddSiblingOffset();
+
+  // Record the abbreviation.
+  AssignAbbrevNumber(Die->getAbbrev());
+
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
+
+  // Set DIE offset
+  Die->setOffset(Offset);
+
+  // Start the size with the size of abbreviation code.
+  Offset += TargetAsmInfo::getULEB128Size(AbbrevNumber);
+
+  const SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-  DIE *VariableDie = CreateGlobalVariableDIE(DW_Unit, DI_GV);
+  // Size the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i)
+    // Size attribute value.
+    Offset += Values[i]->SizeOf(TD, AbbrevData[i].getForm());
 
-  // Add address.
-  DIEBlock *Block = new DIEBlock();
-  AddUInt(Block, 0, dwarf::DW_FORM_data1, dwarf::DW_OP_addr);
-  std::string GLN;
-  AddObjectLabel(Block, 0, dwarf::DW_FORM_udata,
-                 Asm->getGlobalLinkName(DI_GV.getGlobal(), GLN));
-  AddBlock(VariableDie, dwarf::DW_AT_location, 0, Block);
+  // Size the DIE children if any.
+  if (!Children.empty()) {
+    assert(Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes &&
+           "Children flag not set");
 
-  // Add to map.
-  Slot = VariableDie;
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      Offset = SizeAndOffsetDie(Children[j], Offset, (j + 1) == M);
 
-  // Add to context owner.
-  DW_Unit->getDie()->AddChild(VariableDie);
+    // End of children marker.
+    Offset += sizeof(int8_t);
+  }
 
-  // Expose as global. FIXME - need to check external flag.
-  std::string Name;
-  DW_Unit->AddGlobal(DI_GV.getName(Name), VariableDie);
-  return true;
+  Die->setSize(Offset - Die->getOffset());
+  return Offset;
 }
 
-/// ConstructGlobalVariableDIEs - Create DIEs for each of the externally visible
-/// global variables. Return true if at least one global DIE is created.
-bool DwarfDebug::ConstructGlobalVariableDIEs() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.global_variables");
-  if (!Root)
-    return false;
+/// SizeAndOffsets - Compute the size and offset of all the DIEs.
+///
+void DwarfDebug::SizeAndOffsets() {
+  // Compute size of compile unit header.
+  static unsigned Offset =
+    sizeof(int32_t) + // Length of Compilation Unit Info
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t);   // Pointer Size (in bytes)
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed global variable descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed global variable descriptor anchor type");
+  // Process base compile unit.
+  if (MainCU) {
+    SizeAndOffsetDie(MainCU->getDie(), Offset, true);
+    CompileUnitOffsets[MainCU] = 0;
+    return;
+  }
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructGlobalVariableDIE(cast<GlobalVariable>(*UUI));
+  // Process all compile units.
+  unsigned PrevOffset = 0;
 
-  return Result;
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i) {
+    CompileUnit *Unit = CompileUnits[i];
+    CompileUnitOffsets[Unit] = PrevOffset;
+    PrevOffset += SizeAndOffsetDie(Unit->getDie(), Offset, true)
+      + sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+  }
 }
 
-bool DwarfDebug::ConstructSubprogram(GlobalVariable *GV) {
-  DISubprogram SP(GV);
-  CompileUnit *Unit = MainCU;
-  if (!Unit)
-    Unit = &FindCompileUnit(SP.getCompileUnit());
-
-  // Check for pre-existence.
-  DIE *&Slot = Unit->getDieMapSlotFor(GV);
-  if (Slot)
-    return false;
+/// EmitInitial - Emit initial Dwarf declarations.  This is necessary for cc
+/// tools to recognize the object file contains Dwarf information.
+void DwarfDebug::EmitInitial() {
+  // Check to see if we already emitted intial headers.
+  if (didInitial) return;
+  didInitial = true;
 
-  if (!SP.isDefinition())
-    // This is a method declaration which will be handled while constructing
-    // class type.
-    return false;
+  // Dwarf sections base addresses.
+  if (TAI->doesDwarfRequireFrameSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
+    EmitLabel("section_debug_frame", 0);
+  }
 
-  DIE *SubprogramDie = CreateSubprogramDIE(Unit, SP);
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
+  EmitLabel("section_info", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
+  EmitLabel("section_abbrev", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
+  EmitLabel("section_aranges", 0);
 
-  // Add to map.
-  Slot = SubprogramDie;
+  if (TAI->doesSupportMacInfoSection()) {
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    EmitLabel("section_macinfo", 0);
+  }
 
-  // Add to context owner.
-  Unit->getDie()->AddChild(SubprogramDie);
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
+  EmitLabel("section_line", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  EmitLabel("section_loc", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
+  EmitLabel("section_pubnames", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfStrSection());
+  EmitLabel("section_str", 0);
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  EmitLabel("section_ranges", 0);
 
-  // Expose as global.
-  std::string Name;
-  Unit->AddGlobal(SP.getName(Name), SubprogramDie);
-  return true;
+  Asm->SwitchToSection(TAI->getTextSection());
+  EmitLabel("text_begin", 0);
+  Asm->SwitchToSection(TAI->getDataSection());
+  EmitLabel("data_begin", 0);
 }
 
-/// ConstructSubprograms - Create DIEs for each of the externally visible
-/// subprograms. Return true if at least one subprogram DIE is created.
-bool DwarfDebug::ConstructSubprograms() {
-  GlobalVariable *Root = M->getGlobalVariable("llvm.dbg.subprograms");
-  if (!Root)
-    return false;
+/// EmitDIE - Recusively Emits a debug information entry.
+///
+void DwarfDebug::EmitDIE(DIE *Die) {
+  // Get the abbreviation for this DIE.
+  unsigned AbbrevNumber = Die->getAbbrevNumber();
+  const DIEAbbrev *Abbrev = Abbreviations[AbbrevNumber - 1];
 
-  assert(Root->hasLinkOnceLinkage() && Root->hasOneUse() &&
-         "Malformed subprogram descriptor anchor type");
-  Constant *RootC = cast<Constant>(*Root->use_begin());
-  assert(RootC->hasNUsesOrMore(1) &&
-         "Malformed subprogram descriptor anchor type");
+  Asm->EOL();
 
-  bool Result = false;
-  for (Value::use_iterator UI = RootC->use_begin(), UE = Root->use_end();
-       UI != UE; ++UI)
-    for (Value::use_iterator UUI = UI->use_begin(), UUE = UI->use_end();
-         UUI != UUE; ++UUI)
-      Result |= ConstructSubprogram(cast<GlobalVariable>(*UUI));
+  // Emit the code (index) for the abbreviation.
+  Asm->EmitULEB128Bytes(AbbrevNumber);
 
-  return Result;
-}
+  if (Asm->isVerbose())
+    Asm->EOL(std::string("Abbrev [" +
+                         utostr(AbbrevNumber) +
+                         "] 0x" + utohexstr(Die->getOffset()) +
+                         ":0x" + utohexstr(Die->getSize()) + " " +
+                         dwarf::TagString(Abbrev->getTag())));
+  else
+    Asm->EOL();
 
-/// SetDebugInfo - Create global DIEs and emit initial debug info sections.
-/// This is inovked by the target AsmPrinter.
-void DwarfDebug::SetDebugInfo(MachineModuleInfo *mmi) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  SmallVector<DIEValue*, 32> &Values = Die->getValues();
+  const SmallVector<DIEAbbrevData, 8> &AbbrevData = Abbrev->getData();
 
-  // Create all the compile unit DIEs.
-  ConstructCompileUnits();
+  // Emit the DIE attribute values.
+  for (unsigned i = 0, N = Values.size(); i < N; ++i) {
+    unsigned Attr = AbbrevData[i].getAttribute();
+    unsigned Form = AbbrevData[i].getForm();
+    assert(Form && "Too many attributes for DIE (check abbreviation)");
 
-  if (CompileUnits.empty()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    switch (Attr) {
+    case dwarf::DW_AT_sibling:
+      Asm->EmitInt32(Die->SiblingOffset());
+      break;
+    case dwarf::DW_AT_abstract_origin: {
+      DIEEntry *E = cast<DIEEntry>(Values[i]);
+      DIE *Origin = E->getEntry();
+      unsigned Addr =
+        CompileUnitOffsets[Die->getAbstractCompileUnit()] +
+        Origin->getOffset();
 
-    return;
-  }
+      Asm->EmitInt32(Addr);
+      break;
+    }
+    default:
+      // Emit an attribute using the defined form.
+      Values[i]->EmitValue(this, Form);
+      break;
+    }
 
-  // Create DIEs for each of the externally visible global variables.
-  bool globalDIEs = ConstructGlobalVariableDIEs();
+    Asm->EOL(dwarf::AttributeString(Attr));
+  }
 
-  // Create DIEs for each of the externally visible subprograms.
-  bool subprogramDIEs = ConstructSubprograms();
+  // Emit the DIE children if any.
+  if (Abbrev->getChildrenFlag() == dwarf::DW_CHILDREN_yes) {
+    const std::vector<DIE *> &Children = Die->getChildren();
 
-  // If there is not any debug info available for any global variables and any
-  // subprograms then there is not any debug info to emit.
-  if (!globalDIEs && !subprogramDIEs) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    for (unsigned j = 0, M = Children.size(); j < M; ++j)
+      EmitDIE(Children[j]);
 
-    return;
+    Asm->EmitInt8(0); Asm->EOL("End Of Children Mark");
   }
+}
 
-  MMI = mmi;
-  shouldEmit = true;
-  MMI->setDebugInfoAvailability(true);
+/// EmitDebugInfo / EmitDebugInfoPerCU - Emit the debug info section.
+///
+void DwarfDebug::EmitDebugInfoPerCU(CompileUnit *Unit) {
+  DIE *Die = Unit->getDie();
+
+  // Emit the compile units header.
+  EmitLabel("info_begin", Unit->getID());
+
+  // Emit size of content not including length itself
+  unsigned ContentSize = Die->getSize() +
+    sizeof(int16_t) + // DWARF version number
+    sizeof(int32_t) + // Offset Into Abbrev. Section
+    sizeof(int8_t) +  // Pointer Size (in bytes)
+    sizeof(int32_t);  // FIXME - extra pad for gdb bug.
+
+  Asm->EmitInt32(ContentSize);  Asm->EOL("Length of Compilation Unit Info");
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
+  EmitSectionOffset("abbrev_begin", "section_abbrev", 0, 0, true, false);
+  Asm->EOL("Offset Into Abbrev. Section");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
+
+  EmitDIE(Die);
+  // FIXME - extra padding for gdb bug.
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  Asm->EmitInt8(0); Asm->EOL("Extra Pad For GDB");
+  EmitLabel("info_end", Unit->getID());
+
+  Asm->EOL();
+}
 
-  // Prime section data.
-  SectionMap.insert(TAI->getTextSection());
+void DwarfDebug::EmitDebugInfo() {
+  // Start debug info section.
+  Asm->SwitchToDataSection(TAI->getDwarfInfoSection());
 
-  // Print out .file directives to specify files for .loc directives. These are
-  // printed out early so that they precede any .loc directives.
-  if (TAI->hasDotLocAndDotFile()) {
-    for (unsigned i = 1, e = getNumSourceIds()+1; i != e; ++i) {
-      // Remember source id starts at 1.
-      std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(i);
-      sys::Path FullPath(getSourceDirectoryName(Id.first));
-      bool AppendOk =
-        FullPath.appendComponent(getSourceFileName(Id.second));
-      assert(AppendOk && "Could not append filename to directory!");
-      AppendOk = false;
-      Asm->EmitFile(i, FullPath.toString());
-      Asm->EOL();
-    }
+  if (MainCU) {
+    EmitDebugInfoPerCU(MainCU);
+    return;
   }
 
-  // Emit initial sections
-  EmitInitial();
-
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugInfoPerCU(CompileUnits[i]);
 }
 
-/// EndModule - Emit all Dwarf sections that should come after the content.
+/// EmitAbbreviations - Emit the abbreviation section.
 ///
-void DwarfDebug::EndModule() {
-  if (!ShouldEmitDwarfDebug())
-    return;
+void DwarfDebug::EmitAbbreviations() const {
+  // Check to see if it is worth the effort.
+  if (!Abbreviations.empty()) {
+    // Start the debug abbrev section.
+    Asm->SwitchToDataSection(TAI->getDwarfAbbrevSection());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    EmitLabel("abbrev_begin", 0);
 
-  // Standard sections final addresses.
-  Asm->SwitchToSection(TAI->getTextSection());
-  EmitLabel("text_end", 0);
-  Asm->SwitchToSection(TAI->getDataSection());
-  EmitLabel("data_end", 0);
+    // For each abbrevation.
+    for (unsigned i = 0, N = Abbreviations.size(); i < N; ++i) {
+      // Get abbreviation data
+      const DIEAbbrev *Abbrev = Abbreviations[i];
 
-  // End text sections.
-  for (unsigned i = 1, N = SectionMap.size(); i <= N; ++i) {
-    Asm->SwitchToSection(SectionMap[i]);
-    EmitLabel("section_end", i);
-  }
+      // Emit the abbrevations code (base 1 index.)
+      Asm->EmitULEB128Bytes(Abbrev->getNumber());
+      Asm->EOL("Abbreviation Code");
 
-  // Emit common frame information.
-  EmitCommonDebugFrame();
+      // Emit the abbreviations data.
+      Abbrev->Emit(Asm);
 
-  // Emit function debug frame information
-  for (std::vector<FunctionDebugFrameInfo>::iterator I = DebugFrames.begin(),
-         E = DebugFrames.end(); I != E; ++I)
-    EmitFunctionDebugFrame(*I);
+      Asm->EOL();
+    }
 
-  // Compute DIE offsets and sizes.
-  SizeAndOffsets();
+    // Mark end of abbreviations.
+    Asm->EmitULEB128Bytes(0); Asm->EOL("EOM(3)");
 
-  // Emit all the DIEs into a debug info section
-  EmitDebugInfo();
+    EmitLabel("abbrev_end", 0);
+    Asm->EOL();
+  }
+}
 
-  // Corresponding abbreviations into a abbrev section.
-  EmitAbbreviations();
+/// EmitEndOfLineMatrix - Emit the last address of the section and the end of
+/// the line matrix.
+///
+void DwarfDebug::EmitEndOfLineMatrix(unsigned SectionEnd) {
+  // Define last address of section.
+  Asm->EmitInt8(0); Asm->EOL("Extended Op");
+  Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+  Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+  EmitReference("section_end", SectionEnd); Asm->EOL("Section end label");
 
-  // Emit source line correspondence into a debug line section.
-  EmitDebugLines();
+  // Mark end of matrix.
+  Asm->EmitInt8(0); Asm->EOL("DW_LNE_end_sequence");
+  Asm->EmitULEB128Bytes(1); Asm->EOL();
+  Asm->EmitInt8(1); Asm->EOL();
+}
 
-  // Emit info into a debug pubnames section.
-  EmitDebugPubNames();
+/// EmitDebugLines - Emit source line information.
+///
+void DwarfDebug::EmitDebugLines() {
+  // If the target is using .loc/.file, the assembler will be emitting the
+  // .debug_line table automatically.
+  if (TAI->hasDotLocAndDotFile())
+    return;
 
-  // Emit info into a debug str section.
-  EmitDebugStr();
+  // Minimum line delta, thus ranging from -10..(255-10).
+  const int MinLineDelta = -(dwarf::DW_LNS_fixed_advance_pc + 1);
+  // Maximum line delta, thus ranging from -10..(255-10).
+  const int MaxLineDelta = 255 + MinLineDelta;
 
-  // Emit info into a debug loc section.
-  EmitDebugLoc();
+  // Start the dwarf line section.
+  Asm->SwitchToDataSection(TAI->getDwarfLineSection());
 
-  // Emit info into a debug aranges section.
-  EmitDebugARanges();
+  // Construct the section header.
+  EmitDifference("line_end", 0, "line_begin", 0, true);
+  Asm->EOL("Length of Source Line Info");
+  EmitLabel("line_begin", 0);
 
-  // Emit info into a debug ranges section.
-  EmitDebugRanges();
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF version number");
 
-  // Emit info into a debug macinfo section.
-  EmitDebugMacInfo();
+  EmitDifference("line_prolog_end", 0, "line_prolog_begin", 0, true);
+  Asm->EOL("Prolog Length");
+  EmitLabel("line_prolog_begin", 0);
 
-  // Emit inline info.
-  EmitDebugInlineInfo();
+  Asm->EmitInt8(1); Asm->EOL("Minimum Instruction Length");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+  Asm->EmitInt8(1); Asm->EOL("Default is_stmt_start flag");
 
-/// BeginFunction - Gather pre-function debug information.  Assumes being
-/// emitted immediately after the function entry point.
-void DwarfDebug::BeginFunction(MachineFunction *MF) {
-  this->MF = MF;
+  Asm->EmitInt8(MinLineDelta); Asm->EOL("Line Base Value (Special Opcodes)");
 
-  if (!ShouldEmitDwarfDebug()) return;
+  Asm->EmitInt8(MaxLineDelta); Asm->EOL("Line Range Value (Special Opcodes)");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EmitInt8(-MinLineDelta); Asm->EOL("Special Opcode Base");
 
-  // Begin accumulating function debug information.
-  MMI->BeginFunction(MF);
+  // Line number standard opcode encodings argument count
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_copy arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_advance_line arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_file arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_set_column arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_negate_stmt arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_set_basic_block arg count");
+  Asm->EmitInt8(0); Asm->EOL("DW_LNS_const_add_pc arg count");
+  Asm->EmitInt8(1); Asm->EOL("DW_LNS_fixed_advance_pc arg count");
 
-  // Assumes in correct section after the entry point.
-  EmitLabel("func_begin", ++SubprogramCount);
+  // Emit directories.
+  for (unsigned DI = 1, DE = getNumSourceDirectories()+1; DI != DE; ++DI) {
+    Asm->EmitString(getSourceDirectoryName(DI));
+    Asm->EOL("Directory");
+  }
 
-  // Emit label for the implicitly defined dbg.stoppoint at the start of the
-  // function.
-  DebugLoc FDL = MF->getDefaultDebugLoc();
-  if (!FDL.isUnknown()) {
-    DebugLocTuple DLT = MF->getDebugLocTuple(FDL);
-    unsigned LabelID = RecordSourceLine(DLT.Line, DLT.Col,
-                                        DICompileUnit(DLT.CompileUnit));
-    Asm->printLabel(LabelID);
+  Asm->EmitInt8(0); Asm->EOL("End of directories");
+
+  // Emit files.
+  for (unsigned SI = 1, SE = getNumSourceIds()+1; SI != SE; ++SI) {
+    // Remember source id starts at 1.
+    std::pair<unsigned, unsigned> Id = getSourceDirectoryAndFileIds(SI);
+    Asm->EmitString(getSourceFileName(Id.second));
+    Asm->EOL("Source");
+    Asm->EmitULEB128Bytes(Id.first);
+    Asm->EOL("Directory #");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("Mod date");
+    Asm->EmitULEB128Bytes(0);
+    Asm->EOL("File size");
   }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+  Asm->EmitInt8(0); Asm->EOL("End of files");
 
-/// EndFunction - Gather and emit post-function debug information.
-///
-void DwarfDebug::EndFunction(MachineFunction *MF) {
-  if (!ShouldEmitDwarfDebug()) return;
+  EmitLabel("line_prolog_end", 0);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  // A sequence for each text section.
+  unsigned SecSrcLinesSize = SectionSourceLines.size();
 
-  // Define end label for subprogram.
-  EmitLabel("func_end", SubprogramCount);
+  for (unsigned j = 0; j < SecSrcLinesSize; ++j) {
+    // Isolate current sections line info.
+    const std::vector<SrcLineInfo> &LineInfos = SectionSourceLines[j];
 
-  // Get function line info.
-  if (!Lines.empty()) {
-    // Get section line info.
-    unsigned ID = SectionMap.insert(Asm->CurrentSection_);
-    if (SectionSourceLines.size() < ID) SectionSourceLines.resize(ID);
-    std::vector<SrcLineInfo> &SectionLineInfos = SectionSourceLines[ID-1];
-    // Append the function info to section info.
-    SectionLineInfos.insert(SectionLineInfos.end(),
-                            Lines.begin(), Lines.end());
-  }
+    if (Asm->isVerbose()) {
+      const Section* S = SectionMap[j + 1];
+      O << '\t' << TAI->getCommentString() << " Section"
+        << S->getName() << '\n';
+    } else {
+      Asm->EOL();
+    }
 
-  // Construct the DbgScope for abstract instances.
-  for (SmallVector<DbgScope *, 32>::iterator
-         I = AbstractInstanceRootList.begin(),
-         E = AbstractInstanceRootList.end(); I != E; ++I)
-    ConstructAbstractDbgScope(*I);
+    // Dwarf assumes we start with first line of first source file.
+    unsigned Source = 1;
+    unsigned Line = 1;
 
-  // Construct scopes for subprogram.
-  if (FunctionDbgScope)
-    ConstructFunctionDbgScope(FunctionDbgScope);
-  else
-    // FIXME: This is wrong. We are essentially getting past a problem with
-    // debug information not being able to handle unreachable blocks that have
-    // debug information in them. In particular, those unreachable blocks that
-    // have "region end" info in them. That situation results in the "root
-    // scope" not being created. If that's the case, then emit a "default"
-    // scope, i.e., one that encompasses the whole function. This isn't
-    // desirable. And a better way of handling this (and all of the debugging
-    // information) needs to be explored.
-    ConstructDefaultDbgScope(MF);
+    // Construct rows of the address, source, line, column matrix.
+    for (unsigned i = 0, N = LineInfos.size(); i < N; ++i) {
+      const SrcLineInfo &LineInfo = LineInfos[i];
+      unsigned LabelID = MMI->MappedLabel(LineInfo.getLabelID());
+      if (!LabelID) continue;
+
+      if (!Asm->isVerbose())
+        Asm->EOL();
+      else {
+        std::pair<unsigned, unsigned> SourceID =
+          getSourceDirectoryAndFileIds(LineInfo.getSourceID());
+        O << '\t' << TAI->getCommentString() << ' '
+          << getSourceDirectoryName(SourceID.first) << ' '
+          << getSourceFileName(SourceID.second)
+          <<" :" << utostr_32(LineInfo.getLine()) << '\n';
+      }
 
-  DebugFrames.push_back(FunctionDebugFrameInfo(SubprogramCount,
-                                               MMI->getFrameMoves()));
+      // Define the line address.
+      Asm->EmitInt8(0); Asm->EOL("Extended Op");
+      Asm->EmitInt8(TD->getPointerSize() + 1); Asm->EOL("Op size");
+      Asm->EmitInt8(dwarf::DW_LNE_set_address); Asm->EOL("DW_LNE_set_address");
+      EmitReference("label",  LabelID); Asm->EOL("Location label");
 
-  // Clear debug info
-  if (FunctionDbgScope) {
-    delete FunctionDbgScope;
-    DbgScopeMap.clear();
-    DbgAbstractScopeMap.clear();
-    DbgConcreteScopeMap.clear();
-    InlinedVariableScopes.clear();
-    FunctionDbgScope = NULL;
-    LexicalScopeStack.clear();
-    AbstractInstanceRootList.clear();
-  }
+      // If change of source, then switch to the new source.
+      if (Source != LineInfo.getSourceID()) {
+        Source = LineInfo.getSourceID();
+        Asm->EmitInt8(dwarf::DW_LNS_set_file); Asm->EOL("DW_LNS_set_file");
+        Asm->EmitULEB128Bytes(Source); Asm->EOL("New Source");
+      }
 
-  Lines.clear();
+      // If change of line.
+      if (Line != LineInfo.getLine()) {
+        // Determine offset.
+        int Offset = LineInfo.getLine() - Line;
+        int Delta = Offset - MinLineDelta;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
-}
+        // Update line.
+        Line = LineInfo.getLine();
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(Value *V, unsigned Line, unsigned Col) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+        // If delta is small enough and in range...
+        if (Delta >= 0 && Delta < (MaxLineDelta - 1)) {
+          // ... then use fast opcode.
+          Asm->EmitInt8(Delta - MinLineDelta); Asm->EOL("Line Delta");
+        } else {
+          // ... otherwise use long hand.
+          Asm->EmitInt8(dwarf::DW_LNS_advance_line);
+          Asm->EOL("DW_LNS_advance_line");
+          Asm->EmitSLEB128Bytes(Offset); Asm->EOL("Line Offset");
+          Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+        }
+      } else {
+        // Copy the previous row (different address or source)
+        Asm->EmitInt8(dwarf::DW_LNS_copy); Asm->EOL("DW_LNS_copy");
+      }
+    }
 
-  CompileUnit *Unit = CompileUnitMap[V];
-  assert(Unit && "Unable to find CompileUnit");
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Unit->getID(), ID));
+    EmitEndOfLineMatrix(j + 1);
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  if (SecSrcLinesSize == 0)
+    // Because we're emitting a debug_line section, we still need a line
+    // table. The linker and friends expect it to exist. If there's nothing to
+    // put into it, emit an empty table.
+    EmitEndOfLineMatrix(1);
 
-  return ID;
+  EmitLabel("line_end", 0);
+  Asm->EOL();
 }
 
-/// RecordSourceLine - Records location information and associates it with a
-/// label. Returns a unique label ID used to generate a label and provide
-/// correspondence to the source line list.
-unsigned DwarfDebug::RecordSourceLine(unsigned Line, unsigned Col,
-                                      DICompileUnit CU) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+/// EmitCommonDebugFrame - Emit common frame info into a debug frame section.
+///
+void DwarfDebug::EmitCommonDebugFrame() {
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  std::string Dir, Fn;
-  unsigned Src = GetOrCreateSourceID(CU.getDirectory(Dir),
-                                     CU.getFilename(Fn));
-  unsigned ID = MMI->NextLabelID();
-  Lines.push_back(SrcLineInfo(Line, Col, Src, ID));
+  int stackGrowth =
+    Asm->TM.getFrameInfo()->getStackGrowthDirection() ==
+      TargetFrameInfo::StackGrowsUp ?
+    TD->getPointerSize() : -TD->getPointerSize();
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  return ID;
-}
+  EmitLabel("debug_frame_common", 0);
+  EmitDifference("debug_frame_common_end", 0,
+                 "debug_frame_common_begin", 0, true);
+  Asm->EOL("Length of Common Information Entry");
 
-/// getOrCreateSourceID - Public version of GetOrCreateSourceID. This can be
-/// timed. Look up the source id with the given directory and source file
-/// names. If none currently exists, create a new id and insert it in the
-/// SourceIds map. This can update DirectoryNames and SourceFileNames maps as
-/// well.
-unsigned DwarfDebug::getOrCreateSourceID(const std::string &DirName,
-                                         const std::string &FileName) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("debug_frame_common_begin", 0);
+  Asm->EmitInt32((int)dwarf::DW_CIE_ID);
+  Asm->EOL("CIE Identifier Tag");
+  Asm->EmitInt8(dwarf::DW_CIE_VERSION);
+  Asm->EOL("CIE Version");
+  Asm->EmitString("");
+  Asm->EOL("CIE Augmentation");
+  Asm->EmitULEB128Bytes(1);
+  Asm->EOL("CIE Code Alignment Factor");
+  Asm->EmitSLEB128Bytes(stackGrowth);
+  Asm->EOL("CIE Data Alignment Factor");
+  Asm->EmitInt8(RI->getDwarfRegNum(RI->getRARegister(), false));
+  Asm->EOL("CIE RA Column");
 
-  unsigned SrcId = GetOrCreateSourceID(DirName, FileName);
+  std::vector<MachineMove> Moves;
+  RI->getInitialFrameState(Moves);
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitFrameMoves(NULL, 0, Moves, false);
 
-  return SrcId;
-}
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_common_end", 0);
 
-/// RecordRegionStart - Indicate the start of a region.
-unsigned DwarfDebug::RecordRegionStart(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EOL();
+}
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  if (!Scope->getStartLabelID()) Scope->setStartLabelID(ID);
-  LexicalScopeStack.push_back(Scope);
+/// EmitFunctionDebugFrame - Emit per function frame info into a debug frame
+/// section.
+void
+DwarfDebug::EmitFunctionDebugFrame(const FunctionDebugFrameInfo&DebugFrameInfo){
+  if (!TAI->doesDwarfRequireFrameSection())
+    return;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  // Start the dwarf frame section.
+  Asm->SwitchToDataSection(TAI->getDwarfFrameSection());
 
-  return ID;
-}
+  EmitDifference("debug_frame_end", DebugFrameInfo.Number,
+                 "debug_frame_begin", DebugFrameInfo.Number, true);
+  Asm->EOL("Length of Frame Information Entry");
 
-/// RecordRegionEnd - Indicate the end of a region.
-unsigned DwarfDebug::RecordRegionEnd(GlobalVariable *V) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  EmitLabel("debug_frame_begin", DebugFrameInfo.Number);
 
-  DbgScope *Scope = getOrCreateScope(V);
-  unsigned ID = MMI->NextLabelID();
-  Scope->setEndLabelID(ID);
-  if (LexicalScopeStack.size() != 0)
-    LexicalScopeStack.pop_back();
+  EmitSectionOffset("debug_frame_common", "section_debug_frame",
+                    0, 0, true, false);
+  Asm->EOL("FDE CIE offset");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitReference("func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE initial location");
+  EmitDifference("func_end", DebugFrameInfo.Number,
+                 "func_begin", DebugFrameInfo.Number);
+  Asm->EOL("FDE address range");
 
-  return ID;
-}
+  EmitFrameMoves("func_begin", DebugFrameInfo.Number, DebugFrameInfo.Moves,
+                 false);
 
-/// RecordVariable - Indicate the declaration of a local variable.
-void DwarfDebug::RecordVariable(GlobalVariable *GV, unsigned FrameIndex,
-                                const MachineInstr *MI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  Asm->EmitAlignment(2, 0, 0, false);
+  EmitLabel("debug_frame_end", DebugFrameInfo.Number);
 
-  DIDescriptor Desc(GV);
-  DbgScope *Scope = NULL;
-  bool InlinedFnVar = false;
+  Asm->EOL();
+}
 
-  if (Desc.getTag() == dwarf::DW_TAG_variable) {
-    // GV is a global variable.
-    DIGlobalVariable DG(GV);
-    Scope = getOrCreateScope(DG.getContext().getGV());
-  } else {
-    DenseMap<const MachineInstr *, DbgScope *>::iterator
-      SI = InlinedVariableScopes.find(MI);
+void DwarfDebug::EmitDebugPubNamesPerCU(CompileUnit *Unit) {
+  EmitDifference("pubnames_end", Unit->getID(),
+                 "pubnames_begin", Unit->getID(), true);
+  Asm->EOL("Length of Public Names Info");
 
-    if (SI != InlinedVariableScopes.end()) {
-      // or GV is an inlined local variable.
-      Scope = SI->second;
-    } else {
-      DIVariable DV(GV);
-      GlobalVariable *V = DV.getContext().getGV();
+  EmitLabel("pubnames_begin", Unit->getID());
 
-      // FIXME: The code that checks for the inlined local variable is a hack!
-      DenseMap<const GlobalVariable *, DbgScope *>::iterator
-        AI = AbstractInstanceRootMap.find(V);
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("DWARF Version");
 
-      if (AI != AbstractInstanceRootMap.end()) {
-        // This method is called each time a DECLARE node is encountered. For an
-        // inlined function, this could be many, many times. We don't want to
-        // re-add variables to that DIE for each time. We just want to add them
-        // once. Check to make sure that we haven't added them already.
-        DenseMap<const GlobalVariable *,
-          SmallSet<const GlobalVariable *, 32> >::iterator
-          IP = InlinedParamMap.find(V);
+  EmitSectionOffset("info_begin", "section_info",
+                    Unit->getID(), 0, true, false);
+  Asm->EOL("Offset of Compilation Unit Info");
 
-        if (IP != InlinedParamMap.end() && IP->second.count(GV) > 0) {
-          if (TimePassesIsEnabled)
-            DebugTimer->stopTimer();
-          return;
-        }
+  EmitDifference("info_end", Unit->getID(), "info_begin", Unit->getID(),
+                 true);
+  Asm->EOL("Compilation Unit Length");
 
-        // or GV is an inlined local variable.
-        Scope = AI->second;
-        InlinedParamMap[V].insert(GV);
-        InlinedFnVar = true;
-      } else {
-        // or GV is a local variable.
-        Scope = getOrCreateScope(V);
-      }
-    }
+  StringMap<DIE*> &Globals = Unit->getGlobals();
+  for (StringMap<DIE*>::const_iterator
+         GI = Globals.begin(), GE = Globals.end(); GI != GE; ++GI) {
+    const char *Name = GI->getKeyData();
+    DIE * Entity = GI->second;
+
+    Asm->EmitInt32(Entity->getOffset()); Asm->EOL("DIE offset");
+    Asm->EmitString(Name, strlen(Name)); Asm->EOL("External Name");
   }
 
-  assert(Scope && "Unable to find the variable's scope");
-  DbgVariable *DV = new DbgVariable(DIVariable(GV), FrameIndex, InlinedFnVar);
-  Scope->AddVariable(DV);
+  Asm->EmitInt32(0); Asm->EOL("End Mark");
+  EmitLabel("pubnames_end", Unit->getID());
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  Asm->EOL();
 }
 
-//// RecordInlinedFnStart - Indicate the start of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnStart(DISubprogram &SP, DICompileUnit CU,
-                                          unsigned Line, unsigned Col) {
-  unsigned LabelID = MMI->NextLabelID();
+/// EmitDebugPubNames - Emit visible names into a debug pubnames section.
+///
+void DwarfDebug::EmitDebugPubNames() {
+  // Start the dwarf pubnames section.
+  Asm->SwitchToDataSection(TAI->getDwarfPubNamesSection());
 
-  if (!TAI->doesDwarfUsesInlineInfoSection())
-    return LabelID;
+  if (MainCU) {
+    EmitDebugPubNamesPerCU(MainCU);
+    return;
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  for (unsigned i = 0, e = CompileUnits.size(); i != e; ++i)
+    EmitDebugPubNamesPerCU(CompileUnits[i]);
+}
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<const GlobalVariable *, DbgScope *>::iterator
-    II = AbstractInstanceRootMap.find(GV);
+/// EmitDebugStr - Emit visible names into a debug str section.
+///
+void DwarfDebug::EmitDebugStr() {
+  // Check to see if it is worth the effort.
+  if (!StringPool.empty()) {
+    // Start the dwarf str section.
+    Asm->SwitchToDataSection(TAI->getDwarfStrSection());
 
-  if (II == AbstractInstanceRootMap.end()) {
-    // Create an abstract instance entry for this inlined function if it doesn't
-    // already exist.
-    DbgScope *Scope = new DbgScope(NULL, DIDescriptor(GV));
+    // For each of strings in the string pool.
+    for (unsigned StringID = 1, N = StringPool.size();
+         StringID <= N; ++StringID) {
+      // Emit a label for reference from debug information entries.
+      EmitLabel("string", StringID);
 
-    // Get the compile unit context.
-    CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-    DIE *SPDie = Unit->getDieMapSlotFor(GV);
-    if (!SPDie)
-      SPDie = CreateSubprogramDIE(Unit, SP, false, true);
+      // Emit the string itself.
+      const std::string &String = StringPool[StringID];
+      Asm->EmitString(String); Asm->EOL();
+    }
 
-    // Mark as being inlined. This makes this subprogram entry an abstract
-    // instance root.
-    // FIXME: Our debugger doesn't care about the value of DW_AT_inline, only
-    // that it's defined. That probably won't change in the future. However,
-    // this could be more elegant.
-    AddUInt(SPDie, dwarf::DW_AT_inline, 0, dwarf::DW_INL_declared_not_inlined);
+    Asm->EOL();
+  }
+}
 
-    // Keep track of the abstract scope for this function.
-    DbgAbstractScopeMap[GV] = Scope;
+/// EmitDebugLoc - Emit visible names into a debug loc section.
+///
+void DwarfDebug::EmitDebugLoc() {
+  // Start the dwarf loc section.
+  Asm->SwitchToDataSection(TAI->getDwarfLocSection());
+  Asm->EOL();
+}
 
-    AbstractInstanceRootMap[GV] = Scope;
-    AbstractInstanceRootList.push_back(Scope);
-  }
+/// EmitDebugARanges - Emit visible names into a debug aranges section.
+///
+void DwarfDebug::EmitDebugARanges() {
+  // Start the dwarf aranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfARangesSection());
 
-  // Create a concrete inlined instance for this inlined function.
-  DbgConcreteScope *ConcreteScope = new DbgConcreteScope(DIDescriptor(GV));
-  DIE *ScopeDie = new DIE(dwarf::DW_TAG_inlined_subroutine);
-  CompileUnit *Unit = &FindCompileUnit(SP.getCompileUnit());
-  ScopeDie->setAbstractCompileUnit(Unit);
+  // FIXME - Mock up
+#if 0
+  CompileUnit *Unit = GetBaseCompileUnit();
 
-  DIE *Origin = Unit->getDieMapSlotFor(GV);
-  AddDIEEntry(ScopeDie, dwarf::DW_AT_abstract_origin,
-              dwarf::DW_FORM_ref4, Origin);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_file, 0, Unit->getID());
-  AddUInt(ScopeDie, dwarf::DW_AT_call_line, 0, Line);
-  AddUInt(ScopeDie, dwarf::DW_AT_call_column, 0, Col);
+  // Don't include size of length
+  Asm->EmitInt32(0x1c); Asm->EOL("Length of Address Ranges Info");
 
-  ConcreteScope->setDie(ScopeDie);
-  ConcreteScope->setStartLabelID(LabelID);
-  MMI->RecordUsedDbgLabel(LabelID);
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
 
-  LexicalScopeStack.back()->AddConcreteInst(ConcreteScope);
+  EmitReference("info_begin", Unit->getID());
+  Asm->EOL("Offset of Compilation Unit Info");
 
-  // Keep track of the concrete scope that's inlined into this function.
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    SI = DbgConcreteScopeMap.find(GV);
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Size of Address");
 
-  if (SI == DbgConcreteScopeMap.end())
-    DbgConcreteScopeMap[GV].push_back(ConcreteScope);
-  else
-    SI->second.push_back(ConcreteScope);
+  Asm->EmitInt8(0); Asm->EOL("Size of Segment Descriptor");
 
-  // Track the start label for this inlined function.
-  DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
-    I = InlineInfo.find(GV);
+  Asm->EmitInt16(0);  Asm->EOL("Pad (1)");
+  Asm->EmitInt16(0);  Asm->EOL("Pad (2)");
 
-  if (I == InlineInfo.end())
-    InlineInfo[GV].push_back(LabelID);
-  else
-    I->second.push_back(LabelID);
+  // Range 1
+  EmitReference("text_begin", 0); Asm->EOL("Address");
+  EmitDifference("text_end", 0, "text_begin", 0, true); Asm->EOL("Length");
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  Asm->EmitInt32(0); Asm->EOL("EOM (1)");
+  Asm->EmitInt32(0); Asm->EOL("EOM (2)");
+#endif
 
-  return LabelID;
+  Asm->EOL();
 }
 
-/// RecordInlinedFnEnd - Indicate the end of inlined subroutine.
-unsigned DwarfDebug::RecordInlinedFnEnd(DISubprogram &SP) {
+/// EmitDebugRanges - Emit visible names into a debug ranges section.
+///
+void DwarfDebug::EmitDebugRanges() {
+  // Start the dwarf ranges section.
+  Asm->SwitchToDataSection(TAI->getDwarfRangesSection());
+  Asm->EOL();
+}
+
+/// EmitDebugMacInfo - Emit visible names into a debug macinfo section.
+///
+void DwarfDebug::EmitDebugMacInfo() {
+  if (TAI->doesSupportMacInfoSection()) {
+    // Start the dwarf macinfo section.
+    Asm->SwitchToDataSection(TAI->getDwarfMacInfoSection());
+    Asm->EOL();
+  }
+}
+
+/// EmitDebugInlineInfo - Emit inline info using following format.
+/// Section Header:
+/// 1. length of section
+/// 2. Dwarf version number
+/// 3. address size.
+///
+/// Entries (one "entry" for each function that was inlined):
+///
+/// 1. offset into __debug_str section for MIPS linkage name, if exists;
+///   otherwise offset into __debug_str for regular function name.
+/// 2. offset into __debug_str section for regular function name.
+/// 3. an unsigned LEB128 number indicating the number of distinct inlining
+/// instances for the function.
+///
+/// The rest of the entry consists of a {die_offset, low_pc} pair for each
+/// inlined instance; the die_offset points to the inlined_subroutine die in the
+/// __debug_info section, and the low_pc is the starting address for the
+/// inlining instance.
+void DwarfDebug::EmitDebugInlineInfo() {
   if (!TAI->doesDwarfUsesInlineInfoSection())
-    return 0;
+    return;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+  if (!MainCU)
+    return;
 
-  GlobalVariable *GV = SP.getGV();
-  DenseMap<GlobalVariable *, SmallVector<DbgScope *, 8> >::iterator
-    I = DbgConcreteScopeMap.find(GV);
+  Asm->SwitchToDataSection(TAI->getDwarfDebugInlineSection());
+  Asm->EOL();
+  EmitDifference("debug_inlined_end", 1,
+                 "debug_inlined_begin", 1, true);
+  Asm->EOL("Length of Debug Inlined Information Entry");
 
-  if (I == DbgConcreteScopeMap.end()) {
-    // FIXME: Can this situation actually happen? And if so, should it?
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_begin", 1);
 
-    return 0;
-  }
+  Asm->EmitInt16(dwarf::DWARF_VERSION); Asm->EOL("Dwarf Version");
+  Asm->EmitInt8(TD->getPointerSize()); Asm->EOL("Address Size (in bytes)");
 
-  SmallVector<DbgScope *, 8> &Scopes = I->second;
-  assert(!Scopes.empty() && "We should have at least one debug scope!");
-  DbgScope *Scope = Scopes.back(); Scopes.pop_back();
-  unsigned ID = MMI->NextLabelID();
-  MMI->RecordUsedDbgLabel(ID);
-  Scope->setEndLabelID(ID);
+  for (DenseMap<GlobalVariable *, SmallVector<unsigned, 4> >::iterator
+         I = InlineInfo.begin(), E = InlineInfo.end(); I != E; ++I) {
+    GlobalVariable *GV = I->first;
+    SmallVector<unsigned, 4> &Labels = I->second;
+    DISubprogram SP(GV);
+    std::string Name;
+    std::string LName;
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+    SP.getLinkageName(LName);
+    SP.getName(Name);
 
-  return ID;
-}
+    Asm->EmitString(LName.empty() ? Name : LName);
+    Asm->EOL("MIPS linkage name");
 
-/// RecordVariableScope - Record scope for the variable declared by
-/// DeclareMI. DeclareMI must describe TargetInstrInfo::DECLARE. Record scopes
-/// for only inlined subroutine variables. Other variables's scopes are
-/// determined during RecordVariable().
-void DwarfDebug::RecordVariableScope(DIVariable &DV,
-                                     const MachineInstr *DeclareMI) {
-  if (TimePassesIsEnabled)
-    DebugTimer->startTimer();
+    Asm->EmitString(Name); Asm->EOL("Function name");
 
-  DISubprogram SP(DV.getContext().getGV());
+    Asm->EmitULEB128Bytes(Labels.size()); Asm->EOL("Inline count");
 
-  if (SP.isNull()) {
-    if (TimePassesIsEnabled)
-      DebugTimer->stopTimer();
+    for (SmallVector<unsigned, 4>::iterator LI = Labels.begin(),
+           LE = Labels.end(); LI != LE; ++LI) {
+      DIE *SP = MainCU->getDieMapSlotFor(GV);
+      Asm->EmitInt32(SP->getOffset()); Asm->EOL("DIE offset");
 
-    return;
-  }
+      if (TD->getPointerSize() == sizeof(int32_t))
+        O << TAI->getData32bitsDirective();
+      else
+        O << TAI->getData64bitsDirective();
 
-  DenseMap<GlobalVariable *, DbgScope *>::iterator
-    I = DbgAbstractScopeMap.find(SP.getGV());
-  if (I != DbgAbstractScopeMap.end())
-    InlinedVariableScopes[DeclareMI] = I->second;
+      PrintLabelName("label", *LI); Asm->EOL("low_pc");
+    }
+  }
 
-  if (TimePassesIsEnabled)
-    DebugTimer->stopTimer();
+  EmitLabel("debug_inlined_end", 1);
+  Asm->EOL();
 }




From isanbard at gmail.com  Wed May 20 18:22:41 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:22:41 -0000
Subject: [llvm-commits] [llvm] r72195 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <200905202322.n4KNMfib029133@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:22:40 2009
New Revision: 72195

URL: http://llvm.org/viewvc/llvm-project?rev=72195&view=rev
Log:
Add comment for emit section.

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=72195&r1=72194&r2=72195&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:22:40 2009
@@ -1919,6 +1919,10 @@
     DebugTimer->stopTimer();
 }
 
+//===----------------------------------------------------------------------===//
+// Emit Methods
+//===----------------------------------------------------------------------===//
+
 /// SizeAndOffsetDie - Compute the size and offset of a DIE.
 ///
 unsigned DwarfDebug::SizeAndOffsetDie(DIE *Die, unsigned Offset, bool Last) {




From isanbard at gmail.com  Wed May 20 18:24:48 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:24:48 -0000
Subject: [llvm-commits] [llvm] r72196 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <200905202324.n4KNOnmN029198@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:24:48 2009
New Revision: 72196

URL: http://llvm.org/viewvc/llvm-project?rev=72196&view=rev
Log:
Rename 'New*' methods to 'Create*' to be consistent. 'NewString' isn't used.

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=72196&r1=72195&r2=72196&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:24:48 2009
@@ -275,9 +275,9 @@
   }
 }
 
-/// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
-/// entry.
-DIEEntry *DwarfDebug::NewDIEEntry(DIE *Entry) {
+/// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
+/// information entry.
+DIEEntry *DwarfDebug::CreateDIEEntry(DIE *Entry) {
   DIEEntry *Value;
 
   if (Entry) {
@@ -545,7 +545,7 @@
   }
 
   // Set up proxy.
-  Slot = NewDIEEntry();
+  Slot = CreateDIEEntry();
 
   // Construct type.
   DIE Buffer(dwarf::DW_TAG_base_type);
@@ -925,9 +925,9 @@
   return *I->second;
 }
 
-/// NewDbgScopeVariable - Create a new scope variable.
+/// CreateDbgScopeVariable - Create a new scope variable.
 ///
-DIE *DwarfDebug::NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
+DIE *DwarfDebug::CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit) {
   // Get the descriptor.
   const DIVariable &VD = DV->getVariable();
 
@@ -1012,7 +1012,7 @@
   // Add variables to scope.
   SmallVector<DbgVariable *, 8> &Variables = ParentScope->getVariables();
   for (unsigned i = 0, N = Variables.size(); i < N; ++i) {
-    DIE *VariableDie = NewDbgScopeVariable(Variables[i], Unit);
+    DIE *VariableDie = CreateDbgScopeVariable(Variables[i], Unit);
     if (VariableDie) ParentDie->AddChild(VariableDie);
   }
 

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=72196&r1=72195&r2=72196&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Wed May 20 18:24:48 2009
@@ -230,16 +230,9 @@
   ///
   void AssignAbbrevNumber(DIEAbbrev &Abbrev);
 
-  /// NewString - Add a string to the constant pool and returns a label.
-  ///
-  DWLabel NewString(const std::string &String) {
-    unsigned StringID = StringPool.insert(String);
-    return DWLabel("string", StringID);
-  }
-
-  /// NewDIEEntry - Creates a new DIEEntry to be a proxy for a debug information
-  /// entry.
-  DIEEntry *NewDIEEntry(DIE *Entry = NULL);
+  /// CreateDIEEntry - Creates a new DIEEntry to be a proxy for a debug
+  /// information entry.
+  DIEEntry *CreateDIEEntry(DIE *Entry = NULL);
 
   /// SetDIEEntry - Set a DIEEntry once the debug information entry is defined.
   ///
@@ -282,7 +275,7 @@
   /// AddDIEEntry - Add a DIE attribute data and value.
   ///
   void AddDIEEntry(DIE *Die, unsigned Attribute, unsigned Form, DIE *Entry) {
-    Die->AddValue(Attribute, Form, NewDIEEntry(Entry));
+    Die->AddValue(Attribute, Form, CreateDIEEntry(Entry));
   }
 
   /// AddBlock - Add block data.
@@ -346,9 +339,9 @@
   ///
   CompileUnit &FindCompileUnit(DICompileUnit Unit) const;
 
-  /// NewDbgScopeVariable - Create a new scope variable.
+  /// CreateDbgScopeVariable - Create a new scope variable.
   ///
-  DIE *NewDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
+  DIE *CreateDbgScopeVariable(DbgVariable *DV, CompileUnit *Unit);
 
   /// getOrCreateScope - Returns the scope associated with the given descriptor.
   ///




From isanbard at gmail.com  Wed May 20 18:28:57 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:28:57 -0000
Subject: [llvm-commits] [llvm] r72197 - in
 /llvm/trunk/lib/CodeGen/AsmPrinter: DwarfDebug.cpp DwarfDebug.h
Message-ID: <200905202328.n4KNSwh3029315@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:28:48 2009
New Revision: 72197

URL: http://llvm.org/viewvc/llvm-project?rev=72197&view=rev
Log:
Merge 'ConstructFunctionDbgScope' and 'ConstructAbstractDbgScope'.

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=72197&r1=72196&r2=72197&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:28:48 2009
@@ -1092,7 +1092,8 @@
 
 /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
 ///
-void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope) {
+void DwarfDebug::ConstructFunctionDbgScope(DbgScope *RootScope,
+                                           bool AbstractScope) {
   // Exit if there is no root scope.
   if (!RootScope) return;
   DIDescriptor Desc = RootScope->getDesc();
@@ -1111,43 +1112,19 @@
   DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
   assert(SPDie && "Missing subprogram descriptor");
 
-  // Add the function bounds.
-  AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
-           DWLabel("func_begin", SubprogramCount));
-  AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
-           DWLabel("func_end", SubprogramCount));
-  MachineLocation Location(RI->getFrameRegister(*MF));
-  AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
+  if (!AbstractScope) {
+    // Add the function bounds.
+    AddLabel(SPDie, dwarf::DW_AT_low_pc, dwarf::DW_FORM_addr,
+             DWLabel("func_begin", SubprogramCount));
+    AddLabel(SPDie, dwarf::DW_AT_high_pc, dwarf::DW_FORM_addr,
+             DWLabel("func_end", SubprogramCount));
+    MachineLocation Location(RI->getFrameRegister(*MF));
+    AddAddress(SPDie, dwarf::DW_AT_frame_base, Location);
+  }
 
   ConstructDbgScope(RootScope, 0, 0, SPDie, Unit);
 }
 
-/// ConstructFunctionDbgScope - Construct the scope for the abstract debug
-/// scope.
-///
-void DwarfDebug::ConstructAbstractDbgScope(DbgScope *AbsScope) {
-  // Exit if there is no root scope.
-  if (!AbsScope) return;
-
-  DIDescriptor Desc = AbsScope->getDesc();
-  if (Desc.isNull())
-    return;
-
-  // Get the subprogram debug information entry.
-  DISubprogram SPD(Desc.getGV());
-
-  // Get the compile unit context.
-  CompileUnit *Unit = MainCU;
-  if (!Unit)
-    Unit = &FindCompileUnit(SPD.getCompileUnit());
-
-  // Get the subprogram die.
-  DIE *SPDie = Unit->getDieMapSlotFor(SPD.getGV());
-  assert(SPDie && "Missing subprogram descriptor");
-
-  ConstructDbgScope(AbsScope, 0, 0, SPDie, Unit);
-}
-
 /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
 ///
 void DwarfDebug::ConstructDefaultDbgScope(MachineFunction *MF) {
@@ -1587,7 +1564,7 @@
   for (SmallVector<DbgScope *, 32>::iterator
          I = AbstractInstanceRootList.begin(),
          E = AbstractInstanceRootList.end(); I != E; ++I)
-    ConstructAbstractDbgScope(*I);
+    ConstructFunctionDbgScope(*I);
 
   // Construct scopes for subprogram.
   if (FunctionDbgScope)

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h?rev=72197&r1=72196&r2=72197&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.h Wed May 20 18:28:48 2009
@@ -355,12 +355,8 @@
 
   /// ConstructFunctionDbgScope - Construct the scope for the subprogram.
   ///
-  void ConstructFunctionDbgScope(DbgScope *RootScope);
-
-  /// ConstructFunctionDbgScope - Construct the scope for the abstract debug
-  /// scope.
-  ///
-  void ConstructAbstractDbgScope(DbgScope *AbsScope);
+  void ConstructFunctionDbgScope(DbgScope *RootScope,
+                                 bool AbstractScope = false);
 
   /// ConstructDefaultDbgScope - Construct a default scope for the subprogram.
   ///




From isanbard at gmail.com  Wed May 20 18:31:55 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 23:31:55 -0000
Subject: [llvm-commits] [llvm] r72198 -
	/llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
Message-ID: <200905202331.n4KNVuoI029406@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 18:31:45 2009
New Revision: 72198

URL: http://llvm.org/viewvc/llvm-project?rev=72198&view=rev
Log:
Minor code cleanup. No functionality change.

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=72198&r1=72197&r2=72198&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/DwarfDebug.cpp Wed May 20 18:31:45 2009
@@ -67,36 +67,25 @@
   FoldingSet<DIE> DiesSet;
 public:
   CompileUnit(unsigned I, DIE *D)
-    : ID(I), Die(D), GVToDieMap(),
-      GVToDIEEntryMap(), Globals(), DiesSet(InitDiesSetSize)
-  {}
-
-  ~CompileUnit() {
-    delete Die;
-  }
+    : ID(I), Die(D), DiesSet(InitDiesSetSize) {}
+  ~CompileUnit() { delete Die; }
 
   // Accessors.
-  unsigned getID()           const { return ID; }
-  DIE* getDie()              const { return Die; }
+  unsigned getID() const { return ID; }
+  DIE* getDie() const { return Die; }
   StringMap<DIE*> &getGlobals() { return Globals; }
 
   /// hasContent - Return true if this compile unit has something to write out.
   ///
-  bool hasContent() const {
-    return !Die->getChildren().empty();
-  }
+  bool hasContent() const { return !Die->getChildren().empty(); }
 
   /// AddGlobal - Add a new global entity to the compile unit.
   ///
-  void AddGlobal(const std::string &Name, DIE *Die) {
-    Globals[Name] = Die;
-  }
+  void AddGlobal(const std::string &Name, DIE *Die) { Globals[Name] = Die; }
 
   /// getDieMapSlotFor - Returns the debug information entry map slot for the
   /// specified debug variable.
-  DIE *&getDieMapSlotFor(GlobalVariable *GV) {
-    return GVToDieMap[GV];
-  }
+  DIE *&getDieMapSlotFor(GlobalVariable *GV) { return GVToDieMap[GV]; }
 
   /// getDIEEntrySlotFor - Returns the debug information entry proxy slot for the
   /// specified debug variable.




From isanbard at gmail.com  Wed May 20 18:54:12 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 16:54:12 -0700
Subject: [llvm-commits] [llvm] r72191 - in /llvm/trunk:
	include/llvm/CodeGen/DebugLoc.h include/llvm/CodeGen/FastISel.h
	include/llvm/CodeGen/MachineFunction.h
	lib/CodeGen/AsmPrinter/AsmPrinter.cpp
	lib/CodeGen/DebugLoc.cpp lib/CodeGen/MachineFunction.cpp l
Message-ID: <16e5fdf90905201654u39cbcb62t7787674728a5d87c@mail.gmail.com>

Argiris,

This is causing this failure during an llvm-gcc bootstrap. Please investigate.

Assertion failed: (!CurScope.isInvalid() && "Mismatched region.end
?"), function ExitDebugScope, file
/Volumes/Sandbox/Buildbot/llvm/full-llvm/build/llvm.src/lib/CodeGen/DebugLoc.cpp,
line 30.
../../llvm-gcc.src/gcc/libgcc2.c:1147: internal compiler error: Abort trap
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://llvm.org/bugs/> for instructions.
make[4]: *** [libgcc/./_umoddi3.o] Error 1
make[4]: *** Waiting for unfinished jobs....
make[3]: *** [stmp-multilib] Error 2
rm fsf-funding.pod gcov.pod gfdl.pod cpp.pod gpl.pod gcc.pod
make[2]: *** [all-stage1-gcc] Error 2
make[1]: *** [stage1-bubble] Error 2
make: *** [all] Error 2

-bw

On Wed, May 20, 2009 at 3:57 PM, Argiris Kirtzidis <akyrtzi at gmail.com> wrote:
> Author: akirtzidis
> Date: Wed May 20 17:57:17 2009
> New Revision: 72191
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72191&view=rev
> Log:
> Introduce DebugScope which gets embedded into the machine instructions' DebugLoc.
> DebugScope refers to a debug region, function or block.
>
> Added:
> ? ?llvm/trunk/lib/CodeGen/DebugLoc.cpp
> Modified:
> ? ?llvm/trunk/include/llvm/CodeGen/DebugLoc.h
> ? ?llvm/trunk/include/llvm/CodeGen/FastISel.h
> ? ?llvm/trunk/include/llvm/CodeGen/MachineFunction.h
> ? ?llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
> ? ?llvm/trunk/lib/CodeGen/MachineFunction.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
>
> Modified: llvm/trunk/include/llvm/CodeGen/DebugLoc.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DebugLoc.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/DebugLoc.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/DebugLoc.h Wed May 20 17:57:17 2009
> @@ -20,18 +20,77 @@
>
> ?namespace llvm {
> ? class GlobalVariable;
> + ?class MachineFunction;
>
> - ?/// DebugLocTuple - Debug location tuple of filename id, line and column.
> + ?/// DebugScope - Debug scope id. This is carried into DebugLocTuple to index
> + ?/// into a vector of DebugScopeInfos. Provides the ability to associate a
> + ?/// SDNode/MachineInstr with the debug scope that it belongs to.
> + ?///
> + ?class DebugScope {
> + ? ?unsigned Idx;
> +
> + ?public:
> + ? ?DebugScope() : Idx(~0U) {} ?// Defaults to null.
> +
> + ? ?static DebugScope getInvalid() ? { DebugScope S; S.Idx = ~0U; return S; }
> + ? ?static DebugScope get(unsigned idx) { DebugScope S; S.Idx = idx; return S; }
> +
> + ? ?unsigned getIndex() const { return Idx; }
> +
> + ? ?/// isInvalid - Return true if it doesn't refer to any scope.
> + ? ?bool isInvalid() const { return Idx == ~0U; }
> +
> + ? ?bool operator==(const DebugScope &DS) const { return Idx == DS.Idx; }
> + ? ?bool operator!=(const DebugScope &DS) const { return !(*this == DS); }
> + ?};
> +
> + ?/// DebugScopeInfo - Contains info about the scope global variable and the
> + ?/// parent debug scope. DebugScope is only a "cookie" that can point to a
> + ?/// specific DebugScopeInfo.
> + ?///
> + ?struct DebugScopeInfo {
> + ? ?GlobalVariable *GV;
> + ? ?DebugScope Parent;
> +
> + ? ?DebugScopeInfo(GlobalVariable *gv, DebugScope parent)
> + ? ? ?: GV(gv), Parent(parent) {}
> + ? ?DebugScopeInfo()
> + ? ? ?: GV(0), Parent(DebugScope::getInvalid()) {}
> + ?};
> +
> + ?/// DebugScopeTracker - Create and keep track of the debug scope while
> + ?/// entering/exiting subprograms and blocks. Used by the instruction
> + ?/// selectors.
> + ?///
> + ?class DebugScopeTracker {
> + ? ?DebugScope CurScope;
> +
> + ?public:
> + ? ?/// getCurScope - The current debug scope that we "entered" through
> + ? ?/// EnterDebugScope.
> + ? ?DebugScope getCurScope() const { return CurScope; }
> +
> + ? ?/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
> + ? ?/// or a DIBlock.
> + ? ?void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
> +
> + ? ?/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
> + ? ?void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
> + ?};
> +
> + ?/// DebugLocTuple - Debug location tuple of a DICompileUnit global variable,
> + ?/// debug scope, line and column.
> ? ///
> ? struct DebugLocTuple {
> ? ? GlobalVariable *CompileUnit;
> + ? ?DebugScope Scope;
> ? ? unsigned Line, Col;
>
> - ? ?DebugLocTuple(GlobalVariable *v, unsigned l, unsigned c)
> - ? ? ?: CompileUnit(v), Line(l), Col(c) {};
> + ? ?DebugLocTuple(GlobalVariable *v, DebugScope s, unsigned l, unsigned c)
> + ? ? ?: CompileUnit(v), Scope(s), Line(l), Col(c) {};
>
> ? ? bool operator==(const DebugLocTuple &DLT) const {
> - ? ? ?return CompileUnit == DLT.CompileUnit &&
> + ? ? ?return CompileUnit == DLT.CompileUnit && Scope == DLT.Scope &&
> ? ? ? ? ? ? ?Line == DLT.Line && Col == DLT.Col;
> ? ? }
> ? ? bool operator!=(const DebugLocTuple &DLT) const {
> @@ -63,18 +122,20 @@
> ? // Partially specialize DenseMapInfo for DebugLocTyple.
> ? template<> ?struct DenseMapInfo<DebugLocTuple> {
> ? ? static inline DebugLocTuple getEmptyKey() {
> - ? ? ?return DebugLocTuple(0, ~0U, ~0U);
> + ? ? ?return DebugLocTuple(0, DebugScope::getInvalid(), ~0U, ~0U);
> ? ? }
> ? ? static inline DebugLocTuple getTombstoneKey() {
> - ? ? ?return DebugLocTuple((GlobalVariable*)~1U, ~1U, ~1U);
> + ? ? ?return DebugLocTuple((GlobalVariable*)~1U,DebugScope::get(~1U), ~1U, ~1U);
> ? ? }
> ? ? static unsigned getHashValue(const DebugLocTuple &Val) {
> ? ? ? return DenseMapInfo<GlobalVariable*>::getHashValue(Val.CompileUnit) ^
> + ? ? ? ? ? ? DenseMapInfo<unsigned>::getHashValue(Val.Scope.getIndex()) ^
> ? ? ? ? ? ? ?DenseMapInfo<unsigned>::getHashValue(Val.Line) ^
> ? ? ? ? ? ? ?DenseMapInfo<unsigned>::getHashValue(Val.Col);
> ? ? }
> ? ? static bool isEqual(const DebugLocTuple &LHS, const DebugLocTuple &RHS) {
> ? ? ? return LHS.CompileUnit == RHS.CompileUnit &&
> + ? ? ? ? ? ? LHS.Scope ? ? ? == RHS.Scope &&
> ? ? ? ? ? ? ?LHS.Line ? ? ? ?== RHS.Line &&
> ? ? ? ? ? ? ?LHS.Col ? ? ? ? == RHS.Col;
> ? ? }
>
> Modified: llvm/trunk/include/llvm/CodeGen/FastISel.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/FastISel.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/FastISel.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/FastISel.h Wed May 20 17:57:17 2009
> @@ -57,6 +57,7 @@
> ? MachineFrameInfo &MFI;
> ? MachineConstantPool &MCP;
> ? DebugLoc DL;
> + ?DebugScopeTracker DbgScopeTrack;
> ? const TargetMachine &TM;
> ? const TargetData &TD;
> ? const TargetInstrInfo &TII;
>
> Modified: llvm/trunk/include/llvm/CodeGen/MachineFunction.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineFunction.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/MachineFunction.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/MachineFunction.h Wed May 20 17:57:17 2009
> @@ -111,6 +111,8 @@
> ? // Tracks debug locations.
> ? DebugLocTracker DebugLocInfo;
>
> + ?std::vector<DebugScopeInfo> DbgScopeInfos;
> +
> ?public:
> ? MachineFunction(const Function *Fn, const TargetMachine &TM);
> ? ~MachineFunction();
> @@ -332,7 +334,7 @@
> ? /// getOrCreateDebugLocID - Look up the DebugLocTuple index with the given
> ? /// source file, line, and column. If none currently exists, create a new
> ? /// DebugLocTuple, and insert it into the DebugIdMap.
> - ?unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit,
> + ?unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit, DebugScope Scope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?unsigned Line, unsigned Col);
>
> ? /// getDebugLocTuple - Get the DebugLocTuple for a given DebugLoc object.
> @@ -345,6 +347,12 @@
> ? /// setDefaultDebugLoc - Get the default debug location for the machine
> ? /// function.
> ? void setDefaultDebugLoc(DebugLoc DL) { DefaultDebugLoc = DL; }
> +
> + ?/// CreateDebugScope - Create a new debug scope.
> + ?DebugScope CreateDebugScope(GlobalVariable *ScopeGV, DebugScope Parent);
> +
> + ?/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
> + ?const DebugScopeInfo &getDebugScopeInfo(DebugScope DS) const;
> ?};
>
> ?//===--------------------------------------------------------------------===//
>
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Wed May 20 17:57:17 2009
> @@ -1311,7 +1311,7 @@
> ?void AsmPrinter::processDebugLoc(DebugLoc DL) {
> ? if (TAI->doesSupportDebugInformation() && DW->ShouldEmitDwarfDebug()) {
> ? ? if (!DL.isUnknown()) {
> - ? ? ?static DebugLocTuple PrevDLT(0, ~0U, ~0U);
> + ? ? ?static DebugLocTuple PrevDLT(0, DebugScope::getInvalid(), ~0U, ~0U);
> ? ? ? DebugLocTuple CurDLT = MF->getDebugLocTuple(DL);
>
> ? ? ? if (CurDLT.CompileUnit != 0 && PrevDLT != CurDLT)
>
> Added: llvm/trunk/lib/CodeGen/DebugLoc.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DebugLoc.cpp?rev=72191&view=auto
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/DebugLoc.cpp (added)
> +++ llvm/trunk/lib/CodeGen/DebugLoc.cpp Wed May 20 17:57:17 2009
> @@ -0,0 +1,38 @@
> +//===-- DebugLoc.cpp ------------------------------------------------------===//
> +//
> +// ? ? ? ? ? ? ? ? ? ? The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +//
> +// Implementation for DebugScopeTracker.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#include "llvm/CodeGen/DebugLoc.h"
> +#include "llvm/CodeGen/MachineFunction.h"
> +using namespace llvm;
> +
> +/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
> +/// or a DIBlock.
> +void DebugScopeTracker::EnterDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?MachineFunction &MF) {
> + ?assert(ScopeGV && "GlobalVariable for scope is null!");
> + ?CurScope = MF.CreateDebugScope(ScopeGV, CurScope);
> +}
> +
> +/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
> +void DebugScopeTracker::ExitDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? MachineFunction &MF) {
> + ?assert(ScopeGV && "GlobalVariable for scope is null!");
> + ?assert(!CurScope.isInvalid() && "Mismatched region.end ?");
> + ?// We may have skipped a region.end because it was in an unreachable block.
> + ?// Go up the scope chain until we reach the scope that ScopeGV points to.
> + ?DebugScopeInfo DSI;
> + ?do {
> + ? ?DSI = ?MF.getDebugScopeInfo(CurScope);
> + ? ?CurScope = DSI.Parent;
> + ?} while (!DSI.Parent.isInvalid() && DSI.GV != ScopeGV);
> +}
>
> Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
> +++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Wed May 20 17:57:17 2009
> @@ -403,8 +403,9 @@
> ?/// source file, line, and column. If none currently exists, create a new
> ?/// DebugLocTuple, and insert it into the DebugIdMap.
> ?unsigned MachineFunction::getOrCreateDebugLocID(GlobalVariable *CompileUnit,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DebugScope Scope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? unsigned Line, unsigned Col) {
> - ?DebugLocTuple Tuple(CompileUnit, Line, Col);
> + ?DebugLocTuple Tuple(CompileUnit, Scope, Line, Col);
> ? DenseMap<DebugLocTuple, unsigned>::iterator II
> ? ? = DebugLocInfo.DebugIdMap.find(Tuple);
> ? if (II != DebugLocInfo.DebugIdMap.end())
> @@ -424,6 +425,21 @@
> ? return DebugLocInfo.DebugLocations[Idx];
> ?}
>
> +/// CreateDebugScope - Create a new debug scope.
> +DebugScope MachineFunction::CreateDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DebugScope Parent) {
> + ?DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
> + ?return DebugScope::get(DbgScopeInfos.size() - 1);
> +}
> +
> +/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
> +const DebugScopeInfo &MachineFunction::getDebugScopeInfo(DebugScope DS) const {
> + ?unsigned Idx = DS.getIndex();
> + ?assert(Idx < DbgScopeInfos.size() && "Invalid index into debug scopes!");
> + ?return DbgScopeInfos[Idx];
> +}
> +
> +
> ?//===----------------------------------------------------------------------===//
> ?// ?MachineFrameInfo implementation
> ?//===----------------------------------------------------------------------===//
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Wed May 20 17:57:17 2009
> @@ -331,17 +331,22 @@
> ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
> ? ? ? unsigned Line = SPI->getLine();
> ? ? ? unsigned Col = SPI->getColumn();
> - ? ? ?unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(), Line, Col);
> + ? ? ?unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, Col);
> ? ? ? setCurDebugLoc(DebugLoc::get(Idx));
> ? ? }
> ? ? return true;
> ? }
> ? case Intrinsic::dbg_region_start: {
> ? ? DbgRegionStartInst *RSI = cast<DbgRegionStartInst>(I);
> - ? ?if (DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?unsigned ID =
> - ? ? ? ?DW->RecordRegionStart(cast<GlobalVariable>(RSI->getContext()));
> + ? ?if (!DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None))
> + ? ? ?return true;
> +
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(RSI->getContext());
> + ? ?DbgScopeTrack.EnterDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?unsigned ID = DW->RecordRegionStart(Rgn);
> ? ? ? const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
> ? ? ? BuildMI(MBB, DL, II).addImm(ID);
> ? ? }
> @@ -349,10 +354,14 @@
> ? }
> ? case Intrinsic::dbg_region_end: {
> ? ? DbgRegionEndInst *REI = cast<DbgRegionEndInst>(I);
> - ? ?if (DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> + ? ?if (!DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None))
> + ? ? ?return true;
> +
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(REI->getContext());
> + ? ?DbgScopeTrack.ExitDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ?unsigned ID = 0;
> - ? ? DISubprogram Subprogram(cast<GlobalVariable>(REI->getContext()));
> + ? ? DISubprogram Subprogram(Rgn);
> ? ? ?if (!Subprogram.isNull() && !Subprogram.describes(MF.getFunction())) {
> ? ? ? ? // This is end of an inlined function.
> ? ? ? ? const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
> @@ -382,6 +391,7 @@
> ? ? DebugLoc PrevLoc = DL;
> ? ? DISubprogram Subprogram(cast<GlobalVariable>(SP));
> ? ? DICompileUnit CompileUnit = Subprogram.getCompileUnit();
> + ? ?DbgScopeTrack.EnterDebugScope(Subprogram.getGV(), MF);
>
> ? ? if (!Subprogram.describes(MF.getFunction())) {
> ? ? ? // This is a beginning of an inlined function.
> @@ -393,8 +403,10 @@
> ? ? ? ? return true;
> ? ? ? // Record the source line.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ?setCurDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CompileUnit.getGV(), Line, 0)));
> + ? ? ?setCurDebugLoc(
> + ? ? ? ?DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Line, 0)));
>
> ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
> @@ -408,8 +420,10 @@
> ? ? } else {
> ? ? ? // Record the source line.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ?MF.setDefaultDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CompileUnit.getGV(), Line, 0)));
> + ? ? ?MF.setDefaultDebugLoc(
> + ? ? ? ?DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Line, 0)));
> ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? // llvm.dbg.func_start also defines beginning of function scope.
> ? ? ? ? DW->RecordRegionStart(cast<GlobalVariable>(FSI->getSubprogram()));
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Wed May 20 17:57:17 2009
> @@ -312,6 +312,17 @@
> ? ? ? ? ? ? !StaticAllocaMap.count(cast<AllocaInst>(I)))
> ? ? ? ? ? InitializeRegForValue(I);
>
> + ?// FIXME: PHI instructions currently get an invalid scope. This is because
> + ?// tracking of debug scopes is done through a simple mechanism where
> + ?// "entering" a scope implies that the scope is entered for the first time.
> + ?// If we keep track of debug scopes for the following loop, the PHI
> + ?// instructions would get scopes that will not be used again later by the
> + ?// instruction selectors.
> + ?// Either provide a mechanism to re-enter a previously created scope or wait
> + ?// for when the DebugLoc is retrieved from Instruction, in which case the
> + ?// current debug scope tracking mechanism will be obsolete.
> + ?DebugScope DbgScope;
> +
> ? // Create an initial MachineBasicBlock for each LLVM BasicBlock in F. ?This
> ? // also creates the initial PHI MachineInstrs, though none of the input
> ? // operands are populated.
> @@ -337,6 +348,7 @@
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CodeGenOpt::Default)) {
> ? ? ? ? ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
> ? ? ? ? ? ? ? unsigned idx = MF->getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI->getLine(),
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI->getColumn());
> ? ? ? ? ? ? ? DL = DebugLoc::get(idx);
> @@ -352,7 +364,7 @@
> ? ? ? ? ? ? ? DISubprogram Subprogram(cast<GlobalVariable>(SP));
> ? ? ? ? ? ? ? DICompileUnit CU(Subprogram.getCompileUnit());
> ? ? ? ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ? ? ? ? ?DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ?DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(), DbgScope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0));
> ? ? ? ? ? ? }
>
> @@ -3909,7 +3921,9 @@
> ? ? ? MachineFunction &MF = DAG.getMachineFunction();
> ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI.getContext()));
> ? ? ? DebugLoc Loc = DebugLoc::get(MF.getOrCreateDebugLocID(CU.getGV(),
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI.getLine(), SPI.getColumn()));
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? SPI.getLine(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? SPI.getColumn()));
> ? ? ? setCurDebugLoc(Loc);
>
> ? ? ? if (OptLevel == CodeGenOpt::None)
> @@ -3923,11 +3937,14 @@
> ? case Intrinsic::dbg_region_start: {
> ? ? DwarfWriter *DW = DAG.getDwarfWriter();
> ? ? DbgRegionStartInst &RSI = cast<DbgRegionStartInst>(I);
> + ? ?if (!DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel))
> + ? ? ?return 0;
>
> - ? ?if (DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?unsigned LabelID =
> - ? ? ? ?DW->RecordRegionStart(cast<GlobalVariable>(RSI.getContext()));
> + ? ?MachineFunction &MF = DAG.getMachineFunction();
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(RSI.getContext());
> + ? ?DbgScopeTrack.EnterDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?unsigned LabelID = DW->RecordRegionStart(Rgn);
> ? ? ? DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?getRoot(), LabelID));
> ? ? }
> @@ -3937,11 +3954,14 @@
> ? case Intrinsic::dbg_region_end: {
> ? ? DwarfWriter *DW = DAG.getDwarfWriter();
> ? ? DbgRegionEndInst &REI = cast<DbgRegionEndInst>(I);
> + ? ?if (!DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel))
> + ? ? ?return 0;
>
> - ? ?if (DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?MachineFunction &MF = DAG.getMachineFunction();
> - ? ? ?DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
> + ? ?MachineFunction &MF = DAG.getMachineFunction();
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(REI.getContext());
> + ? ?DbgScopeTrack.ExitDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?DISubprogram Subprogram(Rgn);
>
> ? ? ? if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
> ? ? ? ? unsigned LabelID =
> @@ -3974,6 +3994,7 @@
> ? ? ? return 0;
>
> ? ? MachineFunction &MF = DAG.getMachineFunction();
> + ? ?DbgScopeTrack.EnterDebugScope(cast<GlobalVariable>(SP), MF);
> ? ? if (OptLevel == CodeGenOpt::None) {
> ? ? ? // llvm.dbg.func.start implicitly defines a dbg_stoppoint which is what
> ? ? ? // (most?) gdb expects.
> @@ -3993,7 +4014,9 @@
> ? ? ? ? // Record the source line.
> ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? ? setCurDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
>
> ? ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? ? DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
> @@ -4008,7 +4031,9 @@
> ? ? ? ? // Record the source line.
> ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? ? MF.setDefaultDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
> ? ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? ? // llvm.dbg.func_start also defines beginning of function scope.
> ? ? ? ? ? DW->RecordRegionStart(cast<GlobalVariable>(FSI.getSubprogram()));
> @@ -4035,7 +4060,9 @@
> ? ? ? // create a label if this is a beginning of inlined function.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? setCurDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
> ? ? ? // FIXME - ?Start new region because llvm.dbg.func_start also defines
> ? ? ? // beginning of function scope.
> ? ? }
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h Wed May 20 17:57:17 2009
> @@ -169,6 +169,8 @@
> ? /// CurDebugLoc - current file + line number. ?Changes as we build the DAG.
> ? DebugLoc CurDebugLoc;
>
> + ?DebugScopeTracker DbgScopeTrack;
> +
> ? DenseMap<const Value*, SDValue> NodeMap;
>
> ? /// PendingLoads - Loads are not emitted to the program immediately. ?We bunch
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From howard0su at gmail.com  Wed May 20 19:03:37 2009
From: howard0su at gmail.com (Howard Su)
Date: Thu, 21 May 2009 08:03:37 +0800
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
	include/llvm/System/Atomic.h lib/System/Atomic.cpp
In-Reply-To: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
References: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
Message-ID: <f126fae00905201703u137104feg29e30f608d16344@mail.gmail.com>

don't inlcude platform dependent header here. please include
"llvm/Support/DataTypes.h" instead.

Also, please update CMakeLists file also to add Atomic.cpp.

On Thu, May 21, 2009 at 2:26 AM, Owen Anderson <resistor at mac.com> wrote:

> Author: resistor
> Date: Wed May 20 13:26:15 2009
> New Revision: 72174
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72174&view=rev
> Log:
> Move atomic operations' definitions out of line.  While this seems kind of
> silly,
> all kinds of problems caused by including windows.h and/or config.h in an
> LLVM header.
>
> Added:
>    llvm/trunk/lib/System/Atomic.cpp
> Modified:
>    llvm/trunk/include/llvm/System/Atomic.h
>
> Modified: llvm/trunk/include/llvm/System/Atomic.h
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72174&r1=72173&r2=72174&view=diff
>
>
> ==============================================================================
> --- llvm/trunk/include/llvm/System/Atomic.h (original)
> +++ llvm/trunk/include/llvm/System/Atomic.h Wed May 20 13:26:15 2009
> @@ -14,75 +14,16 @@
>  #ifndef LLVM_SYSTEM_ATOMIC_H
>  #define LLVM_SYSTEM_ATOMIC_H
>
> -#include "llvm/Config/config.h"
> -
> -#if defined(_MSC_VER)
> -#define NOMINMAX
> -#include <windows.h>
> -#endif
> -
> +#include <stdint.h>
>
>  namespace llvm {
>   namespace sys {
> -
> -    inline void MemoryFence() {
> -#if LLVM_MULTITHREADED==0
> -      return;
> -#else
> -#  if defined(__GNUC__)
> -      __sync_synchronize();
> -#  elif defined(_MSC_VER)
> -      MemoryBarrier();
> -#  else
> -#    error No memory fence implementation for your platform!
> -#  endif
> -#endif
> -}
> -
> -#if LLVM_MULTITHREADED==0
> -    typedef unsigned long cas_flag;
> -    template<typename T>
> -    inline T CompareAndSwap(volatile T* dest,
> -                           T exc, T c) {
> -      T result = *dest;
> -      if (result == c)
> -        *dest = exc;
> -      return result;
> -    }
> -#elif defined(__GNUC__)
> -    typedef unsigned long cas_flag;
> -    template<typename T>
> -    inline T CompareAndSwap(volatile T* ptr,
> -                           T new_value,
> -                           T old_value) {
> -      return __sync_val_compare_and_swap(ptr, old_value, new_value);
> -    }
> -#elif defined(_MSC_VER)
> -    typedef LONG cas_flag;
> -    template<typename T>
> -    inline T CompareAndSwap(volatile T* ptr,
> -                           T new_value,
> -                           T old_value) {
> -      if (sizeof(T) == 4)
> -       return InterlockedCompareExchange(ptr, new_value, old_value);
> -      else if (sizeof(T) == 8)
> -       return InterlockedCompareExchange64(ptr, new_value, old_value);
> -      else
> -       assert(0 && "Unsupported compare-and-swap size!");
> -    }
> -
> -    template<typename T>
> -    inline T* CompareAndSwap<T*>(volatile T** ptr,
> -                                T* new_value,
> -                                T* old_value) {
> -      return InterlockedCompareExchangePtr(ptr, new_value, old_value);
> -    }
> -
> -
> -#else
> -#  error No compare-and-swap implementation for your platform!
> -#endif
> +    void MemoryFence();
>
> +    typedef uint32_t cas_flag;
> +    cas_flag CompareAndSwap(volatile cas_flag* ptr,
> +                             cas_flag new_value,
> +                            cas_flag old_value);
>   }
>  }
>
>
> Added: llvm/trunk/lib/System/Atomic.cpp
> URL:
> http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/Atomic.cpp?rev=72174&view=auto
>
>
> ==============================================================================
> --- llvm/trunk/lib/System/Atomic.cpp (added)
> +++ llvm/trunk/lib/System/Atomic.cpp Wed May 20 13:26:15 2009
> @@ -0,0 +1,52 @@
> +//===-- Atomic.cpp - Atomic Operations --------------------------*- C++
> -*-===//
> +//
> +//                     The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
>
> +//===----------------------------------------------------------------------===//
> +//
> +//  This header file implements atomic operations.
> +//
>
> +//===----------------------------------------------------------------------===//
> +
> +#include "llvm/System/Atomic.h"
> +#include "llvm/Config/config.h"
> +
> +using namespace llvm;
> +
> +#if defined(_MSC_VER)
> +#include <windows.h>
> +#endif
> +
> +void sys::MemoryFence() {
> +#if LLVM_MULTITHREADED==0
> +  return;
> +#else
> +#  if defined(__GNUC__)
> +  __sync_synchronize();
> +#  elif defined(_MSC_VER)
> +  MemoryBarrier();
> +#  else
> +# error No memory fence implementation for your platform!
> +#  endif
> +#endif
> +}
> +
> +sys::cas_flag sys::CompareAndSwap(volatile sys::cas_flag* ptr,
> +                                       sys::cas_flag new_value,
> +                                       sys::cas_flag old_value) {
> +#if LLVM_MULTITHREADED==0
> +  T result = *dest;
> +  if (result == c)
> +    *dest = exc;
> +  return result;
> +#elif defined(__GNUC__)
> +  return __sync_val_compare_and_swap(ptr, old_value, new_value);
> +#elif defined(_MSC_VER)
> +       return InterlockedCompareExchange(ptr, new_value, old_value);
> +#else
> +#  error No compare-and-swap implementation for your platform!
> +#endif
> +}
> \ 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
>



-- 
-Howard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090521/fb14577b/attachment.html 

From isanbard at gmail.com  Wed May 20 19:04:56 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Thu, 21 May 2009 00:04:56 -0000
Subject: [llvm-commits] [llvm] r72200 - in /llvm/trunk:
 include/llvm/CodeGen/DebugLoc.h include/llvm/CodeGen/FastISel.h
 include/llvm/CodeGen/MachineFunction.h
 lib/CodeGen/AsmPrinter/AsmPrinter.cpp lib/CodeGen/DebugLoc.cpp
 lib/CodeGen/MachineFunction.cpp lib/CodeGen/SelectionDAG/FastISel.cpp
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
 lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
Message-ID: <200905210004.n4L04xfX030417@zion.cs.uiuc.edu>

Author: void
Date: Wed May 20 19:04:55 2009
New Revision: 72200

URL: http://llvm.org/viewvc/llvm-project?rev=72200&view=rev
Log:
Temporarily revert r72191. It was causing an assert during llvm-gcc
bootstrapping.

Removed:
    llvm/trunk/lib/CodeGen/DebugLoc.cpp
Modified:
    llvm/trunk/include/llvm/CodeGen/DebugLoc.h
    llvm/trunk/include/llvm/CodeGen/FastISel.h
    llvm/trunk/include/llvm/CodeGen/MachineFunction.h
    llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
    llvm/trunk/lib/CodeGen/MachineFunction.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
    llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h

Modified: llvm/trunk/include/llvm/CodeGen/DebugLoc.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DebugLoc.h?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/DebugLoc.h (original)
+++ llvm/trunk/include/llvm/CodeGen/DebugLoc.h Wed May 20 19:04:55 2009
@@ -20,77 +20,18 @@
 
 namespace llvm {
   class GlobalVariable;
-  class MachineFunction;
 
-  /// DebugScope - Debug scope id. This is carried into DebugLocTuple to index
-  /// into a vector of DebugScopeInfos. Provides the ability to associate a
-  /// SDNode/MachineInstr with the debug scope that it belongs to.
-  ///
-  class DebugScope {
-    unsigned Idx;
-
-  public:
-    DebugScope() : Idx(~0U) {}  // Defaults to null.
-
-    static DebugScope getInvalid()   { DebugScope S; S.Idx = ~0U; return S; }
-    static DebugScope get(unsigned idx) { DebugScope S; S.Idx = idx; return S; }
-
-    unsigned getIndex() const { return Idx; }
-
-    /// isInvalid - Return true if it doesn't refer to any scope.
-    bool isInvalid() const { return Idx == ~0U; }
-
-    bool operator==(const DebugScope &DS) const { return Idx == DS.Idx; }
-    bool operator!=(const DebugScope &DS) const { return !(*this == DS); }
-  };
-
-  /// DebugScopeInfo - Contains info about the scope global variable and the
-  /// parent debug scope. DebugScope is only a "cookie" that can point to a
-  /// specific DebugScopeInfo.
-  ///
-  struct DebugScopeInfo {
-    GlobalVariable *GV;
-    DebugScope Parent;
-
-    DebugScopeInfo(GlobalVariable *gv, DebugScope parent)
-      : GV(gv), Parent(parent) {}
-    DebugScopeInfo()
-      : GV(0), Parent(DebugScope::getInvalid()) {}
-  };
-
-  /// DebugScopeTracker - Create and keep track of the debug scope while
-  /// entering/exiting subprograms and blocks. Used by the instruction
-  /// selectors.
-  ///
-  class DebugScopeTracker {
-    DebugScope CurScope;
-
-  public:
-    /// getCurScope - The current debug scope that we "entered" through
-    /// EnterDebugScope.
-    DebugScope getCurScope() const { return CurScope; }
-
-    /// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
-    /// or a DIBlock.
-    void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
-
-    /// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
-    void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
-  };
-
-  /// DebugLocTuple - Debug location tuple of a DICompileUnit global variable,
-  /// debug scope, line and column.
+  /// DebugLocTuple - Debug location tuple of filename id, line and column.
   ///
   struct DebugLocTuple {
     GlobalVariable *CompileUnit;
-    DebugScope Scope;
     unsigned Line, Col;
 
-    DebugLocTuple(GlobalVariable *v, DebugScope s, unsigned l, unsigned c)
-      : CompileUnit(v), Scope(s), Line(l), Col(c) {};
+    DebugLocTuple(GlobalVariable *v, unsigned l, unsigned c)
+      : CompileUnit(v), Line(l), Col(c) {};
 
     bool operator==(const DebugLocTuple &DLT) const {
-      return CompileUnit == DLT.CompileUnit && Scope == DLT.Scope &&
+      return CompileUnit == DLT.CompileUnit &&
              Line == DLT.Line && Col == DLT.Col;
     }
     bool operator!=(const DebugLocTuple &DLT) const {
@@ -122,20 +63,18 @@
   // Partially specialize DenseMapInfo for DebugLocTyple.
   template<>  struct DenseMapInfo<DebugLocTuple> {
     static inline DebugLocTuple getEmptyKey() {
-      return DebugLocTuple(0, DebugScope::getInvalid(), ~0U, ~0U);
+      return DebugLocTuple(0, ~0U, ~0U);
     }
     static inline DebugLocTuple getTombstoneKey() {
-      return DebugLocTuple((GlobalVariable*)~1U,DebugScope::get(~1U), ~1U, ~1U);
+      return DebugLocTuple((GlobalVariable*)~1U, ~1U, ~1U);
     }
     static unsigned getHashValue(const DebugLocTuple &Val) {
       return DenseMapInfo<GlobalVariable*>::getHashValue(Val.CompileUnit) ^
-             DenseMapInfo<unsigned>::getHashValue(Val.Scope.getIndex()) ^
              DenseMapInfo<unsigned>::getHashValue(Val.Line) ^
              DenseMapInfo<unsigned>::getHashValue(Val.Col);
     }
     static bool isEqual(const DebugLocTuple &LHS, const DebugLocTuple &RHS) {
       return LHS.CompileUnit == RHS.CompileUnit &&
-             LHS.Scope       == RHS.Scope &&
              LHS.Line        == RHS.Line &&
              LHS.Col         == RHS.Col;
     }

Modified: llvm/trunk/include/llvm/CodeGen/FastISel.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/FastISel.h?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/FastISel.h (original)
+++ llvm/trunk/include/llvm/CodeGen/FastISel.h Wed May 20 19:04:55 2009
@@ -57,7 +57,6 @@
   MachineFrameInfo &MFI;
   MachineConstantPool &MCP;
   DebugLoc DL;
-  DebugScopeTracker DbgScopeTrack;
   const TargetMachine &TM;
   const TargetData &TD;
   const TargetInstrInfo &TII;

Modified: llvm/trunk/include/llvm/CodeGen/MachineFunction.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineFunction.h?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/include/llvm/CodeGen/MachineFunction.h (original)
+++ llvm/trunk/include/llvm/CodeGen/MachineFunction.h Wed May 20 19:04:55 2009
@@ -111,8 +111,6 @@
   // Tracks debug locations.
   DebugLocTracker DebugLocInfo;
 
-  std::vector<DebugScopeInfo> DbgScopeInfos;
-
 public:
   MachineFunction(const Function *Fn, const TargetMachine &TM);
   ~MachineFunction();
@@ -334,7 +332,7 @@
   /// getOrCreateDebugLocID - Look up the DebugLocTuple index with the given
   /// source file, line, and column. If none currently exists, create a new
   /// DebugLocTuple, and insert it into the DebugIdMap.
-  unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit, DebugScope Scope,
+  unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit,
                                  unsigned Line, unsigned Col);
 
   /// getDebugLocTuple - Get the DebugLocTuple for a given DebugLoc object.
@@ -347,12 +345,6 @@
   /// setDefaultDebugLoc - Get the default debug location for the machine
   /// function.
   void setDefaultDebugLoc(DebugLoc DL) { DefaultDebugLoc = DL; }
-
-  /// CreateDebugScope - Create a new debug scope.
-  DebugScope CreateDebugScope(GlobalVariable *ScopeGV, DebugScope Parent);
-
-  /// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
-  const DebugScopeInfo &getDebugScopeInfo(DebugScope DS) const;
 };
 
 //===--------------------------------------------------------------------===//

Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
+++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Wed May 20 19:04:55 2009
@@ -1311,7 +1311,7 @@
 void AsmPrinter::processDebugLoc(DebugLoc DL) {
   if (TAI->doesSupportDebugInformation() && DW->ShouldEmitDwarfDebug()) {
     if (!DL.isUnknown()) {
-      static DebugLocTuple PrevDLT(0, DebugScope::getInvalid(), ~0U, ~0U);
+      static DebugLocTuple PrevDLT(0, ~0U, ~0U);
       DebugLocTuple CurDLT = MF->getDebugLocTuple(DL);
 
       if (CurDLT.CompileUnit != 0 && PrevDLT != CurDLT)

Removed: llvm/trunk/lib/CodeGen/DebugLoc.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DebugLoc.cpp?rev=72199&view=auto

==============================================================================
--- llvm/trunk/lib/CodeGen/DebugLoc.cpp (original)
+++ llvm/trunk/lib/CodeGen/DebugLoc.cpp (removed)
@@ -1,38 +0,0 @@
-//===-- DebugLoc.cpp ------------------------------------------------------===//
-//
-//                     The LLVM Compiler Infrastructure
-//
-// This file is distributed under the University of Illinois Open Source
-// License. See LICENSE.TXT for details.
-//
-//===----------------------------------------------------------------------===//
-//
-// Implementation for DebugScopeTracker.
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/CodeGen/DebugLoc.h"
-#include "llvm/CodeGen/MachineFunction.h"
-using namespace llvm;
-
-/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
-/// or a DIBlock.
-void DebugScopeTracker::EnterDebugScope(GlobalVariable *ScopeGV,
-                                        MachineFunction &MF) {
-  assert(ScopeGV && "GlobalVariable for scope is null!");
-  CurScope = MF.CreateDebugScope(ScopeGV, CurScope);
-}
-
-/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
-void DebugScopeTracker::ExitDebugScope(GlobalVariable *ScopeGV,
-                                       MachineFunction &MF) {
-  assert(ScopeGV && "GlobalVariable for scope is null!");
-  assert(!CurScope.isInvalid() && "Mismatched region.end ?");
-  // We may have skipped a region.end because it was in an unreachable block.
-  // Go up the scope chain until we reach the scope that ScopeGV points to.
-  DebugScopeInfo DSI;
-  do {
-    DSI =  MF.getDebugScopeInfo(CurScope);
-    CurScope = DSI.Parent;
-  } while (!DSI.Parent.isInvalid() && DSI.GV != ScopeGV);
-}

Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
+++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Wed May 20 19:04:55 2009
@@ -403,9 +403,8 @@
 /// source file, line, and column. If none currently exists, create a new
 /// DebugLocTuple, and insert it into the DebugIdMap.
 unsigned MachineFunction::getOrCreateDebugLocID(GlobalVariable *CompileUnit,
-                                                DebugScope Scope,
                                                 unsigned Line, unsigned Col) {
-  DebugLocTuple Tuple(CompileUnit, Scope, Line, Col);
+  DebugLocTuple Tuple(CompileUnit, Line, Col);
   DenseMap<DebugLocTuple, unsigned>::iterator II
     = DebugLocInfo.DebugIdMap.find(Tuple);
   if (II != DebugLocInfo.DebugIdMap.end())
@@ -425,21 +424,6 @@
   return DebugLocInfo.DebugLocations[Idx];
 }
 
-/// CreateDebugScope - Create a new debug scope.
-DebugScope MachineFunction::CreateDebugScope(GlobalVariable *ScopeGV,
-                                             DebugScope Parent) {
-  DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
-  return DebugScope::get(DbgScopeInfos.size() - 1);
-}
-
-/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
-const DebugScopeInfo &MachineFunction::getDebugScopeInfo(DebugScope DS) const {
-  unsigned Idx = DS.getIndex();
-  assert(Idx < DbgScopeInfos.size() && "Invalid index into debug scopes!");
-  return DbgScopeInfos[Idx];
-}
-
-
 //===----------------------------------------------------------------------===//
 //  MachineFrameInfo implementation
 //===----------------------------------------------------------------------===//

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Wed May 20 19:04:55 2009
@@ -331,22 +331,17 @@
       DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
       unsigned Line = SPI->getLine();
       unsigned Col = SPI->getColumn();
-      unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(),
-                                              DbgScopeTrack.getCurScope(),
-                                              Line, Col);
+      unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(), Line, Col);
       setCurDebugLoc(DebugLoc::get(Idx));
     }
     return true;
   }
   case Intrinsic::dbg_region_start: {
     DbgRegionStartInst *RSI = cast<DbgRegionStartInst>(I);
-    if (!DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None))
-      return true;
-
-    GlobalVariable *Rgn = cast<GlobalVariable>(RSI->getContext());
-    DbgScopeTrack.EnterDebugScope(Rgn, MF);
-    if (DW && DW->ShouldEmitDwarfDebug()) {
-      unsigned ID = DW->RecordRegionStart(Rgn);
+    if (DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None) &&
+        DW && DW->ShouldEmitDwarfDebug()) {
+      unsigned ID = 
+        DW->RecordRegionStart(cast<GlobalVariable>(RSI->getContext()));
       const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
       BuildMI(MBB, DL, II).addImm(ID);
     }
@@ -354,14 +349,10 @@
   }
   case Intrinsic::dbg_region_end: {
     DbgRegionEndInst *REI = cast<DbgRegionEndInst>(I);
-    if (!DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None))
-      return true;
-
-    GlobalVariable *Rgn = cast<GlobalVariable>(REI->getContext());
-    DbgScopeTrack.ExitDebugScope(Rgn, MF);
-    if (DW && DW->ShouldEmitDwarfDebug()) {
+    if (DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None) &&
+        DW && DW->ShouldEmitDwarfDebug()) {
      unsigned ID = 0;
-     DISubprogram Subprogram(Rgn);
+     DISubprogram Subprogram(cast<GlobalVariable>(REI->getContext()));
      if (!Subprogram.isNull() && !Subprogram.describes(MF.getFunction())) {
         // This is end of an inlined function.
         const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
@@ -391,7 +382,6 @@
     DebugLoc PrevLoc = DL;
     DISubprogram Subprogram(cast<GlobalVariable>(SP));
     DICompileUnit CompileUnit = Subprogram.getCompileUnit();
-    DbgScopeTrack.EnterDebugScope(Subprogram.getGV(), MF);
 
     if (!Subprogram.describes(MF.getFunction())) {
       // This is a beginning of an inlined function.
@@ -403,10 +393,8 @@
         return true;
       // Record the source line.
       unsigned Line = Subprogram.getLineNumber();
-      setCurDebugLoc(
-        DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
-                                               DbgScopeTrack.getCurScope(),
-                                               Line, 0)));
+      setCurDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
+                                              CompileUnit.getGV(), Line, 0)));
 
       if (DW && DW->ShouldEmitDwarfDebug()) {
         DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
@@ -420,10 +408,8 @@
     } else {
       // Record the source line.
       unsigned Line = Subprogram.getLineNumber();
-      MF.setDefaultDebugLoc(
-        DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
-                                               DbgScopeTrack.getCurScope(),
-                                               Line, 0)));
+      MF.setDefaultDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
+                                              CompileUnit.getGV(), Line, 0)));
       if (DW && DW->ShouldEmitDwarfDebug()) {
         // llvm.dbg.func_start also defines beginning of function scope.
         DW->RecordRegionStart(cast<GlobalVariable>(FSI->getSubprogram()));

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Wed May 20 19:04:55 2009
@@ -312,17 +312,6 @@
             !StaticAllocaMap.count(cast<AllocaInst>(I)))
           InitializeRegForValue(I);
 
-  // FIXME: PHI instructions currently get an invalid scope. This is because
-  // tracking of debug scopes is done through a simple mechanism where
-  // "entering" a scope implies that the scope is entered for the first time.
-  // If we keep track of debug scopes for the following loop, the PHI
-  // instructions would get scopes that will not be used again later by the
-  // instruction selectors.
-  // Either provide a mechanism to re-enter a previously created scope or wait
-  // for when the DebugLoc is retrieved from Instruction, in which case the
-  // current debug scope tracking mechanism will be obsolete.
-  DebugScope DbgScope;
-
   // Create an initial MachineBasicBlock for each LLVM BasicBlock in F.  This
   // also creates the initial PHI MachineInstrs, though none of the input
   // operands are populated.
@@ -348,7 +337,6 @@
                                              CodeGenOpt::Default)) {
               DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
               unsigned idx = MF->getOrCreateDebugLocID(CU.getGV(),
-                                                       DbgScope,
                                                        SPI->getLine(),
                                                        SPI->getColumn());
               DL = DebugLoc::get(idx);
@@ -364,7 +352,7 @@
               DISubprogram Subprogram(cast<GlobalVariable>(SP));
               DICompileUnit CU(Subprogram.getCompileUnit());
               unsigned Line = Subprogram.getLineNumber();
-              DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(), DbgScope,
+              DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(),
                                                            Line, 0));
             }
 
@@ -3921,9 +3909,7 @@
       MachineFunction &MF = DAG.getMachineFunction();
       DICompileUnit CU(cast<GlobalVariable>(SPI.getContext()));
       DebugLoc Loc = DebugLoc::get(MF.getOrCreateDebugLocID(CU.getGV(),
-                                                 DbgScopeTrack.getCurScope(),
-                                                           SPI.getLine(),
-                                                           SPI.getColumn()));
+                                              SPI.getLine(), SPI.getColumn()));
       setCurDebugLoc(Loc);
       
       if (OptLevel == CodeGenOpt::None)
@@ -3937,14 +3923,11 @@
   case Intrinsic::dbg_region_start: {
     DwarfWriter *DW = DAG.getDwarfWriter();
     DbgRegionStartInst &RSI = cast<DbgRegionStartInst>(I);
-    if (!DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel))
-      return 0;
 
-    MachineFunction &MF = DAG.getMachineFunction();
-    GlobalVariable *Rgn = cast<GlobalVariable>(RSI.getContext());
-    DbgScopeTrack.EnterDebugScope(Rgn, MF);
-    if (DW && DW->ShouldEmitDwarfDebug()) {
-      unsigned LabelID = DW->RecordRegionStart(Rgn);
+    if (DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel) &&
+        DW && DW->ShouldEmitDwarfDebug()) {
+      unsigned LabelID =
+        DW->RecordRegionStart(cast<GlobalVariable>(RSI.getContext()));
       DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),
                                getRoot(), LabelID));
     }
@@ -3954,14 +3937,11 @@
   case Intrinsic::dbg_region_end: {
     DwarfWriter *DW = DAG.getDwarfWriter();
     DbgRegionEndInst &REI = cast<DbgRegionEndInst>(I);
-    if (!DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel))
-      return 0;
 
-    MachineFunction &MF = DAG.getMachineFunction();
-    GlobalVariable *Rgn = cast<GlobalVariable>(REI.getContext());
-    DbgScopeTrack.ExitDebugScope(Rgn, MF);
-    if (DW && DW->ShouldEmitDwarfDebug()) {
-      DISubprogram Subprogram(Rgn);
+    if (DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel) &&
+        DW && DW->ShouldEmitDwarfDebug()) {
+      MachineFunction &MF = DAG.getMachineFunction();
+      DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
 
       if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
         unsigned LabelID =
@@ -3994,7 +3974,6 @@
       return 0;
 
     MachineFunction &MF = DAG.getMachineFunction();
-    DbgScopeTrack.EnterDebugScope(cast<GlobalVariable>(SP), MF);
     if (OptLevel == CodeGenOpt::None) {
       // llvm.dbg.func.start implicitly defines a dbg_stoppoint which is what
       // (most?) gdb expects.
@@ -4014,9 +3993,7 @@
         // Record the source line.
         unsigned Line = Subprogram.getLineNumber();
         setCurDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
-                                              DbgScopeTrack.getCurScope(),
-                                              Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
 
         if (DW && DW->ShouldEmitDwarfDebug()) {
           DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
@@ -4031,9 +4008,7 @@
         // Record the source line.
         unsigned Line = Subprogram.getLineNumber();
         MF.setDefaultDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
-                                              DbgScopeTrack.getCurScope(),
-                                              Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
         if (DW && DW->ShouldEmitDwarfDebug()) {
           // llvm.dbg.func_start also defines beginning of function scope.
           DW->RecordRegionStart(cast<GlobalVariable>(FSI.getSubprogram()));
@@ -4060,9 +4035,7 @@
       // create a label if this is a beginning of inlined function.
       unsigned Line = Subprogram.getLineNumber();
       setCurDebugLoc(DebugLoc::get(
-                     MF.getOrCreateDebugLocID(CompileUnit.getGV(),
-                                              DbgScopeTrack.getCurScope(),
-                                              Line, 0)));
+                     MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
       // FIXME -  Start new region because llvm.dbg.func_start also defines
       // beginning of function scope.
     }

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h?rev=72200&r1=72199&r2=72200&view=diff

==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h Wed May 20 19:04:55 2009
@@ -169,8 +169,6 @@
   /// CurDebugLoc - current file + line number.  Changes as we build the DAG.
   DebugLoc CurDebugLoc;
 
-  DebugScopeTracker DbgScopeTrack;
-
   DenseMap<const Value*, SDValue> NodeMap;
 
   /// PendingLoads - Loads are not emitted to the program immediately.  We bunch




From isanbard at gmail.com  Wed May 20 19:05:10 2009
From: isanbard at gmail.com (Bill Wendling)
Date: Wed, 20 May 2009 17:05:10 -0700
Subject: [llvm-commits] [llvm] r72191 - in /llvm/trunk:
	include/llvm/CodeGen/DebugLoc.h include/llvm/CodeGen/FastISel.h
	include/llvm/CodeGen/MachineFunction.h
	lib/CodeGen/AsmPrinter/AsmPrinter.cpp
	lib/CodeGen/DebugLoc.cpp lib/CodeGen/MachineFunction.cpp l
Message-ID: <16e5fdf90905201705x221ff162l4b050ef06c9c7545@mail.gmail.com>

I temporarily reverted this until you can look at this. Sorry :-(

-bw

On Wed, May 20, 2009 at 3:57 PM, Argiris Kirtzidis <akyrtzi at gmail.com> wrote:
> Author: akirtzidis
> Date: Wed May 20 17:57:17 2009
> New Revision: 72191
>
> URL: http://llvm.org/viewvc/llvm-project?rev=72191&view=rev
> Log:
> Introduce DebugScope which gets embedded into the machine instructions' DebugLoc.
> DebugScope refers to a debug region, function or block.
>
> Added:
> ? ?llvm/trunk/lib/CodeGen/DebugLoc.cpp
> Modified:
> ? ?llvm/trunk/include/llvm/CodeGen/DebugLoc.h
> ? ?llvm/trunk/include/llvm/CodeGen/FastISel.h
> ? ?llvm/trunk/include/llvm/CodeGen/MachineFunction.h
> ? ?llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
> ? ?llvm/trunk/lib/CodeGen/MachineFunction.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
> ? ?llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
>
> Modified: llvm/trunk/include/llvm/CodeGen/DebugLoc.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/DebugLoc.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/DebugLoc.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/DebugLoc.h Wed May 20 17:57:17 2009
> @@ -20,18 +20,77 @@
>
> ?namespace llvm {
> ? class GlobalVariable;
> + ?class MachineFunction;
>
> - ?/// DebugLocTuple - Debug location tuple of filename id, line and column.
> + ?/// DebugScope - Debug scope id. This is carried into DebugLocTuple to index
> + ?/// into a vector of DebugScopeInfos. Provides the ability to associate a
> + ?/// SDNode/MachineInstr with the debug scope that it belongs to.
> + ?///
> + ?class DebugScope {
> + ? ?unsigned Idx;
> +
> + ?public:
> + ? ?DebugScope() : Idx(~0U) {} ?// Defaults to null.
> +
> + ? ?static DebugScope getInvalid() ? { DebugScope S; S.Idx = ~0U; return S; }
> + ? ?static DebugScope get(unsigned idx) { DebugScope S; S.Idx = idx; return S; }
> +
> + ? ?unsigned getIndex() const { return Idx; }
> +
> + ? ?/// isInvalid - Return true if it doesn't refer to any scope.
> + ? ?bool isInvalid() const { return Idx == ~0U; }
> +
> + ? ?bool operator==(const DebugScope &DS) const { return Idx == DS.Idx; }
> + ? ?bool operator!=(const DebugScope &DS) const { return !(*this == DS); }
> + ?};
> +
> + ?/// DebugScopeInfo - Contains info about the scope global variable and the
> + ?/// parent debug scope. DebugScope is only a "cookie" that can point to a
> + ?/// specific DebugScopeInfo.
> + ?///
> + ?struct DebugScopeInfo {
> + ? ?GlobalVariable *GV;
> + ? ?DebugScope Parent;
> +
> + ? ?DebugScopeInfo(GlobalVariable *gv, DebugScope parent)
> + ? ? ?: GV(gv), Parent(parent) {}
> + ? ?DebugScopeInfo()
> + ? ? ?: GV(0), Parent(DebugScope::getInvalid()) {}
> + ?};
> +
> + ?/// DebugScopeTracker - Create and keep track of the debug scope while
> + ?/// entering/exiting subprograms and blocks. Used by the instruction
> + ?/// selectors.
> + ?///
> + ?class DebugScopeTracker {
> + ? ?DebugScope CurScope;
> +
> + ?public:
> + ? ?/// getCurScope - The current debug scope that we "entered" through
> + ? ?/// EnterDebugScope.
> + ? ?DebugScope getCurScope() const { return CurScope; }
> +
> + ? ?/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
> + ? ?/// or a DIBlock.
> + ? ?void EnterDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
> +
> + ? ?/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
> + ? ?void ExitDebugScope(GlobalVariable *ScopeGV, MachineFunction &MF);
> + ?};
> +
> + ?/// DebugLocTuple - Debug location tuple of a DICompileUnit global variable,
> + ?/// debug scope, line and column.
> ? ///
> ? struct DebugLocTuple {
> ? ? GlobalVariable *CompileUnit;
> + ? ?DebugScope Scope;
> ? ? unsigned Line, Col;
>
> - ? ?DebugLocTuple(GlobalVariable *v, unsigned l, unsigned c)
> - ? ? ?: CompileUnit(v), Line(l), Col(c) {};
> + ? ?DebugLocTuple(GlobalVariable *v, DebugScope s, unsigned l, unsigned c)
> + ? ? ?: CompileUnit(v), Scope(s), Line(l), Col(c) {};
>
> ? ? bool operator==(const DebugLocTuple &DLT) const {
> - ? ? ?return CompileUnit == DLT.CompileUnit &&
> + ? ? ?return CompileUnit == DLT.CompileUnit && Scope == DLT.Scope &&
> ? ? ? ? ? ? ?Line == DLT.Line && Col == DLT.Col;
> ? ? }
> ? ? bool operator!=(const DebugLocTuple &DLT) const {
> @@ -63,18 +122,20 @@
> ? // Partially specialize DenseMapInfo for DebugLocTyple.
> ? template<> ?struct DenseMapInfo<DebugLocTuple> {
> ? ? static inline DebugLocTuple getEmptyKey() {
> - ? ? ?return DebugLocTuple(0, ~0U, ~0U);
> + ? ? ?return DebugLocTuple(0, DebugScope::getInvalid(), ~0U, ~0U);
> ? ? }
> ? ? static inline DebugLocTuple getTombstoneKey() {
> - ? ? ?return DebugLocTuple((GlobalVariable*)~1U, ~1U, ~1U);
> + ? ? ?return DebugLocTuple((GlobalVariable*)~1U,DebugScope::get(~1U), ~1U, ~1U);
> ? ? }
> ? ? static unsigned getHashValue(const DebugLocTuple &Val) {
> ? ? ? return DenseMapInfo<GlobalVariable*>::getHashValue(Val.CompileUnit) ^
> + ? ? ? ? ? ? DenseMapInfo<unsigned>::getHashValue(Val.Scope.getIndex()) ^
> ? ? ? ? ? ? ?DenseMapInfo<unsigned>::getHashValue(Val.Line) ^
> ? ? ? ? ? ? ?DenseMapInfo<unsigned>::getHashValue(Val.Col);
> ? ? }
> ? ? static bool isEqual(const DebugLocTuple &LHS, const DebugLocTuple &RHS) {
> ? ? ? return LHS.CompileUnit == RHS.CompileUnit &&
> + ? ? ? ? ? ? LHS.Scope ? ? ? == RHS.Scope &&
> ? ? ? ? ? ? ?LHS.Line ? ? ? ?== RHS.Line &&
> ? ? ? ? ? ? ?LHS.Col ? ? ? ? == RHS.Col;
> ? ? }
>
> Modified: llvm/trunk/include/llvm/CodeGen/FastISel.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/FastISel.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/FastISel.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/FastISel.h Wed May 20 17:57:17 2009
> @@ -57,6 +57,7 @@
> ? MachineFrameInfo &MFI;
> ? MachineConstantPool &MCP;
> ? DebugLoc DL;
> + ?DebugScopeTracker DbgScopeTrack;
> ? const TargetMachine &TM;
> ? const TargetData &TD;
> ? const TargetInstrInfo &TII;
>
> Modified: llvm/trunk/include/llvm/CodeGen/MachineFunction.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/CodeGen/MachineFunction.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/include/llvm/CodeGen/MachineFunction.h (original)
> +++ llvm/trunk/include/llvm/CodeGen/MachineFunction.h Wed May 20 17:57:17 2009
> @@ -111,6 +111,8 @@
> ? // Tracks debug locations.
> ? DebugLocTracker DebugLocInfo;
>
> + ?std::vector<DebugScopeInfo> DbgScopeInfos;
> +
> ?public:
> ? MachineFunction(const Function *Fn, const TargetMachine &TM);
> ? ~MachineFunction();
> @@ -332,7 +334,7 @@
> ? /// getOrCreateDebugLocID - Look up the DebugLocTuple index with the given
> ? /// source file, line, and column. If none currently exists, create a new
> ? /// DebugLocTuple, and insert it into the DebugIdMap.
> - ?unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit,
> + ?unsigned getOrCreateDebugLocID(GlobalVariable *CompileUnit, DebugScope Scope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?unsigned Line, unsigned Col);
>
> ? /// getDebugLocTuple - Get the DebugLocTuple for a given DebugLoc object.
> @@ -345,6 +347,12 @@
> ? /// setDefaultDebugLoc - Get the default debug location for the machine
> ? /// function.
> ? void setDefaultDebugLoc(DebugLoc DL) { DefaultDebugLoc = DL; }
> +
> + ?/// CreateDebugScope - Create a new debug scope.
> + ?DebugScope CreateDebugScope(GlobalVariable *ScopeGV, DebugScope Parent);
> +
> + ?/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
> + ?const DebugScopeInfo &getDebugScopeInfo(DebugScope DS) const;
> ?};
>
> ?//===--------------------------------------------------------------------===//
>
> Modified: llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp (original)
> +++ llvm/trunk/lib/CodeGen/AsmPrinter/AsmPrinter.cpp Wed May 20 17:57:17 2009
> @@ -1311,7 +1311,7 @@
> ?void AsmPrinter::processDebugLoc(DebugLoc DL) {
> ? if (TAI->doesSupportDebugInformation() && DW->ShouldEmitDwarfDebug()) {
> ? ? if (!DL.isUnknown()) {
> - ? ? ?static DebugLocTuple PrevDLT(0, ~0U, ~0U);
> + ? ? ?static DebugLocTuple PrevDLT(0, DebugScope::getInvalid(), ~0U, ~0U);
> ? ? ? DebugLocTuple CurDLT = MF->getDebugLocTuple(DL);
>
> ? ? ? if (CurDLT.CompileUnit != 0 && PrevDLT != CurDLT)
>
> Added: llvm/trunk/lib/CodeGen/DebugLoc.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/DebugLoc.cpp?rev=72191&view=auto
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/DebugLoc.cpp (added)
> +++ llvm/trunk/lib/CodeGen/DebugLoc.cpp Wed May 20 17:57:17 2009
> @@ -0,0 +1,38 @@
> +//===-- DebugLoc.cpp ------------------------------------------------------===//
> +//
> +// ? ? ? ? ? ? ? ? ? ? The LLVM Compiler Infrastructure
> +//
> +// This file is distributed under the University of Illinois Open Source
> +// License. See LICENSE.TXT for details.
> +//
> +//===----------------------------------------------------------------------===//
> +//
> +// Implementation for DebugScopeTracker.
> +//
> +//===----------------------------------------------------------------------===//
> +
> +#include "llvm/CodeGen/DebugLoc.h"
> +#include "llvm/CodeGen/MachineFunction.h"
> +using namespace llvm;
> +
> +/// EnterDebugScope - Start a new debug scope. ScopeGV can be a DISubprogram
> +/// or a DIBlock.
> +void DebugScopeTracker::EnterDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?MachineFunction &MF) {
> + ?assert(ScopeGV && "GlobalVariable for scope is null!");
> + ?CurScope = MF.CreateDebugScope(ScopeGV, CurScope);
> +}
> +
> +/// ExitDebugScope - "Pop" a DISubprogram or a DIBlock.
> +void DebugScopeTracker::ExitDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? MachineFunction &MF) {
> + ?assert(ScopeGV && "GlobalVariable for scope is null!");
> + ?assert(!CurScope.isInvalid() && "Mismatched region.end ?");
> + ?// We may have skipped a region.end because it was in an unreachable block.
> + ?// Go up the scope chain until we reach the scope that ScopeGV points to.
> + ?DebugScopeInfo DSI;
> + ?do {
> + ? ?DSI = ?MF.getDebugScopeInfo(CurScope);
> + ? ?CurScope = DSI.Parent;
> + ?} while (!DSI.Parent.isInvalid() && DSI.GV != ScopeGV);
> +}
>
> Modified: llvm/trunk/lib/CodeGen/MachineFunction.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/MachineFunction.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/MachineFunction.cpp (original)
> +++ llvm/trunk/lib/CodeGen/MachineFunction.cpp Wed May 20 17:57:17 2009
> @@ -403,8 +403,9 @@
> ?/// source file, line, and column. If none currently exists, create a new
> ?/// DebugLocTuple, and insert it into the DebugIdMap.
> ?unsigned MachineFunction::getOrCreateDebugLocID(GlobalVariable *CompileUnit,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DebugScope Scope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? unsigned Line, unsigned Col) {
> - ?DebugLocTuple Tuple(CompileUnit, Line, Col);
> + ?DebugLocTuple Tuple(CompileUnit, Scope, Line, Col);
> ? DenseMap<DebugLocTuple, unsigned>::iterator II
> ? ? = DebugLocInfo.DebugIdMap.find(Tuple);
> ? if (II != DebugLocInfo.DebugIdMap.end())
> @@ -424,6 +425,21 @@
> ? return DebugLocInfo.DebugLocations[Idx];
> ?}
>
> +/// CreateDebugScope - Create a new debug scope.
> +DebugScope MachineFunction::CreateDebugScope(GlobalVariable *ScopeGV,
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DebugScope Parent) {
> + ?DbgScopeInfos.push_back(DebugScopeInfo(ScopeGV, Parent));
> + ?return DebugScope::get(DbgScopeInfos.size() - 1);
> +}
> +
> +/// getDebugScopeInfo - Get the DebugScopeInfo for a given DebugScope object.
> +const DebugScopeInfo &MachineFunction::getDebugScopeInfo(DebugScope DS) const {
> + ?unsigned Idx = DS.getIndex();
> + ?assert(Idx < DbgScopeInfos.size() && "Invalid index into debug scopes!");
> + ?return DbgScopeInfos[Idx];
> +}
> +
> +
> ?//===----------------------------------------------------------------------===//
> ?// ?MachineFrameInfo implementation
> ?//===----------------------------------------------------------------------===//
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Wed May 20 17:57:17 2009
> @@ -331,17 +331,22 @@
> ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
> ? ? ? unsigned Line = SPI->getLine();
> ? ? ? unsigned Col = SPI->getColumn();
> - ? ? ?unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(), Line, Col);
> + ? ? ?unsigned Idx = MF.getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, Col);
> ? ? ? setCurDebugLoc(DebugLoc::get(Idx));
> ? ? }
> ? ? return true;
> ? }
> ? case Intrinsic::dbg_region_start: {
> ? ? DbgRegionStartInst *RSI = cast<DbgRegionStartInst>(I);
> - ? ?if (DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?unsigned ID =
> - ? ? ? ?DW->RecordRegionStart(cast<GlobalVariable>(RSI->getContext()));
> + ? ?if (!DIDescriptor::ValidDebugInfo(RSI->getContext(), CodeGenOpt::None))
> + ? ? ?return true;
> +
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(RSI->getContext());
> + ? ?DbgScopeTrack.EnterDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?unsigned ID = DW->RecordRegionStart(Rgn);
> ? ? ? const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
> ? ? ? BuildMI(MBB, DL, II).addImm(ID);
> ? ? }
> @@ -349,10 +354,14 @@
> ? }
> ? case Intrinsic::dbg_region_end: {
> ? ? DbgRegionEndInst *REI = cast<DbgRegionEndInst>(I);
> - ? ?if (DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> + ? ?if (!DIDescriptor::ValidDebugInfo(REI->getContext(), CodeGenOpt::None))
> + ? ? ?return true;
> +
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(REI->getContext());
> + ? ?DbgScopeTrack.ExitDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ?unsigned ID = 0;
> - ? ? DISubprogram Subprogram(cast<GlobalVariable>(REI->getContext()));
> + ? ? DISubprogram Subprogram(Rgn);
> ? ? ?if (!Subprogram.isNull() && !Subprogram.describes(MF.getFunction())) {
> ? ? ? ? // This is end of an inlined function.
> ? ? ? ? const TargetInstrDesc &II = TII.get(TargetInstrInfo::DBG_LABEL);
> @@ -382,6 +391,7 @@
> ? ? DebugLoc PrevLoc = DL;
> ? ? DISubprogram Subprogram(cast<GlobalVariable>(SP));
> ? ? DICompileUnit CompileUnit = Subprogram.getCompileUnit();
> + ? ?DbgScopeTrack.EnterDebugScope(Subprogram.getGV(), MF);
>
> ? ? if (!Subprogram.describes(MF.getFunction())) {
> ? ? ? // This is a beginning of an inlined function.
> @@ -393,8 +403,10 @@
> ? ? ? ? return true;
> ? ? ? // Record the source line.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ?setCurDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CompileUnit.getGV(), Line, 0)));
> + ? ? ?setCurDebugLoc(
> + ? ? ? ?DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Line, 0)));
>
> ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
> @@ -408,8 +420,10 @@
> ? ? } else {
> ? ? ? // Record the source line.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ?MF.setDefaultDebugLoc(DebugLoc::get(MF.getOrCreateDebugLocID(
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CompileUnit.getGV(), Line, 0)));
> + ? ? ?MF.setDefaultDebugLoc(
> + ? ? ? ?DebugLoc::get(MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? Line, 0)));
> ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? // llvm.dbg.func_start also defines beginning of function scope.
> ? ? ? ? DW->RecordRegionStart(cast<GlobalVariable>(FSI->getSubprogram()));
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.cpp Wed May 20 17:57:17 2009
> @@ -312,6 +312,17 @@
> ? ? ? ? ? ? !StaticAllocaMap.count(cast<AllocaInst>(I)))
> ? ? ? ? ? InitializeRegForValue(I);
>
> + ?// FIXME: PHI instructions currently get an invalid scope. This is because
> + ?// tracking of debug scopes is done through a simple mechanism where
> + ?// "entering" a scope implies that the scope is entered for the first time.
> + ?// If we keep track of debug scopes for the following loop, the PHI
> + ?// instructions would get scopes that will not be used again later by the
> + ?// instruction selectors.
> + ?// Either provide a mechanism to re-enter a previously created scope or wait
> + ?// for when the DebugLoc is retrieved from Instruction, in which case the
> + ?// current debug scope tracking mechanism will be obsolete.
> + ?DebugScope DbgScope;
> +
> ? // Create an initial MachineBasicBlock for each LLVM BasicBlock in F. ?This
> ? // also creates the initial PHI MachineInstrs, though none of the input
> ? // operands are populated.
> @@ -337,6 +348,7 @@
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?CodeGenOpt::Default)) {
> ? ? ? ? ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI->getContext()));
> ? ? ? ? ? ? ? unsigned idx = MF->getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI->getLine(),
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI->getColumn());
> ? ? ? ? ? ? ? DL = DebugLoc::get(idx);
> @@ -352,7 +364,7 @@
> ? ? ? ? ? ? ? DISubprogram Subprogram(cast<GlobalVariable>(SP));
> ? ? ? ? ? ? ? DICompileUnit CU(Subprogram.getCompileUnit());
> ? ? ? ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> - ? ? ? ? ? ? ?DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(),
> + ? ? ? ? ? ? ?DL = DebugLoc::get(MF->getOrCreateDebugLocID(CU.getGV(), DbgScope,
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0));
> ? ? ? ? ? ? }
>
> @@ -3909,7 +3921,9 @@
> ? ? ? MachineFunction &MF = DAG.getMachineFunction();
> ? ? ? DICompileUnit CU(cast<GlobalVariable>(SPI.getContext()));
> ? ? ? DebugLoc Loc = DebugLoc::get(MF.getOrCreateDebugLocID(CU.getGV(),
> - ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?SPI.getLine(), SPI.getColumn()));
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? SPI.getLine(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? SPI.getColumn()));
> ? ? ? setCurDebugLoc(Loc);
>
> ? ? ? if (OptLevel == CodeGenOpt::None)
> @@ -3923,11 +3937,14 @@
> ? case Intrinsic::dbg_region_start: {
> ? ? DwarfWriter *DW = DAG.getDwarfWriter();
> ? ? DbgRegionStartInst &RSI = cast<DbgRegionStartInst>(I);
> + ? ?if (!DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel))
> + ? ? ?return 0;
>
> - ? ?if (DIDescriptor::ValidDebugInfo(RSI.getContext(), OptLevel) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?unsigned LabelID =
> - ? ? ? ?DW->RecordRegionStart(cast<GlobalVariable>(RSI.getContext()));
> + ? ?MachineFunction &MF = DAG.getMachineFunction();
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(RSI.getContext());
> + ? ?DbgScopeTrack.EnterDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?unsigned LabelID = DW->RecordRegionStart(Rgn);
> ? ? ? DAG.setRoot(DAG.getLabel(ISD::DBG_LABEL, getCurDebugLoc(),
> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?getRoot(), LabelID));
> ? ? }
> @@ -3937,11 +3954,14 @@
> ? case Intrinsic::dbg_region_end: {
> ? ? DwarfWriter *DW = DAG.getDwarfWriter();
> ? ? DbgRegionEndInst &REI = cast<DbgRegionEndInst>(I);
> + ? ?if (!DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel))
> + ? ? ?return 0;
>
> - ? ?if (DIDescriptor::ValidDebugInfo(REI.getContext(), OptLevel) &&
> - ? ? ? ?DW && DW->ShouldEmitDwarfDebug()) {
> - ? ? ?MachineFunction &MF = DAG.getMachineFunction();
> - ? ? ?DISubprogram Subprogram(cast<GlobalVariable>(REI.getContext()));
> + ? ?MachineFunction &MF = DAG.getMachineFunction();
> + ? ?GlobalVariable *Rgn = cast<GlobalVariable>(REI.getContext());
> + ? ?DbgScopeTrack.ExitDebugScope(Rgn, MF);
> + ? ?if (DW && DW->ShouldEmitDwarfDebug()) {
> + ? ? ?DISubprogram Subprogram(Rgn);
>
> ? ? ? if (Subprogram.isNull() || Subprogram.describes(MF.getFunction())) {
> ? ? ? ? unsigned LabelID =
> @@ -3974,6 +3994,7 @@
> ? ? ? return 0;
>
> ? ? MachineFunction &MF = DAG.getMachineFunction();
> + ? ?DbgScopeTrack.EnterDebugScope(cast<GlobalVariable>(SP), MF);
> ? ? if (OptLevel == CodeGenOpt::None) {
> ? ? ? // llvm.dbg.func.start implicitly defines a dbg_stoppoint which is what
> ? ? ? // (most?) gdb expects.
> @@ -3993,7 +4014,9 @@
> ? ? ? ? // Record the source line.
> ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? ? setCurDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
>
> ? ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? ? DebugLocTuple PrevLocTpl = MF.getDebugLocTuple(PrevLoc);
> @@ -4008,7 +4031,9 @@
> ? ? ? ? // Record the source line.
> ? ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? ? MF.setDefaultDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
> ? ? ? ? if (DW && DW->ShouldEmitDwarfDebug()) {
> ? ? ? ? ? // llvm.dbg.func_start also defines beginning of function scope.
> ? ? ? ? ? DW->RecordRegionStart(cast<GlobalVariable>(FSI.getSubprogram()));
> @@ -4035,7 +4060,9 @@
> ? ? ? // create a label if this is a beginning of inlined function.
> ? ? ? unsigned Line = Subprogram.getLineNumber();
> ? ? ? setCurDebugLoc(DebugLoc::get(
> - ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(), Line, 0)));
> + ? ? ? ? ? ? ? ? ? ? MF.getOrCreateDebugLocID(CompileUnit.getGV(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DbgScopeTrack.getCurScope(),
> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?Line, 0)));
> ? ? ? // FIXME - ?Start new region because llvm.dbg.func_start also defines
> ? ? ? // beginning of function scope.
> ? ? }
>
> Modified: llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h?rev=72191&r1=72190&r2=72191&view=diff
>
> ==============================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/SelectionDAGBuild.h Wed May 20 17:57:17 2009
> @@ -169,6 +169,8 @@
> ? /// CurDebugLoc - current file + line number. ?Changes as we build the DAG.
> ? DebugLoc CurDebugLoc;
>
> + ?DebugScopeTracker DbgScopeTrack;
> +
> ? DenseMap<const Value*, SDValue> NodeMap;
>
> ? /// PendingLoads - Loads are not emitted to the program immediately. ?We bunch
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



From resistor at mac.com  Wed May 20 19:25:52 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 17:25:52 -0700
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
 include/llvm/System/Atomic.h lib/System/Atomic.cpp
In-Reply-To: <f126fae00905201703u137104feg29e30f608d16344@mail.gmail.com>
References: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
	<f126fae00905201703u137104feg29e30f608d16344@mail.gmail.com>
Message-ID: <93A3BB88-3C5D-4E39-84FA-720334B5E88A@mac.com>


On May 20, 2009, at 5:03 PM, Howard Su wrote:

> don't inlcude platform dependent header here. please include "llvm/ 
> Support/DataTypes.h" instead.

That is not sufficient.  It needs windows.h in order to get the  
MemoryBarrier() macro.  Plus, files in lib/System/ are allowed to  
include platform dependent headers.  Take a look at Windows/Mutex.inc,  
for instance.

> Also, please update CMakeLists file also to add Atomic.cpp.

I don't speak CMake.  Please send a patch for it.

--Owen

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2620 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/83d410e5/attachment.bin 

From howard0su at gmail.com  Wed May 20 19:32:02 2009
From: howard0su at gmail.com (Howard Su)
Date: Thu, 21 May 2009 08:32:02 +0800
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
	include/llvm/System/Atomic.h lib/System/Atomic.cpp
In-Reply-To: <93A3BB88-3C5D-4E39-84FA-720334B5E88A@mac.com>
References: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
	<f126fae00905201703u137104feg29e30f608d16344@mail.gmail.com>
	<93A3BB88-3C5D-4E39-84FA-720334B5E88A@mac.com>
Message-ID: <f126fae00905201732l7defc2d9ua437bc8d803162e5@mail.gmail.com>

On Thu, May 21, 2009 at 8:25 AM, Owen Anderson <resistor at mac.com> wrote:

>
> On May 20, 2009, at 5:03 PM, Howard Su wrote:
>
> don't inlcude platform dependent header here. please include
>> "llvm/Support/DataTypes.h" instead.
>>
>
> That is not sufficient.  It needs windows.h in order to get the
> MemoryBarrier() macro.  Plus, files in lib/System/ are allowed to include
> platform dependent headers.  Take a look at Windows/Mutex.inc, for instance.
>

Sorry I didn't express myself clearly. I mean you can not include stdint.h
in atomic.h since windows don't have it. In the atomic header file, what we
need it uint_t define which is in  "llvm/Support/DataTypes.h"

>
>
> Also, please update CMakeLists file also to add Atomic.cpp.
>>
>
> I don't speak CMake.  Please send a patch for it.
>

Index: include/llvm/System/Atomic.h
===================================================================
--- include/llvm/System/Atomic.h (revision 72198)
+++ include/llvm/System/Atomic.h (working copy)
@@ -14,7 +14,7 @@
 #ifndef LLVM_SYSTEM_ATOMIC_H
 #define LLVM_SYSTEM_ATOMIC_H

-#include <stdint.h>
+#include "llvm/Support/DataTypes.h"

 namespace llvm {
   namespace sys {
Index: lib/System/CMakeLists.txt
===================================================================
--- lib/System/CMakeLists.txt (revision 72198)
+++ lib/System/CMakeLists.txt (working copy)
@@ -1,5 +1,6 @@
 add_llvm_library(LLVMSystem
   Alarm.cpp
+  Atomic.cpp
   Disassembler.cpp
   DynamicLibrary.cpp
   Host.cpp

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


-- 
-Howard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090521/e17d5c87/attachment.html 

From resistor at mac.com  Wed May 20 19:48:18 2009
From: resistor at mac.com (Owen Anderson)
Date: Thu, 21 May 2009 00:48:18 -0000
Subject: [llvm-commits] [llvm] r72201 -
	/llvm/trunk/include/llvm/System/Atomic.h
Message-ID: <200905210048.n4L0mK7k031688@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 19:48:13 2009
New Revision: 72201

URL: http://llvm.org/viewvc/llvm-project?rev=72201&view=rev
Log:
Use DataTypes.h instead of stdint.h.

Modified:
    llvm/trunk/include/llvm/System/Atomic.h

Modified: llvm/trunk/include/llvm/System/Atomic.h
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/System/Atomic.h?rev=72201&r1=72200&r2=72201&view=diff

==============================================================================
--- llvm/trunk/include/llvm/System/Atomic.h (original)
+++ llvm/trunk/include/llvm/System/Atomic.h Wed May 20 19:48:13 2009
@@ -14,7 +14,7 @@
 #ifndef LLVM_SYSTEM_ATOMIC_H
 #define LLVM_SYSTEM_ATOMIC_H
 
-#include <stdint.h>
+#include "llvm/Support/DataTypes.h"
 
 namespace llvm {
   namespace sys {




From resistor at mac.com  Wed May 20 19:48:57 2009
From: resistor at mac.com (Owen Anderson)
Date: Thu, 21 May 2009 00:48:57 -0000
Subject: [llvm-commits] [llvm] r72202 - /llvm/trunk/lib/System/CMakeLists.txt
Message-ID: <200905210048.n4L0mvRw031714@zion.cs.uiuc.edu>

Author: resistor
Date: Wed May 20 19:48:56 2009
New Revision: 72202

URL: http://llvm.org/viewvc/llvm-project?rev=72202&view=rev
Log:
Add Atomic.cpp to the CMake build system.

Modified:
    llvm/trunk/lib/System/CMakeLists.txt

Modified: llvm/trunk/lib/System/CMakeLists.txt
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/System/CMakeLists.txt?rev=72202&r1=72201&r2=72202&view=diff

==============================================================================
--- llvm/trunk/lib/System/CMakeLists.txt (original)
+++ llvm/trunk/lib/System/CMakeLists.txt Wed May 20 19:48:56 2009
@@ -1,5 +1,6 @@
 add_llvm_library(LLVMSystem
   Alarm.cpp
+  Atomic.cpp
   Disassembler.cpp
   DynamicLibrary.cpp
   Host.cpp




From resistor at mac.com  Wed May 20 19:49:11 2009
From: resistor at mac.com (Owen Anderson)
Date: Wed, 20 May 2009 17:49:11 -0700
Subject: [llvm-commits] [llvm] r72174 - in /llvm/trunk:
 include/llvm/System/Atomic.h lib/System/Atomic.cpp
In-Reply-To: <f126fae00905201732l7defc2d9ua437bc8d803162e5@mail.gmail.com>
References: <200905201826.n4KIQG0K019017@zion.cs.uiuc.edu>
	<f126fae00905201703u137104feg29e30f608d16344@mail.gmail.com>
	<93A3BB88-3C5D-4E39-84FA-720334B5E88A@mac.com>
	<f126fae00905201732l7defc2d9ua437bc8d803162e5@mail.gmail.com>
Message-ID: <13288AFD-8890-42C0-8360-3E6E7B4B50CF@mac.com>

Done.

--Owen

On May 20, 2009, at 5:32 PM, Howard Su wrote:

>
>
> On Thu, May 21, 2009 at 8:25 AM, Owen Anderson <resistor at mac.com>  
> wrote:
>
> On May 20, 2009, at 5:03 PM, Howard Su wrote:
>
> don't inlcude platform dependent header here. please include "llvm/ 
> Support/DataTypes.h" instead.
>
> That is not sufficient.  It needs windows.h in order to get the  
> MemoryBarrier() macro.  Plus, files in lib/System/ are allowed to  
> include platform dependent headers.  Take a look at Windows/ 
> Mutex.inc, for instance.
> Sorry I didn't express myself clearly. I mean you can not include  
> stdint.h in atomic.h since windows don't have it. In the atomic  
> header file, what we need it uint_t define which is in  "llvm/ 
> Support/DataTypes.h"
>
>
> Also, please update CMakeLists file also to add Atomic.cpp.
>
> I don't speak CMake.  Please send a patch for it.
>
> Index: include/llvm/System/Atomic.h
> ===================================================================
> --- include/llvm/System/Atomic.h (revision 72198)
> +++ include/llvm/System/Atomic.h (working copy)
> @@ -14,7 +14,7 @@
>  #ifndef LLVM_SYSTEM_ATOMIC_H
>  #define LLVM_SYSTEM_ATOMIC_H
>
> -#include <stdint.h>
> +#include "llvm/Support/DataTypes.h"
>
>  namespace llvm {
>    namespace sys {
> Index: lib/System/CMakeLists.txt
> ===================================================================
> --- lib/System/CMakeLists.txt (revision 72198)
> +++ lib/System/CMakeLists.txt (working copy)
> @@ -1,5 +1,6 @@
>  add_llvm_library(LLVMSystem
>    Alarm.cpp
> +  Atomic.cpp
>    Disassembler.cpp
>    DynamicLibrary.cpp
>    Host.cpp
>
> --Owen
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>
>
>
>
> -- 
> -Howard
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/852473e8/attachment.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 2620 bytes
Desc: not available
Url : http://lists.cs.uiuc.edu/pipermail/llvm-commits/attachments/20090520/852473e8/attachment.bin 

From gohman at apple.com  Wed May 20 21:28:34 2009
From: gohman at apple.com (Dan Gohman)
Date: Thu, 21 May 2009 02:28:34 -0000
Subject: [llvm-commits] [llvm] r72204 - in /llvm/trunk:
 lib/Analysis/ValueTracking.cpp
 lib/Transforms/Scalar/InstructionCombining.cpp
 test/Transforms/InstCombine/lshr-phi.ll
Message-ID: <200905210228.n4L2SZjf001928@zion.cs.uiuc.edu>

Author: djg
Date: Wed May 20 21:28:33 2009
New Revision: 72204

URL: http://llvm.org/viewvc/llvm-project?rev=72204&view=rev
Log:
Teach ValueTracking a new way to analyze PHI nodes, and and teach
Instcombine to be more aggressive about using SimplifyDemandedBits
on shift nodes. This allows a shift to be simplified to zero in the
included test case.

Added:
    llvm/trunk/test/Transforms/InstCombine/lshr-phi.ll
Modified:
    llvm/trunk/lib/Analysis/ValueTracking.cpp
    llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp

Modified: llvm/trunk/lib/Analysis/ValueTracking.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Analysis/ValueTracking.cpp?rev=72204&r1=72203&r2=72204&view=diff

==============================================================================
--- llvm/trunk/lib/Analysis/ValueTracking.cpp (original)
+++ llvm/trunk/lib/Analysis/ValueTracking.cpp Wed May 20 21:28:33 2009
@@ -48,8 +48,9 @@
 void llvm::ComputeMaskedBits(Value *V, const APInt &Mask,
                              APInt &KnownZero, APInt &KnownOne,
                              TargetData *TD, unsigned Depth) {
+  const unsigned MaxDepth = 6;
   assert(V && "No Value?");
-  assert(Depth <= 6 && "Limit Search Depth");
+  assert(Depth <= MaxDepth && "Limit Search Depth");
   unsigned BitWidth = Mask.getBitWidth();
   assert((V->getType()->isInteger() || isa<PointerType>(V->getType())) &&
          "Not integer or pointer type!");
@@ -88,7 +89,7 @@
 
   KnownZero.clear(); KnownOne.clear();   // Start out not knowing anything.
 
-  if (Depth == 6 || Mask == 0)
+  if (Depth == MaxDepth || Mask == 0)
     return;  // Limit search depth.
 
   User *I = dyn_cast<User>(V);
@@ -522,6 +523,30 @@
         }
       }
     }
+
+    // Otherwise take the unions of the known bit sets of the operands,
+    // taking conservative care to avoid excessive recursion.
+    if (Depth < MaxDepth - 1 && !KnownZero && !KnownOne) {
+      KnownZero = APInt::getAllOnesValue(BitWidth);
+      KnownOne = APInt::getAllOnesValue(BitWidth);
+      for (unsigned i = 0, e = P->getNumIncomingValues(); i != e; ++i) {
+        // Skip direct self references.
+        if (P->getIncomingValue(i) == P) continue;
+
+        KnownZero2 = APInt(BitWidth, 0);
+        KnownOne2 = APInt(BitWidth, 0);
+        // Recurse, but cap the recursion to one level, because we don't
+        // want to waste time spinning around in loops.
+        ComputeMaskedBits(P->getIncomingValue(i), KnownZero | KnownOne,
+                          KnownZero2, KnownOne2, TD, MaxDepth-1);
+        KnownZero &= KnownZero2;
+        KnownOne &= KnownOne2;
+        // If all bits have been ruled out, there's no need to check
+        // more operands.
+        if (!KnownZero && !KnownOne)
+          break;
+      }
+    }
     break;
   }
   case Instruction::Call:

Modified: llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp?rev=72204&r1=72203&r2=72204&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp Wed May 20 21:28:33 2009
@@ -7152,6 +7152,10 @@
       return ReplaceInstUsesWith(I, Constant::getNullValue(I.getType()));
   }
 
+  // See if we can fold away this shift.
+  if (!isa<VectorType>(I.getType()) && SimplifyDemandedInstructionBits(I))
+    return &I;
+
   // Try to fold constant and into select arguments.
   if (isa<Constant>(Op0))
     if (SelectInst *SI = dyn_cast<SelectInst>(Op1))
@@ -7171,8 +7175,6 @@
   // See if we can simplify any instructions used by the instruction whose sole 
   // purpose is to compute bits we don't care about.
   uint32_t TypeBits = Op0->getType()->getPrimitiveSizeInBits();
-  if (SimplifyDemandedInstructionBits(I))
-    return &I;
   
   // shl uint X, 32 = 0 and shr ubyte Y, 9 = 0, ... just don't eliminate shr
   // of a signed value.

Added: llvm/trunk/test/Transforms/InstCombine/lshr-phi.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/lshr-phi.ll?rev=72204&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/lshr-phi.ll (added)
+++ llvm/trunk/test/Transforms/InstCombine/lshr-phi.ll Wed May 20 21:28:33 2009
@@ -0,0 +1,35 @@
+; RUN: llvm-as < %s | opt -instcombine | llvm-dis > %t
+; RUN: not grep lshr %t
+; RUN: grep add %t | count 1
+
+; Instcombine should be able to eliminate the lshr, because only
+; bits in the operand which might be non-zero will be shifted
+; off the end.
+
+define i32 @hash_string(i8* nocapture %key) nounwind readonly {
+entry:
+	%t0 = load i8* %key, align 1		; <i8> [#uses=1]
+	%t1 = icmp eq i8 %t0, 0		; <i1> [#uses=1]
+	br i1 %t1, label %bb2, label %bb
+
+bb:		; preds = %bb, %entry
+	%indvar = phi i64 [ 0, %entry ], [ %tmp, %bb ]		; <i64> [#uses=2]
+	%k.04 = phi i32 [ 0, %entry ], [ %t8, %bb ]		; <i32> [#uses=2]
+	%cp.05 = getelementptr i8* %key, i64 %indvar		; <i8*> [#uses=1]
+	%t2 = shl i32 %k.04, 1		; <i32> [#uses=1]
+	%t3 = lshr i32 %k.04, 14		; <i32> [#uses=1]
+	%t4 = add i32 %t2, %t3		; <i32> [#uses=1]
+	%t5 = load i8* %cp.05, align 1		; <i8> [#uses=1]
+	%t6 = sext i8 %t5 to i32		; <i32> [#uses=1]
+	%t7 = xor i32 %t6, %t4		; <i32> [#uses=1]
+	%t8 = and i32 %t7, 16383		; <i32> [#uses=2]
+	%tmp = add i64 %indvar, 1		; <i64> [#uses=2]
+	%scevgep = getelementptr i8* %key, i64 %tmp		; <i8*> [#uses=1]
+	%t9 = load i8* %scevgep, align 1		; <i8> [#uses=1]
+	%t10 = icmp eq i8 %t9, 0		; <i1> [#uses=1]
+	br i1 %t10, label %bb2, label %bb
+
+bb2:		; preds = %bb, %entry
+	%k.0.lcssa = phi i32 [ 0, %entry ], [ %t8, %bb ]		; <i32> [#uses=1]
+	ret i32 %k.0.lcssa
+}




From nicholas at mxc.ca  Wed May 20 22:09:45 2009
From: nicholas at mxc.ca (Nick Lewycky)
Date: Wed, 20 May 2009 20:09:45 -0700
Subject: [llvm-commits] RFC: initial union syntax support
In-Reply-To: <200905142101.25748.baldrick@free.fr>
References: <85dfcd7f0905121101y62ac0c4esd27a518adbabd3b9@mail.gmail.com>
	<85dfcd7f0905140458l7f039716me3addf865df8c87@mail.gmail.com>
	<4A0C475C.8010608@mxc.ca> <200905142101.25748.baldrick@free.fr>
Message-ID: <4A14C5F9.2050802@mxc.ca>

Duncan Sands wrote:
> Hi Nick,
> 
>> Then it would seem I misunderstood the purpose of unions. I thought the 
>> problem was that it was impossible to declare a type which would be "as 
>> large as the largest of any of these" without having accurate 
>> TargetData. The union type was supposed to do that and nothing more.
> 
> I sent an example earlier showing that you can do this already without
> union types.

Close. Your trick does perform a ptrtoint which requires knowing what
int size is large enough. Fortunately in your case it's indexing off of
null so it's very unlikely that it won't fit in 16 bits or less, but
it's still not as good as a first-class union.

I've been thinking about the original suggestion and the reasons I 
objected to it. It seems that the original suggest was to think about a 
union as a structure where the offset into each element is zero instead 
of being contiguous to each other. That makes the original proposal make 
a whole lot more sense to me than it did originally.

Despite Chris' message to the contrary, I still think u{i32, i32} 
shouldn't be allowed (rather, it should be folded to u{i32} by the 
getter). We could provide an accessor that returns the element number 
for a given Type* and the only drawback is that it means doing an extra 
lookup through a small list. Allowing GEP makes sense, and unions should 
certainly be first class aggregates.

Nick



From daniel at zuster.org  Wed May 20 23:36:47 2009
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 21 May 2009 04:36:47 -0000
Subject: [llvm-commits] [klee] r72205 [1/3] - in /klee/trunk: ./ autoconf/
 docs/ docs/SMT-COMP/ examples/ examples/regexp/ examples/sort/ include/
 include/expr/ include/klee/ include/klee/Config/ include/klee/Internal/
 include/klee/Internal/ADT/ include/klee/Internal/Module/
 include/klee/Internal/Support/ include/klee/Internal/System/
 include/klee/util/ lib/ lib/Basic/ lib/Core/ lib/Expr/ lib/Module/
 lib/Solver/ lib/Support/ runtime/ runtime/Intrinsic/ runtime/POSIX/
 runtime/POSIX/testing-dir/ runtime/POSIX/testing-dir/e/ runti...
Message-ID: <200905210437.n4L4b0oL006290@zion.cs.uiuc.edu>

Author: ddunbar
Date: Wed May 20 23:36:41 2009
New Revision: 72205

URL: http://llvm.org/viewvc/llvm-project?rev=72205&view=rev
Log:
Initial KLEE checkin.
 - Lots more tweaks, documentation, and web page content is needed,
   but this should compile & work on OS X & Linux.

Added:
    klee/trunk/Makefile
    klee/trunk/Makefile.common
    klee/trunk/Makefile.config.in
    klee/trunk/TODO.txt
    klee/trunk/autoconf/
    klee/trunk/autoconf/AutoRegen.sh   (with props)
    klee/trunk/autoconf/config.guess   (with props)
    klee/trunk/autoconf/config.sub   (with props)
    klee/trunk/autoconf/configure.ac
    klee/trunk/autoconf/install-sh   (with props)
    klee/trunk/configure   (with props)
    klee/trunk/docs/
    klee/trunk/docs/SMT-COMP/
    klee/trunk/docs/SMT-COMP/BitVector_ArraysEx.smt
    klee/trunk/docs/SMT-COMP/BitVectors.smt
    klee/trunk/docs/SMT-COMP/QF_AUFBV.smt
    klee/trunk/docs/SMT-COMP/QF_BV.smt
    klee/trunk/docs/doxygen.cfg
    klee/trunk/docs/intro
    klee/trunk/docs/overview
    klee/trunk/examples/
    klee/trunk/examples/regexp/
    klee/trunk/examples/regexp/Regexp.c
    klee/trunk/examples/regexp/notes.txt
    klee/trunk/examples/sort/
    klee/trunk/examples/sort/sort.c
    klee/trunk/include/
    klee/trunk/include/expr/
    klee/trunk/include/expr/Lexer.h
    klee/trunk/include/expr/Parser.h
    klee/trunk/include/klee/
    klee/trunk/include/klee/Config/
    klee/trunk/include/klee/Config/config.h.in
    klee/trunk/include/klee/Constraints.h
    klee/trunk/include/klee/ExecutionState.h
    klee/trunk/include/klee/Expr.h
    klee/trunk/include/klee/IncompleteSolver.h
    klee/trunk/include/klee/Internal/
    klee/trunk/include/klee/Internal/ADT/
    klee/trunk/include/klee/Internal/ADT/BOut.h
    klee/trunk/include/klee/Internal/ADT/DiscretePDF.h
    klee/trunk/include/klee/Internal/ADT/DiscretePDF.inc
    klee/trunk/include/klee/Internal/ADT/ImmutableMap.h
    klee/trunk/include/klee/Internal/ADT/ImmutableSet.h
    klee/trunk/include/klee/Internal/ADT/ImmutableTree.h
    klee/trunk/include/klee/Internal/ADT/MapOfSets.h
    klee/trunk/include/klee/Internal/ADT/RNG.h
    klee/trunk/include/klee/Internal/ADT/TreeStream.h
    klee/trunk/include/klee/Internal/Module/
    klee/trunk/include/klee/Internal/Module/Cell.h
    klee/trunk/include/klee/Internal/Module/InstructionInfoTable.h
    klee/trunk/include/klee/Internal/Module/KInstIterator.h
    klee/trunk/include/klee/Internal/Module/KInstruction.h
    klee/trunk/include/klee/Internal/Module/KModule.h
    klee/trunk/include/klee/Internal/README.txt
    klee/trunk/include/klee/Internal/Support/
    klee/trunk/include/klee/Internal/Support/FloatEvaluation.h
    klee/trunk/include/klee/Internal/Support/IntEvaluation.h
    klee/trunk/include/klee/Internal/Support/ModuleUtil.h
    klee/trunk/include/klee/Internal/Support/QueryLog.h
    klee/trunk/include/klee/Internal/Support/Timer.h
    klee/trunk/include/klee/Internal/System/
    klee/trunk/include/klee/Internal/System/Time.h
    klee/trunk/include/klee/Interpreter.h
    klee/trunk/include/klee/Machine.h
    klee/trunk/include/klee/Solver.h
    klee/trunk/include/klee/SolverImpl.h
    klee/trunk/include/klee/Statistic.h
    klee/trunk/include/klee/Statistics.h
    klee/trunk/include/klee/TimerStatIncrementer.h
    klee/trunk/include/klee/klee.h
    klee/trunk/include/klee/util/
    klee/trunk/include/klee/util/Assignment.h
    klee/trunk/include/klee/util/BitArray.h
    klee/trunk/include/klee/util/Bits.h
    klee/trunk/include/klee/util/ExprEvaluator.h
    klee/trunk/include/klee/util/ExprHashMap.h
    klee/trunk/include/klee/util/ExprPPrinter.h
    klee/trunk/include/klee/util/ExprRangeEvaluator.h
    klee/trunk/include/klee/util/ExprUtil.h
    klee/trunk/include/klee/util/ExprVisitor.h
    klee/trunk/include/klee/util/Ref.h
    klee/trunk/lib/
    klee/trunk/lib/Basic/
    klee/trunk/lib/Basic/BOut.cpp
    klee/trunk/lib/Basic/Makefile
    klee/trunk/lib/Basic/README.txt
    klee/trunk/lib/Basic/Statistics.cpp
    klee/trunk/lib/Core/
    klee/trunk/lib/Core/AddressSpace.cpp
    klee/trunk/lib/Core/AddressSpace.h
    klee/trunk/lib/Core/CallPathManager.cpp
    klee/trunk/lib/Core/CallPathManager.h
    klee/trunk/lib/Core/Common.cpp
    klee/trunk/lib/Core/Common.h
    klee/trunk/lib/Core/CoreStats.cpp
    klee/trunk/lib/Core/CoreStats.h
    klee/trunk/lib/Core/ExecutionState.cpp
    klee/trunk/lib/Core/Executor.cpp
    klee/trunk/lib/Core/Executor.h
    klee/trunk/lib/Core/ExecutorTimers.cpp
    klee/trunk/lib/Core/ExecutorUtil.cpp
    klee/trunk/lib/Core/ExternalDispatcher.cpp
    klee/trunk/lib/Core/ExternalDispatcher.h
    klee/trunk/lib/Core/ImpliedValue.cpp
    klee/trunk/lib/Core/ImpliedValue.h
    klee/trunk/lib/Core/Makefile   (with props)
    klee/trunk/lib/Core/Memory.cpp
    klee/trunk/lib/Core/Memory.h
    klee/trunk/lib/Core/MemoryManager.cpp
    klee/trunk/lib/Core/MemoryManager.h
    klee/trunk/lib/Core/ObjectHolder.h
    klee/trunk/lib/Core/PTree.cpp
    klee/trunk/lib/Core/PTree.h
    klee/trunk/lib/Core/Searcher.cpp
    klee/trunk/lib/Core/Searcher.h
    klee/trunk/lib/Core/SeedInfo.cpp
    klee/trunk/lib/Core/SeedInfo.h
    klee/trunk/lib/Core/SpecialFunctionHandler.cpp
    klee/trunk/lib/Core/SpecialFunctionHandler.h
    klee/trunk/lib/Core/StatsTracker.cpp
    klee/trunk/lib/Core/StatsTracker.h
    klee/trunk/lib/Core/TimingSolver.cpp
    klee/trunk/lib/Core/TimingSolver.h
    klee/trunk/lib/Core/UserSearcher.cpp
    klee/trunk/lib/Core/UserSearcher.h
    klee/trunk/lib/Expr/
    klee/trunk/lib/Expr/Constraints.cpp
    klee/trunk/lib/Expr/Expr.cpp
    klee/trunk/lib/Expr/ExprEvaluator.cpp
    klee/trunk/lib/Expr/ExprPPrinter.cpp
    klee/trunk/lib/Expr/ExprUtil.cpp
    klee/trunk/lib/Expr/ExprVisitor.cpp
    klee/trunk/lib/Expr/Lexer.cpp
    klee/trunk/lib/Expr/Makefile
    klee/trunk/lib/Expr/Parser.cpp
    klee/trunk/lib/Expr/Updates.cpp
    klee/trunk/lib/Makefile
    klee/trunk/lib/Module/
    klee/trunk/lib/Module/Checks.cpp
    klee/trunk/lib/Module/InstructionInfoTable.cpp
    klee/trunk/lib/Module/IntrinsicCleaner.cpp
    klee/trunk/lib/Module/KInstruction.cpp
    klee/trunk/lib/Module/KModule.cpp
    klee/trunk/lib/Module/LowerSwitch.cpp
    klee/trunk/lib/Module/Makefile   (with props)
    klee/trunk/lib/Module/ModuleUtil.cpp
    klee/trunk/lib/Module/Optimize.cpp
    klee/trunk/lib/Module/Passes.h
    klee/trunk/lib/Module/PhiCleaner.cpp
    klee/trunk/lib/Module/RaiseAsm.cpp
    klee/trunk/lib/README.txt
    klee/trunk/lib/Solver/
    klee/trunk/lib/Solver/CachingSolver.cpp
    klee/trunk/lib/Solver/CexCachingSolver.cpp
    klee/trunk/lib/Solver/ConstantDivision.cpp
    klee/trunk/lib/Solver/ConstantDivision.h
    klee/trunk/lib/Solver/FastCexSolver.cpp
    klee/trunk/lib/Solver/IncompleteSolver.cpp
    klee/trunk/lib/Solver/IndependentSolver.cpp
    klee/trunk/lib/Solver/Makefile   (with props)
    klee/trunk/lib/Solver/PCLoggingSolver.cpp
    klee/trunk/lib/Solver/STPBuilder.cpp
    klee/trunk/lib/Solver/STPBuilder.h
    klee/trunk/lib/Solver/Solver.cpp
    klee/trunk/lib/Solver/SolverStats.cpp
    klee/trunk/lib/Solver/SolverStats.h
    klee/trunk/lib/Support/
    klee/trunk/lib/Support/Makefile
    klee/trunk/lib/Support/README.txt
    klee/trunk/lib/Support/RNG.cpp
    klee/trunk/lib/Support/Time.cpp
    klee/trunk/lib/Support/Timer.cpp
    klee/trunk/lib/Support/TreeStream.cpp
    klee/trunk/runtime/
    klee/trunk/runtime/Intrinsic/
    klee/trunk/runtime/Intrinsic/Makefile
    klee/trunk/runtime/Intrinsic/klee_div_zero_check.c
    klee/trunk/runtime/Intrinsic/klee_int.c
    klee/trunk/runtime/Intrinsic/klee_make_symbolic.c
    klee/trunk/runtime/Intrinsic/klee_range.c
    klee/trunk/runtime/Intrinsic/memcpy.c
    klee/trunk/runtime/Intrinsic/memmove.c
    klee/trunk/runtime/Intrinsic/mempcpy.c
    klee/trunk/runtime/Intrinsic/memset.c
    klee/trunk/runtime/Makefile   (with props)
    klee/trunk/runtime/POSIX/
    klee/trunk/runtime/POSIX/Makefile
    klee/trunk/runtime/POSIX/fd.c
    klee/trunk/runtime/POSIX/fd.h
    klee/trunk/runtime/POSIX/fd_32.c
    klee/trunk/runtime/POSIX/fd_64.c
    klee/trunk/runtime/POSIX/fd_init.c
    klee/trunk/runtime/POSIX/illegal.c
    klee/trunk/runtime/POSIX/klee_init_env.c
    klee/trunk/runtime/POSIX/misc.c
    klee/trunk/runtime/POSIX/selinux.c
    klee/trunk/runtime/POSIX/stubs.c
    klee/trunk/runtime/POSIX/testing-dir/
    klee/trunk/runtime/POSIX/testing-dir/a   (with props)
    klee/trunk/runtime/POSIX/testing-dir/b   (with props)
    klee/trunk/runtime/POSIX/testing-dir/c   (with props)
    klee/trunk/runtime/POSIX/testing-dir/d
    klee/trunk/runtime/POSIX/testing-dir/e/
    klee/trunk/runtime/POSIX/testing-env
    klee/trunk/runtime/Runtest/
    klee/trunk/runtime/Runtest/Makefile
    klee/trunk/runtime/Runtest/intrinsics.c
    klee/trunk/runtime/klee-libc/
    klee/trunk/runtime/klee-libc/Makefile   (with props)
    klee/trunk/runtime/klee-libc/__cxa_atexit.c
    klee/trunk/runtime/klee-libc/abort.c
    klee/trunk/runtime/klee-libc/atexit.c
    klee/trunk/runtime/klee-libc/atoi.c
    klee/trunk/runtime/klee-libc/calloc.c
    klee/trunk/runtime/klee-libc/htonl.c
    klee/trunk/runtime/klee-libc/klee-choose.c
    klee/trunk/runtime/klee-libc/memchr.c
    klee/trunk/runtime/klee-libc/memcmp.c
    klee/trunk/runtime/klee-libc/memcpy.c
    klee/trunk/runtime/klee-libc/memmove.c
    klee/trunk/runtime/klee-libc/mempcpy.c
    klee/trunk/runtime/klee-libc/memset.c
    klee/trunk/runtime/klee-libc/putchar.c
    klee/trunk/runtime/klee-libc/stpcpy.c
    klee/trunk/runtime/klee-libc/strcat.c
    klee/trunk/runtime/klee-libc/strchr.c
    klee/trunk/runtime/klee-libc/strcmp.c
    klee/trunk/runtime/klee-libc/strcoll.c
    klee/trunk/runtime/klee-libc/strcpy.c
    klee/trunk/runtime/klee-libc/strlen.c
    klee/trunk/runtime/klee-libc/strncmp.c
    klee/trunk/runtime/klee-libc/strncpy.c
    klee/trunk/runtime/klee-libc/strrchr.c
    klee/trunk/runtime/klee-libc/strtol.c
    klee/trunk/runtime/klee-libc/strtoul.c
    klee/trunk/runtime/klee-libc/tolower.c
    klee/trunk/runtime/klee-libc/toupper.c
    klee/trunk/scripts/
    klee/trunk/scripts/IStatsMerge.py   (with props)
    klee/trunk/scripts/IStatsSum.py   (with props)
    klee/trunk/scripts/PrintStats.py   (with props)
    klee/trunk/scripts/klee-control   (with props)
    klee/trunk/scripts/objdump   (with props)
    klee/trunk/stp/
    klee/trunk/stp/AST/
    klee/trunk/stp/AST/AST.cpp
    klee/trunk/stp/AST/AST.h
    klee/trunk/stp/AST/ASTKind.kinds
    klee/trunk/stp/AST/ASTUtil.cpp
    klee/trunk/stp/AST/ASTUtil.h
    klee/trunk/stp/AST/BitBlast.cpp
    klee/trunk/stp/AST/Makefile
    klee/trunk/stp/AST/STLport_config.h
    klee/trunk/stp/AST/SimpBool.cpp
    klee/trunk/stp/AST/ToCNF.cpp
    klee/trunk/stp/AST/ToSAT.cpp
    klee/trunk/stp/AST/Transform.cpp
    klee/trunk/stp/AST/asttest.cpp
    klee/trunk/stp/AST/bbtest.cpp
    klee/trunk/stp/AST/cnftest.cpp
    klee/trunk/stp/AST/genkinds.pl   (with props)
    klee/trunk/stp/INSTALL
    klee/trunk/stp/LICENSE
    klee/trunk/stp/Makefile
    klee/trunk/stp/Makefile.common.in
    klee/trunk/stp/README
    klee/trunk/stp/bitvec/
    klee/trunk/stp/bitvec/Makefile
    klee/trunk/stp/bitvec/consteval.cpp
    klee/trunk/stp/c_interface/
    klee/trunk/stp/c_interface/Makefile
    klee/trunk/stp/c_interface/c_interface.cpp
    klee/trunk/stp/c_interface/c_interface.h
    klee/trunk/stp/c_interface/fdstream.h
    klee/trunk/stp/constantbv/
    klee/trunk/stp/constantbv/Makefile
    klee/trunk/stp/constantbv/constantbv.cpp
    klee/trunk/stp/constantbv/constantbv.h
    klee/trunk/stp/parser/
    klee/trunk/stp/parser/Makefile
    klee/trunk/stp/parser/PL.lex
    klee/trunk/stp/parser/PL.y
    klee/trunk/stp/parser/let-funcs.cpp
    klee/trunk/stp/parser/main.cpp
    klee/trunk/stp/parser/smtlib.lex
    klee/trunk/stp/parser/smtlib.y
    klee/trunk/stp/sat/
    klee/trunk/stp/sat/Global.h
    klee/trunk/stp/sat/Heap.h
    klee/trunk/stp/sat/LICENSE
    klee/trunk/stp/sat/Makefile
    klee/trunk/stp/sat/Simplifier.C
    klee/trunk/stp/sat/Solver.C
    klee/trunk/stp/sat/Solver.h
    klee/trunk/stp/sat/SolverTypes.h
    klee/trunk/stp/sat/Sort.h
    klee/trunk/stp/sat/VarOrder.h
    klee/trunk/stp/simplifier/
    klee/trunk/stp/simplifier/Makefile
    klee/trunk/stp/simplifier/bvsolver.cpp
    klee/trunk/stp/simplifier/bvsolver.h
    klee/trunk/stp/simplifier/simplifier.cpp
    klee/trunk/test/
    klee/trunk/test/CXX/
    klee/trunk/test/CXX/ArrayNew.cpp
    klee/trunk/test/CXX/New.cpp
    klee/trunk/test/CXX/SimpleVirtual.cpp
    klee/trunk/test/CXX/StaticConstructor.cpp
    klee/trunk/test/CXX/StaticDestructor.cpp
    klee/trunk/test/CXX/Trivial.cpp
    klee/trunk/test/CXX/dg.exp
    klee/trunk/test/Concrete/
    klee/trunk/test/Concrete/BitwiseOps.ll
    klee/trunk/test/Concrete/BoolReadWrite.ll
    klee/trunk/test/Concrete/Casts.ll   (with props)
    klee/trunk/test/Concrete/CmpEq.ll
    klee/trunk/test/Concrete/ConcreteTest.py   (with props)
    klee/trunk/test/Concrete/ConstantExpr.ll
    klee/trunk/test/Concrete/FloatingPointOps.ll
    klee/trunk/test/Concrete/GlobalInitializers.ll   (with props)
    klee/trunk/test/Concrete/GlobalVariable.ll
    klee/trunk/test/Concrete/ICmp.ll
    klee/trunk/test/Concrete/InvokeAndReturn.ll
    klee/trunk/test/Concrete/InvokeAndUnwindOnce.ll
    klee/trunk/test/Concrete/InvokeAndUnwindTwice.ll
    klee/trunk/test/Concrete/Makefile
    klee/trunk/test/Concrete/OneCall.ll
    klee/trunk/test/Concrete/OverlappingPhiNodes.ll
    klee/trunk/test/Concrete/Select.ll
    klee/trunk/test/Concrete/Shifts.ll
    klee/trunk/test/Concrete/SimpleStoreAndLoad.ll
    klee/trunk/test/Concrete/UnconditionalBranch.ll
    klee/trunk/test/Concrete/UnconditionalBranchWithSimplePhi.ll
    klee/trunk/test/Concrete/UnorderedPhiNodes.ll
    klee/trunk/test/Concrete/_testingUtils.c
    klee/trunk/test/Concrete/ackermann.c
    klee/trunk/test/Concrete/arith_test.ll
    klee/trunk/test/Coverage/
    klee/trunk/test/Coverage/ReadArgs.c
    klee/trunk/test/Coverage/ReplayOutDir.c
    klee/trunk/test/Coverage/dg.exp
    klee/trunk/test/Dogfood/
    klee/trunk/test/Dogfood/ImmutableSet.cpp
    klee/trunk/test/Dogfood/dg.exp
    klee/trunk/test/Expr/
    klee/trunk/test/Expr/Lexer/
    klee/trunk/test/Expr/Parser/
    klee/trunk/test/Expr/dg.exp
    klee/trunk/test/Feature/
    klee/trunk/test/Feature/Alias.c
    klee/trunk/test/Feature/AliasFunction.c
    klee/trunk/test/Feature/AliasFunctionExit.c
    klee/trunk/test/Feature/AsmAddresses.c
    klee/trunk/test/Feature/ByteSwap.c
    klee/trunk/test/Feature/CallToUndefinedExternal.cpp
    klee/trunk/test/Feature/CheckForImpliedValue.c.failing
    klee/trunk/test/Feature/CheckMemoryAccess.c
    klee/trunk/test/Feature/CopyOnWrite.c
    klee/trunk/test/Feature/DanglingConcreteReadExpr.c
    klee/trunk/test/Feature/DefineFixedObject.c
    klee/trunk/test/Feature/DoubleFree.c
    klee/trunk/test/Feature/DumpStatesOnHalt.c
    klee/trunk/test/Feature/Envp.c
    klee/trunk/test/Feature/ExprLogging.c
    klee/trunk/test/Feature/ExternalWeakLinkage.c
    klee/trunk/test/Feature/FunctionPointer.c
    klee/trunk/test/Feature/GetValue.c
    klee/trunk/test/Feature/ImpliedValue.c.failing
    klee/trunk/test/Feature/InAndOutOfBounds.c
    klee/trunk/test/Feature/IndirectCallToBuiltin.c
    klee/trunk/test/Feature/IndirectCallToExternal.c
    klee/trunk/test/Feature/InvalidBitfieldAccess.c.failing
    klee/trunk/test/Feature/IsSymbolic.c
    klee/trunk/test/Feature/KleeReportError.c
    klee/trunk/test/Feature/LongDoubleSupport.c
    klee/trunk/test/Feature/LowerSwitch.c
    klee/trunk/test/Feature/MakeConcreteSymbolic.c
    klee/trunk/test/Feature/MakeSymbolicName.c
    klee/trunk/test/Feature/MemoryLimit.c
    klee/trunk/test/Feature/MultipleFreeResolution.c
    klee/trunk/test/Feature/MultipleReadResolution.c
    klee/trunk/test/Feature/MultipleReallocResolution.c
    klee/trunk/test/Feature/MultipleWriteResolution.c
    klee/trunk/test/Feature/NamedSeedMatching.c
    klee/trunk/test/Feature/OneFreeError.c
    klee/trunk/test/Feature/OneOutOfBounds.c
    klee/trunk/test/Feature/Optimize.c
    klee/trunk/test/Feature/OverlappedError.c
    klee/trunk/test/Feature/PreferCex.c
    klee/trunk/test/Feature/RaiseAsm.c
    klee/trunk/test/Feature/ReallocFailure.c
    klee/trunk/test/Feature/ReplayPath.c
    klee/trunk/test/Feature/Searchers.c
    klee/trunk/test/Feature/SetForking.c
    klee/trunk/test/Feature/Vararg.c
    klee/trunk/test/Feature/WithLibc.c
    klee/trunk/test/Feature/WriteCov.c
    klee/trunk/test/Feature/_utils._ll
    klee/trunk/test/Feature/const_array_opt1.c
    klee/trunk/test/Feature/dg.exp
    klee/trunk/test/Feature/utils.h
    klee/trunk/test/Makefile
    klee/trunk/test/Makefile.tests
    klee/trunk/test/README
    klee/trunk/test/Runtime/
    klee/trunk/test/Runtime/POSIX/
    klee/trunk/test/Runtime/POSIX/DirConsistency.c
    klee/trunk/test/Runtime/POSIX/DirSeek.c
    klee/trunk/test/Runtime/POSIX/FDNumbers.c
    klee/trunk/test/Runtime/POSIX/FD_Fail.c
    klee/trunk/test/Runtime/POSIX/FD_Fail2.c
    klee/trunk/test/Runtime/POSIX/Fcntl.c
    klee/trunk/test/Runtime/POSIX/FilePerm.c
    klee/trunk/test/Runtime/POSIX/FreeArgv.c
    klee/trunk/test/Runtime/POSIX/Getenv.c
    klee/trunk/test/Runtime/POSIX/Ioctl.c
    klee/trunk/test/Runtime/POSIX/Isatty.c
    klee/trunk/test/Runtime/POSIX/PrgName.c
    klee/trunk/test/Runtime/POSIX/Read1.c
    klee/trunk/test/Runtime/POSIX/SELinux.c
    klee/trunk/test/Runtime/POSIX/SeedAndFail.c
    klee/trunk/test/Runtime/POSIX/Stdin.c
    klee/trunk/test/Runtime/POSIX/Write1.c
    klee/trunk/test/Runtime/POSIX/Write2.c
    klee/trunk/test/Runtime/POSIX/dg.exp
    klee/trunk/test/Runtime/Uclibc/
    klee/trunk/test/Runtime/Uclibc/2007-10-08-optimization-calls-wrong-libc-functions.c
    klee/trunk/test/Runtime/Uclibc/2008-03-04-libc-atexit-uses-dso-handle.c
    klee/trunk/test/Runtime/Uclibc/Environ.c
    klee/trunk/test/Runtime/Uclibc/dg.exp
    klee/trunk/test/TestRunner.sh   (with props)
    klee/trunk/test/lib/
    klee/trunk/test/lib/llvm.exp
    klee/trunk/test/regression/
    klee/trunk/test/regression/2007-07-25-invalid-stp-array-binding-to-objectstate.c
    klee/trunk/test/regression/2007-07-30-unflushed-byte.c
    klee/trunk/test/regression/2007-08-01-bool-zext-in-call.ll
    klee/trunk/test/regression/2007-08-01-cache-unclear-on-overwrite-flushed.c
    klee/trunk/test/regression/2007-08-06-64bit-shift.c
    klee/trunk/test/regression/2007-08-06-access-after-free.c
    klee/trunk/test/regression/2007-08-08-free-zero.c
    klee/trunk/test/regression/2007-08-16-invalid-constant-value.c
    klee/trunk/test/regression/2007-08-16-valid-write-to-freed-object.c
    klee/trunk/test/regression/2007-10-11-free-of-alloca.c
    klee/trunk/test/regression/2007-10-11-illegal-access-after-free-and-branch.c
    klee/trunk/test/regression/2007-10-12-failed-make-symbolic-after-copy.c
    klee/trunk/test/regression/2008-02-11-phi-nodes-after-invoke.ll
    klee/trunk/test/regression/2008-03-04-free-of-global.c
    klee/trunk/test/regression/2008-03-11-free-of-malloc-zero.c
    klee/trunk/test/regression/2008-04-10-bad-alloca-free.c
    klee/trunk/test/regression/2008-05-23-gep-with-global-const.c
    klee/trunk/test/regression/dg.exp
    klee/trunk/tools/
    klee/trunk/tools/Makefile
    klee/trunk/tools/gen-random-bout/
    klee/trunk/tools/gen-random-bout/Makefile
    klee/trunk/tools/gen-random-bout/gen-random-bout.cpp
    klee/trunk/tools/kleaver/
    klee/trunk/tools/kleaver/Makefile
    klee/trunk/tools/kleaver/main.cpp
    klee/trunk/tools/klee/
    klee/trunk/tools/klee-bout-tool/
    klee/trunk/tools/klee-bout-tool/Makefile
    klee/trunk/tools/klee-bout-tool/klee-bout-tool   (with props)
    klee/trunk/tools/klee/Debug.cpp
    klee/trunk/tools/klee/Makefile
    klee/trunk/tools/klee/main.cpp
    klee/trunk/unittests/
    klee/trunk/unittests/Expr/
    klee/trunk/unittests/Expr/ExprTest.cpp
    klee/trunk/unittests/Expr/Makefile
    klee/trunk/unittests/Makefile   (with props)
    klee/trunk/unittests/Solver/
    klee/trunk/unittests/Solver/Makefile
    klee/trunk/unittests/Solver/SolverTest.cpp
    klee/trunk/unittests/TestMain.cpp
    klee/trunk/utils/
    klee/trunk/utils/emacs/
    klee/trunk/utils/emacs/klee-pc-mode.el
    klee/trunk/utils/valgrind/
    klee/trunk/utils/valgrind/README.txt
    klee/trunk/utils/valgrind/valgrind-llvm.supp
    klee/trunk/utils/valgrind/valgrind-stp.supp
    klee/trunk/www/Examples.html
    klee/trunk/www/GetInvolved.html
    klee/trunk/www/GetStarted.html
Modified:
    klee/trunk/LICENSE.TXT
    klee/trunk/README.txt
    klee/trunk/www/bugs.html
    klee/trunk/www/content.css
    klee/trunk/www/index.html
    klee/trunk/www/install.html
    klee/trunk/www/menu.html.incl
    klee/trunk/www/tutorials.html

Modified: klee/trunk/LICENSE.TXT
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/LICENSE.TXT?rev=72205&r1=72204&r2=72205&view=diff

==============================================================================
--- klee/trunk/LICENSE.TXT (original)
+++ klee/trunk/LICENSE.TXT Wed May 20 23:36:41 2009
@@ -11,8 +11,7 @@
 
     klee Team
 
-    Stanford Checking Group: Daniel Dunbar, Cristian Cadar, Peter
-    Pawlowki, Dawson Engler.
+    Stanford Checking Group
 
     http://klee.llvm.org
 
@@ -30,10 +29,10 @@
       this list of conditions and the following disclaimers in the
       documentation and/or other materials provided with the distribution.
 
-    * Neither the names of the LLVM Team, University of Illinois at
-      Urbana-Champaign, nor the names of its contributors may be used to
-      endorse or promote products derived from this Software without specific
-      prior written permission.
+    * Neither the names of the klee Team, Stanford University, nor the
+      names of its contributors may be used to endorse or promote
+      products derived from this Software without specific prior
+      written permission.
 
 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
@@ -50,10 +49,9 @@
 to that code.
 
 The disclaimer of warranty in the University of Illinois Open Source License
-applies to all code in the klee Distribution, and nothing in any of the
-other licenses gives permission to use the names of the klee Team or the
-University of Illinois to endorse or promote products derived from this
-Software.
+applies to all code in the klee Distribution, and nothing in any of the other 
+licenses gives permission to use the names of the klee Team or Stanford 
+University to endorse or promote products derived from this Software.
 
 The following pieces of software have additional or alternate copyrights,
 licenses, and/or restrictions:
@@ -61,4 +59,5 @@
 Program             Directory
 -------             ---------
 STP                 klee/stp
+klee-libc           runtime/klee-libc
 

Added: klee/trunk/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/Makefile (added)
+++ klee/trunk/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,57 @@
+#===-- klee/Makefile ---------------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+#
+# Indicates our relative path to the top of the project's root directory.
+#
+LEVEL = .
+
+DIRS = stp lib tools runtime
+EXTRA_DIST = include
+
+# Only build support directories when building unittests.
+ifeq ($(MAKECMDGOALS),unittests)
+  DIRS := $(filter-out tools runtime, $(DIRS)) unittests
+  OPTIONAL_DIRS :=
+endif
+
+#
+# Include the Master Makefile that knows how to build all.
+#
+include $(LEVEL)/Makefile.common
+
+.PHONY: doxygen
+doxygen:
+	doxygen docs/doxygen.cfg
+
+.PHONY: cscope.files
+cscope.files:
+	find \
+          lib include stp tools runtime examples unittests \
+          -name Makefile -or \
+          -name \*.in -or \
+          -name \*.c -or \
+          -name \*.cpp -or \
+          -name \*.exp -or \
+          -name \*.inc -or \
+          -name \*.h | sort > cscope.files
+
+test::
+	-(cd test/ && make)
+
+.PHONY: klee-cov
+klee-cov:
+	rm -rf klee-cov
+	zcov-scan --look-up-dirs=1 klee.zcov .
+	zcov-genhtml --root $$(pwd) klee.zcov klee-cov
+
+clean::
+	$(MAKE) -C test clean 
+	$(MAKE) -C unittests clean
+	rm -rf docs/doxygen

Added: klee/trunk/Makefile.common
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/Makefile.common?rev=72205&view=auto

==============================================================================
--- klee/trunk/Makefile.common (added)
+++ klee/trunk/Makefile.common Wed May 20 23:36:41 2009
@@ -0,0 +1,26 @@
+# -*- Makefile -*-
+
+include $(LEVEL)/Makefile.config
+
+# Include LLVM's Master Makefile config and rules.
+include $(LLVM_OBJ_ROOT)/Makefile.config
+
+ifeq ($(BYTECODE_LIBRARY), 1)
+#
+# Override make variables based on the runtime configuration. We want
+# to override whatever the user may have said on the command line,
+# hence the use of override.
+#
+
+override ENABLE_OPTIMIZED := $(RUNTIME_ENABLE_OPTIMIZED)
+override DISABLE_ASSERTIONS := $(RUNTIME_DISABLE_ASSERTIONS)
+override ENABLE_PROFILING := $(RUNTIME_ENABLE_PROFILING)
+override ENABLE_COVERAGE := $(RUNTIME_ENABLE_COVERAGE)
+endif
+
+include $(LLVM_OBJ_ROOT)/Makefile.rules
+
+LD.Flags += -L$(PROJ_SRC_ROOT)/stp/lib
+CXX.Flags += -DLLVM_23
+CXX.Flags += -I$(PROJ_SRC_ROOT)/stp/include
+CXX.Flags += -DKLEE_DIR=\"$(PROJ_SRC_ROOT)\"

Added: klee/trunk/Makefile.config.in
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/Makefile.config.in?rev=72205&view=auto

==============================================================================
--- klee/trunk/Makefile.config.in (added)
+++ klee/trunk/Makefile.config.in Wed May 20 23:36:41 2009
@@ -0,0 +1,45 @@
+# -*- Makefile -*-
+
+# Set the name of the project here
+PROJECT_NAME := klee
+PROJ_VERSION := 0.01
+
+# Set this variable to the top of the LLVM source tree.
+LLVM_SRC_ROOT = @LLVM_SRC@
+
+# Set this variable to the top level directory where LLVM was built
+# (this is *not* the same as OBJ_ROOT as defined in LLVM's Makefile.config).
+LLVM_OBJ_ROOT = @LLVM_OBJ@
+
+# Set the directory root of this project's source files
+PROJ_SRC_ROOT := $(subst //,/, at abs_top_srcdir@)
+
+# Set the root directory of this project's object files
+PROJ_OBJ_ROOT := $(subst //,/, at abs_top_objdir@)
+
+# Set the root directory of this project's install prefix
+PROJ_INSTALL_ROOT := @prefix@
+
+ENABLE_POSIX_RUNTIME := @ENABLE_POSIX_RUNTIME@
+ENABLE_STPLOG := @ENABLE_STPLOG@
+ENABLE_UCLIBC := @ENABLE_UCLIBC@
+
+HAVE_SELINUX := @HAVE_SELINUX@
+
+RUNTIME_ENABLE_OPTIMIZED := @RUNTIME_ENABLE_OPTIMIZED@
+RUNTIME_DISABLE_ASSERTIONS :=
+RUNTIME_ENABLE_COVERAGE :=
+RUNTIME_ENABLE_PROFILING :=
+
+# A list of "features" which tests can check for in XFAIL:
+TEST_FEATURE_LIST :=
+
+ifeq ($(HAVE_SELINUX_SELINUX_H),1)
+  TEST_FEATURE_LIST += have-selinux
+else
+  TEST_FEATURE_LIST += no-selinux
+endif
+
+CFLAGS := @CFLAGS@
+CXXFLAGS := @CXXFLAGS@
+LDFLAGS := @LDFLAGS@

Modified: klee/trunk/README.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/README.txt?rev=72205&r1=72204&r2=72205&view=diff

==============================================================================
--- klee/trunk/README.txt (original)
+++ klee/trunk/README.txt Wed May 20 23:36:41 2009
@@ -1,24 +1,24 @@
 //===----------------------------------------------------------------------===//
 // Klee Symbolic Virtual Machine
 //===----------------------------------------------------------------------===//
-                                                             Daniel Dunbar
 
 klee is a symbolic virtual machine built on top of the LLVM compiler
-infrastructure. Currently, there are two primary components.
+infrastructure. Currently, there are two primary components:
 
-1. The core symbolic virtual machine engine; this is responsible for
-executing LLVM bitcode modules with support for symbolic values. This
-is comprised of the code in lib/.
+  1. The core symbolic virtual machine engine; this is responsible for
+     executing LLVM bitcode modules with support for symbolic
+     values. This is comprised of the code in lib/.
 
-2. An emulation layer for the Linux system call interface, with
-additional support for making parts of the operating environment
-symbolic. This is found in models/simple.
+  2. A POSIX/Linux emulation layer oriented towards supporting uClibc,
+     with additional support for making parts of the operating system
+     environment symbolic.
 
-Additionally, there is a simple library in runtime/ which supports
-replaying computed inputs on native code. There is a more complicated
-library in replay/ which supports running inputs computed as part of
-the system call emulation layer natively -- setting up files, pipes,
-etc. on the native system to match the inputs that the emulation layer
-provided.
+Additionally, there is a simple library for replaying computed inputs
+on native code (for closed programs). There is also a more complicated
+infrastructure for replaying the inputs generated for the POSIX/Linux
+emulation layer, which handles running native programs in an
+environment that matches a computed test input, including setting up
+files, pipes, environment variables, and passing command line
+arguments.
 
-For further information, see the docs in www/.
+For further information, see the webpage or docs in www/.

Added: klee/trunk/TODO.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/TODO.txt?rev=72205&view=auto

==============================================================================
--- klee/trunk/TODO.txt (added)
+++ klee/trunk/TODO.txt Wed May 20 23:36:41 2009
@@ -0,0 +1,59 @@
+TODO
+--
+
+Build System / Configure / Release Cleanups
+--
+ o Rename .bout to .ktest (klee test)
+
+ o Rename .pc to .kquery (kleaver query)
+
+ o Configure doesn't check for bison / flex, we don't really use these
+   for anything important (just the command line STP tool), it would
+   be nice if they weren't required.
+
+ o Need a way to hide LLVM options in "klee --help".
+
+Klee Internal
+--
+ o Make sure that namespaces and .cpp locations match with reorganized
+   include locations.
+
+ o Add replay framework for POSIX model tests.
+
+Kleaver Internal
+--
+ o We need to fix the constants-in-exprs problem, this makes
+   separating out a Kleaver expr library much more difficult. There
+   are two parts:
+   
+     1. Pull fast (pure constant) path operations out of Expr.cpp,
+        into Executor.cpp.
+
+     2. Lift constants-are-immediate optimization out of ref<Expr>
+        into Cell. Expressions in memory already have the concrete
+        cache, so we get that part for free. 
+
+        We will need a way to distinguish if a cell has an expr or a
+        constant. Incidentally, this gives us an extra sentinel value
+        (is-expr == true and Expr* == null) we can use to mark
+        uninitialized-value of a register.
+
+   It may be worth sinking Expr construction into a Builder class
+   while we are at it.
+
+   There is a also a nice cleanup/perf win where we can work with
+   registers (Cells) directly, now that we build the constant table,
+   it might be worth doing this at the same time. This exposes a win
+   for IVC where it can write back a constant value into a register,
+   which needs to be done with care but would be a big improvement for
+   IVC.
+
+ o The stpArray field of an UpdateNode needs to die. This isn't as
+   easy as dropping it from the map, because we also need a
+   notification to free it. I think probably what we should do is
+   introduce an ExprContext can be used to deal with such things.
+     o The ExprContext could also have the default builder, for
+       example, which would make it easy to swap in an optimizing
+       builder.
+
+

Added: klee/trunk/autoconf/AutoRegen.sh
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/autoconf/AutoRegen.sh?rev=72205&view=auto

==============================================================================
--- klee/trunk/autoconf/AutoRegen.sh (added)
+++ klee/trunk/autoconf/AutoRegen.sh Wed May 20 23:36:41 2009
@@ -0,0 +1,41 @@
+#!/bin/sh
+
+if ([ "$#" != 1 ] || 
+    [ ! -d "$1" ] ||
+    [ ! -d "$1/autoconf/m4" ]); then
+    echo "usage: $0 <llvmsrc-dir>" 1>& 2
+    exit 1
+fi
+
+llvm_src_root=$(cd $1; pwd)
+llvm_m4=$llvm_src_root/autoconf/m4
+die () {
+	echo "$@" 1>&2
+	exit 1
+}
+test -d autoconf && test -f autoconf/configure.ac && cd autoconf
+test -f configure.ac || die "Can't find 'autoconf' dir; please cd into it first"
+autoconf --version | egrep '2\.60' > /dev/null
+if test $? -ne 0 ; then
+  die "Your autoconf was not detected as being 2.60"
+fi
+# Patch LLVM_SRC_ROOT in configure.ac
+sed -e "s#^LLVM_SRC_ROOT=.*#LLVM_SRC_ROOT=\"$llvm_src_root\"#" \
+    configure.ac > configure.tmp.ac
+echo "Regenerating aclocal.m4 with aclocal"
+rm -f aclocal.m4
+echo aclocal -I $llvm_m4 -I "$llvm_m4/.." || die "aclocal failed"
+aclocal -I $llvm_m4 -I "$llvm_m4/.." || die "aclocal failed"
+echo "Regenerating configure with autoconf 2.60"
+echo autoconf --warnings=all -o ../configure configure.tmp.ac || die "autoconf failed"
+autoconf --warnings=all -o ../configure configure.tmp.ac || die "autoconf failed"
+cp ../configure ../configure.bak
+sed -e "s#^LLVM_SRC_ROOT=.*#LLVM_SRC_ROOT=\".\"#" \
+    ../configure.bak > ../configure
+cd ..
+echo "Regenerating config.h.in with autoheader"
+autoheader --warnings=all \
+    -I autoconf -I autoconf/m4 \
+    autoconf/configure.tmp.ac || die "autoheader failed"
+rm -f autoconf/configure.tmp.ac configure.bak
+exit 0

Propchange: klee/trunk/autoconf/AutoRegen.sh

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/autoconf/config.guess
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/autoconf/config.guess?rev=72205&view=auto

==============================================================================
--- klee/trunk/autoconf/config.guess (added)
+++ klee/trunk/autoconf/config.guess Wed May 20 23:36:41 2009
@@ -0,0 +1,1447 @@
+#! /bin/sh
+# Attempt to guess a canonical system name.
+#   Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
+#   2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
+
+timestamp='2004-09-07'
+
+# This file is free software; you can redistribute it and/or modify it
+# under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+# General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
+#
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# Originally written by Per Bothner <per at bothner.com>.
+# Please send patches to <config-patches at gnu.org>.  Submit a context
+# diff and a properly formatted ChangeLog entry.
+#
+# This script attempts to guess a canonical system name similar to
+# config.sub.  If it succeeds, it prints the system name on stdout, and
+# exits with 0.  Otherwise, it exits with 1.
+#
+# The plan is that this can be called by configure scripts if you
+# don't specify an explicit build system type.
+
+me=`echo "$0" | sed -e 's,.*/,,'`
+
+usage="\
+Usage: $0 [OPTION]
+
+Output the configuration name of the system \`$me' is run on.
+
+Operation modes:
+  -h, --help         print this help, then exit
+  -t, --time-stamp   print date of last modification, then exit
+  -v, --version      print version number, then exit
+
+Report bugs and patches to <config-patches at gnu.org>."
+
+version="\
+GNU config.guess ($timestamp)
+
+Originally written by Per Bothner.
+Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004
+Free Software Foundation, Inc.
+
+This is free software; see the source for copying conditions.  There is NO
+warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE."
+
+help="
+Try \`$me --help' for more information."
+
+# Parse command line
+while test $# -gt 0 ; do
+  case $1 in
+    --time-stamp | --time* | -t )
+       echo "$timestamp" ; exit 0 ;;
+    --version | -v )
+       echo "$version" ; exit 0 ;;
+    --help | --h* | -h )
+       echo "$usage"; exit 0 ;;
+    -- )     # Stop option processing
+       shift; break ;;
+    - )	# Use stdin as input.
+       break ;;
+    -* )
+       echo "$me: invalid option $1$help" >&2
+       exit 1 ;;
+    * )
+       break ;;
+  esac
+done
+
+if test $# != 0; then
+  echo "$me: too many arguments$help" >&2
+  exit 1
+fi
+
+trap 'exit 1' 1 2 15
+
+# CC_FOR_BUILD -- compiler used by this script. Note that the use of a
+# compiler to aid in system detection is discouraged as it requires
+# temporary files to be created and, as you can see below, it is a
+# headache to deal with in a portable fashion.
+
+# Historically, `CC_FOR_BUILD' used to be named `HOST_CC'. We still
+# use `HOST_CC' if defined, but it is deprecated.
+
+# Portable tmp directory creation inspired by the Autoconf team.
+
+set_cc_for_build='
+trap "exitcode=\$?; (rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null) && exit \$exitcode" 0 ;
+trap "rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null; exit 1" 1 2 13 15 ;
+: ${TMPDIR=/tmp} ;
+ { tmp=`(umask 077 && mktemp -d -q "$TMPDIR/cgXXXXXX") 2>/dev/null` && test -n "$tmp" && test -d "$tmp" ; } ||
+ { test -n "$RANDOM" && tmp=$TMPDIR/cg$$-$RANDOM && (umask 077 && mkdir $tmp) ; } ||
+ { tmp=$TMPDIR/cg-$$ && (umask 077 && mkdir $tmp) && echo "Warning: creating insecure temp directory" >&2 ; } ||
+ { echo "$me: cannot create a temporary directory in $TMPDIR" >&2 ; exit 1 ; } ;
+dummy=$tmp/dummy ;
+tmpfiles="$dummy.c $dummy.o $dummy.rel $dummy" ;
+case $CC_FOR_BUILD,$HOST_CC,$CC in
+ ,,)    echo "int x;" > $dummy.c ;
+	for c in cc gcc c89 c99 ; do
+	  if ($c -c -o $dummy.o $dummy.c) >/dev/null 2>&1 ; then
+	     CC_FOR_BUILD="$c"; break ;
+	  fi ;
+	done ;
+	if test x"$CC_FOR_BUILD" = x ; then
+	  CC_FOR_BUILD=no_compiler_found ;
+	fi
+	;;
+ ,,*)   CC_FOR_BUILD=$CC ;;
+ ,*,*)  CC_FOR_BUILD=$HOST_CC ;;
+esac ;'
+
+# This is needed to find uname on a Pyramid OSx when run in the BSD universe.
+# (ghazi at noc.rutgers.edu 1994-08-24)
+if (test -f /.attbin/uname) >/dev/null 2>&1 ; then
+	PATH=$PATH:/.attbin ; export PATH
+fi
+
+UNAME_MACHINE=`(uname -m) 2>/dev/null` || UNAME_MACHINE=unknown
+UNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown
+UNAME_SYSTEM=`(uname -s) 2>/dev/null`  || UNAME_SYSTEM=unknown
+UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown
+
+# Note: order is significant - the case branches are not exclusive.
+
+case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
+    *:NetBSD:*:*)
+	# NetBSD (nbsd) targets should (where applicable) match one or
+	# more of the tupples: *-*-netbsdelf*, *-*-netbsdaout*,
+	# *-*-netbsdecoff* and *-*-netbsd*.  For targets that recently
+	# switched to ELF, *-*-netbsd* would select the old
+	# object file format.  This provides both forward
+	# compatibility and a consistent mechanism for selecting the
+	# object file format.
+	#
+	# Note: NetBSD doesn't particularly care about the vendor
+	# portion of the name.  We always set it to "unknown".
+	sysctl="sysctl -n hw.machine_arch"
+	UNAME_MACHINE_ARCH=`(/sbin/$sysctl 2>/dev/null || \
+	    /usr/sbin/$sysctl 2>/dev/null || echo unknown)`
+	case "${UNAME_MACHINE_ARCH}" in
+	    armeb) machine=armeb-unknown ;;
+	    arm*) machine=arm-unknown ;;
+	    sh3el) machine=shl-unknown ;;
+	    sh3eb) machine=sh-unknown ;;
+	    *) machine=${UNAME_MACHINE_ARCH}-unknown ;;
+	esac
+	# The Operating System including object format, if it has switched
+	# to ELF recently, or will in the future.
+	case "${UNAME_MACHINE_ARCH}" in
+	    arm*|i386|m68k|ns32k|sh3*|sparc|vax)
+		eval $set_cc_for_build
+		if echo __ELF__ | $CC_FOR_BUILD -E - 2>/dev/null \
+			| grep __ELF__ >/dev/null
+		then
+		    # Once all utilities can be ECOFF (netbsdecoff) or a.out (netbsdaout).
+		    # Return netbsd for either.  FIX?
+		    os=netbsd
+		else
+		    os=netbsdelf
+		fi
+		;;
+	    *)
+	        os=netbsd
+		;;
+	esac
+	# The OS release
+	# Debian GNU/NetBSD machines have a different userland, and
+	# thus, need a distinct triplet. However, they do not need
+	# kernel version information, so it can be replaced with a
+	# suitable tag, in the style of linux-gnu.
+	case "${UNAME_VERSION}" in
+	    Debian*)
+		release='-gnu'
+		;;
+	    *)
+		release=`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\./'`
+		;;
+	esac
+	# Since CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM:
+	# contains redundant information, the shorter form:
+	# CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM is used.
+	echo "${machine}-${os}${release}"
+	exit 0 ;;
+    amd64:OpenBSD:*:*)
+	echo x86_64-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    amiga:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    cats:OpenBSD:*:*)
+	echo arm-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    hp300:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    luna88k:OpenBSD:*:*)
+    	echo m88k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mac68k:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    macppc:OpenBSD:*:*)
+	echo powerpc-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mvme68k:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mvme88k:OpenBSD:*:*)
+	echo m88k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    mvmeppc:OpenBSD:*:*)
+	echo powerpc-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    sgi:OpenBSD:*:*)
+	echo mips64-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    sun3:OpenBSD:*:*)
+	echo m68k-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    *:OpenBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-openbsd${UNAME_RELEASE}
+	exit 0 ;;
+    *:ekkoBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE}
+	exit 0 ;;
+    macppc:MirBSD:*:*)
+	echo powerppc-unknown-mirbsd${UNAME_RELEASE}
+	exit 0 ;;
+    *:MirBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-mirbsd${UNAME_RELEASE}
+	exit 0 ;;
+    alpha:OSF1:*:*)
+	case $UNAME_RELEASE in
+	*4.0)
+		UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'`
+		;;
+	*5.*)
+	        UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'`
+		;;
+	esac
+	# According to Compaq, /usr/sbin/psrinfo has been available on
+	# OSF/1 and Tru64 systems produced since 1995.  I hope that
+	# covers most systems running today.  This code pipes the CPU
+	# types through head -n 1, so we only detect the type of CPU 0.
+	ALPHA_CPU_TYPE=`/usr/sbin/psrinfo -v | sed -n -e 's/^  The alpha \(.*\) processor.*$/\1/p' | head -n 1`
+	case "$ALPHA_CPU_TYPE" in
+	    "EV4 (21064)")
+		UNAME_MACHINE="alpha" ;;
+	    "EV4.5 (21064)")
+		UNAME_MACHINE="alpha" ;;
+	    "LCA4 (21066/21068)")
+		UNAME_MACHINE="alpha" ;;
+	    "EV5 (21164)")
+		UNAME_MACHINE="alphaev5" ;;
+	    "EV5.6 (21164A)")
+		UNAME_MACHINE="alphaev56" ;;
+	    "EV5.6 (21164PC)")
+		UNAME_MACHINE="alphapca56" ;;
+	    "EV5.7 (21164PC)")
+		UNAME_MACHINE="alphapca57" ;;
+	    "EV6 (21264)")
+		UNAME_MACHINE="alphaev6" ;;
+	    "EV6.7 (21264A)")
+		UNAME_MACHINE="alphaev67" ;;
+	    "EV6.8CB (21264C)")
+		UNAME_MACHINE="alphaev68" ;;
+	    "EV6.8AL (21264B)")
+		UNAME_MACHINE="alphaev68" ;;
+	    "EV6.8CX (21264D)")
+		UNAME_MACHINE="alphaev68" ;;
+	    "EV6.9A (21264/EV69A)")
+		UNAME_MACHINE="alphaev69" ;;
+	    "EV7 (21364)")
+		UNAME_MACHINE="alphaev7" ;;
+	    "EV7.9 (21364A)")
+		UNAME_MACHINE="alphaev79" ;;
+	esac
+	# A Pn.n version is a patched version.
+	# A Vn.n version is a released version.
+	# A Tn.n version is a released field test version.
+	# A Xn.n version is an unreleased experimental baselevel.
+	# 1.2 uses "1.2" for uname -r.
+	echo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[PVTX]//' | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`
+	exit 0 ;;
+    Alpha\ *:Windows_NT*:*)
+	# How do we know it's Interix rather than the generic POSIX subsystem?
+	# Should we change UNAME_MACHINE based on the output of uname instead
+	# of the specific Alpha model?
+	echo alpha-pc-interix
+	exit 0 ;;
+    21064:Windows_NT:50:3)
+	echo alpha-dec-winnt3.5
+	exit 0 ;;
+    Amiga*:UNIX_System_V:4.0:*)
+	echo m68k-unknown-sysv4
+	exit 0;;
+    *:[Aa]miga[Oo][Ss]:*:*)
+	echo ${UNAME_MACHINE}-unknown-amigaos
+	exit 0 ;;
+    *:[Mm]orph[Oo][Ss]:*:*)
+	echo ${UNAME_MACHINE}-unknown-morphos
+	exit 0 ;;
+    *:OS/390:*:*)
+	echo i370-ibm-openedition
+	exit 0 ;;
+    *:OS400:*:*)
+        echo powerpc-ibm-os400
+	exit 0 ;;
+    arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*)
+	echo arm-acorn-riscix${UNAME_RELEASE}
+	exit 0;;
+    SR2?01:HI-UX/MPP:*:* | SR8000:HI-UX/MPP:*:*)
+	echo hppa1.1-hitachi-hiuxmpp
+	exit 0;;
+    Pyramid*:OSx*:*:* | MIS*:OSx*:*:* | MIS*:SMP_DC-OSx*:*:*)
+	# akee at wpdis03.wpafb.af.mil (Earle F. Ake) contributed MIS and NILE.
+	if test "`(/bin/universe) 2>/dev/null`" = att ; then
+		echo pyramid-pyramid-sysv3
+	else
+		echo pyramid-pyramid-bsd
+	fi
+	exit 0 ;;
+    NILE*:*:*:dcosx)
+	echo pyramid-pyramid-svr4
+	exit 0 ;;
+    DRS?6000:unix:4.0:6*)
+	echo sparc-icl-nx6
+	exit 0 ;;
+    DRS?6000:UNIX_SV:4.2*:7*)
+	case `/usr/bin/uname -p` in
+	    sparc) echo sparc-icl-nx7 && exit 0 ;;
+	esac ;;
+    sun4H:SunOS:5.*:*)
+	echo sparc-hal-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*)
+	echo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    i86pc:SunOS:5.*:*)
+	echo i386-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    sun4*:SunOS:6*:*)
+	# According to config.sub, this is the proper way to canonicalize
+	# SunOS6.  Hard to guess exactly what SunOS6 will be like, but
+	# it's likely to be more like Solaris than SunOS4.
+	echo sparc-sun-solaris3`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    sun4*:SunOS:*:*)
+	case "`/usr/bin/arch -k`" in
+	    Series*|S4*)
+		UNAME_RELEASE=`uname -v`
+		;;
+	esac
+	# Japanese Language versions have a version number like `4.1.3-JL'.
+	echo sparc-sun-sunos`echo ${UNAME_RELEASE}|sed -e 's/-/_/'`
+	exit 0 ;;
+    sun3*:SunOS:*:*)
+	echo m68k-sun-sunos${UNAME_RELEASE}
+	exit 0 ;;
+    sun*:*:4.2BSD:*)
+	UNAME_RELEASE=`(sed 1q /etc/motd | awk '{print substr($5,1,3)}') 2>/dev/null`
+	test "x${UNAME_RELEASE}" = "x" && UNAME_RELEASE=3
+	case "`/bin/arch`" in
+	    sun3)
+		echo m68k-sun-sunos${UNAME_RELEASE}
+		;;
+	    sun4)
+		echo sparc-sun-sunos${UNAME_RELEASE}
+		;;
+	esac
+	exit 0 ;;
+    aushp:SunOS:*:*)
+	echo sparc-auspex-sunos${UNAME_RELEASE}
+	exit 0 ;;
+    # The situation for MiNT is a little confusing.  The machine name
+    # can be virtually everything (everything which is not
+    # "atarist" or "atariste" at least should have a processor
+    # > m68000).  The system name ranges from "MiNT" over "FreeMiNT"
+    # to the lowercase version "mint" (or "freemint").  Finally
+    # the system name "TOS" denotes a system which is actually not
+    # MiNT.  But MiNT is downward compatible to TOS, so this should
+    # be no problem.
+    atarist[e]:*MiNT:*:* | atarist[e]:*mint:*:* | atarist[e]:*TOS:*:*)
+        echo m68k-atari-mint${UNAME_RELEASE}
+	exit 0 ;;
+    atari*:*MiNT:*:* | atari*:*mint:*:* | atarist[e]:*TOS:*:*)
+	echo m68k-atari-mint${UNAME_RELEASE}
+        exit 0 ;;
+    *falcon*:*MiNT:*:* | *falcon*:*mint:*:* | *falcon*:*TOS:*:*)
+        echo m68k-atari-mint${UNAME_RELEASE}
+	exit 0 ;;
+    milan*:*MiNT:*:* | milan*:*mint:*:* | *milan*:*TOS:*:*)
+        echo m68k-milan-mint${UNAME_RELEASE}
+        exit 0 ;;
+    hades*:*MiNT:*:* | hades*:*mint:*:* | *hades*:*TOS:*:*)
+        echo m68k-hades-mint${UNAME_RELEASE}
+        exit 0 ;;
+    *:*MiNT:*:* | *:*mint:*:* | *:*TOS:*:*)
+        echo m68k-unknown-mint${UNAME_RELEASE}
+        exit 0 ;;
+    m68k:machten:*:*)
+	echo m68k-apple-machten${UNAME_RELEASE}
+	exit 0 ;;
+    powerpc:machten:*:*)
+	echo powerpc-apple-machten${UNAME_RELEASE}
+	exit 0 ;;
+    RISC*:Mach:*:*)
+	echo mips-dec-mach_bsd4.3
+	exit 0 ;;
+    RISC*:ULTRIX:*:*)
+	echo mips-dec-ultrix${UNAME_RELEASE}
+	exit 0 ;;
+    VAX*:ULTRIX*:*:*)
+	echo vax-dec-ultrix${UNAME_RELEASE}
+	exit 0 ;;
+    2020:CLIX:*:* | 2430:CLIX:*:*)
+	echo clipper-intergraph-clix${UNAME_RELEASE}
+	exit 0 ;;
+    mips:*:*:UMIPS | mips:*:*:RISCos)
+	eval $set_cc_for_build
+	sed 's/^	//' << EOF >$dummy.c
+#ifdef __cplusplus
+#include <stdio.h>  /* for printf() prototype */
+	int main (int argc, char *argv[]) {
+#else
+	int main (argc, argv) int argc; char *argv[]; {
+#endif
+	#if defined (host_mips) && defined (MIPSEB)
+	#if defined (SYSTYPE_SYSV)
+	  printf ("mips-mips-riscos%ssysv\n", argv[1]); exit (0);
+	#endif
+	#if defined (SYSTYPE_SVR4)
+	  printf ("mips-mips-riscos%ssvr4\n", argv[1]); exit (0);
+	#endif
+	#if defined (SYSTYPE_BSD43) || defined(SYSTYPE_BSD)
+	  printf ("mips-mips-riscos%sbsd\n", argv[1]); exit (0);
+	#endif
+	#endif
+	  exit (-1);
+	}
+EOF
+	$CC_FOR_BUILD -o $dummy $dummy.c \
+	  && $dummy `echo "${UNAME_RELEASE}" | sed -n 's/\([0-9]*\).*/\1/p'` \
+	  && exit 0
+	echo mips-mips-riscos${UNAME_RELEASE}
+	exit 0 ;;
+    Motorola:PowerMAX_OS:*:*)
+	echo powerpc-motorola-powermax
+	exit 0 ;;
+    Motorola:*:4.3:PL8-*)
+	echo powerpc-harris-powermax
+	exit 0 ;;
+    Night_Hawk:*:*:PowerMAX_OS | Synergy:PowerMAX_OS:*:*)
+	echo powerpc-harris-powermax
+	exit 0 ;;
+    Night_Hawk:Power_UNIX:*:*)
+	echo powerpc-harris-powerunix
+	exit 0 ;;
+    m88k:CX/UX:7*:*)
+	echo m88k-harris-cxux7
+	exit 0 ;;
+    m88k:*:4*:R4*)
+	echo m88k-motorola-sysv4
+	exit 0 ;;
+    m88k:*:3*:R3*)
+	echo m88k-motorola-sysv3
+	exit 0 ;;
+    AViiON:dgux:*:*)
+        # DG/UX returns AViiON for all architectures
+        UNAME_PROCESSOR=`/usr/bin/uname -p`
+	if [ $UNAME_PROCESSOR = mc88100 ] || [ $UNAME_PROCESSOR = mc88110 ]
+	then
+	    if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx ] || \
+	       [ ${TARGET_BINARY_INTERFACE}x = x ]
+	    then
+		echo m88k-dg-dgux${UNAME_RELEASE}
+	    else
+		echo m88k-dg-dguxbcs${UNAME_RELEASE}
+	    fi
+	else
+	    echo i586-dg-dgux${UNAME_RELEASE}
+	fi
+ 	exit 0 ;;
+    M88*:DolphinOS:*:*)	# DolphinOS (SVR3)
+	echo m88k-dolphin-sysv3
+	exit 0 ;;
+    M88*:*:R3*:*)
+	# Delta 88k system running SVR3
+	echo m88k-motorola-sysv3
+	exit 0 ;;
+    XD88*:*:*:*) # Tektronix XD88 system running UTekV (SVR3)
+	echo m88k-tektronix-sysv3
+	exit 0 ;;
+    Tek43[0-9][0-9]:UTek:*:*) # Tektronix 4300 system running UTek (BSD)
+	echo m68k-tektronix-bsd
+	exit 0 ;;
+    *:IRIX*:*:*)
+	echo mips-sgi-irix`echo ${UNAME_RELEASE}|sed -e 's/-/_/g'`
+	exit 0 ;;
+    ????????:AIX?:[12].1:2)   # AIX 2.2.1 or AIX 2.1.1 is RT/PC AIX.
+	echo romp-ibm-aix      # uname -m gives an 8 hex-code CPU id
+	exit 0 ;;              # Note that: echo "'`uname -s`'" gives 'AIX '
+    i*86:AIX:*:*)
+	echo i386-ibm-aix
+	exit 0 ;;
+    ia64:AIX:*:*)
+	if [ -x /usr/bin/oslevel ] ; then
+		IBM_REV=`/usr/bin/oslevel`
+	else
+		IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE}
+	fi
+	echo ${UNAME_MACHINE}-ibm-aix${IBM_REV}
+	exit 0 ;;
+    *:AIX:2:3)
+	if grep bos325 /usr/include/stdio.h >/dev/null 2>&1; then
+		eval $set_cc_for_build
+		sed 's/^		//' << EOF >$dummy.c
+		#include <sys/systemcfg.h>
+
+		main()
+			{
+			if (!__power_pc())
+				exit(1);
+			puts("powerpc-ibm-aix3.2.5");
+			exit(0);
+			}
+EOF
+		$CC_FOR_BUILD -o $dummy $dummy.c && $dummy && exit 0
+		echo rs6000-ibm-aix3.2.5
+	elif grep bos324 /usr/include/stdio.h >/dev/null 2>&1; then
+		echo rs6000-ibm-aix3.2.4
+	else
+		echo rs6000-ibm-aix3.2
+	fi
+	exit 0 ;;
+    *:AIX:*:[45])
+	IBM_CPU_ID=`/usr/sbin/lsdev -C -c processor -S available | sed 1q | awk '{ print $1 }'`
+	if /usr/sbin/lsattr -El ${IBM_CPU_ID} | grep ' POWER' >/dev/null 2>&1; then
+		IBM_ARCH=rs6000
+	else
+		IBM_ARCH=powerpc
+	fi
+	if [ -x /usr/bin/oslevel ] ; then
+		IBM_REV=`/usr/bin/oslevel`
+	else
+		IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE}
+	fi
+	echo ${IBM_ARCH}-ibm-aix${IBM_REV}
+	exit 0 ;;
+    *:AIX:*:*)
+	echo rs6000-ibm-aix
+	exit 0 ;;
+    ibmrt:4.4BSD:*|romp-ibm:BSD:*)
+	echo romp-ibm-bsd4.4
+	exit 0 ;;
+    ibmrt:*BSD:*|romp-ibm:BSD:*)            # covers RT/PC BSD and
+	echo romp-ibm-bsd${UNAME_RELEASE}   # 4.3 with uname added to
+	exit 0 ;;                           # report: romp-ibm BSD 4.3
+    *:BOSX:*:*)
+	echo rs6000-bull-bosx
+	exit 0 ;;
+    DPX/2?00:B.O.S.:*:*)
+	echo m68k-bull-sysv3
+	exit 0 ;;
+    9000/[34]??:4.3bsd:1.*:*)
+	echo m68k-hp-bsd
+	exit 0 ;;
+    hp300:4.4BSD:*:* | 9000/[34]??:4.3bsd:2.*:*)
+	echo m68k-hp-bsd4.4
+	exit 0 ;;
+    9000/[34678]??:HP-UX:*:*)
+	HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'`
+	case "${UNAME_MACHINE}" in
+	    9000/31? )            HP_ARCH=m68000 ;;
+	    9000/[34]?? )         HP_ARCH=m68k ;;
+	    9000/[678][0-9][0-9])
+		if [ -x /usr/bin/getconf ]; then
+		    sc_cpu_version=`/usr/bin/getconf SC_CPU_VERSION 2>/dev/null`
+                    sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null`
+                    case "${sc_cpu_version}" in
+                      523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0
+                      528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1
+                      532)                      # CPU_PA_RISC2_0
+                        case "${sc_kernel_bits}" in
+                          32) HP_ARCH="hppa2.0n" ;;
+                          64) HP_ARCH="hppa2.0w" ;;
+			  '') HP_ARCH="hppa2.0" ;;   # HP-UX 10.20
+                        esac ;;
+                    esac
+		fi
+		if [ "${HP_ARCH}" = "" ]; then
+		    eval $set_cc_for_build
+		    sed 's/^              //' << EOF >$dummy.c
+
+              #define _HPUX_SOURCE
+              #include <stdlib.h>
+              #include <unistd.h>
+
+              int main ()
+              {
+              #if defined(_SC_KERNEL_BITS)
+                  long bits = sysconf(_SC_KERNEL_BITS);
+              #endif
+                  long cpu  = sysconf (_SC_CPU_VERSION);
+
+                  switch (cpu)
+              	{
+              	case CPU_PA_RISC1_0: puts ("hppa1.0"); break;
+              	case CPU_PA_RISC1_1: puts ("hppa1.1"); break;
+              	case CPU_PA_RISC2_0:
+              #if defined(_SC_KERNEL_BITS)
+              	    switch (bits)
+              		{
+              		case 64: puts ("hppa2.0w"); break;
+              		case 32: puts ("hppa2.0n"); break;
+              		default: puts ("hppa2.0"); break;
+              		} break;
+              #else  /* !defined(_SC_KERNEL_BITS) */
+              	    puts ("hppa2.0"); break;
+              #endif
+              	default: puts ("hppa1.0"); break;
+              	}
+                  exit (0);
+              }
+EOF
+		    (CCOPTS= $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null) && HP_ARCH=`$dummy`
+		    test -z "$HP_ARCH" && HP_ARCH=hppa
+		fi ;;
+	esac
+	if [ ${HP_ARCH} = "hppa2.0w" ]
+	then
+	    # avoid double evaluation of $set_cc_for_build
+	    test -n "$CC_FOR_BUILD" || eval $set_cc_for_build
+	    if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E -) | grep __LP64__ >/dev/null
+	    then
+		HP_ARCH="hppa2.0w"
+	    else
+		HP_ARCH="hppa64"
+	    fi
+	fi
+	echo ${HP_ARCH}-hp-hpux${HPUX_REV}
+	exit 0 ;;
+    ia64:HP-UX:*:*)
+	HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'`
+	echo ia64-hp-hpux${HPUX_REV}
+	exit 0 ;;
+    3050*:HI-UX:*:*)
+	eval $set_cc_for_build
+	sed 's/^	//' << EOF >$dummy.c
+	#include <unistd.h>
+	int
+	main ()
+	{
+	  long cpu = sysconf (_SC_CPU_VERSION);
+	  /* The order matters, because CPU_IS_HP_MC68K erroneously returns
+	     true for CPU_PA_RISC1_0.  CPU_IS_PA_RISC returns correct
+	     results, however.  */
+	  if (CPU_IS_PA_RISC (cpu))
+	    {
+	      switch (cpu)
+		{
+		  case CPU_PA_RISC1_0: puts ("hppa1.0-hitachi-hiuxwe2"); break;
+		  case CPU_PA_RISC1_1: puts ("hppa1.1-hitachi-hiuxwe2"); break;
+		  case CPU_PA_RISC2_0: puts ("hppa2.0-hitachi-hiuxwe2"); break;
+		  default: puts ("hppa-hitachi-hiuxwe2"); break;
+		}
+	    }
+	  else if (CPU_IS_HP_MC68K (cpu))
+	    puts ("m68k-hitachi-hiuxwe2");
+	  else puts ("unknown-hitachi-hiuxwe2");
+	  exit (0);
+	}
+EOF
+	$CC_FOR_BUILD -o $dummy $dummy.c && $dummy && exit 0
+	echo unknown-hitachi-hiuxwe2
+	exit 0 ;;
+    9000/7??:4.3bsd:*:* | 9000/8?[79]:4.3bsd:*:* )
+	echo hppa1.1-hp-bsd
+	exit 0 ;;
+    9000/8??:4.3bsd:*:*)
+	echo hppa1.0-hp-bsd
+	exit 0 ;;
+    *9??*:MPE/iX:*:* | *3000*:MPE/iX:*:*)
+	echo hppa1.0-hp-mpeix
+	exit 0 ;;
+    hp7??:OSF1:*:* | hp8?[79]:OSF1:*:* )
+	echo hppa1.1-hp-osf
+	exit 0 ;;
+    hp8??:OSF1:*:*)
+	echo hppa1.0-hp-osf
+	exit 0 ;;
+    i*86:OSF1:*:*)
+	if [ -x /usr/sbin/sysversion ] ; then
+	    echo ${UNAME_MACHINE}-unknown-osf1mk
+	else
+	    echo ${UNAME_MACHINE}-unknown-osf1
+	fi
+	exit 0 ;;
+    parisc*:Lites*:*:*)
+	echo hppa1.1-hp-lites
+	exit 0 ;;
+    C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*)
+	echo c1-convex-bsd
+        exit 0 ;;
+    C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*)
+	if getsysinfo -f scalar_acc
+	then echo c32-convex-bsd
+	else echo c2-convex-bsd
+	fi
+        exit 0 ;;
+    C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*)
+	echo c34-convex-bsd
+        exit 0 ;;
+    C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*)
+	echo c38-convex-bsd
+        exit 0 ;;
+    C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*)
+	echo c4-convex-bsd
+        exit 0 ;;
+    CRAY*Y-MP:*:*:*)
+	echo ymp-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    CRAY*[A-Z]90:*:*:*)
+	echo ${UNAME_MACHINE}-cray-unicos${UNAME_RELEASE} \
+	| sed -e 's/CRAY.*\([A-Z]90\)/\1/' \
+	      -e y/ABCDEFGHIJKLMNOPQRSTUVWXYZ/abcdefghijklmnopqrstuvwxyz/ \
+	      -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    CRAY*TS:*:*:*)
+	echo t90-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    CRAY*T3E:*:*:*)
+	echo alphaev5-cray-unicosmk${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    CRAY*SV1:*:*:*)
+	echo sv1-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    *:UNICOS/mp:*:*)
+	echo craynv-cray-unicosmp${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
+	exit 0 ;;
+    F30[01]:UNIX_System_V:*:* | F700:UNIX_System_V:*:*)
+	FUJITSU_PROC=`uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`
+        FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
+        FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'`
+        echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
+        exit 0 ;;
+    5000:UNIX_System_V:4.*:*)
+        FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
+        FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'`
+        echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
+	exit 0 ;;
+    i*86:BSD/386:*:* | i*86:BSD/OS:*:* | *:Ascend\ Embedded/OS:*:*)
+	echo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE}
+	exit 0 ;;
+    sparc*:BSD/OS:*:*)
+	echo sparc-unknown-bsdi${UNAME_RELEASE}
+	exit 0 ;;
+    *:BSD/OS:*:*)
+	echo ${UNAME_MACHINE}-unknown-bsdi${UNAME_RELEASE}
+	exit 0 ;;
+    *:FreeBSD:*:*)
+	echo ${UNAME_MACHINE}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`
+	exit 0 ;;
+    i*:CYGWIN*:*)
+	echo ${UNAME_MACHINE}-pc-cygwin
+	exit 0 ;;
+    i*:MINGW*:*)
+	echo ${UNAME_MACHINE}-pc-mingw32
+	exit 0 ;;
+    i*:PW*:*)
+	echo ${UNAME_MACHINE}-pc-pw32
+	exit 0 ;;
+    x86:Interix*:[34]*)
+	echo i586-pc-interix${UNAME_RELEASE}|sed -e 's/\..*//'
+	exit 0 ;;
+    [345]86:Windows_95:* | [345]86:Windows_98:* | [345]86:Windows_NT:*)
+	echo i${UNAME_MACHINE}-pc-mks
+	exit 0 ;;
+    i*:Windows_NT*:* | Pentium*:Windows_NT*:*)
+	# How do we know it's Interix rather than the generic POSIX subsystem?
+	# It also conflicts with pre-2.0 versions of AT&T UWIN. Should we
+	# UNAME_MACHINE based on the output of uname instead of i386?
+	echo i586-pc-interix
+	exit 0 ;;
+    i*:UWIN*:*)
+	echo ${UNAME_MACHINE}-pc-uwin
+	exit 0 ;;
+    p*:CYGWIN*:*)
+	echo powerpcle-unknown-cygwin
+	exit 0 ;;
+    prep*:SunOS:5.*:*)
+	echo powerpcle-unknown-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
+	exit 0 ;;
+    *:GNU:*:*)
+	# the GNU system
+	echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-gnu`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'`
+	exit 0 ;;
+    *:GNU/*:*:*)
+	# other systems with GNU libc and userland
+	echo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-gnu
+	exit 0 ;;
+    i*86:Minix:*:*)
+	echo ${UNAME_MACHINE}-pc-minix
+	exit 0 ;;
+    arm*:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    cris:Linux:*:*)
+	echo cris-axis-linux-gnu
+	exit 0 ;;
+    crisv32:Linux:*:*)
+	echo crisv32-axis-linux-gnu
+	exit 0 ;;
+    frv:Linux:*:*)
+    	echo frv-unknown-linux-gnu
+	exit 0 ;;
+    ia64:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    m32r*:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    m68*:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    mips:Linux:*:*)
+	eval $set_cc_for_build
+	sed 's/^	//' << EOF >$dummy.c
+	#undef CPU
+	#undef mips
+	#undef mipsel
+	#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)
+	CPU=mipsel
+	#else
+	#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)
+	CPU=mips
+	#else
+	CPU=
+	#endif
+	#endif
+EOF
+	eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^CPU=`
+	test x"${CPU}" != x && echo "${CPU}-unknown-linux-gnu" && exit 0
+	;;
+    mips64:Linux:*:*)
+	eval $set_cc_for_build
+	sed 's/^	//' << EOF >$dummy.c
+	#undef CPU
+	#undef mips64
+	#undef mips64el
+	#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)
+	CPU=mips64el
+	#else
+	#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)
+	CPU=mips64
+	#else
+	CPU=
+	#endif
+	#endif
+EOF
+	eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^CPU=`
+	test x"${CPU}" != x && echo "${CPU}-unknown-linux-gnu" && exit 0
+	;;
+    ppc:Linux:*:*)
+	echo powerpc-unknown-linux-gnu
+	exit 0 ;;
+    ppc64:Linux:*:*)
+	echo powerpc64-unknown-linux-gnu
+	exit 0 ;;
+    alpha:Linux:*:*)
+	case `sed -n '/^cpu model/s/^.*: \(.*\)/\1/p' < /proc/cpuinfo` in
+	  EV5)   UNAME_MACHINE=alphaev5 ;;
+	  EV56)  UNAME_MACHINE=alphaev56 ;;
+	  PCA56) UNAME_MACHINE=alphapca56 ;;
+	  PCA57) UNAME_MACHINE=alphapca56 ;;
+	  EV6)   UNAME_MACHINE=alphaev6 ;;
+	  EV67)  UNAME_MACHINE=alphaev67 ;;
+	  EV68*) UNAME_MACHINE=alphaev68 ;;
+        esac
+	objdump --private-headers /bin/sh | grep ld.so.1 >/dev/null
+	if test "$?" = 0 ; then LIBC="libc1" ; else LIBC="" ; fi
+	echo ${UNAME_MACHINE}-unknown-linux-gnu${LIBC}
+	exit 0 ;;
+    parisc:Linux:*:* | hppa:Linux:*:*)
+	# Look for CPU level
+	case `grep '^cpu[^a-z]*:' /proc/cpuinfo 2>/dev/null | cut -d' ' -f2` in
+	  PA7*) echo hppa1.1-unknown-linux-gnu ;;
+	  PA8*) echo hppa2.0-unknown-linux-gnu ;;
+	  *)    echo hppa-unknown-linux-gnu ;;
+	esac
+	exit 0 ;;
+    parisc64:Linux:*:* | hppa64:Linux:*:*)
+	echo hppa64-unknown-linux-gnu
+	exit 0 ;;
+    s390:Linux:*:* | s390x:Linux:*:*)
+	echo ${UNAME_MACHINE}-ibm-linux
+	exit 0 ;;
+    sh64*:Linux:*:*)
+    	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    sh*:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    sparc:Linux:*:* | sparc64:Linux:*:*)
+	echo ${UNAME_MACHINE}-unknown-linux-gnu
+	exit 0 ;;
+    x86_64:Linux:*:*)
+	echo x86_64-unknown-linux-gnu
+	exit 0 ;;
+    i*86:Linux:*:*)
+	# The BFD linker knows what the default object file format is, so
+	# first see if it will tell us. cd to the root directory to prevent
+	# problems with other programs or directories called `ld' in the path.
+	# Set LC_ALL=C to ensure ld outputs messages in English.
+	ld_supported_targets=`cd /; LC_ALL=C ld --help 2>&1 \
+			 | sed -ne '/supported targets:/!d
+				    s/[ 	][ 	]*/ /g
+				    s/.*supported targets: *//
+				    s/ .*//
+				    p'`
+        case "$ld_supported_targets" in
+	  elf32-i386)
+		TENTATIVE="${UNAME_MACHINE}-pc-linux-gnu"
+		;;
+	  a.out-i386-linux)
+		echo "${UNAME_MACHINE}-pc-linux-gnuaout"
+		exit 0 ;;
+	  coff-i386)
+		echo "${UNAME_MACHINE}-pc-linux-gnucoff"
+		exit 0 ;;
+	  "")
+		# Either a pre-BFD a.out linker (linux-gnuoldld) or
+		# one that does not give us useful --help.
+		echo "${UNAME_MACHINE}-pc-linux-gnuoldld"
+		exit 0 ;;
+	esac
+	# Determine whether the default compiler is a.out or elf
+	eval $set_cc_for_build
+	sed 's/^	//' << EOF >$dummy.c
+	#include <features.h>
+	#ifdef __ELF__
+	# ifdef __GLIBC__
+	#  if __GLIBC__ >= 2
+	LIBC=gnu
+	#  else
+	LIBC=gnulibc1
+	#  endif
+	# else
+	LIBC=gnulibc1
+	# endif
+	#else
+	#ifdef __INTEL_COMPILER
+	LIBC=gnu
+	#else
+	LIBC=gnuaout
+	#endif
+	#endif
+	#ifdef __dietlibc__
+	LIBC=dietlibc
+	#endif
+EOF
+	eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep ^LIBC=`
+	test x"${LIBC}" != x && echo "${UNAME_MACHINE}-pc-linux-${LIBC}" && exit 0
+	test x"${TENTATIVE}" != x && echo "${TENTATIVE}" && exit 0
+	;;
+    i*86:DYNIX/ptx:4*:*)
+	# ptx 4.0 does uname -s correctly, with DYNIX/ptx in there.
+	# earlier versions are messed up and put the nodename in both
+	# sysname and nodename.
+	echo i386-sequent-sysv4
+	exit 0 ;;
+    i*86:UNIX_SV:4.2MP:2.*)
+        # Unixware is an offshoot of SVR4, but it has its own version
+        # number series starting with 2...
+        # I am not positive that other SVR4 systems won't match this,
+	# I just have to hope.  -- rms.
+        # Use sysv4.2uw... so that sysv4* matches it.
+	echo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION}
+	exit 0 ;;
+    i*86:OS/2:*:*)
+	# If we were able to find `uname', then EMX Unix compatibility
+	# is probably installed.
+	echo ${UNAME_MACHINE}-pc-os2-emx
+	exit 0 ;;
+    i*86:XTS-300:*:STOP)
+	echo ${UNAME_MACHINE}-unknown-stop
+	exit 0 ;;
+    i*86:atheos:*:*)
+	echo ${UNAME_MACHINE}-unknown-atheos
+	exit 0 ;;
+	i*86:syllable:*:*)
+	echo ${UNAME_MACHINE}-pc-syllable
+	exit 0 ;;
+    i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.0*:*)
+	echo i386-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    i*86:*DOS:*:*)
+	echo ${UNAME_MACHINE}-pc-msdosdjgpp
+	exit 0 ;;
+    i*86:*:4.*:* | i*86:SYSTEM_V:4.*:*)
+	UNAME_REL=`echo ${UNAME_RELEASE} | sed 's/\/MP$//'`
+	if grep Novell /usr/include/link.h >/dev/null 2>/dev/null; then
+		echo ${UNAME_MACHINE}-univel-sysv${UNAME_REL}
+	else
+		echo ${UNAME_MACHINE}-pc-sysv${UNAME_REL}
+	fi
+	exit 0 ;;
+    i*86:*:5:[78]*)
+	case `/bin/uname -X | grep "^Machine"` in
+	    *486*)	     UNAME_MACHINE=i486 ;;
+	    *Pentium)	     UNAME_MACHINE=i586 ;;
+	    *Pent*|*Celeron) UNAME_MACHINE=i686 ;;
+	esac
+	echo ${UNAME_MACHINE}-unknown-sysv${UNAME_RELEASE}${UNAME_SYSTEM}${UNAME_VERSION}
+	exit 0 ;;
+    i*86:*:3.2:*)
+	if test -f /usr/options/cb.name; then
+		UNAME_REL=`sed -n 's/.*Version //p' </usr/options/cb.name`
+		echo ${UNAME_MACHINE}-pc-isc$UNAME_REL
+	elif /bin/uname -X 2>/dev/null >/dev/null ; then
+		UNAME_REL=`(/bin/uname -X|grep Release|sed -e 's/.*= //')`
+		(/bin/uname -X|grep i80486 >/dev/null) && UNAME_MACHINE=i486
+		(/bin/uname -X|grep '^Machine.*Pentium' >/dev/null) \
+			&& UNAME_MACHINE=i586
+		(/bin/uname -X|grep '^Machine.*Pent *II' >/dev/null) \
+			&& UNAME_MACHINE=i686
+		(/bin/uname -X|grep '^Machine.*Pentium Pro' >/dev/null) \
+			&& UNAME_MACHINE=i686
+		echo ${UNAME_MACHINE}-pc-sco$UNAME_REL
+	else
+		echo ${UNAME_MACHINE}-pc-sysv32
+	fi
+	exit 0 ;;
+    pc:*:*:*)
+	# Left here for compatibility:
+        # uname -m prints for DJGPP always 'pc', but it prints nothing about
+        # the processor, so we play safe by assuming i386.
+	echo i386-pc-msdosdjgpp
+        exit 0 ;;
+    Intel:Mach:3*:*)
+	echo i386-pc-mach3
+	exit 0 ;;
+    paragon:*:*:*)
+	echo i860-intel-osf1
+	exit 0 ;;
+    i860:*:4.*:*) # i860-SVR4
+	if grep Stardent /usr/include/sys/uadmin.h >/dev/null 2>&1 ; then
+	  echo i860-stardent-sysv${UNAME_RELEASE} # Stardent Vistra i860-SVR4
+	else # Add other i860-SVR4 vendors below as they are discovered.
+	  echo i860-unknown-sysv${UNAME_RELEASE}  # Unknown i860-SVR4
+	fi
+	exit 0 ;;
+    mini*:CTIX:SYS*5:*)
+	# "miniframe"
+	echo m68010-convergent-sysv
+	exit 0 ;;
+    mc68k:UNIX:SYSTEM5:3.51m)
+	echo m68k-convergent-sysv
+	exit 0 ;;
+    M680?0:D-NIX:5.3:*)
+	echo m68k-diab-dnix
+	exit 0 ;;
+    M68*:*:R3V[5678]*:*)
+	test -r /sysV68 && echo 'm68k-motorola-sysv' && exit 0 ;;
+    3[345]??:*:4.0:3.0 | 3[34]??A:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 3[34]??/*:*:4.0:3.0 | 4400:*:4.0:3.0 | 4850:*:4.0:3.0 | SKA40:*:4.0:3.0 | SDS2:*:4.0:3.0 | SHG2:*:4.0:3.0 | S7501*:*:4.0:3.0)
+	OS_REL=''
+	test -r /etc/.relid \
+	&& OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid`
+	/bin/uname -p 2>/dev/null | grep 86 >/dev/null \
+	  && echo i486-ncr-sysv4.3${OS_REL} && exit 0
+	/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \
+	  && echo i586-ncr-sysv4.3${OS_REL} && exit 0 ;;
+    3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*)
+        /bin/uname -p 2>/dev/null | grep 86 >/dev/null \
+          && echo i486-ncr-sysv4 && exit 0 ;;
+    m68*:LynxOS:2.*:* | m68*:LynxOS:3.0*:*)
+	echo m68k-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    mc68030:UNIX_System_V:4.*:*)
+	echo m68k-atari-sysv4
+	exit 0 ;;
+    TSUNAMI:LynxOS:2.*:*)
+	echo sparc-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    rs6000:LynxOS:2.*:*)
+	echo rs6000-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.0*:*)
+	echo powerpc-unknown-lynxos${UNAME_RELEASE}
+	exit 0 ;;
+    SM[BE]S:UNIX_SV:*:*)
+	echo mips-dde-sysv${UNAME_RELEASE}
+	exit 0 ;;
+    RM*:ReliantUNIX-*:*:*)
+	echo mips-sni-sysv4
+	exit 0 ;;
+    RM*:SINIX-*:*:*)
+	echo mips-sni-sysv4
+	exit 0 ;;
+    *:SINIX-*:*:*)
+	if uname -p 2>/dev/null >/dev/null ; then
+		UNAME_MACHINE=`(uname -p) 2>/dev/null`
+		echo ${UNAME_MACHINE}-sni-sysv4
+	else
+		echo ns32k-sni-sysv
+	fi
+	exit 0 ;;
+    PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort
+                      # says <Richard.M.Bartel at ccMail.Census.GOV>
+        echo i586-unisys-sysv4
+        exit 0 ;;
+    *:UNIX_System_V:4*:FTX*)
+	# From Gerald Hewes <hewes at openmarket.com>.
+	# How about differentiating between stratus architectures? -djm
+	echo hppa1.1-stratus-sysv4
+	exit 0 ;;
+    *:*:*:FTX*)
+	# From seanf at swdc.stratus.com.
+	echo i860-stratus-sysv4
+	exit 0 ;;
+    *:VOS:*:*)
+	# From Paul.Green at stratus.com.
+	echo hppa1.1-stratus-vos
+	exit 0 ;;
+    mc68*:A/UX:*:*)
+	echo m68k-apple-aux${UNAME_RELEASE}
+	exit 0 ;;
+    news*:NEWS-OS:6*:*)
+	echo mips-sony-newsos6
+	exit 0 ;;
+    R[34]000:*System_V*:*:* | R4000:UNIX_SYSV:*:* | R*000:UNIX_SV:*:*)
+	if [ -d /usr/nec ]; then
+	        echo mips-nec-sysv${UNAME_RELEASE}
+	else
+	        echo mips-unknown-sysv${UNAME_RELEASE}
+	fi
+        exit 0 ;;
+    BeBox:BeOS:*:*)	# BeOS running on hardware made by Be, PPC only.
+	echo powerpc-be-beos
+	exit 0 ;;
+    BeMac:BeOS:*:*)	# BeOS running on Mac or Mac clone, PPC only.
+	echo powerpc-apple-beos
+	exit 0 ;;
+    BePC:BeOS:*:*)	# BeOS running on Intel PC compatible.
+	echo i586-pc-beos
+	exit 0 ;;
+    SX-4:SUPER-UX:*:*)
+	echo sx4-nec-superux${UNAME_RELEASE}
+	exit 0 ;;
+    SX-5:SUPER-UX:*:*)
+	echo sx5-nec-superux${UNAME_RELEASE}
+	exit 0 ;;
+    SX-6:SUPER-UX:*:*)
+	echo sx6-nec-superux${UNAME_RELEASE}
+	exit 0 ;;
+    Power*:Rhapsody:*:*)
+	echo powerpc-apple-rhapsody${UNAME_RELEASE}
+	exit 0 ;;
+    *:Rhapsody:*:*)
+	echo ${UNAME_MACHINE}-apple-rhapsody${UNAME_RELEASE}
+	exit 0 ;;
+    *:Darwin:*:*)
+	UNAME_PROCESSOR=`uname -p` || UNAME_PROCESSOR=unknown
+	case $UNAME_PROCESSOR in
+	    *86) UNAME_PROCESSOR=i686 ;;
+	    unknown) UNAME_PROCESSOR=powerpc ;;
+	esac
+	echo ${UNAME_PROCESSOR}-apple-darwin${UNAME_RELEASE}
+	exit 0 ;;
+    *:procnto*:*:* | *:QNX:[0123456789]*:*)
+	UNAME_PROCESSOR=`uname -p`
+	if test "$UNAME_PROCESSOR" = "x86"; then
+		UNAME_PROCESSOR=i386
+		UNAME_MACHINE=pc
+	fi
+	echo ${UNAME_PROCESSOR}-${UNAME_MACHINE}-nto-qnx${UNAME_RELEASE}
+	exit 0 ;;
+    *:QNX:*:4*)
+	echo i386-pc-qnx
+	exit 0 ;;
+    NSR-?:NONSTOP_KERNEL:*:*)
+	echo nsr-tandem-nsk${UNAME_RELEASE}
+	exit 0 ;;
+    *:NonStop-UX:*:*)
+	echo mips-compaq-nonstopux
+	exit 0 ;;
+    BS2000:POSIX*:*:*)
+	echo bs2000-siemens-sysv
+	exit 0 ;;
+    DS/*:UNIX_System_V:*:*)
+	echo ${UNAME_MACHINE}-${UNAME_SYSTEM}-${UNAME_RELEASE}
+	exit 0 ;;
+    *:Plan9:*:*)
+	# "uname -m" is not consistent, so use $cputype instead. 386
+	# is converted to i386 for consistency with other x86
+	# operating systems.
+	if test "$cputype" = "386"; then
+	    UNAME_MACHINE=i386
+	else
+	    UNAME_MACHINE="$cputype"
+	fi
+	echo ${UNAME_MACHINE}-unknown-plan9
+	exit 0 ;;
+    *:TOPS-10:*:*)
+	echo pdp10-unknown-tops10
+	exit 0 ;;
+    *:TENEX:*:*)
+	echo pdp10-unknown-tenex
+	exit 0 ;;
+    KS10:TOPS-20:*:* | KL10:TOPS-20:*:* | TYPE4:TOPS-20:*:*)
+	echo pdp10-dec-tops20
+	exit 0 ;;
+    XKL-1:TOPS-20:*:* | TYPE5:TOPS-20:*:*)
+	echo pdp10-xkl-tops20
+	exit 0 ;;
+    *:TOPS-20:*:*)
+	echo pdp10-unknown-tops20
+	exit 0 ;;
+    *:ITS:*:*)
+	echo pdp10-unknown-its
+	exit 0 ;;
+    SEI:*:*:SEIUX)
+        echo mips-sei-seiux${UNAME_RELEASE}
+	exit 0 ;;
+    *:DragonFly:*:*)
+	echo ${UNAME_MACHINE}-unknown-dragonfly`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`
+	exit 0 ;;
+    *:*VMS:*:*)
+    	UNAME_MACHINE=`(uname -p) 2>/dev/null`
+	case "${UNAME_MACHINE}" in
+	    A*) echo alpha-dec-vms && exit 0 ;;
+	    I*) echo ia64-dec-vms && exit 0 ;;
+	    V*) echo vax-dec-vms && exit 0 ;;
+	esac
+esac
+
+#echo '(No uname command or uname output not recognized.)' 1>&2
+#echo "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" 1>&2
+
+eval $set_cc_for_build
+cat >$dummy.c <<EOF
+#ifdef _SEQUENT_
+# include <sys/types.h>
+# include <sys/utsname.h>
+#endif
+main ()
+{
+#if defined (sony)
+#if defined (MIPSEB)
+  /* BFD wants "bsd" instead of "newsos".  Perhaps BFD should be changed,
+     I don't know....  */
+  printf ("mips-sony-bsd\n"); exit (0);
+#else
+#include <sys/param.h>
+  printf ("m68k-sony-newsos%s\n",
+#ifdef NEWSOS4
+          "4"
+#else
+	  ""
+#endif
+         ); exit (0);
+#endif
+#endif
+
+#if defined (__arm) && defined (__acorn) && defined (__unix)
+  printf ("arm-acorn-riscix"); exit (0);
+#endif
+
+#if defined (hp300) && !defined (hpux)
+  printf ("m68k-hp-bsd\n"); exit (0);
+#endif
+
+#if defined (NeXT)
+#if !defined (__ARCHITECTURE__)
+#define __ARCHITECTURE__ "m68k"
+#endif
+  int version;
+  version=`(hostinfo | sed -n 's/.*NeXT Mach \([0-9]*\).*/\1/p') 2>/dev/null`;
+  if (version < 4)
+    printf ("%s-next-nextstep%d\n", __ARCHITECTURE__, version);
+  else
+    printf ("%s-next-openstep%d\n", __ARCHITECTURE__, version);
+  exit (0);
+#endif
+
+#if defined (MULTIMAX) || defined (n16)
+#if defined (UMAXV)
+  printf ("ns32k-encore-sysv\n"); exit (0);
+#else
+#if defined (CMU)
+  printf ("ns32k-encore-mach\n"); exit (0);
+#else
+  printf ("ns32k-encore-bsd\n"); exit (0);
+#endif
+#endif
+#endif
+
+#if defined (__386BSD__)
+  printf ("i386-pc-bsd\n"); exit (0);
+#endif
+
+#if defined (sequent)
+#if defined (i386)
+  printf ("i386-sequent-dynix\n"); exit (0);
+#endif
+#if defined (ns32000)
+  printf ("ns32k-sequent-dynix\n"); exit (0);
+#endif
+#endif
+
+#if defined (_SEQUENT_)
+    struct utsname un;
+
+    uname(&un);
+
+    if (strncmp(un.version, "V2", 2) == 0) {
+	printf ("i386-sequent-ptx2\n"); exit (0);
+    }
+    if (strncmp(un.version, "V1", 2) == 0) { /* XXX is V1 correct? */
+	printf ("i386-sequent-ptx1\n"); exit (0);
+    }
+    printf ("i386-sequent-ptx\n"); exit (0);
+
+#endif
+
+#if defined (vax)
+# if !defined (ultrix)
+#  include <sys/param.h>
+#  if defined (BSD)
+#   if BSD == 43
+      printf ("vax-dec-bsd4.3\n"); exit (0);
+#   else
+#    if BSD == 199006
+      printf ("vax-dec-bsd4.3reno\n"); exit (0);
+#    else
+      printf ("vax-dec-bsd\n"); exit (0);
+#    endif
+#   endif
+#  else
+    printf ("vax-dec-bsd\n"); exit (0);
+#  endif
+# else
+    printf ("vax-dec-ultrix\n"); exit (0);
+# endif
+#endif
+
+#if defined (alliant) && defined (i860)
+  printf ("i860-alliant-bsd\n"); exit (0);
+#endif
+
+  exit (1);
+}
+EOF
+
+$CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null && $dummy && exit 0
+
+# Apollos put the system type in the environment.
+
+test -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit 0; }
+
+# Convex versions that predate uname can use getsysinfo(1)
+
+if [ -x /usr/convex/getsysinfo ]
+then
+    case `getsysinfo -f cpu_type` in
+    c1*)
+	echo c1-convex-bsd
+	exit 0 ;;
+    c2*)
+	if getsysinfo -f scalar_acc
+	then echo c32-convex-bsd
+	else echo c2-convex-bsd
+	fi
+	exit 0 ;;
+    c34*)
+	echo c34-convex-bsd
+	exit 0 ;;
+    c38*)
+	echo c38-convex-bsd
+	exit 0 ;;
+    c4*)
+	echo c4-convex-bsd
+	exit 0 ;;
+    esac
+fi
+
+cat >&2 <<EOF
+$0: unable to guess system type
+
+This script, last modified $timestamp, has failed to recognize
+the operating system you are using. It is advised that you
+download the most up to date version of the config scripts from
+
+    ftp://ftp.gnu.org/pub/gnu/config/
+
+If the version you run ($0) is already up to date, please
+send the following data and any information you think might be
+pertinent to <config-patches at gnu.org> in order to provide the needed
+information to handle your system.
+
+config.guess timestamp = $timestamp
+
+uname -m = `(uname -m) 2>/dev/null || echo unknown`
+uname -r = `(uname -r) 2>/dev/null || echo unknown`
+uname -s = `(uname -s) 2>/dev/null || echo unknown`
+uname -v = `(uname -v) 2>/dev/null || echo unknown`
+
+/usr/bin/uname -p = `(/usr/bin/uname -p) 2>/dev/null`
+/bin/uname -X     = `(/bin/uname -X) 2>/dev/null`
+
+hostinfo               = `(hostinfo) 2>/dev/null`
+/bin/universe          = `(/bin/universe) 2>/dev/null`
+/usr/bin/arch -k       = `(/usr/bin/arch -k) 2>/dev/null`
+/bin/arch              = `(/bin/arch) 2>/dev/null`
+/usr/bin/oslevel       = `(/usr/bin/oslevel) 2>/dev/null`
+/usr/convex/getsysinfo = `(/usr/convex/getsysinfo) 2>/dev/null`
+
+UNAME_MACHINE = ${UNAME_MACHINE}
+UNAME_RELEASE = ${UNAME_RELEASE}
+UNAME_SYSTEM  = ${UNAME_SYSTEM}
+UNAME_VERSION = ${UNAME_VERSION}
+EOF
+
+exit 1
+
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "timestamp='"
+# time-stamp-format: "%:y-%02m-%02d"
+# time-stamp-end: "'"
+# End:

Propchange: klee/trunk/autoconf/config.guess

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/autoconf/config.sub
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/autoconf/config.sub?rev=72205&view=auto

==============================================================================
--- klee/trunk/autoconf/config.sub (added)
+++ klee/trunk/autoconf/config.sub Wed May 20 23:36:41 2009
@@ -0,0 +1,1555 @@
+#! /bin/sh
+# Configuration validation subroutine script.
+#   Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
+#   2000, 2001, 2002, 2003, 2004 Free Software Foundation, Inc.
+
+timestamp='2004-08-29'
+
+# This file is (in principle) common to ALL GNU software.
+# The presence of a machine in this file suggests that SOME GNU software
+# can handle that machine.  It does not imply ALL GNU software can.
+#
+# This file is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 2 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software
+# Foundation, Inc., 59 Temple Place - Suite 330,
+# Boston, MA 02111-1307, USA.
+
+# As a special exception to the GNU General Public License, if you
+# distribute this file as part of a program that contains a
+# configuration script generated by Autoconf, you may include it under
+# the same distribution terms that you use for the rest of that program.
+
+# Please send patches to <config-patches at gnu.org>.  Submit a context
+# diff and a properly formatted ChangeLog entry.
+#
+# Configuration subroutine to validate and canonicalize a configuration type.
+# Supply the specified configuration type as an argument.
+# If it is invalid, we print an error message on stderr and exit with code 1.
+# Otherwise, we print the canonical config type on stdout and succeed.
+
+# This file is supposed to be the same for all GNU packages
+# and recognize all the CPU types, system types and aliases
+# that are meaningful with *any* GNU software.
+# Each package is responsible for reporting which valid configurations
+# it does not support.  The user should be able to distinguish
+# a failure to support a valid configuration from a meaningless
+# configuration.
+
+# The goal of this file is to map all the various variations of a given
+# machine specification into a single specification in the form:
+#	CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM
+# or in some cases, the newer four-part form:
+#	CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM
+# It is wrong to echo any other type of specification.
+
+me=`echo "$0" | sed -e 's,.*/,,'`
+
+usage="\
+Usage: $0 [OPTION] CPU-MFR-OPSYS
+       $0 [OPTION] ALIAS
+
+Canonicalize a configuration name.
+
+Operation modes:
+  -h, --help         print this help, then exit
+  -t, --time-stamp   print date of last modification, then exit
+  -v, --version      print version number, then exit
+
+Report bugs and patches to <config-patches at gnu.org>."
+
+version="\
+GNU config.sub ($timestamp)
+
+Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004
+Free Software Foundation, Inc.
+
+This is free software; see the source for copying conditions.  There is NO
+warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE."
+
+help="
+Try \`$me --help' for more information."
+
+# Parse command line
+while test $# -gt 0 ; do
+  case $1 in
+    --time-stamp | --time* | -t )
+       echo "$timestamp" ; exit 0 ;;
+    --version | -v )
+       echo "$version" ; exit 0 ;;
+    --help | --h* | -h )
+       echo "$usage"; exit 0 ;;
+    -- )     # Stop option processing
+       shift; break ;;
+    - )	# Use stdin as input.
+       break ;;
+    -* )
+       echo "$me: invalid option $1$help"
+       exit 1 ;;
+
+    *local*)
+       # First pass through any local machine types.
+       echo $1
+       exit 0;;
+
+    * )
+       break ;;
+  esac
+done
+
+case $# in
+ 0) echo "$me: missing argument$help" >&2
+    exit 1;;
+ 1) ;;
+ *) echo "$me: too many arguments$help" >&2
+    exit 1;;
+esac
+
+# Separate what the user gave into CPU-COMPANY and OS or KERNEL-OS (if any).
+# Here we must recognize all the valid KERNEL-OS combinations.
+maybe_os=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\2/'`
+case $maybe_os in
+  nto-qnx* | linux-gnu* | linux-dietlibc | linux-uclibc* | uclinux-uclibc* | uclinux-gnu* | \
+  kfreebsd*-gnu* | knetbsd*-gnu* | netbsd*-gnu* | storm-chaos* | os2-emx* | rtmk-nova*)
+    os=-$maybe_os
+    basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'`
+    ;;
+  *)
+    basic_machine=`echo $1 | sed 's/-[^-]*$//'`
+    if [ $basic_machine != $1 ]
+    then os=`echo $1 | sed 's/.*-/-/'`
+    else os=; fi
+    ;;
+esac
+
+### Let's recognize common machines as not being operating systems so
+### that things like config.sub decstation-3100 work.  We also
+### recognize some manufacturers as not being operating systems, so we
+### can provide default operating systems below.
+case $os in
+	-sun*os*)
+		# Prevent following clause from handling this invalid input.
+		;;
+	-dec* | -mips* | -sequent* | -encore* | -pc532* | -sgi* | -sony* | \
+	-att* | -7300* | -3300* | -delta* | -motorola* | -sun[234]* | \
+	-unicom* | -ibm* | -next | -hp | -isi* | -apollo | -altos* | \
+	-convergent* | -ncr* | -news | -32* | -3600* | -3100* | -hitachi* |\
+	-c[123]* | -convex* | -sun | -crds | -omron* | -dg | -ultra | -tti* | \
+	-harris | -dolphin | -highlevel | -gould | -cbm | -ns | -masscomp | \
+	-apple | -axis | -knuth | -cray)
+		os=
+		basic_machine=$1
+		;;
+	-sim | -cisco | -oki | -wec | -winbond)
+		os=
+		basic_machine=$1
+		;;
+	-scout)
+		;;
+	-wrs)
+		os=-vxworks
+		basic_machine=$1
+		;;
+	-chorusos*)
+		os=-chorusos
+		basic_machine=$1
+		;;
+ 	-chorusrdb)
+ 		os=-chorusrdb
+		basic_machine=$1
+ 		;;
+	-hiux*)
+		os=-hiuxwe2
+		;;
+	-sco5)
+		os=-sco3.2v5
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco4)
+		os=-sco3.2v4
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco3.2.[4-9]*)
+		os=`echo $os | sed -e 's/sco3.2./sco3.2v/'`
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco3.2v[4-9]*)
+		# Don't forget version if it is 3.2v4 or newer.
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-sco*)
+		os=-sco3.2v2
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-udk*)
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-isc)
+		os=-isc2.2
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-clix*)
+		basic_machine=clipper-intergraph
+		;;
+	-isc*)
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'`
+		;;
+	-lynx*)
+		os=-lynxos
+		;;
+	-ptx*)
+		basic_machine=`echo $1 | sed -e 's/86-.*/86-sequent/'`
+		;;
+	-windowsnt*)
+		os=`echo $os | sed -e 's/windowsnt/winnt/'`
+		;;
+	-psos*)
+		os=-psos
+		;;
+	-mint | -mint[0-9]*)
+		basic_machine=m68k-atari
+		os=-mint
+		;;
+esac
+
+# Decode aliases for certain CPU-COMPANY combinations.
+case $basic_machine in
+	# Recognize the basic CPU types without company name.
+	# Some are omitted here because they have special meanings below.
+	1750a | 580 \
+	| a29k \
+	| alpha | alphaev[4-8] | alphaev56 | alphaev6[78] | alphapca5[67] \
+	| alpha64 | alpha64ev[4-8] | alpha64ev56 | alpha64ev6[78] | alpha64pca5[67] \
+	| am33_2.0 \
+	| arc | arm | arm[bl]e | arme[lb] | armv[2345] | armv[345][lb] | avr \
+	| c4x | clipper \
+	| d10v | d30v | dlx | dsp16xx \
+	| fr30 | frv \
+	| h8300 | h8500 | hppa | hppa1.[01] | hppa2.0 | hppa2.0[nw] | hppa64 \
+	| i370 | i860 | i960 | ia64 \
+	| ip2k | iq2000 \
+	| m32r | m32rle | m68000 | m68k | m88k | mcore \
+	| mips | mipsbe | mipseb | mipsel | mipsle \
+	| mips16 \
+	| mips64 | mips64el \
+	| mips64vr | mips64vrel \
+	| mips64orion | mips64orionel \
+	| mips64vr4100 | mips64vr4100el \
+	| mips64vr4300 | mips64vr4300el \
+	| mips64vr5000 | mips64vr5000el \
+	| mipsisa32 | mipsisa32el \
+	| mipsisa32r2 | mipsisa32r2el \
+	| mipsisa64 | mipsisa64el \
+	| mipsisa64r2 | mipsisa64r2el \
+	| mipsisa64sb1 | mipsisa64sb1el \
+	| mipsisa64sr71k | mipsisa64sr71kel \
+	| mipstx39 | mipstx39el \
+	| mn10200 | mn10300 \
+	| msp430 \
+	| ns16k | ns32k \
+	| openrisc | or32 \
+	| pdp10 | pdp11 | pj | pjl \
+	| powerpc | powerpc64 | powerpc64le | powerpcle | ppcbe \
+	| pyramid \
+	| sh | sh[1234] | sh[23]e | sh[34]eb | shbe | shle | sh[1234]le | sh3ele \
+	| sh64 | sh64le \
+	| sparc | sparc64 | sparc86x | sparclet | sparclite | sparcv8 | sparcv9 | sparcv9b \
+	| strongarm \
+	| tahoe | thumb | tic4x | tic80 | tron \
+	| v850 | v850e \
+	| we32k \
+	| x86 | xscale | xstormy16 | xtensa \
+	| z8k)
+		basic_machine=$basic_machine-unknown
+		;;
+	m6811 | m68hc11 | m6812 | m68hc12)
+		# Motorola 68HC11/12.
+		basic_machine=$basic_machine-unknown
+		os=-none
+		;;
+	m88110 | m680[12346]0 | m683?2 | m68360 | m5200 | v70 | w65 | z8k)
+		;;
+
+	# We use `pc' rather than `unknown'
+	# because (1) that's what they normally are, and
+	# (2) the word "unknown" tends to confuse beginning users.
+	i*86 | x86_64)
+	  basic_machine=$basic_machine-pc
+	  ;;
+	# Object if more than one company name word.
+	*-*-*)
+		echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2
+		exit 1
+		;;
+	# Recognize the basic CPU types with company name.
+	580-* \
+	| a29k-* \
+	| alpha-* | alphaev[4-8]-* | alphaev56-* | alphaev6[78]-* \
+	| alpha64-* | alpha64ev[4-8]-* | alpha64ev56-* | alpha64ev6[78]-* \
+	| alphapca5[67]-* | alpha64pca5[67]-* | arc-* \
+	| arm-*  | armbe-* | armle-* | armeb-* | armv*-* \
+	| avr-* \
+	| bs2000-* \
+	| c[123]* | c30-* | [cjt]90-* | c4x-* | c54x-* | c55x-* | c6x-* \
+	| clipper-* | craynv-* | cydra-* \
+	| d10v-* | d30v-* | dlx-* \
+	| elxsi-* \
+	| f30[01]-* | f700-* | fr30-* | frv-* | fx80-* \
+	| h8300-* | h8500-* \
+	| hppa-* | hppa1.[01]-* | hppa2.0-* | hppa2.0[nw]-* | hppa64-* \
+	| i*86-* | i860-* | i960-* | ia64-* \
+	| ip2k-* | iq2000-* \
+	| m32r-* | m32rle-* \
+	| m68000-* | m680[012346]0-* | m68360-* | m683?2-* | m68k-* \
+	| m88110-* | m88k-* | mcore-* \
+	| mips-* | mipsbe-* | mipseb-* | mipsel-* | mipsle-* \
+	| mips16-* \
+	| mips64-* | mips64el-* \
+	| mips64vr-* | mips64vrel-* \
+	| mips64orion-* | mips64orionel-* \
+	| mips64vr4100-* | mips64vr4100el-* \
+	| mips64vr4300-* | mips64vr4300el-* \
+	| mips64vr5000-* | mips64vr5000el-* \
+	| mipsisa32-* | mipsisa32el-* \
+	| mipsisa32r2-* | mipsisa32r2el-* \
+	| mipsisa64-* | mipsisa64el-* \
+	| mipsisa64r2-* | mipsisa64r2el-* \
+	| mipsisa64sb1-* | mipsisa64sb1el-* \
+	| mipsisa64sr71k-* | mipsisa64sr71kel-* \
+	| mipstx39-* | mipstx39el-* \
+	| mmix-* \
+	| msp430-* \
+	| none-* | np1-* | ns16k-* | ns32k-* \
+	| orion-* \
+	| pdp10-* | pdp11-* | pj-* | pjl-* | pn-* | power-* \
+	| powerpc-* | powerpc64-* | powerpc64le-* | powerpcle-* | ppcbe-* \
+	| pyramid-* \
+	| romp-* | rs6000-* \
+	| sh-* | sh[1234]-* | sh[23]e-* | sh[34]eb-* | shbe-* \
+	| shle-* | sh[1234]le-* | sh3ele-* | sh64-* | sh64le-* \
+	| sparc-* | sparc64-* | sparc86x-* | sparclet-* | sparclite-* \
+	| sparcv8-* | sparcv9-* | sparcv9b-* | strongarm-* | sv1-* | sx?-* \
+	| tahoe-* | thumb-* \
+	| tic30-* | tic4x-* | tic54x-* | tic55x-* | tic6x-* | tic80-* \
+	| tron-* \
+	| v850-* | v850e-* | vax-* \
+	| we32k-* \
+	| x86-* | x86_64-* | xps100-* | xscale-* | xstormy16-* \
+	| xtensa-* \
+	| ymp-* \
+	| z8k-*)
+		;;
+	# Recognize the various machine names and aliases which stand
+	# for a CPU type and a company and sometimes even an OS.
+	386bsd)
+		basic_machine=i386-unknown
+		os=-bsd
+		;;
+	3b1 | 7300 | 7300-att | att-7300 | pc7300 | safari | unixpc)
+		basic_machine=m68000-att
+		;;
+	3b*)
+		basic_machine=we32k-att
+		;;
+	a29khif)
+		basic_machine=a29k-amd
+		os=-udi
+		;;
+    	abacus)
+		basic_machine=abacus-unknown
+		;;
+	adobe68k)
+		basic_machine=m68010-adobe
+		os=-scout
+		;;
+	alliant | fx80)
+		basic_machine=fx80-alliant
+		;;
+	altos | altos3068)
+		basic_machine=m68k-altos
+		;;
+	am29k)
+		basic_machine=a29k-none
+		os=-bsd
+		;;
+	amd64)
+		basic_machine=x86_64-pc
+		;;
+	amd64-*)
+		basic_machine=x86_64-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	amdahl)
+		basic_machine=580-amdahl
+		os=-sysv
+		;;
+	amiga | amiga-*)
+		basic_machine=m68k-unknown
+		;;
+	amigaos | amigados)
+		basic_machine=m68k-unknown
+		os=-amigaos
+		;;
+	amigaunix | amix)
+		basic_machine=m68k-unknown
+		os=-sysv4
+		;;
+	apollo68)
+		basic_machine=m68k-apollo
+		os=-sysv
+		;;
+	apollo68bsd)
+		basic_machine=m68k-apollo
+		os=-bsd
+		;;
+	aux)
+		basic_machine=m68k-apple
+		os=-aux
+		;;
+	balance)
+		basic_machine=ns32k-sequent
+		os=-dynix
+		;;
+	c90)
+		basic_machine=c90-cray
+		os=-unicos
+		;;
+	convex-c1)
+		basic_machine=c1-convex
+		os=-bsd
+		;;
+	convex-c2)
+		basic_machine=c2-convex
+		os=-bsd
+		;;
+	convex-c32)
+		basic_machine=c32-convex
+		os=-bsd
+		;;
+	convex-c34)
+		basic_machine=c34-convex
+		os=-bsd
+		;;
+	convex-c38)
+		basic_machine=c38-convex
+		os=-bsd
+		;;
+	cray | j90)
+		basic_machine=j90-cray
+		os=-unicos
+		;;
+	craynv)
+		basic_machine=craynv-cray
+		os=-unicosmp
+		;;
+	cr16c)
+		basic_machine=cr16c-unknown
+		os=-elf
+		;;
+	crds | unos)
+		basic_machine=m68k-crds
+		;;
+	crisv32 | crisv32-* | etraxfs*)
+		basic_machine=crisv32-axis
+		;;
+	cris | cris-* | etrax*)
+		basic_machine=cris-axis
+		;;
+	crx)
+		basic_machine=crx-unknown
+		os=-elf
+		;;
+	da30 | da30-*)
+		basic_machine=m68k-da30
+		;;
+	decstation | decstation-3100 | pmax | pmax-* | pmin | dec3100 | decstatn)
+		basic_machine=mips-dec
+		;;
+	decsystem10* | dec10*)
+		basic_machine=pdp10-dec
+		os=-tops10
+		;;
+	decsystem20* | dec20*)
+		basic_machine=pdp10-dec
+		os=-tops20
+		;;
+	delta | 3300 | motorola-3300 | motorola-delta \
+	      | 3300-motorola | delta-motorola)
+		basic_machine=m68k-motorola
+		;;
+	delta88)
+		basic_machine=m88k-motorola
+		os=-sysv3
+		;;
+	dpx20 | dpx20-*)
+		basic_machine=rs6000-bull
+		os=-bosx
+		;;
+	dpx2* | dpx2*-bull)
+		basic_machine=m68k-bull
+		os=-sysv3
+		;;
+	ebmon29k)
+		basic_machine=a29k-amd
+		os=-ebmon
+		;;
+	elxsi)
+		basic_machine=elxsi-elxsi
+		os=-bsd
+		;;
+	encore | umax | mmax)
+		basic_machine=ns32k-encore
+		;;
+	es1800 | OSE68k | ose68k | ose | OSE)
+		basic_machine=m68k-ericsson
+		os=-ose
+		;;
+	fx2800)
+		basic_machine=i860-alliant
+		;;
+	genix)
+		basic_machine=ns32k-ns
+		;;
+	gmicro)
+		basic_machine=tron-gmicro
+		os=-sysv
+		;;
+	go32)
+		basic_machine=i386-pc
+		os=-go32
+		;;
+	h3050r* | hiux*)
+		basic_machine=hppa1.1-hitachi
+		os=-hiuxwe2
+		;;
+	h8300hms)
+		basic_machine=h8300-hitachi
+		os=-hms
+		;;
+	h8300xray)
+		basic_machine=h8300-hitachi
+		os=-xray
+		;;
+	h8500hms)
+		basic_machine=h8500-hitachi
+		os=-hms
+		;;
+	harris)
+		basic_machine=m88k-harris
+		os=-sysv3
+		;;
+	hp300-*)
+		basic_machine=m68k-hp
+		;;
+	hp300bsd)
+		basic_machine=m68k-hp
+		os=-bsd
+		;;
+	hp300hpux)
+		basic_machine=m68k-hp
+		os=-hpux
+		;;
+	hp3k9[0-9][0-9] | hp9[0-9][0-9])
+		basic_machine=hppa1.0-hp
+		;;
+	hp9k2[0-9][0-9] | hp9k31[0-9])
+		basic_machine=m68000-hp
+		;;
+	hp9k3[2-9][0-9])
+		basic_machine=m68k-hp
+		;;
+	hp9k6[0-9][0-9] | hp6[0-9][0-9])
+		basic_machine=hppa1.0-hp
+		;;
+	hp9k7[0-79][0-9] | hp7[0-79][0-9])
+		basic_machine=hppa1.1-hp
+		;;
+	hp9k78[0-9] | hp78[0-9])
+		# FIXME: really hppa2.0-hp
+		basic_machine=hppa1.1-hp
+		;;
+	hp9k8[67]1 | hp8[67]1 | hp9k80[24] | hp80[24] | hp9k8[78]9 | hp8[78]9 | hp9k893 | hp893)
+		# FIXME: really hppa2.0-hp
+		basic_machine=hppa1.1-hp
+		;;
+	hp9k8[0-9][13679] | hp8[0-9][13679])
+		basic_machine=hppa1.1-hp
+		;;
+	hp9k8[0-9][0-9] | hp8[0-9][0-9])
+		basic_machine=hppa1.0-hp
+		;;
+	hppa-next)
+		os=-nextstep3
+		;;
+	hppaosf)
+		basic_machine=hppa1.1-hp
+		os=-osf
+		;;
+	hppro)
+		basic_machine=hppa1.1-hp
+		os=-proelf
+		;;
+	i370-ibm* | ibm*)
+		basic_machine=i370-ibm
+		;;
+# I'm not sure what "Sysv32" means.  Should this be sysv3.2?
+	i*86v32)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv32
+		;;
+	i*86v4*)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv4
+		;;
+	i*86v)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-sysv
+		;;
+	i*86sol2)
+		basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'`
+		os=-solaris2
+		;;
+	i386mach)
+		basic_machine=i386-mach
+		os=-mach
+		;;
+	i386-vsta | vsta)
+		basic_machine=i386-unknown
+		os=-vsta
+		;;
+	iris | iris4d)
+		basic_machine=mips-sgi
+		case $os in
+		    -irix*)
+			;;
+		    *)
+			os=-irix4
+			;;
+		esac
+		;;
+	isi68 | isi)
+		basic_machine=m68k-isi
+		os=-sysv
+		;;
+	m88k-omron*)
+		basic_machine=m88k-omron
+		;;
+	magnum | m3230)
+		basic_machine=mips-mips
+		os=-sysv
+		;;
+	merlin)
+		basic_machine=ns32k-utek
+		os=-sysv
+		;;
+	mingw32)
+		basic_machine=i386-pc
+		os=-mingw32
+		;;
+	miniframe)
+		basic_machine=m68000-convergent
+		;;
+	*mint | -mint[0-9]* | *MiNT | *MiNT[0-9]*)
+		basic_machine=m68k-atari
+		os=-mint
+		;;
+	mips3*-*)
+		basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`
+		;;
+	mips3*)
+		basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`-unknown
+		;;
+	monitor)
+		basic_machine=m68k-rom68k
+		os=-coff
+		;;
+	morphos)
+		basic_machine=powerpc-unknown
+		os=-morphos
+		;;
+	msdos)
+		basic_machine=i386-pc
+		os=-msdos
+		;;
+	mvs)
+		basic_machine=i370-ibm
+		os=-mvs
+		;;
+	ncr3000)
+		basic_machine=i486-ncr
+		os=-sysv4
+		;;
+	netbsd386)
+		basic_machine=i386-unknown
+		os=-netbsd
+		;;
+	netwinder)
+		basic_machine=armv4l-rebel
+		os=-linux
+		;;
+	news | news700 | news800 | news900)
+		basic_machine=m68k-sony
+		os=-newsos
+		;;
+	news1000)
+		basic_machine=m68030-sony
+		os=-newsos
+		;;
+	news-3600 | risc-news)
+		basic_machine=mips-sony
+		os=-newsos
+		;;
+	necv70)
+		basic_machine=v70-nec
+		os=-sysv
+		;;
+	next | m*-next )
+		basic_machine=m68k-next
+		case $os in
+		    -nextstep* )
+			;;
+		    -ns2*)
+		      os=-nextstep2
+			;;
+		    *)
+		      os=-nextstep3
+			;;
+		esac
+		;;
+	nh3000)
+		basic_machine=m68k-harris
+		os=-cxux
+		;;
+	nh[45]000)
+		basic_machine=m88k-harris
+		os=-cxux
+		;;
+	nindy960)
+		basic_machine=i960-intel
+		os=-nindy
+		;;
+	mon960)
+		basic_machine=i960-intel
+		os=-mon960
+		;;
+	nonstopux)
+		basic_machine=mips-compaq
+		os=-nonstopux
+		;;
+	np1)
+		basic_machine=np1-gould
+		;;
+	nsr-tandem)
+		basic_machine=nsr-tandem
+		;;
+	op50n-* | op60c-*)
+		basic_machine=hppa1.1-oki
+		os=-proelf
+		;;
+	or32 | or32-*)
+		basic_machine=or32-unknown
+		os=-coff
+		;;
+	os400)
+		basic_machine=powerpc-ibm
+		os=-os400
+		;;
+	OSE68000 | ose68000)
+		basic_machine=m68000-ericsson
+		os=-ose
+		;;
+	os68k)
+		basic_machine=m68k-none
+		os=-os68k
+		;;
+	pa-hitachi)
+		basic_machine=hppa1.1-hitachi
+		os=-hiuxwe2
+		;;
+	paragon)
+		basic_machine=i860-intel
+		os=-osf
+		;;
+	pbd)
+		basic_machine=sparc-tti
+		;;
+	pbb)
+		basic_machine=m68k-tti
+		;;
+	pc532 | pc532-*)
+		basic_machine=ns32k-pc532
+		;;
+	pentium | p5 | k5 | k6 | nexgen | viac3)
+		basic_machine=i586-pc
+		;;
+	pentiumpro | p6 | 6x86 | athlon | athlon_*)
+		basic_machine=i686-pc
+		;;
+	pentiumii | pentium2 | pentiumiii | pentium3)
+		basic_machine=i686-pc
+		;;
+	pentium4)
+		basic_machine=i786-pc
+		;;
+	pentium-* | p5-* | k5-* | k6-* | nexgen-* | viac3-*)
+		basic_machine=i586-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	pentiumpro-* | p6-* | 6x86-* | athlon-*)
+		basic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	pentiumii-* | pentium2-* | pentiumiii-* | pentium3-*)
+		basic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	pentium4-*)
+		basic_machine=i786-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	pn)
+		basic_machine=pn-gould
+		;;
+	power)	basic_machine=power-ibm
+		;;
+	ppc)	basic_machine=powerpc-unknown
+		;;
+	ppc-*)	basic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ppcle | powerpclittle | ppc-le | powerpc-little)
+		basic_machine=powerpcle-unknown
+		;;
+	ppcle-* | powerpclittle-*)
+		basic_machine=powerpcle-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ppc64)	basic_machine=powerpc64-unknown
+		;;
+	ppc64-*) basic_machine=powerpc64-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ppc64le | powerpc64little | ppc64-le | powerpc64-little)
+		basic_machine=powerpc64le-unknown
+		;;
+	ppc64le-* | powerpc64little-*)
+		basic_machine=powerpc64le-`echo $basic_machine | sed 's/^[^-]*-//'`
+		;;
+	ps2)
+		basic_machine=i386-ibm
+		;;
+	pw32)
+		basic_machine=i586-unknown
+		os=-pw32
+		;;
+	rom68k)
+		basic_machine=m68k-rom68k
+		os=-coff
+		;;
+	rm[46]00)
+		basic_machine=mips-siemens
+		;;
+	rtpc | rtpc-*)
+		basic_machine=romp-ibm
+		;;
+	s390 | s390-*)
+		basic_machine=s390-ibm
+		;;
+	s390x | s390x-*)
+		basic_machine=s390x-ibm
+		;;
+	sa29200)
+		basic_machine=a29k-amd
+		os=-udi
+		;;
+	sb1)
+		basic_machine=mipsisa64sb1-unknown
+		;;
+	sb1el)
+		basic_machine=mipsisa64sb1el-unknown
+		;;
+	sei)
+		basic_machine=mips-sei
+		os=-seiux
+		;;
+	sequent)
+		basic_machine=i386-sequent
+		;;
+	sh)
+		basic_machine=sh-hitachi
+		os=-hms
+		;;
+	sh64)
+		basic_machine=sh64-unknown
+		;;
+	sparclite-wrs | simso-wrs)
+		basic_machine=sparclite-wrs
+		os=-vxworks
+		;;
+	sps7)
+		basic_machine=m68k-bull
+		os=-sysv2
+		;;
+	spur)
+		basic_machine=spur-unknown
+		;;
+	st2000)
+		basic_machine=m68k-tandem
+		;;
+	stratus)
+		basic_machine=i860-stratus
+		os=-sysv4
+		;;
+	sun2)
+		basic_machine=m68000-sun
+		;;
+	sun2os3)
+		basic_machine=m68000-sun
+		os=-sunos3
+		;;
+	sun2os4)
+		basic_machine=m68000-sun
+		os=-sunos4
+		;;
+	sun3os3)
+		basic_machine=m68k-sun
+		os=-sunos3
+		;;
+	sun3os4)
+		basic_machine=m68k-sun
+		os=-sunos4
+		;;
+	sun4os3)
+		basic_machine=sparc-sun
+		os=-sunos3
+		;;
+	sun4os4)
+		basic_machine=sparc-sun
+		os=-sunos4
+		;;
+	sun4sol2)
+		basic_machine=sparc-sun
+		os=-solaris2
+		;;
+	sun3 | sun3-*)
+		basic_machine=m68k-sun
+		;;
+	sun4)
+		basic_machine=sparc-sun
+		;;
+	sun386 | sun386i | roadrunner)
+		basic_machine=i386-sun
+		;;
+	sv1)
+		basic_machine=sv1-cray
+		os=-unicos
+		;;
+	symmetry)
+		basic_machine=i386-sequent
+		os=-dynix
+		;;
+	t3e)
+		basic_machine=alphaev5-cray
+		os=-unicos
+		;;
+	t90)
+		basic_machine=t90-cray
+		os=-unicos
+		;;
+	tic54x | c54x*)
+		basic_machine=tic54x-unknown
+		os=-coff
+		;;
+	tic55x | c55x*)
+		basic_machine=tic55x-unknown
+		os=-coff
+		;;
+	tic6x | c6x*)
+		basic_machine=tic6x-unknown
+		os=-coff
+		;;
+	tx39)
+		basic_machine=mipstx39-unknown
+		;;
+	tx39el)
+		basic_machine=mipstx39el-unknown
+		;;
+	toad1)
+		basic_machine=pdp10-xkl
+		os=-tops20
+		;;
+	tower | tower-32)
+		basic_machine=m68k-ncr
+		;;
+	tpf)
+		basic_machine=s390x-ibm
+		os=-tpf
+		;;
+	udi29k)
+		basic_machine=a29k-amd
+		os=-udi
+		;;
+	ultra3)
+		basic_machine=a29k-nyu
+		os=-sym1
+		;;
+	v810 | necv810)
+		basic_machine=v810-nec
+		os=-none
+		;;
+	vaxv)
+		basic_machine=vax-dec
+		os=-sysv
+		;;
+	vms)
+		basic_machine=vax-dec
+		os=-vms
+		;;
+	vpp*|vx|vx-*)
+		basic_machine=f301-fujitsu
+		;;
+	vxworks960)
+		basic_machine=i960-wrs
+		os=-vxworks
+		;;
+	vxworks68)
+		basic_machine=m68k-wrs
+		os=-vxworks
+		;;
+	vxworks29k)
+		basic_machine=a29k-wrs
+		os=-vxworks
+		;;
+	w65*)
+		basic_machine=w65-wdc
+		os=-none
+		;;
+	w89k-*)
+		basic_machine=hppa1.1-winbond
+		os=-proelf
+		;;
+	xps | xps100)
+		basic_machine=xps100-honeywell
+		;;
+	ymp)
+		basic_machine=ymp-cray
+		os=-unicos
+		;;
+	z8k-*-coff)
+		basic_machine=z8k-unknown
+		os=-sim
+		;;
+	none)
+		basic_machine=none-none
+		os=-none
+		;;
+
+# Here we handle the default manufacturer of certain CPU types.  It is in
+# some cases the only manufacturer, in others, it is the most popular.
+	w89k)
+		basic_machine=hppa1.1-winbond
+		;;
+	op50n)
+		basic_machine=hppa1.1-oki
+		;;
+	op60c)
+		basic_machine=hppa1.1-oki
+		;;
+	romp)
+		basic_machine=romp-ibm
+		;;
+	mmix)
+		basic_machine=mmix-knuth
+		;;
+	rs6000)
+		basic_machine=rs6000-ibm
+		;;
+	vax)
+		basic_machine=vax-dec
+		;;
+	pdp10)
+		# there are many clones, so DEC is not a safe bet
+		basic_machine=pdp10-unknown
+		;;
+	pdp11)
+		basic_machine=pdp11-dec
+		;;
+	we32k)
+		basic_machine=we32k-att
+		;;
+	sh3 | sh4 | sh[34]eb | sh[1234]le | sh[23]ele)
+		basic_machine=sh-unknown
+		;;
+	sh64)
+		basic_machine=sh64-unknown
+		;;
+	sparc | sparcv8 | sparcv9 | sparcv9b)
+		basic_machine=sparc-sun
+		;;
+	cydra)
+		basic_machine=cydra-cydrome
+		;;
+	orion)
+		basic_machine=orion-highlevel
+		;;
+	orion105)
+		basic_machine=clipper-highlevel
+		;;
+	mac | mpw | mac-mpw)
+		basic_machine=m68k-apple
+		;;
+	pmac | pmac-mpw)
+		basic_machine=powerpc-apple
+		;;
+	*-unknown)
+		# Make sure to match an already-canonicalized machine name.
+		;;
+	*)
+		echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2
+		exit 1
+		;;
+esac
+
+# Here we canonicalize certain aliases for manufacturers.
+case $basic_machine in
+	*-digital*)
+		basic_machine=`echo $basic_machine | sed 's/digital.*/dec/'`
+		;;
+	*-commodore*)
+		basic_machine=`echo $basic_machine | sed 's/commodore.*/cbm/'`
+		;;
+	*)
+		;;
+esac
+
+# Decode manufacturer-specific aliases for certain operating systems.
+
+if [ x"$os" != x"" ]
+then
+case $os in
+        # First match some system type aliases
+        # that might get confused with valid system types.
+	# -solaris* is a basic system type, with this one exception.
+	-solaris1 | -solaris1.*)
+		os=`echo $os | sed -e 's|solaris1|sunos4|'`
+		;;
+	-solaris)
+		os=-solaris2
+		;;
+	-svr4*)
+		os=-sysv4
+		;;
+	-unixware*)
+		os=-sysv4.2uw
+		;;
+	-gnu/linux*)
+		os=`echo $os | sed -e 's|gnu/linux|linux-gnu|'`
+		;;
+	# First accept the basic system types.
+	# The portable systems comes first.
+	# Each alternative MUST END IN A *, to match a version number.
+	# -sysv* is not here because it comes later, after sysvr4.
+	-gnu* | -bsd* | -mach* | -minix* | -genix* | -ultrix* | -irix* \
+	      | -*vms* | -sco* | -esix* | -isc* | -aix* | -sunos | -sunos[34]*\
+	      | -hpux* | -unos* | -osf* | -luna* | -dgux* | -solaris* | -sym* \
+	      | -amigaos* | -amigados* | -msdos* | -newsos* | -unicos* | -aof* \
+	      | -aos* \
+	      | -nindy* | -vxsim* | -vxworks* | -ebmon* | -hms* | -mvs* \
+	      | -clix* | -riscos* | -uniplus* | -iris* | -rtu* | -xenix* \
+	      | -hiux* | -386bsd* | -knetbsd* | -mirbsd* | -netbsd* | -openbsd* \
+	      | -ekkobsd* | -kfreebsd* | -freebsd* | -riscix* | -lynxos* \
+	      | -bosx* | -nextstep* | -cxux* | -aout* | -elf* | -oabi* \
+	      | -ptx* | -coff* | -ecoff* | -winnt* | -domain* | -vsta* \
+	      | -udi* | -eabi* | -lites* | -ieee* | -go32* | -aux* \
+	      | -chorusos* | -chorusrdb* \
+	      | -cygwin* | -pe* | -psos* | -moss* | -proelf* | -rtems* \
+	      | -mingw32* | -linux-gnu* | -linux-uclibc* | -uxpv* | -beos* | -mpeix* | -udk* \
+	      | -interix* | -uwin* | -mks* | -rhapsody* | -darwin* | -opened* \
+	      | -openstep* | -oskit* | -conix* | -pw32* | -nonstopux* \
+	      | -storm-chaos* | -tops10* | -tenex* | -tops20* | -its* \
+	      | -os2* | -vos* | -palmos* | -uclinux* | -nucleus* \
+	      | -morphos* | -superux* | -rtmk* | -rtmk-nova* | -windiss* \
+	      | -powermax* | -dnix* | -nx6 | -nx7 | -sei* | -dragonfly*)
+	# Remember, each alternative MUST END IN *, to match a version number.
+		;;
+	-qnx*)
+		case $basic_machine in
+		    x86-* | i*86-*)
+			;;
+		    *)
+			os=-nto$os
+			;;
+		esac
+		;;
+	-nto-qnx*)
+		;;
+	-nto*)
+		os=`echo $os | sed -e 's|nto|nto-qnx|'`
+		;;
+	-sim | -es1800* | -hms* | -xray | -os68k* | -none* | -v88r* \
+	      | -windows* | -osx | -abug | -netware* | -os9* | -beos* \
+	      | -macos* | -mpw* | -magic* | -mmixware* | -mon960* | -lnews*)
+		;;
+	-mac*)
+		os=`echo $os | sed -e 's|mac|macos|'`
+		;;
+	-linux-dietlibc)
+		os=-linux-dietlibc
+		;;
+	-linux*)
+		os=`echo $os | sed -e 's|linux|linux-gnu|'`
+		;;
+	-sunos5*)
+		os=`echo $os | sed -e 's|sunos5|solaris2|'`
+		;;
+	-sunos6*)
+		os=`echo $os | sed -e 's|sunos6|solaris3|'`
+		;;
+	-opened*)
+		os=-openedition
+		;;
+        -os400*)
+		os=-os400
+		;;
+	-wince*)
+		os=-wince
+		;;
+	-osfrose*)
+		os=-osfrose
+		;;
+	-osf*)
+		os=-osf
+		;;
+	-utek*)
+		os=-bsd
+		;;
+	-dynix*)
+		os=-bsd
+		;;
+	-acis*)
+		os=-aos
+		;;
+	-atheos*)
+		os=-atheos
+		;;
+	-syllable*)
+		os=-syllable
+		;;
+	-386bsd)
+		os=-bsd
+		;;
+	-ctix* | -uts*)
+		os=-sysv
+		;;
+	-nova*)
+		os=-rtmk-nova
+		;;
+	-ns2 )
+		os=-nextstep2
+		;;
+	-nsk*)
+		os=-nsk
+		;;
+	# Preserve the version number of sinix5.
+	-sinix5.*)
+		os=`echo $os | sed -e 's|sinix|sysv|'`
+		;;
+	-sinix*)
+		os=-sysv4
+		;;
+        -tpf*)
+		os=-tpf
+		;;
+	-triton*)
+		os=-sysv3
+		;;
+	-oss*)
+		os=-sysv3
+		;;
+	-svr4)
+		os=-sysv4
+		;;
+	-svr3)
+		os=-sysv3
+		;;
+	-sysvr4)
+		os=-sysv4
+		;;
+	# This must come after -sysvr4.
+	-sysv*)
+		;;
+	-ose*)
+		os=-ose
+		;;
+	-es1800*)
+		os=-ose
+		;;
+	-xenix)
+		os=-xenix
+		;;
+	-*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*)
+		os=-mint
+		;;
+	-aros*)
+		os=-aros
+		;;
+	-kaos*)
+		os=-kaos
+		;;
+	-none)
+		;;
+	*)
+		# Get rid of the `-' at the beginning of $os.
+		os=`echo $os | sed 's/[^-]*-//'`
+		echo Invalid configuration \`$1\': system \`$os\' not recognized 1>&2
+		exit 1
+		;;
+esac
+else
+
+# Here we handle the default operating systems that come with various machines.
+# The value should be what the vendor currently ships out the door with their
+# machine or put another way, the most popular os provided with the machine.
+
+# Note that if you're going to try to match "-MANUFACTURER" here (say,
+# "-sun"), then you have to tell the case statement up towards the top
+# that MANUFACTURER isn't an operating system.  Otherwise, code above
+# will signal an error saying that MANUFACTURER isn't an operating
+# system, and we'll never get to this point.
+
+case $basic_machine in
+	*-acorn)
+		os=-riscix1.2
+		;;
+	arm*-rebel)
+		os=-linux
+		;;
+	arm*-semi)
+		os=-aout
+		;;
+    c4x-* | tic4x-*)
+        os=-coff
+        ;;
+	# This must come before the *-dec entry.
+	pdp10-*)
+		os=-tops20
+		;;
+	pdp11-*)
+		os=-none
+		;;
+	*-dec | vax-*)
+		os=-ultrix4.2
+		;;
+	m68*-apollo)
+		os=-domain
+		;;
+	i386-sun)
+		os=-sunos4.0.2
+		;;
+	m68000-sun)
+		os=-sunos3
+		# This also exists in the configure program, but was not the
+		# default.
+		# os=-sunos4
+		;;
+	m68*-cisco)
+		os=-aout
+		;;
+	mips*-cisco)
+		os=-elf
+		;;
+	mips*-*)
+		os=-elf
+		;;
+	or32-*)
+		os=-coff
+		;;
+	*-tti)	# must be before sparc entry or we get the wrong os.
+		os=-sysv3
+		;;
+	sparc-* | *-sun)
+		os=-sunos4.1.1
+		;;
+	*-be)
+		os=-beos
+		;;
+	*-ibm)
+		os=-aix
+		;;
+    	*-knuth)
+		os=-mmixware
+		;;
+	*-wec)
+		os=-proelf
+		;;
+	*-winbond)
+		os=-proelf
+		;;
+	*-oki)
+		os=-proelf
+		;;
+	*-hp)
+		os=-hpux
+		;;
+	*-hitachi)
+		os=-hiux
+		;;
+	i860-* | *-att | *-ncr | *-altos | *-motorola | *-convergent)
+		os=-sysv
+		;;
+	*-cbm)
+		os=-amigaos
+		;;
+	*-dg)
+		os=-dgux
+		;;
+	*-dolphin)
+		os=-sysv3
+		;;
+	m68k-ccur)
+		os=-rtu
+		;;
+	m88k-omron*)
+		os=-luna
+		;;
+	*-next )
+		os=-nextstep
+		;;
+	*-sequent)
+		os=-ptx
+		;;
+	*-crds)
+		os=-unos
+		;;
+	*-ns)
+		os=-genix
+		;;
+	i370-*)
+		os=-mvs
+		;;
+	*-next)
+		os=-nextstep3
+		;;
+	*-gould)
+		os=-sysv
+		;;
+	*-highlevel)
+		os=-bsd
+		;;
+	*-encore)
+		os=-bsd
+		;;
+	*-sgi)
+		os=-irix
+		;;
+	*-siemens)
+		os=-sysv4
+		;;
+	*-masscomp)
+		os=-rtu
+		;;
+	f30[01]-fujitsu | f700-fujitsu)
+		os=-uxpv
+		;;
+	*-rom68k)
+		os=-coff
+		;;
+	*-*bug)
+		os=-coff
+		;;
+	*-apple)
+		os=-macos
+		;;
+	*-atari*)
+		os=-mint
+		;;
+	*)
+		os=-none
+		;;
+esac
+fi
+
+# Here we handle the case where we know the os, and the CPU type, but not the
+# manufacturer.  We pick the logical manufacturer.
+vendor=unknown
+case $basic_machine in
+	*-unknown)
+		case $os in
+			-riscix*)
+				vendor=acorn
+				;;
+			-sunos*)
+				vendor=sun
+				;;
+			-aix*)
+				vendor=ibm
+				;;
+			-beos*)
+				vendor=be
+				;;
+			-hpux*)
+				vendor=hp
+				;;
+			-mpeix*)
+				vendor=hp
+				;;
+			-hiux*)
+				vendor=hitachi
+				;;
+			-unos*)
+				vendor=crds
+				;;
+			-dgux*)
+				vendor=dg
+				;;
+			-luna*)
+				vendor=omron
+				;;
+			-genix*)
+				vendor=ns
+				;;
+			-mvs* | -opened*)
+				vendor=ibm
+				;;
+			-os400*)
+				vendor=ibm
+				;;
+			-ptx*)
+				vendor=sequent
+				;;
+			-tpf*)
+				vendor=ibm
+				;;
+			-vxsim* | -vxworks* | -windiss*)
+				vendor=wrs
+				;;
+			-aux*)
+				vendor=apple
+				;;
+			-hms*)
+				vendor=hitachi
+				;;
+			-mpw* | -macos*)
+				vendor=apple
+				;;
+			-*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*)
+				vendor=atari
+				;;
+			-vos*)
+				vendor=stratus
+				;;
+		esac
+		basic_machine=`echo $basic_machine | sed "s/unknown/$vendor/"`
+		;;
+esac
+
+echo $basic_machine$os
+exit 0
+
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "timestamp='"
+# time-stamp-format: "%:y-%02m-%02d"
+# time-stamp-end: "'"
+# End:

Propchange: klee/trunk/autoconf/config.sub

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/autoconf/configure.ac
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/autoconf/configure.ac?rev=72205&view=auto

==============================================================================
--- klee/trunk/autoconf/configure.ac (added)
+++ klee/trunk/autoconf/configure.ac Wed May 20 23:36:41 2009
@@ -0,0 +1,253 @@
+dnl **************************************************************************
+dnl * Initialize
+dnl **************************************************************************
+AC_INIT([[KLEE]],[[0.01]],[daniel at minormatter.com])
+
+dnl Identify where LLVM source tree is (this is patched by
+dnl AutoRegen.sh)
+LLVM_SRC_ROOT=XXX
+
+dnl Tell autoconf that the auxilliary files are actually located in
+dnl the LLVM autoconf directory, not here.
+AC_CONFIG_AUX_DIR($LLVM_SRC_ROOT/autoconf)
+
+dnl Tell autoconf that this is an LLVM project being configured
+dnl This provides the --with-llvmsrc and --with-llvmobj options
+LLVM_CONFIG_PROJECT("","")
+
+dnl Verify that the source directory is valid
+AC_CONFIG_SRCDIR(["Makefile.config.in"])
+
+dnl Configure a common Makefile
+AC_CONFIG_FILES(Makefile.config)
+AC_CONFIG_FILES(stp/Makefile.common)
+
+dnl Configure project makefiles
+dnl List every Makefile that exists within your source tree
+AC_CONFIG_HEADERS([include/klee/Config/config.h])
+
+dnl FIXME: Make out of tree builds work.
+
+AC_LANG([C++])
+
+dnl **************************************************************************
+dnl Find the host
+
+AC_CANONICAL_TARGET
+
+dnl Determine the platform type and cache its value. This helps us configure
+dnl the System library to the correct build platform.
+AC_CACHE_CHECK([type of operating system we're going to host on],
+               [klee_cv_os_type],
+[case $host in
+  *-*-linux*)
+    host_supports_posix_runtime=yes ;;
+  *)
+    host_supports_posix_runtime=no ;;
+esac])
+
+dnl **************************************************************************
+dnl Verify that we can find llvm
+
+dnl --with-llvm is a shortcut for setting srcdir and objdir.
+AC_ARG_WITH(llvm,
+  AS_HELP_STRING([--with-llvm],
+    [Location of LLVM Source and Object code]),,)
+
+AC_MSG_CHECKING([llvm source dir])
+
+if test X${with_llvm} != X; then
+    dnl Verify that --with-llvm{src,obj} were not given.
+    if test X${with_llvmsrc} != X; then
+       AC_MSG_ERROR([--with-llvmsrc cannot be specified when using --with-llvm])
+    fi   
+    if test X${with_llvmobj} != X; then
+       AC_MSG_ERROR([--with-llvmobj cannot be specified when using --with-llvm])
+    fi   
+    with_llvmsrc=$with_llvm
+    with_llvmobj=$with_llvm
+fi
+
+dnl If one of with_llvmsrc or with_llvmobj was given, we must have both.
+if (test X${with_llvmsrc} != X || test X${with_llvmobj} != X); then
+    dnl Verify that with_llvmobj was given as well.
+    if test X${with_llvmsrc} = X; then
+       AC_MSG_ERROR([--with-llvmsrc must be specified when using --with-llvmobj])
+    fi      
+    if test X${with_llvmobj} = X; then
+       AC_MSG_ERROR([--with-llvmobj must be specified when using --with-llvmsrc])
+    fi      
+else
+    dnl Otherwise try and use llvm-config to find.
+    llvm_version=`llvm-config --version`
+    if test X${llvm_version} = X; then
+       AC_MSG_ERROR([unable to find llvm, use --with-llvmsrc and --with-llvmobj])
+    fi
+    
+    with_llvmsrc=`llvm-config --src-root`
+    with_llvmobj=`llvm-config --obj-root`
+fi
+
+dnl Try to validate directories
+if test ! -f ${with_llvmsrc}/Makefile.rules; then
+   AC_MSG_ERROR([invalid llvmsrc directory: ${with_llvmsrc}])
+fi
+if test ! -f ${with_llvmobj}/Makefile.config; then
+   AC_MSG_ERROR([invalid llvmobj directory: ${with_llvmobj}])
+fi
+
+dnl Make the paths absolute
+llvm_src=`cd $with_llvmsrc 2> /dev/null; pwd`
+llvm_obj=`cd $with_llvmobj 2> /dev/null; pwd`
+   
+AC_MSG_RESULT([$llvm_src])
+
+dnl Report obj dir as well.
+AC_MSG_CHECKING([llvm obj dir])
+AC_MSG_RESULT([$llvm_obj])
+
+AC_SUBST(LLVM_SRC,$llvm_src)
+AC_SUBST(LLVM_OBJ,$llvm_obj)
+
+dnl **************************************************************************
+dnl User option to enable uClibc support.
+
+AC_ARG_WITH(uclibc,
+  AS_HELP_STRING([--with-uclibc],
+    [Enable use of the klee uclibc at the given path]),,)
+
+dnl If uclibc wasn't given, check for a uclibc in the current
+dnl directory.
+if (test X${with_uclibc} = X && test -d uclibc); then
+   with_uclibc=uclibc
+fi
+
+dnl Validate uclibc if given.
+
+AC_MSG_CHECKING([uclibc])
+if (test X${with_uclibc} != X); then
+   if test ! -d ${with_uclibc}; then
+       AC_MSG_ERROR([invalid uclibc directory: ${with_uclibc}])
+   fi
+
+   dnl Make the path absolute
+   with_uclibc=`cd $with_uclibc 2> /dev/null; pwd`
+
+   AC_MSG_RESULT([$with_uclibc])
+else
+   AC_MSG_RESULT([no])
+fi
+
+AC_DEFINE_UNQUOTED(KLEE_UCLIBC, "$with_uclibc", [Path to KLEE's uClibc])
+AC_SUBST(KLEE_UCLIBC)         
+
+if test X${with_uclibc} != X ; then
+  AC_SUBST(ENABLE_UCLIBC,[[1]])
+else
+  AC_SUBST(ENABLE_UCLIBC,[[0]])
+fi
+
+dnl **************************************************************************
+dnl User option to enable the POSIX runtime
+
+AC_ARG_ENABLE(posix-runtime,
+              AS_HELP_STRING([--enable-posix-runtime],
+                             [Enable the POSIX runtime]),
+                             ,enableval=default)
+
+AC_MSG_CHECKING([POSIX runtime])
+if test ${enableval} = "default" ; then
+  if test X${with_uclibc} != X; then
+    enableval=$host_supports_posix_runtime
+    if test ${enableval} = "yes"; then
+      AC_MSG_RESULT([default (enabled)])
+    else
+      AC_MSG_RESULT([default (disabled, unsupported target)])
+    fi
+  else
+    enableval="no"
+    AC_MSG_RESULT([default (disabled, no uclibc)])
+  fi
+else
+  if test ${enableval} = "yes" ; then
+    AC_MSG_RESULT([yes])
+  else
+    AC_MSG_RESULT([no])
+  fi
+fi
+
+if test ${enableval} = "yes" ; then
+  AC_SUBST(ENABLE_POSIX_RUNTIME,[[1]])
+else
+  AC_SUBST(ENABLE_POSIX_RUNTIME,[[0]])
+fi
+
+dnl **************************************************************************
+dnl User option to select runtime version
+
+AC_ARG_WITH(runtime,
+  AS_HELP_STRING([--with-runtime],
+    [Select build configuration for runtime libraries (default [Release])]),,
+    withval=default)
+
+if test X"${withval}" = Xdefault; then
+   with_runtime=Release
+fi
+
+AC_MSG_CHECKING([runtime configuration])
+if test X${with_runtime} = XRelease; then
+    AC_MSG_RESULT([Release])
+    AC_SUBST(RUNTIME_ENABLE_OPTIMIZED,[[1]])
+elif test X${with_runtime} = XDebug; then
+   AC_MSG_RESULT([Debug])
+   AC_SUBST(RUNTIME_ENABLE_OPTIMIZED,[[0]])
+else
+   AC_MSG_ERROR([invalid configuration: ${with_runtime}])
+fi
+
+AC_DEFINE_UNQUOTED(RUNTIME_CONFIGURATION, "$with_runtime", [Configuration for runtime libraries])
+AC_SUBST(RUNTIME_CONFIGURATION)
+
+dnl **************************************************************************
+dnl See if we should support __ctype_b_loc externals.
+
+dnl FIXME: Do the proper test if we continue to need this.
+case $host in
+  *-*-linux*)
+    AC_DEFINE_UNQUOTED(HAVE_CTYPE_EXTERNALS, 1, [Does the platform use __ctype_b_loc, etc.])
+esac
+
+dnl **************************************************************************
+dnl Checks for header files.
+
+dnl NOTE: This is mostly just to force autoconf to make CFLAGS defines
+dnl for us.
+AC_LANG_PUSH([C])
+
+AC_CHECK_HEADERS([sys/acl.h])
+
+AC_LANG_POP([C])
+
+AC_CHECK_HEADERS([selinux/selinux.h],
+        AC_SUBST(HAVE_SELINUX, 1),
+        AC_SUBST(HAVE_SELINUX, 0))
+
+dnl User option to use stplog.
+
+AC_ARG_ENABLE(stplog,
+              AS_HELP_STRING([--enable-stplog],
+                             [Compile with the stplog library [[disabled]]]),
+                             ,enableval=no)
+if test ${enableval} = "yes" ; then
+  AC_SUBST(ENABLE_STPLOG,[[1]])
+else
+  AC_SUBST(ENABLE_STPLOG,[[0]])
+fi
+AC_DEFINE_UNQUOTED([ENABLE_STPLOG],$ENABLE_STPLOG,[Define if stplog enabled])
+
+dnl **************************************************************************
+dnl * Create the output files
+dnl **************************************************************************
+
+dnl This must be last
+AC_OUTPUT

Added: klee/trunk/autoconf/install-sh
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/autoconf/install-sh?rev=72205&view=auto

==============================================================================
--- klee/trunk/autoconf/install-sh (added)
+++ klee/trunk/autoconf/install-sh Wed May 20 23:36:41 2009
@@ -0,0 +1,322 @@
+#!/bin/sh
+# install - install a program, script, or datafile
+
+scriptversion=2004-09-10.20
+
+# This originates from X11R5 (mit/util/scripts/install.sh), which was
+# later released in X11R6 (xc/config/util/install.sh) with the
+# following copyright and license.
+#
+# Copyright (C) 1994 X Consortium
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to
+# deal in the Software without restriction, including without limitation the
+# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+# sell copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+# X CONSORTIUM BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNEC-
+# TION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+#
+# Except as contained in this notice, the name of the X Consortium shall not
+# be used in advertising or otherwise to promote the sale, use or other deal-
+# ings in this Software without prior written authorization from the X Consor-
+# tium.
+#
+#
+# FSF changes to this file are in the public domain.
+#
+# Calling this script install-sh is preferred over install.sh, to prevent
+# `make' implicit rules from creating a file called install from it
+# when there is no Makefile.
+#
+# This script is compatible with the BSD install script, but was written
+# from scratch.  It can only install one file at a time, a restriction
+# shared with many OS's install programs.
+
+# set DOITPROG to echo to test this script
+
+# Don't use :- since 4.3BSD and earlier shells don't like it.
+doit="${DOITPROG-}"
+
+# put in absolute paths if you don't have them in your path; or use env. vars.
+
+mvprog="${MVPROG-mv}"
+cpprog="${CPPROG-cp}"
+chmodprog="${CHMODPROG-chmod}"
+chownprog="${CHOWNPROG-chown}"
+chgrpprog="${CHGRPPROG-chgrp}"
+stripprog="${STRIPPROG-strip}"
+rmprog="${RMPROG-rm}"
+mkdirprog="${MKDIRPROG-mkdir}"
+
+chmodcmd="$chmodprog 0755"
+chowncmd=
+chgrpcmd=
+stripcmd=
+rmcmd="$rmprog -f"
+mvcmd="$mvprog"
+src=
+dst=
+dir_arg=
+dstarg=
+no_target_directory=
+
+usage="Usage: $0 [OPTION]... [-T] SRCFILE DSTFILE
+   or: $0 [OPTION]... SRCFILES... DIRECTORY
+   or: $0 [OPTION]... -t DIRECTORY SRCFILES...
+   or: $0 [OPTION]... -d DIRECTORIES...
+
+In the 1st form, copy SRCFILE to DSTFILE.
+In the 2nd and 3rd, copy all SRCFILES to DIRECTORY.
+In the 4th, create DIRECTORIES.
+
+Options:
+-c         (ignored)
+-d         create directories instead of installing files.
+-g GROUP   $chgrpprog installed files to GROUP.
+-m MODE    $chmodprog installed files to MODE.
+-o USER    $chownprog installed files to USER.
+-s         $stripprog installed files.
+-t DIRECTORY  install into DIRECTORY.
+-T         report an error if DSTFILE is a directory.
+--help     display this help and exit.
+--version  display version info and exit.
+
+Environment variables override the default commands:
+  CHGRPPROG CHMODPROG CHOWNPROG CPPROG MKDIRPROG MVPROG RMPROG STRIPPROG
+"
+
+while test -n "$1"; do
+  case $1 in
+    -c) shift
+        continue;;
+
+    -d) dir_arg=true
+        shift
+        continue;;
+
+    -g) chgrpcmd="$chgrpprog $2"
+        shift
+        shift
+        continue;;
+
+    --help) echo "$usage"; exit 0;;
+
+    -m) chmodcmd="$chmodprog $2"
+        shift
+        shift
+        continue;;
+
+    -o) chowncmd="$chownprog $2"
+        shift
+        shift
+        continue;;
+
+    -s) stripcmd=$stripprog
+        shift
+        continue;;
+
+    -t) dstarg=$2
+	shift
+	shift
+	continue;;
+
+    -T) no_target_directory=true
+	shift
+	continue;;
+
+    --version) echo "$0 $scriptversion"; exit 0;;
+
+    *)  # When -d is used, all remaining arguments are directories to create.
+	# When -t is used, the destination is already specified.
+	test -n "$dir_arg$dstarg" && break
+        # Otherwise, the last argument is the destination.  Remove it from $@.
+	for arg
+	do
+          if test -n "$dstarg"; then
+	    # $@ is not empty: it contains at least $arg.
+	    set fnord "$@" "$dstarg"
+	    shift # fnord
+	  fi
+	  shift # arg
+	  dstarg=$arg
+	done
+	break;;
+  esac
+done
+
+if test -z "$1"; then
+  if test -z "$dir_arg"; then
+    echo "$0: no input file specified." >&2
+    exit 1
+  fi
+  # It's OK to call `install-sh -d' without argument.
+  # This can happen when creating conditional directories.
+  exit 0
+fi
+
+for src
+do
+  # Protect names starting with `-'.
+  case $src in
+    -*) src=./$src ;;
+  esac
+
+  if test -n "$dir_arg"; then
+    dst=$src
+    src=
+
+    if test -d "$dst"; then
+      mkdircmd=:
+      chmodcmd=
+    else
+      mkdircmd=$mkdirprog
+    fi
+  else
+    # Waiting for this to be detected by the "$cpprog $src $dsttmp" command
+    # might cause directories to be created, which would be especially bad
+    # if $src (and thus $dsttmp) contains '*'.
+    if test ! -f "$src" && test ! -d "$src"; then
+      echo "$0: $src does not exist." >&2
+      exit 1
+    fi
+
+    if test -z "$dstarg"; then
+      echo "$0: no destination specified." >&2
+      exit 1
+    fi
+
+    dst=$dstarg
+    # Protect names starting with `-'.
+    case $dst in
+      -*) dst=./$dst ;;
+    esac
+
+    # If destination is a directory, append the input filename; won't work
+    # if double slashes aren't ignored.
+    if test -d "$dst"; then
+      if test -n "$no_target_directory"; then
+	echo "$0: $dstarg: Is a directory" >&2
+	exit 1
+      fi
+      dst=$dst/`basename "$src"`
+    fi
+  fi
+
+  # This sed command emulates the dirname command.
+  dstdir=`echo "$dst" | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'`
+
+  # Make sure that the destination directory exists.
+
+  # Skip lots of stat calls in the usual case.
+  if test ! -d "$dstdir"; then
+    defaultIFS='
+	 '
+    IFS="${IFS-$defaultIFS}"
+
+    oIFS=$IFS
+    # Some sh's can't handle IFS=/ for some reason.
+    IFS='%'
+    set - `echo "$dstdir" | sed -e 's@/@%@g' -e 's@^%@/@'`
+    IFS=$oIFS
+
+    pathcomp=
+
+    while test $# -ne 0 ; do
+      pathcomp=$pathcomp$1
+      shift
+      if test ! -d "$pathcomp"; then
+        $mkdirprog "$pathcomp"
+	# mkdir can fail with a `File exist' error in case several
+	# install-sh are creating the directory concurrently.  This
+	# is OK.
+	test -d "$pathcomp" || exit
+      fi
+      pathcomp=$pathcomp/
+    done
+  fi
+
+  if test -n "$dir_arg"; then
+    $doit $mkdircmd "$dst" \
+      && { test -z "$chowncmd" || $doit $chowncmd "$dst"; } \
+      && { test -z "$chgrpcmd" || $doit $chgrpcmd "$dst"; } \
+      && { test -z "$stripcmd" || $doit $stripcmd "$dst"; } \
+      && { test -z "$chmodcmd" || $doit $chmodcmd "$dst"; }
+
+  else
+    dstfile=`basename "$dst"`
+
+    # Make a couple of temp file names in the proper directory.
+    dsttmp=$dstdir/_inst.$$_
+    rmtmp=$dstdir/_rm.$$_
+
+    # Trap to clean up those temp files at exit.
+    trap 'ret=$?; rm -f "$dsttmp" "$rmtmp" && exit $ret' 0
+    trap '(exit $?); exit' 1 2 13 15
+
+    # Copy the file name to the temp name.
+    $doit $cpprog "$src" "$dsttmp" &&
+
+    # and set any options; do chmod last to preserve setuid bits.
+    #
+    # If any of these fail, we abort the whole thing.  If we want to
+    # ignore errors from any of these, just make sure not to ignore
+    # errors from the above "$doit $cpprog $src $dsttmp" command.
+    #
+    { test -z "$chowncmd" || $doit $chowncmd "$dsttmp"; } \
+      && { test -z "$chgrpcmd" || $doit $chgrpcmd "$dsttmp"; } \
+      && { test -z "$stripcmd" || $doit $stripcmd "$dsttmp"; } \
+      && { test -z "$chmodcmd" || $doit $chmodcmd "$dsttmp"; } &&
+
+    # Now rename the file to the real destination.
+    { $doit $mvcmd -f "$dsttmp" "$dstdir/$dstfile" 2>/dev/null \
+      || {
+	   # The rename failed, perhaps because mv can't rename something else
+	   # to itself, or perhaps because mv is so ancient that it does not
+	   # support -f.
+
+	   # Now remove or move aside any old file at destination location.
+	   # We try this two ways since rm can't unlink itself on some
+	   # systems and the destination file might be busy for other
+	   # reasons.  In this case, the final cleanup might fail but the new
+	   # file should still install successfully.
+	   {
+	     if test -f "$dstdir/$dstfile"; then
+	       $doit $rmcmd -f "$dstdir/$dstfile" 2>/dev/null \
+	       || $doit $mvcmd -f "$dstdir/$dstfile" "$rmtmp" 2>/dev/null \
+	       || {
+		 echo "$0: cannot unlink or rename $dstdir/$dstfile" >&2
+		 (exit 1); exit
+	       }
+	     else
+	       :
+	     fi
+	   } &&
+
+	   # Now rename the file to the real destination.
+	   $doit $mvcmd "$dsttmp" "$dstdir/$dstfile"
+	 }
+    }
+  fi || { (exit 1); exit; }
+done
+
+# The final little trick to "correctly" pass the exit status to the exit trap.
+{
+  (exit 0); exit
+}
+
+# Local variables:
+# eval: (add-hook 'write-file-hooks 'time-stamp)
+# time-stamp-start: "scriptversion="
+# time-stamp-format: "%:y-%02m-%02d.%02H"
+# time-stamp-end: "$"
+# End:

Propchange: klee/trunk/autoconf/install-sh

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/configure
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/configure?rev=72205&view=auto

==============================================================================
--- klee/trunk/configure (added)
+++ klee/trunk/configure Wed May 20 23:36:41 2009
@@ -0,0 +1,6039 @@
+#! /bin/sh
+# Guess values for system-dependent variables and create Makefiles.
+# Generated by GNU Autoconf 2.60 for KLEE 0.01.
+#
+# Report bugs to <daniel at minormatter.com>.
+#
+# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
+# 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
+# This configure script is free software; the Free Software Foundation
+# gives unlimited permission to copy, distribute and modify it.
+## --------------------- ##
+## M4sh Initialization.  ##
+## --------------------- ##
+
+# Be Bourne compatible
+if test -n "${ZSH_VERSION+set}" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on ${1+"$@"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '${1+"$@"}'='"$@"'
+  setopt NO_GLOB_SUBST
+else
+  case `(set -o) 2>/dev/null` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+
+# PATH needs CR
+# Avoid depending upon Character Ranges.
+as_cr_letters='abcdefghijklmnopqrstuvwxyz'
+as_cr_LETTERS='ABCDEFGHIJKLMNOPQRSTUVWXYZ'
+as_cr_Letters=$as_cr_letters$as_cr_LETTERS
+as_cr_digits='0123456789'
+as_cr_alnum=$as_cr_Letters$as_cr_digits
+
+# The user is always right.
+if test "${PATH_SEPARATOR+set}" != set; then
+  echo "#! /bin/sh" >conf$$.sh
+  echo  "exit 0"   >>conf$$.sh
+  chmod +x conf$$.sh
+  if (PATH="/nonexistent;."; conf$$.sh) >/dev/null 2>&1; then
+    PATH_SEPARATOR=';'
+  else
+    PATH_SEPARATOR=:
+  fi
+  rm -f conf$$.sh
+fi
+
+# Support unset when possible.
+if ( (MAIL=60; unset MAIL) || exit) >/dev/null 2>&1; then
+  as_unset=unset
+else
+  as_unset=false
+fi
+
+
+# IFS
+# We need space, tab and new line, in precisely that order.  Quoting is
+# there to prevent editors from complaining about space-tab.
+# (If _AS_PATH_WALK were called with IFS unset, it would disable word
+# splitting by setting IFS to empty value.)
+as_nl='
+'
+IFS=" ""	$as_nl"
+
+# Find who we are.  Look in the path if we contain no directory separator.
+case $0 in
+  *[\\/]* ) as_myself=$0 ;;
+  *) as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  test -r "$as_dir/$0" && as_myself=$as_dir/$0 && break
+done
+IFS=$as_save_IFS
+
+     ;;
+esac
+# We did not find ourselves, most probably we were run as `sh COMMAND'
+# in which case we are not to be found in the path.
+if test "x$as_myself" = x; then
+  as_myself=$0
+fi
+if test ! -f "$as_myself"; then
+  echo "$as_myself: error: cannot find myself; rerun with an absolute file name" >&2
+  { (exit 1); exit 1; }
+fi
+
+# Work around bugs in pre-3.0 UWIN ksh.
+for as_var in ENV MAIL MAILPATH
+do ($as_unset $as_var) >/dev/null 2>&1 && $as_unset $as_var
+done
+PS1='$ '
+PS2='> '
+PS4='+ '
+
+# NLS nuisances.
+for as_var in \
+  LANG LANGUAGE LC_ADDRESS LC_ALL LC_COLLATE LC_CTYPE LC_IDENTIFICATION \
+  LC_MEASUREMENT LC_MESSAGES LC_MONETARY LC_NAME LC_NUMERIC LC_PAPER \
+  LC_TELEPHONE LC_TIME
+do
+  if (set +x; test -z "`(eval $as_var=C; export $as_var) 2>&1`"); then
+    eval $as_var=C; export $as_var
+  else
+    ($as_unset $as_var) >/dev/null 2>&1 && $as_unset $as_var
+  fi
+done
+
+# Required to use basename.
+if expr a : '\(a\)' >/dev/null 2>&1 &&
+   test "X`expr 00001 : '.*\(...\)'`" = X001; then
+  as_expr=expr
+else
+  as_expr=false
+fi
+
+if (basename -- /) >/dev/null 2>&1 && test "X`basename -- / 2>&1`" = "X/"; then
+  as_basename=basename
+else
+  as_basename=false
+fi
+
+
+# Name of the executable.
+as_me=`$as_basename -- "$0" ||
+$as_expr X/"$0" : '.*/\([^/][^/]*\)/*$' \| \
+	 X"$0" : 'X\(//\)$' \| \
+	 X"$0" : 'X\(/\)' \| . 2>/dev/null ||
+echo X/"$0" |
+    sed '/^.*\/\([^/][^/]*\)\/*$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\/\(\/\/\)$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\/\(\/\).*/{
+	    s//\1/
+	    q
+	  }
+	  s/.*/./; q'`
+
+# CDPATH.
+$as_unset CDPATH
+
+
+if test "x$CONFIG_SHELL" = x; then
+  if (eval ":") 2>/dev/null; then
+  as_have_required=yes
+else
+  as_have_required=no
+fi
+
+  if test $as_have_required = yes && 	 (eval ":
+(as_func_return () {
+  (exit \$1)
+}
+as_func_success () {
+  as_func_return 0
+}
+as_func_failure () {
+  as_func_return 1
+}
+as_func_ret_success () {
+  return 0
+}
+as_func_ret_failure () {
+  return 1
+}
+
+exitcode=0
+if as_func_success; then
+  :
+else
+  exitcode=1
+  echo as_func_success failed.
+fi
+
+if as_func_failure; then
+  exitcode=1
+  echo as_func_failure succeeded.
+fi
+
+if as_func_ret_success; then
+  :
+else
+  exitcode=1
+  echo as_func_ret_success failed.
+fi
+
+if as_func_ret_failure; then
+  exitcode=1
+  echo as_func_ret_failure succeeded.
+fi
+
+if ( set x; as_func_ret_success y && test x = \"\$1\" ); then
+  :
+else
+  exitcode=1
+  echo positional parameters were not saved.
+fi
+
+test \$exitcode = 0) || { (exit 1); exit 1; }
+
+(
+  as_lineno_1=\$LINENO
+  as_lineno_2=\$LINENO
+  test \"x\$as_lineno_1\" != \"x\$as_lineno_2\" &&
+  test \"x\`expr \$as_lineno_1 + 1\`\" = \"x\$as_lineno_2\") || { (exit 1); exit 1; }
+") 2> /dev/null; then
+  :
+else
+  as_candidate_shells=
+    as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in /usr/bin/posix$PATH_SEPARATOR/bin$PATH_SEPARATOR/usr/bin$PATH_SEPARATOR$PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  case $as_dir in
+	 /*)
+	   for as_base in sh bash ksh sh5; do
+	     as_candidate_shells="$as_candidate_shells $as_dir/$as_base"
+	   done;;
+       esac
+done
+IFS=$as_save_IFS
+
+
+      for as_shell in $as_candidate_shells $SHELL; do
+	 # Try only shells that exist, to save several forks.
+	 if { test -f "$as_shell" || test -f "$as_shell.exe"; } &&
+		{ ("$as_shell") 2> /dev/null <<\_ASEOF
+# Be Bourne compatible
+if test -n "${ZSH_VERSION+set}" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on ${1+"$@"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '${1+"$@"}'='"$@"'
+  setopt NO_GLOB_SUBST
+else
+  case `(set -o) 2>/dev/null` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+:
+_ASEOF
+}; then
+  CONFIG_SHELL=$as_shell
+	       as_have_required=yes
+	       if { "$as_shell" 2> /dev/null <<\_ASEOF
+# Be Bourne compatible
+if test -n "${ZSH_VERSION+set}" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on ${1+"$@"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '${1+"$@"}'='"$@"'
+  setopt NO_GLOB_SUBST
+else
+  case `(set -o) 2>/dev/null` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+:
+(as_func_return () {
+  (exit $1)
+}
+as_func_success () {
+  as_func_return 0
+}
+as_func_failure () {
+  as_func_return 1
+}
+as_func_ret_success () {
+  return 0
+}
+as_func_ret_failure () {
+  return 1
+}
+
+exitcode=0
+if as_func_success; then
+  :
+else
+  exitcode=1
+  echo as_func_success failed.
+fi
+
+if as_func_failure; then
+  exitcode=1
+  echo as_func_failure succeeded.
+fi
+
+if as_func_ret_success; then
+  :
+else
+  exitcode=1
+  echo as_func_ret_success failed.
+fi
+
+if as_func_ret_failure; then
+  exitcode=1
+  echo as_func_ret_failure succeeded.
+fi
+
+if ( set x; as_func_ret_success y && test x = "$1" ); then
+  :
+else
+  exitcode=1
+  echo positional parameters were not saved.
+fi
+
+test $exitcode = 0) || { (exit 1); exit 1; }
+
+(
+  as_lineno_1=$LINENO
+  as_lineno_2=$LINENO
+  test "x$as_lineno_1" != "x$as_lineno_2" &&
+  test "x`expr $as_lineno_1 + 1`" = "x$as_lineno_2") || { (exit 1); exit 1; }
+
+_ASEOF
+}; then
+  break
+fi
+
+fi
+
+      done
+
+      if test "x$CONFIG_SHELL" != x; then
+  for as_var in BASH_ENV ENV
+        do ($as_unset $as_var) >/dev/null 2>&1 && $as_unset $as_var
+        done
+        export CONFIG_SHELL
+        exec "$CONFIG_SHELL" "$as_myself" ${1+"$@"}
+fi
+
+
+    if test $as_have_required = no; then
+  echo This script requires a shell more modern than all the
+      echo shells that I found on your system.  Please install a
+      echo modern shell, or manually run the script under such a
+      echo shell if you do have one.
+      { (exit 1); exit 1; }
+fi
+
+
+fi
+
+fi
+
+
+
+(eval "as_func_return () {
+  (exit \$1)
+}
+as_func_success () {
+  as_func_return 0
+}
+as_func_failure () {
+  as_func_return 1
+}
+as_func_ret_success () {
+  return 0
+}
+as_func_ret_failure () {
+  return 1
+}
+
+exitcode=0
+if as_func_success; then
+  :
+else
+  exitcode=1
+  echo as_func_success failed.
+fi
+
+if as_func_failure; then
+  exitcode=1
+  echo as_func_failure succeeded.
+fi
+
+if as_func_ret_success; then
+  :
+else
+  exitcode=1
+  echo as_func_ret_success failed.
+fi
+
+if as_func_ret_failure; then
+  exitcode=1
+  echo as_func_ret_failure succeeded.
+fi
+
+if ( set x; as_func_ret_success y && test x = \"\$1\" ); then
+  :
+else
+  exitcode=1
+  echo positional parameters were not saved.
+fi
+
+test \$exitcode = 0") || {
+  echo No shell found that supports shell functions.
+  echo Please tell autoconf at gnu.org about your system,
+  echo including any error possibly output before this
+  echo message
+}
+
+
+
+  as_lineno_1=$LINENO
+  as_lineno_2=$LINENO
+  test "x$as_lineno_1" != "x$as_lineno_2" &&
+  test "x`expr $as_lineno_1 + 1`" = "x$as_lineno_2" || {
+
+  # Create $as_me.lineno as a copy of $as_myself, but with $LINENO
+  # uniformly replaced by the line number.  The first 'sed' inserts a
+  # line-number line after each line using $LINENO; the second 'sed'
+  # does the real work.  The second script uses 'N' to pair each
+  # line-number line with the line containing $LINENO, and appends
+  # trailing '-' during substitution so that $LINENO is not a special
+  # case at line end.
+  # (Raja R Harinath suggested sed '=', and Paul Eggert wrote the
+  # scripts with optimization help from Paolo Bonzini.  Blame Lee
+  # E. McMahon (1931-1989) for sed's syntax.  :-)
+  sed -n '
+    p
+    /[$]LINENO/=
+  ' <$as_myself |
+    sed '
+      s/[$]LINENO.*/&-/
+      t lineno
+      b
+      :lineno
+      N
+      :loop
+      s/[$]LINENO\([^'$as_cr_alnum'_].*\n\)\(.*\)/\2\1\2/
+      t loop
+      s/-\n.*//
+    ' >$as_me.lineno &&
+  chmod +x "$as_me.lineno" ||
+    { echo "$as_me: error: cannot create $as_me.lineno; rerun with a POSIX shell" >&2
+   { (exit 1); exit 1; }; }
+
+  # Don't try to exec as it changes $[0], causing all sort of problems
+  # (the dirname of $[0] is not the place where we might find the
+  # original and so on.  Autoconf is especially sensitive to this).
+  . "./$as_me.lineno"
+  # Exit status is that of the last command.
+  exit
+}
+
+
+if (as_dir=`dirname -- /` && test "X$as_dir" = X/) >/dev/null 2>&1; then
+  as_dirname=dirname
+else
+  as_dirname=false
+fi
+
+ECHO_C= ECHO_N= ECHO_T=
+case `echo -n x` in
+-n*)
+  case `echo 'x\c'` in
+  *c*) ECHO_T='	';;	# ECHO_T is single tab character.
+  *)   ECHO_C='\c';;
+  esac;;
+*)
+  ECHO_N='-n';;
+esac
+
+if expr a : '\(a\)' >/dev/null 2>&1 &&
+   test "X`expr 00001 : '.*\(...\)'`" = X001; then
+  as_expr=expr
+else
+  as_expr=false
+fi
+
+rm -f conf$$ conf$$.exe conf$$.file
+if test -d conf$$.dir; then
+  rm -f conf$$.dir/conf$$.file
+else
+  rm -f conf$$.dir
+  mkdir conf$$.dir
+fi
+echo >conf$$.file
+if ln -s conf$$.file conf$$ 2>/dev/null; then
+  as_ln_s='ln -s'
+  # ... but there are two gotchas:
+  # 1) On MSYS, both `ln -s file dir' and `ln file dir' fail.
+  # 2) DJGPP < 2.04 has no symlinks; `ln -s' creates a wrapper executable.
+  # In both cases, we have to default to `cp -p'.
+  ln -s conf$$.file conf$$.dir 2>/dev/null && test ! -f conf$$.exe ||
+    as_ln_s='cp -p'
+elif ln conf$$.file conf$$ 2>/dev/null; then
+  as_ln_s=ln
+else
+  as_ln_s='cp -p'
+fi
+rm -f conf$$ conf$$.exe conf$$.dir/conf$$.file conf$$.file
+rmdir conf$$.dir 2>/dev/null
+
+if mkdir -p . 2>/dev/null; then
+  as_mkdir_p=:
+else
+  test -d ./-p && rmdir ./-p
+  as_mkdir_p=false
+fi
+
+# Find out whether ``test -x'' works.  Don't use a zero-byte file, as
+# systems may use methods other than mode bits to determine executability.
+cat >conf$$.file <<_ASEOF
+#! /bin/sh
+exit 0
+_ASEOF
+chmod +x conf$$.file
+if test -x conf$$.file >/dev/null 2>&1; then
+  as_executable_p="test -x"
+else
+  as_executable_p=:
+fi
+rm -f conf$$.file
+
+# Sed expression to map a string onto a valid CPP name.
+as_tr_cpp="eval sed 'y%*$as_cr_letters%P$as_cr_LETTERS%;s%[^_$as_cr_alnum]%_%g'"
+
+# Sed expression to map a string onto a valid variable name.
+as_tr_sh="eval sed 'y%*+%pp%;s%[^_$as_cr_alnum]%_%g'"
+
+
+
+exec 7<&0 </dev/null 6>&1
+
+# Name of the host.
+# hostname on some systems (SVR3.2, Linux) returns a bogus exit status,
+# so uname gets run too.
+ac_hostname=`(hostname || uname -n) 2>/dev/null | sed 1q`
+
+#
+# Initializations.
+#
+ac_default_prefix=/usr/local
+ac_clean_files=
+ac_config_libobj_dir=.
+LIBOBJS=
+cross_compiling=no
+subdirs=
+MFLAGS=
+MAKEFLAGS=
+SHELL=${CONFIG_SHELL-/bin/sh}
+
+# Identity of this package.
+PACKAGE_NAME='KLEE'
+PACKAGE_TARNAME='-klee-'
+PACKAGE_VERSION='0.01'
+PACKAGE_STRING='KLEE 0.01'
+PACKAGE_BUGREPORT='daniel at minormatter.com'
+
+ac_unique_file=""Makefile.config.in""
+# Factoring default headers for most tests.
+ac_includes_default="\
+#include <stdio.h>
+#if HAVE_SYS_TYPES_H
+# include <sys/types.h>
+#endif
+#if HAVE_SYS_STAT_H
+# include <sys/stat.h>
+#endif
+#if STDC_HEADERS
+# include <stdlib.h>
+# include <stddef.h>
+#else
+# if HAVE_STDLIB_H
+#  include <stdlib.h>
+# endif
+#endif
+#if HAVE_STRING_H
+# if !STDC_HEADERS && HAVE_MEMORY_H
+#  include <memory.h>
+# endif
+# include <string.h>
+#endif
+#if HAVE_STRINGS_H
+# include <strings.h>
+#endif
+#if HAVE_INTTYPES_H
+# include <inttypes.h>
+#endif
+#if HAVE_STDINT_H
+# include <stdint.h>
+#endif
+#if HAVE_UNISTD_H
+# include <unistd.h>
+#endif"
+
+ac_subst_vars='SHELL
+PATH_SEPARATOR
+PACKAGE_NAME
+PACKAGE_TARNAME
+PACKAGE_VERSION
+PACKAGE_STRING
+PACKAGE_BUGREPORT
+exec_prefix
+prefix
+program_transform_name
+bindir
+sbindir
+libexecdir
+datarootdir
+datadir
+sysconfdir
+sharedstatedir
+localstatedir
+includedir
+oldincludedir
+docdir
+infodir
+htmldir
+dvidir
+pdfdir
+psdir
+libdir
+localedir
+mandir
+DEFS
+ECHO_C
+ECHO_N
+ECHO_T
+LIBS
+build_alias
+host_alias
+target_alias
+LLVM_SRC
+LLVM_OBJ
+build
+build_cpu
+build_vendor
+build_os
+host
+host_cpu
+host_vendor
+host_os
+target
+target_cpu
+target_vendor
+target_os
+KLEE_UCLIBC
+ENABLE_UCLIBC
+ENABLE_POSIX_RUNTIME
+RUNTIME_ENABLE_OPTIMIZED
+RUNTIME_CONFIGURATION
+CC
+CFLAGS
+LDFLAGS
+CPPFLAGS
+ac_ct_CC
+EXEEXT
+OBJEXT
+CPP
+GREP
+EGREP
+HAVE_SELINUX
+CXX
+CXXFLAGS
+ac_ct_CXX
+CXXCPP
+ENABLE_STPLOG
+LIBOBJS
+LTLIBOBJS'
+ac_subst_files=''
+      ac_precious_vars='build_alias
+host_alias
+target_alias
+CC
+CFLAGS
+LDFLAGS
+CPPFLAGS
+CPP
+CXX
+CXXFLAGS
+CCC
+CXXCPP'
+
+
+# Initialize some variables set by options.
+ac_init_help=
+ac_init_version=false
+# The variables have the same names as the options, with
+# dashes changed to underlines.
+cache_file=/dev/null
+exec_prefix=NONE
+no_create=
+no_recursion=
+prefix=NONE
+program_prefix=NONE
+program_suffix=NONE
+program_transform_name=s,x,x,
+silent=
+site=
+srcdir=
+verbose=
+x_includes=NONE
+x_libraries=NONE
+
+# Installation directory options.
+# These are left unexpanded so users can "make install exec_prefix=/foo"
+# and all the variables that are supposed to be based on exec_prefix
+# by default will actually change.
+# Use braces instead of parens because sh, perl, etc. also accept them.
+# (The list follows the same order as the GNU Coding Standards.)
+bindir='${exec_prefix}/bin'
+sbindir='${exec_prefix}/sbin'
+libexecdir='${exec_prefix}/libexec'
+datarootdir='${prefix}/share'
+datadir='${datarootdir}'
+sysconfdir='${prefix}/etc'
+sharedstatedir='${prefix}/com'
+localstatedir='${prefix}/var'
+includedir='${prefix}/include'
+oldincludedir='/usr/include'
+docdir='${datarootdir}/doc/${PACKAGE_TARNAME}'
+infodir='${datarootdir}/info'
+htmldir='${docdir}'
+dvidir='${docdir}'
+pdfdir='${docdir}'
+psdir='${docdir}'
+libdir='${exec_prefix}/lib'
+localedir='${datarootdir}/locale'
+mandir='${datarootdir}/man'
+
+ac_prev=
+ac_dashdash=
+for ac_option
+do
+  # If the previous option needs an argument, assign it.
+  if test -n "$ac_prev"; then
+    eval $ac_prev=\$ac_option
+    ac_prev=
+    continue
+  fi
+
+  case $ac_option in
+  *=*)	ac_optarg=`expr "X$ac_option" : '[^=]*=\(.*\)'` ;;
+  *)	ac_optarg=yes ;;
+  esac
+
+  # Accept the important Cygnus configure options, so we can diagnose typos.
+
+  case $ac_dashdash$ac_option in
+  --)
+    ac_dashdash=yes ;;
+
+  -bindir | --bindir | --bindi | --bind | --bin | --bi)
+    ac_prev=bindir ;;
+  -bindir=* | --bindir=* | --bindi=* | --bind=* | --bin=* | --bi=*)
+    bindir=$ac_optarg ;;
+
+  -build | --build | --buil | --bui | --bu)
+    ac_prev=build_alias ;;
+  -build=* | --build=* | --buil=* | --bui=* | --bu=*)
+    build_alias=$ac_optarg ;;
+
+  -cache-file | --cache-file | --cache-fil | --cache-fi \
+  | --cache-f | --cache- | --cache | --cach | --cac | --ca | --c)
+    ac_prev=cache_file ;;
+  -cache-file=* | --cache-file=* | --cache-fil=* | --cache-fi=* \
+  | --cache-f=* | --cache-=* | --cache=* | --cach=* | --cac=* | --ca=* | --c=*)
+    cache_file=$ac_optarg ;;
+
+  --config-cache | -C)
+    cache_file=config.cache ;;
+
+  -datadir | --datadir | --datadi | --datad)
+    ac_prev=datadir ;;
+  -datadir=* | --datadir=* | --datadi=* | --datad=*)
+    datadir=$ac_optarg ;;
+
+  -datarootdir | --datarootdir | --datarootdi | --datarootd | --dataroot \
+  | --dataroo | --dataro | --datar)
+    ac_prev=datarootdir ;;
+  -datarootdir=* | --datarootdir=* | --datarootdi=* | --datarootd=* \
+  | --dataroot=* | --dataroo=* | --dataro=* | --datar=*)
+    datarootdir=$ac_optarg ;;
+
+  -disable-* | --disable-*)
+    ac_feature=`expr "x$ac_option" : 'x-*disable-\(.*\)'`
+    # Reject names that are not valid shell variable names.
+    expr "x$ac_feature" : ".*[^-_$as_cr_alnum]" >/dev/null &&
+      { echo "$as_me: error: invalid feature name: $ac_feature" >&2
+   { (exit 1); exit 1; }; }
+    ac_feature=`echo $ac_feature | sed 's/-/_/g'`
+    eval enable_$ac_feature=no ;;
+
+  -docdir | --docdir | --docdi | --doc | --do)
+    ac_prev=docdir ;;
+  -docdir=* | --docdir=* | --docdi=* | --doc=* | --do=*)
+    docdir=$ac_optarg ;;
+
+  -dvidir | --dvidir | --dvidi | --dvid | --dvi | --dv)
+    ac_prev=dvidir ;;
+  -dvidir=* | --dvidir=* | --dvidi=* | --dvid=* | --dvi=* | --dv=*)
+    dvidir=$ac_optarg ;;
+
+  -enable-* | --enable-*)
+    ac_feature=`expr "x$ac_option" : 'x-*enable-\([^=]*\)'`
+    # Reject names that are not valid shell variable names.
+    expr "x$ac_feature" : ".*[^-_$as_cr_alnum]" >/dev/null &&
+      { echo "$as_me: error: invalid feature name: $ac_feature" >&2
+   { (exit 1); exit 1; }; }
+    ac_feature=`echo $ac_feature | sed 's/-/_/g'`
+    eval enable_$ac_feature=\$ac_optarg ;;
+
+  -exec-prefix | --exec_prefix | --exec-prefix | --exec-prefi \
+  | --exec-pref | --exec-pre | --exec-pr | --exec-p | --exec- \
+  | --exec | --exe | --ex)
+    ac_prev=exec_prefix ;;
+  -exec-prefix=* | --exec_prefix=* | --exec-prefix=* | --exec-prefi=* \
+  | --exec-pref=* | --exec-pre=* | --exec-pr=* | --exec-p=* | --exec-=* \
+  | --exec=* | --exe=* | --ex=*)
+    exec_prefix=$ac_optarg ;;
+
+  -gas | --gas | --ga | --g)
+    # Obsolete; use --with-gas.
+    with_gas=yes ;;
+
+  -help | --help | --hel | --he | -h)
+    ac_init_help=long ;;
+  -help=r* | --help=r* | --hel=r* | --he=r* | -hr*)
+    ac_init_help=recursive ;;
+  -help=s* | --help=s* | --hel=s* | --he=s* | -hs*)
+    ac_init_help=short ;;
+
+  -host | --host | --hos | --ho)
+    ac_prev=host_alias ;;
+  -host=* | --host=* | --hos=* | --ho=*)
+    host_alias=$ac_optarg ;;
+
+  -htmldir | --htmldir | --htmldi | --htmld | --html | --htm | --ht)
+    ac_prev=htmldir ;;
+  -htmldir=* | --htmldir=* | --htmldi=* | --htmld=* | --html=* | --htm=* \
+  | --ht=*)
+    htmldir=$ac_optarg ;;
+
+  -includedir | --includedir | --includedi | --included | --include \
+  | --includ | --inclu | --incl | --inc)
+    ac_prev=includedir ;;
+  -includedir=* | --includedir=* | --includedi=* | --included=* | --include=* \
+  | --includ=* | --inclu=* | --incl=* | --inc=*)
+    includedir=$ac_optarg ;;
+
+  -infodir | --infodir | --infodi | --infod | --info | --inf)
+    ac_prev=infodir ;;
+  -infodir=* | --infodir=* | --infodi=* | --infod=* | --info=* | --inf=*)
+    infodir=$ac_optarg ;;
+
+  -libdir | --libdir | --libdi | --libd)
+    ac_prev=libdir ;;
+  -libdir=* | --libdir=* | --libdi=* | --libd=*)
+    libdir=$ac_optarg ;;
+
+  -libexecdir | --libexecdir | --libexecdi | --libexecd | --libexec \
+  | --libexe | --libex | --libe)
+    ac_prev=libexecdir ;;
+  -libexecdir=* | --libexecdir=* | --libexecdi=* | --libexecd=* | --libexec=* \
+  | --libexe=* | --libex=* | --libe=*)
+    libexecdir=$ac_optarg ;;
+
+  -localedir | --localedir | --localedi | --localed | --locale)
+    ac_prev=localedir ;;
+  -localedir=* | --localedir=* | --localedi=* | --localed=* | --locale=*)
+    localedir=$ac_optarg ;;
+
+  -localstatedir | --localstatedir | --localstatedi | --localstated \
+  | --localstate | --localstat | --localsta | --localst | --locals)
+    ac_prev=localstatedir ;;
+  -localstatedir=* | --localstatedir=* | --localstatedi=* | --localstated=* \
+  | --localstate=* | --localstat=* | --localsta=* | --localst=* | --locals=*)
+    localstatedir=$ac_optarg ;;
+
+  -mandir | --mandir | --mandi | --mand | --man | --ma | --m)
+    ac_prev=mandir ;;
+  -mandir=* | --mandir=* | --mandi=* | --mand=* | --man=* | --ma=* | --m=*)
+    mandir=$ac_optarg ;;
+
+  -nfp | --nfp | --nf)
+    # Obsolete; use --without-fp.
+    with_fp=no ;;
+
+  -no-create | --no-create | --no-creat | --no-crea | --no-cre \
+  | --no-cr | --no-c | -n)
+    no_create=yes ;;
+
+  -no-recursion | --no-recursion | --no-recursio | --no-recursi \
+  | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r)
+    no_recursion=yes ;;
+
+  -oldincludedir | --oldincludedir | --oldincludedi | --oldincluded \
+  | --oldinclude | --oldinclud | --oldinclu | --oldincl | --oldinc \
+  | --oldin | --oldi | --old | --ol | --o)
+    ac_prev=oldincludedir ;;
+  -oldincludedir=* | --oldincludedir=* | --oldincludedi=* | --oldincluded=* \
+  | --oldinclude=* | --oldinclud=* | --oldinclu=* | --oldincl=* | --oldinc=* \
+  | --oldin=* | --oldi=* | --old=* | --ol=* | --o=*)
+    oldincludedir=$ac_optarg ;;
+
+  -prefix | --prefix | --prefi | --pref | --pre | --pr | --p)
+    ac_prev=prefix ;;
+  -prefix=* | --prefix=* | --prefi=* | --pref=* | --pre=* | --pr=* | --p=*)
+    prefix=$ac_optarg ;;
+
+  -program-prefix | --program-prefix | --program-prefi | --program-pref \
+  | --program-pre | --program-pr | --program-p)
+    ac_prev=program_prefix ;;
+  -program-prefix=* | --program-prefix=* | --program-prefi=* \
+  | --program-pref=* | --program-pre=* | --program-pr=* | --program-p=*)
+    program_prefix=$ac_optarg ;;
+
+  -program-suffix | --program-suffix | --program-suffi | --program-suff \
+  | --program-suf | --program-su | --program-s)
+    ac_prev=program_suffix ;;
+  -program-suffix=* | --program-suffix=* | --program-suffi=* \
+  | --program-suff=* | --program-suf=* | --program-su=* | --program-s=*)
+    program_suffix=$ac_optarg ;;
+
+  -program-transform-name | --program-transform-name \
+  | --program-transform-nam | --program-transform-na \
+  | --program-transform-n | --program-transform- \
+  | --program-transform | --program-transfor \
+  | --program-transfo | --program-transf \
+  | --program-trans | --program-tran \
+  | --progr-tra | --program-tr | --program-t)
+    ac_prev=program_transform_name ;;
+  -program-transform-name=* | --program-transform-name=* \
+  | --program-transform-nam=* | --program-transform-na=* \
+  | --program-transform-n=* | --program-transform-=* \
+  | --program-transform=* | --program-transfor=* \
+  | --program-transfo=* | --program-transf=* \
+  | --program-trans=* | --program-tran=* \
+  | --progr-tra=* | --program-tr=* | --program-t=*)
+    program_transform_name=$ac_optarg ;;
+
+  -pdfdir | --pdfdir | --pdfdi | --pdfd | --pdf | --pd)
+    ac_prev=pdfdir ;;
+  -pdfdir=* | --pdfdir=* | --pdfdi=* | --pdfd=* | --pdf=* | --pd=*)
+    pdfdir=$ac_optarg ;;
+
+  -psdir | --psdir | --psdi | --psd | --ps)
+    ac_prev=psdir ;;
+  -psdir=* | --psdir=* | --psdi=* | --psd=* | --ps=*)
+    psdir=$ac_optarg ;;
+
+  -q | -quiet | --quiet | --quie | --qui | --qu | --q \
+  | -silent | --silent | --silen | --sile | --sil)
+    silent=yes ;;
+
+  -sbindir | --sbindir | --sbindi | --sbind | --sbin | --sbi | --sb)
+    ac_prev=sbindir ;;
+  -sbindir=* | --sbindir=* | --sbindi=* | --sbind=* | --sbin=* \
+  | --sbi=* | --sb=*)
+    sbindir=$ac_optarg ;;
+
+  -sharedstatedir | --sharedstatedir | --sharedstatedi \
+  | --sharedstated | --sharedstate | --sharedstat | --sharedsta \
+  | --sharedst | --shareds | --shared | --share | --shar \
+  | --sha | --sh)
+    ac_prev=sharedstatedir ;;
+  -sharedstatedir=* | --sharedstatedir=* | --sharedstatedi=* \
+  | --sharedstated=* | --sharedstate=* | --sharedstat=* | --sharedsta=* \
+  | --sharedst=* | --shareds=* | --shared=* | --share=* | --shar=* \
+  | --sha=* | --sh=*)
+    sharedstatedir=$ac_optarg ;;
+
+  -site | --site | --sit)
+    ac_prev=site ;;
+  -site=* | --site=* | --sit=*)
+    site=$ac_optarg ;;
+
+  -srcdir | --srcdir | --srcdi | --srcd | --src | --sr)
+    ac_prev=srcdir ;;
+  -srcdir=* | --srcdir=* | --srcdi=* | --srcd=* | --src=* | --sr=*)
+    srcdir=$ac_optarg ;;
+
+  -sysconfdir | --sysconfdir | --sysconfdi | --sysconfd | --sysconf \
+  | --syscon | --sysco | --sysc | --sys | --sy)
+    ac_prev=sysconfdir ;;
+  -sysconfdir=* | --sysconfdir=* | --sysconfdi=* | --sysconfd=* | --sysconf=* \
+  | --syscon=* | --sysco=* | --sysc=* | --sys=* | --sy=*)
+    sysconfdir=$ac_optarg ;;
+
+  -target | --target | --targe | --targ | --tar | --ta | --t)
+    ac_prev=target_alias ;;
+  -target=* | --target=* | --targe=* | --targ=* | --tar=* | --ta=* | --t=*)
+    target_alias=$ac_optarg ;;
+
+  -v | -verbose | --verbose | --verbos | --verbo | --verb)
+    verbose=yes ;;
+
+  -version | --version | --versio | --versi | --vers | -V)
+    ac_init_version=: ;;
+
+  -with-* | --with-*)
+    ac_package=`expr "x$ac_option" : 'x-*with-\([^=]*\)'`
+    # Reject names that are not valid shell variable names.
+    expr "x$ac_package" : ".*[^-_$as_cr_alnum]" >/dev/null &&
+      { echo "$as_me: error: invalid package name: $ac_package" >&2
+   { (exit 1); exit 1; }; }
+    ac_package=`echo $ac_package| sed 's/-/_/g'`
+    eval with_$ac_package=\$ac_optarg ;;
+
+  -without-* | --without-*)
+    ac_package=`expr "x$ac_option" : 'x-*without-\(.*\)'`
+    # Reject names that are not valid shell variable names.
+    expr "x$ac_package" : ".*[^-_$as_cr_alnum]" >/dev/null &&
+      { echo "$as_me: error: invalid package name: $ac_package" >&2
+   { (exit 1); exit 1; }; }
+    ac_package=`echo $ac_package | sed 's/-/_/g'`
+    eval with_$ac_package=no ;;
+
+  --x)
+    # Obsolete; use --with-x.
+    with_x=yes ;;
+
+  -x-includes | --x-includes | --x-include | --x-includ | --x-inclu \
+  | --x-incl | --x-inc | --x-in | --x-i)
+    ac_prev=x_includes ;;
+  -x-includes=* | --x-includes=* | --x-include=* | --x-includ=* | --x-inclu=* \
+  | --x-incl=* | --x-inc=* | --x-in=* | --x-i=*)
+    x_includes=$ac_optarg ;;
+
+  -x-libraries | --x-libraries | --x-librarie | --x-librari \
+  | --x-librar | --x-libra | --x-libr | --x-lib | --x-li | --x-l)
+    ac_prev=x_libraries ;;
+  -x-libraries=* | --x-libraries=* | --x-librarie=* | --x-librari=* \
+  | --x-librar=* | --x-libra=* | --x-libr=* | --x-lib=* | --x-li=* | --x-l=*)
+    x_libraries=$ac_optarg ;;
+
+  -*) { echo "$as_me: error: unrecognized option: $ac_option
+Try \`$0 --help' for more information." >&2
+   { (exit 1); exit 1; }; }
+    ;;
+
+  *=*)
+    ac_envvar=`expr "x$ac_option" : 'x\([^=]*\)='`
+    # Reject names that are not valid shell variable names.
+    expr "x$ac_envvar" : ".*[^_$as_cr_alnum]" >/dev/null &&
+      { echo "$as_me: error: invalid variable name: $ac_envvar" >&2
+   { (exit 1); exit 1; }; }
+    eval $ac_envvar=\$ac_optarg
+    export $ac_envvar ;;
+
+  *)
+    # FIXME: should be removed in autoconf 3.0.
+    echo "$as_me: WARNING: you should use --build, --host, --target" >&2
+    expr "x$ac_option" : ".*[^-._$as_cr_alnum]" >/dev/null &&
+      echo "$as_me: WARNING: invalid host type: $ac_option" >&2
+    : ${build_alias=$ac_option} ${host_alias=$ac_option} ${target_alias=$ac_option}
+    ;;
+
+  esac
+done
+
+if test -n "$ac_prev"; then
+  ac_option=--`echo $ac_prev | sed 's/_/-/g'`
+  { echo "$as_me: error: missing argument to $ac_option" >&2
+   { (exit 1); exit 1; }; }
+fi
+
+# Be sure to have absolute directory names.
+for ac_var in	exec_prefix prefix bindir sbindir libexecdir datarootdir \
+		datadir sysconfdir sharedstatedir localstatedir includedir \
+		oldincludedir docdir infodir htmldir dvidir pdfdir psdir \
+		libdir localedir mandir
+do
+  eval ac_val=\$$ac_var
+  case $ac_val in
+    [\\/$]* | ?:[\\/]* )  continue;;
+    NONE | '' ) case $ac_var in *prefix ) continue;; esac;;
+  esac
+  { echo "$as_me: error: expected an absolute directory name for --$ac_var: $ac_val" >&2
+   { (exit 1); exit 1; }; }
+done
+
+# There might be people who depend on the old broken behavior: `$host'
+# used to hold the argument of --host etc.
+# FIXME: To remove some day.
+build=$build_alias
+host=$host_alias
+target=$target_alias
+
+# FIXME: To remove some day.
+if test "x$host_alias" != x; then
+  if test "x$build_alias" = x; then
+    cross_compiling=maybe
+    echo "$as_me: WARNING: If you wanted to set the --build type, don't use --host.
+    If a cross compiler is detected then cross compile mode will be used." >&2
+  elif test "x$build_alias" != "x$host_alias"; then
+    cross_compiling=yes
+  fi
+fi
+
+ac_tool_prefix=
+test -n "$host_alias" && ac_tool_prefix=$host_alias-
+
+test "$silent" = yes && exec 6>/dev/null
+
+
+ac_pwd=`pwd` && test -n "$ac_pwd" &&
+ac_ls_di=`ls -di .` &&
+ac_pwd_ls_di=`cd "$ac_pwd" && ls -di .` ||
+  { echo "$as_me: error: Working directory cannot be determined" >&2
+   { (exit 1); exit 1; }; }
+test "X$ac_ls_di" = "X$ac_pwd_ls_di" ||
+  { echo "$as_me: error: pwd does not report name of working directory" >&2
+   { (exit 1); exit 1; }; }
+
+
+# Find the source files, if location was not specified.
+if test -z "$srcdir"; then
+  ac_srcdir_defaulted=yes
+  # Try the directory containing this script, then the parent directory.
+  ac_confdir=`$as_dirname -- "$0" ||
+$as_expr X"$0" : 'X\(.*[^/]\)//*[^/][^/]*/*$' \| \
+	 X"$0" : 'X\(//\)[^/]' \| \
+	 X"$0" : 'X\(//\)$' \| \
+	 X"$0" : 'X\(/\)' \| . 2>/dev/null ||
+echo X"$0" |
+    sed '/^X\(.*[^/]\)\/\/*[^/][^/]*\/*$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)[^/].*/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\).*/{
+	    s//\1/
+	    q
+	  }
+	  s/.*/./; q'`
+  srcdir=$ac_confdir
+  if test ! -r "$srcdir/$ac_unique_file"; then
+    srcdir=..
+  fi
+else
+  ac_srcdir_defaulted=no
+fi
+if test ! -r "$srcdir/$ac_unique_file"; then
+  test "$ac_srcdir_defaulted" = yes && srcdir="$ac_confdir or .."
+  { echo "$as_me: error: cannot find sources ($ac_unique_file) in $srcdir" >&2
+   { (exit 1); exit 1; }; }
+fi
+ac_msg="sources are in $srcdir, but \`cd $srcdir' does not work"
+ac_abs_confdir=`(
+	cd "$srcdir" && test -r "./$ac_unique_file" || { echo "$as_me: error: $ac_msg" >&2
+   { (exit 1); exit 1; }; }
+	pwd)`
+# When building in place, set srcdir=.
+if test "$ac_abs_confdir" = "$ac_pwd"; then
+  srcdir=.
+fi
+# Remove unnecessary trailing slashes from srcdir.
+# Double slashes in file names in object file debugging info
+# mess up M-x gdb in Emacs.
+case $srcdir in
+*/) srcdir=`expr "X$srcdir" : 'X\(.*[^/]\)' \| "X$srcdir" : 'X\(.*\)'`;;
+esac
+for ac_var in $ac_precious_vars; do
+  eval ac_env_${ac_var}_set=\${${ac_var}+set}
+  eval ac_env_${ac_var}_value=\$${ac_var}
+  eval ac_cv_env_${ac_var}_set=\${${ac_var}+set}
+  eval ac_cv_env_${ac_var}_value=\$${ac_var}
+done
+
+#
+# Report the --help message.
+#
+if test "$ac_init_help" = "long"; then
+  # Omit some internal or obsolete options to make the list less imposing.
+  # This message is too long to be a string in the A/UX 3.1 sh.
+  cat <<_ACEOF
+\`configure' configures KLEE 0.01 to adapt to many kinds of systems.
+
+Usage: $0 [OPTION]... [VAR=VALUE]...
+
+To assign environment variables (e.g., CC, CFLAGS...), specify them as
+VAR=VALUE.  See below for descriptions of some of the useful variables.
+
+Defaults for the options are specified in brackets.
+
+Configuration:
+  -h, --help              display this help and exit
+      --help=short        display options specific to this package
+      --help=recursive    display the short help of all the included packages
+  -V, --version           display version information and exit
+  -q, --quiet, --silent   do not print \`checking...' messages
+      --cache-file=FILE   cache test results in FILE [disabled]
+  -C, --config-cache      alias for \`--cache-file=config.cache'
+  -n, --no-create         do not create output files
+      --srcdir=DIR        find the sources in DIR [configure dir or \`..']
+
+Installation directories:
+  --prefix=PREFIX         install architecture-independent files in PREFIX
+			  [$ac_default_prefix]
+  --exec-prefix=EPREFIX   install architecture-dependent files in EPREFIX
+			  [PREFIX]
+
+By default, \`make install' will install all the files in
+\`$ac_default_prefix/bin', \`$ac_default_prefix/lib' etc.  You can specify
+an installation prefix other than \`$ac_default_prefix' using \`--prefix',
+for instance \`--prefix=\$HOME'.
+
+For better control, use the options below.
+
+Fine tuning of the installation directories:
+  --bindir=DIR           user executables [EPREFIX/bin]
+  --sbindir=DIR          system admin executables [EPREFIX/sbin]
+  --libexecdir=DIR       program executables [EPREFIX/libexec]
+  --sysconfdir=DIR       read-only single-machine data [PREFIX/etc]
+  --sharedstatedir=DIR   modifiable architecture-independent data [PREFIX/com]
+  --localstatedir=DIR    modifiable single-machine data [PREFIX/var]
+  --libdir=DIR           object code libraries [EPREFIX/lib]
+  --includedir=DIR       C header files [PREFIX/include]
+  --oldincludedir=DIR    C header files for non-gcc [/usr/include]
+  --datarootdir=DIR      read-only arch.-independent data root [PREFIX/share]
+  --datadir=DIR          read-only architecture-independent data [DATAROOTDIR]
+  --infodir=DIR          info documentation [DATAROOTDIR/info]
+  --localedir=DIR        locale-dependent data [DATAROOTDIR/locale]
+  --mandir=DIR           man documentation [DATAROOTDIR/man]
+  --docdir=DIR           documentation root [DATAROOTDIR/doc/-klee-]
+  --htmldir=DIR          html documentation [DOCDIR]
+  --dvidir=DIR           dvi documentation [DOCDIR]
+  --pdfdir=DIR           pdf documentation [DOCDIR]
+  --psdir=DIR            ps documentation [DOCDIR]
+_ACEOF
+
+  cat <<\_ACEOF
+
+System types:
+  --build=BUILD     configure for building on BUILD [guessed]
+  --host=HOST       cross-compile to build programs to run on HOST [BUILD]
+  --target=TARGET   configure for building compilers for TARGET [HOST]
+_ACEOF
+fi
+
+if test -n "$ac_init_help"; then
+  case $ac_init_help in
+     short | recursive ) echo "Configuration of KLEE 0.01:";;
+   esac
+  cat <<\_ACEOF
+
+Optional Features:
+  --disable-FEATURE       do not include FEATURE (same as --enable-FEATURE=no)
+  --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
+  --enable-posix-runtime  Enable the POSIX runtime
+  --enable-stplog         Compile with the stplog library [disabled]
+
+Optional Packages:
+  --with-PACKAGE[=ARG]    use PACKAGE [ARG=yes]
+  --without-PACKAGE       do not use PACKAGE (same as --with-PACKAGE=no)
+  --with-llvmsrc          Location of LLVM Source Code
+  --with-llvmobj          Location of LLVM Object Code
+  --with-llvm             Location of LLVM Source and Object code
+  --with-uclibc           Enable use of the klee uclibc at the given path
+  --with-runtime          Select build configuration for runtime libraries
+                          (default Release)
+
+Some influential environment variables:
+  CC          C compiler command
+  CFLAGS      C compiler flags
+  LDFLAGS     linker flags, e.g. -L<lib dir> if you have libraries in a
+              nonstandard directory <lib dir>
+  CPPFLAGS    C/C++/Objective C preprocessor flags, e.g. -I<include dir> if
+              you have headers in a nonstandard directory <include dir>
+  CPP         C preprocessor
+  CXX         C++ compiler command
+  CXXFLAGS    C++ compiler flags
+  CXXCPP      C++ preprocessor
+
+Use these variables to override the choices made by `configure' or to help
+it to find libraries and programs with nonstandard names/locations.
+
+Report bugs to <daniel at minormatter.com>.
+_ACEOF
+ac_status=$?
+fi
+
+if test "$ac_init_help" = "recursive"; then
+  # If there are subdirs, report their specific --help.
+  for ac_dir in : $ac_subdirs_all; do test "x$ac_dir" = x: && continue
+    test -d "$ac_dir" || continue
+    ac_builddir=.
+
+case "$ac_dir" in
+.) ac_dir_suffix= ac_top_builddir_sub=. ac_top_build_prefix= ;;
+*)
+  ac_dir_suffix=/`echo "$ac_dir" | sed 's,^\.[\\/],,'`
+  # A ".." for each directory in $ac_dir_suffix.
+  ac_top_builddir_sub=`echo "$ac_dir_suffix" | sed 's,/[^\\/]*,/..,g;s,/,,'`
+  case $ac_top_builddir_sub in
+  "") ac_top_builddir_sub=. ac_top_build_prefix= ;;
+  *)  ac_top_build_prefix=$ac_top_builddir_sub/ ;;
+  esac ;;
+esac
+ac_abs_top_builddir=$ac_pwd
+ac_abs_builddir=$ac_pwd$ac_dir_suffix
+# for backward compatibility:
+ac_top_builddir=$ac_top_build_prefix
+
+case $srcdir in
+  .)  # We are building in place.
+    ac_srcdir=.
+    ac_top_srcdir=$ac_top_builddir_sub
+    ac_abs_top_srcdir=$ac_pwd ;;
+  [\\/]* | ?:[\\/]* )  # Absolute name.
+    ac_srcdir=$srcdir$ac_dir_suffix;
+    ac_top_srcdir=$srcdir
+    ac_abs_top_srcdir=$srcdir ;;
+  *) # Relative name.
+    ac_srcdir=$ac_top_build_prefix$srcdir$ac_dir_suffix
+    ac_top_srcdir=$ac_top_build_prefix$srcdir
+    ac_abs_top_srcdir=$ac_pwd/$srcdir ;;
+esac
+ac_abs_srcdir=$ac_abs_top_srcdir$ac_dir_suffix
+
+    cd "$ac_dir" || { ac_status=$?; continue; }
+    # Check for guested configure.
+    if test -f "$ac_srcdir/configure.gnu"; then
+      echo &&
+      $SHELL "$ac_srcdir/configure.gnu" --help=recursive
+    elif test -f "$ac_srcdir/configure"; then
+      echo &&
+      $SHELL "$ac_srcdir/configure" --help=recursive
+    else
+      echo "$as_me: WARNING: no configuration information is in $ac_dir" >&2
+    fi || ac_status=$?
+    cd "$ac_pwd" || { ac_status=$?; break; }
+  done
+fi
+
+test -n "$ac_init_help" && exit $ac_status
+if $ac_init_version; then
+  cat <<\_ACEOF
+KLEE configure 0.01
+generated by GNU Autoconf 2.60
+
+Copyright (C) 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001,
+2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
+This configure script is free software; the Free Software Foundation
+gives unlimited permission to copy, distribute and modify it.
+_ACEOF
+  exit
+fi
+cat >config.log <<_ACEOF
+This file contains any messages produced by compilers while
+running configure, to aid debugging if configure makes a mistake.
+
+It was created by KLEE $as_me 0.01, which was
+generated by GNU Autoconf 2.60.  Invocation command line was
+
+  $ $0 $@
+
+_ACEOF
+exec 5>>config.log
+{
+cat <<_ASUNAME
+## --------- ##
+## Platform. ##
+## --------- ##
+
+hostname = `(hostname || uname -n) 2>/dev/null | sed 1q`
+uname -m = `(uname -m) 2>/dev/null || echo unknown`
+uname -r = `(uname -r) 2>/dev/null || echo unknown`
+uname -s = `(uname -s) 2>/dev/null || echo unknown`
+uname -v = `(uname -v) 2>/dev/null || echo unknown`
+
+/usr/bin/uname -p = `(/usr/bin/uname -p) 2>/dev/null || echo unknown`
+/bin/uname -X     = `(/bin/uname -X) 2>/dev/null     || echo unknown`
+
+/bin/arch              = `(/bin/arch) 2>/dev/null              || echo unknown`
+/usr/bin/arch -k       = `(/usr/bin/arch -k) 2>/dev/null       || echo unknown`
+/usr/convex/getsysinfo = `(/usr/convex/getsysinfo) 2>/dev/null || echo unknown`
+/usr/bin/hostinfo      = `(/usr/bin/hostinfo) 2>/dev/null      || echo unknown`
+/bin/machine           = `(/bin/machine) 2>/dev/null           || echo unknown`
+/usr/bin/oslevel       = `(/usr/bin/oslevel) 2>/dev/null       || echo unknown`
+/bin/universe          = `(/bin/universe) 2>/dev/null          || echo unknown`
+
+_ASUNAME
+
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  echo "PATH: $as_dir"
+done
+IFS=$as_save_IFS
+
+} >&5
+
+cat >&5 <<_ACEOF
+
+
+## ----------- ##
+## Core tests. ##
+## ----------- ##
+
+_ACEOF
+
+
+# Keep a trace of the command line.
+# Strip out --no-create and --no-recursion so they do not pile up.
+# Strip out --silent because we don't want to record it for future runs.
+# Also quote any args containing shell meta-characters.
+# Make two passes to allow for proper duplicate-argument suppression.
+ac_configure_args=
+ac_configure_args0=
+ac_configure_args1=
+ac_must_keep_next=false
+for ac_pass in 1 2
+do
+  for ac_arg
+  do
+    case $ac_arg in
+    -no-create | --no-c* | -n | -no-recursion | --no-r*) continue ;;
+    -q | -quiet | --quiet | --quie | --qui | --qu | --q \
+    | -silent | --silent | --silen | --sile | --sil)
+      continue ;;
+    *\'*)
+      ac_arg=`echo "$ac_arg" | sed "s/'/'\\\\\\\\''/g"` ;;
+    esac
+    case $ac_pass in
+    1) ac_configure_args0="$ac_configure_args0 '$ac_arg'" ;;
+    2)
+      ac_configure_args1="$ac_configure_args1 '$ac_arg'"
+      if test $ac_must_keep_next = true; then
+	ac_must_keep_next=false # Got value, back to normal.
+      else
+	case $ac_arg in
+	  *=* | --config-cache | -C | -disable-* | --disable-* \
+	  | -enable-* | --enable-* | -gas | --g* | -nfp | --nf* \
+	  | -q | -quiet | --q* | -silent | --sil* | -v | -verb* \
+	  | -with-* | --with-* | -without-* | --without-* | --x)
+	    case "$ac_configure_args0 " in
+	      "$ac_configure_args1"*" '$ac_arg' "* ) continue ;;
+	    esac
+	    ;;
+	  -* ) ac_must_keep_next=true ;;
+	esac
+      fi
+      ac_configure_args="$ac_configure_args '$ac_arg'"
+      ;;
+    esac
+  done
+done
+$as_unset ac_configure_args0 || test "${ac_configure_args0+set}" != set || { ac_configure_args0=; export ac_configure_args0; }
+$as_unset ac_configure_args1 || test "${ac_configure_args1+set}" != set || { ac_configure_args1=; export ac_configure_args1; }
+
+# When interrupted or exit'd, cleanup temporary files, and complete
+# config.log.  We remove comments because anyway the quotes in there
+# would cause problems or look ugly.
+# WARNING: Use '\'' to represent an apostrophe within the trap.
+# WARNING: Do not start the trap code with a newline, due to a FreeBSD 4.0 bug.
+trap 'exit_status=$?
+  # Save into config.log some information that might help in debugging.
+  {
+    echo
+
+    cat <<\_ASBOX
+## ---------------- ##
+## Cache variables. ##
+## ---------------- ##
+_ASBOX
+    echo
+    # The following way of writing the cache mishandles newlines in values,
+(
+  for ac_var in `(set) 2>&1 | sed -n '\''s/^\([a-zA-Z_][a-zA-Z0-9_]*\)=.*/\1/p'\''`; do
+    eval ac_val=\$$ac_var
+    case $ac_val in #(
+    *${as_nl}*)
+      case $ac_var in #(
+      *_cv_*) { echo "$as_me:$LINENO: WARNING: Cache variable $ac_var contains a newline." >&5
+echo "$as_me: WARNING: Cache variable $ac_var contains a newline." >&2;} ;;
+      esac
+      case $ac_var in #(
+      _ | IFS | as_nl) ;; #(
+      *) $as_unset $ac_var ;;
+      esac ;;
+    esac
+  done
+  (set) 2>&1 |
+    case $as_nl`(ac_space='\'' '\''; set) 2>&1` in #(
+    *${as_nl}ac_space=\ *)
+      sed -n \
+	"s/'\''/'\''\\\\'\'''\''/g;
+	  s/^\\([_$as_cr_alnum]*_cv_[_$as_cr_alnum]*\\)=\\(.*\\)/\\1='\''\\2'\''/p"
+      ;; #(
+    *)
+      sed -n "/^[_$as_cr_alnum]*_cv_[_$as_cr_alnum]*=/p"
+      ;;
+    esac |
+    sort
+)
+    echo
+
+    cat <<\_ASBOX
+## ----------------- ##
+## Output variables. ##
+## ----------------- ##
+_ASBOX
+    echo
+    for ac_var in $ac_subst_vars
+    do
+      eval ac_val=\$$ac_var
+      case $ac_val in
+      *\'\''*) ac_val=`echo "$ac_val" | sed "s/'\''/'\''\\\\\\\\'\'''\''/g"`;;
+      esac
+      echo "$ac_var='\''$ac_val'\''"
+    done | sort
+    echo
+
+    if test -n "$ac_subst_files"; then
+      cat <<\_ASBOX
+## ------------------- ##
+## File substitutions. ##
+## ------------------- ##
+_ASBOX
+      echo
+      for ac_var in $ac_subst_files
+      do
+	eval ac_val=\$$ac_var
+	case $ac_val in
+	*\'\''*) ac_val=`echo "$ac_val" | sed "s/'\''/'\''\\\\\\\\'\'''\''/g"`;;
+	esac
+	echo "$ac_var='\''$ac_val'\''"
+      done | sort
+      echo
+    fi
+
+    if test -s confdefs.h; then
+      cat <<\_ASBOX
+## ----------- ##
+## confdefs.h. ##
+## ----------- ##
+_ASBOX
+      echo
+      cat confdefs.h
+      echo
+    fi
+    test "$ac_signal" != 0 &&
+      echo "$as_me: caught signal $ac_signal"
+    echo "$as_me: exit $exit_status"
+  } >&5
+  rm -f core *.core core.conftest.* &&
+    rm -f -r conftest* confdefs* conf$$* $ac_clean_files &&
+    exit $exit_status
+' 0
+for ac_signal in 1 2 13 15; do
+  trap 'ac_signal='$ac_signal'; { (exit 1); exit 1; }' $ac_signal
+done
+ac_signal=0
+
+# confdefs.h avoids OS command line length limits that DEFS can exceed.
+rm -f -r conftest* confdefs.h
+
+# Predefined preprocessor variables.
+
+cat >>confdefs.h <<_ACEOF
+#define PACKAGE_NAME "$PACKAGE_NAME"
+_ACEOF
+
+
+cat >>confdefs.h <<_ACEOF
+#define PACKAGE_TARNAME "$PACKAGE_TARNAME"
+_ACEOF
+
+
+cat >>confdefs.h <<_ACEOF
+#define PACKAGE_VERSION "$PACKAGE_VERSION"
+_ACEOF
+
+
+cat >>confdefs.h <<_ACEOF
+#define PACKAGE_STRING "$PACKAGE_STRING"
+_ACEOF
+
+
+cat >>confdefs.h <<_ACEOF
+#define PACKAGE_BUGREPORT "$PACKAGE_BUGREPORT"
+_ACEOF
+
+
+# Let the site file select an alternate cache file if it wants to.
+# Prefer explicitly selected file to automatically selected ones.
+if test -n "$CONFIG_SITE"; then
+  set x "$CONFIG_SITE"
+elif test "x$prefix" != xNONE; then
+  set x "$prefix/share/config.site" "$prefix/etc/config.site"
+else
+  set x "$ac_default_prefix/share/config.site" \
+	"$ac_default_prefix/etc/config.site"
+fi
+shift
+for ac_site_file
+do
+  if test -r "$ac_site_file"; then
+    { echo "$as_me:$LINENO: loading site script $ac_site_file" >&5
+echo "$as_me: loading site script $ac_site_file" >&6;}
+    sed 's/^/| /' "$ac_site_file" >&5
+    . "$ac_site_file"
+  fi
+done
+
+if test -r "$cache_file"; then
+  # Some versions of bash will fail to source /dev/null (special
+  # files actually), so we avoid doing that.
+  if test -f "$cache_file"; then
+    { echo "$as_me:$LINENO: loading cache $cache_file" >&5
+echo "$as_me: loading cache $cache_file" >&6;}
+    case $cache_file in
+      [\\/]* | ?:[\\/]* ) . "$cache_file";;
+      *)                      . "./$cache_file";;
+    esac
+  fi
+else
+  { echo "$as_me:$LINENO: creating cache $cache_file" >&5
+echo "$as_me: creating cache $cache_file" >&6;}
+  >$cache_file
+fi
+
+# Check that the precious variables saved in the cache have kept the same
+# value.
+ac_cache_corrupted=false
+for ac_var in $ac_precious_vars; do
+  eval ac_old_set=\$ac_cv_env_${ac_var}_set
+  eval ac_new_set=\$ac_env_${ac_var}_set
+  eval ac_old_val=\$ac_cv_env_${ac_var}_value
+  eval ac_new_val=\$ac_env_${ac_var}_value
+  case $ac_old_set,$ac_new_set in
+    set,)
+      { echo "$as_me:$LINENO: error: \`$ac_var' was set to \`$ac_old_val' in the previous run" >&5
+echo "$as_me: error: \`$ac_var' was set to \`$ac_old_val' in the previous run" >&2;}
+      ac_cache_corrupted=: ;;
+    ,set)
+      { echo "$as_me:$LINENO: error: \`$ac_var' was not set in the previous run" >&5
+echo "$as_me: error: \`$ac_var' was not set in the previous run" >&2;}
+      ac_cache_corrupted=: ;;
+    ,);;
+    *)
+      if test "x$ac_old_val" != "x$ac_new_val"; then
+	{ echo "$as_me:$LINENO: error: \`$ac_var' has changed since the previous run:" >&5
+echo "$as_me: error: \`$ac_var' has changed since the previous run:" >&2;}
+	{ echo "$as_me:$LINENO:   former value:  $ac_old_val" >&5
+echo "$as_me:   former value:  $ac_old_val" >&2;}
+	{ echo "$as_me:$LINENO:   current value: $ac_new_val" >&5
+echo "$as_me:   current value: $ac_new_val" >&2;}
+	ac_cache_corrupted=:
+      fi;;
+  esac
+  # Pass precious variables to config.status.
+  if test "$ac_new_set" = set; then
+    case $ac_new_val in
+    *\'*) ac_arg=$ac_var=`echo "$ac_new_val" | sed "s/'/'\\\\\\\\''/g"` ;;
+    *) ac_arg=$ac_var=$ac_new_val ;;
+    esac
+    case " $ac_configure_args " in
+      *" '$ac_arg' "*) ;; # Avoid dups.  Use of quotes ensures accuracy.
+      *) ac_configure_args="$ac_configure_args '$ac_arg'" ;;
+    esac
+  fi
+done
+if $ac_cache_corrupted; then
+  { echo "$as_me:$LINENO: error: changes in the environment can compromise the build" >&5
+echo "$as_me: error: changes in the environment can compromise the build" >&2;}
+  { { echo "$as_me:$LINENO: error: run \`make distclean' and/or \`rm $cache_file' and start over" >&5
+echo "$as_me: error: run \`make distclean' and/or \`rm $cache_file' and start over" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ac_ext=c
+ac_cpp='$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
+
+
+
+LLVM_SRC_ROOT="."
+
+ac_aux_dir=
+for ac_dir in $LLVM_SRC_ROOT/autoconf "$srcdir"/$LLVM_SRC_ROOT/autoconf; do
+  if test -f "$ac_dir/install-sh"; then
+    ac_aux_dir=$ac_dir
+    ac_install_sh="$ac_aux_dir/install-sh -c"
+    break
+  elif test -f "$ac_dir/install.sh"; then
+    ac_aux_dir=$ac_dir
+    ac_install_sh="$ac_aux_dir/install.sh -c"
+    break
+  elif test -f "$ac_dir/shtool"; then
+    ac_aux_dir=$ac_dir
+    ac_install_sh="$ac_aux_dir/shtool install -c"
+    break
+  fi
+done
+if test -z "$ac_aux_dir"; then
+  { { echo "$as_me:$LINENO: error: cannot find install-sh or install.sh in $LLVM_SRC_ROOT/autoconf \"$srcdir\"/$LLVM_SRC_ROOT/autoconf" >&5
+echo "$as_me: error: cannot find install-sh or install.sh in $LLVM_SRC_ROOT/autoconf \"$srcdir\"/$LLVM_SRC_ROOT/autoconf" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+# These three variables are undocumented and unsupported,
+# and are intended to be withdrawn in a future Autoconf release.
+# They can cause serious problems if a builder's source tree is in a directory
+# whose full name contains unusual characters.
+ac_config_guess="$SHELL $ac_aux_dir/config.guess"  # Please don't use this var.
+ac_config_sub="$SHELL $ac_aux_dir/config.sub"  # Please don't use this var.
+ac_configure="$SHELL $ac_aux_dir/configure"  # Please don't use this var.
+
+
+
+
+# Check whether --with-llvmsrc was given.
+if test "${with_llvmsrc+set}" = set; then
+  withval=$with_llvmsrc; llvm_src="$withval"
+else
+  llvm_src=""""
+fi
+
+  LLVM_SRC=$llvm_src
+
+
+# Check whether --with-llvmobj was given.
+if test "${with_llvmobj+set}" = set; then
+  withval=$with_llvmobj; llvm_obj="$withval"
+else
+  llvm_obj=""""
+fi
+
+  LLVM_OBJ=$llvm_obj
+
+  ac_config_commands="$ac_config_commands setup"
+
+
+
+
+
+ac_config_files="$ac_config_files Makefile.config"
+
+ac_config_files="$ac_config_files stp/Makefile.common"
+
+
+ac_config_headers="$ac_config_headers include/klee/Config/config.h"
+
+
+
+ac_ext=cpp
+ac_cpp='$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
+
+
+
+# Make sure we can run config.sub.
+$SHELL "$ac_aux_dir/config.sub" sun4 >/dev/null 2>&1 ||
+  { { echo "$as_me:$LINENO: error: cannot run $SHELL $ac_aux_dir/config.sub" >&5
+echo "$as_me: error: cannot run $SHELL $ac_aux_dir/config.sub" >&2;}
+   { (exit 1); exit 1; }; }
+
+{ echo "$as_me:$LINENO: checking build system type" >&5
+echo $ECHO_N "checking build system type... $ECHO_C" >&6; }
+if test "${ac_cv_build+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_build_alias=$build_alias
+test "x$ac_build_alias" = x &&
+  ac_build_alias=`$SHELL "$ac_aux_dir/config.guess"`
+test "x$ac_build_alias" = x &&
+  { { echo "$as_me:$LINENO: error: cannot guess build type; you must specify one" >&5
+echo "$as_me: error: cannot guess build type; you must specify one" >&2;}
+   { (exit 1); exit 1; }; }
+ac_cv_build=`$SHELL "$ac_aux_dir/config.sub" $ac_build_alias` ||
+  { { echo "$as_me:$LINENO: error: $SHELL $ac_aux_dir/config.sub $ac_build_alias failed" >&5
+echo "$as_me: error: $SHELL $ac_aux_dir/config.sub $ac_build_alias failed" >&2;}
+   { (exit 1); exit 1; }; }
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_build" >&5
+echo "${ECHO_T}$ac_cv_build" >&6; }
+case $ac_cv_build in
+*-*-*) ;;
+*) { { echo "$as_me:$LINENO: error: invalid value of canonical build" >&5
+echo "$as_me: error: invalid value of canonical build" >&2;}
+   { (exit 1); exit 1; }; };;
+esac
+build=$ac_cv_build
+ac_save_IFS=$IFS; IFS='-'
+set x $ac_cv_build
+shift
+build_cpu=$1
+build_vendor=$2
+shift; shift
+# Remember, the first character of IFS is used to create $*,
+# except with old shells:
+build_os=$*
+IFS=$ac_save_IFS
+case $build_os in *\ *) build_os=`echo "$build_os" | sed 's/ /-/g'`;; esac
+
+
+{ echo "$as_me:$LINENO: checking host system type" >&5
+echo $ECHO_N "checking host system type... $ECHO_C" >&6; }
+if test "${ac_cv_host+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test "x$host_alias" = x; then
+  ac_cv_host=$ac_cv_build
+else
+  ac_cv_host=`$SHELL "$ac_aux_dir/config.sub" $host_alias` ||
+    { { echo "$as_me:$LINENO: error: $SHELL $ac_aux_dir/config.sub $host_alias failed" >&5
+echo "$as_me: error: $SHELL $ac_aux_dir/config.sub $host_alias failed" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_host" >&5
+echo "${ECHO_T}$ac_cv_host" >&6; }
+case $ac_cv_host in
+*-*-*) ;;
+*) { { echo "$as_me:$LINENO: error: invalid value of canonical host" >&5
+echo "$as_me: error: invalid value of canonical host" >&2;}
+   { (exit 1); exit 1; }; };;
+esac
+host=$ac_cv_host
+ac_save_IFS=$IFS; IFS='-'
+set x $ac_cv_host
+shift
+host_cpu=$1
+host_vendor=$2
+shift; shift
+# Remember, the first character of IFS is used to create $*,
+# except with old shells:
+host_os=$*
+IFS=$ac_save_IFS
+case $host_os in *\ *) host_os=`echo "$host_os" | sed 's/ /-/g'`;; esac
+
+
+{ echo "$as_me:$LINENO: checking target system type" >&5
+echo $ECHO_N "checking target system type... $ECHO_C" >&6; }
+if test "${ac_cv_target+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test "x$target_alias" = x; then
+  ac_cv_target=$ac_cv_host
+else
+  ac_cv_target=`$SHELL "$ac_aux_dir/config.sub" $target_alias` ||
+    { { echo "$as_me:$LINENO: error: $SHELL $ac_aux_dir/config.sub $target_alias failed" >&5
+echo "$as_me: error: $SHELL $ac_aux_dir/config.sub $target_alias failed" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_target" >&5
+echo "${ECHO_T}$ac_cv_target" >&6; }
+case $ac_cv_target in
+*-*-*) ;;
+*) { { echo "$as_me:$LINENO: error: invalid value of canonical target" >&5
+echo "$as_me: error: invalid value of canonical target" >&2;}
+   { (exit 1); exit 1; }; };;
+esac
+target=$ac_cv_target
+ac_save_IFS=$IFS; IFS='-'
+set x $ac_cv_target
+shift
+target_cpu=$1
+target_vendor=$2
+shift; shift
+# Remember, the first character of IFS is used to create $*,
+# except with old shells:
+target_os=$*
+IFS=$ac_save_IFS
+case $target_os in *\ *) target_os=`echo "$target_os" | sed 's/ /-/g'`;; esac
+
+
+# The aliases save the names the user supplied, while $host etc.
+# will get canonicalized.
+test -n "$target_alias" &&
+  test "$program_prefix$program_suffix$program_transform_name" = \
+    NONENONEs,x,x, &&
+  program_prefix=${target_alias}-
+
+{ echo "$as_me:$LINENO: checking type of operating system we're going to host on" >&5
+echo $ECHO_N "checking type of operating system we're going to host on... $ECHO_C" >&6; }
+if test "${klee_cv_os_type+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  case $host in
+  *-*-linux*)
+    host_supports_posix_runtime=yes ;;
+  *)
+    host_supports_posix_runtime=no ;;
+esac
+fi
+{ echo "$as_me:$LINENO: result: $klee_cv_os_type" >&5
+echo "${ECHO_T}$klee_cv_os_type" >&6; }
+
+
+
+# Check whether --with-llvm was given.
+if test "${with_llvm+set}" = set; then
+  withval=$with_llvm;
+fi
+
+
+{ echo "$as_me:$LINENO: checking llvm source dir" >&5
+echo $ECHO_N "checking llvm source dir... $ECHO_C" >&6; }
+
+if test X${with_llvm} != X; then
+        if test X${with_llvmsrc} != X; then
+       { { echo "$as_me:$LINENO: error: --with-llvmsrc cannot be specified when using --with-llvm" >&5
+echo "$as_me: error: --with-llvmsrc cannot be specified when using --with-llvm" >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+    if test X${with_llvmobj} != X; then
+       { { echo "$as_me:$LINENO: error: --with-llvmobj cannot be specified when using --with-llvm" >&5
+echo "$as_me: error: --with-llvmobj cannot be specified when using --with-llvm" >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+    with_llvmsrc=$with_llvm
+    with_llvmobj=$with_llvm
+fi
+
+if (test X${with_llvmsrc} != X || test X${with_llvmobj} != X); then
+        if test X${with_llvmsrc} = X; then
+       { { echo "$as_me:$LINENO: error: --with-llvmsrc must be specified when using --with-llvmobj" >&5
+echo "$as_me: error: --with-llvmsrc must be specified when using --with-llvmobj" >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+    if test X${with_llvmobj} = X; then
+       { { echo "$as_me:$LINENO: error: --with-llvmobj must be specified when using --with-llvmsrc" >&5
+echo "$as_me: error: --with-llvmobj must be specified when using --with-llvmsrc" >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+else
+        llvm_version=`llvm-config --version`
+    if test X${llvm_version} = X; then
+       { { echo "$as_me:$LINENO: error: unable to find llvm, use --with-llvmsrc and --with-llvmobj" >&5
+echo "$as_me: error: unable to find llvm, use --with-llvmsrc and --with-llvmobj" >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+
+    with_llvmsrc=`llvm-config --src-root`
+    with_llvmobj=`llvm-config --obj-root`
+fi
+
+if test ! -f ${with_llvmsrc}/Makefile.rules; then
+   { { echo "$as_me:$LINENO: error: invalid llvmsrc directory: ${with_llvmsrc}" >&5
+echo "$as_me: error: invalid llvmsrc directory: ${with_llvmsrc}" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+if test ! -f ${with_llvmobj}/Makefile.config; then
+   { { echo "$as_me:$LINENO: error: invalid llvmobj directory: ${with_llvmobj}" >&5
+echo "$as_me: error: invalid llvmobj directory: ${with_llvmobj}" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+llvm_src=`cd $with_llvmsrc 2> /dev/null; pwd`
+llvm_obj=`cd $with_llvmobj 2> /dev/null; pwd`
+
+{ echo "$as_me:$LINENO: result: $llvm_src" >&5
+echo "${ECHO_T}$llvm_src" >&6; }
+
+{ echo "$as_me:$LINENO: checking llvm obj dir" >&5
+echo $ECHO_N "checking llvm obj dir... $ECHO_C" >&6; }
+{ echo "$as_me:$LINENO: result: $llvm_obj" >&5
+echo "${ECHO_T}$llvm_obj" >&6; }
+
+LLVM_SRC=$llvm_src
+
+LLVM_OBJ=$llvm_obj
+
+
+
+
+# Check whether --with-uclibc was given.
+if test "${with_uclibc+set}" = set; then
+  withval=$with_uclibc;
+fi
+
+
+if (test X${with_uclibc} = X && test -d uclibc); then
+   with_uclibc=uclibc
+fi
+
+
+{ echo "$as_me:$LINENO: checking uclibc" >&5
+echo $ECHO_N "checking uclibc... $ECHO_C" >&6; }
+if (test X${with_uclibc} != X); then
+   if test ! -d ${with_uclibc}; then
+       { { echo "$as_me:$LINENO: error: invalid uclibc directory: ${with_uclibc}" >&5
+echo "$as_me: error: invalid uclibc directory: ${with_uclibc}" >&2;}
+   { (exit 1); exit 1; }; }
+   fi
+
+      with_uclibc=`cd $with_uclibc 2> /dev/null; pwd`
+
+   { echo "$as_me:$LINENO: result: $with_uclibc" >&5
+echo "${ECHO_T}$with_uclibc" >&6; }
+else
+   { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+cat >>confdefs.h <<_ACEOF
+#define KLEE_UCLIBC "$with_uclibc"
+_ACEOF
+
+
+
+if test X${with_uclibc} != X ; then
+  ENABLE_UCLIBC=1
+
+else
+  ENABLE_UCLIBC=0
+
+fi
+
+
+# Check whether --enable-posix-runtime was given.
+if test "${enable_posix_runtime+set}" = set; then
+  enableval=$enable_posix_runtime;
+else
+  enableval=default
+fi
+
+
+{ echo "$as_me:$LINENO: checking POSIX runtime" >&5
+echo $ECHO_N "checking POSIX runtime... $ECHO_C" >&6; }
+if test ${enableval} = "default" ; then
+  if test X${with_uclibc} != X; then
+    enableval=$host_supports_posix_runtime
+    if test ${enableval} = "yes"; then
+      { echo "$as_me:$LINENO: result: default (enabled)" >&5
+echo "${ECHO_T}default (enabled)" >&6; }
+    else
+      { echo "$as_me:$LINENO: result: default (disabled, unsupported target)" >&5
+echo "${ECHO_T}default (disabled, unsupported target)" >&6; }
+    fi
+  else
+    enableval="no"
+    { echo "$as_me:$LINENO: result: default (disabled, no uclibc)" >&5
+echo "${ECHO_T}default (disabled, no uclibc)" >&6; }
+  fi
+else
+  if test ${enableval} = "yes" ; then
+    { echo "$as_me:$LINENO: result: yes" >&5
+echo "${ECHO_T}yes" >&6; }
+  else
+    { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+  fi
+fi
+
+if test ${enableval} = "yes" ; then
+  ENABLE_POSIX_RUNTIME=1
+
+else
+  ENABLE_POSIX_RUNTIME=0
+
+fi
+
+
+
+# Check whether --with-runtime was given.
+if test "${with_runtime+set}" = set; then
+  withval=$with_runtime;
+else
+  withval=default
+fi
+
+
+if test X"${withval}" = Xdefault; then
+   with_runtime=Release
+fi
+
+{ echo "$as_me:$LINENO: checking runtime configuration" >&5
+echo $ECHO_N "checking runtime configuration... $ECHO_C" >&6; }
+if test X${with_runtime} = XRelease; then
+    { echo "$as_me:$LINENO: result: Release" >&5
+echo "${ECHO_T}Release" >&6; }
+    RUNTIME_ENABLE_OPTIMIZED=1
+
+elif test X${with_runtime} = XDebug; then
+   { echo "$as_me:$LINENO: result: Debug" >&5
+echo "${ECHO_T}Debug" >&6; }
+   RUNTIME_ENABLE_OPTIMIZED=0
+
+else
+   { { echo "$as_me:$LINENO: error: invalid configuration: ${with_runtime}" >&5
+echo "$as_me: error: invalid configuration: ${with_runtime}" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+
+cat >>confdefs.h <<_ACEOF
+#define RUNTIME_CONFIGURATION "$with_runtime"
+_ACEOF
+
+
+
+
+case $host in
+  *-*-linux*)
+
+cat >>confdefs.h <<_ACEOF
+#define HAVE_CTYPE_EXTERNALS 1
+_ACEOF
+
+esac
+
+
+ac_ext=c
+ac_cpp='$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
+
+
+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
+if test -n "$ac_tool_prefix"; then
+  # Extract the first word of "${ac_tool_prefix}gcc", so it can be a program name with args.
+set dummy ${ac_tool_prefix}gcc; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_CC="${ac_tool_prefix}gcc"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+CC=$ac_cv_prog_CC
+if test -n "$CC"; then
+  { echo "$as_me:$LINENO: result: $CC" >&5
+echo "${ECHO_T}$CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+fi
+if test -z "$ac_cv_prog_CC"; then
+  ac_ct_CC=$CC
+  # Extract the first word of "gcc", so it can be a program name with args.
+set dummy gcc; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_ac_ct_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$ac_ct_CC"; then
+  ac_cv_prog_ac_ct_CC="$ac_ct_CC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_ac_ct_CC="gcc"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+ac_ct_CC=$ac_cv_prog_ac_ct_CC
+if test -n "$ac_ct_CC"; then
+  { echo "$as_me:$LINENO: result: $ac_ct_CC" >&5
+echo "${ECHO_T}$ac_ct_CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+  if test "x$ac_ct_CC" = x; then
+    CC=""
+  else
+    case $cross_compiling:$ac_tool_warned in
+yes:)
+{ echo "$as_me:$LINENO: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&5
+echo "$as_me: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&2;}
+ac_tool_warned=yes ;;
+esac
+    CC=$ac_ct_CC
+  fi
+else
+  CC="$ac_cv_prog_CC"
+fi
+
+if test -z "$CC"; then
+          if test -n "$ac_tool_prefix"; then
+    # Extract the first word of "${ac_tool_prefix}cc", so it can be a program name with args.
+set dummy ${ac_tool_prefix}cc; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_CC="${ac_tool_prefix}cc"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+CC=$ac_cv_prog_CC
+if test -n "$CC"; then
+  { echo "$as_me:$LINENO: result: $CC" >&5
+echo "${ECHO_T}$CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+  fi
+fi
+if test -z "$CC"; then
+  # Extract the first word of "cc", so it can be a program name with args.
+set dummy cc; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+  ac_prog_rejected=no
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    if test "$as_dir/$ac_word$ac_exec_ext" = "/usr/ucb/cc"; then
+       ac_prog_rejected=yes
+       continue
+     fi
+    ac_cv_prog_CC="cc"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+if test $ac_prog_rejected = yes; then
+  # We found a bogon in the path, so make sure we never use it.
+  set dummy $ac_cv_prog_CC
+  shift
+  if test $# != 0; then
+    # We chose a different compiler from the bogus one.
+    # However, it has the same basename, so the bogon will be chosen
+    # first if we set CC to just the basename; use the full file name.
+    shift
+    ac_cv_prog_CC="$as_dir/$ac_word${1+' '}$@"
+  fi
+fi
+fi
+fi
+CC=$ac_cv_prog_CC
+if test -n "$CC"; then
+  { echo "$as_me:$LINENO: result: $CC" >&5
+echo "${ECHO_T}$CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+fi
+if test -z "$CC"; then
+  if test -n "$ac_tool_prefix"; then
+  for ac_prog in cl.exe
+  do
+    # Extract the first word of "$ac_tool_prefix$ac_prog", so it can be a program name with args.
+set dummy $ac_tool_prefix$ac_prog; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$CC"; then
+  ac_cv_prog_CC="$CC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_CC="$ac_tool_prefix$ac_prog"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+CC=$ac_cv_prog_CC
+if test -n "$CC"; then
+  { echo "$as_me:$LINENO: result: $CC" >&5
+echo "${ECHO_T}$CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+    test -n "$CC" && break
+  done
+fi
+if test -z "$CC"; then
+  ac_ct_CC=$CC
+  for ac_prog in cl.exe
+do
+  # Extract the first word of "$ac_prog", so it can be a program name with args.
+set dummy $ac_prog; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_ac_ct_CC+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$ac_ct_CC"; then
+  ac_cv_prog_ac_ct_CC="$ac_ct_CC" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_ac_ct_CC="$ac_prog"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+ac_ct_CC=$ac_cv_prog_ac_ct_CC
+if test -n "$ac_ct_CC"; then
+  { echo "$as_me:$LINENO: result: $ac_ct_CC" >&5
+echo "${ECHO_T}$ac_ct_CC" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+  test -n "$ac_ct_CC" && break
+done
+
+  if test "x$ac_ct_CC" = x; then
+    CC=""
+  else
+    case $cross_compiling:$ac_tool_warned in
+yes:)
+{ echo "$as_me:$LINENO: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&5
+echo "$as_me: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&2;}
+ac_tool_warned=yes ;;
+esac
+    CC=$ac_ct_CC
+  fi
+fi
+
+fi
+
+
+test -z "$CC" && { { echo "$as_me:$LINENO: error: no acceptable C compiler found in \$PATH
+See \`config.log' for more details." >&5
+echo "$as_me: error: no acceptable C compiler found in \$PATH
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+
+# Provide some information about the compiler.
+echo "$as_me:$LINENO: checking for C compiler version" >&5
+ac_compiler=`set X $ac_compile; echo $2`
+{ (ac_try="$ac_compiler --version >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler --version >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+{ (ac_try="$ac_compiler -v >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler -v >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+{ (ac_try="$ac_compiler -V >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler -V >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+ac_clean_files_save=$ac_clean_files
+ac_clean_files="$ac_clean_files a.out a.exe b.out"
+# Try to create an executable without -o first, disregard a.out.
+# It will help us diagnose broken compilers, and finding out an intuition
+# of exeext.
+{ echo "$as_me:$LINENO: checking for C compiler default output file name" >&5
+echo $ECHO_N "checking for C compiler default output file name... $ECHO_C" >&6; }
+ac_link_default=`echo "$ac_link" | sed 's/ -o *conftest[^ ]*//'`
+#
+# List of possible output files, starting from the most likely.
+# The algorithm is not robust to junk in `.', hence go to wildcards (a.*)
+# only as a last resort.  b.out is created by i960 compilers.
+ac_files='a_out.exe a.exe conftest.exe a.out conftest a.* conftest.* b.out'
+#
+# The IRIX 6 linker writes into existing files which may not be
+# executable, retaining their permissions.  Remove them first so a
+# subsequent execution test works.
+ac_rmfiles=
+for ac_file in $ac_files
+do
+  case $ac_file in
+    *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.o | *.obj ) ;;
+    * ) ac_rmfiles="$ac_rmfiles $ac_file";;
+  esac
+done
+rm -f $ac_rmfiles
+
+if { (ac_try="$ac_link_default"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_link_default") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; then
+  # Autoconf-2.13 could set the ac_cv_exeext variable to `no'.
+# So ignore a value of `no', otherwise this would lead to `EXEEXT = no'
+# in a Makefile.  We should not override ac_cv_exeext if it was cached,
+# so that the user can short-circuit this test for compilers unknown to
+# Autoconf.
+for ac_file in $ac_files
+do
+  test -f "$ac_file" || continue
+  case $ac_file in
+    *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.o | *.obj )
+	;;
+    [ab].out )
+	# We found the default executable, but exeext='' is most
+	# certainly right.
+	break;;
+    *.* )
+        if test "${ac_cv_exeext+set}" = set && test "$ac_cv_exeext" != no;
+	then :; else
+	   ac_cv_exeext=`expr "$ac_file" : '[^.]*\(\..*\)'`
+	fi
+	# We set ac_cv_exeext here because the later test for it is not
+	# safe: cross compilers may not add the suffix if given an `-o'
+	# argument, so we may need to know it at that point already.
+	# Even if this section looks crufty: it has the advantage of
+	# actually working.
+	break;;
+    * )
+	break;;
+  esac
+done
+test "$ac_cv_exeext" = no && ac_cv_exeext=
+
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+{ { echo "$as_me:$LINENO: error: C compiler cannot create executables
+See \`config.log' for more details." >&5
+echo "$as_me: error: C compiler cannot create executables
+See \`config.log' for more details." >&2;}
+   { (exit 77); exit 77; }; }
+fi
+
+ac_exeext=$ac_cv_exeext
+{ echo "$as_me:$LINENO: result: $ac_file" >&5
+echo "${ECHO_T}$ac_file" >&6; }
+
+# Check that the compiler produces executables we can run.  If not, either
+# the compiler is broken, or we cross compile.
+{ echo "$as_me:$LINENO: checking whether the C compiler works" >&5
+echo $ECHO_N "checking whether the C compiler works... $ECHO_C" >&6; }
+# FIXME: These cross compiler hacks should be removed for Autoconf 3.0
+# If not cross compiling, check that we can run a simple program.
+if test "$cross_compiling" != yes; then
+  if { ac_try='./$ac_file'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+    cross_compiling=no
+  else
+    if test "$cross_compiling" = maybe; then
+	cross_compiling=yes
+    else
+	{ { echo "$as_me:$LINENO: error: cannot run C compiled programs.
+If you meant to cross compile, use \`--host'.
+See \`config.log' for more details." >&5
+echo "$as_me: error: cannot run C compiled programs.
+If you meant to cross compile, use \`--host'.
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+    fi
+  fi
+fi
+{ echo "$as_me:$LINENO: result: yes" >&5
+echo "${ECHO_T}yes" >&6; }
+
+rm -f a.out a.exe conftest$ac_cv_exeext b.out
+ac_clean_files=$ac_clean_files_save
+# Check that the compiler produces executables we can run.  If not, either
+# the compiler is broken, or we cross compile.
+{ echo "$as_me:$LINENO: checking whether we are cross compiling" >&5
+echo $ECHO_N "checking whether we are cross compiling... $ECHO_C" >&6; }
+{ echo "$as_me:$LINENO: result: $cross_compiling" >&5
+echo "${ECHO_T}$cross_compiling" >&6; }
+
+{ echo "$as_me:$LINENO: checking for suffix of executables" >&5
+echo $ECHO_N "checking for suffix of executables... $ECHO_C" >&6; }
+if { (ac_try="$ac_link"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_link") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; then
+  # If both `conftest.exe' and `conftest' are `present' (well, observable)
+# catch `conftest.exe'.  For instance with Cygwin, `ls conftest' will
+# work properly (i.e., refer to `conftest.exe'), while it won't with
+# `rm'.
+for ac_file in conftest.exe conftest conftest.*; do
+  test -f "$ac_file" || continue
+  case $ac_file in
+    *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf | *.o | *.obj ) ;;
+    *.* ) ac_cv_exeext=`expr "$ac_file" : '[^.]*\(\..*\)'`
+	  break;;
+    * ) break;;
+  esac
+done
+else
+  { { echo "$as_me:$LINENO: error: cannot compute suffix of executables: cannot compile and link
+See \`config.log' for more details." >&5
+echo "$as_me: error: cannot compute suffix of executables: cannot compile and link
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+rm -f conftest$ac_cv_exeext
+{ echo "$as_me:$LINENO: result: $ac_cv_exeext" >&5
+echo "${ECHO_T}$ac_cv_exeext" >&6; }
+
+rm -f conftest.$ac_ext
+EXEEXT=$ac_cv_exeext
+ac_exeext=$EXEEXT
+{ echo "$as_me:$LINENO: checking for suffix of object files" >&5
+echo $ECHO_N "checking for suffix of object files... $ECHO_C" >&6; }
+if test "${ac_cv_objext+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.o conftest.obj
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; then
+  for ac_file in conftest.o conftest.obj conftest.*; do
+  test -f "$ac_file" || continue;
+  case $ac_file in
+    *.$ac_ext | *.xcoff | *.tds | *.d | *.pdb | *.xSYM | *.bb | *.bbg | *.map | *.inf ) ;;
+    *) ac_cv_objext=`expr "$ac_file" : '.*\.\(.*\)'`
+       break;;
+  esac
+done
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+{ { echo "$as_me:$LINENO: error: cannot compute suffix of object files: cannot compile
+See \`config.log' for more details." >&5
+echo "$as_me: error: cannot compute suffix of object files: cannot compile
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+rm -f conftest.$ac_cv_objext conftest.$ac_ext
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_objext" >&5
+echo "${ECHO_T}$ac_cv_objext" >&6; }
+OBJEXT=$ac_cv_objext
+ac_objext=$OBJEXT
+{ echo "$as_me:$LINENO: checking whether we are using the GNU C compiler" >&5
+echo $ECHO_N "checking whether we are using the GNU C compiler... $ECHO_C" >&6; }
+if test "${ac_cv_c_compiler_gnu+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+#ifndef __GNUC__
+       choke me
+#endif
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_compiler_gnu=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_compiler_gnu=no
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+ac_cv_c_compiler_gnu=$ac_compiler_gnu
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_c_compiler_gnu" >&5
+echo "${ECHO_T}$ac_cv_c_compiler_gnu" >&6; }
+GCC=`test $ac_compiler_gnu = yes && echo yes`
+ac_test_CFLAGS=${CFLAGS+set}
+ac_save_CFLAGS=$CFLAGS
+{ echo "$as_me:$LINENO: checking whether $CC accepts -g" >&5
+echo $ECHO_N "checking whether $CC accepts -g... $ECHO_C" >&6; }
+if test "${ac_cv_prog_cc_g+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_save_c_werror_flag=$ac_c_werror_flag
+   ac_c_werror_flag=yes
+   ac_cv_prog_cc_g=no
+   CFLAGS="-g"
+   cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_prog_cc_g=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	CFLAGS=""
+      cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_c_werror_flag=$ac_save_c_werror_flag
+	 CFLAGS="-g"
+	 cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_prog_cc_g=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+   ac_c_werror_flag=$ac_save_c_werror_flag
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_prog_cc_g" >&5
+echo "${ECHO_T}$ac_cv_prog_cc_g" >&6; }
+if test "$ac_test_CFLAGS" = set; then
+  CFLAGS=$ac_save_CFLAGS
+elif test $ac_cv_prog_cc_g = yes; then
+  if test "$GCC" = yes; then
+    CFLAGS="-g -O2"
+  else
+    CFLAGS="-g"
+  fi
+else
+  if test "$GCC" = yes; then
+    CFLAGS="-O2"
+  else
+    CFLAGS=
+  fi
+fi
+{ echo "$as_me:$LINENO: checking for $CC option to accept ISO C89" >&5
+echo $ECHO_N "checking for $CC option to accept ISO C89... $ECHO_C" >&6; }
+if test "${ac_cv_prog_cc_c89+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_cv_prog_cc_c89=no
+ac_save_CC=$CC
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <stdarg.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+/* Most of the following tests are stolen from RCS 5.7's src/conf.sh.  */
+struct buf { int x; };
+FILE * (*rcsopen) (struct buf *, struct stat *, int);
+static char *e (p, i)
+     char **p;
+     int i;
+{
+  return p[i];
+}
+static char *f (char * (*g) (char **, int), char **p, ...)
+{
+  char *s;
+  va_list v;
+  va_start (v,p);
+  s = g (p, va_arg (v,int));
+  va_end (v);
+  return s;
+}
+
+/* OSF 4.0 Compaq cc is some sort of almost-ANSI by default.  It has
+   function prototypes and stuff, but not '\xHH' hex character constants.
+   These don't provoke an error unfortunately, instead are silently treated
+   as 'x'.  The following induces an error, until -std is added to get
+   proper ANSI mode.  Curiously '\x00'!='x' always comes out true, for an
+   array size at least.  It's necessary to write '\x00'==0 to get something
+   that's true only with -std.  */
+int osf4_cc_array ['\x00' == 0 ? 1 : -1];
+
+/* IBM C 6 for AIX is almost-ANSI by default, but it replaces macro parameters
+   inside strings and character constants.  */
+#define FOO(x) 'x'
+int xlc6_cc_array[FOO(a) == 'x' ? 1 : -1];
+
+int test (int i, double x);
+struct s1 {int (*f) (int a);};
+struct s2 {int (*f) (double a);};
+int pairnames (int, char **, FILE *(*)(struct buf *, struct stat *, int), int, int);
+int argc;
+char **argv;
+int
+main ()
+{
+return f (e, argv, 0) != argv[0]  ||  f (e, argv, 1) != argv[1];
+  ;
+  return 0;
+}
+_ACEOF
+for ac_arg in '' -qlanglvl=extc89 -qlanglvl=ansi -std \
+	-Ae "-Aa -D_HPUX_SOURCE" "-Xc -D__EXTENSIONS__"
+do
+  CC="$ac_save_CC $ac_arg"
+  rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_prog_cc_c89=$ac_arg
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+
+fi
+
+rm -f core conftest.err conftest.$ac_objext
+  test "x$ac_cv_prog_cc_c89" != "xno" && break
+done
+rm -f conftest.$ac_ext
+CC=$ac_save_CC
+
+fi
+# AC_CACHE_VAL
+case "x$ac_cv_prog_cc_c89" in
+  x)
+    { echo "$as_me:$LINENO: result: none needed" >&5
+echo "${ECHO_T}none needed" >&6; } ;;
+  xno)
+    { echo "$as_me:$LINENO: result: unsupported" >&5
+echo "${ECHO_T}unsupported" >&6; } ;;
+  *)
+    CC="$CC $ac_cv_prog_cc_c89"
+    { echo "$as_me:$LINENO: result: $ac_cv_prog_cc_c89" >&5
+echo "${ECHO_T}$ac_cv_prog_cc_c89" >&6; } ;;
+esac
+
+
+ac_ext=c
+ac_cpp='$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
+
+
+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: checking how to run the C preprocessor" >&5
+echo $ECHO_N "checking how to run the C preprocessor... $ECHO_C" >&6; }
+# On Suns, sometimes $CPP names a directory.
+if test -n "$CPP" && test -d "$CPP"; then
+  CPP=
+fi
+if test -z "$CPP"; then
+  if test "${ac_cv_prog_CPP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+      # Double quotes because CPP needs to be expanded
+    for CPP in "$CC -E" "$CC -E -traditional-cpp" "/lib/cpp"
+    do
+      ac_preproc_ok=false
+for ac_c_preproc_warn_flag in '' yes
+do
+  # Use a header file that comes with gcc, so configuring glibc
+  # with a fresh cross-compiler works.
+  # Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
+  # <limits.h> exists even on freestanding compilers.
+  # On the NeXT, cc -E runs the code through the compiler's parser,
+  # not just through cpp. "Syntax error" is here to catch this case.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#ifdef __STDC__
+# include <limits.h>
+#else
+# include <assert.h>
+#endif
+		     Syntax error
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_c_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_c_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Broken: fails on valid input.
+continue
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+  # OK, works on sane cases.  Now check whether nonexistent headers
+  # can be detected and how.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <ac_nonexistent.h>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_c_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_c_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  # Broken: success on invalid input.
+continue
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Passes both tests.
+ac_preproc_ok=:
+break
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+done
+# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.
+rm -f conftest.err conftest.$ac_ext
+if $ac_preproc_ok; then
+  break
+fi
+
+    done
+    ac_cv_prog_CPP=$CPP
+
+fi
+  CPP=$ac_cv_prog_CPP
+else
+  ac_cv_prog_CPP=$CPP
+fi
+{ echo "$as_me:$LINENO: result: $CPP" >&5
+echo "${ECHO_T}$CPP" >&6; }
+ac_preproc_ok=false
+for ac_c_preproc_warn_flag in '' yes
+do
+  # Use a header file that comes with gcc, so configuring glibc
+  # with a fresh cross-compiler works.
+  # Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
+  # <limits.h> exists even on freestanding compilers.
+  # On the NeXT, cc -E runs the code through the compiler's parser,
+  # not just through cpp. "Syntax error" is here to catch this case.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#ifdef __STDC__
+# include <limits.h>
+#else
+# include <assert.h>
+#endif
+		     Syntax error
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_c_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_c_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Broken: fails on valid input.
+continue
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+  # OK, works on sane cases.  Now check whether nonexistent headers
+  # can be detected and how.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <ac_nonexistent.h>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_c_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_c_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  # Broken: success on invalid input.
+continue
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Passes both tests.
+ac_preproc_ok=:
+break
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+done
+# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.
+rm -f conftest.err conftest.$ac_ext
+if $ac_preproc_ok; then
+  :
+else
+  { { echo "$as_me:$LINENO: error: C preprocessor \"$CPP\" fails sanity check
+See \`config.log' for more details." >&5
+echo "$as_me: error: C preprocessor \"$CPP\" fails sanity check
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+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: checking for grep that handles long lines and -e" >&5
+echo $ECHO_N "checking for grep that handles long lines and -e... $ECHO_C" >&6; }
+if test "${ac_cv_path_GREP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  # Extract the first word of "grep ggrep" to use in msg output
+if test -z "$GREP"; then
+set dummy grep ggrep; ac_prog_name=$2
+if test "${ac_cv_path_GREP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_path_GREP_found=false
+# Loop through the user's path and test for each of PROGNAME-LIST
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH$PATH_SEPARATOR/usr/xpg4/bin
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_prog in grep ggrep; do
+  for ac_exec_ext in '' $ac_executable_extensions; do
+    ac_path_GREP="$as_dir/$ac_prog$ac_exec_ext"
+    { test -f "$ac_path_GREP" && $as_executable_p "$ac_path_GREP"; } || continue
+    # Check for GNU ac_path_GREP and select it if it is found.
+  # Check for GNU $ac_path_GREP
+case `"$ac_path_GREP" --version 2>&1` in
+*GNU*)
+  ac_cv_path_GREP="$ac_path_GREP" ac_path_GREP_found=:;;
+*)
+  ac_count=0
+  echo $ECHO_N "0123456789$ECHO_C" >"conftest.in"
+  while :
+  do
+    cat "conftest.in" "conftest.in" >"conftest.tmp"
+    mv "conftest.tmp" "conftest.in"
+    cp "conftest.in" "conftest.nl"
+    echo 'GREP' >> "conftest.nl"
+    "$ac_path_GREP" -e 'GREP$' -e '-(cannot match)-' < "conftest.nl" >"conftest.out" 2>/dev/null || break
+    diff "conftest.out" "conftest.nl" >/dev/null 2>&1 || break
+    ac_count=`expr $ac_count + 1`
+    if test $ac_count -gt ${ac_path_GREP_max-0}; then
+      # Best one so far, save it but keep looking for a better one
+      ac_cv_path_GREP="$ac_path_GREP"
+      ac_path_GREP_max=$ac_count
+    fi
+    # 10*(2^10) chars as input seems more than enough
+    test $ac_count -gt 10 && break
+  done
+  rm -f conftest.in conftest.tmp conftest.nl conftest.out;;
+esac
+
+
+    $ac_path_GREP_found && break 3
+  done
+done
+
+done
+IFS=$as_save_IFS
+
+
+fi
+
+GREP="$ac_cv_path_GREP"
+if test -z "$GREP"; then
+  { { echo "$as_me:$LINENO: error: no acceptable $ac_prog_name could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin" >&5
+echo "$as_me: error: no acceptable $ac_prog_name could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+else
+  ac_cv_path_GREP=$GREP
+fi
+
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_path_GREP" >&5
+echo "${ECHO_T}$ac_cv_path_GREP" >&6; }
+ GREP="$ac_cv_path_GREP"
+
+
+{ echo "$as_me:$LINENO: checking for egrep" >&5
+echo $ECHO_N "checking for egrep... $ECHO_C" >&6; }
+if test "${ac_cv_path_EGREP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if echo a | $GREP -E '(a|b)' >/dev/null 2>&1
+   then ac_cv_path_EGREP="$GREP -E"
+   else
+     # Extract the first word of "egrep" to use in msg output
+if test -z "$EGREP"; then
+set dummy egrep; ac_prog_name=$2
+if test "${ac_cv_path_EGREP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_path_EGREP_found=false
+# Loop through the user's path and test for each of PROGNAME-LIST
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH$PATH_SEPARATOR/usr/xpg4/bin
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_prog in egrep; do
+  for ac_exec_ext in '' $ac_executable_extensions; do
+    ac_path_EGREP="$as_dir/$ac_prog$ac_exec_ext"
+    { test -f "$ac_path_EGREP" && $as_executable_p "$ac_path_EGREP"; } || continue
+    # Check for GNU ac_path_EGREP and select it if it is found.
+  # Check for GNU $ac_path_EGREP
+case `"$ac_path_EGREP" --version 2>&1` in
+*GNU*)
+  ac_cv_path_EGREP="$ac_path_EGREP" ac_path_EGREP_found=:;;
+*)
+  ac_count=0
+  echo $ECHO_N "0123456789$ECHO_C" >"conftest.in"
+  while :
+  do
+    cat "conftest.in" "conftest.in" >"conftest.tmp"
+    mv "conftest.tmp" "conftest.in"
+    cp "conftest.in" "conftest.nl"
+    echo 'EGREP' >> "conftest.nl"
+    "$ac_path_EGREP" 'EGREP$' < "conftest.nl" >"conftest.out" 2>/dev/null || break
+    diff "conftest.out" "conftest.nl" >/dev/null 2>&1 || break
+    ac_count=`expr $ac_count + 1`
+    if test $ac_count -gt ${ac_path_EGREP_max-0}; then
+      # Best one so far, save it but keep looking for a better one
+      ac_cv_path_EGREP="$ac_path_EGREP"
+      ac_path_EGREP_max=$ac_count
+    fi
+    # 10*(2^10) chars as input seems more than enough
+    test $ac_count -gt 10 && break
+  done
+  rm -f conftest.in conftest.tmp conftest.nl conftest.out;;
+esac
+
+
+    $ac_path_EGREP_found && break 3
+  done
+done
+
+done
+IFS=$as_save_IFS
+
+
+fi
+
+EGREP="$ac_cv_path_EGREP"
+if test -z "$EGREP"; then
+  { { echo "$as_me:$LINENO: error: no acceptable $ac_prog_name could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin" >&5
+echo "$as_me: error: no acceptable $ac_prog_name could be found in $PATH$PATH_SEPARATOR/usr/xpg4/bin" >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+else
+  ac_cv_path_EGREP=$EGREP
+fi
+
+
+   fi
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_path_EGREP" >&5
+echo "${ECHO_T}$ac_cv_path_EGREP" >&6; }
+ EGREP="$ac_cv_path_EGREP"
+
+
+{ echo "$as_me:$LINENO: checking for ANSI C header files" >&5
+echo $ECHO_N "checking for ANSI C header files... $ECHO_C" >&6; }
+if test "${ac_cv_header_stdc+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <float.h>
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_header_stdc=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_cv_header_stdc=no
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+
+if test $ac_cv_header_stdc = yes; then
+  # SunOS 4.x string.h does not declare mem*, contrary to ANSI.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <string.h>
+
+_ACEOF
+if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
+  $EGREP "memchr" >/dev/null 2>&1; then
+  :
+else
+  ac_cv_header_stdc=no
+fi
+rm -f conftest*
+
+fi
+
+if test $ac_cv_header_stdc = yes; then
+  # ISC 2.0.2 stdlib.h does not declare free, contrary to ANSI.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <stdlib.h>
+
+_ACEOF
+if (eval "$ac_cpp conftest.$ac_ext") 2>&5 |
+  $EGREP "free" >/dev/null 2>&1; then
+  :
+else
+  ac_cv_header_stdc=no
+fi
+rm -f conftest*
+
+fi
+
+if test $ac_cv_header_stdc = yes; then
+  # /bin/cc in Irix-4.0.5 gets non-ANSI ctype macros unless using -ansi.
+  if test "$cross_compiling" = yes; then
+  :
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <ctype.h>
+#include <stdlib.h>
+#if ((' ' & 0x0FF) == 0x020)
+# define ISLOWER(c) ('a' <= (c) && (c) <= 'z')
+# define TOUPPER(c) (ISLOWER(c) ? 'A' + ((c) - 'a') : (c))
+#else
+# define ISLOWER(c) \
+		   (('a' <= (c) && (c) <= 'i') \
+		     || ('j' <= (c) && (c) <= 'r') \
+		     || ('s' <= (c) && (c) <= 'z'))
+# define TOUPPER(c) (ISLOWER(c) ? ((c) | 0x40) : (c))
+#endif
+
+#define XOR(e, f) (((e) && !(f)) || (!(e) && (f)))
+int
+main ()
+{
+  int i;
+  for (i = 0; i < 256; i++)
+    if (XOR (islower (i), ISLOWER (i))
+	|| toupper (i) != TOUPPER (i))
+      return 2;
+  return 0;
+}
+_ACEOF
+rm -f conftest$ac_exeext
+if { (ac_try="$ac_link"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_link") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } && { ac_try='./conftest$ac_exeext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  :
+else
+  echo "$as_me: program exited with status $ac_status" >&5
+echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+( exit $ac_status )
+ac_cv_header_stdc=no
+fi
+rm -f core *.core core.conftest.* gmon.out bb.out conftest$ac_exeext conftest.$ac_objext conftest.$ac_ext
+fi
+
+
+fi
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_header_stdc" >&5
+echo "${ECHO_T}$ac_cv_header_stdc" >&6; }
+if test $ac_cv_header_stdc = yes; then
+
+cat >>confdefs.h <<\_ACEOF
+#define STDC_HEADERS 1
+_ACEOF
+
+fi
+
+# On IRIX 5.3, sys/types and inttypes.h are conflicting.
+
+
+
+
+
+
+
+
+
+for ac_header in sys/types.h sys/stat.h stdlib.h string.h memory.h strings.h \
+		  inttypes.h stdint.h unistd.h
+do
+as_ac_Header=`echo "ac_cv_header_$ac_header" | $as_tr_sh`
+{ echo "$as_me:$LINENO: checking for $ac_header" >&5
+echo $ECHO_N "checking for $ac_header... $ECHO_C" >&6; }
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+$ac_includes_default
+
+#include <$ac_header>
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  eval "$as_ac_Header=yes"
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	eval "$as_ac_Header=no"
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+ac_res=`eval echo '${'$as_ac_Header'}'`
+	       { echo "$as_me:$LINENO: result: $ac_res" >&5
+echo "${ECHO_T}$ac_res" >&6; }
+if test `eval echo '${'$as_ac_Header'}'` = yes; then
+  cat >>confdefs.h <<_ACEOF
+#define `echo "HAVE_$ac_header" | $as_tr_cpp` 1
+_ACEOF
+
+fi
+
+done
+
+
+
+for ac_header in sys/acl.h
+do
+as_ac_Header=`echo "ac_cv_header_$ac_header" | $as_tr_sh`
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  { echo "$as_me:$LINENO: checking for $ac_header" >&5
+echo $ECHO_N "checking for $ac_header... $ECHO_C" >&6; }
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+fi
+ac_res=`eval echo '${'$as_ac_Header'}'`
+	       { echo "$as_me:$LINENO: result: $ac_res" >&5
+echo "${ECHO_T}$ac_res" >&6; }
+else
+  # Is the header compilable?
+{ echo "$as_me:$LINENO: checking $ac_header usability" >&5
+echo $ECHO_N "checking $ac_header usability... $ECHO_C" >&6; }
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+$ac_includes_default
+#include <$ac_header>
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_c_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_header_compiler=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_header_compiler=no
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+{ echo "$as_me:$LINENO: result: $ac_header_compiler" >&5
+echo "${ECHO_T}$ac_header_compiler" >&6; }
+
+# Is the header present?
+{ echo "$as_me:$LINENO: checking $ac_header presence" >&5
+echo $ECHO_N "checking $ac_header presence... $ECHO_C" >&6; }
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <$ac_header>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_c_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_c_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  ac_header_preproc=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  ac_header_preproc=no
+fi
+
+rm -f conftest.err conftest.$ac_ext
+{ echo "$as_me:$LINENO: result: $ac_header_preproc" >&5
+echo "${ECHO_T}$ac_header_preproc" >&6; }
+
+# So?  What about this header?
+case $ac_header_compiler:$ac_header_preproc:$ac_c_preproc_warn_flag in
+  yes:no: )
+    { echo "$as_me:$LINENO: WARNING: $ac_header: accepted by the compiler, rejected by the preprocessor!" >&5
+echo "$as_me: WARNING: $ac_header: accepted by the compiler, rejected by the preprocessor!" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: proceeding with the compiler's result" >&5
+echo "$as_me: WARNING: $ac_header: proceeding with the compiler's result" >&2;}
+    ac_header_preproc=yes
+    ;;
+  no:yes:* )
+    { echo "$as_me:$LINENO: WARNING: $ac_header: present but cannot be compiled" >&5
+echo "$as_me: WARNING: $ac_header: present but cannot be compiled" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header:     check for missing prerequisite headers?" >&5
+echo "$as_me: WARNING: $ac_header:     check for missing prerequisite headers?" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: see the Autoconf documentation" >&5
+echo "$as_me: WARNING: $ac_header: see the Autoconf documentation" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header:     section \"Present But Cannot Be Compiled\"" >&5
+echo "$as_me: WARNING: $ac_header:     section \"Present But Cannot Be Compiled\"" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: proceeding with the preprocessor's result" >&5
+echo "$as_me: WARNING: $ac_header: proceeding with the preprocessor's result" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: in the future, the compiler will take precedence" >&5
+echo "$as_me: WARNING: $ac_header: in the future, the compiler will take precedence" >&2;}
+    ( cat <<\_ASBOX
+## ------------------------------------- ##
+## Report this to daniel at minormatter.com ##
+## ------------------------------------- ##
+_ASBOX
+     ) | sed "s/^/$as_me: WARNING:     /" >&2
+    ;;
+esac
+{ echo "$as_me:$LINENO: checking for $ac_header" >&5
+echo $ECHO_N "checking for $ac_header... $ECHO_C" >&6; }
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  eval "$as_ac_Header=\$ac_header_preproc"
+fi
+ac_res=`eval echo '${'$as_ac_Header'}'`
+	       { echo "$as_me:$LINENO: result: $ac_res" >&5
+echo "${ECHO_T}$ac_res" >&6; }
+
+fi
+if test `eval echo '${'$as_ac_Header'}'` = yes; then
+  cat >>confdefs.h <<_ACEOF
+#define `echo "HAVE_$ac_header" | $as_tr_cpp` 1
+_ACEOF
+
+fi
+
+done
+
+
+ac_ext=cpp
+ac_cpp='$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
+
+
+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
+if test -z "$CXX"; then
+  if test -n "$CCC"; then
+    CXX=$CCC
+  else
+    if test -n "$ac_tool_prefix"; then
+  for ac_prog in g++ c++ gpp aCC CC cxx cc++ cl.exe FCC KCC RCC xlC_r xlC
+  do
+    # Extract the first word of "$ac_tool_prefix$ac_prog", so it can be a program name with args.
+set dummy $ac_tool_prefix$ac_prog; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_CXX+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$CXX"; then
+  ac_cv_prog_CXX="$CXX" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_CXX="$ac_tool_prefix$ac_prog"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+CXX=$ac_cv_prog_CXX
+if test -n "$CXX"; then
+  { echo "$as_me:$LINENO: result: $CXX" >&5
+echo "${ECHO_T}$CXX" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+    test -n "$CXX" && break
+  done
+fi
+if test -z "$CXX"; then
+  ac_ct_CXX=$CXX
+  for ac_prog in g++ c++ gpp aCC CC cxx cc++ cl.exe FCC KCC RCC xlC_r xlC
+do
+  # Extract the first word of "$ac_prog", so it can be a program name with args.
+set dummy $ac_prog; ac_word=$2
+{ echo "$as_me:$LINENO: checking for $ac_word" >&5
+echo $ECHO_N "checking for $ac_word... $ECHO_C" >&6; }
+if test "${ac_cv_prog_ac_ct_CXX+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  if test -n "$ac_ct_CXX"; then
+  ac_cv_prog_ac_ct_CXX="$ac_ct_CXX" # Let the user override the test.
+else
+as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  for ac_exec_ext in '' $ac_executable_extensions; do
+  if { test -f "$as_dir/$ac_word$ac_exec_ext" && $as_executable_p "$as_dir/$ac_word$ac_exec_ext"; }; then
+    ac_cv_prog_ac_ct_CXX="$ac_prog"
+    echo "$as_me:$LINENO: found $as_dir/$ac_word$ac_exec_ext" >&5
+    break 2
+  fi
+done
+done
+IFS=$as_save_IFS
+
+fi
+fi
+ac_ct_CXX=$ac_cv_prog_ac_ct_CXX
+if test -n "$ac_ct_CXX"; then
+  { echo "$as_me:$LINENO: result: $ac_ct_CXX" >&5
+echo "${ECHO_T}$ac_ct_CXX" >&6; }
+else
+  { echo "$as_me:$LINENO: result: no" >&5
+echo "${ECHO_T}no" >&6; }
+fi
+
+
+  test -n "$ac_ct_CXX" && break
+done
+
+  if test "x$ac_ct_CXX" = x; then
+    CXX="g++"
+  else
+    case $cross_compiling:$ac_tool_warned in
+yes:)
+{ echo "$as_me:$LINENO: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&5
+echo "$as_me: WARNING: In the future, Autoconf will not detect cross-tools
+whose name does not start with the host triplet.  If you think this
+configuration is useful to you, please write to autoconf at gnu.org." >&2;}
+ac_tool_warned=yes ;;
+esac
+    CXX=$ac_ct_CXX
+  fi
+fi
+
+  fi
+fi
+# Provide some information about the compiler.
+echo "$as_me:$LINENO: checking for C++ compiler version" >&5
+ac_compiler=`set X $ac_compile; echo $2`
+{ (ac_try="$ac_compiler --version >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler --version >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+{ (ac_try="$ac_compiler -v >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler -v >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+{ (ac_try="$ac_compiler -V >&5"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compiler -V >&5") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }
+
+{ echo "$as_me:$LINENO: checking whether we are using the GNU C++ compiler" >&5
+echo $ECHO_N "checking whether we are using the GNU C++ compiler... $ECHO_C" >&6; }
+if test "${ac_cv_cxx_compiler_gnu+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+#ifndef __GNUC__
+       choke me
+#endif
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_cxx_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_compiler_gnu=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_compiler_gnu=no
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+ac_cv_cxx_compiler_gnu=$ac_compiler_gnu
+
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_cxx_compiler_gnu" >&5
+echo "${ECHO_T}$ac_cv_cxx_compiler_gnu" >&6; }
+GXX=`test $ac_compiler_gnu = yes && echo yes`
+ac_test_CXXFLAGS=${CXXFLAGS+set}
+ac_save_CXXFLAGS=$CXXFLAGS
+{ echo "$as_me:$LINENO: checking whether $CXX accepts -g" >&5
+echo $ECHO_N "checking whether $CXX accepts -g... $ECHO_C" >&6; }
+if test "${ac_cv_prog_cxx_g+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  ac_save_cxx_werror_flag=$ac_cxx_werror_flag
+   ac_cxx_werror_flag=yes
+   ac_cv_prog_cxx_g=no
+   CXXFLAGS="-g"
+   cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_cxx_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_prog_cxx_g=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	CXXFLAGS=""
+      cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_cxx_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_cxx_werror_flag=$ac_save_cxx_werror_flag
+	 CXXFLAGS="-g"
+	 cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+
+int
+main ()
+{
+
+  ;
+  return 0;
+}
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_cxx_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_cv_prog_cxx_g=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+   ac_cxx_werror_flag=$ac_save_cxx_werror_flag
+fi
+{ echo "$as_me:$LINENO: result: $ac_cv_prog_cxx_g" >&5
+echo "${ECHO_T}$ac_cv_prog_cxx_g" >&6; }
+if test "$ac_test_CXXFLAGS" = set; then
+  CXXFLAGS=$ac_save_CXXFLAGS
+elif test $ac_cv_prog_cxx_g = yes; then
+  if test "$GXX" = yes; then
+    CXXFLAGS="-g -O2"
+  else
+    CXXFLAGS="-g"
+  fi
+else
+  if test "$GXX" = yes; then
+    CXXFLAGS="-O2"
+  else
+    CXXFLAGS=
+  fi
+fi
+ac_ext=cpp
+ac_cpp='$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
+
+
+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
+{ echo "$as_me:$LINENO: checking how to run the C++ preprocessor" >&5
+echo $ECHO_N "checking how to run the C++ preprocessor... $ECHO_C" >&6; }
+if test -z "$CXXCPP"; then
+  if test "${ac_cv_prog_CXXCPP+set}" = set; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+      # Double quotes because CXXCPP needs to be expanded
+    for CXXCPP in "$CXX -E" "/lib/cpp"
+    do
+      ac_preproc_ok=false
+for ac_cxx_preproc_warn_flag in '' yes
+do
+  # Use a header file that comes with gcc, so configuring glibc
+  # with a fresh cross-compiler works.
+  # Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
+  # <limits.h> exists even on freestanding compilers.
+  # On the NeXT, cc -E runs the code through the compiler's parser,
+  # not just through cpp. "Syntax error" is here to catch this case.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#ifdef __STDC__
+# include <limits.h>
+#else
+# include <assert.h>
+#endif
+		     Syntax error
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_cxx_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_cxx_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Broken: fails on valid input.
+continue
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+  # OK, works on sane cases.  Now check whether nonexistent headers
+  # can be detected and how.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <ac_nonexistent.h>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_cxx_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_cxx_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  # Broken: success on invalid input.
+continue
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Passes both tests.
+ac_preproc_ok=:
+break
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+done
+# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.
+rm -f conftest.err conftest.$ac_ext
+if $ac_preproc_ok; then
+  break
+fi
+
+    done
+    ac_cv_prog_CXXCPP=$CXXCPP
+
+fi
+  CXXCPP=$ac_cv_prog_CXXCPP
+else
+  ac_cv_prog_CXXCPP=$CXXCPP
+fi
+{ echo "$as_me:$LINENO: result: $CXXCPP" >&5
+echo "${ECHO_T}$CXXCPP" >&6; }
+ac_preproc_ok=false
+for ac_cxx_preproc_warn_flag in '' yes
+do
+  # Use a header file that comes with gcc, so configuring glibc
+  # with a fresh cross-compiler works.
+  # Prefer <limits.h> to <assert.h> if __STDC__ is defined, since
+  # <limits.h> exists even on freestanding compilers.
+  # On the NeXT, cc -E runs the code through the compiler's parser,
+  # not just through cpp. "Syntax error" is here to catch this case.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#ifdef __STDC__
+# include <limits.h>
+#else
+# include <assert.h>
+#endif
+		     Syntax error
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_cxx_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_cxx_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  :
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Broken: fails on valid input.
+continue
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+  # OK, works on sane cases.  Now check whether nonexistent headers
+  # can be detected and how.
+  cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <ac_nonexistent.h>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_cxx_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_cxx_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  # Broken: success on invalid input.
+continue
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  # Passes both tests.
+ac_preproc_ok=:
+break
+fi
+
+rm -f conftest.err conftest.$ac_ext
+
+done
+# Because of `break', _AC_PREPROC_IFELSE's cleaning code was skipped.
+rm -f conftest.err conftest.$ac_ext
+if $ac_preproc_ok; then
+  :
+else
+  { { echo "$as_me:$LINENO: error: C++ preprocessor \"$CXXCPP\" fails sanity check
+See \`config.log' for more details." >&5
+echo "$as_me: error: C++ preprocessor \"$CXXCPP\" fails sanity check
+See \`config.log' for more details." >&2;}
+   { (exit 1); exit 1; }; }
+fi
+
+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
+
+
+
+for ac_header in selinux/selinux.h
+do
+as_ac_Header=`echo "ac_cv_header_$ac_header" | $as_tr_sh`
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  { echo "$as_me:$LINENO: checking for $ac_header" >&5
+echo $ECHO_N "checking for $ac_header... $ECHO_C" >&6; }
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+fi
+ac_res=`eval echo '${'$as_ac_Header'}'`
+	       { echo "$as_me:$LINENO: result: $ac_res" >&5
+echo "${ECHO_T}$ac_res" >&6; }
+else
+  # Is the header compilable?
+{ echo "$as_me:$LINENO: checking $ac_header usability" >&5
+echo $ECHO_N "checking $ac_header usability... $ECHO_C" >&6; }
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+$ac_includes_default
+#include <$ac_header>
+_ACEOF
+rm -f conftest.$ac_objext
+if { (ac_try="$ac_compile"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_compile") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } &&
+	 { ac_try='test -z "$ac_cxx_werror_flag" || test ! -s conftest.err'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; } &&
+	 { ac_try='test -s conftest.$ac_objext'
+  { (case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_try") 2>&5
+  ac_status=$?
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); }; }; then
+  ac_header_compiler=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+	ac_header_compiler=no
+fi
+
+rm -f core conftest.err conftest.$ac_objext conftest.$ac_ext
+{ echo "$as_me:$LINENO: result: $ac_header_compiler" >&5
+echo "${ECHO_T}$ac_header_compiler" >&6; }
+
+# Is the header present?
+{ echo "$as_me:$LINENO: checking $ac_header presence" >&5
+echo $ECHO_N "checking $ac_header presence... $ECHO_C" >&6; }
+cat >conftest.$ac_ext <<_ACEOF
+/* confdefs.h.  */
+_ACEOF
+cat confdefs.h >>conftest.$ac_ext
+cat >>conftest.$ac_ext <<_ACEOF
+/* end confdefs.h.  */
+#include <$ac_header>
+_ACEOF
+if { (ac_try="$ac_cpp conftest.$ac_ext"
+case "(($ac_try" in
+  *\"* | *\`* | *\\*) ac_try_echo=\$ac_try;;
+  *) ac_try_echo=$ac_try;;
+esac
+eval "echo \"\$as_me:$LINENO: $ac_try_echo\"") >&5
+  (eval "$ac_cpp conftest.$ac_ext") 2>conftest.er1
+  ac_status=$?
+  grep -v '^ *+' conftest.er1 >conftest.err
+  rm -f conftest.er1
+  cat conftest.err >&5
+  echo "$as_me:$LINENO: \$? = $ac_status" >&5
+  (exit $ac_status); } >/dev/null; then
+  if test -s conftest.err; then
+    ac_cpp_err=$ac_cxx_preproc_warn_flag
+    ac_cpp_err=$ac_cpp_err$ac_cxx_werror_flag
+  else
+    ac_cpp_err=
+  fi
+else
+  ac_cpp_err=yes
+fi
+if test -z "$ac_cpp_err"; then
+  ac_header_preproc=yes
+else
+  echo "$as_me: failed program was:" >&5
+sed 's/^/| /' conftest.$ac_ext >&5
+
+  ac_header_preproc=no
+fi
+
+rm -f conftest.err conftest.$ac_ext
+{ echo "$as_me:$LINENO: result: $ac_header_preproc" >&5
+echo "${ECHO_T}$ac_header_preproc" >&6; }
+
+# So?  What about this header?
+case $ac_header_compiler:$ac_header_preproc:$ac_cxx_preproc_warn_flag in
+  yes:no: )
+    { echo "$as_me:$LINENO: WARNING: $ac_header: accepted by the compiler, rejected by the preprocessor!" >&5
+echo "$as_me: WARNING: $ac_header: accepted by the compiler, rejected by the preprocessor!" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: proceeding with the compiler's result" >&5
+echo "$as_me: WARNING: $ac_header: proceeding with the compiler's result" >&2;}
+    ac_header_preproc=yes
+    ;;
+  no:yes:* )
+    { echo "$as_me:$LINENO: WARNING: $ac_header: present but cannot be compiled" >&5
+echo "$as_me: WARNING: $ac_header: present but cannot be compiled" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header:     check for missing prerequisite headers?" >&5
+echo "$as_me: WARNING: $ac_header:     check for missing prerequisite headers?" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: see the Autoconf documentation" >&5
+echo "$as_me: WARNING: $ac_header: see the Autoconf documentation" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header:     section \"Present But Cannot Be Compiled\"" >&5
+echo "$as_me: WARNING: $ac_header:     section \"Present But Cannot Be Compiled\"" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: proceeding with the preprocessor's result" >&5
+echo "$as_me: WARNING: $ac_header: proceeding with the preprocessor's result" >&2;}
+    { echo "$as_me:$LINENO: WARNING: $ac_header: in the future, the compiler will take precedence" >&5
+echo "$as_me: WARNING: $ac_header: in the future, the compiler will take precedence" >&2;}
+    ( cat <<\_ASBOX
+## ------------------------------------- ##
+## Report this to daniel at minormatter.com ##
+## ------------------------------------- ##
+_ASBOX
+     ) | sed "s/^/$as_me: WARNING:     /" >&2
+    ;;
+esac
+{ echo "$as_me:$LINENO: checking for $ac_header" >&5
+echo $ECHO_N "checking for $ac_header... $ECHO_C" >&6; }
+if { as_var=$as_ac_Header; eval "test \"\${$as_var+set}\" = set"; }; then
+  echo $ECHO_N "(cached) $ECHO_C" >&6
+else
+  eval "$as_ac_Header=\$ac_header_preproc"
+fi
+ac_res=`eval echo '${'$as_ac_Header'}'`
+	       { echo "$as_me:$LINENO: result: $ac_res" >&5
+echo "${ECHO_T}$ac_res" >&6; }
+
+fi
+if test `eval echo '${'$as_ac_Header'}'` = yes; then
+  cat >>confdefs.h <<_ACEOF
+#define `echo "HAVE_$ac_header" | $as_tr_cpp` 1
+_ACEOF
+ HAVE_SELINUX=1
+
+else
+  HAVE_SELINUX=0
+
+fi
+
+done
+
+
+
+# Check whether --enable-stplog was given.
+if test "${enable_stplog+set}" = set; then
+  enableval=$enable_stplog;
+else
+  enableval=no
+fi
+
+if test ${enableval} = "yes" ; then
+  ENABLE_STPLOG=1
+
+else
+  ENABLE_STPLOG=0
+
+fi
+
+cat >>confdefs.h <<_ACEOF
+#define ENABLE_STPLOG $ENABLE_STPLOG
+_ACEOF
+
+
+
+cat >confcache <<\_ACEOF
+# This file is a shell script that caches the results of configure
+# tests run on this system so they can be shared between configure
+# scripts and configure runs, see configure's option --config-cache.
+# It is not useful on other systems.  If it contains results you don't
+# want to keep, you may remove or edit it.
+#
+# config.status only pays attention to the cache file if you give it
+# the --recheck option to rerun configure.
+#
+# `ac_cv_env_foo' variables (set or unset) will be overridden when
+# loading this file, other *unset* `ac_cv_foo' will be assigned the
+# following values.
+
+_ACEOF
+
+# The following way of writing the cache mishandles newlines in values,
+# but we know of no workaround that is simple, portable, and efficient.
+# So, we kill variables containing newlines.
+# Ultrix sh set writes to stderr and can't be redirected directly,
+# and sets the high bit in the cache file unless we assign to the vars.
+(
+  for ac_var in `(set) 2>&1 | sed -n 's/^\([a-zA-Z_][a-zA-Z0-9_]*\)=.*/\1/p'`; do
+    eval ac_val=\$$ac_var
+    case $ac_val in #(
+    *${as_nl}*)
+      case $ac_var in #(
+      *_cv_*) { echo "$as_me:$LINENO: WARNING: Cache variable $ac_var contains a newline." >&5
+echo "$as_me: WARNING: Cache variable $ac_var contains a newline." >&2;} ;;
+      esac
+      case $ac_var in #(
+      _ | IFS | as_nl) ;; #(
+      *) $as_unset $ac_var ;;
+      esac ;;
+    esac
+  done
+
+  (set) 2>&1 |
+    case $as_nl`(ac_space=' '; set) 2>&1` in #(
+    *${as_nl}ac_space=\ *)
+      # `set' does not quote correctly, so add quotes (double-quote
+      # substitution turns \\\\ into \\, and sed turns \\ into \).
+      sed -n \
+	"s/'/'\\\\''/g;
+	  s/^\\([_$as_cr_alnum]*_cv_[_$as_cr_alnum]*\\)=\\(.*\\)/\\1='\\2'/p"
+      ;; #(
+    *)
+      # `set' quotes correctly as required by POSIX, so do not add quotes.
+      sed -n "/^[_$as_cr_alnum]*_cv_[_$as_cr_alnum]*=/p"
+      ;;
+    esac |
+    sort
+) |
+  sed '
+     /^ac_cv_env_/b end
+     t clear
+     :clear
+     s/^\([^=]*\)=\(.*[{}].*\)$/test "${\1+set}" = set || &/
+     t end
+     s/^\([^=]*\)=\(.*\)$/\1=${\1=\2}/
+     :end' >>confcache
+if diff "$cache_file" confcache >/dev/null 2>&1; then :; else
+  if test -w "$cache_file"; then
+    test "x$cache_file" != "x/dev/null" &&
+      { echo "$as_me:$LINENO: updating cache $cache_file" >&5
+echo "$as_me: updating cache $cache_file" >&6;}
+    cat confcache >$cache_file
+  else
+    { echo "$as_me:$LINENO: not updating unwritable cache $cache_file" >&5
+echo "$as_me: not updating unwritable cache $cache_file" >&6;}
+  fi
+fi
+rm -f confcache
+
+test "x$prefix" = xNONE && prefix=$ac_default_prefix
+# Let make expand exec_prefix.
+test "x$exec_prefix" = xNONE && exec_prefix='${prefix}'
+
+DEFS=-DHAVE_CONFIG_H
+
+ac_libobjs=
+ac_ltlibobjs=
+for ac_i in : $LIBOBJS; do test "x$ac_i" = x: && continue
+  # 1. Remove the extension, and $U if already installed.
+  ac_script='s/\$U\././;s/\.o$//;s/\.obj$//'
+  ac_i=`echo "$ac_i" | sed "$ac_script"`
+  # 2. Prepend LIBOBJDIR.  When used with automake>=1.10 LIBOBJDIR
+  #    will be set to the directory where LIBOBJS objects are built.
+  ac_libobjs="$ac_libobjs \${LIBOBJDIR}$ac_i\$U.$ac_objext"
+  ac_ltlibobjs="$ac_ltlibobjs \${LIBOBJDIR}$ac_i"'$U.lo'
+done
+LIBOBJS=$ac_libobjs
+
+LTLIBOBJS=$ac_ltlibobjs
+
+
+
+: ${CONFIG_STATUS=./config.status}
+ac_clean_files_save=$ac_clean_files
+ac_clean_files="$ac_clean_files $CONFIG_STATUS"
+{ echo "$as_me:$LINENO: creating $CONFIG_STATUS" >&5
+echo "$as_me: creating $CONFIG_STATUS" >&6;}
+cat >$CONFIG_STATUS <<_ACEOF
+#! $SHELL
+# Generated by $as_me.
+# Run this file to recreate the current configuration.
+# Compiler output produced by configure, useful for debugging
+# configure, is in config.log if it exists.
+
+debug=false
+ac_cs_recheck=false
+ac_cs_silent=false
+SHELL=\${CONFIG_SHELL-$SHELL}
+_ACEOF
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+## --------------------- ##
+## M4sh Initialization.  ##
+## --------------------- ##
+
+# Be Bourne compatible
+if test -n "${ZSH_VERSION+set}" && (emulate sh) >/dev/null 2>&1; then
+  emulate sh
+  NULLCMD=:
+  # Zsh 3.x and 4.x performs word splitting on ${1+"$@"}, which
+  # is contrary to our usage.  Disable this feature.
+  alias -g '${1+"$@"}'='"$@"'
+  setopt NO_GLOB_SUBST
+else
+  case `(set -o) 2>/dev/null` in *posix*) set -o posix;; esac
+fi
+BIN_SH=xpg4; export BIN_SH # for Tru64
+DUALCASE=1; export DUALCASE # for MKS sh
+
+
+# PATH needs CR
+# Avoid depending upon Character Ranges.
+as_cr_letters='abcdefghijklmnopqrstuvwxyz'
+as_cr_LETTERS='ABCDEFGHIJKLMNOPQRSTUVWXYZ'
+as_cr_Letters=$as_cr_letters$as_cr_LETTERS
+as_cr_digits='0123456789'
+as_cr_alnum=$as_cr_Letters$as_cr_digits
+
+# The user is always right.
+if test "${PATH_SEPARATOR+set}" != set; then
+  echo "#! /bin/sh" >conf$$.sh
+  echo  "exit 0"   >>conf$$.sh
+  chmod +x conf$$.sh
+  if (PATH="/nonexistent;."; conf$$.sh) >/dev/null 2>&1; then
+    PATH_SEPARATOR=';'
+  else
+    PATH_SEPARATOR=:
+  fi
+  rm -f conf$$.sh
+fi
+
+# Support unset when possible.
+if ( (MAIL=60; unset MAIL) || exit) >/dev/null 2>&1; then
+  as_unset=unset
+else
+  as_unset=false
+fi
+
+
+# IFS
+# We need space, tab and new line, in precisely that order.  Quoting is
+# there to prevent editors from complaining about space-tab.
+# (If _AS_PATH_WALK were called with IFS unset, it would disable word
+# splitting by setting IFS to empty value.)
+as_nl='
+'
+IFS=" ""	$as_nl"
+
+# Find who we are.  Look in the path if we contain no directory separator.
+case $0 in
+  *[\\/]* ) as_myself=$0 ;;
+  *) as_save_IFS=$IFS; IFS=$PATH_SEPARATOR
+for as_dir in $PATH
+do
+  IFS=$as_save_IFS
+  test -z "$as_dir" && as_dir=.
+  test -r "$as_dir/$0" && as_myself=$as_dir/$0 && break
+done
+IFS=$as_save_IFS
+
+     ;;
+esac
+# We did not find ourselves, most probably we were run as `sh COMMAND'
+# in which case we are not to be found in the path.
+if test "x$as_myself" = x; then
+  as_myself=$0
+fi
+if test ! -f "$as_myself"; then
+  echo "$as_myself: error: cannot find myself; rerun with an absolute file name" >&2
+  { (exit 1); exit 1; }
+fi
+
+# Work around bugs in pre-3.0 UWIN ksh.
+for as_var in ENV MAIL MAILPATH
+do ($as_unset $as_var) >/dev/null 2>&1 && $as_unset $as_var
+done
+PS1='$ '
+PS2='> '
+PS4='+ '
+
+# NLS nuisances.
+for as_var in \
+  LANG LANGUAGE LC_ADDRESS LC_ALL LC_COLLATE LC_CTYPE LC_IDENTIFICATION \
+  LC_MEASUREMENT LC_MESSAGES LC_MONETARY LC_NAME LC_NUMERIC LC_PAPER \
+  LC_TELEPHONE LC_TIME
+do
+  if (set +x; test -z "`(eval $as_var=C; export $as_var) 2>&1`"); then
+    eval $as_var=C; export $as_var
+  else
+    ($as_unset $as_var) >/dev/null 2>&1 && $as_unset $as_var
+  fi
+done
+
+# Required to use basename.
+if expr a : '\(a\)' >/dev/null 2>&1 &&
+   test "X`expr 00001 : '.*\(...\)'`" = X001; then
+  as_expr=expr
+else
+  as_expr=false
+fi
+
+if (basename -- /) >/dev/null 2>&1 && test "X`basename -- / 2>&1`" = "X/"; then
+  as_basename=basename
+else
+  as_basename=false
+fi
+
+
+# Name of the executable.
+as_me=`$as_basename -- "$0" ||
+$as_expr X/"$0" : '.*/\([^/][^/]*\)/*$' \| \
+	 X"$0" : 'X\(//\)$' \| \
+	 X"$0" : 'X\(/\)' \| . 2>/dev/null ||
+echo X/"$0" |
+    sed '/^.*\/\([^/][^/]*\)\/*$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\/\(\/\/\)$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\/\(\/\).*/{
+	    s//\1/
+	    q
+	  }
+	  s/.*/./; q'`
+
+# CDPATH.
+$as_unset CDPATH
+
+
+
+  as_lineno_1=$LINENO
+  as_lineno_2=$LINENO
+  test "x$as_lineno_1" != "x$as_lineno_2" &&
+  test "x`expr $as_lineno_1 + 1`" = "x$as_lineno_2" || {
+
+  # Create $as_me.lineno as a copy of $as_myself, but with $LINENO
+  # uniformly replaced by the line number.  The first 'sed' inserts a
+  # line-number line after each line using $LINENO; the second 'sed'
+  # does the real work.  The second script uses 'N' to pair each
+  # line-number line with the line containing $LINENO, and appends
+  # trailing '-' during substitution so that $LINENO is not a special
+  # case at line end.
+  # (Raja R Harinath suggested sed '=', and Paul Eggert wrote the
+  # scripts with optimization help from Paolo Bonzini.  Blame Lee
+  # E. McMahon (1931-1989) for sed's syntax.  :-)
+  sed -n '
+    p
+    /[$]LINENO/=
+  ' <$as_myself |
+    sed '
+      s/[$]LINENO.*/&-/
+      t lineno
+      b
+      :lineno
+      N
+      :loop
+      s/[$]LINENO\([^'$as_cr_alnum'_].*\n\)\(.*\)/\2\1\2/
+      t loop
+      s/-\n.*//
+    ' >$as_me.lineno &&
+  chmod +x "$as_me.lineno" ||
+    { echo "$as_me: error: cannot create $as_me.lineno; rerun with a POSIX shell" >&2
+   { (exit 1); exit 1; }; }
+
+  # Don't try to exec as it changes $[0], causing all sort of problems
+  # (the dirname of $[0] is not the place where we might find the
+  # original and so on.  Autoconf is especially sensitive to this).
+  . "./$as_me.lineno"
+  # Exit status is that of the last command.
+  exit
+}
+
+
+if (as_dir=`dirname -- /` && test "X$as_dir" = X/) >/dev/null 2>&1; then
+  as_dirname=dirname
+else
+  as_dirname=false
+fi
+
+ECHO_C= ECHO_N= ECHO_T=
+case `echo -n x` in
+-n*)
+  case `echo 'x\c'` in
+  *c*) ECHO_T='	';;	# ECHO_T is single tab character.
+  *)   ECHO_C='\c';;
+  esac;;
+*)
+  ECHO_N='-n';;
+esac
+
+if expr a : '\(a\)' >/dev/null 2>&1 &&
+   test "X`expr 00001 : '.*\(...\)'`" = X001; then
+  as_expr=expr
+else
+  as_expr=false
+fi
+
+rm -f conf$$ conf$$.exe conf$$.file
+if test -d conf$$.dir; then
+  rm -f conf$$.dir/conf$$.file
+else
+  rm -f conf$$.dir
+  mkdir conf$$.dir
+fi
+echo >conf$$.file
+if ln -s conf$$.file conf$$ 2>/dev/null; then
+  as_ln_s='ln -s'
+  # ... but there are two gotchas:
+  # 1) On MSYS, both `ln -s file dir' and `ln file dir' fail.
+  # 2) DJGPP < 2.04 has no symlinks; `ln -s' creates a wrapper executable.
+  # In both cases, we have to default to `cp -p'.
+  ln -s conf$$.file conf$$.dir 2>/dev/null && test ! -f conf$$.exe ||
+    as_ln_s='cp -p'
+elif ln conf$$.file conf$$ 2>/dev/null; then
+  as_ln_s=ln
+else
+  as_ln_s='cp -p'
+fi
+rm -f conf$$ conf$$.exe conf$$.dir/conf$$.file conf$$.file
+rmdir conf$$.dir 2>/dev/null
+
+if mkdir -p . 2>/dev/null; then
+  as_mkdir_p=:
+else
+  test -d ./-p && rmdir ./-p
+  as_mkdir_p=false
+fi
+
+# Find out whether ``test -x'' works.  Don't use a zero-byte file, as
+# systems may use methods other than mode bits to determine executability.
+cat >conf$$.file <<_ASEOF
+#! /bin/sh
+exit 0
+_ASEOF
+chmod +x conf$$.file
+if test -x conf$$.file >/dev/null 2>&1; then
+  as_executable_p="test -x"
+else
+  as_executable_p=:
+fi
+rm -f conf$$.file
+
+# Sed expression to map a string onto a valid CPP name.
+as_tr_cpp="eval sed 'y%*$as_cr_letters%P$as_cr_LETTERS%;s%[^_$as_cr_alnum]%_%g'"
+
+# Sed expression to map a string onto a valid variable name.
+as_tr_sh="eval sed 'y%*+%pp%;s%[^_$as_cr_alnum]%_%g'"
+
+
+exec 6>&1
+
+# Save the log message, to keep $[0] and so on meaningful, and to
+# report actual input values of CONFIG_FILES etc. instead of their
+# values after options handling.
+ac_log="
+This file was extended by KLEE $as_me 0.01, which was
+generated by GNU Autoconf 2.60.  Invocation command line was
+
+  CONFIG_FILES    = $CONFIG_FILES
+  CONFIG_HEADERS  = $CONFIG_HEADERS
+  CONFIG_LINKS    = $CONFIG_LINKS
+  CONFIG_COMMANDS = $CONFIG_COMMANDS
+  $ $0 $@
+
+on `(hostname || uname -n) 2>/dev/null | sed 1q`
+"
+
+_ACEOF
+
+cat >>$CONFIG_STATUS <<_ACEOF
+# Files that config.status was made for.
+config_files="$ac_config_files"
+config_headers="$ac_config_headers"
+config_commands="$ac_config_commands"
+
+_ACEOF
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+ac_cs_usage="\
+\`$as_me' instantiates files from templates according to the
+current configuration.
+
+Usage: $0 [OPTIONS] [FILE]...
+
+  -h, --help       print this help, then exit
+  -V, --version    print version number, then exit
+  -q, --quiet      do not print progress messages
+  -d, --debug      don't remove temporary files
+      --recheck    update $as_me by reconfiguring in the same conditions
+  --file=FILE[:TEMPLATE]
+		   instantiate the configuration file FILE
+  --header=FILE[:TEMPLATE]
+		   instantiate the configuration header FILE
+
+Configuration files:
+$config_files
+
+Configuration headers:
+$config_headers
+
+Configuration commands:
+$config_commands
+
+Report bugs to <bug-autoconf at gnu.org>."
+
+_ACEOF
+cat >>$CONFIG_STATUS <<_ACEOF
+ac_cs_version="\\
+KLEE config.status 0.01
+configured by $0, generated by GNU Autoconf 2.60,
+  with options \\"`echo "$ac_configure_args" | sed 's/^ //; s/[\\""\`\$]/\\\\&/g'`\\"
+
+Copyright (C) 2006 Free Software Foundation, Inc.
+This config.status script is free software; the Free Software Foundation
+gives unlimited permission to copy, distribute and modify it."
+
+ac_pwd='$ac_pwd'
+srcdir='$srcdir'
+_ACEOF
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+# If no file are specified by the user, then we need to provide default
+# value.  By we need to know if files were specified by the user.
+ac_need_defaults=:
+while test $# != 0
+do
+  case $1 in
+  --*=*)
+    ac_option=`expr "X$1" : 'X\([^=]*\)='`
+    ac_optarg=`expr "X$1" : 'X[^=]*=\(.*\)'`
+    ac_shift=:
+    ;;
+  *)
+    ac_option=$1
+    ac_optarg=$2
+    ac_shift=shift
+    ;;
+  esac
+
+  case $ac_option in
+  # Handling of the options.
+  -recheck | --recheck | --rechec | --reche | --rech | --rec | --re | --r)
+    ac_cs_recheck=: ;;
+  --version | --versio | --versi | --vers | --ver | --ve | --v | -V )
+    echo "$ac_cs_version"; exit ;;
+  --debug | --debu | --deb | --de | --d | -d )
+    debug=: ;;
+  --file | --fil | --fi | --f )
+    $ac_shift
+    CONFIG_FILES="$CONFIG_FILES $ac_optarg"
+    ac_need_defaults=false;;
+  --header | --heade | --head | --hea )
+    $ac_shift
+    CONFIG_HEADERS="$CONFIG_HEADERS $ac_optarg"
+    ac_need_defaults=false;;
+  --he | --h)
+    # Conflict between --help and --header
+    { echo "$as_me: error: ambiguous option: $1
+Try \`$0 --help' for more information." >&2
+   { (exit 1); exit 1; }; };;
+  --help | --hel | -h )
+    echo "$ac_cs_usage"; exit ;;
+  -q | -quiet | --quiet | --quie | --qui | --qu | --q \
+  | -silent | --silent | --silen | --sile | --sil | --si | --s)
+    ac_cs_silent=: ;;
+
+  # This is an error.
+  -*) { echo "$as_me: error: unrecognized option: $1
+Try \`$0 --help' for more information." >&2
+   { (exit 1); exit 1; }; } ;;
+
+  *) ac_config_targets="$ac_config_targets $1"
+     ac_need_defaults=false ;;
+
+  esac
+  shift
+done
+
+ac_configure_extra_args=
+
+if $ac_cs_silent; then
+  exec 6>/dev/null
+  ac_configure_extra_args="$ac_configure_extra_args --silent"
+fi
+
+_ACEOF
+cat >>$CONFIG_STATUS <<_ACEOF
+if \$ac_cs_recheck; then
+  echo "running CONFIG_SHELL=$SHELL $SHELL $0 "$ac_configure_args \$ac_configure_extra_args " --no-create --no-recursion" >&6
+  CONFIG_SHELL=$SHELL
+  export CONFIG_SHELL
+  exec $SHELL "$0"$ac_configure_args \$ac_configure_extra_args --no-create --no-recursion
+fi
+
+_ACEOF
+cat >>$CONFIG_STATUS <<\_ACEOF
+exec 5>>config.log
+{
+  echo
+  sed 'h;s/./-/g;s/^.../## /;s/...$/ ##/;p;x;p;x' <<_ASBOX
+## Running $as_me. ##
+_ASBOX
+  echo "$ac_log"
+} >&5
+
+_ACEOF
+cat >>$CONFIG_STATUS <<_ACEOF
+#
+# INIT-COMMANDS
+#
+llvm_src="${LLVM_SRC}"
+
+_ACEOF
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+
+# Handling of arguments.
+for ac_config_target in $ac_config_targets
+do
+  case $ac_config_target in
+    "setup") CONFIG_COMMANDS="$CONFIG_COMMANDS setup" ;;
+    "Makefile.config") CONFIG_FILES="$CONFIG_FILES Makefile.config" ;;
+    "stp/Makefile.common") CONFIG_FILES="$CONFIG_FILES stp/Makefile.common" ;;
+    "include/klee/Config/config.h") CONFIG_HEADERS="$CONFIG_HEADERS include/klee/Config/config.h" ;;
+
+  *) { { echo "$as_me:$LINENO: error: invalid argument: $ac_config_target" >&5
+echo "$as_me: error: invalid argument: $ac_config_target" >&2;}
+   { (exit 1); exit 1; }; };;
+  esac
+done
+
+
+# If the user did not use the arguments to specify the items to instantiate,
+# then the envvar interface is used.  Set only those that are not.
+# We use the long form for the default assignment because of an extremely
+# bizarre bug on SunOS 4.1.3.
+if $ac_need_defaults; then
+  test "${CONFIG_FILES+set}" = set || CONFIG_FILES=$config_files
+  test "${CONFIG_HEADERS+set}" = set || CONFIG_HEADERS=$config_headers
+  test "${CONFIG_COMMANDS+set}" = set || CONFIG_COMMANDS=$config_commands
+fi
+
+# Have a temporary directory for convenience.  Make it in the build tree
+# simply because there is no reason against having it here, and in addition,
+# creating and moving files from /tmp can sometimes cause problems.
+# Hook for its removal unless debugging.
+# Note that there is a small window in which the directory will not be cleaned:
+# after its creation but before its name has been assigned to `$tmp'.
+$debug ||
+{
+  tmp=
+  trap 'exit_status=$?
+  { test -z "$tmp" || test ! -d "$tmp" || rm -fr "$tmp"; } && exit $exit_status
+' 0
+  trap '{ (exit 1); exit 1; }' 1 2 13 15
+}
+# Create a (secure) tmp directory for tmp files.
+
+{
+  tmp=`(umask 077 && mktemp -d "./confXXXXXX") 2>/dev/null` &&
+  test -n "$tmp" && test -d "$tmp"
+}  ||
+{
+  tmp=./conf$$-$RANDOM
+  (umask 077 && mkdir "$tmp")
+} ||
+{
+   echo "$me: cannot create a temporary directory in ." >&2
+   { (exit 1); exit 1; }
+}
+
+#
+# Set up the sed scripts for CONFIG_FILES section.
+#
+
+# No need to generate the scripts if there are no CONFIG_FILES.
+# This happens for instance when ./config.status config.h
+if test -n "$CONFIG_FILES"; then
+
+_ACEOF
+
+
+
+ac_delim='%!_!# '
+for ac_last_try in false false false false false :; do
+  cat >conf$$subs.sed <<_ACEOF
+SHELL!$SHELL$ac_delim
+PATH_SEPARATOR!$PATH_SEPARATOR$ac_delim
+PACKAGE_NAME!$PACKAGE_NAME$ac_delim
+PACKAGE_TARNAME!$PACKAGE_TARNAME$ac_delim
+PACKAGE_VERSION!$PACKAGE_VERSION$ac_delim
+PACKAGE_STRING!$PACKAGE_STRING$ac_delim
+PACKAGE_BUGREPORT!$PACKAGE_BUGREPORT$ac_delim
+exec_prefix!$exec_prefix$ac_delim
+prefix!$prefix$ac_delim
+program_transform_name!$program_transform_name$ac_delim
+bindir!$bindir$ac_delim
+sbindir!$sbindir$ac_delim
+libexecdir!$libexecdir$ac_delim
+datarootdir!$datarootdir$ac_delim
+datadir!$datadir$ac_delim
+sysconfdir!$sysconfdir$ac_delim
+sharedstatedir!$sharedstatedir$ac_delim
+localstatedir!$localstatedir$ac_delim
+includedir!$includedir$ac_delim
+oldincludedir!$oldincludedir$ac_delim
+docdir!$docdir$ac_delim
+infodir!$infodir$ac_delim
+htmldir!$htmldir$ac_delim
+dvidir!$dvidir$ac_delim
+pdfdir!$pdfdir$ac_delim
+psdir!$psdir$ac_delim
+libdir!$libdir$ac_delim
+localedir!$localedir$ac_delim
+mandir!$mandir$ac_delim
+DEFS!$DEFS$ac_delim
+ECHO_C!$ECHO_C$ac_delim
+ECHO_N!$ECHO_N$ac_delim
+ECHO_T!$ECHO_T$ac_delim
+LIBS!$LIBS$ac_delim
+build_alias!$build_alias$ac_delim
+host_alias!$host_alias$ac_delim
+target_alias!$target_alias$ac_delim
+LLVM_SRC!$LLVM_SRC$ac_delim
+LLVM_OBJ!$LLVM_OBJ$ac_delim
+build!$build$ac_delim
+build_cpu!$build_cpu$ac_delim
+build_vendor!$build_vendor$ac_delim
+build_os!$build_os$ac_delim
+host!$host$ac_delim
+host_cpu!$host_cpu$ac_delim
+host_vendor!$host_vendor$ac_delim
+host_os!$host_os$ac_delim
+target!$target$ac_delim
+target_cpu!$target_cpu$ac_delim
+target_vendor!$target_vendor$ac_delim
+target_os!$target_os$ac_delim
+KLEE_UCLIBC!$KLEE_UCLIBC$ac_delim
+ENABLE_UCLIBC!$ENABLE_UCLIBC$ac_delim
+ENABLE_POSIX_RUNTIME!$ENABLE_POSIX_RUNTIME$ac_delim
+RUNTIME_ENABLE_OPTIMIZED!$RUNTIME_ENABLE_OPTIMIZED$ac_delim
+RUNTIME_CONFIGURATION!$RUNTIME_CONFIGURATION$ac_delim
+CC!$CC$ac_delim
+CFLAGS!$CFLAGS$ac_delim
+LDFLAGS!$LDFLAGS$ac_delim
+CPPFLAGS!$CPPFLAGS$ac_delim
+ac_ct_CC!$ac_ct_CC$ac_delim
+EXEEXT!$EXEEXT$ac_delim
+OBJEXT!$OBJEXT$ac_delim
+CPP!$CPP$ac_delim
+GREP!$GREP$ac_delim
+EGREP!$EGREP$ac_delim
+HAVE_SELINUX!$HAVE_SELINUX$ac_delim
+CXX!$CXX$ac_delim
+CXXFLAGS!$CXXFLAGS$ac_delim
+ac_ct_CXX!$ac_ct_CXX$ac_delim
+CXXCPP!$CXXCPP$ac_delim
+ENABLE_STPLOG!$ENABLE_STPLOG$ac_delim
+LIBOBJS!$LIBOBJS$ac_delim
+LTLIBOBJS!$LTLIBOBJS$ac_delim
+_ACEOF
+
+  if test `sed -n "s/.*$ac_delim\$/X/p" conf$$subs.sed | grep -c X` = 74; then
+    break
+  elif $ac_last_try; then
+    { { echo "$as_me:$LINENO: error: could not make $CONFIG_STATUS" >&5
+echo "$as_me: error: could not make $CONFIG_STATUS" >&2;}
+   { (exit 1); exit 1; }; }
+  else
+    ac_delim="$ac_delim!$ac_delim _$ac_delim!! "
+  fi
+done
+
+ac_eof=`sed -n '/^CEOF[0-9]*$/s/CEOF/0/p' conf$$subs.sed`
+if test -n "$ac_eof"; then
+  ac_eof=`echo "$ac_eof" | sort -nru | sed 1q`
+  ac_eof=`expr $ac_eof + 1`
+fi
+
+cat >>$CONFIG_STATUS <<_ACEOF
+cat >"\$tmp/subs-1.sed" <<\CEOF$ac_eof
+/@[a-zA-Z_][a-zA-Z_0-9]*@/!b end
+_ACEOF
+sed '
+s/[,\\&]/\\&/g; s/@/@|#_!!_#|/g
+s/^/s,@/; s/!/@,|#_!!_#|/
+:n
+t n
+s/'"$ac_delim"'$/,g/; t
+s/$/\\/; p
+N; s/^.*\n//; s/[,\\&]/\\&/g; s/@/@|#_!!_#|/g; b n
+' >>$CONFIG_STATUS <conf$$subs.sed
+rm -f conf$$subs.sed
+cat >>$CONFIG_STATUS <<_ACEOF
+:end
+s/|#_!!_#|//g
+CEOF$ac_eof
+_ACEOF
+
+
+# VPATH may cause trouble with some makes, so we remove $(srcdir),
+# ${srcdir} and @srcdir@ from VPATH if srcdir is ".", strip leading and
+# trailing colons and then remove the whole line if VPATH becomes empty
+# (actually we leave an empty line to preserve line numbers).
+if test "x$srcdir" = x.; then
+  ac_vpsub='/^[	 ]*VPATH[	 ]*=/{
+s/:*\$(srcdir):*/:/
+s/:*\${srcdir}:*/:/
+s/:*@srcdir@:*/:/
+s/^\([^=]*=[	 ]*\):*/\1/
+s/:*$//
+s/^[^=]*=[	 ]*$//
+}'
+fi
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+fi # test -n "$CONFIG_FILES"
+
+
+for ac_tag in  :F $CONFIG_FILES  :H $CONFIG_HEADERS    :C $CONFIG_COMMANDS
+do
+  case $ac_tag in
+  :[FHLC]) ac_mode=$ac_tag; continue;;
+  esac
+  case $ac_mode$ac_tag in
+  :[FHL]*:*);;
+  :L* | :C*:*) { { echo "$as_me:$LINENO: error: Invalid tag $ac_tag." >&5
+echo "$as_me: error: Invalid tag $ac_tag." >&2;}
+   { (exit 1); exit 1; }; };;
+  :[FH]-) ac_tag=-:-;;
+  :[FH]*) ac_tag=$ac_tag:$ac_tag.in;;
+  esac
+  ac_save_IFS=$IFS
+  IFS=:
+  set x $ac_tag
+  IFS=$ac_save_IFS
+  shift
+  ac_file=$1
+  shift
+
+  case $ac_mode in
+  :L) ac_source=$1;;
+  :[FH])
+    ac_file_inputs=
+    for ac_f
+    do
+      case $ac_f in
+      -) ac_f="$tmp/stdin";;
+      *) # Look for the file first in the build tree, then in the source tree
+	 # (if the path is not absolute).  The absolute path cannot be DOS-style,
+	 # because $ac_f cannot contain `:'.
+	 test -f "$ac_f" ||
+	   case $ac_f in
+	   [\\/$]*) false;;
+	   *) test -f "$srcdir/$ac_f" && ac_f="$srcdir/$ac_f";;
+	   esac ||
+	   { { echo "$as_me:$LINENO: error: cannot find input file: $ac_f" >&5
+echo "$as_me: error: cannot find input file: $ac_f" >&2;}
+   { (exit 1); exit 1; }; };;
+      esac
+      ac_file_inputs="$ac_file_inputs $ac_f"
+    done
+
+    # Let's still pretend it is `configure' which instantiates (i.e., don't
+    # use $as_me), people would be surprised to read:
+    #    /* config.h.  Generated by config.status.  */
+    configure_input="Generated from "`IFS=:
+	  echo $* | sed 's|^[^:]*/||;s|:[^:]*/|, |g'`" by configure."
+    if test x"$ac_file" != x-; then
+      configure_input="$ac_file.  $configure_input"
+      { echo "$as_me:$LINENO: creating $ac_file" >&5
+echo "$as_me: creating $ac_file" >&6;}
+    fi
+
+    case $ac_tag in
+    *:-:* | *:-) cat >"$tmp/stdin";;
+    esac
+    ;;
+  esac
+
+  ac_dir=`$as_dirname -- "$ac_file" ||
+$as_expr X"$ac_file" : 'X\(.*[^/]\)//*[^/][^/]*/*$' \| \
+	 X"$ac_file" : 'X\(//\)[^/]' \| \
+	 X"$ac_file" : 'X\(//\)$' \| \
+	 X"$ac_file" : 'X\(/\)' \| . 2>/dev/null ||
+echo X"$ac_file" |
+    sed '/^X\(.*[^/]\)\/\/*[^/][^/]*\/*$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)[^/].*/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\).*/{
+	    s//\1/
+	    q
+	  }
+	  s/.*/./; q'`
+  { as_dir="$ac_dir"
+  case $as_dir in #(
+  -*) as_dir=./$as_dir;;
+  esac
+  test -d "$as_dir" || { $as_mkdir_p && mkdir -p "$as_dir"; } || {
+    as_dirs=
+    while :; do
+      case $as_dir in #(
+      *\'*) as_qdir=`echo "$as_dir" | sed "s/'/'\\\\\\\\''/g"`;; #(
+      *) as_qdir=$as_dir;;
+      esac
+      as_dirs="'$as_qdir' $as_dirs"
+      as_dir=`$as_dirname -- "$as_dir" ||
+$as_expr X"$as_dir" : 'X\(.*[^/]\)//*[^/][^/]*/*$' \| \
+	 X"$as_dir" : 'X\(//\)[^/]' \| \
+	 X"$as_dir" : 'X\(//\)$' \| \
+	 X"$as_dir" : 'X\(/\)' \| . 2>/dev/null ||
+echo X"$as_dir" |
+    sed '/^X\(.*[^/]\)\/\/*[^/][^/]*\/*$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)[^/].*/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\/\)$/{
+	    s//\1/
+	    q
+	  }
+	  /^X\(\/\).*/{
+	    s//\1/
+	    q
+	  }
+	  s/.*/./; q'`
+      test -d "$as_dir" && break
+    done
+    test -z "$as_dirs" || eval "mkdir $as_dirs"
+  } || test -d "$as_dir" || { { echo "$as_me:$LINENO: error: cannot create directory $as_dir" >&5
+echo "$as_me: error: cannot create directory $as_dir" >&2;}
+   { (exit 1); exit 1; }; }; }
+  ac_builddir=.
+
+case "$ac_dir" in
+.) ac_dir_suffix= ac_top_builddir_sub=. ac_top_build_prefix= ;;
+*)
+  ac_dir_suffix=/`echo "$ac_dir" | sed 's,^\.[\\/],,'`
+  # A ".." for each directory in $ac_dir_suffix.
+  ac_top_builddir_sub=`echo "$ac_dir_suffix" | sed 's,/[^\\/]*,/..,g;s,/,,'`
+  case $ac_top_builddir_sub in
+  "") ac_top_builddir_sub=. ac_top_build_prefix= ;;
+  *)  ac_top_build_prefix=$ac_top_builddir_sub/ ;;
+  esac ;;
+esac
+ac_abs_top_builddir=$ac_pwd
+ac_abs_builddir=$ac_pwd$ac_dir_suffix
+# for backward compatibility:
+ac_top_builddir=$ac_top_build_prefix
+
+case $srcdir in
+  .)  # We are building in place.
+    ac_srcdir=.
+    ac_top_srcdir=$ac_top_builddir_sub
+    ac_abs_top_srcdir=$ac_pwd ;;
+  [\\/]* | ?:[\\/]* )  # Absolute name.
+    ac_srcdir=$srcdir$ac_dir_suffix;
+    ac_top_srcdir=$srcdir
+    ac_abs_top_srcdir=$srcdir ;;
+  *) # Relative name.
+    ac_srcdir=$ac_top_build_prefix$srcdir$ac_dir_suffix
+    ac_top_srcdir=$ac_top_build_prefix$srcdir
+    ac_abs_top_srcdir=$ac_pwd/$srcdir ;;
+esac
+ac_abs_srcdir=$ac_abs_top_srcdir$ac_dir_suffix
+
+
+  case $ac_mode in
+  :F)
+  #
+  # CONFIG_FILE
+  #
+
+_ACEOF
+
+cat >>$CONFIG_STATUS <<\_ACEOF
+# If the template does not know about datarootdir, expand it.
+# FIXME: This hack should be removed a few years after 2.60.
+ac_datarootdir_hack=; ac_datarootdir_seen=
+
+case `sed -n '/datarootdir/ {
+  p
+  q
+}
+/@datadir@/p
+/@docdir@/p
+/@infodir@/p
+/@localedir@/p
+/@mandir@/p
+' $ac_file_inputs` in
+*datarootdir*) ac_datarootdir_seen=yes;;
+*@datadir@*|*@docdir@*|*@infodir@*|*@localedir@*|*@mandir@*)
+  { echo "$as_me:$LINENO: WARNING: $ac_file_inputs seems to ignore the --datarootdir setting" >&5
+echo "$as_me: WARNING: $ac_file_inputs seems to ignore the --datarootdir setting" >&2;}
+_ACEOF
+cat >>$CONFIG_STATUS <<_ACEOF
+  ac_datarootdir_hack='
+  s&@datadir@&$datadir&g
+  s&@docdir@&$docdir&g
+  s&@infodir@&$infodir&g
+  s&@localedir@&$localedir&g
+  s&@mandir@&$mandir&g
+    s&\\\${datarootdir}&$datarootdir&g' ;;
+esac
+_ACEOF
+
+# Neutralize VPATH when `$srcdir' = `.'.
+# Shell code in configure.ac might set extrasub.
+# FIXME: do we really want to maintain this feature?
+cat >>$CONFIG_STATUS <<_ACEOF
+  sed "$ac_vpsub
+$extrasub
+_ACEOF
+cat >>$CONFIG_STATUS <<\_ACEOF
+:t
+/@[a-zA-Z_][a-zA-Z_0-9]*@/!b
+s&@configure_input@&$configure_input&;t t
+s&@top_builddir@&$ac_top_builddir_sub&;t t
+s&@srcdir@&$ac_srcdir&;t t
+s&@abs_srcdir@&$ac_abs_srcdir&;t t
+s&@top_srcdir@&$ac_top_srcdir&;t t
+s&@abs_top_srcdir@&$ac_abs_top_srcdir&;t t
+s&@builddir@&$ac_builddir&;t t
+s&@abs_builddir@&$ac_abs_builddir&;t t
+s&@abs_top_builddir@&$ac_abs_top_builddir&;t t
+$ac_datarootdir_hack
+" $ac_file_inputs | sed -f "$tmp/subs-1.sed" >$tmp/out
+
+test -z "$ac_datarootdir_hack$ac_datarootdir_seen" &&
+  { ac_out=`sed -n '/\${datarootdir}/p' "$tmp/out"`; test -n "$ac_out"; } &&
+  { ac_out=`sed -n '/^[	 ]*datarootdir[	 ]*:*=/p' "$tmp/out"`; test -z "$ac_out"; } &&
+  { echo "$as_me:$LINENO: WARNING: $ac_file contains a reference to the variable \`datarootdir'
+which seems to be undefined.  Please make sure it is defined." >&5
+echo "$as_me: WARNING: $ac_file contains a reference to the variable \`datarootdir'
+which seems to be undefined.  Please make sure it is defined." >&2;}
+
+  rm -f "$tmp/stdin"
+  case $ac_file in
+  -) cat "$tmp/out"; rm -f "$tmp/out";;
+  *) rm -f "$ac_file"; mv "$tmp/out" $ac_file;;
+  esac
+ ;;
+  :H)
+  #
+  # CONFIG_HEADER
+  #
+_ACEOF
+
+# Transform confdefs.h into a sed script `conftest.defines', that
+# substitutes the proper values into config.h.in to produce config.h.
+rm -f conftest.defines conftest.tail
+# First, append a space to every undef/define line, to ease matching.
+echo 's/$/ /' >conftest.defines
+# Then, protect against being on the right side of a sed subst, or in
+# an unquoted here document, in config.status.  If some macros were
+# called several times there might be several #defines for the same
+# symbol, which is useless.  But do not sort them, since the last
+# AC_DEFINE must be honored.
+ac_word_re=[_$as_cr_Letters][_$as_cr_alnum]*
+# These sed commands are passed to sed as "A NAME B PARAMS C VALUE D", where
+# NAME is the cpp macro being defined, VALUE is the value it is being given.
+# PARAMS is the parameter list in the macro definition--in most cases, it's
+# just an empty string.
+ac_dA='s,^\\([	 #]*\\)[^	 ]*\\([	 ]*'
+ac_dB='\\)[	 (].*,\\1define\\2'
+ac_dC=' '
+ac_dD=' ,'
+
+uniq confdefs.h |
+  sed -n '
+	t rset
+	:rset
+	s/^[	 ]*#[	 ]*define[	 ][	 ]*//
+	t ok
+	d
+	:ok
+	s/[\\&,]/\\&/g
+	s/^\('"$ac_word_re"'\)\(([^()]*)\)[	 ]*\(.*\)/ '"$ac_dA"'\1'"$ac_dB"'\2'"${ac_dC}"'\3'"$ac_dD"'/p
+	s/^\('"$ac_word_re"'\)[	 ]*\(.*\)/'"$ac_dA"'\1'"$ac_dB$ac_dC"'\2'"$ac_dD"'/p
+  ' >>conftest.defines
+
+# Remove the space that was appended to ease matching.
+# Then replace #undef with comments.  This is necessary, for
+# example, in the case of _POSIX_SOURCE, which is predefined and required
+# on some systems where configure will not decide to define it.
+# (The regexp can be short, since the line contains either #define or #undef.)
+echo 's/ $//
+s,^[	 #]*u.*,/* & */,' >>conftest.defines
+
+# Break up conftest.defines:
+ac_max_sed_lines=50
+
+# First sed command is:	 sed -f defines.sed $ac_file_inputs >"$tmp/out1"
+# Second one is:	 sed -f defines.sed "$tmp/out1" >"$tmp/out2"
+# Third one will be:	 sed -f defines.sed "$tmp/out2" >"$tmp/out1"
+# et cetera.
+ac_in='$ac_file_inputs'
+ac_out='"$tmp/out1"'
+ac_nxt='"$tmp/out2"'
+
+while :
+do
+  # Write a here document:
+    cat >>$CONFIG_STATUS <<_ACEOF
+    # First, check the format of the line:
+    cat >"\$tmp/defines.sed" <<\\CEOF
+/^[	 ]*#[	 ]*undef[	 ][	 ]*$ac_word_re[	 ]*\$/b def
+/^[	 ]*#[	 ]*define[	 ][	 ]*$ac_word_re[(	 ]/b def
+b
+:def
+_ACEOF
+  sed ${ac_max_sed_lines}q conftest.defines >>$CONFIG_STATUS
+  echo 'CEOF
+    sed -f "$tmp/defines.sed"' "$ac_in >$ac_out" >>$CONFIG_STATUS
+  ac_in=$ac_out; ac_out=$ac_nxt; ac_nxt=$ac_in
+  sed 1,${ac_max_sed_lines}d conftest.defines >conftest.tail
+  grep . conftest.tail >/dev/null || break
+  rm -f conftest.defines
+  mv conftest.tail conftest.defines
+done
+rm -f conftest.defines conftest.tail
+
+echo "ac_result=$ac_in" >>$CONFIG_STATUS
+cat >>$CONFIG_STATUS <<\_ACEOF
+  if test x"$ac_file" != x-; then
+    echo "/* $configure_input  */" >"$tmp/config.h"
+    cat "$ac_result" >>"$tmp/config.h"
+    if diff $ac_file "$tmp/config.h" >/dev/null 2>&1; then
+      { echo "$as_me:$LINENO: $ac_file is unchanged" >&5
+echo "$as_me: $ac_file is unchanged" >&6;}
+    else
+      rm -f $ac_file
+      mv "$tmp/config.h" $ac_file
+    fi
+  else
+    echo "/* $configure_input  */"
+    cat "$ac_result"
+  fi
+  rm -f "$tmp/out12"
+ ;;
+
+  :C)  { echo "$as_me:$LINENO: executing $ac_file commands" >&5
+echo "$as_me: executing $ac_file commands" >&6;}
+ ;;
+  esac
+
+done # for ac_tag
+
+
+{ (exit 0); exit 0; }
+_ACEOF
+chmod +x $CONFIG_STATUS
+ac_clean_files=$ac_clean_files_save
+
+
+# configure is writing to config.log, and then calls config.status.
+# config.status does its own redirection, appending to config.log.
+# Unfortunately, on DOS this fails, as config.log is still kept open
+# by configure, so config.status won't be able to write to it; its
+# output is simply discarded.  So we exec the FD to /dev/null,
+# effectively closing config.log, so it can be properly (re)opened and
+# appended to by config.status.  When coming back to configure, we
+# need to make the FD available again.
+if test "$no_create" != yes; then
+  ac_cs_success=:
+  ac_config_status_args=
+  test "$silent" = yes &&
+    ac_config_status_args="$ac_config_status_args --quiet"
+  exec 5>/dev/null
+  $SHELL $CONFIG_STATUS $ac_config_status_args || ac_cs_success=false
+  exec 5>>config.log
+  # Use ||, not &&, to avoid exiting from the if with $? = 1, which
+  # would make configure fail if this is the last instruction.
+  $ac_cs_success || { (exit 1); exit 1; }
+fi
+

Propchange: klee/trunk/configure

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/docs/SMT-COMP/BitVector_ArraysEx.smt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/SMT-COMP/BitVector_ArraysEx.smt?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/SMT-COMP/BitVector_ArraysEx.smt (added)
+++ klee/trunk/docs/SMT-COMP/BitVector_ArraysEx.smt Wed May 20 23:36:41 2009
@@ -0,0 +1,74 @@
+(theory BitVector_ArraysEx
+
+ :written_by {Clark Barrett}
+ :date {May 7, 2007}
+ 
+:sorts_description {
+    All sort symbols of the form BitVec[m],
+    where m is a numeral greater than 0.
+
+    All sort symbols of the form Array[m:n],
+    where m and n are numerals with m > 0 and n > 0.
+}
+
+:funs_description {
+    All functions from the theory Fixed_Size_Bitvectors.
+}
+
+:funs_description {
+    All function symbols with arity of the form
+
+      (select Array[m:n] BitVec[m] BitVec[n])
+
+    where
+    - m,n are numerals
+    - m > 0, n > 0
+}
+
+:funs_description {
+    All function symbols with arity of the form
+
+      (store Array[m:n] BitVec[m] BitVec[n] Array[m:n])
+
+    where
+    - m,n are numerals
+    - m > 0, n > 0
+}
+
+:preds_description {
+    All predicates from the theory Fixed_Size_Bitvectors.
+}
+
+
+ :definition 
+ "This is a theory containing an infinite number of copies of the theory of
+  functional arrays with extensionality: one for each pair of bitvector sorts.
+  It can be formally defined as the union of the SMT-LIB theory
+  Fixed_Size_Bitvectors and an infinite number of variants of the SMT-LIB
+  theory ArraysEx: one for each distinct signature morphism mapping the sort Index
+  to BitVec[m] and the sort Element to Bitvec[n] where m and n range over all
+  positive numerals.  In each of the copies of ArraysEx, the sort Array is
+  renamed to Array[m:n] and each copy of ArraysEx contributes exactly one select
+  function and one store function to the infinite polymorphic family of select
+  and store functions described above.
+ "
+
+ :notes
+ "As in the theory Fixed_Size_Bitvectors, this theory does not
+  provide a value for the formal attributes :sorts, :funs, and :preds because
+  there are an infinite number of them.  See the notes in theory
+  Fixed_Size_Bitvectors for details.
+
+  If for i=1,2, T_i is an SMT-LIB theory with sorts S_i, function symbols F_i,
+  predicate symbols P_i, and axioms A_i, by \"union\" of T_1 and T_2 
+  we mean the theory T with sorts S_1 U S_2, function symbols F_1 U F_2,
+  predicate symbols P_1 U P_2, and axioms A_1 U A_2 (where U stands for set
+  theoretic union).
+
+  The theory T is a well-defined SMT-LIB theory whenever S_1, S_2, F_1, F_2, 
+  P_1, P_2 are all pairwise disjoint, as is the case for the component theories
+  considered here. 
+ "
+)
+
+

Added: klee/trunk/docs/SMT-COMP/BitVectors.smt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/SMT-COMP/BitVectors.smt?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/SMT-COMP/BitVectors.smt (added)
+++ klee/trunk/docs/SMT-COMP/BitVectors.smt Wed May 20 23:36:41 2009
@@ -0,0 +1,187 @@
+(theory Fixed_Size_BitVectors
+
+:written_by {Silvio Ranise, Cesare Tinelli, and Clark Barrett}
+
+:date {May 7, 2007}
+
+:notes
+   "Against the requirements of the current SMT-LIB standard this theory does
+    not provide a value for the formal attributes :sorts, :funs, and :preds.
+    The reason is that the theory has an infinite number of sort, function, and
+    predicate symbols, and so they cannot be specified formally in the current
+    SMT-LIB language.  While extending SMT-LIB's type system with dependent
+    types would allow a finitary formal specification of all the symbols in
+    this theory's signature, such an extension does not seem to be worth the
+    trouble at the moment.  As a temporary ad-hoc solution, this theory
+    declaration specifies the signature, in English, in the user-defined
+    attributes :sorts_description, :funs_description, and :preds_description.
+    "
+
+:sorts_description {
+    All sort symbols of the form BitVec[m],
+    where m is a numeral greater than 0.
+}
+
+:funs_description {
+    Constant symbols bit0 and bit1 of sort BitVec[1]
+}
+
+:funs_description {
+    All function symbols with arity of the form
+
+      (concat BitVec[i] BitVec[j] BitVec[m])
+
+    where
+    - i,j,m are numerals
+    - i,j > 0
+    - i + j = m
+}
+
+:funs_description {
+    All function symbols with arity of the form
+
+      (extract[i:j] BitVec[m] BitVec[n])
+
+    where
+    - i,j,m,n are numerals
+    - m > i >= j >= 0,
+    - n = i-j+1.
+}
+
+:funs_description {
+    All function symbols with arity of the form
+
+       (op1 BitVec[m] BitVec[m])
+    or
+       (op2 BitVec[m] BitVec[m] BitVec[m])
+
+    where
+    - op1 is from {bvnot, bvneg}
+    - op2 is from {bvand, bvor, bvadd, bvmul, bvudiv, bvurem, bvshl, bvlshr}
+    - m is a numeral greater than 0
+}
+
+:preds_description {
+    All predicate symbols with arity of the form
+
+       (pred BitVec[m] BitVec[m])
+
+    where
+    - pred is from {bvult}
+    - m is a numeral greater than 0
+}
+
+:definition
+  "This is a core theory for fixed-size bitvectors where the operations
+   of concatenation and extraction of bitvectors as well as the usual
+   logical and arithmetic operations are overloaded.
+   The theory is defined semantically as follows.
+
+   The sort BitVec[m] (for m > 0) is the set of finite functions
+   whose domain is the initial segment of the naturals [0...m), meaning
+   that 0 is included and m is excluded, and the co-domain is {0,1}.
+
+   The semantic interpretation [[_]] of well-sorted BitVec-terms is
+   inductively defined as follows.
+
+   - Variables
+
+   If v is a variable of sort BitVec[m] with 0 < m, then
+   [[v]] is some element of [{0,...,m-1} -> {0,1}], the set of total
+   functions from {0,...,m-1} to {0,1}.
+
+   - Constant symbols bit0 and bit1 of sort BitVec[1]
+
+   [[bit0]] := \lambda x : [0,1). 0
+   [[bit1]] := \lambda x : [0,1). 1
+
+   - Function symbols for concatenation
+
+   [[(concat s t)]] := \lambda x : [0...n+m).
+                          if (x<m) then [[t]](x) else [[s]](x-m)
+   where
+   s and t are terms of sort BitVec[n] and BitVec[m], respectively,
+   0 < n, 0 < m.
+
+   - Function symbols for extraction
+
+   [[(extract[i:j] s)]] := \lambda x : [0...i-j+1). [[s]](j+x)
+   where s is of sort BitVec[l], 0 <= j <= i < l.
+
+   - Bit-wise operations
+
+   [[(bvnot s)]] := \lambda x : [0...m). if [[s]](x) = 0 then 1 else 0
+
+   [[(bvand s t)]] := \lambda x : [0...m).
+                         if [[s]](x) = 0 then 0 else [[t]](x)
+
+   [[(bvor s t)]] := \lambda x : [0...m).
+                         if [[s]](x) = 1 then 1 else [[t]](x)
+
+   where s and t are both of sort BitVec[m] and 0 < m.
+
+   - Arithmetic operations
+
+   To define the semantics of the bitvector arithmetic operators, we first
+   introduce some additional definitions:
+
+   o (x div y) where x and y are integers with x >= 0 and y > 0 returns the
+     integer part of x divided by y (i.e., truncated integer division).
+
+   o (x rem y) where x and y are integers with x >= 0 and y > 0 returns the
+     remainder when x is divided by y.  Note that we always have the following
+     equivalence (for y > 0): (x div y) * y + (x rem y) = x.
+
+   o bv2nat which takes a bitvector b: [0...m) --> {0,1}
+     with 0 < m, and returns an integer in the range [0...2^m),
+     and is defined as follows:
+
+       bv2nat(b) := b(m-1)*2^{m-1} + b(m-2)*2^{m-2} + ... + b(0)*2^0
+
+   o nat2bv[m], with 0 < m, which takes a non-negative integer
+     n and returns the (unique) bitvector b: [0,...,m) -> {0,1}
+     such that
+
+       b(m-1)*2^{m-1} + ... + b(0)*2^0 = n rem 2^m
+
+   Now, we can define the following operations.  Suppose s and t are both terms
+   of sort BitVec[m], m > 0.
+
+   [[(bvneg s)]] := nat2bv[m](2^m - bv2nat([[s]]))
+
+   [[(bvadd s t)]] := nat2bv[m](bv2nat([[s]]) + bv2nat([[t]]))
+
+   [[(bvmul s t)]] := nat2bv[m](bv2nat([[s]]) * bv2nat([[t]]))
+
+   [[(bvudiv s t)]] := if bv2nat([[t]]) != 0 then
+                          nat2bv[m](bv2nat([[s]]) div bv2nat([[t]]))
+
+   [[(bvurem s t)]] := if bv2nat([[t]]) != 0 then
+                          nat2bv[m](bv2nat([[s]]) rem bv2nat([[t]]))
+
+   - Shift operations
+
+   Suppose s and t are both terms of sort BitVec[m], m > 0.  We make use of the
+   definitions given for the arithmetic operations, above.
+
+   [[(bvshl s t)]] := nat2bv[m](bv2nat([[s]]) * 2^(bv2nat([[t]])))
+
+   [[(bvlshr s t)]] := nat2bv[m](bv2nat([[s]]) div 2^(bv2nat([[t]])))
+
+   Finally, we can define the binary predicate bvult:
+
+   (bvult s t) is interpreted to be true iff bv2nat([[s]]) < bv2nat([[t]])
+
+   Note that the semantic interpretation above is underspecified because it
+   does not specify the meaning of (bvudiv s t) or (bvurem s t) in case
+   bv2nat([[t]]) is 0.  Since the semantics of SMT-LIB's underlying logic
+   associates *total* functions to function symbols, we then consider as models
+   of this theory *any* interpretation conforming to the specifications above
+   (and defining bvudiv and bvurem arbitrarily when the second argument
+   evaluates to 0).  Benchmarks using this theory should only include a
+   :status sat or :status unsat attribute if the status is independent of
+   the particular choice of model for the theory.
+
+  "
+
+)

Added: klee/trunk/docs/SMT-COMP/QF_AUFBV.smt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/SMT-COMP/QF_AUFBV.smt?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/SMT-COMP/QF_AUFBV.smt (added)
+++ klee/trunk/docs/SMT-COMP/QF_AUFBV.smt Wed May 20 23:36:41 2009
@@ -0,0 +1,19 @@
+(logic QF_AUFBV
+
+ :written_by {Clark Barrett}
+ :date {May 7, 2007}
+ 
+ :theory BV_ArraysEx
+
+ :language 
+ "Closed quantifier-free formulas built over an arbitrary expansion of the
+  BV_ArraysEx signature with free function and predicate symbols over
+  the sorts of BV_ArraysEx.  Formulas in ite terms must satisfy the same
+  restriction as well, with the exception that they need not be closed (because
+  they may be in the scope of a let expression).
+ "
+ :extensions
+ "As in the logic QF_BV."
+)
+
+

Added: klee/trunk/docs/SMT-COMP/QF_BV.smt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/SMT-COMP/QF_BV.smt?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/SMT-COMP/QF_BV.smt (added)
+++ klee/trunk/docs/SMT-COMP/QF_BV.smt Wed May 20 23:36:41 2009
@@ -0,0 +1,261 @@
+(logic QF_BV
+
+:written_by {Silvio Ranise, Cesare Tinelli, and Clark Barrett}
+:date {May 7, 2007}
+
+:theory Fixed_Size_BitVectors
+
+:language
+
+ "Closed quantifier-free formulas built over an arbitrary expansion of the
+  Fixed_Size_BitVectors signature with free constant symbols over the sorts
+  BitVec[m] for 0 < m.  Formulas in ite terms must satisfy the same restriction
+  as well, with the exception that they need not be closed (because they may be
+  in the scope of a let expression).
+ "
+
+:notes
+ "For quick reference, the following is a brief and informal summary of the
+  legal symbols in this logic and their meaning (formal definitions are found
+  either in the Fixed_Size_Bitvectors theory, or in the extensions below).
+
+  Defined in theory Fixed_Size_Bitvectors:
+
+    Functions/Constants:
+
+    (bit0 BitVec[1])
+      - the constant consisting of a single bit with value 0
+    (bit1 BitVec[1])
+      - the constant consisting of a single bit with value 1
+    (concat BitVec[i] BitVec[j] BitVec[m])
+      - concatenation of bitvectors of size i and j to get a new bitvector of
+        size m, where m = i + j
+    (extract[i:j] BitVec[m] BitVec[n])
+      - extraction of bits i down to j from a bitvector of size m to yield a
+        new bitvector of size n, where n = i - j + 1
+    (bvnot BitVec[m] BitVec[m])
+      - bitwise negation
+    (bvand BitVec[m] BitVec[m] BitVec[m])
+      - bitwise and
+    (bvor BitVec[m] BitVec[m] BitVec[m])
+      - bitwise or
+    (bvneg BitVec[m] BitVec[m])
+      - 2's complement unary minus
+    (bvadd BitVec[m] BitVec[m] BitVec[m])
+      - addition modulo 2^m
+    (bvmul BitVec[m] BitVec[m] BitVec[m])
+      - multiplication modulo 2^m
+    (bvudiv BitVec[m] BitVec[m] BitVec[m])
+      - unsigned division, truncating towards 0 (undefined if divisor is 0)
+    (bvurem BitVec[m] BitVec[m] BitVec[m])
+      - unsigned remainder from truncating division (undefined if divisor is 0)
+    (bvshl BitVec[m] BitVec[m] BitVec[m])
+      - shift left (equivalent to multiplication by 2^x where x is the value of
+        the second argument)
+    (bvlshr BitVec[m] BitVec[m] BitVec[m])
+      - logical shift right (equivalent to unsigned division by 2^x where x is
+        the value of the second argument)
+
+    Predicates:
+
+    (bvult BitVec[m] BitVec[m])
+      - binary predicate for unsigned less than
+
+  Defined below:
+
+    Functions/Constants:
+
+    Bitvector constants:
+      - bvX[m] where X is a numeral in base 10 defines the bitvector constant
+        with numeric value X of size m.
+      - bvbinX where X is a binary numeral of length m defines the
+        bitvector constant with value X and size m.
+      - bvhexX where X is a hexadecimal numeral of length m defines the
+        bitvector constant with value X and size 4*m.
+    (bvnand BitVec[m] BitVec[m] BitVec[m])
+      - bitwise nand (negation of and)
+    (bvnor BitVec[m] BitVec[m] BitVec[m])
+      - bitwise nor (negation of or)
+    (bvxor BitVec[m] BitVec[m] BitVec[m])
+      - bitwise exclusive or
+    (bvxnor BitVec[m] BitVec[m] BitVec[m])
+      - bitwise equivalence (equivalently, negation of bitwise exclusive or)
+    (bvcomp BitVec[m] BitVec[m] BitVec[1])
+      - bit comparator: equals bit1 iff all bits are equal
+    (bvsub BitVec[m] BitVec[m] BitVec[m])
+      - 2's complement subtraction modulo 2^m
+    (bvsdiv BitVec[m] BitVec[m] BitVec[m])
+      - 2's complement signed division
+    (bvsrem BitVec[m] BitVec[m] BitVec[m])
+      - 2's complement signed remainder (sign follows dividend)
+    (bvsmod BitVec[m] BitVec[m] BitVec[m])
+      - 2's complement signed remainder (sign follows divisor)
+    (bvashr BitVec[m] BitVec[m] BitVec[m])
+      - Arithmetic shift right, like logical shift right except that the most
+        significant bits of the result always copy the most significant
+        bit of the first argument.
+
+    The following symbols are parameterized by the numeral i, where i >= 0.
+
+    (repeat[i] BitVec[m] BitVec[i*m])
+      - (repeat[i] x) means concatenate i copies of x
+    (zero_extend[i] BitVec[m] BitVec[m+i])
+      - (zero_extend[i] x) means extend x with zeroes to the (unsigned)
+        equivalent bitvector of size m+i
+    (sign_extend[i] BitVec[m] BitVec[m+i])
+      - (sign_extend[i] x) means extend x to the (signed) equivalent bitvector
+        of size m+i
+    (rotate_left[i] BitVec[m] BitVec[m])
+      - (rotate_left[i] x) means rotate bits of x to the left i times
+    (rotate_right[i] BitVec[m] BitVec[m])
+      - (rotate_right[i] x) means rotate bits of x to the right y times
+
+    Predicates:
+
+    (bvule BitVec[m] BitVec[m])
+      - binary predicate for unsigned less than or equal
+    (bvugt BitVec[m] BitVec[m])
+      - binary predicate for unsigned greater than
+    (bvuge BitVec[m] BitVec[m])
+      - binary predicate for unsigned greater than or equal
+    (bvslt BitVec[m] BitVec[m])
+      - binary predicate for signed less than
+    (bvsle BitVec[m] BitVec[m])
+      - binary predicate for signed less than or equal
+    (bvsgt BitVec[m] BitVec[m])
+      - binary predicate for signed greater than
+    (bvsge BitVec[m] BitVec[m])
+      - binary predicate for signed greater than or equal
+
+ "
+
+:extensions
+ "Below, let |exp| denote the integer resulting from the evaluation
+  of the arithmetic expression exp.
+
+  - Bitvector Constants:
+    The string bv followed by the numeral n and a size [m] (as in bv13[32])
+    abbreviates any term t of sort BitVec[m] built only out of the symbols in
+    {concat, bit0, bit1} such that
+
+    [[t]] = nat2bv[m](n) for n=0, ..., 2^m - 1.
+
+    See the specification of the theory's semantics for a definition
+    of the functions [[_]] and nat2bv.  Note that this convention implicitly
+    considers the numeral n as a number written in base 10.
+
+    For backward compatibility, if the size [m] is omitted, then the size is
+    assumed to be 32.
+
+    The string bvbin followed by a sequence of 0's and 1's abbreviates the
+    concatenation of a similar sequence of bit0 and bit1 terms.  Thus,
+    if n is the numeral represented in base 2 by the sequence of 0's and 1's
+    and m is the length of the sequence, then the term represents
+    nat2bv[m](n).  For example bvbin0101 is equivalent to bv5[4].
+
+    The string bvhex followed by a sequence of digits and/or letters from A to
+    F is interpreted similarly as a concatenation of bit0 and bit1 as follows.
+    If n is the numeral represented in hexadecimal (base 16) by the sequence of
+    digits and letters from A to F and m is four times the length of the
+    sequence, then the term represents nat2bv[m](n).  For example, bvbinFF is
+    equivalent to bv255[8].  Letters in the hexadecimal sequence may be in
+    either upper or lower case.
+
+  - Bitwise operators
+
+    For all terms s,t of sort BitVec[m], where 0 < m,
+
+    (bvnand s t) abbreviates (bvnot (bvand s t))
+    (bvnor s t) abbreviates (bvnot (bvor s t))
+    (bvxor s t) abbreviates (bvor (bvand s (bvnot t)) (bvand (bvnot s) t))
+    (bvxnor s t) abbreviates (bvor (bvand s t) (bvand (bvnot s) (bvnot t)))
+    (bvcomp s t) abbreviates (bvxnor s t) if m = 1, and
+       (bvand (bvxnor (extract[|m-1|:|m-1|] s) (extract[|m-1|:|m-1|] t))
+              (bvcomp (extract[|m-2|:0] s) (extract[|m-2|:0] t))) otherwise
+
+  - Arithmetic operators
+
+    For all terms s,t of sort BitVec[m], where 0 < m,
+
+    (bvsub s t) abbreviates (bvadd s (bvneg t))
+    (bvsdiv s t) abbreviates
+      (let (?msb_s (extract[|m-1|:|m-1|] s))
+      (let (?msb_t (extract[|m-1|:|m-1|] t))
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit0))
+           (bvudiv s t)
+      (ite (and (= ?msb_s bit1) (= ?msb_t bit0))
+           (bvneg (bvudiv (bvneg s) t))
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit1))
+           (bvneg (bvudiv s (bvneg t)))
+           (bvudiv (bvneg s) (bvneg t)))))))
+    (bvsrem s t) abbreviates
+      (let (?msb_s (extract[|m-1|:|m-1|] s))
+      (let (?msb_t (extract[|m-1|:|m-1|] t))
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit0))
+           (bvurem s t)
+      (ite (and (= ?msb_s bit1) (= ?msb_t bit0))
+           (bvneg (bvurem (bvneg s) t))
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit1))
+           (bvurem s (bvneg t)))
+           (bvneg (bvurem (bvneg s) (bvneg t)))))))
+    (bvsmod s t) abbreviates
+      (let (?msb_s (extract[|m-1|:|m-1|] s))
+      (let (?msb_t (extract[|m-1|:|m-1|] t))
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit0))
+           (bvurem s t)
+      (ite (and (= ?msb_s bit1) (= ?msb_t bit0))
+           (bvadd (bvneg (bvurem (bvneg s) t)) t)
+      (ite (and (= ?msb_s bit0) (= ?msb_t bit1))
+           (bvadd (bvurem s (bvneg t)) t)
+           (bvneg (bvurem (bvneg s) (bvneg t)))))))
+    (bvule s t) abbreviates (or (bvult s t) (= s t))
+    (bvugt s t) abbreviates (bvult t s)
+    (bvuge s t) abbreviates (or (bvult t s) (= s t))
+    (bvslt s t) abbreviates:
+      (or (and (= (extract[|m-1|:|m-1|] s) bit1)
+               (= (extract[|m-1|:|m-1|] t) bit0))
+          (and (= (extract[|m-1|:|m-1|] s) (extract[|m-1|:|m-1|] t))
+               (bvult s t)))
+    (bvsle s t) abbreviates:
+      (or (and (= (extract[|m-1|:|m-1|] s) bit1)
+               (= (extract[|m-1|:|m-1|] t) bit0))
+          (and (= (extract[|m-1|:|m-1|] s) (extract[|m-1|:|m-1|] t))
+               (bvule s t)))
+    (bvsgt s t) abbreviates (bvslt t s)
+    (bvsge s t) abbreviates (bvsle t s)
+
+  - Other operations
+
+    For all numerals j > 1 and 0 < m, and all terms of s and t of
+    sort BitVec[m],
+
+    (bvashr s t) abbreviates 
+      (ite (= (extract[|m-1|:|m-1|] s) bit0)
+           (bvlshr s t)
+           (bvnot (bvlshr (bvnot s) t)))
+
+    (repeat[1] t) stands for t
+    (repeat[j] t) abbreviates (concat t (repeat[|j-1|] t))
+
+    (zero_extend[0] t) stands for t
+    (zero_extend[j] t) abbreviates (concat (repeat[j] bit0) t)
+
+    (sign_extend[0] t) stands for t
+    (sign_extend[j] t) abbreviates
+      (concat (repeat[j] (extract[|m-1|:|m-1|] t)) t)
+
+    (rotate_left[0] t) stands for t
+    (rotate_left[j] t) abbreviates t if m = 1, and
+      (rotate_left[|j-1|]
+        (concat (extract[|m-2|:0] t) (extract[|m-1|:|m-1|] t))
+      otherwise
+
+    (rotate_right[0] t) stands for t
+    (rotate_right[j] t) abbreviates t if m = 1, and
+      (rotate_right[|j-1|]
+        (concat (extract[0:0] t) (extract[|m-1|:1] t)))
+      otherwise
+
+ "
+)
+

Added: klee/trunk/docs/doxygen.cfg
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/doxygen.cfg?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/doxygen.cfg (added)
+++ klee/trunk/docs/doxygen.cfg Wed May 20 23:36:41 2009
@@ -0,0 +1,1291 @@
+# Doxyfile 1.5.3
+
+# This file describes the settings to be used by the documentation system
+# doxygen (www.doxygen.org) for a project
+#
+# All text after a hash (#) is considered a comment and will be ignored
+# The format is:
+#       TAG = value [value, ...]
+# For lists items can also be appended using:
+#       TAG += value [value, ...]
+# Values that contain spaces should be placed between quotes (" ")
+
+#---------------------------------------------------------------------------
+# Project related configuration options
+#---------------------------------------------------------------------------
+
+# This tag specifies the encoding used for all characters in the config file that 
+# follow. The default is UTF-8 which is also the encoding used for all text before 
+# the first occurrence of this tag. Doxygen uses libiconv (or the iconv built into 
+# libc) for the transcoding. See http://www.gnu.org/software/libiconv for the list of 
+# possible encodings.
+
+DOXYFILE_ENCODING      = UTF-8
+
+# The PROJECT_NAME tag is a single word (or a sequence of words surrounded 
+# by quotes) that should identify the project.
+
+PROJECT_NAME           = klee
+
+# The PROJECT_NUMBER tag can be used to enter a project or revision number. 
+# This could be handy for archiving the generated documentation or 
+# if some version control system is used.
+
+PROJECT_NUMBER         = 
+
+# The OUTPUT_DIRECTORY tag is used to specify the (relative or absolute) 
+# base path where the generated documentation will be put. 
+# If a relative path is entered, it will be relative to the location 
+# where doxygen was started. If left blank the current directory will be used.
+
+OUTPUT_DIRECTORY       = docs/doxygen
+
+# If the CREATE_SUBDIRS tag is set to YES, then doxygen will create 
+# 4096 sub-directories (in 2 levels) under the output directory of each output 
+# format and will distribute the generated files over these directories. 
+# Enabling this option can be useful when feeding doxygen a huge amount of 
+# source files, where putting all generated files in the same directory would 
+# otherwise cause performance problems for the file system.
+
+CREATE_SUBDIRS         = NO
+
+# The OUTPUT_LANGUAGE tag is used to specify the language in which all 
+# documentation generated by doxygen is written. Doxygen will use this 
+# information to generate all constant output in the proper language. 
+# The default language is English, other supported languages are: 
+# Afrikaans, Arabic, Brazilian, Catalan, Chinese, Chinese-Traditional, 
+# Croatian, Czech, Danish, Dutch, Finnish, French, German, Greek, Hungarian, 
+# Italian, Japanese, Japanese-en (Japanese with English messages), Korean, 
+# Korean-en, Lithuanian, Norwegian, Polish, Portuguese, Romanian, Russian, 
+# Serbian, Slovak, Slovene, Spanish, Swedish, and Ukrainian.
+
+OUTPUT_LANGUAGE        = English
+
+# If the BRIEF_MEMBER_DESC tag is set to YES (the default) Doxygen will 
+# include brief member descriptions after the members that are listed in 
+# the file and class documentation (similar to JavaDoc). 
+# Set to NO to disable this.
+
+BRIEF_MEMBER_DESC      = YES
+
+# If the REPEAT_BRIEF tag is set to YES (the default) Doxygen will prepend 
+# the brief description of a member or function before the detailed description. 
+# Note: if both HIDE_UNDOC_MEMBERS and BRIEF_MEMBER_DESC are set to NO, the 
+# brief descriptions will be completely suppressed.
+
+REPEAT_BRIEF           = YES
+
+# This tag implements a quasi-intelligent brief description abbreviator 
+# that is used to form the text in various listings. Each string 
+# in this list, if found as the leading text of the brief description, will be 
+# stripped from the text and the result after processing the whole list, is 
+# used as the annotated text. Otherwise, the brief description is used as-is. 
+# If left blank, the following values are used ("$name" is automatically 
+# replaced with the name of the entity): "The $name class" "The $name widget" 
+# "The $name file" "is" "provides" "specifies" "contains" 
+# "represents" "a" "an" "the"
+
+ABBREVIATE_BRIEF       = 
+
+# If the ALWAYS_DETAILED_SEC and REPEAT_BRIEF tags are both set to YES then 
+# Doxygen will generate a detailed section even if there is only a brief 
+# description.
+
+ALWAYS_DETAILED_SEC    = NO
+
+# If the INLINE_INHERITED_MEMB tag is set to YES, doxygen will show all 
+# inherited members of a class in the documentation of that class as if those 
+# members were ordinary class members. Constructors, destructors and assignment 
+# operators of the base classes will not be shown.
+
+INLINE_INHERITED_MEMB  = NO
+
+# If the FULL_PATH_NAMES tag is set to YES then Doxygen will prepend the full 
+# path before files name in the file list and in the header files. If set 
+# to NO the shortest path that makes the file name unique will be used.
+
+FULL_PATH_NAMES        = YES
+
+# If the FULL_PATH_NAMES tag is set to YES then the STRIP_FROM_PATH tag 
+# can be used to strip a user-defined part of the path. Stripping is 
+# only done if one of the specified strings matches the left-hand part of 
+# the path. The tag can be used to show relative paths in the file list. 
+# If left blank the directory from which doxygen is run is used as the 
+# path to strip.
+
+STRIP_FROM_PATH        = 
+
+# The STRIP_FROM_INC_PATH tag can be used to strip a user-defined part of 
+# the path mentioned in the documentation of a class, which tells 
+# the reader which header file to include in order to use a class. 
+# If left blank only the name of the header file containing the class 
+# definition is used. Otherwise one should specify the include paths that 
+# are normally passed to the compiler using the -I flag.
+
+STRIP_FROM_INC_PATH    = 
+
+# If the SHORT_NAMES tag is set to YES, doxygen will generate much shorter 
+# (but less readable) file names. This can be useful is your file systems 
+# doesn't support long names like on DOS, Mac, or CD-ROM.
+
+SHORT_NAMES            = NO
+
+# If the JAVADOC_AUTOBRIEF tag is set to YES then Doxygen 
+# will interpret the first line (until the first dot) of a JavaDoc-style 
+# comment as the brief description. If set to NO, the JavaDoc 
+# comments will behave just like regular Qt-style comments 
+# (thus requiring an explicit @brief command for a brief description.)
+
+JAVADOC_AUTOBRIEF      = NO
+
+# If the QT_AUTOBRIEF tag is set to YES then Doxygen will 
+# interpret the first line (until the first dot) of a Qt-style 
+# comment as the brief description. If set to NO, the comments 
+# will behave just like regular Qt-style comments (thus requiring 
+# an explicit \brief command for a brief description.)
+
+QT_AUTOBRIEF           = NO
+
+# The MULTILINE_CPP_IS_BRIEF tag can be set to YES to make Doxygen 
+# treat a multi-line C++ special comment block (i.e. a block of //! or /// 
+# comments) as a brief description. This used to be the default behaviour. 
+# The new default is to treat a multi-line C++ comment block as a detailed 
+# description. Set this tag to YES if you prefer the old behaviour instead.
+
+MULTILINE_CPP_IS_BRIEF = NO
+
+# If the INHERIT_DOCS tag is set to YES (the default) then an undocumented 
+# member inherits the documentation from any documented member that it 
+# re-implements.
+
+INHERIT_DOCS           = YES
+
+# If the SEPARATE_MEMBER_PAGES tag is set to YES, then doxygen will produce 
+# a new page for each member. If set to NO, the documentation of a member will 
+# be part of the file/class/namespace that contains it.
+
+SEPARATE_MEMBER_PAGES  = NO
+
+# The TAB_SIZE tag can be used to set the number of spaces in a tab. 
+# Doxygen uses this value to replace tabs by spaces in code fragments.
+
+TAB_SIZE               = 8
+
+# This tag can be used to specify a number of aliases that acts 
+# as commands in the documentation. An alias has the form "name=value". 
+# For example adding "sideeffect=\par Side Effects:\n" will allow you to 
+# put the command \sideeffect (or @sideeffect) in the documentation, which 
+# will result in a user-defined paragraph with heading "Side Effects:". 
+# You can put \n's in the value part of an alias to insert newlines.
+
+ALIASES                = 
+
+# Set the OPTIMIZE_OUTPUT_FOR_C tag to YES if your project consists of C 
+# sources only. Doxygen will then generate output that is more tailored for C. 
+# For instance, some of the names that are used will be different. The list 
+# of all members will be omitted, etc.
+
+OPTIMIZE_OUTPUT_FOR_C  = NO
+
+# Set the OPTIMIZE_OUTPUT_JAVA tag to YES if your project consists of Java 
+# sources only. Doxygen will then generate output that is more tailored for Java. 
+# For instance, namespaces will be presented as packages, qualified scopes 
+# will look different, etc.
+
+OPTIMIZE_OUTPUT_JAVA   = NO
+
+# If you use STL classes (i.e. std::string, std::vector, etc.) but do not want to 
+# include (a tag file for) the STL sources as input, then you should 
+# set this tag to YES in order to let doxygen match functions declarations and 
+# definitions whose arguments contain STL classes (e.g. func(std::string); v.s. 
+# func(std::string) {}). This also make the inheritance and collaboration 
+# diagrams that involve STL classes more complete and accurate.
+
+BUILTIN_STL_SUPPORT    = NO
+
+# If you use Microsoft's C++/CLI language, you should set this option to YES to
+# enable parsing support.
+
+CPP_CLI_SUPPORT        = NO
+
+# If member grouping is used in the documentation and the DISTRIBUTE_GROUP_DOC 
+# tag is set to YES, then doxygen will reuse the documentation of the first 
+# member in the group (if any) for the other members of the group. By default 
+# all members of a group must be documented explicitly.
+
+DISTRIBUTE_GROUP_DOC   = NO
+
+# Set the SUBGROUPING tag to YES (the default) to allow class member groups of 
+# the same type (for instance a group of public functions) to be put as a 
+# subgroup of that type (e.g. under the Public Functions section). Set it to 
+# NO to prevent subgrouping. Alternatively, this can be done per class using 
+# the \nosubgrouping command.
+
+SUBGROUPING            = YES
+
+#---------------------------------------------------------------------------
+# Build related configuration options
+#---------------------------------------------------------------------------
+
+# If the EXTRACT_ALL tag is set to YES doxygen will assume all entities in 
+# documentation are documented, even if no documentation was available. 
+# Private class members and static file members will be hidden unless 
+# the EXTRACT_PRIVATE and EXTRACT_STATIC tags are set to YES
+
+EXTRACT_ALL            = YES
+
+# If the EXTRACT_PRIVATE tag is set to YES all private members of a class 
+# will be included in the documentation.
+
+EXTRACT_PRIVATE        = YES
+
+# If the EXTRACT_STATIC tag is set to YES all static members of a file 
+# will be included in the documentation.
+
+EXTRACT_STATIC         = YES
+
+# If the EXTRACT_LOCAL_CLASSES tag is set to YES classes (and structs) 
+# defined locally in source files will be included in the documentation. 
+# If set to NO only classes defined in header files are included.
+
+EXTRACT_LOCAL_CLASSES  = YES
+
+# This flag is only useful for Objective-C code. When set to YES local 
+# methods, which are defined in the implementation section but not in 
+# the interface are included in the documentation. 
+# If set to NO (the default) only methods in the interface are included.
+
+EXTRACT_LOCAL_METHODS  = NO
+
+# If this flag is set to YES, the members of anonymous namespaces will be extracted 
+# and appear in the documentation as a namespace called 'anonymous_namespace{file}', 
+# where file will be replaced with the base name of the file that contains the anonymous 
+# namespace. By default anonymous namespace are hidden.
+
+EXTRACT_ANON_NSPACES   = NO
+
+# If the HIDE_UNDOC_MEMBERS tag is set to YES, Doxygen will hide all 
+# undocumented members of documented classes, files or namespaces. 
+# If set to NO (the default) these members will be included in the 
+# various overviews, but no documentation section is generated. 
+# This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_MEMBERS     = YES
+
+# If the HIDE_UNDOC_CLASSES tag is set to YES, Doxygen will hide all 
+# undocumented classes that are normally visible in the class hierarchy. 
+# If set to NO (the default) these classes will be included in the various 
+# overviews. This option has no effect if EXTRACT_ALL is enabled.
+
+HIDE_UNDOC_CLASSES     = YES
+
+# If the HIDE_FRIEND_COMPOUNDS tag is set to YES, Doxygen will hide all 
+# friend (class|struct|union) declarations. 
+# If set to NO (the default) these declarations will be included in the 
+# documentation.
+
+HIDE_FRIEND_COMPOUNDS  = NO
+
+# If the HIDE_IN_BODY_DOCS tag is set to YES, Doxygen will hide any 
+# documentation blocks found inside the body of a function. 
+# If set to NO (the default) these blocks will be appended to the 
+# function's detailed documentation block.
+
+HIDE_IN_BODY_DOCS      = NO
+
+# The INTERNAL_DOCS tag determines if documentation 
+# that is typed after a \internal command is included. If the tag is set 
+# to NO (the default) then the documentation will be excluded. 
+# Set it to YES to include the internal documentation.
+
+INTERNAL_DOCS          = NO
+
+# If the CASE_SENSE_NAMES tag is set to NO then Doxygen will only generate 
+# file names in lower-case letters. If set to YES upper-case letters are also 
+# allowed. This is useful if you have classes or files whose names only differ 
+# in case and if your file system supports case sensitive file names. Windows 
+# and Mac users are advised to set this option to NO.
+
+CASE_SENSE_NAMES       = YES
+
+# If the HIDE_SCOPE_NAMES tag is set to NO (the default) then Doxygen 
+# will show members with their full class and namespace scopes in the 
+# documentation. If set to YES the scope will be hidden.
+
+HIDE_SCOPE_NAMES       = NO
+
+# If the SHOW_INCLUDE_FILES tag is set to YES (the default) then Doxygen 
+# will put a list of the files that are included by a file in the documentation 
+# of that file.
+
+SHOW_INCLUDE_FILES     = YES
+
+# If the INLINE_INFO tag is set to YES (the default) then a tag [inline] 
+# is inserted in the documentation for inline members.
+
+INLINE_INFO            = YES
+
+# If the SORT_MEMBER_DOCS tag is set to YES (the default) then doxygen 
+# will sort the (detailed) documentation of file and class members 
+# alphabetically by member name. If set to NO the members will appear in 
+# declaration order.
+
+SORT_MEMBER_DOCS       = YES
+
+# If the SORT_BRIEF_DOCS tag is set to YES then doxygen will sort the 
+# brief documentation of file, namespace and class members alphabetically 
+# by member name. If set to NO (the default) the members will appear in 
+# declaration order.
+
+SORT_BRIEF_DOCS        = NO
+
+# If the SORT_BY_SCOPE_NAME tag is set to YES, the class list will be 
+# sorted by fully-qualified names, including namespaces. If set to 
+# NO (the default), the class list will be sorted only by class name, 
+# not including the namespace part. 
+# Note: This option is not very useful if HIDE_SCOPE_NAMES is set to YES.
+# Note: This option applies only to the class list, not to the 
+# alphabetical list.
+
+SORT_BY_SCOPE_NAME     = NO
+
+# The GENERATE_TODOLIST tag can be used to enable (YES) or 
+# disable (NO) the todo list. This list is created by putting \todo 
+# commands in the documentation.
+
+GENERATE_TODOLIST      = YES
+
+# The GENERATE_TESTLIST tag can be used to enable (YES) or 
+# disable (NO) the test list. This list is created by putting \test 
+# commands in the documentation.
+
+GENERATE_TESTLIST      = YES
+
+# The GENERATE_BUGLIST tag can be used to enable (YES) or 
+# disable (NO) the bug list. This list is created by putting \bug 
+# commands in the documentation.
+
+GENERATE_BUGLIST       = YES
+
+# The GENERATE_DEPRECATEDLIST tag can be used to enable (YES) or 
+# disable (NO) the deprecated list. This list is created by putting 
+# \deprecated commands in the documentation.
+
+GENERATE_DEPRECATEDLIST= YES
+
+# The ENABLED_SECTIONS tag can be used to enable conditional 
+# documentation sections, marked by \if sectionname ... \endif.
+
+ENABLED_SECTIONS       = 
+
+# The MAX_INITIALIZER_LINES tag determines the maximum number of lines 
+# the initial value of a variable or define consists of for it to appear in 
+# the documentation. If the initializer consists of more lines than specified 
+# here it will be hidden. Use a value of 0 to hide initializers completely. 
+# The appearance of the initializer of individual variables and defines in the 
+# documentation can be controlled using \showinitializer or \hideinitializer 
+# command in the documentation regardless of this setting.
+
+MAX_INITIALIZER_LINES  = 30
+
+# Set the SHOW_USED_FILES tag to NO to disable the list of files generated 
+# at the bottom of the documentation of classes and structs. If set to YES the 
+# list will mention the files that were used to generate the documentation.
+
+SHOW_USED_FILES        = YES
+
+# If the sources in your project are distributed over multiple directories 
+# then setting the SHOW_DIRECTORIES tag to YES will show the directory hierarchy 
+# in the documentation. The default is NO.
+
+SHOW_DIRECTORIES       = YES
+
+# The FILE_VERSION_FILTER tag can be used to specify a program or script that 
+# doxygen should invoke to get the current version for each file (typically from the 
+# version control system). Doxygen will invoke the program by executing (via 
+# popen()) the command <command> <input-file>, where <command> is the value of 
+# the FILE_VERSION_FILTER tag, and <input-file> is the name of an input file 
+# provided by doxygen. Whatever the program writes to standard output 
+# is used as the file version. See the manual for examples.
+
+FILE_VERSION_FILTER    = 
+
+#---------------------------------------------------------------------------
+# configuration options related to warning and progress messages
+#---------------------------------------------------------------------------
+
+# The QUIET tag can be used to turn on/off the messages that are generated 
+# by doxygen. Possible values are YES and NO. If left blank NO is used.
+
+QUIET                  = NO
+
+# The WARNINGS tag can be used to turn on/off the warning messages that are 
+# generated by doxygen. Possible values are YES and NO. If left blank 
+# NO is used.
+
+WARNINGS               = YES
+
+# If WARN_IF_UNDOCUMENTED is set to YES, then doxygen will generate warnings 
+# for undocumented members. If EXTRACT_ALL is set to YES then this flag will 
+# automatically be disabled.
+
+WARN_IF_UNDOCUMENTED   = YES
+
+# If WARN_IF_DOC_ERROR is set to YES, doxygen will generate warnings for 
+# potential errors in the documentation, such as not documenting some 
+# parameters in a documented function, or documenting parameters that 
+# don't exist or using markup commands wrongly.
+
+WARN_IF_DOC_ERROR      = YES
+
+# This WARN_NO_PARAMDOC option can be abled to get warnings for 
+# functions that are documented, but have no documentation for their parameters 
+# or return value. If set to NO (the default) doxygen will only warn about 
+# wrong or incomplete parameter documentation, but not about the absence of 
+# documentation.
+
+WARN_NO_PARAMDOC       = NO
+
+# The WARN_FORMAT tag determines the format of the warning messages that 
+# doxygen can produce. The string should contain the $file, $line, and $text 
+# tags, which will be replaced by the file and line number from which the 
+# warning originated and the warning text. Optionally the format may contain 
+# $version, which will be replaced by the version of the file (if it could 
+# be obtained via FILE_VERSION_FILTER)
+
+WARN_FORMAT            = "$file:$line: $text"
+
+# The WARN_LOGFILE tag can be used to specify a file to which warning 
+# and error messages should be written. If left blank the output is written 
+# to stderr.
+
+WARN_LOGFILE           = 
+
+#---------------------------------------------------------------------------
+# configuration options related to the input files
+#---------------------------------------------------------------------------
+
+# The INPUT tag can be used to specify the files and/or directories that contain 
+# documented source files. You may enter file names like "myfile.cpp" or 
+# directories like "/usr/src/myproject". Separate the files or directories 
+# with spaces.
+
+INPUT                  = tools/ \
+                         lib/ \
+                         include/klee \
+                         docs/intro \
+                         docs/overview
+
+# This tag can be used to specify the character encoding of the source files that 
+# doxygen parses. Internally doxygen uses the UTF-8 encoding, which is also the default 
+# input encoding. Doxygen uses libiconv (or the iconv built into libc) for the transcoding. 
+# See http://www.gnu.org/software/libiconv for the list of possible encodings.
+
+INPUT_ENCODING         = UTF-8
+
+# If the value of the INPUT tag contains directories, you can use the 
+# FILE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp 
+# and *.h) to filter out the source-files in the directories. If left 
+# blank the following patterns are tested: 
+# *.c *.cc *.cxx *.cpp *.c++ *.java *.ii *.ixx *.ipp *.i++ *.inl *.h *.hh *.hxx 
+# *.hpp *.h++ *.idl *.odl *.cs *.php *.php3 *.inc *.m *.mm *.py
+
+FILE_PATTERNS          = 
+
+# The RECURSIVE tag can be used to turn specify whether or not subdirectories 
+# should be searched for input files as well. Possible values are YES and NO. 
+# If left blank NO is used.
+
+RECURSIVE              = YES
+
+# The EXCLUDE tag can be used to specify files and/or directories that should 
+# excluded from the INPUT source files. This way you can easily exclude a 
+# subdirectory from a directory tree whose root is specified with the INPUT tag.
+
+EXCLUDE                = 
+
+# The EXCLUDE_SYMLINKS tag can be used select whether or not files or 
+# directories that are symbolic links (a Unix filesystem feature) are excluded 
+# from the input.
+
+EXCLUDE_SYMLINKS       = NO
+
+# If the value of the INPUT tag contains directories, you can use the 
+# EXCLUDE_PATTERNS tag to specify one or more wildcard patterns to exclude 
+# certain files from those directories. Note that the wildcards are matched 
+# against the file with absolute path, so to exclude all test directories 
+# for example use the pattern */test/*
+
+EXCLUDE_PATTERNS       =  */.svn* */Debug*
+
+# The EXCLUDE_SYMBOLS tag can be used to specify one or more symbol names 
+# (namespaces, classes, functions, etc.) that should be excluded from the output. 
+# The symbol name can be a fully qualified name, a word, or if the wildcard * is used, 
+# a substring. Examples: ANamespace, AClass, AClass::ANamespace, ANamespace::*Test
+
+EXCLUDE_SYMBOLS        = 
+
+# The EXAMPLE_PATH tag can be used to specify one or more files or 
+# directories that contain example code fragments that are included (see 
+# the \include command).
+
+EXAMPLE_PATH           = 
+
+# If the value of the EXAMPLE_PATH tag contains directories, you can use the 
+# EXAMPLE_PATTERNS tag to specify one or more wildcard pattern (like *.cpp 
+# and *.h) to filter out the source-files in the directories. If left 
+# blank all files are included.
+
+EXAMPLE_PATTERNS       = 
+
+# If the EXAMPLE_RECURSIVE tag is set to YES then subdirectories will be 
+# searched for input files to be used with the \include or \dontinclude 
+# commands irrespective of the value of the RECURSIVE tag. 
+# Possible values are YES and NO. If left blank NO is used.
+
+EXAMPLE_RECURSIVE      = NO
+
+# The IMAGE_PATH tag can be used to specify one or more files or 
+# directories that contain image that are included in the documentation (see 
+# the \image command).
+
+IMAGE_PATH             = 
+
+# The INPUT_FILTER tag can be used to specify a program that doxygen should 
+# invoke to filter for each input file. Doxygen will invoke the filter program 
+# by executing (via popen()) the command <filter> <input-file>, where <filter> 
+# is the value of the INPUT_FILTER tag, and <input-file> is the name of an 
+# input file. Doxygen will then use the output that the filter program writes 
+# to standard output.  If FILTER_PATTERNS is specified, this tag will be 
+# ignored.
+
+INPUT_FILTER           = 
+
+# The FILTER_PATTERNS tag can be used to specify filters on a per file pattern 
+# basis.  Doxygen will compare the file name with each pattern and apply the 
+# filter if there is a match.  The filters are a list of the form: 
+# pattern=filter (like *.cpp=my_cpp_filter). See INPUT_FILTER for further 
+# info on how filters are used. If FILTER_PATTERNS is empty, INPUT_FILTER 
+# is applied to all files.
+
+FILTER_PATTERNS        = 
+
+# If the FILTER_SOURCE_FILES tag is set to YES, the input filter (if set using 
+# INPUT_FILTER) will be used to filter the input files when producing source 
+# files to browse (i.e. when SOURCE_BROWSER is set to YES).
+
+FILTER_SOURCE_FILES    = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to source browsing
+#---------------------------------------------------------------------------
+
+# If the SOURCE_BROWSER tag is set to YES then a list of source files will 
+# be generated. Documented entities will be cross-referenced with these sources. 
+# Note: To get rid of all source code in the generated output, make sure also 
+# VERBATIM_HEADERS is set to NO. If you have enabled CALL_GRAPH or CALLER_GRAPH 
+# then you must also enable this option. If you don't then doxygen will produce 
+# a warning and turn it on anyway
+
+SOURCE_BROWSER         = YES
+
+# Setting the INLINE_SOURCES tag to YES will include the body 
+# of functions and classes directly in the documentation.
+
+INLINE_SOURCES         = NO
+
+# Setting the STRIP_CODE_COMMENTS tag to YES (the default) will instruct 
+# doxygen to hide any special comment blocks from generated source code 
+# fragments. Normal C and C++ comments will always remain visible.
+
+STRIP_CODE_COMMENTS    = YES
+
+# If the REFERENCED_BY_RELATION tag is set to YES (the default) 
+# then for each documented function all documented 
+# functions referencing it will be listed.
+
+REFERENCED_BY_RELATION = YES
+
+# If the REFERENCES_RELATION tag is set to YES (the default) 
+# then for each documented function all documented entities 
+# called/used by that function will be listed.
+
+REFERENCES_RELATION    = YES
+
+# If the REFERENCES_LINK_SOURCE tag is set to YES (the default)
+# and SOURCE_BROWSER tag is set to YES, then the hyperlinks from
+# functions in REFERENCES_RELATION and REFERENCED_BY_RELATION lists will
+# link to the source code.  Otherwise they will link to the documentstion.
+
+REFERENCES_LINK_SOURCE = YES
+
+# If the USE_HTAGS tag is set to YES then the references to source code 
+# will point to the HTML generated by the htags(1) tool instead of doxygen 
+# built-in source browser. The htags tool is part of GNU's global source 
+# tagging system (see http://www.gnu.org/software/global/global.html). You 
+# will need version 4.8.6 or higher.
+
+USE_HTAGS              = NO
+
+# If the VERBATIM_HEADERS tag is set to YES (the default) then Doxygen 
+# will generate a verbatim copy of the header file for each class for 
+# which an include is specified. Set to NO to disable this.
+
+VERBATIM_HEADERS       = YES
+
+#---------------------------------------------------------------------------
+# configuration options related to the alphabetical class index
+#---------------------------------------------------------------------------
+
+# If the ALPHABETICAL_INDEX tag is set to YES, an alphabetical index 
+# of all compounds will be generated. Enable this if the project 
+# contains a lot of classes, structs, unions or interfaces.
+
+ALPHABETICAL_INDEX     = YES
+
+# If the alphabetical index is enabled (see ALPHABETICAL_INDEX) then 
+# the COLS_IN_ALPHA_INDEX tag can be used to specify the number of columns 
+# in which this list will be split (can be a number in the range [1..20])
+
+COLS_IN_ALPHA_INDEX    = 5
+
+# In case all classes in a project start with a common prefix, all 
+# classes will be put under the same header in the alphabetical index. 
+# The IGNORE_PREFIX tag can be used to specify one or more prefixes that 
+# should be ignored while generating the index headers.
+
+IGNORE_PREFIX          = 
+
+#---------------------------------------------------------------------------
+# configuration options related to the HTML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_HTML tag is set to YES (the default) Doxygen will 
+# generate HTML output.
+
+GENERATE_HTML          = YES
+
+# The HTML_OUTPUT tag is used to specify where the HTML docs will be put. 
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be 
+# put in front of it. If left blank `html' will be used as the default path.
+
+HTML_OUTPUT            = html
+
+# The HTML_FILE_EXTENSION tag can be used to specify the file extension for 
+# each generated HTML page (for example: .htm,.php,.asp). If it is left blank 
+# doxygen will generate files with .html extension.
+
+HTML_FILE_EXTENSION    = .html
+
+# The HTML_HEADER tag can be used to specify a personal HTML header for 
+# each generated HTML page. If it is left blank doxygen will generate a 
+# standard header.
+
+HTML_HEADER            = 
+
+# The HTML_FOOTER tag can be used to specify a personal HTML footer for 
+# each generated HTML page. If it is left blank doxygen will generate a 
+# standard footer.
+
+HTML_FOOTER            = 
+
+# The HTML_STYLESHEET tag can be used to specify a user-defined cascading 
+# style sheet that is used by each HTML page. It can be used to 
+# fine-tune the look of the HTML output. If the tag is left blank doxygen 
+# will generate a default style sheet. Note that doxygen will try to copy 
+# the style sheet file to the HTML output directory, so don't put your own 
+# stylesheet in the HTML output directory as well, or it will be erased!
+
+HTML_STYLESHEET        = 
+
+# If the HTML_ALIGN_MEMBERS tag is set to YES, the members of classes, 
+# files or namespaces will be aligned in HTML using tables. If set to 
+# NO a bullet list will be used.
+
+HTML_ALIGN_MEMBERS     = YES
+
+# If the GENERATE_HTMLHELP tag is set to YES, additional index files 
+# will be generated that can be used as input for tools like the 
+# Microsoft HTML help workshop to generate a compressed HTML help file (.chm) 
+# of the generated HTML documentation.
+
+GENERATE_HTMLHELP      = NO
+
+# If the HTML_DYNAMIC_SECTIONS tag is set to YES then the generated HTML 
+# documentation will contain sections that can be hidden and shown after the 
+# page has loaded. For this to work a browser that supports 
+# JavaScript and DHTML is required (for instance Mozilla 1.0+, Firefox 
+# Netscape 6.0+, Internet explorer 5.0+, Konqueror, or Safari).
+
+HTML_DYNAMIC_SECTIONS  = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the CHM_FILE tag can 
+# be used to specify the file name of the resulting .chm file. You 
+# can add a path in front of the file if the result should not be 
+# written to the html output directory.
+
+CHM_FILE               = 
+
+# If the GENERATE_HTMLHELP tag is set to YES, the HHC_LOCATION tag can 
+# be used to specify the location (absolute path including file name) of 
+# the HTML help compiler (hhc.exe). If non-empty doxygen will try to run 
+# the HTML help compiler on the generated index.hhp.
+
+HHC_LOCATION           = 
+
+# If the GENERATE_HTMLHELP tag is set to YES, the GENERATE_CHI flag 
+# controls if a separate .chi index file is generated (YES) or that 
+# it should be included in the master .chm file (NO).
+
+GENERATE_CHI           = NO
+
+# If the GENERATE_HTMLHELP tag is set to YES, the BINARY_TOC flag 
+# controls whether a binary table of contents is generated (YES) or a 
+# normal table of contents (NO) in the .chm file.
+
+BINARY_TOC             = NO
+
+# The TOC_EXPAND flag can be set to YES to add extra items for group members 
+# to the contents of the HTML help documentation and to the tree view.
+
+TOC_EXPAND             = NO
+
+# The DISABLE_INDEX tag can be used to turn on/off the condensed index at 
+# top of each HTML page. The value NO (the default) enables the index and 
+# the value YES disables it.
+
+DISABLE_INDEX          = NO
+
+# This tag can be used to set the number of enum values (range [1..20]) 
+# that doxygen will group on one line in the generated HTML documentation.
+
+ENUM_VALUES_PER_LINE   = 4
+
+# If the GENERATE_TREEVIEW tag is set to YES, a side panel will be
+# generated containing a tree-like index structure (just like the one that 
+# is generated for HTML Help). For this to work a browser that supports 
+# JavaScript, DHTML, CSS and frames is required (for instance Mozilla 1.0+, 
+# Netscape 6.0+, Internet explorer 5.0+, or Konqueror). Windows users are 
+# probably better off using the HTML help feature.
+
+GENERATE_TREEVIEW      = YES
+
+# If the treeview is enabled (see GENERATE_TREEVIEW) then this tag can be 
+# used to set the initial width (in pixels) of the frame in which the tree 
+# is shown.
+
+TREEVIEW_WIDTH         = 250
+
+#---------------------------------------------------------------------------
+# configuration options related to the LaTeX output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_LATEX tag is set to YES (the default) Doxygen will 
+# generate Latex output.
+
+GENERATE_LATEX         = NO
+
+# The LATEX_OUTPUT tag is used to specify where the LaTeX docs will be put. 
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be 
+# put in front of it. If left blank `latex' will be used as the default path.
+
+LATEX_OUTPUT           = latex
+
+# The LATEX_CMD_NAME tag can be used to specify the LaTeX command name to be 
+# invoked. If left blank `latex' will be used as the default command name.
+
+LATEX_CMD_NAME         = latex
+
+# The MAKEINDEX_CMD_NAME tag can be used to specify the command name to 
+# generate index for LaTeX. If left blank `makeindex' will be used as the 
+# default command name.
+
+MAKEINDEX_CMD_NAME     = makeindex
+
+# If the COMPACT_LATEX tag is set to YES Doxygen generates more compact 
+# LaTeX documents. This may be useful for small projects and may help to 
+# save some trees in general.
+
+COMPACT_LATEX          = NO
+
+# The PAPER_TYPE tag can be used to set the paper type that is used 
+# by the printer. Possible values are: a4, a4wide, letter, legal and 
+# executive. If left blank a4wide will be used.
+
+PAPER_TYPE             = a4wide
+
+# The EXTRA_PACKAGES tag can be to specify one or more names of LaTeX 
+# packages that should be included in the LaTeX output.
+
+EXTRA_PACKAGES         = 
+
+# The LATEX_HEADER tag can be used to specify a personal LaTeX header for 
+# the generated latex document. The header should contain everything until 
+# the first chapter. If it is left blank doxygen will generate a 
+# standard header. Notice: only use this tag if you know what you are doing!
+
+LATEX_HEADER           = 
+
+# If the PDF_HYPERLINKS tag is set to YES, the LaTeX that is generated 
+# is prepared for conversion to pdf (using ps2pdf). The pdf file will 
+# contain links (just like the HTML output) instead of page references 
+# This makes the output suitable for online browsing using a pdf viewer.
+
+PDF_HYPERLINKS         = NO
+
+# If the USE_PDFLATEX tag is set to YES, pdflatex will be used instead of 
+# plain latex in the generated Makefile. Set this option to YES to get a 
+# higher quality PDF documentation.
+
+USE_PDFLATEX           = NO
+
+# If the LATEX_BATCHMODE tag is set to YES, doxygen will add the \\batchmode. 
+# command to the generated LaTeX files. This will instruct LaTeX to keep 
+# running if errors occur, instead of asking the user for help. 
+# This option is also used when generating formulas in HTML.
+
+LATEX_BATCHMODE        = NO
+
+# If LATEX_HIDE_INDICES is set to YES then doxygen will not 
+# include the index chapters (such as File Index, Compound Index, etc.) 
+# in the output.
+
+LATEX_HIDE_INDICES     = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the RTF output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_RTF tag is set to YES Doxygen will generate RTF output 
+# The RTF output is optimized for Word 97 and may not look very pretty with 
+# other RTF readers or editors.
+
+GENERATE_RTF           = NO
+
+# The RTF_OUTPUT tag is used to specify where the RTF docs will be put. 
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be 
+# put in front of it. If left blank `rtf' will be used as the default path.
+
+RTF_OUTPUT             = rtf
+
+# If the COMPACT_RTF tag is set to YES Doxygen generates more compact 
+# RTF documents. This may be useful for small projects and may help to 
+# save some trees in general.
+
+COMPACT_RTF            = NO
+
+# If the RTF_HYPERLINKS tag is set to YES, the RTF that is generated 
+# will contain hyperlink fields. The RTF file will 
+# contain links (just like the HTML output) instead of page references. 
+# This makes the output suitable for online browsing using WORD or other 
+# programs which support those fields. 
+# Note: wordpad (write) and others do not support links.
+
+RTF_HYPERLINKS         = NO
+
+# Load stylesheet definitions from file. Syntax is similar to doxygen's 
+# config file, i.e. a series of assignments. You only have to provide 
+# replacements, missing definitions are set to their default value.
+
+RTF_STYLESHEET_FILE    = 
+
+# Set optional variables used in the generation of an rtf document. 
+# Syntax is similar to doxygen's config file.
+
+RTF_EXTENSIONS_FILE    = 
+
+#---------------------------------------------------------------------------
+# configuration options related to the man page output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_MAN tag is set to YES (the default) Doxygen will 
+# generate man pages
+
+GENERATE_MAN           = NO
+
+# The MAN_OUTPUT tag is used to specify where the man pages will be put. 
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be 
+# put in front of it. If left blank `man' will be used as the default path.
+
+MAN_OUTPUT             = man
+
+# The MAN_EXTENSION tag determines the extension that is added to 
+# the generated man pages (default is the subroutine's section .3)
+
+MAN_EXTENSION          = .3
+
+# If the MAN_LINKS tag is set to YES and Doxygen generates man output, 
+# then it will generate one additional man file for each entity 
+# documented in the real man page(s). These additional files 
+# only source the real man page, but without them the man command 
+# would be unable to find the correct page. The default is NO.
+
+MAN_LINKS              = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the XML output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_XML tag is set to YES Doxygen will 
+# generate an XML file that captures the structure of 
+# the code including all documentation.
+
+GENERATE_XML           = NO
+
+# The XML_OUTPUT tag is used to specify where the XML pages will be put. 
+# If a relative path is entered the value of OUTPUT_DIRECTORY will be 
+# put in front of it. If left blank `xml' will be used as the default path.
+
+XML_OUTPUT             = xml
+
+# The XML_SCHEMA tag can be used to specify an XML schema, 
+# which can be used by a validating XML parser to check the 
+# syntax of the XML files.
+
+XML_SCHEMA             = 
+
+# The XML_DTD tag can be used to specify an XML DTD, 
+# which can be used by a validating XML parser to check the 
+# syntax of the XML files.
+
+XML_DTD                = 
+
+# If the XML_PROGRAMLISTING tag is set to YES Doxygen will 
+# dump the program listings (including syntax highlighting 
+# and cross-referencing information) to the XML output. Note that 
+# enabling this will significantly increase the size of the XML output.
+
+XML_PROGRAMLISTING     = YES
+
+#---------------------------------------------------------------------------
+# configuration options for the AutoGen Definitions output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_AUTOGEN_DEF tag is set to YES Doxygen will 
+# generate an AutoGen Definitions (see autogen.sf.net) file 
+# that captures the structure of the code including all 
+# documentation. Note that this feature is still experimental 
+# and incomplete at the moment.
+
+GENERATE_AUTOGEN_DEF   = NO
+
+#---------------------------------------------------------------------------
+# configuration options related to the Perl module output
+#---------------------------------------------------------------------------
+
+# If the GENERATE_PERLMOD tag is set to YES Doxygen will 
+# generate a Perl module file that captures the structure of 
+# the code including all documentation. Note that this 
+# feature is still experimental and incomplete at the 
+# moment.
+
+GENERATE_PERLMOD       = NO
+
+# If the PERLMOD_LATEX tag is set to YES Doxygen will generate 
+# the necessary Makefile rules, Perl scripts and LaTeX code to be able 
+# to generate PDF and DVI output from the Perl module output.
+
+PERLMOD_LATEX          = NO
+
+# If the PERLMOD_PRETTY tag is set to YES the Perl module output will be 
+# nicely formatted so it can be parsed by a human reader.  This is useful 
+# if you want to understand what is going on.  On the other hand, if this 
+# tag is set to NO the size of the Perl module output will be much smaller 
+# and Perl will parse it just the same.
+
+PERLMOD_PRETTY         = YES
+
+# The names of the make variables in the generated doxyrules.make file 
+# are prefixed with the string contained in PERLMOD_MAKEVAR_PREFIX. 
+# This is useful so different doxyrules.make files included by the same 
+# Makefile don't overwrite each other's variables.
+
+PERLMOD_MAKEVAR_PREFIX = 
+
+#---------------------------------------------------------------------------
+# Configuration options related to the preprocessor   
+#---------------------------------------------------------------------------
+
+# If the ENABLE_PREPROCESSING tag is set to YES (the default) Doxygen will 
+# evaluate all C-preprocessor directives found in the sources and include 
+# files.
+
+ENABLE_PREPROCESSING   = YES
+
+# If the MACRO_EXPANSION tag is set to YES Doxygen will expand all macro 
+# names in the source code. If set to NO (the default) only conditional 
+# compilation will be performed. Macro expansion can be done in a controlled 
+# way by setting EXPAND_ONLY_PREDEF to YES.
+
+MACRO_EXPANSION        = NO
+
+# If the EXPAND_ONLY_PREDEF and MACRO_EXPANSION tags are both set to YES 
+# then the macro expansion is limited to the macros specified with the 
+# PREDEFINED and EXPAND_AS_DEFINED tags.
+
+EXPAND_ONLY_PREDEF     = NO
+
+# If the SEARCH_INCLUDES tag is set to YES (the default) the includes files 
+# in the INCLUDE_PATH (see below) will be search if a #include is found.
+
+SEARCH_INCLUDES        = YES
+
+# The INCLUDE_PATH tag can be used to specify one or more directories that 
+# contain include files that are not input files but should be processed by 
+# the preprocessor.
+
+INCLUDE_PATH           =
+
+# You can use the INCLUDE_FILE_PATTERNS tag to specify one or more wildcard 
+# patterns (like *.h and *.hpp) to filter out the header-files in the 
+# directories. If left blank, the patterns specified with FILE_PATTERNS will 
+# be used.
+
+INCLUDE_FILE_PATTERNS  = 
+
+# The PREDEFINED tag can be used to specify one or more macro names that 
+# are defined before the preprocessor is started (similar to the -D option of 
+# gcc). The argument of the tag is a list of macros of the form: name 
+# or name=definition (no spaces). If the definition and the = are 
+# omitted =1 is assumed. To prevent a macro definition from being 
+# undefined via #undef or recursively expanded use the := operator 
+# instead of the = operator.
+
+PREDEFINED             = 
+
+# If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then 
+# this tag can be used to specify a list of macro names that should be expanded. 
+# The macro definition that is found in the sources will be used. 
+# Use the PREDEFINED tag if you want to use a different macro definition.
+
+EXPAND_AS_DEFINED      = 
+
+# If the SKIP_FUNCTION_MACROS tag is set to YES (the default) then 
+# doxygen's preprocessor will remove all function-like macros that are alone 
+# on a line, have an all uppercase name, and do not end with a semicolon. Such 
+# function macros are typically used for boiler-plate code, and will confuse 
+# the parser if not removed.
+
+SKIP_FUNCTION_MACROS   = YES
+
+#---------------------------------------------------------------------------
+# Configuration::additions related to external references   
+#---------------------------------------------------------------------------
+
+# The TAGFILES option can be used to specify one or more tagfiles. 
+# Optionally an initial location of the external documentation 
+# can be added for each tagfile. The format of a tag file without 
+# this location is as follows: 
+#   TAGFILES = file1 file2 ... 
+# Adding location for the tag files is done as follows: 
+#   TAGFILES = file1=loc1 "file2 = loc2" ... 
+# where "loc1" and "loc2" can be relative or absolute paths or 
+# URLs. If a location is present for each tag, the installdox tool 
+# does not have to be run to correct the links.
+# Note that each tag file must have a unique name
+# (where the name does NOT include the path)
+# If a tag file is not located in the directory in which doxygen 
+# is run, you must also specify the path to the tagfile here.
+
+TAGFILES               = 
+
+# When a file name is specified after GENERATE_TAGFILE, doxygen will create 
+# a tag file that is based on the input files it reads.
+
+GENERATE_TAGFILE       = 
+
+# If the ALLEXTERNALS tag is set to YES all external classes will be listed 
+# in the class index. If set to NO only the inherited external classes 
+# will be listed.
+
+ALLEXTERNALS           = NO
+
+# If the EXTERNAL_GROUPS tag is set to YES all external groups will be listed 
+# in the modules index. If set to NO, only the current project's groups will 
+# be listed.
+
+EXTERNAL_GROUPS        = YES
+
+# The PERL_PATH should be the absolute path and name of the perl script 
+# interpreter (i.e. the result of `which perl').
+
+PERL_PATH              = /usr/bin/perl
+
+#---------------------------------------------------------------------------
+# Configuration options related to the dot tool   
+#---------------------------------------------------------------------------
+
+# If the CLASS_DIAGRAMS tag is set to YES (the default) Doxygen will 
+# generate a inheritance diagram (in HTML, RTF and LaTeX) for classes with base 
+# or super classes. Setting the tag to NO turns the diagrams off. Note that 
+# this option is superseded by the HAVE_DOT option below. This is only a 
+# fallback. It is recommended to install and use dot, since it yields more 
+# powerful graphs.
+
+CLASS_DIAGRAMS         = YES
+
+# You can define message sequence charts within doxygen comments using the \msc 
+# command. Doxygen will then run the mscgen tool (see http://www.mcternan.me.uk/mscgen/) to 
+# produce the chart and insert it in the documentation. The MSCGEN_PATH tag allows you to 
+# specify the directory where the mscgen tool resides. If left empty the tool is assumed to 
+# be found in the default search path.
+
+MSCGEN_PATH            = 
+
+# If set to YES, the inheritance and collaboration graphs will hide 
+# inheritance and usage relations if the target is undocumented 
+# or is not a class.
+
+HIDE_UNDOC_RELATIONS   = YES
+
+# If you set the HAVE_DOT tag to YES then doxygen will assume the dot tool is 
+# available from the path. This tool is part of Graphviz, a graph visualization 
+# toolkit from AT&T and Lucent Bell Labs. The other options in this section 
+# have no effect if this option is set to NO (the default)
+
+HAVE_DOT               = YES
+
+# If the CLASS_GRAPH and HAVE_DOT tags are set to YES then doxygen 
+# will generate a graph for each documented class showing the direct and 
+# indirect inheritance relations. Setting this tag to YES will force the 
+# the CLASS_DIAGRAMS tag to NO.
+
+CLASS_GRAPH            = YES
+
+# If the COLLABORATION_GRAPH and HAVE_DOT tags are set to YES then doxygen 
+# will generate a graph for each documented class showing the direct and 
+# indirect implementation dependencies (inheritance, containment, and 
+# class references variables) of the class with other documented classes.
+
+COLLABORATION_GRAPH    = YES
+
+# If the GROUP_GRAPHS and HAVE_DOT tags are set to YES then doxygen 
+# will generate a graph for groups, showing the direct groups dependencies
+
+GROUP_GRAPHS           = YES
+
+# If the UML_LOOK tag is set to YES doxygen will generate inheritance and 
+# collaboration diagrams in a style similar to the OMG's Unified Modeling 
+# Language.
+
+UML_LOOK               = NO
+
+# If set to YES, the inheritance and collaboration graphs will show the 
+# relations between templates and their instances.
+
+TEMPLATE_RELATIONS     = YES
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDE_GRAPH, and HAVE_DOT 
+# tags are set to YES then doxygen will generate a graph for each documented 
+# file showing the direct and indirect include dependencies of the file with 
+# other documented files.
+
+INCLUDE_GRAPH          = YES
+
+# If the ENABLE_PREPROCESSING, SEARCH_INCLUDES, INCLUDED_BY_GRAPH, and 
+# HAVE_DOT tags are set to YES then doxygen will generate a graph for each 
+# documented header file showing the documented files that directly or 
+# indirectly include this file.
+
+INCLUDED_BY_GRAPH      = YES
+
+# If the CALL_GRAPH, SOURCE_BROWSER and HAVE_DOT tags are set to YES then doxygen will 
+# generate a call dependency graph for every global function or class method. 
+# Note that enabling this option will significantly increase the time of a run. 
+# So in most cases it will be better to enable call graphs for selected 
+# functions only using the \callgraph command.
+
+CALL_GRAPH             = YES
+
+# If the CALLER_GRAPH, SOURCE_BROWSER and HAVE_DOT tags are set to YES then doxygen will 
+# generate a caller dependency graph for every global function or class method. 
+# Note that enabling this option will significantly increase the time of a run. 
+# So in most cases it will be better to enable caller graphs for selected 
+# functions only using the \callergraph command.
+
+CALLER_GRAPH           = YES
+
+# If the GRAPHICAL_HIERARCHY and HAVE_DOT tags are set to YES then doxygen 
+# will graphical hierarchy of all classes instead of a textual one.
+
+GRAPHICAL_HIERARCHY    = YES
+
+# If the DIRECTORY_GRAPH, SHOW_DIRECTORIES and HAVE_DOT tags are set to YES 
+# then doxygen will show the dependencies a directory has on other directories 
+# in a graphical way. The dependency relations are determined by the #include
+# relations between the files in the directories.
+
+DIRECTORY_GRAPH        = YES
+
+# The DOT_IMAGE_FORMAT tag can be used to set the image format of the images 
+# generated by dot. Possible values are png, jpg, or gif
+# If left blank png will be used.
+
+DOT_IMAGE_FORMAT       = png
+
+# The tag DOT_PATH can be used to specify the path where the dot tool can be 
+# found. If left blank, it is assumed the dot tool can be found in the path.
+
+DOT_PATH               = 
+
+# The DOTFILE_DIRS tag can be used to specify one or more directories that 
+# contain dot files that are included in the documentation (see the 
+# \dotfile command).
+
+DOTFILE_DIRS           = 
+
+# The MAX_DOT_GRAPH_MAX_NODES tag can be used to set the maximum number of 
+# nodes that will be shown in the graph. If the number of nodes in a graph 
+# becomes larger than this value, doxygen will truncate the graph, which is 
+# visualized by representing a node as a red box. Note that doxygen if the number 
+# of direct children of the root node in a graph is already larger than 
+# MAX_DOT_GRAPH_NOTES then the graph will not be shown at all. Also note 
+# that the size of a graph can be further restricted by MAX_DOT_GRAPH_DEPTH.
+
+DOT_GRAPH_MAX_NODES    = 50
+
+# The MAX_DOT_GRAPH_DEPTH tag can be used to set the maximum depth of the 
+# graphs generated by dot. A depth value of 3 means that only nodes reachable 
+# from the root by following a path via at most 3 edges will be shown. Nodes 
+# that lay further from the root node will be omitted. Note that setting this 
+# option to 1 or 2 may greatly reduce the computation time needed for large 
+# code bases. Also note that the size of a graph can be further restricted by 
+# DOT_GRAPH_MAX_NODES. Using a depth of 0 means no depth restriction.
+
+MAX_DOT_GRAPH_DEPTH    = 0
+
+# Set the DOT_TRANSPARENT tag to YES to generate images with a transparent 
+# background. This is disabled by default, which results in a white background. 
+# Warning: Depending on the platform used, enabling this option may lead to 
+# badly anti-aliased labels on the edges of a graph (i.e. they become hard to 
+# read).
+
+DOT_TRANSPARENT        = NO
+
+# Set the DOT_MULTI_TARGETS tag to YES allow dot to generate multiple output 
+# files in one run (i.e. multiple -o and -T options on the command line). This 
+# makes dot run faster, but since only newer versions of dot (>1.8.10) 
+# support this, this feature is disabled by default.
+
+DOT_MULTI_TARGETS      = NO
+
+# If the GENERATE_LEGEND tag is set to YES (the default) Doxygen will 
+# generate a legend page explaining the meaning of the various boxes and 
+# arrows in the dot generated graphs.
+
+GENERATE_LEGEND        = YES
+
+# If the DOT_CLEANUP tag is set to YES (the default) Doxygen will 
+# remove the intermediate dot files that are used to generate 
+# the various graphs.
+
+DOT_CLEANUP            = YES
+
+#---------------------------------------------------------------------------
+# Configuration::additions related to the search engine   
+#---------------------------------------------------------------------------
+
+# The SEARCHENGINE tag specifies whether or not a search engine should be 
+# used. If set to NO the values of all tags below this one will be ignored.
+
+SEARCHENGINE           = YES

Added: klee/trunk/docs/intro
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/intro?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/intro (added)
+++ klee/trunk/docs/intro Wed May 20 23:36:41 2009
@@ -0,0 +1,10 @@
+/// @mainpage KLEE
+///
+/// @section main_intro Introduction
+/// Welcome to KLEE. KLEE is a symbolic execution engine that works on LLVM
+/// bitcode.
+/// 
+/// @section Documentation
+/// The documentation of KLEE is composed of the Doxygen documentation
+/// of the code as well as the following documents:
+/// - @subpage overview

Added: klee/trunk/docs/overview
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/docs/overview?rev=72205&view=auto

==============================================================================
--- klee/trunk/docs/overview (added)
+++ klee/trunk/docs/overview Wed May 20 23:36:41 2009
@@ -0,0 +1,104 @@
+/// @page overview High level overview of KLEE.
+/// This document contains a high level overview of the inner workings of KLEE.
+///
+/// KLEE implements symbolic execution by interpreting LLVM bitcode. Symbolic
+/// memory is defined by inserting special calls to KLEE (namely
+/// klee_make_symbolic)
+/// During execution, KLEE tracks all uses of symbolic memory. Constraints
+/// on symbolic memory usage are collected. Memory
+/// that is defined using previously declared symbolic memory become
+/// symbolic as well.
+/// Whenever a branch refering to symbolic memory is encountered, KLEE forks
+/// the entire states and explores each side of the branch for which a possible
+/// solution to the symbolic constraints can be found.
+/// KLEE makes queries to STP to solve symbolic constraints.
+///
+/// The rest of this document describes some of the important components of KLEE
+/// 
+/// @section executor Interpreter
+/// klee::Interpreter is the main abstract class defining the interface of the
+/// bitcode interpreter. klee::Executor is the main concrete instance of this
+/// class.
+/// Application states (i.e. memory, registers and PC) are stored in instances of
+/// class klee::ExecutionState. There is one such instance for each path beeing
+/// executed (except when some states are merged together).
+/// On a branch, if condition is symbolic, klee::Executor::fork returns a
+/// klee::ExecutionState::StatePair which is a pair of ExecutionState to be
+/// executed.
+/// 
+/// @section memory Memory model
+/// MemoryObject's represent allocation sites in the program (calls to malloc, stack
+/// objects, global variables)
+/// and, at least conceptually, can be thought of as the unique name for the object
+/// allocated at that site.
+/// ObjectState's are used to store the actual contents of a MemoryObject in a
+/// particular ExecutionState (but
+/// can be shared). I need better names for these two things. 
+/// 
+/// Each ExecutionState stores a mapping of MemoryObjects -> ObjectState using the
+/// AddressSpace data
+/// structure (implemented as an immutable tree so that copying is cheap and the
+/// shared structure is exploited).
+/// Each AddressSpace may "own" some subset of the ObjectStates in the mapping. When
+/// an AddressSpace
+/// is duplicated it loses ownership of the ObjectState in the map. Any subsequent
+/// write to an ObjectState will
+/// create a copy of the object (AddressSpace::getWriteable). This is the COW
+/// mechanism (which gets used
+/// for all objects, not just globals).
+/// 
+/// From the point of view of the state and this mapping there is no distinction
+/// between stack, heap, and global
+/// objects. The only special handling for stack objects is that the MemoryObject is
+/// marked as isLocal and the
+/// MemoryObject is stored in the StackFrame alloca list. When the StackFrame is
+/// popped these objects are
+/// then unbound so that the state can no longer access the memory directly
+/// (references to the memory object
+/// may still remain in ReadExprs, but conceptually the actual memory is no longer
+/// addressable).
+/// 
+/// It is also important that the AddressSpace mapping is ordered. We use this when
+/// we need to resolve a symbolic
+/// address to an ObjectState by first getting a particular value for the symbolic
+/// address, and using that value to start
+/// looking for objects that the pointer can fall within.
+/// Difference betweens MemoryObjects and ObjectStates ?
+///
+/// @section expression Expressions
+/// The various Expr classes mostly model the llvm instruction set. ref<Expr> is
+/// used to maintain the reference count
+/// but also embeds any constant expressions. In fact in the current code base
+/// ConstantExprs should almost never be
+/// created. Most of the Expr's are straightforward. Some of the most important ones
+/// are Concat?Expr, which join
+/// some number of bytes into a larger type, ExtractExpr which extracts smaller
+/// types from larger ones, and ReadExpr
+/// which is a symbolic array access.
+///
+/// The way memory is implemented all accesses are broken down into byte level
+/// operations. This means that the
+/// memory system (by which I mean the ObjectState data structure) tends to use a
+/// lot of ExtractExpr and Concat?Expr,
+/// so it is very important that these expressions fold their operands when
+/// possible.
+///
+/// The ReadExpr is probably the most important one. Conceptually it is simply an
+/// index and a list of (index, value)
+/// updates (writes). The ReadExpr evaluates to all the values for which the two
+/// indices can be equal. The ObjectState
+/// structure uses a cache for concrete writes and for symbolic writes at concrete
+/// indices, but for writes at symbolic
+/// indices it must construct a list of such updates. These are stored in the
+/// UpdateList and UpdateNode structures
+/// which are again immutable data structures so that copy is cheap and the sharing
+/// is exploited.
+/// 
+/// @section searcher Searcher
+/// Base classe: klee::Searcher. The Executor uses a Searcher to select the next
+/// state (i.e. program instance following a single path) for which an
+/// instruction
+/// will be executed. There are multiple implementations of Searcher in klee,
+/// implementing different search policies. klee::RandomSearcher selects the next state randomly.
+/// klee::DFSSearcher uses a depth first approach. klee::MergingSearcher tries
+/// to merge states ?

Added: klee/trunk/examples/regexp/Regexp.c
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/examples/regexp/Regexp.c?rev=72205&view=auto

==============================================================================
--- klee/trunk/examples/regexp/Regexp.c (added)
+++ klee/trunk/examples/regexp/Regexp.c Wed May 20 23:36:41 2009
@@ -0,0 +1,62 @@
+/* 
+ * Simple regular expression matching.
+ *
+ * From:
+ *   The Practice of Programming
+ *   Brian W. Kernighan, Rob Pike
+ *
+ */ 
+
+#include <klee/klee.h>
+
+static int matchhere(char*,char*);
+
+static int matchstar(int c, char *re, char *text) {
+  do {
+    if (matchhere(re, text))
+      return 1;
+  } while (*text != '\0' && (*text++ == c || c== '.'));
+  return 0;
+}
+
+static int matchhere(char *re, char *text) {
+  if (re[0] == '\0')
+     return 0;
+  if (re[1] == '*')
+    return matchstar(re[0], re+2, text);
+  if (re[0] == '$' && re[1]=='\0')
+    return *text == '\0';
+  if (*text!='\0' && (re[0]=='.' || re[0]==*text))
+    return matchhere(re+1, text+1);
+  return 0;
+}
+
+int match(char *re, char *text) {
+  if (re[0] == '^')
+    return matchhere(re+1, text);
+  do {
+    if (matchhere(re, text))
+      return 1;
+  } while (*text++ != '\0');
+  return 0;
+}
+
+/*
+ * Harness for testing with KLEE.
+ */
+
+// The size of the buffer to test with.
+#define SIZE 7
+
+int main() {
+  // The input regular expression.
+  char re[SIZE];
+  
+  // Make the input symbolic. 
+  klee_make_symbolic_name(re, sizeof re, "re");
+
+  // Try to match against a constant string "hello".
+  match(re, "hello");
+
+  return 0;
+}

Added: klee/trunk/examples/regexp/notes.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/examples/regexp/notes.txt?rev=72205&view=auto

==============================================================================
--- klee/trunk/examples/regexp/notes.txt (added)
+++ klee/trunk/examples/regexp/notes.txt Wed May 20 23:36:41 2009
@@ -0,0 +1,26 @@
+clang -m32 -I ~/private/klee/include -c -emit-llvm Regexp.c
+
+klee Regexp.o
+klee --only-output-states-covering-new Regexp.o
+ls -l klee-out-0
+ls -l klee-out-1
+ls -l klee-last
+
+cd klee-last
+klee-bout-tool *.bout
+klee-bout-tool --trim-zeros *.bout
+
+Stuff to show:
+Adding klee_prefer_cex
+
+PrintStats.py klee-last
+
+PrintStats.py klee-last
+Why not 100% coverage?
+
+clang -g -m32 -I ~/private/klee/include -c -emit-llvm Regexp.c
+
+KCachegrind?
+
+Disable klee_assume, show coverage again (why is klee-check-div getting pulled
+in?)

Added: klee/trunk/examples/sort/sort.c
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/examples/sort/sort.c?rev=72205&view=auto

==============================================================================
--- klee/trunk/examples/sort/sort.c (added)
+++ klee/trunk/examples/sort/sort.c Wed May 20 23:36:41 2009
@@ -0,0 +1,78 @@
+#include <klee/klee.h>
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <assert.h>
+
+static void insert_ordered(int *array, unsigned nelem, int item) {
+  unsigned i = 0;
+
+  for (; i != nelem; ++i) {
+    if (item < array[i]) {
+      memmove(&array[i+1], &array[i], sizeof(*array) * (nelem - i));
+      break;
+    }
+  }
+
+  array[i] = item;
+}
+
+void bubble_sort(int *array, unsigned nelem) {
+  for (;;) {
+    int done = 1;
+
+    for (unsigned i = 0; i + 1 < nelem; ++i) {
+      if (array[i+1] < array[i]) {
+        int t = array[i + 1];
+        array[i + 1] = array[i];
+        array[i] = t;
+        done = 0;
+      }
+    }
+
+    break;
+  }
+}
+
+void insertion_sort(int *array, unsigned nelem) {
+  int *temp = malloc(sizeof(*temp) * nelem);
+
+  for (unsigned i = 0; i != nelem; ++i)
+    insert_ordered(temp, i, array[i]);
+
+  memcpy(array, temp, sizeof(*array) * nelem);
+  free(temp);
+}
+
+void test(int *array, unsigned nelem) {
+  int *temp1 = malloc(sizeof(*array) * nelem);
+  int *temp2 = malloc(sizeof(*array) * nelem);
+
+  printf("input: [%d, %d, %d, %d]\n",
+         array[0], array[1], array[2], array[3]);
+
+  memcpy(temp1, array, sizeof(*array) * 4);
+  memcpy(temp2, array, sizeof(*array) * 4);
+
+  insertion_sort(temp1, 4);
+  bubble_sort(temp2, 4);
+
+  printf("insertion_sort: [%d, %d, %d, %d]\n",
+         temp1[0], temp1[1], temp1[2], temp1[3]);
+
+  printf("bubble_sort   : [%d, %d, %d, %d]\n",
+         temp2[0], temp2[1], temp2[2], temp2[3]);
+
+  for (unsigned i = 0; i != nelem; ++i)
+    assert(temp1[i] == temp2[i]);
+}
+
+int main() {
+  int input[4] = { 4, 3, 2, 1};
+
+  klee_make_symbolic(&input, sizeof(input));
+  test(input, 4);
+
+  return 0;
+}

Added: klee/trunk/include/expr/Lexer.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/expr/Lexer.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/expr/Lexer.h (added)
+++ klee/trunk/include/expr/Lexer.h Wed May 20 23:36:41 2009
@@ -0,0 +1,114 @@
+//===-- Lexer.h -------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPR_LEXER_H
+#define KLEE_EXPR_LEXER_H
+
+#include <string>
+
+namespace llvm {
+  class MemoryBuffer;
+}
+
+namespace klee {
+namespace expr {
+  struct Token {
+    enum Kind {
+      At,                       /// '@'
+      Arrow,                    /// '->'
+      Colon,                    /// ':'
+      Comma,                    /// ','
+      Comment,                  /// #[^\n]+
+      EndOfFile,                /// <end of file>
+      Equals,                   /// ' = '
+      Identifier,               /// [a-zA-Z_][a-zA-Z0-9._]*
+      KWFalse,                  /// 'false'
+      KWQuery,                  /// 'query'
+      KWReserved,               /// fp[0-9]+([.].*)?, i[0-9]+
+      KWTrue,                   /// 'true'
+      KWWidth,                  /// w[0-9]+
+      LBrace,                   /// '{'
+      LParen,                   /// '('
+      LSquare,                  /// '['
+      Number,                   /// [+-]?[0-9][a-zA-Z0-9_]+
+      RBrace,                   /// '}'
+      RParen,                   /// ')'
+      RSquare,                  /// ']'
+      Semicolon,                /// ';'
+      Unknown                   /// <other>
+    };
+
+    Kind        kind;           /// The token kind.
+    const char *start;          /// The beginning of the token string. 
+    unsigned    length;         /// The length of the token.
+    unsigned    line;           /// The line number of the start of this token.
+    unsigned    column;         /// The column number at the start of
+                                /// this token.
+
+    /// getKindName - The name of this token's kind.
+    const char *getKindName() const;
+
+    /// getString - The string spanned by this token. This is not
+    /// particularly efficient, use start and length when reasonable.
+    std::string getString() const { return std::string(start, length); }
+
+    /// isKeyword - True if this token is a keyword.
+    bool isKeyword() const { 
+      return kind >= KWFalse && kind <= KWTrue; 
+    }
+
+    // dump - Dump the token to stderr.
+    void dump();
+  };
+
+  /// Lexer - Interface for lexing tokens from a .pc language file.
+  class Lexer {
+    const char *BufferPos;      /// The current lexer position.
+    const char *BufferEnd;      /// The buffer end position.
+    unsigned    LineNumber;     /// The current line.
+    unsigned    ColumnNumber;   /// The current column.
+
+    /// GetNextChar - Eat a character or -1 from the stream.
+    int GetNextChar();
+
+    /// PeekNextChar - Return the next character without consuming it
+    /// from the stream. This does not perform newline
+    /// canonicalization.
+    int PeekNextChar();
+
+    /// SetTokenKind - Set the token kind and length (using the
+    /// token's start pointer, which must have been initialized).
+    Token &SetTokenKind(Token &Result, Token::Kind k);
+
+    /// SetTokenKind - Set an identifiers token kind. This has the
+    /// same requirements as SetTokenKind and additionally takes care
+    /// of keyword recognition.
+    Token &SetIdentifierTokenKind(Token &Result);
+  
+    void SkipToEndOfLine();
+
+    /// LexNumber - Lex a number which does not have a base specifier.
+    Token &LexNumber(Token &Result);
+
+    /// LexIdentifier - Lex an identifier.
+    Token &LexIdentifier(Token &Result);
+
+  public:
+    explicit Lexer(const llvm::MemoryBuffer *_buf);
+    ~Lexer();
+
+    /// Lex - Return the next token from the file or EOF continually
+    /// when the end of the file is reached. The input argument is
+    /// used as the result, for convenience.
+    Token &Lex(Token &Result);
+  };
+}
+}
+
+#endif

Added: klee/trunk/include/expr/Parser.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/expr/Parser.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/expr/Parser.h (added)
+++ klee/trunk/include/expr/Parser.h Wed May 20 23:36:41 2009
@@ -0,0 +1,178 @@
+//===-- Parser.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPR_PARSER_H
+#define KLEE_EXPR_PARSER_H
+
+#include "klee/Expr.h"
+
+#include <vector>
+#include <string>
+
+namespace llvm {
+  class MemoryBuffer;
+}
+
+namespace klee {
+namespace expr {
+  // These are the language types we manipulate.
+  typedef ref<Expr> ExprHandle;
+  typedef UpdateList VersionHandle;
+  
+  /// Identifier - Wrapper for a uniqued string.
+  struct Identifier {
+    const std::string Name;
+
+  public:
+    Identifier(const std::string _Name) : Name(_Name) {}
+  };
+
+  // FIXME: Do we have a use for tracking source locations?
+
+  /// Decl - Base class for top level declarations.
+  class Decl {
+  public:
+    Decl();
+    virtual ~Decl() {}
+
+    /// dump - Dump the AST node to stderr.
+    virtual void dump() = 0;
+  };
+
+  /// ArrayDecl - Array declarations.
+  ///
+  /// For example:
+  ///   array obj : 32 -> 8 = symbolic
+  ///   array obj[32] : 32 -> 8 = { ... }
+  class ArrayDecl : public Decl {
+  public:
+    /// Name - The name of this array.
+    const Identifier *Name;
+
+    /// Size - The maximum array size (or 0 if unspecified). Concrete
+    /// arrays always are specified with a size.
+    const unsigned Size;
+
+    /// Domain - The width of indices.
+    const unsigned Domain;
+
+    /// Range - The width of array contents.
+    const unsigned Range;
+    
+    /// Contents - The initial contents of the array. The array is
+    /// symbolic if no contents are specified. The contained
+    /// expressions are guaranteed to be constants.
+    const std::vector<ExprHandle> Contents;
+
+  public:
+    template<typename InputIterator>
+    ArrayDecl(const Identifier *_Name, unsigned _Size, 
+              unsigned _Domain, unsigned _Range,
+              InputIterator ContentsBegin=InputIterator(), 
+              InputIterator ContentsEnd=InputIterator()) 
+      : Name(_Name), Size(_Size), Domain(_Domain), Range(_Range),
+        Contents(ContentsBegin, ContentsEnd) {}
+  };
+
+  /// VarDecl - Variable declarations, used to associate names to
+  /// expressions or array versions outside of expressions.
+  /// 
+  /// For example:
+  // FIXME: What syntax are we going to use for this? We need it.
+  class VarDecl : public Decl {
+  public:
+    const Identifier *Name;    
+  };
+
+  /// ExprVarDecl - Expression variable declarations.
+  class ExprVarDecl : public VarDecl {
+  public:
+    ExprHandle Value;
+  };
+
+  /// VersionVarDecl - Array version variable declarations.
+  class VersionVarDecl : public VarDecl {
+  public:
+    VersionHandle Value;
+  };
+
+  /// CommandDecl - Base class for language commands.
+  class CommandDecl : public Decl {
+  public:
+    const Identifier *Name;
+  };
+
+  /// QueryCommand - Query commands.
+  ///
+  /// (query [ ... constraints ... ] expression)
+  /// (query [ ... constraints ... ] expression values)
+  /// (query [ ... constraints ... ] expression values objects)
+  class QueryCommand : public CommandDecl {
+  public:
+    // FIXME: One issue with STP... should we require the FE to
+    // guarantee that these are consistent? That is a cornerstone of
+    // being able to do independence. We may want this as a flag, if
+    // we are to interface with things like SMT.
+
+    /// Constraints - The list of constraints to assume for this
+    /// expression.
+    const std::vector<ExprHandle> Constraints;
+    
+    /// Query - The expression being queried.
+    ExprHandle Query;
+
+    /// Values - The expressions for which counterexamples should be
+    /// given if the query is invalid.
+    const std::vector<ExprHandle> Values;
+
+    /// Objects - Symbolic arrays whose initial contents should be
+    /// given if the query is invalid.
+    const std::vector<ArrayDecl> Objects;
+
+  public:
+    template<typename InputIterator>
+    QueryCommand(InputIterator ConstraintsBegin, 
+                 InputIterator ConstraintsEnd,
+                 ExprHandle _Query)
+      : Constraints(ConstraintsBegin, ConstraintsEnd),
+        Query(_Query) {}
+
+    virtual void dump();
+  };
+  
+  /// Parser - Public interface for parsing a .pc language file.
+  class Parser {
+  protected:
+    Parser();
+  public:
+    virtual ~Parser();
+
+    /// SetMaxErrors - Suppress anything beyond the first N errors.
+    virtual void SetMaxErrors(unsigned N) = 0;
+    
+    /// GetNumErrors - Return the number of encountered errors.
+    virtual unsigned GetNumErrors() const = 0;
+
+    /// ParseTopLevelDecl - Parse and return a top level declaration,
+    /// which the caller assumes ownership of.
+    ///
+    /// \return NULL indicates the end of the file has been reached.
+    virtual Decl *ParseTopLevelDecl() = 0;
+
+    /// CreateParser - Create a parser implementation for the given
+    /// MemoryBuffer.
+    ///
+    /// \arg Name - The name to use in diagnostic messages.
+    static Parser *Create(const std::string Name,
+                          const llvm::MemoryBuffer *MB);
+  };
+}
+}
+
+#endif

Added: klee/trunk/include/klee/Config/config.h.in
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Config/config.h.in?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Config/config.h.in (added)
+++ klee/trunk/include/klee/Config/config.h.in Wed May 20 23:36:41 2009
@@ -0,0 +1,64 @@
+/* include/klee/Config/config.h.in.  Generated from autoconf/configure.tmp.ac by autoheader.  */
+
+/* Define if stplog enabled */
+#undef ENABLE_STPLOG
+
+/* Does the platform use __ctype_b_loc, etc. */
+#undef HAVE_CTYPE_EXTERNALS
+
+/* Define to 1 if you have the <inttypes.h> header file. */
+#undef HAVE_INTTYPES_H
+
+/* Define to 1 if you have the <memory.h> header file. */
+#undef HAVE_MEMORY_H
+
+/* Define to 1 if you have the <selinux/selinux.h> header file. */
+#undef HAVE_SELINUX_SELINUX_H
+
+/* Define to 1 if you have the <stdint.h> header file. */
+#undef HAVE_STDINT_H
+
+/* Define to 1 if you have the <stdlib.h> header file. */
+#undef HAVE_STDLIB_H
+
+/* Define to 1 if you have the <strings.h> header file. */
+#undef HAVE_STRINGS_H
+
+/* Define to 1 if you have the <string.h> header file. */
+#undef HAVE_STRING_H
+
+/* Define to 1 if you have the <sys/acl.h> header file. */
+#undef HAVE_SYS_ACL_H
+
+/* Define to 1 if you have the <sys/stat.h> header file. */
+#undef HAVE_SYS_STAT_H
+
+/* Define to 1 if you have the <sys/types.h> header file. */
+#undef HAVE_SYS_TYPES_H
+
+/* Define to 1 if you have the <unistd.h> header file. */
+#undef HAVE_UNISTD_H
+
+/* Path to KLEE's uClibc */
+#undef KLEE_UCLIBC
+
+/* Define to the address where bug reports for this package should be sent. */
+#undef PACKAGE_BUGREPORT
+
+/* Define to the full name of this package. */
+#undef PACKAGE_NAME
+
+/* Define to the full name and version of this package. */
+#undef PACKAGE_STRING
+
+/* Define to the one symbol short name of this package. */
+#undef PACKAGE_TARNAME
+
+/* Define to the version of this package. */
+#undef PACKAGE_VERSION
+
+/* Configuration for runtime libraries */
+#undef RUNTIME_CONFIGURATION
+
+/* Define to 1 if you have the ANSI C header files. */
+#undef STDC_HEADERS

Added: klee/trunk/include/klee/Constraints.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Constraints.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Constraints.h (added)
+++ klee/trunk/include/klee/Constraints.h Wed May 20 23:36:41 2009
@@ -0,0 +1,79 @@
+//===-- Constraints.h -------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_CONSTRAINTS_H
+#define KLEE_CONSTRAINTS_H
+
+#include "klee/Expr.h"
+
+// FIXME: Currently we use ConstraintManager for two things: to pass
+// sets of constraints around, and to optimize constraints. We should
+// move the first usage into a separate data structure
+// (ConstraintSet?) which ConstraintManager could embed if it likes.
+namespace klee {
+
+class ExprVisitor;
+  
+class ConstraintManager {
+public:
+  typedef std::vector< ref<Expr> > constraints_ty;
+  typedef constraints_ty::iterator iterator;
+  typedef constraints_ty::const_iterator const_iterator;
+
+  ConstraintManager() {}
+
+  // create from constraints with no optimization
+  explicit
+  ConstraintManager(const std::vector< ref<Expr> > &_constraints) :
+    constraints(_constraints) {}
+
+  ConstraintManager(const ConstraintManager &cs) : constraints(cs.constraints) {}
+
+  typedef std::vector< ref<Expr> >::const_iterator constraint_iterator;
+
+  // given a constraint which is known to be valid, attempt to 
+  // simplify the existing constraint set
+  void simplifyForValidConstraint(ref<Expr> e);
+
+  ref<Expr> simplifyExpr(ref<Expr> e) const;
+
+  void addConstraint(ref<Expr> e);
+  
+  bool empty() const {
+    return constraints.empty();
+  }
+  ref<Expr> back() const {
+    return constraints.back();
+  }
+  constraint_iterator begin() const {
+    return constraints.begin();
+  }
+  constraint_iterator end() const {
+    return constraints.end();
+  }
+  size_t size() const {
+    return constraints.size();
+  }
+
+  bool operator==(const ConstraintManager &other) const {
+    return constraints == other.constraints;
+  }
+  
+private:
+  std::vector< ref<Expr> > constraints;
+
+  // returns true iff the constraints were modified
+  bool rewriteConstraints(ExprVisitor &visitor);
+
+  void addConstraintInternal(ref<Expr> e);
+};
+
+}
+
+#endif /* KLEE_CONSTRAINTS_H */

Added: klee/trunk/include/klee/ExecutionState.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/ExecutionState.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/ExecutionState.h (added)
+++ klee/trunk/include/klee/ExecutionState.h Wed May 20 23:36:41 2009
@@ -0,0 +1,250 @@
+//===-- ExecutionState.h ----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXECUTIONSTATE_H
+#define KLEE_EXECUTIONSTATE_H
+
+#include "klee/Constraints.h"
+#include "klee/Expr.h"
+#include "klee/Internal/ADT/TreeStream.h"
+
+// FIXME: We do not want to be exposing these? :(
+#include "../../lib/Core/AddressSpace.h"
+#include "klee/Internal/Module/KInstIterator.h"
+
+#include <map>
+#include <set>
+#include <vector>
+
+namespace klee {
+  class CallPathNode;
+  class Cell;
+  class KFunction;
+  class KInstruction;
+  class MemoryObject;
+  class PTreeNode;
+  class InstructionInfo;
+  class ExecutionTraceEvent;
+
+std::ostream &operator<<(std::ostream &os, const MemoryMap &mm);
+
+struct StackFrame {
+  KInstIterator caller;
+  KFunction *kf;
+  CallPathNode *callPathNode;
+
+  std::vector<const MemoryObject*> allocas;
+  Cell *locals;
+
+  /// Minimum distance to an uncovered instruction once the function
+  /// returns. This is not a good place for this but is used to
+  /// quickly compute the context sensitive minimum distance to an
+  /// uncovered instruction. This value is updated by the StatsTracker
+  /// periodically.
+  unsigned minDistToUncoveredOnReturn;
+
+  // For vararg functions: arguments not passed via parameter are
+  // stored (packed tightly) in a local (alloca) memory object. This
+  // is setup to match the way the front-end generates vaarg code (it
+  // does not pass vaarg through as expected). VACopy is lowered inside
+  // of intrinsic lowering.
+  MemoryObject *varargs;
+
+  StackFrame(KInstIterator caller, KFunction *kf);
+  StackFrame(const StackFrame &s);
+  ~StackFrame();
+};
+
+// FIXME: Redo execution trace stuff to use a TreeStream, there is no
+// need to keep this stuff in memory as far as I can tell.
+
+// each state should have only one of these guys ...
+class ExecutionTraceManager {
+public:
+  ExecutionTraceManager() : hasSeenUserMain(false) {}
+
+  void addEvent(ExecutionTraceEvent* evt);
+  void printAllEvents(std::ostream &os) const;
+
+private:
+  // have we seen a call to __user_main() yet?
+  // don't bother tracing anything until we see this,
+  // or else we'll get junky prologue shit
+  bool hasSeenUserMain;
+
+  // ugh C++ only supports polymorphic calls thru pointers
+  //
+  // WARNING: these are NEVER FREED, because they are shared
+  // across different states (when states fork), so we have 
+  // an *intentional* memory leak, but oh wellz ;)
+  std::vector<ExecutionTraceEvent*> events;
+};
+
+
+class ExecutionState {
+public:
+  typedef std::vector<StackFrame> stack_ty;
+
+private:
+  // unsupported, use copy constructor
+  ExecutionState &operator=(const ExecutionState&); 
+  std::map< std::string, std::string > fnAliases;
+
+public:
+  bool fakeState;
+  // Are we currently underconstrained?  Hack: value is size to make fake
+  // objects.
+  unsigned underConstrained;
+  unsigned depth;
+  
+  // pc - pointer to current instruction stream
+  KInstIterator pc, prevPC;
+  stack_ty stack;
+  ConstraintManager constraints;
+  mutable double queryCost;
+  double weight;
+  AddressSpace addressSpace;
+  TreeOStream pathOS, symPathOS;
+  unsigned instsSinceCovNew;
+  bool coveredNew;
+
+  // for printing execution traces when this state terminates
+  ExecutionTraceManager exeTraceMgr;
+
+  /// Disables forking, set by user code.
+  bool forkDisabled;
+
+  std::map<const std::string*, std::set<unsigned> > coveredLines;
+  PTreeNode *ptreeNode;
+
+  /// ordered list of symbolics: used to generate test cases. 
+  //
+  // FIXME: Move to a shared list structure (not critical).
+  std::vector<const MemoryObject*> symbolics;
+
+  // Used by the checkpoint/rollback methods for fake objects.
+  // FIXME: not freeing things on branch deletion.
+  MemoryMap shadowObjects;
+
+  unsigned incomingBBIndex;
+
+  std::string getFnAlias(std::string fn);
+  void addFnAlias(std::string old_fn, std::string new_fn);
+  void removeFnAlias(std::string fn);
+  
+private:
+  ExecutionState() : fakeState(false), underConstrained(0), ptreeNode(0) {};
+
+public:
+  ExecutionState(KFunction *kf);
+
+  // XXX total hack, just used to make a state so solver can
+  // use on structure
+  ExecutionState(const std::vector<ref<Expr> > &assumptions);
+
+  ~ExecutionState();
+  
+  ExecutionState *branch();
+
+  void pushFrame(KInstIterator caller, KFunction *kf);
+  void popFrame();
+
+  void addSymbolic(const MemoryObject *mo) { 
+    symbolics.push_back(mo);
+  }
+  void addConstraint(ref<Expr> e) { 
+    constraints.addConstraint(e); 
+  }
+
+  // Used for checkpoint/rollback of fake objects created during tainting.
+  ObjectState *cloneObject(ObjectState *os, MemoryObject *mo);
+
+  //
+
+  bool merge(const ExecutionState &b);
+};
+
+
+// for producing abbreviated execution traces to help visualize
+// paths and diagnose bugs
+
+class ExecutionTraceEvent {
+public:
+  // the location of the instruction:
+  std::string file;
+  unsigned line;
+  std::string funcName;
+  unsigned stackDepth;
+
+  unsigned consecutiveCount; // init to 1, increase for CONSECUTIVE
+                             // repetitions of the SAME event
+
+  ExecutionTraceEvent()
+    : file("global"), line(0), funcName("global_def"),
+      consecutiveCount(1) {}
+
+  ExecutionTraceEvent(ExecutionState& state, KInstruction* ki);
+
+  virtual ~ExecutionTraceEvent() {}
+
+  void print(std::ostream &os) const;
+
+  // return true if it shouldn't be added to ExecutionTraceManager
+  //
+  virtual bool ignoreMe() const;
+
+private:
+  virtual void printDetails(std::ostream &os) const = 0;
+};
+
+
+class FunctionCallTraceEvent : public ExecutionTraceEvent {
+public:
+  std::string calleeFuncName;
+
+  FunctionCallTraceEvent(ExecutionState& state, KInstruction* ki,
+                         const std::string& _calleeFuncName)
+    : ExecutionTraceEvent(state, ki), calleeFuncName(_calleeFuncName) {}
+
+private:
+  virtual void printDetails(std::ostream &os) const {
+    os << "CALL " << calleeFuncName;
+  }
+
+};
+
+class FunctionReturnTraceEvent : public ExecutionTraceEvent {
+public:
+  FunctionReturnTraceEvent(ExecutionState& state, KInstruction* ki)
+    : ExecutionTraceEvent(state, ki) {}
+
+private:
+  virtual void printDetails(std::ostream &os) const {
+    os << "RETURN";
+  }
+};
+
+class BranchTraceEvent : public ExecutionTraceEvent {
+public:
+  bool trueTaken;         // which side taken?
+  bool canForkGoBothWays;
+
+  BranchTraceEvent(ExecutionState& state, KInstruction* ki,
+                   bool _trueTaken, bool _isTwoWay)
+    : ExecutionTraceEvent(state, ki),
+      trueTaken(_trueTaken),
+      canForkGoBothWays(_isTwoWay) {}
+
+private:
+  virtual void printDetails(std::ostream &os) const;
+};
+
+}
+
+#endif

Added: klee/trunk/include/klee/Expr.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Expr.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Expr.h (added)
+++ klee/trunk/include/klee/Expr.h Wed May 20 23:36:41 2009
@@ -0,0 +1,808 @@
+//===-- Expr.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPR_H
+#define KLEE_EXPR_H
+
+#include "Machine.h"
+#include "klee/util/Bits.h"
+
+#include "llvm/Support/Streams.h"
+#include "llvm/ADT/SmallVector.h"
+
+#include <set>
+#include <vector>
+
+namespace llvm {
+  class Type;
+}
+
+namespace klee {
+
+class Array;
+class ConstantExpr;
+class ObjectState;
+class MemoryObject;
+
+template<class T> class ref;
+
+
+/// Class representing symbolic expressions.
+/**
+
+<b>Expression canonicalization</b>: we define certain rules for
+canonicalization rules for Exprs in order to simplify code that
+pattern matches Exprs (since the number of forms are reduced), to open
+up further chances for optimization, and to increase similarity for
+caching and other purposes.
+
+The general rules are:
+<ol>
+<li> No Expr has all constant arguments.</li>
+
+<li> Booleans:
+    <ol type="a">
+    <li> Boolean not is written as <tt>(false == ?)</tt> </li>
+     <li> \c Ne, \c Ugt, \c Uge, \c Sgt, \c Sge are not used </li>
+     <li> The only acceptable operations with boolean arguments are \c And, 
+          \c Or, \c Xor, \c Eq, as well as \c SExt, \c ZExt,
+          \c Select and \c NotOptimized. </li>
+     <li> The only boolean operation which may involve a constant is boolean not (<tt>== false</tt>). </li>
+     </ol>
+</li>
+
+<li> Linear Formulas: 
+   <ol type="a">
+   <li> For any subtree representing a linear formula, a constant
+   term must be on the LHS of the root node of the subtree.  In particular, 
+   in a BinaryExpr a constant must always be on the LHS.  For example, subtraction 
+   by a constant c is written as <tt>add(-c, ?)</tt>.  </li>
+    </ol>
+</li>
+
+
+<li> Chains are unbalanced to the right </li>
+
+</ol>
+
+
+<b>Steps required for adding an expr</b>:
+   -# Add case to printKind
+   -# Add to ExprVisitor
+   -# Add to IVC (implied value concretization) if possible
+
+Todo: Shouldn't bool \c Xor just be written as not equal?
+
+*/
+
+class Expr {
+public:
+  static unsigned count;
+  static const unsigned MAGIC_HASH_CONSTANT = 39;
+
+  /// The type of an expression is simply its width, in bits. 
+  typedef unsigned Width; 
+  
+  static const Width InvalidWidth = 0;
+  static const Width Bool = 1;
+  static const Width Int8 = 8;
+  static const Width Int16 = 16;
+  static const Width Int32 = 32;
+  static const Width Int64 = 64;
+  
+
+  enum Kind {
+    InvalidKind = -1,
+
+    // Primitive
+
+    Constant = 0,
+
+    // Special
+
+    /// Prevents optimization below the given expression.  Used for
+    /// testing: make equality constraints that KLEE will not use to
+    /// optimize to concretes.
+    NotOptimized,
+
+    //// Skip old varexpr, just for deserialization, purge at some point
+    Read=NotOptimized+2, 
+    Select,
+    Concat,
+    Extract,
+
+    // Casting,
+    
+    ZExt,
+    SExt,
+
+    // Arithmetic
+    Add,
+    Sub,
+    Mul,
+    UDiv,
+    SDiv,
+    URem,
+    SRem,
+
+    // Bit
+    And,
+    Or,
+    Xor,
+    Shl,
+    LShr,
+    AShr,
+    
+    // Compare
+    Eq,
+    Ne,  /// Not used in canonical form
+    Ult,
+    Ule,
+    Ugt, /// Not used in canonical form
+    Uge, /// Not used in canonical form
+    Slt,
+    Sle,
+    Sgt, /// Not used in canonical form
+    Sge, /// Not used in canonical form
+
+    LastKind=Sge
+  };
+
+  unsigned refCount;
+
+protected:  
+  unsigned hashValue;
+  
+public:
+  Expr() : refCount(0) { Expr::count++; }
+  virtual ~Expr() { Expr::count--; } 
+
+  virtual Kind getKind() const = 0;
+  virtual Width getWidth() const = 0;
+  
+  virtual unsigned getNumKids() const = 0;
+  virtual ref<Expr> getKid(unsigned i) const = 0;
+    
+  virtual void print(std::ostream &os) const;
+
+  /// Returns the pre-computed hash of the current expression
+  virtual unsigned hash() const { return hashValue; }
+
+  /// (Re)computes the hash of the current expression.
+  /// Returns the hash value. 
+  virtual unsigned computeHash();
+  
+  static unsigned hashConstant(uint64_t val, Width w) {
+    return val ^ (w * MAGIC_HASH_CONSTANT);
+  }
+  
+  /// Returns 0 iff b is structuraly equivalent to *this
+  int compare(const Expr &b) const;
+  virtual int compareContents(const Expr &b) const { return 0; }
+
+  // Given an array of new kids return a copy of the expression
+  // but using those children. 
+  virtual ref<Expr> rebuild(ref<Expr> kids[/* getNumKids() */]) const = 0;
+
+  ///
+
+  uint64_t getConstantValue() const;
+
+  /* Static utility methods */
+
+  static void printKind(std::ostream &os, Kind k);
+  static void printWidth(std::ostream &os, Expr::Width w);
+  static Width getWidthForLLVMType(const llvm::Type *type);
+
+  /// returns the smallest number of bytes in which the given width fits
+  static inline unsigned getMinBytesForWidth(Width w) {
+      return (w + 7) / 8;
+  }
+
+  /* Kind utilities */
+
+  /* Utility creation functions */
+  static ref<Expr> createCoerceToPointerType(ref<Expr> e);
+  static ref<Expr> createNot(ref<Expr> e);
+  static ref<Expr> createImplies(ref<Expr> hyp, ref<Expr> conc);
+  static ref<Expr> createIsZero(ref<Expr> e);
+
+  /// Create a little endian read of the given type at offset 0 of the
+  /// given object.
+  static ref<Expr> createTempRead(const Array *array, Expr::Width w);
+  
+  static ref<Expr> createPointer(uint64_t v);
+
+  // do not use
+  static Expr *createConstant(uint64_t val, Width w);
+  
+  struct CreateArg;
+  static ref<Expr> createFromKind(Kind k, std::vector<CreateArg> args);
+
+  static bool isValidKidWidth(unsigned kid, Width w) { return true; }
+  static bool needsResultType() { return false; }
+};
+// END class Expr
+
+
+
+#include "klee/util/Ref.h"
+
+struct Expr::CreateArg {
+  ref<Expr> expr;
+  Width width;
+  
+  CreateArg(Width w = Bool) : expr(0, Expr::Bool), width(w) {}
+  CreateArg(ref<Expr> e) : expr(e), width(Expr::InvalidWidth) {}
+  
+  bool isExpr() { return !isWidth(); }
+  bool isWidth() { return width != Expr::InvalidWidth; }
+};
+
+// Comparison operators
+
+inline bool operator==(const Expr &lhs, const Expr &rhs) {
+  return lhs.compare(rhs) == 0;
+}
+
+inline bool operator<(const Expr &lhs, const Expr &rhs) {
+  return lhs.compare(rhs) < 0;
+}
+
+inline bool operator!=(const Expr &lhs, const Expr &rhs) {
+  return !(lhs == rhs);
+}
+
+inline bool operator>(const Expr &lhs, const Expr &rhs) {
+  return rhs < lhs;
+}
+
+inline bool operator<=(const Expr &lhs, const Expr &rhs) {
+  return !(lhs > rhs);
+}
+
+inline bool operator>=(const Expr &lhs, const Expr &rhs) {
+  return !(lhs < rhs);
+}
+
+// Printing operators
+
+inline std::ostream &operator<<(std::ostream &os, const Expr &e) {
+  e.print(os);
+  return os;
+}
+
+inline std::ostream &operator<<(std::ostream &os, const Expr::Kind kind) {
+  Expr::printKind(os, kind);
+  return os;
+}
+
+// Terminal Exprs
+
+class ConstantExpr : public Expr {
+public:
+  static const Kind kind = Constant;
+  static const unsigned numKids = 0;
+  
+public:
+  union {
+    uint64_t asUInt64;
+  };
+  Width width;
+
+public:
+  ~ConstantExpr() {};
+  // should change the code to make this private
+  ConstantExpr(uint64_t v, Width w) : asUInt64(v), width(w) {}
+  
+  Width getWidth() const { return width; }
+  Kind getKind() const { return Constant; }
+
+  unsigned getNumKids() const { return 0; }
+  ref<Expr> getKid(unsigned i) const { return 0; }
+
+  int compareContents(const Expr &b) const { 
+    const ConstantExpr &cb = static_cast<const ConstantExpr&>(b);
+    if (width != cb.width) return width < cb.width ? -1 : 1;
+    if (asUInt64 < cb.asUInt64) {
+      return -1;
+    } else if (asUInt64 > cb.asUInt64) {
+      return 1;
+    } else {
+      return 0;
+    }
+  }
+
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { 
+    assert(0 && "rebuild() on ConstantExpr"); 
+  }
+
+  virtual unsigned computeHash();
+  
+  static ref<ConstantExpr> fromMemory(void *address, Width w);
+  void toMemory(void *address);
+
+  static ref<ConstantExpr> alloc(uint64_t v, Width w) {
+    // constructs an "optimized" ConstantExpr
+    return ref<ConstantExpr>(v, w);
+  }
+  
+  static ref<ConstantExpr> create(uint64_t v, Width w) {
+    assert(v == bits64::truncateToNBits(v, w) &&
+           "invalid constant");
+    return alloc(v, w);
+  }
+};
+
+  
+// Utility classes
+
+class BinaryExpr : public Expr {
+public:
+  ref<Expr> left, right;
+
+public:
+  unsigned getNumKids() const { return 2; }
+  ref<Expr> getKid(unsigned i) const { 
+    if(i == 0)
+      return left;
+    if(i == 1)
+      return right;
+    return 0;
+  }
+ 
+protected:
+  BinaryExpr(const ref<Expr> &l, const ref<Expr> &r) : left(l), right(r) {}
+};
+
+
+class CmpExpr : public BinaryExpr {
+
+protected:
+  CmpExpr(ref<Expr> l, ref<Expr> r) : BinaryExpr(l,r) {}
+  
+public:                                                       
+  Width getWidth() const { return Bool; }
+};
+
+// Special
+
+class NotOptimizedExpr : public Expr {
+public:
+  static const Kind kind = NotOptimized;
+  static const unsigned numKids = 1;
+  ref<Expr> src;
+
+  static ref<Expr> alloc(const ref<Expr> &src) {
+    ref<Expr> r(new NotOptimizedExpr(src));
+    r.computeHash();
+    return r;
+  }
+  
+  static ref<Expr> create(ref<Expr> src);
+  
+  Width getWidth() const { return src.getWidth(); }
+  Kind getKind() const { return NotOptimized; }
+
+  unsigned getNumKids() const { return 1; }
+  ref<Expr> getKid(unsigned i) const { return src; }
+
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { return create(kids[0]); }
+
+private:
+  NotOptimizedExpr(const ref<Expr> &_src) : src(_src) {}
+};
+
+
+/// Class representing a byte update of an array.
+class UpdateNode {
+  friend class UpdateList;
+  friend class STPBuilder; // for setting STPArray
+
+  mutable unsigned refCount;
+  // gross
+  mutable void *stpArray;
+  // cache instead of recalc
+  unsigned hashValue;
+
+public:
+  const UpdateNode *next;
+  ref<Expr> index, value;
+  
+private:
+  /// size of this update sequence, including this update
+  unsigned size;
+  
+public:
+  UpdateNode(const UpdateNode *_next, 
+             const ref<Expr> &_index, 
+             const ref<Expr> &_value);
+
+  unsigned getSize() const { return size; }
+
+  int compare(const UpdateNode &b) const;  
+  unsigned hash() const { return hashValue; }
+
+private:
+  UpdateNode() : refCount(0), stpArray(0) {}
+  ~UpdateNode();
+
+  unsigned computeHash();
+};
+
+class Array {
+public:
+  const MemoryObject *object;
+  unsigned id;
+  unsigned size;
+
+  // FIXME: This does not belong here.
+  mutable void *stpInitialArray;
+
+public:
+  // NOTE: id's ***MUST*** be unique to ensure sanity w.r.t. STP,
+  // which hashes different arrays with the same id to the same
+  // object! We should probably use the pointer for talking to STP, as
+  // long as we can guarantee that it won't be a "stale" reference
+  // once we have freed it.
+  Array(const MemoryObject *_object, unsigned _id, uint64_t _size) 
+    : object(_object), id(_id), size(_size), stpInitialArray(0) {}
+  ~Array() {
+    // FIXME: This relies on caller to delete the STP array.
+    assert(!stpInitialArray && "Array must be deleted by caller!");
+  }
+};
+
+/// Class representing a complete list of updates into an array. 
+/** The main trick is the isRooted bit, which enables important optimizations. 
+    ...
+ */
+class UpdateList { 
+  friend class ReadExpr; // for default constructor
+
+public:
+  const Array *root;
+  
+  /// pointer to the most recent update node
+  const UpdateNode *head;
+  
+  // shouldn't this be part of the ReadExpr? 
+  bool isRooted;
+
+public:
+  UpdateList(const Array *_root, bool isRooted, const UpdateNode *_head);
+  UpdateList(const UpdateList &b);
+  ~UpdateList();
+  
+  UpdateList &operator=(const UpdateList &b);
+
+  /// size of this update list
+  unsigned getSize() const { return (head ? head->getSize() : 0); }
+  
+  void extend(const ref<Expr> &index, const ref<Expr> &value);
+
+  int compare(const UpdateList &b) const;
+  unsigned hash() const;
+};
+
+/// Class representing a one byte read from an array. 
+class ReadExpr : public Expr {
+public:
+  static const Kind kind = Read;
+  static const unsigned numKids = 1;
+  
+public:
+  UpdateList updates;
+  ref<Expr> index;
+
+public:
+  static ref<Expr> alloc(const UpdateList &updates, const ref<Expr> &index) {
+    ref<Expr> r(new ReadExpr(updates, index));
+    r.computeHash();
+    return r;
+  }
+  
+  static ref<Expr> create(const UpdateList &updates, ref<Expr> i);
+  
+  Width getWidth() const { return Expr::Int8; }
+  Kind getKind() const { return Read; }
+  
+  unsigned getNumKids() const { return numKids; }
+  ref<Expr> getKid(unsigned i) const { return !i ? index : 0; }  
+  
+  int compareContents(const Expr &b) const;
+
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { 
+    return create(updates, kids[0]);
+  }
+
+  virtual unsigned computeHash();
+
+private:
+  ReadExpr(const UpdateList &_updates, const ref<Expr> &_index) : 
+    updates(_updates), index(_index) {}
+};
+
+
+/// Class representing an if-then-else expression.
+class SelectExpr : public Expr {
+public:
+  static const Kind kind = Select;
+  static const unsigned numKids = 3;
+  
+public:
+  ref<Expr> cond, trueExpr, falseExpr;
+
+public:
+  static ref<Expr> alloc(const ref<Expr> &c, const ref<Expr> &t, const ref<Expr> &f) {
+    ref<Expr> r(new SelectExpr(c, t, f));
+    r.computeHash();
+    return r;
+  }
+  
+  static ref<Expr> create(ref<Expr> c, ref<Expr> t, ref<Expr> f);
+
+  Width getWidth() const { return trueExpr.getWidth(); }
+  Kind getKind() const { return Select; }
+
+  unsigned getNumKids() const { return numKids; }
+  ref<Expr> getKid(unsigned i) const { 
+        switch(i) {
+        case 0: return cond;
+        case 1: return trueExpr;
+        case 2: return falseExpr;
+        default: return 0;
+        }
+   }
+
+  static bool isValidKidWidth(unsigned kid, Width w) {
+    if (kid == 0)
+      return w == Bool;
+    else
+      return true;
+  }
+    
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { 
+    return create(kids[0], kids[1], kids[2]);
+  }
+
+private:
+  SelectExpr(const ref<Expr> &c, const ref<Expr> &t, const ref<Expr> &f) 
+    : cond(c), trueExpr(t), falseExpr(f) {}
+};
+
+
+/** Children of a concat expression can have arbitrary widths.  
+    Kid 0 is the left kid, kid 1 is the right kid.
+*/
+class ConcatExpr : public Expr { 
+public: 
+  static const Kind kind = Concat;
+  static const unsigned numKids = 2;
+
+private:
+  Width width;
+  ref<Expr> left, right;  
+
+public:
+  static ref<Expr> alloc(const ref<Expr> &l, const ref<Expr> &r) {
+    ref<Expr> c(new ConcatExpr(l, r));
+    c.computeHash();
+    return c;
+  }
+  
+  static ref<Expr> create(const ref<Expr> &l, const ref<Expr> &r);
+
+  Width getWidth() const { return width; }
+  Kind getKind() const { return kind; }
+  ref<Expr> getLeft() const { return left; }
+  ref<Expr> getRight() const { return right; }
+
+  unsigned getNumKids() const { return numKids; }
+  ref<Expr> getKid(unsigned i) const { 
+    if (i == 0) return left; 
+    else if (i == 1) return right;
+    else return NULL;
+  }
+
+  /// Shortcuts to create larger concats.  The chain returned is unbalanced to the right
+  static ref<Expr> createN(unsigned nKids, const ref<Expr> kids[]);
+  static ref<Expr> create4(const ref<Expr> &kid1, const ref<Expr> &kid2,
+			   const ref<Expr> &kid3, const ref<Expr> &kid4);
+  static ref<Expr> create8(const ref<Expr> &kid1, const ref<Expr> &kid2,
+			   const ref<Expr> &kid3, const ref<Expr> &kid4,
+			   const ref<Expr> &kid5, const ref<Expr> &kid6,
+			   const ref<Expr> &kid7, const ref<Expr> &kid8);
+  
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { return create(kids[0], kids[1]); }
+
+
+  /* These will be eliminated */
+  bool is2ByteConcat() const { return false; }
+  bool is4ByteConcat() const { return false; }
+  bool is8ByteConcat() const { return false; }
+  
+private:
+  ConcatExpr(const ref<Expr> &l, const ref<Expr> &r) : left(l), right(r) {
+    width = l.getWidth() + r.getWidth();
+  }
+};
+
+
+/** This class represents an extract from expression {\tt expr}, at
+    bit offset {\tt offset} of width {\tt width}.  Bit 0 is the right most 
+    bit of the expression.
+ */
+class ExtractExpr : public Expr { 
+public:
+  static const Kind kind = Extract;
+  static const unsigned numKids = 1;
+  
+public:
+  ref<Expr> expr;
+  unsigned offset;
+  Width width;
+
+public:  
+  static ref<Expr> alloc(const ref<Expr> &e, unsigned o, Width w) {
+    ref<Expr> r(new ExtractExpr(e, o, w));
+    r.computeHash();
+    return r;
+  }
+  
+  /// Creates an ExtractExpr with the given bit offset and width
+  static ref<Expr> create(ref<Expr> e, unsigned bitOff, Width w);
+
+  /// Creates an ExtractExpr with the given byte offset and width
+  static ref<Expr> createByteOff(ref<Expr> e, unsigned byteOff, Width w=Expr::Int8);
+
+  Width getWidth() const { return width; }
+  Kind getKind() const { return Extract; }
+
+  unsigned getNumKids() const { return numKids; }
+  ref<Expr> getKid(unsigned i) const { return expr; }
+
+  int compareContents(const Expr &b) const {
+    const ExtractExpr &eb = static_cast<const ExtractExpr&>(b);
+    if (offset != eb.offset) return offset < eb.offset ? -1 : 1;
+    if (width != eb.width) return width < eb.width ? -1 : 1;
+    return 0;
+  }
+
+  virtual ref<Expr> rebuild(ref<Expr> kids[]) const { 
+    return create(kids[0], offset, width);
+  }
+
+  virtual unsigned computeHash();
+
+private:
+  ExtractExpr(const ref<Expr> &e, unsigned b, Width w) 
+    : expr(e),offset(b),width(w) {}
+};
+
+
+// Casting
+
+class CastExpr : public Expr {
+public:
+  ref<Expr> src;
+  Width width;
+
+public:
+  CastExpr(const ref<Expr> &e, Width w) : src(e), width(w) {}
+
+  Width getWidth() const { return width; }
+
+  unsigned getNumKids() const { return 1; }
+  ref<Expr> getKid(unsigned i) const { return (i==0) ? src : 0; }
+  
+  static bool needsResultType() { return true; }
+  
+  int compareContents(const Expr &b) const {
+    const CastExpr &eb = static_cast<const CastExpr&>(b);
+    if (width != eb.width) return width < eb.width ? -1 : 1;
+    return 0;
+  }
+
+  virtual unsigned computeHash();
+};
+
+#define CAST_EXPR_CLASS(_class_kind)                        \
+class _class_kind ## Expr : public CastExpr {               \
+public:                                                     \
+  static const Kind kind = _class_kind;                     \
+  static const unsigned numKids = 1;                        \
+public:                                                     \
+    _class_kind ## Expr(ref<Expr> e, Width w) : CastExpr(e,w) {} \
+    static ref<Expr> alloc(const ref<Expr> &e, Width w) {                   \
+      ref<Expr> r(new _class_kind ## Expr(e, w));                     \
+      r.computeHash();                                                    \
+      return r;                                                       \
+    }                                                       \
+    static ref<Expr> create(const ref<Expr> &e, Width w);                   \
+    Kind getKind() const { return _class_kind; }            \
+    virtual ref<Expr> rebuild(ref<Expr> kids[]) const { \
+      return create(kids[0], width); \
+    } \
+};                                                          \
+
+CAST_EXPR_CLASS(SExt)
+CAST_EXPR_CLASS(ZExt)
+
+// Arithmetic/Bit Exprs
+
+#define ARITHMETIC_EXPR_CLASS(_class_kind)                      \
+class _class_kind ## Expr : public BinaryExpr {                 \
+public:                                                         \
+  static const Kind kind = _class_kind;                         \
+  static const unsigned numKids = 2;                            \
+public:                                                         \
+    _class_kind ## Expr(const ref<Expr> &l, const ref<Expr> &r) : BinaryExpr(l,r) {}  \
+    static ref<Expr> alloc(const ref<Expr> &l, const ref<Expr> &r) {          \
+      ref<Expr> res(new _class_kind ## Expr (l, r));                      \
+      res.computeHash();                                                      \
+      return res;                                                         \
+    }                                                           \
+    static ref<Expr> create(const ref<Expr> &l, const ref<Expr> &r);                      \
+    Width getWidth() const { return left.getWidth(); }            \
+    Kind getKind() const { return _class_kind; }                \
+    virtual ref<Expr> rebuild(ref<Expr> kids[]) const { \
+      return create(kids[0], kids[1]); \
+    } \
+};                                                              \
+
+ARITHMETIC_EXPR_CLASS(Add)
+ARITHMETIC_EXPR_CLASS(Sub)
+ARITHMETIC_EXPR_CLASS(Mul)
+ARITHMETIC_EXPR_CLASS(UDiv)
+ARITHMETIC_EXPR_CLASS(SDiv)
+ARITHMETIC_EXPR_CLASS(URem)
+ARITHMETIC_EXPR_CLASS(SRem)
+ARITHMETIC_EXPR_CLASS(And)
+ARITHMETIC_EXPR_CLASS(Or)
+ARITHMETIC_EXPR_CLASS(Xor)
+ARITHMETIC_EXPR_CLASS(Shl)
+ARITHMETIC_EXPR_CLASS(LShr)
+ARITHMETIC_EXPR_CLASS(AShr)
+
+// Comparison Exprs
+
+#define COMPARISON_EXPR_CLASS(_class_kind)                  \
+class _class_kind ## Expr : public CmpExpr {                \
+public:                                                     \
+  static const Kind kind = _class_kind;                     \
+  static const unsigned numKids = 2;                        \
+public:                                                     \
+    _class_kind ## Expr(const ref<Expr> &l, const ref<Expr> &r) : CmpExpr(l,r) {} \
+    static ref<Expr> alloc(const ref<Expr> &l, const ref<Expr> &r) {                  \
+      ref<Expr> res(new _class_kind ## Expr (l, r));                    \
+      res.computeHash();                                                    \
+      return res;                                                       \
+    }                                                       \
+    static ref<Expr> create(const ref<Expr> &l, const ref<Expr> &r);                  \
+    Kind getKind() const { return _class_kind; }            \
+    virtual ref<Expr> rebuild(ref<Expr> kids[]) const { \
+      return create(kids[0], kids[1]); \
+    } \
+};                                                          \
+
+COMPARISON_EXPR_CLASS(Eq)
+COMPARISON_EXPR_CLASS(Ne)
+COMPARISON_EXPR_CLASS(Ult)
+COMPARISON_EXPR_CLASS(Ule)
+COMPARISON_EXPR_CLASS(Ugt)
+COMPARISON_EXPR_CLASS(Uge)
+COMPARISON_EXPR_CLASS(Slt)
+COMPARISON_EXPR_CLASS(Sle)
+COMPARISON_EXPR_CLASS(Sgt)
+COMPARISON_EXPR_CLASS(Sge)
+
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/IncompleteSolver.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/IncompleteSolver.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/IncompleteSolver.h (added)
+++ klee/trunk/include/klee/IncompleteSolver.h Wed May 20 23:36:41 2009
@@ -0,0 +1,108 @@
+//===-- IncompleteSolver.h --------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_INCOMPLETESOLVER_H
+#define KLEE_INCOMPLETESOLVER_H
+
+#include "klee/Solver.h"
+#include "klee/SolverImpl.h"
+
+namespace klee {
+
+/// IncompleteSolver - Base class for incomplete solver
+/// implementations.
+///
+/// Incomplete solvers are useful for implementing optimizations which
+/// may quickly compute an answer, but cannot always compute the
+/// correct answer. They can be used with the StagedSolver to provide
+/// a complete Solver implementation.
+class IncompleteSolver {
+public:
+  /// PartialValidity - Represent a possibility incomplete query
+  /// validity.
+  enum PartialValidity {
+    /// The query is provably true.
+    MustBeTrue = 1,
+
+    /// The query is provably false.
+    MustBeFalse = -1,
+
+    /// The query is not provably false (a true assignment is known to
+    /// exist).
+    MayBeTrue = 2,
+
+    /// The query is not provably true (a false assignment is known to
+    /// exist).
+    MayBeFalse = -2,
+
+    /// The query is known to have both true and false assignments.
+    TrueOrFalse = 0,
+
+    /// The validity of the query is unknown.
+    None = 3
+  };
+
+  static PartialValidity negatePartialValidity(PartialValidity pv);
+
+public:
+  IncompleteSolver() {};
+  virtual ~IncompleteSolver() {};
+
+  /// computeValidity - Compute a partial validity for the given query.
+  ///
+  /// The passed expression is non-constant with bool type.
+  ///
+  /// The IncompleteSolver class provides an implementation of
+  /// computeValidity using computeTruth. Sub-classes may override
+  /// this if a more efficient implementation is available.
+  virtual IncompleteSolver::PartialValidity computeValidity(const Query&);
+
+  /// computeValidity - Compute a partial validity for the given query.
+  ///
+  /// The passed expression is non-constant with bool type.
+  virtual IncompleteSolver::PartialValidity computeTruth(const Query&) = 0;
+  
+  /// computeValue - Attempt to compute a value for the given expression.
+  virtual bool computeValue(const Query&, ref<Expr> &result) = 0;
+
+  /// computeInitialValues - Attempt to compute the constant values
+  /// for the initial state of each given object. If a correct result
+  /// is not found, then the values array must be unmodified.
+  virtual bool computeInitialValues(const Query&,
+                                    const std::vector<const Array*> 
+                                      &objects,
+                                    std::vector< std::vector<unsigned char> > 
+                                      &values,
+                                    bool &hasSolution) = 0;
+};
+
+/// StagedSolver - Adapter class for staging an incomplete solver with
+/// a complete secondary solver, to form an (optimized) complete
+/// solver.
+class StagedSolverImpl : public SolverImpl {
+private:
+  IncompleteSolver *primary;
+  Solver *secondary;
+  
+public:
+  StagedSolverImpl(IncompleteSolver *_primary, Solver *_secondary);
+  ~StagedSolverImpl();
+    
+  bool computeTruth(const Query&, bool &isValid);
+  bool computeValidity(const Query&, Solver::Validity &result);
+  bool computeValue(const Query&, ref<Expr> &result);
+  bool computeInitialValues(const Query&,
+                            const std::vector<const Array*> &objects,
+                            std::vector< std::vector<unsigned char> > &values,
+                            bool &hasSolution);
+};
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/BOut.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/BOut.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/BOut.h (added)
+++ klee/trunk/include/klee/Internal/ADT/BOut.h Wed May 20 23:36:41 2009
@@ -0,0 +1,62 @@
+//===-- BOut.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __COMMON_BOUT_H__
+#define __COMMON_BOUT_H__
+
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+  typedef struct BOutObject BOutObject;
+  struct BOutObject {
+    char *name;
+    unsigned numBytes;
+    unsigned char *bytes;
+  };
+  
+  typedef struct BOut BOut;
+  struct BOut {
+    /* file format version */
+    unsigned version; 
+    
+    unsigned numArgs;
+    char **args;
+
+    unsigned symArgvs;
+    unsigned symArgvLen;
+
+    unsigned numObjects;
+    BOutObject *objects;
+  };
+
+  
+  /* returns the current .bout file format version */
+  unsigned bOut_getCurrentVersion();
+  
+  /* return true iff file at path matches BOut header */
+  int   bOut_isBOutFile(const char *path);
+
+  /* returns NULL on (unspecified) error */
+  BOut* bOut_fromFile(const char *path);
+
+  /* returns 1 on success, 0 on (unspecified) error */
+  int   bOut_toFile(BOut *, const char *path);
+  
+  /* returns total number of object bytes */
+  unsigned bOut_numBytes(BOut *);
+
+  void  bOut_free(BOut *);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/DiscretePDF.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/DiscretePDF.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/DiscretePDF.h (added)
+++ klee/trunk/include/klee/Internal/ADT/DiscretePDF.h Wed May 20 23:36:41 2009
@@ -0,0 +1,47 @@
+//===-- DiscretePDF.h -------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+namespace klee {
+  template <class T>
+  class DiscretePDF {
+    // not perfectly parameterized, but float/double/int should work ok,
+    // although it would be better to have choose argument range from 0
+    // to queryable max.
+    typedef double weight_type;
+
+  public:
+    DiscretePDF();
+    ~DiscretePDF();
+
+    bool empty() const;
+    void insert(T item, weight_type weight);
+    void update(T item, weight_type newWeight);
+    void remove(T item);
+    bool inTree(T item);
+    weight_type getWeight(T item);
+	
+    /* pick a tree element according to its
+     * weight. p should be in [0,1).
+     */
+    T choose(double p);
+    
+  private:
+    class Node;
+    Node *m_root;
+    
+    Node **lookup(T item, Node **parent_out);
+    void split(Node *node);
+    void rotate(Node *node);
+    void lengthen(Node *node);
+    void propogateSumsUp(Node *n);
+  };
+
+}
+
+#include "DiscretePDF.inc"

Added: klee/trunk/include/klee/Internal/ADT/DiscretePDF.inc
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/DiscretePDF.inc?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/DiscretePDF.inc (added)
+++ klee/trunk/include/klee/Internal/ADT/DiscretePDF.inc Wed May 20 23:36:41 2009
@@ -0,0 +1,342 @@
+//===- DiscretePDF.inc - --*- C++ -*-===//
+
+//
+
+namespace klee {
+
+template <class T>
+class DiscretePDF<T>::Node
+{
+private:
+  bool m_mark;
+
+public:
+  Node *parent, *left, *right;
+  T key;
+  weight_type weight, sumWeights;
+
+public:
+  Node(T key_, weight_type weight_, Node *parent_);
+  ~Node();
+
+  Node *sibling() { return this==parent->left?parent->right:parent->left; }
+
+  void markRed() { m_mark = true; }
+  void markBlack() { m_mark = false; }
+  bool isBlack() { return !m_mark; }
+  bool leftIsBlack() { return !left || left->isBlack(); }
+  bool rightIsBlack() { return !right || right->isBlack(); }
+  void setSum() { 
+    sumWeights = weight;
+    if (left) sumWeights += left->sumWeights;
+    if (right) sumWeights += right->sumWeights;
+  }
+};
+
+  ///
+
+template <class T>
+DiscretePDF<T>::Node::Node(T key_, weight_type weight_, Node *parent_) {
+  m_mark = false;
+
+  key = key_;
+  weight = weight_;
+  sumWeights = 0;
+  left = right = 0;
+  parent = parent_;
+}
+
+template <class T>
+DiscretePDF<T>::Node::~Node() {
+  if (left) delete left;
+  if (right) delete right;
+}
+
+//
+
+template <class T>
+DiscretePDF<T>::DiscretePDF() {
+  m_root = 0;
+}
+
+template <class T>
+DiscretePDF<T>::~DiscretePDF() {
+  if (m_root) delete m_root;
+}
+
+template <class T>
+bool DiscretePDF<T>::empty() const {
+  return m_root == 0;
+}
+
+template <class T>
+void DiscretePDF<T>::insert(T item, weight_type weight) {
+  Node *p=0, *n=m_root;
+
+  while (n) {
+    if (!n->leftIsBlack() && !n->rightIsBlack())
+      split(n);
+
+    p = n;
+    if (n->key==item) {
+      assert(0 && "insert: argument(item) already in tree");
+    } else {
+      n = (item<n->key)?n->left:n->right;
+    }
+  }
+
+  n = new Node(item, weight, p);
+
+  if (!p) {
+    m_root = n;
+  } else {
+    if (item<p->key) {
+      p->left = n;
+    } else {
+      p->right = n;
+    }
+
+    split(n);
+  }
+
+  propogateSumsUp(n);
+}
+
+template <class T>
+void DiscretePDF<T>::remove(T item) {
+  Node **np = lookup(item, 0);
+  Node *child, *n = *np;
+
+  if (!n) {
+    assert(0 && "remove: argument(item) not in tree");
+  } else {
+    if (n->left) {
+      Node **leftMaxp = &n->left;
+
+      while ((*leftMaxp)->right)
+        leftMaxp = &(*leftMaxp)->right;
+
+      n->key = (*leftMaxp)->key;
+      n->weight = (*leftMaxp)->weight;
+
+      np = leftMaxp;
+      n = *np;
+    }
+
+    // node now has at most one child
+
+    child = n->left?n->left:n->right;
+    *np = child;
+
+    if (child) {
+      child->parent = n->parent;
+
+      if (n->isBlack()) {
+        lengthen(child);
+      }
+    }
+
+    propogateSumsUp(n->parent);
+
+    n->left = n->right = 0;
+    delete n;
+  }
+}
+
+template <class T>
+void DiscretePDF<T>::update(T item, weight_type weight) {
+  Node *n = *lookup(item, 0);
+
+  if (!n) {
+    assert(0 && "update: argument(item) not in tree");
+  } else {
+    n->weight = weight;
+    propogateSumsUp(n);
+  }
+}
+
+template <class T>
+T DiscretePDF<T>::choose(double p) {
+  if (p<0.0 || p>=1.0) {
+    assert(0 && "choose: argument(p) outside valid range");
+  } else if (!m_root) {
+    assert(0 && "choose: choose() called on empty tree");
+  } else {
+    weight_type w = (weight_type) (m_root->sumWeights * p);
+    Node *n = m_root;
+
+    while (1) {
+      if (n->left) {
+        if (w<n->left->sumWeights) {
+          n = n->left;
+          continue;
+        } else {
+          w -= n->left->sumWeights;
+        }
+      }
+      if (w<n->weight || !n->right) {
+        break; // !n->right condition shouldn't be necessary, just sanity check
+      }
+      w -= n->weight;
+      n = n->right;
+    }
+
+    return n->key;
+  }
+}
+
+template <class T>
+bool DiscretePDF<T>::inTree(T item) {
+  Node *n = *lookup(item, 0);
+
+  return !!n;
+}
+
+template <class T>
+typename DiscretePDF<T>::weight_type DiscretePDF<T>::getWeight(T item) {
+  Node *n = *lookup(item, 0);
+  assert(n);
+  return n->weight;
+}
+
+//
+
+template <class T>
+typename DiscretePDF<T>::Node **
+DiscretePDF<T>::lookup(T item, Node **parent_out) {
+  Node *n, *p=0, **np=&m_root;
+
+  while ((n = *np)) {
+    if (n->key==item) {
+      break;
+    } else {
+      p = n;
+      if (item<n->key) {
+        np = &n->left;
+      } else {
+        np = &n->right;
+      }
+    }
+  }
+
+  if (parent_out)
+    *parent_out = p;
+  return np;
+}
+
+template <class T>
+void DiscretePDF<T>::split(Node *n) {
+  if (n->left) n->left->markBlack();
+  if (n->right) n->right->markBlack();
+
+  if (n->parent) {
+    Node *p = n->parent;
+
+    n->markRed();
+
+    if (!p->isBlack()) {
+      p->parent->markRed();
+
+      // not same direction
+      if (!((n==p->left && p==p->parent->left) || 
+            (n==p->right && p==p->parent->right))) {
+        rotate(n);
+        p = n;
+      }
+
+      rotate(p);
+      p->markBlack();
+    }
+  }
+}
+
+template <class T>
+void DiscretePDF<T>::rotate(Node *n) {
+  Node *p=n->parent, *pp=p->parent;
+
+  n->parent = pp;
+  p->parent = n;
+
+  if (n==p->left) {
+    p->left = n->right;
+    n->right = p;
+    if (p->left) p->left->parent = p;
+  } else {
+    p->right = n->left;
+    n->left = p;
+    if (p->right) p->right->parent = p;
+  }
+
+  n->setSum();
+  p->setSum();
+	
+  if (!pp) {
+    m_root = n;
+  } else {
+    if (p==pp->left) {
+      pp->left = n;
+    } else {
+      pp->right = n;
+    }
+  }
+}
+
+template <class T>
+void DiscretePDF<T>::lengthen(Node *n) {
+  if (!n->isBlack()) {
+    n->markBlack();
+  } else if (n->parent) {
+    Node *sibling = n->sibling();
+
+    if (sibling && !sibling->isBlack()) {
+      n->parent->markRed();
+      sibling->markBlack();
+
+      rotate(sibling); // node sibling is now old sibling child, must be black
+      sibling = n->sibling();
+    }
+
+    // sibling is black
+
+    if (!sibling) {
+      lengthen(n->parent);
+    } else if (sibling->leftIsBlack() && sibling->rightIsBlack()) {
+      if (n->parent->isBlack()) {
+        sibling->markRed();
+        lengthen(n->parent);
+      } else {
+        sibling->markRed();
+        n->parent->markBlack();
+      }
+    } else {
+      if (n==n->parent->left && sibling->rightIsBlack()) {
+        rotate(sibling->left); // sibling->left must be red
+        sibling->markRed();
+        sibling->parent->markBlack();
+        sibling = sibling->parent;
+      } else if (n==n->parent->right && sibling->leftIsBlack()) {
+        rotate(sibling->right); // sibling->right must be red
+        sibling->markRed();
+        sibling->parent->markBlack();
+        sibling = sibling->parent;
+      }
+
+      // sibling is black, and sibling's far child is red
+
+      rotate(sibling);
+      if (!n->parent->isBlack()) 
+        sibling->markRed();
+      sibling->left->markBlack();
+      sibling->right->markBlack();
+    }
+  }
+}
+
+template <class T>
+void DiscretePDF<T>::propogateSumsUp(Node *n) {
+  for (; n; n=n->parent)
+    n->setSum();
+}
+
+}
+

Added: klee/trunk/include/klee/Internal/ADT/ImmutableMap.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/ImmutableMap.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/ImmutableMap.h (added)
+++ klee/trunk/include/klee/Internal/ADT/ImmutableMap.h Wed May 20 23:36:41 2009
@@ -0,0 +1,104 @@
+//===-- ImmutableMap.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_IMMUTABLEMAP_H__
+#define __UTIL_IMMUTABLEMAP_H__
+
+#include <functional>
+
+#include "ImmutableTree.h"
+
+namespace klee {
+  template<class V, class D>
+  struct _Select1st {
+    D &operator()(V &a) const { return a.first; }
+    const D &operator()(const V &a) const { return a.first; }
+  };
+  
+  template<class K, class D, class CMP=std::less<K> >
+  class ImmutableMap {
+  public:
+    typedef K key_type;
+    typedef std::pair<K,D> value_type;
+
+    typedef ImmutableTree<K, value_type, _Select1st<value_type,key_type>, CMP> Tree;
+    typedef typename Tree::iterator iterator;
+
+  private:
+    Tree elts;
+
+    ImmutableMap(const Tree &b): elts(b) {}
+
+  public:
+    ImmutableMap() {}
+    ImmutableMap(const ImmutableMap &b) : elts(b.elts) {}
+    ~ImmutableMap() {}
+
+    ImmutableMap &operator=(const ImmutableMap &b) { elts = b.elts; return *this; }
+    
+    bool empty() const { 
+      return elts.empty(); 
+    }
+    unsigned count(const key_type &key) const { 
+      return elts.count(key); 
+    }
+    const value_type *lookup(const key_type &key) const { 
+      return elts.lookup(key); 
+    }
+    const value_type *lookup_previous(const key_type &key) const { 
+      return elts.lookup_previous(key); 
+    }
+    const value_type &min() const { 
+      return elts.min(); 
+    }
+    const value_type &max() const { 
+      return elts.max(); 
+    }
+    unsigned size() const { 
+      return elts.size(); 
+    }
+
+    ImmutableMap insert(const value_type &value) const { 
+      return elts.insert(value); 
+    }
+    ImmutableMap replace(const value_type &value) const { 
+      return elts.replace(value); 
+    }
+    ImmutableMap remove(const key_type &key) const { 
+      return elts.remove(key); 
+    }
+    ImmutableMap popMin(const value_type &valueOut) const { 
+      return elts.popMin(valueOut); 
+    }
+    ImmutableMap popMax(const value_type &valueOut) const { 
+      return elts.popMax(valueOut); 
+    }
+
+    iterator begin() const { 
+      return elts.begin(); 
+    }
+    iterator end() const { 
+      return elts.end(); 
+    }
+    iterator find(const key_type &key) const { 
+      return elts.find(key); 
+    }
+    iterator lower_bound(const key_type &key) const { 
+      return elts.lower_bound(key); 
+    }
+    iterator upper_bound(const key_type &key) const { 
+      return elts.upper_bound(key); 
+    }
+
+    static unsigned getAllocated() { return Tree::allocated; }
+  };
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/ImmutableSet.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/ImmutableSet.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/ImmutableSet.h (added)
+++ klee/trunk/include/klee/Internal/ADT/ImmutableSet.h Wed May 20 23:36:41 2009
@@ -0,0 +1,101 @@
+//===-- ImmutableSet.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_IMMUTABLESET_H__
+#define __UTIL_IMMUTABLESET_H__
+
+#include <functional>
+
+#include "ImmutableTree.h"
+
+namespace klee {
+  template<class T>
+  struct _Identity {
+    T &operator()(T &a) const { return a; }
+    const T &operator()(const T &a) const { return a; }
+  };
+  
+  template<class T, class CMP=std::less<T> >
+  class ImmutableSet {
+  public:
+    typedef T key_type;
+    typedef T value_type;
+
+    typedef ImmutableTree<T, T, _Identity<T>, CMP> Tree;
+    typedef typename Tree::iterator iterator;
+
+  private:
+    Tree elts;
+
+    ImmutableSet(const Tree &b): elts(b) {}
+
+  public:
+    ImmutableSet() {}
+    ImmutableSet(const ImmutableSet &b) : elts(b.elts) {}
+    ~ImmutableSet() {}
+
+    ImmutableSet &operator=(const ImmutableSet &b) { elts = b.elts; return *this; }
+    
+    bool empty() const { 
+      return elts.empty(); 
+    }
+    unsigned count(const key_type &key) const { 
+      return elts.count(key); 
+    }
+    const value_type *lookup(const key_type &key) const { 
+      return elts.lookup(key); 
+    }
+    const value_type &min() const { 
+      return elts.min(); 
+    }
+    const value_type &max() const { 
+      return elts.max(); 
+    }
+    unsigned size() { 
+      return elts.size(); 
+    }
+
+    ImmutableSet insert(const value_type &value) const {
+      return elts.insert(value); 
+    }
+    ImmutableSet replace(const value_type &value) const {
+      return elts.replace(value); 
+    }
+    ImmutableSet remove(const key_type &key) const { 
+      return elts.remove(key); 
+    }
+    ImmutableSet popMin(const value_type &valueOut) const { 
+      return elts.popMin(valueOut); 
+    }
+    ImmutableSet popMax(const value_type &valueOut) const { 
+      return elts.popMax(valueOut); 
+    }
+
+    iterator begin() const { 
+      return elts.begin(); 
+    }
+    iterator end() const { 
+      return elts.end(); 
+    }
+    iterator find(const key_type &key) const { 
+      return elts.find(key); 
+    }
+    iterator lower_bound(const key_type &key) const { 
+      return elts.lower_bound(key); 
+    }
+    iterator upper_bound(const key_type &key) const { 
+      return elts.upper_bound(key); 
+    }
+
+    static unsigned getAllocated() { return Tree::allocated; }
+  };
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/ImmutableTree.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/ImmutableTree.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/ImmutableTree.h (added)
+++ klee/trunk/include/klee/Internal/ADT/ImmutableTree.h Wed May 20 23:36:41 2009
@@ -0,0 +1,619 @@
+//===-- ImmutableTree.h -----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_IMMUTABLETREE_H__
+#define __UTIL_IMMUTABLETREE_H__
+
+#include <cassert>
+#include <vector>
+
+namespace klee {
+  template<class K, class V, class KOV, class CMP>
+  class ImmutableTree {
+  public:
+    static unsigned allocated;
+    class iterator;
+
+    typedef K key_type;
+    typedef V value_type;
+    typedef KOV key_of_value;
+    typedef CMP key_compare;
+
+  public:
+    ImmutableTree();
+    ImmutableTree(const ImmutableTree &s);
+    ~ImmutableTree();
+
+    ImmutableTree &operator=(const ImmutableTree &s);
+
+    bool empty() const;
+
+    unsigned count(const key_type &key) const; // always 0 or 1
+    const value_type *lookup(const key_type &key) const;
+
+    // find the last value less than or equal to key, or null if
+    // no such value exists
+    const value_type *lookup_previous(const key_type &key) const;
+
+    const value_type &min() const;
+    const value_type &max() const;
+    unsigned size() const;
+
+    ImmutableTree insert(const value_type &value) const;
+    ImmutableTree replace(const value_type &value) const;
+    ImmutableTree remove(const key_type &key) const;
+    ImmutableTree popMin(value_type &valueOut) const;
+    ImmutableTree popMax(value_type &valueOut) const;
+
+    iterator begin() const;
+    iterator end() const;
+    iterator find(const key_type &key) const;
+    iterator lower_bound(const key_type &key) const;
+    iterator upper_bound(const key_type &key) const;
+
+    static unsigned getAllocated() { return allocated; }
+
+  private:
+    class Node;
+
+    Node *node;
+    ImmutableTree(Node *_node);
+  };
+
+  /***/
+
+  template<class K, class V, class KOV, class CMP>
+  class ImmutableTree<K,V,KOV,CMP>::Node {
+  public:
+    static Node terminator;
+    Node *left, *right;
+    value_type value;
+    unsigned height, references;
+
+  protected:
+    Node(); // solely for creating the terminator node
+    static Node *balance(Node *left, const value_type &value, Node *right);
+
+  public:
+
+    Node(Node *_left, Node *_right, const value_type &_value);
+    ~Node();
+
+    void decref();
+    Node *incref();
+
+    bool isTerminator();
+
+    unsigned size();
+    Node *popMin(value_type &valueOut);
+    Node *popMax(value_type &valueOut);
+    Node *insert(const value_type &v);
+    Node *replace(const value_type &v);
+    Node *remove(const key_type &k);
+  };
+
+  // Should live somewhere else, this is a simple stack with maximum size.
+  template<typename T>
+  class FixedStack {
+    unsigned pos, max;
+    T *elts;
+
+  public:
+    FixedStack(unsigned _max) : pos(0),
+                                max(_max),
+                                elts(new T[max]) {}
+    FixedStack(const FixedStack &b) : pos(b.pos),
+                                      max(b.max),
+                                      elts(new T[b.max]) {
+      std::copy(b.elts, b.elts+pos, elts);
+    }
+    ~FixedStack() { delete[] elts; }
+
+    void push_back(const T &elt) { elts[pos++] = elt; }
+    void pop_back() { --pos; }
+    bool empty() { return pos==0; }
+    T &back() { return elts[pos-1]; }
+
+
+    FixedStack &operator=(const FixedStack &b) {
+      assert(max == b.max); 
+      pos = b.pos;
+      std::copy(b.elts, b.elts+pos, elts);
+      return *this;
+    }
+
+    bool operator==(const FixedStack &b) {
+      return (pos == b.pos &&
+              std::equal(elts, elts+pos, b.elts));
+    }
+    bool operator!=(const FixedStack &b) { return !(*this==b); }
+  };
+
+  template<class K, class V, class KOV, class CMP>
+  class ImmutableTree<K,V,KOV,CMP>::iterator {
+    friend class ImmutableTree<K,V,KOV,CMP>;
+  private:
+    Node *root; // so can back up from end
+    FixedStack<Node*> stack;
+    
+  public:
+    iterator(Node *_root, bool atBeginning) : root(_root->incref()),
+                                              stack(root->height) {
+      if (atBeginning) {
+        for (Node *n=root; !n->isTerminator(); n=n->left)
+          stack.push_back(n);
+      }
+    }
+    iterator(const iterator &i) : root(i.root->incref()),
+                                  stack(i.stack) {
+    }
+    ~iterator() {
+      root->decref();
+    }
+
+    iterator &operator=(const iterator &b) {
+      b.root->incref();
+      root->decref();
+      root = b.root;
+      stack = b.stack;
+      return *this;
+    }
+
+    const value_type &operator*() {
+      Node *n = stack.back();
+      return n->value;
+    }
+
+    const value_type *operator->() {
+      Node *n = stack.back();
+      return &n->value;
+    }
+
+    bool operator==(const iterator &b) {
+      return stack==b.stack;
+    }
+    bool operator!=(const iterator &b) {
+      return stack!=b.stack;
+    }
+    
+    iterator &operator--() {
+      if (stack.empty()) {
+        for (Node *n=root; !n->isTerminator(); n=n->right)
+          stack.push_back(n);
+      } else {
+        Node *n = stack.back();
+        if (n->left->isTerminator()) {
+          for (;;) {
+            Node *prev = n;
+            stack.pop_back();
+            if (stack.empty()) {
+              break;
+            } else {
+              n = stack.back();
+              if (prev==n->right)
+                break;
+            }
+          }
+        } else {
+          stack.push_back(n->left);
+          for (n=n->left->right; !n->isTerminator(); n=n->right)
+            stack.push_back(n);
+        }
+      }
+      return *this;
+    }
+
+    iterator &operator++() {
+      assert(!stack.empty());
+      Node *n = stack.back();
+      if (n->right->isTerminator()) {
+        for (;;) {
+          Node *prev = n;
+          stack.pop_back();
+          if (stack.empty()) {
+            break;
+          } else {
+            n = stack.back();
+            if (prev==n->left)
+              break;
+          }
+        }
+      } else {
+        stack.push_back(n->right);
+        for (n=n->right->left; !n->isTerminator(); n=n->left)
+          stack.push_back(n);
+      }
+      return *this;
+    }
+  };
+
+  /***/
+
+  template<class K, class V, class KOV, class CMP> 
+  typename ImmutableTree<K,V,KOV,CMP>::Node 
+  ImmutableTree<K,V,KOV,CMP>::Node::terminator;
+
+  template<class K, class V, class KOV, class CMP> 
+  unsigned ImmutableTree<K,V,KOV,CMP>::allocated = 0;
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::Node::Node() 
+    : left(&terminator), 
+      right(&terminator), 
+      height(0), 
+      references(3) { 
+    assert(this==&terminator);
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::Node::Node(Node *_left, Node *_right, const value_type &_value)
+    : left(_left), 
+      right(_right), 
+      value(_value), 
+      height(std::max(left->height, right->height) + 1),
+      references(1) 
+  {
+    ++allocated;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::Node::~Node() {
+    left->decref();
+    right->decref();
+    --allocated;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline void ImmutableTree<K,V,KOV,CMP>::Node::decref() {
+    --references;
+    if (references==0) delete this;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline typename ImmutableTree<K,V,KOV,CMP>::Node *ImmutableTree<K,V,KOV,CMP>::Node::incref() {
+    ++references;
+    return this;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline bool ImmutableTree<K,V,KOV,CMP>::Node::isTerminator() {
+    return this==&terminator;
+  }
+
+  /***/
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::balance(Node *left, const value_type &value, Node *right) {
+    if (left->height > right->height + 2) {
+      Node *ll = left->left;
+      Node *lr = left->right;
+      if (ll->height >= lr->height) {
+        Node *nlr = new Node(lr->incref(), right, value);
+        Node *res = new Node(ll->incref(), nlr, left->value);
+        left->decref();
+        return res;
+      } else {
+        Node *lrl = lr->left;
+        Node *lrr = lr->right;
+        Node *nll = new Node(ll->incref(), lrl->incref(), left->value);
+        Node *nlr = new Node(lrr->incref(), right, value);
+        Node *res = new Node(nll, nlr, lr->value);
+        left->decref();
+        return res;
+      }
+    } else if (right->height > left->height + 2) {
+      Node *rl = right->left;
+      Node *rr = right->right;
+      if (rr->height >= rl->height) {
+        Node *nrl = new Node(left, rl->incref(), value);
+        Node *res = new Node(nrl, rr->incref(), right->value);
+        right->decref();
+        return res;
+      } else {
+        Node *rll = rl->left;
+        Node *rlr = rl->right;
+        Node *nrl = new Node(left, rll->incref(), value);
+        Node *nrr = new Node(rlr->incref(), rr->incref(), right->value);
+        Node *res = new Node(nrl, nrr, rl->value);
+        right->decref();
+        return res;
+      }
+    } else {
+      return new Node(left, right, value);
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  unsigned ImmutableTree<K,V,KOV,CMP>::Node::size() {
+    if (isTerminator()) {
+      return 0;
+    } else {
+      return left->size() + 1 + right->size();
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::popMin(value_type &valueOut) {
+    if (left->isTerminator()) {
+      valueOut = value;
+      return right->incref();
+    } else {
+      return balance(left->popMin(valueOut), value, right->incref());
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::popMax(value_type &valueOut) {
+    if (right->isTerminator()) {
+      valueOut = value;
+      return left->incref();
+    } else {
+      return balance(left->incref(), value, right->popMax(valueOut));
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::insert(const value_type &v) {
+    if (isTerminator()) {
+      return new Node(terminator.incref(), terminator.incref(), v);
+    } else {
+      if (key_compare()(key_of_value()(v), key_of_value()(value))) {
+        return balance(left->insert(v), value, right->incref());
+      } else if (key_compare()(key_of_value()(value), key_of_value()(v))) {
+        return balance(left->incref(), value, right->insert(v));
+      } else {
+        return incref();
+      }
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::replace(const value_type &v) {
+    if (isTerminator()) {
+      return new Node(terminator.incref(), terminator.incref(), v);
+    } else {
+      if (key_compare()(key_of_value()(v), key_of_value()(value))) {
+        return balance(left->replace(v), value, right->incref());
+      } else if (key_compare()(key_of_value()(value), key_of_value()(v))) {
+        return balance(left->incref(), value, right->replace(v));
+      } else {
+        return new Node(left->incref(), right->incref(), v);
+      }
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::Node *
+  ImmutableTree<K,V,KOV,CMP>::Node::remove(const key_type &k) {
+    if (isTerminator()) {
+      return incref();
+    } else {
+      if (key_compare()(k, key_of_value()(value))) {
+        return balance(left->remove(k), value, right->incref());
+      } else if (key_compare()(key_of_value()(value), k)) {
+        return balance(left->incref(), value, right->remove(k));
+      } else {
+        if (left->isTerminator()) {
+          return right->incref();
+        } else if (right->isTerminator()) {
+          return left->incref();
+        } else {
+          value_type min;
+          Node *nr = right->popMin(min);
+          return balance(left->incref(), min, nr);
+        }
+      }
+    }
+  }
+
+  /***/
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::ImmutableTree() 
+    : node(Node::terminator.incref()) {
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::ImmutableTree(Node *_node) 
+    : node(_node) {
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::ImmutableTree(const ImmutableTree &s) 
+    : node(s.node->incref()) {
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP>::~ImmutableTree() {
+    node->decref(); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> &ImmutableTree<K,V,KOV,CMP>::operator=(const ImmutableTree &s) {
+    Node *n = s.node->incref();
+    node->decref();
+    node = n;
+    return *this;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  bool ImmutableTree<K,V,KOV,CMP>::empty() const {
+    return node->isTerminator();
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  unsigned ImmutableTree<K,V,KOV,CMP>::count(const key_type &k) const {
+    Node *n = node;
+    while (!n->isTerminator()) {
+      key_type key = key_of_value()(n->value);
+      if (key_compare()(k, key)) {
+        n = n->left;
+      } else if (key_compare()(key, k)) {
+        n = n->right;
+      } else {
+        return 1;
+      }
+    }
+    return 0;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  const typename ImmutableTree<K,V,KOV,CMP>::value_type *
+  ImmutableTree<K,V,KOV,CMP>::lookup(const key_type &k) const {
+    Node *n = node;
+    while (!n->isTerminator()) {
+      key_type key = key_of_value()(n->value);
+      if (key_compare()(k, key)) {
+        n = n->left;
+      } else if (key_compare()(key, k)) {
+        n = n->right;
+      } else {
+        return &n->value;
+      }
+    }
+    return 0;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  const typename ImmutableTree<K,V,KOV,CMP>::value_type *
+  ImmutableTree<K,V,KOV,CMP>::lookup_previous(const key_type &k) const {
+    Node *n = node;
+    Node *result = 0;
+    while (!n->isTerminator()) {
+      key_type key = key_of_value()(n->value);
+      if (key_compare()(k, key)) {
+        n = n->left;
+      } else if (key_compare()(key, k)) {
+        result = n;
+        n = n->right;
+      } else {
+        return &n->value;
+      }
+    }
+    return result ? &result->value : 0;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  const typename ImmutableTree<K,V,KOV,CMP>::value_type &
+  ImmutableTree<K,V,KOV,CMP>::min() const { 
+    Node *n = node;
+    assert(!n->isTerminator());
+    while (!n->left->isTerminator()) n = n->left;
+    return n->value;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  const typename ImmutableTree<K,V,KOV,CMP>::value_type &
+  ImmutableTree<K,V,KOV,CMP>::max() const {
+    Node *n = node;
+    assert(!n->isTerminator());
+    while (!n->right->isTerminator()) n = n->right;
+    return n->value;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  unsigned ImmutableTree<K,V,KOV,CMP>::size() const {
+    return node->size();
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> 
+  ImmutableTree<K,V,KOV,CMP>::insert(const value_type &value) const { 
+    return ImmutableTree(node->insert(value)); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> 
+  ImmutableTree<K,V,KOV,CMP>::replace(const value_type &value) const { 
+    return ImmutableTree(node->replace(value)); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> 
+  ImmutableTree<K,V,KOV,CMP>::remove(const key_type &key) const { 
+    return ImmutableTree(node->remove(key)); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> 
+  ImmutableTree<K,V,KOV,CMP>::popMin(value_type &valueOut) const { 
+    return ImmutableTree(node->popMin(valueOut)); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  ImmutableTree<K,V,KOV,CMP> 
+  ImmutableTree<K,V,KOV,CMP>::popMax(value_type &valueOut) const { 
+    return ImmutableTree(node->popMax(valueOut)); 
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline typename ImmutableTree<K,V,KOV,CMP>::iterator 
+  ImmutableTree<K,V,KOV,CMP>::begin() const {
+    return iterator(node, true);
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline typename ImmutableTree<K,V,KOV,CMP>::iterator 
+  ImmutableTree<K,V,KOV,CMP>::end() const {
+    return iterator(node, false);
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline typename ImmutableTree<K,V,KOV,CMP>::iterator 
+  ImmutableTree<K,V,KOV,CMP>::find(const key_type &key) const {
+    iterator end(node,false), it = lower_bound(key);
+    if (it==end || key_compare()(key,key_of_value()(*it))) {
+      return end;
+    } else {
+      return it;
+    }
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  inline typename ImmutableTree<K,V,KOV,CMP>::iterator 
+  ImmutableTree<K,V,KOV,CMP>::lower_bound(const key_type &k) const {
+    // XXX ugh this doesn't have to be so ugly does it?
+    iterator it(node,false);
+    for (Node *root=node; !root->isTerminator();) {
+      it.stack.push_back(root);
+      if (key_compare()(k, key_of_value()(root->value))) {
+        root = root->left;
+      } else if (key_compare()(key_of_value()(root->value), k)) {
+        root = root->right;
+      } else {
+        return it;
+      }
+    }
+    // it is now beginning or first element < k
+    if (!it.stack.empty()) {
+      Node *last = it.stack.back();
+      if (key_compare()(key_of_value()(last->value), k))
+        ++it;
+    }
+    return it;
+  }
+
+  template<class K, class V, class KOV, class CMP>
+  typename ImmutableTree<K,V,KOV,CMP>::iterator 
+  ImmutableTree<K,V,KOV,CMP>::upper_bound(const key_type &key) const {
+    iterator end(node,false),it = lower_bound(key);
+    if (it!=end && 
+        !key_compare()(key,key_of_value()(*it))) // no need to loop, no duplicates
+      ++it;
+    return it;
+  }
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/MapOfSets.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/MapOfSets.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/MapOfSets.h (added)
+++ klee/trunk/include/klee/Internal/ADT/MapOfSets.h Wed May 20 23:36:41 2009
@@ -0,0 +1,385 @@
+//===-- MapOfSets.h ---------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_MAPOFSETS_H__
+#define __UTIL_MAPOFSETS_H__
+
+#include <cassert>
+#include <vector>
+#include <set>
+#include <map>
+
+// This should really be broken down into TreeOfSets on top of which
+// SetOfSets and MapOfSets are easily implemeted. It should also be
+// parameterized on the underlying set type. Neither are hard to do,
+// just not worth it at the moment.
+//
+// I also broke some STLish conventions because I was bored.
+
+namespace klee {
+
+  /** This implements the UBTree data structure (see Hoffmann and
+      Koehler, "A New Method to Index and Query Sets", IJCAI 1999) */
+  template<class K, class V>
+  class MapOfSets {
+  public:
+    class iterator;
+
+  public:
+    MapOfSets();
+
+    void clear();
+
+    void insert(const std::set<K> &set, const V &value);
+
+    V *lookup(const std::set<K> &set);
+
+    iterator begin();
+    iterator end();
+
+    void subsets(const std::set<K> &set, 
+                 std::vector< std::pair<std::set<K>, V> > &resultOut);
+    void supersets(const std::set<K> &set, 
+                   std::vector< std::pair<std::set<K>, V> > &resultOut);
+    
+    template<class Predicate>
+    V *findSuperset(const std::set<K> &set, const Predicate &p);
+    template<class Predicate>
+    V *findSubset(const std::set<K> &set, const Predicate &p);
+
+  private:
+    class Node;
+
+    Node root;
+
+    template<class Iterator, class Vector>
+    void findSubsets(Node *n, 
+                     const std::set<K> &accum,
+                     Iterator begin, 
+                     Iterator end,
+                     Vector &resultsOut);
+    template<class Iterator, class Vector>
+    void findSupersets(Node *n, 
+                       const std::set<K> &accum,
+                       Iterator begin, 
+                       Iterator end,
+                       Vector &resultsOut);
+    template<class Predicate>
+    V *findSuperset(Node *n, 
+                    typename std::set<K>::iterator begin, 
+                    typename std::set<K>::iterator end,
+                    const Predicate &p);
+    template<class Predicate>
+    V *findSubset(Node *n, 
+                  typename std::set<K>::iterator begin, 
+                  typename std::set<K>::iterator end,
+                  const Predicate &p);
+  };
+
+  /***/
+
+  template<class K, class V>
+  class MapOfSets<K,V>::Node {
+    friend class MapOfSets<K,V>;
+    friend class MapOfSets<K,V>::iterator;
+
+  public:
+    typedef std::map<K, Node> children_ty;
+
+    V value;
+
+  private:
+    bool isEndOfSet;
+    std::map<K, Node> children;
+    
+  public:
+    Node() : isEndOfSet(false) {}
+  };
+  
+  template<class K, class V>
+  class MapOfSets<K,V>::iterator {
+    typedef std::vector< typename std::map<K, Node>::iterator > stack_ty;
+    friend class MapOfSets<K,V>;
+  private:
+    Node *root;
+    bool onEntry;
+    stack_ty stack;
+
+    void step() {
+      if (onEntry) {
+        onEntry = false;
+        Node *n = stack.empty() ? root : &stack.back()->second;
+        while (!n->children.empty()) {
+          stack.push_back(n->children.begin());
+          n = &stack.back()->second;
+          if (n->isEndOfSet) {
+            onEntry = true;
+            return;
+          }
+        } 
+      }
+
+      while (!stack.empty()) {
+        unsigned size = stack.size();
+        Node *at = size==1 ? root : &stack[size-2]->second;
+        typename std::map<K,Node>::iterator &cur = stack.back();
+        ++cur;
+        if (cur==at->children.end()) {
+          stack.pop_back();
+        } else {
+          Node *n = &cur->second;
+
+          while (!n->isEndOfSet) {
+            assert(!n->children.empty());
+            stack.push_back(n->children.begin());
+            n = &stack.back()->second;
+          }
+
+          onEntry = true;
+          break;
+        }
+      } 
+    }
+
+  public:
+    // end()
+    iterator() : onEntry(false) {} 
+    // begin()
+    iterator(Node *_n) : root(_n), onEntry(true) {
+      if (!root->isEndOfSet)
+        step();
+    }
+
+    const std::pair<const std::set<K>, const V> operator*() {
+      assert(onEntry || !stack.empty());
+      std::set<K> s;
+      for (typename stack_ty::iterator it = stack.begin(), ie = stack.end();
+           it != ie; ++it)
+        s.insert((*it)->first);
+      Node *n = stack.empty() ? root : &stack.back()->second;
+      return std::make_pair(s, n->value);
+    }
+
+    bool operator==(const iterator &b) {
+      return onEntry==b.onEntry && stack==b.stack;
+    }
+    bool operator!=(const iterator &b) {
+      return !(*this==b);
+    }
+
+    iterator &operator++() {
+      step();
+      return *this;
+    }
+  };
+
+  /***/
+
+  template<class K, class V>
+  MapOfSets<K,V>::MapOfSets() {}  
+
+  template<class K, class V>
+  void MapOfSets<K,V>::insert(const std::set<K> &set, const V &value) {
+    Node *n = &root;
+    for (typename std::set<K>::const_iterator it = set.begin(), ie = set.end();
+         it != ie; ++it)
+      n = &n->children.insert(std::make_pair(*it, Node())).first->second;
+    n->isEndOfSet = true;
+    n->value = value;
+  }
+
+  template<class K, class V>
+  V *MapOfSets<K,V>::lookup(const std::set<K> &set) {
+    Node *n = &root;
+    for (typename std::set<K>::const_iterator it = set.begin(), ie = set.end();
+         it != ie; ++it) {
+      typename Node::children_ty::iterator kit = n->children.find(*it);
+      if (kit==n->children.end()) {
+        return 0;
+      } else {
+        n = &kit->second;
+      }
+    }
+    if (n->isEndOfSet) {
+      return &n->value;
+    } else {
+      return 0;
+    }
+  }
+
+  template<class K, class V>
+  typename MapOfSets<K,V>::iterator 
+  MapOfSets<K,V>::begin() { return iterator(&root); }
+  
+  template<class K, class V>
+  typename MapOfSets<K,V>::iterator 
+  MapOfSets<K,V>::end() { return iterator(); }
+
+  template<class K, class V>
+  template<class Iterator, class Vector>
+  void MapOfSets<K,V>::findSubsets(Node *n, 
+                                  const std::set<K> &accum,
+                                  Iterator begin, 
+                                  Iterator end,
+                                  Vector &resultsOut) {
+    if (n->isEndOfSet) {
+      resultsOut.push_back(std::make_pair(accum, n->value));
+    }
+    
+    for (Iterator it=begin; it!=end;) {
+      K elt = *it;
+      typename Node::children_ty::iterator kit = n->children.find(elt);
+      it++;
+      if (kit!=n->children.end()) {
+        std::set<K> nacc = accum;
+        nacc.insert(elt);
+        findSubsets(&kit->second, nacc, it, end, resultsOut);
+      }
+    }
+  }
+
+  template<class K, class V>
+  void MapOfSets<K,V>::subsets(const std::set<K> &set,
+                               std::vector< std::pair<std::set<K>, 
+                                                      V> > &resultOut) {
+    findSubsets(&root, std::set<K>(), set.begin(), set.end(), resultOut);
+  }
+
+  template<class K, class V>
+  template<class Iterator, class Vector>
+  void MapOfSets<K,V>::findSupersets(Node *n, 
+                                     const std::set<K> &accum,
+                                     Iterator begin, 
+                                     Iterator end,
+                                     Vector &resultsOut) {
+    if (begin==end) {
+      if (n->isEndOfSet)
+        resultsOut.push_back(std::make_pair(accum, n->value));
+      for (typename Node::children_ty::iterator it = n->children.begin(),
+             ie = n->children.end(); it != ie; ++it) {
+        std::set<K> nacc = accum;
+        nacc.insert(it->first);
+        findSupersets(&it->second, nacc, begin, end, resultsOut);
+      }
+    } else {
+      K elt = *begin;
+      Iterator next = begin;
+      ++next;
+      for (typename Node::children_ty::iterator it = n->children.begin(),
+             ie = n->children.end(); it != ie; ++it) {
+        std::set<K> nacc = accum;
+        nacc.insert(it->first);
+        if (elt==it->first) {
+          findSupersets(&it->second, nacc, next, end, resultsOut);
+        } else if (it->first<elt) {
+          findSupersets(&it->second, nacc, begin, end, resultsOut);
+        } else {
+          break;
+        }
+      }
+    }
+  }
+
+  template<class K, class V>
+  void MapOfSets<K,V>::supersets(const std::set<K> &set,
+                               std::vector< std::pair<std::set<K>, V> > &resultOut) {
+    findSupersets(&root, std::set<K>(), set.begin(), set.end(), resultOut);
+  }
+
+  template<class K, class V>
+  template<class Predicate>
+  V *MapOfSets<K,V>::findSubset(Node *n, 
+                                typename std::set<K>::iterator begin, 
+                                typename std::set<K>::iterator end,
+                                const Predicate &p) {   
+    if (n->isEndOfSet && p(n->value)) {
+      return &n->value;
+    } else if (begin==end) {
+      return 0;
+    } else {
+      typename Node::children_ty::iterator kend = n->children.end();
+      typename Node::children_ty::iterator 
+        kbegin = n->children.lower_bound(*begin);
+      typename std::set<K>::iterator it = begin;
+      if (kbegin==kend)
+        return 0;
+      for (;;) { // kbegin!=kend && *it <= kbegin->first
+        while (*it < kbegin->first) {
+          ++it;
+          if (it==end)
+            return 0;
+        }
+        if (*it == kbegin->first) {
+          ++it;
+          V *res = findSubset(&kbegin->second, it, end, p);
+          if (res || it==end)
+            return res;
+        }
+        while (kbegin->first < *it) {
+          ++kbegin;
+          if (kbegin==kend)
+            return 0;
+        }
+      }
+    }
+  }
+  
+  template<class K, class V>
+  template<class Predicate>
+  V *MapOfSets<K,V>::findSuperset(Node *n, 
+                                  typename std::set<K>::iterator begin, 
+                                  typename std::set<K>::iterator end,
+                                  const Predicate &p) {   
+    if (begin==end) {
+      if (n->isEndOfSet && p(n->value))
+        return &n->value;
+      for (typename Node::children_ty::iterator it = n->children.begin(),
+             ie = n->children.end(); it != ie; ++it) {
+        V *res = findSuperset(&it->second, begin, end, p);
+        if (res) return res;
+      }
+    } else {
+      typename Node::children_ty::iterator kbegin = n->children.begin();
+      typename Node::children_ty::iterator kmid = 
+        n->children.lower_bound(*begin);
+      for (typename Node::children_ty::iterator it = n->children.begin(),
+             ie = n->children.end(); it != ie; ++it) {
+        V *res = findSuperset(&it->second, begin, end, p);
+        if (res) return res;
+      }
+      if (kmid!=n->children.end() && *begin==kmid->first) {
+        V *res = findSuperset(&kmid->second, ++begin, end, p);
+        if (res) return res;
+      }
+    }
+    return 0;
+  }
+
+  template<class K, class V>
+  template<class Predicate>
+  V *MapOfSets<K,V>::findSuperset(const std::set<K> &set, const Predicate &p) {    
+    return findSuperset(&root, set.begin(), set.end(), p);
+  }
+
+  template<class K, class V>
+  template<class Predicate>
+  V *MapOfSets<K,V>::findSubset(const std::set<K> &set, const Predicate &p) {    
+    return findSubset(&root, set.begin(), set.end(), p);
+  }
+
+  template<class K, class V>
+  void MapOfSets<K,V>::clear() {
+    root.isEndOfSet = false;
+    root.value = V();
+    root.children.clear();
+  }
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/RNG.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/RNG.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/RNG.h (added)
+++ klee/trunk/include/klee/Internal/ADT/RNG.h Wed May 20 23:36:41 2009
@@ -0,0 +1,50 @@
+//===-- RNG.h ---------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_RNG_H
+#define KLEE_UTIL_RNG_H
+
+namespace klee {
+  class RNG {
+  private:
+    /* Period parameters */  
+    static const int N = 624;
+    static const int M = 397;
+    static const unsigned int MATRIX_A = 0x9908b0dfUL;   /* constant vector a */
+    static const unsigned int UPPER_MASK = 0x80000000UL; /* most significant w-r bits */
+    static const unsigned int LOWER_MASK = 0x7fffffffUL; /* least significant r bits */
+      
+  private:
+    unsigned int mt[N]; /* the array for the state vector  */
+    int mti;
+    
+  public:
+    RNG(unsigned int seed=5489UL);
+  
+    void seed(unsigned int seed);
+    
+    /* generates a random number on [0,0xffffffff]-interval */
+    unsigned int getInt32();
+    /* generates a random number on [0,0x7fffffff]-interval */
+    int getInt31();
+    /* generates a random number on [0,1]-real-interval */
+    double getDoubleLR();
+    float getFloatLR();
+    /* generates a random number on [0,1)-real-interval */
+    double getDoubleL();
+    float getFloatL();
+    /* generates a random number on (0,1)-real-interval */
+    double getDouble();
+    float getFloat();
+    /* generators a random flop */
+    bool getBool();
+  };
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/ADT/TreeStream.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/ADT/TreeStream.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/ADT/TreeStream.h (added)
+++ klee/trunk/include/klee/Internal/ADT/TreeStream.h Wed May 20 23:36:41 2009
@@ -0,0 +1,77 @@
+//===-- TreeStream.h --------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_TREESTREAM_H__
+#define __UTIL_TREESTREAM_H__
+
+#include <string>
+#include <iostream>
+#include <vector>
+
+namespace klee {
+
+  typedef unsigned TreeStreamID;
+  class TreeOStream;
+
+  class TreeStreamWriter {
+    static const unsigned bufferSize = 4*4096;
+
+    friend class TreeOStream;
+
+  private:
+    char buffer[bufferSize];
+    unsigned lastID, bufferCount;
+
+    std::string path;
+    std::ofstream *output;
+    unsigned ids;
+
+    void write(TreeOStream &os, const char *s, unsigned size);
+    void flushBuffer();
+
+  public:
+    TreeStreamWriter(const std::string &_path);
+    ~TreeStreamWriter();
+
+    bool good();
+
+    TreeOStream open();
+    TreeOStream open(const TreeOStream &node);
+
+    void flush();
+
+    // hack, to be replace by proper stream capabilities
+    void readStream(TreeStreamID id,
+                    std::vector<unsigned char> &out);
+  };
+
+  class TreeOStream {
+    friend class TreeStreamWriter;
+
+  private:
+    TreeStreamWriter *writer;
+    unsigned id;
+    
+    TreeOStream(TreeStreamWriter &_writer, unsigned _id);
+
+  public:
+    TreeOStream();
+    ~TreeOStream();
+
+    unsigned getID() const;
+
+    void write(const char *buffer, unsigned size);
+
+    TreeOStream &operator<<(const std::string &s);
+
+    void flush();
+  };
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/Module/Cell.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Module/Cell.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Module/Cell.h (added)
+++ klee/trunk/include/klee/Internal/Module/Cell.h Wed May 20 23:36:41 2009
@@ -0,0 +1,23 @@
+//===-- Cell.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_CELL_H
+#define KLEE_CELL_H
+
+#include <klee/Expr.h>
+
+namespace klee {
+  class MemoryObject;
+
+  struct Cell {
+    ref<Expr> value;
+  };
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/Module/InstructionInfoTable.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Module/InstructionInfoTable.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Module/InstructionInfoTable.h (added)
+++ klee/trunk/include/klee/Internal/Module/InstructionInfoTable.h Wed May 20 23:36:41 2009
@@ -0,0 +1,70 @@
+//===-- InstructionInfoTable.h ----------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_LIB_INSTRUCTIONINFOTABLE_H
+#define KLEE_LIB_INSTRUCTIONINFOTABLE_H
+
+#include <map>
+#include <string>
+#include <set>
+
+namespace llvm {
+  class Function;
+  class Instruction;
+  class Module; 
+}
+
+namespace klee {
+
+  /* Stores debug information for a KInstruction */
+  struct InstructionInfo {
+    unsigned id;
+    const std::string &file;
+    unsigned line;
+    unsigned assemblyLine;
+
+  public:
+    InstructionInfo(unsigned _id,
+                    const std::string &_file,
+                    unsigned _line,
+                    unsigned _assemblyLine)
+      : id(_id), 
+        file(_file),
+        line(_line),
+        assemblyLine(_assemblyLine) {
+    }
+  };
+
+  class InstructionInfoTable {
+    struct ltstr { 
+      bool operator()(const std::string *a, const std::string *b) const {
+        return *a<*b;
+      }
+    };
+
+    std::string dummyString;
+    InstructionInfo dummyInfo;
+    std::map<const llvm::Instruction*, InstructionInfo> infos;
+    std::set<const std::string *, ltstr> internedStrings;
+
+  private:
+    const std::string *internString(std::string s);
+
+  public:
+    InstructionInfoTable(llvm::Module *m);
+    ~InstructionInfoTable();
+
+    unsigned getMaxID() const;
+    const InstructionInfo &getInfo(const llvm::Instruction*) const;
+    const InstructionInfo &getFunctionInfo(const llvm::Function*) const;
+  };
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/Module/KInstIterator.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Module/KInstIterator.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Module/KInstIterator.h (added)
+++ klee/trunk/include/klee/Internal/Module/KInstIterator.h Wed May 20 23:36:41 2009
@@ -0,0 +1,49 @@
+//===-- KInstIterator.h -----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_KINSTITERATOR_H
+#define KLEE_KINSTITERATOR_H
+
+namespace klee {
+  class KInstruction;
+
+  class KInstIterator {
+    KInstruction **it;
+
+  public:
+    KInstIterator() : it(0) {}
+    KInstIterator(KInstruction **_it) : it(_it) {}
+    KInstIterator(const KInstIterator &b) : it(b.it) {}
+    ~KInstIterator() {}
+
+    KInstIterator &operator=(const KInstIterator &b) {
+      it = b.it;
+      return *this;
+    }
+
+    bool operator==(const KInstIterator &b) const {
+      return it==b.it;
+    }
+    bool operator!=(const KInstIterator &b) const {
+      return !(*this == b);
+    }
+
+    KInstIterator &operator++() {
+      ++it;
+      return *this;
+    }
+
+    operator KInstruction*() const { return it ? *it : 0;}
+    operator bool() const { return it != 0; }
+
+    KInstruction *operator ->() const { return *it; }
+  };
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/Internal/Module/KInstruction.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Module/KInstruction.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Module/KInstruction.h (added)
+++ klee/trunk/include/klee/Internal/Module/KInstruction.h Wed May 20 23:36:41 2009
@@ -0,0 +1,50 @@
+//===-- KInstruction.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_KINSTRUCTION_H
+#define KLEE_KINSTRUCTION_H
+
+#include <vector>
+
+namespace llvm {
+  class Instruction;
+}
+
+namespace klee {
+  class Executor;
+  struct InstructionInfo;
+  class KModule;
+
+
+  /// KInstruction - Intermediate instruction representation used
+  /// during execution.
+  struct KInstruction {
+    llvm::Instruction *inst;    
+    const InstructionInfo *info;
+
+    /// Value numbers for each operand. -1 is an invalid value,
+    /// otherwise negative numbers are indices (negated and offset by
+    /// 2) into the module constant table and positive numbers are
+    /// register indices.
+    int *operands;
+    /// Destination register index.
+    unsigned dest;
+
+  public:
+    virtual ~KInstruction(); 
+  };
+
+  struct KGEPInstruction : KInstruction {
+    std::vector< std::pair<unsigned, unsigned> > indices;
+    unsigned offset;
+  };
+}
+
+#endif
+

Added: klee/trunk/include/klee/Internal/Module/KModule.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Module/KModule.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Module/KModule.h (added)
+++ klee/trunk/include/klee/Internal/Module/KModule.h Wed May 20 23:36:41 2009
@@ -0,0 +1,119 @@
+//===-- KModule.h -----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_KMODULE_H
+#define KLEE_KMODULE_H
+
+#include "klee/Interpreter.h"
+
+#include <map>
+#include <set>
+#include <vector>
+
+namespace llvm {
+  class BasicBlock;
+  class Constant;
+  class Function;
+  class Instruction;
+  class Module;
+  class TargetData;
+}
+
+namespace klee {
+  class Cell;
+  class Executor;
+  class Expr;
+  class InterpreterHandler;
+  class InstructionInfoTable;
+  class KInstruction;
+  class KModule;
+  template<class T> class ref;
+
+  struct KFunction {
+    llvm::Function *function;
+
+    unsigned numArgs, numRegisters;
+
+    unsigned numInstructions;
+    KInstruction **instructions;
+
+    std::map<llvm::BasicBlock*, unsigned> basicBlockEntry;
+
+    /// Whether instructions in this function should count as
+    /// "coverable" for statistics and search heuristics.
+    bool trackCoverage;
+
+  private:
+    KFunction(const KFunction&);
+    KFunction &operator=(const KFunction&);
+
+  public:
+    explicit KFunction(llvm::Function*, KModule *);
+    ~KFunction();
+
+    unsigned getArgRegister(unsigned index) { return index; }
+  };
+
+
+  class KConstant {
+  public:
+    /// Actual LLVM constant this represents.
+    llvm::Constant* ct;
+
+    /// The constant ID.
+    unsigned id;
+
+    /// First instruction where this constant was encountered, or NULL
+    /// if not applicable/unavailable.
+    KInstruction *ki;
+
+    KConstant(llvm::Constant*, unsigned, KInstruction*);
+  };
+
+
+  class KModule {
+  public:
+    llvm::Module *module;
+    llvm::TargetData *targetData;
+    
+    // Some useful functions to know the address of
+    llvm::Function *dbgStopPointFn, *kleeMergeFn;
+
+    // Our shadow versions of LLVM structures.
+    std::vector<KFunction*> functions;
+    std::map<llvm::Function*, KFunction*> functionMap;
+
+    // Functions which escape (may be called indirectly)
+    // XXX change to KFunction
+    std::set<llvm::Function*> escapingFunctions;
+
+    InstructionInfoTable *infos;
+
+    std::vector<llvm::Constant*> constants;
+    std::map<llvm::Constant*, KConstant*> constantMap;
+    KConstant* getKConstant(llvm::Constant *c);
+
+    Cell *constantTable;
+
+  public:
+    KModule(llvm::Module *_module);
+    ~KModule();
+
+    /// Initialize local data structures.
+    //
+    // FIXME: ihandler should not be here
+    void prepare(const Interpreter::ModuleOptions &opts, 
+                 InterpreterHandler *ihandler);
+
+    /// Return an id for the given constant, creating a new one if necessary.
+    unsigned getConstantID(llvm::Constant *c, KInstruction* ki);
+  };
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/Internal/README.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/README.txt?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/README.txt (added)
+++ klee/trunk/include/klee/Internal/README.txt Wed May 20 23:36:41 2009
@@ -0,0 +1,3 @@
+This directory holds header files for things which are exposed as part
+of the internal API of a library, but shouldn't be exposed to
+externally.

Added: klee/trunk/include/klee/Internal/Support/FloatEvaluation.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Support/FloatEvaluation.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Support/FloatEvaluation.h (added)
+++ klee/trunk/include/klee/Internal/Support/FloatEvaluation.h Wed May 20 23:36:41 2009
@@ -0,0 +1,258 @@
+//===-- FloatEvaluation.h ---------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+// FIXME: Ditch this and use APFloat.
+
+#ifndef KLEE_UTIL_FLOATS_H
+#define KLEE_UTIL_FLOATS_H
+
+#include "klee/util/Bits.h"     //bits64::truncateToNBits
+#include "IntEvaluation.h" //ints::sext
+
+#include "llvm/Support/MathExtras.h"
+
+namespace klee {
+namespace floats {
+
+// ********************************** //
+// *** Pack uint64_t into FP types ** //
+// ********************************** //
+
+// interpret the 64 bits as a double instead of a uint64_t
+inline double UInt64AsDouble( uint64_t bits ) {
+  double ret;
+  assert( sizeof(bits) == sizeof(ret) );
+  memcpy( &ret, &bits, sizeof bits );
+  return ret;
+}
+
+// interpret the first 32 bits as a float instead of a uint64_t
+inline float UInt64AsFloat( uint64_t bits ) {
+  uint32_t tmp = bits; // ensure that we read correct bytes
+  float ret;
+  assert( sizeof(tmp) == sizeof(ret) );
+  memcpy( &ret, &tmp, sizeof tmp );
+  return ret;
+}
+
+
+// ********************************** //
+// *** Pack FP types into uint64_t ** //
+// ********************************** //
+
+// interpret the 64 bits as a uint64_t instead of a double
+inline uint64_t DoubleAsUInt64( double d ) {
+  uint64_t ret;
+  assert( sizeof(d) == sizeof(ret) );
+  memcpy( &ret, &d, sizeof d );
+  return ret;
+}
+
+// interpret the 32 bits as a uint64_t instead of as a float (add 32 0s)
+inline uint64_t FloatAsUInt64( float f ) {
+  uint32_t tmp;
+  assert( sizeof(tmp) == sizeof(f) );
+  memcpy( &tmp, &f, sizeof f );
+  return (uint64_t)tmp;
+}
+
+
+// ********************************** //
+// ************ Constants *********** //
+// ********************************** //
+
+const unsigned FLT_BITS = 32;
+const unsigned DBL_BITS = 64;
+
+
+
+// ********************************** //
+// ***** LLVM Binary Operations ***** //
+// ********************************** //
+
+// add of l and r
+inline uint64_t add(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64(UInt64AsFloat(l)  + UInt64AsFloat(r)),  FLT_BITS);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64(UInt64AsDouble(l) + UInt64AsDouble(r)), DBL_BITS);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// difference of l and r
+inline uint64_t sub(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64(UInt64AsFloat(l)  - UInt64AsFloat(r)),  FLT_BITS);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64(UInt64AsDouble(l) - UInt64AsDouble(r)), DBL_BITS);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// product of l and r
+inline uint64_t mul(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64(UInt64AsFloat(l)  * UInt64AsFloat(r)),  FLT_BITS);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64(UInt64AsDouble(l) * UInt64AsDouble(r)), DBL_BITS);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// signed divide of l by r
+inline uint64_t div(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64(UInt64AsFloat(l)  / UInt64AsFloat(r)),  FLT_BITS);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64(UInt64AsDouble(l) / UInt64AsDouble(r)), DBL_BITS);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// signed modulo of l by r
+inline uint64_t mod(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64( fmod(UInt64AsFloat(l),  UInt64AsFloat(r)) ), FLT_BITS);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64( fmod(UInt64AsDouble(l), UInt64AsDouble(r)) ), DBL_BITS);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+
+// ********************************** //
+// *** LLVM Comparison Operations *** //
+// ********************************** //
+
+// determine if l represents NaN
+inline bool isNaN(uint64_t l, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return llvm::IsNAN( UInt64AsFloat(l) );
+  case DBL_BITS: return llvm::IsNAN( UInt64AsDouble(l) );
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t eq(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  == UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) == UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t ne(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  != UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) != UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t lt(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  < UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) < UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t le(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  <= UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) <= UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t gt(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  > UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) > UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+inline uint64_t ge(uint64_t l, uint64_t r, unsigned inWidth) {
+  switch( inWidth ) {
+  case FLT_BITS: return UInt64AsFloat(l)  >= UInt64AsFloat(r);
+  case DBL_BITS: return UInt64AsDouble(l) >= UInt64AsDouble(r);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+
+// ********************************** //
+// *** LLVM Conversion Operations *** //
+// ********************************** //
+
+// truncation of l (which must be a double) to float (casts a double to a float)
+inline uint64_t trunc(uint64_t l, unsigned outWidth, unsigned inWidth) {
+  // FIXME: Investigate this, should this not happen? Was a quick
+  // patch for busybox.
+  if (inWidth==64 && outWidth==64) {
+    return l;
+  } else {
+    assert(inWidth==64 && "can only truncate from a 64-bit double");
+    assert(outWidth==32 && "can only truncate to a 32-bit float");
+    float trunc = (float)UInt64AsDouble(l);
+    return bits64::truncateToNBits(FloatAsUInt64(trunc), outWidth);
+  }
+}
+
+// extension of l (which must be a float) to double (casts a float to a double)
+inline uint64_t ext(uint64_t l, unsigned outWidth, unsigned inWidth) {
+  // FIXME: Investigate this, should this not happen? Was a quick
+  // patch for busybox.
+  if (inWidth==64 && outWidth==64) {
+    return l;
+  } else {
+    assert(inWidth==32 && "can only extend from a 32-bit float");
+    assert(outWidth==64 && "can only extend to a 64-bit double");
+    double ext = (double)UInt64AsFloat(l);
+    return bits64::truncateToNBits(DoubleAsUInt64(ext), outWidth);
+  }
+}
+
+// conversion of l to an unsigned integer, rounding towards zero
+inline uint64_t toUnsignedInt( uint64_t l, unsigned outWidth, unsigned inWidth ) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits((uint64_t)UInt64AsFloat(l),  outWidth );
+  case DBL_BITS: return bits64::truncateToNBits((uint64_t)UInt64AsDouble(l), outWidth );
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// conversion of l to a signed integer, rounding towards zero
+inline uint64_t toSignedInt( uint64_t l, unsigned outWidth, unsigned inWidth ) {
+  switch( inWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits((int64_t)UInt64AsFloat(l),  outWidth);
+  case DBL_BITS: return bits64::truncateToNBits((int64_t)UInt64AsDouble(l), outWidth);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// conversion of l, interpreted as an unsigned int, to a floating point number
+inline uint64_t UnsignedIntToFP( uint64_t l, unsigned outWidth ) {
+  switch( outWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64((float)l),  outWidth);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64((double)l), outWidth);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+// conversion of l, interpreted as a signed int, to a floating point number
+inline uint64_t SignedIntToFP( uint64_t l, unsigned outWidth, unsigned inWidth ) {
+  switch( outWidth ) {
+  case FLT_BITS: return bits64::truncateToNBits(FloatAsUInt64((float)(int64_t)ints::sext(l, 64, inWidth)), outWidth);
+  case DBL_BITS: return bits64::truncateToNBits(DoubleAsUInt64((double)(int64_t)ints::sext(l,64, inWidth)), outWidth);
+  default: assert(0 && "invalid floating point width");
+  }
+}
+
+} // end namespace ints
+} // end namespace klee
+
+#endif //KLEE_UTIL_FLOATS_H

Added: klee/trunk/include/klee/Internal/Support/IntEvaluation.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Support/IntEvaluation.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Support/IntEvaluation.h (added)
+++ klee/trunk/include/klee/Internal/Support/IntEvaluation.h Wed May 20 23:36:41 2009
@@ -0,0 +1,164 @@
+//===-- IntEvaluation.h -----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_INTEVALUATION_H
+#define KLEE_UTIL_INTEVALUATION_H
+
+#include "klee/util/Bits.h"
+
+#define MAX_BITS (sizeof(uint64_t) * 8)
+
+// ASSUMPTION: invalid bits in each uint64_t are 0. the trade-off here is
+// between making trunc/zext/sext fast and making operations that depend
+// on the invalid bits being 0 fast. 
+
+namespace klee {
+namespace ints {
+
+// add of l and r
+inline uint64_t add(uint64_t l, uint64_t r, unsigned inWidth) {
+  return bits64::truncateToNBits(l + r, inWidth);
+}
+
+// difference of l and r
+inline uint64_t sub(uint64_t l, uint64_t r, unsigned inWidth) {
+  return bits64::truncateToNBits(l - r, inWidth);
+}
+
+// product of l and r
+inline uint64_t mul(uint64_t l, uint64_t r, unsigned inWidth) {
+  return bits64::truncateToNBits(l * r, inWidth);
+}
+
+// truncation of l to outWidth bits
+inline uint64_t trunc(uint64_t l, unsigned outWidth, unsigned inWidth) {
+  return bits64::truncateToNBits(l, outWidth);
+}
+
+// zero-extension of l from inWidth to outWidth bits
+inline uint64_t zext(uint64_t l, unsigned outWidth, unsigned inWidth) {
+  return l;
+}
+
+// sign-extension of l from inWidth to outWidth bits
+inline uint64_t sext(uint64_t l, unsigned outWidth, unsigned inWidth) {
+  uint32_t numInvalidBits = MAX_BITS - inWidth;
+  return bits64::truncateToNBits(((int64_t)(l << numInvalidBits)) >> numInvalidBits, outWidth);
+}
+
+// unsigned divide of l by r
+inline uint64_t udiv(uint64_t l, uint64_t r, unsigned inWidth) {
+  return bits64::truncateToNBits(l / r, inWidth);
+}
+
+// unsigned mod of l by r
+inline uint64_t urem(uint64_t l, uint64_t r, unsigned inWidth) {
+  return bits64::truncateToNBits(l % r, inWidth);
+}
+
+// signed divide of l by r
+inline uint64_t sdiv(uint64_t l, uint64_t r, unsigned inWidth) {
+  // sign extend operands so that signed operation on 64-bits works correctly
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return bits64::truncateToNBits(sl / sr, inWidth);
+}
+
+// signed mod of l by r
+inline uint64_t srem(uint64_t l, uint64_t r, unsigned inWidth) {
+  // sign extend operands so that signed operation on 64-bits works correctly
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return bits64::truncateToNBits(sl % sr, inWidth);
+}
+
+// arithmetic shift right of l by r bits
+inline uint64_t ashr(uint64_t l, uint64_t shift, unsigned inWidth) {
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  return bits64::truncateToNBits(sl >> shift, inWidth);
+}
+
+// logical shift right of l by r bits
+inline uint64_t lshr(uint64_t l, uint64_t shift, unsigned inWidth) {
+  return l >> shift;
+}
+
+// shift left of l by r bits
+inline uint64_t shl(uint64_t l, uint64_t shift, unsigned inWidth) {
+  return bits64::truncateToNBits(l << shift, inWidth);
+}
+
+// logical AND of l and r
+inline uint64_t land(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l & r;
+}
+
+// logical OR of l and r
+inline uint64_t lor(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l | r;
+}
+
+// logical XOR of l and r
+inline uint64_t lxor(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l ^ r;
+}
+
+// comparison operations
+inline uint64_t eq(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l == r;
+}
+
+inline uint64_t ne(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l != r;
+}
+
+inline uint64_t ult(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l < r;
+}
+
+inline uint64_t ule(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l <= r;
+}
+
+inline uint64_t ugt(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l > r;
+}
+
+inline uint64_t uge(uint64_t l, uint64_t r, unsigned inWidth) {
+  return l >= r;
+}
+
+inline uint64_t slt(uint64_t l, uint64_t r, unsigned inWidth) {
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return sl < sr;
+}
+
+inline uint64_t sle(uint64_t l, uint64_t r, unsigned inWidth) {
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return sl <= sr;
+}
+
+inline uint64_t sgt(uint64_t l, uint64_t r, unsigned inWidth) {
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return sl > sr;
+}
+
+inline uint64_t sge(uint64_t l, uint64_t r, unsigned inWidth) {
+  int64_t sl = sext(l, MAX_BITS, inWidth);
+  int64_t sr = sext(r, MAX_BITS, inWidth);
+  return sl >= sr;
+}
+
+} // end namespace ints
+} // end namespace klee
+
+#endif 

Added: klee/trunk/include/klee/Internal/Support/ModuleUtil.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Support/ModuleUtil.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Support/ModuleUtil.h (added)
+++ klee/trunk/include/klee/Internal/Support/ModuleUtil.h Wed May 20 23:36:41 2009
@@ -0,0 +1,40 @@
+//===-- ModuleUtil.h --------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_TRANSFORM_UTIL_H
+#define KLEE_TRANSFORM_UTIL_H
+
+#include <string>
+
+namespace llvm {
+  class Function;
+  class Instruction;
+  class Module; 
+}
+
+namespace klee {
+ 
+  /// Link a module with a specified bitcode archive.
+  llvm::Module *linkWithLibrary(llvm::Module *module, 
+                                const std::string &libraryName);
+
+  /// Return the Function* target of a Call or Invoke instruction, or
+  /// null if it cannot be determined (should be only for indirect
+  /// calls, although complicated constant expressions might be
+  /// another possibility).
+  llvm::Function *getDirectCallTarget(const llvm::Instruction*);
+
+  /// Return true iff the given Function value is used in something
+  /// other than a direct call (or a constant expression that
+  /// terminates in a direct call).
+  bool functionEscapes(const llvm::Function *f);
+
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/Support/QueryLog.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Support/QueryLog.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Support/QueryLog.h (added)
+++ klee/trunk/include/klee/Internal/Support/QueryLog.h Wed May 20 23:36:41 2009
@@ -0,0 +1,63 @@
+//===-- QueryLog.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_OPT_LOGGINGSOLVER_H
+#define KLEE_OPT_LOGGINGSOLVER_H
+
+#include "klee/Expr.h"
+
+#include <vector>
+
+namespace klee {
+  struct Query;
+
+  class QueryLogEntry {
+  public:
+    enum Type {
+      Validity,
+      Truth,
+      Value,
+      Cex
+    };
+    
+    typedef std::vector< ref<Expr> > exprs_ty;
+    exprs_ty exprs;
+
+    Type type;
+    ref<Expr> query;
+    unsigned instruction;
+    std::vector<const Array*> objects;
+    
+  public:
+    QueryLogEntry() : query(0,Expr::Bool) {}
+    QueryLogEntry(const QueryLogEntry &b);
+    QueryLogEntry(const Query &_query, 
+                  Type _type,
+                  const std::vector<const Array*> *objects = 0);
+  };
+
+  class QueryLogResult {
+  public:
+    uint64_t result;
+    double time;
+
+  public:
+    QueryLogResult() {}
+    QueryLogResult(bool _success, uint64_t _result, double _time) 
+      : result(_result), time(_time) {
+      if (!_success) { // la la la
+        result = 0;
+        time = -1;
+      }
+    }
+  };
+  
+}
+
+#endif

Added: klee/trunk/include/klee/Internal/Support/Timer.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/Support/Timer.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/Support/Timer.h (added)
+++ klee/trunk/include/klee/Internal/Support/Timer.h Wed May 20 23:36:41 2009
@@ -0,0 +1,28 @@
+//===-- Timer.h -------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_TIMER_H
+#define KLEE_TIMER_H
+
+#include <stdint.h>
+
+namespace klee {
+  class WallTimer {
+    uint64_t startMicroseconds;
+    
+  public:
+    WallTimer();
+
+    /// check - Return the delta since the timer was created, in microseconds.
+    uint64_t check();
+  };
+}
+
+#endif
+

Added: klee/trunk/include/klee/Internal/System/Time.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Internal/System/Time.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Internal/System/Time.h (added)
+++ klee/trunk/include/klee/Internal/System/Time.h Wed May 20 23:36:41 2009
@@ -0,0 +1,20 @@
+//===-- Time.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_TIME_H
+#define KLEE_UTIL_TIME_H
+
+namespace klee {
+  namespace util {
+    double getUserTime();
+    double getWallTime();
+  }
+}
+
+#endif

Added: klee/trunk/include/klee/Interpreter.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Interpreter.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Interpreter.h (added)
+++ klee/trunk/include/klee/Interpreter.h Wed May 20 23:36:41 2009
@@ -0,0 +1,149 @@
+//===-- Interpreter.h - Abstract Execution Engine Interface -----*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_INTERPRETER_H
+#define KLEE_INTERPRETER_H
+
+#include <vector>
+#include <string>
+#include <map>
+#include <set>
+
+struct BOut;
+
+namespace llvm {
+class Function;
+class Module;
+}
+
+namespace klee {
+class ExecutionState;
+class Interpreter;
+class TreeStreamWriter;
+
+class InterpreterHandler {
+public:
+  InterpreterHandler() {}
+  virtual ~InterpreterHandler() {};
+
+  virtual std::ostream &getInfoStream() const = 0;
+
+  virtual std::string getOutputFilename(const std::string &filename) = 0;
+  virtual std::ostream *openOutputFile(const std::string &filename) = 0;
+
+  virtual void incPathsExplored() = 0;
+
+  virtual void processTestCase(const ExecutionState &state,
+                               const char *err, 
+                               const char *suffix) = 0;
+};
+
+class Interpreter {
+public:
+  /// ModuleOptions - Module level options which can be set when
+  /// registering a module with the interpreter.
+  struct ModuleOptions {
+    std::string LibraryDir;
+    bool Optimize;
+    bool CheckDivZero;
+
+    ModuleOptions(const std::string& _LibraryDir, 
+                  bool _Optimize, bool _CheckDivZero)
+      : LibraryDir(_LibraryDir), Optimize(_Optimize), 
+        CheckDivZero(_CheckDivZero) {}
+  };
+
+  /// InterpreterOptions - Options varying the runtime behavior during
+  /// interpretation.
+  struct InterpreterOptions {
+    /// A frequency at which to make concrete reads return constrained
+    /// symbolic values. This is used to test the correctness of the
+    /// symbolic execution on concrete programs.
+    unsigned MakeConcreteSymbolic;
+
+    InterpreterOptions()
+      : MakeConcreteSymbolic(false)
+    {}
+  };
+
+protected:
+  const InterpreterOptions interpreterOpts;
+
+  Interpreter(const InterpreterOptions &_interpreterOpts)
+    : interpreterOpts(_interpreterOpts)
+  {};
+
+public:
+  virtual ~Interpreter() {};
+
+  static Interpreter *create(const InterpreterOptions &_interpreterOpts,
+                             InterpreterHandler *ih);
+
+  /// Register the module to be executed.  
+  ///
+  /// \return The final module after it has been optimized, checks
+  /// inserted, and modified for interpretation.
+  virtual const llvm::Module * 
+  setModule(llvm::Module *module, 
+            const ModuleOptions &opts) = 0;
+
+  // supply a tree stream writer which the interpreter will use
+  // to record the concrete path (as a stream of '0' and '1' bytes).
+  virtual void setPathWriter(TreeStreamWriter *tsw) = 0;
+
+  // supply a tree stream writer which the interpreter will use
+  // to record the symbolic path (as a stream of '0' and '1' bytes).
+  virtual void setSymbolicPathWriter(TreeStreamWriter *tsw) = 0;
+
+  // supply a test case to replay from. this can be used to drive the
+  // interpretation down a user specified path. use null to reset.
+  virtual void setReplayOut(const struct BOut *out) = 0;
+
+  // supply a list of branch decisions specifying which direction to
+  // take on forks. this can be used to drive the interpretation down
+  // a user specified path. use null to reset.
+  virtual void setReplayPath(const std::vector<bool> *path) = 0;
+
+  // supply a set of symbolic bindings that will be used as "seeds"
+  // for the search. use null to reset.
+  virtual void useSeeds(const std::vector<struct BOut *> *seeds) = 0;
+
+  virtual void runFunctionAsMain(llvm::Function *f,
+                                 int argc,
+                                 char **argv,
+                                 char **envp) = 0;
+
+  /*** Runtime options ***/
+
+  virtual void setHaltExecution(bool value) = 0;
+
+  virtual void setInhibitForking(bool value) = 0;
+
+  /*** State accessor methods ***/
+
+  virtual unsigned getPathStreamID(const ExecutionState &state) = 0;
+
+  virtual unsigned getSymbolicPathStreamID(const ExecutionState &state) = 0;
+  
+  virtual void getConstraintLog(const ExecutionState &state,
+                                std::string &res,
+                                bool asCVC = false) = 0;
+
+  virtual bool getSymbolicSolution(const ExecutionState &state, 
+                                   std::vector< 
+                                   std::pair<std::string,
+                                   std::vector<unsigned char> > >
+                                   &res) = 0;
+
+  virtual void getCoveredLines(const ExecutionState &state,
+                               std::map<const std::string*, std::set<unsigned> > &res) = 0;
+};
+
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/Machine.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Machine.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Machine.h (added)
+++ klee/trunk/include/klee/Machine.h Wed May 20 23:36:41 2009
@@ -0,0 +1,28 @@
+//===-- Machine.h -----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __KLEE_MACHINE_H__
+#define __KLEE_MACHINE_H__
+
+#include "klee/Expr.h"
+
+namespace klee {
+  namespace machine {
+    enum ByteOrder {
+      LSB = 0,
+      MSB = 1
+    };
+  }
+}
+
+#define kMachineByteOrder      klee::machine::LSB
+#define kMachinePointerType    Expr::Int32
+#define kMachinePointerSize    4
+
+#endif /* __KLEE_MACHINE_H__ */

Added: klee/trunk/include/klee/Solver.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Solver.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Solver.h (added)
+++ klee/trunk/include/klee/Solver.h Wed May 20 23:36:41 2009
@@ -0,0 +1,216 @@
+//===-- Solver.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_SOLVER_H
+#define KLEE_SOLVER_H
+
+#include "klee/Expr.h"
+
+#include <vector>
+
+namespace klee {
+  class ConstraintManager;
+  class Expr;
+  class SolverImpl;
+
+  struct Query {
+  public:
+    const ConstraintManager &constraints;
+    ref<Expr> expr;
+
+    Query(const ConstraintManager& _constraints, ref<Expr> _expr)
+      : constraints(_constraints), expr(_expr) {
+    }
+
+    /// withExpr - Return a copy of the query with the given expression.
+    Query withExpr(ref<Expr> _expr) const {
+      return Query(constraints, _expr);
+    }
+
+    /// withFalse - Return a copy of the query with a false expression.
+    Query withFalse() const {
+      return Query(constraints, ref<Expr>(0, Expr::Bool));
+    }
+
+    /// negateExpr - Return a copy of the query with the expression negated.
+    Query negateExpr() const {
+      return withExpr(Expr::createNot(expr));
+    }
+  };
+
+  class Solver {
+    // DO NOT IMPLEMENT.
+    Solver(const Solver&);
+    void operator=(const Solver&);
+
+  public:
+    enum Validity {
+      True = 1,
+      False = -1,
+      Unknown = 0
+    };
+  
+  public:
+    /// validity_to_str - Return the name of given Validity enum value.
+    static const char *validity_to_str(Validity v);
+
+  public:
+    SolverImpl *impl;
+
+  public:
+    Solver(SolverImpl *_impl) : impl(_impl) {};
+    ~Solver();
+
+    /// evaluate - Determine the full validity of an expression in particular
+    /// state.
+    ////
+    /// \param [out] result - The validity of the given expression (provably
+    /// true, provably false, or neither).
+    ///
+    /// \return True on success.
+    bool evaluate(const Query&, Validity &result);
+  
+    /// mustBeTrue - Determine if the expression is provably true.
+    ///
+    /// \param [out] result - On success, true iff the expresssion is provably
+    /// false.
+    ///
+    /// \return True on success.
+    bool mustBeTrue(const Query&, bool &result);
+
+    /// mustBeFalse - Determine if the expression is provably false.
+    ///
+    /// \param [out] result - On success, true iff the expresssion is provably
+    /// false.
+    ///
+    /// \return True on success.
+    bool mustBeFalse(const Query&, bool &result);
+
+    /// mayBeTrue - Determine if there is a valid assignment for the given state
+    /// in which the expression evaluates to false.
+    ///
+    /// \param [out] result - On success, true iff the expresssion is true for
+    /// some satisfying assignment.
+    ///
+    /// \return True on success.
+    bool mayBeTrue(const Query&, bool &result);
+
+    /// mayBeFalse - Determine if there is a valid assignment for the given
+    /// state in which the expression evaluates to false.
+    ///
+    /// \param [out] result - On success, true iff the expresssion is false for
+    /// some satisfying assignment.
+    ///
+    /// \return True on success.
+    bool mayBeFalse(const Query&, bool &result);
+
+    /// getValue - Compute one possible value for the given expression.
+    ///
+    /// \param [out] result - On success, a value for the expression in some
+    /// satisying assignment.
+    ///
+    /// \return True on success.
+    bool getValue(const Query&, ref<Expr> &result);
+
+    /// getInitialValues - Compute the initial values for a list of objects.
+    ///
+    /// \param [out] result - On success, this vector will be filled in with an
+    /// array of bytes for each given object (with length matching the object
+    /// size). The bytes correspond to the initial values for the objects for
+    /// some satisying assignment.
+    ///
+    /// \return True on success.
+    ///
+    /// NOTE: This function returns failure if there is no satisfying
+    /// assignment.
+    //
+    // FIXME: This API is lame. We should probably just provide an API which
+    // returns an Assignment object, then clients can get out whatever values
+    // they want. This also allows us to optimize the representation.
+    bool getInitialValues(const Query&, 
+                          const std::vector<const Array*> &objects,
+                          std::vector< std::vector<unsigned char> > &result);
+
+    /// getRange - Compute a tight range of possible values for a given
+    /// expression.
+    ///
+    /// \return - A pair with (min, max) values for the expression.
+    ///
+    /// \post(mustBeTrue(min <= e <= max) && 
+    ///       mayBeTrue(min == e) &&
+    ///       mayBeTrue(max == e))
+    //
+    // FIXME: This should go into a helper class, and should handle failure.
+    virtual std::pair< ref<Expr>, ref<Expr> > getRange(const Query&);
+  };
+
+  /// STPSolver - A complete solver based on STP.
+  class STPSolver : public Solver {
+  public:
+    /// STPSolver - Construct a new STPSolver.
+    ///
+    /// \param useForkedSTP - Whether STP should be run in a separate process
+    /// (required for using timeouts).
+    STPSolver(bool useForkedSTP);
+    
+    /// getConstraintLog - Return the constraint log for the given state in CVC
+    /// format.
+    char *getConstraintLog(const Query&);
+    
+    /// setTimeout - Set constraint solver timeout delay to the given value; 0
+    /// is off.
+    void setTimeout(double timeout);
+  };
+
+  /* *** */
+
+  /// createValidatingSolver - Create a solver which will validate all query
+  /// results against an oracle, used for testing that an optimized solver has
+  /// the same results as an unoptimized one. This solver will assert on any
+  /// mismatches.
+  ///
+  /// \param s - The primary underlying solver to use.
+  /// \param oracle - The solver to check query results against.
+  Solver *createValidatingSolver(Solver *s, Solver *oracle);
+
+  /// createCachingSolver - Create a solver which will cache the queries in
+  /// memory (without eviction).
+  ///
+  /// \param s - The underlying solver to use.
+  Solver *createCachingSolver(Solver *s);
+
+  /// createCexCachingSolver - Create a counterexample caching solver. This is a
+  /// more sophisticated cache which records counterexamples for a constraint
+  /// set and uses subset/superset relations among constraints to try and
+  /// quickly find satisfying assignments.
+  ///
+  /// \param s - The underlying solver to use.
+  Solver *createCexCachingSolver(Solver *s);
+
+  /// createFastCexSolver - Create a "fast counterexample solver", which tries
+  /// to quickly compute a satisfying assignment for a constraint set using
+  /// value propogation and range analysis.
+  ///
+  /// \param s - The underlying solver to use.
+  Solver *createFastCexSolver(Solver *s);
+
+  /// createIndependentSolver - Create a solver which will eliminate any
+  /// unnecessary constraints before propogating the query to the underlying
+  /// solver.
+  ///
+  /// \param s - The underlying solver to use.
+  Solver *createIndependentSolver(Solver *s);
+  
+  /// createPCLoggingSolver - Create a solver which will forward all queries
+  /// after writing them to the given path in .pc format.
+  Solver *createPCLoggingSolver(Solver *s, std::string path);
+
+}
+
+#endif

Added: klee/trunk/include/klee/SolverImpl.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/SolverImpl.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/SolverImpl.h (added)
+++ klee/trunk/include/klee/SolverImpl.h Wed May 20 23:36:41 2009
@@ -0,0 +1,63 @@
+//===-- SolverImpl.h --------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_SOLVERIMPL_H
+#define KLEE_SOLVERIMPL_H
+
+#include <vector>
+
+namespace klee {
+  class Array;
+  class ExecutionState;
+  class Expr;
+  struct Query;
+
+  /// SolverImpl - Abstract base clase for solver implementations.
+  class SolverImpl {
+    // DO NOT IMPLEMENT.
+    SolverImpl(const SolverImpl&);
+    void operator=(const SolverImpl&);
+    
+  public:
+    SolverImpl() {}
+    virtual ~SolverImpl();
+
+    /// computeValidity - Compute a full validity result for the
+    /// query.
+    ///
+    /// The query expression is guaranteed to be non-constant and have
+    /// bool type.
+    ///
+    /// SolverImpl provides a default implementation which uses
+    /// computeTruth. Clients should override this if a more efficient
+    /// implementation is available.
+    virtual bool computeValidity(const Query& query, Solver::Validity &result);
+    
+    /// computeTruth - Determine whether the given query is provable.
+    ///
+    /// The query expression is guaranteed to be non-constant and have
+    /// bool type.
+    virtual bool computeTruth(const Query& query, bool &isValid) = 0;
+
+    /// computeValue - Compute a feasible value for the expression.
+    ///
+    /// The query expression is guaranteed to be non-constant.
+    virtual bool computeValue(const Query& query, ref<Expr> &result) = 0;
+    
+    virtual bool computeInitialValues(const Query& query,
+                                      const std::vector<const Array*> 
+                                        &objects,
+                                      std::vector< std::vector<unsigned char> > 
+                                        &values,
+                                      bool &hasSolution) = 0;  
+};
+
+}
+
+#endif

Added: klee/trunk/include/klee/Statistic.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Statistic.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Statistic.h (added)
+++ klee/trunk/include/klee/Statistic.h Wed May 20 23:36:41 2009
@@ -0,0 +1,65 @@
+//===-- Statistic.h ---------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_STATISTIC_H
+#define KLEE_STATISTIC_H
+
+#include <string>
+
+namespace klee {
+  class Statistic;
+  class StatisticManager;
+  class StatisticRecord;
+
+  /// Statistic - A named statistic instance.
+  ///
+  /// The Statistic class holds information about the statistic, but
+  /// not the actual values. Values are managed by the global
+  /// StatisticManager to enable transparent support for instruction
+  /// level and call path level statistics.
+  class Statistic {
+    friend class StatisticManager;
+    friend class StatisticRecord;
+
+  private:
+    unsigned id;
+    const std::string name;
+    const std::string shortName;
+
+  public:
+    Statistic(const std::string &_name, 
+              const std::string &_shortName);
+    ~Statistic();
+
+    /// getID - Get the unique statistic ID.
+    unsigned getID() { return id; }
+
+    /// getName - Get the statistic name.
+    const std::string &getName() const { return name; }
+
+    /// getShortName - Get the "short" statistic name, used in
+    /// callgrind output for example.
+    const std::string &getShortName() const { return shortName; }
+
+    /// getValue - Get the current primary statistic value.
+    uint64_t getValue() const;
+
+    /// operator uint64_t - Get the current primary statistic value.
+    operator uint64_t () const { return getValue(); }
+
+    /// operator++ - Increment the statistic by 1.
+    Statistic &operator ++() { return (*this += 1); }
+
+    /// operator+= - Increment the statistic by \arg addend.
+    Statistic &operator +=(const uint64_t addend);
+  };
+}
+
+#endif
+

Added: klee/trunk/include/klee/Statistics.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/Statistics.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/Statistics.h (added)
+++ klee/trunk/include/klee/Statistics.h Wed May 20 23:36:41 2009
@@ -0,0 +1,154 @@
+//===-- Statistics.h --------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_STATISTICS_H
+#define KLEE_STATISTICS_H
+
+#include "Statistic.h"
+
+#include <vector>
+#include <string>
+#include <string.h>
+
+namespace klee {
+  class Statistic;
+  class StatisticRecord {
+    friend class StatisticManager;
+
+  private:
+    uint64_t *data;
+
+  public:    
+    StatisticRecord();
+    StatisticRecord(const StatisticRecord &s);
+    ~StatisticRecord() { delete[] data; }
+    
+    void zero();
+
+    uint64_t getValue(const Statistic &s) const; 
+    void incrementValue(const Statistic &s, uint64_t addend) const;
+    StatisticRecord &operator =(const StatisticRecord &s);
+    StatisticRecord &operator +=(const StatisticRecord &sr);
+  };
+
+  class StatisticManager {
+  private:
+    bool enabled;
+    std::vector<Statistic*> stats;
+    uint64_t *globalStats;
+    uint64_t *indexedStats;
+    StatisticRecord *contextStats;
+    unsigned index;
+
+  public:
+    StatisticManager();
+    ~StatisticManager();
+
+    void useIndexedStats(unsigned totalIndices);
+
+    StatisticRecord *getContext();
+    void setContext(StatisticRecord *sr); /* null to reset */
+
+    void setIndex(unsigned i) { index = i; }
+    unsigned getIndex() { return index; }
+    unsigned getNumStatistics() { return stats.size(); }
+    Statistic &getStatistic(unsigned i) { return *stats[i]; }
+    
+    void registerStatistic(Statistic &s);
+    void incrementStatistic(Statistic &s, uint64_t addend);
+    uint64_t getValue(const Statistic &s) const;
+    void incrementIndexedValue(const Statistic &s, unsigned index, 
+                               uint64_t addend) const;
+    uint64_t getIndexedValue(const Statistic &s, unsigned index) const;
+    void setIndexedValue(const Statistic &s, unsigned index, uint64_t value);
+    int getStatisticID(const std::string &name) const;
+    Statistic *getStatisticByName(const std::string &name) const;
+  };
+
+  extern StatisticManager *theStatisticManager;
+
+  inline void StatisticManager::incrementStatistic(Statistic &s, 
+                                                   uint64_t addend) {
+    if (enabled) {
+      globalStats[s.id] += addend;
+      if (indexedStats) {
+        indexedStats[index*stats.size() + s.id] += addend;
+        if (contextStats)
+          contextStats->data[s.id] += addend;
+      }
+    }
+  }
+
+  inline StatisticRecord *StatisticManager::getContext() {
+    return contextStats;
+  }
+  inline void StatisticManager::setContext(StatisticRecord *sr) {
+    contextStats = sr;
+  }
+
+  inline void StatisticRecord::zero() {
+    ::memset(data, 0, sizeof(*data)*theStatisticManager->getNumStatistics());
+  }
+
+  inline StatisticRecord::StatisticRecord() 
+    : data(new uint64_t[theStatisticManager->getNumStatistics()]) {
+    zero();
+  }
+
+  inline StatisticRecord::StatisticRecord(const StatisticRecord &s) 
+    : data(new uint64_t[theStatisticManager->getNumStatistics()]) {
+    ::memcpy(data, s.data, 
+             sizeof(*data)*theStatisticManager->getNumStatistics());
+  }
+
+  inline StatisticRecord &StatisticRecord::operator=(const StatisticRecord &s) {
+    ::memcpy(data, s.data, 
+             sizeof(*data)*theStatisticManager->getNumStatistics());
+    return *this;
+  }
+
+  inline void StatisticRecord::incrementValue(const Statistic &s, 
+                                              uint64_t addend) const {
+    data[s.id] += addend;
+  }
+  inline uint64_t StatisticRecord::getValue(const Statistic &s) const { 
+    return data[s.id]; 
+  }
+
+  inline StatisticRecord &
+  StatisticRecord::operator +=(const StatisticRecord &sr) {
+    unsigned nStats = theStatisticManager->getNumStatistics();
+    for (unsigned i=0; i<nStats; i++)
+      data[i] += sr.data[i];
+    return *this;
+  }
+
+  inline uint64_t StatisticManager::getValue(const Statistic &s) const {
+    return globalStats[s.id];
+  }
+
+  inline void StatisticManager::incrementIndexedValue(const Statistic &s, 
+                                                      unsigned index,
+                                                      uint64_t addend) const {
+    indexedStats[index*stats.size() + s.id] += addend;
+  }
+
+  inline uint64_t StatisticManager::getIndexedValue(const Statistic &s, 
+                                                    unsigned index) const {
+    return indexedStats[index*stats.size() + s.id];
+  }
+
+  inline void StatisticManager::setIndexedValue(const Statistic &s, 
+                                                unsigned index,
+                                                uint64_t value) {
+    indexedStats[index*stats.size() + s.id] = value;
+  }
+}
+
+#endif

Added: klee/trunk/include/klee/TimerStatIncrementer.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/TimerStatIncrementer.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/TimerStatIncrementer.h (added)
+++ klee/trunk/include/klee/TimerStatIncrementer.h Wed May 20 23:36:41 2009
@@ -0,0 +1,32 @@
+//===-- TimerStatIncrementer.h ----------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_TIMERSTATINCREMENTER_H
+#define KLEE_TIMERSTATINCREMENTER_H
+
+#include "klee/Statistics.h"
+#include "klee/Internal/Support/Timer.h"
+
+namespace klee {
+  class TimerStatIncrementer {
+  private:
+    WallTimer timer;
+    Statistic &statistic;
+
+  public:
+    TimerStatIncrementer(Statistic &_statistic) : statistic(_statistic) {}
+    ~TimerStatIncrementer() {
+      statistic += timer.check(); 
+    };
+
+    uint64_t check() { return timer.check(); }
+  };
+}
+
+#endif

Added: klee/trunk/include/klee/klee.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/klee.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/klee.h (added)
+++ klee/trunk/include/klee/klee.h Wed May 20 23:36:41 2009
@@ -0,0 +1,120 @@
+//===-- klee.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __KLEE_H__
+#define __KLEE_H__
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+  
+  /* Add an accesible memory object at a user specified location. It
+     is the users responsibility to make sure that these memory
+     objects do not overlap. These memory objects will also
+     (obviously) not correctly interact with external function
+     calls. */
+  void klee_define_fixed_object(void *addr, unsigned nbytes);
+
+  /* make the contents of the object pointed to by addr symbolic. addr
+   * should always be the start of the object and nbytes must be its
+   * entire size. name is an optional name (can be the empty string)
+   */
+  void klee_make_symbolic_name(void *addr, unsigned nbytes, 
+			       const char *name);
+
+  void klee_make_symbolic(void *addr, unsigned nbytes);  
+  
+  /* Return symbolic value in the (signed) interval [begin,end) */
+  int klee_range(int begin, int end, const char *name);
+
+  /* Return a symbolic integer */
+  int klee_int(const char *name);
+
+  void *klee_malloc_n(unsigned nelems, unsigned size, unsigned alignment);
+
+  /* terminate the state without generating a test file */
+  __attribute__((noreturn))
+  void klee_silent_exit(int status);
+  
+  __attribute__((noreturn))
+  void klee_abort(void);  
+
+  /** Report an error and terminate the state. */
+  __attribute__((noreturn))
+  void klee_report_error(const char *file, 
+			 int line, 
+			 const char *message, 
+			 const char *suffix);
+  
+  /* called by checking code to get size of memory. */
+  unsigned klee_get_obj_size(void *ptr);
+  
+  /* print the tree associated w/ a given expression. */
+  void klee_print_expr(const char *msg, ...);
+  
+  /* NB: this *does not* fork n times and return [0,n) in children.
+   * It makes n be symbolic and returns: caller must compare N times.
+   */
+  unsigned klee_choose(unsigned n);
+  
+  /* special klee assert macro. this assert should be used when path consistency
+   * across platforms is desired (e.g., in tests).
+   * NB: __assert_fail is a klee "special" function
+   */
+# define klee_assert(expr)                                              \
+  ((expr)                                                               \
+   ? (void) (0)                                                         \
+   : __assert_fail (#expr, __FILE__, __LINE__, __PRETTY_FUNCTION__))    \
+
+  /* Return true if the given value is symbolic (represented by an
+   * expression) in the current state. This is primarily for debugging
+   * and writing tests but can also be used to enable prints in replay
+   * mode.
+   */
+  unsigned klee_is_symbolic(unsigned n);
+
+
+  /* The following intrinsics are primarily intended for internal use
+     and may have peculiar semantics. */
+
+  void klee_assume(unsigned condition);
+  void klee_warning(const char *message);
+  void klee_warning_once(const char *message);
+  void klee_prefer_cex(void *object, unsigned condition);
+  void klee_mark_global(void *object);
+
+  /* Return a possible constant value for the input expression. This
+     allows programs to forcibly concretize values on their own. */
+  unsigned klee_get_value(unsigned expr);
+
+  /* Ensure that memory in the range [address, address+size) is
+     accessible to the program. If some byte in the range is not
+     accessible an error will be generated and the state
+     terminated. 
+  
+     The current implementation requires both address and size to be
+     constants and that the range lie within a single object. */
+  void klee_check_memory_access(const void *address, unsigned size);
+
+  /* Enable/disable forking. */
+  void klee_set_forking(unsigned enable);
+
+  /* klee_alias_function("foo", "bar") will replace, at runtime (on
+     the current path and all paths spawned on the current path), all
+     calls to foo() by calls to bar().  foo() and bar() have to exist
+     and have identical types.  Use klee_alias_function("foo", "foo")
+     to undo.  Be aware that some special functions, such as exit(),
+     may not always work. */
+  void klee_alias_function(const char* fn_name, const char* new_fn_name);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __KLEE_H__ */

Added: klee/trunk/include/klee/util/Assignment.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/Assignment.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/Assignment.h (added)
+++ klee/trunk/include/klee/util/Assignment.h Wed May 20 23:36:41 2009
@@ -0,0 +1,100 @@
+//===-- Assignment.h --------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_ASSIGNMENT_H
+#define KLEE_UTIL_ASSIGNMENT_H
+
+#include <map>
+
+#include "klee/util/ExprEvaluator.h"
+
+// FIXME: Rename?
+
+namespace klee {
+  class Array;
+
+  class Assignment {
+  public:
+    typedef std::map<const Array*, std::vector<unsigned char> > bindings_ty;
+
+    bool allowFreeValues;
+    bindings_ty bindings;
+    
+  public:
+    Assignment(bool _allowFreeValues=false) 
+      : allowFreeValues(_allowFreeValues) {}
+    Assignment(std::vector<const Array*> &objects, 
+               std::vector< std::vector<unsigned char> > &values,
+               bool _allowFreeValues=false) 
+      : allowFreeValues(_allowFreeValues){
+      std::vector< std::vector<unsigned char> >::iterator valIt = 
+        values.begin();
+      for (std::vector<const Array*>::iterator it = objects.begin(),
+             ie = objects.end(); it != ie; ++it) {
+        const Array *os = *it;
+        std::vector<unsigned char> &arr = *valIt;
+        bindings.insert(std::make_pair(os, arr));
+        ++valIt;
+      }
+    }
+    
+    ref<Expr> evaluate(const Array *mo, unsigned index) const;
+    ref<Expr> evaluate(ref<Expr> e);
+
+    template<typename InputIterator>
+    bool satisfies(InputIterator begin, InputIterator end);
+  };
+  
+  class AssignmentEvaluator : public ExprEvaluator {
+    const Assignment &a;
+
+  protected:
+    ref<Expr> getInitialValue(const Array &mo, unsigned index) {
+      return a.evaluate(&mo, index);
+    }
+    
+  public:
+    AssignmentEvaluator(const Assignment &_a) : a(_a) {}    
+  };
+
+  /***/
+
+  inline ref<Expr> Assignment::evaluate(const Array *array, 
+                                        unsigned index) const {
+    bindings_ty::const_iterator it = bindings.find(array);
+    if (it!=bindings.end() && index<it->second.size()) {
+      return ref<Expr>(it->second[index], Expr::Int8);
+    } else {
+      if (allowFreeValues) {
+        return ReadExpr::create(UpdateList(array, true, 0), 
+                                ref<Expr>(index, Expr::Int32));
+      } else {
+        return ref<Expr>(0, Expr::Int8);
+      }
+    }
+  }
+
+  inline ref<Expr> Assignment::evaluate(ref<Expr> e) { 
+    AssignmentEvaluator v(*this);
+    return v.visit(e); 
+  }
+
+  template<typename InputIterator>
+  inline bool Assignment::satisfies(InputIterator begin, InputIterator end) {
+    AssignmentEvaluator v(*this);
+    for (; begin!=end; ++begin) {
+      ref<Expr> res = v.visit(*begin);
+      if (!res.isConstant() || !res.getConstantValue())
+        return false;
+    }
+    return true;
+  }
+}
+
+#endif

Added: klee/trunk/include/klee/util/BitArray.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/BitArray.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/BitArray.h (added)
+++ klee/trunk/include/klee/util/BitArray.h Wed May 20 23:36:41 2009
@@ -0,0 +1,42 @@
+//===-- BitArray.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_BITARRAY_H
+#define KLEE_UTIL_BITARRAY_H
+
+namespace klee {
+
+  // XXX would be nice not to have
+  // two allocations here for allocated
+  // BitArrays
+class BitArray {
+private:
+  uint32_t *bits;
+  
+protected:
+  static uint32_t length(unsigned size) { return (size+31)/32; }
+
+public:
+  BitArray(unsigned size, bool value = false) : bits(new uint32_t[length(size)]) {
+    memset(bits, value?0xFF:0, sizeof(*bits)*length(size));
+  }
+  BitArray(const BitArray &b, unsigned size) : bits(new uint32_t[length(size)]) {
+    memcpy(bits, b.bits, sizeof(*bits)*length(size));
+  }
+  ~BitArray() { delete[] bits; }
+
+  bool get(unsigned idx) { return (bool) ((bits[idx/32]>>(idx&0x1F))&1); }
+  void set(unsigned idx) { bits[idx/32] |= 1<<(idx&0x1F); }
+  void unset(unsigned idx) { bits[idx/32] &= ~(1<<(idx&0x1F)); }
+  void set(unsigned idx, bool value) { if (value) set(idx); else unset(idx); }
+};
+
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/util/Bits.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/Bits.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/Bits.h (added)
+++ klee/trunk/include/klee/util/Bits.h Wed May 20 23:36:41 2009
@@ -0,0 +1,102 @@
+//===-- Bits.h --------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_UTIL_BITS_H
+#define KLEE_UTIL_BITS_H
+
+#include "llvm/Support/DataTypes.h"
+
+namespace klee {
+  namespace bits32 {
+    // @pre(0 <= N <= 32)
+    // @post(retval = max([truncateToNBits(i,N) for i in naturals()]))
+    inline unsigned maxValueOfNBits(unsigned N) {
+      if (N==0)
+        return 0;
+      return ((unsigned) -1) >> (32 - N);
+    }
+
+    // @pre(0 < N <= 32)
+    inline unsigned truncateToNBits(unsigned x, unsigned N) {
+      return x&(((unsigned) -1) >> (32 - N));
+    }
+
+    inline unsigned withoutRightmostBit(unsigned x) {
+      return x&(x-1);
+    }
+
+    inline unsigned isolateRightmostBit(unsigned x) {
+      return x&-x;
+    }
+
+    inline unsigned isPowerOfTwo(unsigned x) {
+      if (x==0) return 0;
+      return !(x&(x-1));
+    }
+
+    // @pre(withoutRightmostBit(x) == 0)
+    // @post((1 << retval) == x)
+    inline unsigned indexOfSingleBit(unsigned x) {
+      unsigned res = 0;
+      if (x&0xFFFF0000) res += 16;
+      if (x&0xFF00FF00) res += 8;
+      if (x&0xF0F0F0F0) res += 4;
+      if (x&0xCCCCCCCC) res += 2;
+      if (x&0xAAAAAAAA) res += 1;
+      return res;
+    } 
+
+    inline unsigned indexOfRightmostBit(unsigned x) {
+      return indexOfSingleBit(isolateRightmostBit(x));
+    }
+  }
+
+  namespace bits64 {
+    // @pre(0 <= N <= 32)
+    // @post(retval = max([truncateToNBits(i,N) for i in naturals()]))
+    inline uint64_t maxValueOfNBits(unsigned N) {
+      if (N==0)
+        return 0;
+      return ((uint64_t) (int64_t) -1) >> (64 - N);
+    }
+    
+    // @pre(0 < N <= 64)
+    inline uint64_t truncateToNBits(uint64_t x, unsigned N) {
+      return x&(((uint64_t) (int64_t) -1) >> (64 - N));
+    }
+
+    inline uint64_t withoutRightmostBit(uint64_t x) {
+      return x&(x-1);
+    }
+
+    inline uint64_t isolateRightmostBit(uint64_t x) {
+      return x&-x;
+    }
+
+    inline uint64_t isPowerOfTwo(uint64_t x) {
+      if (x==0) return 0;
+      return !(x&(x-1));
+    }
+
+    // @pre((x&(x-1)) == 0)
+    // @post((1 << retval) == x)
+    inline unsigned indexOfSingleBit(uint64_t x) {
+      unsigned res = bits32::indexOfSingleBit((unsigned) (x | (x>>32)));
+      if (x&((uint64_t) 0xFFFFFFFF << 32))
+	  res += 32;
+      return res;
+    } 
+
+    inline uint64_t indexOfRightmostBit(uint64_t x) {
+      return indexOfSingleBit(isolateRightmostBit(x));
+    }
+  }
+} // End klee namespace
+
+#endif

Added: klee/trunk/include/klee/util/ExprEvaluator.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprEvaluator.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprEvaluator.h (added)
+++ klee/trunk/include/klee/util/ExprEvaluator.h Wed May 20 23:36:41 2009
@@ -0,0 +1,39 @@
+//===-- ExprEvaluator.h -----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPREVALUATOR_H
+#define KLEE_EXPREVALUATOR_H
+
+#include "klee/Expr.h"
+#include "klee/util/ExprVisitor.h"
+
+namespace klee {
+  class ExprEvaluator : public ExprVisitor {
+  protected:
+    Action evalRead(const UpdateList &ul, unsigned index);
+    Action visitRead(const ReadExpr &re);
+      
+    Action protectedDivOperation(const BinaryExpr &e);
+    Action visitUDiv(const UDivExpr &e);
+    Action visitSDiv(const SDivExpr &e);
+    Action visitURem(const URemExpr &e);
+    Action visitSRem(const SRemExpr &e);
+      
+  public:
+    ExprEvaluator() {}
+
+    // override to implement evaluation, this function is called to
+    // get the initial value for a symbolic byte. if the value is
+    // unknown then the user can simply return a ReadExpr at version 0
+    // of this MemoryObject.
+    virtual ref<Expr> getInitialValue(const Array& os, unsigned index) = 0;
+  };
+}
+
+#endif

Added: klee/trunk/include/klee/util/ExprHashMap.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprHashMap.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprHashMap.h (added)
+++ klee/trunk/include/klee/util/ExprHashMap.h Wed May 20 23:36:41 2009
@@ -0,0 +1,48 @@
+//===-- ExprHashMap.h -------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPRHASHMAP_H
+#define KLEE_EXPRHASHMAP_H
+
+#include "klee/Expr.h"
+#include <tr1/unordered_map>
+#include <tr1/unordered_set>
+
+namespace klee {
+
+  namespace util {
+    struct ExprHash  {
+      unsigned operator()(const ref<Expr> e) const {
+        return e.hash();
+      }
+    };
+    
+    struct ExprCmp {
+      bool operator()(const ref<Expr> &a, const ref<Expr> &b) const {
+        return a==b;
+      }
+    };
+  }
+  
+  template<class T> 
+  class ExprHashMap : 
+
+    public std::tr1::unordered_map<ref<Expr>,
+				   T,
+				   klee::util::ExprHash,
+				   klee::util::ExprCmp> {
+  };
+  
+  typedef std::tr1::unordered_set<ref<Expr>,
+				  klee::util::ExprHash,
+				  klee::util::ExprCmp> ExprHashSet;
+
+}
+
+#endif

Added: klee/trunk/include/klee/util/ExprPPrinter.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprPPrinter.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprPPrinter.h (added)
+++ klee/trunk/include/klee/util/ExprPPrinter.h Wed May 20 23:36:41 2009
@@ -0,0 +1,58 @@
+//===-- ExprPPrinter.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPRPPRINTER_H
+#define KLEE_EXPRPPRINTER_H
+
+#include "klee/Expr.h"
+
+namespace klee {
+  class ConstraintManager;
+
+  class ExprPPrinter {
+  protected:
+    ExprPPrinter() {}
+    
+  public:
+    static ExprPPrinter *create(std::ostream &os);
+
+    virtual ~ExprPPrinter() {}
+
+    virtual void setNewline(const std::string &newline) = 0;
+    virtual void reset() = 0;
+    virtual void scan(const ref<Expr> &e) = 0;
+    virtual void print(const ref<Expr> &e, unsigned indent=0) = 0;
+
+    // utility methods
+
+    template<class Container>
+    void scan(Container c) {
+      scan(c.begin(), c.end());
+    }
+
+    template<class InputIterator>
+    void scan(InputIterator it, InputIterator end) {
+      for (; it!=end; ++it)
+        scan(*it);
+    }
+
+    static void printOne(std::ostream &os, const char *message, 
+                         const ref<Expr> &e);
+
+    static void printConstraints(std::ostream &os,
+                                 const ConstraintManager &constraints);
+
+    static void printQuery(std::ostream &os,
+                           const ConstraintManager &constraints,
+                           const ref<Expr> &q);
+  };
+
+}
+
+#endif

Added: klee/trunk/include/klee/util/ExprRangeEvaluator.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprRangeEvaluator.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprRangeEvaluator.h (added)
+++ klee/trunk/include/klee/util/ExprRangeEvaluator.h Wed May 20 23:36:41 2009
@@ -0,0 +1,283 @@
+//===-- ExprRangeEvaluator.h ------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPRRANGEEVALUATOR_H
+#define KLEE_EXPRRANGEEVALUATOR_H
+
+#include "klee/Expr.h"
+#include "klee/util/Bits.h"
+
+namespace klee {
+
+/*
+class ValueType {
+public:
+  ValueType(); // empty range
+  ValueType(uint64_t value);
+  ValueType(uint64_t min, uint64_t max);
+  
+  bool mustEqual(const uint64_t b);
+  bool mustEqual(const ValueType &b);
+  bool mayEqual(const uint64_t b);  
+  bool mayEqual(const ValueType &b);
+
+  bool isFullRange(unsigned width);
+
+  ValueType set_union(ValueType &);
+  ValueType set_intersection(ValueType &);
+  ValueType set_difference(ValueType &);
+
+  ValueType binaryAnd(ValueType &);
+  ValueType binaryOr(ValueType &);
+  ValueType binaryXor(ValueType &);
+  ValueType concat(ValueType &, unsigned width);
+  ValueType add(ValueType &, unsigned width);
+  ValueType sub(ValueType &, unsigned width);
+  ValueType mul(ValueType &, unsigned width);
+  ValueType udiv(ValueType &, unsigned width);
+  ValueType sdiv(ValueType &, unsigned width);
+  ValueType urem(ValueType &, unsigned width);
+  ValueType srem(ValueType &, unsigned width);
+
+  uint64_t min();
+  uint64_t max();
+  int64_t minSigned(unsigned width);
+  int64_t maxSigned(unsigned width);
+}
+*/
+
+template<class T>
+class ExprRangeEvaluator {
+protected:
+  virtual T getInitialReadRange(const Array &os, T index) = 0;
+
+  T evalRead(const UpdateList &ul, T index);
+
+public:
+  ExprRangeEvaluator() {}
+  virtual ~ExprRangeEvaluator() {}
+
+  T evaluate(const ref<Expr> &e);
+};
+
+template<class T>
+T ExprRangeEvaluator<T>::evalRead(const UpdateList &ul,
+                                  T index) {
+  T res;
+
+  for (const UpdateNode *un=ul.head; un; un=un->next) {
+    T ui = evaluate(un->index);
+
+    if (ui.mustEqual(index)) {
+      return res.set_union(evaluate(un->value));
+    } else if (ui.mayEqual(index)) {
+      res = res.set_union(evaluate(un->value));
+      if (res.isFullRange(8)) {
+        return res;
+      } 
+    }
+  }
+
+  return res.set_union(getInitialReadRange(*ul.root, index));
+}
+
+template<class T>
+T ExprRangeEvaluator<T>::evaluate(const ref<Expr> &e) {
+  switch (e.getKind()) {
+  case Expr::Constant:
+    return T(e.getConstantValue());
+
+  case Expr::NotOptimized: 
+    break;
+
+  case Expr::Read: {
+    const ReadExpr *re = static_ref_cast<const ReadExpr>(e);
+    T index = evaluate(re->index);
+
+    assert(re->getWidth()==Expr::Int8 && "unexpected multibyte read");
+
+    return evalRead(re->updates, index);
+  }
+
+  case Expr::Select: {
+    const SelectExpr *se = static_ref_cast<const SelectExpr>(e);
+    T cond = evaluate(se->cond);
+      
+    if (cond.mustEqual(1)) {
+      return evaluate(se->trueExpr);
+    } else if (cond.mustEqual(0)) {
+      return evaluate(se->falseExpr);
+    } else {
+      return evaluate(se->trueExpr).set_union(evaluate(se->falseExpr));
+    }
+  }
+
+    // XXX these should be unrolled to ensure nice inline
+  case Expr::Concat: {
+    const Expr *ep = e.get();
+    T res(0);
+    for (unsigned i=0; i<ep->getNumKids(); i++)
+      res = res.concat(evaluate(ep->getKid(i)),8);
+    return res;
+  }
+
+    // Arithmetic
+
+  case Expr::Add: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).add(evaluate(be->right), width);
+  }
+  case Expr::Sub: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).sub(evaluate(be->right), width);
+  }
+  case Expr::Mul: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).mul(evaluate(be->right), width);
+  }
+  case Expr::UDiv: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).udiv(evaluate(be->right), width);
+  }
+  case Expr::SDiv: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).sdiv(evaluate(be->right), width);
+  }
+  case Expr::URem: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).urem(evaluate(be->right), width);
+  }
+  case Expr::SRem: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    unsigned width = be->left.getWidth();
+    return evaluate(be->left).srem(evaluate(be->right), width);
+  }
+
+    // Binary
+
+  case Expr::And: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    return evaluate(be->left).binaryAnd(evaluate(be->right));
+  }
+  case Expr::Or: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    return evaluate(be->left).binaryOr(evaluate(be->right));
+  }
+  case Expr::Xor: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    return evaluate(be->left).binaryXor(evaluate(be->right));
+  }
+  case Expr::Shl: {
+    //    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    //    unsigned width = be->left.getWidth();
+    //    return evaluate(be->left).shl(evaluate(be->right), width);
+    break;
+  }
+  case Expr::LShr: {
+    //    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    //    unsigned width = be->left.getWidth();
+    //    return evaluate(be->left).lshr(evaluate(be->right), width);
+    break;
+  }
+  case Expr::AShr: {
+    //    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    //    unsigned width = be->left.getWidth();
+    //    return evaluate(be->left).ashr(evaluate(be->right), width);
+    break;
+  }
+
+    // Comparison
+
+  case Expr::Eq: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    T left = evaluate(be->left);
+    T right = evaluate(be->right);
+      
+    if (left.mustEqual(right)) {
+      return T(1);
+    } else if (!left.mayEqual(right)) {
+      return T(0);
+    }
+    break;
+  }
+
+  case Expr::Ult: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    T left = evaluate(be->left);
+    T right = evaluate(be->right);
+      
+    if (left.max() < right.min()) {
+      return T(1);
+    } else if (left.min() >= right.max()) {
+      return T(0);
+    }
+    break;
+  }
+  case Expr::Ule: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    T left = evaluate(be->left);
+    T right = evaluate(be->right);
+      
+    if (left.max() <= right.min()) {
+      return T(1);
+    } else if (left.min() > right.max()) {
+      return T(0);
+    }
+    break;
+  }
+  case Expr::Slt: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    T left = evaluate(be->left);
+    T right = evaluate(be->right);
+    unsigned bits = be->left.getWidth();
+
+    if (left.maxSigned(bits) < right.minSigned(bits)) {
+      return T(1);
+    } else if (left.minSigned(bits) >= right.maxSigned(bits)) {
+      return T(0);
+    }
+    break;
+  }
+  case Expr::Sle: {
+    const BinaryExpr *be = static_ref_cast<const BinaryExpr>(e);
+    T left = evaluate(be->left);
+    T right = evaluate(be->right);
+    unsigned bits = be->left.getWidth();
+      
+    if (left.maxSigned(bits) <= right.minSigned(bits)) {
+      return T(1);
+    } else if (left.minSigned(bits) > right.maxSigned(bits)) {
+      return T(0);
+    }
+    break;
+  }
+
+  case Expr::Ne:
+  case Expr::Ugt:
+  case Expr::Uge:
+  case Expr::Sgt:
+  case Expr::Sge:
+    assert(0 && "invalid expressions (uncanonicalized)");
+
+  default:
+    break;
+  }
+
+  return T(0, bits64::maxValueOfNBits(e.getWidth()));
+}
+
+}
+
+#endif

Added: klee/trunk/include/klee/util/ExprUtil.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprUtil.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprUtil.h (added)
+++ klee/trunk/include/klee/util/ExprUtil.h Wed May 20 23:36:41 2009
@@ -0,0 +1,43 @@
+//===-- ExprUtil.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPRUTIL_H
+#define KLEE_EXPRUTIL_H
+
+#include <vector>
+
+namespace klee {
+  class Array;
+  class Expr;
+  class ReadExpr;
+  template<typename T> class ref;
+
+  /// Find all ReadExprs used in the expression DAG. If visitUpdates
+  /// is true then this will including those reachable by traversing
+  /// update lists. Note that this may be slow and return a large
+  /// number of results.
+  void findReads(ref<Expr> e, 
+                 bool visitUpdates,
+                 std::vector< ref<ReadExpr> > &result);
+  
+  /// Return a list of all unique symbolic objects referenced by the given
+  /// expression.
+  void findSymbolicObjects(ref<Expr> e,
+                           std::vector<const Array*> &results);
+
+  /// Return a list of all unique symbolic objects referenced by the
+  /// given expression range.
+  template<typename InputIterator>
+  void findSymbolicObjects(InputIterator begin, 
+                           InputIterator end,
+                           std::vector<const Array*> &results);
+
+}
+
+#endif

Added: klee/trunk/include/klee/util/ExprVisitor.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/ExprVisitor.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/ExprVisitor.h (added)
+++ klee/trunk/include/klee/util/ExprVisitor.h Wed May 20 23:36:41 2009
@@ -0,0 +1,95 @@
+//===-- ExprVisitor.h -------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXPRVISITOR_H
+#define KLEE_EXPRVISITOR_H
+
+#include "ExprHashMap.h"
+
+namespace klee {
+  class ExprVisitor {
+  protected:
+    // typed variant, but non-virtual for efficiency
+    class Action {
+    public:
+      enum Kind { SkipChildren, DoChildren, ChangeTo };
+
+    private:
+      //      Action() {}
+      Action(Kind _kind) 
+        : kind(_kind), argument(0,Expr::Bool) {}
+      Action(Kind _kind, const ref<Expr> &_argument) 
+        : kind(_kind), argument(_argument) {}
+
+      friend class ExprVisitor;
+
+    public:
+      Kind kind;
+      ref<Expr> argument;
+
+      static Action changeTo(const ref<Expr> &expr) { return Action(ChangeTo,expr); }
+      static Action doChildren() { return Action(DoChildren); }
+      static Action skipChildren() { return Action(SkipChildren); }
+    };
+
+  protected:
+    explicit
+    ExprVisitor(bool _recursive=false) : recursive(_recursive) {}
+    virtual ~ExprVisitor() {}
+
+    virtual Action visitExpr(const Expr&);
+    virtual Action visitExprPost(const Expr&);
+
+    virtual Action visitNotOptimized(const NotOptimizedExpr&);
+    virtual Action visitRead(const ReadExpr&);
+    virtual Action visitSelect(const SelectExpr&);
+    virtual Action visitConcat(const ConcatExpr&);
+    virtual Action visitExtract(const ExtractExpr&);
+    virtual Action visitZExt(const ZExtExpr&);
+    virtual Action visitSExt(const SExtExpr&);
+    virtual Action visitAdd(const AddExpr&);
+    virtual Action visitSub(const SubExpr&);
+    virtual Action visitMul(const MulExpr&);
+    virtual Action visitUDiv(const UDivExpr&);
+    virtual Action visitSDiv(const SDivExpr&);
+    virtual Action visitURem(const URemExpr&);
+    virtual Action visitSRem(const SRemExpr&);
+    virtual Action visitAnd(const AndExpr&);
+    virtual Action visitOr(const OrExpr&);
+    virtual Action visitXor(const XorExpr&);
+    virtual Action visitShl(const ShlExpr&);
+    virtual Action visitLShr(const LShrExpr&);
+    virtual Action visitAShr(const AShrExpr&);
+    virtual Action visitEq(const EqExpr&);
+    virtual Action visitNe(const NeExpr&);
+    virtual Action visitUlt(const UltExpr&);
+    virtual Action visitUle(const UleExpr&);
+    virtual Action visitUgt(const UgtExpr&);
+    virtual Action visitUge(const UgeExpr&);
+    virtual Action visitSlt(const SltExpr&);
+    virtual Action visitSle(const SleExpr&);
+    virtual Action visitSgt(const SgtExpr&);
+    virtual Action visitSge(const SgeExpr&);
+
+  private:
+    typedef ExprHashMap< ref<Expr> > visited_ty;
+    visited_ty visited;
+    bool recursive;
+
+    ref<Expr> visitActual(const ref<Expr> &e);
+    
+  public:
+    // apply the visitor to the expression and return a possibly
+    // modified new expression.
+    ref<Expr> visit(const ref<Expr> &e);
+  };
+
+}
+
+#endif

Added: klee/trunk/include/klee/util/Ref.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/include/klee/util/Ref.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/include/klee/util/Ref.h (added)
+++ klee/trunk/include/klee/util/Ref.h Wed May 20 23:36:41 2009
@@ -0,0 +1,303 @@
+//===-- Ref.h ---------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_REF_H
+#define KLEE_REF_H
+
+#include <assert.h>
+
+class Expr;
+class BinaryExpr;
+class CastExpr;
+class CmpExpr;
+
+class ConstantExpr;
+class ReadExpr;
+class UpdateNode;
+class NotOptimizedExpr;
+class ReadExpr;
+class SelectExpr;
+class ConcatExpr;
+class ExtractExpr;
+class ZExtExpr;
+class SExtExpr;
+class AddExpr;
+class SubExpr;
+class MulExpr;
+class UDivExpr;
+class SDivExpr;
+class URemExpr;
+class SRemExpr;
+class AndExpr;
+class OrExpr;
+class XorExpr;
+class ShlExpr;
+class LShrExpr;
+class AShrExpr;
+class EqExpr;
+class NeExpr;
+class UltExpr;
+class UleExpr;
+class UgtExpr;
+class UgeExpr;
+class SltExpr;
+class SleExpr;
+class SgtExpr;
+class SgeExpr;
+class KModule;
+
+  class ExprVisitor;
+  class StackFrame;
+  class ObjectState;
+
+template<class T>
+class ref {
+public:
+  // default constructor: create a NULL reference
+  ref() : constantWidth(Expr::InvalidWidth) {
+    contents.ptr = 0;
+  }
+
+private:
+  // if NotConstant, not this ref is not a constant.
+  // otherwise, it's the width of the constant.
+  Expr::Width constantWidth;
+  static const Expr::Width NotConstant = (Expr::Width) 0;
+  
+  union {
+    T *ptr;
+    uint64_t val;
+  } contents;
+  
+  void inc() {
+    if (constantWidth == NotConstant &&
+        contents.ptr) {
+      contents.ptr->refCount++;
+    }
+  }
+  
+  void dec() {
+    if (constantWidth == NotConstant &&
+        contents.ptr &&
+        --contents.ptr->refCount == 0) {
+      delete contents.ptr;
+    }
+  }  
+
+  friend class ExprVisitor;
+  friend class Cell;
+  friend class ObjectState;
+  friend class KModule;
+
+public:
+  template<class U> friend class ref;
+  template<class U> friend U* dyn_ref_cast(ref &src);
+  template<class U> friend const U* dyn_ref_cast(const ref &src);
+  template<class U> friend U* static_ref_cast(ref &src);
+  template<class U> friend const U* static_ref_cast(const ref &src);
+
+  // constructor from pointer
+  ref(T *p) : constantWidth(NotConstant) {
+    contents.ptr = p;
+    inc();
+  }
+
+  // construct from constant
+  ref(uint64_t val, Expr::Width w) : constantWidth(w) {
+    contents.val = val;
+  }
+  
+  // normal copy constructor
+  ref (const ref<T> &r)
+    : constantWidth(r.constantWidth), contents(r.contents) {
+    inc();
+  }
+  
+  // conversion constructor
+  template<class U>
+  ref (const ref<U> &r) {
+    constantWidth = r.constantWidth;
+    contents.val = r.contents.val;
+    inc();
+  }
+  
+  // pointer operations
+  T *get () {
+    // demand(constantWidth == NotConstant, "deref of constant");
+
+    // allocate 
+    if (constantWidth != NotConstant) {
+      contents.ptr = dynamic_cast<T*>(Expr::createConstant(contents.val, constantWidth));
+      assert(contents.ptr && "error with lazy constant initialization");
+      constantWidth = NotConstant;
+      
+      inc();
+    }
+    return contents.ptr;
+  }
+
+  T *get () const {
+    assert(constantWidth == NotConstant && "deref of constant");
+    return contents.ptr;
+  }
+
+  // method calls for the constant optimization
+  bool isConstant() const {
+    if (constantWidth != NotConstant) {
+      return true;
+    } else if (contents.ptr) {
+      return contents.ptr->getKind() == Expr::Constant;
+    } else {
+      return false; // should never happen, but nice check
+    }
+  }
+
+  uint64_t getConstantValue() const {
+    if (constantWidth) {
+      return contents.val;
+    } else {
+      return contents.ptr->getConstantValue();
+    }
+  }
+
+  unsigned hash() const {
+    if (constantWidth) {
+      return Expr::hashConstant(contents.val, constantWidth);
+    } else {
+      return contents.ptr->hash();
+    }
+  }
+
+  unsigned computeHash() const {
+    if (isConstant()) {
+      return Expr::hashConstant(contents.val, constantWidth);
+    } else {
+      return contents.ptr->computeHash();
+    }
+  }
+
+  void rehash() const {
+    if (!isConstant())
+      contents.ptr->computeHash();
+  }
+  
+  Expr::Width getWidth() const {
+    if (constantWidth != NotConstant)
+      return constantWidth;
+    return contents.ptr->getWidth();
+  }
+
+  Expr::Kind getKind() const {
+    if (constantWidth != NotConstant)
+      return Expr::Constant;
+    return contents.ptr->getKind();
+  }
+
+  unsigned getNumKids() const {
+    if (constantWidth != NotConstant)
+      return 0;
+    return contents.ptr->getNumKids();
+  }
+
+  ref<Expr> getKid(unsigned k) {
+    if (constantWidth != NotConstant)
+      return 0;
+    return contents.ptr->getKid(k);
+  }
+
+  ~ref () { dec (); }
+
+  /* The copy assignment operator must also explicitly be defined,
+   * despite a redundant template. */
+  ref<T> &operator= (const ref<T> &r) {
+    dec();
+    constantWidth = r.constantWidth;
+    contents.val = r.contents.val;
+    inc();
+    
+    return *this;
+  }
+  
+  template<class U> ref<T> &operator= (const ref<U> &r) {
+    dec();
+    constantWidth = r.constantWidth;
+    contents.val = r.contents.val;
+    inc();
+    
+    return *this;
+  }
+
+  bool isNull() const { return !constantWidth && !contents.ptr; }
+
+  // assumes non-null arguments
+  int compare(const ref &rhs) const {
+    Expr::Kind ak = getKind(), bk = rhs.getKind();
+    if (ak!=bk) 
+      return (ak < bk) ? -1 : 1;
+    if (ak==Expr::Constant) {
+      Expr::Width at = getWidth(), bt = rhs.getWidth();
+      if (at!=bt) 
+        return (at < bt) ? -1 : 1;
+      uint64_t av = getConstantValue(), bv = rhs.getConstantValue();
+      if (av<bv) {
+        return -1;
+      } else if (av>bv) {
+        return 1;
+      } else {
+        return 0;
+      }
+    } else {
+      return get()->compare(*rhs.get());
+    }
+  }
+
+  // assumes non-null arguments
+  bool operator<(const ref &rhs) const { return compare(rhs)<0; }
+  bool operator==(const ref &rhs) const { return compare(rhs)==0; }
+  bool operator!=(const ref &rhs) const { return compare(rhs)!=0; }
+};
+
+
+template<class T>
+inline std::ostream &operator<<(std::ostream &os, const ref<T> &e) {
+  if (e.isConstant()) {
+    os << e.getConstantValue();
+  } else {
+    os << *e.get();
+  }
+  return os;
+}
+
+template<class U>
+U* dyn_ref_cast(ref<Expr> &src) {
+  if (src.constantWidth != ref<Expr>::NotConstant)
+    return 0;
+    
+  return dynamic_cast<U*>(src.contents.ptr);
+}
+
+template<class U>
+const U* dyn_ref_cast(const ref<Expr> &src) {
+  if (src.constantWidth != ref<Expr>::NotConstant)
+    return 0;
+    
+  return dynamic_cast<const U*>(src.contents.ptr);
+}
+
+template<class U>
+U* static_ref_cast(ref<Expr> &src) {
+  return static_cast<U*>(src.contents.ptr);
+}
+
+template<class U>
+const U* static_ref_cast(const ref<Expr> &src) {
+  return static_cast<const U*>(src.contents.ptr);
+}
+
+#endif /* KLEE_REF_H */

Added: klee/trunk/lib/Basic/BOut.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Basic/BOut.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Basic/BOut.cpp (added)
+++ klee/trunk/lib/Basic/BOut.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,236 @@
+//===-- BOut.c ------------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Internal/ADT/BOut.h"
+
+#include <stdlib.h>
+#include <string.h>
+#include <stdio.h>
+
+#define BOUT_MAGIC "BOUT\n"
+#define BOUT_MAGIC_SIZE 5
+#define BOUT_VERSION 2
+
+/***/
+
+static int read_uint32(FILE *f, unsigned *value_out) {
+  unsigned char data[4];
+  if (fread(data, 4, 1, f)!=1)
+    return 0;
+  *value_out = (((((data[0]<<8) + data[1])<<8) + data[2])<<8) + data[3];
+  return 1;
+}
+
+static int write_uint32(FILE *f, unsigned value) {
+  unsigned char data[4];
+  data[0] = value>>24;
+  data[1] = value>>16;
+  data[2] = value>> 8;
+  data[3] = value>> 0;
+  return fwrite(data, 1, 4, f)==4;
+}
+
+static int read_string(FILE *f, char **value_out) {
+  unsigned len;
+  if (!read_uint32(f, &len))
+    return 0;
+  *value_out = (char*) malloc(len+1);
+  if (!*value_out)
+    return 0;
+  if (fread(*value_out, len, 1, f)!=1)
+    return 0;
+  (*value_out)[len] = 0;
+  return 1;
+}
+
+static int write_string(FILE *f, const char *value) {
+  unsigned len = strlen(value);
+  if (!write_uint32(f, len))
+    return 0;
+  if (fwrite(value, len, 1, f)!=1)
+    return 0;
+  return 1;
+}
+
+/***/
+
+
+unsigned bOut_getCurrentVersion() {
+  return BOUT_VERSION;
+}
+
+
+static int bOut_checkHeader(FILE *f) {
+  char header[BOUT_MAGIC_SIZE];
+  if (fread(header, BOUT_MAGIC_SIZE, 1, f)!=1)
+    return 0;
+  if (memcmp(header, BOUT_MAGIC, BOUT_MAGIC_SIZE))
+    return 0;
+  return 1;
+}
+
+int bOut_isBOutFile(const char *path) {
+  FILE *f = fopen(path, "rb");
+  int res;
+
+  if (!f)
+    return 0;
+  res = bOut_checkHeader(f);
+  fclose(f);
+  
+  return res;
+}
+
+BOut *bOut_fromFile(const char *path) {
+  FILE *f = fopen(path, "rb");
+  BOut *res = 0;
+  unsigned i, version;
+
+  if (!f) 
+    goto error;
+  if (!bOut_checkHeader(f)) 
+    goto error;
+
+  res = (BOut*) calloc(1, sizeof(*res));
+  if (!res) 
+    goto error;
+
+  if (!read_uint32(f, &version)) 
+    goto error;
+  
+  if (version > bOut_getCurrentVersion())
+    goto error;
+
+  res->version = version;
+
+  if (!read_uint32(f, &res->numArgs)) 
+    goto error;
+  res->args = (char**) calloc(res->numArgs, sizeof(*res->args));
+  if (!res->args) 
+    goto error;
+  
+  for (i=0; i<res->numArgs; i++)
+    if (!read_string(f, &res->args[i]))
+      goto error;
+
+  if (version >= 2) {
+    if (!read_uint32(f, &res->symArgvs)) 
+      goto error;
+    if (!read_uint32(f, &res->symArgvLen)) 
+      goto error;
+  }
+
+  if (!read_uint32(f, &res->numObjects))
+    goto error;
+  res->objects = (BOutObject*) calloc(res->numObjects, sizeof(*res->objects));
+  if (!res->objects)
+    goto error;
+  for (i=0; i<res->numObjects; i++) {
+    BOutObject *o = &res->objects[i];
+    if (!read_string(f, &o->name))
+      goto error;
+    if (!read_uint32(f, &o->numBytes))
+      goto error;
+    o->bytes = (unsigned char*) malloc(o->numBytes);
+    if (fread(o->bytes, o->numBytes, 1, f)!=1)
+      goto error;
+  }
+
+  fclose(f);
+
+  return res;
+ error:
+  if (res) {
+    if (res->args) {
+      for (i=0; i<res->numArgs; i++)
+        if (res->args[i])
+          free(res->args[i]);
+      free(res->args);
+    }
+    if (res->objects) {
+      for (i=0; i<res->numObjects; i++) {
+        BOutObject *bo = &res->objects[i];
+        if (bo->name)
+          free(bo->name);
+        if (bo->bytes)
+          free(bo->bytes);
+      }
+      free(res->objects);
+    }
+    free(res);
+  }
+
+  if (f) fclose(f);
+
+  return 0;
+}
+
+int bOut_toFile(BOut *bo, const char *path) {
+  FILE *f = fopen(path, "wb");
+  unsigned i;
+
+  if (!f) 
+    goto error;
+  if (fwrite(BOUT_MAGIC, strlen(BOUT_MAGIC), 1, f)!=1)
+    goto error;
+  if (!write_uint32(f, BOUT_VERSION))
+    goto error;
+      
+  if (!write_uint32(f, bo->numArgs))
+    goto error;
+  for (i=0; i<bo->numArgs; i++) {
+    if (!write_string(f, bo->args[i]))
+      goto error;
+  }
+
+  if (!write_uint32(f, bo->symArgvs))
+    goto error;
+  if (!write_uint32(f, bo->symArgvLen))
+    goto error;
+  
+  if (!write_uint32(f, bo->numObjects))
+    goto error;
+  for (i=0; i<bo->numObjects; i++) {
+    BOutObject *o = &bo->objects[i];
+    if (!write_string(f, o->name))
+      goto error;
+    if (!write_uint32(f, o->numBytes))
+      goto error;
+    if (fwrite(o->bytes, o->numBytes, 1, f)!=1)
+      goto error;
+  }
+
+  fclose(f);
+
+  return 1;
+ error:
+  if (f) fclose(f);
+  
+  return 0;
+}
+
+unsigned bOut_numBytes(BOut *bo) {
+  unsigned i, res = 0;
+  for (i=0; i<bo->numObjects; i++)
+    res += bo->objects[i].numBytes;
+  return res;
+}
+
+void bOut_free(BOut *bo) {
+  unsigned i;
+  for (i=0; i<bo->numArgs; i++)
+    free(bo->args[i]);
+  free(bo->args);
+  for (i=0; i<bo->numObjects; i++) {
+    free(bo->objects[i].name);
+    free(bo->objects[i].bytes);
+  }
+  free(bo->objects);
+  free(bo);
+}

Added: klee/trunk/lib/Basic/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Basic/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Basic/Makefile (added)
+++ klee/trunk/lib/Basic/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,16 @@
+#===-- lib/Basic/Makefile ----------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+LEVEL=../..
+
+LIBRARYNAME=kleeBasic
+DONT_BUILD_RELINKED=1
+BUILD_ARCHIVE=1
+
+include $(LEVEL)/Makefile.common

Added: klee/trunk/lib/Basic/README.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Basic/README.txt?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Basic/README.txt (added)
+++ klee/trunk/lib/Basic/README.txt Wed May 20 23:36:41 2009
@@ -0,0 +1,3 @@
+This directory holds the most basic support facilities provided for
+both the klee and kleaver libraries. The code in this directory should
+have no dependencies on LLVM or any other klee libraries.

Added: klee/trunk/lib/Basic/Statistics.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Basic/Statistics.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Basic/Statistics.cpp (added)
+++ klee/trunk/lib/Basic/Statistics.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,84 @@
+//===-- Statistics.cpp ----------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Statistics.h"
+
+#include <vector>
+
+using namespace klee;
+
+StatisticManager::StatisticManager()
+  : enabled(true),
+    globalStats(0),
+    indexedStats(0),
+    contextStats(0),
+    index(0) {
+}
+
+StatisticManager::~StatisticManager() {
+  if (globalStats) delete[] globalStats;
+  if (indexedStats) delete[] indexedStats;
+}
+
+void StatisticManager::useIndexedStats(unsigned totalIndices) {  
+  if (indexedStats) delete[] indexedStats;
+  indexedStats = new uint64_t[totalIndices * stats.size()];
+  memset(indexedStats, 0, sizeof(*indexedStats) * totalIndices * stats.size());
+}
+
+void StatisticManager::registerStatistic(Statistic &s) {
+  if (globalStats) delete[] globalStats;
+  s.id = stats.size();
+  stats.push_back(&s);
+  globalStats = new uint64_t[stats.size()];
+  memset(globalStats, 0, sizeof(*globalStats)*stats.size());
+}
+
+int StatisticManager::getStatisticID(const std::string &name) const {
+  for (unsigned i=0; i<stats.size(); i++)
+    if (stats[i]->getName() == name)
+      return i;
+  return -1;
+}
+
+Statistic *StatisticManager::getStatisticByName(const std::string &name) const {
+  for (unsigned i=0; i<stats.size(); i++)
+    if (stats[i]->getName() == name)
+      return stats[i];
+  return 0;
+}
+
+StatisticManager *klee::theStatisticManager = 0;
+
+static StatisticManager &getStatisticManager() {
+  static StatisticManager sm;
+  theStatisticManager = &sm;
+  return sm;
+}
+
+/* *** */
+
+Statistic::Statistic(const std::string &_name, 
+                     const std::string &_shortName) 
+  : name(_name), 
+    shortName(_shortName) {
+  getStatisticManager().registerStatistic(*this);
+}
+
+Statistic::~Statistic() {
+}
+
+Statistic &Statistic::operator +=(const uint64_t addend) {
+  theStatisticManager->incrementStatistic(*this, addend);
+  return *this;
+}
+
+uint64_t Statistic::getValue() const {
+  return theStatisticManager->getValue(*this);
+}

Added: klee/trunk/lib/Core/AddressSpace.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/AddressSpace.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/AddressSpace.cpp (added)
+++ klee/trunk/lib/Core/AddressSpace.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,334 @@
+//===-- AddressSpace.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "AddressSpace.h"
+#include "CoreStats.h"
+#include "Memory.h"
+#include "TimingSolver.h"
+
+#include "klee/Expr.h"
+#include "klee/TimerStatIncrementer.h"
+
+using namespace klee;
+
+///
+
+void AddressSpace::bindObject(const MemoryObject *mo, ObjectState *os) {
+  assert(os->copyOnWriteOwner==0 && "object already has owner");
+  os->copyOnWriteOwner = cowKey;
+  objects = objects.replace(std::make_pair(mo, os));
+}
+
+void AddressSpace::unbindObject(const MemoryObject *mo) {
+  objects = objects.remove(mo);
+}
+
+const ObjectState *AddressSpace::findObject(const MemoryObject *mo) const {
+  const MemoryMap::value_type *res = objects.lookup(mo);
+  
+  return res ? res->second : 0;
+}
+
+ObjectState *AddressSpace::getWriteable(const MemoryObject *mo,
+                                        const ObjectState *os) {
+  assert(!os->readOnly);
+
+  if (cowKey==os->copyOnWriteOwner) {
+    return const_cast<ObjectState*>(os);
+  } else {
+    ObjectState *n = new ObjectState(*os);
+    n->copyOnWriteOwner = cowKey;
+    objects = objects.replace(std::make_pair(mo, n));
+    return n;    
+  }
+}
+
+/// 
+
+bool AddressSpace::resolveOne(uint64_t addr64, ObjectPair &result) {
+  unsigned address = (unsigned) addr64;
+  MemoryObject hack(address);
+
+  if (const MemoryMap::value_type *res = objects.lookup_previous(&hack)) {
+    const MemoryObject *mo = res->first;
+    if ((mo->size==0 && address==mo->address) ||
+        (address - mo->address < mo->size)) {
+      result = *res;
+      return true;
+    }
+  }
+
+  return false;
+}
+
+bool AddressSpace::resolveOne(ExecutionState &state,
+                              TimingSolver *solver,
+                              ref<Expr> address,
+                              ObjectPair &result,
+                              bool &success) {
+  if (address.isConstant()) {
+    success = resolveOne(address.getConstantValue(), result);
+    return true;
+  } else {
+    TimerStatIncrementer timer(stats::resolveTime);
+
+    // try cheap search, will succeed for any inbounds pointer
+
+    ref<Expr> cex(0);
+    if (!solver->getValue(state, address, cex))
+      return false;
+    unsigned example = (unsigned) cex.getConstantValue();
+    MemoryObject hack(example);
+    const MemoryMap::value_type *res = objects.lookup_previous(&hack);
+    
+    if (res) {
+      const MemoryObject *mo = res->first;
+      if (example - mo->address < mo->size) {
+        result = *res;
+        success = true;
+        return true;
+      }
+    }
+
+    // didn't work, now we have to search
+       
+    MemoryMap::iterator oi = objects.upper_bound(&hack);
+    MemoryMap::iterator begin = objects.begin();
+    MemoryMap::iterator end = objects.end();
+      
+    MemoryMap::iterator start = oi;
+    while (oi!=begin) {
+      --oi;
+      const MemoryObject *mo = oi->first;
+        
+      bool mayBeTrue;
+      if (!solver->mayBeTrue(state, 
+                             mo->getBoundsCheckPointer(address), mayBeTrue))
+        return false;
+      if (mayBeTrue) {
+        result = *oi;
+        success = true;
+        return true;
+      } else {
+        bool mustBeTrue;
+        if (!solver->mustBeTrue(state, 
+                                UgeExpr::create(address, mo->getBaseExpr()),
+                                mustBeTrue))
+          return false;
+        if (mustBeTrue)
+          break;
+      }
+    }
+
+    // search forwards
+    for (oi=start; oi!=end; ++oi) {
+      const MemoryObject *mo = oi->first;
+
+      bool mustBeTrue;
+      if (!solver->mustBeTrue(state, 
+                              UltExpr::create(address, mo->getBaseExpr()),
+                              mustBeTrue))
+        return false;
+      if (mustBeTrue) {
+        break;
+      } else {
+        bool mayBeTrue;
+
+        if (!solver->mayBeTrue(state, 
+                               mo->getBoundsCheckPointer(address),
+                               mayBeTrue))
+          return false;
+        if (mayBeTrue) {
+          result = *oi;
+          success = true;
+          return true;
+        }
+      }
+    }
+
+    success = false;
+    return true;
+  }
+}
+
+bool AddressSpace::resolve(ExecutionState &state,
+                           TimingSolver *solver, 
+                           ref<Expr> p, 
+                           ResolutionList &rl, 
+                           unsigned maxResolutions,
+                           double timeout) {
+  if (p.isConstant()) {
+    ObjectPair res;
+    if (resolveOne(p.getConstantValue(), res))
+      rl.push_back(res);
+    return false;
+  } else {
+    TimerStatIncrementer timer(stats::resolveTime);
+    uint64_t timeout_us = (uint64_t) (timeout*1000000.);
+
+    // XXX in general this isn't exactly what we want... for
+    // a multiple resolution case (or for example, a \in {b,c,0})
+    // we want to find the first object, find a cex assuming
+    // not the first, find a cex assuming not the second...
+    // etc.
+    
+    // XXX how do we smartly amortize the cost of checking to
+    // see if we need to keep searching up/down, in bad cases?
+    // maybe we don't care?
+    
+    // XXX we really just need a smart place to start (although
+    // if its a known solution then the code below is guaranteed
+    // to hit the fast path with exactly 2 queries). we could also
+    // just get this by inspection of the expr.
+    
+    ref<Expr> cex(0);
+    if (!solver->getValue(state, p, cex))
+      return true;
+    unsigned example = (unsigned) cex.getConstantValue();
+    MemoryObject hack(example);
+    
+    MemoryMap::iterator oi = objects.upper_bound(&hack);
+    MemoryMap::iterator begin = objects.begin();
+    MemoryMap::iterator end = objects.end();
+      
+    MemoryMap::iterator start = oi;
+      
+    // XXX in the common case we can save one query if we ask
+    // mustBeTrue before mayBeTrue for the first result. easy
+    // to add I just want to have a nice symbolic test case first.
+      
+    // search backwards, start with one minus because this
+    // is the object that p *should* be within, which means we
+    // get write off the end with 4 queries (XXX can be better,
+    // no?)
+    while (oi!=begin) {
+      --oi;
+      const MemoryObject *mo = oi->first;
+      if (timeout_us && timeout_us < timer.check())
+        return true;
+
+      // XXX I think there is some query wasteage here?
+      ref<Expr> inBounds = mo->getBoundsCheckPointer(p);
+      bool mayBeTrue;
+      if (!solver->mayBeTrue(state, inBounds, mayBeTrue))
+        return true;
+      if (mayBeTrue) {
+        rl.push_back(*oi);
+        
+        // fast path check
+        unsigned size = rl.size();
+        if (size==1) {
+          bool mustBeTrue;
+          if (!solver->mustBeTrue(state, inBounds, mustBeTrue))
+            return true;
+          if (mustBeTrue)
+            return false;
+        } else if (size==maxResolutions) {
+          return true;
+        }
+      }
+        
+      bool mustBeTrue;
+      if (!solver->mustBeTrue(state, 
+                              UgeExpr::create(p, mo->getBaseExpr()),
+                              mustBeTrue))
+        return true;
+      if (mustBeTrue)
+        break;
+    }
+    // search forwards
+    for (oi=start; oi!=end; ++oi) {
+      const MemoryObject *mo = oi->first;
+      if (timeout_us && timeout_us < timer.check())
+        return true;
+
+      bool mustBeTrue;
+      if (!solver->mustBeTrue(state, 
+                              UltExpr::create(p, mo->getBaseExpr()),
+                              mustBeTrue))
+        return true;
+      if (mustBeTrue)
+        break;
+      
+      // XXX I think there is some query wasteage here?
+      ref<Expr> inBounds = mo->getBoundsCheckPointer(p);
+      bool mayBeTrue;
+      if (!solver->mayBeTrue(state, inBounds, mayBeTrue))
+        return true;
+      if (mayBeTrue) {
+        rl.push_back(*oi);
+        
+        // fast path check
+        unsigned size = rl.size();
+        if (size==1) {
+          bool mustBeTrue;
+          if (!solver->mustBeTrue(state, inBounds, mustBeTrue))
+            return true;
+          if (mustBeTrue)
+            return false;
+        } else if (size==maxResolutions) {
+          return true;
+        }
+      }
+    }
+  }
+
+  return false;
+}
+
+// These two are pretty big hack so we can sort of pass memory back
+// and forth to externals. They work by abusing the concrete cache
+// store inside of the object states, which allows them to
+// transparently avoid screwing up symbolics (if the byte is symbolic
+// then its concrete cache byte isn't being used) but is just a hack.
+
+void AddressSpace::copyOutConcretes() {
+  for (MemoryMap::iterator it = objects.begin(), ie = objects.end(); 
+       it != ie; ++it) {
+    const MemoryObject *mo = it->first;
+
+    if (!mo->isUserSpecified) {
+      ObjectState *os = it->second;
+      uint8_t *address = (uint8_t*) (unsigned long) mo->address;
+
+      if (!os->readOnly)
+        memcpy(address, os->concreteStore, mo->size);
+    }
+  }
+}
+
+bool AddressSpace::copyInConcretes() {
+  for (MemoryMap::iterator it = objects.begin(), ie = objects.end(); 
+       it != ie; ++it) {
+    const MemoryObject *mo = it->first;
+
+    if (!mo->isUserSpecified) {
+      const ObjectState *os = it->second;
+      uint8_t *address = (uint8_t*) (unsigned long) mo->address;
+
+      if (memcmp(address, os->concreteStore, mo->size)!=0) {
+        if (os->readOnly) {
+          return false;
+        } else {
+          ObjectState *wos = getWriteable(mo, os);
+          memcpy(wos->concreteStore, address, mo->size);
+        }
+      }
+    }
+  }
+
+  return true;
+}
+
+/***/
+
+bool MemoryObjectLT::operator()(const MemoryObject *a, const MemoryObject *b) const {
+  return a->address < b->address;
+}
+

Added: klee/trunk/lib/Core/AddressSpace.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/AddressSpace.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/AddressSpace.h (added)
+++ klee/trunk/lib/Core/AddressSpace.h Wed May 20 23:36:41 2009
@@ -0,0 +1,131 @@
+//===-- AddressSpace.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_ADDRESSSPACE_H
+#define KLEE_ADDRESSSPACE_H
+
+#include "ObjectHolder.h"
+
+#include "klee/Expr.h"
+#include "klee/Internal/ADT/ImmutableMap.h"
+
+namespace klee {
+  class ExecutionState;
+  class MemoryObject;
+  class ObjectState;
+  class TimingSolver;
+
+  template<class T> class ref;
+
+  typedef std::pair<const MemoryObject*, const ObjectState*> ObjectPair;
+  typedef std::vector<ObjectPair> ResolutionList;  
+
+  /// Function object ordering MemoryObject's by address.
+  struct MemoryObjectLT {
+    bool operator()(const MemoryObject *a, const MemoryObject *b) const;
+  };
+  
+  typedef ImmutableMap<const MemoryObject*, ObjectHolder, MemoryObjectLT> MemoryMap;
+  
+  class AddressSpace {
+  private:
+    /// Epoch counter used to control ownership of objects.
+    mutable unsigned cowKey;
+
+    /// Unsupported, use copy constructor
+    AddressSpace &operator=(const AddressSpace&); 
+    
+  public:
+    /// The MemoryObject -> ObjectState map that constitutes the
+    /// address space.
+    ///
+    /// The set of objects where o->copyOnWriteOwner == cowKey are the
+    /// objects that we own.
+    ///
+    /// \invariant forall o in objects, o->copyOnWriteOwner <= cowKey
+    MemoryMap objects;
+    
+  public:
+    AddressSpace() : cowKey(1) {}
+    AddressSpace(const AddressSpace &b) : cowKey(++b.cowKey), objects(b.objects) { }
+    ~AddressSpace() {}
+
+    /// Resolve address to an ObjectPair in result.
+    /// \return true iff an object was found.
+    bool resolveOne(uint64_t address, 
+                    ObjectPair &result);
+
+    /// Resolve address to an ObjectPair in result.
+    ///
+    /// \param state The state this address space is part of.
+    /// \param solver A solver used to determine possible 
+    ///               locations of the \a address.
+    /// \param address The address to search for.
+    /// \param[out] result An ObjectPair this address can resolve to 
+    ///               (when returning true).
+    /// \return true iff an object was found at \a address.
+    bool resolveOne(ExecutionState &state, 
+                    TimingSolver *solver,
+                    ref<Expr> address,
+                    ObjectPair &result,
+                    bool &success);
+
+    /// Resolve address to a list of ObjectPairs it can point to. If
+    /// maxResolutions is non-zero then no more than that many pairs
+    /// will be returned. 
+    ///
+    /// \return true iff the resolution is incomplete (maxResolutions
+    /// is non-zero and the search terminated early, or a query timed out).
+    bool resolve(ExecutionState &state,
+                 TimingSolver *solver,
+                 ref<Expr> address, 
+                 ResolutionList &rl, 
+                 unsigned maxResolutions=0,
+                 double timeout=0.);
+
+    /***/
+
+    /// Add a binding to the address space.
+    void bindObject(const MemoryObject *mo, ObjectState *os);
+
+    /// Remove a binding from the address space.
+    void unbindObject(const MemoryObject *mo);
+
+    /// Lookup a binding from a MemoryObject.
+    const ObjectState *findObject(const MemoryObject *mo) const;
+
+    /// \brief Obtain an ObjectState suitable for writing.
+    ///
+    /// This returns a writeable object state, creating a new copy of
+    /// the given ObjectState if necessary. If the address space owns
+    /// the ObjectState then this routine effectively just strips the
+    /// const qualifier it.
+    ///
+    /// \param mo The MemoryObject to get a writeable ObjectState for.
+    /// \param os The current binding of the MemoryObject.
+    /// \return A writeable ObjectState (\a os or a copy).
+    ObjectState *getWriteable(const MemoryObject *mo, const ObjectState *os);
+
+    /// Copy the concrete values of all managed ObjectStates into the
+    /// actual system memory location they were allocated at.
+    void copyOutConcretes();
+
+    /// Copy the concrete values of all managed ObjectStates back from
+    /// the actual system memory location they were allocated
+    /// at. ObjectStates will only be written to (and thus,
+    /// potentially copied) if the memory values are different from
+    /// the current concrete values.
+    ///
+    /// \retval true The copy succeeded. 
+    /// \retval false The copy failed because a read-only object was modified.
+    bool copyInConcretes();
+  };
+} // End klee namespace
+
+#endif

Added: klee/trunk/lib/Core/CallPathManager.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/CallPathManager.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/CallPathManager.cpp (added)
+++ klee/trunk/lib/Core/CallPathManager.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,103 @@
+//===-- CallPathManager.cpp -----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "CallPathManager.h"
+
+#include "klee/Statistics.h"
+
+#include <map>
+#include <vector>
+#include "llvm/Function.h"
+#include "llvm/Support/Streams.h"
+
+using namespace llvm;
+using namespace klee;
+
+///
+
+CallPathNode::CallPathNode(CallPathNode *_parent, 
+                           Instruction *_callSite,
+                           Function *_function)
+  : parent(_parent),
+    callSite(_callSite),
+    function(_function),
+    count(0) {
+}
+
+void CallPathNode::print() {
+  llvm::cerr << "  (Function: " << this->function->getName() << ", "
+	     << "Callsite: " << callSite << ", "
+	     << "Count: " << this->count << ")";
+  if (parent && parent->callSite) {
+    llvm::cerr << ";\n";
+    parent->print();
+  }
+  else llvm::cerr << "\n";
+}
+
+///
+
+CallPathManager::CallPathManager() : root(0, 0, 0) {
+}
+
+CallPathManager::~CallPathManager() {
+  for (std::vector<CallPathNode*>::iterator it = paths.begin(),
+         ie = paths.end(); it != ie; ++it)
+    delete *it;
+}
+
+void CallPathManager::getSummaryStatistics(CallSiteSummaryTable &results) {
+  results.clear();
+
+  for (std::vector<CallPathNode*>::iterator it = paths.begin(),
+         ie = paths.end(); it != ie; ++it)
+    (*it)->summaryStatistics = (*it)->statistics;
+
+  // compute summary bottom up, while building result table
+  for (std::vector<CallPathNode*>::reverse_iterator it = paths.rbegin(),
+         ie = paths.rend(); it != ie; ++it) {
+    CallPathNode *cp = *it;
+    cp->parent->summaryStatistics += cp->summaryStatistics;
+
+    CallSiteInfo &csi = results[cp->callSite][cp->function];
+    csi.count += cp->count;
+    csi.statistics += cp->summaryStatistics;
+  }
+}
+
+
+CallPathNode *CallPathManager::computeCallPath(CallPathNode *parent, 
+                                               Instruction *cs,
+                                               Function *f) {
+  for (CallPathNode *p=parent; p; p=p->parent)
+    if (cs==p->callSite && f==p->function)
+      return p;
+  
+  CallPathNode *cp = new CallPathNode(parent, cs, f);
+  paths.push_back(cp);
+  return cp;
+}
+
+CallPathNode *CallPathManager::getCallPath(CallPathNode *parent, 
+                                           Instruction *cs,
+                                           Function *f) {
+  std::pair<Instruction*,Function*> key(cs, f);
+  if (!parent)
+    parent = &root;
+  
+  CallPathNode::children_ty::iterator it = parent->children.find(key);
+  if (it==parent->children.end()) {
+    CallPathNode *cp = computeCallPath(parent, cs, f);
+    parent->children.insert(std::make_pair(key, cp));
+    return cp;
+  } else {
+    return it->second;
+  }
+}
+

Added: klee/trunk/lib/Core/CallPathManager.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/CallPathManager.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/CallPathManager.h (added)
+++ klee/trunk/lib/Core/CallPathManager.h Wed May 20 23:36:41 2009
@@ -0,0 +1,83 @@
+//===-- CallPathManager.h ---------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_CALLPATHMANAGER_H__
+#define __UTIL_CALLPATHMANAGER_H__
+
+#include "klee/Statistics.h"
+
+#include <map>
+#include <vector>
+
+namespace llvm {
+  class Instruction;
+  class Function;
+}
+
+namespace klee {
+  class StatisticRecord;
+
+  struct CallSiteInfo {
+    unsigned count;
+    StatisticRecord statistics;
+
+  public:
+    CallSiteInfo() : count(0) {}
+  };
+
+  typedef std::map<llvm::Instruction*,
+                   std::map<llvm::Function*, CallSiteInfo> > CallSiteSummaryTable;    
+  
+  class CallPathNode {
+    friend class CallPathManager;
+
+  public:
+    typedef std::map<std::pair<llvm::Instruction*, 
+                               llvm::Function*>, CallPathNode*> children_ty;
+
+    // form list of (callSite,function) path
+    CallPathNode *parent;
+    llvm::Instruction *callSite;
+    llvm::Function *function;
+    children_ty children;
+
+    StatisticRecord statistics;
+    StatisticRecord summaryStatistics;
+    unsigned count;
+
+  public:
+    CallPathNode(CallPathNode *parent, 
+                 llvm::Instruction *callSite,
+                 llvm::Function *function);
+
+    void print();
+  };
+
+  class CallPathManager {
+    CallPathNode root;
+    std::vector<CallPathNode*> paths;
+
+  private:
+    CallPathNode *computeCallPath(CallPathNode *parent, 
+                                  llvm::Instruction *callSite,
+                                  llvm::Function *f);
+    
+  public:
+    CallPathManager();
+    ~CallPathManager();
+
+    void getSummaryStatistics(CallSiteSummaryTable &result);
+    
+    CallPathNode *getCallPath(CallPathNode *parent, 
+                              llvm::Instruction *callSite,
+                              llvm::Function *f);
+  };
+}
+
+#endif

Added: klee/trunk/lib/Core/Common.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Common.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Common.cpp (added)
+++ klee/trunk/lib/Core/Common.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,110 @@
+//===-- Common.cpp --------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <assert.h>
+#include <string.h>
+
+#include <set>
+
+using namespace klee;
+
+FILE* klee::klee_warning_file = NULL;
+FILE* klee::klee_message_file = NULL;
+
+
+/* Prints a message/warning.
+   
+   If pfx is NULL, this is a regular message, and it's sent to
+   klee_message_file (messages.txt).  Otherwise, it is sent to 
+   klee_warning_file (warnings.txt).
+
+   Iff onlyToFile is false, the message is also printed on stderr.
+*/
+static void klee_vmessage(const char *pfx, bool onlyToFile, const char *msg, va_list ap) {
+  FILE *f = stderr;
+  if (!onlyToFile) {
+    fprintf(f, "KLEE: ");
+    if (pfx) fprintf(f, "%s: ", pfx);
+    vfprintf(f, msg, ap);
+    fprintf(f, "\n");
+    fflush(f);
+  }
+
+  if (pfx == NULL)
+    f = klee_message_file;
+  else f = klee_warning_file;
+    
+  if (f) {
+    fprintf(f, "KLEE: ");
+    if (pfx) fprintf(f, "%s: ", pfx);
+    vfprintf(f, msg, ap);
+    fprintf(f, "\n");
+    fflush(f);
+  }
+}
+
+
+void klee::klee_message(const char *msg, ...) {
+  va_list ap;
+  va_start(ap, msg);
+  klee_vmessage(NULL, false, msg, ap);
+  va_end(ap);
+}
+
+/* Message to be written only to file */
+void klee::klee_message_to_file(const char *msg, ...) {
+  va_list ap;
+  va_start(ap, msg);
+  klee_vmessage(NULL, true, msg, ap);
+  va_end(ap);
+}
+
+void klee::klee_error(const char *msg, ...) {
+  va_list ap;
+  va_start(ap, msg);
+  klee_vmessage("ERROR", false, msg, ap);
+  va_end(ap);
+  exit(1);
+}
+
+void klee::klee_warning(const char *msg, ...) {
+  va_list ap;
+  va_start(ap, msg);
+  klee_vmessage("WARNING", false, msg, ap);
+  va_end(ap);
+}
+
+
+/* Prints a warning once per message. */
+void klee::klee_warning_once(const void *id, const char *msg, ...) {
+  static std::set< std::pair<const void*, const char*> > keys;
+  std::pair<const void*, const char*> key;
+
+
+  /* "calling external" messages contain the actual arguments with
+     which we called the external function, so we need to ignore them
+     when computing the key. */
+  if (strncmp(msg, "calling external", strlen("calling external")) != 0)
+    key = std::make_pair(id, msg);
+  else key = std::make_pair(id, "calling external");
+  
+  if (!keys.count(key)) {
+    keys.insert(key);
+    
+    va_list ap;
+    va_start(ap, msg);
+    klee_vmessage("WARNING", false, msg, ap);
+    va_end(ap);
+  }
+}

Added: klee/trunk/lib/Core/Common.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Common.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Common.h (added)
+++ klee/trunk/lib/Core/Common.h Wed May 20 23:36:41 2009
@@ -0,0 +1,56 @@
+//===-- Common.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __KLEE_COMMON_H__
+#define __KLEE_COMMON_H__
+
+#ifdef __CYGWIN__
+#ifndef WINDOWS
+#define WINDOWS
+#endif
+#endif
+
+#include <stdio.h>
+
+// XXX ugh
+namespace klee {
+  class Solver;
+
+  extern FILE* klee_warning_file;
+  extern FILE* klee_message_file;
+
+  /// Print "KLEE: ERROR" followed by the msg in printf format and a
+  /// newline on stderr and to warnings.txt, then exit with an error.
+  void klee_error(const char *msg, ...)
+    __attribute__ ((format (printf, 1, 2), noreturn));
+
+  /// Print "KLEE: " followed by the msg in printf format and a
+  /// newline on stderr and to messages.txt.
+  void klee_message(const char *msg, ...)
+    __attribute__ ((format (printf, 1, 2)));
+
+  /// Print "KLEE: " followed by the msg in printf format and a
+  /// newline to messages.txt.
+  void klee_message_to_file(const char *msg, ...)
+    __attribute__ ((format (printf, 1, 2)));
+
+  /// Print "KLEE: WARNING" followed by the msg in printf format and a
+  /// newline on stderr and to warnings.txt.
+  void klee_warning(const char *msg, ...)
+    __attribute__ ((format (printf, 1, 2)));
+
+  /// Print "KLEE: WARNING" followed by the msg in printf format and a
+  /// newline on stderr and to warnings.txt. However, the warning is only 
+  /// printed once for each unique (id, msg) pair (as pointers).
+  void klee_warning_once(const void *id,
+                         const char *msg, ...)
+    __attribute__ ((format (printf, 2, 3)));
+}
+
+#endif /* __KLEE_COMMON_H__ */

Added: klee/trunk/lib/Core/CoreStats.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/CoreStats.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/CoreStats.cpp (added)
+++ klee/trunk/lib/Core/CoreStats.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,29 @@
+//===-- CoreStats.cpp -----------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "CoreStats.h"
+
+using namespace klee;
+
+Statistic stats::allocations("Allocations", "Alloc");
+Statistic stats::coveredInstructions("CoveredInstructions", "Icov");
+Statistic stats::falseBranches("FalseBranches", "Bf");
+Statistic stats::forkTime("ForkTime", "Ftime");
+Statistic stats::forks("Forks", "Forks");
+Statistic stats::instructionRealTime("InstructionRealTimes", "Ireal");
+Statistic stats::instructionTime("InstructionTimes", "Itime");
+Statistic stats::instructions("Instructions", "I");
+Statistic stats::minDistToReturn("MinDistToReturn", "Rdist");
+Statistic stats::minDistToUncovered("MinDistToUncovered", "UCdist");
+Statistic stats::reachableUncovered("ReachableUncovered", "IuncovReach");
+Statistic stats::resolveTime("ResolveTime", "Rtime");
+Statistic stats::solverTime("SolverTime", "Stime");
+Statistic stats::states("States", "States");
+Statistic stats::trueBranches("TrueBranches", "Bt");
+Statistic stats::uncoveredInstructions("UncoveredInstructions", "Iuncov");

Added: klee/trunk/lib/Core/CoreStats.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/CoreStats.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/CoreStats.h (added)
+++ klee/trunk/lib/Core/CoreStats.h Wed May 20 23:36:41 2009
@@ -0,0 +1,53 @@
+//===-- CoreStats.h ---------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_CORESTATS_H
+#define KLEE_CORESTATS_H
+
+#include "klee/Statistic.h"
+
+namespace klee {
+namespace stats {
+
+  extern Statistic allocations;
+  extern Statistic resolveTime;
+  extern Statistic instructions;
+  extern Statistic instructionTime;
+  extern Statistic instructionRealTime;
+  extern Statistic coveredInstructions;
+  extern Statistic uncoveredInstructions;  
+  extern Statistic trueBranches;
+  extern Statistic falseBranches;
+  extern Statistic forkTime;
+  extern Statistic solverTime;
+
+  /// The number of process forks.
+  extern Statistic forks;
+
+  /// Number of states, this is a "fake" statistic used by istats, it
+  /// isn't normally up-to-date.
+  extern Statistic states;
+
+  /// Instruction level statistic for tracking number of reachable
+  /// uncovered instructions.
+  extern Statistic reachableUncovered;
+
+  /// Instruction level statistic tracking the minimum intraprocedural
+  /// distance to an uncovered instruction; this is only periodically
+  /// updated.
+  extern Statistic minDistToUncovered;
+
+  /// Instruction level statistic tracking the minimum intraprocedural
+  /// distance to a function return.
+  extern Statistic minDistToReturn;
+
+}
+}
+
+#endif

Added: klee/trunk/lib/Core/ExecutionState.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ExecutionState.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ExecutionState.cpp (added)
+++ klee/trunk/lib/Core/ExecutionState.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,417 @@
+//===-- ExecutionState.cpp ------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/ExecutionState.h"
+
+#include "klee/Internal/Module/Cell.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/KModule.h"
+
+#include "klee/Expr.h"
+
+#include "Memory.h"
+
+#include "llvm/Function.h"
+#include "llvm/Support/CommandLine.h"
+
+#include <iostream>
+#include <cassert>
+#include <map>
+#include <set>
+#include <stdarg.h>
+
+using namespace llvm;
+using namespace klee;
+
+namespace { 
+  cl::opt<bool>
+  DebugLogStateMerge("debug-log-state-merge");
+}
+
+/***/
+
+StackFrame::StackFrame(KInstIterator _caller, KFunction *_kf)
+  : caller(_caller), kf(_kf), callPathNode(0), 
+    minDistToUncoveredOnReturn(0), varargs(0) {
+  locals = new Cell[kf->numRegisters];
+}
+
+StackFrame::StackFrame(const StackFrame &s) 
+  : caller(s.caller),
+    kf(s.kf),
+    callPathNode(s.callPathNode),
+    allocas(s.allocas),
+    minDistToUncoveredOnReturn(s.minDistToUncoveredOnReturn),
+    varargs(s.varargs) {
+  locals = new Cell[s.kf->numRegisters];
+  for (unsigned i=0; i<s.kf->numRegisters; i++)
+    locals[i] = s.locals[i];
+}
+
+StackFrame::~StackFrame() { 
+  delete[] locals; 
+}
+
+/***/
+
+ExecutionState::ExecutionState(KFunction *kf) 
+  : fakeState(false),
+    underConstrained(false),
+    depth(0),
+    pc(kf->instructions),
+    prevPC(pc),
+    queryCost(0.), 
+    weight(1),
+    instsSinceCovNew(0),
+    coveredNew(false),
+    forkDisabled(false),
+    ptreeNode(0) {
+  pushFrame(0, kf);
+}
+
+ExecutionState::ExecutionState(const std::vector<ref<Expr> > &assumptions) 
+  : fakeState(true),
+    underConstrained(false),
+    constraints(assumptions),
+    queryCost(0.),
+    ptreeNode(0) {
+}
+
+ExecutionState::~ExecutionState() {
+  while (!stack.empty()) popFrame();
+}
+
+ExecutionState *ExecutionState::branch() {
+  depth++;
+
+  ExecutionState *falseState = new ExecutionState(*this);
+  falseState->coveredNew = false;
+  falseState->coveredLines.clear();
+
+  weight *= .5;
+  falseState->weight -= weight;
+
+  return falseState;
+}
+
+void ExecutionState::pushFrame(KInstIterator caller, KFunction *kf) {
+  stack.push_back(StackFrame(caller,kf));
+}
+
+void ExecutionState::popFrame() {
+  StackFrame &sf = stack.back();
+  for (std::vector<const MemoryObject*>::iterator it = sf.allocas.begin(), 
+         ie = sf.allocas.end(); it != ie; ++it)
+    addressSpace.unbindObject(*it);
+  stack.pop_back();
+}
+
+///
+
+std::string ExecutionState::getFnAlias(std::string fn) {
+  std::map < std::string, std::string >::iterator it = fnAliases.find(fn);
+  if (it != fnAliases.end())
+    return it->second;
+  else return "";
+}
+
+void ExecutionState::addFnAlias(std::string old_fn, std::string new_fn) {
+  fnAliases[old_fn] = new_fn;
+}
+
+void ExecutionState::removeFnAlias(std::string fn) {
+  fnAliases.erase(fn);
+}
+
+/**/
+
+std::ostream &klee::operator<<(std::ostream &os, const MemoryMap &mm) {
+  os << "{";
+  MemoryMap::iterator it = mm.begin();
+  MemoryMap::iterator ie = mm.end();
+  if (it!=ie) {
+    os << "MO" << it->first->id << ":" << it->second;
+    for (++it; it!=ie; ++it)
+      os << ", MO" << it->first->id << ":" << it->second;
+  }
+  os << "}";
+  return os;
+}
+
+bool ExecutionState::merge(const ExecutionState &b) {
+  if (DebugLogStateMerge)
+    llvm::cerr << "-- attempting merge of A:" 
+               << this << " with B:" << &b << "--\n";
+  if (pc != b.pc)
+    return false;
+
+  // XXX is it even possible for these to differ? does it matter? probably
+  // implies difference in object states?
+  if (symbolics!=b.symbolics)
+    return false;
+
+  {
+    std::vector<StackFrame>::const_iterator itA = stack.begin();
+    std::vector<StackFrame>::const_iterator itB = b.stack.begin();
+    while (itA!=stack.end() && itB!=b.stack.end()) {
+      // XXX vaargs?
+      if (itA->caller!=itB->caller || itA->kf!=itB->kf)
+        return false;
+      ++itA;
+      ++itB;
+    }
+    if (itA!=stack.end() || itB!=b.stack.end())
+      return false;
+  }
+
+  std::set< ref<Expr> > aConstraints(constraints.begin(), constraints.end());
+  std::set< ref<Expr> > bConstraints(b.constraints.begin(), 
+                                     b.constraints.end());
+  std::set< ref<Expr> > commonConstraints, aSuffix, bSuffix;
+  std::set_intersection(aConstraints.begin(), aConstraints.end(),
+                        bConstraints.begin(), bConstraints.end(),
+                        std::inserter(commonConstraints, commonConstraints.begin()));
+  std::set_difference(aConstraints.begin(), aConstraints.end(),
+                      commonConstraints.begin(), commonConstraints.end(),
+                      std::inserter(aSuffix, aSuffix.end()));
+  std::set_difference(bConstraints.begin(), bConstraints.end(),
+                      commonConstraints.begin(), commonConstraints.end(),
+                      std::inserter(bSuffix, bSuffix.end()));
+  if (DebugLogStateMerge) {
+    llvm::cerr << "\tconstraint prefix: [";
+    for (std::set< ref<Expr> >::iterator it = commonConstraints.begin(), 
+           ie = commonConstraints.end(); it != ie; ++it)
+      llvm::cerr << *it << ", ";
+    llvm::cerr << "]\n";
+    llvm::cerr << "\tA suffix: [";
+    for (std::set< ref<Expr> >::iterator it = aSuffix.begin(), 
+           ie = aSuffix.end(); it != ie; ++it)
+      llvm::cerr << *it << ", ";
+    llvm::cerr << "]\n";
+    llvm::cerr << "\tB suffix: [";
+    for (std::set< ref<Expr> >::iterator it = bSuffix.begin(), 
+           ie = bSuffix.end(); it != ie; ++it)
+      llvm::cerr << *it << ", ";
+    llvm::cerr << "]\n";
+  }
+
+  // We cannot merge if addresses would resolve differently in the
+  // states. This means:
+  // 
+  // 1. Any objects created since the branch in either object must
+  // have been free'd.
+  //
+  // 2. We cannot have free'd any pre-existing object in one state
+  // and not the other
+
+  if (DebugLogStateMerge) {
+    llvm::cerr << "\tchecking object states\n";
+    llvm::cerr << "A: " << addressSpace.objects << "\n";
+    llvm::cerr << "B: " << b.addressSpace.objects << "\n";
+  }
+    
+  std::set<const MemoryObject*> mutated;
+  MemoryMap::iterator ai = addressSpace.objects.begin();
+  MemoryMap::iterator bi = b.addressSpace.objects.begin();
+  MemoryMap::iterator ae = addressSpace.objects.end();
+  MemoryMap::iterator be = b.addressSpace.objects.end();
+  for (; ai!=ae && bi!=be; ++ai, ++bi) {
+    if (ai->first != bi->first) {
+      if (DebugLogStateMerge) {
+        if (ai->first < bi->first) {
+          llvm::cerr << "\t\tB misses binding for: " << ai->first->id << "\n";
+        } else {
+          llvm::cerr << "\t\tA misses binding for: " << bi->first->id << "\n";
+        }
+      }
+      return false;
+    }
+    if (ai->second != bi->second) {
+      if (DebugLogStateMerge)
+        llvm::cerr << "\t\tmutated: " << ai->first->id << "\n";
+      mutated.insert(ai->first);
+    }
+  }
+  if (ai!=ae || bi!=be) {
+    if (DebugLogStateMerge)
+      llvm::cerr << "\t\tmappings differ\n";
+    return false;
+  }
+  
+  // merge stack
+
+  ref<Expr> inA(1, Expr::Bool), inB(1, Expr::Bool);
+  for (std::set< ref<Expr> >::iterator it = aSuffix.begin(), 
+         ie = aSuffix.end(); it != ie; ++it)
+    inA = AndExpr::create(inA, *it);
+  for (std::set< ref<Expr> >::iterator it = bSuffix.begin(), 
+         ie = bSuffix.end(); it != ie; ++it)
+    inB = AndExpr::create(inB, *it);
+
+  // XXX should we have a preference as to which predicate to use?
+  // it seems like it can make a difference, even though logically
+  // they must contradict each other and so inA => !inB
+
+  std::vector<StackFrame>::iterator itA = stack.begin();
+  std::vector<StackFrame>::const_iterator itB = b.stack.begin();
+  for (; itA!=stack.end(); ++itA, ++itB) {
+    StackFrame &af = *itA;
+    const StackFrame &bf = *itB;
+    for (unsigned i=0; i<af.kf->numRegisters; i++) {
+      ref<Expr> &av = af.locals[i].value;
+      const ref<Expr> &bv = bf.locals[i].value;
+      if (av.isNull() || bv.isNull()) {
+        // if one is null then by implication (we are at same pc)
+        // we cannot reuse this local, so just ignore
+      } else {
+        av = SelectExpr::create(inA, av, bv);
+      }
+    }
+  }
+
+  for (std::set<const MemoryObject*>::iterator it = mutated.begin(), 
+         ie = mutated.end(); it != ie; ++it) {
+    const MemoryObject *mo = *it;
+    const ObjectState *os = addressSpace.findObject(mo);
+    const ObjectState *otherOS = b.addressSpace.findObject(mo);
+    assert(os && !os->readOnly && 
+           "objects mutated but not writable in merging state");
+    assert(otherOS);
+
+    ObjectState *wos = addressSpace.getWriteable(mo, os);
+    for (unsigned i=0; i<mo->size; i++) {
+      ref<Expr> av = wos->read8(i);
+      ref<Expr> bv = otherOS->read8(i);
+      wos->write(i, SelectExpr::create(inA, av, bv));
+    }
+  }
+
+  constraints = ConstraintManager();
+  for (std::set< ref<Expr> >::iterator it = commonConstraints.begin(), 
+         ie = commonConstraints.end(); it != ie; ++it)
+    constraints.addConstraint(*it);
+  constraints.addConstraint(OrExpr::create(inA, inB));
+
+  return true;
+}
+
+/**/
+
+/*
+   Used for tainting: create a clone of os that we can revirt to with
+   the behavior that all constraints are preserved, but writes are 
+   discarded.  When we revirt it will be at the same address.
+ */
+ObjectState *ExecutionState::cloneObject(ObjectState *os, 
+                                         MemoryObject *mo) {
+  MemoryMap::iterator it = shadowObjects.find(mo);
+  if (it != shadowObjects.end())
+    assert(0 && "Cannot exist already!");
+
+  llvm::cerr << "DRE: Inserting a cloned object: " << mo << "\n";
+  shadowObjects = shadowObjects.replace(std::make_pair(mo, os));
+  os = new ObjectState(*os);
+  addressSpace.bindObject(mo, os);
+  return os;
+}
+
+/***/
+
+
+ExecutionTraceEvent::ExecutionTraceEvent(ExecutionState& state, 
+                                         KInstruction* ki)
+  : consecutiveCount(1) 
+{
+  file = ki->info->file;
+  line = ki->info->line;
+  funcName = state.stack.back().kf->function->getName();
+  stackDepth = state.stack.size();
+}
+
+bool ExecutionTraceEvent::ignoreMe() const {
+  // ignore all events occurring in certain pesky uclibc files:
+  if (file.find("libc/stdio/") != std::string::npos) {
+    return true;
+  }
+
+  return false;
+}
+
+void ExecutionTraceEvent::print(std::ostream &os) const {
+  os.width(stackDepth);
+  os << ' ';
+  printDetails(os);
+  os << ' ' << file << ':' << line << ':' << funcName;
+  if (consecutiveCount > 1)
+    os << " (" << consecutiveCount << "x)\n";
+  else
+    os << '\n';
+}
+
+
+bool ExecutionTraceEventEquals(ExecutionTraceEvent* e1, ExecutionTraceEvent* e2) {
+  // first see if their base class members are identical:
+  if (!((e1->file == e2->file) &&
+        (e1->line == e2->line) &&
+        (e1->funcName == e2->funcName)))
+    return false;
+
+  // fairly ugly, but i'm no OOP master, so this is the way i'm
+  // doing it for now ... lemme know if there's a cleaner way:
+  BranchTraceEvent* be1 = dynamic_cast<BranchTraceEvent*>(e1);
+  BranchTraceEvent* be2 = dynamic_cast<BranchTraceEvent*>(e2);
+  if (be1 && be2) {
+    return ((be1->trueTaken == be2->trueTaken) &&
+            (be1->canForkGoBothWays == be2->canForkGoBothWays));
+  }
+
+  // don't tolerate duplicates in anything else:
+  return false;
+}
+
+
+void BranchTraceEvent::printDetails(std::ostream &os) const {
+  os << "BRANCH " << (trueTaken ? "T" : "F") << ' ' <<
+        (canForkGoBothWays ? "2-way" : "1-way");
+}
+
+void ExecutionTraceManager::addEvent(ExecutionTraceEvent* evt) {
+  // don't trace anything before __user_main, except for global events
+  if (!hasSeenUserMain) {
+    if (evt->funcName == "__user_main") {
+      hasSeenUserMain = true;
+    }
+    else if (evt->funcName != "global_def") {
+      return;
+    }
+  }
+
+  // custom ignore events:
+  if (evt->ignoreMe())
+    return;
+
+  if (events.size() > 0) {
+    // compress consecutive duplicates:
+    ExecutionTraceEvent* last = events.back();
+    if (ExecutionTraceEventEquals(last, evt)) {
+      last->consecutiveCount++;
+      return;
+    }
+  }
+
+  events.push_back(evt);
+}
+
+void ExecutionTraceManager::printAllEvents(std::ostream &os) const {
+  for (unsigned i = 0; i != events.size(); ++i)
+    events[i]->print(os);
+}
+
+/***/

Added: klee/trunk/lib/Core/Executor.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Executor.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Executor.cpp (added)
+++ klee/trunk/lib/Core/Executor.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,3260 @@
+//===-- Executor.cpp ------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "Executor.h"
+ 
+#include "CoreStats.h"
+#include "ExternalDispatcher.h"
+#include "ImpliedValue.h"
+#include "Memory.h"
+#include "MemoryManager.h"
+#include "PTree.h"
+#include "Searcher.h"
+#include "SeedInfo.h"
+#include "SpecialFunctionHandler.h"
+#include "StatsTracker.h"
+#include "TimingSolver.h"
+#include "UserSearcher.h"
+#include "../Solver/SolverStats.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Expr.h"
+#include "klee/Interpreter.h"
+#include "klee/Machine.h"
+#include "klee/TimerStatIncrementer.h"
+#include "klee/util/Assignment.h"
+#include "klee/util/ExprPPrinter.h"
+#include "klee/util/ExprUtil.h"
+#include "klee/Config/config.h"
+#include "klee/Internal/ADT/BOut.h"
+#include "klee/Internal/ADT/RNG.h"
+#include "klee/Internal/Module/Cell.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/KModule.h"
+#include "klee/Internal/Support/FloatEvaluation.h"
+#include "klee/Internal/System/Time.h"
+
+#include "llvm/Attributes.h"
+#include "llvm/BasicBlock.h"
+#include "llvm/Constants.h"
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/Module.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/GetElementPtrTypeIterator.h"
+#include "llvm/System/Process.h"
+#include "llvm/Target/TargetData.h"
+
+#include <cassert>
+#include <algorithm>
+#include <iostream>
+#include <iomanip>
+#include <fstream>
+#include <sstream>
+#include <vector>
+#include <string>
+
+#include <sys/mman.h>
+
+#include <errno.h>
+#include <cxxabi.h>
+
+using namespace llvm;
+using namespace klee;
+
+// omg really hard to share cl opts across files ...
+bool WriteTraces = false;
+
+namespace {
+  cl::opt<bool>
+  DumpStatesOnHalt("dump-states-on-halt",
+                   cl::init(true));
+ 
+  cl::opt<bool>
+  NoPreferCex("no-prefer-cex",
+              cl::init(false));
+ 
+  cl::opt<bool>
+  UseAsmAddresses("use-asm-addresses",
+                  cl::init(false));
+ 
+  cl::opt<bool>
+  RandomizeFork("randomize-fork",
+                cl::init(false));
+ 
+  cl::opt<bool>
+  AllowExternalSymCalls("allow-external-sym-calls",
+                        cl::init(false));
+
+  cl::opt<bool>
+  DebugPrintInstructions("debug-print-instructions", 
+                         cl::desc("Print instructions during execution."));
+
+  cl::opt<bool>
+  DebugCheckForImpliedValues("debug-check-for-implied-values");
+
+
+  cl::opt<bool>
+  SimplifySymIndices("simplify-sym-indices",
+                     cl::init(false));
+
+  cl::opt<unsigned>
+  MaxSymArraySize("max-sym-array-size",
+                  cl::init(0));
+
+  cl::opt<bool>
+  DebugValidateSolver("debug-validate-solver",
+		      cl::init(false));
+
+  cl::opt<bool>
+  SuppressExternalWarnings("suppress-external-warnings");
+
+  cl::opt<bool>
+  AllExternalWarnings("all-external-warnings");
+
+  cl::opt<bool>
+  OnlyOutputStatesCoveringNew("only-output-states-covering-new",
+                              cl::init(false));
+
+  cl::opt<bool>
+  AlwaysOutputSeeds("always-output-seeds",
+                              cl::init(true));
+
+  cl::opt<bool>
+  UseFastCexSolver("use-fast-cex-solver",
+		   cl::init(false));
+
+  cl::opt<bool>
+  UseIndependentSolver("use-independent-solver",
+                       cl::init(true),
+		       cl::desc("Use constraint independence"));
+
+  cl::opt<bool>
+  EmitAllErrors("emit-all-errors",
+                cl::init(false),
+                cl::desc("Generate tests cases for all errors "
+                         "(default=one per (error,instruction) pair)"));
+
+  cl::opt<bool>
+  UseCexCache("use-cex-cache",
+              cl::init(true),
+	      cl::desc("Use counterexample caching"));
+
+  cl::opt<bool>
+  UseQueryLog("use-query-log",
+              cl::init(false));
+
+  cl::opt<bool>
+  UseQueryPCLog("use-query-pc-log",
+                cl::init(false));
+  
+  cl::opt<bool>
+  UseSTPQueryPCLog("use-stp-query-pc-log",
+                   cl::init(false));
+
+  cl::opt<bool>
+  NoExternals("no-externals", 
+           cl::desc("Do not allow external functin calls"));
+
+  cl::opt<bool>
+  UseCache("use-cache",
+	   cl::init(true),
+	   cl::desc("Use validity caching"));
+
+  cl::opt<bool>
+  OnlyReplaySeeds("only-replay-seeds", 
+                  cl::desc("Discard states that do not have a seed."));
+ 
+  cl::opt<bool>
+  OnlySeed("only-seed", 
+           cl::desc("Stop execution after seeding is done without doing regular search."));
+ 
+  cl::opt<bool>
+  AllowSeedExtension("allow-seed-extension", 
+                     cl::desc("Allow extra (unbound) values to become symbolic during seeding."));
+ 
+  cl::opt<bool>
+  ZeroSeedExtension("zero-seed-extension");
+ 
+  cl::opt<bool>
+  AllowSeedTruncation("allow-seed-truncation", 
+                      cl::desc("Allow smaller buffers than in seeds."));
+ 
+  cl::opt<bool>
+  NamedSeedMatching("named-seed-matching",
+                    cl::desc("Use names to match symbolic objects to inputs."));
+
+  cl::opt<double>
+  MaxStaticForkPct("max-static-fork-pct", cl::init(1.));
+  cl::opt<double>
+  MaxStaticSolvePct("max-static-solve-pct", cl::init(1.));
+  cl::opt<double>
+  MaxStaticCPForkPct("max-static-cpfork-pct", cl::init(1.));
+  cl::opt<double>
+  MaxStaticCPSolvePct("max-static-cpsolve-pct", cl::init(1.));
+
+  cl::opt<double>
+  MaxInstructionTime("max-instruction-time",
+                     cl::desc("Only allow a single instruction to take this much time (default=0 (off))"),
+                     cl::init(0));
+  
+  cl::opt<double>
+  SeedTime("seed-time",
+           cl::desc("Amount of time to dedicate to seeds, before normal search (default=0 (off))"),
+           cl::init(0));
+  
+  cl::opt<double>
+  MaxSTPTime("max-stp-time",
+             cl::desc("Maximum amount of time for a single query (default=120s)"),
+             cl::init(120.0));
+  
+  cl::opt<unsigned int>
+  StopAfterNInstructions("stop-after-n-instructions",
+                         cl::desc("Stop execution after specified number of instructions (0=off)"),
+                         cl::init(0));
+  
+  cl::opt<unsigned>
+  MaxForks("max-forks",
+           cl::desc("Only fork this many times (-1=off)"),
+           cl::init(~0u));
+  
+  cl::opt<unsigned>
+  MaxDepth("max-depth",
+           cl::desc("Only allow this many symbolic branches (0=off)"),
+           cl::init(0));
+  
+  cl::opt<unsigned>
+  MaxMemory("max-memory",
+            cl::desc("Refuse to fork when more above this about of memory (in MB, 0=off)"),
+            cl::init(0));
+
+  cl::opt<bool>
+  MaxMemoryInhibit("max-memory-inhibit",
+            cl::desc("Inhibit forking at memory cap (vs. random terminat)"),
+            cl::init(true));
+
+  // use 'external storage' because also needed by tools/klee/main.cpp
+  cl::opt<bool, true>
+  WriteTracesProxy("write-traces", 
+           cl::desc("Write .trace file for each terminated state"),
+           cl::location(WriteTraces),
+           cl::init(false));
+
+  cl::opt<bool>
+  UseForkedSTP("use-forked-stp", 
+                 cl::desc("Run STP in forked process"));
+}
+
+
+static void *theMMap = 0;
+static unsigned theMMapSize = 0;
+
+namespace klee {
+  RNG theRNG;
+}
+
+Solver *constructSolverChain(STPSolver *stpSolver,
+                             std::string queryLogPath,
+                             std::string stpQueryLogPath,
+                             std::string queryPCLogPath,
+                             std::string stpQueryPCLogPath) {
+  Solver *solver = stpSolver;
+
+  if (UseSTPQueryPCLog)
+    solver = createPCLoggingSolver(solver, 
+                                   stpQueryLogPath);
+
+  if (UseFastCexSolver)
+    solver = createFastCexSolver(solver);
+
+  if (UseCexCache)
+    solver = createCexCachingSolver(solver);
+
+  if (UseCache)
+    solver = createCachingSolver(solver);
+
+  if (UseIndependentSolver)
+    solver = createIndependentSolver(solver);
+
+  if (DebugValidateSolver)
+    solver = createValidatingSolver(solver, stpSolver);
+
+  if (UseQueryPCLog)
+    solver = createPCLoggingSolver(solver, 
+                                   queryPCLogPath);
+  
+  return solver;
+}
+
+Executor::Executor(const InterpreterOptions &opts,
+                   InterpreterHandler *ih) 
+  : Interpreter(opts),
+    kmodule(0),
+    interpreterHandler(ih),
+    searcher(0),
+    externalDispatcher(new ExternalDispatcher()),
+    statsTracker(0),
+    pathWriter(0),
+    symPathWriter(0),
+    specialFunctionHandler(0),
+    processTree(0),
+    replayOut(0),
+    replayPath(0),    
+    usingSeeds(0),
+    atMemoryLimit(false),
+    inhibitForking(false),
+    haltExecution(false),
+    ivcEnabled(false),
+    stpTimeout(std::min(MaxSTPTime,MaxInstructionTime)) {
+  STPSolver *stpSolver = new STPSolver(UseForkedSTP);
+  Solver *solver = 
+    constructSolverChain(stpSolver,
+                         interpreterHandler->getOutputFilename("queries.qlog"),
+                         interpreterHandler->getOutputFilename("stp-queries.qlog"),
+                         interpreterHandler->getOutputFilename("queries.pc"),
+                         interpreterHandler->getOutputFilename("stp-queries.pc"));
+  
+  this->solver = new TimingSolver(solver, stpSolver);
+
+  memory = new MemoryManager();
+}
+
+
+const Module *Executor::setModule(llvm::Module *module, 
+                                  const ModuleOptions &opts) {
+  assert(!kmodule && module && "can only register one module"); // XXX gross
+
+  kmodule = new KModule(module);
+
+  specialFunctionHandler = new SpecialFunctionHandler(*this);
+
+  specialFunctionHandler->prepare();
+  kmodule->prepare(opts, interpreterHandler);
+  specialFunctionHandler->bind();
+
+  if (StatsTracker::useStatistics()) {
+    statsTracker = 
+      new StatsTracker(*this,
+                       interpreterHandler->getOutputFilename("assembly.ll"),
+                       userSearcherRequiresMD2U());
+  }
+  
+  return module;
+}
+
+Executor::~Executor() {
+  delete memory;
+  delete externalDispatcher;
+  if (processTree)
+    delete processTree;
+  if (specialFunctionHandler)
+    delete specialFunctionHandler;
+  if (statsTracker)
+    delete statsTracker;
+  delete solver;
+  delete kmodule;
+}
+
+/***/
+
+void Executor::initializeGlobalObject(ExecutionState &state, ObjectState *os,
+                                      Constant *c, 
+                                      unsigned offset) {
+  TargetData *targetData = kmodule->targetData;
+  if (ConstantVector *cp = dyn_cast<ConstantVector>(c)) {
+    unsigned elementSize =
+      targetData->getTypeStoreSize(cp->getType()->getElementType());
+    for (unsigned i=0, e=cp->getNumOperands(); i != e; ++i)
+      initializeGlobalObject(state, os, cp->getOperand(i), 
+			     offset + i*elementSize);
+  } else if (isa<ConstantAggregateZero>(c)) {
+    unsigned i, size = targetData->getTypeStoreSize(c->getType());
+    for (i=0; i<size; i++)
+      os->write8(offset+i, (uint8_t) 0);
+  } else if (ConstantArray *ca = dyn_cast<ConstantArray>(c)) {
+    unsigned elementSize =
+      targetData->getTypeStoreSize(ca->getType()->getElementType());
+    for (unsigned i=0, e=ca->getNumOperands(); i != e; ++i)
+      initializeGlobalObject(state, os, ca->getOperand(i), 
+			     offset + i*elementSize);
+  } else if (ConstantStruct *cs = dyn_cast<ConstantStruct>(c)) {
+    const StructLayout *sl =
+      targetData->getStructLayout(cast<StructType>(cs->getType()));
+    for (unsigned i=0, e=cs->getNumOperands(); i != e; ++i)
+      initializeGlobalObject(state, os, cs->getOperand(i), 
+			     offset + sl->getElementOffset(i));
+  } else {
+    os->write(offset, evalConstant(c));
+  }
+}
+
+MemoryObject * Executor::addExternalObject(ExecutionState &state, 
+                                           void *addr, unsigned size, 
+                                           bool isReadOnly) {
+  MemoryObject *mo = memory->allocateFixed((uint64_t) (unsigned long) addr, 
+                                           size, 0);
+  ObjectState *os = bindObjectInState(state, mo, false);
+  for(unsigned i = 0; i < size; i++)
+    os->write8(i, ((uint8_t*)addr)[i]);
+  if(isReadOnly)
+    os->setReadOnly(true);  
+  return mo;
+}
+
+void Executor::initializeGlobals(ExecutionState &state) {
+  Module *m = kmodule->module;
+
+  if (m->getModuleInlineAsm() != "")
+    klee_warning("executable has module level assembly (ignoring)");
+
+  assert(m->lib_begin() == m->lib_end() &&
+         "XXX do not support dependent libraries");
+
+  // represent function globals using the address of the actual llvm function
+  // object. given that we use malloc to allocate memory in states this also
+  // ensures that we won't conflict. we don't need to allocate a memory object
+  // since reading/writing via a function pointer is unsupported anyway.
+  for (Module::iterator i = m->begin(), ie = m->end(); i != ie; ++i) {
+    Function *f = i;
+    ref<Expr> addr(0);
+
+    // If the symbol has external weak linkage then it is implicitly
+    // not defined in this module; if it isn't resolvable then it
+    // should be null.
+    if (f->hasExternalWeakLinkage() && 
+        !externalDispatcher->resolveSymbol(f->getName())) {
+      addr = Expr::createPointer(0);
+    } else {
+      addr = Expr::createPointer((unsigned long) (void*) f);
+      legalFunctions.insert(f);
+    }
+    
+    globalAddresses.insert(std::make_pair(f, addr));
+  }
+
+  // Disabled, we don't want to promote use of live externals.
+#ifdef HAVE_CTYPE_EXTERNALS
+#ifndef WINDOWS
+#ifndef DARWIN
+  /* From /usr/include/errno.h: it [errno] is a per-thread variable. */
+  int *errno_addr = __errno_location();
+  addExternalObject(state, (void *)errno_addr, sizeof *errno_addr, false);
+
+  /* from /usr/include/ctype.h:
+       These point into arrays of 384, so they can be indexed by any `unsigned
+       char' value [0,255]; by EOF (-1); or by any `signed char' value
+       [-128,-1).  ISO C requires that the ctype functions work for `unsigned */
+  const uint16_t **addr = __ctype_b_loc();
+  addExternalObject(state, (void *)(*addr-128), 
+                    384 * sizeof **addr, true);
+  addExternalObject(state, addr, 4, true);
+    
+  const int32_t **lower_addr = __ctype_tolower_loc();
+  addExternalObject(state, (void *)(*lower_addr-128), 
+                    384 * sizeof **lower_addr, true);
+  addExternalObject(state, lower_addr, 4, true);
+  
+  const int32_t **upper_addr = __ctype_toupper_loc();
+  addExternalObject(state, (void *)(*upper_addr-128), 
+                    384 * sizeof **upper_addr, true);
+  addExternalObject(state, upper_addr, 4, true);
+#endif
+#endif
+#endif
+
+  // allocate and initialize globals, done in two passes since we may
+  // need address of a global in order to initialize some other one.
+
+  // allocate memory objects for all globals
+  for (Module::const_global_iterator i = m->global_begin(),
+         e = m->global_end();
+       i != e; ++i) {
+    if (i->isDeclaration()) {
+      // FIXME: We have no general way of handling unknown external
+      // symbols. If we really cared about making external stuff work
+      // better we could support user definition, or use the EXE style
+      // hack where we check the object file information.
+
+      const Type *ty = i->getType()->getElementType();
+      const std::string &name = i->getName();
+      uint64_t size = kmodule->targetData->getTypeStoreSize(ty);
+
+      // XXX - DWD - hardcode some things until we decide how to fix.
+#ifndef WINDOWS
+      if (name == "_ZTVN10__cxxabiv117__class_type_infoE") {
+        size = 0x2C;
+      } else if (name == "_ZTVN10__cxxabiv120__si_class_type_infoE") {
+        size = 0x2C;
+      } else if (name == "_ZTVN10__cxxabiv121__vmi_class_type_infoE") {
+        size = 0x2C;
+      }
+#endif
+
+      if (size == 0) {
+        llvm::cerr << "Unable to find size for global variable: " << i->getName() 
+                   << " (use will result in out of bounds access)\n";
+      }
+
+      MemoryObject *mo = memory->allocate(size, false, true, i);
+      ObjectState *os = bindObjectInState(state, mo, false);
+      globalObjects.insert(std::make_pair(i, mo));
+      globalAddresses.insert(std::make_pair(i, mo->getBaseExpr()));
+
+      // Program already running = object already initialized.  Read
+      // concrete value and write it to our copy.
+      if (size) {
+        void *addr;
+        if (name=="__dso_handle") {
+          extern void *__dso_handle __attribute__ ((__weak__));
+          addr = &__dso_handle; // wtf ?
+        } else {
+          addr = externalDispatcher->resolveSymbol(name);
+        }
+        if (!addr)
+          klee_error("unable to load symbol(%s) while initializing globals.", 
+                     name.c_str());
+
+        for (unsigned offset=0; offset<mo->size; offset++)
+          os->write8(offset, ((unsigned char*)addr)[offset]);
+      }
+    } else {
+      const std::string &name = i->getName();
+      const Type *ty = i->getType()->getElementType();
+      uint64_t size = kmodule->targetData->getTypeStoreSize(ty);
+      MemoryObject *mo = 0;
+
+      if (UseAsmAddresses && name[0]=='\01') {
+        char *end;
+        uint64_t address = ::strtoll(name.c_str()+1, &end, 0);
+
+        if (end && *end == '\0') {
+          klee_message("NOTE: allocated global at asm specified address: %#08llx"
+                       " (%llu bytes)",
+                       address, size);
+          mo = memory->allocateFixed(address, size, &*i);
+          mo->isUserSpecified = true; // XXX hack;
+        }
+      }
+
+      if (!mo)
+        mo = memory->allocate(size, false, true, &*i);
+      assert(mo && "out of memory");
+      ObjectState *os = bindObjectInState(state, mo, false);
+      globalObjects.insert(std::make_pair(i, mo));
+      globalAddresses.insert(std::make_pair(i, mo->getBaseExpr()));
+
+      if (!i->hasInitializer())
+          os->initializeToRandom();
+    }
+  }
+  
+  // link aliases to their definitions (if bound)
+  for (Module::alias_iterator i = m->alias_begin(), ie = m->alias_end(); 
+       i != ie; ++i) {
+    // Map the alias to its aliasee's address. This works because we have
+    // addresses for everything, even undefined functions. 
+    globalAddresses.insert(std::make_pair(i, evalConstant(i->getAliasee())));
+  }
+
+  // once all objects are allocated, do the actual initialization
+  for (Module::const_global_iterator i = m->global_begin(),
+         e = m->global_end();
+       i != e; ++i) {
+    if (i->hasInitializer()) {
+      MemoryObject *mo = globalObjects.find(i)->second;
+      const ObjectState *os = state.addressSpace.findObject(mo);
+      assert(os);
+      ObjectState *wos = state.addressSpace.getWriteable(mo, os);
+      
+      initializeGlobalObject(state, wos, i->getInitializer(), 0);
+      // if(i->isConstant()) os->setReadOnly(true);
+    }
+  }
+}
+
+void Executor::branch(ExecutionState &state, 
+                      const std::vector< ref<Expr> > &conditions,
+                      std::vector<ExecutionState*> &result) {
+  TimerStatIncrementer timer(stats::forkTime);
+  unsigned N = conditions.size();
+  assert(N);
+
+  stats::forks += N-1;
+
+  // XXX do proper balance or keep random?
+  result.push_back(&state);
+  for (unsigned i=1; i<N; ++i) {
+    ExecutionState *es = result[theRNG.getInt32() % i];
+    ExecutionState *ns = es->branch();
+    addedStates.insert(ns);
+    result.push_back(ns);
+    es->ptreeNode->data = 0;
+    std::pair<PTree::Node*,PTree::Node*> res = 
+      processTree->split(es->ptreeNode, ns, es);
+    ns->ptreeNode = res.first;
+    es->ptreeNode = res.second;
+  }
+
+  // If necessary redistribute seeds to match conditions, killing
+  // states if necessary due to OnlyReplaySeeds (inefficient but
+  // simple).
+  
+  std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+    seedMap.find(&state);
+  if (it != seedMap.end()) {
+    std::vector<SeedInfo> seeds = it->second;
+    seedMap.erase(it);
+
+    // Assume each seed only satisfies one condition (necessarily true
+    // when conditions are mutually exclusive and their conjunction is
+    // a tautology).
+    for (std::vector<SeedInfo>::iterator siit = seeds.begin(), 
+           siie = seeds.end(); siit != siie; ++siit) {
+      unsigned i;
+      for (i=0; i<N; ++i) {
+        ref<Expr> res;
+        bool success = 
+          solver->getValue(state, siit->assignment.evaluate(conditions[i]), 
+                           res);
+        assert(success && "FIXME: Unhandled solver failure");
+        if (res.getConstantValue())
+          break;
+      }
+      
+      // If we didn't find a satisfying condition randomly pick one
+      // (the seed will be patched).
+      if (i==N)
+        i = theRNG.getInt32() % N;
+
+      seedMap[result[i]].push_back(*siit);
+    }
+
+    if (OnlyReplaySeeds) {
+      for (unsigned i=0; i<N; ++i) {
+        if (!seedMap.count(result[i])) {
+          terminateState(*result[i]);
+          result[i] = NULL;
+        }
+      } 
+    }
+  }
+
+  for (unsigned i=0; i<N; ++i)
+    if (result[i])
+      addConstraint(*result[i], conditions[i]);
+}
+
+Executor::StatePair 
+Executor::fork(ExecutionState &current, ref<Expr> condition, bool isInternal) {
+  Solver::Validity res;
+  std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+    seedMap.find(&current);
+  bool isSeeding = it != seedMap.end();
+
+  if (!isSeeding &&
+      !condition.isConstant() && 
+      (MaxStaticForkPct!=1. || MaxStaticSolvePct != 1. ||
+       MaxStaticCPForkPct!=1. || MaxStaticCPSolvePct != 1.) &&
+      statsTracker->elapsed() > 60.) {
+    StatisticManager &sm = *theStatisticManager;
+    CallPathNode *cpn = current.stack.back().callPathNode;
+    if ((MaxStaticForkPct<1. &&
+         sm.getIndexedValue(stats::forks, sm.getIndex()) > 
+         stats::forks*MaxStaticForkPct) ||
+        (MaxStaticCPForkPct<1. &&
+         cpn && (cpn->statistics.getValue(stats::forks) > 
+                 stats::forks*MaxStaticCPForkPct)) ||
+        (MaxStaticSolvePct<1 &&
+         sm.getIndexedValue(stats::solverTime, sm.getIndex()) > 
+         stats::solverTime*MaxStaticSolvePct) ||
+        (MaxStaticCPForkPct<1. &&
+         cpn && (cpn->statistics.getValue(stats::solverTime) > 
+                 stats::solverTime*MaxStaticCPSolvePct))) {
+      ref<Expr> value; 
+      bool success = solver->getValue(current, condition, value);
+      assert(success && "FIXME: Unhandled solver failure");
+      addConstraint(current, EqExpr::create(value, condition));
+      condition = value;
+    }      
+  }
+
+  double timeout = stpTimeout;
+  if (isSeeding)
+    timeout *= it->second.size();
+  solver->setTimeout(timeout);
+  bool success = solver->evaluate(current, condition, res);
+  solver->setTimeout(0);
+  if (!success) {
+    current.pc = current.prevPC;
+    terminateStateEarly(current, "query timed out");
+    return StatePair(0, 0);
+  }
+
+  if (!isSeeding) {
+    if (replayPath && !isInternal) {
+      assert(replayPosition<replayPath->size() &&
+             "ran out of branches in replay path mode");
+      bool branch = (*replayPath)[replayPosition++];
+      
+      if (res==Solver::True) {
+        assert(branch && "hit invalid branch in replay path mode");
+      } else if (res==Solver::False) {
+        assert(!branch && "hit invalid branch in replay path mode");
+      } else {
+        // add constraints
+        if(branch) {
+          res = Solver::True;
+          addConstraint(current, condition);
+        } else  {
+          res = Solver::False;
+          addConstraint(current, Expr::createNot(condition));
+        }
+      }
+    } else if (res==Solver::Unknown) {
+      assert(!replayOut && "in replay mode, only one branch can be true.");
+      
+      if ((MaxMemoryInhibit && atMemoryLimit) || 
+          current.forkDisabled ||
+          inhibitForking || 
+          (MaxForks!=~0u && stats::forks >= MaxForks)) {
+        TimerStatIncrementer timer(stats::forkTime);
+        if (theRNG.getBool()) {
+          addConstraint(current, condition);
+          res = Solver::True;        
+        } else {
+          addConstraint(current, Expr::createNot(condition));
+          res = Solver::False;
+        }
+      }
+    }
+  }
+
+  // Fix branch in only-replay-seed mode, if we don't have both true
+  // and false seeds.
+  if (isSeeding && 
+      (current.forkDisabled || OnlyReplaySeeds) && 
+      res == Solver::Unknown) {
+    bool trueSeed=false, falseSeed=false;
+    // Is seed extension still ok here?
+    for (std::vector<SeedInfo>::iterator siit = it->second.begin(), 
+           siie = it->second.end(); siit != siie; ++siit) {
+      ref<Expr> res;
+      bool success = 
+        solver->getValue(current, siit->assignment.evaluate(condition), res);
+      assert(success && "FIXME: Unhandled solver failure");
+      if (res.isConstant()) {
+        if (res.getConstantValue()) {
+          trueSeed = true;
+        } else {
+          falseSeed = true;
+        }
+        if (trueSeed && falseSeed)
+          break;
+      }
+    }
+    if (!(trueSeed && falseSeed)) {
+      assert(trueSeed || falseSeed);
+      
+      res = trueSeed ? Solver::True : Solver::False;
+      addConstraint(current, trueSeed ? condition : Expr::createNot(condition));
+    }
+  }
+
+
+  // XXX - even if the constraint is provable one way or the other we
+  // can probably benefit by adding this constraint and allowing it to
+  // reduce the other constraints. For example, if we do a binary
+  // search on a particular value, and then see a comparison against
+  // the value it has been fixed at, we should take this as a nice
+  // hint to just use the single constraint instead of all the binary
+  // search ones. If that makes sense.
+  if (res==Solver::True) {
+    if (!isInternal) {
+      if (pathWriter) {
+        current.pathOS << "1";
+      }
+    }
+
+    return StatePair(&current, 0);
+  } else if (res==Solver::False) {
+    if (!isInternal) {
+      if (pathWriter) {
+        current.pathOS << "0";
+      }
+    }
+
+    return StatePair(0, &current);
+  } else {
+    TimerStatIncrementer timer(stats::forkTime);
+    ExecutionState *falseState, *trueState = &current;
+
+    ++stats::forks;
+
+    falseState = trueState->branch();
+    addedStates.insert(falseState);
+
+    if (RandomizeFork && theRNG.getBool())
+      std::swap(trueState, falseState);
+
+    if (it != seedMap.end()) {
+      std::vector<SeedInfo> seeds = it->second;
+      it->second.clear();
+      std::vector<SeedInfo> &trueSeeds = seedMap[trueState];
+      std::vector<SeedInfo> &falseSeeds = seedMap[falseState];
+      for (std::vector<SeedInfo>::iterator siit = seeds.begin(), 
+             siie = seeds.end(); siit != siie; ++siit) {
+        ref<Expr> res;
+        bool success = 
+          solver->getValue(current, siit->assignment.evaluate(condition), res);
+        assert(success && "FIXME: Unhandled solver failure");
+        if (res.getConstantValue()) {
+          trueSeeds.push_back(*siit);
+        } else {
+          falseSeeds.push_back(*siit);
+        }
+      }
+      
+      bool swapInfo = false;
+      if (trueSeeds.empty()) {
+        if (&current == trueState) swapInfo = true;
+        seedMap.erase(trueState);
+      }
+      if (falseSeeds.empty()) {
+        if (&current == falseState) swapInfo = true;
+        seedMap.erase(falseState);
+      }
+      if (swapInfo) {
+        std::swap(trueState->coveredNew, falseState->coveredNew);
+        std::swap(trueState->coveredLines, falseState->coveredLines);
+      }
+    }
+
+    current.ptreeNode->data = 0;
+    std::pair<PTree::Node*, PTree::Node*> res =
+      processTree->split(current.ptreeNode, falseState, trueState);
+    falseState->ptreeNode = res.first;
+    trueState->ptreeNode = res.second;
+
+    if (!isInternal) {
+      if (pathWriter) {
+        falseState->pathOS = pathWriter->open(current.pathOS);
+        trueState->pathOS << "1";
+        falseState->pathOS << "0";
+      }      
+      if (symPathWriter) {
+        falseState->symPathOS = symPathWriter->open(current.symPathOS);
+        trueState->symPathOS << "1";
+        falseState->symPathOS << "0";
+      }
+    }
+
+    addConstraint(*trueState, condition);
+    addConstraint(*falseState, Expr::createNot(condition));
+
+    // Kinda gross, do we even really still want this option?
+    if (MaxDepth && MaxDepth<=trueState->depth) {
+      terminateStateEarly(*trueState, "max-depth exceeded");
+      terminateStateEarly(*falseState, "max-depth exceeded");
+      return StatePair(0, 0);
+    }
+
+    return StatePair(trueState, falseState);
+  }
+}
+
+void Executor::addConstraint(ExecutionState &state, ref<Expr> condition) {
+  if (condition.isConstant()) {
+    assert(condition.getConstantValue() &&
+           "attempt to add invalid constraint");
+    return;
+  }
+
+  // Check to see if this constraint violates seeds.
+  std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+    seedMap.find(&state);
+  if (it != seedMap.end()) {
+    bool warn = false;
+    for (std::vector<SeedInfo>::iterator siit = it->second.begin(), 
+           siie = it->second.end(); siit != siie; ++siit) {
+      bool res;
+      bool success = 
+        solver->mustBeFalse(state, siit->assignment.evaluate(condition), res);
+      assert(success && "FIXME: Unhandled solver failure");
+      if (res) {
+        siit->patchSeed(state, condition, solver);
+        warn = true;
+      }
+    }
+    if (warn)
+      klee_warning("seeds patched for violating constraint"); 
+  }
+
+  state.addConstraint(condition);
+  if (ivcEnabled)
+    doImpliedValueConcretization(state, condition, ref<Expr>(1, Expr::Bool));
+}
+
+ref<Expr> Executor::evalConstant(Constant *c) {
+  if (llvm::ConstantExpr *ce = dyn_cast<llvm::ConstantExpr>(c)) {
+    return evalConstantExpr(ce);
+  } else {
+    if (const ConstantInt *ci = dyn_cast<ConstantInt>(c)) {
+      switch(ci->getBitWidth()) {
+      case  1: return ConstantExpr::create(ci->getZExtValue(), Expr::Bool);
+      case  8: return ConstantExpr::create(ci->getZExtValue(), Expr::Int8);
+      case 16: return ConstantExpr::create(ci->getZExtValue(), Expr::Int16);
+      case 32: return ConstantExpr::create(ci->getZExtValue(), Expr::Int32);
+      case 64: return ConstantExpr::create(ci->getZExtValue(), Expr::Int64);
+      default:
+        assert(0 && "XXX arbitrary bit width constants unhandled");
+      }
+    } else if (const ConstantFP *cf = dyn_cast<ConstantFP>(c)) {      
+      switch(cf->getType()->getTypeID()) {
+      case Type::FloatTyID: {
+	float f = cf->getValueAPF().convertToFloat();
+	return ConstantExpr::create(floats::FloatAsUInt64(f), Expr::Int32);
+      }
+      case Type::DoubleTyID: {
+	double d = cf->getValueAPF().convertToDouble();
+	return ConstantExpr::create(floats::DoubleAsUInt64(d), Expr::Int64);
+      }	
+      case Type::X86_FP80TyID: {
+        // FIXME: This is really broken, but for now we just convert
+	// to a double. This isn't going to work at all in general,
+	// but we need support for wide constants.
+	APFloat apf = cf->getValueAPF();
+        bool ignored;
+	APFloat::opStatus r = apf.convert(APFloat::IEEEdouble, 
+                                          APFloat::rmNearestTiesToAway,
+                                          &ignored);
+        (void) r;
+        //assert(!(r & APFloat::opOverflow) && !(r & APFloat::opUnderflow) &&
+        //       "Overflow/underflow while converting from FP80 (x87) to 64-bit double");
+	double d = apf.convertToDouble();
+	return ConstantExpr::create(floats::DoubleAsUInt64(d), Expr::Int64);
+      }
+      default:		
+	llvm::cerr << "Constant of type " << cf->getType()->getDescription() 
+                   << " not supported\n";
+	llvm::cerr << "Constant used at ";
+	KConstant *kc = kmodule->getKConstant((Constant*) cf);
+	if (kc && kc->ki && kc->ki->info)
+	  llvm::cerr << kc->ki->info->file << ":" << kc->ki->info->line << "\n";
+	else llvm::cerr << "<unknown>\n";
+	  
+	assert(0 && "Arbitrary bit width floating point constants unsupported");
+      }
+    } else if (const GlobalValue *gv = dyn_cast<GlobalValue>(c)) {
+      return globalAddresses.find(gv)->second;
+    } else if (isa<ConstantPointerNull>(c)) {
+      return Expr::createPointer(0);
+    } else if (isa<UndefValue>(c)) {
+      return ConstantExpr::create(0, Expr::getWidthForLLVMType(c->getType()));
+    } else {
+      // Constant{AggregateZero,Array,Struct,Vector}
+      assert(0 && "invalid argument to evalConstant()");
+    }
+  }
+}
+
+ref<Expr> Executor::eval(KInstruction *ki,
+                         unsigned index,
+                         ExecutionState &state) {
+  assert(index < ki->inst->getNumOperands());
+  int vnumber = ki->operands[index];
+
+  // Determine if this is a constant or not.
+  if (vnumber < 0) {
+    unsigned index = -vnumber - 2;
+    Cell &c = kmodule->constantTable[index];
+    return c.value;
+  } else {
+    unsigned index = vnumber;
+    StackFrame &sf = state.stack.back();
+    Cell &c = sf.locals[index];
+    return c.value;
+  }
+}
+
+void Executor::bindLocal(KInstruction *target, ExecutionState &state, 
+                         ref<Expr> value) {
+  StackFrame &sf = state.stack.back();
+  unsigned reg = target->dest;
+  Cell &c = sf.locals[reg];
+  c.value = value;  
+}
+
+void Executor::bindArgument(KFunction *kf, unsigned index, 
+                            ExecutionState &state, ref<Expr> value) {
+  StackFrame &sf = state.stack.back();
+  unsigned reg = kf->getArgRegister(index);
+  Cell &c = sf.locals[reg];
+  c.value = value;
+}
+
+ref<Expr> Executor::toUnique(const ExecutionState &state, 
+                             ref<Expr> &e) {
+  ref<Expr> result = e;
+
+  if (!e.isConstant()) {
+    ref<Expr> value(0);
+    bool isTrue = false;
+
+    solver->setTimeout(stpTimeout);      
+    if (solver->getValue(state, e, value) &&
+        solver->mustBeTrue(state, EqExpr::create(e, value), isTrue) &&
+        isTrue)
+      result = value;
+    solver->setTimeout(0);
+  }
+  
+  return result;
+}
+
+
+/* Concretize the given expression, and return a possible constant value. 
+   'reason' is just a documentation string stating the reason for concretization. */
+ref<Expr> Executor::toConstant(ExecutionState &state, 
+                               ref<Expr> e,
+                               const char *reason) {
+  e = state.constraints.simplifyExpr(e);
+  if (!e.isConstant()) {
+    ref<Expr> value;
+    bool success = solver->getValue(state, e, value);
+    assert(success && "FIXME: Unhandled solver failure");
+    
+    std::ostringstream os;
+    os << "silently concretizing (reason: " << reason << ") expression " << e 
+       << " to value " << value 
+       << " (" << (*(state.pc)).info->file << ":" << (*(state.pc)).info->line << ")";
+      
+    if (AllExternalWarnings)
+      klee_warning(reason, os.str().c_str());
+    else
+      klee_warning_once(reason, "%s", os.str().c_str());
+
+    addConstraint(state, EqExpr::create(e, value));
+    
+    return value;
+  } else {
+    return e;
+  }
+}
+
+void Executor::executeGetValue(ExecutionState &state,
+                               ref<Expr> e,
+                               KInstruction *target) {
+  e = state.constraints.simplifyExpr(e);
+  std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+    seedMap.find(&state);
+  if (it==seedMap.end() || e.isConstant()) {
+    ref<Expr> value;
+    bool success = solver->getValue(state, e, value);
+    assert(success && "FIXME: Unhandled solver failure");
+    bindLocal(target, state, value);
+  } else {
+    std::set< ref<Expr> > values;
+    for (std::vector<SeedInfo>::iterator siit = it->second.begin(), 
+           siie = it->second.end(); siit != siie; ++siit) {
+      ref<Expr> value;
+      bool success = 
+        solver->getValue(state, siit->assignment.evaluate(e), value);
+      assert(success && "FIXME: Unhandled solver failure");
+      values.insert(value);
+    }
+    
+    std::vector< ref<Expr> > conditions;
+    for (std::set< ref<Expr> >::iterator vit = values.begin(), 
+           vie = values.end(); vit != vie; ++vit)
+      conditions.push_back(EqExpr::create(e, *vit));
+
+    std::vector<ExecutionState*> branches;
+    branch(state, conditions, branches);
+    
+    std::vector<ExecutionState*>::iterator bit = branches.begin();
+    for (std::set< ref<Expr> >::iterator vit = values.begin(), 
+           vie = values.end(); vit != vie; ++vit) {
+      ExecutionState *es = *bit;
+      if (es)
+        bindLocal(target, *es, *vit);
+      ++bit;
+    }
+  }
+}
+
+void Executor::stepInstruction(ExecutionState &state) {
+  if (DebugPrintInstructions) {
+    printFileLine(state, state.pc);
+    llvm::cerr << std::setw(10) << stats::instructions << " " << *state.pc->inst;
+  }
+
+  if (statsTracker)
+    statsTracker->stepInstruction(state);
+
+  ++stats::instructions;
+  state.prevPC = state.pc;
+  ++state.pc;
+
+  if (stats::instructions==StopAfterNInstructions)
+    haltExecution = true;
+}
+
+void Executor::executeCall(ExecutionState &state, 
+                           KInstruction *ki,
+                           Function *f,
+                           std::vector< ref<Expr> > &arguments) {
+  if (WriteTraces) {
+    // don't print out special debug stop point 'function' calls
+    if (f->getIntrinsicID() != Intrinsic::dbg_stoppoint) {
+      const std::string& calleeFuncName = f->getName();
+      state.exeTraceMgr.addEvent(new FunctionCallTraceEvent(state, ki, calleeFuncName));
+    }
+  }
+
+  Instruction *i = ki->inst;
+  if (f && f->isDeclaration()) {
+    if (f!=kmodule->dbgStopPointFn) { // special case speed hack
+      switch(f->getIntrinsicID()) {
+      case Intrinsic::dbg_stoppoint:
+      case Intrinsic::dbg_region_start:
+      case Intrinsic::dbg_region_end:
+      case Intrinsic::dbg_func_start:
+      case Intrinsic::dbg_declare:
+      case Intrinsic::not_intrinsic:
+        // state may be destroyed by this call, cannot touch
+        callExternalFunction(state, ki, f, arguments);
+        break;
+          
+        // vararg is handled by caller and intrinsic lowering,
+        // see comment for ExecutionState::varargs
+      case Intrinsic::vastart:  {
+        StackFrame &sf = state.stack.back();
+        assert(sf.varargs && 
+               "vastart called in function with no vararg object");
+        executeMemoryOperation(state, true, arguments[0], 
+                               sf.varargs->getBaseExpr(), 0);
+        break;
+      }
+      case Intrinsic::vaend:    // va_end is a noop for the interpreter
+        break;
+          
+      case Intrinsic::vacopy: // should be lowered
+      default:
+        klee_error("unknown intrinsic: %s", f->getName().c_str());
+      }
+    }
+
+    if (InvokeInst *ii = dyn_cast<InvokeInst>(i)) {
+      transferToBasicBlock(ii->getNormalDest(), i->getParent(), state);
+    }
+  } else {
+    // XXX not really happy about this reliance on prevPC but is ok I
+    // guess. This just done to avoid having to pass KInstIterator
+    // everywhere instead of the actual instruction, since we can't
+    // make a KInstIterator from just an instruction (unlike LLVM).
+    KFunction *kf = kmodule->functionMap[f];
+    state.pushFrame(state.prevPC, kf);
+    state.pc = kf->instructions;
+        
+    if (statsTracker)
+      statsTracker->framePushed(state, &state.stack[state.stack.size()-2]);
+        
+    unsigned callingArgs = arguments.size();
+    unsigned funcArgs = f->arg_size();
+    if (!f->isVarArg()) {
+      if (callingArgs > funcArgs) {
+        klee_warning_once(f, "calling %s with extra arguments.", 
+                          f->getName().c_str());
+      } else if (callingArgs < funcArgs) {
+        terminateStateOnError(state, "calling function with too few arguments", 
+                              "user.err");
+        return;
+      }
+    } else {
+      if (callingArgs < funcArgs) {
+        terminateStateOnError(state, "calling function with too few arguments", 
+                              "user.err");
+        return;
+      }
+            
+      StackFrame &sf = state.stack.back();
+      unsigned size = 0;
+      for (unsigned i = funcArgs; i < callingArgs; i++)
+        size += Expr::getMinBytesForWidth(arguments[i].getWidth());
+
+      MemoryObject *mo = sf.varargs = memory->allocate(size, true, false, 
+                                                       state.prevPC->inst);
+      if (!mo) {
+        terminateStateOnExecError(state, "out of memory (varargs)");
+        return;
+      }
+      ObjectState *os = bindObjectInState(state, mo, true);
+      unsigned offset = 0;
+      for (unsigned i = funcArgs; i < callingArgs; i++) {
+        // XXX: DRE: i think we bind memory objects here?
+        os->write(offset, arguments[i]);
+        offset += Expr::getMinBytesForWidth(arguments[i].getWidth());
+      }
+    }
+
+    unsigned numFormals = f->arg_size();
+    for (unsigned i=0; i<numFormals; ++i) 
+      bindArgument(kf, i, state, arguments[i]);
+  }
+}
+
+void Executor::transferToBasicBlock(BasicBlock *dst, BasicBlock *src, 
+                                    ExecutionState &state) {
+  // Note that in general phi nodes can reuse phi values from the same
+  // block but the incoming value is the eval() result *before* the
+  // execution of any phi nodes. this is pathological and doesn't
+  // really seem to occur, but just in case we run the PhiCleanerPass
+  // which makes sure this cannot happen and so it is safe to just
+  // eval things in order. The PhiCleanerPass also makes sure that all
+  // incoming blocks have the same order for each PHINode so we only
+  // have to compute the index once.
+  //
+  // With that done we simply set an index in the state so that PHI
+  // instructions know which argument to eval, set the pc, and continue.
+  
+  // XXX this lookup has to go ?
+  KFunction *kf = state.stack.back().kf;
+  unsigned entry = kf->basicBlockEntry[dst];
+  state.pc = &kf->instructions[entry];
+  if (state.pc->inst->getOpcode() == Instruction::PHI) {
+    PHINode *first = static_cast<PHINode*>(state.pc->inst);
+    state.incomingBBIndex = first->getBasicBlockIndex(src);
+  }
+}
+
+void Executor::printFileLine(ExecutionState &state, KInstruction *ki) {
+  const InstructionInfo &ii = *ki->info;
+  if (ii.file != "") 
+    llvm::cerr << "     " << ii.file << ":" << ii.line << ":";
+  else
+    llvm::cerr << "     [no debug info]:";
+}
+
+
+Function* Executor::getCalledFunction(CallSite &cs, ExecutionState &state) {
+  Function *f = cs.getCalledFunction();
+  
+  if (f) {
+    std::string alias = state.getFnAlias(f->getName());
+    if (alias != "") {
+      //llvm::cerr << f->getName() << "() is aliased with " << alias << "()\n";
+      llvm::Module* currModule = kmodule->module;
+      Function* old_f = f;
+      f = currModule->getFunction(alias);
+      if (!f) {
+	llvm::cerr << "Function " << alias << "(), alias for " << old_f->getName() << " not found!\n";
+	assert(f && "function alias not found");
+      }
+    }
+  }
+  
+  return f;
+}
+
+
+void Executor::executeInstruction(ExecutionState &state, KInstruction *ki) {
+  Instruction *i = ki->inst;
+  switch (i->getOpcode()) {
+    // Control flow
+  case Instruction::Ret: {
+    ReturnInst *ri = cast<ReturnInst>(i);
+    KInstIterator kcaller = state.stack.back().caller;
+    Instruction *caller = kcaller ? kcaller->inst : 0;
+    bool isVoidReturn = (ri->getNumOperands() == 0);
+    ref<Expr> result(0,Expr::Bool);
+
+    if (WriteTraces) {
+      state.exeTraceMgr.addEvent(new FunctionReturnTraceEvent(state, ki));
+    }
+    
+    if (!isVoidReturn) {
+      result = eval(ki, 0, state);
+    }
+    
+    if (state.stack.size() <= 1) {
+      assert(!caller && "caller set on initial stack frame");
+      terminateStateOnExit(state);
+    } else {
+      state.popFrame();
+
+      if (statsTracker)
+        statsTracker->framePopped(state);
+
+      if (InvokeInst *ii = dyn_cast<InvokeInst>(caller)) {
+        transferToBasicBlock(ii->getNormalDest(), caller->getParent(), state);
+      } else {
+        state.pc = kcaller;
+        ++state.pc;
+      }
+
+      if (!isVoidReturn) {
+        const Type *t = caller->getType();
+        if (t != Type::VoidTy) {
+          // may need to do coercion due to bitcasts
+          Expr::Width from = result.getWidth();
+          Expr::Width to = Expr::getWidthForLLVMType(t);
+            
+          if (from != to) {
+            CallSite cs = (isa<InvokeInst>(caller) ? CallSite(cast<InvokeInst>(caller)) : 
+                           CallSite(cast<CallInst>(caller)));
+
+            // XXX need to check other param attrs ?
+            if (cs.paramHasAttr(0, llvm::Attribute::SExt)) {
+              result = SExtExpr::create(result, to);
+            } else {
+              result = ZExtExpr::create(result, to);
+            }
+          }
+
+          bindLocal(kcaller, state, result);
+        }
+      } else {
+        // We check that the return value has no users instead of
+        // checking the type, since C defaults to returning int for
+        // undeclared functions.
+        if (!caller->use_empty()) {
+          terminateStateOnExecError(state, "return void when caller expected a result");
+        }
+      }
+    }      
+    break;
+  }
+  case Instruction::Unwind: {
+    for (;;) {
+      KInstruction *kcaller = state.stack.back().caller;
+      state.popFrame();
+
+      if (statsTracker)
+        statsTracker->framePopped(state);
+
+      if (state.stack.empty()) {
+        terminateStateOnExecError(state, "unwind from initial stack frame");
+        break;
+      } else {
+        Instruction *caller = kcaller->inst;
+        if (InvokeInst *ii = dyn_cast<InvokeInst>(caller)) {
+          transferToBasicBlock(ii->getUnwindDest(), caller->getParent(), state);
+          break;
+        }
+      }
+    }
+    break;
+  }
+  case Instruction::Br: {
+    BranchInst *bi = cast<BranchInst>(i);
+    if (bi->isUnconditional()) {
+      transferToBasicBlock(bi->getSuccessor(0), bi->getParent(), state);
+    } else {
+      // FIXME: Find a way that we don't have this hidden dependency.
+      assert(bi->getCondition() == bi->getOperand(0) &&
+             "Wrong operand index!");
+      ref<Expr> cond = eval(ki, 0, state);
+      Executor::StatePair branches = fork(state, cond, false);
+
+      if (WriteTraces) {
+        bool isTwoWay = (branches.first && branches.second);
+
+        if (branches.first) {
+          branches.first->exeTraceMgr.addEvent(
+            new BranchTraceEvent(state, ki, true, isTwoWay));
+        }
+
+        if (branches.second) {
+          branches.second->exeTraceMgr.addEvent(
+            new BranchTraceEvent(state, ki, false, isTwoWay));
+        }
+      }
+
+      // NOTE: There is a hidden dependency here, markBranchVisited
+      // requires that we still be in the context of the branch
+      // instruction (it reuses its statistic id). Should be cleaned
+      // up with convenient instruction specific data.
+      if (statsTracker && state.stack.back().kf->trackCoverage)
+        statsTracker->markBranchVisited(branches.first, branches.second);
+
+      if (branches.first)
+        transferToBasicBlock(bi->getSuccessor(0), bi->getParent(), *branches.first);
+      if (branches.second)
+        transferToBasicBlock(bi->getSuccessor(1), bi->getParent(), *branches.second);
+    }
+    break;
+  }
+  case Instruction::Switch: {
+    SwitchInst *si = cast<SwitchInst>(i);
+    ref<Expr> cond = eval(ki, 0, state);
+    unsigned cases = si->getNumCases();
+    BasicBlock *bb = si->getParent();
+
+    cond = toUnique(state, cond);
+    if (cond.isConstant()) {
+      // Somewhat gross to create these all the time, but fine till we
+      // switch to an internal rep.
+      ConstantInt *ci = ConstantInt::get(si->getCondition()->getType(),
+                                         cond.getConstantValue());
+      unsigned index = si->findCaseValue(ci);
+      transferToBasicBlock(si->getSuccessor(index), si->getParent(), state);
+    } else {
+      std::map<BasicBlock*, ref<Expr> > targets;
+      ref<Expr> isDefault(1,Expr::Bool);
+      for (unsigned i=1; i<cases; ++i) {
+        ref<Expr> value = evalConstant(si->getCaseValue(i));
+        ref<Expr> match = EqExpr::create(cond, value);
+        isDefault = AndExpr::create(isDefault, Expr::createNot(match));
+        bool result;
+        bool success = solver->mayBeTrue(state, match, result);
+        assert(success && "FIXME: Unhandled solver failure");
+        if (result) {
+          std::map<BasicBlock*, ref<Expr> >::iterator it =
+            targets.insert(std::make_pair(si->getSuccessor(i),
+                                          ref<Expr>(0,Expr::Bool))).first;
+          it->second = OrExpr::create(match, it->second);
+        }
+      }
+      bool res;
+      bool success = solver->mayBeTrue(state, isDefault, res);
+      assert(success && "FIXME: Unhandled solver failure");
+      if (res)
+        targets.insert(std::make_pair(si->getSuccessor(0), isDefault));
+      
+      std::vector< ref<Expr> > conditions;
+      for (std::map<BasicBlock*, ref<Expr> >::iterator it = 
+             targets.begin(), ie = targets.end();
+           it != ie; ++it)
+        conditions.push_back(it->second);
+      
+      std::vector<ExecutionState*> branches;
+      branch(state, conditions, branches);
+        
+      std::vector<ExecutionState*>::iterator bit = branches.begin();
+      for (std::map<BasicBlock*, ref<Expr> >::iterator it = 
+             targets.begin(), ie = targets.end();
+           it != ie; ++it) {
+        ExecutionState *es = *bit;
+        if (es)
+          transferToBasicBlock(it->first, bb, *es);
+        ++bit;
+      }
+    }
+    break;
+ }
+  case Instruction::Unreachable:
+    // Note that this is not necessarily an internal bug, llvm will
+    // generate unreachable instructions in cases where it knows the
+    // program will crash. So it is effectively a SEGV or internal
+    // error.
+    terminateStateOnExecError(state, "reached \"unreachable\" instruction");
+    break;
+
+  case Instruction::Invoke:
+  case Instruction::Call: {
+    CallSite cs;
+    unsigned argStart;
+    if (i->getOpcode()==Instruction::Call) {
+      cs = CallSite(cast<CallInst>(i));
+      argStart = 1;
+    } else {
+      cs = CallSite(cast<InvokeInst>(i));
+      argStart = 3;
+    }
+
+    unsigned numArgs = cs.arg_size();
+    Function *f = getCalledFunction(cs, state);
+      
+    // evaluate arguments
+    std::vector< ref<Expr> > arguments;
+    arguments.reserve(numArgs);
+
+    for (unsigned j=0; j<numArgs; ++j)
+      arguments.push_back(eval(ki, argStart+j, state));
+
+    if (!f) {
+      // special case the call with a bitcast case
+      Value *fp = cs.getCalledValue();
+      llvm::ConstantExpr *ce = dyn_cast<llvm::ConstantExpr>(fp);
+        
+      if (ce && ce->getOpcode()==Instruction::BitCast) {
+        f = dyn_cast<Function>(ce->getOperand(0));
+        assert(f && "XXX unrecognized constant expression in call");
+        const FunctionType *fType = 
+          dyn_cast<FunctionType>(cast<PointerType>(f->getType())->getElementType());
+        const FunctionType *ceType =
+          dyn_cast<FunctionType>(cast<PointerType>(ce->getType())->getElementType());
+        assert(fType && ceType && "unable to get function type");
+
+        // XXX check result coercion
+
+        // XXX this really needs thought and validation
+        unsigned i=0;
+        for (std::vector< ref<Expr> >::iterator
+               ai = arguments.begin(), ie = arguments.end();
+             ai != ie; ++ai) {
+          Expr::Width to, from = (*ai).getWidth();
+            
+          if (i<fType->getNumParams()) {
+            to = Expr::getWidthForLLVMType(fType->getParamType(i));
+
+            if (from != to) {
+              // XXX need to check other param attrs ?
+              if (cs.paramHasAttr(i+1, llvm::Attribute::SExt)) {
+                arguments[i] = SExtExpr::create(arguments[i], to);
+              } else {
+                arguments[i] = ZExtExpr::create(arguments[i], to);
+              }
+            }
+          }
+            
+          i++;
+        }
+      } else if (isa<InlineAsm>(fp)) {
+        terminateStateOnExecError(state, "inline assembly is unsupported");
+        break;
+      }
+    }
+
+    if (f) {
+      executeCall(state, ki, f, arguments);
+    } else {
+      ref<Expr> v = eval(ki, 0, state);
+
+      ExecutionState *free = &state;
+      bool hasInvalid = false, first = true;
+
+      /* XXX This is wasteful, no need to do a full evaluate since we
+         have already got a value. But in the end the caches should
+         handle it for us, albeit with some overhead. */
+      do {
+        ref<Expr> value;
+        bool success = solver->getValue(*free, v, value);
+        assert(success && "FIXME: Unhandled solver failure");
+        StatePair res = fork(*free, EqExpr::create(v, value), true);
+        if (res.first) {
+          void *addr = (void*) (unsigned long) value.getConstantValue();
+          std::set<void*>::iterator it = legalFunctions.find(addr);
+          if (it != legalFunctions.end()) {
+            f = (Function*) addr;
+
+            // Don't give warning on unique resolution
+            if (res.second || !first)
+              klee_warning_once(addr, 
+                                "resolved symbolic function pointer to: %s",
+                                f->getName().c_str());
+
+            executeCall(*res.first, ki, f, arguments);
+          } else {
+            if (!hasInvalid) {
+              terminateStateOnExecError(state, "invalid function pointer");
+              hasInvalid = true;
+            }
+          }
+        }
+
+        first = false;
+        free = res.second;
+      } while (free);
+    }
+    break;
+  }
+  case Instruction::PHI: {
+    ref<Expr> result = eval(ki, state.incomingBBIndex * 2, state);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+    // Special instructions
+  case Instruction::Select: {
+    SelectInst *SI = cast<SelectInst>(ki->inst);
+    assert(SI->getCondition() == SI->getOperand(0) &&
+           "Wrong operand index!");
+    ref<Expr> cond = eval(ki, 0, state);
+    ref<Expr> tExpr = eval(ki, 1, state);
+    ref<Expr> fExpr = eval(ki, 2, state);
+    ref<Expr> result = SelectExpr::create(cond, tExpr, fExpr);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::VAArg:
+    terminateStateOnExecError(state, "unexpected VAArg instruction");
+    break;
+
+    // Arithmetic / logical
+#define FP_CONSTANT_BINOP(op, type, l, r, target, state) \
+        bindLocal(target, state, \
+                  ref<Expr>(op(toConstant(state, l, "floating point").getConstantValue(), \
+                               toConstant(state, r, "floating point").getConstantValue(), \
+                               type), type))
+  case Instruction::Add: {
+    BinaryOperator *bi = cast<BinaryOperator>(i);
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+
+    if( bi->getType()->getTypeID() == llvm::Type::IntegerTyID ) {
+      bindLocal(ki, state, AddExpr::create(left, right));
+    } else {
+      Expr::Width type = Expr::getWidthForLLVMType(bi->getType());
+      FP_CONSTANT_BINOP(floats::add, type, left, right, ki, state);
+    }
+
+    break;
+  }
+
+  case Instruction::Sub: {
+    BinaryOperator *bi = cast<BinaryOperator>(i);
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+
+    if( bi->getType()->getTypeID() == llvm::Type::IntegerTyID ) {
+      bindLocal(ki, state, SubExpr::create(left, right));
+    } else {
+      Expr::Width type = Expr::getWidthForLLVMType(bi->getType());
+      FP_CONSTANT_BINOP(floats::sub, type, left, right, ki, state);
+    }
+
+    break;
+  }
+ 
+  case Instruction::Mul: {
+    BinaryOperator *bi = cast<BinaryOperator>(i);
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+
+    if( bi->getType()->getTypeID() == llvm::Type::IntegerTyID ) {
+      bindLocal(ki, state, MulExpr::create(left, right));
+    } else {
+      Expr::Width type = Expr::getWidthForLLVMType(bi->getType());
+      FP_CONSTANT_BINOP(floats::mul, type, left, right, ki, state);
+    }
+
+    break;
+  }
+
+  case Instruction::UDiv: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = UDivExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::SDiv: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = SDivExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::URem: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = URemExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+ 
+  case Instruction::SRem: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = SRemExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::And: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = AndExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::Or: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = OrExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::Xor: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = XorExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::Shl: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = ShlExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::LShr: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = LShrExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::AShr: {
+    ref<Expr> left = eval(ki, 0, state);
+    ref<Expr> right = eval(ki, 1, state);
+    ref<Expr> result = AShrExpr::create(left, right);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+    // Compare
+
+  case Instruction::ICmp: {
+    CmpInst *ci = cast<CmpInst>(i);
+    ICmpInst *ii = cast<ICmpInst>(ci);
+ 
+    switch(ii->getPredicate()) {
+    case ICmpInst::ICMP_EQ: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = EqExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_NE: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = NeExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_UGT: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = UgtExpr::create(left, right);
+      bindLocal(ki, state,result);
+      break;
+    }
+
+    case ICmpInst::ICMP_UGE: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = UgeExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_ULT: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = UltExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_ULE: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = UleExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_SGT: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = SgtExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_SGE: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = SgeExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_SLT: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = SltExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    case ICmpInst::ICMP_SLE: {
+      ref<Expr> left = eval(ki, 0, state);
+      ref<Expr> right = eval(ki, 1, state);
+      ref<Expr> result = SleExpr::create(left, right);
+      bindLocal(ki, state, result);
+      break;
+    }
+
+    default:
+      terminateStateOnExecError(state, "invalid ICmp predicate");
+    }
+    break;
+  }
+ 
+    // Memory instructions...
+  case Instruction::Alloca:
+  case Instruction::Malloc: {
+    AllocationInst *ai = cast<AllocationInst>(i);
+    unsigned elementSize = 
+      kmodule->targetData->getTypeStoreSize(ai->getAllocatedType());
+    ref<Expr> size = Expr::createPointer(elementSize);
+    if (ai->isArrayAllocation()) {
+      // XXX coerce?
+      ref<Expr> count = eval(ki, 0, state);
+      size = MulExpr::create(count, size);
+    }
+    bool isLocal = i->getOpcode()==Instruction::Alloca;
+    executeAlloc(state, size, isLocal, ki);
+    break;
+  }
+  case Instruction::Free: {
+    executeFree(state, eval(ki, 0, state));
+    break;
+  }
+
+  case Instruction::Load: {
+    ref<Expr> base = eval(ki, 0, state);
+    executeMemoryOperation(state, false, base, 0, ki);
+    break;
+  }
+  case Instruction::Store: {
+    ref<Expr> base = eval(ki, 1, state);
+    ref<Expr> value = eval(ki, 0, state);
+    executeMemoryOperation(state, true, base, value, 0);
+    break;
+  }
+
+  case Instruction::GetElementPtr: {
+    KGEPInstruction *kgepi = static_cast<KGEPInstruction*>(ki);
+    ref<Expr> base = eval(ki, 0, state);
+
+    for (std::vector< std::pair<unsigned, unsigned> >::iterator 
+           it = kgepi->indices.begin(), ie = kgepi->indices.end(); 
+         it != ie; ++it) {
+      unsigned elementSize = it->second;
+      ref<Expr> index = eval(ki, it->first, state);
+      base = AddExpr::create(base,
+                             MulExpr::create(Expr::createCoerceToPointerType(index),
+                                             Expr::createPointer(elementSize)));
+    }
+    if (kgepi->offset)
+      base = AddExpr::create(base,
+                             Expr::createPointer(kgepi->offset));
+    bindLocal(ki, state, base);
+    break;
+  }
+
+    // Conversion
+  case Instruction::Trunc: {
+    CastInst *ci = cast<CastInst>(i);
+    ref<Expr> result = ExtractExpr::createByteOff(eval(ki, 0, state),
+						  0,
+						  Expr::getWidthForLLVMType(ci->getType()));
+    bindLocal(ki, state, result);
+    break;
+  }
+  case Instruction::ZExt: {
+    CastInst *ci = cast<CastInst>(i);
+    ref<Expr> result = ZExtExpr::create(eval(ki, 0, state),
+                                        Expr::getWidthForLLVMType(ci->getType()));
+    bindLocal(ki, state, result);
+    break;
+  }
+  case Instruction::SExt: {
+    CastInst *ci = cast<CastInst>(i);
+    ref<Expr> result = SExtExpr::create(eval(ki, 0, state),
+                                        Expr::getWidthForLLVMType(ci->getType()));
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::IntToPtr: {
+    CastInst *ci = cast<CastInst>(i);
+    Expr::Width pType = Expr::getWidthForLLVMType(ci->getType());
+    ref<Expr> arg = eval(ki, 0, state);
+    bindLocal(ki, state, ZExtExpr::create(arg, pType));
+    break;
+  } 
+  case Instruction::PtrToInt: {
+    CastInst *ci = cast<CastInst>(i);
+    Expr::Width iType = Expr::getWidthForLLVMType(ci->getType());
+    ref<Expr> arg = eval(ki, 0, state);
+    bindLocal(ki, state, ZExtExpr::create(arg, iType));
+    break;
+  }
+
+  case Instruction::BitCast: {
+    ref<Expr> result = eval(ki, 0, state);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+    // Floating Point specific instructions
+  case Instruction::FPTrunc: {
+    FPTruncInst *fi = cast<FPTruncInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::trunc(arg.getConstantValue(),
+                                   resultType,
+                                   arg.getWidth());
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::FPExt: {
+    FPExtInst *fi = cast<FPExtInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::ext(arg.getConstantValue(),
+                                 resultType,
+                                 arg.getWidth());
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::FPToUI: {
+    FPToUIInst *fi = cast<FPToUIInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::toUnsignedInt(arg.getConstantValue(),
+                                           resultType,
+                                           arg.getWidth());
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::FPToSI: {
+    FPToSIInst *fi = cast<FPToSIInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::toSignedInt(arg.getConstantValue(),
+                                         resultType,
+                                         arg.getWidth());
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::UIToFP: {
+    UIToFPInst *fi = cast<UIToFPInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::UnsignedIntToFP(arg.getConstantValue(),
+                                             resultType);
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::SIToFP: {
+    SIToFPInst *fi = cast<SIToFPInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> arg = toConstant(state, eval(ki, 0, state),
+                               "floating point");
+    uint64_t value = floats::SignedIntToFP(arg.getConstantValue(),
+                                           resultType,
+                                           arg.getWidth());
+    ref<Expr> result(value, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::FCmp: {
+    FCmpInst *fi = cast<FCmpInst>(i);
+    Expr::Width resultType = Expr::getWidthForLLVMType(fi->getType());
+    ref<Expr> left  = toConstant(state, eval(ki, 0, state), 
+                                 "floating point");
+    ref<Expr> right = toConstant(state, eval(ki, 1, state),
+                                 "floating point");
+    uint64_t leftVal = left.getConstantValue();
+    uint64_t rightVal = right.getConstantValue();
+ 
+    //determine whether the operands are NANs
+    unsigned inWidth = left.getWidth();
+    bool leftIsNaN   = floats::isNaN( leftVal,  inWidth );
+    bool rightIsNaN  = floats::isNaN( rightVal, inWidth );
+
+    //handle NAN based on whether the predicate is "ordered" or "unordered"
+    uint64_t ret = (uint64_t)-1;
+    bool done = false;
+    switch( fi->getPredicate() ) {
+      //predicates which only care about whether or not the operands are NaNs
+    case FCmpInst::FCMP_ORD:
+      done = true;
+      ret = !leftIsNaN && !rightIsNaN;
+      break;
+
+    case FCmpInst::FCMP_UNO:
+      done = true;
+      ret = leftIsNaN || rightIsNaN;
+      break;
+
+      //ordered comparisons return false if either operand is NaN
+    case FCmpInst::FCMP_OEQ:
+    case FCmpInst::FCMP_OGT:
+    case FCmpInst::FCMP_OGE:
+    case FCmpInst::FCMP_OLT:
+    case FCmpInst::FCMP_OLE:
+    case FCmpInst::FCMP_ONE:
+      if( !leftIsNaN && !rightIsNaN) //only fall through and return false if there are NaN(s)
+        break;
+
+    case FCmpInst::FCMP_FALSE: { //always return false for this predicate
+      done = true;
+      ret  = false;
+      break;
+    }
+
+      //unordered comparisons return true if either operand is NaN
+    case FCmpInst::FCMP_UEQ:
+    case FCmpInst::FCMP_UGT:
+    case FCmpInst::FCMP_UGE:
+    case FCmpInst::FCMP_ULT:
+    case FCmpInst::FCMP_ULE:
+    case FCmpInst::FCMP_UNE:
+      if( !leftIsNaN && !rightIsNaN) //only fall through and return true if there are NaN(s)
+        break;
+
+    case FCmpInst::FCMP_TRUE: //always return true for this predicate
+      done = true;
+      ret  = true;
+
+    default:
+    case FCmpInst::BAD_FCMP_PREDICATE: /* will fall through and trigger fatal in the next switch */
+      break;
+    }
+
+    //if not done, then we need to actually do a comparison to get the result
+    if( !done ) {
+      switch( fi->getPredicate() ) {
+        //ordered comparisons return false if either operand is NaN
+      case FCmpInst::FCMP_OEQ:
+      case FCmpInst::FCMP_UEQ:
+        ret = floats::eq( leftVal, rightVal, inWidth );
+        break;
+
+      case FCmpInst::FCMP_OGT:
+      case FCmpInst::FCMP_UGT:
+        ret = floats::gt( leftVal, rightVal, inWidth );
+        break;
+
+      case FCmpInst::FCMP_OGE:
+      case FCmpInst::FCMP_UGE:
+        ret = floats::ge( leftVal, rightVal, inWidth );
+        break;
+
+      case FCmpInst::FCMP_OLT:
+      case FCmpInst::FCMP_ULT:
+        ret = floats::lt( leftVal, rightVal, inWidth );
+        break;
+
+      case FCmpInst::FCMP_OLE:
+      case FCmpInst::FCMP_ULE:
+        ret = floats::le( leftVal, rightVal, inWidth );
+        break;
+
+      case FCmpInst::FCMP_ONE:
+      case FCmpInst::FCMP_UNE:
+        ret = floats::ne( leftVal, rightVal, inWidth );
+        break;
+      
+      default:
+        terminateStateOnExecError(state, "invalid FCmp predicate");
+      }
+    }
+
+    ref<Expr> result(ret, resultType);
+    bindLocal(ki, state, result);
+    break;
+  }
+
+  case Instruction::FDiv: {
+    BinaryOperator *bi = cast<BinaryOperator>(i);
+
+    ref<Expr> dividend = eval(ki, 0, state);
+    ref<Expr> divisor = eval(ki, 1, state);
+    Expr::Width type = Expr::getWidthForLLVMType(bi->getType());
+    FP_CONSTANT_BINOP(floats::div, type, dividend, divisor, ki, state);
+    break;
+  }
+
+  case Instruction::FRem: {
+    BinaryOperator *bi = cast<BinaryOperator>(i);
+
+    ref<Expr> dividend = eval(ki, 0, state);
+    ref<Expr> divisor = eval(ki, 1, state);
+    Expr::Width type = Expr::getWidthForLLVMType(bi->getType());
+    FP_CONSTANT_BINOP(floats::mod, type, dividend, divisor, ki, state);
+    break;
+  }
+
+ 
+    // Other instructions...
+    // Unhandled
+  case Instruction::ExtractElement:
+  case Instruction::InsertElement:
+  case Instruction::ShuffleVector:
+    terminateStateOnError(state, "XXX vector instructions unhandled",
+                          "xxx.err");
+    break;
+ 
+  default:
+    terminateStateOnExecError(state, "invalid instruction");
+    break;
+  }
+}
+
+void Executor::updateStates(ExecutionState *current) {
+  if (searcher) {
+    searcher->update(current, addedStates, removedStates);
+  }
+  
+  states.insert(addedStates.begin(), addedStates.end());
+  addedStates.clear();
+  
+  for (std::set<ExecutionState*>::iterator
+         it = removedStates.begin(), ie = removedStates.end();
+       it != ie; ++it) {
+    ExecutionState *es = *it;
+    std::set<ExecutionState*>::iterator it2 = states.find(es);
+    assert(it2!=states.end());
+    states.erase(it2);
+    std::map<ExecutionState*, std::vector<SeedInfo> >::iterator it3 = 
+      seedMap.find(es);
+    if (it3 != seedMap.end())
+      seedMap.erase(it3);
+    processTree->remove(es->ptreeNode);
+    delete es;
+  }
+  removedStates.clear();
+}
+
+void Executor::bindInstructionConstants(KInstruction *KI) {
+  GetElementPtrInst *gepi = dyn_cast<GetElementPtrInst>(KI->inst);
+  if (!gepi)
+    return;
+
+  KGEPInstruction *kgepi = static_cast<KGEPInstruction*>(KI);
+  ref<Expr> constantOffset = Expr::createPointer(0);
+  unsigned index = 1;
+  for (gep_type_iterator ii = gep_type_begin(gepi), ie = gep_type_end(gepi);
+       ii != ie; ++ii) {
+    if (const StructType *st = dyn_cast<StructType>(*ii)) {
+      const StructLayout *sl = 
+        kmodule->targetData->getStructLayout(st);
+      const ConstantInt *ci = cast<ConstantInt>(ii.getOperand());
+      ref<Expr> addend = Expr::createPointer(sl->getElementOffset((unsigned) 
+                                                                  ci->getZExtValue()));
+      constantOffset = AddExpr::create(constantOffset, addend);
+    } else {
+      const SequentialType *st = cast<SequentialType>(*ii);
+      unsigned elementSize = 
+        kmodule->targetData->getTypeStoreSize(st->getElementType());
+      Value *operand = ii.getOperand();
+      if (Constant *c = dyn_cast<Constant>(operand)) {
+        ref<Expr> index = evalConstant(c);
+        ref<Expr> addend = MulExpr::create(Expr::createCoerceToPointerType(index), 
+                                           Expr::createPointer(elementSize));
+        constantOffset = AddExpr::create(constantOffset, addend);
+      } else {
+        kgepi->indices.push_back(std::make_pair(index, elementSize));
+      }
+    }
+    index++;
+  }
+  assert(constantOffset.isConstant());
+  kgepi->offset = constantOffset.getConstantValue();
+}
+
+void Executor::bindModuleConstants() {
+  for (std::vector<KFunction*>::iterator it = kmodule->functions.begin(), 
+         ie = kmodule->functions.end(); it != ie; ++it) {
+    KFunction *kf = *it;
+    for (unsigned i=0; i<kf->numInstructions; ++i)
+      bindInstructionConstants(kf->instructions[i]);
+  }
+
+  kmodule->constantTable = new Cell[kmodule->constants.size()];
+  for (unsigned i=0; i<kmodule->constants.size(); ++i) {
+    Cell &c = kmodule->constantTable[i];
+    c.value = evalConstant(kmodule->constants[i]);
+  }
+}
+
+void Executor::run(ExecutionState &initialState) {
+  bindModuleConstants();
+
+  // Delay init till now so that ticks don't accrue during
+  // optimization and such.
+  initTimers();
+
+  states.insert(&initialState);
+
+  if (usingSeeds) {
+    std::vector<SeedInfo> &v = seedMap[&initialState];
+    
+    for (std::vector<BOut*>::const_iterator it = usingSeeds->begin(), 
+           ie = usingSeeds->end(); it != ie; ++it)
+      v.push_back(SeedInfo(*it));
+
+    int lastNumSeeds = usingSeeds->size()+10;
+    double lastTime, startTime = lastTime = util::getWallTime();
+    ExecutionState *lastState = 0;
+    while (!seedMap.empty()) {
+      if (haltExecution) goto dump;
+
+      std::map<ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+        seedMap.upper_bound(lastState);
+      if (it == seedMap.end())
+        it = seedMap.begin();
+      lastState = it->first;
+      unsigned numSeeds = it->second.size();
+      ExecutionState &state = *lastState;
+      KInstruction *ki = state.pc;
+      stepInstruction(state);
+
+      executeInstruction(state, ki);
+      processTimers(&state, MaxInstructionTime * numSeeds);
+      updateStates(&state);
+
+      if ((stats::instructions % 1000) == 0) {
+        int numSeeds = 0, numStates = 0;
+        for (std::map<ExecutionState*, std::vector<SeedInfo> >::iterator
+               it = seedMap.begin(), ie = seedMap.end();
+             it != ie; ++it) {
+          numSeeds += it->second.size();
+          numStates++;
+        }
+        double time = util::getWallTime();
+        if (SeedTime>0. && time > startTime + SeedTime) {
+          klee_warning("seed time expired, %d seeds remain over %d states",
+                       numSeeds, numStates);
+          break;
+        } else if (numSeeds<=lastNumSeeds-10 ||
+                   time >= lastTime+10) {
+          lastTime = time;
+          lastNumSeeds = numSeeds;          
+          klee_message("%d seeds remaining over: %d states", 
+                       numSeeds, numStates);
+        }
+      }
+    }
+
+    klee_message("seeding done (%d states remain)", (int) states.size());
+
+    // XXX total hack, just because I like non uniform better but want
+    // seed results to be equally weighted.
+    for (std::set<ExecutionState*>::iterator
+           it = states.begin(), ie = states.end();
+         it != ie; ++it) {
+      (*it)->weight = 1.;
+    }
+
+    if (OnlySeed)
+      goto dump;
+  }
+
+  searcher = constructUserSearcher(*this);
+
+  searcher->update(0, states, std::set<ExecutionState*>());
+
+  while (!states.empty() && !haltExecution) {
+    ExecutionState &state = searcher->selectState();
+    KInstruction *ki = state.pc;
+    stepInstruction(state);
+
+    executeInstruction(state, ki);
+    processTimers(&state, MaxInstructionTime);
+
+    if (MaxMemory) {
+      if ((stats::instructions & 0xFFFF) == 0) {
+        // We need to avoid calling GetMallocUsage() often because it
+        // is O(elts on freelist). This is really bad since we start
+        // to pummel the freelist once we hit the memory cap.
+        unsigned mbs = sys::Process::GetTotalMemoryUsage() >> 20;
+        
+        if (mbs > MaxMemory) {
+          if (mbs > MaxMemory + 100) {
+            // just guess at how many to kill
+            unsigned numStates = states.size();
+            unsigned toKill = std::max(1U, numStates - numStates*MaxMemory/mbs);
+
+            if (MaxMemoryInhibit)
+              klee_warning("killing %d states (over memory cap)",
+                           toKill);
+
+            std::vector<ExecutionState*> arr(states.begin(), states.end());
+            for (unsigned i=0,N=arr.size(); N && i<toKill; ++i,--N) {
+              unsigned idx = rand() % N;
+
+              // Make two pulls to try and not hit a state that
+              // covered new code.
+              if (arr[idx]->coveredNew)
+                idx = rand() % N;
+
+              std::swap(arr[idx], arr[N-1]);
+              terminateStateEarly(*arr[N-1], "memory limit");
+            }
+          }
+          atMemoryLimit = true;
+        } else {
+          atMemoryLimit = false;
+        }
+      }
+    }
+
+    updateStates(&state);
+  }
+
+  delete searcher;
+  searcher = 0;
+  
+ dump:
+  if (DumpStatesOnHalt && !states.empty()) {
+    llvm::cerr << "KLEE: halting execution, dumping remaining states\n";
+    for (std::set<ExecutionState*>::iterator
+           it = states.begin(), ie = states.end();
+         it != ie; ++it) {
+      ExecutionState &state = **it;
+      stepInstruction(state); // keep stats rolling
+      terminateStateEarly(state, "execution halting");
+    }
+    updateStates(0);
+  }
+}
+
+std::string Executor::getAddressInfo(ExecutionState &state, 
+                                     ref<Expr> address) const{
+  std::ostringstream info;
+  info << "\taddress: " << address << "\n";
+  uint64_t example;
+  if (address.isConstant()) {
+    example = address.getConstantValue();
+  } else {
+    ref<Expr> value;
+    bool success = solver->getValue(state, address, value);
+    assert(success && "FIXME: Unhandled solver failure");
+    example = value.getConstantValue();
+    info << "\texample: " << example << "\n";
+    std::pair< ref<Expr>, ref<Expr> > res = solver->getRange(state, address);
+    info << "\trange: [" << res.first << ", " << res.second <<"]\n";
+  }
+  
+  MemoryObject hack((unsigned) example);    
+  MemoryMap::iterator lower = state.addressSpace.objects.upper_bound(&hack);
+  info << "\tnext: ";
+  if (lower==state.addressSpace.objects.end()) {
+    info << "none\n";
+  } else {
+    const MemoryObject *mo = lower->first;
+    info << "object at " << mo->address
+         << " of size " << mo->size << "\n";
+  }
+  if (lower!=state.addressSpace.objects.begin()) {
+    --lower;
+    info << "\tprev: ";
+    if (lower==state.addressSpace.objects.end()) {
+      info << "none\n";
+    } else {
+      const MemoryObject *mo = lower->first;
+      info << "object at " << mo->address 
+           << " of size " << mo->size << "\n";
+    }
+  }
+
+  return info.str();
+}
+
+void Executor::terminateState(ExecutionState &state) {
+  if (replayOut && replayPosition!=replayOut->numObjects) {
+    klee_warning_once(replayOut, "replay did not consume all objects in .bout input.");
+  }
+
+  interpreterHandler->incPathsExplored();
+
+  std::set<ExecutionState*>::iterator it = addedStates.find(&state);
+  if (it==addedStates.end()) {
+    state.pc = state.prevPC;
+
+    removedStates.insert(&state);
+  } else {
+    // never reached searcher, just delete immediately
+    std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it3 = 
+      seedMap.find(&state);
+    if (it3 != seedMap.end())
+      seedMap.erase(it3);
+    addedStates.erase(it);
+    processTree->remove(state.ptreeNode);
+    delete &state;
+  }
+}
+
+void Executor::terminateStateEarly(ExecutionState &state, std::string message) {
+  if (!OnlyOutputStatesCoveringNew || state.coveredNew ||
+      (AlwaysOutputSeeds && seedMap.count(&state)))
+    interpreterHandler->processTestCase(state, (message + "\n").c_str(), "early");
+  terminateState(state);
+}
+
+void Executor::terminateStateOnExit(ExecutionState &state) {
+  if (!OnlyOutputStatesCoveringNew || state.coveredNew || 
+      (AlwaysOutputSeeds && seedMap.count(&state)))
+    interpreterHandler->processTestCase(state, 0, 0);
+  terminateState(state);
+}
+
+void Executor::terminateStateOnError(ExecutionState &state,
+                                     const std::string &message,
+                                     const std::string &suffix,
+                                     const std::string &info) {
+  static std::set< std::pair<Instruction*, std::string> > emittedErrors;
+  const InstructionInfo &ii = *state.prevPC->info;
+  
+  if (EmitAllErrors ||
+      emittedErrors.insert(std::make_pair(state.prevPC->inst,message)).second) {
+    if (ii.file != "") {
+      klee_message("ERROR: %s:%d: %s", ii.file.c_str(), ii.line, message.c_str());
+    } else {
+      klee_message("ERROR: %s", message.c_str());
+    }
+    if (!EmitAllErrors)
+      klee_message("NOTE: now ignoring this error at this location");
+    
+    std::ostringstream msg;
+    msg << "Error: " << message << "\n";
+    if (ii.file != "") {
+      msg << "File: " << ii.file << "\n";
+      msg << "Line: " << ii.line << "\n";
+    }
+    msg << "Stack: \n";
+    unsigned idx = 0;
+    const KInstruction *target = state.prevPC;
+    for (ExecutionState::stack_ty::reverse_iterator
+           it = state.stack.rbegin(), ie = state.stack.rend();
+         it != ie; ++it) {
+      StackFrame &sf = *it;
+      Function *f = sf.kf->function;
+      const InstructionInfo &ii = *target->info;
+      msg << "\t#" << idx++ 
+          << " " << std::setw(8) << std::setfill('0') << ii.assemblyLine
+          << " in " << f->getName() << " (";
+      // Yawn, we could go up and print varargs if we wanted to.
+      unsigned index = 0;
+      for (Function::arg_iterator ai = f->arg_begin(), ae = f->arg_end();
+           ai != ae; ++ai) {
+        if (ai!=f->arg_begin()) msg << ", ";
+
+        msg << ai->getName();
+        // XXX should go through function
+        ref<Expr> value = sf.locals[sf.kf->getArgRegister(index++)].value; 
+        if (value.isConstant())
+          msg << "=" << value;
+      }
+      msg << ")";
+      if (ii.file != "")
+        msg << " at " << ii.file << ":" << ii.line;
+      msg << "\n";
+      target = sf.caller;
+    }
+
+    if (info != "")
+      msg << "Info: \n" << info;
+    interpreterHandler->processTestCase(state, msg.str().c_str(), suffix.c_str());
+  }
+    
+  terminateState(state);
+}
+
+// XXX shoot me
+static const char *okExternalsList[] = { "printf", 
+                                         "fprintf", 
+                                         "puts",
+                                         "getpid" };
+static std::set<std::string> okExternals(okExternalsList,
+                                         okExternalsList + 
+                                         (sizeof(okExternalsList)/sizeof(okExternalsList[0])));
+
+void Executor::callExternalFunction(ExecutionState &state,
+                                    KInstruction *target,
+                                    Function *function,
+                                    std::vector< ref<Expr> > &arguments) {
+  // check if specialFunctionHandler wants it
+  if (specialFunctionHandler->handle(state, function, target, arguments))
+    return;
+  
+  if (NoExternals && !okExternals.count(function->getName())) {
+    llvm::cerr << "KLEE:ERROR: Calling not-OK external function : " << function->getName() << "\n";
+    terminateStateOnError(state, "externals disallowed", "user.err");
+    return;
+  }
+
+  // normal external function handling path
+  uint64_t *args = (uint64_t*) alloca(sizeof(*args) * (arguments.size() + 1));
+  memset(args, 0, sizeof(*args) * (arguments.size() + 1));
+
+  unsigned i = 1;
+  for (std::vector<ref<Expr> >::iterator ai = arguments.begin(), ae = arguments.end();
+       ai!=ae; ++ai, ++i) {
+    if (AllowExternalSymCalls) { // don't bother checking uniqueness
+      ref<Expr> ce;
+      bool success = solver->getValue(state, *ai, ce);
+      assert(success && "FIXME: Unhandled solver failure");
+      static_cast<ConstantExpr*>(ce.get())->toMemory((void*) &args[i]);
+    } else {
+      ref<Expr> arg = toUnique(state, *ai);
+      if (arg.isConstant()) {
+        // XXX kick toMemory functions from here
+        static_cast<ConstantExpr*>(arg.get())->toMemory((void*) &args[i]);
+      } else {
+        std::string msg = "external call with symbolic argument: " + function->getName();
+        terminateStateOnExecError(state, msg);
+        return;
+      }
+    }
+  }
+
+  state.addressSpace.copyOutConcretes();
+
+  if (!SuppressExternalWarnings) {
+    std::ostringstream os;
+    os << "calling external: " << function->getName().c_str() << "(";
+    for (unsigned i=0; i<arguments.size(); i++) {
+      os << arguments[i];
+      if (i != arguments.size()-1)
+	os << ", ";
+    }
+    os << ")";
+    
+    if (AllExternalWarnings)
+      klee_warning("%s", os.str().c_str());
+    else
+      klee_warning_once(function, "%s", os.str().c_str());
+  }
+  
+  bool success = externalDispatcher->executeCall(function, target->inst, args);
+  if (!success) {
+    terminateStateOnError(state, "failed external call: " + function->getName(), "external.err");
+    return;
+  }
+
+  if (!state.addressSpace.copyInConcretes()) {
+    terminateStateOnError(state, "external modified read-only object", "external.err");
+    return;
+  }
+
+  const Type *resultType = target->inst->getType();
+  if (resultType != Type::VoidTy) {
+    ref<Expr> e = ConstantExpr::fromMemory((void*) args, 
+                                           Expr::getWidthForLLVMType(resultType));
+    bindLocal(target, state, e);
+  }
+}
+
+/***/
+
+ref<Expr> Executor::replaceReadWithSymbolic(ExecutionState &state, 
+                                            ref<Expr> e) {
+  unsigned n = interpreterOpts.MakeConcreteSymbolic;
+  if (!n || replayOut || replayPath)
+    return e;
+
+  // right now, we don't replace symbolics (is there any reason too?)
+  if (!e.isConstant())
+    return e;
+
+  if (n != 1 && random() %  n)
+    return e;
+
+  // create a new fresh location, assert it is equal to concrete value in e
+  // and return it.
+  
+  const MemoryObject *mo = memory->allocate(Expr::getMinBytesForWidth(e.getWidth()), 
+                                            false, false,
+                                            state.prevPC->inst);
+  assert(mo && "out of memory");
+  ref<Expr> res = Expr::createTempRead(mo->array, e.getWidth());
+  ref<Expr> eq = NotOptimizedExpr::create(EqExpr::create(e, res));
+  llvm::cerr << "Making symbolic: " << eq << "\n";
+  state.addConstraint(eq);
+  return res;
+}
+
+ObjectState *Executor::bindObjectInState(ExecutionState &state, const MemoryObject *mo,
+                                         bool isLocal) {
+  ObjectState *os = new ObjectState(mo, mo->size);
+  state.addressSpace.bindObject(mo, os);
+
+  // Its possible that multiple bindings of the same mo in the state
+  // will put multiple copies on this list, but it doesn't really
+  // matter because all we use this list for is to unbind the object
+  // on function return.
+  if (isLocal)
+    state.stack.back().allocas.push_back(mo);
+
+  return os;
+}
+
+void Executor::executeAllocN(ExecutionState &state,
+                             uint64_t nelems,
+                             uint64_t size,
+                             uint64_t alignment,
+                             bool isLocal,
+                             KInstruction *target) {
+#if 0  
+  // over-allocate so that we can properly align the whole buffer
+  uint64_t address = (uint64_t) (unsigned) malloc(nelems * size + alignment - 1);
+  address += (alignment - address % alignment);
+#else
+  theMMap =   
+    mmap((void*) 0x90000000, 
+         nelems*size, PROT_READ|PROT_WRITE, 
+         MAP_PRIVATE
+#ifdef MAP_ANONYMOUS
+         |MAP_ANONYMOUS
+#endif
+         , 0, 0);
+  uint64_t address = (uintptr_t) theMMap;
+  theMMapSize = nelems*size;
+#endif
+
+  for (unsigned i = 0; i < nelems; i++) {
+    MemoryObject *mo = memory->allocateFixed(address + i*size, size, state.prevPC->inst);
+    ObjectState *os = bindObjectInState(state, mo, isLocal);
+    os->initializeToRandom();
+
+    // bind the local to the first memory object in the whole array
+    if (i == 0)
+      bindLocal(target, state, mo->getBaseExpr());
+  }
+
+  llvm::cerr << "KLEE: allocN at: " << address << "\n";
+}
+
+void Executor::executeAlloc(ExecutionState &state,
+                            ref<Expr> size,
+                            bool isLocal,
+                            KInstruction *target,
+                            bool zeroMemory,
+                            const ObjectState *reallocFrom) {
+  size = toUnique(state, size);
+  if (size.isConstant()) {
+    MemoryObject *mo = memory->allocate(size.getConstantValue(), isLocal, false,
+                                        state.prevPC->inst);
+    if (!mo) {
+      bindLocal(target, state, ref<Expr>(0, kMachinePointerType));
+    } else {
+      ObjectState *os = bindObjectInState(state, mo, isLocal);
+      if (zeroMemory) {
+        os->initializeToZero();
+      } else {
+        os->initializeToRandom();
+      }
+      bindLocal(target, state, mo->getBaseExpr());
+      
+      if (reallocFrom) {
+        unsigned count = std::min(reallocFrom->size, os->size);
+        for (unsigned i=0; i<count; i++)
+          os->write(i, reallocFrom->read8(i));
+        state.addressSpace.unbindObject(reallocFrom->getObject());
+      }
+    }
+  } else {
+    // XXX For now we just pick a size. Ideally we would support
+    // symbolic sizes fully but even if we don't it would be better to
+    // "smartly" pick a value, for example we could fork and pick the
+    // min and max values and perhaps some intermediate (reasonable
+    // value).
+    // 
+    // It would also be nice to recognize the case when size has
+    // exactly two values and just fork (but we need to get rid of
+    // return argument first). This shows up in pcre when llvm
+    // collapses the size expression with a select.
+
+    ref<Expr> example;
+    bool success = solver->getValue(state, size, example);
+    assert(success && "FIXME: Unhandled solver failure");
+    
+    // Try and start with a small example
+    while (example.getConstantValue()>128) {
+      ref<Expr> tmp = ref<Expr>(example.getConstantValue() >> 1, 
+                                example.getWidth());
+      bool res;
+      bool success = solver->mayBeTrue(state, EqExpr::create(tmp, size), res);
+      assert(success && "FIXME: Unhandled solver failure");      
+      if (!res)
+        break;
+      example = tmp;
+    }
+
+    StatePair fixedSize = fork(state, EqExpr::create(example, size), true);
+    
+    if (fixedSize.second) { 
+      // Check for exactly two values
+      ref<Expr> tmp;
+      bool success = solver->getValue(*fixedSize.second, size, tmp);
+      assert(success && "FIXME: Unhandled solver failure");      
+      bool res;
+      success = solver->mustBeTrue(*fixedSize.second, 
+                                   EqExpr::create(tmp, size),
+                                   res);
+      assert(success && "FIXME: Unhandled solver failure");      
+      if (res) {
+        executeAlloc(*fixedSize.second, tmp, isLocal,
+                     target, zeroMemory, reallocFrom);
+      } else {
+        // See if a *really* big value is possible. If so assume
+        // malloc will fail for it, so lets fork and return 0.
+        StatePair hugeSize = fork(*fixedSize.second, 
+                                  UltExpr::create(ref<Expr>(1<<31, Expr::Int32), size), 
+                                  true);
+        if (hugeSize.first) {
+          klee_message("NOTE: found huge malloc, returing 0");
+          bindLocal(target, *hugeSize.first, ref<Expr>(0,kMachinePointerType));
+        }
+        
+        if (hugeSize.second) {
+          std::ostringstream info;
+          ExprPPrinter::printOne(info, "  size expr", size);
+          info << "  concretization : " << example << "\n";
+          info << "  unbound example: " << tmp << "\n";
+          terminateStateOnError(*hugeSize.second, 
+                                "concretized symbolic size", 
+                                "model.err", 
+                                info.str());
+        }
+      }
+    }
+
+    if (fixedSize.first) // can be zero when fork fails
+      executeAlloc(*fixedSize.first, example, isLocal, 
+                   target, zeroMemory, reallocFrom);
+  }
+}
+
+void Executor::executeFree(ExecutionState &state,
+                           ref<Expr> address,
+                           KInstruction *target) {
+  StatePair zeroPointer = fork(state, Expr::createIsZero(address), true);
+  if (zeroPointer.first) {
+    if (target)
+      bindLocal(target, *zeroPointer.first, Expr::createPointer(0));
+  }
+  if (zeroPointer.second) { // address != 0
+    ExactResolutionList rl;
+    resolveExact(*zeroPointer.second, address, rl, "free");
+    
+    for (Executor::ExactResolutionList::iterator it = rl.begin(), 
+           ie = rl.end(); it != ie; ++it) {
+      const MemoryObject *mo = it->first.first;
+      if (mo->isLocal) {
+        terminateStateOnError(*it->second, 
+                              "free of alloca", 
+                              "free.err",
+                              getAddressInfo(*it->second, address));
+      } else if (mo->isGlobal) {
+        terminateStateOnError(*it->second, 
+                              "free of global", 
+                              "free.err",
+                              getAddressInfo(*it->second, address));
+      } else {
+        it->second->addressSpace.unbindObject(mo);
+        if (target)
+          bindLocal(target, *it->second, Expr::createPointer(0));
+      }
+    }
+  }
+}
+
+void Executor::resolveExact(ExecutionState &state,
+                            ref<Expr> p,
+                            ExactResolutionList &results, 
+                            const std::string &name) {
+  // XXX we may want to be capping this?
+  ResolutionList rl;
+  state.addressSpace.resolve(state, solver, p, rl);
+  
+  ExecutionState *unbound = &state;
+  for (ResolutionList::iterator it = rl.begin(), ie = rl.end(); 
+       it != ie; ++it) {
+    ref<Expr> inBounds = EqExpr::create(p, it->first->getBaseExpr());
+    
+    StatePair branches = fork(*unbound, inBounds, true);
+    
+    if (branches.first)
+      results.push_back(std::make_pair(*it, branches.first));
+
+    unbound = branches.second;
+    if (!unbound) // Fork failure
+      break;
+  }
+
+  if (unbound) {
+    terminateStateOnError(*unbound,
+                          "memory error: invalid pointer: " + name,
+                          "ptr.err",
+                          getAddressInfo(*unbound, p));
+  }
+}
+
+void Executor::executeMemoryOperation(ExecutionState &state,
+                                      bool isWrite,
+                                      ref<Expr> address,
+                                      ref<Expr> value /* undef if read */,
+                                      KInstruction *target /* undef if write */) {
+  Expr::Width type = (isWrite ? value.getWidth() : 
+                     Expr::getWidthForLLVMType(target->inst->getType()));
+  unsigned bytes = Expr::getMinBytesForWidth(type);
+
+  if (SimplifySymIndices) {
+    if (!address.isConstant())
+      address = state.constraints.simplifyExpr(address);
+    if (isWrite && !value.isConstant())
+      value = state.constraints.simplifyExpr(value);
+  }
+
+  // fast path: single in-bounds resolution
+  ObjectPair op;
+  bool success;
+  solver->setTimeout(stpTimeout);
+  if (!state.addressSpace.resolveOne(state, solver, address, op, success)) {
+    address = toConstant(state, address, "resolveOne failure");
+    success = state.addressSpace.resolveOne(address.getConstantValue(), op);
+  }
+  solver->setTimeout(0);
+
+  if (success) {
+    const MemoryObject *mo = op.first;
+
+    if (MaxSymArraySize && mo->size>=MaxSymArraySize) {
+      address = toConstant(state, address, "max-sym-array-size");
+    }
+    
+    ref<Expr> offset = mo->getOffsetExpr(address);
+
+    bool inBounds;
+    solver->setTimeout(stpTimeout);
+    bool success = solver->mustBeTrue(state, 
+                                      mo->getBoundsCheckOffset(offset, bytes),
+                                      inBounds);
+    solver->setTimeout(0);
+    if (!success) {
+      state.pc = state.prevPC;
+      terminateStateEarly(state, "query timed out");
+      return;
+    }
+
+    if (inBounds) {
+      const ObjectState *os = op.second;
+      if (isWrite) {
+        if (os->readOnly) {
+          terminateStateOnError(state,
+                                "memory error: object read only",
+                                "readonly.err");
+        } else {
+          ObjectState *wos = state.addressSpace.getWriteable(mo, os);
+          wos->write(offset, value);
+        }          
+      } else {
+        ref<Expr> result = os->read(offset, type);
+        
+        if (interpreterOpts.MakeConcreteSymbolic)
+          result = replaceReadWithSymbolic(state, result);
+        
+        bindLocal(target, state, result);
+      }
+
+      return;
+    }
+  } 
+
+  // we are on an error path (no resolution, multiple resolution, one
+  // resolution with out of bounds)
+  
+  ResolutionList rl;  
+  solver->setTimeout(stpTimeout);
+  bool incomplete = state.addressSpace.resolve(state, solver, address, rl,
+                                               0, stpTimeout);
+  solver->setTimeout(0);
+  
+  // XXX there is some query wasteage here. who cares?
+  ExecutionState *unbound = &state;
+  
+  for (ResolutionList::iterator i = rl.begin(), ie = rl.end(); i != ie; ++i) {
+    const MemoryObject *mo = i->first;
+    const ObjectState *os = i->second;
+    ref<Expr> inBounds = mo->getBoundsCheckPointer(address, bytes);
+    
+    StatePair branches = fork(*unbound, inBounds, true);
+    ExecutionState *bound = branches.first;
+
+    // bound can be 0 on failure or overlapped 
+    if (bound) {
+      if (isWrite) {
+        if (os->readOnly) {
+          terminateStateOnError(*bound,
+                                "memory error: object read only",
+                                "readonly.err");
+        } else {
+          ObjectState *wos = bound->addressSpace.getWriteable(mo, os);
+          wos->write(mo->getOffsetExpr(address), value);
+        }
+      } else {
+        ref<Expr> result = os->read(mo->getOffsetExpr(address), type);
+        bindLocal(target, *bound, result);
+      }
+    }
+
+    unbound = branches.second;
+    if (!unbound)
+      break;
+  }
+  
+  // XXX should we distinguish out of bounds and overlapped cases?
+  if (unbound) {
+    if (incomplete) {
+      terminateStateEarly(*unbound, "query timed out (resolve)");
+    } else {
+      terminateStateOnError(*unbound,
+                            "memory error: out of bound pointer",
+                            "ptr.err",
+                            getAddressInfo(*unbound, address));
+    }
+  }
+}
+
+void Executor::executeMakeSymbolic(ExecutionState &state, 
+                                   const MemoryObject *mo) {
+  // make a new one and rebind, we use bind here because we want to
+  // create a flat out new state, not a copy. although I'm not really
+  // sure it matters.
+  ObjectState *os = bindObjectInState(state, mo, false);
+  if (!replayOut) {
+    os->makeSymbolic();
+    state.addSymbolic(mo);
+    
+    std::map< ExecutionState*, std::vector<SeedInfo> >::iterator it = 
+      seedMap.find(&state);
+    if (it!=seedMap.end()) { // In seed mode we need to add this as a
+                             // binding.
+      for (std::vector<SeedInfo>::iterator siit = it->second.begin(), 
+             siie = it->second.end(); siit != siie; ++siit) {
+        SeedInfo &si = *siit;
+        BOutObject *obj = si.getNextInput(mo,
+                                          NamedSeedMatching);
+
+        if (!obj) {
+          if (ZeroSeedExtension) {
+            std::vector<unsigned char> &values = 
+              si.assignment.bindings[mo->array];
+            values = std::vector<unsigned char>(mo->size, '\0');
+          } else if (!AllowSeedExtension) {
+            terminateStateOnError(state, 
+                                  "ran out of inputs during seeding",
+                                  "user.err");
+            break;
+          }
+        } else {
+          if (obj->numBytes != mo->size &&
+              ((!(AllowSeedExtension || ZeroSeedExtension)
+                && obj->numBytes < mo->size) ||
+               (!AllowSeedTruncation && obj->numBytes > mo->size))) {
+	    std::stringstream msg;
+	    msg << "replace size mismatch: "
+		<< mo->name << "[" << mo->size << "]"
+		<< " vs " << obj->name << "[" << obj->numBytes << "]"
+		<< " in bout\n";
+
+            terminateStateOnError(state,
+                                  msg.str(),
+                                  "user.err");
+            break;
+          } else {
+            std::vector<unsigned char> &values = 
+              si.assignment.bindings[mo->array];
+            values.insert(values.begin(), obj->bytes, 
+                          obj->bytes + std::min(obj->numBytes, mo->size));            
+            if (ZeroSeedExtension) {
+              for (unsigned i=obj->numBytes; i<mo->size; ++i)
+                values.push_back('\0');
+            }
+          }
+        }
+      }
+    }
+  } else {
+    if (replayPosition >= replayOut->numObjects) {
+      terminateStateOnError(state, "replay count mismatch", "user.err");
+    } else {
+      BOutObject *obj = &replayOut->objects[replayPosition++];
+      if (obj->numBytes != mo->size) {
+        terminateStateOnError(state, "replay size mismatch", "user.err");
+      } else {
+        for (unsigned i=0; i<mo->size; i++)
+          os->write8(i, obj->bytes[i]);
+      }
+    }
+  }
+}
+
+/***/
+
+void Executor::runFunctionAsMain(Function *f,
+				 int argc,
+				 char **argv,
+				 char **envp) {
+  std::vector<ref<Expr> > arguments;
+
+  // force deterministic initialization of memory objects
+  srand(1);
+  srandom(1);
+  
+  MemoryObject *argvMO = 0;
+
+  // In order to make uclibc happy and be closer to what the system is
+  // doing we lay out the environments at the end of the argv array
+  // (both are terminated by a null). There is also a final terminating
+  // null that uclibc seems to expect, possibly the ELF header?
+
+  int envc;
+  for (envc=0; envp[envc]; ++envc) ;
+
+  KFunction *kf = kmodule->functionMap[f];
+  assert(kf);
+  Function::arg_iterator ai = f->arg_begin(), ae = f->arg_end();
+  if (ai!=ae) {
+    arguments.push_back(ref<Expr>(argc, Expr::Int32));
+
+    if (++ai!=ae) {
+      argvMO = memory->allocate((argc+1+envc+1+1) * kMachinePointerSize, false, true,
+                                f->begin()->begin());
+      
+      arguments.push_back(argvMO->getBaseExpr());
+
+      if (++ai!=ae) {
+        uint64_t envp_start = argvMO->address + (argc+1)*kMachinePointerSize;
+        arguments.push_back(Expr::createPointer(envp_start));
+
+        if (++ai!=ae)
+          klee_error("invalid main function (expect 0-3 arguments)");
+      }
+    }
+  }
+
+  ExecutionState *state = new ExecutionState(kmodule->functionMap[f]);
+  
+  if (pathWriter) 
+    state->pathOS = pathWriter->open();
+  if (symPathWriter) 
+    state->symPathOS = symPathWriter->open();
+
+
+  if (statsTracker)
+    statsTracker->framePushed(*state, 0);
+
+  assert(arguments.size() == f->arg_size() && "wrong number of arguments");
+  for (unsigned i = 0, e = f->arg_size(); i != e; ++i)
+    bindArgument(kf, i, *state, arguments[i]);
+
+  if (argvMO) {
+    ObjectState *argvOS = bindObjectInState(*state, argvMO, false);
+
+    for (int i=0; i<argc+1+envc+1+1; i++) {
+      MemoryObject *arg;
+      
+      if (i==argc || i>=argc+1+envc) {
+        arg = 0;
+      } else {
+        char *s = i<argc ? argv[i] : envp[i-(argc+1)];
+        int j, len = strlen(s);
+        
+        arg = memory->allocate(len+1, false, true, state->pc->inst);
+        ObjectState *os = bindObjectInState(*state, arg, false);
+        for (j=0; j<len+1; j++)
+          os->write8(j, s[j]);
+      }
+
+      if (arg) {
+        argvOS->write(i * kMachinePointerSize, arg->getBaseExpr());
+      } else {
+        argvOS->write(i * kMachinePointerSize, Expr::createPointer(0));
+      }
+    }
+  }
+  
+  initializeGlobals(*state);
+
+  processTree = new PTree(state);
+  state->ptreeNode = processTree->root;
+  run(*state);
+  delete processTree;
+  processTree = 0;
+
+  // hack to clear memory objects
+  delete memory;
+  memory = new MemoryManager();
+  
+  globalObjects.clear();
+  globalAddresses.clear();
+
+  if (statsTracker)
+    statsTracker->done();
+
+  if (theMMap) {
+    munmap(theMMap, theMMapSize);
+    theMMap = 0;
+  }
+}
+
+unsigned Executor::getPathStreamID(const ExecutionState &state) {
+  assert(pathWriter);
+  return state.pathOS.getID();
+}
+
+unsigned Executor::getSymbolicPathStreamID(const ExecutionState &state) {
+  assert(symPathWriter);
+  return state.symPathOS.getID();
+}
+
+void Executor::getConstraintLog(const ExecutionState &state,
+                                std::string &res,
+                                bool asCVC) {
+  if (asCVC) {
+    Query query(state.constraints, ref<Expr>(0, Expr::Bool));
+    char *log = solver->stpSolver->getConstraintLog(query);
+    res = std::string(log);
+    free(log);
+  } else {
+    std::ostringstream info;
+    ExprPPrinter::printConstraints(info, state.constraints);
+    res = info.str();    
+  }
+}
+
+bool Executor::getSymbolicSolution(const ExecutionState &state,
+                                   std::vector< 
+                                   std::pair<std::string,
+                                   std::vector<unsigned char> > >
+                                   &res) {
+  solver->setTimeout(stpTimeout);
+
+  ExecutionState tmp(state);
+  if (!NoPreferCex) {
+    for (std::vector<const MemoryObject*>::const_iterator 
+           it = state.symbolics.begin(), ie = state.symbolics.end(); 
+         it != ie; ++it) {
+      const MemoryObject *mo = *it;
+      std::vector< ref<Expr> >::const_iterator pi = 
+        mo->cexPreferences.begin(), pie = mo->cexPreferences.end();
+      for (; pi != pie; ++pi) {
+        bool mustBeTrue;
+        bool success = solver->mustBeTrue(tmp, Expr::createNot(*pi), 
+                                          mustBeTrue);
+        if (!success) break;
+        if (!mustBeTrue) tmp.addConstraint(*pi);
+      }
+      if (pi!=pie) break;
+    }
+  }
+
+  std::vector< std::vector<unsigned char> > values;
+  std::vector<const Array*> objects;
+  for (unsigned i = 0; i != state.symbolics.size(); ++i)
+    objects.push_back(state.symbolics[i]->array);
+  bool success = solver->getInitialValues(tmp, objects, values);
+  solver->setTimeout(0);
+  if (!success) {
+    klee_warning("unable to compute initial values (invalid constraints?)!");
+    ExprPPrinter::printQuery(std::cerr,
+                             state.constraints, 
+                             ref<Expr>(0,Expr::Bool));
+    return false;
+  }
+  
+  unsigned i = 0;
+  for (std::vector<const MemoryObject*>::const_iterator 
+         it = state.symbolics.begin(), ie = state.symbolics.end(); 
+       it != ie; ++it) {
+    res.push_back(std::make_pair((*it)->name, values[i]));
+    ++i;
+  }
+  return true;
+}
+
+void Executor::getCoveredLines(const ExecutionState &state,
+                               std::map<const std::string*, std::set<unsigned> > &res) {
+  res = state.coveredLines;
+}
+
+void Executor::doImpliedValueConcretization(ExecutionState &state,
+                                            ref<Expr> e,
+                                            ref<Expr> value) {
+  assert(value.isConstant() && "non-constant passed in place of constant");
+  
+  if (DebugCheckForImpliedValues)
+    ImpliedValue::checkForImpliedValues(solver->solver, e, value);
+
+  ImpliedValueList results;
+  ImpliedValue::getImpliedValues(e, value, results);
+  for (ImpliedValueList::iterator it = results.begin(), ie = results.end();
+       it != ie; ++it) {
+    ReadExpr *re = it->first.get();
+    
+    if (re->index.isConstant()) {
+      // FIXME: This is the sole remaining usage of the Array object
+      // variable. Kill me.
+      const MemoryObject *mo = re->updates.root->object;
+      const ObjectState *os = state.addressSpace.findObject(mo);
+
+      if (!os) {
+        // object has been free'd, no need to concretize (although as
+        // in other cases we would like to concretize the outstanding
+        // reads, but we have no facility for that yet)
+      } else {
+        assert(!os->readOnly && "not possible? read only object with static read?");
+        ObjectState *wos = state.addressSpace.getWriteable(mo, os);
+        wos->write(re->index.getConstantValue(), it->second);
+      }
+    }
+  }
+}
+
+///
+
+Interpreter *Interpreter::create(const InterpreterOptions &opts,
+                                 InterpreterHandler *ih) {
+  return new Executor(opts, ih);
+}

Added: klee/trunk/lib/Core/Executor.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Executor.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Executor.h (added)
+++ klee/trunk/lib/Core/Executor.h Wed May 20 23:36:41 2009
@@ -0,0 +1,445 @@
+//===-- Executor.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Class to perform actual execution, hides implementation details from external
+// interpreter.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXECUTOR_H
+#define KLEE_EXECUTOR_H
+
+#include "klee/Interpreter.h"
+#include "llvm/Support/CallSite.h"
+#include <vector>
+#include <string>
+#include <map>
+#include <set>
+
+struct BOut;
+
+namespace llvm {
+  class BasicBlock;
+  class BranchInst;
+  class CallInst;
+  class Constant;
+  class ConstantExpr;
+  class Function;
+  class GlobalValue;
+  class Instruction;
+  class TargetData;
+  class Value;
+}
+
+namespace klee {  
+  class ExecutionState;
+  class ExternalDispatcher;
+  class Expr;
+  class InstructionInfoTable;
+  class KFunction;
+  class KInstruction;
+  class KInstIterator;
+  class KModule;
+  class MemoryManager;
+  class MemoryObject;
+  class ObjectState;
+  class PTree;
+  class Searcher;
+  class SeedInfo;
+  class SpecialFunctionHandler;
+  class StackFrame;
+  class StatsTracker;
+  class TimingSolver;
+  class TreeStreamWriter;
+  template<class T> class ref;
+
+  /// \todo Add a context object to keep track of data only live
+  /// during an instruction step. Should contain addedStates,
+  /// removedStates, and haltExecution, among others.
+
+class Executor : public Interpreter {
+  friend class BumpMergingSearcher;
+  friend class MergingSearcher;
+  friend class RandomPathSearcher;
+  friend class OwningSearcher;
+  friend class WeightedRandomSearcher;
+  friend class SpecialFunctionHandler;
+  friend class StatsTracker;
+
+public:
+  class Timer {
+  public:
+    Timer();
+    virtual ~Timer();
+
+    /// The event callback.
+    virtual void run() = 0;
+  };
+
+  typedef std::pair<ExecutionState*,ExecutionState*> StatePair;
+
+private:
+  class TimerInfo;
+
+  KModule *kmodule;
+  InterpreterHandler *interpreterHandler;
+  Searcher *searcher;
+
+  ExternalDispatcher *externalDispatcher;
+  TimingSolver *solver;
+  MemoryManager *memory;
+  std::set<ExecutionState*> states;
+  StatsTracker *statsTracker;
+  TreeStreamWriter *pathWriter, *symPathWriter;
+  SpecialFunctionHandler *specialFunctionHandler;
+  std::vector<TimerInfo*> timers;
+  PTree *processTree;
+
+  /// Used to track states that have been added during the current
+  /// instructions step. 
+  /// \invariant \ref addedStates is a subset of \ref states. 
+  /// \invariant \ref addedStates and \ref removedStates are disjoint.
+  std::set<ExecutionState*> addedStates;
+  /// Used to track states that have been removed during the current
+  /// instructions step. 
+  /// \invariant \ref removedStates is a subset of \ref states. 
+  /// \invariant \ref addedStates and \ref removedStates are disjoint.
+  std::set<ExecutionState*> removedStates;
+
+  /// When non-empty the Executor is running in "seed" mode. The
+  /// states in this map will be executed in an arbitrary order
+  /// (outside the normal search interface) until they terminate. When
+  /// the states reach a symbolic branch then either direction that
+  /// satisfies one or more seeds will be added to this map. What
+  /// happens with other states (that don't satisfy the seeds) depends
+  /// on as-yet-to-be-determined flags.
+  std::map<ExecutionState*, std::vector<SeedInfo> > seedMap;
+  
+  /// Map of globals to their representative memory object.
+  std::map<const llvm::GlobalValue*, MemoryObject*> globalObjects;
+
+  /// Map of globals to their bound address. This also includes
+  /// globals that have no representative object (i.e. functions).
+  std::map<const llvm::GlobalValue*, ref<Expr> > globalAddresses;
+
+  /// The set of legal function addresses, used to validate function
+  /// pointers.
+  std::set<void*> legalFunctions;
+
+  /// When non-null the bindings that will be used for calls to
+  /// klee_make_symbolic in order replay.
+  const struct BOut *replayOut;
+  /// When non-null a list of branch decisions to be used for replay.
+  const std::vector<bool> *replayPath;
+  /// The index into the current \ref replayOut or \ref replayPath
+  /// object.
+  unsigned replayPosition;
+
+  /// When non-null a list of "seed" inputs which will be used to
+  /// drive execution.
+  const std::vector<struct BOut *> *usingSeeds;  
+
+  /// Disables forking, instead a random path is chosen. Enabled as
+  /// needed to control memory usage. \see fork()
+  bool atMemoryLimit;
+
+  /// Disables forking, set by client. \see setInhibitForking()
+  bool inhibitForking;
+
+  /// Signals the executor to halt execution at the next instruction
+  /// step.
+  bool haltExecution;  
+
+  /// Whether implied-value concretization is enabled. Currently
+  /// false, it is buggy (it needs to validate its writes).
+  bool ivcEnabled;
+
+  /// The maximum time to allow for a single stp query.
+  double stpTimeout;  
+
+  llvm::Function* getCalledFunction(llvm::CallSite &cs, ExecutionState &state);
+  
+  void executeInstruction(ExecutionState &state, KInstruction *ki);
+
+  void printFileLine(ExecutionState &state, KInstruction *ki);
+
+  void run(ExecutionState &initialState);
+
+  // Given a concrete object in our [klee's] address space, add it to 
+  // objects checked code can reference.
+  MemoryObject *addExternalObject(ExecutionState &state, void *addr, 
+                                  unsigned size, bool isReadOnly);
+
+  void initializeGlobalObject(ExecutionState &state, ObjectState *os, 
+			      llvm::Constant *c,
+			      unsigned offset);
+  void initializeGlobals(ExecutionState &state);
+
+  void stepInstruction(ExecutionState &state);
+  void updateStates(ExecutionState *current);
+  void transferToBasicBlock(llvm::BasicBlock *dst, 
+			    llvm::BasicBlock *src,
+			    ExecutionState &state);
+
+  void callExternalFunction(ExecutionState &state,
+                            KInstruction *target,
+                            llvm::Function *function,
+                            std::vector< ref<Expr> > &arguments);
+
+  ObjectState *bindObjectInState(ExecutionState &state, const MemoryObject *mo,
+                                 bool isLocal);
+
+  /// Resolve a pointer to the memory objects it could point to the
+  /// start of, forking execution when necessary and generating errors
+  /// for pointers to invalid locations (either out of bounds or
+  /// address inside the middle of objects).
+  ///
+  /// \param results[out] A list of ((MemoryObject,ObjectState),
+  /// state) pairs for each object the given address can point to the
+  /// beginning of.
+  typedef std::vector< std::pair<std::pair<const MemoryObject*, const ObjectState*>, 
+                                 ExecutionState*> > ExactResolutionList;
+  void resolveExact(ExecutionState &state,
+                    ref<Expr> p,
+                    ExactResolutionList &results,
+                    const std::string &name);
+
+  /// Allocate and bind a new object in a particular state. NOTE: This
+  /// function may fork.
+  ///
+  /// \param isLocal Flag to indicate if the object should be
+  /// automatically deallocated on function return (this also makes it
+  /// illegal to free directly).
+  ///
+  /// \param target Value at which to bind the base address of the new
+  /// object.
+  ///
+  /// \param reallocFrom If non-zero and the allocation succeeds,
+  /// initialize the new object from the given one and unbind it when
+  /// done (realloc semantics). The initialized bytes will be the
+  /// minimum of the size of the old and new objects, with remaining
+  /// bytes initialized as specified by zeroMemory.
+  void executeAlloc(ExecutionState &state,
+                    ref<Expr> size,
+                    bool isLocal,
+                    KInstruction *target,
+                    bool zeroMemory=false,
+                    const ObjectState *reallocFrom=0);
+
+  /// XXX not for public use (this is for histar, it allocations a
+  /// contiguous set of objects, while guaranteeing page alignment)
+  void executeAllocN(ExecutionState &state,
+                     uint64_t nelems,
+                     uint64_t size,
+                     uint64_t alignment,
+                     bool isLocal,
+                     KInstruction *target);
+
+  /// Free the given address with checking for errors. If target is
+  /// given it will be bound to 0 in the resulting states (this is a
+  /// convenience for realloc). Note that this function can cause the
+  /// state to fork and that \ref state cannot be safely accessed
+  /// afterwards.
+  void executeFree(ExecutionState &state,
+                   ref<Expr> address,
+                   KInstruction *target = 0);
+  
+  void executeCall(ExecutionState &state, 
+                   KInstruction *ki,
+                   llvm::Function *f,
+                   std::vector< ref<Expr> > &arguments);
+                   
+  // do address resolution / object binding / out of bounds checking
+  // and perform the operation
+  void executeMemoryOperation(ExecutionState &state,
+                              bool isWrite,
+                              ref<Expr> address,
+                              ref<Expr> value /* undef if read */,
+                              KInstruction *target /* undef if write */);
+
+  void executeMakeSymbolic(ExecutionState &state, const MemoryObject *mo);
+
+  /// Create a new state where each input condition has been added as
+  /// a constraint and return the results. The input state is included
+  /// as one of the results. Note that the output vector may included
+  /// NULL pointers for states which were unable to be created.
+  void branch(ExecutionState &state, 
+              const std::vector< ref<Expr> > &conditions,
+              std::vector<ExecutionState*> &result);
+
+  // Fork current and return states in which condition holds / does
+  // not hold, respectively. One of the states is necessarily the
+  // current state, and one of the states may be null.
+  StatePair fork(ExecutionState &current, ref<Expr> condition, bool isInternal);
+
+  /// Add the given (boolean) condition as a constraint on state. This
+  /// function is a wrapper around the state's addConstraint function
+  /// which also manages manages propogation of implied values,
+  /// validity checks, and seed patching.
+  void addConstraint(ExecutionState &state, ref<Expr> condition);
+
+  // Called on [for now] concrete reads, replaces constant with a symbolic
+  // Used for testing.
+  ref<Expr> replaceReadWithSymbolic(ExecutionState &state, ref<Expr> e);
+
+  ref<Expr> eval(KInstruction *ki,
+                 unsigned index, 
+                 ExecutionState &state);
+  
+  void bindLocal(KInstruction *target, 
+                 ExecutionState &state, 
+                 ref<Expr> value);
+  void bindArgument(KFunction *kf, 
+                    unsigned index,
+                    ExecutionState &state,
+                    ref<Expr> value);
+
+  ref<Expr> evalConstantExpr(llvm::ConstantExpr *ce);
+
+  /// Return a unique constant value for the given expression in the
+  /// given state, if it has one (i.e. it provably only has a single
+  /// value). Otherwise return the original expression.
+  ref<Expr> toUnique(const ExecutionState &state, ref<Expr> &e);
+
+  /// Return a constant value for the given expression, forcing it to
+  /// be constant in the given state by adding a constraint if
+  /// necessary. Note that this function breaks completeness and
+  /// should generally be avoided.
+  ///
+  /// \param purpose An identify string to printed in case of concretization.
+  ref<Expr> toConstant(ExecutionState &state, ref<Expr> e, const char *purpose);
+
+  /// Bind a constant value for e to the given target. NOTE: This
+  /// function may fork state if the state has multiple seeds.
+  void executeGetValue(ExecutionState &state, ref<Expr> e, KInstruction *target);
+
+  /// Get textual information regarding a memory address.
+  std::string getAddressInfo(ExecutionState &state, ref<Expr> address) const;
+
+  // remove state from queue and delete
+  void terminateState(ExecutionState &state);
+  // call exit handler and terminate state
+  void terminateStateEarly(ExecutionState &state, std::string message);
+  // call exit handler and terminate state
+  void terminateStateOnExit(ExecutionState &state);
+  // call error handler and terminate state
+  void terminateStateOnError(ExecutionState &state, 
+                             const std::string &message,
+                             const std::string &suffix,
+                             const std::string &longMessage="");
+
+  // call error handler and terminate state, for execution errors
+  // (things that should not be possible, like illegal instruction or
+  // unlowered instrinsic, or are unsupported, like inline assembly)
+  void terminateStateOnExecError(ExecutionState &state, 
+                                 const std::string &message,
+                                 const std::string &info="") {
+    terminateStateOnError(state, message, "exec.err", info);
+  }
+
+  /// bindModuleConstants - Initialize the module constant table.
+  void bindModuleConstants();
+
+  /// bindInstructionConstants - Initialize any necessary per instruction
+  /// constant values.
+  void bindInstructionConstants(KInstruction *KI);
+
+  void handlePointsToObj(ExecutionState &state, 
+                         KInstruction *target, 
+                         const std::vector<ref<Expr> > &arguments);
+
+  void doImpliedValueConcretization(ExecutionState &state,
+                                    ref<Expr> e,
+                                    ref<Expr> value);
+
+  /// Add a timer to be executed periodically.
+  ///
+  /// \param timer The timer object to run on firings.
+  /// \param rate The approximate delay (in seconds) between firings.
+  void addTimer(Timer *timer, double rate);
+
+  void initTimers();
+  void processTimers(ExecutionState *current,
+                     double maxInstTime);
+                
+public:
+  Executor(const InterpreterOptions &opts, InterpreterHandler *ie);
+  virtual ~Executor();
+
+  const InterpreterHandler& getHandler() {
+    return *interpreterHandler;
+  }
+
+  // XXX should just be moved out to utility module
+  ref<Expr> evalConstant(llvm::Constant *c);
+
+  virtual void setPathWriter(TreeStreamWriter *tsw) {
+    pathWriter = tsw;
+  }
+  virtual void setSymbolicPathWriter(TreeStreamWriter *tsw) {
+    symPathWriter = tsw;
+  }      
+
+  virtual void setReplayOut(const struct BOut *out) {
+    assert(!replayPath && "cannot replay both buffer and path");
+    replayOut = out;
+    replayPosition = 0;
+  }
+
+  virtual void setReplayPath(const std::vector<bool> *path) {
+    assert(!replayOut && "cannot replay both buffer and path");
+    replayPath = path;
+    replayPosition = 0;
+  }
+
+  virtual const llvm::Module *
+  setModule(llvm::Module *module, const ModuleOptions &opts);
+
+  virtual void useSeeds(const std::vector<struct BOut *> *seeds) { 
+    usingSeeds = seeds;
+  }
+
+  virtual void runFunctionAsMain(llvm::Function *f,
+                                 int argc,
+                                 char **argv,
+                                 char **envp);
+
+  /*** Runtime options ***/
+  
+  virtual void setHaltExecution(bool value) {
+    haltExecution = value;
+  }
+
+  virtual void setInhibitForking(bool value) {
+    inhibitForking = value;
+  }
+
+  /*** State accessor methods ***/
+
+  virtual unsigned getPathStreamID(const ExecutionState &state);
+
+  virtual unsigned getSymbolicPathStreamID(const ExecutionState &state);
+
+  virtual void getConstraintLog(const ExecutionState &state,
+                                std::string &res,
+                                bool asCVC = false);
+
+  virtual bool getSymbolicSolution(const ExecutionState &state, 
+                                   std::vector< 
+                                   std::pair<std::string,
+                                   std::vector<unsigned char> > >
+                                   &res);
+
+  virtual void getCoveredLines(const ExecutionState &state,
+                               std::map<const std::string*, std::set<unsigned> > &res);
+};
+  
+} // End klee namespace
+
+#endif

Added: klee/trunk/lib/Core/ExecutorTimers.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ExecutorTimers.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ExecutorTimers.cpp (added)
+++ klee/trunk/lib/Core/ExecutorTimers.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,220 @@
+//===-- ExecutorTimers.cpp ------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "CoreStats.h"
+#include "Executor.h"
+#include "PTree.h"
+#include "StatsTracker.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/KModule.h"
+#include "klee/Internal/System/Time.h"
+
+#include "llvm/Function.h"
+#include "llvm/Support/CommandLine.h"
+
+#include <unistd.h>
+#include <signal.h>
+#include <sys/time.h>
+#include <math.h>
+
+
+using namespace llvm;
+using namespace klee;
+
+cl::opt<double>
+MaxTime("max-time",
+        cl::desc("Halt execution after the specified number of seconds (0=off)"),
+        cl::init(0));
+
+///
+
+class HaltTimer : public Executor::Timer {
+  Executor *executor;
+
+public:
+  HaltTimer(Executor *_executor) : executor(_executor) {}
+  ~HaltTimer() {}
+
+  void run() {
+    llvm::cerr << "KLEE: HaltTimer invoked\n";
+    executor->setHaltExecution(true);
+  }
+};
+
+///
+
+static const double kSecondsPerTick = .1;
+static volatile unsigned timerTicks = 0;
+
+// XXX hack
+extern "C" unsigned dumpStates, dumpPTree;
+unsigned dumpStates = 0, dumpPTree = 0;
+
+static void onAlarm(int) {
+  ++timerTicks;
+}
+
+// oooogalay
+static void setupHandler() {
+  struct itimerval t;
+  struct timeval tv;
+  
+  tv.tv_sec = (long) kSecondsPerTick;
+  tv.tv_usec = (long) (fmod(kSecondsPerTick, 1.)*1000000);
+  
+  t.it_interval = t.it_value = tv;
+  
+  ::setitimer(ITIMER_REAL, &t, 0);
+  ::signal(SIGALRM, onAlarm);
+}
+
+void Executor::initTimers() {
+  static bool first = true;
+
+  if (first) {
+    first = false;
+    setupHandler();
+  }
+
+  if (MaxTime) {
+    addTimer(new HaltTimer(this), MaxTime);
+  }
+}
+
+///
+
+Executor::Timer::Timer() {}
+
+Executor::Timer::~Timer() {}
+
+class Executor::TimerInfo {
+public:
+  Timer *timer;
+  
+  /// Approximate delay per timer firing.
+  double rate;
+  /// Wall time for next firing.
+  double nextFireTime;
+  
+public:
+  TimerInfo(Timer *_timer, double _rate) 
+    : timer(_timer),
+      rate(_rate),
+      nextFireTime(util::getWallTime() + rate) {}
+  ~TimerInfo() { delete timer; }
+};
+
+void Executor::addTimer(Timer *timer, double rate) {
+  timers.push_back(new TimerInfo(timer, rate));
+}
+
+void Executor::processTimers(ExecutionState *current,
+                             double maxInstTime) {
+  static unsigned callsWithoutCheck = 0;
+  unsigned ticks = timerTicks;
+
+  if (!ticks && ++callsWithoutCheck > 1000) {
+    setupHandler();
+    ticks = 1;
+  }
+
+  if (ticks || dumpPTree || dumpStates) {
+    if (dumpPTree) {
+      char name[32];
+      sprintf(name, "ptree%08d.dot", (int) stats::instructions);
+      std::ostream *os = interpreterHandler->openOutputFile(name);
+      if (os) {
+        processTree->dump(*os);
+        delete os;
+      }
+      
+      dumpPTree = 0;
+    }
+
+    if (dumpStates) {
+      std::ostream *os = interpreterHandler->openOutputFile("states.txt");
+      
+      if (os) {
+        for (std::set<ExecutionState*>::const_iterator it = states.begin(), 
+               ie = states.end(); it != ie; ++it) {
+          ExecutionState *es = *it;
+          *os << "(" << es << ",";
+          *os << "[";
+          ExecutionState::stack_ty::iterator next = es->stack.begin();
+          ++next;
+          for (ExecutionState::stack_ty::iterator sfIt = es->stack.begin(),
+                 sf_ie = es->stack.end(); sfIt != sf_ie; ++sfIt) {
+            *os << "('" << sfIt->kf->function->getName() << "',";
+            if (next == es->stack.end()) {
+              *os << es->prevPC->info->line << "), ";
+            } else {
+              *os << next->caller->info->line << "), ";
+              ++next;
+            }
+          }
+          *os << "], ";
+
+          StackFrame &sf = es->stack.back();
+          uint64_t md2u = computeMinDistToUncovered(es->pc,
+                                                    sf.minDistToUncoveredOnReturn);
+          uint64_t icnt = theStatisticManager->getIndexedValue(stats::instructions,
+                                                               es->pc->info->id);
+          uint64_t cpicnt = sf.callPathNode->statistics.getValue(stats::instructions);
+
+          *os << "{";
+          *os << "'depth' : " << es->depth << ", ";
+          *os << "'weight' : " << es->weight << ", ";
+          *os << "'queryCost' : " << es->queryCost << ", ";
+          *os << "'coveredNew' : " << es->coveredNew << ", ";
+          *os << "'instsSinceCovNew' : " << es->instsSinceCovNew << ", ";
+          *os << "'md2u' : " << md2u << ", ";
+          *os << "'icnt' : " << icnt << ", ";
+          *os << "'CPicnt' : " << cpicnt << ", ";
+          *os << "}";
+          *os << ")\n";
+        }
+        
+        delete os;
+      }
+
+      dumpStates = 0;
+    }
+
+    if (maxInstTime>0 && current && !removedStates.count(current)) {
+      if (timerTicks*kSecondsPerTick > maxInstTime) {
+        klee_warning("max-instruction-time exceeded: %.2fs",
+                     timerTicks*kSecondsPerTick);
+        terminateStateEarly(*current, "max-instruction-time exceeded");
+      }
+    }
+
+    if (!timers.empty()) {
+      double time = util::getWallTime();
+
+      for (std::vector<TimerInfo*>::iterator it = timers.begin(), 
+             ie = timers.end(); it != ie; ++it) {
+        TimerInfo *ti = *it;
+        
+        if (time >= ti->nextFireTime) {
+          ti->timer->run();
+          ti->nextFireTime = time + ti->rate;
+        }
+      }
+    }
+
+    timerTicks = 0;
+    callsWithoutCheck = 0;
+  }
+}
+

Added: klee/trunk/lib/Core/ExecutorUtil.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ExecutorUtil.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ExecutorUtil.cpp (added)
+++ klee/trunk/lib/Core/ExecutorUtil.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,144 @@
+//===-- ExecutorUtil.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Executor.h"
+
+#include "klee/Expr.h"
+#include "klee/Interpreter.h"
+#include "klee/Machine.h"
+#include "klee/Solver.h"
+
+#include "klee/Internal/Module/KModule.h"
+
+#include "llvm/Constants.h"
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+#include "llvm/ModuleProvider.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/Support/GetElementPtrTypeIterator.h"
+#include "llvm/Support/Streams.h"
+#include "llvm/Target/TargetData.h"
+#include <iostream>
+#include <cassert>
+
+using namespace klee;
+using namespace llvm;
+
+namespace klee {
+
+ref<Expr> 
+Executor::evalConstantExpr(llvm::ConstantExpr *ce) {
+  const llvm::Type *type = ce->getType();
+
+  ref<Expr> op1(0,Expr::Bool), op2(0,Expr::Bool), op3(0,Expr::Bool);
+  int numOperands = ce->getNumOperands();
+
+  if (numOperands > 0) op1 = evalConstant(ce->getOperand(0));
+  if (numOperands > 1) op2 = evalConstant(ce->getOperand(1));
+  if (numOperands > 2) op3 = evalConstant(ce->getOperand(2));
+  
+  switch (ce->getOpcode()) {
+    case Instruction::Trunc: return ExtractExpr::createByteOff(op1, 
+							       0,
+							       Expr::getWidthForLLVMType(type));
+    case Instruction::ZExt:  return  ZExtExpr::create(op1, 
+                                                      Expr::getWidthForLLVMType(type));
+    case Instruction::SExt:  return  SExtExpr::create(op1, 
+                                                      Expr::getWidthForLLVMType(type));
+    case Instruction::Add:   return   AddExpr::create(op1, op2);
+    case Instruction::Sub:   return   SubExpr::create(op1, op2);
+    case Instruction::Mul:   return   MulExpr::create(op1, op2);
+    case Instruction::SDiv:  return  SDivExpr::create(op1, op2);
+    case Instruction::UDiv:  return  UDivExpr::create(op1, op2);
+    case Instruction::SRem:  return  SRemExpr::create(op1, op2);
+    case Instruction::URem:  return  URemExpr::create(op1, op2);
+    case Instruction::And:   return   AndExpr::create(op1, op2);
+    case Instruction::Or:    return    OrExpr::create(op1, op2);
+    case Instruction::Xor:   return   XorExpr::create(op1, op2);
+    case Instruction::Shl:   return   ShlExpr::create(op1, op2);
+    case Instruction::LShr:  return  LShrExpr::create(op1, op2);
+    case Instruction::AShr:  return  AShrExpr::create(op1, op2);
+    case Instruction::BitCast:  return op1;
+
+    case Instruction::IntToPtr: {
+      return ZExtExpr::create(op1, Expr::getWidthForLLVMType(type));
+    } 
+
+    case Instruction::PtrToInt: {
+      return ZExtExpr::create(op1, Expr::getWidthForLLVMType(type));
+    }
+      
+    case Instruction::GetElementPtr: {
+      ref<Expr> base = op1;
+
+      for (gep_type_iterator ii = gep_type_begin(ce), ie = gep_type_end(ce);
+           ii != ie; ++ii) {
+        ref<Expr> addend(0, kMachinePointerType);
+        
+        if (const StructType *st = dyn_cast<StructType>(*ii)) {
+          const StructLayout *sl = kmodule->targetData->getStructLayout(st);
+          const ConstantInt *ci = cast<ConstantInt>(ii.getOperand());
+          
+          addend = Expr::createPointer(sl->getElementOffset((unsigned)
+                                                            ci->getZExtValue()));
+        } else {
+          const SequentialType *st = cast<SequentialType>(*ii);
+          ref<Expr> index = evalConstant(cast<Constant>(ii.getOperand()));
+          unsigned elementSize = kmodule->targetData->getTypeStoreSize(st->getElementType());
+          
+          index = Expr::createCoerceToPointerType(index);
+          addend = MulExpr::create(index,
+                                   Expr::createPointer(elementSize));
+        }
+        
+        base = AddExpr::create(base, addend);
+      }
+      
+      return base;
+    }
+      
+    case Instruction::ICmp: {
+      switch(ce->getPredicate()) {
+        case ICmpInst::ICMP_EQ:  return  EqExpr::create(op1, op2);
+        case ICmpInst::ICMP_NE:  return  NeExpr::create(op1, op2);
+        case ICmpInst::ICMP_UGT: return UgtExpr::create(op1, op2);
+        case ICmpInst::ICMP_UGE: return UgeExpr::create(op1, op2);
+        case ICmpInst::ICMP_ULT: return UltExpr::create(op1, op2);
+        case ICmpInst::ICMP_ULE: return UleExpr::create(op1, op2);
+        case ICmpInst::ICMP_SGT: return SgtExpr::create(op1, op2);
+        case ICmpInst::ICMP_SGE: return SgeExpr::create(op1, op2);
+        case ICmpInst::ICMP_SLT: return SltExpr::create(op1, op2);
+        case ICmpInst::ICMP_SLE: return SleExpr::create(op1, op2);
+        default:
+          assert(0 && "unhandled ICmp predicate");
+      }
+    }
+      
+    case Instruction::Select: {
+      return SelectExpr::create(op1, op2, op3);
+    }
+
+    case Instruction::FDiv:
+    case Instruction::FRem:
+    case Instruction::FPTrunc:
+    case Instruction::FPExt:
+    case Instruction::UIToFP:
+    case Instruction::SIToFP:
+    case Instruction::FPToUI:
+    case Instruction::FPToSI:
+    case Instruction::FCmp:
+      assert(0 && "floating point ConstantExprs unsupported");
+
+    default :
+      assert(0 && "unknown ConstantExpr type");
+  }
+}
+
+}

Added: klee/trunk/lib/Core/ExternalDispatcher.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ExternalDispatcher.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ExternalDispatcher.cpp (added)
+++ klee/trunk/lib/Core/ExternalDispatcher.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,230 @@
+//===-- ExternalDispatcher.cpp --------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "ExternalDispatcher.h"
+
+#include "llvm/Module.h"
+#include "llvm/Constants.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/Instructions.h"
+#include "llvm/ModuleProvider.h"
+#include "llvm/ExecutionEngine/JIT.h"
+#include "llvm/ExecutionEngine/GenericValue.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/System/DynamicLibrary.h"
+#include "llvm/Support/Streams.h"
+#include "llvm/Support/raw_ostream.h"
+#include <setjmp.h>
+#include <signal.h>
+
+using namespace llvm;
+using namespace klee;
+
+/***/
+
+static jmp_buf escapeCallJmpBuf;
+
+extern "C" {
+
+static void sigsegv_handler(int signal, siginfo_t *info, void *context) {
+  longjmp(escapeCallJmpBuf, 1);
+}
+
+}
+
+void *ExternalDispatcher::resolveSymbol(const std::string &name) {
+  assert(executionEngine);
+  
+  const char *str = name.c_str();
+
+  // We use this to validate that function names can be resolved so we
+  // need to match how the JIT does it. Unfortunately we can't
+  // directly access the JIT resolution function
+  // JIT::getPointerToNamedFunction so we emulate the important points.
+
+  if (str[0] == 1) // asm specifier, skipped
+    ++str;
+
+  void *addr = dl_symbols.SearchForAddressOfSymbol(str);
+  if (addr)
+    return addr;
+  
+  // If it has an asm specifier and starts with an underscore we retry
+  // without the underscore. I (DWD) don't know why.
+  if (name[0] == 1 && str[0]=='_') { 
+    ++str;
+    addr = dl_symbols.SearchForAddressOfSymbol(str);
+  }
+
+  return addr;
+}
+
+ExternalDispatcher::ExternalDispatcher() {
+  dispatchModule = new Module("ExternalDispatcher");
+  ExistingModuleProvider* MP = new ExistingModuleProvider(dispatchModule);
+  
+  std::string error;
+  executionEngine = ExecutionEngine::createJIT(MP, &error);
+  if (!executionEngine) {
+    llvm::cerr << "unable to make jit: " << error << "\n";
+    abort();
+  }
+
+  // from ExecutionEngine::create
+  if (executionEngine) {
+    // Make sure we can resolve symbols in the program as well. The zero arg
+    // to the function tells DynamicLibrary to load the program, not a library.
+    try {
+      dl_symbols.LoadLibraryPermanently(0);
+    } catch (...) {
+      assert(0 && "Exception in LoadLibraryPermantently.\n");
+    }
+  }
+
+#ifdef WINDOWS
+  preboundFunctions["getpid"] = (void*) (long) getpid;
+  preboundFunctions["putchar"] = (void*) (long) putchar;
+  preboundFunctions["printf"] = (void*) (long) printf;
+  preboundFunctions["fprintf"] = (void*) (long) fprintf;
+  preboundFunctions["sprintf"] = (void*) (long) sprintf;
+#endif
+}
+
+ExternalDispatcher::~ExternalDispatcher() {
+  delete executionEngine;
+}
+
+bool ExternalDispatcher::executeCall(Function *f, Instruction *i, uint64_t *args) {
+  dispatchers_ty::iterator it = dispatchers.find(i);
+  Function *dispatcher;
+
+  if (it == dispatchers.end()) {
+#ifdef WINDOWS
+    std::map<std::string, void*>::iterator it2 = 
+      preboundFunctions.find(f->getName()));
+
+    if (it2 != preboundFunctions.end()) {
+      // only bind once
+      if (it2->second) {
+        executionEngine->addGlobalMapping(f, it2->second);
+        it2->second = 0;
+      }
+    }
+#endif
+
+    dispatcher = createDispatcher(f,i);
+
+    dispatchers.insert(std::make_pair(i, dispatcher));
+
+    if (dispatcher) {
+      // force the JIT execution engine to go ahead and build the
+      // function. this ensures that any errors or assertions in the
+      // compilation process will trigger crashes instead of being
+      // caught as aborts in the external function.
+      executionEngine->recompileAndRelinkFunction(dispatcher);
+    }
+  } else {
+    dispatcher = it->second;
+  }
+
+  return runProtectedCall(dispatcher, args);
+}
+
+// XXX not reentrant
+static uint64_t *gTheArgsP;
+
+bool ExternalDispatcher::runProtectedCall(Function *f, uint64_t *args) {
+  struct sigaction segvAction, segvActionOld;
+  bool res;
+  
+  if (!f)
+    return false;
+
+  std::vector<GenericValue> gvArgs;
+  gTheArgsP = args;
+
+  segvAction.sa_handler = 0;
+  memset(&segvAction.sa_mask, 0, sizeof(segvAction.sa_mask));
+  segvAction.sa_flags = SA_SIGINFO;
+  segvAction.sa_sigaction = ::sigsegv_handler;
+  sigaction(SIGSEGV, &segvAction, &segvActionOld);
+
+  if (setjmp(escapeCallJmpBuf)) {
+    res = false;
+  } else {
+    executionEngine->runFunction(f, gvArgs);
+    res = true;
+  }
+
+  sigaction(SIGSEGV, &segvActionOld, 0);
+  return res;
+}
+
+// for performance purposes we construct the stub in such a way that
+// the arguments pointer is passed through the static global variable
+// gTheArgsP in this file. This is done so that the stub function
+// prototype trivially matches the special cases that the JIT knows
+// how to directly call. If this is not done, then the jit will end up
+// generating a nullary stub just to call our stub, for every single
+// function call.
+Function *ExternalDispatcher::createDispatcher(Function *target, Instruction *inst) {
+  if (!resolveSymbol(target->getName()))
+    return 0;
+
+  CallSite cs;
+  if (inst->getOpcode()==Instruction::Call) {
+    cs = CallSite(cast<CallInst>(inst));
+  } else {
+    cs = CallSite(cast<InvokeInst>(inst));
+  }
+
+  Value **args = new Value*[cs.arg_size()];
+
+  std::vector<const Type*> nullary;
+  
+  Function *dispatcher = Function::Create(FunctionType::get(Type::VoidTy, 
+							    nullary, false),
+					  GlobalVariable::ExternalLinkage, 
+					  "",
+					  dispatchModule);
+
+
+  BasicBlock *dBB = BasicBlock::Create("entry", dispatcher);
+
+  Instruction *argI64sp = new IntToPtrInst(ConstantInt::get(Type::Int64Ty, (long) (void*) &gTheArgsP),
+					   PointerType::getUnqual(PointerType::getUnqual(Type::Int64Ty)),
+					   "argsp",
+					   dBB);
+  Instruction *argI64s = new LoadInst(argI64sp, "args", dBB); 
+
+  unsigned i = 0;
+  for (CallSite::arg_iterator ai = cs.arg_begin(), ae = cs.arg_end();
+       ai!=ae; ++ai, ++i) {
+    Value *index = ConstantInt::get(Type::Int32Ty, i+1);
+
+    Instruction *argI64p = GetElementPtrInst::Create(argI64s, index, "", dBB);
+    Instruction *argp = new BitCastInst(argI64p, 
+                                        PointerType::getUnqual((*ai)->getType()), "", dBB);
+    args[i] = new LoadInst(argp, "", dBB);
+  }
+
+  Instruction *result = CallInst::Create(target, args, args+i, "", dBB);
+
+  if (result->getType() != Type::VoidTy) {
+    Instruction *resp = new BitCastInst(argI64s, 
+                                        PointerType::getUnqual(result->getType()), "", dBB);
+    new StoreInst(result, resp, dBB);
+  }
+
+  ReturnInst::Create(dBB);
+
+  delete[] args;
+
+  return dispatcher;
+}

Added: klee/trunk/lib/Core/ExternalDispatcher.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ExternalDispatcher.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ExternalDispatcher.h (added)
+++ klee/trunk/lib/Core/ExternalDispatcher.h Wed May 20 23:36:41 2009
@@ -0,0 +1,50 @@
+//===-- ExternalDispatcher.h ------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_EXTERNALDISPATCHER_H
+#define KLEE_EXTERNALDISPATCHER_H
+
+#include <map>
+#include "llvm/System/DynamicLibrary.h"
+
+namespace llvm {
+  class ExecutionEngine;
+  class Instruction;
+  class Function;
+  class FunctionType;
+  class Module;
+}
+
+namespace klee {
+  class ExternalDispatcher {
+  private:
+    typedef std::map<const llvm::Instruction*,llvm::Function*> dispatchers_ty;
+    dispatchers_ty dispatchers;
+    llvm::Module *dispatchModule;
+    llvm::ExecutionEngine *executionEngine;
+    llvm::sys::DynamicLibrary dl_symbols;
+    std::map<std::string, void*> preboundFunctions;
+    
+    llvm::Function *createDispatcher(llvm::Function *f, llvm::Instruction *i);
+    bool runProtectedCall(llvm::Function *f, uint64_t *args);
+    
+  public:
+    ExternalDispatcher();
+    ~ExternalDispatcher();
+
+    /* Call the given function using the parameter passing convention of
+     * ci with arguments in args[1], args[2], ... and writing the result
+     * into args[0].
+     */
+    bool executeCall(llvm::Function *function, llvm::Instruction *i, uint64_t *args);
+    void *resolveSymbol(const std::string &name);
+  };  
+}
+
+#endif

Added: klee/trunk/lib/Core/ImpliedValue.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ImpliedValue.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ImpliedValue.cpp (added)
+++ klee/trunk/lib/Core/ImpliedValue.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,274 @@
+//===-- ImpliedValue.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "ImpliedValue.h"
+
+#include "klee/Constraints.h"
+#include "klee/Expr.h"
+#include "klee/Solver.h"
+// FIXME: Use APInt.
+#include "klee/Internal/Support/IntEvaluation.h"
+
+#include "klee/util/ExprUtil.h"
+
+#include <iostream>
+#include <map>
+#include <set>
+
+using namespace klee;
+
+// XXX we really want to do some sort of canonicalization of exprs
+// globally so that cases below become simpler
+static void _getImpliedValue(ref<Expr> e,
+                             uint64_t value,
+                             ImpliedValueList &results) {
+  switch (e.getKind()) {
+    
+  case Expr::Constant: {
+    assert(value == e.getConstantValue() && "error in implied value calculation");
+    break;
+  }
+
+    // Special
+
+  case Expr::NotOptimized: break;
+
+  case Expr::Read: {
+    // XXX in theory it is possible to descend into a symbolic index
+    // under certain circumstances (all values known, known value
+    // unique, or range known, max / min hit). Seems unlikely this
+    // would work often enough to be worth the effort.
+    ReadExpr *re = static_ref_cast<ReadExpr>(e);
+    results.push_back(std::make_pair(re, 
+                                     ConstantExpr::create(value, e.getWidth())));
+    break;
+  }
+    
+  case Expr::Select: {
+    // not much to do, could improve with range analysis
+    SelectExpr *se = static_ref_cast<SelectExpr>(e);
+    
+    if (se->trueExpr.isConstant()) {
+      if (se->falseExpr.isConstant()) {
+        if (se->trueExpr.getConstantValue() != se->falseExpr.getConstantValue()) {
+          if (value == se->trueExpr.getConstantValue()) {
+            _getImpliedValue(se->cond, 1, results);
+          } else {
+            assert(value == se->falseExpr.getConstantValue() &&
+                   "err in implied value calculation");
+            _getImpliedValue(se->cond, 0, results);
+          }
+        }
+      }
+    }
+    break;
+  }
+
+  case Expr::Concat: {
+    ConcatExpr *ce = static_ref_cast<ConcatExpr>(e);
+    _getImpliedValue(ce->getKid(0), (value >> ce->getKid(1).getWidth()) & ((1 << ce->getKid(0).getWidth()) - 1), results);
+    _getImpliedValue(ce->getKid(1), value & ((1 << ce->getKid(1).getWidth()) - 1), results);
+    break;
+  }
+    
+  case Expr::Extract: {
+    // XXX, could do more here with "some bits" mask
+    break;
+  }
+
+    // Casting
+
+  case Expr::ZExt: 
+  case Expr::SExt: {
+    CastExpr *ce = static_ref_cast<CastExpr>(e);
+    _getImpliedValue(ce->src, 
+                     bits64::truncateToNBits(value, 
+					     ce->src.getWidth()),
+                     results);
+    break;
+  }
+
+    // Arithmetic
+
+  case Expr::Add: { // constants on left
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (be->left.isConstant()) {
+      uint64_t nvalue = ints::sub(value,
+                                  be->left.getConstantValue(),
+                                  be->left.getWidth());
+      _getImpliedValue(be->right, nvalue, results);
+    }
+    break;
+  }
+  case Expr::Sub: { // constants on left
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (be->left.isConstant()) {
+      uint64_t nvalue = ints::sub(be->left.getConstantValue(),
+                                  value,
+                                  be->left.getWidth());
+      _getImpliedValue(be->right, nvalue, results);
+    }
+    break;
+  }
+  case Expr::Mul: {
+    // XXX can do stuff here, but need valid mask and other things
+    // because of bits that might be lost
+    break;
+  }
+
+  case Expr::UDiv:
+  case Expr::SDiv:
+  case Expr::URem:
+  case Expr::SRem:
+    // no, no, no
+    break;
+
+    // Binary
+
+  case Expr::And: {
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (be->getWidth() == Expr::Bool) {
+      if (value) {
+        _getImpliedValue(be->left, value, results);
+        _getImpliedValue(be->right, value, results);
+      }
+    } else {
+      // XXX, we can basically propogate a mask here
+      // where we know "some bits". may or may not be
+      // useful.
+    }
+    break;
+  }
+  case Expr::Or: {
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (!value) {
+      _getImpliedValue(be->left, 0, results);
+      _getImpliedValue(be->right, 0, results);
+    } else {
+      // XXX, can do more?
+    }
+    break;
+  }
+  case Expr::Xor: { // constants on left
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (be->left.isConstant()) {
+      _getImpliedValue(be->right, value ^ be->left.getConstantValue(), results);
+    }
+    break;
+  }
+
+    // Comparison
+  case Expr::Ne: 
+    value = !value;
+    /* fallthru */
+  case Expr::Eq: {
+    EqExpr *ee = static_ref_cast<EqExpr>(e);
+    if (value) {
+      if (ee->left.isConstant())
+        _getImpliedValue(ee->right, ee->left.getConstantValue(), results);
+    } else {
+      // look for limited value range, woohoo
+      //
+      // in general anytime one side was restricted to two values we
+      // can apply this trick. the only obvious case where this
+      // occurs, aside from booleans, is as the result of a select
+      // expression where the true and false branches are single
+      // valued and distinct.
+      
+      if (ee->left.isConstant()) {
+        if (ee->left.getWidth() == Expr::Bool) {
+          _getImpliedValue(ee->right, !ee->left.getConstantValue(), results);
+        }
+      }
+    }
+    break;
+  }
+    
+  default:
+    break;
+  }
+}
+    
+void ImpliedValue::getImpliedValues(ref<Expr> e, 
+                                    ref<Expr> value, 
+                                    ImpliedValueList &results) {
+  assert(value.isConstant() && "non-constant in place of constant");
+  _getImpliedValue(e, value.getConstantValue(), results);
+}
+
+void ImpliedValue::checkForImpliedValues(Solver *S, ref<Expr> e, 
+                                         ref<Expr> value) {
+  assert(value.isConstant() && "non-constant in place of constant");
+
+  std::vector<ref<ReadExpr> > reads;
+  std::map<ref<ReadExpr>, ref<Expr> > found;
+  ImpliedValueList results;
+
+  getImpliedValues(e, value, results);
+
+  for (ImpliedValueList::iterator i = results.begin(), ie = results.end();
+       i != ie; ++i) {
+    std::map<ref<ReadExpr>, ref<Expr> >::iterator it = found.find(i->first);
+    if (it != found.end()) {
+      assert(it->second.getConstantValue() == i->second.getConstantValue() &&
+             "I don't think so Scott");
+    } else {
+      found.insert(std::make_pair(i->first, i->second));
+    }
+  }
+
+  findReads(e, false, reads);
+  std::set< ref<ReadExpr> > readsSet(reads.begin(), reads.end());
+  reads = std::vector< ref<ReadExpr> >(readsSet.begin(), readsSet.end());
+
+  std::vector<ref<Expr> > assumption;
+  assumption.push_back(EqExpr::create(e, value));
+
+  // obscure... we need to make sure that all the read indices are
+  // bounds checked. if we don't do this we can end up constructing
+  // invalid counterexamples because STP will happily make out of
+  // bounds indices which will not get picked up. this is of utmost
+  // importance if we are being backed by the CexCachingSolver.
+
+  for (std::vector< ref<ReadExpr> >::iterator i = reads.begin(), 
+         ie = reads.end(); i != ie; ++i) {
+    ReadExpr *re = i->get();
+    ref<Expr> size = ref<Expr>(re->updates.root->size, kMachinePointerType);
+    assumption.push_back(UltExpr::create(re->index, size));
+  }
+
+  ConstraintManager assume(assumption);
+  for (std::vector< ref<ReadExpr> >::iterator i = reads.begin(), 
+         ie = reads.end(); i != ie; ++i) {
+    ref<ReadExpr> var = *i;
+    ref<Expr> possible;
+    bool success = S->getValue(Query(assume, var), possible);
+    assert(success && "FIXME: Unhandled solver failure");    
+    std::map<ref<ReadExpr>, ref<Expr> >::iterator it = found.find(var);
+    bool res;
+    success = S->mustBeTrue(Query(assume, EqExpr::create(var, possible)), res);
+    assert(success && "FIXME: Unhandled solver failure");    
+    if (res) {
+      if (it != found.end()) {
+        assert(possible.getConstantValue() == it->second.getConstantValue());
+        found.erase(it);
+      }
+    } else {
+      if (it!=found.end()) {
+        ref<Expr> binding = it->second;
+        llvm::cerr << "checkForImpliedValues: " << e  << " = " << value << "\n"
+                   << "\t\t implies " << var << " == " << binding
+                   << " (error)\n";
+        assert(0);
+      }
+    }
+  }
+
+  assert(found.empty());
+}

Added: klee/trunk/lib/Core/ImpliedValue.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ImpliedValue.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ImpliedValue.h (added)
+++ klee/trunk/lib/Core/ImpliedValue.h Wed May 20 23:36:41 2009
@@ -0,0 +1,38 @@
+//===-- ImpliedValue.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_IMPLIEDVALUE_H
+#define KLEE_IMPLIEDVALUE_H
+
+#include "klee/Expr.h"
+
+#include <vector>
+
+// The idea of implied values is that often we know the result of some
+// expression e is a concrete value C. In many cases this directly
+// implies that some variable x embedded in e is also a concrete value
+// (derived from C). This module is used for finding such variables
+// and their computed values.
+
+namespace klee {
+  class ConstantExpr;
+  class Expr;
+  class ReadExpr;
+  class Solver;
+
+  typedef std::vector< std::pair<ref<ReadExpr>, ref<Expr> > > ImpliedValueList;
+  
+  namespace ImpliedValue {        
+    void getImpliedValues(ref<Expr> e, ref<Expr> cvalue, ImpliedValueList &result);
+    void checkForImpliedValues(Solver *S, ref<Expr> e, ref<Expr> cvalue);    
+  }
+
+}
+
+#endif

Added: klee/trunk/lib/Core/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Makefile (added)
+++ klee/trunk/lib/Core/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,16 @@
+#===-- lib/Core/Makefile -----------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+LEVEL=../..
+
+LIBRARYNAME=kleeCore
+DONT_BUILD_RELINKED=1
+BUILD_ARCHIVE=1
+
+include $(LEVEL)/Makefile.common

Propchange: klee/trunk/lib/Core/Makefile

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/lib/Core/Memory.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Memory.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Memory.cpp (added)
+++ klee/trunk/lib/Core/Memory.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,812 @@
+//===-- Memory.cpp --------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "Memory.h"
+
+#include "klee/Expr.h"
+#include "klee/Machine.h"
+#include "klee/Solver.h"
+#include "klee/util/BitArray.h"
+
+#include "ObjectHolder.h"
+
+#include <llvm/Function.h>
+#include <llvm/Instruction.h>
+#include <llvm/Value.h>
+
+#include <iostream>
+#include <cassert>
+#include <sstream>
+
+using namespace llvm;
+using namespace klee;
+
+/***/
+
+ObjectHolder::ObjectHolder(const ObjectHolder &b) : os(b.os) { 
+  if (os) ++os->refCount; 
+}
+
+ObjectHolder::ObjectHolder(ObjectState *_os) : os(_os) { 
+  if (os) ++os->refCount; 
+}
+
+ObjectHolder::~ObjectHolder() { 
+  if (os && --os->refCount==0) delete os; 
+}
+  
+ObjectHolder &ObjectHolder::operator=(const ObjectHolder &b) {
+  if (b.os) ++b.os->refCount;
+  if (os && --os->refCount==0) delete os;
+  os = b.os;
+  return *this;
+}
+
+/***/
+
+int MemoryObject::counter = 0;
+
+extern "C" void vc_DeleteExpr(void*);
+
+MemoryObject::~MemoryObject() {
+  // FIXME: This shouldn't be necessary. Array's should be ref-counted
+  // just like everything else, and the interaction with the STP array
+  // should hide at least inside the Expr/Solver layers.
+  if (array) {
+    if (array->stpInitialArray) {
+      ::vc_DeleteExpr(array->stpInitialArray);
+      array->stpInitialArray = 0;
+    }
+    delete array;
+  }
+}
+
+void MemoryObject::getAllocInfo(std::string &result) const {
+  std::ostringstream info;
+
+  info << "MO" << id << "[" << size << "]";
+
+  if (allocSite) {
+    info << " allocated at ";
+    if (const Instruction *i = dyn_cast<Instruction>(allocSite)) {
+      info << i->getParent()->getParent()->getName() << "():";
+      info << *i;
+    } else if (const GlobalValue *gv = dyn_cast<GlobalValue>(allocSite)) {
+      info << "global:" << gv->getName();
+    } else {
+      info << "value:" << *allocSite;
+    }
+  } else {
+    info << " (no allocation info)";
+  }
+  
+  result = info.str();
+}
+
+/***/
+
+ObjectState::ObjectState(const MemoryObject *mo, unsigned _size)
+  : copyOnWriteOwner(0),
+    refCount(0),
+    object(mo),
+    concreteStore(new uint8_t[_size]),
+    concreteMask(0),
+    flushMask(0),
+    knownSymbolics(0),
+    size(_size),
+    updates(mo->array, false, 0),
+    readOnly(false) {
+}
+
+ObjectState::ObjectState(const ObjectState &os) 
+  : copyOnWriteOwner(0),
+    refCount(0),
+    object(os.object),
+    concreteStore(new uint8_t[os.size]),
+    concreteMask(os.concreteMask ? new BitArray(*os.concreteMask, os.size) : 0),
+    flushMask(os.flushMask ? new BitArray(*os.flushMask, os.size) : 0),
+    knownSymbolics(0),
+    size(os.size),
+    updates(os.updates),
+    readOnly(false) {
+  assert(!os.readOnly && "no need to copy read only object?");
+
+  if (os.knownSymbolics) {
+    knownSymbolics = new ref<Expr>[size];
+    for (unsigned i=0; i<size; i++)
+      knownSymbolics[i] = os.knownSymbolics[i];
+  }
+
+  memcpy(concreteStore, os.concreteStore, size*sizeof(*concreteStore));
+}
+
+ObjectState::~ObjectState() {
+  if (concreteMask) delete concreteMask;
+  if (flushMask) delete flushMask;
+  if (knownSymbolics) delete[] knownSymbolics;
+  delete[] concreteStore;
+}
+
+/***/
+
+void ObjectState::makeConcrete() {
+  if (concreteMask) delete concreteMask;
+  if (flushMask) delete flushMask;
+  if (knownSymbolics) delete[] knownSymbolics;
+  concreteMask = 0;
+  flushMask = 0;
+  knownSymbolics = 0;
+}
+
+void ObjectState::makeSymbolic() {
+  assert(!updates.head &&
+         "XXX makeSymbolic of objects with symbolic values is unsupported");
+  updates.isRooted = true;
+
+  // XXX simplify this, can just delete various arrays I guess
+  for (unsigned i=0; i<size; i++) {
+    markByteSymbolic(i);
+    setKnownSymbolic(i, 0);
+    markByteFlushed(i);
+  }
+}
+
+void ObjectState::initializeToZero() {
+  makeConcrete();
+  memset(concreteStore, 0, size);
+}
+
+void ObjectState::initializeToRandom() {  
+  makeConcrete();
+  for (unsigned i=0; i<size; i++) {
+    // randomly selected by 256 sided die
+    concreteStore[i] = 0xAB;
+  }
+}
+
+/*
+Cache Invariants
+--
+isByteKnownSymbolic(i) => !isByteConcrete(i)
+isByteConcrete(i) => !isByteKnownSymbolic(i)
+!isByteFlushed(i) => (isByteConcrete(i) || isByteKnownSymbolic(i))
+ */
+
+void ObjectState::fastRangeCheckOffset(ref<Expr> offset,
+                                       unsigned *base_r,
+                                       unsigned *size_r) const {
+  *base_r = 0;
+  *size_r = size;
+}
+
+void ObjectState::flushRangeForRead(unsigned rangeBase, 
+                                    unsigned rangeSize) const {
+  if (!flushMask) flushMask = new BitArray(size, true);
+ 
+  for (unsigned offset=rangeBase; offset<rangeBase+rangeSize; offset++) {
+    if (!isByteFlushed(offset)) {
+      if (isByteConcrete(offset)) {
+        updates.extend(ConstantExpr::create(offset, kMachinePointerType),
+                       ConstantExpr::create(concreteStore[offset], Expr::Int8));
+      } else {
+        assert(isByteKnownSymbolic(offset) && "invalid bit set in flushMask");
+        updates.extend(ConstantExpr::create(offset, kMachinePointerType),
+                       knownSymbolics[offset]);
+      }
+
+      flushMask->unset(offset);
+    }
+  } 
+}
+
+void ObjectState::flushRangeForWrite(unsigned rangeBase, 
+                                     unsigned rangeSize) {
+  if (!flushMask) flushMask = new BitArray(size, true);
+
+  for (unsigned offset=rangeBase; offset<rangeBase+rangeSize; offset++) {
+    if (!isByteFlushed(offset)) {
+      if (isByteConcrete(offset)) {
+        updates.extend(ConstantExpr::create(offset, kMachinePointerType),
+                       ConstantExpr::create(concreteStore[offset], Expr::Int8));
+        markByteSymbolic(offset);
+      } else {
+        assert(isByteKnownSymbolic(offset) && "invalid bit set in flushMask");
+        updates.extend(ConstantExpr::create(offset, kMachinePointerType),
+                       knownSymbolics[offset]);
+        setKnownSymbolic(offset, 0);
+      }
+
+      flushMask->unset(offset);
+    } else {
+      // flushed bytes that are written over still need
+      // to be marked out
+      if (isByteConcrete(offset)) {
+        markByteSymbolic(offset);
+      } else if (isByteKnownSymbolic(offset)) {
+        setKnownSymbolic(offset, 0);
+      }
+    }
+  } 
+}
+
+bool ObjectState::isByteConcrete(unsigned offset) const {
+  return !concreteMask || concreteMask->get(offset);
+}
+
+bool ObjectState::isByteFlushed(unsigned offset) const {
+  return flushMask && !flushMask->get(offset);
+}
+
+bool ObjectState::isByteKnownSymbolic(unsigned offset) const {
+  return knownSymbolics && knownSymbolics[offset].get();
+}
+
+void ObjectState::markByteConcrete(unsigned offset) {
+  if (concreteMask)
+    concreteMask->set(offset);
+}
+
+void ObjectState::markByteSymbolic(unsigned offset) {
+  if (!concreteMask)
+    concreteMask = new BitArray(size, true);
+  concreteMask->unset(offset);
+}
+
+void ObjectState::markByteUnflushed(unsigned offset) {
+  if (flushMask)
+    flushMask->set(offset);
+}
+
+void ObjectState::markByteFlushed(unsigned offset) {
+  if (!flushMask) {
+    flushMask = new BitArray(size, false);
+  } else {
+    flushMask->unset(offset);
+  }
+}
+
+void ObjectState::setKnownSymbolic(unsigned offset, 
+                                   Expr *value /* can be null */) {
+  if (knownSymbolics) {
+    knownSymbolics[offset] = value;
+  } else {
+    if (value) {
+      knownSymbolics = new ref<Expr>[size];
+      knownSymbolics[offset] = value;
+    }
+  }
+}
+
+/***/
+
+ref<Expr> ObjectState::read8(unsigned offset) const {
+  if (isByteConcrete(offset)) {
+    return ConstantExpr::create(concreteStore[offset], Expr::Int8);
+  } else if (isByteKnownSymbolic(offset)) {
+    return knownSymbolics[offset];
+  } else {
+    assert(isByteFlushed(offset) && "unflushed byte without cache value");
+    
+    return ReadExpr::create(updates, 
+                            ConstantExpr::create(offset, kMachinePointerType));
+  }    
+}
+
+ref<Expr> ObjectState::read8(ref<Expr> offset) const {
+  assert(!offset.isConstant() && "constant offset passed to symbolic read8");
+  unsigned base, size;
+  fastRangeCheckOffset(offset, &base, &size);
+  flushRangeForRead(base, size);
+
+  if (size>4096) {
+    std::string allocInfo;
+    object->getAllocInfo(allocInfo);
+    klee_warning_once(0, "flushing %d bytes on read, may be slow and/or crash: %s", 
+                      size,
+                      allocInfo.c_str());
+  }
+  
+  return ReadExpr::create(updates, offset);
+}
+
+void ObjectState::write8(unsigned offset, uint8_t value) {
+  //assert(read_only == false && "writing to read-only object!");
+  concreteStore[offset] = value;
+  setKnownSymbolic(offset, 0);
+
+  markByteConcrete(offset);
+  markByteUnflushed(offset);
+}
+
+void ObjectState::write8(unsigned offset, ref<Expr> value) {
+  // can happen when ExtractExpr special cases
+  if (value.isConstant()) {
+    write8(offset, (uint8_t) value.getConstantValue());
+  } else {
+    setKnownSymbolic(offset, value.get());
+      
+    markByteSymbolic(offset);
+    markByteUnflushed(offset);
+  }
+}
+
+void ObjectState::write8(ref<Expr> offset, ref<Expr> value) {
+  assert(!offset.isConstant() && "constant offset passed to symbolic write8");
+  unsigned base, size;
+  fastRangeCheckOffset(offset, &base, &size);
+  flushRangeForWrite(base, size);
+
+  if (size>4096) {
+    std::string allocInfo;
+    object->getAllocInfo(allocInfo);
+    klee_warning_once(0, "flushing %d bytes on read, may be slow and/or crash: %s", 
+                      size,
+                      allocInfo.c_str());
+  }
+  
+  updates.extend(offset, value);
+}
+
+/***/
+
+ref<Expr> ObjectState::read(ref<Expr> offset, Expr::Width width) const {
+  if (offset.isConstant()) {
+    return read((unsigned) offset.getConstantValue(), width);
+  } else { 
+    switch (width) {
+    case  Expr::Bool: return  read1(offset);
+    case  Expr::Int8: return  read8(offset);
+    case Expr::Int16: return read16(offset);
+    case Expr::Int32: return read32(offset);
+    case Expr::Int64: return read64(offset);
+    default: assert(0 && "invalid type");
+    }
+  }
+}
+
+ref<Expr> ObjectState::read(unsigned offset, Expr::Width width) const {
+  switch (width) {
+  case  Expr::Bool: return  read1(offset);
+  case  Expr::Int8: return  read8(offset);
+  case Expr::Int16: return read16(offset);
+  case Expr::Int32: return read32(offset);
+  case Expr::Int64: return read64(offset);
+  default: assert(0 && "invalid type");
+  }
+}
+
+ref<Expr> ObjectState::read1(unsigned offset) const {
+  return ExtractExpr::createByteOff(read8(offset), 0, Expr::Bool);
+}
+
+ref<Expr> ObjectState::read1(ref<Expr> offset) const {
+  return ExtractExpr::createByteOff(read8(offset), 0, Expr::Bool);
+}
+
+ref<Expr> ObjectState::read16(unsigned offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create(read8(offset+0),
+			      read8(offset+1));
+  } else {
+    return ConcatExpr::create(read8(offset+1),
+			      read8(offset+0));
+  }
+}
+
+ref<Expr> ObjectState::read16(ref<Expr> offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))));
+  } else {
+    return ConcatExpr::create
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))));
+  }
+}
+
+ref<Expr> ObjectState::read32(unsigned offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create4(read8(offset+0),
+                               read8(offset+1),
+                               read8(offset+2),
+                               read8(offset+3));
+  } else {
+    return ConcatExpr::create4(read8(offset+3),
+                               read8(offset+2),
+                               read8(offset+1),
+                               read8(offset+0));
+  }
+}
+
+ref<Expr> ObjectState::read32(ref<Expr> offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create4
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(2,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(3,
+                                                  kMachinePointerType))));
+  } else {
+    return ConcatExpr::create4
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(3,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(2,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))));
+  }
+}
+
+ref<Expr> ObjectState::read64(unsigned offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create8(read8(offset+0),
+                               read8(offset+1),
+                               read8(offset+2),
+                               read8(offset+3),
+                               read8(offset+4),
+                               read8(offset+5),
+                               read8(offset+6),
+                               read8(offset+7));
+  } else {
+    return ConcatExpr::create8(read8(offset+7),
+                               read8(offset+6),
+                               read8(offset+5),
+                               read8(offset+4),
+                               read8(offset+3),
+                               read8(offset+2),
+                               read8(offset+1),
+                               read8(offset+0));
+  }
+}
+
+ref<Expr> ObjectState::read64(ref<Expr> offset) const {
+  if (kMachineByteOrder == machine::MSB) {
+    return ConcatExpr::create8
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(2,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(3,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(4,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(5,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(6,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(7,
+                                                  kMachinePointerType))));
+  } else {
+    return ConcatExpr::create8
+      (read8(AddExpr::create(offset,
+                             ConstantExpr::create(7,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(6,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(5,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(4,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(3,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(2,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(1,
+                                                  kMachinePointerType))),
+       read8(AddExpr::create(offset,
+                             ConstantExpr::create(0,
+                                                  kMachinePointerType))));
+  }
+}
+
+void ObjectState::write(ref<Expr> offset, ref<Expr> value) {
+  Expr::Width w = value.getWidth();
+  if (offset.isConstant()) {
+    write(offset.getConstantValue(), value);
+  } else {
+    switch(w) {
+    case  Expr::Bool:  write1(offset, value); break;
+    case  Expr::Int8:  write8(offset, value); break;
+    case Expr::Int16: write16(offset, value); break;
+    case Expr::Int32: write32(offset, value); break;
+    case Expr::Int64: write64(offset, value); break;
+    default: assert(0 && "invalid number of bytes in write");
+    }
+  }
+}
+
+void ObjectState::write(unsigned offset, ref<Expr> value) {
+  Expr::Width w = value.getWidth();
+  if (value.isConstant()) {
+    uint64_t val = value.getConstantValue();
+    switch(w) {
+    case  Expr::Bool:
+    case  Expr::Int8:  write8(offset, val); break;
+    case Expr::Int16: write16(offset, val); break;
+    case Expr::Int32: write32(offset, val); break;
+    case Expr::Int64: write64(offset, val); break;
+    default: assert(0 && "invalid number of bytes in write");
+    }
+  } else {
+    switch(w) {
+    case  Expr::Bool:  write1(offset, value); break;
+    case  Expr::Int8:  write8(offset, value); break;
+    case Expr::Int16: write16(offset, value); break;
+    case Expr::Int32: write32(offset, value); break;
+    case Expr::Int64: write64(offset, value); break;
+    default: assert(0 && "invalid number of bytes in write");
+    }
+  }
+}
+
+void ObjectState::write1(unsigned offset, ref<Expr> value) {
+  write8(offset, ZExtExpr::create(value, Expr::Int8));
+}
+
+void ObjectState::write1(ref<Expr> offset, ref<Expr> value) {
+  write8(offset, ZExtExpr::create(value, Expr::Int8));
+}
+
+void ObjectState::write16(unsigned offset, uint16_t value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, (uint8_t) (value >>  8));
+    write8(offset+1, (uint8_t) (value >>  0));
+  } else {
+    write8(offset+1, (uint8_t) (value >>  8));
+    write8(offset+0, (uint8_t) (value >>  0));
+  }
+}
+
+void ObjectState::write16(unsigned offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, ExtractExpr::createByteOff(value, 1));
+    write8(offset+1, ExtractExpr::createByteOff(value, 0));
+  } else {
+    write8(offset+1, ExtractExpr::createByteOff(value, 1));
+    write8(offset+0, ExtractExpr::createByteOff(value, 0));
+  }
+}
+
+
+void ObjectState::write16(ref<Expr> offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  } else {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(1, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  }
+}
+
+void ObjectState::write32(unsigned offset, uint32_t value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, (uint8_t) (value >>  24));
+    write8(offset+1, (uint8_t) (value >>  16));
+    write8(offset+2, (uint8_t) (value >>  8));
+    write8(offset+3, (uint8_t) (value >>  0));
+  } else {
+    write8(offset+3, (uint8_t) (value >>  24));
+    write8(offset+2, (uint8_t) (value >>  16));
+    write8(offset+1, (uint8_t) (value >>  8));
+    write8(offset+0, (uint8_t) (value >>  0));
+  }
+}
+
+void ObjectState::write32(unsigned offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, ExtractExpr::createByteOff(value, 3));
+    write8(offset+1, ExtractExpr::createByteOff(value, 2));
+    write8(offset+2, ExtractExpr::createByteOff(value, 1));
+    write8(offset+3, ExtractExpr::createByteOff(value, 0));
+  } else {
+    write8(offset+3, ExtractExpr::createByteOff(value, 3));
+    write8(offset+2, ExtractExpr::createByteOff(value, 2));
+    write8(offset+1, ExtractExpr::createByteOff(value, 1));
+    write8(offset+0, ExtractExpr::createByteOff(value, 0));
+  }
+}
+
+void ObjectState::write32(ref<Expr> offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,3));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(1, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,2));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(2, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(3, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  } else {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(3, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,3));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(2, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,2));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(1, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  }
+}
+
+void ObjectState::write64(unsigned offset, uint64_t value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, (uint8_t) (value >>  56));
+    write8(offset+1, (uint8_t) (value >>  48));
+    write8(offset+2, (uint8_t) (value >>  40));
+    write8(offset+3, (uint8_t) (value >>  32));
+    write8(offset+4, (uint8_t) (value >>  24));
+    write8(offset+5, (uint8_t) (value >>  16));
+    write8(offset+6, (uint8_t) (value >>  8));
+    write8(offset+7, (uint8_t) (value >>  0));
+  } else {
+    write8(offset+7, (uint8_t) (value >>  56));
+    write8(offset+6, (uint8_t) (value >>  48));
+    write8(offset+5, (uint8_t) (value >>  40));
+    write8(offset+4, (uint8_t) (value >>  32));
+    write8(offset+3, (uint8_t) (value >>  24));
+    write8(offset+2, (uint8_t) (value >>  16));
+    write8(offset+1, (uint8_t) (value >>  8));
+    write8(offset+0, (uint8_t) (value >>  0));
+  }
+}
+
+void ObjectState::write64(unsigned offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(offset+0, ExtractExpr::createByteOff(value, 7));
+    write8(offset+1, ExtractExpr::createByteOff(value, 6));
+    write8(offset+2, ExtractExpr::createByteOff(value, 5));
+    write8(offset+3, ExtractExpr::createByteOff(value, 4));
+    write8(offset+4, ExtractExpr::createByteOff(value, 3));
+    write8(offset+5, ExtractExpr::createByteOff(value, 2));
+    write8(offset+6, ExtractExpr::createByteOff(value, 1));
+    write8(offset+7, ExtractExpr::createByteOff(value, 0));
+  } else {
+    write8(offset+7, ExtractExpr::createByteOff(value, 7));
+    write8(offset+6, ExtractExpr::createByteOff(value, 6));
+    write8(offset+5, ExtractExpr::createByteOff(value, 5));
+    write8(offset+4, ExtractExpr::createByteOff(value, 4));
+    write8(offset+3, ExtractExpr::createByteOff(value, 3));
+    write8(offset+2, ExtractExpr::createByteOff(value, 2));
+    write8(offset+1, ExtractExpr::createByteOff(value, 1));
+    write8(offset+0, ExtractExpr::createByteOff(value, 0));
+  }
+}
+
+void ObjectState::write64(ref<Expr> offset, ref<Expr> value) {
+  if (kMachineByteOrder == machine::MSB) {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,7));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(1, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,6));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(2, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,5));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(3, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,4));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(4, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,3));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(5, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,2));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(6, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(7, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  } else {
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(7, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,7));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(6, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,6));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(5, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,5));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(4, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,4));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(3, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,3));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(2, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,2));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(1, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,1));
+    write8(AddExpr::create(offset,
+                           ConstantExpr::create(0, kMachinePointerType)), 
+           ExtractExpr::createByteOff(value,0));
+  }
+}
+
+void ObjectState::print() {
+  llvm::cerr << "-- ObjectState --\n";
+  llvm::cerr << "\tMemoryObject ID: " << object->id << "\n";
+  llvm::cerr << "\tRoot Object: " << updates.root << "\n";
+  llvm::cerr << "\tIs Rooted? " << updates.isRooted << "\n";
+  llvm::cerr << "\tSize: " << size << "\n";
+
+  llvm::cerr << "\tBytes:\n";
+  for (unsigned i=0; i<size; i++) {
+    llvm::cerr << "\t\t["<<i<<"]"
+               << " concrete? " << isByteConcrete(i)
+               << " known-sym? " << isByteKnownSymbolic(i)
+               << " flushed? " << isByteFlushed(i) << " = ";
+    ref<Expr> e = read8(i);
+    llvm::cerr << e << "\n";
+  }
+
+  llvm::cerr << "\tUpdates:\n";
+  for (const UpdateNode *un=updates.head; un; un=un->next) {
+    llvm::cerr << "\t\t[" << un->index << "] = " << un->value << "\n";
+  }
+}

Added: klee/trunk/lib/Core/Memory.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Memory.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Memory.h (added)
+++ klee/trunk/lib/Core/Memory.h Wed May 20 23:36:41 2009
@@ -0,0 +1,239 @@
+//===-- Memory.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_MEMORY_H
+#define KLEE_MEMORY_H
+
+#include "klee/Expr.h"
+
+#include <vector>
+#include <string>
+
+namespace llvm {
+  class Value;
+}
+
+namespace klee {
+
+class BitArray;
+class MemoryManager;
+class Solver;
+
+class MemoryObject {
+  friend class STPBuilder;
+
+private:
+  static int counter;
+
+public:
+  unsigned id;
+  uint64_t address;
+  Array *array;
+
+  /// size in bytes
+  unsigned size;
+  std::string name;
+
+  bool isLocal;
+  bool isGlobal;
+  bool isFixed;
+
+  /// true if created by us.
+  bool fake_object;
+  bool isUserSpecified;
+
+  /// "Location" for which this memory object was allocated. This
+  /// should be either the allocating instruction or the global object
+  /// it was allocated for (or whatever else makes sense).
+  const llvm::Value *allocSite;
+  
+  /// A list of boolean expressions the user has requested be true of
+  /// a counterexample. Mutable since we play a little fast and loose
+  /// with allowing it to be added to during execution (although
+  /// should sensibly be only at creation time).
+  mutable std::vector< ref<Expr> > cexPreferences;
+
+  // DO NOT IMPLEMENT
+  MemoryObject(const MemoryObject &b);
+  MemoryObject &operator=(const MemoryObject &b);
+
+public:
+  // XXX this is just a temp hack, should be removed
+  explicit
+  MemoryObject(uint64_t _address) 
+    : id(counter++),
+      address(_address),
+      array(new Array(this, 0, id)),
+      size(0),
+      isFixed(true),
+      allocSite(0) {
+  }
+
+  MemoryObject(uint64_t _address, unsigned _size, 
+               bool _isLocal, bool _isGlobal, bool _isFixed,
+               const llvm::Value *_allocSite) 
+    : id(counter++),
+      address(_address),
+      array(new Array(this, id, _size)),
+      size(_size),
+      name("unnamed"),
+      isLocal(_isLocal),
+      isGlobal(_isGlobal),
+      isFixed(_isFixed),
+      fake_object(false),
+      isUserSpecified(false),
+      allocSite(_allocSite) {
+  }
+
+  ~MemoryObject();
+
+  /// Get an identifying string for this allocation.
+  void getAllocInfo(std::string &result) const;
+
+  void setName(std::string name) {
+    this->name = name;
+  }
+
+  ref<Expr> getBaseExpr() const { 
+    return ConstantExpr::create(address, kMachinePointerType);
+  }
+  ref<Expr> getSizeExpr() const { 
+    return ConstantExpr::create(size, kMachinePointerType);
+  }
+  ref<Expr> getOffsetExpr(ref<Expr> pointer) const {
+    return SubExpr::create(pointer, getBaseExpr());
+  }
+  ref<Expr> getBoundsCheckPointer(ref<Expr> pointer) const {
+    return getBoundsCheckOffset(getOffsetExpr(pointer));
+  }
+  ref<Expr> getBoundsCheckPointer(ref<Expr> pointer, unsigned bytes) const {
+    return getBoundsCheckOffset(getOffsetExpr(pointer), bytes);
+  }
+
+  ref<Expr> getBoundsCheckOffset(ref<Expr> offset) const {
+    if (size==0) {
+      return EqExpr::create(offset, ref<Expr>(0, kMachinePointerType));
+    } else {
+      return UltExpr::create(offset, getSizeExpr());
+    }
+  }
+  ref<Expr> getBoundsCheckOffset(ref<Expr> offset, unsigned bytes) const {
+    if (bytes<=size) {
+      return UltExpr::create(offset, 
+                             ref<Expr>(size - bytes + 1, kMachinePointerType));
+    } else {
+      return ref<Expr>(0, Expr::Bool);
+    }
+  }
+};
+
+class ObjectState {
+private:
+  friend class AddressSpace;
+  unsigned copyOnWriteOwner; // exclusively for AddressSpace
+
+  friend class ObjectHolder;
+  unsigned refCount;
+
+  const MemoryObject *object;
+
+  uint8_t *concreteStore;
+  // XXX cleanup name of flushMask (its backwards or something)
+  BitArray *concreteMask;
+
+  // mutable because may need flushed during read of const
+  mutable BitArray *flushMask;
+
+  ref<Expr> *knownSymbolics;
+
+public:
+  unsigned size;
+
+  // mutable because we may need flush during read of const
+  mutable UpdateList updates;
+
+  bool readOnly;
+
+public:
+  // initial contents are undefined but concrete, it is the creators
+  // responsibility to initialize the object contents appropriate
+  ObjectState(const MemoryObject *mo, unsigned size);
+  ObjectState(const ObjectState &os);
+  ~ObjectState();
+
+  const MemoryObject *getObject() const { return object; }
+
+  void setReadOnly(bool ro) { readOnly = ro; }
+
+  // make all bytes are concrete with undefined values
+  void makeConcrete();
+
+  void makeSymbolic();
+
+  // make contents all concrete and zero
+  void initializeToZero();
+  // make contents all concrete and random
+  void initializeToRandom();
+
+  ref<Expr> read(ref<Expr> offset, Expr::Width width) const;
+  ref<Expr> read(unsigned offset, Expr::Width width) const;
+  ref<Expr> read1(unsigned offset) const;
+  ref<Expr> read8(unsigned offset) const;
+  ref<Expr> read16(unsigned offset) const;
+  ref<Expr> read32(unsigned offset) const;
+  ref<Expr> read64(unsigned offset) const;
+
+  // return bytes written.
+  void write(unsigned offset, ref<Expr> value);
+  void write(ref<Expr> offset, ref<Expr> value);
+
+  void write8(unsigned offset, uint8_t value);
+  void write16(unsigned offset, uint16_t value);
+  void write32(unsigned offset, uint32_t value);
+  void write64(unsigned offset, uint64_t value);
+
+private:
+  ref<Expr> read1(ref<Expr> offset) const;
+  ref<Expr> read8(ref<Expr> offset) const;
+  ref<Expr> read16(ref<Expr> offset) const;
+  ref<Expr> read32(ref<Expr> offset) const;
+  ref<Expr> read64(ref<Expr> offset) const;
+
+  void write1(unsigned offset, ref<Expr> value);
+  void write1(ref<Expr> offset, ref<Expr> value);
+  void write8(unsigned offset, ref<Expr> value);
+  void write8(ref<Expr> offset, ref<Expr> value);
+  void write16(unsigned offset, ref<Expr> value);
+  void write16(ref<Expr> offset, ref<Expr> value);
+  void write32(unsigned offset, ref<Expr> value);
+  void write32(ref<Expr> offset, ref<Expr> value);
+  void write64(unsigned offset, ref<Expr> value);
+  void write64(ref<Expr> offset, ref<Expr> value);
+
+  
+  void fastRangeCheckOffset(ref<Expr> offset, unsigned *base_r, unsigned *size_r) const;
+  void flushRangeForRead(unsigned rangeBase, unsigned rangeSize) const;
+  void flushRangeForWrite(unsigned rangeBase, unsigned rangeSize);
+
+  bool isByteConcrete(unsigned offset) const;
+  bool isByteFlushed(unsigned offset) const;
+  bool isByteKnownSymbolic(unsigned offset) const;
+
+  void markByteConcrete(unsigned offset);
+  void markByteSymbolic(unsigned offset);
+  void markByteFlushed(unsigned offset);
+  void markByteUnflushed(unsigned offset);
+  void setKnownSymbolic(unsigned offset, Expr *value);
+
+  void print();
+};
+  
+} // End klee namespace
+
+#endif

Added: klee/trunk/lib/Core/MemoryManager.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/MemoryManager.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/MemoryManager.cpp (added)
+++ klee/trunk/lib/Core/MemoryManager.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,69 @@
+//===-- MemoryManager.cpp -------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "CoreStats.h"
+#include "Memory.h"
+#include "MemoryManager.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Expr.h"
+#include "klee/Solver.h"
+
+#include "llvm/Support/CommandLine.h"
+
+using namespace klee;
+
+/***/
+
+MemoryManager::~MemoryManager() { 
+  while (!objects.empty()) {
+    MemoryObject *mo = objects.back();
+    objects.pop_back();
+    delete mo;
+  }
+}
+
+MemoryObject *MemoryManager::allocate(uint64_t size, bool isLocal, bool isGlobal,
+                                      const llvm::Value *allocSite) {
+  if (size>10*1024*1024) {
+    klee_warning_once(0, "failing large alloc: %u bytes", (unsigned) size);
+    return 0;
+  }
+  uint64_t address = (uint64_t) (unsigned long) malloc((unsigned) size);
+  if (!address)
+    return 0;
+  
+  ++stats::allocations;
+  MemoryObject *res = new MemoryObject(address, size, isLocal, isGlobal, false,
+                                       allocSite);
+  objects.push_back(res);
+  return res;
+}
+
+MemoryObject *MemoryManager::allocateFixed(uint64_t address, uint64_t size,
+                                           const llvm::Value *allocSite) {
+  for (objects_ty::iterator it = objects.begin(), ie = objects.end();
+       it != ie; ++it) {
+    MemoryObject *mo = *it;
+    assert(!(address+size > mo->address && address < mo->address+mo->size) &&
+           "allocated an overlapping object");
+  }
+
+  ++stats::allocations;
+  MemoryObject *res = new MemoryObject(address, size, false, true, true,
+                                       allocSite);
+  objects.push_back(res);
+  return res;
+}
+
+void MemoryManager::deallocate(const MemoryObject *mo) {
+  assert(0);
+}

Added: klee/trunk/lib/Core/MemoryManager.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/MemoryManager.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/MemoryManager.h (added)
+++ klee/trunk/lib/Core/MemoryManager.h Wed May 20 23:36:41 2009
@@ -0,0 +1,41 @@
+//===-- MemoryManager.h -----------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_MEMORYMANAGER_H
+#define KLEE_MEMORYMANAGER_H
+
+#include <vector>
+#include <stdint.h>
+
+namespace llvm {
+  class Value;
+}
+
+namespace klee {
+  class MemoryObject;
+
+  class MemoryManager {
+  private:
+    typedef std::vector<MemoryObject*> objects_ty;
+    objects_ty objects;
+
+  public:
+    MemoryManager() {}
+    ~MemoryManager();
+
+    MemoryObject *allocate(uint64_t size, bool isLocal, bool isGlobal,
+                           const llvm::Value *allocSite);
+    MemoryObject *allocateFixed(uint64_t address, uint64_t size,
+                                const llvm::Value *allocSite);
+    void deallocate(const MemoryObject *mo);
+  };
+
+} // End klee namespace
+
+#endif

Added: klee/trunk/lib/Core/ObjectHolder.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/ObjectHolder.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/ObjectHolder.h (added)
+++ klee/trunk/lib/Core/ObjectHolder.h Wed May 20 23:36:41 2009
@@ -0,0 +1,33 @@
+//===-- ObjectHolder.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_OBJECTHOLDER_H
+#define KLEE_OBJECTHOLDER_H
+
+namespace klee {
+  class ObjectState;
+
+  class ObjectHolder {
+    ObjectState *os;
+    
+  public:
+    ObjectHolder() : os(0) {}
+    ObjectHolder(ObjectState *_os);
+    ObjectHolder(const ObjectHolder &b);
+    ~ObjectHolder(); 
+    
+    ObjectHolder &operator=(const ObjectHolder &b);
+
+    operator class ObjectState *() { return os; }
+    operator class ObjectState *() const { return (ObjectState*) os; }
+  };
+}
+
+#endif
+

Added: klee/trunk/lib/Core/PTree.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/PTree.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/PTree.cpp (added)
+++ klee/trunk/lib/Core/PTree.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,103 @@
+//===-- PTree.cpp ---------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "PTree.h"
+
+#include <klee/Expr.h>
+#include <klee/util/ExprPPrinter.h>
+
+#include <vector>
+#include <iostream>
+
+using namespace klee;
+
+  /* *** */
+
+PTree::PTree(const data_type &_root) : root(new Node(0,_root)) {
+}
+
+PTree::~PTree() {}
+
+std::pair<PTreeNode*, PTreeNode*>
+PTree::split(Node *n, 
+             const data_type &leftData, 
+             const data_type &rightData) {
+  assert(n && !n->left && !n->right);
+  n->left = new Node(n, leftData);
+  n->right = new Node(n, rightData);
+  return std::make_pair(n->left, n->right);
+}
+
+void PTree::remove(Node *n) {
+  assert(!n->left && !n->right);
+  do {
+    Node *p = n->parent;
+    delete n;
+    if (p) {
+      if (n == p->left) {
+        p->left = 0;
+      } else {
+        assert(n == p->right);
+        p->right = 0;
+      }
+    }
+    n = p;
+  } while (n && !n->left && !n->right);
+}
+
+void PTree::dump(std::ostream &os) {
+  ExprPPrinter *pp = ExprPPrinter::create(os);
+  pp->setNewline("\\l");
+  os << "digraph G {\n";
+  os << "\tsize=\"10,7.5\";\n";
+  os << "\tratio=fill;\n";
+  os << "\trotate=90;\n";
+  os << "\tcenter = \"true\";\n";
+  os << "\tnode [style=\"filled\",width=.1,height=.1,fontname=\"Terminus\"]\n";
+  os << "\tedge [arrowsize=.3]\n";
+  std::vector<PTree::Node*> stack;
+  stack.push_back(root);
+  while (!stack.empty()) {
+    PTree::Node *n = stack.back();
+    stack.pop_back();
+    if (n->condition.isNull()) {
+      os << "\tn" << n << " [label=\"\"";
+    } else {
+      os << "\tn" << n << " [label=\"";
+      pp->print(n->condition);
+      os << "\",shape=diamond";
+    }
+    if (n->data)
+      os << ",fillcolor=green";
+    os << "];\n";
+    if (n->left) {
+      os << "\tn" << n << " -> n" << n->left << ";\n";
+      stack.push_back(n->left);
+    }
+    if (n->right) {
+      os << "\tn" << n << " -> n" << n->right << ";\n";
+      stack.push_back(n->right);
+    }
+  }
+  os << "}\n";
+  delete pp;
+}
+
+PTreeNode::PTreeNode(PTreeNode *_parent, 
+                     ExecutionState *_data) 
+  : parent(_parent),
+    left(0),
+    right(0),
+    data(_data),
+    condition(0) {
+}
+
+PTreeNode::~PTreeNode() {
+}
+

Added: klee/trunk/lib/Core/PTree.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/PTree.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/PTree.h (added)
+++ klee/trunk/lib/Core/PTree.h Wed May 20 23:36:41 2009
@@ -0,0 +1,53 @@
+//===-- PTree.h -------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_PTREE_H__
+#define __UTIL_PTREE_H__
+
+#include <klee/Expr.h>
+
+#include <utility>
+#include <cassert>
+#include <iostream>
+
+namespace klee {
+  class ExecutionState;
+
+  class PTree { 
+    typedef ExecutionState* data_type;
+
+  public:
+    typedef class PTreeNode Node;
+    Node *root;
+
+    PTree(const data_type &_root);
+    ~PTree();
+    
+    std::pair<Node*,Node*> split(Node *n,
+                                 const data_type &leftData,
+                                 const data_type &rightData);
+    void remove(Node *n);
+
+    void dump(std::ostream &os);
+  };
+
+  class PTreeNode {
+    friend class PTree;
+  public:
+    PTreeNode *parent, *left, *right;
+    ExecutionState *data;
+    ref<Expr> condition;
+
+  private:
+    PTreeNode(PTreeNode *_parent, ExecutionState *_data);
+    ~PTreeNode();
+  };
+}
+
+#endif

Added: klee/trunk/lib/Core/Searcher.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Searcher.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Searcher.cpp (added)
+++ klee/trunk/lib/Core/Searcher.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,575 @@
+//===-- Searcher.cpp ------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "Searcher.h"
+
+#include "CoreStats.h"
+#include "Executor.h"
+#include "PTree.h"
+#include "StatsTracker.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Statistics.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/KModule.h"
+#include "klee/Internal/ADT/DiscretePDF.h"
+#include "klee/Internal/ADT/RNG.h"
+#include "klee/Internal/Support/ModuleUtil.h"
+#include "klee/Internal/System/Time.h"
+
+#include "llvm/Constants.h"
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+#include "llvm/Support/CallSite.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/Support/CommandLine.h"
+
+#include <cassert>
+#include <fstream>
+#include <climits>
+
+using namespace klee;
+using namespace llvm;
+
+namespace {
+  cl::opt<bool>
+  DebugLogMerge("debug-log-merge");
+}
+
+namespace klee {
+  extern RNG theRNG;
+}
+
+Searcher::~Searcher() {
+}
+
+///
+
+ExecutionState &DFSSearcher::selectState() {
+  return *states.back();
+}
+
+void DFSSearcher::update(ExecutionState *current,
+                         const std::set<ExecutionState*> &addedStates,
+                         const std::set<ExecutionState*> &removedStates) {
+  states.insert(states.end(),
+                addedStates.begin(),
+                addedStates.end());
+  for (std::set<ExecutionState*>::const_iterator it = removedStates.begin(),
+         ie = removedStates.end(); it != ie; ++it) {
+    ExecutionState *es = *it;
+    if (es == states.back()) {
+      states.pop_back();
+    } else {
+      bool ok = false;
+
+      for (std::vector<ExecutionState*>::iterator it = states.begin(),
+             ie = states.end(); it != ie; ++it) {
+        if (es==*it) {
+          states.erase(it);
+          ok = true;
+          break;
+        }
+      }
+
+      assert(ok && "invalid state removed");
+    }
+  }
+}
+
+///
+
+ExecutionState &RandomSearcher::selectState() {
+  return *states[theRNG.getInt32()%states.size()];
+}
+
+void RandomSearcher::update(ExecutionState *current,
+                            const std::set<ExecutionState*> &addedStates,
+                            const std::set<ExecutionState*> &removedStates) {
+  states.insert(states.end(),
+                addedStates.begin(),
+                addedStates.end());
+  for (std::set<ExecutionState*>::const_iterator it = removedStates.begin(),
+         ie = removedStates.end(); it != ie; ++it) {
+    ExecutionState *es = *it;
+    bool ok = false;
+
+    for (std::vector<ExecutionState*>::iterator it = states.begin(),
+           ie = states.end(); it != ie; ++it) {
+      if (es==*it) {
+        states.erase(it);
+        ok = true;
+        break;
+      }
+    }
+    
+    assert(ok && "invalid state removed");
+  }
+}
+
+///
+
+WeightedRandomSearcher::WeightedRandomSearcher(Executor &_executor,
+                                               WeightType _type) 
+  : executor(_executor),
+    states(new DiscretePDF<ExecutionState*>()),
+    type(_type) {
+  switch(type) {
+  case Depth: 
+    updateWeights = false;
+    break;
+  case InstCount:
+  case CPInstCount:
+  case QueryCost:
+  case MinDistToUncovered:
+  case CoveringNew:
+    updateWeights = true;
+    break;
+  default:
+    assert(0 && "invalid weight type");
+  }
+}
+
+WeightedRandomSearcher::~WeightedRandomSearcher() {
+  delete states;
+}
+
+ExecutionState &WeightedRandomSearcher::selectState() {
+  return *states->choose(theRNG.getDoubleL());
+}
+
+double WeightedRandomSearcher::getWeight(ExecutionState *es) {
+  switch(type) {
+  default:
+  case Depth: 
+    return es->weight;
+  case InstCount: {
+    uint64_t count = theStatisticManager->getIndexedValue(stats::instructions,
+                                                          es->pc->info->id);
+    double inv = 1. / std::max((uint64_t) 1, count);
+    return inv * inv;
+  }
+  case CPInstCount: {
+    StackFrame &sf = es->stack.back();
+    uint64_t count = sf.callPathNode->statistics.getValue(stats::instructions);
+    double inv = 1. / std::max((uint64_t) 1, count);
+    return inv;
+  }
+  case QueryCost:
+    return (es->queryCost < .1) ? 1. : 1./es->queryCost;
+  case CoveringNew:
+  case MinDistToUncovered: {
+    uint64_t md2u = computeMinDistToUncovered(es->pc,
+                                              es->stack.back().minDistToUncoveredOnReturn);
+
+    double invMD2U = 1. / (md2u ? md2u : 10000);
+    if (type==CoveringNew) {
+      double invCovNew = 0.;
+      if (es->instsSinceCovNew)
+        invCovNew = 1. / std::max(1, (int) es->instsSinceCovNew - 1000);
+      return (invCovNew * invCovNew + invMD2U * invMD2U);
+    } else {
+      return invMD2U * invMD2U;
+    }
+  }
+  }
+}
+
+void WeightedRandomSearcher::update(ExecutionState *current,
+                                    const std::set<ExecutionState*> &addedStates,
+                                    const std::set<ExecutionState*> &removedStates) {
+  if (current && updateWeights && !removedStates.count(current))
+    states->update(current, getWeight(current));
+  
+  for (std::set<ExecutionState*>::const_iterator it = addedStates.begin(),
+         ie = addedStates.end(); it != ie; ++it) {
+    ExecutionState *es = *it;
+    states->insert(es, getWeight(es));
+  }
+
+  for (std::set<ExecutionState*>::const_iterator it = removedStates.begin(),
+         ie = removedStates.end(); it != ie; ++it) {
+    states->remove(*it);
+  }
+}
+
+bool WeightedRandomSearcher::empty() { 
+  return states->empty(); 
+}
+
+///
+
+RandomPathSearcher::RandomPathSearcher(Executor &_executor)
+  : executor(_executor) {
+}
+
+RandomPathSearcher::~RandomPathSearcher() {
+}
+
+ExecutionState &RandomPathSearcher::selectState() {
+  unsigned flips=0, bits=0;
+  PTree::Node *n = executor.processTree->root;
+  
+  while (!n->data) {
+    if (!n->left) {
+      n = n->right;
+    } else if (!n->right) {
+      n = n->left;
+    } else {
+      if (bits==0) {
+        flips = theRNG.getInt32();
+        bits = 32;
+      }
+      --bits;
+      n = (flips&(1<<bits)) ? n->left : n->right;
+    }
+  }
+
+  return *n->data;
+}
+
+void RandomPathSearcher::update(ExecutionState *current,
+                                const std::set<ExecutionState*> &addedStates,
+                                const std::set<ExecutionState*> &removedStates) {
+}
+
+bool RandomPathSearcher::empty() { 
+  return executor.states.empty(); 
+}
+
+///
+
+BumpMergingSearcher::BumpMergingSearcher(Executor &_executor, Searcher *_baseSearcher) 
+  : executor(_executor),
+    baseSearcher(_baseSearcher),
+    mergeFunction(executor.kmodule->kleeMergeFn) {
+}
+
+BumpMergingSearcher::~BumpMergingSearcher() {
+  delete baseSearcher;
+}
+
+///
+
+Instruction *BumpMergingSearcher::getMergePoint(ExecutionState &es) {  
+  if (mergeFunction) {
+    Instruction *i = es.pc->inst;
+
+    if (i->getOpcode()==Instruction::Call) {
+      CallSite cs(cast<CallInst>(i));
+      if (mergeFunction==cs.getCalledFunction())
+        return i;
+    }
+  }
+
+  return 0;
+}
+
+ExecutionState &BumpMergingSearcher::selectState() {
+entry:
+  // out of base states, pick one to pop
+  if (baseSearcher->empty()) {
+    std::map<llvm::Instruction*, ExecutionState*>::iterator it = 
+      statesAtMerge.begin();
+    ExecutionState *es = it->second;
+    statesAtMerge.erase(it);
+    ++es->pc;
+
+    baseSearcher->addState(es);
+  }
+
+  ExecutionState &es = baseSearcher->selectState();
+
+  if (Instruction *mp = getMergePoint(es)) {
+    std::map<llvm::Instruction*, ExecutionState*>::iterator it = 
+      statesAtMerge.find(mp);
+
+    baseSearcher->removeState(&es);
+
+    if (it==statesAtMerge.end()) {
+      statesAtMerge.insert(std::make_pair(mp, &es));
+    } else {
+      ExecutionState *mergeWith = it->second;
+      if (mergeWith->merge(es)) {
+        // hack, because we are terminating the state we need to let
+        // the baseSearcher know about it again
+        baseSearcher->addState(&es);
+        executor.terminateState(es);
+      } else {
+        it->second = &es; // the bump
+        ++mergeWith->pc;
+
+        baseSearcher->addState(mergeWith);
+      }
+    }
+
+    goto entry;
+  } else {
+    return es;
+  }
+}
+
+void BumpMergingSearcher::update(ExecutionState *current,
+                                 const std::set<ExecutionState*> &addedStates,
+                                 const std::set<ExecutionState*> &removedStates) {
+  baseSearcher->update(current, addedStates, removedStates);
+}
+
+///
+
+MergingSearcher::MergingSearcher(Executor &_executor, Searcher *_baseSearcher) 
+  : executor(_executor),
+    baseSearcher(_baseSearcher),
+    mergeFunction(executor.kmodule->kleeMergeFn) {
+}
+
+MergingSearcher::~MergingSearcher() {
+  delete baseSearcher;
+}
+
+///
+
+Instruction *MergingSearcher::getMergePoint(ExecutionState &es) {
+  if (mergeFunction) {
+    Instruction *i = es.pc->inst;
+
+    if (i->getOpcode()==Instruction::Call) {
+      CallSite cs(cast<CallInst>(i));
+      if (mergeFunction==cs.getCalledFunction())
+        return i;
+    }
+  }
+
+  return 0;
+}
+
+ExecutionState &MergingSearcher::selectState() {
+  while (!baseSearcher->empty()) {
+    ExecutionState &es = baseSearcher->selectState();
+    if (getMergePoint(es)) {
+      baseSearcher->removeState(&es, &es);
+      statesAtMerge.insert(&es);
+    } else {
+      return es;
+    }
+  }
+  
+  // build map of merge point -> state list
+  std::map<Instruction*, std::vector<ExecutionState*> > merges;
+  for (std::set<ExecutionState*>::const_iterator it = statesAtMerge.begin(),
+         ie = statesAtMerge.end(); it != ie; ++it) {
+    ExecutionState &state = **it;
+    Instruction *mp = getMergePoint(state);
+    
+    merges[mp].push_back(&state);
+  }
+  
+  if (DebugLogMerge)
+    llvm::cerr << "-- all at merge --\n";
+  for (std::map<Instruction*, std::vector<ExecutionState*> >::iterator
+         it = merges.begin(), ie = merges.end(); it != ie; ++it) {
+    if (DebugLogMerge) {
+      llvm::cerr << "\tmerge: " << it->first << " [";
+      for (std::vector<ExecutionState*>::iterator it2 = it->second.begin(),
+             ie2 = it->second.end(); it2 != ie2; ++it2) {
+        ExecutionState *state = *it2;
+        llvm::cerr << state << ", ";
+      }
+      llvm::cerr << "]\n";
+    }
+
+    // merge states
+    std::set<ExecutionState*> toMerge(it->second.begin(), it->second.end());
+    while (!toMerge.empty()) {
+      ExecutionState *base = *toMerge.begin();
+      toMerge.erase(toMerge.begin());
+      
+      std::set<ExecutionState*> toErase;
+      for (std::set<ExecutionState*>::iterator it = toMerge.begin(),
+             ie = toMerge.end(); it != ie; ++it) {
+        ExecutionState *mergeWith = *it;
+        
+        if (base->merge(*mergeWith)) {
+          toErase.insert(mergeWith);
+        }
+      }
+      if (DebugLogMerge && !toErase.empty()) {
+        llvm::cerr << "\t\tmerged: " << base << " with [";
+        for (std::set<ExecutionState*>::iterator it = toErase.begin(),
+               ie = toErase.end(); it != ie; ++it) {
+          if (it!=toErase.begin()) llvm::cerr << ", ";
+          llvm::cerr << *it;
+        }
+        llvm::cerr << "]\n";
+      }
+      for (std::set<ExecutionState*>::iterator it = toErase.begin(),
+             ie = toErase.end(); it != ie; ++it) {
+        std::set<ExecutionState*>::iterator it2 = toMerge.find(*it);
+        assert(it2!=toMerge.end());
+        executor.terminateState(**it);
+        toMerge.erase(it2);
+      }
+
+      // step past merge and toss base back in pool
+      statesAtMerge.erase(statesAtMerge.find(base));
+      ++base->pc;
+      baseSearcher->addState(base);
+    }  
+  }
+  
+  if (DebugLogMerge)
+    llvm::cerr << "-- merge complete, continuing --\n";
+  
+  return selectState();
+}
+
+void MergingSearcher::update(ExecutionState *current,
+                             const std::set<ExecutionState*> &addedStates,
+                             const std::set<ExecutionState*> &removedStates) {
+  if (!removedStates.empty()) {
+    std::set<ExecutionState *> alt = removedStates;
+    for (std::set<ExecutionState*>::const_iterator it = removedStates.begin(),
+           ie = removedStates.end(); it != ie; ++it) {
+      ExecutionState *es = *it;
+      std::set<ExecutionState*>::const_iterator it = statesAtMerge.find(es);
+      if (it!=statesAtMerge.end()) {
+        statesAtMerge.erase(it);
+        alt.erase(alt.find(es));
+      }
+    }    
+    baseSearcher->update(current, addedStates, alt);
+  } else {
+    baseSearcher->update(current, addedStates, removedStates);
+  }
+}
+
+///
+
+BatchingSearcher::BatchingSearcher(Searcher *_baseSearcher,
+                                   double _timeBudget,
+                                   unsigned _instructionBudget) 
+  : baseSearcher(_baseSearcher),
+    timeBudget(_timeBudget),
+    instructionBudget(_instructionBudget),
+    lastState(0) {
+  
+}
+
+BatchingSearcher::~BatchingSearcher() {
+  delete baseSearcher;
+}
+
+ExecutionState &BatchingSearcher::selectState() {
+  if (!lastState || 
+      (util::getWallTime()-lastStartTime)>timeBudget ||
+      (stats::instructions-lastStartInstructions)>instructionBudget) {
+    if (lastState) {
+      double delta = util::getWallTime()-lastStartTime;
+      if (delta>timeBudget*1.1) {
+        llvm::cerr << "KLEE: increased time budget from " << timeBudget << " to " << delta << "\n";
+        timeBudget = delta;
+      }
+    }
+    lastState = &baseSearcher->selectState();
+    lastStartTime = util::getWallTime();
+    lastStartInstructions = stats::instructions;
+    return *lastState;
+  } else {
+    return *lastState;
+  }
+}
+
+void BatchingSearcher::update(ExecutionState *current,
+                              const std::set<ExecutionState*> &addedStates,
+                              const std::set<ExecutionState*> &removedStates) {
+  if (removedStates.count(lastState))
+    lastState = 0;
+  baseSearcher->update(current, addedStates, removedStates);
+}
+
+/***/
+
+IterativeDeepeningTimeSearcher::IterativeDeepeningTimeSearcher(Searcher *_baseSearcher)
+  : baseSearcher(_baseSearcher),
+    time(1.) {
+}
+
+IterativeDeepeningTimeSearcher::~IterativeDeepeningTimeSearcher() {
+  delete baseSearcher;
+}
+
+ExecutionState &IterativeDeepeningTimeSearcher::selectState() {
+  ExecutionState &res = baseSearcher->selectState();
+  startTime = util::getWallTime();
+  return res;
+}
+
+void IterativeDeepeningTimeSearcher::update(ExecutionState *current,
+                                            const std::set<ExecutionState*> &addedStates,
+                                            const std::set<ExecutionState*> &removedStates) {
+  double elapsed = util::getWallTime() - startTime;
+
+  if (!removedStates.empty()) {
+    std::set<ExecutionState *> alt = removedStates;
+    for (std::set<ExecutionState*>::const_iterator it = removedStates.begin(),
+           ie = removedStates.end(); it != ie; ++it) {
+      ExecutionState *es = *it;
+      std::set<ExecutionState*>::const_iterator it = pausedStates.find(es);
+      if (it!=pausedStates.end()) {
+        pausedStates.erase(it);
+        alt.erase(alt.find(es));
+      }
+    }    
+    baseSearcher->update(current, addedStates, alt);
+  } else {
+    baseSearcher->update(current, addedStates, removedStates);
+  }
+
+  if (current && !removedStates.count(current) && elapsed>time) {
+    pausedStates.insert(current);
+    baseSearcher->removeState(current);
+  }
+
+  if (baseSearcher->empty()) {
+    time *= 2;
+    llvm::cerr << "KLEE: increasing time budget to: " << time << "\n";
+    baseSearcher->update(0, pausedStates, std::set<ExecutionState*>());
+    pausedStates.clear();
+  }
+}
+
+/***/
+
+InterleavedSearcher::InterleavedSearcher(const std::vector<Searcher*> &_searchers)
+  : searchers(_searchers),
+    index(1) {
+}
+
+InterleavedSearcher::~InterleavedSearcher() {
+  for (std::vector<Searcher*>::const_iterator it = searchers.begin(),
+         ie = searchers.end(); it != ie; ++it)
+    delete *it;
+}
+
+ExecutionState &InterleavedSearcher::selectState() {
+  Searcher *s = searchers[--index];
+  if (index==0) index = searchers.size();
+  return s->selectState();
+}
+
+void InterleavedSearcher::update(ExecutionState *current,
+                                 const std::set<ExecutionState*> &addedStates,
+                                 const std::set<ExecutionState*> &removedStates) {
+  for (std::vector<Searcher*>::const_iterator it = searchers.begin(),
+         ie = searchers.end(); it != ie; ++it)
+    (*it)->update(current, addedStates, removedStates);
+}

Added: klee/trunk/lib/Core/Searcher.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/Searcher.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/Searcher.h (added)
+++ klee/trunk/lib/Core/Searcher.h Wed May 20 23:36:41 2009
@@ -0,0 +1,279 @@
+//===-- Searcher.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_SEARCHER_H
+#define KLEE_SEARCHER_H
+
+#include <vector>
+#include <set>
+#include <map>
+#include <queue>
+
+// FIXME: Move out of header, use llvm streams.
+#include <ostream>
+
+namespace llvm {
+  class BasicBlock;
+  class Function;
+  class Instruction;
+}
+
+namespace klee {
+  template<class T> class DiscretePDF;
+  class ExecutionState;
+  class Executor;
+
+  class Searcher {
+  public:
+    virtual ~Searcher();
+
+    virtual ExecutionState &selectState() = 0;
+
+    virtual void update(ExecutionState *current,
+                        const std::set<ExecutionState*> &addedStates,
+                        const std::set<ExecutionState*> &removedStates) = 0;
+
+    virtual bool empty() = 0;
+
+    // prints name of searcher as a klee_message()
+    // TODO: could probably make prettier or more flexible
+    virtual void printName(std::ostream &os) { 
+      os << "<unnamed searcher>\n";
+    }
+
+    // pgbovine - to be called when a searcher gets activated and
+    // deactivated, say, by a higher-level searcher; most searchers
+    // don't need this functionality, so don't have to override.
+    virtual void activate() {};
+    virtual void deactivate() {};
+
+    // utility functions
+
+    void addState(ExecutionState *es, ExecutionState *current = 0) {
+      std::set<ExecutionState*> tmp;
+      tmp.insert(es);
+      update(current, tmp, std::set<ExecutionState*>());
+    }
+
+    void removeState(ExecutionState *es, ExecutionState *current = 0) {
+      std::set<ExecutionState*> tmp;
+      tmp.insert(es);
+      update(current, std::set<ExecutionState*>(), tmp);
+    }
+  };
+
+  class DFSSearcher : public Searcher {
+    std::vector<ExecutionState*> states;
+
+  public:
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return states.empty(); }
+    void printName(std::ostream &os) {
+      os << "DFSSearcher\n";
+    }
+  };
+
+  class RandomSearcher : public Searcher {
+    std::vector<ExecutionState*> states;
+
+  public:
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return states.empty(); }
+    void printName(std::ostream &os) {
+      os << "RandomSearcher\n";
+    }
+  };
+
+  class WeightedRandomSearcher : public Searcher {
+  public:
+    enum WeightType {
+      Depth,
+      QueryCost,
+      InstCount,
+      CPInstCount,
+      MinDistToUncovered,
+      CoveringNew
+    };
+
+  private:
+    Executor &executor;
+    DiscretePDF<ExecutionState*> *states;
+    WeightType type;
+    bool updateWeights;
+    
+    double getWeight(ExecutionState*);
+
+  public:
+    WeightedRandomSearcher(Executor &executor, WeightType type);
+    ~WeightedRandomSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty();
+    void printName(std::ostream &os) {
+      os << "WeightedRandomSearcher::";
+      switch(type) {
+      case Depth              : os << "Depth\n"; return;
+      case QueryCost          : os << "QueryCost\n"; return;
+      case InstCount          : os << "InstCount\n"; return;
+      case CPInstCount        : os << "CPInstCount\n"; return;
+      case MinDistToUncovered : os << "MinDistToUncovered\n"; return;
+      case CoveringNew        : os << "CoveringNew\n"; return;
+      default                 : os << "<unknown type>\n"; return;
+      }
+    }
+  };
+
+  class RandomPathSearcher : public Searcher {
+    Executor &executor;
+
+  public:
+    RandomPathSearcher(Executor &_executor);
+    ~RandomPathSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty();
+    void printName(std::ostream &os) {
+      os << "RandomPathSearcher\n";
+    }
+  };
+
+  class MergingSearcher : public Searcher {
+    Executor &executor;
+    std::set<ExecutionState*> statesAtMerge;
+    Searcher *baseSearcher;
+    llvm::Function *mergeFunction;
+
+  private:
+    llvm::Instruction *getMergePoint(ExecutionState &es);
+
+  public:
+    MergingSearcher(Executor &executor, Searcher *baseSearcher);
+    ~MergingSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return baseSearcher->empty() && statesAtMerge.empty(); }
+    void printName(std::ostream &os) {
+      os << "MergingSearcher\n";
+    }
+  };
+
+  class BumpMergingSearcher : public Searcher {
+    Executor &executor;
+    std::map<llvm::Instruction*, ExecutionState*> statesAtMerge;
+    Searcher *baseSearcher;
+    llvm::Function *mergeFunction;
+
+  private:
+    llvm::Instruction *getMergePoint(ExecutionState &es);
+
+  public:
+    BumpMergingSearcher(Executor &executor, Searcher *baseSearcher);
+    ~BumpMergingSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return baseSearcher->empty() && statesAtMerge.empty(); }
+    void printName(std::ostream &os) {
+      os << "BumpMergingSearcher\n";
+    }
+  };
+
+  class BatchingSearcher : public Searcher {
+    Searcher *baseSearcher;
+    double timeBudget;
+    unsigned instructionBudget;
+
+    ExecutionState *lastState;
+    double lastStartTime;
+    unsigned lastStartInstructions;
+
+  public:
+    BatchingSearcher(Searcher *baseSearcher, 
+                     double _timeBudget,
+                     unsigned _instructionBudget);
+    ~BatchingSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return baseSearcher->empty(); }
+    void printName(std::ostream &os) {
+      os << "<BatchingSearcher> timeBudget: " << timeBudget
+         << ", instructionBudget: " << instructionBudget
+         << ", baseSearcher:\n";
+      baseSearcher->printName(os);
+      os << "</BatchingSearcher>\n";
+    }
+  };
+
+  class IterativeDeepeningTimeSearcher : public Searcher {
+    Searcher *baseSearcher;
+    double time, startTime;
+    std::set<ExecutionState*> pausedStates;
+
+  public:
+    IterativeDeepeningTimeSearcher(Searcher *baseSearcher);
+    ~IterativeDeepeningTimeSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return baseSearcher->empty() && pausedStates.empty(); }
+    void printName(std::ostream &os) {
+      os << "IterativeDeepeningTimeSearcher\n";
+    }
+  };
+
+  class InterleavedSearcher : public Searcher {
+    typedef std::vector<Searcher*> searchers_ty;
+
+    searchers_ty searchers;
+    unsigned index;
+
+  public:
+    explicit InterleavedSearcher(const searchers_ty &_searchers);
+    ~InterleavedSearcher();
+
+    ExecutionState &selectState();
+    void update(ExecutionState *current,
+                const std::set<ExecutionState*> &addedStates,
+                const std::set<ExecutionState*> &removedStates);
+    bool empty() { return searchers[0]->empty(); }
+    void printName(std::ostream &os) {
+      os << "<InterleavedSearcher> containing "
+         << searchers.size() << " searchers:\n";
+      for (searchers_ty::iterator it = searchers.begin(), ie = searchers.end();
+           it != ie; ++it)
+        (*it)->printName(os);
+      os << "</InterleavedSearcher>\n";
+    }
+  };
+
+}
+
+#endif

Added: klee/trunk/lib/Core/SeedInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/SeedInfo.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/SeedInfo.cpp (added)
+++ klee/trunk/lib/Core/SeedInfo.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,151 @@
+//===-- SeedInfo.cpp ------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "Memory.h"
+#include "SeedInfo.h"
+#include "TimingSolver.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Expr.h"
+#include "klee/util/ExprUtil.h"
+#include "klee/Internal/ADT/BOut.h"
+
+using namespace klee;
+
+BOutObject *SeedInfo::getNextInput(const MemoryObject *mo,
+                                   bool byName) {
+  if (byName) {
+    unsigned i;
+    
+    for (i=0; i<input->numObjects; ++i) {
+      BOutObject *obj = &input->objects[i];
+      if (std::string(obj->name) == mo->name)
+        if (used.insert(obj).second)
+          return obj;
+    }
+    
+    // If first unused input matches in size then accept that as
+    // well.
+    for (i=0; i<input->numObjects; ++i)
+      if (!used.count(&input->objects[i]))
+        break;
+    if (i<input->numObjects) {
+      BOutObject *obj = &input->objects[i];
+      if (obj->numBytes == mo->size) {
+        used.insert(obj);
+        klee_warning_once(mo, "using seed input %s[%d] for: %s (no name match)",
+                          obj->name, obj->numBytes, mo->name.c_str());
+        return obj;
+      }
+    }
+    
+    klee_warning_once(mo, "no seed input for: %s", mo->name.c_str());
+    return 0;
+  } else {
+    if (inputPosition >= input->numObjects) {
+      return 0; 
+    } else {
+      return &input->objects[inputPosition++];
+    }
+  }
+}
+
+void SeedInfo::patchSeed(const ExecutionState &state, 
+                         ref<Expr> condition,
+                         TimingSolver *solver) {
+  std::vector< ref<Expr> > required(state.constraints.begin(),
+                                    state.constraints.end());
+  ExecutionState tmp(required);
+  tmp.addConstraint(condition);
+
+  // Try and patch direct reads first, this is likely to resolve the
+  // problem quickly and avoids long traversal of all seed
+  // values. There are other smart ways to do this, the nicest is if
+  // we got a minimal counterexample from STP, in which case we would
+  // just inject those values back into the seed.
+  std::set< std::pair<const Array*, unsigned> > directReads;
+  std::vector< ref<ReadExpr> > reads;
+  findReads(condition, false, reads);
+  for (std::vector< ref<ReadExpr> >::iterator it = reads.begin(), 
+         ie = reads.end(); it != ie; ++it) {
+    ReadExpr *re = it->get();
+    if (re->index.isConstant()) {
+      unsigned index = (unsigned) re->index.getConstantValue();
+      directReads.insert(std::make_pair(re->updates.root, index));
+    }
+  }
+  
+  for (std::set< std::pair<const Array*, unsigned> >::iterator
+         it = directReads.begin(), ie = directReads.end(); it != ie; ++it) {
+    const Array *array = it->first;
+    unsigned i = it->second;
+    ref<Expr> read = ReadExpr::create(UpdateList(array, true, 0),
+                                      ref<Expr>(i, Expr::Int32));
+    
+    // If not in bindings then this can't be a violation?
+    Assignment::bindings_ty::iterator it2 = assignment.bindings.find(array);
+    if (it2 != assignment.bindings.end()) {
+      ref<Expr> isSeed = EqExpr::create(read, ref<Expr>(it2->second[i], Expr::Int8));
+      bool res;
+      bool success = solver->mustBeFalse(tmp, isSeed, res);
+      assert(success && "FIXME: Unhandled solver failure");
+      if (res) {
+        ref<Expr> value;
+        bool success = solver->getValue(tmp, read, value);
+        assert(success && "FIXME: Unhandled solver failure");            
+        it2->second[i] = value.getConstantValue();
+        tmp.addConstraint(EqExpr::create(read, ref<Expr>(it2->second[i], Expr::Int8)));
+      } else {
+        tmp.addConstraint(isSeed);
+      }
+    }
+  }
+
+  bool res;
+  bool success = solver->mayBeTrue(state, assignment.evaluate(condition), res);
+  assert(success && "FIXME: Unhandled solver failure");
+  if (res)
+    return;
+  
+  // We could still do a lot better than this, for example by looking at
+  // independence. But really, this shouldn't be happening often.
+  for (Assignment::bindings_ty::iterator it = assignment.bindings.begin(), 
+         ie = assignment.bindings.end(); it != ie; ++it) {
+    const Array *array = it->first;
+    for (unsigned i=0; i<array->size; ++i) {
+      ref<Expr> read = ReadExpr::create(UpdateList(array, true, 0),
+                                        ref<Expr>(i, Expr::Int32));
+      ref<Expr> isSeed = EqExpr::create(read, ref<Expr>(it->second[i], Expr::Int8));
+      bool res;
+      bool success = solver->mustBeFalse(tmp, isSeed, res);
+      assert(success && "FIXME: Unhandled solver failure");
+      if (res) {
+        ref<Expr> value;
+        bool success = solver->getValue(tmp, read, value);
+        assert(success && "FIXME: Unhandled solver failure");            
+        it->second[i] = value.getConstantValue();
+        tmp.addConstraint(EqExpr::create(read, ref<Expr>(it->second[i], Expr::Int8)));
+      } else {
+        tmp.addConstraint(isSeed);
+      }
+    }
+  }
+
+#ifndef NDEBUG
+  {
+    bool res;
+    bool success = 
+      solver->mayBeTrue(state, assignment.evaluate(condition), res);
+    assert(success && "FIXME: Unhandled solver failure");            
+    assert(res && "seed patching failed");
+  }
+#endif
+}

Added: klee/trunk/lib/Core/SeedInfo.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/SeedInfo.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/SeedInfo.h (added)
+++ klee/trunk/lib/Core/SeedInfo.h Wed May 20 23:36:41 2009
@@ -0,0 +1,48 @@
+//===-- SeedInfo.h ----------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_SEEDINFO_H
+#define KLEE_SEEDINFO_H
+
+#include "klee/util/Assignment.h"
+
+extern "C" {
+  struct BOut;
+  struct BOutObject;
+}
+
+namespace klee {
+  class ExecutionState;
+  class TimingSolver;
+
+  class SeedInfo {
+  public:
+    Assignment assignment;
+    BOut *input;
+    unsigned inputPosition;
+    std::set<struct BOutObject*> used;
+    
+  public:
+    explicit
+    SeedInfo(BOut *_input) : assignment(true),
+                             input(_input),
+                             inputPosition(0) {}
+    
+    BOutObject *getNextInput(const MemoryObject *mo,
+                             bool byName);
+    
+    /// Patch the seed so that condition is satisfied while retaining as
+    /// many of the seed values as possible.
+    void patchSeed(const ExecutionState &state, 
+                   ref<Expr> condition,
+                   TimingSolver *solver);
+  };
+}
+
+#endif

Added: klee/trunk/lib/Core/SpecialFunctionHandler.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/SpecialFunctionHandler.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/SpecialFunctionHandler.cpp (added)
+++ klee/trunk/lib/Core/SpecialFunctionHandler.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,727 @@
+//===-- SpecialFunctionHandler.cpp ----------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "Memory.h"
+#include "SpecialFunctionHandler.h"
+#include "TimingSolver.h"
+
+#include "klee/ExecutionState.h"
+
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/KModule.h"
+
+#include "Executor.h"
+#include "MemoryManager.h"
+
+#include "llvm/Module.h"
+
+#include <errno.h>
+
+using namespace llvm;
+using namespace klee;
+
+/// \todo Almost all of the demands in this file should be replaced
+/// with terminateState calls.
+
+///
+
+struct HandlerInfo {
+  const char *name;
+  SpecialFunctionHandler::Handler handler;
+  bool doesNotReturn; /// Intrinsic terminates the process
+  bool hasReturnValue; /// Intrinsic has a return value
+  bool doNotOverride; /// Intrinsic should not be used if already defined
+};
+
+// FIXME: We are more or less committed to requiring an intrinsic
+// library these days. We can move some of this stuff there,
+// especially things like realloc which have complicated semantics
+// w.r.t. forking. Among other things this makes delayed query
+// dispatch easier to implement.
+HandlerInfo handlerInfo[] = {
+#define add(name, handler, ret) { name, \
+                                  &SpecialFunctionHandler::handler, \
+                                  false, ret, false }
+#define addDNR(name, handler) { name, \
+                                &SpecialFunctionHandler::handler, \
+                                true, false, false }
+  addDNR("__assert_rtn", handleAssertFail),
+  addDNR("__assert_fail", handleAssertFail),
+  addDNR("_assert", handleAssert),
+  addDNR("abort", handleAbort),
+  addDNR("_exit", handleExit),
+  { "exit", &SpecialFunctionHandler::handleExit, true, false, true },
+  addDNR("klee_abort", handleAbort),
+  addDNR("klee_silent_exit", handleSilentExit),  
+  addDNR("klee_report_error", handleReportError),
+
+  add("calloc", handleCalloc, true),
+  add("free", handleFree, false),
+  add("klee_assume", handleAssume, false),
+  add("klee_check_memory_access", handleCheckMemoryAccess, false),
+  add("klee_get_value", handleGetValue, true),
+  add("klee_define_fixed_object", handleDefineFixedObject, false),
+  add("klee_get_obj_size", handleGetObjSize, true),
+  add("klee_get_errno", handleGetErrno, true),
+  add("klee_is_symbolic", handleIsSymbolic, true),
+  add("klee_make_symbolic_name", handleMakeSymbolic, false),
+  add("klee_mark_global", handleMarkGlobal, false),
+  add("klee_malloc_n", handleMallocN, true),
+  add("klee_merge", handleMerge, false),
+  add("klee_prefer_cex", handlePreferCex, false),
+  add("klee_print_expr", handlePrintExpr, false),
+  add("klee_print_range", handlePrintRange, false),
+  add("klee_set_forking", handleSetForking, false),
+  add("klee_warning", handleWarning, false),
+  add("klee_warning_once", handleWarningOnce, false),
+  add("klee_under_constrained", handleUnderConstrained, false),
+  add("klee_alias_function", handleAliasFunction, false),
+  add("malloc", handleMalloc, true),
+  add("realloc", handleRealloc, true),
+
+  // operator delete[](void*)
+  add("_ZdaPv", handleDeleteArray, false),
+  // operator delete(void*)
+  add("_ZdlPv", handleDelete, false),
+
+  // operator new[](unsigned int)
+  add("_Znaj", handleNewArray, true),
+  // operator new(unsigned int)
+  add("_Znwj", handleNew, true),
+
+  // FIXME-64: This is wrong for 64-bit long...
+
+  // operator new[](unsigned long)
+  add("_Znam", handleNewArray, true),
+  // operator new(unsigned long)
+  add("_Znwm", handleNew, true),
+
+#undef addDNR
+#undef add  
+};
+
+SpecialFunctionHandler::SpecialFunctionHandler(Executor &_executor) 
+  : executor(_executor) {}
+
+
+void SpecialFunctionHandler::prepare() {
+  unsigned N = sizeof(handlerInfo)/sizeof(handlerInfo[0]);
+
+  for (unsigned i=0; i<N; ++i) {
+    HandlerInfo &hi = handlerInfo[i];
+    Function *f = executor.kmodule->module->getFunction(hi.name);
+    
+    // No need to create if the function doesn't exist, since it cannot
+    // be called in that case.
+  
+    if (f && (!hi.doNotOverride || f->isDeclaration())) {
+      // Make sure NoReturn attribute is set, for optimization and
+      // coverage counting.
+      if (hi.doesNotReturn)
+        f->addFnAttr(Attribute::NoReturn);
+
+      // Change to a declaration since we handle internally (simplifies
+      // module and allows deleting dead code).
+      if (!f->isDeclaration())
+        f->deleteBody();
+    }
+  }
+}
+
+void SpecialFunctionHandler::bind() {
+  unsigned N = sizeof(handlerInfo)/sizeof(handlerInfo[0]);
+
+  for (unsigned i=0; i<N; ++i) {
+    HandlerInfo &hi = handlerInfo[i];
+    Function *f = executor.kmodule->module->getFunction(hi.name);
+    
+    if (f && (!hi.doNotOverride || f->isDeclaration()))
+      handlers[f] = std::make_pair(hi.handler, hi.hasReturnValue);
+  }
+}
+
+
+bool SpecialFunctionHandler::handle(ExecutionState &state, 
+                                    Function *f,
+                                    KInstruction *target,
+                                    std::vector< ref<Expr> > &arguments) {
+  handlers_ty::iterator it = handlers.find(f);
+  if (it != handlers.end()) {    
+    Handler h = it->second.first;
+    bool hasReturnValue = it->second.second;
+     // FIXME: Check this... add test?
+    if (!hasReturnValue && !target->inst->use_empty()) {
+      executor.terminateStateOnExecError(state, 
+                                         "expected return value from void special function");
+    } else {
+      (this->*h)(state, target, arguments);
+    }
+    return true;
+  } else {
+    return false;
+  }
+}
+
+/****/
+
+// reads a concrete string from memory
+std::string SpecialFunctionHandler::readStringAtAddress(ExecutionState &state, 
+                                                        ref<Expr> address) {
+  ObjectPair op;
+  address = executor.toUnique(state, address);
+  assert(address.isConstant() && "symbolic string arg to intrinsic");  
+  if (!state.addressSpace.resolveOne(address.getConstantValue(), op))
+    assert(0 && "XXX out of bounds / multiple resolution unhandled");
+  bool res;
+  assert(executor.solver->mustBeTrue(state, 
+                                     EqExpr::create(address, 
+                                                    op.first->getBaseExpr()),
+                                     res) &&
+         res &&
+         "XXX interior pointer unhandled");
+  const MemoryObject *mo = op.first;
+  const ObjectState *os = op.second;
+
+  char *buf = new char[mo->size];
+
+  unsigned i;
+  for (i = 0; i < mo->size - 1; i++) {
+    ref<Expr> cur = os->read8(i);
+    cur = executor.toUnique(state, cur);
+    assert(cur.isConstant() && 
+           "hit symbolic char while reading concrete string");
+    buf[i] = cur.getConstantValue();
+  }
+  buf[i] = 0;
+  
+  std::string result(buf);
+  delete[] buf;
+  return result;
+}
+
+/****/
+
+void SpecialFunctionHandler::handleAbort(ExecutionState &state,
+                           KInstruction *target,
+                           std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==0 && "invalid number of arguments to abort");
+
+  //XXX:DRE:TAINT
+  if(state.underConstrained) {
+    llvm::cerr << "TAINT: skipping abort fail\n";
+    executor.terminateState(state);
+  } else {
+    executor.terminateStateOnError(state, "abort failure", "abort.err");
+  }
+}
+
+void SpecialFunctionHandler::handleExit(ExecutionState &state,
+                           KInstruction *target,
+                           std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 && "invalid number of arguments to exit");
+  executor.terminateStateOnExit(state);
+}
+
+void SpecialFunctionHandler::handleSilentExit(ExecutionState &state,
+                                              KInstruction *target,
+                                              std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 && "invalid number of arguments to exit");
+  executor.terminateState(state);
+}
+
+void SpecialFunctionHandler::handleAliasFunction(ExecutionState &state,
+						 KInstruction *target,
+						 std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 && 
+         "invalid number of arguments to klee_alias_function");
+  std::string old_fn = readStringAtAddress(state, arguments[0]);
+  std::string new_fn = readStringAtAddress(state, arguments[1]);
+  //llvm::cerr << "Replacing " << old_fn << "() with " << new_fn << "()\n";
+  if (old_fn == new_fn)
+    state.removeFnAlias(old_fn);
+  else state.addFnAlias(old_fn, new_fn);
+}
+
+void SpecialFunctionHandler::handleAssert(ExecutionState &state,
+                                          KInstruction *target,
+                                          std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==3 && "invalid number of arguments to _assert");  
+  
+  //XXX:DRE:TAINT
+  if(state.underConstrained) {
+    llvm::cerr << "TAINT: skipping assertion:" 
+               << readStringAtAddress(state, arguments[0]) << "\n";
+    executor.terminateState(state);
+  } else
+    executor.terminateStateOnError(state, 
+                                   "ASSERTION FAIL: " + readStringAtAddress(state, arguments[0]),
+                                   "assert.err");
+}
+
+void SpecialFunctionHandler::handleAssertFail(ExecutionState &state,
+                                              KInstruction *target,
+                                              std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==4 && "invalid number of arguments to __assert_fail");
+  
+  //XXX:DRE:TAINT
+  if(state.underConstrained) {
+    llvm::cerr << "TAINT: skipping assertion:" 
+               << readStringAtAddress(state, arguments[0]) << "\n";
+    executor.terminateState(state);
+  } else
+    executor.terminateStateOnError(state, 
+                                   "ASSERTION FAIL: " + readStringAtAddress(state, arguments[0]),
+                                   "assert.err");
+}
+
+void SpecialFunctionHandler::handleReportError(ExecutionState &state,
+                                               KInstruction *target,
+                                               std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==4 && "invalid number of arguments to klee_report_error");
+  
+  // arguments[0], arguments[1] are file, line
+  
+  //XXX:DRE:TAINT
+  if(state.underConstrained) {
+    llvm::cerr << "TAINT: skipping klee_report_error:"
+               << readStringAtAddress(state, arguments[2]) << ":"
+               << readStringAtAddress(state, arguments[3]) << "\n";
+    executor.terminateState(state);
+  } else
+    executor.terminateStateOnError(state, 
+                                   readStringAtAddress(state, arguments[2]),
+                                   readStringAtAddress(state, arguments[3]));
+}
+
+void SpecialFunctionHandler::handleMerge(ExecutionState &state,
+                           KInstruction *target,
+                           std::vector<ref<Expr> > &arguments) {
+  // nop
+}
+
+void SpecialFunctionHandler::handleNew(ExecutionState &state,
+                         KInstruction *target,
+                         std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 && "invalid number of arguments to new");
+
+  executor.executeAlloc(state, arguments[0], false, target);
+}
+
+void SpecialFunctionHandler::handleDelete(ExecutionState &state,
+                            KInstruction *target,
+                            std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 && "invalid number of arguments to delete");
+  executor.executeFree(state, arguments[0]);
+}
+
+void SpecialFunctionHandler::handleNewArray(ExecutionState &state,
+                              KInstruction *target,
+                              std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 && "invalid number of arguments to new[]");
+  executor.executeAlloc(state, arguments[0], false, target);
+}
+
+void SpecialFunctionHandler::handleDeleteArray(ExecutionState &state,
+                                 KInstruction *target,
+                                 std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 && "invalid number of arguments to delete[]");
+  executor.executeFree(state, arguments[0]);
+}
+
+void SpecialFunctionHandler::handleMalloc(ExecutionState &state,
+                                  KInstruction *target,
+                                  std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 && "invalid number of arguments to malloc");
+  executor.executeAlloc(state, arguments[0], false, target);
+}
+
+void SpecialFunctionHandler::handleMallocN(ExecutionState &state,
+                             KInstruction *target,
+                             std::vector<ref<Expr> > &arguments) {
+
+  // XXX should type check args
+  assert(arguments.size() == 3 && "invalid number of arguments to malloc");
+
+  // mallocn(number, size, alignment)
+  ref<Expr> numElems = executor.toUnique(state, arguments[0]);
+  ref<Expr> elemSize = executor.toUnique(state, arguments[1]);
+  ref<Expr> elemAlignment = executor.toUnique(state, arguments[2]);
+
+  assert(numElems.isConstant() &&
+         elemSize.isConstant() &&
+         elemAlignment.isConstant() &&
+         "symbolic arguments passed to klee_mallocn");
+  
+  executor.executeAllocN(state,
+                         numElems.getConstantValue(),
+                         elemSize.getConstantValue(),
+                         elemAlignment.getConstantValue(),
+                         false,
+                         target);
+}
+
+void SpecialFunctionHandler::handleAssume(ExecutionState &state,
+                            KInstruction *target,
+                            std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 && "invalid number of arguments to klee_assume");
+  
+  ref<Expr> e = arguments[0];
+  
+  if(e.getWidth() != Expr::Bool)
+    e = NeExpr::create(e, ConstantExpr::create(0, e.getWidth()));
+  
+  bool res;
+  bool success = executor.solver->mustBeFalse(state, e, res);
+  assert(success && "FIXME: Unhandled solver failure");
+  if (res) {
+    executor.terminateStateOnError(state, 
+                                   "invalid klee_assume call (provably false)",
+                                   "user.err");
+  } else {
+    executor.addConstraint(state, e);
+  }
+}
+
+void SpecialFunctionHandler::handleIsSymbolic(ExecutionState &state,
+                                KInstruction *target,
+                                std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 && "invalid number of arguments to klee_is_symbolic");
+
+  executor.bindLocal(target, state, 
+                     ConstantExpr::create(!arguments[0].isConstant(), Expr::Int32));
+}
+
+void SpecialFunctionHandler::handlePreferCex(ExecutionState &state,
+                                             KInstruction *target,
+                                             std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to klee_prefex_cex");
+
+  ref<Expr> cond = arguments[1];
+  if (cond.getWidth() != Expr::Bool)
+    cond = NeExpr::create(cond, ref<Expr>(0, cond.getWidth()));
+
+  Executor::ExactResolutionList rl;
+  executor.resolveExact(state, arguments[0], rl, "prefex_cex");
+  
+  assert(rl.size() == 1 &&
+         "prefer_cex target must resolve to precisely one object");
+
+  rl[0].first.first->cexPreferences.push_back(cond);
+}
+
+void SpecialFunctionHandler::handlePrintExpr(ExecutionState &state,
+                                  KInstruction *target,
+                                  std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to klee_print_expr");
+
+  std::string msg_str = readStringAtAddress(state, arguments[0]);
+  llvm::cerr << msg_str << ":" << arguments[1] << "\n";
+}
+
+
+void SpecialFunctionHandler::handleUnderConstrained(ExecutionState &state,
+                                  KInstruction *target,
+                                  std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_under_constrained().");
+  assert(arguments[0].isConstant() &&
+   	 "symbolic argument given to klee_under_constrained!");
+
+  unsigned v = arguments[0].getConstantValue();
+  llvm::cerr << "argument = " << v << " under=" << state.underConstrained << "\n";
+  if(v) {
+    assert(state.underConstrained == false &&
+         "Bogus call to klee_under_constrained().");
+    state.underConstrained = v;
+    llvm::cerr << "turning on under!\n";
+  } else {
+    assert(state.underConstrained != 0 && "Bogus call to klee_taint_end()");
+    state.underConstrained = 0;
+    llvm::cerr << "turning off under!\n";
+  }
+}
+
+void SpecialFunctionHandler::handleSetForking(ExecutionState &state,
+                                              KInstruction *target,
+                                              std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_set_forking");
+  ref<Expr> value = executor.toUnique(state, arguments[0]);
+  
+  if (!value.isConstant()) {
+    executor.terminateStateOnError(state, 
+                                   "klee_set_forking requires a constant arg",
+                                   "user.err");
+  } else {
+    state.forkDisabled = !value.getConstantValue();
+  }
+}
+
+void SpecialFunctionHandler::handleWarning(ExecutionState &state,
+                                           KInstruction *target,
+                                           std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 && "invalid number of arguments to klee_warning");
+
+  std::string msg_str = readStringAtAddress(state, arguments[0]);
+  klee_warning("%s: %s", state.stack.back().kf->function->getName().c_str(), 
+               msg_str.c_str());
+}
+
+void SpecialFunctionHandler::handleWarningOnce(ExecutionState &state,
+                                               KInstruction *target,
+                                               std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_warning_once");
+
+  std::string msg_str = readStringAtAddress(state, arguments[0]);
+  klee_warning_once(0, "%s: %s", state.stack.back().kf->function->getName().c_str(), 
+                    msg_str.c_str());
+}
+
+void SpecialFunctionHandler::handlePrintRange(ExecutionState &state,
+                                  KInstruction *target,
+                                  std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to klee_print_range");
+
+  std::string msg_str = readStringAtAddress(state, arguments[0]);
+  llvm::cerr << msg_str << ":" << arguments[1];
+  if (!arguments[1].isConstant()) {
+    // FIXME: Pull into a unique value method?
+    ref<Expr> value;
+    bool success = executor.solver->getValue(state, arguments[1], value);
+    assert(success && "FIXME: Unhandled solver failure");
+    bool res;
+    success = executor.solver->mustBeTrue(state, 
+                                          EqExpr::create(arguments[1], value), 
+                                          res);
+    assert(success && "FIXME: Unhandled solver failure");
+    if (res) {
+      llvm::cerr << " == " << value;
+    } else { 
+      llvm::cerr << " ~= " << value;
+      std::pair< ref<Expr>, ref<Expr> > res =
+        executor.solver->getRange(state, arguments[1]);
+      llvm::cerr << " (in [" << res.first << ", " << res.second <<"])";
+    }
+  }
+  llvm::cerr << "\n";
+}
+
+void SpecialFunctionHandler::handleGetObjSize(ExecutionState &state,
+                                  KInstruction *target,
+                                  std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_get_obj_size");
+  Executor::ExactResolutionList rl;
+  executor.resolveExact(state, arguments[0], rl, "klee_get_obj_size");
+  for (Executor::ExactResolutionList::iterator it = rl.begin(), 
+         ie = rl.end(); it != ie; ++it) {
+    executor.bindLocal(target, *it->second, 
+                       ConstantExpr::create(it->first.first->size, Expr::Int32));
+  }
+}
+
+void SpecialFunctionHandler::handleGetErrno(ExecutionState &state,
+                                            KInstruction *target,
+                                            std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==0 &&
+         "invalid number of arguments to klee_get_obj_size");
+  executor.bindLocal(target, state,
+                     ConstantExpr::create(errno, Expr::Int32));
+}
+
+void SpecialFunctionHandler::handleCalloc(ExecutionState &state,
+                            KInstruction *target,
+                            std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to calloc");
+
+  ref<Expr> size = MulExpr::create(arguments[0],
+                                   arguments[1]);
+  executor.executeAlloc(state, size, false, target, true);
+}
+
+void SpecialFunctionHandler::handleRealloc(ExecutionState &state,
+                            KInstruction *target,
+                            std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to realloc");
+  ref<Expr> address = arguments[0];
+  ref<Expr> size = arguments[1];
+
+  Executor::StatePair zeroSize = executor.fork(state, 
+                                               Expr::createIsZero(size), 
+                                               true);
+  
+  if (zeroSize.first) { // size == 0
+    executor.executeFree(*zeroSize.first, address, target);   
+  }
+  if (zeroSize.second) { // size != 0
+    Executor::StatePair zeroPointer = executor.fork(*zeroSize.second, 
+                                                    Expr::createIsZero(address), 
+                                                    true);
+    
+    if (zeroPointer.first) { // address == 0
+      executor.executeAlloc(*zeroPointer.first, size, false, target);
+    } 
+    if (zeroPointer.second) { // address != 0
+      Executor::ExactResolutionList rl;
+      executor.resolveExact(*zeroPointer.second, address, rl, "realloc");
+      
+      for (Executor::ExactResolutionList::iterator it = rl.begin(), 
+             ie = rl.end(); it != ie; ++it) {
+        executor.executeAlloc(*it->second, size, false, target, false, 
+                              it->first.second);
+      }
+    }
+  }
+}
+
+void SpecialFunctionHandler::handleFree(ExecutionState &state,
+                          KInstruction *target,
+                          std::vector<ref<Expr> > &arguments) {
+  // XXX should type check args
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to free");
+  executor.executeFree(state, arguments[0]);
+}
+
+void SpecialFunctionHandler::handleCheckMemoryAccess(ExecutionState &state,
+                                            KInstruction *target,
+                                            std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to klee_check_memory_access");
+
+  ref<Expr> address = executor.toUnique(state, arguments[0]);
+  ref<Expr> size = executor.toUnique(state, arguments[1]);
+  if (!address.isConstant() || !size.isConstant()) {
+    executor.terminateStateOnError(state, 
+                                   "check_memory_access requires constant args",
+                                   "user.err");
+  } else {
+    ObjectPair op;
+
+    if (!state.addressSpace.resolveOne(address.getConstantValue(), op)) {
+      executor.terminateStateOnError(state,
+                                     "check_memory_access: memory error",
+                                     "ptr.err",
+                                     executor.getAddressInfo(state, address));
+    } else {
+      ref<Expr> chk = op.first->getBoundsCheckPointer(address, 
+                                                      size.getConstantValue());
+      assert(chk.isConstant());
+      if (!chk.getConstantValue()) {
+        executor.terminateStateOnError(state,
+                                       "check_memory_access: memory error",
+                                       "ptr.err",
+                                       executor.getAddressInfo(state, address));
+      }
+    }
+  }
+}
+
+void SpecialFunctionHandler::handleGetValue(ExecutionState &state,
+                                            KInstruction *target,
+                                            std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_get_value");
+
+  executor.executeGetValue(state, arguments[0], target);
+}
+
+void SpecialFunctionHandler::handleDefineFixedObject(ExecutionState &state,
+                                                     KInstruction *target,
+                                                     std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==2 &&
+         "invalid number of arguments to klee_define_fixed_object");
+  assert(arguments[0].isConstant() &&
+         "expect constant address argument to klee_define_fixed_object");
+  assert(arguments[1].isConstant() &&
+         "expect constant size argument to klee_define_fixed_object");
+  
+  uint64_t address = arguments[0].getConstantValue();
+  uint64_t size = arguments[1].getConstantValue();
+  MemoryObject *mo = executor.memory->allocateFixed(address, size, state.prevPC->inst);
+  executor.bindObjectInState(state, mo, false);
+  mo->isUserSpecified = true; // XXX hack;
+}
+
+void SpecialFunctionHandler::handleMakeSymbolic(ExecutionState &state,
+                                                KInstruction *target,
+                                                std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==3 &&
+         "invalid number of arguments to klee_make_symbolic[_name]");  
+
+  Executor::ExactResolutionList rl;
+  executor.resolveExact(state, arguments[0], rl, "make_symbolic");
+  
+  for (Executor::ExactResolutionList::iterator it = rl.begin(), 
+         ie = rl.end(); it != ie; ++it) {
+    MemoryObject *mo = (MemoryObject*) it->first.first;
+    std::string name = readStringAtAddress(state, arguments[2]);
+    mo->setName(name);
+    
+    const ObjectState *old = it->first.second;
+    ExecutionState *s = it->second;
+    
+    if (old->readOnly) {
+      executor.terminateStateOnError(*s, 
+                                     "cannot make readonly object symbolic", 
+                                     "user.err");
+      return;
+    } 
+
+    bool res;
+    bool success =
+      executor.solver->mustBeTrue(*s, EqExpr::create(arguments[1],
+                                                     mo->getSizeExpr()),
+                                  res);
+    assert(success && "FIXME: Unhandled solver failure");
+    
+    if (res) {
+      executor.executeMakeSymbolic(*s, mo);
+    } else {      
+      executor.terminateStateOnError(*s, 
+                                     "wrong size given to klee_make_symbolic[_name]", 
+                                     "user.err");
+    }
+  }
+}
+
+void SpecialFunctionHandler::handleMarkGlobal(ExecutionState &state,
+                                              KInstruction *target,
+                                              std::vector<ref<Expr> > &arguments) {
+  assert(arguments.size()==1 &&
+         "invalid number of arguments to klee_mark_global");  
+
+  Executor::ExactResolutionList rl;
+  executor.resolveExact(state, arguments[0], rl, "mark_global");
+  
+  for (Executor::ExactResolutionList::iterator it = rl.begin(), 
+         ie = rl.end(); it != ie; ++it) {
+    MemoryObject *mo = (MemoryObject*) it->first.first;
+    assert(!mo->isLocal);
+    mo->isGlobal = true;
+  }
+}

Added: klee/trunk/lib/Core/SpecialFunctionHandler.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/SpecialFunctionHandler.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/SpecialFunctionHandler.h (added)
+++ klee/trunk/lib/Core/SpecialFunctionHandler.h Wed May 20 23:36:41 2009
@@ -0,0 +1,106 @@
+//===-- SpecialFunctionHandler.h --------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_SPECIALFUNCTIONHANDLER_H
+#define KLEE_SPECIALFUNCTIONHANDLER_H
+
+#include <map>
+#include <vector>
+#include <string>
+
+namespace llvm {
+  class Function;
+}
+
+namespace klee {
+  class Executor;
+  class Expr;
+  class ExecutionState;
+  class KInstruction;
+  template<typename T> class ref;
+  
+  class SpecialFunctionHandler {
+  public:
+    typedef void (SpecialFunctionHandler::*Handler)(ExecutionState &state,
+                                                    KInstruction *target, 
+                                                    std::vector<ref<Expr> > 
+                                                      &arguments);
+    typedef std::map<const llvm::Function*, 
+                     std::pair<Handler,bool> > handlers_ty;
+
+    handlers_ty handlers;
+    class Executor &executor;
+
+  public:
+    SpecialFunctionHandler(Executor &_executor);
+
+    /// Perform any modifications on the LLVM module before it is
+    /// prepared for execution. At the moment this involves deleting
+    /// unused function bodies and marking intrinsics with appropriate
+    /// flags for use in optimizations.
+    void prepare();
+
+    /// Initialize the internal handler map after the module has been
+    /// prepared for execution.
+    void bind();
+
+    bool handle(ExecutionState &state, 
+                llvm::Function *f,
+                KInstruction *target,
+                std::vector< ref<Expr> > &arguments);
+
+    /* Convenience routines */
+
+    std::string readStringAtAddress(ExecutionState &state, ref<Expr> address);
+    
+    /* Handlers */
+
+#define HANDLER(name) void name(ExecutionState &state, \
+                                KInstruction *target, \
+                                std::vector< ref<Expr> > &arguments)
+    HANDLER(handleAbort);
+    HANDLER(handleAssert);
+    HANDLER(handleAssertFail);
+    HANDLER(handleAssume);
+    HANDLER(handleCalloc);
+    HANDLER(handleCheckMemoryAccess);
+    HANDLER(handleDefineFixedObject);
+    HANDLER(handleDelete);    
+    HANDLER(handleDeleteArray);
+    HANDLER(handleExit);
+    HANDLER(handleAliasFunction);
+    HANDLER(handleFree);
+    HANDLER(handleGetErrno);
+    HANDLER(handleGetObjSize);
+    HANDLER(handleGetValue);
+    HANDLER(handleIsSymbolic);
+    HANDLER(handleMakeSymbolic);
+    HANDLER(handleMalloc);
+    HANDLER(handleMallocN);
+    HANDLER(handleMarkGlobal);
+    HANDLER(handleMerge);
+    HANDLER(handleNew);
+    HANDLER(handleNewArray);
+    HANDLER(handlePreferCex);
+    HANDLER(handlePrintExpr);
+    HANDLER(handlePrintRange);
+    HANDLER(handleRange);
+    HANDLER(handleRealloc);
+    HANDLER(handleReportError);
+    HANDLER(handleRevirtObjects);
+    HANDLER(handleSetForking);
+    HANDLER(handleSilentExit);
+    HANDLER(handleUnderConstrained);
+    HANDLER(handleWarning);
+    HANDLER(handleWarningOnce);
+#undef HANDLER
+  };
+} // End klee namespace
+
+#endif

Added: klee/trunk/lib/Core/StatsTracker.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/StatsTracker.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/StatsTracker.cpp (added)
+++ klee/trunk/lib/Core/StatsTracker.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,814 @@
+//===-- StatsTracker.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "StatsTracker.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Statistics.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Module/KModule.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Support/ModuleUtil.h"
+#include "klee/Internal/System/Time.h"
+
+#include "CallPathManager.h"
+#include "CoreStats.h"
+#include "Executor.h"
+#include "MemoryManager.h"
+#include "UserSearcher.h"
+#include "../Solver/SolverStats.h"
+
+#include "llvm/BasicBlock.h"
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/InlineAsm.h"
+#include "llvm/Module.h"
+#include "llvm/Type.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/System/Process.h"
+#include "llvm/System/Path.h"
+
+#include <iostream>
+#include <fstream>
+
+using namespace klee;
+using namespace llvm;
+
+///
+
+namespace {  
+  cl::opt<bool>
+  TrackInstructionTime("track-instruction-time",
+                       cl::desc("Enable tracking of time for individual instructions"),
+                       cl::init(false));
+
+  cl::opt<bool>
+  OutputStats("output-stats",
+              cl::desc("Write running stats trace file"),
+              cl::init(true));
+
+  cl::opt<bool>
+  OutputIStats("output-istats",
+               cl::desc("Write instruction level statistics (in callgrind format)"),
+               cl::init(true));
+
+  cl::opt<double>
+  StatsWriteInterval("stats-write-interval",
+                     cl::desc("Approximate number of seconds between stats writes (default: 1.0)"),
+                     cl::init(1.));
+
+  cl::opt<double>
+  IStatsWriteInterval("istats-write-interval",
+                      cl::desc("Approximate number of seconds between istats writes (default: 10.0)"),
+                      cl::init(10.));
+
+  /*
+  cl::opt<double>
+  BranchCovCountsWriteInterval("branch-cov-counts-write-interval",
+                     cl::desc("Approximate number of seconds between run.branches writes (default: 5.0)"),
+                     cl::init(5.));
+  */
+
+  // XXX I really would like to have dynamic rate control for something like this.
+  cl::opt<double>
+  UncoveredUpdateInterval("uncovered-update-interval",
+                          cl::init(30.));
+  
+  cl::opt<bool>
+  UseCallPaths("use-call-paths",
+               cl::desc("Enable calltree tracking for instruction level statistics"),
+               cl::init(true));
+  
+}
+
+///
+
+bool StatsTracker::useStatistics() {
+  return OutputStats || OutputIStats;
+}
+
+namespace klee {
+  class WriteIStatsTimer : public Executor::Timer {
+    StatsTracker *statsTracker;
+    
+  public:
+    WriteIStatsTimer(StatsTracker *_statsTracker) : statsTracker(_statsTracker) {}
+    ~WriteIStatsTimer() {}
+    
+    void run() { statsTracker->writeIStats(); }
+  };
+  
+  class WriteStatsTimer : public Executor::Timer {
+    StatsTracker *statsTracker;
+    
+  public:
+    WriteStatsTimer(StatsTracker *_statsTracker) : statsTracker(_statsTracker) {}
+    ~WriteStatsTimer() {}
+    
+    void run() { statsTracker->writeStatsLine(); }
+  };
+
+  class UpdateReachableTimer : public Executor::Timer {
+    StatsTracker *statsTracker;
+    
+  public:
+    UpdateReachableTimer(StatsTracker *_statsTracker) : statsTracker(_statsTracker) {}
+    
+    void run() { statsTracker->computeReachableUncovered(); }
+  };
+ 
+}
+
+//
+
+/// Check for special cases where we statically know an instruction is
+/// uncoverable. Currently the case is an unreachable instruction
+/// following a noreturn call; the instruction is really only there to
+/// satisfy LLVM's termination requirement.
+static bool instructionIsCoverable(Instruction *i) {
+  if (i->getOpcode() == Instruction::Unreachable) {
+    BasicBlock *bb = i->getParent();
+    BasicBlock::iterator it(i);
+    if (it==bb->begin()) {
+      return true;
+    } else {
+      Instruction *prev = --it;
+      if (isa<CallInst>(prev) || isa<InvokeInst>(prev)) {
+        Function *target = getDirectCallTarget(prev);
+        if (target && target->doesNotReturn())
+          return false;
+      }
+    }
+  }
+
+  return true;
+}
+
+StatsTracker::StatsTracker(Executor &_executor, std::string _objectFilename,
+                           bool _updateMinDistToUncovered)
+  : executor(_executor),
+    objectFilename(_objectFilename),
+    statsFile(0),
+    istatsFile(0),
+    startWallTime(util::getWallTime()),
+    numBranches(0),
+    fullBranches(0),
+    partialBranches(0),
+    updateMinDistToUncovered(_updateMinDistToUncovered) {
+  KModule *km = executor.kmodule;
+
+  sys::Path module(objectFilename);
+  if (!sys::Path(objectFilename).isAbsolute()) {
+    sys::Path current = sys::Path::GetCurrentDirectory();
+    current.appendComponent(objectFilename);
+    if (current.exists())
+      objectFilename = current.c_str();
+  }
+
+  if (OutputIStats)
+    theStatisticManager->useIndexedStats(km->infos->getMaxID());
+
+  for (std::vector<KFunction*>::iterator it = km->functions.begin(), 
+         ie = km->functions.end(); it != ie; ++it) {
+    KFunction *kf = *it;
+    kf->trackCoverage = 1;
+
+    for (unsigned i=0; i<kf->numInstructions; ++i) {
+      KInstruction *ki = kf->instructions[i];
+
+      if (OutputIStats) {
+        unsigned id = ki->info->id;
+        theStatisticManager->setIndex(id);
+        if (kf->trackCoverage && instructionIsCoverable(ki->inst))
+          ++stats::uncoveredInstructions;
+      }
+      
+      if (kf->trackCoverage) {
+        if (BranchInst *bi = dyn_cast<BranchInst>(ki->inst))
+          if (!bi->isUnconditional())
+            numBranches++;
+      }
+    }
+  }
+
+  if (OutputStats) {
+    statsFile = executor.interpreterHandler->openOutputFile("run.stats");
+    assert(statsFile && "unable to open statistics trace file");
+    writeStatsHeader();
+    writeStatsLine();
+
+    executor.addTimer(new WriteStatsTimer(this), StatsWriteInterval);
+
+    if (updateMinDistToUncovered)
+      executor.addTimer(new UpdateReachableTimer(this), UncoveredUpdateInterval);
+  }
+
+  if (OutputIStats) {
+    istatsFile = executor.interpreterHandler->openOutputFile("run.istats");
+    assert(istatsFile && "unable to open istats file");
+
+    executor.addTimer(new WriteIStatsTimer(this), IStatsWriteInterval);
+  }
+}
+
+StatsTracker::~StatsTracker() {  
+  if (statsFile)
+    delete statsFile;
+  if (istatsFile)
+    delete istatsFile;
+}
+
+void StatsTracker::done() {
+  if (statsFile)
+    writeStatsLine();
+  if (OutputIStats)
+    writeIStats();
+}
+
+void StatsTracker::stepInstruction(ExecutionState &es) {
+  if (OutputIStats) {
+    if (TrackInstructionTime) {
+      static sys::TimeValue lastNowTime(0,0),lastUserTime(0,0);
+    
+      if (lastUserTime.seconds()==0 && lastUserTime.nanoseconds()==0) {
+        sys::TimeValue sys(0,0);
+        sys::Process::GetTimeUsage(lastNowTime,lastUserTime,sys);
+      } else {
+        sys::TimeValue now(0,0),user(0,0),sys(0,0);
+        sys::Process::GetTimeUsage(now,user,sys);
+        sys::TimeValue delta = user - lastUserTime;
+        sys::TimeValue deltaNow = now - lastNowTime;
+        stats::instructionTime += delta.usec();
+        stats::instructionRealTime += deltaNow.usec();
+        lastUserTime = user;
+        lastNowTime = now;
+      }
+    }
+
+    Instruction *inst = es.pc->inst;
+    const InstructionInfo &ii = *es.pc->info;
+    StackFrame &sf = es.stack.back();
+    theStatisticManager->setIndex(ii.id);
+    if (UseCallPaths)
+      theStatisticManager->setContext(&sf.callPathNode->statistics);
+
+    if (es.instsSinceCovNew)
+      ++es.instsSinceCovNew;
+
+    if (sf.kf->trackCoverage && instructionIsCoverable(inst)) {
+      if (!theStatisticManager->getIndexedValue(stats::coveredInstructions, ii.id)) {
+        // Checking for actual stoppoints avoids inconsistencies due
+        // to line number propogation.
+        if (isa<DbgStopPointInst>(inst))
+          es.coveredLines[&ii.file].insert(ii.line);
+	es.coveredNew = true;
+        es.instsSinceCovNew = 1;
+	++stats::coveredInstructions;
+	stats::uncoveredInstructions += (uint64_t)-1;
+      }
+    }
+  }
+}
+
+///
+
+/* Should be called _after_ the es->pushFrame() */
+void StatsTracker::framePushed(ExecutionState &es, StackFrame *parentFrame) {
+  if (OutputIStats) {
+    StackFrame &sf = es.stack.back();
+
+    if (UseCallPaths) {
+      CallPathNode *parent = parentFrame ? parentFrame->callPathNode : 0;
+      CallPathNode *cp = callPathManager.getCallPath(parent, 
+                                                     sf.caller ? sf.caller->inst : 0, 
+                                                     sf.kf->function);
+      sf.callPathNode = cp;
+      cp->count++;
+    }
+
+    if (updateMinDistToUncovered) {
+      uint64_t minDistAtRA = 0;
+      if (parentFrame)
+        minDistAtRA = parentFrame->minDistToUncoveredOnReturn;
+      
+      sf.minDistToUncoveredOnReturn = sf.caller ?
+        computeMinDistToUncovered(sf.caller, minDistAtRA) : 0;
+    }
+  }
+}
+
+/* Should be called _after_ the es->popFrame() */
+void StatsTracker::framePopped(ExecutionState &es) {
+  // XXX remove me?
+}
+
+
+void StatsTracker::markBranchVisited(ExecutionState *visitedTrue, 
+                                     ExecutionState *visitedFalse) {
+  if (OutputIStats) {
+    unsigned id = theStatisticManager->getIndex();
+    uint64_t hasTrue = theStatisticManager->getIndexedValue(stats::trueBranches, id);
+    uint64_t hasFalse = theStatisticManager->getIndexedValue(stats::falseBranches, id);
+    if (visitedTrue && !hasTrue) {
+      visitedTrue->coveredNew = true;
+      visitedTrue->instsSinceCovNew = 1;
+      ++stats::trueBranches;
+      if (hasFalse) { ++fullBranches; --partialBranches; }
+      else ++partialBranches;
+      hasTrue = 1;
+    }
+    if (visitedFalse && !hasFalse) {
+      visitedFalse->coveredNew = true;
+      visitedFalse->instsSinceCovNew = 1;
+      ++stats::falseBranches;
+      if (hasTrue) { ++fullBranches; --partialBranches; }
+      else ++partialBranches;
+    }
+  }
+}
+
+void StatsTracker::writeStatsHeader() {
+  *statsFile << "('Instructions',"
+             << "'FullBranches',"
+             << "'PartialBranches',"
+             << "'NumBranches',"
+             << "'UserTime',"
+             << "'NumStates',"
+             << "'MallocUsage',"
+             << "'NumQueries',"
+             << "'NumQueryConstructs',"
+             << "'NumObjects',"
+             << "'WallTime',"
+             << "'CoveredInstructions',"
+             << "'UncoveredInstructions',"
+             << "'QueryTime',"
+             << "'SolverTime',"
+             << "'CexCacheTime',"
+             << "'ForkTime',"
+             << "'ResolveTime',"
+             << ")\n";
+  statsFile->flush();
+}
+
+double StatsTracker::elapsed() {
+  return util::getWallTime() - startWallTime;
+}
+
+void StatsTracker::writeStatsLine() {
+  *statsFile << "(" << stats::instructions
+             << "," << fullBranches
+             << "," << partialBranches
+             << "," << numBranches
+             << "," << util::getUserTime()
+             << "," << executor.states.size()
+             << "," << sys::Process::GetTotalMemoryUsage()
+             << "," << stats::queries
+             << "," << stats::queryConstructs
+             << "," << 0 // was numObjects
+             << "," << elapsed()
+             << "," << stats::coveredInstructions
+             << "," << stats::uncoveredInstructions
+             << "," << stats::queryTime / 1000000.
+             << "," << stats::solverTime / 1000000.
+             << "," << stats::cexCacheTime / 1000000.
+             << "," << stats::forkTime / 1000000.
+             << "," << stats::resolveTime / 1000000.
+             << ")\n";
+  statsFile->flush();
+}
+
+void StatsTracker::updateStateStatistics(uint64_t addend) {
+  for (std::set<ExecutionState*>::iterator it = executor.states.begin(),
+         ie = executor.states.end(); it != ie; ++it) {
+    ExecutionState &state = **it;
+    const InstructionInfo &ii = *state.pc->info;
+    theStatisticManager->incrementIndexedValue(stats::states, ii.id, addend);
+    if (UseCallPaths)
+      state.stack.back().callPathNode->statistics.incrementValue(stats::states, addend);
+  }
+}
+
+void StatsTracker::writeIStats() {
+  Module *m = executor.kmodule->module;
+  uint64_t istatsMask = 0;
+  std::ostream &of = *istatsFile;
+  
+  of.seekp(0, std::ios::end);
+  unsigned istatsSize = of.tellp();
+  of.seekp(0);
+
+  of << "version: 1\n";
+  of << "creator: klee\n";
+  of << "pid: " << sys::Process::GetCurrentUserId() << "\n";
+  of << "cmd: " << m->getModuleIdentifier() << "\n\n";
+  of << "\n";
+  
+  StatisticManager &sm = *theStatisticManager;
+  unsigned nStats = sm.getNumStatistics();
+
+  // Max is 13, sadly
+  istatsMask |= 1<<sm.getStatisticID("Queries");
+  istatsMask |= 1<<sm.getStatisticID("QueriesValid");
+  istatsMask |= 1<<sm.getStatisticID("QueriesInvalid");
+  istatsMask |= 1<<sm.getStatisticID("QueryTime");
+  istatsMask |= 1<<sm.getStatisticID("ResolveTime");
+  istatsMask |= 1<<sm.getStatisticID("Instructions");
+  istatsMask |= 1<<sm.getStatisticID("InstructionTimes");
+  istatsMask |= 1<<sm.getStatisticID("InstructionRealTimes");
+  istatsMask |= 1<<sm.getStatisticID("Forks");
+  istatsMask |= 1<<sm.getStatisticID("CoveredInstructions");
+  istatsMask |= 1<<sm.getStatisticID("UncoveredInstructions");
+  istatsMask |= 1<<sm.getStatisticID("States");
+  istatsMask |= 1<<sm.getStatisticID("MinDistToUncovered");
+
+  of << "positions: instr line\n";
+
+  for (unsigned i=0; i<nStats; i++) {
+    if (istatsMask & (1<<i)) {
+      Statistic &s = sm.getStatistic(i);
+      of << "event: " << s.getShortName() << " : " 
+         << s.getName() << "\n";
+    }
+  }
+
+  of << "events: ";
+  for (unsigned i=0; i<nStats; i++) {
+    if (istatsMask & (1<<i))
+      of << sm.getStatistic(i).getShortName() << " ";
+  }
+  of << "\n";
+  
+  // set state counts, decremented after we process so that we don't
+  // have to zero all records each time.
+  if (istatsMask & (1<<stats::states.getID()))
+    updateStateStatistics(1);
+
+  std::string sourceFile = "";
+
+  CallSiteSummaryTable callSiteStats;
+  if (UseCallPaths)
+    callPathManager.getSummaryStatistics(callSiteStats);
+
+  of << "ob=" << objectFilename << "\n";
+
+  for (Module::iterator fnIt = m->begin(), fn_ie = m->end(); 
+       fnIt != fn_ie; ++fnIt) {
+    if (!fnIt->isDeclaration()) {
+      of << "fn=" << fnIt->getName() << "\n";
+      for (Function::iterator bbIt = fnIt->begin(), bb_ie = fnIt->end(); 
+           bbIt != bb_ie; ++bbIt) {
+        for (BasicBlock::iterator it = bbIt->begin(), ie = bbIt->end(); 
+             it != it; ++it) {
+          Instruction *instr = &*it;
+          const InstructionInfo &ii = executor.kmodule->infos->getInfo(instr);
+          unsigned index = ii.id;
+          if (ii.file!=sourceFile) {
+            of << "fl=" << ii.file << "\n";
+            sourceFile = ii.file;
+          }
+          of << ii.assemblyLine << " ";
+          of << ii.line << " ";
+          for (unsigned i=0; i<nStats; i++)
+            if (istatsMask&(1<<i))
+              of << sm.getIndexedValue(sm.getStatistic(i), index) << " ";
+          of << "\n";
+
+          if (UseCallPaths && 
+              (isa<CallInst>(instr) || isa<InvokeInst>(instr))) {
+            CallSiteSummaryTable::iterator it = callSiteStats.find(instr);
+            if (it!=callSiteStats.end()) {
+              for (std::map<llvm::Function*, CallSiteInfo>::iterator
+                     fit = it->second.begin(), fie = it->second.end(); 
+                   fit != fie; ++fit) {
+                Function *f = fit->first;
+                CallSiteInfo &csi = fit->second;
+                const InstructionInfo &fii = 
+                  executor.kmodule->infos->getFunctionInfo(f);
+  
+                if (fii.file!="" && fii.file!=sourceFile)
+                  of << "cfl=" << fii.file << "\n";
+                of << "cfn=" << f->getName() << "\n";
+                of << "calls=" << csi.count << " ";
+                of << fii.assemblyLine << " ";
+                of << fii.line << "\n";
+
+                of << ii.assemblyLine << " ";
+                of << ii.line << " ";
+                for (unsigned i=0; i<nStats; i++) {
+                  if (istatsMask&(1<<i)) {
+                    Statistic &s = sm.getStatistic(i);
+                    uint64_t value;
+
+                    // Hack, ignore things that don't make sense on
+                    // call paths.
+                    if (&s == &stats::uncoveredInstructions) {
+                      value = 0;
+                    } else {
+                      value = csi.statistics.getValue(s);
+                    }
+
+                    of << value << " ";
+                  }
+                }
+                of << "\n";
+              }
+            }
+          }
+        }
+      }
+    }
+  }
+
+  if (istatsMask & (1<<stats::states.getID()))
+    updateStateStatistics((uint64_t)-1);
+  
+  // Clear then end of the file if necessary (no truncate op?).
+  unsigned pos = of.tellp();
+  for (unsigned i=pos; i<istatsSize; ++i)
+    of << '\n';
+  
+  of.flush();
+}
+
+///
+
+typedef std::map<Instruction*, std::vector<Function*> > calltargets_ty;
+
+static calltargets_ty callTargets;
+static std::map<Function*, std::vector<Instruction*> > functionCallers;
+static std::map<Function*, unsigned> functionShortestPath;
+
+static std::vector<Instruction*> getSuccs(Instruction *i) {
+  BasicBlock *bb = i->getParent();
+  std::vector<Instruction*> res;
+
+  if (i==bb->getTerminator()) {
+    for (succ_iterator it = succ_begin(bb), ie = succ_end(bb); it != ie; ++it)
+      res.push_back(it->begin());
+  } else {
+    res.push_back(++BasicBlock::iterator(i));
+  }
+
+  return res;
+}
+
+uint64_t klee::computeMinDistToUncovered(const KInstruction *ki,
+                                         uint64_t minDistAtRA) {
+  StatisticManager &sm = *theStatisticManager;
+  if (minDistAtRA==0) { // unreachable on return, best is local
+    return sm.getIndexedValue(stats::minDistToUncovered,
+                              ki->info->id);
+  } else {
+    uint64_t minDistLocal = sm.getIndexedValue(stats::minDistToUncovered,
+                                               ki->info->id);
+    uint64_t distToReturn = sm.getIndexedValue(stats::minDistToReturn,
+                                               ki->info->id);
+
+    if (distToReturn==0) { // return unreachable, best is local
+      return minDistLocal;
+    } else if (!minDistLocal) { // no local reachable
+      return distToReturn + minDistAtRA;
+    } else {
+      return std::min(minDistLocal, distToReturn + minDistAtRA);
+    }
+  }
+}
+
+void StatsTracker::computeReachableUncovered() {
+  KModule *km = executor.kmodule;
+  Module *m = km->module;
+  static bool init = true;
+  const InstructionInfoTable &infos = *km->infos;
+  StatisticManager &sm = *theStatisticManager;
+  
+  if (init) {
+    init = false;
+
+    // Compute call targets. It would be nice to use alias information
+    // instead of assuming all indirect calls hit all escaping
+    // functions, eh?
+    for (Module::iterator fnIt = m->begin(), fn_ie = m->end(); 
+         fnIt != fn_ie; ++fnIt) {
+      for (Function::iterator bbIt = fnIt->begin(), bb_ie = fnIt->end(); 
+           bbIt != bb_ie; ++bbIt) {
+        for (BasicBlock::iterator it = bbIt->begin(), ie = bbIt->end(); 
+             it != it; ++it) {
+          if (isa<CallInst>(it) || isa<InvokeInst>(it)) {
+            if (isa<InlineAsm>(it->getOperand(0))) {
+              // We can never call through here so assume no targets
+              // (which should be correct anyhow).
+              callTargets.insert(std::make_pair(it,
+                                                std::vector<Function*>()));
+            } else if (Function *target = getDirectCallTarget(it)) {
+              callTargets[it].push_back(target);
+            } else {
+              callTargets[it] = 
+                std::vector<Function*>(km->escapingFunctions.begin(),
+                                       km->escapingFunctions.end());
+            }
+          }
+        }
+      }
+    }
+
+    // Compute function callers as reflexion of callTargets.
+    for (calltargets_ty::iterator it = callTargets.begin(), 
+           ie = callTargets.end(); it != ie; ++it)
+      for (std::vector<Function*>::iterator fit = it->second.begin(), 
+             fie = it->second.end(); fit != fie; ++fit) 
+        functionCallers[*fit].push_back(it->first);
+
+    // Initialize minDistToReturn to shortest paths through
+    // functions. 0 is unreachable.
+    std::vector<Instruction *> instructions;
+    for (Module::iterator fnIt = m->begin(), fn_ie = m->end(); 
+         fnIt != fn_ie; ++fnIt) {
+      if (fnIt->isDeclaration()) {
+        if (fnIt->doesNotReturn()) {
+          functionShortestPath[fnIt] = 0;
+        } else {
+          functionShortestPath[fnIt] = 1; // whatever
+        }
+      } else {
+        functionShortestPath[fnIt] = 0;
+      }
+
+      // Not sure if I should bother to preorder here. XXX I should.
+      for (Function::iterator bbIt = fnIt->begin(), bb_ie = fnIt->end(); 
+           bbIt != bb_ie; ++bbIt) {
+        for (BasicBlock::iterator it = bbIt->begin(), ie = bbIt->end(); 
+             it != it; ++it) {
+          instructions.push_back(it);
+          unsigned id = infos.getInfo(it).id;
+          sm.setIndexedValue(stats::minDistToReturn, 
+                             id, 
+                             isa<ReturnInst>(it) || isa<UnwindInst>(it));
+        }
+      }
+    }
+  
+    std::reverse(instructions.begin(), instructions.end());
+    
+    // I'm so lazy it's not even worklisted.
+    bool changed;
+    do {
+      changed = false;
+      for (std::vector<Instruction*>::iterator it = instructions.begin(),
+             ie = instructions.end(); it != ie; ++it) {
+        Instruction *inst = *it;
+        unsigned bestThrough = 0;
+
+        if (isa<CallInst>(inst) || isa<InvokeInst>(inst)) {
+          std::vector<Function*> &targets = callTargets[inst];
+          for (std::vector<Function*>::iterator fnIt = targets.begin(),
+                 ie = targets.end(); fnIt != ie; ++fnIt) {
+            uint64_t dist = functionShortestPath[*fnIt];
+            if (dist) {
+              dist = 1+dist; // count instruction itself
+              if (bestThrough==0 || dist<bestThrough)
+                bestThrough = dist;
+            }
+          }
+        } else {
+          bestThrough = 1;
+        }
+       
+        if (bestThrough) {
+          unsigned id = infos.getInfo(*it).id;
+          uint64_t best, cur = best = sm.getIndexedValue(stats::minDistToReturn, id);
+          std::vector<Instruction*> succs = getSuccs(*it);
+          for (std::vector<Instruction*>::iterator it2 = succs.begin(),
+                 ie = succs.end(); it2 != ie; ++it2) {
+            uint64_t dist = sm.getIndexedValue(stats::minDistToReturn,
+                                               infos.getInfo(*it2).id);
+            if (dist) {
+              uint64_t val = bestThrough + dist;
+              if (best==0 || val<best)
+                best = val;
+            }
+          }
+          if (best != cur) {
+            sm.setIndexedValue(stats::minDistToReturn, id, best);
+            changed = true;
+
+            // Update shortest path if this is the entry point.
+            Function *f = inst->getParent()->getParent();
+            if (inst==f->begin()->begin())
+              functionShortestPath[f] = best;
+          }
+        }
+      }
+    } while (changed);
+  }
+
+  // compute minDistToUncovered, 0 is unreachable
+  std::vector<Instruction *> instructions;
+  for (Module::iterator fnIt = m->begin(), fn_ie = m->end(); 
+       fnIt != fn_ie; ++fnIt) {
+    // Not sure if I should bother to preorder here.
+    for (Function::iterator bbIt = fnIt->begin(), bb_ie = fnIt->end(); 
+         bbIt != bb_ie; ++bbIt) {
+      for (BasicBlock::iterator it = bbIt->begin(), ie = bbIt->end(); 
+           it != it; ++it) {
+        unsigned id = infos.getInfo(it).id;
+        instructions.push_back(&*it);
+        sm.setIndexedValue(stats::minDistToUncovered, 
+                           id, 
+                           sm.getIndexedValue(stats::uncoveredInstructions, id));
+      }
+    }
+  }
+  
+  std::reverse(instructions.begin(), instructions.end());
+  
+  // I'm so lazy it's not even worklisted.
+  bool changed;
+  do {
+    changed = false;
+    for (std::vector<Instruction*>::iterator it = instructions.begin(),
+           ie = instructions.end(); it != ie; ++it) {
+      Instruction *inst = *it;
+      uint64_t best, cur = best = sm.getIndexedValue(stats::minDistToUncovered, 
+                                                     infos.getInfo(inst).id);
+      unsigned bestThrough = 0;
+      
+      if (isa<CallInst>(inst) || isa<InvokeInst>(inst)) {
+        std::vector<Function*> &targets = callTargets[inst];
+        for (std::vector<Function*>::iterator fnIt = targets.begin(),
+               ie = targets.end(); fnIt != ie; ++fnIt) {
+          uint64_t dist = functionShortestPath[*fnIt];
+          if (dist) {
+            dist = 1+dist; // count instruction itself
+            if (bestThrough==0 || dist<bestThrough)
+              bestThrough = dist;
+          }
+
+          if (!(*fnIt)->isDeclaration()) {
+            uint64_t calleeDist = sm.getIndexedValue(stats::minDistToUncovered,
+                                                     infos.getFunctionInfo(*fnIt).id);
+            if (calleeDist) {
+              calleeDist = 1+calleeDist; // count instruction itself
+              if (best==0 || calleeDist<best)
+                best = calleeDist;
+            }
+          }
+        }
+      } else {
+        bestThrough = 1;
+      }
+      
+      if (bestThrough) {
+        std::vector<Instruction*> succs = getSuccs(inst);
+        for (std::vector<Instruction*>::iterator it2 = succs.begin(),
+               ie = succs.end(); it2 != ie; ++it2) {
+          uint64_t dist = sm.getIndexedValue(stats::minDistToUncovered,
+                                             infos.getInfo(*it2).id);
+          if (dist) {
+            uint64_t val = bestThrough + dist;
+            if (best==0 || val<best)
+              best = val;
+          }
+        }
+      }
+
+      if (best != cur) {
+        sm.setIndexedValue(stats::minDistToUncovered, 
+                           infos.getInfo(inst).id, 
+                           best);
+        changed = true;
+      }
+    }
+  } while (changed);
+
+  for (std::set<ExecutionState*>::iterator it = executor.states.begin(),
+         ie = executor.states.end(); it != ie; ++it) {
+    ExecutionState *es = *it;
+    uint64_t currentFrameMinDist = 0;
+    for (ExecutionState::stack_ty::iterator sfIt = es->stack.begin(),
+           sf_ie = es->stack.end(); sfIt != sf_ie; ++sfIt) {
+      ExecutionState::stack_ty::iterator next = sfIt + 1;
+      KInstIterator kii;
+
+      if (next==es->stack.end()) {
+        kii = es->pc;
+      } else {
+        kii = next->caller;
+        ++kii;
+      }
+      
+      sfIt->minDistToUncoveredOnReturn = currentFrameMinDist;
+      
+      currentFrameMinDist = computeMinDistToUncovered(kii, currentFrameMinDist);
+    }
+  }
+}

Added: klee/trunk/lib/Core/StatsTracker.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/StatsTracker.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/StatsTracker.h (added)
+++ klee/trunk/lib/Core/StatsTracker.h Wed May 20 23:36:41 2009
@@ -0,0 +1,93 @@
+//===-- StatsTracker.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_STATSTRACKER_H
+#define KLEE_STATSTRACKER_H
+
+#include "CallPathManager.h"
+
+#include <iostream>
+#include <set>
+
+namespace llvm {
+  class BranchInst;
+  class Function;
+  class Instruction;
+}
+
+namespace klee {
+  class ExecutionState;
+  class Executor;  
+  class InstructionInfoTable;
+  class InterpreterHandler;
+  class KInstruction;
+  class StackFrame;
+
+  class StatsTracker {
+    friend class WriteStatsTimer;
+    friend class WriteIStatsTimer;
+
+    Executor &executor;
+    std::string objectFilename;
+
+    std::ostream *statsFile, *istatsFile;
+    double startWallTime;
+    
+    unsigned numBranches;
+    unsigned fullBranches, partialBranches;
+
+    CallPathManager callPathManager;    
+
+    bool updateMinDistToUncovered;
+
+  public:
+    static bool useStatistics();
+
+  private:
+    void updateStateStatistics(uint64_t addend);
+    void writeStatsHeader();
+    void writeStatsLine();
+    void writeIStats();
+
+  public:
+    StatsTracker(Executor &_executor, std::string _objectFilename,
+                 bool _updateMinDistToUncovered);
+    ~StatsTracker();
+
+    // called after a new StackFrame has been pushed (for callpath tracing)
+    void framePushed(ExecutionState &es, StackFrame *parentFrame);
+
+    // called after a StackFrame has been popped 
+    void framePopped(ExecutionState &es);
+
+    // called when some side of a branch has been visited. it is
+    // imperative that this be called when the statistics index is at
+    // the index for the branch itself.
+    void markBranchVisited(ExecutionState *visitedTrue, 
+                           ExecutionState *visitedFalse);
+    
+    // called when execution is done and stats files should be flushed
+    void done();
+
+    // process stats for a single instruction step, es is the state
+    // about to be stepped
+    void stepInstruction(ExecutionState &es);
+
+    /// Return time in seconds since execution start.
+    double elapsed();
+
+    void computeReachableUncovered();
+  };
+
+  uint64_t computeMinDistToUncovered(const KInstruction *ki,
+                                     uint64_t minDistAtRA);
+
+}
+
+#endif

Added: klee/trunk/lib/Core/TimingSolver.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/TimingSolver.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/TimingSolver.cpp (added)
+++ klee/trunk/lib/Core/TimingSolver.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,147 @@
+//===-- TimingSolver.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "TimingSolver.h"
+
+#include "klee/ExecutionState.h"
+#include "klee/Solver.h"
+#include "klee/Statistics.h"
+
+#include "CoreStats.h"
+
+#include "llvm/System/Process.h"
+
+using namespace klee;
+using namespace llvm;
+
+/***/
+
+bool TimingSolver::evaluate(const ExecutionState& state, ref<Expr> expr,
+                            Solver::Validity &result) {
+  // Fast path, to avoid timer and OS overhead.
+  if (expr.isConstant()) {
+    result = expr.getConstantValue() ? Solver::True : Solver::False;
+    return true;
+  }
+
+  sys::TimeValue now(0,0),user(0,0),delta(0,0),sys(0,0);
+  sys::Process::GetTimeUsage(now,user,sys);
+
+  if (simplifyExprs)
+    expr = state.constraints.simplifyExpr(expr);
+
+  bool success = solver->evaluate(Query(state.constraints, expr), result);
+
+  sys::Process::GetTimeUsage(delta,user,sys);
+  delta -= now;
+  stats::solverTime += delta.usec();
+  state.queryCost += delta.usec()/1000000.;
+
+  return success;
+}
+
+bool TimingSolver::mustBeTrue(const ExecutionState& state, ref<Expr> expr, 
+                              bool &result) {
+  // Fast path, to avoid timer and OS overhead.
+  if (expr.isConstant()) {
+    result = expr.getConstantValue() ? true : false;
+    return true;
+  }
+
+  sys::TimeValue now(0,0),user(0,0),delta(0,0),sys(0,0);
+  sys::Process::GetTimeUsage(now,user,sys);
+
+  if (simplifyExprs)
+    expr = state.constraints.simplifyExpr(expr);
+
+  bool success = solver->mustBeTrue(Query(state.constraints, expr), result);
+
+  sys::Process::GetTimeUsage(delta,user,sys);
+  delta -= now;
+  stats::solverTime += delta.usec();
+  state.queryCost += delta.usec()/1000000.;
+
+  return success;
+}
+
+bool TimingSolver::mustBeFalse(const ExecutionState& state, ref<Expr> expr,
+                               bool &result) {
+  return mustBeTrue(state, Expr::createNot(expr), result);
+}
+
+bool TimingSolver::mayBeTrue(const ExecutionState& state, ref<Expr> expr, 
+                             bool &result) {
+  bool res;
+  if (!mustBeFalse(state, expr, res))
+    return false;
+  result = !res;
+  return true;
+}
+
+bool TimingSolver::mayBeFalse(const ExecutionState& state, ref<Expr> expr, 
+                              bool &result) {
+  bool res;
+  if (!mustBeTrue(state, expr, res))
+    return false;
+  result = !res;
+  return true;
+}
+
+bool TimingSolver::getValue(const ExecutionState& state, ref<Expr> expr, 
+                            ref<Expr> &result) {
+  // Fast path, to avoid timer and OS overhead.
+  if (expr.isConstant()) {
+    result = expr;
+    return true;
+  }
+  
+  sys::TimeValue now(0,0),user(0,0),delta(0,0),sys(0,0);
+  sys::Process::GetTimeUsage(now,user,sys);
+
+  if (simplifyExprs)
+    expr = state.constraints.simplifyExpr(expr);
+
+  bool success = solver->getValue(Query(state.constraints, expr), result);
+
+  sys::Process::GetTimeUsage(delta,user,sys);
+  delta -= now;
+  stats::solverTime += delta.usec();
+  state.queryCost += delta.usec()/1000000.;
+
+  return success;
+}
+
+bool 
+TimingSolver::getInitialValues(const ExecutionState& state, 
+                               const std::vector<const Array*>
+                                 &objects,
+                               std::vector< std::vector<unsigned char> >
+                                 &result) {
+  if (objects.empty())
+    return true;
+
+  sys::TimeValue now(0,0),user(0,0),delta(0,0),sys(0,0);
+  sys::Process::GetTimeUsage(now,user,sys);
+
+  bool success = solver->getInitialValues(Query(state.constraints,
+                                                ref<Expr>(0, Expr::Bool)), 
+                                          objects, result);
+  
+  sys::Process::GetTimeUsage(delta,user,sys);
+  delta -= now;
+  stats::solverTime += delta.usec();
+  state.queryCost += delta.usec()/1000000.;
+  
+  return success;
+}
+
+std::pair< ref<Expr>, ref<Expr> >
+TimingSolver::getRange(const ExecutionState& state, ref<Expr> expr) {
+  return solver->getRange(Query(state.constraints, expr));
+}

Added: klee/trunk/lib/Core/TimingSolver.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/TimingSolver.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/TimingSolver.h (added)
+++ klee/trunk/lib/Core/TimingSolver.h Wed May 20 23:36:41 2009
@@ -0,0 +1,70 @@
+//===-- TimingSolver.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_TIMINGSOLVER_H
+#define KLEE_TIMINGSOLVER_H
+
+#include "klee/Expr.h"
+#include "klee/Solver.h"
+
+#include <vector>
+
+namespace klee {
+  class ExecutionState;
+  class Solver;
+  class STPSolver;
+
+  /// TimingSolver - A simple class which wraps a solver and handles
+  /// tracking the statistics that we care about.
+  class TimingSolver {
+  public:
+    Solver *solver;
+    STPSolver *stpSolver;
+    bool simplifyExprs;
+
+  public:
+    /// TimingSolver - Construct a new timing solver.
+    ///
+    /// \param _simplifyExprs - Whether expressions should be
+    /// simplified (via the constraint manager interface) prior to
+    /// querying.
+    TimingSolver(Solver *_solver, STPSolver *_stpSolver, 
+                 bool _simplifyExprs = true) 
+      : solver(_solver), stpSolver(_stpSolver), simplifyExprs(_simplifyExprs) {}
+    ~TimingSolver() {
+      delete solver;
+    }
+
+    void setTimeout(double t) {
+      stpSolver->setTimeout(t);
+    }
+
+    bool evaluate(const ExecutionState&, ref<Expr>, Solver::Validity &result);
+
+    bool mustBeTrue(const ExecutionState&, ref<Expr>, bool &result);
+
+    bool mustBeFalse(const ExecutionState&, ref<Expr>, bool &result);
+
+    bool mayBeTrue(const ExecutionState&, ref<Expr>, bool &result);
+
+    bool mayBeFalse(const ExecutionState&, ref<Expr>, bool &result);
+
+    bool getValue(const ExecutionState &, ref<Expr> expr, ref<Expr> &result);
+
+    bool getInitialValues(const ExecutionState&, 
+                          const std::vector<const Array*> &objects,
+                          std::vector< std::vector<unsigned char> > &result);
+
+    virtual std::pair< ref<Expr>, ref<Expr> >
+    getRange(const ExecutionState&, ref<Expr> query);
+  };
+
+}
+
+#endif

Added: klee/trunk/lib/Core/UserSearcher.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/UserSearcher.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/UserSearcher.cpp (added)
+++ klee/trunk/lib/Core/UserSearcher.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,175 @@
+//===-- UserSearcher.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Common.h"
+
+#include "UserSearcher.h"
+
+#include "Searcher.h"
+#include "Executor.h"
+
+#include "llvm/Support/CommandLine.h"
+
+using namespace llvm;
+using namespace klee;
+
+namespace {
+  cl::opt<bool>
+  UseRandomSearch("use-random-search");
+
+  cl::opt<bool>
+  UseInterleavedRS("use-interleaved-RS");
+
+  cl::opt<bool>
+  UseInterleavedNURS("use-interleaved-NURS");
+
+  cl::opt<bool>
+  UseInterleavedMD2UNURS("use-interleaved-MD2U-NURS");
+
+  cl::opt<bool>
+  UseInterleavedInstCountNURS("use-interleaved-icnt-NURS");
+
+  cl::opt<bool>
+  UseInterleavedCPInstCountNURS("use-interleaved-cpicnt-NURS");
+
+  cl::opt<bool>
+  UseInterleavedQueryCostNURS("use-interleaved-query-cost-NURS");
+
+  cl::opt<bool>
+  UseInterleavedCovNewNURS("use-interleaved-covnew-NURS");
+
+  cl::opt<bool>
+  UseNonUniformRandomSearch("use-non-uniform-random-search");
+
+  cl::opt<bool>
+  UseRandomPathSearch("use-random-path");
+
+  cl::opt<WeightedRandomSearcher::WeightType>
+  WeightType("weight-type", cl::desc("Set the weight type for --use-non-uniform-random-search"),
+             cl::values(clEnumValN(WeightedRandomSearcher::Depth, "none", "use (2^depth)"),
+                        clEnumValN(WeightedRandomSearcher::InstCount, "icnt", "use current pc exec count"),
+                        clEnumValN(WeightedRandomSearcher::CPInstCount, "cpicnt", "use current pc exec count"),
+                        clEnumValN(WeightedRandomSearcher::QueryCost, "query-cost", "use query cost"),
+                        clEnumValN(WeightedRandomSearcher::MinDistToUncovered, "md2u", "use min dist to uncovered"),
+                        clEnumValN(WeightedRandomSearcher::CoveringNew, "covnew", "use min dist to uncovered + coveringNew flag"),
+                        clEnumValEnd));
+  
+  cl::opt<bool>
+  UseMerge("use-merge", 
+           cl::desc("Enable support for klee_merge() (experimental)"));
+ 
+  cl::opt<bool>
+  UseBumpMerge("use-bump-merge", 
+           cl::desc("Enable support for klee_merge() (extra experimental)"));
+ 
+  cl::opt<bool>
+  UseIterativeDeepeningTimeSearch("use-iterative-deepening-time-search", 
+                                    cl::desc("(experimental)"));
+
+  cl::opt<bool>
+  UseBatchingSearch("use-batching-search", 
+           cl::desc("Use batching searcher (keep running selected state for N instructions/time, see --batch-instructions and --batch-time"));
+
+  cl::opt<unsigned>
+  BatchInstructions("batch-instructions",
+                    cl::desc("Number of instructions to batch when using --use-batching-search"),
+                    cl::init(10000));
+  
+  cl::opt<double>
+  BatchTime("batch-time",
+            cl::desc("Amount of time to batch when using --use-batching-search"),
+            cl::init(5.0));
+}
+
+bool klee::userSearcherRequiresMD2U() {
+  return (WeightType==WeightedRandomSearcher::MinDistToUncovered ||
+          WeightType==WeightedRandomSearcher::CoveringNew ||
+          UseInterleavedMD2UNURS ||
+          UseInterleavedCovNewNURS || 
+          UseInterleavedInstCountNURS || 
+          UseInterleavedCPInstCountNURS || 
+          UseInterleavedQueryCostNURS);
+}
+
+// FIXME: Remove.
+bool klee::userSearcherRequiresBranchSequences() {
+  return false;
+}
+
+Searcher *klee::constructUserSearcher(Executor &executor) {
+  Searcher *searcher = 0;
+
+  if (UseRandomPathSearch) {
+    searcher = new RandomPathSearcher(executor);
+  } else if (UseNonUniformRandomSearch) {
+    searcher = new WeightedRandomSearcher(executor, WeightType);
+  } else if (UseRandomSearch) {
+    searcher = new RandomSearcher();
+  } else {
+    searcher = new DFSSearcher();
+  }
+
+  if (UseInterleavedNURS || UseInterleavedMD2UNURS || UseInterleavedRS ||
+      UseInterleavedCovNewNURS || UseInterleavedInstCountNURS ||
+      UseInterleavedCPInstCountNURS || UseInterleavedQueryCostNURS) {
+    std::vector<Searcher *> s;
+    s.push_back(searcher);
+    
+    if (UseInterleavedNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::Depth));
+    if (UseInterleavedMD2UNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::MinDistToUncovered));
+
+    if (UseInterleavedCovNewNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::CoveringNew));
+   
+    if (UseInterleavedInstCountNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::InstCount));
+    
+    if (UseInterleavedCPInstCountNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::CPInstCount));
+    
+    if (UseInterleavedQueryCostNURS)
+      s.push_back(new WeightedRandomSearcher(executor, 
+                                             WeightedRandomSearcher::QueryCost));
+
+    if (UseInterleavedRS) 
+      s.push_back(new RandomSearcher());
+
+    searcher = new InterleavedSearcher(s);
+  }
+
+  if (UseBatchingSearch) {
+    searcher = new BatchingSearcher(searcher, BatchTime, BatchInstructions);
+  }
+
+  if (UseMerge) {
+    assert(!UseBumpMerge);
+    searcher = new MergingSearcher(executor, searcher);
+  } else if (UseBumpMerge) {    
+    searcher = new BumpMergingSearcher(executor, searcher);
+  }
+  
+  if (UseIterativeDeepeningTimeSearch) {
+    searcher = new IterativeDeepeningTimeSearcher(searcher);
+  }
+
+  std::ostream &os = executor.getHandler().getInfoStream();
+
+  os << "BEGIN searcher description\n";
+  searcher->printName(os);
+  os << "END searcher description\n";
+
+  return searcher;
+}

Added: klee/trunk/lib/Core/UserSearcher.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Core/UserSearcher.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Core/UserSearcher.h (added)
+++ klee/trunk/lib/Core/UserSearcher.h Wed May 20 23:36:41 2009
@@ -0,0 +1,25 @@
+//===-- UserSearcher.h ------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_USERSEARCHER_H
+#define KLEE_USERSEARCHER_H
+
+namespace klee {
+  class Executor;
+  class Searcher;
+
+  // XXX gross, should be on demand?
+  bool userSearcherRequiresMD2U();
+
+  bool userSearcherRequiresBranchSequences();
+
+  Searcher *constructUserSearcher(Executor &executor);
+}
+
+#endif

Added: klee/trunk/lib/Expr/Constraints.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Constraints.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Constraints.cpp (added)
+++ klee/trunk/lib/Expr/Constraints.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,155 @@
+//===-- Constraints.cpp ---------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Constraints.h"
+
+#include "klee/util/ExprPPrinter.h"
+#include "klee/util/ExprVisitor.h"
+
+#include <iostream>
+#include <map>
+
+using namespace klee;
+
+class ExprReplaceVisitor : public ExprVisitor {
+private:
+  ref<Expr> src, dst;
+
+public:
+  ExprReplaceVisitor(ref<Expr> _src, ref<Expr> _dst) : src(_src), dst(_dst) {}
+
+  Action visitExpr(const Expr &e) {
+    if (e == *src.get()) {
+      return Action::changeTo(dst);
+    } else {
+      return Action::doChildren();
+    }
+  }
+
+  Action visitExprPost(const Expr &e) {
+    if (e == *src.get()) {
+      return Action::changeTo(dst);
+    } else {
+      return Action::doChildren();
+    }
+  }
+};
+
+class ExprReplaceVisitor2 : public ExprVisitor {
+private:
+  const std::map< ref<Expr>, ref<Expr> > &replacements;
+
+public:
+  ExprReplaceVisitor2(const std::map< ref<Expr>, ref<Expr> > &_replacements) 
+    : ExprVisitor(true),
+      replacements(_replacements) {}
+
+  Action visitExprPost(const Expr &e) {
+    std::map< ref<Expr>, ref<Expr> >::const_iterator it =
+      replacements.find(ref<Expr>((Expr*) &e));
+    if (it!=replacements.end()) {
+      return Action::changeTo(it->second);
+    } else {
+      return Action::doChildren();
+    }
+  }
+};
+
+bool ConstraintManager::rewriteConstraints(ExprVisitor &visitor) {
+  ConstraintManager::constraints_ty old;
+  bool changed = false;
+
+  constraints.swap(old);
+  for (ConstraintManager::constraints_ty::iterator 
+         it = old.begin(), ie = old.end(); it != ie; ++it) {
+    ref<Expr> &ce = *it;
+    ref<Expr> e = visitor.visit(ce);
+
+    if (e!=ce) {
+      addConstraintInternal(e); // enable further reductions
+      changed = true;
+    } else {
+      constraints.push_back(ce);
+    }
+  }
+
+  return changed;
+}
+
+void ConstraintManager::simplifyForValidConstraint(ref<Expr> e) {
+  // XXX 
+}
+
+ref<Expr> ConstraintManager::simplifyExpr(ref<Expr> e) const {
+  if (e.isConstant())
+    return e;
+
+  std::map< ref<Expr>, ref<Expr> > equalities;
+  
+  for (ConstraintManager::constraints_ty::const_iterator 
+         it = constraints.begin(), ie = constraints.end(); it != ie; ++it) {
+    if (const EqExpr *ee = dyn_ref_cast<EqExpr>(*it)) {
+      if (ee->left.isConstant()) {
+        equalities.insert(std::make_pair(ee->right,
+                                         ee->left));
+      } else {
+        equalities.insert(std::make_pair(*it,
+                                         ref<Expr>(1,Expr::Bool)));
+      }
+    } else {
+      equalities.insert(std::make_pair(*it,
+                                       ref<Expr>(1,Expr::Bool)));
+    }
+  }
+
+  return ExprReplaceVisitor2(equalities).visit(e);
+}
+
+void ConstraintManager::addConstraintInternal(ref<Expr> e) {
+  // rewrite any known equalities 
+
+  // XXX should profile the effects of this and the overhead.
+  // traversing the constraints looking for equalities is hardly the
+  // slowest thing we do, but it is probably nicer to have a
+  // ConstraintSet ADT which efficiently remembers obvious patterns
+  // (byte-constant comparison).
+
+  switch (e.getKind()) {
+  case Expr::Constant:
+    assert(e.getConstantValue() && "attempt to add invalid (false) constraint");
+    break;
+    
+    // split to enable finer grained independence and other optimizations
+  case Expr::And: {
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    addConstraintInternal(be->left);
+    addConstraintInternal(be->right);
+    break;
+  }
+
+  case Expr::Eq: {
+    BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+    if (be->left.isConstant()) {
+      ExprReplaceVisitor visitor(be->right, be->left);
+      rewriteConstraints(visitor);
+    }
+    constraints.push_back(e);
+    break;
+  }
+    
+  default:
+    constraints.push_back(e);
+    break;
+  }
+}
+
+void ConstraintManager::addConstraint(ref<Expr> e) {
+  e = simplifyExpr(e);
+  addConstraintInternal(e);
+}




From daniel at zuster.org  Wed May 20 23:36:47 2009
From: daniel at zuster.org (Daniel Dunbar)
Date: Thu, 21 May 2009 04:36:47 -0000
Subject: [llvm-commits] [klee] r72205 [2/3] - in /klee/trunk: ./ autoconf/
 docs/ docs/SMT-COMP/ examples/ examples/regexp/ examples/sort/ include/
 include/expr/ include/klee/ include/klee/Config/ include/klee/Internal/
 include/klee/Internal/ADT/ include/klee/Internal/Module/
 include/klee/Internal/Support/ include/klee/Internal/System/
 include/klee/util/ lib/ lib/Basic/ lib/Core/ lib/Expr/ lib/Module/
 lib/Solver/ lib/Support/ runtime/ runtime/Intrinsic/ runtime/POSIX/
 runtime/POSIX/testing-dir/ runtime/POSIX/testing-dir/e/ runti...
Message-ID: <200905210437.n4L4b1av006294@zion.cs.uiuc.edu>

Added: klee/trunk/lib/Expr/Expr.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Expr.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Expr.cpp (added)
+++ klee/trunk/lib/Expr/Expr.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,1122 @@
+//===-- Expr.cpp ----------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Expr.h"
+
+
+#include "klee/Machine.h"
+// FIXME: This shouldn't be here.
+//#include "klee/Memory.h"
+#include "llvm/Type.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Streams.h"
+// FIXME: We shouldn't need this once fast constant support moves into
+// Core. If we need to do arithmetic, we probably want to use APInt.
+#include "klee/Internal/Support/IntEvaluation.h"
+
+#include "klee/util/ExprPPrinter.h"
+
+using namespace klee;
+using namespace llvm;
+
+namespace {
+  cl::opt<bool>
+  ConstArrayOpt("const-array-opt",
+	 cl::init(false),
+	 cl::desc("Enable various optimizations involving all-constant arrays."));
+}
+
+/***/
+
+unsigned Expr::count = 0;
+
+ref<Expr> Expr::createTempRead(const Array *array, Expr::Width w) {
+  UpdateList ul(array, true, 0);
+
+  switch (w) {
+  case Expr::Bool: 
+    return ZExtExpr::create(ReadExpr::create(ul, 
+                                             ref<Expr>(0,kMachinePointerType)),
+                            Expr::Bool);
+  case Expr::Int8: 
+    return ReadExpr::create(ul, 
+                            ref<Expr>(0,kMachinePointerType));
+  case Expr::Int16: 
+    return ConcatExpr::create(ReadExpr::create(ul, 
+                                               ref<Expr>(1,kMachinePointerType)),
+                              ReadExpr::create(ul, 
+                                               ref<Expr>(0,kMachinePointerType)));
+  case Expr::Int32: 
+    return ConcatExpr::create4(ReadExpr::create(ul, 
+                                                ref<Expr>(3,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(2,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(1,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(0,kMachinePointerType)));
+  case Expr::Int64: 
+    return ConcatExpr::create8(ReadExpr::create(ul, 
+                                                ref<Expr>(7,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(6,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(5,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(4,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(3,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(2,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(1,kMachinePointerType)),
+                               ReadExpr::create(ul, 
+                                                ref<Expr>(0,kMachinePointerType)));
+  default: assert(0 && "invalid width");
+  }
+}
+
+// returns 0 if b is structurally equal to *this
+int Expr::compare(const Expr &b) const {
+  if (this == &b) return 0;
+
+  Kind ak = getKind(), bk = b.getKind();
+  if (ak!=bk)
+    return (ak < bk) ? -1 : 1;
+
+  if (hashValue != b.hashValue) 
+    return (hashValue < b.hashValue) ? -1 : 1;
+
+  if (int res = compareContents(b)) 
+    return res;
+
+  unsigned aN = getNumKids();
+  for (unsigned i=0; i<aN; i++)
+    if (int res = getKid(i).compare(b.getKid(i)))
+      return res;
+
+  return 0;
+}
+
+void Expr::printKind(std::ostream &os, Kind k) {
+  switch(k) {
+#define X(C) case C: os << #C; break
+    X(Constant);
+    X(NotOptimized);
+    X(Read);
+    X(Select);
+    X(Concat);
+    X(Extract);
+    X(ZExt);
+    X(SExt);
+    X(Add);
+    X(Sub);
+    X(Mul);
+    X(UDiv);
+    X(SDiv);
+    X(URem);
+    X(SRem);
+    X(And);
+    X(Or);
+    X(Xor);
+    X(Shl);
+    X(LShr);
+    X(AShr);
+    X(Eq);
+    X(Ne);
+    X(Ult);
+    X(Ule);
+    X(Ugt);
+    X(Uge);
+    X(Slt);
+    X(Sle);
+    X(Sgt);
+    X(Sge);
+#undef X
+  default:
+    assert(0 && "invalid kind");
+    }
+}
+
+////////
+//
+// Simple hash functions for various kinds of Exprs
+//
+///////
+
+unsigned Expr::computeHash() {
+  unsigned res = getKind() * Expr::MAGIC_HASH_CONSTANT;
+
+  int n = getNumKids();
+  for (int i = 0; i < n; i++) {
+    res <<= 1;
+    res ^= getKid(i).hash() * Expr::MAGIC_HASH_CONSTANT;
+  }
+  
+  hashValue = res;
+  return hashValue;
+}
+
+unsigned ConstantExpr::computeHash() {
+  hashValue = Expr::hashConstant(asUInt64, width);
+  return hashValue;
+}
+
+unsigned CastExpr::computeHash() {
+  unsigned res = getWidth() * Expr::MAGIC_HASH_CONSTANT;
+  hashValue = res ^ src.hash() * Expr::MAGIC_HASH_CONSTANT;
+  return hashValue;
+}
+
+unsigned ExtractExpr::computeHash() {
+  unsigned res = offset * Expr::MAGIC_HASH_CONSTANT;
+  res ^= getWidth() * Expr::MAGIC_HASH_CONSTANT;
+  hashValue = res ^ expr.hash() * Expr::MAGIC_HASH_CONSTANT;
+  return hashValue;
+}
+
+unsigned ReadExpr::computeHash() {
+  unsigned res = index.hash() * Expr::MAGIC_HASH_CONSTANT;
+  res ^= updates.hash();
+  hashValue = res;
+  return hashValue;
+}
+
+uint64_t Expr::getConstantValue() const {
+  assert(getKind() == Constant);
+  return static_cast<const ConstantExpr*>(this)->asUInt64;
+}
+
+ref<Expr> Expr::createFromKind(Kind k, std::vector<CreateArg> args) {
+  unsigned numArgs = args.size();
+  
+  switch(k) {
+    case NotOptimized:
+      assert(numArgs == 1 && args[0].isExpr() &&
+             "invalid args array for given opcode");
+      return NotOptimizedExpr::create(args[0].expr);
+      
+    case Select:
+      assert(numArgs == 3 && args[0].isExpr() &&
+             args[1].isExpr() && args[2].isExpr() &&
+             "invalid args array for Select opcode");
+      return SelectExpr::create(args[0].expr,
+                                args[1].expr,
+                                args[2].expr);
+
+    case Concat: {
+      assert(numArgs == 2 && args[0].isExpr() && args[1].isExpr() && 
+             "invalid args array for Concat opcode");
+      
+      return ConcatExpr::create(args[0].expr, args[1].expr);
+    }
+      
+#define CAST_EXPR_CASE(T)                                    \
+      case T:                                                \
+        assert(numArgs == 2 &&				     \
+               args[0].isExpr() && args[1].isWidth() &&      \
+               "invalid args array for given opcode");       \
+      return T ## Expr::create(args[0].expr, args[1].width); \
+      
+#define BINARY_EXPR_CASE(T)                                 \
+      case T:                                               \
+        assert(numArgs == 2 &&                              \
+               args[0].isExpr() && args[1].isExpr() &&      \
+               "invalid args array for given opcode");      \
+      return T ## Expr::create(args[0].expr, args[1].expr); \
+
+      CAST_EXPR_CASE(ZExt);
+      CAST_EXPR_CASE(SExt);
+      
+      BINARY_EXPR_CASE(Add);
+      BINARY_EXPR_CASE(Sub);
+      BINARY_EXPR_CASE(Mul);
+      BINARY_EXPR_CASE(UDiv);
+      BINARY_EXPR_CASE(SDiv);
+      BINARY_EXPR_CASE(URem);
+      BINARY_EXPR_CASE(SRem);
+      BINARY_EXPR_CASE(And);
+      BINARY_EXPR_CASE(Or);
+      BINARY_EXPR_CASE(Xor);
+      BINARY_EXPR_CASE(Shl);
+      BINARY_EXPR_CASE(LShr);
+      BINARY_EXPR_CASE(AShr);
+      
+      BINARY_EXPR_CASE(Eq);
+      BINARY_EXPR_CASE(Ne);
+      BINARY_EXPR_CASE(Ult);
+      BINARY_EXPR_CASE(Ule);
+      BINARY_EXPR_CASE(Ugt);
+      BINARY_EXPR_CASE(Uge);
+      BINARY_EXPR_CASE(Slt);
+      BINARY_EXPR_CASE(Sle);
+      BINARY_EXPR_CASE(Sgt);
+      BINARY_EXPR_CASE(Sge);
+
+    case Constant:
+    case Extract:
+    case Read:
+    default:
+      assert(0 && "invalid kind");
+  }
+
+}
+
+
+void Expr::printWidth(std::ostream &os, Width width) {
+  switch(width) {
+  case Expr::Bool: os << "Expr::Bool"; break;
+  case Expr::Int8: os << "Expr::Int8"; break;
+  case Expr::Int16: os << "Expr::Int16"; break;
+  case Expr::Int32: os << "Expr::Int32"; break;
+  case Expr::Int64: os << "Expr::Int64"; break;
+  default: os << "<invalid type: " << (unsigned) width << ">";
+  }
+}
+
+Expr::Width Expr::getWidthForLLVMType(const llvm::Type *t) {
+  switch (t->getTypeID()) {
+  case llvm::Type::IntegerTyID: {
+    Width w = cast<IntegerType>(t)->getBitWidth();
+
+    // should remove this limitation soon
+    if (w == 1 || w == 8 || w == 16 || w == 32 || w == 64)
+      return w;
+    else {
+      assert(0 && "XXX arbitrary bit widths unsupported");
+      abort();
+    }
+  }
+  case llvm::Type::FloatTyID: return Expr::Int32;
+  case llvm::Type::DoubleTyID: return Expr::Int64;
+  case llvm::Type::X86_FP80TyID: return Expr::Int64; // XXX: needs to be fixed
+  case llvm::Type::PointerTyID: return kMachinePointerType;
+  default:
+    cerr << "non-primitive type argument to Expr::getTypeForLLVMType()\n";
+    abort();
+  }
+}
+
+ref<Expr> Expr::createImplies(ref<Expr> hyp, ref<Expr> conc) {
+  return OrExpr::create(Expr::createNot(hyp), conc);
+}
+
+ref<Expr> Expr::createIsZero(ref<Expr> e) {
+  return EqExpr::create(e, ConstantExpr::create(0, e.getWidth()));
+}
+
+ref<Expr> Expr::createCoerceToPointerType(ref<Expr> e) {
+  return ZExtExpr::create(e, kMachinePointerType);
+}
+
+ref<Expr> Expr::createNot(ref<Expr> e) {
+  return createIsZero(e);
+}
+
+ref<Expr> Expr::createPointer(uint64_t v) {
+  return ConstantExpr::create(v, kMachinePointerType);
+}
+
+Expr* Expr::createConstant(uint64_t val, Width w) {
+  Expr *r = new ConstantExpr(val, w);
+  r->computeHash();
+  return r;
+}
+
+void Expr::print(std::ostream &os) const {
+  const ref<Expr> tmp((Expr*)this);
+  ExprPPrinter::printOne(os, "", tmp);
+}
+
+/***/
+
+ref<ConstantExpr> ConstantExpr::fromMemory(void *address, Width width) {
+  switch (width) {
+  case  Expr::Bool: return ConstantExpr::create(*(( uint8_t*) address), width);
+  case  Expr::Int8: return ConstantExpr::create(*(( uint8_t*) address), width);
+  case Expr::Int16: return ConstantExpr::create(*((uint16_t*) address), width);
+  case Expr::Int32: return ConstantExpr::create(*((uint32_t*) address), width);
+  case Expr::Int64: return ConstantExpr::create(*((uint64_t*) address), width);
+  default: assert(0 && "invalid type");
+  }
+}
+
+void ConstantExpr::toMemory(void *address) {
+  switch (width) {
+  case  Expr::Bool: *(( uint8_t*) address) = asUInt64; break;
+  case  Expr::Int8: *(( uint8_t*) address) = asUInt64; break;
+  case Expr::Int16: *((uint16_t*) address) = asUInt64; break;
+  case Expr::Int32: *((uint32_t*) address) = asUInt64; break;
+  case Expr::Int64: *((uint64_t*) address) = asUInt64; break;
+  default: assert(0 && "invalid type");
+  }
+}
+
+/***/
+
+ref<Expr>  NotOptimizedExpr::create(ref<Expr> src) {
+  return NotOptimizedExpr::alloc(src);
+}
+
+ref<Expr> ReadExpr::create(const UpdateList &ul, ref<Expr> index) {
+  // rollback index when possible... 
+
+  // XXX this doesn't really belong here... there are basically two
+  // cases, one is rebuild, where we want to optimistically try various
+  // optimizations when the index has changed, and the other is 
+  // initial creation, where we expect the ObjectState to have constructed
+  // a smart UpdateList so it is not worth rescanning.
+
+  const UpdateNode *un = ul.head;
+  for (; un; un=un->next) {
+    ref<Expr> cond = EqExpr::create(index, un->index);
+    
+    if (cond.isConstant()) {
+      if (cond.getConstantValue())
+        return un->value;
+    } else {
+      break;
+    }
+  }
+
+  return ReadExpr::alloc(ul, index);
+}
+
+int ReadExpr::compareContents(const Expr &b) const { 
+  return updates.compare(static_cast<const ReadExpr&>(b).updates);
+}
+
+ref<Expr> SelectExpr::create(ref<Expr> c, ref<Expr> t, ref<Expr> f) {
+  Expr::Width kt = t.getWidth();
+
+  assert(c.getWidth()==Bool && "type mismatch");
+  assert(kt==f.getWidth() && "type mismatch");
+
+  if (c.isConstant()) {
+    return c.getConstantValue() ? t : f;
+  } else if (t==f) {
+    return t;
+  } else if (kt==Expr::Bool) { // c ? t : f  <=> (c and t) or (not c and f)
+    if (t.isConstant()) {      
+      if (t.getConstantValue()) {
+        return OrExpr::create(c, f);
+      } else {
+        return AndExpr::create(Expr::createNot(c), f);
+      }
+    } else if (f.isConstant()) {
+      if (f.getConstantValue()) {
+        return OrExpr::create(Expr::createNot(c), t);
+      } else {
+        return AndExpr::create(c, t);
+      }
+    }
+  }
+  
+  return SelectExpr::alloc(c, t, f);
+}
+
+/***/
+
+
+ref<Expr> ConcatExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  Expr::Width w = l.getWidth() + r.getWidth();
+  
+  /* Constant folding */
+  if (l.getKind() == Expr::Constant && r.getKind() == Expr::Constant) {
+    // XXX: should fix this constant limitation soon
+    assert(w <= 64 && "ConcatExpr::create(): don't support concats describing constants greater than 64 bits yet");
+    
+    uint64_t res = (l.getConstantValue() << r.getWidth()) + r.getConstantValue();
+    return ConstantExpr::create(res, w);
+  }
+
+  // Merge contiguous Extracts
+  if (l.getKind() == Expr::Extract && r.getKind() == Expr::Extract) {
+    const ExtractExpr* ee_left = static_ref_cast<ExtractExpr>(l);
+    const ExtractExpr* ee_right = static_ref_cast<ExtractExpr>(r);
+    if (ee_left->expr == ee_right->expr &&
+	ee_right->offset + ee_right->width == ee_left->offset) {
+      return ExtractExpr::create(ee_left->expr, ee_right->offset, w);
+    }
+  }
+
+  return ConcatExpr::alloc(l, r);
+}
+
+/// Shortcut to concat N kids.  The chain returned is unbalanced to the right
+ref<Expr> ConcatExpr::createN(unsigned n_kids, const ref<Expr> kids[]) {
+  assert(n_kids > 0);
+  if (n_kids == 1)
+    return kids[0];
+  
+  ref<Expr> r = ConcatExpr::create(kids[n_kids-2], kids[n_kids-1]);
+  for (int i=n_kids-3; i>=0; i--)
+    r = ConcatExpr::create(kids[i], r);
+  return r;
+}
+
+/// Shortcut to concat 4 kids.  The chain returned is unbalanced to the right
+ref<Expr> ConcatExpr::create4(const ref<Expr> &kid1, const ref<Expr> &kid2,
+				     const ref<Expr> &kid3, const ref<Expr> &kid4) {
+  return ConcatExpr::create(kid1, ConcatExpr::create(kid2, ConcatExpr::create(kid3, kid4)));
+}
+
+/// Shortcut to concat 8 kids.  The chain returned is unbalanced to the right
+ref<Expr> ConcatExpr::create8(const ref<Expr> &kid1, const ref<Expr> &kid2,
+			      const ref<Expr> &kid3, const ref<Expr> &kid4,
+			      const ref<Expr> &kid5, const ref<Expr> &kid6,
+			      const ref<Expr> &kid7, const ref<Expr> &kid8) {
+  return ConcatExpr::create(kid1, ConcatExpr::create(kid2, ConcatExpr::create(kid3, 
+			      ConcatExpr::create(kid4, ConcatExpr::create4(kid5, kid6, kid7, kid8)))));
+}
+
+/***/
+
+ref<Expr> ExtractExpr::create(ref<Expr> expr, unsigned off, Width w) {
+  unsigned kw = expr.getWidth();
+  assert(w > 0 && off + w <= kw && "invalid extract");
+  
+  if (w == kw)
+    return expr;
+  else if (expr.isConstant()) {
+    return ConstantExpr::create(ints::trunc(expr.getConstantValue() >> off, w, kw), w);
+  } 
+  else 
+    // Extract(Concat)
+    if (ConcatExpr *ce = dyn_ref_cast<ConcatExpr>(expr)) {
+      // if the extract skips the right side of the concat
+      if (off >= ce->getRight().getWidth())
+	return ExtractExpr::create(ce->getLeft(), off - ce->getRight().getWidth(), w);
+      
+      // if the extract skips the left side of the concat
+      if (off + w <= ce->getRight().getWidth())
+	return ExtractExpr::create(ce->getRight(), off, w);
+
+      // E(C(x,y)) = C(E(x), E(y))
+      return ConcatExpr::create(ExtractExpr::create(ce->getKid(0), 0, w - ce->getKid(1).getWidth() + off),
+				ExtractExpr::create(ce->getKid(1), off, ce->getKid(1).getWidth() - off));
+    }
+  
+  return ExtractExpr::alloc(expr, off, w);
+}
+
+
+ref<Expr> ExtractExpr::createByteOff(ref<Expr> expr, unsigned offset, Width bits) {
+  return ExtractExpr::create(expr, 8*offset, bits);
+}
+
+/***/
+
+ref<Expr> ZExtExpr::create(const ref<Expr> &e, Width w) {
+  unsigned kBits = e.getWidth();
+  if (w == kBits) {
+    return e;
+  } else if (w < kBits) { // trunc
+    return ExtractExpr::createByteOff(e, 0, w);
+  } else {
+    if (e.isConstant()) {
+      return ConstantExpr::create(ints::zext(e.getConstantValue(), w, kBits),
+                                  w);
+    }
+    
+    return ZExtExpr::alloc(e, w);
+  }
+}
+
+ref<Expr> SExtExpr::create(const ref<Expr> &e, Width w) {
+  unsigned kBits = e.getWidth();
+  if (w == kBits) {
+    return e;
+  } else if (w < kBits) { // trunc
+    return ExtractExpr::createByteOff(e, 0, w);
+  } else {
+    if (e.isConstant()) {
+      return ConstantExpr::create(ints::sext(e.getConstantValue(), w, kBits),
+                                  w);
+    }
+    
+    return SExtExpr::alloc(e, w);
+  }
+}
+
+/***/
+
+static ref<Expr> AndExpr_create(Expr *l, Expr *r);
+static ref<Expr> XorExpr_create(Expr *l, Expr *r);
+
+static ref<Expr> EqExpr_createPartial(Expr *l, const ref<Expr> &cr);
+static ref<Expr> AndExpr_createPartialR(const ref<Expr> &cl, Expr *r);
+static ref<Expr> SubExpr_createPartialR(const ref<Expr> &cl, Expr *r);
+static ref<Expr> XorExpr_createPartialR(const ref<Expr> &cl, Expr *r);
+
+static ref<Expr> AddExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  assert(cl.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cl.getConstantValue();
+  Expr::Width type = cl.getWidth();
+
+  if (type==Expr::Bool) {
+    return XorExpr_createPartialR(cl, r);
+  } else if (!value) {
+    return r;
+  } else {
+    Expr::Kind rk = r->getKind();
+    if (rk==Expr::Add && r->getKid(0).isConstant()) { // A + (B+c) == (A+B) + c
+      return AddExpr::create(AddExpr::create(cl, r->getKid(0)),
+                             r->getKid(1));
+    } else if (rk==Expr::Sub && r->getKid(0).isConstant()) { // A + (B-c) == (A+B) - c
+      return SubExpr::create(AddExpr::create(cl, r->getKid(0)),
+                             r->getKid(1));
+    } else {
+      return AddExpr::alloc(cl, r);
+    }
+  }
+}
+static ref<Expr> AddExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  return AddExpr_createPartialR(cr, l);
+}
+static ref<Expr> AddExpr_create(Expr *l, Expr *r) {
+  Expr::Width type = l->getWidth();
+
+  if (type == Expr::Bool) {
+    return XorExpr_create(l, r);
+  } else {
+    Expr::Kind lk = l->getKind(), rk = r->getKind();
+    if (lk==Expr::Add && l->getKid(0).isConstant()) { // (k+a)+b = k+(a+b)
+      return AddExpr::create(l->getKid(0),
+                             AddExpr::create(l->getKid(1), r));
+    } else if (lk==Expr::Sub && l->getKid(0).isConstant()) { // (k-a)+b = k+(b-a)
+      return AddExpr::create(l->getKid(0),
+                             SubExpr::create(r, l->getKid(1)));
+    } else if (rk==Expr::Add && r->getKid(0).isConstant()) { // a + (k+b) = k+(a+b)
+      return AddExpr::create(r->getKid(0),
+                             AddExpr::create(l, r->getKid(1)));
+    } else if (rk==Expr::Sub && r->getKid(0).isConstant()) { // a + (k-b) = k+(a-b)
+      return AddExpr::create(r->getKid(0),
+                             SubExpr::create(l, r->getKid(1)));
+    } else {
+      return AddExpr::alloc(l, r);
+    }
+  }  
+}
+
+static ref<Expr> SubExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  assert(cl.isConstant() && "non-constant passed in place of constant");
+  Expr::Width type = cl.getWidth();
+
+  if (type==Expr::Bool) {
+    return XorExpr_createPartialR(cl, r);
+  } else {
+    Expr::Kind rk = r->getKind();
+    if (rk==Expr::Add && r->getKid(0).isConstant()) { // A - (B+c) == (A-B) - c
+      return SubExpr::create(SubExpr::create(cl, r->getKid(0)),
+                             r->getKid(1));
+    } else if (rk==Expr::Sub && r->getKid(0).isConstant()) { // A - (B-c) == (A-B) + c
+      return AddExpr::create(SubExpr::create(cl, r->getKid(0)),
+                             r->getKid(1));
+    } else {
+      return SubExpr::alloc(cl, r);
+    }
+  }
+}
+static ref<Expr> SubExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  assert(cr.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cr.getConstantValue();
+  Expr::Width width = cr.getWidth();
+  uint64_t nvalue = ints::sub(0, value, width);
+
+  return AddExpr_createPartial(l, ConstantExpr::create(nvalue, width));
+}
+static ref<Expr> SubExpr_create(Expr *l, Expr *r) {
+  Expr::Width type = l->getWidth();
+
+  if (type == Expr::Bool) {
+    return XorExpr_create(l, r);
+  } else if (*l==*r) {
+    return ref<Expr>(0, type);
+  } else {
+    Expr::Kind lk = l->getKind(), rk = r->getKind();
+    if (lk==Expr::Add && l->getKid(0).isConstant()) { // (k+a)-b = k+(a-b)
+      return AddExpr::create(l->getKid(0),
+                             SubExpr::create(l->getKid(1), r));
+    } else if (lk==Expr::Sub && l->getKid(0).isConstant()) { // (k-a)-b = k-(a+b)
+      return SubExpr::create(l->getKid(0),
+                             AddExpr::create(l->getKid(1), r));
+    } else if (rk==Expr::Add && r->getKid(0).isConstant()) { // a - (k+b) = (a-c) - k
+      return SubExpr::create(SubExpr::create(l, r->getKid(1)),
+                             r->getKid(0));
+    } else if (rk==Expr::Sub && r->getKid(0).isConstant()) { // a - (k-b) = (a+b) - k
+      return SubExpr::create(AddExpr::create(l, r->getKid(1)),
+                             r->getKid(0));
+    } else {
+      return SubExpr::alloc(l, r);
+    }
+  }  
+}
+
+static ref<Expr> MulExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  assert(cl.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cl.getConstantValue();
+  Expr::Width type = cl.getWidth();
+
+  if (type == Expr::Bool) {
+    return AndExpr_createPartialR(cl, r);
+  } else if (value == 1) {
+    return r;
+  } else if (!value) {
+    return cl;
+  } else {
+    return MulExpr::alloc(cl, r);
+  }
+}
+static ref<Expr> MulExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  return MulExpr_createPartialR(cr, l);
+}
+static ref<Expr> MulExpr_create(Expr *l, Expr *r) {
+  Expr::Width type = l->getWidth();
+  
+  if (type == Expr::Bool) {
+    return AndExpr::alloc(l, r);
+  } else {
+    return MulExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> AndExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  assert(cr.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cr.getConstantValue();
+  Expr::Width width = cr.getWidth();;
+
+  if (value==ints::sext(1, width, 1)) {
+    return l;
+  } else if (!value) {
+    return cr;
+  } else {
+    return AndExpr::alloc(l, cr);
+  }
+}
+static ref<Expr> AndExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  return AndExpr_createPartial(r, cl);
+}
+static ref<Expr> AndExpr_create(Expr *l, Expr *r) {
+  return AndExpr::alloc(l, r);
+}
+
+static ref<Expr> OrExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  assert(cr.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cr.getConstantValue();
+  Expr::Width width = cr.getWidth();
+
+  if (value == ints::sext(1, width, 1)) {
+    return cr;
+  } else if (!value) {
+    return l;
+  } else {
+    return OrExpr::alloc(l, cr);
+  }
+}
+static ref<Expr> OrExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  return OrExpr_createPartial(r, cl);
+}
+static ref<Expr> OrExpr_create(Expr *l, Expr *r) {
+  return OrExpr::alloc(l, r);
+}
+
+static ref<Expr> XorExpr_createPartialR(const ref<Expr> &cl, Expr *r) {
+  assert(cl.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cl.getConstantValue();
+  Expr::Width type = cl.getWidth();
+
+  if (type==Expr::Bool) {
+    if (value) {
+      return EqExpr_createPartial(r, ConstantExpr::create(0, Expr::Bool));
+    } else {
+      return r;
+    }
+  } else if (!value) {
+    return r;
+  } else {
+    return XorExpr::alloc(cl, r);
+  }
+}
+
+static ref<Expr> XorExpr_createPartial(Expr *l, const ref<Expr> &cr) {
+  return XorExpr_createPartialR(cr, l);
+}
+static ref<Expr> XorExpr_create(Expr *l, Expr *r) {
+  return XorExpr::alloc(l, r);
+}
+
+static ref<Expr> UDivExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // r must be 1
+    return l;
+  } else{
+    return UDivExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> SDivExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // r must be 1
+    return l;
+  } else{
+    return SDivExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> URemExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // r must be 1
+    return ConstantExpr::create(0, Expr::Bool);
+  } else{
+    return URemExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> SRemExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // r must be 1
+    return ConstantExpr::create(0, Expr::Bool);
+  } else{
+    return SRemExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> ShlExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // l & !r
+    return AndExpr::create(l, Expr::createNot(r));
+  } else{
+    return ShlExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> LShrExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // l & !r
+    return AndExpr::create(l, Expr::createNot(r));
+  } else{
+    return LShrExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> AShrExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // l
+    return l;
+  } else{
+    return AShrExpr::alloc(l, r);
+  }
+}
+
+#define BCREATE_R(_e_op, _op, partialL, partialR) \
+ref<Expr>  _e_op ::create(const ref<Expr> &l, const ref<Expr> &r) { \
+  assert(l.getWidth()==r.getWidth() && "type mismatch"); \
+  if (l.isConstant()) {                                \
+    if (r.isConstant()) {                              \
+      Expr::Width width = l.getWidth(); \
+      uint64_t val = ints::_op(l.getConstantValue(),  \
+                               r.getConstantValue(), width);  \
+      return ConstantExpr::create(val, width); \
+    } else { \
+      return _e_op ## _createPartialR(l, r.get()); \
+    } \
+  } else if (r.isConstant()) {             \
+    return _e_op ## _createPartial(l.get(), r); \
+  } \
+  return _e_op ## _create(l.get(), r.get()); \
+}
+
+#define BCREATE(_e_op, _op) \
+ref<Expr>  _e_op ::create(const ref<Expr> &l, const ref<Expr> &r) { \
+  assert(l.getWidth()==r.getWidth() && "type mismatch"); \
+  if (l.isConstant()) {                                \
+    if (r.isConstant()) {                              \
+      Expr::Width width = l.getWidth(); \
+      uint64_t val = ints::_op(l.getConstantValue(), \
+                               r.getConstantValue(), width);  \
+      return ConstantExpr::create(val, width); \
+    } \
+  } \
+  return _e_op ## _create(l, r);                    \
+}
+
+BCREATE_R(AddExpr, add, AddExpr_createPartial, AddExpr_createPartialR)
+BCREATE_R(SubExpr, sub, SubExpr_createPartial, SubExpr_createPartialR)
+BCREATE_R(MulExpr, mul, MulExpr_createPartial, MulExpr_createPartialR)
+BCREATE_R(AndExpr, land, AndExpr_createPartial, AndExpr_createPartialR)
+BCREATE_R(OrExpr, lor, OrExpr_createPartial, OrExpr_createPartialR)
+BCREATE_R(XorExpr, lxor, XorExpr_createPartial, XorExpr_createPartialR)
+BCREATE(UDivExpr, udiv)
+BCREATE(SDivExpr, sdiv)
+BCREATE(URemExpr, urem)
+BCREATE(SRemExpr, srem)
+BCREATE(ShlExpr, shl)
+BCREATE(LShrExpr, lshr)
+BCREATE(AShrExpr, ashr)
+
+#define CMPCREATE(_e_op, _op) \
+ref<Expr>  _e_op ::create(const ref<Expr> &l, const ref<Expr> &r) { \
+  assert(l.getWidth()==r.getWidth() && "type mismatch"); \
+  if (l.isConstant()) {                                \
+    if (r.isConstant()) {                              \
+      Expr::Width width = l.getWidth(); \
+      uint64_t val = ints::_op(l.getConstantValue(), \
+                               r.getConstantValue(), width);  \
+      return ConstantExpr::create(val, Expr::Bool); \
+    } \
+  } \
+  return _e_op ## _create(l, r);                    \
+}
+
+#define CMPCREATE_T(_e_op, _op, _reflexive_e_op, partialL, partialR) \
+ref<Expr>  _e_op ::create(const ref<Expr> &l, const ref<Expr> &r) { \
+  assert(l.getWidth()==r.getWidth() && "type mismatch"); \
+  if (l.isConstant()) {                                \
+    if (r.isConstant()) {                              \
+      Expr::Width width = l.getWidth(); \
+      uint64_t val = ints::_op(l.getConstantValue(), \
+                               r.getConstantValue(), width);  \
+      return ConstantExpr::create(val, Expr::Bool); \
+    } else { \
+      return partialR(l, r.get()); \
+    } \
+  } else if (r.isConstant()) {                  \
+    return partialL(l.get(), r); \
+  } else { \
+    return _e_op ## _create(l.get(), r.get()); \
+  } \
+}
+  
+
+static ref<Expr> EqExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l == r) {
+    return ref<Expr>(1, Expr::Bool);
+  } else {
+    return EqExpr::alloc(l, r);
+  }
+}
+
+
+/// Tries to optimize EqExpr cl == rd, where cl is a ConstantExpr and
+/// rd a ReadExpr.  If rd is a read into an all-constant array,
+/// returns a disjunction of equalities on the index.  Otherwise,
+/// returns the initial equality expression. 
+static ref<Expr> TryConstArrayOpt(const ref<Expr> &cl, 
+				  ReadExpr *rd) {
+  assert(cl.isConstant() && "constant expression required");
+  assert(rd->getKind() == Expr::Read && "read expression required");
+  
+  uint64_t ct = cl.getConstantValue();
+  ref<Expr> first_idx_match;
+
+  // number of positions in the array that contain value ct
+  unsigned matches = 0;
+
+  //llvm::cerr << "Size updates/root: " << rd->updates.getSize() << " / " << (rd->updates.root)->size << "\n";
+
+  // for now, just assume standard "flushing" of a concrete array,
+  // where the concrete array has one update for each index, in order
+  bool all_const = true;
+  if (rd->updates.getSize() == rd->updates.root->size) {
+    unsigned k = rd->updates.getSize();
+    for (const UpdateNode *un = rd->updates.head; un; un = un->next) {
+      assert(k > 0);
+      k--;
+
+      ref<Expr> idx = un->index;
+      ref<Expr> val = un->value;
+      if (!idx.isConstant() || !val.isConstant()) {
+	all_const = false;
+	//llvm::cerr << "Idx or val not constant\n";
+	break;
+      }
+      else {
+	if (idx.getConstantValue() != k) {
+	  all_const = false;
+	  //llvm::cerr << "Wrong constant\n";
+	  break;
+	}
+	if (val.getConstantValue() == ct) {
+	  matches++;
+	  if (matches == 1)
+	    first_idx_match = un->index;
+	}
+      }
+    }
+  }
+  else all_const = false;
+  
+  if (all_const && matches <= 100) {
+    // apply optimization
+    //llvm::cerr << "\n\n=== Applying const array optimization ===\n\n";
+
+    if (matches == 0)
+      return ref<Expr>(0, Expr::Bool);
+
+    ref<Expr> res = EqExpr::create(first_idx_match, rd->index);
+    if (matches == 1)
+      return res;
+    
+    for (const UpdateNode *un = rd->updates.head; un; un = un->next) {
+      if (un->index != first_idx_match && un->value.getConstantValue() == ct) {
+	ref<Expr> curr_eq = EqExpr::create(un->index, rd->index);
+	res = OrExpr::create(curr_eq, res);
+      }
+    }
+    
+    return res;
+  }
+
+  return EqExpr_create(cl, ref<Expr>(rd));
+}
+
+
+static ref<Expr> EqExpr_createPartialR(const ref<Expr> &cl, Expr *r) {  
+  assert(cl.isConstant() && "non-constant passed in place of constant");
+  uint64_t value = cl.getConstantValue();
+  Expr::Width width = cl.getWidth();
+
+  Expr::Kind rk = r->getKind();
+  if (width == Expr::Bool) {
+    if (value) {
+      return r;
+    } else {
+      // 0 != ...
+      
+      if (rk == Expr::Eq) {
+        const EqExpr *ree = static_ref_cast<EqExpr>(r);
+
+        // eliminate double negation
+        if (ree->left.isConstant() &&
+            ree->left.getWidth()==Expr::Bool) {
+          assert(!ree->left.getConstantValue());
+          return ree->right;
+        }
+      } else if (rk == Expr::Or) {
+        const OrExpr *roe = static_ref_cast<OrExpr>(r);
+
+        // transform not(or(a,b)) to and(not a, not b)
+        return AndExpr::create(Expr::createNot(roe->left),
+                               Expr::createNot(roe->right));
+      }
+    }
+  } else if (rk == Expr::SExt) {
+    // (sext(a,T)==c) == (a==c)
+    const SExtExpr *see = static_ref_cast<SExtExpr>(r);
+    Expr::Width fromBits = see->src.getWidth();
+    uint64_t trunc = bits64::truncateToNBits(value, fromBits);
+
+    // pathological check, make sure it is possible to
+    // sext to this value *from any value*
+    if (value == ints::sext(trunc, width, fromBits)) {
+      return EqExpr::create(see->src, ConstantExpr::create(trunc, fromBits));
+    } else {
+      return ConstantExpr::create(0, Expr::Bool);
+    }
+  } else if (rk == Expr::ZExt) {
+    // (zext(a,T)==c) == (a==c)
+    const ZExtExpr *zee = static_ref_cast<ZExtExpr>(r);
+    Expr::Width fromBits = zee->src.getWidth();
+    uint64_t trunc = bits64::truncateToNBits(value, fromBits);
+    
+    // pathological check, make sure it is possible to
+    // zext to this value *from any value*
+    if (value == ints::zext(trunc, width, fromBits)) {
+      return EqExpr::create(zee->src, ConstantExpr::create(trunc, fromBits));
+    } else {
+      return ConstantExpr::create(0, Expr::Bool);
+    }
+  } else if (rk==Expr::Add) {
+    const AddExpr *ae = static_ref_cast<AddExpr>(r);
+    if (ae->left.isConstant()) {
+      // c0 = c1 + b => c0 - c1 = b
+      return EqExpr_createPartialR(SubExpr::create(cl, ae->left),
+                                   ae->right.get());
+    }
+  } else if (rk==Expr::Sub) {
+    const SubExpr *se = static_ref_cast<SubExpr>(r);
+    if (se->left.isConstant()) {
+      // c0 = c1 - b => c1 - c0 = b
+      return EqExpr_createPartialR(SubExpr::create(se->left, cl),
+                                   se->right.get());
+    }
+  } else if (rk == Expr::Read && ConstArrayOpt) {
+    return TryConstArrayOpt(cl, static_cast<ReadExpr*>(r));
+  }
+    
+  return EqExpr_create(cl, r);
+}
+
+static ref<Expr> EqExpr_createPartial(Expr *l, const ref<Expr> &cr) {  
+  return EqExpr_createPartialR(cr, l);
+}
+  
+ref<Expr> NeExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  return EqExpr::create(ConstantExpr::create(0, Expr::Bool),
+                        EqExpr::create(l, r));
+}
+
+ref<Expr> UgtExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  return UltExpr::create(r, l);
+}
+ref<Expr> UgeExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  return UleExpr::create(r, l);
+}
+
+ref<Expr> SgtExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  return SltExpr::create(r, l);
+}
+ref<Expr> SgeExpr::create(const ref<Expr> &l, const ref<Expr> &r) {
+  return SleExpr::create(r, l);
+}
+
+static ref<Expr> UltExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  Expr::Width t = l.getWidth();
+  if (t == Expr::Bool) { // !l && r
+    return AndExpr::create(Expr::createNot(l), r);
+  } else {
+    if (r.isConstant()) {      
+      uint64_t value = r.getConstantValue();
+      if (value <= 8) {
+        ref<Expr> res(0,Expr::Bool);
+        for (unsigned i=0; i<value; i++) {
+          res = OrExpr::create(EqExpr::create(l, ref<Expr>(i,t)), res);
+        }
+        //        llvm::cerr << l << "<" << r << "  <=>  " << res << "\n";
+        return res;
+      }
+    }
+    return UltExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> UleExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // !(l && !r)
+    return OrExpr::create(Expr::createNot(l), r);
+  } else {
+    return UleExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> SltExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // l && !r
+    return AndExpr::create(l, Expr::createNot(r));
+  } else {
+    return SltExpr::alloc(l, r);
+  }
+}
+
+static ref<Expr> SleExpr_create(const ref<Expr> &l, const ref<Expr> &r) {
+  if (l.getWidth() == Expr::Bool) { // !(!l && r)
+    return OrExpr::create(l, Expr::createNot(r));
+  } else {
+    return SleExpr::alloc(l, r);
+  }
+}
+
+CMPCREATE_T(EqExpr, eq, EqExpr, EqExpr_createPartial, EqExpr_createPartialR)
+CMPCREATE(UltExpr, ult)
+CMPCREATE(UleExpr, ule)
+CMPCREATE(SltExpr, slt)
+CMPCREATE(SleExpr, sle)

Added: klee/trunk/lib/Expr/ExprEvaluator.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/ExprEvaluator.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/ExprEvaluator.cpp (added)
+++ klee/trunk/lib/Expr/ExprEvaluator.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,74 @@
+//===-- ExprEvaluator.cpp -------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/util/ExprEvaluator.h"
+
+using namespace klee;
+
+ExprVisitor::Action ExprEvaluator::evalRead(const UpdateList &ul,
+                                            unsigned index) {
+  for (const UpdateNode *un=ul.head; un; un=un->next) {
+    ref<Expr> ui = visit(un->index);
+    
+    if (ui.isConstant()) {
+      if (ui.getConstantValue() == index)
+        return Action::changeTo(visit(un->value));
+    } else {
+      // update index is unknown, so may or may not be index, we
+      // cannot guarantee value. we can rewrite to read at this
+      // version though (mostly for debugging).
+      
+      UpdateList fwd(ul.root, un, 0);
+      return Action::changeTo(ReadExpr::create(fwd, 
+                                               ref<Expr>(index,Expr::Int32)));
+    }
+  }
+  
+  return Action::changeTo(getInitialValue(*ul.root, index));
+}
+
+ExprVisitor::Action ExprEvaluator::visitRead(const ReadExpr &re) {
+  ref<Expr> v = visit(re.index);
+  
+  if (v.isConstant()) {
+    return evalRead(re.updates, v.getConstantValue());
+  } else {
+    return Action::doChildren();
+  }
+}
+
+// we need to check for div by zero during partial evaluation,
+// if this occurs then simply ignore the 0 divisor and use the
+// original expression.
+ExprVisitor::Action ExprEvaluator::protectedDivOperation(const BinaryExpr &e) {
+  ref<Expr> kids[2] = { visit(e.left),
+                        visit(e.right) };
+
+  if (kids[1].isConstant() && !kids[1].getConstantValue())
+    kids[1] = e.right;
+
+  if (kids[0]!=e.left || kids[1]!=e.right) {
+    return Action::changeTo(e.rebuild(kids));
+  } else {
+    return Action::skipChildren();
+  }
+}
+
+ExprVisitor::Action ExprEvaluator::visitUDiv(const UDivExpr &e) { 
+  return protectedDivOperation(e); 
+}
+ExprVisitor::Action ExprEvaluator::visitSDiv(const SDivExpr &e) { 
+  return protectedDivOperation(e); 
+}
+ExprVisitor::Action ExprEvaluator::visitURem(const URemExpr &e) { 
+  return protectedDivOperation(e); 
+}
+ExprVisitor::Action ExprEvaluator::visitSRem(const SRemExpr &e) { 
+  return protectedDivOperation(e); 
+}

Added: klee/trunk/lib/Expr/ExprPPrinter.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/ExprPPrinter.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/ExprPPrinter.cpp (added)
+++ klee/trunk/lib/Expr/ExprPPrinter.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,478 @@
+//===-- ExprPPrinter.cpp -   ----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/util/ExprPPrinter.h"
+
+#include "klee/Constraints.h"
+
+#include "llvm/Support/CommandLine.h"
+
+#include <map>
+#include <vector>
+#include <iostream>
+#include <sstream>
+#include <iomanip>
+
+using namespace klee;
+
+namespace {
+  llvm::cl::opt<bool>
+  PCWidthAsArg("pc-width-as-arg", llvm::cl::init(true));
+
+  llvm::cl::opt<bool>
+  PCAllWidths("pc-all-widths", llvm::cl::init(false));
+
+  llvm::cl::opt<bool>
+  PCPrefixWidth("pc-prefix-width", llvm::cl::init(true));
+
+  llvm::cl::opt<bool>
+  PCMultibyteReads("pc-multibyte-reads", llvm::cl::init(true));
+
+  llvm::cl::opt<bool>
+  PCAllConstWidths("pc-all-const-widths",  llvm::cl::init(false));
+}
+
+/// PrintContext - Helper class for storing extra information for
+/// the pretty printer.
+class PrintContext {
+private:
+  std::ostream &os;
+  std::stringstream ss;
+  std::string newline;
+
+public:
+  /// Number of characters on the current line.
+  unsigned pos;
+
+public:
+  PrintContext(std::ostream &_os) : os(_os), newline("\n"), pos(0) {}
+
+  void setNewline(const std::string &_newline) {
+    newline = _newline;
+  }
+
+  void breakLine(unsigned indent=0) {
+    os << newline;
+    if (indent)
+      os << std::setw(indent) << ' ';
+    pos = indent;
+  }
+
+  /// write - Output a string to the stream and update the
+  /// position. The stream should not have any newlines.
+  void write(const std::string &s) {
+    os << s;
+    pos += s.length();
+  }
+
+  template <typename T>
+  PrintContext &operator<<(T elt) {
+    ss.str("");
+    ss << elt;
+    write(ss.str());
+    return *this;
+  }
+};
+
+class PPrinter : public ExprPPrinter {
+  std::map<ref<Expr>, unsigned> bindings;
+  std::map<const UpdateNode*, unsigned> updateBindings;
+  std::set< ref<Expr> > couldPrint, shouldPrint;
+  std::set<const UpdateNode*> couldPrintUpdates, shouldPrintUpdates;
+  std::ostream &os;
+  unsigned counter;
+  unsigned updateCounter;
+  bool hasScan;
+  std::string newline;
+
+  /// shouldPrintWidth - Predicate for whether this expression should
+  /// be printed with its width.
+  bool shouldPrintWidth(ref<Expr> e) {
+    if (PCAllWidths)
+      return true;
+    return e.getWidth() != Expr::Bool;
+  }
+
+  bool isVerySimple(const ref<Expr> &e) { 
+    return e.isConstant() || bindings.find(e)!=bindings.end();
+  }
+
+  bool isVerySimpleUpdate(const UpdateNode *un) {
+    return !un || updateBindings.find(un)!=updateBindings.end();
+  }
+
+
+  // document me!
+  bool isSimple(const ref<Expr> &e) { 
+    if (isVerySimple(e)) {
+      return true;
+    } else if (const ReadExpr *re = dyn_ref_cast<ReadExpr>(e)) {
+      return isVerySimple(re->index) && isVerySimpleUpdate(re->updates.head);
+    } else {
+      Expr *ep = e.get();
+      for (unsigned i=0; i<ep->getNumKids(); i++)
+        if (!isVerySimple(ep->getKid(i)))
+          return false;
+      return true;
+    }
+  }
+
+  bool hasSimpleKids(const Expr *ep) {
+      for (unsigned i=0; i<ep->getNumKids(); i++)
+        if (!isSimple(ep->getKid(i)))
+          return false;
+      return true;
+  }
+  
+  void scanUpdate(const UpdateNode *un) {
+    if (un) {
+      if (couldPrintUpdates.insert(un).second) {
+        scanUpdate(un->next);
+        scan1(un->index);
+        scan1(un->value);
+      } else {
+        shouldPrintUpdates.insert(un);
+      }
+    }
+  }
+
+  void scan1(const ref<Expr> &e) {
+    if (!e.isConstant()) {
+      if (couldPrint.insert(e).second) {
+        Expr *ep = e.get();
+        for (unsigned i=0; i<ep->getNumKids(); i++)
+          scan1(ep->getKid(i));
+        if (const ReadExpr *re = dyn_ref_cast<ReadExpr>(e)) 
+          scanUpdate(re->updates.head);
+      } else {
+        shouldPrint.insert(e);
+      }
+    }
+  }
+
+  void printUpdateList(const UpdateList &updates, PrintContext &PC) {
+    const UpdateNode *head = updates.head;
+
+    // Special case empty list.
+    if (!head) {
+      if (updates.isRooted) {
+        PC << "arr" << updates.root->id;
+      } else {
+        PC << "[]";
+      }
+      return;
+    }
+
+    // FIXME: Explain this breaking policy.
+    bool openedList = false, nextShouldBreak = false;
+    unsigned outerIndent = PC.pos;
+    unsigned middleIndent = 0;
+    for (const UpdateNode *un = head; un; un = un->next) {      
+      // We are done if we hit the cache.
+      std::map<const UpdateNode*, unsigned>::iterator it = 
+        updateBindings.find(un);
+      if (it!=updateBindings.end()) {
+        if (openedList)
+          PC << "] @ ";
+        PC << "U" << it->second;
+        return;
+      } else if (!hasScan || shouldPrintUpdates.count(un)) {
+        if (openedList)
+          PC << "] @";
+        if (un != head)
+          PC.breakLine(outerIndent);
+        PC << "U" << updateCounter << ":"; 
+        updateBindings.insert(std::make_pair(un, updateCounter++));
+        openedList = nextShouldBreak = false;
+     }
+    
+      if (!openedList) {
+        openedList = 1;
+        PC << '[';
+        middleIndent = PC.pos;
+      } else {
+        PC << ',';
+        printSeparator(PC, !nextShouldBreak, middleIndent);
+      }
+      //PC << "(=";
+      //unsigned innerIndent = PC.pos;
+      print(un->index, PC);
+      //printSeparator(PC, isSimple(un->index), innerIndent);
+      PC << "=";
+      print(un->value, PC);
+      //PC << ')';
+      
+      nextShouldBreak = !(un->index.isConstant() && un->value.isConstant());
+    }
+
+    if (openedList)
+      PC << ']';
+
+    if (updates.isRooted)
+      PC << " @ arr" << updates.root->id;
+  }
+
+  void printWidth(PrintContext &PC, ref<Expr> e) {
+    if (!shouldPrintWidth(e))
+      return;
+
+    if (PCWidthAsArg) {
+      PC << ' ';
+      if (PCPrefixWidth)
+        PC << 'w';
+    }
+
+    PC << e.getWidth();
+  }
+
+  /// hasOrderedReads - True iff all children are reads with
+  /// consecutive offsets according to the given \arg stride.
+  bool hasOrderedReads(const Expr *ep, int stride) {
+    const ReadExpr *base = dyn_ref_cast<ReadExpr>(ep->getKid(0));
+    if (!base)
+      return false;
+
+    // Get stride expr in proper index width.
+    Expr::Width idxWidth = base->index.getWidth();
+    ref<Expr> strideExpr(stride, idxWidth), offset(0, idxWidth);
+    for (unsigned i=1; i<ep->getNumKids(); ++i) {
+      const ReadExpr *re = dyn_ref_cast<ReadExpr>(ep->getKid(i));
+      if (!re) 
+        return false;
+
+      // Check if the index follows the stride. 
+      // FIXME: How aggressive should this be simplified. The
+      // canonicalizing builder is probably the right choice, but this
+      // is yet another area where we would really prefer it to be
+      // global or else use static methods.
+      offset = AddExpr::create(offset, strideExpr);
+      if (SubExpr::create(re->index, base->index) != offset)
+        return false;
+    }
+
+    return true;
+  }
+
+  /// hasAllByteReads - True iff all children are byte level reads.
+  bool hasAllByteReads(const Expr *ep) {
+    for (unsigned i=0; i<ep->getNumKids(); ++i) {
+      const ReadExpr *re = dyn_ref_cast<ReadExpr>(ep->getKid(i));
+      if (!re || re->getWidth() != Expr::Int8)
+        return false;
+    }
+    return true;
+  }
+
+  void printRead(const ReadExpr *re, PrintContext &PC, unsigned indent) {
+    print(re->index, PC);
+    printSeparator(PC, isVerySimple(re->index), indent);
+    printUpdateList(re->updates, PC);
+  }
+
+  void printExtract(const ExtractExpr *ee, PrintContext &PC, unsigned indent) {
+    PC << ee->offset << ' ';
+    print(ee->expr, PC);
+  }
+
+  void printExpr(const Expr *ep, PrintContext &PC, unsigned indent, bool printConstWidth=false) {
+    bool simple = hasSimpleKids(ep);
+    
+    print(ep->getKid(0), PC);
+    for (unsigned i=1; i<ep->getNumKids(); i++) {
+      printSeparator(PC, simple, indent);
+      print(ep->getKid(i), PC, printConstWidth);
+    }
+  }
+
+public:
+  PPrinter(std::ostream &_os) : os(_os), newline("\n") {
+    reset();
+  }
+
+  void setNewline(const std::string &_newline) {
+    newline = _newline;
+  }
+
+  void reset() {
+    counter = 0;
+    updateCounter = 0;
+    hasScan = false;
+    bindings.clear();
+    updateBindings.clear();
+    couldPrint.clear();
+    shouldPrint.clear();
+    couldPrintUpdates.clear();
+    shouldPrintUpdates.clear();
+  }
+
+  void scan(const ref<Expr> &e) {
+    hasScan = true;
+    scan1(e);
+  }
+
+  void print(const ref<Expr> &e, unsigned level=0) {
+    PrintContext PC(os);
+    PC.pos = level;
+    print(e, PC);
+  }
+
+  void printConst(const ref<Expr> &e, PrintContext &PC, bool printWidth) {
+    assert(e.isConstant());
+
+    if (e.getWidth() == Expr::Bool)
+      PC << (e.getConstantValue() ? "true" : "false");
+    else {
+      if (PCAllConstWidths)
+	printWidth = true;
+    
+      if (printWidth)
+	PC << "(w" << e.getWidth() << " ";
+
+      PC << e.getConstantValue();
+
+      if (printWidth)
+	PC << ")";
+    }    
+  }
+
+  void print(const ref<Expr> &e, PrintContext &PC, bool printConstWidth=false) {
+    if (e.isConstant()) 
+      printConst(e, PC, printConstWidth);
+    else {
+      std::map<ref<Expr>, unsigned>::iterator it = bindings.find(e);
+      if (it!=bindings.end()) {
+        PC << 'N' << it->second;
+      } else {
+        if (!hasScan || shouldPrint.count(e)) {
+          PC << 'N' << counter << ':';
+          bindings.insert(std::make_pair(e, counter++));
+        }
+
+        // Detect Not.
+        // FIXME: This should be in common code.
+        if (const EqExpr *ee = dyn_ref_cast<EqExpr>(e)) {
+          if (ee->left == ref<Expr>(false, Expr::Bool)) {
+            PC << "(Not";
+            printWidth(PC, e);
+            PC << ' ';
+            // FIXME: This is a boom if right is a constant.
+            print(ee->right, PC);
+            PC << ')';
+            return;
+          }
+        }
+
+        // Detect multibyte reads.
+        // FIXME: Hrm. One problem with doing this is that we are
+        // masking the sharing of the indices which aren't
+        // visible. Need to think if this matters... probably not
+        // because if they are offset reads then its either constant,
+        // or they are (base + offset) and base will get printed with
+        // a declaration.
+        if (PCMultibyteReads && e.getKind() == Expr::Concat) {
+          const Expr *ep = e.get();
+          if (hasAllByteReads(ep)) {
+            bool isMSB = hasOrderedReads(ep, 1);
+            if (isMSB || hasOrderedReads(ep, -1)) {
+              PC << "(Read" << (isMSB ? "MSB" : "LSB");
+              printWidth(PC, e);
+              PC << ' ';
+              unsigned firstIdx = isMSB ? 0 : ep->getNumKids()-1;
+              printRead(static_ref_cast<ReadExpr>(ep->getKid(firstIdx)), 
+                        PC, PC.pos);
+              PC << ')';
+              return;
+            }
+          }
+        }
+
+	PC << '(' << e.getKind();
+        printWidth(PC, e);
+        PC << ' ';
+
+        // Indent at first argument and dispatch to appropriate print
+        // routine for exprs which require special handling.
+        unsigned indent = PC.pos;
+        if (const ReadExpr *re = dyn_ref_cast<ReadExpr>(e)) {
+          printRead(re, PC, indent);
+        } else if (const ExtractExpr *ee = dyn_ref_cast<ExtractExpr>(e)) {
+          printExtract(ee, PC, indent);
+        } else if (e.getKind() == Expr::Concat || e.getKind() == Expr::SExt)
+	  printExpr(e.get(), PC, indent, true);
+	else
+          printExpr(e.get(), PC, indent);	
+        PC << ")";
+      }
+    }
+  }
+
+  /* Public utility functions */
+
+  void printSeparator(PrintContext &PC, bool simple, unsigned indent) {
+    if (simple) {
+      PC << ' ';
+    } else {
+      PC.breakLine(indent);
+    }
+  }
+};
+
+ExprPPrinter *klee::ExprPPrinter::create(std::ostream &os) {
+  return new PPrinter(os);
+}
+
+void ExprPPrinter::printOne(std::ostream &os,
+                            const char *message, 
+                            const ref<Expr> &e) {
+  PPrinter p(os);
+  p.scan(e);
+
+  // FIXME: Need to figure out what to do here. Probably print as a
+  // "forward declaration" with whatever syntax we pick for that.
+  PrintContext PC(os);
+  PC << message << ": ";
+  p.print(e, PC);
+  PC.breakLine();
+}
+
+void ExprPPrinter::printConstraints(std::ostream &os,
+                                    const ConstraintManager &constraints) {
+  printQuery(os, constraints, ref<Expr>(false, Expr::Bool));
+}
+
+void ExprPPrinter::printQuery(std::ostream &os,
+                              const ConstraintManager &constraints,
+                              const ref<Expr> &q) {
+  PPrinter p(os);
+  
+  for (ConstraintManager::const_iterator it = constraints.begin(),
+         ie = constraints.end(); it != ie; ++it)
+    p.scan(*it);
+  p.scan(q);
+
+  PrintContext PC(os);
+  PC << "(query [";
+  
+  // Ident at constraint list;
+  unsigned indent = PC.pos;
+  for (ConstraintManager::const_iterator it = constraints.begin(),
+         ie = constraints.end(); it != ie;) {
+    p.print(*it, PC);
+    ++it;
+    if (it != ie)
+      PC.breakLine(indent);
+  }
+  PC << ']';
+
+  p.printSeparator(PC, constraints.empty(), indent-1);
+  p.print(q, PC);
+
+  PC << ')';
+  PC.breakLine();
+}

Added: klee/trunk/lib/Expr/ExprUtil.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/ExprUtil.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/ExprUtil.cpp (added)
+++ klee/trunk/lib/Expr/ExprUtil.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,127 @@
+//===-- ExprUtil.cpp ------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/util/ExprUtil.h"
+#include "klee/util/ExprHashMap.h"
+
+#include "klee/Expr.h"
+
+#include "klee/util/ExprVisitor.h"
+
+#include <set>
+
+using namespace klee;
+
+void klee::findReads(ref<Expr> e, 
+                     bool visitUpdates,
+                     std::vector< ref<ReadExpr> > &results) {
+  // Invariant: \forall_{i \in stack} !i.isConstant() && i \in visited 
+  std::vector< ref<Expr> > stack;
+  ExprHashSet visited;
+  std::set<const UpdateNode *> updates;
+  
+  if (!e.isConstant()) {
+    visited.insert(e);
+    stack.push_back(e);
+  }
+
+  while (!stack.empty()) {
+    ref<Expr> top = stack.back();
+    stack.pop_back();
+
+    if (ReadExpr *re = dyn_ref_cast<ReadExpr>(top)) {
+      // We memoized so can just add to list without worrying about
+      // repeats.
+      results.push_back(re);
+
+      if (!re->index.isConstant() &&
+          visited.insert(re->index).second)
+        stack.push_back(re->index);
+      
+      if (visitUpdates) {
+        // XXX this is probably suboptimal. We want to avoid a potential
+        // explosion traversing update lists which can be quite
+        // long. However, it seems silly to hash all of the update nodes
+        // especially since we memoize all the expr results anyway. So
+        // we take a simple approach of memoizing the results for the
+        // head, which often will be shared among multiple nodes.
+        if (updates.insert(re->updates.head).second) {
+          for (const UpdateNode *un=re->updates.head; un; un=un->next) {
+            if (!un->index.isConstant() &&
+                visited.insert(un->index).second)
+              stack.push_back(un->index);
+            if (!un->value.isConstant() &&
+                visited.insert(un->value).second)
+              stack.push_back(un->value);
+          }
+        }
+      }
+    } else if (!top.isConstant()) {
+      Expr *e = top.get();
+      for (unsigned i=0; i<e->getNumKids(); i++) {
+        ref<Expr> k = e->getKid(i);
+        if (!k.isConstant() &&
+            visited.insert(k).second)
+          stack.push_back(k);
+      }
+    }
+  }
+}
+
+///
+
+namespace klee {
+
+class SymbolicObjectFinder : public ExprVisitor {
+protected:
+  Action visitRead(const ReadExpr &re) {
+    const UpdateList &ul = re.updates;
+
+    // XXX should we memo better than what ExprVisitor is doing for us?
+    for (const UpdateNode *un=ul.head; un; un=un->next) {
+      visit(un->index);
+      visit(un->value);
+    }
+
+    if (ul.isRooted)
+      if (results.insert(ul.root).second)
+        objects.push_back(ul.root);
+
+    return Action::doChildren();
+  }
+
+public:
+  std::set<const Array*> results;
+  std::vector<const Array*> &objects;
+  
+  SymbolicObjectFinder(std::vector<const Array*> &_objects)
+    : objects(_objects) {}
+};
+
+}
+
+template<typename InputIterator>
+void klee::findSymbolicObjects(InputIterator begin, 
+                               InputIterator end,
+                               std::vector<const Array*> &results) {
+  SymbolicObjectFinder of(results);
+  for (; begin!=end; ++begin)
+    of.visit(*begin);
+}
+
+void klee::findSymbolicObjects(ref<Expr> e,
+                               std::vector<const Array*> &results) {
+  findSymbolicObjects(&e, &e+1, results);
+}
+
+typedef std::vector< ref<Expr> >::iterator A;
+template void klee::findSymbolicObjects<A>(A, A, std::vector<const Array*> &);
+
+typedef std::set< ref<Expr> >::iterator B;
+template void klee::findSymbolicObjects<B>(B, B, std::vector<const Array*> &);

Added: klee/trunk/lib/Expr/ExprVisitor.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/ExprVisitor.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/ExprVisitor.cpp (added)
+++ klee/trunk/lib/Expr/ExprVisitor.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,253 @@
+//===-- ExprVisitor.cpp ---------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Expr.h"
+#include "klee/util/ExprVisitor.h"
+
+#include "llvm/Support/CommandLine.h"
+
+namespace {
+  llvm::cl::opt<bool>
+  UseVisitorHash("use-visitor-hash", 
+                 llvm::cl::desc("Use hash-consing during expr visitation."),
+                 llvm::cl::init(true));
+}
+
+using namespace klee;
+
+ref<Expr> ExprVisitor::visit(const ref<Expr> &e) {
+  if (!UseVisitorHash || e.isConstant()) {
+    return visitActual(e);
+  } else {
+    visited_ty::iterator it = visited.find(e);
+
+    if (it!=visited.end()) {
+      return it->second;
+    } else {
+      ref<Expr> res = visitActual(e);
+      visited.insert(std::make_pair(e, res));
+      return res;
+    }
+  }
+}
+
+ref<Expr> ExprVisitor::visitActual(const ref<Expr> &e) {
+  if (e.isConstant()) {    
+    return e;
+  } else {
+    Expr &ep = *e.get();
+
+    Action res = visitExpr(ep);
+    switch(res.kind) {
+    case Action::DoChildren:
+      // continue with normal action
+      break;
+    case Action::SkipChildren:
+      return e;
+    case Action::ChangeTo:
+      return res.argument;
+    }
+
+    switch(ep.getKind()) {
+    case Expr::NotOptimized: res = visitNotOptimized(static_cast<NotOptimizedExpr&>(ep)); break;
+    case Expr::Read: res = visitRead(static_cast<ReadExpr&>(ep)); break;
+    case Expr::Select: res = visitSelect(static_cast<SelectExpr&>(ep)); break;
+    case Expr::Concat: res = visitConcat(static_cast<ConcatExpr&>(ep)); break;
+    case Expr::Extract: res = visitExtract(static_cast<ExtractExpr&>(ep)); break;
+    case Expr::ZExt: res = visitZExt(static_cast<ZExtExpr&>(ep)); break;
+    case Expr::SExt: res = visitSExt(static_cast<SExtExpr&>(ep)); break;
+    case Expr::Add: res = visitAdd(static_cast<AddExpr&>(ep)); break;
+    case Expr::Sub: res = visitSub(static_cast<SubExpr&>(ep)); break;
+    case Expr::Mul: res = visitMul(static_cast<MulExpr&>(ep)); break;
+    case Expr::UDiv: res = visitUDiv(static_cast<UDivExpr&>(ep)); break;
+    case Expr::SDiv: res = visitSDiv(static_cast<SDivExpr&>(ep)); break;
+    case Expr::URem: res = visitURem(static_cast<URemExpr&>(ep)); break;
+    case Expr::SRem: res = visitSRem(static_cast<SRemExpr&>(ep)); break;
+    case Expr::And: res = visitAnd(static_cast<AndExpr&>(ep)); break;
+    case Expr::Or: res = visitOr(static_cast<OrExpr&>(ep)); break;
+    case Expr::Xor: res = visitXor(static_cast<XorExpr&>(ep)); break;
+    case Expr::Shl: res = visitShl(static_cast<ShlExpr&>(ep)); break;
+    case Expr::LShr: res = visitLShr(static_cast<LShrExpr&>(ep)); break;
+    case Expr::AShr: res = visitAShr(static_cast<AShrExpr&>(ep)); break;
+    case Expr::Eq: res = visitEq(static_cast<EqExpr&>(ep)); break;
+    case Expr::Ne: res = visitNe(static_cast<NeExpr&>(ep)); break;
+    case Expr::Ult: res = visitUlt(static_cast<UltExpr&>(ep)); break;
+    case Expr::Ule: res = visitUle(static_cast<UleExpr&>(ep)); break;
+    case Expr::Ugt: res = visitUgt(static_cast<UgtExpr&>(ep)); break;
+    case Expr::Uge: res = visitUge(static_cast<UgeExpr&>(ep)); break;
+    case Expr::Slt: res = visitSlt(static_cast<SltExpr&>(ep)); break;
+    case Expr::Sle: res = visitSle(static_cast<SleExpr&>(ep)); break;
+    case Expr::Sgt: res = visitSgt(static_cast<SgtExpr&>(ep)); break;
+    case Expr::Sge: res = visitSge(static_cast<SgeExpr&>(ep)); break;
+    case Expr::Constant:
+    default:
+      assert(0 && "invalid expression kind");
+    }
+
+    switch(res.kind) {
+    case Action::DoChildren: {  
+      bool rebuild = false;
+      ref<Expr> e(&ep), kids[8];
+      unsigned count = ep.getNumKids();
+      for (unsigned i=0; i<count; i++) {
+        ref<Expr> kid = ep.getKid(i);
+        kids[i] = visit(kid);
+        if (kids[i] != kid)
+          rebuild = true;
+      }
+      if (rebuild) {
+        e = ep.rebuild(kids);
+        if (recursive)
+          e = visit(e);
+      }
+      if (!e.isConstant()) {
+        res = visitExprPost(*e.get());
+        if (res.kind==Action::ChangeTo)
+          e = res.argument;
+      }
+      return e;
+    }
+    case Action::SkipChildren:
+      return e;
+    case Action::ChangeTo:
+      return res.argument;
+    default:
+      assert(0 && "invalid kind");
+    }
+  }
+}
+
+ExprVisitor::Action ExprVisitor::visitExpr(const Expr&) {
+  return Action::doChildren();
+}
+
+ExprVisitor::Action ExprVisitor::visitExprPost(const Expr&) {
+  return Action::skipChildren();
+}
+
+ExprVisitor::Action ExprVisitor::visitNotOptimized(const NotOptimizedExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitRead(const ReadExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSelect(const SelectExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitConcat(const ConcatExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitExtract(const ExtractExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitZExt(const ZExtExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSExt(const SExtExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitAdd(const AddExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSub(const SubExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitMul(const MulExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitUDiv(const UDivExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSDiv(const SDivExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitURem(const URemExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSRem(const SRemExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitAnd(const AndExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitOr(const OrExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitXor(const XorExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitShl(const ShlExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitLShr(const LShrExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitAShr(const AShrExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitEq(const EqExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitNe(const NeExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitUlt(const UltExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitUle(const UleExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitUgt(const UgtExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitUge(const UgeExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSlt(const SltExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSle(const SleExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSgt(const SgtExpr&) {
+  return Action::doChildren(); 
+}
+
+ExprVisitor::Action ExprVisitor::visitSge(const SgeExpr&) {
+  return Action::doChildren(); 
+}
+

Added: klee/trunk/lib/Expr/Lexer.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Lexer.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Lexer.cpp (added)
+++ klee/trunk/lib/Expr/Lexer.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,261 @@
+//===-- Lexer.cpp ---------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "expr/Lexer.h"
+
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Streams.h"
+
+#include <iomanip>
+#include <iostream>
+#include <string.h>
+
+using namespace llvm;
+using namespace klee;
+using namespace klee::expr;
+
+///
+
+const char *Token::getKindName() const {
+  switch (kind) {
+  default:
+  case Unknown:    return "Unknown";
+  case Arrow:      return "Arrow";
+  case At:         return "At";
+  case Colon:      return "Colon";
+  case Comma:      return "Comma";
+  case Comment:    return "Comment";
+  case EndOfFile:  return "EndOfFile";
+  case Equals:     return "Equals";
+  case Identifier: return "Identifier";
+  case KWFalse:    return "KWFalse";
+  case KWQuery:    return "KWQuery";
+  case KWReserved: return "KWReserved";
+  case KWTrue:     return "KWTrue";
+  case KWWidth:    return "KWWidth";
+  case LBrace:     return "LBrace";
+  case LParen:     return "LParen";
+  case LSquare:    return "LSquare";
+  case Number:     return "Number";
+  case RBrace:     return "RBrace";
+  case RParen:     return "RParen";
+  case RSquare:    return "RSquare";
+  case Semicolon:  return "Semicolon";
+  }
+}
+
+void Token::dump() {
+  llvm::cerr << "(Token \"" << getKindName() << "\" "
+             << (void*) start << " " << length << " "
+             << line << " " << column << ")";
+}
+
+///
+
+static inline bool isInternalIdentifierChar(int Char) {
+  return isalnum(Char) || Char == '_' || Char == '.';
+}
+
+Lexer::Lexer(const llvm::MemoryBuffer *MB) 
+  : BufferPos(MB->getBufferStart()), BufferEnd(MB->getBufferEnd()), 
+    LineNumber(1), ColumnNumber(0) {
+}
+
+Lexer::~Lexer() {
+}
+
+int Lexer::PeekNextChar() {
+  if (BufferPos == BufferEnd)
+    return -1;
+  return *BufferPos;
+}
+
+int Lexer::GetNextChar() {
+  if (BufferPos == BufferEnd)
+    return -1;
+
+  // Handle DOS/Mac newlines here, by stripping duplicates and by
+  // returning '\n' for both.
+  char Result = *BufferPos++;
+  if (Result == '\n' || Result == '\r') {
+    if (BufferPos != BufferEnd && *BufferPos == ('\n' + '\r' - Result))
+      ++BufferPos;
+    Result = '\n';
+  }
+
+  if (Result == '\n') {
+    ++LineNumber;
+    ColumnNumber = 0;
+  } else {
+    ++ColumnNumber;
+  }
+
+  return Result;
+}
+
+Token &Lexer::SetTokenKind(Token &Result, Token::Kind k) {
+  Result.kind = k;
+  Result.length = BufferPos - Result.start;
+  return Result;
+}
+
+static bool isReservedKW(const char *Str, unsigned N) {
+    unsigned i;
+
+  // Check for i[0-9]+
+  if (N>1 && Str[0] == 'i') {
+    for (i=1; i<N; ++i)
+      if (!isdigit(Str[i]))
+        break;
+    if (i==N)
+      return true;
+  }
+
+  // Check for fp[0-9]+([.].*)?$
+  if (N>3 && Str[0]=='f' && Str[1]=='p' && isdigit(Str[2])) {
+    for (i=3; i<N; ++i)
+      if (!isdigit(Str[i]))
+        break;
+    if (i==N || Str[i]=='.')
+      return true;
+  }
+  
+  return false;
+}
+static bool isWidthKW(const char *Str, unsigned N) {
+  if (N<2 || Str[0] != 'w')
+    return false;
+  for (unsigned i=1; i<N; ++i)
+    if (!isdigit(Str[i]))
+      return false;
+  return true;
+}
+Token &Lexer::SetIdentifierTokenKind(Token &Result) {
+  unsigned Length = BufferPos - Result.start;
+  switch (Length) {
+  case 3:
+    if (memcmp("def", Result.start, 3) == 0)
+      return SetTokenKind(Result, Token::KWReserved);
+    if (memcmp("var", Result.start, 3) == 0)
+      return SetTokenKind(Result, Token::KWReserved);
+    break;
+
+  case 4:
+    if (memcmp("true", Result.start, 4) == 0)
+      return SetTokenKind(Result, Token::KWTrue);
+    break;
+
+  case 5:
+    if (memcmp("array", Result.start, 5) == 0)
+      return SetTokenKind(Result, Token::KWReserved);
+    if (memcmp("false", Result.start, 5) == 0)
+      return SetTokenKind(Result, Token::KWFalse);
+    if (memcmp("query", Result.start, 5) == 0)
+      return SetTokenKind(Result, Token::KWQuery);
+    break;      
+    
+  case 6:
+    if (memcmp("define", Result.start, 6) == 0)
+      return SetTokenKind(Result, Token::KWReserved);
+    break;
+
+  case 7:
+    if (memcmp("declare", Result.start, 7) == 0)
+      return SetTokenKind(Result, Token::KWReserved);
+    break;
+  }
+
+  if (isReservedKW(Result.start, Length))
+    return SetTokenKind(Result, Token::KWReserved);
+  if (isWidthKW(Result.start, Length))
+    return SetTokenKind(Result, Token::KWWidth);
+
+  return SetTokenKind(Result, Token::Identifier);
+}
+
+void Lexer::SkipToEndOfLine() {
+  for (;;) {
+    int Char = GetNextChar();
+    if (Char == -1 || Char =='\n')
+      break;
+  }
+}
+
+Token &Lexer::LexNumber(Token &Result) {
+  while (isalnum(PeekNextChar()) || PeekNextChar()=='_')
+    GetNextChar();
+  return SetTokenKind(Result, Token::Number);
+}
+
+Token &Lexer::LexIdentifier(Token &Result) {
+  while (isInternalIdentifierChar(PeekNextChar()))
+    GetNextChar();
+
+  // Recognize keywords specially.
+  return SetIdentifierTokenKind(Result);
+}
+
+Token &Lexer::Lex(Token &Result) {
+  Result.kind = Token::Unknown;
+  Result.length = 0;
+  Result.start = BufferPos;
+  
+  // Skip whitespace.
+  while (isspace(PeekNextChar()))
+    GetNextChar();
+
+  Result.start = BufferPos;
+  Result.line = LineNumber;
+  Result.column = ColumnNumber;
+  int Char = GetNextChar();
+  switch (Char) {
+  case -1:  return SetTokenKind(Result, Token::EndOfFile);
+    
+  case '(': return SetTokenKind(Result, Token::LParen);
+  case ')': return SetTokenKind(Result, Token::RParen);
+  case ',': return SetTokenKind(Result, Token::Comma);
+  case ':': return SetTokenKind(Result, Token::Colon);
+  case ';': return SetTokenKind(Result, Token::Semicolon);
+  case '=': return SetTokenKind(Result, Token::Equals);
+  case '@': return SetTokenKind(Result, Token::At);
+  case '[': return SetTokenKind(Result, Token::LSquare);
+  case ']': return SetTokenKind(Result, Token::RSquare);
+  case '{': return SetTokenKind(Result, Token::LBrace);
+  case '}': return SetTokenKind(Result, Token::RBrace);
+
+  case '#':
+    SkipToEndOfLine();
+    return SetTokenKind(Result, Token::Comment);
+
+  case '+': {
+    if (isdigit(PeekNextChar()))
+      return LexNumber(Result);
+    else
+      return SetTokenKind(Result, Token::Unknown);
+  }
+
+  case '-': {
+    int Next = PeekNextChar();
+    if (Next == '>')
+      return GetNextChar(), SetTokenKind(Result, Token::Arrow);
+    else if (isdigit(Next))
+      return LexNumber(Result);
+    else
+      return SetTokenKind(Result, Token::Unknown);
+    break;
+  }
+
+  default:
+    if (isdigit(Char))
+      return LexNumber(Result);
+    else if (isalpha(Char) || Char == '_')
+      return LexIdentifier(Result);
+    return SetTokenKind(Result, Token::Unknown);
+  }
+}

Added: klee/trunk/lib/Expr/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Makefile (added)
+++ klee/trunk/lib/Expr/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,16 @@
+#===-- lib/Expr/Makefile -----------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+LEVEL=../..
+
+LIBRARYNAME=kleaverExpr
+DONT_BUILD_RELINKED=1
+BUILD_ARCHIVE=1
+
+include $(LEVEL)/Makefile.common

Added: klee/trunk/lib/Expr/Parser.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Parser.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Parser.cpp (added)
+++ klee/trunk/lib/Expr/Parser.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,1310 @@
+//===-- Parser.cpp --------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "expr/Parser.h"
+
+#include "expr/Lexer.h"
+
+#include "klee/Constraints.h"
+#include "klee/Solver.h"
+#include "klee/util/ExprPPrinter.h"
+
+#include "llvm/ADT/APInt.h"
+#include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Streams.h"
+
+#include <cassert>
+#include <iostream>
+#include <map>
+
+using namespace llvm;
+using namespace klee;
+using namespace klee::expr;
+
+namespace {
+  /// ParseResult - Represent a possibly invalid parse result.
+  template<typename T>
+  struct ParseResult {
+    bool IsValid;
+    T Value;
+
+  public:
+    ParseResult() : IsValid(false) {}
+    ParseResult(T _Value) : IsValid(true), Value(_Value) {}
+    ParseResult(bool _IsValid, T _Value) : IsValid(_IsValid), Value(_Value) {}
+
+    bool isValid() { 
+      return IsValid; 
+    }
+    T get() { 
+      assert(IsValid && "get() on invalid ParseResult!");
+      return Value; 
+    }
+  };
+  
+  typedef ParseResult<Decl*> DeclResult;
+  typedef ParseResult<ExprHandle> ExprResult;
+  typedef ParseResult<Expr::Width> TypeResult;
+  typedef ParseResult<VersionHandle> VersionResult;
+
+  /// NumberOrExprResult - Represent a number or expression. This is used to
+  /// wrap an expression production which may be a number, but for
+  /// which the type width is unknown.
+  class NumberOrExprResult {
+    Token AsNumber;
+    ExprResult AsExpr;
+    bool IsNumber;
+
+  public:
+    NumberOrExprResult() : IsNumber(false) {}
+    explicit NumberOrExprResult(Token _AsNumber) : AsNumber(_AsNumber),
+                                                   IsNumber(true) {}
+    explicit NumberOrExprResult(ExprResult _AsExpr) : AsExpr(_AsExpr),
+                                                      IsNumber(false) {}
+    
+    bool isNumber() const { return IsNumber; }
+    const Token &getNumber() const { 
+      assert(IsNumber && "Invalid accessor call.");
+      return AsNumber; 
+    }
+    const ExprResult &getExpr() const {
+      assert(!IsNumber && "Invalid accessor call.");
+      return AsExpr;
+    }
+  };
+
+  /// ParserImpl - Parser implementation.
+  class ParserImpl : public Parser {
+    typedef std::map<const std::string, const Identifier*> IdentifierTabTy;
+    typedef std::map<const Identifier*, ExprHandle> ExprSymTabTy;
+    typedef std::map<const Identifier*, VersionHandle> VersionSymTabTy;
+
+    const std::string Filename;
+    const MemoryBuffer *TheMemoryBuffer;
+    Lexer TheLexer;
+    unsigned MaxErrors;
+    unsigned NumErrors;
+
+    // FIXME: Use LLVM symbol tables?
+    IdentifierTabTy IdentifierTab;
+
+    std::map<const Identifier*, const ArrayDecl*> ArraySymTab;
+    ExprSymTabTy ExprSymTab;
+    VersionSymTabTy VersionSymTab;
+
+    /// Tok - The currently lexed token.
+    Token Tok;
+
+    /// ParenLevel - The current depth of matched '(' tokens. 
+    unsigned ParenLevel;
+    /// SquareLevel - The current depth of matched '[' tokens.
+    unsigned SquareLevel;
+
+    /* Core parsing functionality */
+    
+    const Identifier *GetOrCreateIdentifier(const Token &Tok);
+
+    void GetNextNonCommentToken() {
+      do {
+        TheLexer.Lex(Tok);
+      } while (Tok.kind == Token::Comment);
+    }
+
+    /// ConsumeToken - Consume the current 'peek token' and lex the next one.
+    void ConsumeToken() {
+      assert(Tok.kind != Token::LParen && Tok.kind != Token::RParen);
+      GetNextNonCommentToken();
+    }
+
+    /// ConsumeExpectedToken - Check that the current token is of the
+    /// expected kind and consume it.
+    void ConsumeExpectedToken(Token::Kind k) {
+      assert(Tok.kind == k && "Unexpected token!");
+      GetNextNonCommentToken();
+    }
+
+    void ConsumeLParen() {
+      ++ParenLevel;
+      ConsumeExpectedToken(Token::LParen);
+    }
+
+    void ConsumeRParen() {
+      if (ParenLevel) // Cannot go below zero.
+        --ParenLevel;
+      ConsumeExpectedToken(Token::RParen);
+    }
+
+    void ConsumeLSquare() {
+      ++SquareLevel;
+      ConsumeExpectedToken(Token::LSquare);
+    }
+
+    void ConsumeRSquare() {
+      if (SquareLevel) // Cannot go below zero.
+        --SquareLevel;
+      ConsumeExpectedToken(Token::RSquare);
+    }
+
+    void ConsumeAnyToken() {
+      switch (Tok.kind) {
+      case Token::LParen: return ConsumeLParen();
+      case Token::RParen: return ConsumeRParen();
+      case Token::LSquare: return ConsumeLSquare();
+      case Token::RSquare: return ConsumeRSquare();
+      default: 
+        return ConsumeToken();
+      }
+    }
+
+    /* Utility functions */
+
+    /// SkipUntilRParen - Scan forward to the next token following an
+    /// rparen at the given level, or EOF, whichever is first.
+    void SkipUntilRParen(unsigned Level) {
+      // FIXME: I keep wavering on whether it is an error to call this
+      // with the current token an rparen. In most cases this should
+      // have been handled differently (error reported,
+      // whatever). Audit & resolve.
+      assert(Level <= ParenLevel && 
+             "Refusing to skip until rparen at higher level.");
+      while (Tok.kind != Token::EndOfFile) {
+        if (Tok.kind == Token::RParen && ParenLevel == Level) {
+          ConsumeRParen();
+          break;
+        }
+        ConsumeAnyToken();
+      }
+    }
+
+    /// SkipUntilRParen - Scan forward until reaching an rparen token
+    /// at the current level (or EOF).
+    void SkipUntilRParen() {
+      SkipUntilRParen(ParenLevel);
+    }
+    
+    /// ExpectRParen - Utility method to close an sexp. This expects to
+    /// eat an rparen, and emits a diagnostic and skips to the next one
+    /// (or EOF) if it cannot.
+    void ExpectRParen(const char *Msg) {
+      if (Tok.kind == Token::EndOfFile) {
+        // FIXME: Combine with Msg
+        Error("expected ')' but found end-of-file.", Tok);
+      } else if (Tok.kind != Token::RParen) {
+        Error(Msg, Tok);
+        SkipUntilRParen();
+      } else {
+        ConsumeRParen();
+      }
+    }
+
+    /// SkipUntilRSquare - Scan forward to the next token following an
+    /// rsquare at the given level, or EOF, whichever is first.
+    void SkipUntilRSquare(unsigned Level) {
+      // FIXME: I keep wavering on whether it is an error to call this
+      // with the current token an rparen. In most cases this should
+      // have been handled differently (error reported,
+      // whatever). Audit & resolve.
+      assert(Level <= ParenLevel && 
+             "Refusing to skip until rparen at higher level.");
+      while (Tok.kind != Token::EndOfFile) {
+        if (Tok.kind == Token::RSquare && ParenLevel == Level) {
+          ConsumeRSquare();
+          break;
+        }
+        ConsumeAnyToken();
+      }
+    }
+
+    /// SkipUntilRSquare - Scan forward until reaching an rsquare token
+    /// at the current level (or EOF).
+    void SkipUntilRSquare() {
+      SkipUntilRSquare(ParenLevel);
+    }
+    
+    /// ExpectRSquare - Utility method to close an array. This expects
+    /// to eat an rparen, and emits a diagnostic and skips to the next
+    /// one (or EOF) if it cannot.
+    void ExpectRSquare(const char *Msg) {
+      if (Tok.kind == Token::EndOfFile) {
+        // FIXME: Combine with Msg
+        Error("expected ']' but found end-of-file.", Tok);
+      } else if (Tok.kind != Token::RSquare) {
+        Error(Msg, Tok);
+        SkipUntilRSquare();
+      } else {
+        ConsumeRSquare();
+      }
+    }
+
+    /*** Grammar productions ****/
+
+    /* Top level decls */
+
+    DeclResult ParseArrayDecl();
+    DeclResult ParseExprVarDecl();
+    DeclResult ParseVersionVarDecl();
+    DeclResult ParseCommandDecl();
+
+    /* Commands */
+
+    DeclResult ParseQueryCommand();
+
+    /* Etc. */
+
+    NumberOrExprResult ParseNumberOrExpr();
+
+    ExprResult ParseExpr(TypeResult ExpectedType);
+    ExprResult ParseParenExpr(TypeResult ExpectedType);
+    ExprResult ParseUnaryParenExpr(const Token &Name,
+                                   unsigned Kind, bool IsFixed,
+                                   Expr::Width ResTy);
+    ExprResult ParseBinaryParenExpr(const Token &Name,
+                                    unsigned Kind, bool IsFixed,
+                                    Expr::Width ResTy);
+    ExprResult ParseSelectParenExpr(const Token &Name, Expr::Width ResTy);
+    ExprResult ParseConcatParenExpr(const Token &Name, Expr::Width ResTy);
+    ExprResult ParseExtractParenExpr(const Token &Name, Expr::Width ResTy);
+    ExprResult ParseAnyReadParenExpr(const Token &Name,
+                                     unsigned Kind,
+                                     Expr::Width ResTy);
+    void ParseMatchedBinaryArgs(const Token &Name, 
+                                TypeResult ExpectType,
+                                ExprResult &LHS, ExprResult &RHS);
+    ExprResult ParseNumber(Expr::Width Width);
+    ExprResult ParseNumberToken(Expr::Width Width, const Token &Tok);
+
+    VersionResult ParseVersionSpecifier();
+    VersionResult ParseVersion();
+
+    TypeResult ParseTypeSpecifier();
+
+    /*** Diagnostics ***/
+    
+    void Error(const char *Message, const Token &At);
+    void Error(const char *Message) { Error(Message, Tok); }
+
+  public:
+    ParserImpl(const std::string _Filename,
+               const MemoryBuffer *MB) : Filename(_Filename),
+                                         TheMemoryBuffer(MB),
+                                         TheLexer(MB),
+                                         MaxErrors(~0u),
+                                         NumErrors(0) {}
+
+    /// Initialize - Initialize the parsing state. This must be called
+    /// prior to the start of parsing.
+    void Initialize() {
+      ParenLevel = SquareLevel = 0;
+
+      ConsumeAnyToken();
+    }
+
+    /* Parser interface implementation */
+
+    virtual Decl *ParseTopLevelDecl();
+
+    virtual void SetMaxErrors(unsigned N) {
+      MaxErrors = N;
+    }
+
+    virtual unsigned GetNumErrors() const {
+      return NumErrors; 
+    }
+  };
+}
+
+const Identifier *ParserImpl::GetOrCreateIdentifier(const Token &Tok) {
+  // FIXME: Make not horribly inefficient please.
+  assert(Tok.kind == Token::Identifier && "Expected only identifier tokens.");
+  std::string Name(Tok.start, Tok.length);
+  IdentifierTabTy::iterator it = IdentifierTab.find(Name);
+  if (it != IdentifierTab.end())
+    return it->second;
+
+  Identifier *I = new Identifier(Name);
+  IdentifierTab.insert(std::make_pair(Name, I));
+
+  return I;
+}
+
+Decl *ParserImpl::ParseTopLevelDecl() {
+  // Repeat until success or EOF.
+  while (Tok.kind != Token::EndOfFile) {
+    // Only handle commands for now.
+    if (Tok.kind == Token::LParen) {
+      DeclResult Res = ParseCommandDecl();
+      if (Res.isValid())
+        return Res.get();
+    } else {
+      Error("expected '(' token.");
+      ConsumeAnyToken();
+    }
+  }
+
+  return 0;
+}
+
+/// ParseCommandDecl - Parse a command declaration. The lexer should
+/// be positioned at the opening '('.
+///
+/// command = '(' name ... ')'
+DeclResult ParserImpl::ParseCommandDecl() {
+  ConsumeLParen();
+
+  if (!Tok.isKeyword()) {
+    Error("malformed command.");
+    SkipUntilRParen();
+    return DeclResult();
+  }
+
+  switch (Tok.kind) {
+  case Token::KWQuery:
+    return ParseQueryCommand();
+
+  default:
+    Error("malformed command (unexpected keyword).");
+    SkipUntilRParen();
+    return DeclResult();
+  }
+}
+
+/// ParseQueryCommand - Parse query command. The lexer should be
+/// positioned at the 'query' keyword.
+/// 
+/// 'query' expressions-list expression [expression-list [array-list]]
+DeclResult ParserImpl::ParseQueryCommand() {
+  // FIXME: We need a command for this. Or something.
+  ExprSymTab.clear();
+  VersionSymTab.clear();
+
+  std::vector<ExprHandle> Constraints;
+  ConsumeExpectedToken(Token::KWQuery);
+  if (Tok.kind != Token::LSquare) {
+    Error("malformed query, expected constraint list.");
+    SkipUntilRParen();
+    return DeclResult();
+  }
+
+  ConsumeExpectedToken(Token::LSquare);
+  // FIXME: Should avoid reading past unbalanced parens here.
+  while (Tok.kind != Token::RSquare) {
+    if (Tok.kind == Token::EndOfFile) {
+      Error("unexpected end of file.");
+      return new QueryCommand(Constraints.begin(), Constraints.end(),
+                              ref<Expr>(false, Expr::Bool));
+    }
+
+    ExprResult Res = ParseExpr(TypeResult(Expr::Bool));
+    if (Res.isValid())
+      Constraints.push_back(Res.get());
+  }
+
+  ConsumeRSquare();
+
+  ExprResult Res = ParseExpr(TypeResult());
+  if (!Res.isValid()) // Error emitted by ParseExpr.
+    Res = ExprResult(ref<Expr>(0, Expr::Bool));
+
+  ExpectRParen("unexpected argument to 'query'.");  
+  return new QueryCommand(Constraints.begin(), Constraints.end(),
+                          Res.get());
+}
+
+/// ParseNumberOrExpr - Parse an expression whose type cannot be
+/// predicted.
+NumberOrExprResult ParserImpl::ParseNumberOrExpr() {
+  if (Tok.kind == Token::Number){ 
+    Token Num = Tok;
+    ConsumeToken();
+    return NumberOrExprResult(Num);
+  } else {
+    return NumberOrExprResult(ParseExpr(TypeResult()));
+  }
+}
+
+/// ParseExpr - Parse an expression with the given \arg
+/// ExpectedType. \arg ExpectedType can be invalid if the type cannot
+/// be inferred from the context.
+///
+/// expr = false | true
+/// expr = <constant>
+/// expr = <identifier>
+/// expr = [<identifier>:] paren-expr
+ExprResult ParserImpl::ParseExpr(TypeResult ExpectedType) {
+  // FIXME: Is it right to need to do this here?
+  if (Tok.kind == Token::EndOfFile) {
+    Error("unexpected end of file.");
+    return ExprResult();
+  }
+
+  if (Tok.kind == Token::KWFalse || Tok.kind == Token::KWTrue) {
+    bool Value = Tok.kind == Token::KWTrue;
+    ConsumeToken();
+    return ExprResult(ref<Expr>(Value, Expr::Bool));
+  }
+  
+  if (Tok.kind == Token::Number) {
+    if (!ExpectedType.isValid()) {
+      Error("cannot infer type of number.");
+      ConsumeToken();
+      return ExprResult();
+    }
+    
+    return ParseNumber(ExpectedType.get());
+  }
+
+  const Identifier *Label = 0;
+  if (Tok.kind == Token::Identifier) {
+    Token LTok = Tok;
+    Label = GetOrCreateIdentifier(Tok);
+    ConsumeToken();
+
+    if (Tok.kind != Token::Colon) {
+      ExprSymTabTy::iterator it = ExprSymTab.find(Label);
+
+      if (it == ExprSymTab.end()) {
+        Error("invalid expression label reference.", LTok);
+        return ExprResult();
+      }
+
+      return it->second;
+    }
+
+    ConsumeToken();
+    if (ExprSymTab.count(Label)) {
+      Error("duplicate expression label definition.", LTok);
+      Label = 0;
+    }
+  }
+
+  Token Start = Tok;
+  ExprResult Res = ParseParenExpr(ExpectedType);
+  if (!Res.isValid()) {
+    // If we know the type, define the identifier just so we don't get
+    // use-of-undef errors. 
+    // FIXME: Maybe we should let the symbol table map to invalid
+    // entries?
+    if (Label && ExpectedType.isValid())
+      ExprSymTab.insert(std::make_pair(Label, 
+                                       ref<Expr>(0, ExpectedType.get())));
+    return Res;
+  } else if (ExpectedType.isValid()) {
+    // Type check result.    
+    if (Res.get().getWidth() != ExpectedType.get()) {
+      // FIXME: Need more info, and range
+      Error("expression has incorrect type.", Start);
+      return ExprResult();
+    }
+  }
+
+  if (Label)
+    ExprSymTab.insert(std::make_pair(Label, Res.get()));
+  return Res;
+}
+
+// Additional kinds for macro forms.
+enum MacroKind {
+  eMacroKind_Not = Expr::LastKind + 1,     // false == x
+  eMacroKind_Neg,                          // 0 - x
+  eMacroKind_ReadLSB,                      // Multibyte read
+  eMacroKind_ReadMSB,                      // Multibyte write
+  eMacroKind_Concat,                       // Magic concatenation syntax
+  eMacroKind_LastMacroKind = eMacroKind_ReadMSB
+};
+
+/// LookupExprInfo - Return information on the named token, if it is
+/// recognized.
+///
+/// \param Kind [out] - The Expr::Kind or MacroKind of the identifier.
+/// \param IsFixed [out] - True if the given kinds result and
+/// (expression) arguments are all of the same width.
+/// \param NumArgs [out] - The number of expression arguments for this
+/// kind. -1 indicates the kind is variadic or has non-expression
+/// arguments.
+/// \return True if the token is a valid kind or macro name.
+static bool LookupExprInfo(const Token &Tok, unsigned &Kind, 
+                           bool &IsFixed, int &NumArgs) {
+#define SetOK(kind, isfixed, numargs) (Kind=kind, IsFixed=isfixed,\
+                                       NumArgs=numargs, true)
+  assert(Tok.kind == Token::Identifier && "Unexpected token.");
+  
+  switch (Tok.length) {
+  case 2:
+    if (memcmp(Tok.start, "Eq", 2) == 0)
+      return SetOK(Expr::Eq, false, 2);
+    if (memcmp(Tok.start, "Ne", 2) == 0)
+      return SetOK(Expr::Ne, false, 2);
+
+    if (memcmp(Tok.start, "Or", 2) == 0)
+      return SetOK(Expr::Or, true, 2);
+    break;
+
+  case 3:
+    if (memcmp(Tok.start, "Add", 3) == 0)
+      return SetOK(Expr::Add, true, 2);
+    if (memcmp(Tok.start, "Sub", 3) == 0)
+      return SetOK(Expr::Sub, true, 2);
+    if (memcmp(Tok.start, "Mul", 3) == 0)
+      return SetOK(Expr::Mul, true, 2);
+
+    if (memcmp(Tok.start, "And", 3) == 0)
+      return SetOK(Expr::And, true, 2);
+    if (memcmp(Tok.start, "Shl", 3) == 0)
+      return SetOK(Expr::Shl, true, 2);
+    if (memcmp(Tok.start, "Xor", 3) == 0)
+      return SetOK(Expr::Xor, true, 2);
+
+    if (memcmp(Tok.start, "Not", 3) == 0)
+      return SetOK(eMacroKind_Not, true, 1);
+    if (memcmp(Tok.start, "Neg", 3) == 0)
+      return SetOK(eMacroKind_Neg, true, 1);
+    if (memcmp(Tok.start, "Ult", 3) == 0)
+      return SetOK(Expr::Ult, false, 2);
+    if (memcmp(Tok.start, "Ule", 3) == 0)
+      return SetOK(Expr::Ule, false, 2);
+    if (memcmp(Tok.start, "Ugt", 3) == 0)
+      return SetOK(Expr::Ugt, false, 2);
+    if (memcmp(Tok.start, "Uge", 3) == 0)
+      return SetOK(Expr::Uge, false, 2);
+    if (memcmp(Tok.start, "Slt", 3) == 0)
+      return SetOK(Expr::Slt, false, 2);
+    if (memcmp(Tok.start, "Sle", 3) == 0)
+      return SetOK(Expr::Sle, false, 2);
+    if (memcmp(Tok.start, "Sgt", 3) == 0)
+      return SetOK(Expr::Sgt, false, 2);
+    if (memcmp(Tok.start, "Sge", 3) == 0)
+      return SetOK(Expr::Sge, false, 2);
+    break;
+
+  case 4:
+    if (memcmp(Tok.start, "Read", 4) == 0)
+      return SetOK(Expr::Read, true, -1);
+    if (memcmp(Tok.start, "AShr", 4) == 0)
+      return SetOK(Expr::AShr, true, 2);
+    if (memcmp(Tok.start, "LShr", 4) == 0)
+      return SetOK(Expr::LShr, true, 2);
+
+    if (memcmp(Tok.start, "UDiv", 4) == 0)
+      return SetOK(Expr::UDiv, true, 2);
+    if (memcmp(Tok.start, "SDiv", 4) == 0)
+      return SetOK(Expr::SDiv, true, 2);
+    if (memcmp(Tok.start, "URem", 4) == 0)
+      return SetOK(Expr::URem, true, 2);
+    if (memcmp(Tok.start, "SRem", 4) == 0)
+      return SetOK(Expr::SRem, true, 2);
+    
+    if (memcmp(Tok.start, "SExt", 4) == 0)
+      return SetOK(Expr::SExt, false, 1);
+    if (memcmp(Tok.start, "ZExt", 4) == 0)
+      return SetOK(Expr::ZExt, false, 1);
+    break;
+    
+  case 6:
+    if (memcmp(Tok.start, "Concat", 6) == 0)
+      return SetOK(eMacroKind_Concat, false, -1); 
+    if (memcmp(Tok.start, "Select", 6) == 0)
+      return SetOK(Expr::Select, false, 3);
+    break;
+    
+  case 7:
+    if (memcmp(Tok.start, "Extract", 7) == 0)
+      return SetOK(Expr::Extract, false, -1);
+    if (memcmp(Tok.start, "ReadLSB", 7) == 0)
+      return SetOK(eMacroKind_ReadLSB, true, -1);
+    if (memcmp(Tok.start, "ReadMSB", 7) == 0)
+      return SetOK(eMacroKind_ReadMSB, true, -1);
+    break;
+  }
+
+  return false;
+#undef SetOK
+}
+
+/// ParseParenExpr - Parse a parenthesized expression with the given
+/// \arg ExpectedType. \arg ExpectedType can be invalid if the type
+/// cannot be inferred from the context.
+///
+/// paren-expr = '(' type number ')'
+/// paren-expr = '(' identifier [type] expr+ ')
+/// paren-expr = '(' ('Read' | 'ReadMSB' | 'ReadLSB') type expr update-list ')'
+ExprResult ParserImpl::ParseParenExpr(TypeResult FIXME_UNUSED) {
+  if (Tok.kind != Token::LParen) {
+    Error("unexpected token.");
+    ConsumeAnyToken();
+    return ExprResult();
+  }
+
+  ConsumeLParen();
+  
+  // Check for coercion case (w32 11).
+  if (Tok.kind == Token::KWWidth) {
+    TypeResult ExpectedType = ParseTypeSpecifier();
+
+    if (Tok.kind != Token::Number) {
+      Error("coercion can only apply to a number.");
+      SkipUntilRParen();
+      return ExprResult();
+    }
+    
+    // Make sure this was a type specifier we support.
+    ExprResult Res;
+    if (ExpectedType.isValid()) 
+      Res = ParseNumber(ExpectedType.get());
+    else
+      ConsumeToken();
+
+    ExpectRParen("unexpected argument in coercion.");  
+    return Res;
+  }
+  
+  if (Tok.kind != Token::Identifier) {
+    Error("unexpected token, expected expression.");
+    SkipUntilRParen();
+    return ExprResult();
+  }
+
+  Token Name = Tok;
+  ConsumeToken();
+
+  // FIXME: Use invalid type (i.e. width==0)?
+  Token TypeTok = Tok;
+  bool HasType = TypeTok.kind == Token::KWWidth;
+  TypeResult Type = HasType ? ParseTypeSpecifier() : Expr::Bool;
+
+  // FIXME: For now just skip to rparen on error. It might be nice
+  // to try and actually parse the child nodes though for error
+  // messages & better recovery?
+  if (!Type.isValid()) {
+    SkipUntilRParen();
+    return ExprResult();
+  }
+  Expr::Width ResTy = Type.get();
+
+  unsigned ExprKind;
+  bool IsFixed;
+  int NumArgs;
+  if (!LookupExprInfo(Name, ExprKind, IsFixed, NumArgs)) {
+    // FIXME: For now just skip to rparen on error. It might be nice
+    // to try and actually parse the child nodes though for error
+    // messages & better recovery?
+    Error("unknown expression kind.", Name);
+    SkipUntilRParen();
+    return ExprResult();
+  }
+
+  // See if we have to parse this form specially.
+  if (NumArgs == -1) {
+    switch (ExprKind) {
+    case eMacroKind_Concat:
+      return ParseConcatParenExpr(Name, ResTy);
+
+    case Expr::Extract:
+      return ParseExtractParenExpr(Name, ResTy);
+
+    case eMacroKind_ReadLSB:
+    case eMacroKind_ReadMSB:
+    case Expr::Read:
+      return ParseAnyReadParenExpr(Name, ExprKind, ResTy);
+
+    default:
+      Error("internal error, unimplemented special form.", Name);
+      SkipUntilRParen();
+      return ExprResult(ref<Expr>(0, ResTy));
+    }
+  }
+
+  switch (NumArgs) {
+  case 1:
+    return ParseUnaryParenExpr(Name, ExprKind, IsFixed, ResTy);
+  case 2:
+    return ParseBinaryParenExpr(Name, ExprKind, IsFixed, ResTy);
+  case 3:
+    if (ExprKind == Expr::Select)
+      return ParseSelectParenExpr(Name, ResTy);
+  default:
+    assert(0 && "Invalid argument kind (number of args).");
+    return ExprResult();
+  }
+}
+
+ExprResult ParserImpl::ParseUnaryParenExpr(const Token &Name,
+                                           unsigned Kind, bool IsFixed,
+                                           Expr::Width ResTy) {
+  if (Tok.kind == Token::RParen) {
+    Error("unexpected end of arguments.", Name);
+    ConsumeRParen();
+    return ref<Expr>(0, ResTy);
+  }
+
+  ExprResult Arg = ParseExpr(IsFixed ? ResTy : TypeResult());
+  if (!Arg.isValid())
+    Arg = ref<Expr>(0, ResTy);
+
+  ExpectRParen("unexpected argument in unary expression.");  
+  ExprHandle E = Arg.get();
+  switch (Kind) {
+  case eMacroKind_Not:
+    return EqExpr::alloc(ref<Expr>(0, E.getWidth()), E);
+  case eMacroKind_Neg:
+    return SubExpr::alloc(ref<Expr>(0, E.getWidth()), E);
+  case Expr::SExt:
+    // FIXME: Type check arguments.
+    return SExtExpr::alloc(E, ResTy);
+  case Expr::ZExt:
+    // FIXME: Type check arguments.
+    return ZExtExpr::alloc(E, ResTy);
+  default:
+    Error("internal error, unhandled kind.", Name);
+    return ref<Expr>(0, ResTy);
+  }
+}
+
+/// ParseMatchedBinaryArgs - Parse a pair of arguments who are
+/// expected to be of the same type. Upon return, if both LHS and RHS
+/// are valid then they are guaranteed to have the same type.
+///
+/// Name - The name token of the expression, for diagnostics.
+/// ExpectType - The expected type of the arguments, if known.
+void ParserImpl::ParseMatchedBinaryArgs(const Token &Name, 
+                                        TypeResult ExpectType,
+                                        ExprResult &LHS, ExprResult &RHS) {
+  if (Tok.kind == Token::RParen) {
+    Error("unexpected end of arguments.", Name);
+    ConsumeRParen();
+    return;
+  }
+
+  // Avoid NumberOrExprResult overhead and give more precise
+  // diagnostics when we know the type.
+  if (ExpectType.isValid()) {
+    LHS = ParseExpr(ExpectType);
+    if (Tok.kind == Token::RParen) {
+      Error("unexpected end of arguments.", Name);
+      ConsumeRParen();
+      return;
+    }
+    RHS = ParseExpr(ExpectType);
+  } else {
+    NumberOrExprResult LHS_NOE = ParseNumberOrExpr();
+
+    if (Tok.kind == Token::RParen) {
+      Error("unexpected end of arguments.", Name);
+      ConsumeRParen();
+      return;
+    }
+
+    if (LHS_NOE.isNumber()) {
+      NumberOrExprResult RHS_NOE = ParseNumberOrExpr();
+      
+      if (RHS_NOE.isNumber()) {
+        Error("ambiguous arguments to expression.", Name);
+      } else {
+        RHS = RHS_NOE.getExpr();
+        if (RHS.isValid())
+          LHS = ParseNumberToken(RHS.get().getWidth(), LHS_NOE.getNumber());
+      }
+    } else {
+      LHS = LHS_NOE.getExpr();
+      if (!LHS.isValid()) {
+        // FIXME: Should suppress ambiguity warnings here.
+        RHS = ParseExpr(TypeResult());
+      } else {
+        RHS = ParseExpr(LHS.get().getWidth());
+      }
+    }
+  }
+
+  ExpectRParen("unexpected argument to expression.");
+}
+
+ExprResult ParserImpl::ParseBinaryParenExpr(const Token &Name,
+                                           unsigned Kind, bool IsFixed,
+                                           Expr::Width ResTy) {
+  ExprResult LHS, RHS;
+  ParseMatchedBinaryArgs(Name, IsFixed ? TypeResult(ResTy) : TypeResult(), 
+                         LHS, RHS);
+  if (!LHS.isValid() || !RHS.isValid())
+    return ref<Expr>(0, ResTy);
+
+  ref<Expr> LHS_E = LHS.get(), RHS_E = RHS.get();
+  assert(LHS_E.getWidth() == RHS_E.getWidth() && "Mismatched types!");
+
+  switch (Kind) {    
+  case Expr::Add: return AddExpr::alloc(LHS_E, RHS_E);
+  case Expr::Sub: return SubExpr::alloc(LHS_E, RHS_E);
+  case Expr::Mul: return MulExpr::alloc(LHS_E, RHS_E);
+  case Expr::UDiv: return UDivExpr::alloc(LHS_E, RHS_E);
+  case Expr::SDiv: return SDivExpr::alloc(LHS_E, RHS_E);
+  case Expr::URem: return URemExpr::alloc(LHS_E, RHS_E);
+  case Expr::SRem: return SRemExpr::alloc(LHS_E, RHS_E);
+
+  case Expr::AShr: return AShrExpr::alloc(LHS_E, RHS_E);
+  case Expr::LShr: return LShrExpr::alloc(LHS_E, RHS_E);
+  case Expr::Shl: return AndExpr::alloc(LHS_E, RHS_E);
+
+  case Expr::And: return AndExpr::alloc(LHS_E, RHS_E);
+  case Expr::Or:  return OrExpr::alloc(LHS_E, RHS_E);
+  case Expr::Xor: return XorExpr::alloc(LHS_E, RHS_E);
+
+  case Expr::Eq:  return EqExpr::alloc(LHS_E, RHS_E);
+  case Expr::Ne:  return NeExpr::alloc(LHS_E, RHS_E);
+  case Expr::Ult: return UltExpr::alloc(LHS_E, RHS_E);
+  case Expr::Ule: return UleExpr::alloc(LHS_E, RHS_E);
+  case Expr::Ugt: return UgtExpr::alloc(LHS_E, RHS_E);
+  case Expr::Uge: return UgeExpr::alloc(LHS_E, RHS_E);
+  case Expr::Slt: return SltExpr::alloc(LHS_E, RHS_E);
+  case Expr::Sle: return SleExpr::alloc(LHS_E, RHS_E);
+  case Expr::Sgt: return SgtExpr::alloc(LHS_E, RHS_E);
+  case Expr::Sge: return SgeExpr::alloc(LHS_E, RHS_E);
+  default:
+    Error("FIXME: unhandled kind.", Name);
+    return ref<Expr>(0, ResTy);
+  }  
+}
+
+ExprResult ParserImpl::ParseSelectParenExpr(const Token &Name, 
+                                            Expr::Width ResTy) {
+  // FIXME: Why does this need to be here?
+  if (Tok.kind == Token::RParen) {
+    Error("unexpected end of arguments.", Name);
+    ConsumeRParen();
+    return ref<Expr>(0, ResTy);
+  }
+
+  ExprResult Cond = ParseExpr(Expr::Bool);
+  ExprResult LHS, RHS;
+  ParseMatchedBinaryArgs(Name, ResTy, LHS, RHS);
+  if (!Cond.isValid() || !LHS.isValid() || !RHS.isValid())
+    return ref<Expr>(0, ResTy);
+  return SelectExpr::alloc(Cond.get(), LHS.get(), RHS.get());
+}
+
+
+// need to decide if we want to allow n-ary Concat expressions in the
+// language
+ExprResult ParserImpl::ParseConcatParenExpr(const Token &Name,
+                                            Expr::Width ResTy) {
+  std::vector<ExprHandle> Kids;
+  
+  unsigned Width = 0;
+  while (Tok.kind != Token::RParen) {
+    ExprResult E = ParseExpr(TypeResult());
+
+    // Skip to end of expr on error.
+    if (!E.isValid()) {
+      SkipUntilRParen();
+      return ref<Expr>(0, ResTy);
+    }
+    
+    Kids.push_back(E.get());
+    Width += E.get().getWidth();
+  }
+  
+  ConsumeRParen();
+
+  if (Width != ResTy) {
+    Error("concat does not match expected result size.");
+    return ref<Expr>(0, ResTy);
+  }
+
+  return ConcatExpr::createN(Kids.size(), &Kids[0]);
+}
+
+ExprResult ParserImpl::ParseExtractParenExpr(const Token &Name,
+                                             Expr::Width ResTy) {
+  // FIXME: Pull out parse constant integer expression.
+  ExprResult OffsetExpr = ParseNumber(Expr::Int32);
+  ExprResult Child = ParseExpr(TypeResult());
+
+  ExpectRParen("unexpected argument to expression.");
+
+  if (!OffsetExpr.isValid() || !Child.isValid())
+    return ref<Expr>(0, ResTy);
+
+  assert(OffsetExpr.get().isConstant() && "ParseNumber returned non-constant.");
+  unsigned Offset = (unsigned) OffsetExpr.get().getConstantValue();
+
+  if (Offset + ResTy > Child.get().getWidth()) {
+    Error("extract out-of-range of child expression.", Name);
+    return ref<Expr>(0, ResTy);
+  }
+
+  return ExtractExpr::alloc(Child.get(), Offset, ResTy);
+}
+
+ExprResult ParserImpl::ParseAnyReadParenExpr(const Token &Name,
+                                             unsigned Kind,
+                                             Expr::Width ResTy) {
+  NumberOrExprResult Index = ParseNumberOrExpr();
+  VersionResult Array = ParseVersionSpecifier();
+  ExpectRParen("unexpected argument in read expression.");
+  
+  if (!Array.isValid())
+    return ref<Expr>(0, ResTy);
+
+  // FIXME: Need generic way to get array width. Needs to work with
+  // anonymous arrays.
+  Expr::Width ArrayDomainType = Expr::Int32;
+  Expr::Width ArrayRangeType = Expr::Int8;
+
+  // Coerce number to correct type.
+  ExprResult IndexExpr;
+  if (Index.isNumber())
+    IndexExpr = ParseNumberToken(ArrayDomainType, Index.getNumber());
+  else
+    IndexExpr = Index.getExpr();
+  
+  if (!IndexExpr.isValid())
+    return ref<Expr>(0, ResTy);
+  else if (IndexExpr.get().getWidth() != ArrayDomainType) {
+    Error("index width does not match array domain.");
+    return ref<Expr>(0, ResTy);
+  }
+
+  // FIXME: Check range width.
+
+  switch (Kind) {
+  default:
+    assert(0 && "Invalid kind.");
+    return ref<Expr>(0, ResTy);
+  case eMacroKind_ReadLSB:
+  case eMacroKind_ReadMSB: {
+    unsigned NumReads = ResTy / ArrayRangeType;
+    if (ResTy != NumReads*ArrayRangeType) {
+      Error("invalid ordered read (not multiple of range type).", Name);
+      return ref<Expr>(0, ResTy);
+    }
+    std::vector<ExprHandle> Kids;
+    Kids.reserve(NumReads);
+    ExprHandle Index = IndexExpr.get();
+    for (unsigned i=0; i<NumReads; ++i) {
+      // FIXME: using folding here
+      ExprHandle OffsetIndex = AddExpr::create(IndexExpr.get(),
+                                               ref<Expr>(i, ArrayDomainType));
+      Kids.push_back(ReadExpr::alloc(Array.get(), OffsetIndex));
+    }
+    if (Kind == eMacroKind_ReadLSB)
+      std::reverse(Kids.begin(), Kids.end());
+    return ConcatExpr::createN(NumReads, &Kids[0]);
+  }
+  case Expr::Read:
+    return ReadExpr::alloc(Array.get(), IndexExpr.get());
+  }
+}
+
+/// version-specifier = <identifier>
+/// version-specifier = [<identifier>:] [ version ]
+VersionResult ParserImpl::ParseVersionSpecifier() {
+  const Identifier *Label = 0;
+  if (Tok.kind == Token::Identifier) {
+    Token LTok = Tok;
+    Label = GetOrCreateIdentifier(Tok);
+    ConsumeToken();
+
+    // FIXME: hack: add array declarations and ditch this.
+    if (memcmp(Label->Name.c_str(), "arr", 3) == 0) {
+      // Declare or create array.
+      const ArrayDecl *&A = ArraySymTab[Label];
+      if (!A) {
+        //        Array = new ArrayDecl(Label, 0, 32, 8);
+        unsigned id = atoi(&Label->Name.c_str()[3]);
+        Array *root = new Array(0, id, 0);
+        // Create update list mapping of name -> array.
+        VersionSymTab.insert(std::make_pair(Label,
+                                            UpdateList(root, true, NULL)));
+      }
+    }
+
+    if (Tok.kind != Token::Colon) {
+      VersionSymTabTy::iterator it = VersionSymTab.find(Label);
+      
+      if (it == VersionSymTab.end()) {
+        Error("invalid update list label reference.", LTok);
+        return VersionResult(false,
+                             UpdateList(0, true, NULL));
+      }
+
+      return it->second;
+    }
+
+    ConsumeToken();
+    if (VersionSymTab.count(Label)) {
+      Error("duplicate update list label definition.", LTok);
+      Label = 0;
+    }
+  }
+
+  Token Start = Tok;
+  VersionResult Res = ParseVersion();
+  // Define update list to avoid use-of-undef errors.
+  if (!Res.isValid())
+    Res = VersionResult(false,
+                        UpdateList(0, true, NULL));
+  
+  if (Label)
+    VersionSymTab.insert(std::make_pair(Label, Res.get()));
+  return Res;
+}
+
+/// version - '[' update-list? ']' ['@' version-specifier]
+/// update-list - empty
+/// update-list - lhs '=' rhs [',' update-list]
+VersionResult ParserImpl::ParseVersion() {
+  if (Tok.kind != Token::LSquare)
+    return VersionResult(false, UpdateList(0, false, NULL));
+  
+  std::vector< std::pair<NumberOrExprResult, NumberOrExprResult> > Writes;
+  ConsumeLSquare();
+  for (;;) {
+    // FIXME: Type check exprs.
+
+    // FIXME: We need to do this (the above) anyway just to handle
+    // implicit constants correctly.
+    NumberOrExprResult LHS = ParseNumberOrExpr();
+    
+    if (Tok.kind != Token::Equals) {
+      Error("expected '='.", Tok);
+      break;
+    }
+    
+    ConsumeToken();
+    NumberOrExprResult RHS = ParseNumberOrExpr();
+
+    Writes.push_back(std::make_pair(LHS, RHS));
+    
+    if (Tok.kind == Token::Comma)
+      ConsumeToken();
+    else
+      break;
+  }
+  ExpectRSquare("expected close of update list");
+
+  VersionHandle Base(0, false, NULL);
+
+  // Anonymous array case.
+  if (Tok.kind != Token::At) { 
+    Array *root = new Array(0, 0, 0);
+    Base = UpdateList(root, false, NULL);
+  } else {
+    ConsumeToken();
+
+    VersionResult BaseRes = ParseVersionSpecifier();
+    if (!BaseRes.isValid())
+      return BaseRes;
+
+    Base = BaseRes.get();
+  }
+
+  Expr::Width ArrayDomainType = Expr::Int32;
+  Expr::Width ArrayRangeType = Expr::Int8;
+
+  for (std::vector< std::pair<NumberOrExprResult, NumberOrExprResult> >::reverse_iterator
+         it = Writes.rbegin(), ie = Writes.rend(); it != ie; ++it) {
+    ExprResult LHS, RHS;
+    // FIXME: This can be factored into common helper for coercing a
+    // NumberOrExpr into an Expr.
+    if (it->first.isNumber()) {
+      LHS = ParseNumberToken(ArrayDomainType, it->first.getNumber());
+    } else {
+      LHS = it->first.getExpr(); 
+      if (LHS.isValid() && LHS.get().getWidth() != ArrayDomainType) {
+        // FIXME: bad token location. We should maybe try and know the
+        // array up-front?
+        Error("invalid value in write index (doesn't match domain).", Tok);
+        LHS = ExprResult();
+      }
+    }
+
+    if (it->second.isNumber()) {
+      RHS = ParseNumberToken(ArrayRangeType, it->second.getNumber());
+    } else {
+      RHS = it->second.getExpr();
+      if (RHS.isValid() && RHS.get().getWidth() != ArrayRangeType) {
+        // FIXME: bad token location. We should maybe try and know the
+        // array up-front?
+        Error("invalid value in write assignment (doesn't match range).", Tok);
+        RHS = ExprResult();
+      }
+    }
+    
+    if (LHS.isValid() && RHS.isValid())
+      Base.extend(LHS.get(), RHS.get());
+  }
+
+  return Base;
+}
+
+/// ParseNumber - Parse a number of the given type.
+ExprResult ParserImpl::ParseNumber(Expr::Width Type) {
+  ExprResult Res = ParseNumberToken(Type, Tok);
+  ConsumeExpectedToken(Token::Number);
+  return Res;
+}
+
+/// ParseNumberToken - Parse a number of the given type from the given
+/// token.
+ExprResult ParserImpl::ParseNumberToken(Expr::Width Type, const Token &Tok) {
+  const char *S = Tok.start;
+  unsigned N = Tok.length;
+  unsigned Radix = 10, RadixBits = 4;
+  bool HasMinus = false;
+
+  // Detect +/- (a number token cannot have both).
+  if (S[0] == '+') {
+    ++S;
+    --N;
+  } else if (S[0] == '-') {
+    HasMinus = true;
+    ++S;
+    --N;
+  }
+
+  // Detect 0[box].
+  if ((Tok.length >= 2 && S[0] == '0') &&
+      (S[1] == 'b' || S[1] == 'o' || S[1] == 'x')) {
+    if (S[1] == 'b') {
+      Radix = 2; 
+      RadixBits = 1;
+    } else if (S[1] == 'o') {
+      Radix = 8;
+      RadixBits = 3;
+    } else {
+      Radix = 16;
+      RadixBits = 4;
+    }
+    S += 2;
+    N -= 2;
+
+    // Diagnose 0[box] with no trailing digits.
+    if (!N) {
+      Error("invalid numeric token (no digits).", Tok);
+      return ref<Expr>(0, Type);
+    }
+  }
+
+  // This is a simple but slow way to handle overflow.
+  APInt Val(std::max(64U, RadixBits * N), 0);
+  APInt RadixVal(Val.getBitWidth(), Radix);
+  APInt DigitVal(Val.getBitWidth(), 0);
+  for (unsigned i=0; i<N; ++i) {
+    unsigned Digit, Char = S[i];
+    
+    if (Char == '_')
+      continue;
+    
+    if ('0' <= Char && Char <= '9')
+      Digit = Char - '0';
+    else if ('a' <= Char && Char <= 'z')
+      Digit = Char - 'a' + 10;
+    else if ('A' <= Char && Char <= 'Z')
+      Digit = Char - 'A' + 10;
+    else {
+      Error("invalid character in numeric token.", Tok);
+      return ref<Expr>(0, Type);
+    }
+
+    if (Digit >= Radix) {
+      Error("invalid character in numeric token (out of range).", Tok);
+      return ref<Expr>(0, Type);
+    }
+
+    DigitVal = Digit;
+    Val = Val * RadixVal + DigitVal;
+  }
+
+  // FIXME: Actually do the check for overflow.
+  if (HasMinus)
+    Val = -Val;
+
+  return ExprResult(ref<Expr>(Val.trunc(Type).getZExtValue(), Type));
+}
+
+/// ParseTypeSpecifier - Parse a type specifier.
+///
+/// type = w[0-9]+
+TypeResult ParserImpl::ParseTypeSpecifier() {
+  assert(Tok.kind == Token::KWWidth && "Unexpected token.");
+
+  // FIXME: Need APInt technically.
+  Token TypeTok = Tok;
+  int width = atoi(std::string(Tok.start+1,Tok.length-1).c_str());
+  ConsumeToken();
+
+  // FIXME: We should impose some sort of maximum just for sanity?
+  return TypeResult(width);
+}
+
+void ParserImpl::Error(const char *Message, const Token &At) {
+  ++NumErrors;
+  if (MaxErrors && NumErrors >= MaxErrors)
+    return;
+
+  llvm::cerr << Filename 
+             << ":" << At.line << ":" << At.column 
+             << ": error: " << Message << "\n";
+
+  // Skip carat diagnostics on EOF token.
+  if (At.kind == Token::EndOfFile)
+    return;
+  
+  // Simple caret style diagnostics.
+  const char *LineBegin = At.start, *LineEnd = At.start,
+    *BufferBegin = TheMemoryBuffer->getBufferStart(),
+    *BufferEnd = TheMemoryBuffer->getBufferEnd();
+
+  // Run line pointers forward and back.
+  while (LineBegin > BufferBegin && 
+         LineBegin[-1] != '\r' && LineBegin[-1] != '\n')
+    --LineBegin;
+  while (LineEnd < BufferEnd && 
+         LineEnd[0] != '\r' && LineEnd[0] != '\n')
+    ++LineEnd;
+
+  // Show the line.
+  llvm::cerr << std::string(LineBegin, LineEnd) << "\n";
+
+  // Show the caret or squiggly, making sure to print back spaces the
+  // same.
+  for (const char *S=LineBegin; S != At.start; ++S)
+    llvm::cerr << (isspace(*S) ? *S : ' ');
+  if (At.length > 1) {
+    for (unsigned i=0; i<At.length; ++i)
+      llvm::cerr << '~';
+  } else
+    llvm::cerr << '^';
+  llvm::cerr << '\n';
+}
+
+// AST API
+// FIXME: Move out of parser.
+
+Decl::Decl() {}
+
+void QueryCommand::dump() {
+  // FIXME: This is masking the difference between an actual query and
+  // a query decl.
+  ExprPPrinter::printQuery(std::cerr, 
+                           ConstraintManager(Constraints), 
+                           Query);
+}
+
+// Public parser API
+
+Parser::Parser() {
+}
+
+Parser::~Parser() {
+}
+
+Parser *Parser::Create(const std::string Filename,
+                       const MemoryBuffer *MB) {
+  ParserImpl *P = new ParserImpl(Filename, MB);
+  P->Initialize();
+  return P;
+}

Added: klee/trunk/lib/Expr/Updates.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Expr/Updates.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Expr/Updates.cpp (added)
+++ klee/trunk/lib/Expr/Updates.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,126 @@
+//===-- Updates.cpp -------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Expr.h"
+
+#include <cassert>
+
+using namespace klee;
+
+///
+
+UpdateNode::UpdateNode(const UpdateNode *_next, 
+                       const ref<Expr> &_index, 
+                       const ref<Expr> &_value) 
+  : refCount(0),
+    stpArray(0),
+    next(_next),
+    index(_index),
+    value(_value) {
+  assert(_value.getWidth() == Expr::Int8 && "Update value should be 8-bit wide.");
+  computeHash();
+  if (next) {
+    ++next->refCount;
+    size = 1 + next->size;
+  }
+  else size = 1;
+}
+
+extern "C" void vc_DeleteExpr(void*);
+
+UpdateNode::~UpdateNode() {
+  // XXX gross
+  if (stpArray)
+    ::vc_DeleteExpr(stpArray);
+}
+
+int UpdateNode::compare(const UpdateNode &b) const {
+  if (int i = index.compare(b.index)) 
+    return i;
+  return value.compare(b.value);
+}
+
+unsigned UpdateNode::computeHash() {
+  hashValue = index.hash() ^ value.hash();
+  if (next)
+    hashValue ^= next->hash();
+  return hashValue;
+}
+
+///
+
+UpdateList::UpdateList(const Array *_root, bool _isRooted,
+                       const UpdateNode *_head)
+  : root(_root),
+    head(_head),
+    isRooted(_isRooted) {
+  if (head) ++head->refCount;
+}
+
+UpdateList::UpdateList(const UpdateList &b)
+  : root(b.root),
+    head(b.head),
+    isRooted(b.isRooted) {
+  if (head) ++head->refCount;
+}
+
+UpdateList::~UpdateList() {
+  // We need to be careful and avoid recursion here. We do this in
+  // cooperation with the private dtor of UpdateNode which does not
+  // recursively free its tail.
+  while (head && --head->refCount==0) {
+    const UpdateNode *n = head->next;
+    delete head;
+    head = n;
+  }
+}
+
+UpdateList &UpdateList::operator=(const UpdateList &b) {
+  if (b.head) ++b.head->refCount;
+  if (head && --head->refCount==0) delete head;
+  root = b.root;
+  head = b.head;
+  isRooted = b.isRooted;
+  return *this;
+}
+
+void UpdateList::extend(const ref<Expr> &index, const ref<Expr> &value) {
+  if (head) --head->refCount;
+  head = new UpdateNode(head, index, value);
+  ++head->refCount;
+}
+
+int UpdateList::compare(const UpdateList &b) const {
+  // use object id to increase determinism
+  if (root->id != b.root->id) 
+    return root->id < b.root->id ? -1 : 1;
+  
+  if (getSize() < b.getSize()) return -1;
+  else if (getSize() > b.getSize()) return 1;    
+
+  // XXX build comparison into update, make fast
+  const UpdateNode *an=head, *bn=b.head;
+  for (; an && bn; an=an->next,bn=bn->next) {
+    if (an==bn) { // exploit shared list structure
+      return 0;
+    } else {
+      if (int res = an->compare(*bn))
+        return res;
+    }
+  }
+  assert(!an && !bn);
+  return 0;
+}
+
+unsigned UpdateList::hash() const {
+  unsigned res = root->id * Expr::MAGIC_HASH_CONSTANT;
+  if (head)
+    res ^= head->hash();
+  return res;
+}

Added: klee/trunk/lib/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Makefile (added)
+++ klee/trunk/lib/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,15 @@
+#===-- lib/Makefile ----------------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+LEVEL=..
+
+PARALLEL_DIRS=Basic Support Expr Solver Module Core
+
+include $(LEVEL)/Makefile.common
+

Added: klee/trunk/lib/Module/Checks.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/Checks.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/Checks.cpp (added)
+++ klee/trunk/lib/Module/Checks.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,68 @@
+//===-- Checks.cpp --------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Passes.h"
+
+#include "llvm/Constants.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/Function.h"
+#include "llvm/InstrTypes.h"
+#include "llvm/Instruction.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/Module.h"
+#include "llvm/Pass.h"
+#include "llvm/Type.h"
+#include "llvm/Transforms/Scalar.h"
+#include "llvm/Transforms/Utils/BasicBlockUtils.h"
+#include "llvm/Target/TargetData.h"
+
+using namespace llvm;
+using namespace klee;
+
+char DivCheckPass::ID;
+
+bool DivCheckPass::runOnModule(Module &M) { 
+  Function *divZeroCheckFunction = 0;
+
+  bool moduleChanged = false;
+  
+  for (Module::iterator f = M.begin(), fe = M.end(); f != fe; ++f) {
+    for (Function::iterator b = f->begin(), be = f->end(); b != be; ++b) {
+      for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {     
+          if (BinaryOperator* binOp = dyn_cast<BinaryOperator>(i)) {
+          // find all [s|u][div|mod] instructions
+          Instruction::BinaryOps opcode = binOp->getOpcode();
+          if (opcode == Instruction::SDiv || opcode == Instruction::UDiv ||
+              opcode == Instruction::SRem || opcode == Instruction::URem) {
+            
+            CastInst *denominator =
+              CastInst::CreateIntegerCast(i->getOperand(1),
+                                          (Type*)Type::Int64Ty,
+                                          false,  /* sign doesn't matter */
+                                          "int_cast_to_i64",
+                                          i);
+            
+            // Lazily bind the function to avoid always importing it.
+            if (!divZeroCheckFunction) {
+              Constant *fc = M.getOrInsertFunction("klee_div_zero_check", 
+                                                   Type::VoidTy, 
+                                                   Type::Int64Ty, NULL);
+              divZeroCheckFunction = cast<Function>(fc);
+            }
+
+	    CallInst::Create(divZeroCheckFunction, denominator, "", &*i);
+            moduleChanged = true;
+          }
+        }
+      }
+    }
+  }
+  return moduleChanged;
+}

Added: klee/trunk/lib/Module/InstructionInfoTable.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/InstructionInfoTable.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/InstructionInfoTable.cpp (added)
+++ klee/trunk/lib/Module/InstructionInfoTable.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,196 @@
+//===-- InstructionInfoTable.cpp ------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Internal/Module/InstructionInfoTable.h"
+
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/Linker.h"
+#include "llvm/Module.h"
+#include "llvm/Assembly/AsmAnnotationWriter.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/Support/InstIterator.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Analysis/ValueTracking.h"
+
+#include <map>
+#include <iostream>
+#include <fstream>
+#include <sstream>
+#include <string>
+
+using namespace llvm;
+using namespace klee;
+
+class InstructionToLineAnnotator : public llvm::AssemblyAnnotationWriter {
+public:
+  void emitInstructionAnnot(const Instruction *i, llvm::raw_ostream &os) {
+    os << "%%%" << (uintptr_t) i;
+  }
+};
+        
+static void buildInstructionToLineMap(Module *m,
+                                      std::map<const Instruction*, unsigned> &out) {  
+  InstructionToLineAnnotator a;
+  std::ostringstream buffer;
+  m->print(buffer, &a);
+  std::string str = buffer.str();
+  const char *s;
+
+  unsigned line = 1;
+  for (s=str.c_str(); *s; s++) {
+    if (*s=='\n') {
+      line++;
+      if (s[1]=='%' && s[2]=='%' && s[3]=='%') {
+        s += 4;
+        char *end;
+        unsigned long long value = strtoull(s, &end, 10);
+        if (end!=s) {
+          out.insert(std::make_pair((const Instruction*) value, line));
+        }
+        s = end;
+      }
+    }
+  }
+}
+
+static std::string getDSPIPath(DbgStopPointInst *dspi) {
+  std::string dir, file;
+  bool res = GetConstantStringInfo(dspi->getDirectory(), dir);
+  assert(res && "GetConstantStringInfo failed");
+  res = GetConstantStringInfo(dspi->getFileName(), file);
+  assert(res && "GetConstantStringInfo failed");
+  if (dir.empty()) {
+    return file;
+  } else if (*dir.rbegin() == '/') {
+    return dir + file;
+  } else {
+    return dir + "/" + file;
+  }
+}
+
+InstructionInfoTable::InstructionInfoTable(Module *m) 
+  : dummyString(""), dummyInfo(0, dummyString, 0, 0) {
+  unsigned id = 0;
+  std::map<const Instruction*, unsigned> lineTable;
+  buildInstructionToLineMap(m, lineTable);
+
+  for (Module::iterator fnIt = m->begin(), fn_ie = m->end(); 
+       fnIt != fn_ie; ++fnIt) {
+    const std::string *initialFile = &dummyString;
+    unsigned initialLine = 0;
+
+    // It may be better to look for the closest stoppoint to the entry
+    // following the CFG, but it is not clear that it ever matters in
+    // practice.
+    for (inst_iterator it = inst_begin(fnIt), ie = inst_end(fnIt);
+         it != ie; ++it) {
+      if (DbgStopPointInst *dspi = dyn_cast<DbgStopPointInst>(&*it)) {
+        initialFile = internString(getDSPIPath(dspi));
+        initialLine = dspi->getLine();
+        break;
+      }
+    }
+    
+    typedef std::map<BasicBlock*, std::pair<const std::string*,unsigned> > 
+      sourceinfo_ty;
+    sourceinfo_ty sourceInfo;
+    for (llvm::Function::iterator bbIt = fnIt->begin(), bbie = fnIt->end(); 
+         bbIt != bbie; ++bbIt) {
+      std::pair<sourceinfo_ty::iterator, bool>
+        res = sourceInfo.insert(std::make_pair(bbIt,
+                                               std::make_pair(initialFile,
+                                                              initialLine)));
+      if (!res.second)
+        continue;
+
+      std::vector<BasicBlock*> worklist;
+      worklist.push_back(bbIt);
+
+      do {
+        BasicBlock *bb = worklist.back();
+        worklist.pop_back();
+
+        sourceinfo_ty::iterator si = sourceInfo.find(bb);
+        assert(si != sourceInfo.end());
+        const std::string *file = si->second.first;
+        unsigned line = si->second.second;
+        
+        for (BasicBlock::iterator it = bb->begin(), ie = bb->end();
+             it != ie; ++it) {
+          Instruction *instr = it;
+          unsigned assemblyLine = 0;
+          std::map<const Instruction*, unsigned>::const_iterator ltit = 
+            lineTable.find(instr);
+          if (ltit!=lineTable.end())
+            assemblyLine = ltit->second;
+          if (DbgStopPointInst *dspi = dyn_cast<DbgStopPointInst>(instr)) {
+            file = internString(getDSPIPath(dspi));
+            line = dspi->getLine();
+          }
+          infos.insert(std::make_pair(instr,
+                                      InstructionInfo(id++,
+                                                      *file,
+                                                      line,
+                                                      assemblyLine)));        
+        }
+        
+        for (succ_iterator it = succ_begin(bb), ie = succ_end(bb); 
+             it != ie; ++it) {
+          if (sourceInfo.insert(std::make_pair(*it,
+                                               std::make_pair(file, line))).second)
+            worklist.push_back(*it);
+        }
+      } while (!worklist.empty());
+    }
+  }
+}
+
+InstructionInfoTable::~InstructionInfoTable() {
+  for (std::set<const std::string *, ltstr>::iterator
+         it = internedStrings.begin(), ie = internedStrings.end();
+       it != ie; ++it)
+    delete *it;
+}
+
+const std::string *InstructionInfoTable::internString(std::string s) {
+  std::set<const std::string *, ltstr>::iterator it = internedStrings.find(&s);
+  if (it==internedStrings.end()) {
+    std::string *interned = new std::string(s);
+    internedStrings.insert(interned);
+    return interned;
+  } else {
+    return *it;
+  }
+}
+
+unsigned InstructionInfoTable::getMaxID() const {
+  return infos.size();
+}
+
+const InstructionInfo &
+InstructionInfoTable::getInfo(const Instruction *inst) const {
+  std::map<const llvm::Instruction*, InstructionInfo>::const_iterator it = 
+    infos.find(inst);
+  if (it==infos.end()) {
+    return dummyInfo;
+  } else {
+    return it->second;
+  }
+}
+
+const InstructionInfo &
+InstructionInfoTable::getFunctionInfo(const Function *f) const {
+  if (f->isDeclaration()) {
+    return dummyInfo;
+  } else {
+    return getInfo(f->begin()->begin());
+  }
+}

Added: klee/trunk/lib/Module/IntrinsicCleaner.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/IntrinsicCleaner.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/IntrinsicCleaner.cpp (added)
+++ klee/trunk/lib/Module/IntrinsicCleaner.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,119 @@
+//===-- IntrinsicCleaner.cpp ----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Passes.h"
+
+#include "llvm/Constants.h"
+#include "llvm/DerivedTypes.h"
+#include "llvm/Function.h"
+#include "llvm/InstrTypes.h"
+#include "llvm/Instruction.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/Module.h"
+#include "llvm/Pass.h"
+#include "llvm/Type.h"
+#include "llvm/Transforms/Scalar.h"
+#include "llvm/Transforms/Utils/BasicBlockUtils.h"
+#include "llvm/Target/TargetData.h"
+
+using namespace llvm;
+
+namespace klee {
+
+char IntrinsicCleanerPass::ID;
+
+bool IntrinsicCleanerPass::runOnModule(Module &M) {
+  bool dirty = false;
+  for (Module::iterator f = M.begin(), fe = M.end(); f != fe; ++f)
+    for (Function::iterator b = f->begin(), be = f->end(); b != be; ++b)
+        dirty |= runOnBasicBlock(*b);
+  return dirty;
+}
+
+bool IntrinsicCleanerPass::runOnBasicBlock(BasicBlock &b) { 
+  bool dirty = false;
+
+  for (BasicBlock::iterator i = b.begin(), ie = b.end(); i != ie;) {     
+    IntrinsicInst *ii = dyn_cast<IntrinsicInst>(&*i);
+    // increment now since LowerIntrinsic deletion makes iterator invalid.
+    ++i;  
+    if(ii) {
+      switch (ii->getIntrinsicID()) {
+      case Intrinsic::vastart:
+      case Intrinsic::vaend:
+        break;
+        
+        // Lower vacopy so that object resolution etc is handled by
+        // normal instructions.  FIXME: This is broken for non-x86_32.
+      case Intrinsic::vacopy: { // (dst, src) -> *((i8**) dst) = *((i8**) src)
+        Value *dst = ii->getOperand(1);
+        Value *src = ii->getOperand(2);
+        Type *i8pp = PointerType::getUnqual(PointerType::getUnqual(Type::Int8Ty));
+        Value *castedDst = CastInst::CreatePointerCast(dst, i8pp, "vacopy.cast.dst", ii);
+        Value *castedSrc = CastInst::CreatePointerCast(src, i8pp, "vacopy.cast.src", ii);
+        Value *load = new LoadInst(castedSrc, "vacopy.read", ii);
+        new StoreInst(load, castedDst, false, ii);
+        ii->removeFromParent();
+        delete ii;
+        break;
+      }
+
+      case Intrinsic::dbg_stoppoint: {
+        // We can remove this stoppoint if the next instruction is
+        // sure to be another stoppoint. This is nice for cleanliness
+        // but also important for switch statements where it can allow
+        // the targets to be joined.
+        bool erase = false;
+        if (isa<DbgStopPointInst>(i) ||
+            isa<UnreachableInst>(i)) {
+          erase = true;
+        } else if (isa<BranchInst>(i) ||
+                   isa<SwitchInst>(i)) {
+          BasicBlock *bb = i->getParent();
+          erase = true;
+          for (succ_iterator it=succ_begin(bb), ie=succ_end(bb);
+               it!=ie; ++it) {
+            if (!isa<DbgStopPointInst>(it->getFirstNonPHI())) {
+              erase = false;
+              break;
+            }
+          }
+        }
+
+        if (erase) {
+          ii->eraseFromParent();
+          dirty = true;
+        }
+        break;
+      }
+
+      case Intrinsic::dbg_region_start:
+      case Intrinsic::dbg_region_end:
+      case Intrinsic::dbg_func_start:
+      case Intrinsic::dbg_declare:
+        // Remove these regardless of lower intrinsics flag. This can
+        // be removed once IntrinsicLowering is fixed to not have bad
+        // caches.
+        ii->eraseFromParent();
+        dirty = true;
+        break;
+                    
+      default:
+        if (LowerIntrinsics)
+          IL->LowerIntrinsicCall(ii);
+        dirty = true;
+        break;
+      }
+    }
+  }
+
+  return dirty;
+}
+}

Added: klee/trunk/lib/Module/KInstruction.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/KInstruction.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/KInstruction.cpp (added)
+++ klee/trunk/lib/Module/KInstruction.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,19 @@
+//===-- KInstruction.cpp --------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Internal/Module/KInstruction.h"
+
+using namespace llvm;
+using namespace klee;
+
+/***/
+
+KInstruction::~KInstruction() {
+  delete[] operands;
+}

Added: klee/trunk/lib/Module/KModule.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/KModule.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/KModule.cpp (added)
+++ klee/trunk/lib/Module/KModule.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,506 @@
+//===-- KModule.cpp -------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+// FIXME: This does not belong here.
+#include "../Core/Common.h"
+
+#include "klee/Internal/Module/KModule.h"
+
+#include "Passes.h"
+
+#include "klee/Interpreter.h"
+#include "klee/Internal/Module/Cell.h"
+#include "klee/Internal/Module/KInstruction.h"
+#include "klee/Internal/Module/InstructionInfoTable.h"
+#include "klee/Internal/Support/ModuleUtil.h"
+
+#include "llvm/Bitcode/ReaderWriter.h"
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+#include "llvm/PassManager.h"
+#include "llvm/ValueSymbolTable.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/System/Path.h"
+#include "llvm/Target/TargetData.h"
+#include "llvm/Transforms/Scalar.h"
+
+#include <sstream>
+
+using namespace llvm;
+using namespace klee;
+
+namespace {
+  enum SwitchImplType {
+    eSwitchTypeSimple,
+    eSwitchTypeLLVM,
+    eSwitchTypeInternal
+  };
+
+  cl::list<std::string>
+  MergeAtExit("merge-at-exit");
+    
+  cl::opt<bool>
+  NoTruncateSourceLines("no-truncate-source-lines",
+                        cl::desc("Don't truncate long lines in the output source"));
+
+  cl::opt<bool>
+  OutputSource("output-source",
+               cl::desc("Write the assembly for the final transformed source"),
+               cl::init(true));
+
+  cl::opt<bool>
+  OutputModule("output-module",
+               cl::desc("Write the bitcode for the final transformed module"),
+               cl::init(false));
+
+  cl::opt<SwitchImplType>
+  SwitchType("switch-type", cl::desc("Select the implementation of switch"),
+             cl::values(clEnumValN(eSwitchTypeSimple, "simple", 
+                                   "lower to ordered branches"),
+                        clEnumValN(eSwitchTypeLLVM, "llvm", 
+                                   "lower using LLVM"),
+                        clEnumValN(eSwitchTypeInternal, "internal", 
+                                   "execute switch internally"),
+                        clEnumValEnd),
+             cl::init(eSwitchTypeInternal));
+  
+  cl::opt<bool>
+  DebugPrintEscapingFunctions("debug-print-escaping-functions", 
+                              cl::desc("Print functions whose address is taken."));
+}
+
+KModule::KModule(Module *_module) 
+  : module(_module),
+    targetData(new TargetData(module)),
+    dbgStopPointFn(0),
+    kleeMergeFn(0),
+    infos(0),
+    constantTable(0) {
+}
+
+KModule::~KModule() {
+  delete[] constantTable;
+  delete infos;
+
+  for (std::vector<KFunction*>::iterator it = functions.begin(), 
+         ie = functions.end(); it != ie; ++it)
+    delete *it;
+
+  delete targetData;
+  delete module;
+}
+
+/***/
+
+namespace llvm {
+extern void Optimize(Module*);
+}
+
+// what a hack
+static Function *getStubFunctionForCtorList(Module *m,
+                                            GlobalVariable *gv, 
+                                            std::string name) {
+  assert(!gv->isDeclaration() && !gv->hasInternalLinkage() &&
+         "do not support old LLVM style constructor/destructor lists");
+  
+  std::vector<const Type*> nullary;
+
+  Function *fn = Function::Create(FunctionType::get(Type::VoidTy, 
+						    nullary, false),
+				  GlobalVariable::InternalLinkage, 
+				  name,
+                              m);
+  BasicBlock *bb = BasicBlock::Create("entry", fn);
+  
+  // From lli:
+  // Should be an array of '{ int, void ()* }' structs.  The first value is
+  // the init priority, which we ignore.
+  ConstantArray *arr = dyn_cast<ConstantArray>(gv->getInitializer());
+  if (arr) {
+    for (unsigned i=0; i<arr->getNumOperands(); i++) {
+      ConstantStruct *cs = cast<ConstantStruct>(arr->getOperand(i));
+      assert(cs->getNumOperands()==2 && "unexpected element in ctor initializer list");
+      
+      Constant *fp = cs->getOperand(1);      
+      if (!fp->isNullValue()) {
+        if (llvm::ConstantExpr *ce = dyn_cast<llvm::ConstantExpr>(fp))
+          fp = ce->getOperand(0);
+
+        if (Function *f = dyn_cast<Function>(fp)) {
+	  CallInst::Create(f, "", bb);
+        } else {
+          assert(0 && "unable to get function pointer from ctor initializer list");
+        }
+      }
+    }
+  }
+  
+  ReturnInst::Create(bb);
+
+  return fn;
+}
+
+static void injectStaticConstructorsAndDestructors(Module *m) {
+  GlobalVariable *ctors = m->getNamedGlobal("llvm.global_ctors");
+  GlobalVariable *dtors = m->getNamedGlobal("llvm.global_dtors");
+  
+  if (ctors || dtors) {
+    Function *mainFn = m->getFunction("main");
+    assert(mainFn && "unable to find main function");
+
+    if (ctors)
+    CallInst::Create(getStubFunctionForCtorList(m, ctors, "klee.ctor_stub"),
+		     "", mainFn->begin()->begin());
+    if (dtors) {
+      Function *dtorStub = getStubFunctionForCtorList(m, dtors, "klee.dtor_stub");
+      for (Function::iterator it = mainFn->begin(), ie = mainFn->end();
+           it != ie; ++it) {
+        if (isa<ReturnInst>(it->getTerminator()))
+	  CallInst::Create(dtorStub, "", it->getTerminator());
+      }
+    }
+  }
+}
+
+static void forceImport(Module *m, const char *name, const Type *retType, ...) {
+  // If module lacks an externally visible symbol for the name then we
+  // need to create one. We have to look in the symbol table because
+  // we want to check everything (global variables, functions, and
+  // aliases).
+
+  Value *v = m->getValueSymbolTable().lookup(name);
+  GlobalValue *gv = dyn_cast_or_null<GlobalValue>(v);
+
+  if (!gv || gv->hasInternalLinkage()) {
+    va_list ap;
+
+    va_start(ap, retType);
+    std::vector<const Type *> argTypes;
+    while (const Type *t = va_arg(ap, const Type*))
+      argTypes.push_back(t);
+    va_end(ap);
+
+    m->getOrInsertFunction(name, FunctionType::get(retType, argTypes, false));
+  }
+}
+
+void KModule::prepare(const Interpreter::ModuleOptions &opts,
+                      InterpreterHandler *ih) {
+  if (!MergeAtExit.empty()) {
+    Function *mergeFn = module->getFunction("klee_merge");
+    if (!mergeFn) {
+      const llvm::FunctionType *Ty = 
+        FunctionType::get(Type::VoidTy, std::vector<const Type*>(), false);
+      mergeFn = Function::Create(Ty, GlobalVariable::ExternalLinkage,
+				 "klee_merge",
+				 module);
+    }
+
+    for (cl::list<std::string>::iterator it = MergeAtExit.begin(), 
+           ie = MergeAtExit.end(); it != ie; ++it) {
+      std::string &name = *it;
+      Function *f = module->getFunction(name);
+      if (!f) {
+        klee_error("cannot insert merge-at-exit for: %s (cannot find)",
+                   name.c_str());
+      } else if (f->isDeclaration()) {
+        klee_error("cannot insert merge-at-exit for: %s (external)",
+                   name.c_str());
+      }
+
+      BasicBlock *exit = BasicBlock::Create("exit", f);
+      PHINode *result = 0;
+      if (f->getReturnType() != Type::VoidTy)
+        result = PHINode::Create(f->getReturnType(), "retval", exit);
+      CallInst::Create(mergeFn, "", exit);
+      ReturnInst::Create(result, exit);
+
+      llvm::cerr << "KLEE: adding klee_merge at exit of: " << name << "\n";
+      for (llvm::Function::iterator bbit = f->begin(), bbie = f->end(); 
+           bbit != bbie; ++bbit) {
+        if (&*bbit != exit) {
+          Instruction *i = bbit->getTerminator();
+          if (i->getOpcode()==Instruction::Ret) {
+            if (result) {
+              result->addIncoming(i->getOperand(0), bbit);
+            }
+            i->eraseFromParent();
+	    BranchInst::Create(exit, bbit);
+          }
+        }
+      }
+    }
+  }
+
+  // Inject checks prior to optimization... we also perform the
+  // invariant transformations that we will end up doing later so that
+  // optimize is seeing what is as close as possible to the final
+  // module.
+  PassManager pm;
+  pm.add(new RaiseAsmPass());
+  if (opts.CheckDivZero) pm.add(new DivCheckPass());
+  // FIXME: This false here is to work around a bug in
+  // IntrinsicLowering which caches values which may eventually be
+  // deleted (via RAUW). This can be removed once LLVM fixes this
+  // issue.
+  pm.add(new IntrinsicCleanerPass(*targetData, false));
+  pm.run(*module);
+
+  if (opts.Optimize)
+    Optimize(module);
+
+  // Force importing functions required by intrinsic lowering. Kind of
+  // unfortunate clutter when we don't need them but we won't know
+  // that until after all linking and intrinsic lowering is
+  // done. After linking and passes we just try to manually trim these
+  // by name. We only add them if such a function doesn't exist to
+  // avoid creating stale uses.
+
+  forceImport(module, "memcpy", PointerType::getUnqual(Type::Int8Ty),
+              PointerType::getUnqual(Type::Int8Ty),
+              PointerType::getUnqual(Type::Int8Ty),
+              targetData->getIntPtrType(), (Type*) 0);
+  forceImport(module, "memmove", PointerType::getUnqual(Type::Int8Ty),
+              PointerType::getUnqual(Type::Int8Ty),
+              PointerType::getUnqual(Type::Int8Ty),
+              targetData->getIntPtrType(), (Type*) 0);
+  forceImport(module, "memset", PointerType::getUnqual(Type::Int8Ty),
+              PointerType::getUnqual(Type::Int8Ty),
+              Type::Int32Ty,
+              targetData->getIntPtrType(), (Type*) 0);
+
+  // FIXME: Missing force import for various math functions.
+
+  // FIXME: Find a way that we can test programs without requiring
+  // this to be linked in, it makes low level debugging much more
+  // annoying.
+  llvm::sys::Path path(opts.LibraryDir);
+  path.appendComponent("libintrinsic.bca");
+  module = linkWithLibrary(module, path.c_str());
+
+  // Needs to happen after linking (since ctors/dtors can be modified)
+  // and optimization (since global optimization can rewrite lists).
+  injectStaticConstructorsAndDestructors(module);
+
+  // Finally, run the passes that maintain invariants we expect during
+  // interpretation. We run the intrinsic cleaner just in case we
+  // linked in something with intrinsics but any external calls are
+  // going to be unresolved. We really need to handle the intrinsics
+  // directly I think?
+  PassManager pm3;
+  pm3.add(createCFGSimplificationPass());
+  switch(SwitchType) {
+  case eSwitchTypeInternal: break;
+  case eSwitchTypeSimple: pm3.add(new LowerSwitchPass()); break;
+  case eSwitchTypeLLVM:  pm3.add(createLowerSwitchPass()); break;
+  default: klee_error("invalid --switch-type");
+  }
+  pm3.add(new IntrinsicCleanerPass(*targetData));
+  pm3.add(new PhiCleanerPass());
+  pm3.run(*module);
+
+  // For cleanliness see if we can discard any of the functions we
+  // forced to import.
+  Function *f;
+  f = module->getFunction("memcpy");
+  if (f && f->use_empty()) f->eraseFromParent();
+  f = module->getFunction("memmove");
+  if (f && f->use_empty()) f->eraseFromParent();
+  f = module->getFunction("memset");
+  if (f && f->use_empty()) f->eraseFromParent();
+
+
+  // Write out the .ll assembly file. We truncate long lines to work
+  // around a kcachegrind parsing bug (it puts them on new lines), so
+  // that source browsing works.
+  if (OutputSource) {
+    std::ostream *os = ih->openOutputFile("assembly.ll");
+    assert(os && os->good() && "unable to open source output");
+
+    // We have an option for this in case the user wants a .ll they
+    // can compile.
+    if (NoTruncateSourceLines) {
+      *os << *module;
+    } else {
+      bool truncated = false;
+      std::stringstream buffer;
+      buffer << *module;
+      std::string string = buffer.str();
+      const char *position = string.c_str();
+
+      for (;;) {
+        const char *end = index(position, '\n');
+        if (!end) {
+          *os << position;
+          break;
+        } else {
+          unsigned count = (end - position) + 1;
+          if (count<255) {
+            os->write(position, count);
+          } else {
+            os->write(position, 254);
+            *os << "\n";
+            truncated = true;
+          }
+          position = end+1;
+        }
+      }
+    }
+
+    delete os;
+  }
+
+  if (OutputModule) {
+    std::ostream *f = ih->openOutputFile("final.bc");
+    WriteBitcodeToFile(module, *f);
+    delete f;
+  }
+
+  dbgStopPointFn = module->getFunction("llvm.dbg.stoppoint");
+  kleeMergeFn = module->getFunction("klee_merge");
+
+  /* Build shadow structures */
+
+  infos = new InstructionInfoTable(module);  
+  
+  for (Module::iterator it = module->begin(), ie = module->end();
+       it != ie; ++it) {
+    if (it->isDeclaration())
+      continue;
+
+    KFunction *kf = new KFunction(it, this);
+    
+    for (unsigned i=0; i<kf->numInstructions; ++i) {
+      KInstruction *ki = kf->instructions[i];
+      ki->info = &infos->getInfo(ki->inst);
+    }
+
+    functions.push_back(kf);
+    functionMap.insert(std::make_pair(it, kf));
+  }
+
+  /* Compute various interesting properties */
+
+  for (std::vector<KFunction*>::iterator it = functions.begin(), 
+         ie = functions.end(); it != ie; ++it) {
+    KFunction *kf = *it;
+    if (functionEscapes(kf->function))
+      escapingFunctions.insert(kf->function);
+  }
+
+  if (DebugPrintEscapingFunctions && !escapingFunctions.empty()) {
+    llvm::cerr << "KLEE: escaping functions: [";
+    for (std::set<Function*>::iterator it = escapingFunctions.begin(), 
+         ie = escapingFunctions.end(); it != ie; ++it) {
+      llvm::cerr << (*it)->getName() << ", ";
+    }
+    llvm::cerr << "]\n";
+  }
+}
+
+KConstant* KModule::getKConstant(Constant *c) {
+  std::map<llvm::Constant*, KConstant*>::iterator it = constantMap.find(c);
+  if (it != constantMap.end())
+    return it->second;
+  return NULL;
+}
+
+unsigned KModule::getConstantID(Constant *c, KInstruction* ki) {
+  KConstant *kc = getKConstant(c);
+  if (kc)
+    return kc->id;  
+
+  unsigned id = constants.size();
+  kc = new KConstant(c, id, ki);
+  constantMap.insert(std::make_pair(c, kc));
+  constants.push_back(c);
+  return id;
+}
+
+/***/
+
+KConstant::KConstant(llvm::Constant* _ct, unsigned _id, KInstruction* _ki) {
+  ct = _ct;
+  id = _id;
+  ki = _ki;
+}
+
+/***/
+
+KFunction::KFunction(llvm::Function *_function,
+                     KModule *km) 
+  : function(_function),
+    numArgs(function->arg_size()),
+    numInstructions(0),
+    trackCoverage(true) {
+  for (llvm::Function::iterator bbit = function->begin(), 
+         bbie = function->end(); bbit != bbie; ++bbit) {
+    BasicBlock *bb = bbit;
+    basicBlockEntry[bb] = numInstructions;
+    numInstructions += bb->size();
+  }
+
+  instructions = new KInstruction*[numInstructions];
+
+  std::map<Instruction*, unsigned> registerMap;
+
+  // The first arg_size() registers are reserved for formals.
+  unsigned rnum = numArgs;
+  for (llvm::Function::iterator bbit = function->begin(), 
+         bbie = function->end(); bbit != bbie; ++bbit) {
+    for (llvm::BasicBlock::iterator it = bbit->begin(), ie = bbit->end();
+         it != ie; ++it)
+      registerMap[it] = rnum++;
+  }
+  numRegisters = rnum;
+  
+  unsigned i = 0;
+  for (llvm::Function::iterator bbit = function->begin(), 
+         bbie = function->end(); bbit != bbie; ++bbit) {
+    for (llvm::BasicBlock::iterator it = bbit->begin(), ie = bbit->end();
+         it != ie; ++it) {
+      KInstruction *ki;
+
+      switch(it->getOpcode()) {
+      case Instruction::GetElementPtr:
+        ki = new KGEPInstruction(); break;
+      default:
+        ki = new KInstruction(); break;
+      }
+
+      unsigned numOperands = it->getNumOperands();
+      ki->inst = it;      
+      ki->operands = new int[numOperands];
+      ki->dest = registerMap[it];
+      for (unsigned j=0; j<numOperands; j++) {
+        Value *v = it->getOperand(j);
+        
+        if (Instruction *inst = dyn_cast<Instruction>(v)) {
+          ki->operands[j] = registerMap[inst];
+        } else if (Argument *a = dyn_cast<Argument>(v)) {
+          ki->operands[j] = a->getArgNo();
+        } else if (isa<BasicBlock>(v) || isa<InlineAsm>(v)) {
+          ki->operands[j] = -1;
+        } else {
+          assert(isa<Constant>(v));
+          Constant *c = cast<Constant>(v);
+          ki->operands[j] = -(km->getConstantID(c, ki) + 2);
+        }
+      }
+
+      instructions[i++] = ki;
+    }
+  }
+}
+
+KFunction::~KFunction() {
+  for (unsigned i=0; i<numInstructions; ++i)
+    delete instructions[i];
+  delete[] instructions;
+}

Added: klee/trunk/lib/Module/LowerSwitch.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/LowerSwitch.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/LowerSwitch.cpp (added)
+++ klee/trunk/lib/Module/LowerSwitch.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,134 @@
+//===-- LowerSwitch.cpp - Eliminate Switch instructions -------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// Derived from LowerSwitch.cpp in LLVM, heavily modified by piotrek
+// to get rid of the binary search transform, as it was creating
+// multiple paths through the program (i.e., extra paths that didn't
+// exist in the original program).
+//
+//===----------------------------------------------------------------------===//
+
+#include "Passes.h"
+#include <algorithm>
+
+using namespace llvm;
+
+namespace klee {
+
+char LowerSwitchPass::ID = 0;
+
+// The comparison function for sorting the switch case values in the vector.
+struct SwitchCaseCmp {
+  bool operator () (const LowerSwitchPass::SwitchCase& C1,
+                    const LowerSwitchPass::SwitchCase& C2) {
+    
+    const ConstantInt* CI1 = cast<const ConstantInt>(C1.value);
+    const ConstantInt* CI2 = cast<const ConstantInt>(C2.value);
+    return CI1->getValue().slt(CI2->getValue());
+  }
+};
+
+bool LowerSwitchPass::runOnFunction(Function &F) {
+  bool changed = false;
+
+  for (Function::iterator I = F.begin(), E = F.end(); I != E; ) {
+    BasicBlock *cur = I++; // Advance over block so we don't traverse new blocks
+
+    if (SwitchInst *SI = dyn_cast<SwitchInst>(cur->getTerminator())) {
+      changed = true;
+      processSwitchInst(SI);
+    }
+  }
+
+  return changed;
+}
+
+// switchConvert - Convert the switch statement into a linear scan
+// through all the case values
+void LowerSwitchPass::switchConvert(CaseItr begin, CaseItr end,
+                                    Value* value, BasicBlock* origBlock,
+                                    BasicBlock* defaultBlock)
+{
+  BasicBlock *curHead = defaultBlock;
+  Function *F = origBlock->getParent();
+  
+  // iterate through all the cases, creating a new BasicBlock for each
+  for (CaseItr it = begin; it < end; ++it) {
+    BasicBlock *newBlock = BasicBlock::Create("NodeBlock");
+    Function::iterator FI = origBlock;
+    F->getBasicBlockList().insert(++FI, newBlock);
+    
+    ICmpInst *cmpInst = new ICmpInst(ICmpInst::ICMP_EQ,
+                                     value,
+                                     it->value,
+                                     "Case Comparison");
+    
+    newBlock->getInstList().push_back(cmpInst);
+    BranchInst::Create(it->block, curHead, cmpInst, newBlock);
+
+    // If there were any PHI nodes in this successor, rewrite one entry
+    // from origBlock to come from newBlock.
+    for (BasicBlock::iterator bi = it->block->begin(); isa<PHINode>(bi); ++bi) {
+      PHINode* PN = cast<PHINode>(bi);
+
+      int blockIndex = PN->getBasicBlockIndex(origBlock);
+      assert(blockIndex != -1 && "Switch didn't go to this successor??");
+      PN->setIncomingBlock((unsigned)blockIndex, newBlock);
+    }
+    
+    curHead = newBlock;
+  }
+
+  // Branch to our shiny new if-then stuff...
+  BranchInst::Create(curHead, origBlock);
+}
+
+// processSwitchInst - Replace the specified switch instruction with a sequence
+// of chained if-then instructions.
+//
+void LowerSwitchPass::processSwitchInst(SwitchInst *SI) {
+  BasicBlock *origBlock = SI->getParent();
+  BasicBlock *defaultBlock = SI->getDefaultDest();
+  Function *F = origBlock->getParent();
+  Value *switchValue = SI->getOperand(0);
+
+  // Create a new, empty default block so that the new hierarchy of
+  // if-then statements go to this and the PHI nodes are happy.
+  BasicBlock* newDefault = BasicBlock::Create("newDefault");
+
+  F->getBasicBlockList().insert(defaultBlock, newDefault);
+  BranchInst::Create(defaultBlock, newDefault);
+
+  // If there is an entry in any PHI nodes for the default edge, make sure
+  // to update them as well.
+  for (BasicBlock::iterator I = defaultBlock->begin(); isa<PHINode>(I); ++I) {
+    PHINode *PN = cast<PHINode>(I);
+    int BlockIdx = PN->getBasicBlockIndex(origBlock);
+    assert(BlockIdx != -1 && "Switch didn't go to this successor??");
+    PN->setIncomingBlock((unsigned)BlockIdx, newDefault);
+  }
+  
+  CaseVector cases;
+  for (unsigned i = 1; i < SI->getNumSuccessors(); ++i)
+    cases.push_back(SwitchCase(SI->getSuccessorValue(i),
+                               SI->getSuccessor(i)));
+  
+  // reverse cases, as switchConvert constructs a chain of
+  //   basic blocks by appending to the front. if we reverse,
+  //   the if comparisons will happen in the same order
+  //   as the cases appear in the switch
+  std::reverse(cases.begin(), cases.end());
+  
+  switchConvert(cases.begin(), cases.end(), switchValue, origBlock, newDefault);
+
+  // We are now done with the switch instruction, so delete it
+  origBlock->getInstList().erase(SI);
+}
+
+}

Added: klee/trunk/lib/Module/Makefile
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/Makefile?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/Makefile (added)
+++ klee/trunk/lib/Module/Makefile Wed May 20 23:36:41 2009
@@ -0,0 +1,16 @@
+#===-- lib/Module/Makefile ---------------------------------*- Makefile -*--===#
+#
+#                     The KLEE Symbolic Virtual Machine
+#
+# This file is distributed under the University of Illinois Open Source
+# License. See LICENSE.TXT for details.
+#
+#===------------------------------------------------------------------------===#
+
+LEVEL=../..
+
+LIBRARYNAME=kleeModule
+DONT_BUILD_RELINKED=1
+BUILD_ARCHIVE=1
+
+include $(LEVEL)/Makefile.common

Propchange: klee/trunk/lib/Module/Makefile

------------------------------------------------------------------------------
    svn:executable = *

Added: klee/trunk/lib/Module/ModuleUtil.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/ModuleUtil.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/ModuleUtil.cpp (added)
+++ klee/trunk/lib/Module/ModuleUtil.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,101 @@
+//===-- ModuleUtil.cpp ----------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Internal/Support/ModuleUtil.h"
+
+#include "llvm/Function.h"
+#include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
+#include "llvm/Linker.h"
+#include "llvm/Module.h"
+#include "llvm/Assembly/AsmAnnotationWriter.h"
+#include "llvm/Support/CFG.h"
+#include "llvm/Support/InstIterator.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/Analysis/ValueTracking.h"
+
+#include <map>
+#include <iostream>
+#include <fstream>
+#include <sstream>
+#include <string>
+
+using namespace llvm;
+using namespace klee;
+
+Module *klee::linkWithLibrary(Module *module, 
+                              const std::string &libraryName) {
+  try {
+    Linker linker("klee", module, false);
+
+    llvm::sys::Path libraryPath(libraryName);
+    bool native = false;
+    
+    if (linker.LinkInFile(libraryPath, native)) {
+      assert(0 && "linking in library failed!");
+    }
+    
+    return linker.releaseModule();
+  } catch (...) {
+    assert(0 && "error during linking");
+  }
+}
+
+Function *klee::getDirectCallTarget(const Instruction *i) {
+  assert(isa<CallInst>(i) || isa<InvokeInst>(i));
+
+  Value *v = i->getOperand(0);
+  if (Function *f = dyn_cast<Function>(v)) {
+    return f;
+  } else if (llvm::ConstantExpr *ce = dyn_cast<llvm::ConstantExpr>(v)) {
+    if (ce->getOpcode()==Instruction::BitCast)
+      if (Function *f = dyn_cast<Function>(ce->getOperand(0)))
+        return f;
+
+    // NOTE: This assert may fire, it isn't necessarily a problem and
+    // can be disabled, I just wanted to know when and if it happened.
+    assert(0 && "FIXME: Unresolved direct target for a constant expression.");
+  }
+  
+  return 0;
+}
+
+static bool valueIsOnlyCalled(const Value *v) {
+  for (Value::use_const_iterator it = v->use_begin(), ie = v->use_end();
+       it != ie; ++it) {
+    if (const Instruction *instr = dyn_cast<Instruction>(*it)) {
+      if (instr->getOpcode()==0) continue; // XXX function numbering inst
+      if (!isa<CallInst>(instr) && !isa<InvokeInst>(instr)) return false;
+      
+      // Make sure that the value is only the target of this call and
+      // not an argument.
+      for (unsigned i=1,e=instr->getNumOperands(); i!=e; ++i)
+        if (instr->getOperand(i)==v)
+          return false;
+    } else if (const llvm::ConstantExpr *ce = 
+               dyn_cast<llvm::ConstantExpr>(*it)) {
+      if (ce->getOpcode()==Instruction::BitCast)
+        if (valueIsOnlyCalled(ce))
+          continue;
+      return false;
+    } else if (const GlobalAlias *ga = dyn_cast<GlobalAlias>(*it)) {
+      // XXX what about v is bitcast of aliasee?
+      if (v==ga->getAliasee() && !valueIsOnlyCalled(ga))
+        return false;
+    } else {
+      return false;
+    }
+  }
+
+  return true;
+}
+
+bool klee::functionEscapes(const Function *f) {
+  return !valueIsOnlyCalled(f);
+}

Added: klee/trunk/lib/Module/Optimize.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/Optimize.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/Optimize.cpp (added)
+++ klee/trunk/lib/Module/Optimize.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,272 @@
+// FIXME: This file is a bastard child of opt.cpp and llvm-ld's
+// Optimize.cpp. This stuff should live in common code.
+
+
+//===- Optimize.cpp - Optimize a complete program -------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements all optimization of the linked module for llvm-ld.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/Module.h"
+#include "llvm/PassManager.h"
+#include "llvm/Analysis/Passes.h"
+#include "llvm/Analysis/LoopPass.h"
+#include "llvm/Analysis/Verifier.h"
+#include "llvm/Support/CommandLine.h"
+#include "llvm/System/DynamicLibrary.h"
+#include "llvm/Target/TargetData.h"
+#include "llvm/Target/TargetMachine.h"
+#include "llvm/Transforms/IPO.h"
+#include "llvm/Transforms/Scalar.h"
+#include "llvm/Support/PassNameParser.h"
+#include "llvm/Support/PluginLoader.h"
+#include <iostream>
+using namespace llvm;
+
+#if 0
+// Pass Name Options as generated by the PassNameParser
+static cl::list<const PassInfo*, bool, PassNameParser>
+  OptimizationList(cl::desc("Optimizations available:"));
+#endif
+
+// Don't verify at the end
+static cl::opt<bool> DontVerify("disable-verify", cl::ReallyHidden);
+
+static cl::opt<bool> DisableInline("disable-inlining",
+  cl::desc("Do not run the inliner pass"));
+
+static cl::opt<bool>
+DisableOptimizations("disable-opt",
+  cl::desc("Do not run any optimization passes"));
+
+static cl::opt<bool> DisableInternalize("disable-internalize",
+  cl::desc("Do not mark all symbols as internal"));
+
+static cl::opt<bool> VerifyEach("verify-each",
+ cl::desc("Verify intermediate results of all passes"));
+
+static cl::alias ExportDynamic("export-dynamic",
+  cl::aliasopt(DisableInternalize),
+  cl::desc("Alias for -disable-internalize"));
+
+static cl::opt<bool> Strip("strip-all", 
+  cl::desc("Strip all symbol info from executable"));
+
+static cl::alias A0("s", cl::desc("Alias for --strip-all"), 
+  cl::aliasopt(Strip));
+
+static cl::opt<bool> StripDebug("strip-debug",
+  cl::desc("Strip debugger symbol info from executable"));
+
+static cl::alias A1("S", cl::desc("Alias for --strip-debug"),
+  cl::aliasopt(StripDebug));
+
+// A utility function that adds a pass to the pass manager but will also add
+// a verifier pass after if we're supposed to verify.
+static inline void addPass(PassManager &PM, Pass *P) {
+  // Add the pass to the pass manager...
+  PM.add(P);
+
+  // If we are verifying all of the intermediate steps, add the verifier...
+  if (VerifyEach)
+    PM.add(createVerifierPass());
+}
+
+namespace llvm {
+
+
+static void AddStandardCompilePasses(PassManager &PM) {
+  PM.add(createVerifierPass());                  // Verify that input is correct
+
+  addPass(PM, createLowerSetJmpPass());          // Lower llvm.setjmp/.longjmp
+
+  // If the -strip-debug command line option was specified, do it.
+  if (StripDebug)
+    addPass(PM, createStripSymbolsPass(true));
+
+  if (DisableOptimizations) return;
+
+  addPass(PM, createRaiseAllocationsPass());     // call %malloc -> malloc inst
+  addPass(PM, createCFGSimplificationPass());    // Clean up disgusting code
+  addPass(PM, createPromoteMemoryToRegisterPass());// Kill useless allocas
+  addPass(PM, createGlobalOptimizerPass());      // Optimize out global vars
+  addPass(PM, createGlobalDCEPass());            // Remove unused fns and globs
+  addPass(PM, createIPConstantPropagationPass());// IP Constant Propagation
+  addPass(PM, createDeadArgEliminationPass());   // Dead argument elimination
+  addPass(PM, createInstructionCombiningPass()); // Clean up after IPCP & DAE
+  addPass(PM, createCFGSimplificationPass());    // Clean up after IPCP & DAE
+
+  addPass(PM, createPruneEHPass());              // Remove dead EH info
+  addPass(PM, createFunctionAttrsPass());        // Deduce function attrs
+
+  if (!DisableInline)
+    addPass(PM, createFunctionInliningPass());   // Inline small functions
+  addPass(PM, createArgumentPromotionPass());    // Scalarize uninlined fn args
+
+  addPass(PM, createSimplifyLibCallsPass());     // Library Call Optimizations
+  addPass(PM, createInstructionCombiningPass()); // Cleanup for scalarrepl.
+  addPass(PM, createJumpThreadingPass());        // Thread jumps.
+  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
+  addPass(PM, createScalarReplAggregatesPass()); // Break up aggregate allocas
+  addPass(PM, createInstructionCombiningPass()); // Combine silly seq's
+  addPass(PM, createCondPropagationPass());      // Propagate conditionals
+
+  addPass(PM, createTailCallEliminationPass());  // Eliminate tail calls
+  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
+  addPass(PM, createReassociatePass());          // Reassociate expressions
+  addPass(PM, createLoopRotatePass());
+  addPass(PM, createLICMPass());                 // Hoist loop invariants
+  addPass(PM, createLoopUnswitchPass());         // Unswitch loops.
+  addPass(PM, createLoopIndexSplitPass());       // Index split loops.
+  // FIXME : Removing instcombine causes nestedloop regression.
+  addPass(PM, createInstructionCombiningPass());
+  addPass(PM, createIndVarSimplifyPass());       // Canonicalize indvars
+  addPass(PM, createLoopDeletionPass());         // Delete dead loops
+  addPass(PM, createLoopUnrollPass());           // Unroll small loops
+  addPass(PM, createInstructionCombiningPass()); // Clean up after the unroller
+  addPass(PM, createGVNPass());                  // Remove redundancies
+  addPass(PM, createMemCpyOptPass());            // Remove memcpy / form memset
+  addPass(PM, createSCCPPass());                 // Constant prop with SCCP
+
+  // Run instcombine after redundancy elimination to exploit opportunities
+  // opened up by them.
+  addPass(PM, createInstructionCombiningPass());
+  addPass(PM, createCondPropagationPass());      // Propagate conditionals
+
+  addPass(PM, createDeadStoreEliminationPass()); // Delete dead stores
+  addPass(PM, createAggressiveDCEPass());        // Delete dead instructions
+  addPass(PM, createCFGSimplificationPass());    // Merge & remove BBs
+  addPass(PM, createStripDeadPrototypesPass());  // Get rid of dead prototypes
+  addPass(PM, createDeadTypeEliminationPass());  // Eliminate dead types
+  addPass(PM, createConstantMergePass());        // Merge dup global constants
+}
+
+/// Optimize - Perform link time optimizations. This will run the scalar
+/// optimizations, any loaded plugin-optimization modules, and then the
+/// inter-procedural optimizations if applicable.
+void Optimize(Module* M) {
+
+  // Instantiate the pass manager to organize the passes.
+  PassManager Passes;
+
+  // If we're verifying, start off with a verification pass.
+  if (VerifyEach)
+    Passes.add(createVerifierPass());
+
+  // Add an appropriate TargetData instance for this module...
+  addPass(Passes, new TargetData(M));
+
+  // DWD - Run the opt standard pass list as well.
+  AddStandardCompilePasses(Passes);
+
+  if (!DisableOptimizations) {
+    // Now that composite has been compiled, scan through the module, looking
+    // for a main function.  If main is defined, mark all other functions
+    // internal.
+    if (!DisableInternalize)
+      addPass(Passes, createInternalizePass(true));
+
+    // Propagate constants at call sites into the functions they call.  This
+    // opens opportunities for globalopt (and inlining) by substituting function
+    // pointers passed as arguments to direct uses of functions.  
+    addPass(Passes, createIPSCCPPass());
+
+    // Now that we internalized some globals, see if we can hack on them!
+    addPass(Passes, createGlobalOptimizerPass());
+
+    // Linking modules together can lead to duplicated global constants, only
+    // keep one copy of each constant...
+    addPass(Passes, createConstantMergePass());
+
+    // Remove unused arguments from functions...
+    addPass(Passes, createDeadArgEliminationPass());
+
+    // Reduce the code after globalopt and ipsccp.  Both can open up significant
+    // simplification opportunities, and both can propagate functions through
+    // function pointers.  When this happens, we often have to resolve varargs
+    // calls, etc, so let instcombine do this.
+    addPass(Passes, createInstructionCombiningPass());
+
+    if (!DisableInline)
+      addPass(Passes, createFunctionInliningPass()); // Inline small functions
+
+    addPass(Passes, createPruneEHPass());            // Remove dead EH info
+    addPass(Passes, createGlobalOptimizerPass());    // Optimize globals again.
+    addPass(Passes, createGlobalDCEPass());          // Remove dead functions
+
+    // If we didn't decide to inline a function, check to see if we can
+    // transform it to pass arguments by value instead of by reference.
+    addPass(Passes, createArgumentPromotionPass());
+
+    // The IPO passes may leave cruft around.  Clean up after them.
+    addPass(Passes, createInstructionCombiningPass());
+    addPass(Passes, createJumpThreadingPass());        // Thread jumps.
+    addPass(Passes, createScalarReplAggregatesPass()); // Break up allocas
+
+    // Run a few AA driven optimizations here and now, to cleanup the code.
+    addPass(Passes, createFunctionAttrsPass());      // Add nocapture
+    addPass(Passes, createGlobalsModRefPass());      // IP alias analysis
+
+    addPass(Passes, createLICMPass());               // Hoist loop invariants
+    addPass(Passes, createGVNPass());                // Remove redundancies
+    addPass(Passes, createMemCpyOptPass());          // Remove dead memcpy's
+    addPass(Passes, createDeadStoreEliminationPass()); // Nuke dead stores
+
+    // Cleanup and simplify the code after the scalar optimizations.
+    addPass(Passes, createInstructionCombiningPass());
+
+    addPass(Passes, createJumpThreadingPass());        // Thread jumps.
+    addPass(Passes, createPromoteMemoryToRegisterPass()); // Cleanup jumpthread.
+    
+    // Delete basic blocks, which optimization passes may have killed...
+    addPass(Passes, createCFGSimplificationPass());
+
+    // Now that we have optimized the program, discard unreachable functions...
+    addPass(Passes, createGlobalDCEPass());
+  }
+
+  // If the -s or -S command line options were specified, strip the symbols out
+  // of the resulting program to make it smaller.  -s and -S are GNU ld options
+  // that we are supporting; they alias -strip-all and -strip-debug.
+  if (Strip || StripDebug)
+    addPass(Passes, createStripSymbolsPass(StripDebug && !Strip));
+
+#if 0
+  // Create a new optimization pass for each one specified on the command line
+  std::auto_ptr<TargetMachine> target;
+  for (unsigned i = 0; i < OptimizationList.size(); ++i) {
+    const PassInfo *Opt = OptimizationList[i];
+    if (Opt->getNormalCtor())
+      addPass(Passes, Opt->getNormalCtor()());
+    else
+      std::cerr << "llvm-ld: cannot create pass: " << Opt->getPassName() 
+                << "\n";
+  }
+#endif
+
+  // The user's passes may leave cruft around. Clean up after them them but
+  // only if we haven't got DisableOptimizations set
+  if (!DisableOptimizations) {
+    addPass(Passes, createInstructionCombiningPass());
+    addPass(Passes, createCFGSimplificationPass());
+    addPass(Passes, createAggressiveDCEPass());
+    addPass(Passes, createGlobalDCEPass());
+  }
+
+  // Make sure everything is still good.
+  if (!DontVerify)
+    Passes.add(createVerifierPass());
+
+  // Run our queue of passes all at once now, efficiently.
+  Passes.run(*M);
+}
+
+}

Added: klee/trunk/lib/Module/Passes.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/Passes.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/Passes.h (added)
+++ klee/trunk/lib/Module/Passes.h Wed May 20 23:36:41 2009
@@ -0,0 +1,132 @@
+//===-- Passes.h ------------------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef KLEE_PASSES_H
+#define KLEE_PASSES_H
+
+#include "llvm/Constants.h"
+#include "llvm/Instructions.h"
+#include "llvm/Module.h"
+#include "llvm/Pass.h"
+#include "llvm/CodeGen/IntrinsicLowering.h"
+
+namespace llvm {
+  class Function;
+  class Instruction;
+  class Module;
+  class TargetData;
+  class Type;
+}
+
+namespace klee {
+
+  /// RaiseAsmPass - This pass raises some common occurences of inline
+  /// asm which are used by glibc into normal LLVM IR.
+class RaiseAsmPass : public llvm::ModulePass {
+  static char ID;
+
+  llvm::Function *getIntrinsic(llvm::Module &M,
+                               unsigned IID,
+                               const llvm::Type **Tys,
+                               unsigned NumTys);
+  llvm::Function *getIntrinsic(llvm::Module &M,
+                               unsigned IID, 
+                               const llvm::Type *Ty0) {
+    return getIntrinsic(M, IID, &Ty0, 1);
+  }
+
+  bool runOnInstruction(llvm::Module &M, llvm::Instruction *I);
+
+public:
+  RaiseAsmPass() : llvm::ModulePass((intptr_t) &ID) {}
+  
+  virtual bool runOnModule(llvm::Module &M);
+};
+
+  // This is a module pass because it can add and delete module
+  // variables (via intrinsic lowering).
+class IntrinsicCleanerPass : public llvm::ModulePass {
+  static char ID;
+  llvm::IntrinsicLowering *IL;
+  bool LowerIntrinsics;
+
+  bool runOnBasicBlock(llvm::BasicBlock &b);
+public:
+  IntrinsicCleanerPass(const llvm::TargetData &TD,
+                       bool LI=true)
+    : llvm::ModulePass((intptr_t) &ID),
+      IL(new llvm::IntrinsicLowering(TD)),
+      LowerIntrinsics(LI) {}
+  ~IntrinsicCleanerPass() { delete IL; } 
+  
+  virtual bool runOnModule(llvm::Module &M);
+};
+  
+  // performs two transformations which make interpretation
+  // easier and faster.
+  //
+  // 1) Ensure that all the PHI nodes in a basic block have
+  //    the incoming block list in the same order. Thus the
+  //    incoming block index only needs to be computed once
+  //    for each transfer.
+  // 
+  // 2) Ensure that no PHI node result is used as an argument to
+  //    a subsequent PHI node in the same basic block. This allows
+  //    the transfer to execute the instructions in order instead
+  //    of in two passes.
+class PhiCleanerPass : public llvm::FunctionPass {
+  static char ID;
+
+public:
+  PhiCleanerPass() : llvm::FunctionPass((intptr_t) &ID) {}
+  
+  virtual bool runOnFunction(llvm::Function &f);
+};
+  
+class DivCheckPass : public llvm::ModulePass {
+  static char ID;
+public:
+  DivCheckPass(): ModulePass((intptr_t) &ID) {}
+  virtual bool runOnModule(llvm::Module &M);
+};
+
+/// LowerSwitchPass - Replace all SwitchInst instructions with chained branch
+/// instructions.  Note that this cannot be a BasicBlock pass because it
+/// modifies the CFG!
+class LowerSwitchPass : public llvm::FunctionPass {
+public:
+  static char ID; // Pass identification, replacement for typeid
+  LowerSwitchPass() : FunctionPass((intptr_t) &ID) {} 
+  
+  virtual bool runOnFunction(llvm::Function &F);
+  
+  struct SwitchCase {
+    llvm ::Constant *value;
+    llvm::BasicBlock *block;
+    
+    SwitchCase() : value(0), block(0) { }
+    SwitchCase(llvm::Constant *v, llvm::BasicBlock *b) :
+      value(v), block(b) { }
+  };
+  
+  typedef std::vector<SwitchCase>           CaseVector;
+  typedef std::vector<SwitchCase>::iterator CaseItr;
+  
+private:
+  void processSwitchInst(llvm::SwitchInst *SI);
+  void switchConvert(CaseItr begin,
+                     CaseItr end,
+                     llvm::Value *value,
+                     llvm::BasicBlock *origBlock,
+                     llvm::BasicBlock *defaultBlock);
+};
+
+}
+
+#endif

Added: klee/trunk/lib/Module/PhiCleaner.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/PhiCleaner.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/PhiCleaner.cpp (added)
+++ klee/trunk/lib/Module/PhiCleaner.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,83 @@
+//===-- PhiCleaner.cpp ----------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Passes.h"
+
+#include <set>
+
+using namespace llvm;
+
+char klee::PhiCleanerPass::ID = 0;
+
+bool klee::PhiCleanerPass::runOnFunction(Function &f) {
+  bool changed = false;
+  
+  for (Function::iterator b = f.begin(), be = f.end(); b != be; ++b) {
+    BasicBlock::iterator it = b->begin();
+
+    if (it->getOpcode() == Instruction::PHI) {
+      PHINode *reference = cast<PHINode>(it);
+      
+      std::set<Value*> phis;
+      phis.insert(reference);
+
+      unsigned numBlocks = reference->getNumIncomingValues();
+      for (++it; isa<PHINode>(*it); ++it) {
+        PHINode *pi = cast<PHINode>(it);
+
+        assert(numBlocks == pi->getNumIncomingValues());
+
+        // see if it is out of order
+        unsigned i;
+        for (i=0; i<numBlocks; i++)
+          if (pi->getIncomingBlock(i) != reference->getIncomingBlock(i))
+            break;
+
+        if (i!=numBlocks) {
+          std::vector<Value*> values;
+          values.reserve(numBlocks);
+          for (unsigned i=0; i<numBlocks; i++)
+            values[i] = pi->getIncomingValueForBlock(reference->getIncomingBlock(i));
+          for (unsigned i=0; i<numBlocks; i++) {
+            pi->setIncomingBlock(i, reference->getIncomingBlock(i));
+            pi->setIncomingValue(i, values[i]);
+          }
+          changed = true;
+        }
+
+        // see if it uses any previously defined phi nodes
+        for (i=0; i<numBlocks; i++) {
+          Value *value = pi->getIncomingValue(i);
+
+          if (phis.find(value) != phis.end()) {
+            // fix by making a "move" at the end of the incoming block
+            // to a new temporary, which is thus known not to be a phi
+            // result. we could be somewhat more efficient about this
+            // by sharing temps and by reordering phi instructions so
+            // this isn't completely necessary, but in the end this is
+            // just a pathological case which does not occur very
+            // often.
+            Instruction *tmp = 
+              new BitCastInst(value, 
+                              value->getType(),
+                              value->getName() + ".phiclean",
+                              pi->getIncomingBlock(i)->getTerminator());
+            pi->setIncomingValue(i, tmp);
+          }
+
+          changed = true;
+        }
+        
+        phis.insert(pi);
+      }
+    }
+  }
+
+  return changed;
+}

Added: klee/trunk/lib/Module/RaiseAsm.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Module/RaiseAsm.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Module/RaiseAsm.cpp (added)
+++ klee/trunk/lib/Module/RaiseAsm.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,69 @@
+//===-- RaiseAsm.cpp ------------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "Passes.h"
+
+#include "llvm/InlineAsm.h"
+
+using namespace llvm;
+using namespace klee;
+
+char RaiseAsmPass::ID = 0;
+
+Function *RaiseAsmPass::getIntrinsic(llvm::Module &M,
+                                     unsigned IID,
+                                     const Type **Tys,
+                                     unsigned NumTys) {  
+  return Intrinsic::getDeclaration(&M, (llvm::Intrinsic::ID) IID, Tys, NumTys);
+}
+
+// FIXME: This should just be implemented as a patch to
+// X86TargetAsmInfo.cpp, then everyone will benefit.
+bool RaiseAsmPass::runOnInstruction(Module &M, Instruction *I) {
+  if (CallInst *ci = dyn_cast<CallInst>(I)) {
+    if (InlineAsm *ia = dyn_cast<InlineAsm>(ci->getCalledValue())) {
+      const std::string &as = ia->getAsmString();
+      const std::string &cs = ia->getConstraintString();
+      const llvm::Type *T = ci->getType();
+
+      // bswaps
+      if (ci->getNumOperands() == 2 && 
+          T == ci->getOperand(1)->getType() &&
+          ((T == llvm::Type::Int16Ty && 
+            as == "rorw $$8, ${0:w}" &&
+            cs == "=r,0,~{dirflag},~{fpsr},~{flags},~{cc}") ||
+           (T == llvm::Type::Int32Ty &&
+            as == "rorw $$8, ${0:w};rorl $$16, $0;rorw $$8, ${0:w}" &&
+            cs == "=r,0,~{dirflag},~{fpsr},~{flags},~{cc}"))) {
+        llvm::Value *Arg0 = ci->getOperand(1);
+        Function *F = getIntrinsic(M, Intrinsic::bswap, Arg0->getType());
+        ci->setOperand(0, F);
+        return true;
+      }
+    }
+  }
+
+  return false;
+}
+
+bool RaiseAsmPass::runOnModule(Module &M) {
+  bool changed = false;
+  
+  for (Module::iterator fi = M.begin(), fe = M.end(); fi != fe; ++fi) {
+    for (Function::iterator bi = fi->begin(), be = fi->end(); bi != be; ++bi) {
+      for (BasicBlock::iterator ii = bi->begin(), ie = bi->end(); ii != ie;) {
+        Instruction *i = ii;
+        ++ii;  
+        changed |= runOnInstruction(M, i);
+      }
+    }
+  }
+
+  return changed;
+}

Added: klee/trunk/lib/README.txt
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/README.txt?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/README.txt (added)
+++ klee/trunk/lib/README.txt Wed May 20 23:36:41 2009
@@ -0,0 +1,18 @@
+The klee and kleaver code is organized as follows:
+
+lib/Basic   - Low level support for both klee and kleaver which should
+              be independent of LLVM.
+
+lib/Support - Higher level support, but only used by klee. This can
+              use LLVM facilities.
+
+lib/Expr    - The core kleaver expression library.
+
+lib/Solver  - The kleaver solver library.
+
+lib/Module  - klee facilities for working with LLVM modules, including
+              the shadow module/instruction structures we use during
+              execution.
+
+lib/Core    - The core symbolic virtual machine.
+

Added: klee/trunk/lib/Solver/CachingSolver.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/CachingSolver.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/CachingSolver.cpp (added)
+++ klee/trunk/lib/Solver/CachingSolver.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,241 @@
+//===-- CachingSolver.cpp - Caching expression solver ---------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+
+#include "klee/Solver.h"
+
+#include "klee/Constraints.h"
+#include "klee/Expr.h"
+#include "klee/IncompleteSolver.h"
+#include "klee/SolverImpl.h"
+
+#include "SolverStats.h"
+
+#include <tr1/unordered_map>
+
+using namespace klee;
+
+class CachingSolver : public SolverImpl {
+private:
+  ref<Expr> canonicalizeQuery(ref<Expr> originalQuery,
+                              bool &negationUsed);
+
+  void cacheInsert(const Query& query,
+                   IncompleteSolver::PartialValidity result);
+
+  bool cacheLookup(const Query& query,
+                   IncompleteSolver::PartialValidity &result);
+  
+  struct CacheEntry {
+    CacheEntry(const ConstraintManager &c, ref<Expr> q)
+      : constraints(c), query(q) {}
+
+    CacheEntry(const CacheEntry &ce)
+      : constraints(ce.constraints), query(ce.query) {}
+    
+    ConstraintManager constraints;
+    ref<Expr> query;
+
+    bool operator==(const CacheEntry &b) const {
+      return constraints==b.constraints && *query.get()==*b.query.get();
+    }
+  };
+  
+  struct CacheEntryHash {
+    unsigned operator()(const CacheEntry &ce) const {
+      unsigned result = ce.query.hash();
+      
+      for (ConstraintManager::constraint_iterator it = ce.constraints.begin();
+           it != ce.constraints.end(); ++it)
+        result ^= it->hash();
+      
+      return result;
+    }
+  };
+
+  typedef std::tr1::unordered_map<CacheEntry, 
+                                  IncompleteSolver::PartialValidity, 
+                                  CacheEntryHash> cache_map;
+  
+  Solver *solver;
+  cache_map cache;
+
+public:
+  CachingSolver(Solver *s) : solver(s) {}
+  ~CachingSolver() { cache.clear(); delete solver; }
+
+  bool computeValidity(const Query&, Solver::Validity &result);
+  bool computeTruth(const Query&, bool &isValid);
+  bool computeValue(const Query& query, ref<Expr> &result) {
+    return solver->impl->computeValue(query, result);
+  }
+  bool computeInitialValues(const Query& query,
+                            const std::vector<const Array*> &objects,
+                            std::vector< std::vector<unsigned char> > &values,
+                            bool &hasSolution) {
+    return solver->impl->computeInitialValues(query, objects, values, 
+                                              hasSolution);
+  }
+};
+
+/** @returns the canonical version of the given query.  The reference
+    negationUsed is set to true if the original query was negated in
+    the canonicalization process. */
+ref<Expr> CachingSolver::canonicalizeQuery(ref<Expr> originalQuery,
+                                           bool &negationUsed) {
+  ref<Expr> negatedQuery = Expr::createNot(originalQuery);
+
+  // select the "smaller" query to the be canonical representation
+  if (originalQuery.compare(negatedQuery) < 0) {
+    negationUsed = false;
+    return originalQuery;
+  } else {
+    negationUsed = true;
+    return negatedQuery;
+  }
+}
+
+/** @returns true on a cache hit, false of a cache miss.  Reference
+    value result only valid on a cache hit. */
+bool CachingSolver::cacheLookup(const Query& query,
+                                IncompleteSolver::PartialValidity &result) {
+  bool negationUsed;
+  ref<Expr> canonicalQuery = canonicalizeQuery(query.expr, negationUsed);
+
+  CacheEntry ce(query.constraints, canonicalQuery);
+  cache_map::iterator it = cache.find(ce);
+  
+  if (it != cache.end()) {
+    result = (negationUsed ?
+              IncompleteSolver::negatePartialValidity(it->second) :
+              it->second);
+    return true;
+  }
+  
+  return false;
+}
+
+/// Inserts the given query, result pair into the cache.
+void CachingSolver::cacheInsert(const Query& query,
+                                IncompleteSolver::PartialValidity result) {
+  bool negationUsed;
+  ref<Expr> canonicalQuery = canonicalizeQuery(query.expr, negationUsed);
+
+  CacheEntry ce(query.constraints, canonicalQuery);
+  IncompleteSolver::PartialValidity cachedResult = 
+    (negationUsed ? IncompleteSolver::negatePartialValidity(result) : result);
+  
+  cache.insert(std::make_pair(ce, cachedResult));
+}
+
+bool CachingSolver::computeValidity(const Query& query,
+                                    Solver::Validity &result) {
+  IncompleteSolver::PartialValidity cachedResult;
+  bool tmp, cacheHit = cacheLookup(query, cachedResult);
+  
+  if (cacheHit) {
+    ++stats::queryCacheHits;
+
+    switch(cachedResult) {
+    case IncompleteSolver::MustBeTrue:   
+      result = Solver::True;
+      return true;
+    case IncompleteSolver::MustBeFalse:  
+      result = Solver::False;
+      return true;
+    case IncompleteSolver::TrueOrFalse:  
+      result = Solver::Unknown;
+      return true;
+    case IncompleteSolver::MayBeTrue: {
+      if (!solver->impl->computeTruth(query, tmp))
+        return false;
+      if (tmp) {
+        cacheInsert(query, IncompleteSolver::MustBeTrue);
+        result = Solver::True;
+        return true;
+      } else {
+        cacheInsert(query, IncompleteSolver::TrueOrFalse);
+        result = Solver::Unknown;
+        return true;
+      }
+    }
+    case IncompleteSolver::MayBeFalse: {
+      if (!solver->impl->computeTruth(query.negateExpr(), tmp))
+        return false;
+      if (tmp) {
+        cacheInsert(query, IncompleteSolver::MustBeFalse);
+        result = Solver::False;
+        return true;
+      } else {
+        cacheInsert(query, IncompleteSolver::TrueOrFalse);
+        result = Solver::Unknown;
+        return true;
+      }
+    }
+    default: assert(0 && "unreachable");
+    }
+  }
+
+  ++stats::queryCacheMisses;
+  
+  if (!solver->impl->computeValidity(query, result))
+    return false;
+
+  switch (result) {
+  case Solver::True: 
+    cachedResult = IncompleteSolver::MustBeTrue; break;
+  case Solver::False: 
+    cachedResult = IncompleteSolver::MustBeFalse; break;
+  default: 
+    cachedResult = IncompleteSolver::TrueOrFalse; break;
+  }
+  
+  cacheInsert(query, cachedResult);
+  return true;
+}
+
+bool CachingSolver::computeTruth(const Query& query,
+                                 bool &isValid) {
+  IncompleteSolver::PartialValidity cachedResult;
+  bool cacheHit = cacheLookup(query, cachedResult);
+
+  // a cached result of MayBeTrue forces us to check whether
+  // a False assignment exists.
+  if (cacheHit && cachedResult != IncompleteSolver::MayBeTrue) {
+    ++stats::queryCacheHits;
+    isValid = (cachedResult == IncompleteSolver::MustBeTrue);
+    return true;
+  }
+
+  ++stats::queryCacheMisses;
+  
+  // cache miss: query solver
+  if (!solver->impl->computeTruth(query, isValid))
+    return false;
+
+  if (isValid) {
+    cachedResult = IncompleteSolver::MustBeTrue;
+  } else if (cacheHit) {
+    // We know a true assignment exists, and query isn't valid, so
+    // must be TrueOrFalse.
+    assert(cachedResult == IncompleteSolver::MayBeTrue);
+    cachedResult = IncompleteSolver::TrueOrFalse;
+  } else {
+    cachedResult = IncompleteSolver::MayBeFalse;
+  }
+  
+  cacheInsert(query, cachedResult);
+  return true;
+}
+
+///
+
+Solver *klee::createCachingSolver(Solver *_solver) {
+  return new Solver(new CachingSolver(_solver));
+}

Added: klee/trunk/lib/Solver/CexCachingSolver.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/CexCachingSolver.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/CexCachingSolver.cpp (added)
+++ klee/trunk/lib/Solver/CexCachingSolver.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,313 @@
+//===-- CexCachingSolver.cpp ----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Solver.h"
+
+#include "klee/Constraints.h"
+#include "klee/Expr.h"
+#include "klee/SolverImpl.h"
+#include "klee/TimerStatIncrementer.h"
+#include "klee/util/Assignment.h"
+#include "klee/util/ExprUtil.h"
+#include "klee/util/ExprVisitor.h"
+#include "klee/Internal/ADT/MapOfSets.h"
+
+#include "SolverStats.h"
+
+#include "llvm/Support/CommandLine.h"
+
+using namespace klee;
+using namespace llvm;
+
+namespace {
+  cl::opt<bool>
+  DebugCexCacheCheckBinding("debug-cex-cache-check-binding");
+
+  cl::opt<bool>
+  CexCacheTryAll("cex-cache-try-all",
+                 cl::desc("try substituting all counterexamples before asking STP"),
+                 cl::init(false));
+
+  cl::opt<bool>
+  CexCacheExperimental("cex-cache-exp", cl::init(false));
+
+}
+
+///
+
+typedef std::set< ref<Expr> > KeyType;
+
+struct AssignmentLessThan {
+  bool operator()(const Assignment *a, const Assignment *b) {
+    return a->bindings < b->bindings;
+  }
+};
+
+
+class CexCachingSolver : public SolverImpl {
+  typedef std::set<Assignment*, AssignmentLessThan> assignmentsTable_ty;
+
+  Solver *solver;
+  
+  MapOfSets<ref<Expr>, Assignment*> cache;
+  // memo table
+  assignmentsTable_ty assignmentsTable;
+
+  bool searchForAssignment(KeyType &key, 
+                           Assignment *&result);
+  
+  bool lookupAssignment(const Query& query, Assignment *&result);
+
+  bool getAssignment(const Query& query, Assignment *&result);
+  
+public:
+  CexCachingSolver(Solver *_solver) : solver(_solver) {}
+  ~CexCachingSolver();
+  
+  bool computeTruth(const Query&, bool &isValid);
+  bool computeValidity(const Query&, Solver::Validity &result);
+  bool computeValue(const Query&, ref<Expr> &result);
+  bool computeInitialValues(const Query&,
+                            const std::vector<const Array*> &objects,
+                            std::vector< std::vector<unsigned char> > &values,
+                            bool &hasSolution);
+};
+
+///
+
+struct NullAssignment {
+  bool operator()(Assignment *a) const { return !a; }
+};
+
+struct NonNullAssignment {
+  bool operator()(Assignment *a) const { return a!=0; }
+};
+
+struct NullOrSatisfyingAssignment {
+  KeyType &key;
+  
+  NullOrSatisfyingAssignment(KeyType &_key) : key(_key) {}
+
+  bool operator()(Assignment *a) const { 
+    return !a || a->satisfies(key.begin(), key.end()); 
+  }
+};
+
+bool CexCachingSolver::searchForAssignment(KeyType &key, Assignment *&result) {
+  Assignment * const *lookup = cache.lookup(key);
+  if (lookup) {
+    result = *lookup;
+    return true;
+  }
+
+  if (CexCacheTryAll) {
+    Assignment **lookup = cache.findSuperset(key, NonNullAssignment());
+    if (!lookup) lookup = cache.findSubset(key, NullAssignment());
+    if (lookup) {
+      result = *lookup;
+      return true;
+    }
+    for (assignmentsTable_ty::iterator it = assignmentsTable.begin(), 
+           ie = assignmentsTable.end(); it != ie; ++it) {
+      Assignment *a = *it;
+      if (a->satisfies(key.begin(), key.end())) {
+        result = a;
+        return true;
+      }
+    }
+  } else {
+    // XXX which order? one is sure to be better
+    Assignment **lookup = cache.findSuperset(key, NonNullAssignment());
+    if (!lookup) lookup = cache.findSubset(key, NullOrSatisfyingAssignment(key));
+    if (lookup) {
+      result = *lookup;
+      return true;
+    }
+  }
+  
+  return false;
+}
+
+bool CexCachingSolver::lookupAssignment(const Query &query, 
+                                        Assignment *&result) {
+  KeyType key(query.constraints.begin(), query.constraints.end());
+  ref<Expr> neg = Expr::createNot(query.expr);
+  if (neg.isConstant()) {
+    if (!neg.getConstantValue()) {
+      result = (Assignment*) 0;
+      return true;
+    }
+  } else {
+    key.insert(neg);
+  }
+
+  return searchForAssignment(key, result);
+}
+
+bool CexCachingSolver::getAssignment(const Query& query, Assignment *&result) {
+  KeyType key(query.constraints.begin(), query.constraints.end());
+  ref<Expr> neg = Expr::createNot(query.expr);
+  if (neg.isConstant()) {
+    if (!neg.getConstantValue()) {
+      result = (Assignment*) 0;
+      return true;
+    }
+  } else {
+    key.insert(neg);
+  }
+
+  if (!searchForAssignment(key, result)) {
+    // need to solve
+    
+    std::vector<const Array*> objects;
+    findSymbolicObjects(key.begin(), key.end(), objects);
+
+    std::vector< std::vector<unsigned char> > values;
+    bool hasSolution;
+    if (!solver->impl->computeInitialValues(query, objects, values, 
+                                            hasSolution))
+      return false;
+    
+    Assignment *binding;
+    if (hasSolution) {
+      binding = new Assignment(objects, values);
+
+      // memoization
+      std::pair<assignmentsTable_ty::iterator, bool>
+        res = assignmentsTable.insert(binding);
+      if (!res.second) {
+        delete binding;
+        binding = *res.first;
+      }
+
+      if (DebugCexCacheCheckBinding)
+        assert(binding->satisfies(key.begin(), key.end()));
+    } else {
+      binding = (Assignment*) 0;
+    }
+
+    result = binding;
+    cache.insert(key, binding);
+  }
+
+  return true;
+}
+
+///
+
+CexCachingSolver::~CexCachingSolver() {
+  cache.clear();
+  delete solver;
+  for (assignmentsTable_ty::iterator it = assignmentsTable.begin(), 
+         ie = assignmentsTable.end(); it != ie; ++it)
+    delete *it;
+}
+
+bool CexCachingSolver::computeValidity(const Query& query,
+                                       Solver::Validity &result) {
+  TimerStatIncrementer t(stats::cexCacheTime);
+  Assignment *a;
+  if (!getAssignment(query.withFalse(), a))
+    return false;
+  assert(a && "computeValidity() must have assignment");
+  ref<Expr> q = a->evaluate(query.expr);
+  assert(q.isConstant() && "assignment evaluation did not result in constant");
+
+  if (q.getConstantValue()) {
+    if (!getAssignment(query, a))
+      return false;
+    result = !a ? Solver::True : Solver::Unknown;
+  } else {
+    if (!getAssignment(query.negateExpr(), a))
+      return false;
+    result = !a ? Solver::False : Solver::Unknown;
+  }
+  
+  return true;
+}
+
+bool CexCachingSolver::computeTruth(const Query& query,
+                                    bool &isValid) {
+  TimerStatIncrementer t(stats::cexCacheTime);
+
+  // There is a small amount of redundancy here. We only need to know
+  // truth and do not really need to compute an assignment. This means
+  // that we could check the cache to see if we already know that
+  // state ^ query has no assignment. In that case, by the validity of
+  // state, we know that state ^ !query must have an assignment, and
+  // so query cannot be true (valid). This does get hits, but doesn't
+  // really seem to be worth the overhead.
+
+  if (CexCacheExperimental) {
+    Assignment *a;
+    if (lookupAssignment(query.negateExpr(), a) && !a)
+      return false;
+  }
+
+  Assignment *a;
+  if (!getAssignment(query, a))
+    return false;
+
+  isValid = !a;
+
+  return true;
+}
+
+bool CexCachingSolver::computeValue(const Query& query,
+                                    ref<Expr> &result) {
+  TimerStatIncrementer t(stats::cexCacheTime);
+
+  Assignment *a;
+  if (!getAssignment(query.withFalse(), a))
+    return false;
+  assert(a && "computeValue() must have assignment");
+  result = a->evaluate(query.expr);  
+  assert(result.isConstant() && 
+         "assignment evaluation did not result in constant");
+  return true;
+}
+
+bool 
+CexCachingSolver::computeInitialValues(const Query& query,
+                                       const std::vector<const Array*> 
+                                         &objects,
+                                       std::vector< std::vector<unsigned char> >
+                                         &values,
+                                       bool &hasSolution) {
+  TimerStatIncrementer t(stats::cexCacheTime);
+  Assignment *a;
+  if (!getAssignment(query, a))
+    return false;
+  hasSolution = !!a;
+  
+  if (!a)
+    return true;
+
+  // FIXME: We should use smarter assignment for result so we don't
+  // need redundant copy.
+  values = std::vector< std::vector<unsigned char> >(objects.size());
+  for (unsigned i=0; i < objects.size(); ++i) {
+    const Array *os = objects[i];
+    Assignment::bindings_ty::iterator it = a->bindings.find(os);
+    
+    if (it == a->bindings.end()) {
+      values[i] = std::vector<unsigned char>(os->size, 0);
+    } else {
+      values[i] = it->second;
+    }
+  }
+  
+  return true;
+}
+
+///
+
+Solver *klee::createCexCachingSolver(Solver *_solver) {
+  return new Solver(new CexCachingSolver(_solver));
+}

Added: klee/trunk/lib/Solver/ConstantDivision.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/ConstantDivision.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/ConstantDivision.cpp (added)
+++ klee/trunk/lib/Solver/ConstantDivision.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,146 @@
+//===-- ConstantDivision.cpp ----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "ConstantDivision.h"
+
+#include "klee/util/Bits.h"
+
+#include <algorithm>
+#include <cassert>
+
+namespace klee {
+
+/* Macros and functions which define the basic bit-level operations 
+ * needed to implement quick division operations.
+ *
+ * Based on Hacker's Delight (2003) by Henry S. Warren, Jr.
+ */
+
+/* 32 -- number of bits in the integer type on this architecture */
+
+/* 2^32 -- NUM_BITS=32 requires 64 bits to represent this unsigned value */
+#define TWO_TO_THE_32_U64 (1ULL << 32)
+
+/* 2e31 -- NUM_BITS=32 requires 64 bits to represent this signed value */
+#define TWO_TO_THE_31_S64 (1LL << 31)
+
+/* ABS(x) -- positive x */
+#define ABS(x) ( ((x)>0)?x:-(x) ) /* fails if x is the min value of its type */
+
+/* XSIGN(x) -- -1 if x<0 and 0 otherwise */
+#define XSIGN(x) ( (x) >> 31 )
+
+/* LOG2_CEIL(x) -- logarithm base 2 of x, rounded up */
+#define LOG2_CEIL(x) ( 32 - ldz(x - 1) )
+
+/* ones(x) -- counts the number of bits in x with the value 1 */
+static uint32_t ones( register uint32_t x ) {
+  x -= ((x >> 1) & 0x55555555);
+  x = (((x >> 2) & 0x33333333) + (x & 0x33333333));
+  x = (((x >> 4) + x) & 0x0f0f0f0f);
+  x += (x >> 8);
+  x += (x >> 16);
+
+  return( x & 0x0000003f );
+}
+
+/* ldz(x) -- counts the number of leading zeroes in a 32-bit word */
+static uint32_t ldz( register uint32_t x ) {
+  x |= (x >> 1);
+  x |= (x >> 2);
+  x |= (x >> 4);
+  x |= (x >> 8);
+  x |= (x >> 16);
+
+  return 32 - ones(x);
+}
+
+/* exp_base_2(n) -- 2^n computed as an integer */
+static uint32_t exp_base_2( register int32_t n ) {
+  register uint32_t x = ~n & (n - 32);
+  x = x >> 31;
+  return( x << n );
+}
+
+// A simple algorithm: Iterate over all contiguous regions of 1 bits
+// in x starting with the lowest bits.
+//
+// For a particular range where x is 1 for bits [low,high) then:
+//   1) if the range is just one bit, simple add it
+//   2) if the range is more than one bit, replace with an add
+//      of the high bit and a subtract of the low bit. we apply
+//      one useful optimization: if we were going to add the bit
+//      below the one we wish to subtract, we simply change that
+//      add to a subtract instead of subtracting the low bit itself.
+// Obviously we must take care when high==64.
+void ComputeMultConstants64(uint64_t multiplicand, 
+                            uint64_t &add, uint64_t &sub) {
+  uint64_t x = multiplicand;
+  add = sub = 0;
+
+  while (x) {
+    // Determine rightmost contiguous region of 1s.
+    unsigned low = bits64::indexOfRightmostBit(x);
+    uint64_t lowbit = 1LL << low;
+    uint64_t p = x + lowbit;
+    uint64_t q = bits64::isolateRightmostBit(p);
+    unsigned high = q ? bits64::indexOfSingleBit(q) : 64;
+
+    if (high==low+1) { // Just one bit...
+      add |= lowbit;
+    } else {
+      // Rewrite as +(1<<high) - (1<<low).
+
+      // Optimize +(1<<x) - (1<<(x+1)) to -(1<<x).
+      if (low && (add & (lowbit>>1))) {
+        add ^= lowbit>>1;
+        sub ^= lowbit>>1;
+      } else {
+        sub |= lowbit;
+      }
+
+      if (high!=64)
+        add |= 1LL << high;
+    }
+
+    x = p ^ q;
+  }
+
+  assert(multiplicand == add - sub);
+}
+
+void ComputeUDivConstants32(uint32_t d, uint32_t &mprime, uint32_t &sh1, 
+                            uint32_t &sh2) {
+  int32_t l = LOG2_CEIL( d ); /* signed so l-1 => -1 when l=0 (see sh2) */
+  uint32_t mid = exp_base_2(l) - d;
+
+  mprime = (TWO_TO_THE_32_U64 * mid / d) + 1;
+  sh1    = std::min( l,   1 );
+  sh2    = std::max( l-1, 0 );
+}
+
+void ComputeSDivConstants32(int32_t d, int32_t &mprime, int32_t &dsign, 
+                            int32_t &shpost ) {
+  uint64_t abs_d = ABS( (int64_t)d ); /* use 64-bits in case d is INT32_MIN */
+
+  /* LOG2_CEIL works on 32-bits, so we cast abs_d.  The only possible value
+   * outside the 32-bit rep. is 2^31.  This is special cased to save computer 
+   * time since 64-bit routines would be overkill. */
+  int32_t l = std::max( 1U, LOG2_CEIL((uint32_t)abs_d) );
+  if( abs_d == TWO_TO_THE_31_S64 ) l = 31;
+
+  uint32_t mid = exp_base_2( l - 1 );
+  uint64_t m = TWO_TO_THE_32_U64 * mid / abs_d + 1ULL;
+
+  mprime = m - TWO_TO_THE_32_U64; /* implicit cast to 32-bits signed */
+  dsign  = XSIGN( d );
+  shpost = l - 1;
+}
+
+}

Added: klee/trunk/lib/Solver/ConstantDivision.h
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/ConstantDivision.h?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/ConstantDivision.h (added)
+++ klee/trunk/lib/Solver/ConstantDivision.h Wed May 20 23:36:41 2009
@@ -0,0 +1,51 @@
+//===-- ConstantDivision.h --------------------------------------*- C++ -*-===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __UTIL_CONSTANTDIVISION_H__
+#define __UTIL_CONSTANTDIVISION_H__
+
+#include <stdint.h>
+
+namespace klee {
+
+/// ComputeMultConstants64 - Compute add and sub such that add-sub==x,
+/// while attempting to minimize the number of bits in add and sub
+/// combined.
+void ComputeMultConstants64(uint64_t x, uint64_t &add_out, 
+                            uint64_t &sub_out);
+
+/// Compute the constants to perform a quicker equivalent of a division of some 
+/// 32-bit unsigned integer n by a known constant d (also a 32-bit unsigned 
+/// integer).  The constants to compute n/d without explicit division will be 
+/// stored in mprime, sh1, and sh2 (unsigned 32-bit integers).
+/// 
+/// @param d - denominator (divisor)
+/// 
+/// @param [out] mprime
+/// @param [out] sh1
+/// @param [out] sh2
+void ComputeUDivConstants32(uint32_t d, uint32_t &mprime, uint32_t &sh1, 
+                            uint32_t &sh2);
+
+/// Compute the constants to perform a quicker equivalent of a division of some 
+/// 32-bit signed integer n by a known constant d (also a 32-bit signed 
+/// integer).  The constants to compute n/d without explicit division will be 
+/// stored in mprime, dsign, and shpost (signed 32-bit integers).
+/// 
+/// @param d - denominator (divisor)
+/// 
+/// @param [out] mprime
+/// @param [out] dsign
+/// @param [out] shpost
+void ComputeSDivConstants32(int32_t d, int32_t &mprime, int32_t &dsign, 
+                            int32_t &shpost);
+
+}
+
+#endif

Added: klee/trunk/lib/Solver/FastCexSolver.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/FastCexSolver.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/FastCexSolver.cpp (added)
+++ klee/trunk/lib/Solver/FastCexSolver.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,959 @@
+//===-- FastCexSolver.cpp -------------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/Solver.h"
+
+#include "klee/Constraints.h"
+#include "klee/Expr.h"
+#include "klee/IncompleteSolver.h"
+#include "klee/util/ExprEvaluator.h"
+#include "klee/util/ExprRangeEvaluator.h"
+#include "klee/util/ExprVisitor.h"
+// FIXME: Use APInt.
+#include "klee/Internal/Support/IntEvaluation.h"
+
+#include <iostream>
+#include <sstream>
+#include <cassert>
+#include <map>
+#include <vector>
+
+using namespace klee;
+
+/***/
+
+//#define LOG
+#ifdef LOG
+std::ostream *theLog;
+#endif
+
+      // Hacker's Delight, pgs 58-63
+static uint64_t minOR(uint64_t a, uint64_t b,
+                      uint64_t c, uint64_t d) {
+  uint64_t temp, m = ((uint64_t) 1)<<63;
+  while (m) {
+    if (~a & c & m) {
+      temp = (a | m) & -m;
+      if (temp <= b) { a = temp; break; }
+    } else if (a & ~c & m) {
+      temp = (c | m) & -m;
+      if (temp <= d) { c = temp; break; }
+    }
+    m >>= 1;
+  }
+  
+  return a | c;
+}
+static uint64_t maxOR(uint64_t a, uint64_t b,
+                      uint64_t c, uint64_t d) {
+  uint64_t temp, m = ((uint64_t) 1)<<63;
+
+  while (m) {
+    if (b & d & m) {
+      temp = (b - m) | (m - 1);
+      if (temp >= a) { b = temp; break; }
+      temp = (d - m) | (m -1);
+      if (temp >= c) { d = temp; break; }
+    }
+    m >>= 1;
+  }
+
+  return b | d;
+}
+static uint64_t minAND(uint64_t a, uint64_t b,
+                       uint64_t c, uint64_t d) {
+  uint64_t temp, m = ((uint64_t) 1)<<63;
+  while (m) {
+    if (~a & ~c & m) {
+      temp = (a | m) & -m;
+      if (temp <= b) { a = temp; break; }
+      temp = (c | m) & -m;
+      if (temp <= d) { c = temp; break; }
+    }
+    m >>= 1;
+  }
+  
+  return a & c;
+}
+static uint64_t maxAND(uint64_t a, uint64_t b,
+                       uint64_t c, uint64_t d) {
+  uint64_t temp, m = ((uint64_t) 1)<<63;
+  while (m) {
+    if (b & ~d & m) {
+      temp = (b & ~m) | (m - 1);
+      if (temp >= a) { b = temp; break; }
+    } else if (~b & d & m) {
+      temp = (d & ~m) | (m - 1);
+      if (temp >= c) { d = temp; break; }
+    }
+    m >>= 1;
+  }
+  
+  return b & d;
+}
+
+///
+
+class ValueRange {
+private:
+  uint64_t m_min, m_max;
+
+public:
+  ValueRange() : m_min(1),m_max(0) {}
+  ValueRange(uint64_t value) : m_min(value), m_max(value) {}
+  ValueRange(uint64_t _min, uint64_t _max) : m_min(_min), m_max(_max) {}
+  ValueRange(const ValueRange &b) : m_min(b.m_min), m_max(b.m_max) {}
+
+  void print(std::ostream &os) const {
+    if (isFixed()) {
+      os << m_min;
+    } else {
+      os << "[" << m_min << "," << m_max << "]";
+    }
+  }
+
+  bool isEmpty() const { 
+    return m_min>m_max; 
+  }
+  bool contains(uint64_t value) const { 
+    return this->intersects(ValueRange(value)); 
+  }
+  bool intersects(const ValueRange &b) const { 
+    return !this->set_intersection(b).isEmpty(); 
+  }
+
+  bool isFullRange(unsigned bits) {
+    return m_min==0 && m_max==bits64::maxValueOfNBits(bits);
+  }
+
+  ValueRange set_intersection(const ValueRange &b) const {
+    return ValueRange(std::max(m_min,b.m_min), std::min(m_max,b.m_max));
+  }
+  ValueRange set_union(const ValueRange &b) const {
+    return ValueRange(std::min(m_min,b.m_min), std::max(m_max,b.m_max));
+  }
+  ValueRange set_difference(const ValueRange &b) const {
+    if (b.isEmpty() || b.m_min > m_max || b.m_max < m_min) { // no intersection
+      return *this;
+    } else if (b.m_min <= m_min && b.m_max >= m_max) { // empty
+      return ValueRange(1,0); 
+    } else if (b.m_min <= m_min) { // one range out
+      // cannot overflow because b.m_max < m_max
+      return ValueRange(b.m_max+1, m_max);
+    } else if (b.m_max >= m_max) {
+      // cannot overflow because b.min > m_min
+      return ValueRange(m_min, b.m_min-1);
+    } else {
+      // two ranges, take bottom
+      return ValueRange(m_min, b.m_min-1);
+    }
+  }
+  ValueRange binaryAnd(const ValueRange &b) const {
+    // XXX
+    assert(!isEmpty() && !b.isEmpty() && "XXX");
+    if (isFixed() && b.isFixed()) {
+      return ValueRange(m_min & b.m_min);
+    } else {
+      return ValueRange(minAND(m_min, m_max, b.m_min, b.m_max),
+                        maxAND(m_min, m_max, b.m_min, b.m_max));
+    }
+  }
+  ValueRange binaryAnd(uint64_t b) const { return binaryAnd(ValueRange(b)); }
+  ValueRange binaryOr(ValueRange b) const {
+    // XXX
+    assert(!isEmpty() && !b.isEmpty() && "XXX");
+    if (isFixed() && b.isFixed()) {
+      return ValueRange(m_min | b.m_min);
+    } else {
+      return ValueRange(minOR(m_min, m_max, b.m_min, b.m_max),
+                        maxOR(m_min, m_max, b.m_min, b.m_max));
+    }
+  }
+  ValueRange binaryOr(uint64_t b) const { return binaryOr(ValueRange(b)); }
+  ValueRange binaryXor(ValueRange b) const {
+    if (isFixed() && b.isFixed()) {
+      return ValueRange(m_min ^ b.m_min);
+    } else {
+      uint64_t t = m_max | b.m_max;
+      while (!bits64::isPowerOfTwo(t))
+        t = bits64::withoutRightmostBit(t);
+      return ValueRange(0, (t<<1)-1);
+    }
+  }
+
+  ValueRange binaryShiftLeft(unsigned bits) const {
+    return ValueRange(m_min<<bits, m_max<<bits);
+  }
+  ValueRange binaryShiftRight(unsigned bits) const {
+    return ValueRange(m_min>>bits, m_max>>bits);
+  }
+
+  ValueRange concat(const ValueRange &b, unsigned bits) const {
+    return binaryShiftLeft(bits).binaryOr(b);
+  }
+  ValueRange extract(uint64_t lowBit, uint64_t maxBit) const {
+    return binaryShiftRight(lowBit).binaryAnd(bits64::maxValueOfNBits(maxBit-lowBit));
+  }
+
+  ValueRange add(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange sub(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange mul(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange udiv(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange sdiv(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange urem(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+  ValueRange srem(const ValueRange &b, unsigned width) const {
+    return ValueRange(0, bits64::maxValueOfNBits(width));
+  }
+
+  // use min() to get value if true (XXX should we add a method to
+  // make code clearer?)
+  bool isFixed() const { return m_min==m_max; }
+
+  bool operator==(const ValueRange &b) const { return m_min==b.m_min && m_max==b.m_max; }
+  bool operator!=(const ValueRange &b) const { return !(*this==b); }
+
+  bool mustEqual(const uint64_t b) const { return m_min==m_max && m_min==b; }
+  bool mayEqual(const uint64_t b) const { return m_min<=b && m_max>=b; }
+  
+  bool mustEqual(const ValueRange &b) const { return isFixed() && b.isFixed() && m_min==b.m_min; }
+  bool mayEqual(const ValueRange &b) const { return this->intersects(b); }
+
+  uint64_t min() const { 
+    assert(!isEmpty() && "cannot get minimum of empty range");
+    return m_min; 
+  }
+
+  uint64_t max() const { 
+    assert(!isEmpty() && "cannot get maximum of empty range");
+    return m_max; 
+  }
+  
+  int64_t minSigned(unsigned bits) const {
+    assert((m_min>>bits)==0 && (m_max>>bits)==0 &&
+           "range is outside given number of bits");
+
+    // if max allows sign bit to be set then it can be smallest value,
+    // otherwise since the range is not empty, min cannot have a sign
+    // bit
+
+    uint64_t smallest = ((uint64_t) 1 << (bits-1));
+    if (m_max >= smallest) {
+      return ints::sext(smallest, 64, bits);
+    } else {
+      return m_min;
+    }
+  }
+
+  int64_t maxSigned(unsigned bits) const {
+    assert((m_min>>bits)==0 && (m_max>>bits)==0 &&
+           "range is outside given number of bits");
+
+    uint64_t smallest = ((uint64_t) 1 << (bits-1));
+
+    // if max and min have sign bit then max is max, otherwise if only
+    // max has sign bit then max is largest signed integer, otherwise
+    // max is max
+
+    if (m_min < smallest && m_max >= smallest) {
+      return smallest - 1;
+    } else {
+      return ints::sext(m_max, 64, bits);
+    }
+  }
+};
+
+inline std::ostream &operator<<(std::ostream &os, const ValueRange &vr) {
+  vr.print(os);
+  return os;
+}
+
+// used to find all memory object ids and the maximum size of any
+// object state that references them (for symbolic size).
+class ObjectFinder : public ExprVisitor {
+protected:
+  Action visitRead(const ReadExpr &re) {
+    addUpdates(re.updates);
+    return Action::doChildren();
+  }
+
+  // XXX nice if this information was cached somewhere, used by
+  // independence as well right?
+  void addUpdates(const UpdateList &ul) {
+    for (const UpdateNode *un=ul.head; un; un=un->next) {
+      visit(un->index);
+      visit(un->value);
+    }
+
+    addObject(*ul.root);
+  }
+
+public:
+  void addObject(const Array& array) {
+    unsigned id = array.id;
+    std::map<unsigned,unsigned>::iterator it = results.find(id);
+
+    // FIXME: Not 64-bit size clean.
+    if (it == results.end()) {
+      results[id] = (unsigned) array.size;      
+    } else {
+      it->second = std::max(it->second, (unsigned) array.size);
+    }
+  }
+
+public:
+  std::map<unsigned, unsigned> results;
+};
+
+// XXX waste of space, rather have ByteValueRange
+typedef ValueRange CexValueData;
+
+class CexObjectData {
+public:
+  unsigned size;
+  CexValueData *values;
+
+public:
+  CexObjectData(unsigned _size) : size(_size), values(new CexValueData[size]) {
+    for (unsigned i=0; i<size; i++)
+      values[i] = ValueRange(0, 255);
+  }
+};
+
+class CexRangeEvaluator : public ExprRangeEvaluator<ValueRange> {
+public:
+  std::map<unsigned, CexObjectData> &objectValues;
+  CexRangeEvaluator(std::map<unsigned, CexObjectData> &_objectValues) 
+    : objectValues(_objectValues) {}
+
+  ValueRange getInitialReadRange(const Array &os, ValueRange index) {
+    return ValueRange(0, 255);
+  }
+};
+
+class CexConstifier : public ExprEvaluator {
+protected:
+  ref<Expr> getInitialValue(const Array& array, unsigned index) {
+    std::map<unsigned, CexObjectData>::iterator it = 
+      objectValues.find(array.id);
+    assert(it != objectValues.end() && "missing object?");
+    CexObjectData &cod = it->second;
+    
+    if (index >= cod.size) {
+      return ReadExpr::create(UpdateList(&array, true, 0), 
+                              ref<Expr>(index, Expr::Int32));
+    } else {
+      CexValueData &cvd = cod.values[index];
+      assert(cvd.min() == cvd.max() && "value is not fixed");
+      return ref<Expr>(cvd.min(), Expr::Int8);
+    }
+  }
+
+public:
+  std::map<unsigned, CexObjectData> &objectValues;
+  CexConstifier(std::map<unsigned, CexObjectData> &_objectValues) 
+    : objectValues(_objectValues) {}
+};
+
+class CexData {
+public:
+  std::map<unsigned, CexObjectData> objectValues;
+
+public:
+  CexData(ObjectFinder &finder) {
+    for (std::map<unsigned,unsigned>::iterator it = finder.results.begin(),
+           ie = finder.results.end(); it != ie; ++it) {
+      objectValues.insert(std::pair<unsigned, CexObjectData>(it->first, 
+                                                             CexObjectData(it->second)));
+    }
+  }  
+  ~CexData() {
+    for (std::map<unsigned, CexObjectData>::iterator it = objectValues.begin(),
+           ie = objectValues.end(); it != ie; ++it)
+      delete[] it->second.values;
+  }
+
+  void forceExprToValue(ref<Expr> e, uint64_t value) {
+    forceExprToRange(e, CexValueData(value,value));
+  }
+
+  void forceExprToRange(ref<Expr> e, CexValueData range) {
+#ifdef LOG
+    //    *theLog << "force: " << e << " to " << range << "\n";
+#endif
+    switch (e.getKind()) {
+    case Expr::Constant: {
+      // rather a pity if the constant isn't in the range, but how can
+      // we use this?
+      break;
+    }
+
+      // Special
+
+    case Expr::NotOptimized: break;
+
+    case Expr::Read: {
+      ReadExpr *re = static_ref_cast<ReadExpr>(e);
+      const Array *array = re->updates.root;
+      CexObjectData &cod = objectValues.find(array->id)->second;
+
+      // XXX we need to respect the version here and object state chain
+
+      if (re->index.isConstant() && 
+          re->index.getConstantValue() < array->size) {
+        CexValueData &cvd = cod.values[re->index.getConstantValue()];
+        CexValueData tmp = cvd.set_intersection(range);
+        
+        if (tmp.isEmpty()) {
+          if (range.isFixed()) // ranges conflict, if new one is fixed use that
+            cvd = range;
+        } else {
+          cvd = tmp;
+        }
+      } else {
+        // XXX        fatal("XXX not implemented");
+      }
+      
+      break;
+    }
+
+    case Expr::Select: {
+      SelectExpr *se = static_ref_cast<SelectExpr>(e);
+      ValueRange cond = evalRangeForExpr(se->cond);
+      if (cond.isFixed()) {
+        if (cond.min()) {
+          forceExprToRange(se->trueExpr, range);
+        } else {
+          forceExprToRange(se->falseExpr, range);
+        }
+      } else {
+        // XXX imprecise... we have a choice here. One method is to
+        // simply force both sides into the specified range (since the
+        // condition is indetermined). This may lose in two ways, the
+        // first is that the condition chosen may limit further
+        // restrict the range in each of the children, however this is
+        // less of a problem as the range will be a superset of legal
+        // values. The other is if the condition ends up being forced
+        // by some other constraints, then we needlessly forced one
+        // side into the given range.
+        //
+        // The other method would be to force the condition to one
+        // side and force that side into the given range. This loses
+        // when we force the condition to an unsatisfiable value
+        // (either because the condition cannot be that, or the
+        // resulting range given that condition is not in the required
+        // range).
+        // 
+        // Currently we just force both into the range. A hybrid would
+        // be to evaluate the ranges for each of the children... if
+        // one of the ranges happens to already be a subset of the
+        // required range then it may be preferable to force the
+        // condition to that side.
+        forceExprToRange(se->trueExpr, range);
+        forceExprToRange(se->falseExpr, range);
+      }
+      break;
+    }
+
+      // XXX imprecise... the problem here is that extracting bits
+      // loses information about what bits are connected across the
+      // bytes. if a value can be 1 or 256 then either the top or
+      // lower byte is 0, but just extraction loses this information
+      // and will allow neither,one,or both to be 1.
+      //
+      // we can protect against this in a limited fashion by writing
+      // the extraction a byte at a time, then checking the evaluated
+      // value, isolating for that range, and continuing.
+    case Expr::Concat: {
+      ConcatExpr *ce = static_ref_cast<ConcatExpr>(e);
+      if (ce->is2ByteConcat()) {
+	forceExprToRange(ce->getKid(0), range.extract( 8, 16));
+	forceExprToRange(ce->getKid(1), range.extract( 0,  8));
+      }
+      else if (ce->is4ByteConcat()) {
+	forceExprToRange(ce->getKid(0), range.extract(24, 32));
+	forceExprToRange(ce->getKid(1), range.extract(16, 24));
+	forceExprToRange(ce->getKid(2), range.extract( 8, 16));
+	forceExprToRange(ce->getKid(3), range.extract( 0,  8));
+      }
+      else if (ce->is8ByteConcat()) {
+	forceExprToRange(ce->getKid(0), range.extract(56, 64));
+	forceExprToRange(ce->getKid(1), range.extract(48, 56));
+	forceExprToRange(ce->getKid(2), range.extract(40, 48));
+	forceExprToRange(ce->getKid(3), range.extract(32, 40));
+	forceExprToRange(ce->getKid(4), range.extract(24, 32));
+	forceExprToRange(ce->getKid(5), range.extract(16, 24));
+	forceExprToRange(ce->getKid(6), range.extract( 8, 16));
+	forceExprToRange(ce->getKid(7), range.extract( 0,  8));
+      }
+      
+      break;
+    }
+
+    case Expr::Extract: {
+      // XXX
+      break;
+    }
+
+      // Casting
+
+      // Simply intersect the output range with the range of all
+      // possible outputs and then truncate to the desired number of
+      // bits. 
+
+      // For ZExt this simplifies to just intersection with the
+      // possible input range.
+    case Expr::ZExt: {
+      CastExpr *ce = static_ref_cast<CastExpr>(e);
+      unsigned inBits = ce->src.getWidth();
+      ValueRange input = range.set_intersection(ValueRange(0, bits64::maxValueOfNBits(inBits)));
+      forceExprToRange(ce->src, input);
+      break;
+    }
+      // For SExt instead of doing the intersection we just take the output range
+      // minus the impossible values. This is nicer since it is a single interval.
+    case Expr::SExt: {
+      CastExpr *ce = static_ref_cast<CastExpr>(e);
+      unsigned inBits = ce->src.getWidth();
+      unsigned outBits = ce->width;
+      ValueRange output = range.set_difference(ValueRange(1<<(inBits-1),
+                                                          (bits64::maxValueOfNBits(outBits)-
+                                                           bits64::maxValueOfNBits(inBits-1)-1)));
+      ValueRange input = output.binaryAnd(bits64::maxValueOfNBits(inBits));
+      forceExprToRange(ce->src, input);
+      break;
+    }
+
+      // Binary
+
+    case Expr::And: {
+      BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+      if (be->getWidth()==Expr::Bool) {
+        if (range.isFixed()) {
+          ValueRange left = evalRangeForExpr(be->left);
+          ValueRange right = evalRangeForExpr(be->right);
+
+          if (!range.min()) {
+            if (left.mustEqual(0) || right.mustEqual(0)) {
+              // all is well
+            } else {
+              // XXX heuristic, which order
+
+              forceExprToValue(be->left, 0);
+              left = evalRangeForExpr(be->left);
+
+              // see if that worked
+              if (!left.mustEqual(1))
+                forceExprToValue(be->right, 0);
+            }
+          } else {
+            if (!left.mustEqual(1)) forceExprToValue(be->left, 1);
+            if (!right.mustEqual(1)) forceExprToValue(be->right, 1);
+          }
+        }
+      } else {
+        // XXX
+      }
+      break;
+    }
+
+    case Expr::Or: {
+      BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+      if (be->getWidth()==Expr::Bool) {
+        if (range.isFixed()) {
+          ValueRange left = evalRangeForExpr(be->left);
+          ValueRange right = evalRangeForExpr(be->right);
+
+          if (range.min()) {
+            if (left.mustEqual(1) || right.mustEqual(1)) {
+              // all is well
+            } else {
+              // XXX heuristic, which order?
+              
+              // force left to value we need
+              forceExprToValue(be->left, 1);
+              left = evalRangeForExpr(be->left);
+
+              // see if that worked
+              if (!left.mustEqual(1))
+                forceExprToValue(be->right, 1);
+            }
+          } else {
+            if (!left.mustEqual(0)) forceExprToValue(be->left, 0);
+            if (!right.mustEqual(0)) forceExprToValue(be->right, 0);
+          }
+        }
+      } else {
+        // XXX
+      }
+      break;
+    }
+
+    case Expr::Xor: break;
+
+      // Comparison
+
+    case Expr::Eq: {
+      BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+      if (range.isFixed()) {
+        if (be->left.isConstant()) {
+          uint64_t value = be->left.getConstantValue();
+          if (range.min()) {
+            forceExprToValue(be->right, value);
+          } else {
+            if (value==0) {
+              forceExprToRange(be->right, 
+                               CexValueData(1,
+                                            ints::sext(1, 
+                                                       be->right.getWidth(),
+                                                       1)));
+            } else {
+              // XXX heuristic / lossy, could be better to pick larger range?
+              forceExprToRange(be->right, CexValueData(0, value-1));
+            }
+          }
+        } else {
+          // XXX what now
+        }
+      }
+      break;
+    }
+
+    case Expr::Ult: {
+      BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+      
+      // XXX heuristic / lossy, what order if conflict
+
+      if (range.isFixed()) {
+        ValueRange left = evalRangeForExpr(be->left);
+        ValueRange right = evalRangeForExpr(be->right);
+
+        uint64_t maxValue = bits64::maxValueOfNBits(be->right.getWidth());
+
+        // XXX should deal with overflow (can lead to empty range)
+
+        if (left.isFixed()) {
+          if (range.min()) {
+            forceExprToRange(be->right, CexValueData(left.min()+1, maxValue));
+          } else {
+            forceExprToRange(be->right, CexValueData(0, left.min()));
+          }
+        } else if (right.isFixed()) {
+          if (range.min()) {
+            forceExprToRange(be->left, CexValueData(0, right.min()-1));
+          } else {
+            forceExprToRange(be->left, CexValueData(right.min(), maxValue));
+          }
+        } else {
+          // XXX ???
+        }
+      }
+      break;
+    }
+    case Expr::Ule: {
+      BinaryExpr *be = static_ref_cast<BinaryExpr>(e);
+      
+      // XXX heuristic / lossy, what order if conflict
+
+      if (range.isFixed()) {
+        ValueRange left = evalRangeForExpr(be->left);
+        ValueRange right = evalRangeForExpr(be->right);
+
+        // XXX should deal with overflow (can lead to empty range)
+
+        uint64_t maxValue = bits64::maxValueOfNBits(be->right.getWidth());
+        if (left.isFixed()) {
+          if (range.min()) {
+            forceExprToRange(be->right, CexValueData(left.min(), maxValue));
+          } else {
+            forceExprToRange(be->right, CexValueData(0, left.min()-1));
+          }
+        } else if (right.isFixed()) {
+          if (range.min()) {
+            forceExprToRange(be->left, CexValueData(0, right.min()));
+          } else {
+            forceExprToRange(be->left, CexValueData(right.min()+1, maxValue));
+          }
+        } else {
+          // XXX ???
+        }
+      }
+      break;
+    }
+
+    case Expr::Ne:
+    case Expr::Ugt:
+    case Expr::Uge:
+    case Expr::Sgt:
+    case Expr::Sge:
+      assert(0 && "invalid expressions (uncanonicalized");
+
+    default:
+      break;
+    }
+  }
+
+  void fixValues() {
+    for (std::map<unsigned, CexObjectData>::iterator it = objectValues.begin(),
+           ie = objectValues.end(); it != ie; ++it) {
+      CexObjectData &cod = it->second;
+      for (unsigned i=0; i<cod.size; i++) {
+        CexValueData &cvd = cod.values[i];
+        cvd = CexValueData(cvd.min() + (cvd.max()-cvd.min())/2);
+      }
+    }
+  }
+
+  ValueRange evalRangeForExpr(ref<Expr> &e) {
+    CexRangeEvaluator ce(objectValues);
+    return ce.evaluate(e);
+  }
+
+  bool exprMustBeValue(ref<Expr> e, uint64_t value) {
+    CexConstifier cc(objectValues);
+    ref<Expr> v = cc.visit(e);
+    if (!v.isConstant()) return false;
+    // XXX reenable once all reads and vars are fixed
+    //    assert(v.isConstant() && "not all values have been fixed");
+    return v.getConstantValue()==value;
+  }
+};
+
+/* *** */
+
+
+class FastCexSolver : public IncompleteSolver {
+public:
+  FastCexSolver();
+  ~FastCexSolver();
+
+  IncompleteSolver::PartialValidity computeTruth(const Query&);  
+  bool computeValue(const Query&, ref<Expr> &result);
+  bool computeInitialValues(const Query&,
+                            const std::vector<const Array*> &objects,
+                            std::vector< std::vector<unsigned char> > &values,
+                            bool &hasSolution);
+};
+
+FastCexSolver::FastCexSolver() { }
+
+FastCexSolver::~FastCexSolver() { }
+
+IncompleteSolver::PartialValidity 
+FastCexSolver::computeTruth(const Query& query) {
+#ifdef LOG
+  std::ostringstream log;
+  theLog = &log;
+  //  log << "------ start FastCexSolver::mustBeTrue ------\n";
+  log << "-- QUERY --\n";
+  unsigned i=0;
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    log << "    C" << i++ << ": " << *it << ", \n";
+  log << "    Q : " << query.expr << "\n";
+#endif
+
+  ObjectFinder of;
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    of.visit(*it);
+  of.visit(query.expr);
+  CexData cd(of);
+
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    cd.forceExprToValue(*it, 1);
+  cd.forceExprToValue(query.expr, 0);
+
+#ifdef LOG
+  log << " -- ranges --\n";
+  for (std::map<unsigned, CexObjectData>::iterator it = objectValues.begin(),
+         ie = objectValues.end(); it != ie; ++it) {
+    CexObjectData &cod = it->second;
+    log << "    arr" << it->first << "[" << cod.size << "] = [";
+    unsigned continueFrom=cod.size-1;
+    for (; continueFrom>0; continueFrom--)
+      if (cod.values[continueFrom-1]!=cod.values[continueFrom])
+        break;
+    for (unsigned i=0; i<cod.size; i++) {
+      log << cod.values[i];
+      if (i<cod.size-1) {
+        log << ", ";
+        if (i==continueFrom) {
+          log << "...";
+          break;
+        }
+      }
+    }
+    log << "]\n";
+  }
+#endif
+
+  // this could be done lazily of course
+  cd.fixValues();
+
+#ifdef LOG
+  log << " -- fixed values --\n";
+  for (std::map<unsigned, CexObjectData>::iterator it = objectValues.begin(),
+         ie = objectValues.end(); it != ie; ++it) {
+    CexObjectData &cod = it->second;
+    log << "    arr" << it->first << "[" << cod.size << "] = [";
+    unsigned continueFrom=cod.size-1;
+    for (; continueFrom>0; continueFrom--)
+      if (cod.values[continueFrom-1]!=cod.values[continueFrom])
+        break;
+    for (unsigned i=0; i<cod.size; i++) {
+      log << cod.values[i];
+      if (i<cod.size-1) {
+        log << ", ";
+        if (i==continueFrom) {
+          log << "...";
+          break;
+        }
+      }
+    }
+    log << "]\n";
+  }
+#endif
+
+  // check the result
+
+  bool isGood = true;
+
+  if (!cd.exprMustBeValue(query.expr, 0)) isGood = false;
+
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    if (!cd.exprMustBeValue(*it, 1)) 
+      isGood = false;
+
+#ifdef LOG
+  log << " -- evaluating result --\n";
+  
+  i=0;
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it) {
+    bool res = cd.exprMustBeValue(*it, 1);
+    log << "    C" << i++ << ": " << (res?"true":"false") << "\n";
+  }
+  log << "    Q : " 
+      << (cd.exprMustBeValue(query.expr, 0)?"true":"false") << "\n";
+  
+  log << "\n\n";
+#endif
+
+  return isGood ? IncompleteSolver::MayBeFalse : IncompleteSolver::None;
+}
+
+bool FastCexSolver::computeValue(const Query& query, ref<Expr> &result) {
+  ObjectFinder of;
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    of.visit(*it);
+  of.visit(query.expr);
+  CexData cd(of);
+
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    cd.forceExprToValue(*it, 1);
+
+  // this could be done lazily of course
+  cd.fixValues();
+
+  // check the result
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    if (!cd.exprMustBeValue(*it, 1))
+      return false;
+
+  CexConstifier cc(cd.objectValues);
+  ref<Expr> value = cc.visit(query.expr);
+
+  if (value.isConstant()) {
+    result = value;
+    return true;
+  } else {
+    return false;
+  }
+}
+
+bool
+FastCexSolver::computeInitialValues(const Query& query,
+                                    const std::vector<const Array*>
+                                      &objects,
+                                    std::vector< std::vector<unsigned char> >
+                                      &values,
+                                    bool &hasSolution) {
+  ObjectFinder of;
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    of.visit(*it);
+  of.visit(query.expr);
+  for (unsigned i = 0; i != objects.size(); ++i)
+    of.addObject(*objects[i]);
+  CexData cd(of);
+
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    cd.forceExprToValue(*it, 1);
+  cd.forceExprToValue(query.expr, 0);
+
+  // this could be done lazily of course
+  cd.fixValues();
+
+  // check the result
+  for (ConstraintManager::const_iterator it = query.constraints.begin(), 
+         ie = query.constraints.end(); it != ie; ++it)
+    if (!cd.exprMustBeValue(*it, 1))
+      return false;
+  if (!cd.exprMustBeValue(query.expr, 0))
+    return false;
+
+  hasSolution = true;
+  CexConstifier cc(cd.objectValues);
+  for (unsigned i = 0; i != objects.size(); ++i) {
+    const Array *array = objects[i];
+    std::vector<unsigned char> data;
+    data.reserve(array->size);
+
+    for (unsigned i=0; i < array->size; i++) {
+      ref<Expr> value =
+        cc.visit(ReadExpr::create(UpdateList(array, true, 0),
+                                  ConstantExpr::create(i,
+                                                       kMachinePointerType)));
+      
+      if (value.isConstant()) {
+        data.push_back(value.getConstantValue());
+      } else {
+        // FIXME: When does this happen?
+        return false;
+      }
+    }
+
+    values.push_back(data);
+  }
+
+  return true;
+}
+
+
+Solver *klee::createFastCexSolver(Solver *s) {
+  return new Solver(new StagedSolverImpl(new FastCexSolver(), s));
+}

Added: klee/trunk/lib/Solver/IncompleteSolver.cpp
URL: http://llvm.org/viewvc/llvm-project/klee/trunk/lib/Solver/IncompleteSolver.cpp?rev=72205&view=auto

==============================================================================
--- klee/trunk/lib/Solver/IncompleteSolver.cpp (added)
+++ klee/trunk/lib/Solver/IncompleteSolver.cpp Wed May 20 23:36:41 2009
@@ -0,0 +1,136 @@
+//===-- IncompleteSolver.cpp ----------------------------------------------===//
+//
+//                     The KLEE Symbolic Virtual Machine
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "klee/IncompleteSolver.h"
+
+#include "klee/Constraints.h"
+
+using namespace klee;
+using namespace llvm;
+
+/***/
+
+IncompleteSolver::PartialValidity 
+IncompleteSolver::negatePartialValidity(PartialValidity pv) {
+  switch(pv) {
+    case MustBeTrue:  return MustBeFalse;
+    case MustBeFalse: return MustBeTrue;
+    case MayBeTrue:   return MayBeFalse;
+    case MayBeFalse:  return MayBeTrue;
+    case TrueOrFalse: return TrueOrFalse;
+    default: assert(0 && "invalid partial validity");  
+  }
+}
+
+IncompleteSolver::PartialValidity 
+IncompleteSolver::computeValidity(const Query& query) {
+  PartialValidity trueResult = computeTruth(query);
+
+  if (trueResult == MustBeTrue) {
+    return MustBeTrue;
+  } else {
+    PartialValidity falseResult = computeTruth(query.negateExpr());
+
+    if (falseResult == MustBeTrue) {
+      return MustBeFalse;
+    } else {
+      bool trueCorrect = trueResult != None,
+        falseCorrect = falseResult != None;
+      
+      if (trueCorrect && falseCorrect) {
+        return TrueOrFalse;
+      } else if (trueCorrect) { // ==> trueResult == MayBeFalse
+        return MayBeFalse;
+      } else if (falseCorrect) { // ==> falseRes