thr3ads.net - llvm dev - [LLVMdev] Wrong AliasAnalysis::getModRefInfo result [Feb 2012]

If this information is useful, please help other people find it:
Share via:

Welson Sun

2012-Feb-15 01:12 UTC

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

Just want to test out the LLVM's AliasAnalysis::getModRefInfo API. The
input C code is very simple:

void foo(int *a, int *b)
{
  for(int i=0; i<10; i++)
    b[i] = a[i]*a[i];
}

int main()
{
  int a[10];
  int b[10];

  for(int i=0; i<10; i++)
    a[i] = i;

  foo(a,b);

  return 0;
}

Obviously, for "foo", it only reads from array "a" and only
writes to array
"b".

The LLVM pass:
    virtual bool runOnFunction(Function &F) {
      ++HelloCounter;
      errs() << "Hello: ";
      errs().write_escaped(F.getName()) << '\n';

      AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
      for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
        Instruction *Inst = &*I;
        if ( CallInst *ci = dyn_cast<CallInst>(Inst) ){
          ci->dump();
          for(int i = 0; i < ci->getNumArgOperands(); i++){
            Value *v = ci->getArgOperand(i);
            if (GetElementPtrInst *vi = dyn_cast<GetElementPtrInst>(v)){
              Value *vPtr = vi->getPointerOperand();
              vPtr->dump();
              if ( AllocaInst *allo = dyn_cast<AllocaInst>(vPtr) ) {
                const Type *t = allo->getAllocatedType();
                if ( const ArrayType *at = dyn_cast<ArrayType>(t) ) {
                  int64_t size = at->getNumElements() *
at->getElementType()->getPrimitiveSizeInBits() / 8;
                  ImmutableCallSite cs(ci);
                  AliasAnalysis::Location loc(v, size);
                  errs() << AA.getModRefInfo(ci, loc) <<
"\n";
                }
              }
            }
          }
        }
      }

      return false;
    }


However, the result is "3" for both a and b, which is both read and
write.
What's the problem? I am not quite sure if I get the
AliasAnalysis::Location right, what is exactly "address-units" for the
size
of the location? And did I get the starting address of the Location right?
I tried v, vi and vPtr, same result.


Any insight helps,
Welson
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20120214/fab3ba30/attachment.html>

Welson Sun

2012-Feb-16 19:38 UTC

head link

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

Something must be wrong, more probable on my side. So the C source code is
unchanged, I just did another experiment to first extract all the GEPs in
the code, and call AliasAnalysis::alias on each pair of GEPs. Here is the
code:

      AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
      TargetData &TD = getAnalysis<TargetData>();

      for (Module::iterator it = M.begin(); it != M.end(); it++){
        Function &F = *it;
        errs().write_escaped(F.getName()) << '\n';
        for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
{
          Instruction *Inst = &*I;
          if ( GetElementPtrInst *gep = dyn_cast<GetElementPtrInst>(Inst)
)
            geps.push_back(gep);
        }
      }

      errs() << "List of GEPs:\n";
      for(std::list<GetElementPtrInst* >::iterator itGep1
            = geps.begin(); itGep1 != geps.end(); itGep1++){
        GetElementPtrInst *gep1 = *itGep1;
        Type *t1 = ((PointerType *)gep1->getType())->getElementType();
        int64_t size1 = TD.getTypeAllocSize(t1);
        std::list<GetElementPtrInst* >::iterator itGep2 = itGep1;
        for(; itGep2 != geps.end(); itGep2++){
          GetElementPtrInst *gep2 = *itGep2;
          gep1->dump();
          errs() << "-->";
          gep2->dump();

          Type *t2 = ((PointerType *)gep2->getType())->getElementType();
          int64_t size2 = TD.getTypeAllocSize(t2);
          AliasAnalysis::AliasResult aa = AA.alias(gep1, size1,
          errs() << "   alias: " << aa <<
"\n";
        }
      }


Here is the result:
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %9 = getelementptr inbounds i32* %8, i32 %7
   alias: 1
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %13 = getelementptr inbounds i32* %12, i32 %11
   alias: 1
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %18 = getelementptr inbounds i32* %17, i32 %16
   alias: 1
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
   alias: 1
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
   alias: 1
  %9 = getelementptr inbounds i32* %8, i32 %7
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1
  %13 = getelementptr inbounds i32* %12, i32 %11
-->  %13 = getelementptr inbounds i32* %12, i32 %11
   alias: 1
  %13 = getelementptr inbounds i32* %12, i32 %11
-->  %18 = getelementptr inbounds i32* %17, i32 %16
   alias: 1
  %13 = getelementptr inbounds i32* %12, i32 %11
-->  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
   alias: 1
  %13 = getelementptr inbounds i32* %12, i32 %11
-->  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
   alias: 1
  %13 = getelementptr inbounds i32* %12, i32 %11
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1
  %18 = getelementptr inbounds i32* %17, i32 %16
-->  %18 = getelementptr inbounds i32* %17, i32 %16
   alias: 1
  %18 = getelementptr inbounds i32* %17, i32 %16
-->  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
   alias: 1
  %18 = getelementptr inbounds i32* %17, i32 %16
-->  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
   alias: 1
  %18 = getelementptr inbounds i32* %17, i32 %16
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1
  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
-->  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
   alias: 1
  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
-->  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
   alias: 1
  %8 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 %7
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1
  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
-->  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
   alias: 1
  %13 = getelementptr inbounds [10 x i32]* %a, i32 0, i32 0
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1
  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
-->  %14 = getelementptr inbounds [10 x i32]* %b, i32 0, i32 0
   alias: 1

How can be each pair MayAlias? That's simply wrong! For example:

%13 = getelementptr inbounds [10 x i32]* *%a*, i32 0, i32 0
-->  %14 = getelementptr inbounds [10 x i32]* *%b*, i32 0, i32 0
Must not alias,

while
  %14 = getelementptr inbounds [10 x i32]* *%b*, i32 0, i32 0
-->  %14 = getelementptr inbounds [10 x i32]* *%b*, i32 0, i32 0
Must alias.

Can someone help identifying where is wrong with my code?


Thanks!
Welson









On Tue, Feb 14, 2012 at 5:12 PM, Welson Sun <welson.sun at gmail.com>
wrote:
> Just want to test out the LLVM's AliasAnalysis::getModRefInfo API. The
> input C code is very simple:
>
> void foo(int *a, int *b)
> {
>   for(int i=0; i<10; i++)
>     b[i] = a[i]*a[i];
> }
>
> int main()
> {
>   int a[10];
>   int b[10];
>
>   for(int i=0; i<10; i++)
>     a[i] = i;
>
>   foo(a,b);
>
>   return 0;
> }
>
> Obviously, for "foo", it only reads from array "a" and
only writes to
> array "b".
>
> The LLVM pass:
>     virtual bool runOnFunction(Function &F) {
>       ++HelloCounter;
>       errs() << "Hello: ";
>       errs().write_escaped(F.getName()) << '\n';
>
>       AliasAnalysis &AA = getAnalysis<AliasAnalysis>();
>       for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I) {
>         Instruction *Inst = &*I;
>         if ( CallInst *ci = dyn_cast<CallInst>(Inst) ){
>           ci->dump();
>           for(int i = 0; i < ci->getNumArgOperands(); i++){
>             Value *v = ci->getArgOperand(i);
>             if (GetElementPtrInst *vi =
dyn_cast<GetElementPtrInst>(v)){
>               Value *vPtr = vi->getPointerOperand();
>               vPtr->dump();
>               if ( AllocaInst *allo = dyn_cast<AllocaInst>(vPtr) ) {
>                 const Type *t = allo->getAllocatedType();
>                 if ( const ArrayType *at = dyn_cast<ArrayType>(t) ) {
>                   int64_t size = at->getNumElements() *
> at->getElementType()->getPrimitiveSizeInBits() / 8;
>                   ImmutableCallSite cs(ci);
>                   AliasAnalysis::Location loc(v, size);
>                   errs() << AA.getModRefInfo(ci, loc) <<
"\n";
>                 }
>               }
>             }
>           }
>         }
>       }
>
>       return false;
>     }
>
>
> However, the result is "3" for both a and b, which is both read
and write.
> What's the problem? I am not quite sure if I get the
> AliasAnalysis::Location right, what is exactly "address-units"
for the size
> of the location? And did I get the starting address of the Location right?
> I tried v, vi and vPtr, same result.
>
>
> Any insight helps,
> Welson
>


-- 
Welson

Phone: (408) 418-8385
Email:  welson.sun at gmail.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20120216/900e2c12/attachment.html>

Duncan Sands

2012-Feb-16 20:05 UTC

head link

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

Hi Welson, the default alias analysis is -no-aa.  As the name suggests it
just returns MayAlias for everything.  Maybe you are using that one?

Best wishes, Duncan.

Possibly Parallel Threads

Search for more seemingly similar threads

llvm dev - Feb 2012 - [LLVMdev] Wrong AliasAnalysis::getModRefInfo result

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

[LLVMdev] Wrong AliasAnalysis::getModRefInfo result

Possibly Parallel Threads