[llvm-commits] [llvm] r89421 - /llvm/trunk/lib/Analysis/CaptureTracking.cpp

[Chris Lattner]

On Nov 20, 2009, at 3:53 AM, Duncan Sands wrote:

> Hi Dan,
> 
>> +      // Don't count comparisons of the original value against null as captures.
>> +      // This allows us to ignore comparisons of malloc results with null,
>> +      // for example.
>> +      if (isIdentifiedObject(V))
>> +        if (ConstantPointerNull *CPN =
>> +              dyn_cast<ConstantPointerNull>(I->getOperand(1)))
>> +          if (CPN->getType()->getAddressSpace() == 0)
>> +            break;
> 
> I think this is wrong, consider the following pseudocode example:

While this example is "possible" I really don't think this is worth worrying about.  It is not valid C code, is not likely to exist in practice, etc.  Beyond that, comparison against null is really common and we really do want "nocapture" in this cases.

-Chris

> 
> i1 equals_null (i8* %P noalias) {
>   return %P == 0
> }
> 
> Thanks to your change, this function will be marked nocapture, because
> a noalias parameter counts as an identified object.
> 
> Now consider
> 
> i8* @capture (i8* %P noalias) {
>   n = 0
>   loop:
>     %Q = GEP %P, -n
>     if (equals_null(%Q))
>       break;
>     n = n + 1;
>     goto loop;
>   return GEP null, n
> }
> 
> I think it is correct to pass %Q to equals_null given that %P is noalias.
> So this example shows that %P can be captured even though CaptureTracking
> says it is not with your change.  A variant of this is to have %P be some
> noalias value like the return value of malloc.
> 
> Ciao,
> 
> Duncan.
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits