[llvm-commits] [llvm] r54144 - /llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp

[Chris Lattner]

On Jul 28, 2008, at 1:52 PM, Owen Anderson wrote:

> Author: resistor
> Date: Mon Jul 28 15:52:42 2008
> New Revision: 54144
>
> URL: http://llvm.org/viewvc/llvm-project?rev=54144&view=rev
> Log:
> Don't remove volatile loads.  Thanks to Duncan for noticing this one.

Shouldn't this check happen earlier?  Why even do the dependence check  
for stores that clearly can't be eliminated?

-Chris

>
>
> Modified:
>    llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp
>
> Modified: llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp?rev=54144&r1=54143&r2=54144&view=diff
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp  
> (original)
> +++ llvm/trunk/lib/Transforms/Scalar/DeadStoreElimination.cpp Mon  
> Jul 28 15:52:42 2008
> @@ -183,7 +183,7 @@
>         Instruction* dep = MD.getDependency(S);
>         DominatorTree& DT = getAnalysis<DominatorTree>();
>
> -        if (S->getParent() == L->getParent() &&
> +        if (!S->isVolatile() && S->getParent() == L->getParent() &&
>             S->getPointerOperand() == L->getPointerOperand() &&
>             ( dep == MemoryDependenceAnalysis::None ||
>               dep == MemoryDependenceAnalysis::NonLocal ||
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits