[llvm-commits] [llvm] r145865 - in /llvm/trunk: lib/Transforms/Scalar/MemCpyOptimizer.cpp test/Transforms/MemCpyOpt/form-memset.ll

[Bob Wilson]

I think you meant <rdar://problem/9892684>

On Dec 5, 2011, at 2:37 PM, Chad Rosier wrote:

> Author: mcrosier
> Date: Mon Dec  5 16:37:00 2011
> New Revision: 145865
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=145865&view=rev
> Log:
> Make the MemCpyOptimizer a bit more aggressive.  I can't think of a scenerio
> where this would be bad as the backend shouldn't have a problem inlining small
> memcpys.
> rdar://10510150
> 
> Modified:
>    llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
>    llvm/trunk/test/Transforms/MemCpyOpt/form-memset.ll
> 
> Modified: llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp?rev=145865&r1=145864&r2=145865&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp (original)
> +++ llvm/trunk/lib/Transforms/Scalar/MemCpyOptimizer.cpp Mon Dec  5 16:37:00 2011
> @@ -148,7 +148,7 @@
> 
> bool MemsetRange::isProfitableToUseMemset(const TargetData &TD) const {
>   // If we found more than 8 stores to merge or 64 bytes, use memset.
> -  if (TheStores.size() >= 8 || End-Start >= 64) return true;
> +  if (TheStores.size() >= 4 || End-Start >= 16) return true;
> 
>   // If there is nothing to merge, don't do anything.
>   if (TheStores.size() < 2) return false;
> 
> Modified: llvm/trunk/test/Transforms/MemCpyOpt/form-memset.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/MemCpyOpt/form-memset.ll?rev=145865&r1=145864&r2=145865&view=diff
> ==============================================================================
> --- llvm/trunk/test/Transforms/MemCpyOpt/form-memset.ll (original)
> +++ llvm/trunk/test/Transforms/MemCpyOpt/form-memset.ll Mon Dec  5 16:37:00 2011
> @@ -57,8 +57,8 @@
> 
> declare i32 @bar(...)
> 
> +%struct.MV = type { i16, i16 }
> 
> -	%struct.MV = type { i16, i16 }
> 
> define void @test2() nounwind  {
> entry:
> @@ -220,3 +220,19 @@
> ; CHECK: call void @llvm.memset.p0i8.i64(i8* %2, i8 0, i64 24, i32 1, i1 false)
> }
> 
> +; More aggressive heuristic
> +; rdar://9892684
> +define void @test7(i32* nocapture %c) nounwind optsize {
> +  store i32 -1, i32* %c, align 4
> +  %1 = getelementptr inbounds i32* %c, i32 1
> +  store i32 -1, i32* %1, align 4
> +  %2 = getelementptr inbounds i32* %c, i32 2
> +  store i32 -1, i32* %2, align 4
> +  %3 = getelementptr inbounds i32* %c, i32 3
> +  store i32 -1, i32* %3, align 4
> +  %4 = getelementptr inbounds i32* %c, i32 4
> +  store i32 -1, i32* %4, align 4
> +; CHECK: @test7
> +; CHECK: call void @llvm.memset.p0i8.i64(i8* %5, i8 -1, i64 20, i32 4, i1 false)
> +  ret void
> +}
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits