[llvm] r176058 - Fix PR10499

[Bill Wendling]

On Feb 25, 2013, at 3:01 PM, Michael Liao <michael.liao at intel.com> wrote:

> Author: hliao
> Date: Mon Feb 25 17:01:03 2013
> New Revision: 176058
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=176058&view=rev
> Log:
> Fix PR10499
> 
> - Check whether SSE is available before lowering all 1s vector building with
>  PCMPEQD, which is only available from SSE2
> 
> 
> Added:
>    llvm/trunk/test/CodeGen/X86/pr10499.ll
> Modified:
>    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> 
> Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=176058&r1=176057&r2=176058&view=diff
> ==============================================================================
> --- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
> +++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Mon Feb 25 17:01:03 2013
> @@ -5344,7 +5344,7 @@ X86TargetLowering::LowerBUILD_VECTOR(SDV
>   // Vectors containing all ones can be matched by pcmpeqd on 128-bit width
>   // vectors or broken into v4i32 operations on 256-bit vectors. AVX2 can use
>   // vpcmpeqd on 256-bit vectors.
> -  if (ISD::isBuildVectorAllOnes(Op.getNode())) {
> +  if (ISD::isBuildVectorAllOnes(Op.getNode()) && Subtarget->hasSSE2()) {

Hi Michael,

As a micro-optimization, could you put the "quick" checks first? (Subtarget-hasSSE2() should be quicker, I would think.) I don't think you need to resubmit a patch, but just something to consider for future patches. :)

-bw

>     if (VT == MVT::v4i32 || (VT == MVT::v8i32 && Subtarget->hasInt256()))
>       return Op;
> 
> 
> Added: llvm/trunk/test/CodeGen/X86/pr10499.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/pr10499.ll?rev=176058&view=auto
> ==============================================================================
> --- llvm/trunk/test/CodeGen/X86/pr10499.ll (added)
> +++ llvm/trunk/test/CodeGen/X86/pr10499.ll Mon Feb 25 17:01:03 2013
> @@ -0,0 +1,14 @@
> +; RUN: llc < %s -mtriple=x86_64-pc-linux -mcpu=corei7-avx -mattr=-sse2
> +
> +; No check as PR10499 is a crashing bug.
> +
> +define void @autogen_24438_500() {
> +BB:
> +  %I = insertelement <8 x i32> undef, i32 -1, i32 4
> +  %BC = bitcast <8 x i32> %I to <8 x float>
> +  br label %CF
> +
> +CF:                                               ; preds = %CF, %BB
> +  %ZE = fpext <8 x float> %BC to <8 x double>
> +  br label %CF
> +}
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits