[llvm-commits] [llvm] r131493 - in /llvm/trunk: lib/Transforms/InstCombine/InstCombineCalls.cpp lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp test/CodeGen/X86/2011-05-17-pmovzxwd.ll
Stuart Hastings
stuart at apple.com
Tue May 17 15:13:31 PDT 2011
Author: stuart
Date: Tue May 17 17:13:31 2011
New Revision: 131493
URL: http://llvm.org/viewvc/llvm-project?rev=131493&view=rev
Log:
X86 pmovsx/pmovzx ignore the upper half of their inputs.
rdar://problem/6945110
Added:
llvm/trunk/test/CodeGen/X86/2011-05-17-pmovzxwd.ll
Modified:
llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
llvm/trunk/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp?rev=131493&r1=131492&r2=131493&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineCalls.cpp Tue May 17 17:13:31 2011
@@ -588,6 +588,28 @@
break;
}
+
+ case Intrinsic::x86_sse41_pmovsxbw:
+ case Intrinsic::x86_sse41_pmovsxwd:
+ case Intrinsic::x86_sse41_pmovsxdq:
+ case Intrinsic::x86_sse41_pmovzxbw:
+ case Intrinsic::x86_sse41_pmovzxwd:
+ case Intrinsic::x86_sse41_pmovzxdq: {
+ unsigned VWidth =
+ cast<VectorType>(II->getArgOperand(0)->getType())->getNumElements();
+ unsigned LowHalfElts = VWidth / 2;
+ APInt InputDemandedElts(VWidth, 0);
+ InputDemandedElts = InputDemandedElts.getBitsSet(VWidth, 0, LowHalfElts);
+ APInt UndefElts(VWidth, 0);
+ if (Value *TmpV = SimplifyDemandedVectorElts(II->getArgOperand(0),
+ InputDemandedElts,
+ UndefElts)) {
+ II->setArgOperand(0, TmpV);
+ return II;
+ }
+ break;
+ }
+
case Intrinsic::ppc_altivec_vperm:
// Turn vperm(V1,V2,mask) -> shuffle(V1,V2,mask) if mask is a constant.
if (ConstantVector *Mask = dyn_cast<ConstantVector>(II->getArgOperand(2))) {
Modified: llvm/trunk/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp?rev=131493&r1=131492&r2=131493&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp (original)
+++ llvm/trunk/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp Tue May 17 17:13:31 2011
@@ -867,7 +867,7 @@
if (Depth == 10)
return 0;
- // If multiple users are using the root value, procede with
+ // If multiple users are using the root value, proceed with
// simplification conservatively assuming that all elements
// are needed.
if (!V->hasOneUse()) {
Added: llvm/trunk/test/CodeGen/X86/2011-05-17-pmovzxwd.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/2011-05-17-pmovzxwd.ll?rev=131493&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/2011-05-17-pmovzxwd.ll (added)
+++ llvm/trunk/test/CodeGen/X86/2011-05-17-pmovzxwd.ll Tue May 17 17:13:31 2011
@@ -0,0 +1,15 @@
+; RUN: opt -instcombine -S < %s | FileCheck %s
+; <rdar://problem/6945110>
+
+define <4 x i32> @kernel3_vertical(<4 x i16> * %src, <8 x i16> * %foo) nounwind {
+entry:
+ %tmp = load <4 x i16>* %src
+ %tmp1 = load <8 x i16>* %foo
+; CHECK: shufflevector
+ %tmp2 = shufflevector <4 x i16> %tmp, <4 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NOT: shufflevector
+ %tmp3 = shufflevector <8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7>
+ %0 = call <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16> %tmp3)
+ ret <4 x i32> %0
+}
+declare <4 x i32> @llvm.x86.sse41.pmovzxwd(<8 x i16>) nounwind readnone
More information about the llvm-commits
mailing list