[llvm] r201140 - AVX: fixed a bug in LowerVECTOR_SHUFFLE

[Elena Demikhovsky]

Author: delena
Date: Tue Feb 11 04:21:53 2014
New Revision: 201140

URL: http://llvm.org/viewvc/llvm-project?rev=201140&view=rev
Log:
AVX: fixed a bug in LowerVECTOR_SHUFFLE

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
    llvm/trunk/test/CodeGen/X86/avx-shuffle.ll

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=201140&r1=201139&r2=201140&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Tue Feb 11 04:21:53 2014
@@ -7371,7 +7371,11 @@ X86TargetLowering::LowerVECTOR_SHUFFLE(S
   if (V1IsUndef && V2IsUndef)
     return DAG.getUNDEF(VT);
 
-  assert(!V1IsUndef && "Op 1 of shuffle should not be undef");
+  // When we create a shuffle node we put the UNDEF node to second operand,
+  // but in some cases the first operand may be transformed to UNDEF.
+  // In this case we should just commute the node.
+  if (V1IsUndef)
+    return CommuteVectorShuffle(SVOp, DAG);
 
   // Vector shuffle lowering takes 3 steps:
   //

Modified: llvm/trunk/test/CodeGen/X86/avx-shuffle.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-shuffle.ll?rev=201140&r1=201139&r2=201140&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-shuffle.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-shuffle.ll Tue Feb 11 04:21:53 2014
@@ -297,3 +297,12 @@ entry:
 }
 declare <2 x double> @llvm.x86.avx.vextractf128.pd.256(<4 x double>, i8) nounwind readnone
 declare <4 x double> @llvm.x86.avx.vinsertf128.pd.256(<4 x double>, <2 x double>, i8) nounwind readnone
+
+; this test case just should not fail
+define void @test20() {
+  %a0 = insertelement <3 x double> <double 0.000000e+00, double 0.000000e+00, double undef>, double 0.000000e+00, i32 2
+  store <3 x double> %a0, <3 x double>* undef, align 1
+  %a1 = insertelement <3 x double> <double 0.000000e+00, double 0.000000e+00, double undef>, double undef, i32 2
+  store <3 x double> %a1, <3 x double>* undef, align 1
+  ret void
+}