[X86] Fix pshuflw formation from repeated shuffle mask (PR43230)

Fix for https://bugs.llvm.org/show_bug.cgi?id=43230.

When creating PSHUFLW from a repeated shuffle mask, we have to apply
the checks to the repeated mask, not the original one. For the test
case from PR43230 the inspected part of the original mask is all undef.

Differential Revision: https://reviews.llvm.org/D67314

llvm-svn: 371307

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 235d31bef8caba9727ae1ff8554667a837b53ce4..d3229296ba77b69c5fb4d10a5037dc086dc50948 100644 (file)
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -32041,8 +32041,8 @@ static bool matchUnaryPermuteShuffle(MVT MaskVT, ArrayRef<int> Mask,
   if (!ContainsZeros && AllowIntDomain && MaskScalarSizeInBits == 16) {
     SmallVector<int, 4> RepeatedMask;
     if (is128BitLaneRepeatedShuffleMask(MaskEltVT, Mask, RepeatedMask)) {
-      ArrayRef<int> LoMask(Mask.data() + 0, 4);
-      ArrayRef<int> HiMask(Mask.data() + 4, 4);
+      ArrayRef<int> LoMask(RepeatedMask.data() + 0, 4);
+      ArrayRef<int> HiMask(RepeatedMask.data() + 4, 4);
 
       // PSHUFLW: permute lower 4 elements only.
       if (isUndefOrInRange(LoMask, 0, 4) &&

diff --git a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
index 89e51ab4be05e232a52268f6d8a3a738ca6c0cea..001288f87d770cf6647462cad3f84cf7373e8ebf 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -4777,23 +4777,14 @@ define <16 x i16> @pr43230(<16 x i16> %a, <16 x i16> %b) {
 ; AVX1-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-SLOW-LABEL: pr43230:
-; AVX2-SLOW:       # %bb.0:
-; AVX2-SLOW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
-; AVX2-SLOW-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
-; AVX2-SLOW-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-SLOW-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[26,27],zero,zero
-; AVX2-SLOW-NEXT:    retq
-;
-; AVX2-FAST-LABEL: pr43230:
-; AVX2-FAST:       # %bb.0:
-; AVX2-FAST-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-FAST-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
-; AVX2-FAST-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
-; AVX2-FAST-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0
-; AVX2-FAST-NEXT:    vpand {{.*}}(%rip), %ymm0, %ymm0
-; AVX2-FAST-NEXT:    retq
+; AVX2-LABEL: pr43230:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm1 = ymm1[4],ymm2[4],ymm1[5],ymm2[5],ymm1[6],ymm2[6],ymm1[7],ymm2[7],ymm1[12],ymm2[12],ymm1[13],ymm2[13],ymm1[14],ymm2[14],ymm1[15],ymm2[15]
+; AVX2-NEXT:    vpunpckhwd {{.*#+}} ymm0 = ymm2[4],ymm0[4],ymm2[5],ymm0[5],ymm2[6],ymm0[6],ymm2[7],ymm0[7],ymm2[12],ymm0[12],ymm2[13],ymm0[13],ymm2[14],ymm0[14],ymm2[15],ymm0[15]
+; AVX2-NEXT:    vpsrlvd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm0[26,27],zero,zero
+; AVX2-NEXT:    retq
 ;
 ; AVX512VL-LABEL: pr43230:
 ; AVX512VL:       # %bb.0: