[X86] canCreateUndefOrPoisonForTargetNode/isGuaranteedNotToBeUndefOrPoisonForTargetNode - add X86ISD::VPERMILPV handling

[RKSimon]

X86ISD::VPERMILPV shuffles can't create undef/poison itself, allowing us to fold freeze(vpermilps(x,y)) -> vpermilps(freeze(x),freeze(y))

[llvmbot]

@llvm/pr-subscribers-backend-x86

Author: Simon Pilgrim (RKSimon)

Changes

X86ISD::VPERMILPV shuffles can't create undef/poison itself, allowing us to fold freeze(vpermilps(x,y)) -> vpermilps(freeze(x),freeze(y))

---

Full diff: https://github.com/llvm/llvm-project/pull/160849.diff

2 Files Affected:

• (modified) llvm/lib/Target/X86/X86ISelLowering.cpp (+2)
• (modified) llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll (-2)

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 57cf66e5275e3..633094df297d6 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -45188,6 +45188,7 @@ bool X86TargetLowering::isGuaranteedNotToBeUndefOrPoisonForTargetNode(
   case X86ISD::PSHUFD:
   case X86ISD::UNPCKL:
   case X86ISD::UNPCKH:
+  case X86ISD::VPERMILPV:
   case X86ISD::VPERMILPI:
   case X86ISD::VPERMV3: {
     SmallVector<int, 8> Mask;
@@ -45254,6 +45255,7 @@ bool X86TargetLowering::canCreateUndefOrPoisonForTargetNode(
   case X86ISD::PSHUFD:
   case X86ISD::UNPCKL:
   case X86ISD::UNPCKH:
+  case X86ISD::VPERMILPV:
   case X86ISD::VPERMILPI:
   case X86ISD::VPERMV3:
     return false;
diff --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
index c9d9db6cc9578..3279a50a1265b 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx.ll
@@ -373,8 +373,6 @@ define <8 x float> @constant_fold_vpermilvar_ps_256() {
 define <8 x float> @freeze_vpermilvar_ps_256(<8 x float> %a0) {
 ; CHECK-LABEL: freeze_vpermilvar_ps_256:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,3,1,2,7,6,5,4]
-; CHECK-NEXT:    vpermilps {{.*#+}} ymm0 = ymm0[0,2,3,1,7,6,5,4]
 ; CHECK-NEXT:    ret{{[l|q]}}
   %s0 = call <8 x float> @llvm.x86.avx.vpermilvar.ps.256(<8 x float> %a0, <8 x i32> <i32 0, i32 3, i32 1, i32 2, i32 7, i32 6, i32 5, i32 4>)
   %f0 = freeze <8 x float> %s0