[llvm] 81003f2 - [X86] combineBitcastvxi1 - peek through FREEZE nodes when determining width of the original SETCC node

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2024-05-03T13:22:16+01:00
New Revision: 81003f23f0361aeb05b98489ba659edad40506f8

URL: https://github.com/llvm/llvm-project/commit/81003f23f0361aeb05b98489ba659edad40506f8
DIFF: https://github.com/llvm/llvm-project/commit/81003f23f0361aeb05b98489ba659edad40506f8.diff

LOG: [X86] combineBitcastvxi1 - peek through FREEZE nodes when determining width of the original SETCC node

Fixes #90847

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelLowering.cpp
    llvm/test/CodeGen/X86/pr90847.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 9ac3f6d4445e4f..cf4a64ffded2e8 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -42836,6 +42836,8 @@ static bool checkBitcastSrcVectorSize(SDValue Src, unsigned Size,
     [[fallthrough]];
   case ISD::SETCC:
     return Src.getOperand(0).getValueSizeInBits() == Size;
+  case ISD::FREEZE:
+    return checkBitcastSrcVectorSize(Src.getOperand(0), Size, AllowTruncate);
   case ISD::AND:
   case ISD::XOR:
   case ISD::OR:
@@ -42905,6 +42907,7 @@ static SDValue signExtendBitcastSrcVector(SelectionDAG &DAG, EVT SExtVT,
                                           SDValue Src, const SDLoc &DL) {
   switch (Src.getOpcode()) {
   case ISD::SETCC:
+  case ISD::FREEZE:
   case ISD::TRUNCATE:
   case ISD::BUILD_VECTOR:
     return DAG.getNode(ISD::SIGN_EXTEND, DL, SExtVT, Src);

diff  --git a/llvm/test/CodeGen/X86/pr90847.ll b/llvm/test/CodeGen/X86/pr90847.ll
index 65cebfb88df899..7aa0ceb26e1acb 100644
--- a/llvm/test/CodeGen/X86/pr90847.ll
+++ b/llvm/test/CodeGen/X86/pr90847.ll
@@ -2,7 +2,7 @@
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx  | FileCheck %s --check-prefixes=AVX1
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2 | FileCheck %s --check-prefixes=AVX2
 
-; TODO: PR90847 - failure to peek through FREEZE(SETCC()) results in VPMOVSMSKB(TRUNC()) instead of VMOVMSKPS
+; PR90847 - failure to peek through FREEZE(SETCC()) results in VPMOVSMSKB(TRUNC()) instead of VMOVMSKPS
 
 define i32 @PR90847(<8 x float> %x) nounwind {
 ; AVX1-LABEL: PR90847:
@@ -14,14 +14,10 @@ define i32 @PR90847(<8 x float> %x) nounwind {
 ; AVX1-NEXT:    vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
 ; AVX1-NEXT:    vminps %ymm2, %ymm1, %ymm1
 ; AVX1-NEXT:    vcmpeqps %ymm0, %ymm1, %ymm0
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
-; AVX1-NEXT:    vpmovmskb %xmm0, %eax
-; AVX1-NEXT:    testb %al, %al
+; AVX1-NEXT:    vmovmskps %ymm0, %eax
+; AVX1-NEXT:    testl %eax, %eax
 ; AVX1-NEXT:    je .LBB0_1
 ; AVX1-NEXT:  # %bb.2: # %cond.false
-; AVX1-NEXT:    movzbl %al, %eax
 ; AVX1-NEXT:    rep bsfl %eax, %eax
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
@@ -39,14 +35,10 @@ define i32 @PR90847(<8 x float> %x) nounwind {
 ; AVX2-NEXT:    vpermpd {{.*#+}} ymm2 = ymm1[2,3,0,1]
 ; AVX2-NEXT:    vminps %ymm2, %ymm1, %ymm1
 ; AVX2-NEXT:    vcmpeqps %ymm0, %ymm1, %ymm0
-; AVX2-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpacksswb %xmm0, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
-; AVX2-NEXT:    testb %al, %al
+; AVX2-NEXT:    vmovmskps %ymm0, %eax
+; AVX2-NEXT:    testl %eax, %eax
 ; AVX2-NEXT:    je .LBB0_1
 ; AVX2-NEXT:  # %bb.2: # %cond.false
-; AVX2-NEXT:    movzbl %al, %eax
 ; AVX2-NEXT:    rep bsfl %eax, %eax
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq