[llvm] 74f9839 - [X86][SSE] combineSetCCMOVMSK - allow comparison with upper (known zero) bits in CMP(MOVMSK(PACKSS())) -> CMP(MOVMSK()) fold

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2021-04-13T17:37:24+01:00
New Revision: 74f98391a7a42eccd7aa96546892cb9b65a39edb

URL: https://github.com/llvm/llvm-project/commit/74f98391a7a42eccd7aa96546892cb9b65a39edb
DIFF: https://github.com/llvm/llvm-project/commit/74f98391a7a42eccd7aa96546892cb9b65a39edb.diff

LOG: [X86][SSE] combineSetCCMOVMSK - allow comparison with upper (known zero) bits in CMP(MOVMSK(PACKSS())) -> CMP(MOVMSK()) fold

We already allow the comparison of the upper bits of 'IsAllOf' (allbits) patterns, but we can safely compare the known zero bits for 'IsAnyOf' (zerobits) patterns as well.

This fixes an issues where we are comparing a type wide than the number of vector elements, which avoids a regression mentioned in rGbaadbe04bf75.

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelLowering.cpp
    llvm/test/CodeGen/X86/vector-compare-all_of.ll
    llvm/test/CodeGen/X86/vector-compare-any_of.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 3d6035afa027..6c8051775915 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -42347,7 +42347,7 @@ static SDValue combineSetCCMOVMSK(SDValue EFLAGS, X86::CondCode &CC,
     }
     // PMOVMSKB(PACKSSBW(LO(X), HI(X)))
     // -> PMOVMSKB(BITCAST_v32i8(X)) & 0xAAAAAAAA.
-    if (CmpBits == 16 && Subtarget.hasInt256() &&
+    if (CmpBits >= 16 && Subtarget.hasInt256() &&
         VecOp0.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
         VecOp1.getOpcode() == ISD::EXTRACT_SUBVECTOR &&
         VecOp0.getOperand(0) == VecOp1.getOperand(0) &&

diff  --git a/llvm/test/CodeGen/X86/vector-compare-all_of.ll b/llvm/test/CodeGen/X86/vector-compare-all_of.ll
index e65083294b45..a4a8367f9329 100644
--- a/llvm/test/CodeGen/X86/vector-compare-all_of.ll
+++ b/llvm/test/CodeGen/X86/vector-compare-all_of.ll
@@ -743,10 +743,8 @@ define i16 @test_v16i16_legal_sext(<16 x i16> %a0, <16 x i16> %a1) {
 ; AVX2-LABEL: test_v16i16_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
-; AVX2-NEXT:    cmpl $65535, %eax # imm = 0xFFFF
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
+; AVX2-NEXT:    cmpl $-1, %eax
 ; AVX2-NEXT:    sete %al
 ; AVX2-NEXT:    negb %al
 ; AVX2-NEXT:    movsbl %al, %eax

diff  --git a/llvm/test/CodeGen/X86/vector-compare-any_of.ll b/llvm/test/CodeGen/X86/vector-compare-any_of.ll
index 26566599e201..c31a4c039eb3 100644
--- a/llvm/test/CodeGen/X86/vector-compare-any_of.ll
+++ b/llvm/test/CodeGen/X86/vector-compare-any_of.ll
@@ -645,9 +645,7 @@ define i16 @test_v16i16_legal_sext(<16 x i16> %a0, <16 x i16> %a1) {
 ; AVX2-LABEL: test_v16i16_legal_sext:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
-; AVX2-NEXT:    vpacksswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpmovmskb %xmm0, %eax
+; AVX2-NEXT:    vpmovmskb %ymm0, %eax
 ; AVX2-NEXT:    negl %eax
 ; AVX2-NEXT:    sbbl %eax, %eax
 ; AVX2-NEXT:    # kill: def $ax killed $ax killed $eax