RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 20:51:52 +01:00
Code Issues Projects Releases Wiki Activity

[X86][SSE] Fold MOVMSK(PCMPEQ(X,0)) != -1 -> !PTESTZ(X,X) allof patterns

Browse Source
This commit is contained in:
Simon Pilgrim 2020-06-19 16:05:04 +01:00
parent add62d20bd
commit 92c7d023b2
2 changed files with 21 additions and 29 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
lib/Target/X86/X86ISelLowering.cpp
View File

@ -40561,7 +40561,8 @@ static SDValue combineSetCCMOVMSK(SDValue EFLAGS, X86::CondCode &CC,
}
// MOVMSK(PCMPEQ(X,0)) == -1 -> PTESTZ(X,X).
if ((IsAllOf && CC == X86::COND_E) && Subtarget.hasSSE41()) {
// MOVMSK(PCMPEQ(X,0)) != -1 -> !PTESTZ(X,X).
if (IsAllOf && Subtarget.hasSSE41()) {
SDValue BC = peekThroughBitcasts(Vec);
if (BC.getOpcode() == X86ISD::PCMPEQ &&
ISD::isBuildVectorAllZeros(BC.getOperand(1).getNode())) {

47
test/CodeGen/X86/movmsk-cmp.ll
View File

@ -998,30 +998,30 @@ define i1 @allones_v16i8_and1(<16 x i8> %arg) {
}
define i1 @allzeros_v16i8_not(<16 x i8> %a0) {
; SSE-LABEL: allzeros_v16i8_not:
; SSE: # %bb.0:
; SSE-NEXT: pxor %xmm1, %xmm1
; SSE-NEXT: pcmpeqb %xmm0, %xmm1
; SSE-NEXT: pmovmskb %xmm1, %eax
; SSE-NEXT: cmpw $-1, %ax
; SSE-NEXT: setne %al
; SSE-NEXT: retq
; SSE2-LABEL: allzeros_v16i8_not:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm1, %xmm1
; SSE2-NEXT: pcmpeqb %xmm0, %xmm1
; SSE2-NEXT: pmovmskb %xmm1, %eax
; SSE2-NEXT: cmpw $-1, %ax
; SSE2-NEXT: setne %al
; SSE2-NEXT: retq
;
; SSE41-LABEL: allzeros_v16i8_not:
; SSE41: # %bb.0:
; SSE41-NEXT: ptest %xmm0, %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: allzeros_v16i8_not:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vpmovmskb %xmm0, %eax
; AVX1OR2-NEXT: cmpw $-1, %ax
; AVX1OR2-NEXT: vptest %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
; KNL-LABEL: allzeros_v16i8_not:
; KNL: # %bb.0:
; KNL-NEXT: vpxor %xmm1, %xmm1, %xmm1
; KNL-NEXT: vpcmpeqb %xmm1, %xmm0, %xmm0
; KNL-NEXT: vpmovmskb %xmm0, %eax
; KNL-NEXT: cmpw $-1, %ax
; KNL-NEXT: vptest %xmm0, %xmm0
; KNL-NEXT: setne %al
; KNL-NEXT: retq
;
@ -1051,19 +1051,13 @@ define i1 @allzeros_v2i64_not(<2 x i64> %a0) {
;
; SSE41-LABEL: allzeros_v2i64_not:
; SSE41: # %bb.0:
; SSE41-NEXT: pxor %xmm1, %xmm1
; SSE41-NEXT: pcmpeqq %xmm0, %xmm1
; SSE41-NEXT: movmskpd %xmm1, %eax
; SSE41-NEXT: cmpb $3, %al
; SSE41-NEXT: ptest %xmm0, %xmm0
; SSE41-NEXT: setne %al
; SSE41-NEXT: retq
;
; AVX1OR2-LABEL: allzeros_v2i64_not:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX1OR2-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
; AVX1OR2-NEXT: vmovmskpd %xmm0, %eax
; AVX1OR2-NEXT: cmpb $3, %al
; AVX1OR2-NEXT: vptest %xmm0, %xmm0
; AVX1OR2-NEXT: setne %al
; AVX1OR2-NEXT: retq
;
@ -1119,10 +1113,7 @@ define i1 @allzeros_v8i32_not(<8 x i32> %a0) {
;
; AVX2-LABEL: allzeros_v8i32_not:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
; AVX2-NEXT: vpcmpeqd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vmovmskps %ymm0, %eax
; AVX2-NEXT: cmpb $-1, %al
; AVX2-NEXT: vptest %ymm0, %ymm0
; AVX2-NEXT: setne %al
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq