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] Add a one use check to the setcc inversion code in combineVSelectWithAllOnesOrZeros

Browse Source 
If we're going to generate a new inverted setcc, we should make sure we will be able to remove the old setcc.

Differential Revision: https://reviews.llvm.org/D56765

llvm-svn: 351378
This commit is contained in:
Craig Topper 2019-01-16 21:29:29 +00:00
parent 5d7dc5530c
commit 8e2e7a1d93
2 changed files with 13 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Target/X86/X86ISelLowering.cpp
View File

@ -33923,8 +33923,8 @@ combineVSelectWithAllOnesOrZeros(SDNode *N, SelectionDAG &DAG,
bool FValIsAllZeros = ISD::isBuildVectorAllZeros(RHS.getNode());
// Try to invert the condition if true value is not all 1s and false value is
// not all 0s.
if (!TValIsAllOnes && !FValIsAllZeros &&
// not all 0s. Only do this if the condition has one use.
if (!TValIsAllOnes && !FValIsAllZeros && Cond.hasOneUse() &&
// Check if the selector will be produced by CMPP*/PCMP*.
Cond.getOpcode() == ISD::SETCC &&
// Check if SETCC has already been promoted.

25
test/CodeGen/X86/vselect.ll
View File

@ -617,26 +617,23 @@ define void @vselect_allzeros_LHS_multiple_use_setcc(<4 x i32> %x, <4 x i32> %y,
; SSE: # %bb.0:
; SSE-NEXT: movdqa {{.*#+}} xmm3 = [1,2,4,8]
; SSE-NEXT: pand %xmm3, %xmm0
; SSE-NEXT: pcmpeqd %xmm0, %xmm3
; SSE-NEXT: pxor %xmm4, %xmm4
; SSE-NEXT: pcmpeqd %xmm0, %xmm4
; SSE-NEXT: pand %xmm1, %xmm4
; SSE-NEXT: pand %xmm2, %xmm3
; SSE-NEXT: movdqa %xmm4, (%rdi)
; SSE-NEXT: movdqa %xmm3, (%rsi)
; SSE-NEXT: pcmpeqd %xmm3, %xmm0
; SSE-NEXT: movdqa %xmm0, %xmm3
; SSE-NEXT: pandn %xmm1, %xmm3
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: movdqa %xmm3, (%rdi)
; SSE-NEXT: movdqa %xmm0, (%rsi)
; SSE-NEXT: retq
;
; AVX-LABEL: vselect_allzeros_LHS_multiple_use_setcc:
; AVX: # %bb.0:
; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = [1,2,4,8]
; AVX-NEXT: vpand %xmm3, %xmm0, %xmm0
; AVX-NEXT: vpcmpeqd %xmm3, %xmm0, %xmm3
; AVX-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX-NEXT: vpcmpeqd %xmm4, %xmm0, %xmm0
; AVX-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX-NEXT: vpand %xmm2, %xmm3, %xmm1
; AVX-NEXT: vmovdqa %xmm0, (%rdi)
; AVX-NEXT: vmovdqa %xmm1, (%rsi)
; AVX-NEXT: vpcmpeqd %xmm3, %xmm0, %xmm0
; AVX-NEXT: vpandn %xmm1, %xmm0, %xmm1
; AVX-NEXT: vpand %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmovdqa %xmm1, (%rdi)
; AVX-NEXT: vmovdqa %xmm0, (%rsi)
; AVX-NEXT: retq
%and = and <4 x i32> %x, <i32 1, i32 2, i32 4, i32 8>
%cond = icmp ne <4 x i32> %and, zeroinitializer