Skip to content

Commit eb13ebd

Browse files
topperctopperc
committedJan 29, 2018
[X86] Don't create SHRUNKBLEND when the condition is used by the true or false operand of the vselect.
Fixes PR34592. Differential Revision: https://reviews.llvm.org/D42628 llvm-svn: 323672
1 parent 63db1c1 commit eb13ebd

File tree

2 files changed

+38
-34
lines changed

2 files changed

+38
-34
lines changed
 

‎llvm/lib/Target/X86/X86ISelLowering.cpp

+3-2
Original file line numberDiff line numberDiff line change
@@ -31828,9 +31828,10 @@ static SDValue combineSelect(SDNode *N, SelectionDAG &DAG,
3182831828
// Check all uses of the condition operand to check whether it will be
3182931829
// consumed by non-BLEND instructions. Those may require that all bits
3183031830
// are set properly.
31831-
for (SDNode *U : Cond->uses()) {
31831+
for (SDNode::use_iterator UI = Cond->use_begin(), UE = Cond->use_end();
31832+
UI != UE; ++UI) {
3183231833
// TODO: Add other opcodes eventually lowered into BLEND.
31833-
if (U->getOpcode() != ISD::VSELECT)
31834+
if (UI->getOpcode() != ISD::VSELECT || UI.getOperandNo() != 0)
3183431835
return SDValue();
3183531836
}
3183631837

‎llvm/test/CodeGen/X86/pr34592.ll

+35-32
Original file line numberDiff line numberDiff line change
@@ -10,7 +10,7 @@ define <16 x i64> @pluto(<16 x i64> %arg, <16 x i64> %arg1, <16 x i64> %arg2, <1
1010
; CHECK-NEXT: movq %rsp, %rbp
1111
; CHECK-NEXT: .cfi_def_cfa_register %rbp
1212
; CHECK-NEXT: andq $-32, %rsp
13-
; CHECK-NEXT: subq $128, %rsp
13+
; CHECK-NEXT: subq $288, %rsp # imm = 0x120
1414
; CHECK-NEXT: vmovaps 240(%rbp), %ymm8
1515
; CHECK-NEXT: vmovaps 208(%rbp), %ymm9
1616
; CHECK-NEXT: vmovaps 176(%rbp), %ymm10
@@ -23,39 +23,42 @@ define <16 x i64> @pluto(<16 x i64> %arg, <16 x i64> %arg1, <16 x i64> %arg2, <1
2323
; CHECK-NEXT: vmovaps {{.*#+}} ymm0 = [0,0,18446744071562067968,18446744071562067968]
2424
; CHECK-NEXT: vblendvpd %ymm0, %ymm2, %ymm6, %ymm0
2525
; CHECK-NEXT: vxorps %xmm2, %xmm2, %xmm2
26-
; CHECK-NEXT: vblendvpd %ymm2, %ymm3, %ymm7, %ymm3
27-
; CHECK-NEXT: vblendvpd %ymm2, %ymm1, %ymm5, %ymm1
28-
; CHECK-NEXT: vpblendd {{.*#+}} ymm5 = ymm2[0,1],ymm13[2,3],ymm2[4,5,6,7]
29-
; CHECK-NEXT: vpblendd {{.*#+}} ymm6 = ymm14[0,1],ymm2[2,3,4,5],ymm14[6,7]
30-
; CHECK-NEXT: vpblendd {{.*#+}} ymm7 = ymm15[0,1],ymm11[2,3,4,5,6,7]
31-
; CHECK-NEXT: vblendvpd %ymm2, %ymm10, %ymm6, %ymm6
32-
; CHECK-NEXT: vmovaps {{.*#+}} ymm10 = [18446744071562067968,18446744071562067968,0,0]
33-
; CHECK-NEXT: vblendvpd %ymm10, %ymm9, %ymm5, %ymm5
34-
; CHECK-NEXT: vblendvpd %ymm2, %ymm8, %ymm2, %ymm2
35-
; CHECK-NEXT: vpshufd {{.*#+}} ymm8 = ymm1[0,1,0,1,4,5,4,5]
36-
; CHECK-NEXT: vpblendd {{.*#+}} ymm9 = ymm3[0,1],ymm2[2,3],ymm3[4,5,6,7]
37-
; CHECK-NEXT: vpermq {{.*#+}} ymm9 = ymm9[2,1,1,3]
38-
; CHECK-NEXT: vpblendd {{.*#+}} ymm8 = ymm9[0,1,2,3,4,5],ymm8[6,7]
26+
; CHECK-NEXT: vpblendd {{.*#+}} ymm6 = ymm2[0,1],ymm13[2,3],ymm2[4,5,6,7]
27+
; CHECK-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1],ymm8[2,3,4,5,6,7]
28+
; CHECK-NEXT: vmovaps {{.*#+}} ymm8 = [18446744071562067968,18446744071562067968,0,0]
29+
; CHECK-NEXT: vblendvpd %ymm8, %ymm9, %ymm6, %ymm6
30+
; CHECK-NEXT: vpblendd {{.*#+}} ymm8 = ymm15[0,1],ymm11[2,3,4,5,6,7]
31+
; CHECK-NEXT: vpalignr {{.*#+}} ymm8 = ymm0[8,9,10,11,12,13,14,15],ymm8[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm8[16,17,18,19,20,21,22,23]
32+
; CHECK-NEXT: vpermq {{.*#+}} ymm8 = ymm8[2,3,2,0]
33+
; CHECK-NEXT: vmovaps %xmm6, %xmm9
34+
; CHECK-NEXT: # implicit-def: %ymm11
35+
; CHECK-NEXT: vinserti128 $1, %xmm9, %ymm11, %ymm11
36+
; CHECK-NEXT: vpblendd {{.*#+}} ymm8 = ymm8[0,1,2,3],ymm11[4,5],ymm8[6,7]
3937
; CHECK-NEXT: vmovaps %xmm0, %xmm9
40-
; CHECK-NEXT: # implicit-def: %ymm10
41-
; CHECK-NEXT: vinserti128 $1, %xmm9, %ymm10, %ymm10
42-
; CHECK-NEXT: vpunpcklqdq {{.*#+}} ymm2 = ymm3[0],ymm2[0],ymm3[2],ymm2[2]
43-
; CHECK-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,1,2,3]
44-
; CHECK-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm10[4,5,6,7]
45-
; CHECK-NEXT: vpalignr {{.*#+}} ymm1 = ymm5[8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4,5,6,7],ymm5[24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20,21,22,23]
46-
; CHECK-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,3]
47-
; CHECK-NEXT: vpunpcklqdq {{.*#+}} ymm3 = ymm6[0],ymm3[0],ymm6[2],ymm3[2]
48-
; CHECK-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,1,2,3]
49-
; CHECK-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1,2,3],ymm1[4,5,6,7]
50-
; CHECK-NEXT: vpalignr {{.*#+}} ymm0 = ymm0[8,9,10,11,12,13,14,15],ymm7[0,1,2,3,4,5,6,7],ymm0[24,25,26,27,28,29,30,31],ymm7[16,17,18,19,20,21,22,23]
51-
; CHECK-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,3,2,0]
52-
; CHECK-NEXT: vmovaps %xmm5, %xmm9
53-
; CHECK-NEXT: # implicit-def: %ymm1
54-
; CHECK-NEXT: vinserti128 $1, %xmm9, %ymm1, %ymm1
55-
; CHECK-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5],ymm0[6,7]
56-
; CHECK-NEXT: vmovaps %ymm8, %ymm1
38+
; CHECK-NEXT: # implicit-def: %ymm0
39+
; CHECK-NEXT: vinserti128 $1, %xmm9, %ymm0, %ymm0
40+
; CHECK-NEXT: vpunpcklqdq {{.*#+}} ymm11 = ymm7[0],ymm2[0],ymm7[2],ymm2[2]
41+
; CHECK-NEXT: vpermq {{.*#+}} ymm11 = ymm11[2,1,2,3]
42+
; CHECK-NEXT: vpblendd {{.*#+}} ymm0 = ymm11[0,1,2,3],ymm0[4,5,6,7]
43+
; CHECK-NEXT: vpblendd {{.*#+}} ymm2 = ymm7[0,1],ymm2[2,3],ymm7[4,5,6,7]
44+
; CHECK-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,1,1,3]
45+
; CHECK-NEXT: vpshufd {{.*#+}} ymm11 = ymm5[0,1,0,1,4,5,4,5]
46+
; CHECK-NEXT: vpblendd {{.*#+}} ymm2 = ymm2[0,1,2,3,4,5],ymm11[6,7]
47+
; CHECK-NEXT: vpalignr {{.*#+}} ymm5 = ymm6[8,9,10,11,12,13,14,15],ymm5[0,1,2,3,4,5,6,7],ymm6[24,25,26,27,28,29,30,31],ymm5[16,17,18,19,20,21,22,23]
48+
; CHECK-NEXT: vpermq {{.*#+}} ymm5 = ymm5[0,1,0,3]
49+
; CHECK-NEXT: vpslldq {{.*#+}} ymm6 = zero,zero,zero,zero,zero,zero,zero,zero,ymm7[0,1,2,3,4,5,6,7],zero,zero,zero,zero,zero,zero,zero,zero,ymm7[16,17,18,19,20,21,22,23]
50+
; CHECK-NEXT: vpblendd {{.*#+}} ymm5 = ymm6[0,1,2,3],ymm5[4,5,6,7]
51+
; CHECK-NEXT: vmovaps %ymm0, {{[0-9]+}}(%rsp) # 32-byte Spill
52+
; CHECK-NEXT: vmovaps %ymm8, %ymm0
53+
; CHECK-NEXT: vmovaps %ymm1, {{[0-9]+}}(%rsp) # 32-byte Spill
54+
; CHECK-NEXT: vmovaps %ymm2, %ymm1
55+
; CHECK-NEXT: vmovaps {{[0-9]+}}(%rsp), %ymm2 # 32-byte Reload
56+
; CHECK-NEXT: vmovaps %ymm3, {{[0-9]+}}(%rsp) # 32-byte Spill
57+
; CHECK-NEXT: vmovaps %ymm5, %ymm3
58+
; CHECK-NEXT: vmovaps %ymm10, {{[0-9]+}}(%rsp) # 32-byte Spill
59+
; CHECK-NEXT: vmovaps %ymm12, {{[0-9]+}}(%rsp) # 32-byte Spill
5760
; CHECK-NEXT: vmovaps %ymm4, {{[0-9]+}}(%rsp) # 32-byte Spill
58-
; CHECK-NEXT: vmovaps %ymm12, (%rsp) # 32-byte Spill
61+
; CHECK-NEXT: vmovaps %ymm14, (%rsp) # 32-byte Spill
5962
; CHECK-NEXT: movq %rbp, %rsp
6063
; CHECK-NEXT: popq %rbp
6164
; CHECK-NEXT: retq

0 commit comments

Comments
 (0)
Please sign in to comment.