Skip to content

[X86] combineX86ShuffleChain - always combine to a new VPERMV node if the root shuffle was a VPERMV node #128183

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Feb 21, 2025
Merged

[X86] combineX86ShuffleChain - always combine to a new VPERMV node if the root shuffle was a VPERMV node #128183

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions llvm/lib/Target/X86/X86ISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40090,8 +40090,9 @@ static SDValue combineX86ShuffleChain(ArrayRef<SDValue> Inputs, SDValue Root,
bool AllowBWIVPERMV3 =
(Depth >= BWIVPERMV3ShuffleDepth || HasSlowVariableMask);

// If root was a VPERMV3 node, always allow a variable shuffle.
if (Root.getOpcode() == X86ISD::VPERMV3)
// If root was a VPERMV/VPERMV3 node, always allow a variable shuffle.
if ((UnaryShuffle && Root.getOpcode() == X86ISD::VPERMV) ||
Root.getOpcode() == X86ISD::VPERMV3)
AllowVariableCrossLaneMask = AllowVariablePerLaneMask = true;

bool MaskContainsZeros = isAnyZero(Mask);
Expand Down
57 changes: 27 additions & 30 deletions llvm/test/CodeGen/X86/vector-interleaved-store-i16-stride-7.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -588,20 +588,19 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FCP-NEXT: vmovq {{.*#+}} xmm5 = mem[0],zero
; AVX2-FCP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX2-FCP-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
; AVX2-FCP-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm1 = [5,7,1,3,7,0,0,0]
; AVX2-FCP-NEXT: vpermd %ymm4, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[0,1,4,5],zero,zero,zero,zero,zero,zero,ymm1[10,11,14,15,2,3,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
; AVX2-FCP-NEXT: # ymm5 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermd %ymm0, %ymm5, %ymm5
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm5 = zero,zero,zero,zero,ymm5[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm5[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-FCP-NEXT: vpor %ymm5, %ymm1, %ymm1
; AVX2-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [1,5,3,7,1,5,3,7]
; AVX2-FCP-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermd %ymm0, %ymm1, %ymm1
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-FCP-NEXT: vpmovsxbd {{.*#+}} ymm5 = [5,7,1,3,7,0,0,0]
; AVX2-FCP-NEXT: vpermd %ymm4, %ymm5, %ymm5
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm5 = ymm5[0,1,4,5],zero,zero,zero,zero,zero,zero,ymm5[10,11,14,15,2,3,18,19],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-FCP-NEXT: vpor %ymm1, %ymm5, %ymm1
; AVX2-FCP-NEXT: vinserti128 $1, %xmm2, %ymm3, %ymm2
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,4,5,6,7,4,5],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[24,25,20,21,22,23,20,21,28,29]
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm3 = ymm4[0,1,8,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,ymm4[2,3],zero,zero,ymm4[18,19,26,27],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX2-FCP-NEXT: vpor %ymm2, %ymm3, %ymm2
; AVX2-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [0,2,4,6,0,2,4,6]
; AVX2-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [0,4,2,6,0,4,2,6]
; AVX2-FCP-NEXT: # ymm3 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermd %ymm0, %ymm3, %ymm0
; AVX2-FCP-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[u,u,u,u,u,u,u,u,0,1,4,5,8,9,u,u,u,u,u,u,u,u,18,19,22,23,26,27,u,u,u,u]
Expand Down Expand Up @@ -670,17 +669,16 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512-FCP-NEXT: vinserti128 $1, %xmm3, %ymm2, %ymm4
; AVX512-FCP-NEXT: vmovq {{.*#+}} xmm5 = mem[0],zero
; AVX512-FCP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512-FCP-NEXT: vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
; AVX512-FCP-NEXT: vpermi2q %ymm5, %ymm0, %ymm1
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
; AVX512-FCP-NEXT: # ymm0 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
; AVX512-FCP-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
; AVX512-FCP-NEXT: # ymm1 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpermd %ymm0, %ymm1, %ymm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero
; AVX512-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,5,3,7,1,5,3,7]
; AVX512-FCP-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512-FCP-NEXT: vpermd %ymm1, %ymm5, %ymm1
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512-FCP-NEXT: vpermd %ymm0, %ymm5, %ymm0
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX512-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm4[0,1,8,9],zero,zero,zero,zero,ymm4[u,u,u,u,u,u,2,3],zero,zero,ymm4[18,19,26,27,u,u,u,u,u,u],zero,zero,zero,zero
; AVX512-FCP-NEXT: vinserti128 $1, %xmm2, %ymm3, %ymm2
; AVX512-FCP-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,u,u,u,u,u,u],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[u,u,u,u,u,u,20,21,28,29]
Expand Down Expand Up @@ -753,17 +751,16 @@ define void @store_i16_stride7_vf4(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm3, %ymm2, %ymm4
; AVX512DQ-FCP-NEXT: vmovq {{.*#+}} xmm5 = mem[0],zero
; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT: vpmovsxbq {{.*#+}} ymm1 = [0,2,4,0]
; AVX512DQ-FCP-NEXT: vpermi2q %ymm5, %ymm0, %ymm1
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm0 = [0,2,4,6,0,2,4,6]
; AVX512DQ-FCP-NEXT: # ymm0 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpermd %ymm1, %ymm0, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,3,5,7,1,3,5,7]
; AVX512DQ-FCP-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm5[0],ymm0[2],ymm5[2]
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
; AVX512DQ-FCP-NEXT: # ymm1 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpermd %ymm0, %ymm1, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vbroadcasti128 {{.*#+}} ymm5 = [1,5,3,7,1,5,3,7]
; AVX512DQ-FCP-NEXT: # ymm5 = mem[0,1,0,1]
; AVX512DQ-FCP-NEXT: vpermd %ymm1, %ymm5, %ymm1
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,ymm1[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm1[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm1, %zmm0, %zmm0
; AVX512DQ-FCP-NEXT: vpermd %ymm0, %ymm5, %ymm0
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm0 = zero,zero,zero,zero,ymm0[0,1,4,5,8,9],zero,zero,zero,zero,zero,zero,zero,zero,ymm0[18,19,22,23,26,27],zero,zero,zero,zero,zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vinserti64x4 $1, %ymm0, %zmm1, %zmm0
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm1 = ymm4[0,1,8,9],zero,zero,zero,zero,ymm4[u,u,u,u,u,u,2,3],zero,zero,ymm4[18,19,26,27,u,u,u,u,u,u],zero,zero,zero,zero
; AVX512DQ-FCP-NEXT: vinserti128 $1, %xmm2, %ymm3, %ymm2
; AVX512DQ-FCP-NEXT: vpshufb {{.*#+}} ymm2 = zero,zero,zero,zero,ymm2[0,1,8,9,u,u,u,u,u,u],zero,zero,ymm2[26,27],zero,zero,zero,zero,ymm2[u,u,u,u,u,u,20,21,28,29]
Expand Down
61 changes: 30 additions & 31 deletions llvm/test/CodeGen/X86/vector-interleaved-store-i32-stride-7.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,17 +95,17 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-NEXT: vmovsd {{.*#+}} xmm3 = mem[0],zero
; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
; AVX2-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
; AVX2-NEXT: vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
; AVX2-NEXT: vmovaps {{.*#+}} xmm3 = [3,5,7,u]
; AVX2-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
; AVX2-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
; AVX2-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,2,2,1,4,6,6,5]
; AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; AVX2-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6],ymm3[7]
; AVX2-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [5,3,0,1,5,3,0,1]
; AVX2-NEXT: # ymm3 = mem[0,1,0,1]
; AVX2-NEXT: vpermps %ymm0, %ymm3, %ymm0
; AVX2-NEXT: vmovaps {{.*#+}} xmm3 = [3,5,7,u]
; AVX2-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3]
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovlps %xmm0, 48(%rax)
Expand All @@ -130,17 +130,17 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FP-NEXT: vmovsd {{.*#+}} xmm3 = mem[0],zero
; AVX2-FP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-FP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FP-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,2,2,1,4,6,6,5]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,2,1]
; AVX2-FP-NEXT: vshufps {{.*#+}} ymm3 = ymm0[0,2,2,3,4,6,6,7]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,2]
; AVX2-FP-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm3[4,5,6],ymm1[7]
; AVX2-FP-NEXT: vmovaps {{.*#+}} xmm3 = [3,5,7,u]
; AVX2-FP-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
; AVX2-FP-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [0,4,2,6,0,4,2,6]
; AVX2-FP-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FP-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-FP-NEXT: vshufps {{.*#+}} ymm3 = ymm2[0,2,2,1,4,6,6,5]
; AVX2-FP-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; AVX2-FP-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4,5,6],ymm3[7]
; AVX2-FP-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [5,3,0,1,5,3,0,1]
; AVX2-FP-NEXT: # ymm3 = mem[0,1,0,1]
; AVX2-FP-NEXT: vpermps %ymm0, %ymm3, %ymm0
; AVX2-FP-NEXT: vmovaps {{.*#+}} xmm3 = [3,5,7,u]
; AVX2-FP-NEXT: vpermps %ymm2, %ymm3, %ymm2
; AVX2-FP-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1,2],xmm0[3]
; AVX2-FP-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-FP-NEXT: vmovlps %xmm0, 48(%rax)
Expand All @@ -165,23 +165,22 @@ define void @store_i32_stride7_vf2(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vec
; AVX2-FCP-NEXT: vmovsd {{.*#+}} xmm3 = mem[0],zero
; AVX2-FCP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX2-FCP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
; AVX2-FCP-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-FCP-NEXT: vmovaps {{.*#+}} xmm1 = [3,5,7,u]
; AVX2-FCP-NEXT: vpermps %ymm2, %ymm1, %ymm1
; AVX2-FCP-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [3,5,0,1,3,5,0,1]
; AVX2-FCP-NEXT: # ymm3 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermps %ymm0, %ymm3, %ymm3
; AVX2-FCP-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm3[3]
; AVX2-FCP-NEXT: vmovaps {{.*#+}} ymm4 = [0,2,4,6,u,u,u,1]
; AVX2-FCP-NEXT: vpermps %ymm2, %ymm4, %ymm2
; AVX2-FCP-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,2,4,0,0,2,4,0]
; AVX2-FCP-NEXT: vbroadcastf128 {{.*#+}} ymm1 = [5,3,0,1,5,3,0,1]
; AVX2-FCP-NEXT: # ymm1 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermps %ymm0, %ymm1, %ymm1
; AVX2-FCP-NEXT: vmovaps {{.*#+}} xmm3 = [3,5,7,u]
; AVX2-FCP-NEXT: vpermps %ymm2, %ymm3, %ymm3
; AVX2-FCP-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1,2],xmm1[3]
; AVX2-FCP-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,4,2,0,0,4,2,0]
; AVX2-FCP-NEXT: # ymm4 = mem[0,1,0,1]
; AVX2-FCP-NEXT: vpermps %ymm0, %ymm4, %ymm0
; AVX2-FCP-NEXT: vmovaps {{.*#+}} ymm4 = [0,2,4,6,u,u,u,1]
; AVX2-FCP-NEXT: vpermps %ymm2, %ymm4, %ymm2
; AVX2-FCP-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6],ymm2[7]
; AVX2-FCP-NEXT: vextractf128 $1, %ymm3, %xmm2
; AVX2-FCP-NEXT: vmovlps %xmm2, 48(%rax)
; AVX2-FCP-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX2-FCP-NEXT: vmovlps %xmm1, 48(%rax)
; AVX2-FCP-NEXT: vmovaps %ymm0, (%rax)
; AVX2-FCP-NEXT: vmovaps %xmm1, 32(%rax)
; AVX2-FCP-NEXT: vmovaps %xmm3, 32(%rax)
; AVX2-FCP-NEXT: vzeroupper
; AVX2-FCP-NEXT: retq
;
Expand Down
Loading