Similar to the original VPERMILPSri -> VSHUFPSrri mapping added in D143787, replacing VPERMILPDri -> VSHUFPDrri should never be any slower and saves an encoding byte.
The sibling VPERMILPDmi -> VPSHUFDmi mapping is trickier as we need the same shuffle mask in every lane (and it needs to be adjusted) - I haven't attempted that yet but we can investigate it in the future if there's interest.
Fixes #61060
Differential Revision: https://reviews.llvm.org/D148999
return ReplaceInTie;
};
+ // `vpermilpd r, i` -> `vshufpd r, r, i`
+ // `vpermilpd r, i, k` -> `vshufpd r, r, i, k`
+ // `vshufpd` is always as fast or faster than `vpermilpd` and takes
+ // 1 less byte of code size for VEX and EVEX encoding.
+ auto ProcessVPERMILPDri = [&](unsigned NewOpc) -> bool {
+ if (!NewOpcPreferable(NewOpc))
+ return false;
+ unsigned MaskImm = MI.getOperand(NumOperands - 1).getImm();
+ MI.removeOperand(NumOperands - 1);
+ MI.addOperand(MI.getOperand(NumOperands - 2));
+ MI.setDesc(TII->get(NewOpc));
+ MI.addOperand(MachineOperand::CreateImm(MaskImm));
+ return true;
+ };
+
// `vpermilps r, i` -> `vshufps r, r, i`
// `vpermilps r, i, k` -> `vshufps r, r, i, k`
// `vshufps` is always as fast or faster than `vpermilps` and takes
};
switch (Opc) {
+ case X86::VPERMILPDri:
+ return ProcessVPERMILPDri(X86::VSHUFPDrri);
+ case X86::VPERMILPDYri:
+ return ProcessVPERMILPDri(X86::VSHUFPDYrri);
+ case X86::VPERMILPDZ128ri:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ128rri);
+ case X86::VPERMILPDZ256ri:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ256rri);
+ case X86::VPERMILPDZri:
+ return ProcessVPERMILPDri(X86::VSHUFPDZrri);
+ case X86::VPERMILPDZ128rikz:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ128rrikz);
+ case X86::VPERMILPDZ256rikz:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ256rrikz);
+ case X86::VPERMILPDZrikz:
+ return ProcessVPERMILPDri(X86::VSHUFPDZrrikz);
+ case X86::VPERMILPDZ128rik:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ128rrik);
+ case X86::VPERMILPDZ256rik:
+ return ProcessVPERMILPDri(X86::VSHUFPDZ256rrik);
+ case X86::VPERMILPDZrik:
+ return ProcessVPERMILPDri(X86::VSHUFPDZrrik);
+
case X86::VPERMILPSri:
return ProcessVPERMILPSri(X86::VSHUFPSrri);
case X86::VPERMILPSYri:
define <2 x double> @test_mm_permute_pd(<2 x double> %a0) nounwind {
; CHECK-LABEL: test_mm_permute_pd:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; CHECK-NEXT: ret{{[l|q]}}
%res = shufflevector <2 x double> %a0, <2 x double> %a0, <2 x i32> <i32 1, i32 0>
ret <2 x double> %res
define <4 x double> @test_mm256_permute_pd(<4 x double> %a0) nounwind {
; CHECK-LABEL: test_mm256_permute_pd:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; CHECK-NEXT: ret{{[l|q]}}
%res = shufflevector <4 x double> %a0, <4 x double> %a0, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
ret <4 x double> %res
define <2 x double> @test_x86_avx_vpermil_pd(<2 x double> %a0) {
; AVX-LABEL: test_x86_avx_vpermil_pd:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; AVX-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; AVX-NEXT: # xmm0 = xmm0[1,0]
; AVX-NEXT: ret{{[l|q]}} # encoding: [0xc3]
;
; AVX512VL-LABEL: test_x86_avx_vpermil_pd:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; AVX512VL-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; AVX512VL-NEXT: # xmm0 = xmm0[1,0]
; AVX512VL-NEXT: ret{{[l|q]}} # encoding: [0xc3]
%res = call <2 x double> @llvm.x86.avx.vpermil.pd(<2 x double> %a0, i8 1) ; <<2 x double>> [#uses=1]
define <4 x double> @test_x86_avx_vpermil_pd_256(<4 x double> %a0) {
; AVX-LABEL: test_x86_avx_vpermil_pd_256:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd $7, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x05,0xc0,0x07]
+; AVX-NEXT: vshufpd $7, %ymm0, %ymm0, %ymm0 # encoding: [0xc5,0xfd,0xc6,0xc0,0x07]
; AVX-NEXT: # ymm0 = ymm0[1,1,3,2]
; AVX-NEXT: ret{{[l|q]}} # encoding: [0xc3]
;
; AVX512VL-LABEL: test_x86_avx_vpermil_pd_256:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpermilpd $7, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x05,0xc0,0x07]
+; AVX512VL-NEXT: vshufpd $7, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xc6,0xc0,0x07]
; AVX512VL-NEXT: # ymm0 = ymm0[1,1,3,2]
; AVX512VL-NEXT: ret{{[l|q]}} # encoding: [0xc3]
%res = call <4 x double> @llvm.x86.avx.vpermil.pd.256(<4 x double> %a0, i8 7) ; <<4 x double>> [#uses=1]
define <4 x double> @test_x86_avx_vpermilvar_pd_256_2(<4 x double> %a0) {
; AVX-LABEL: test_x86_avx_vpermilvar_pd_256_2:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd $9, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0x7d,0x05,0xc0,0x09]
+; AVX-NEXT: vshufpd $9, %ymm0, %ymm0, %ymm0 # encoding: [0xc5,0xfd,0xc6,0xc0,0x09]
; AVX-NEXT: # ymm0 = ymm0[1,0,2,3]
; AVX-NEXT: ret{{[l|q]}} # encoding: [0xc3]
;
; AVX512VL-LABEL: test_x86_avx_vpermilvar_pd_256_2:
; AVX512VL: # %bb.0:
-; AVX512VL-NEXT: vpermilpd $9, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x05,0xc0,0x09]
+; AVX512VL-NEXT: vshufpd $9, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xc6,0xc0,0x09]
; AVX512VL-NEXT: # ymm0 = ymm0[1,0,2,3]
; AVX512VL-NEXT: ret{{[l|q]}} # encoding: [0xc3]
%res = call <4 x double> @llvm.x86.avx.vpermilvar.pd.256(<4 x double> %a0, <4 x i64> <i64 2, i64 0, i64 0, i64 2>) ; <<4 x double>> [#uses=1]
; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
; X86-NEXT: vmulpd (%eax), %xmm0, %xmm1
; X86-NEXT: vmulsd 16(%eax), %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; X86-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; X86-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
; X64-NEXT: vmulpd (%rsi), %xmm0, %xmm1
; X64-NEXT: vmulsd 16(%rsi), %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; X64-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; X64-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X64-NEXT: retq
; NODQ-NEXT: vextractf128 $1, %ymm0, %xmm1
; NODQ-NEXT: vcvttsd2si %xmm1, %rax
; NODQ-NEXT: vmovq %rax, %xmm2
-; NODQ-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; NODQ-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; NODQ-NEXT: vcvttsd2si %xmm1, %rax
; NODQ-NEXT: vmovq %rax, %xmm1
; NODQ-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; NODQ-NEXT: vcvttsd2si %xmm0, %rax
; NODQ-NEXT: vmovq %rax, %xmm2
-; NODQ-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; NODQ-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; NODQ-NEXT: vcvttsd2si %xmm0, %rax
; NODQ-NEXT: vmovq %rax, %xmm0
; NODQ-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; NODQ-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; NODQ-NEXT: vcvttss2si %xmm1, %rax
; NODQ-NEXT: vmovq %rax, %xmm1
-; NODQ-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; NODQ-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; NODQ-NEXT: vcvttss2si %xmm2, %rax
; NODQ-NEXT: vmovq %rax, %xmm2
; NODQ-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
define float @fhadd_16(<16 x float> %x225) {
; KNL-LABEL: fhadd_16:
; KNL: # %bb.0:
-; KNL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; KNL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; KNL-NEXT: vaddps %xmm1, %xmm0, %xmm0
; KNL-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; KNL-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; SKX-LABEL: fhadd_16:
; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; SKX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; SKX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; SKX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SKX-NEXT: vaddss %xmm1, %xmm0, %xmm0
define float @fhsub_16(<16 x float> %x225) {
; KNL-LABEL: fhsub_16:
; KNL: # %bb.0:
-; KNL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; KNL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; KNL-NEXT: vaddps %xmm1, %xmm0, %xmm0
; KNL-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; KNL-NEXT: vsubss %xmm1, %xmm0, %xmm0
;
; SKX-LABEL: fhsub_16:
; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; SKX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; SKX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; SKX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SKX-NEXT: vsubss %xmm1, %xmm0, %xmm0
define double @fadd_noundef_eel(<8 x double> %x225, <8 x double> %x227) {
; KNL-LABEL: fadd_noundef_eel:
; KNL: # %bb.0:
-; KNL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; KNL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; KNL-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; KNL-NEXT: retq
;
; SKX-LABEL: fadd_noundef_eel:
; SKX: # %bb.0:
-; SKX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; SKX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; SKX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; SKX-NEXT: vzeroupper
; SKX-NEXT: retq
; KNL-LABEL: fsub_noundef_ee:
; KNL: # %bb.0:
; KNL-NEXT: vextractf32x4 $2, %zmm1, %xmm0
-; KNL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; KNL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; KNL-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; KNL-NEXT: retq
;
; SKX-LABEL: fsub_noundef_ee:
; SKX: # %bb.0:
; SKX-NEXT: vextractf32x4 $2, %zmm1, %xmm0
-; SKX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; SKX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; SKX-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; SKX-NEXT: vzeroupper
; SKX-NEXT: retq
define <8 x double> @test_mm512_permute_pd(<8 x double> %a0) {
; CHECK-LABEL: test_mm512_permute_pd:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,2,2,4,4,6,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,1,2,2,4,4,6,6]
; CHECK-NEXT: ret{{[l|q]}}
%res = shufflevector <8 x double> %a0, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
ret <8 x double> %res
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT: kmovw %eax, %k1
-; X86-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,2,4,4,6,6]
+; X86-NEXT: vshufpd {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,2,4,4,6,6]
; X86-NEXT: retl
;
; X64-LABEL: test_mm512_mask_permute_pd:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1
-; X64-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,2,4,4,6,6]
+; X64-NEXT: vshufpd {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,2,4,4,6,6]
; X64-NEXT: retq
%arg1 = bitcast i8 %a1 to <8 x i1>
%res0 = shufflevector <8 x double> %a2, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax
; X86-NEXT: kmovw %eax, %k1
-; X86-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,6]
+; X86-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,6]
; X86-NEXT: retl
;
; X64-LABEL: test_mm512_maskz_permute_pd:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1
-; X64-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,6]
+; X64-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,6]
; X64-NEXT: retq
%arg0 = bitcast i8 %a0 to <8 x i1>
%res0 = shufflevector <8 x double> %a1, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
; X86-NEXT: vpinsrd $1, {{[0-9]+}}(%esp), %xmm1, %xmm1
; X86-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; X86-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm1, %xmm1
-; X86-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; X86-NEXT: vaddsd %xmm1, %xmm2, %xmm1
; X86-NEXT: vblendpd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
; X86-NEXT: retl
; X86-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vaddps %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vaddps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vaddss %xmm1, %xmm0, %xmm0
; X64-NEXT: vaddps %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vaddps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vaddss %xmm1, %xmm0, %xmm0
; X86-NEXT: vmulps %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmulps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vmulss %xmm1, %xmm0, %xmm0
; X64-NEXT: vmulps %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmulps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vmulss %xmm1, %xmm0, %xmm0
; X86-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vmulpd %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vmulpd %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vaddps %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vaddps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vaddss %xmm1, %xmm0, %xmm0
; X64-NEXT: vaddps %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vaddps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vaddss %xmm1, %xmm0, %xmm0
; X86-NEXT: vmulps %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmulps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vmulss %xmm1, %xmm0, %xmm0
; X64-NEXT: vmulps %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmulps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vmulss %xmm1, %xmm0, %xmm0
; X86-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vminpd %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vminsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vminpd %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vminsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vmaxpd %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vmaxpd %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vminpd %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vminsd %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovsd %xmm0, (%esp)
; X86-NEXT: fldl (%esp)
; X64-NEXT: vminpd %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vminsd %xmm1, %xmm0, %xmm0
; X64-NEXT: vzeroupper
; X64-NEXT: retq
; X86-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; X64-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; X86-NEXT: vminps %ymm1, %ymm0, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vminps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vminps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vminss %xmm1, %xmm0, %xmm0
; X64-NEXT: vminps %ymm1, %ymm0, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vminps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vminps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vminss %xmm1, %xmm0, %xmm0
; X86-NEXT: vmaxps %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; X64-NEXT: vmaxps %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; X86-NEXT: vminps %ymm0, %ymm1, %ymm0
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vminps %xmm1, %xmm0, %xmm0
-; X86-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-NEXT: vminps %xmm1, %xmm0, %xmm0
; X86-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X86-NEXT: vminss %xmm1, %xmm0, %xmm0
; X64-NEXT: vminps %ymm0, %ymm1, %ymm0
; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-NEXT: vminps %xmm1, %xmm0, %xmm0
-; X64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-NEXT: vminps %xmm1, %xmm0, %xmm0
; X64-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; X64-NEXT: vminss %xmm1, %xmm0, %xmm0
define <8 x double>@test_int_x86_avx512_vpermil_pd_512(<8 x double> %x0, <8 x double> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpermil_pd_512:
; CHECK: ## %bb.0:
-; CHECK-NEXT: vpermilpd $22, %zmm0, %zmm0 ## encoding: [0x62,0xf3,0xfd,0x48,0x05,0xc0,0x16]
+; CHECK-NEXT: vshufpd $22, %zmm0, %zmm0, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0xc6,0xc0,0x16]
; CHECK-NEXT: ## zmm0 = zmm0[0,1,3,2,5,4,6,6]
; CHECK-NEXT: ret{{[l|q]}} ## encoding: [0xc3]
%res = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> %x2, i8 -1)
; X86: ## %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $22, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x05,0xc8,0x16]
+; X86-NEXT: vshufpd $22, %zmm0, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xc6,0xc8,0x16]
; X86-NEXT: ## zmm1 {%k1} = zmm0[0,1,3,2,5,4,6,6]
; X86-NEXT: vmovapd %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc1]
; X86-NEXT: retl ## encoding: [0xc3]
; X64-LABEL: test_int_x86_avx512_mask_vpermil_pd_512:
; X64: ## %bb.0:
; X64-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $22, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf3,0xfd,0x49,0x05,0xc8,0x16]
+; X64-NEXT: vshufpd $22, %zmm0, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0xfd,0x49,0xc6,0xc8,0x16]
; X64-NEXT: ## zmm1 {%k1} = zmm0[0,1,3,2,5,4,6,6]
; X64-NEXT: vmovapd %zmm1, %zmm0 ## encoding: [0x62,0xf1,0xfd,0x48,0x28,0xc1]
; X64-NEXT: retq ## encoding: [0xc3]
; X86: ## %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax ## encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 ## encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $22, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xc9,0x05,0xc0,0x16]
+; X86-NEXT: vshufpd $22, %zmm0, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0xc6,0xc0,0x16]
; X86-NEXT: ## zmm0 {%k1} {z} = zmm0[0,1,3,2,5,4,6,6]
; X86-NEXT: retl ## encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermil_pd_512:
; X64: ## %bb.0:
; X64-NEXT: kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $22, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf3,0xfd,0xc9,0x05,0xc0,0x16]
+; X64-NEXT: vshufpd $22, %zmm0, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0xfd,0xc9,0xc6,0xc0,0x16]
; X64-NEXT: ## zmm0 {%k1} {z} = zmm0[0,1,3,2,5,4,6,6]
; X64-NEXT: retq ## encoding: [0xc3]
%res = call <8 x double> @llvm.x86.avx512.mask.vpermil.pd.512(<8 x double> %x0, i32 22, <8 x double> zeroinitializer, i8 %x3)
define <2 x double> @test_2xdouble_perm_mask0(<2 x double> %vec) {
; CHECK-LABEL: test_2xdouble_perm_mask0:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; CHECK-NEXT: retq
%res = shufflevector <2 x double> %vec, <2 x double> undef, <2 x i32> <i32 1, i32 0>
ret <2 x double> %res
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %xmm3, %xmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
; CHECK-NEXT: vmovapd %xmm1, %xmm0
; CHECK-NEXT: retq
%shuf = shufflevector <2 x double> %vec, <2 x double> undef, <2 x i32> <i32 1, i32 0>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %xmm2, %xmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
; CHECK-NEXT: retq
%shuf = shufflevector <2 x double> %vec, <2 x double> undef, <2 x i32> <i32 1, i32 0>
%cmp = fcmp oeq <2 x double> %mask, zeroinitializer
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %xmm3, %xmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
; CHECK-NEXT: vmovapd %xmm1, %xmm0
; CHECK-NEXT: retq
%shuf = shufflevector <2 x double> %vec, <2 x double> undef, <2 x i32> <i32 1, i32 0>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %xmm2, %xmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
; CHECK-NEXT: retq
%shuf = shufflevector <2 x double> %vec, <2 x double> undef, <2 x i32> <i32 1, i32 0>
%cmp = fcmp oeq <2 x double> %mask, zeroinitializer
define <4 x double> @test_4xdouble_perm_mask0(<4 x double> %vec) {
; CHECK-LABEL: test_4xdouble_perm_mask0:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
; CHECK-NEXT: retq
%res = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
ret <4 x double> %res
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %ymm3, %ymm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 {%k1} = ymm0[1,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 {%k1} = ymm0[1,0,2,3]
; CHECK-NEXT: vmovapd %ymm1, %ymm0
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,0,2,3]
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
%cmp = fcmp oeq <4 x double> %mask, zeroinitializer
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %ymm3, %ymm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 {%k1} = ymm0[1,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 {%k1} = ymm0[1,1,2,2]
; CHECK-NEXT: vmovapd %ymm1, %ymm0
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 2>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,1,2,2]
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 2>
%cmp = fcmp oeq <4 x double> %mask, zeroinitializer
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %ymm3, %ymm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 {%k1} = ymm0[0,1,3,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 {%k1} = ymm0[0,1,3,3]
; CHECK-NEXT: vmovapd %ymm1, %ymm0
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 0, i32 1, i32 3, i32 3>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 {%k1} {z} = ymm0[0,1,3,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm0[0,1,3,3]
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 0, i32 1, i32 3, i32 3>
%cmp = fcmp oeq <4 x double> %mask, zeroinitializer
define <4 x double> @test_4xdouble_perm_mask3(<4 x double> %vec) {
; CHECK-LABEL: test_4xdouble_perm_mask3:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; CHECK-NEXT: retq
%res = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 2>
ret <4 x double> %res
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %ymm3, %ymm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 {%k1} = ymm0[1,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 {%k1} = ymm0[1,1,2,2]
; CHECK-NEXT: vmovapd %ymm1, %ymm0
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 2>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %ymm2, %ymm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm0[1,1,2,2]
; CHECK-NEXT: retq
%shuf = shufflevector <4 x double> %vec, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 2>
%cmp = fcmp oeq <4 x double> %mask, zeroinitializer
define <8 x double> @test_8xdouble_perm_mask0(<8 x double> %vec) {
; CHECK-LABEL: test_8xdouble_perm_mask0:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,0,3,2,4,5,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,0,3,2,4,5,7,6]
; CHECK-NEXT: retq
%res = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 0, i32 3, i32 2, i32 4, i32 5, i32 7, i32 6>
ret <8 x double> %res
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %zmm3, %zmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,0,3,2,4,5,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm1 {%k1} = zmm0[0,0,3,2,4,5,7,6]
; CHECK-NEXT: vmovapd %zmm1, %zmm0
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 0, i32 3, i32 2, i32 4, i32 5, i32 7, i32 6>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %zmm2, %zmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,0,3,2,4,5,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,0,3,2,4,5,7,6]
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 0, i32 3, i32 2, i32 4, i32 5, i32 7, i32 6>
%cmp = fcmp oeq <8 x double> %mask, zeroinitializer
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %zmm3, %zmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,2,3,4,4,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,2,3,4,4,7,6]
; CHECK-NEXT: vmovapd %zmm1, %zmm0
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 7, i32 6>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %zmm2, %zmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,3,4,4,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,3,4,4,7,6]
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 4, i32 7, i32 6>
%cmp = fcmp oeq <8 x double> %mask, zeroinitializer
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %zmm3, %zmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,0,2,3,5,5,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm1 {%k1} = zmm0[0,0,2,3,5,5,6,7]
; CHECK-NEXT: vmovapd %zmm1, %zmm0
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 5, i32 5, i32 6, i32 7>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %zmm2, %zmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,0,2,3,5,5,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,0,2,3,5,5,6,7]
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 5, i32 5, i32 6, i32 7>
%cmp = fcmp oeq <8 x double> %mask, zeroinitializer
define <8 x double> @test_8xdouble_perm_mask3(<8 x double> %vec) {
; CHECK-LABEL: test_8xdouble_perm_mask3:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,2,2,4,4,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,1,2,2,4,4,6,7]
; CHECK-NEXT: retq
%res = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 7>
ret <8 x double> %res
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm3, %xmm3, %xmm3
; CHECK-NEXT: vcmpeqpd %zmm3, %zmm2, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,2,2,4,4,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,2,2,4,4,6,7]
; CHECK-NEXT: vmovapd %zmm1, %zmm0
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 7>
; CHECK: # %bb.0:
; CHECK-NEXT: vxorpd %xmm2, %xmm2, %xmm2
; CHECK-NEXT: vcmpeqpd %zmm2, %zmm1, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,2,4,4,6,7]
; CHECK-NEXT: retq
%shuf = shufflevector <8 x double> %vec, <8 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 2, i32 4, i32 4, i32 6, i32 7>
%cmp = fcmp oeq <8 x double> %mask, zeroinitializer
define half @extract_f16_4(<8 x half> %x) {
; CHECK-LABEL: extract_f16_4:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; CHECK-NEXT: ret{{[l|q]}}
%res = extractelement <8 x half> %x, i32 4
ret half %res
define void @extract_store_f16_4(<8 x half> %x, ptr %y) {
; X64-LABEL: extract_store_f16_4:
; X64: # %bb.0:
-; X64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; X64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; X64-NEXT: vmovsh %xmm0, (%rdi)
; X64-NEXT: retq
;
; X86-LABEL: extract_store_f16_4:
; X86: # %bb.0:
; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; X86-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; X86-NEXT: vmovsh %xmm0, (%eax)
; X86-NEXT: retl
%res = extractelement <8 x half> %x, i32 4
; CHECK-NEXT: vpsrlq $48, %xmm1, %xmm3
; CHECK-NEXT: vpextrq $1, %xmm4, %rax
; CHECK-NEXT: vmovsh %xmm3, (%rax)
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; CHECK-NEXT: vextracti32x4 $2, %zmm0, %xmm4
; CHECK-NEXT: vmovq %xmm4, %rax
; CHECK-NEXT: vmovsh %xmm3, (%rax)
; CHECK-NEXT: vpsrlq $48, %xmm0, %xmm1
; CHECK-NEXT: vpextrq $1, %xmm3, %rax
; CHECK-NEXT: vmovsh %xmm1, (%rax)
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; CHECK-NEXT: vextracti32x4 $2, %zmm2, %xmm3
; CHECK-NEXT: vmovq %xmm3, %rax
; CHECK-NEXT: vmovsh %xmm1, (%rax)
define <4 x double>@test_int_x86_avx512_vpermil_pd_256(<4 x double> %x0, <4 x double> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpermil_pd_256:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd $6, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x05,0xc0,0x06]
+; CHECK-NEXT: vshufpd $6, %ymm0, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xc6,0xc0,0x06]
; CHECK-NEXT: # ymm0 = ymm0[0,1,3,2]
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
%res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> %x2, i8 -1)
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $6, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x06]
+; X86-NEXT: vshufpd $6, %ymm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xc6,0xc8,0x06]
; X86-NEXT: # ymm1 {%k1} = ymm0[0,1,3,2]
; X86-NEXT: vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X86-NEXT: retl # encoding: [0xc3]
; X64-LABEL: test_int_x86_avx512_mask_vpermil_pd_256:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $6, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x05,0xc8,0x06]
+; X64-NEXT: vshufpd $6, %ymm0, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x29,0xc6,0xc8,0x06]
; X64-NEXT: # ymm1 {%k1} = ymm0[0,1,3,2]
; X64-NEXT: vmovapd %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x28,0xc1]
; X64-NEXT: retq # encoding: [0xc3]
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $6, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xc0,0x06]
+; X86-NEXT: vshufpd $6, %ymm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xc6,0xc0,0x06]
; X86-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,3,2]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermil_pd_256:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $6, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0xa9,0x05,0xc0,0x06]
+; X64-NEXT: vshufpd $6, %ymm0, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0xa9,0xc6,0xc0,0x06]
; X64-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,3,2]
; X64-NEXT: retq # encoding: [0xc3]
%res = call <4 x double> @llvm.x86.avx512.mask.vpermil.pd.256(<4 x double> %x0, i32 22, <4 x double> zeroinitializer, i8 %x3)
define <2 x double>@test_int_x86_avx512_vpermil_pd_128(<2 x double> %x0, <2 x double> %x2) {
; CHECK-LABEL: test_int_x86_avx512_vpermil_pd_128:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; CHECK-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; CHECK-NEXT: # xmm0 = xmm0[1,0]
; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
%res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> %x2, i8 -1)
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $1, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
+; X86-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xc6,0xc8,0x01]
; X86-NEXT: # xmm1 {%k1} = xmm0[1,0]
; X86-NEXT: vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X86-NEXT: retl # encoding: [0xc3]
; X64-LABEL: test_int_x86_avx512_mask_vpermil_pd_128:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $1, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x05,0xc8,0x01]
+; X64-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0xfd,0x09,0xc6,0xc8,0x01]
; X64-NEXT: # xmm1 {%k1} = xmm0[1,0]
; X64-NEXT: vmovapd %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0xc1]
; X64-NEXT: retq # encoding: [0xc3]
; X86: # %bb.0:
; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
; X86-NEXT: kmovw %eax, %k1 # encoding: [0xc5,0xf8,0x92,0xc8]
-; X86-NEXT: vpermilpd $1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x05,0xc0,0x01]
+; X86-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xc6,0xc0,0x01]
; X86-NEXT: # xmm0 {%k1} {z} = xmm0[1,0]
; X86-NEXT: retl # encoding: [0xc3]
;
; X64-LABEL: test_int_x86_avx512_maskz_vpermil_pd_128:
; X64: # %bb.0:
; X64-NEXT: kmovw %edi, %k1 # encoding: [0xc5,0xf8,0x92,0xcf]
-; X64-NEXT: vpermilpd $1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0xfd,0x89,0x05,0xc0,0x01]
+; X64-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0xfd,0x89,0xc6,0xc0,0x01]
; X64-NEXT: # xmm0 {%k1} {z} = xmm0[1,0]
; X64-NEXT: retq # encoding: [0xc3]
%res = call <2 x double> @llvm.x86.avx512.mask.vpermil.pd.128(<2 x double> %x0, i32 1, <2 x double> zeroinitializer, i8 %x3)
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[0,1,0,1]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,3]
; AVX1-NEXT: vmovq %rdi, %xmm2
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,1,0,1]
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm2, %ymm2
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,0,1]
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,3]
; AVX1-NEXT: vandnpd %ymm0, %ymm1, %ymm0
; AVX1-NEXT: retq
;
;
; AVX1-LABEL: complex_square_f64:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-NEXT: vaddsd %xmm0, %xmm0, %xmm2
; AVX1-NEXT: vmulsd %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vmulsd %xmm0, %xmm0, %xmm0
;
; FMA-LABEL: complex_square_f64:
; FMA: # %bb.0:
-; FMA-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; FMA-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; FMA-NEXT: vaddsd %xmm0, %xmm0, %xmm2
; FMA-NEXT: vmulsd %xmm2, %xmm1, %xmm2
; FMA-NEXT: vmulsd %xmm1, %xmm1, %xmm1
;
; AVX1-LABEL: complex_mul_f64:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX1-NEXT: vmulsd %xmm0, %xmm3, %xmm4
; AVX1-NEXT: vmulsd %xmm2, %xmm1, %xmm5
; AVX1-NEXT: vaddsd %xmm5, %xmm4, %xmm4
;
; FMA-LABEL: complex_mul_f64:
; FMA: # %bb.0:
-; FMA-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
-; FMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; FMA-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
+; FMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; FMA-NEXT: vmulsd %xmm2, %xmm1, %xmm4
; FMA-NEXT: vfmadd231sd {{.*#+}} xmm4 = (xmm3 * xmm0) + xmm4
; FMA-NEXT: vmulsd %xmm2, %xmm3, %xmm2
define <4 x double> @vec256_eltty_double_source_subvec_0_target_subvec_mask_1_unary(<4 x double> %x) nounwind {
; CHECK-LABEL: vec256_eltty_double_source_subvec_0_target_subvec_mask_1_unary:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
; CHECK-NEXT: retq
%r = shufflevector <4 x double> %x, <4 x double> poison, <4 x i32> <i32 0, i32 0, i32 2, i32 3>
ret <4 x double> %r
define <4 x double> @vec256_eltty_double_source_subvec_1_target_subvec_mask_2_unary(<4 x double> %x) nounwind {
; CHECK-LABEL: vec256_eltty_double_source_subvec_1_target_subvec_mask_2_unary:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,1,2,2]
; CHECK-NEXT: retq
%r = shufflevector <4 x double> %x, <4 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 2>
ret <4 x double> %r
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm1, %ymm1
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
; AVX1-NEXT: vmovddup {{.*#+}} ymm2 = ymm0[0,0,2,2]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm3 = ymm0[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm3 = ymm0[1,1,3,3]
; AVX1-NEXT: vmovaps %ymm4, %ymm0
; AVX1-NEXT: retq
;
; NOFMA: # %bb.0: # %bb
; NOFMA-NEXT: vmulps %xmm1, %xmm0, %xmm0
; NOFMA-NEXT: vaddss %xmm2, %xmm0, %xmm1
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm4 = xmm2[1,0]
; NOFMA-NEXT: vaddss %xmm4, %xmm3, %xmm3
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm4 = xmm0[1,1,3,3]
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm5 = xmm2[1,1,3,3]
; NOFMA: # %bb.0: # %bb
; NOFMA-NEXT: vmulpd %xmm1, %xmm0, %xmm0
; NOFMA-NEXT: vaddsd %xmm2, %xmm0, %xmm1
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; NOFMA-NEXT: vsubsd %xmm2, %xmm0, %xmm0
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; NOFMA-NEXT: retq
; NOFMA: # %bb.0: # %bb
; NOFMA-NEXT: vmulps %ymm1, %ymm0, %ymm0
; NOFMA-NEXT: vaddss %xmm2, %xmm0, %xmm1
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm4 = xmm2[1,0]
; NOFMA-NEXT: vaddss %xmm4, %xmm3, %xmm3
; NOFMA-NEXT: vextractf128 $1, %ymm0, %xmm4
; NOFMA-NEXT: vextractf128 $1, %ymm2, %xmm5
; NOFMA-NEXT: vaddss %xmm5, %xmm4, %xmm6
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm7 = xmm4[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm8 = xmm5[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm7 = xmm4[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm8 = xmm5[1,0]
; NOFMA-NEXT: vaddss %xmm7, %xmm8, %xmm7
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm8 = xmm0[1,1,3,3]
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm9 = xmm2[1,1,3,3]
; NOFMA-NEXT: vextractf128 $1, %ymm0, %xmm3
; NOFMA-NEXT: vextractf128 $1, %ymm2, %xmm4
; NOFMA-NEXT: vaddsd %xmm4, %xmm3, %xmm5
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; NOFMA-NEXT: vsubsd %xmm2, %xmm0, %xmm0
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm2 = xmm4[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm2 = xmm4[1,0]
; NOFMA-NEXT: vsubsd %xmm2, %xmm1, %xmm1
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm5[0],xmm1[0]
; NOFMA-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; NOFMA-NEXT: vmulps %ymm3, %ymm1, %ymm1
; NOFMA-NEXT: vmulps %ymm2, %ymm0, %ymm0
; NOFMA-NEXT: vaddss %xmm4, %xmm0, %xmm2
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm6 = xmm4[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm6 = xmm4[1,0]
; NOFMA-NEXT: vaddss %xmm6, %xmm3, %xmm3
; NOFMA-NEXT: vextractf128 $1, %ymm0, %xmm6
; NOFMA-NEXT: vextractf128 $1, %ymm4, %xmm7
; NOFMA-NEXT: vaddss %xmm7, %xmm6, %xmm8
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm9 = xmm6[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm10 = xmm7[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm9 = xmm6[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm10 = xmm7[1,0]
; NOFMA-NEXT: vaddss %xmm10, %xmm9, %xmm9
; NOFMA-NEXT: vinsertps {{.*#+}} xmm8 = xmm8[0,1],xmm9[0],xmm8[3]
; NOFMA-NEXT: vaddss %xmm5, %xmm1, %xmm9
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm10 = xmm1[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm11 = xmm5[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm10 = xmm1[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm11 = xmm5[1,0]
; NOFMA-NEXT: vaddss %xmm11, %xmm10, %xmm10
; NOFMA-NEXT: vextractf128 $1, %ymm1, %xmm11
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm12 = xmm11[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm12 = xmm11[1,0]
; NOFMA-NEXT: vextractf128 $1, %ymm5, %xmm13
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm14 = xmm13[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm14 = xmm13[1,0]
; NOFMA-NEXT: vaddss %xmm14, %xmm12, %xmm12
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm14 = xmm0[1,1,3,3]
; NOFMA-NEXT: vmovshdup {{.*#+}} xmm15 = xmm4[1,1,3,3]
; NOFMA-NEXT: vextractf128 $1, %ymm1, %xmm1
; NOFMA-NEXT: vextractf128 $1, %ymm5, %xmm5
; NOFMA-NEXT: vaddsd %xmm5, %xmm1, %xmm9
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm4 = xmm4[1,0]
; NOFMA-NEXT: vsubsd %xmm4, %xmm0, %xmm0
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm2[0],xmm0[0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm6[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm6[1,0]
; NOFMA-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm7[0],xmm2[0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
-; NOFMA-NEXT: vpermilpd {{.*#+}} xmm3 = xmm5[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
+; NOFMA-NEXT: vshufpd {{.*#+}} xmm3 = xmm5[1,0]
; NOFMA-NEXT: vsubsd %xmm3, %xmm1, %xmm1
; NOFMA-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm9[0],xmm1[0]
; NOFMA-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; CHECK-NEXT: vfmadd213pd {{.*#+}} ymm0 = (ymm1 * ymm0) + ymm2
; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm1
; CHECK-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; CHECK-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; CHECK-NEXT: vzeroupper
; CHECK-NEXT: retq
;
; AVX-LABEL: not_a_hsub_2:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX-NEXT: vsubss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vsubss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[2,3]
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vsubss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
; AVX-NEXT: vsubss %xmm3, %xmm1, %xmm1
;
; AVX-LABEL: not_a_hsub_3:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vsubsd %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vsubsd %xmm0, %xmm2, %xmm0
; AVX-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
; AVX-NEXT: retq
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-SLOW-NEXT: retq
;
; AVX2: # %bb.0:
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddps %xmm0, %xmm1, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX2-NEXT: retq
%1 = shufflevector <4 x float> %a0, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
; AVX2-NEXT: vmovq %rdi, %xmm1
; AVX2-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; AVX2-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vbroadcastsd %xmm1, %ymm1
; AVX2-NEXT: vdivpd %ymm1, %ymm0, %ymm0
; CHECK-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; CHECK-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; CHECK-NEXT: vbroadcastsd %xmm0, %ymm0
; CHECK-NEXT: retl
;
; AVX1_SLOW-LABEL: hadd_v2f64:
; AVX1_SLOW: # %bb.0:
-; AVX1_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1_SLOW-NEXT: retq
;
; AVX2_SLOW-LABEL: hadd_v2f64:
; AVX2_SLOW: # %bb.0:
-; AVX2_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX2_SLOW-NEXT: retq
;
; AVX1_SLOW-LABEL: hadd_v2f64_scalar_splat:
; AVX1_SLOW: # %bb.0:
-; AVX1_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1_SLOW-NEXT: retq
;
; AVX2_SLOW-LABEL: hadd_v2f64_scalar_splat:
; AVX2_SLOW: # %bb.0:
-; AVX2_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX2_SLOW-NEXT: retq
;
; AVX1_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
; AVX1_SLOW: # %bb.0:
-; AVX1_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1_SLOW-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
;
; AVX2_SLOW-LABEL: hadd_v4f64_scalar_broadcast:
; AVX2_SLOW: # %bb.0:
-; AVX2_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2_SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2_SLOW-NEXT: vbroadcastsd %xmm0, %ymm0
; AVX2_SLOW-NEXT: retq
;
; AVX1_SLOW-LABEL: hsub_v2f64:
; AVX1_SLOW: # %bb.0:
-; AVX1_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1_SLOW-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; AVX1_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1_SLOW-NEXT: retq
;
; AVX2_SLOW-LABEL: hsub_v2f64:
; AVX2_SLOW: # %bb.0:
-; AVX2_SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2_SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2_SLOW-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; AVX2_SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX2_SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: test5_undef:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3]
; AVX512-SLOW: # %bb.0:
; AVX512-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX512-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX512-SLOW-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX512-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[2,3]
; AVX512-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm2
; AVX512-SLOW-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1],xmm2[0],xmm1[3]
-; AVX512-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-SLOW-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX512-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0]
;
; AVX-SLOW-LABEL: add_pd_003_2:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddpd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: add_pd_010:
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = xmm0[0,0]
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddpd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX1-SLOW-LABEL: add_pd_011:
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: retq
;
;
; AVX-SLOW-LABEL: PR45747_2:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm0 = xmm1[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,1,1]
; AVX-SLOW-NEXT: vaddps %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-SLOW-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1],xmm1[0]
; AVX-SLOW-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vhaddpd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1],xmm1[0]
; AVX-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX-SLOW-NEXT: vhaddpd %xmm2, %xmm0, %xmm0
; AVX-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vmovddup {{.*#+}} xmm1 = xmm1[0,0]
; AVX-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX-SLOW-NEXT: vhaddpd %xmm2, %xmm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: haddpd3:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
;
; AVX-SLOW-LABEL: hsubpd2:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vsubpd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
;
; AVX-SLOW-LABEL: extract_extract23_v4f32_fadd_f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract23_v4f32_fadd_f32_commute:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v2f64_fadd_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
;
; AVX-SLOW-LABEL: extract_extract01_v2f64_fadd_f64_commute:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
;
;
; AVX-SLOW-LABEL: extract_extract23_v4f32_fsub_f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vsubss %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: retq
;
; AVX-LABEL: extract_extract23_v4f32_fsub_f32_commute:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vsubss %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v2f64_fsub_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: retq
;
;
; AVX-LABEL: extract_extract01_v2f64_fsub_f64_commute:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vsubsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: retq
%x0 = extractelement <2 x double> %x, i32 0
;
; AVX-SLOW-LABEL: extract_extract23_v8f32_fadd_f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-LABEL: extract_extract67_v8f32_fadd_f32:
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: extract_extract23_v8f32_fadd_f32_commute:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-LABEL: extract_extract67_v8f32_fadd_f32_commute:
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: extract_extract01_v4f64_fadd_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
; AVX-SLOW-LABEL: extract_extract23_v4f64_fadd_f64:
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v4f64_fadd_f64_commute:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
; AVX-SLOW-LABEL: extract_extract23_v4f64_fadd_f64_commute:
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract23_v8f32_fsub_f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vsubss %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: extract_extract01_v4f64_fsub_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-LABEL: extract_extract01_v4f64_fsub_f64_commute:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vsubsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v8f64_fadd_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v8f64_fadd_f64_commute:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-SLOW-LABEL: extract_extract01_v8f64_fsub_f64:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vsubsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
; AVX-SLOW-NEXT: retq
;
; AVX-LABEL: extract_extract01_v8f64_fsub_f64_commute:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vsubsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-SLOW-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: hadd32_4:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX-FAST-LABEL: hadd32_4:
; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-FAST-NEXT: retq
;
; AVX-SLOW-LABEL: hadd32_8:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX-FAST-LABEL: hadd32_8:
; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-FAST-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: hadd32_16:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX-FAST-LABEL: hadd32_16:
; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-FAST-NEXT: vzeroupper
;
; AVX-LABEL: hadd32_4_optsize:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX-LABEL: hadd32_8_optsize:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
;
; AVX-LABEL: hadd32_16_optsize:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
;
; AVX-LABEL: hadd32_4_pgso:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX-LABEL: hadd32_8_pgso:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
;
; AVX-LABEL: hadd32_16_pgso:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: partial_reduction_fadd_v8f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX-SLOW-LABEL: partial_reduction_fadd_v8f32_wrong_flags:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX-FAST-LABEL: partial_reduction_fadd_v8f32_wrong_flags:
; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX-FAST-NEXT: vzeroupper
;
; AVX-SLOW-LABEL: partial_reduction_fadd_v16f32:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; BWON-F16C-NEXT: .LBB26_6:
; BWON-F16C-NEXT: vcvtps2ph $4, %xmm3, %xmm2
; BWON-F16C-NEXT: vmovd %xmm2, %edx
-; BWON-F16C-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; BWON-F16C-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; BWON-F16C-NEXT: vpextrw $0, %xmm2, %esi
; BWON-F16C-NEXT: movzwl %si, %esi
; BWON-F16C-NEXT: vmovd %esi, %xmm2
; BWON-F16C-NEXT: vcvtph2ps %xmm2, %xmm2
-; BWON-F16C-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; BWON-F16C-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; BWON-F16C-NEXT: vpextrw $0, %xmm3, %esi
; BWON-F16C-NEXT: movzwl %si, %esi
; BWON-F16C-NEXT: vmovd %esi, %xmm3
;
; AVX1-SLOW-LABEL: PR37890_v4f32:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX2-LABEL: PR37890_v4f32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX1-SLOW-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX1-SLOW-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-SLOW-NEXT: vhaddps %xmm7, %xmm6, %xmm2
; AVX1-SLOW-NEXT: vhaddps %xmm2, %xmm2, %xmm2
; AVX1-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX1-FAST-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX1-FAST-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-FAST-NEXT: vhaddps %xmm7, %xmm6, %xmm2
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm2, %xmm2
; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX2-SLOW-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX2-SLOW-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX2-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-SLOW-NEXT: vhaddps %xmm7, %xmm6, %xmm2
; AVX2-SLOW-NEXT: vhaddps %xmm2, %xmm2, %xmm2
; AVX2-FAST-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1,2],xmm4[3]
; AVX2-FAST-NEXT: vaddps %xmm1, %xmm3, %xmm1
; AVX2-FAST-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
-; AVX2-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-FAST-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX2-FAST-NEXT: vhaddps %xmm7, %xmm6, %xmm2
; AVX2-FAST-NEXT: vhaddps %xmm0, %xmm2, %xmm2
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vhaddps %xmm1, %xmm0, %xmm4
; AVX-SLOW-NEXT: vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[3],xmm1[1],zero,zero
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
; AVX-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX-FAST: # %bb.0:
; AVX-FAST-NEXT: vhaddps %xmm1, %xmm0, %xmm4
; AVX-FAST-NEXT: vunpckhps {{.*#+}} xmm5 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-FAST-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[3],xmm1[1],zero,zero
; AVX-FAST-NEXT: vhaddps %xmm2, %xmm2, %xmm1
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm4[0,2],xmm1[0,1]
; AVX-SLOW: # %bb.0:
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm4 = xmm0[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm4, %xmm0, %xmm4
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm5 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm5 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddss %xmm5, %xmm4, %xmm4
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm0, %xmm4, %xmm0
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm4, %xmm1, %xmm4
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddss %xmm5, %xmm4, %xmm4
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm4, %xmm1
; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm2, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm2[1,0]
; AVX-SLOW-NEXT: vaddss %xmm4, %xmm1, %xmm1
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; AVX-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm3[1,1,3,3]
; AVX-SLOW-NEXT: vaddss %xmm1, %xmm3, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
; AVX-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-FAST-LABEL: reduction_sum_v4f32_v4f32:
; AVX-FAST: # %bb.0:
; AVX-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm4
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm5 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm5 = xmm0[1,0]
; AVX-FAST-NEXT: vaddss %xmm5, %xmm4, %xmm4
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-FAST-NEXT: vaddss %xmm0, %xmm4, %xmm0
; AVX-FAST-NEXT: vhaddps %xmm1, %xmm1, %xmm4
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; AVX-FAST-NEXT: vaddss %xmm5, %xmm4, %xmm4
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-FAST-NEXT: vaddss %xmm1, %xmm4, %xmm1
; AVX-FAST-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[2,3]
; AVX-FAST-NEXT: vhaddps %xmm2, %xmm2, %xmm1
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm4 = xmm2[1,0]
; AVX-FAST-NEXT: vaddss %xmm4, %xmm1, %xmm1
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-FAST-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0],xmm0[3]
; AVX-FAST-NEXT: vhaddps %xmm3, %xmm3, %xmm1
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm3[3,3,3,3]
; AVX-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
;
; AVX-SLOW-LABEL: reduction_sum_v4f32_v4f32_reassoc:
; AVX-SLOW: # %bb.0:
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-SLOW-NEXT: vaddps %xmm4, %xmm0, %xmm0
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX-SLOW-NEXT: vaddps %xmm4, %xmm1, %xmm1
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm2[1,0]
; AVX-SLOW-NEXT: vaddps %xmm4, %xmm2, %xmm2
-; AVX-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX-SLOW-NEXT: vaddps %xmm4, %xmm3, %xmm3
; AVX-SLOW-NEXT: vshufps {{.*#+}} xmm4 = xmm3[1,1],xmm2[1,1]
; AVX-SLOW-NEXT: vinsertps {{.*#+}} xmm5 = xmm0[1],xmm1[1],zero,zero
;
; AVX-FAST-LABEL: reduction_sum_v4f32_v4f32_reassoc:
; AVX-FAST: # %bb.0:
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-FAST-NEXT: vaddps %xmm4, %xmm0, %xmm0
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX-FAST-NEXT: vaddps %xmm4, %xmm1, %xmm1
; AVX-FAST-NEXT: vhaddps %xmm1, %xmm0, %xmm0
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-FAST-NEXT: vaddps %xmm1, %xmm2, %xmm1
-; AVX-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX-FAST-NEXT: vaddps %xmm2, %xmm3, %xmm2
; AVX-FAST-NEXT: vhaddps %xmm2, %xmm1, %xmm1
; AVX-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X86-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero
; X86-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; X86-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; X86-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; X86-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; X86-NEXT: vextractf128 $1, %ymm0, %xmm1
; X86-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X86-NEXT: vcvtdq2pd %xmm0, %ymm0
; X64-AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm0[0],zero,xmm0[1],zero
; X64-AVX1-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,2,3,3]
; X64-AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; X64-AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; X64-AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; X64-AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; X64-AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
; X64-AVX1-NEXT: vcvtdq2pd %xmm0, %ymm0
; AVX-NEXT: vmovups (%rdi), %xmm0
; AVX-NEXT: vmulps (%rsi), %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: retq
; AVX-NEXT: vmovups (%rdi), %xmm0
; AVX-NEXT: vmulps (%rsi), %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: retq
; AVX-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0,1],mem[0],xmm1[3]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX1-LABEL: test_mul2x2_f32:
; AVX1: # %bb.0: # %entry
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-NEXT: vmovsldup {{.*#+}} xmm3 = xmm1[0,0,2,2]
; AVX1-NEXT: vmulps %xmm3, %xmm0, %xmm3
; AVX1-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
;
; AVX2-LABEL: test_mul2x2_f32:
; AVX2: # %bb.0: # %entry
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vbroadcastss %xmm1, %xmm3
; AVX2-NEXT: vmulps %xmm3, %xmm0, %xmm3
; AVX2-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
;
; AVX512-LABEL: test_mul2x2_f32:
; AVX512: # %bb.0: # %entry
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vbroadcastss %xmm1, %xmm3
; AVX512-NEXT: vmulps %xmm3, %xmm0, %xmm3
; AVX512-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX-NEXT: vmovddup {{.*#+}} xmm3 = xmm1[0,0]
; AVX-NEXT: vmulpd %xmm3, %xmm0, %xmm3
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,1]
; AVX-NEXT: vmulpd %xmm4, %xmm2, %xmm4
; AVX-NEXT: vaddpd %xmm4, %xmm3, %xmm3
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmovddup {{.*#+}} xmm4 = xmm1[0,0]
; AVX-NEXT: vmulpd %xmm4, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,1]
; AVX-NEXT: vmulpd %xmm1, %xmm2, %xmm1
; AVX-NEXT: vaddpd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
; AVX512F-NEXT: vmovshdup {{.*#+}} xmm6 = xmm1[1,1,3,3]
; AVX512F-NEXT: vmulps %xmm6, %xmm2, %xmm4
; AVX512F-NEXT: vaddps %xmm4, %xmm3, %xmm4
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm3 = xmm5[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm3 = xmm5[1,0]
; AVX512F-NEXT: vshufps {{.*#+}} xmm7 = xmm1[3,3,3,3]
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm8 = xmm1[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm8 = xmm1[1,0]
; AVX512F-NEXT: vshufps {{.*#+}} xmm9 = xmm1[2,2,2,2]
; AVX512F-NEXT: vmulps %xmm3, %xmm9, %xmm9
; AVX512F-NEXT: vaddps %xmm4, %xmm9, %xmm9
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX512F-NEXT: vmulss %xmm1, %xmm4, %xmm10
; AVX512F-NEXT: vmovshdup {{.*#+}} xmm5 = xmm5[1,1,3,3]
; AVX512F-NEXT: vmulss %xmm6, %xmm5, %xmm6
; AVX512F-NEXT: vaddss %xmm7, %xmm11, %xmm7
; AVX512F-NEXT: vinsertps {{.*#+}} xmm7 = xmm8[0,1],xmm7[0],xmm8[3]
; AVX512F-NEXT: vshufps {{.*#+}} xmm8 = xmm9[3,3,3,3]
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm11 = xmm9[1,0]
; AVX512F-NEXT: vshufps {{.*#+}} xmm9 = xmm9[2,2,2,2]
; AVX512F-NEXT: vmulps %xmm0, %xmm9, %xmm0
; AVX512F-NEXT: vmulps %xmm2, %xmm8, %xmm2
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm5 = xmm1[1,1,3,3]
; AVX512VL-NEXT: vmulps %xmm5, %xmm2, %xmm6
; AVX512VL-NEXT: vaddps %xmm6, %xmm3, %xmm3
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm6 = xmm4[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm6 = xmm4[1,0]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm7 = xmm1[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm8 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm8 = xmm1[1,0]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm9 = xmm1[2,2,2,2]
; AVX512VL-NEXT: vmulps %xmm6, %xmm9, %xmm9
; AVX512VL-NEXT: vaddps %xmm3, %xmm9, %xmm3
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm9 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm9 = xmm0[1,0]
; AVX512VL-NEXT: vmulss %xmm1, %xmm9, %xmm10
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm4 = xmm4[1,1,3,3]
; AVX512VL-NEXT: vmulss %xmm5, %xmm4, %xmm5
; AVX512VL-NEXT: vaddss %xmm7, %xmm11, %xmm7
; AVX512VL-NEXT: vinsertps {{.*#+}} xmm5 = xmm5[0,1],xmm7[0],xmm5[3]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm7 = xmm8[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm11 = xmm8[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm11 = xmm8[1,0]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm8 = xmm8[2,2,2,2]
; AVX512VL-NEXT: vmulps %xmm0, %xmm8, %xmm0
; AVX512VL-NEXT: vmulps %xmm7, %xmm2, %xmm2
; AVX1-NEXT: vmovddup {{.*#+}} xmm8 = xmm4[0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm8, %ymm8
; AVX1-NEXT: vmulpd %ymm0, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm9 = xmm4[1,1]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm9 = xmm4[1,1]
; AVX1-NEXT: vinsertf128 $1, %xmm9, %ymm9, %ymm9
; AVX1-NEXT: vmulpd %ymm1, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
; AVX1-NEXT: vmovddup {{.*#+}} ymm9 = ymm4[0,0,2,2]
; AVX1-NEXT: vmulpd %ymm2, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[1,1,3,3]
; AVX1-NEXT: vmulpd %ymm4, %ymm3, %ymm4
; AVX1-NEXT: vaddpd %ymm4, %ymm8, %ymm4
; AVX1-NEXT: vmovddup {{.*#+}} xmm8 = xmm5[0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm8, %ymm8
; AVX1-NEXT: vmulpd %ymm0, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm9 = xmm5[1,1]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm9 = xmm5[1,1]
; AVX1-NEXT: vinsertf128 $1, %xmm9, %ymm9, %ymm9
; AVX1-NEXT: vmulpd %ymm1, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
; AVX1-NEXT: vmovddup {{.*#+}} ymm9 = ymm5[0,0,2,2]
; AVX1-NEXT: vmulpd %ymm2, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = ymm5[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[1,1,3,3]
; AVX1-NEXT: vmulpd %ymm5, %ymm3, %ymm5
; AVX1-NEXT: vaddpd %ymm5, %ymm8, %ymm5
; AVX1-NEXT: vmovddup {{.*#+}} xmm8 = xmm6[0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm8, %ymm8
; AVX1-NEXT: vmulpd %ymm0, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm9 = xmm6[1,1]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm9 = xmm6[1,1]
; AVX1-NEXT: vinsertf128 $1, %xmm9, %ymm9, %ymm9
; AVX1-NEXT: vmulpd %ymm1, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
; AVX1-NEXT: vmovddup {{.*#+}} ymm9 = ymm6[0,0,2,2]
; AVX1-NEXT: vmulpd %ymm2, %ymm9, %ymm9
; AVX1-NEXT: vaddpd %ymm9, %ymm8, %ymm8
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1,1,3,3]
; AVX1-NEXT: vmulpd %ymm6, %ymm3, %ymm6
; AVX1-NEXT: vaddpd %ymm6, %ymm8, %ymm6
; AVX1-NEXT: vmovddup {{.*#+}} xmm8 = xmm7[0,0]
; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm8, %ymm8
; AVX1-NEXT: vmulpd %ymm0, %ymm8, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm8 = xmm7[1,1]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm8 = xmm7[1,1]
; AVX1-NEXT: vinsertf128 $1, %xmm8, %ymm8, %ymm8
; AVX1-NEXT: vmulpd %ymm1, %ymm8, %ymm1
; AVX1-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-NEXT: vmovddup {{.*#+}} ymm7 = ymm1[0,0,2,2]
; AVX1-NEXT: vmulpd %ymm7, %ymm2, %ymm2
; AVX1-NEXT: vaddpd %ymm2, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,1,3,3]
; AVX1-NEXT: vmulpd %ymm1, %ymm3, %ymm1
; AVX1-NEXT: vaddpd %ymm1, %ymm0, %ymm3
; AVX1-NEXT: vmovapd %ymm4, %ymm0
; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,1],xmm4[0,2]
; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0,0,3,3]
; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3],ymm4[4],ymm2[5,6],ymm4[7]
; AVX1-NEXT: vunpckhpd {{.*#+}} xmm4 = xmm3[1],xmm1[1]
; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm1[1,1],xmm4[0,2]
; AVX1-NEXT: vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7]
; AVX1-NEXT: vpermilpd {{.*#+}} ymm4 = mem[1,0,2,2]
; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2],ymm0[3],ymm3[4,5],ymm0[6],ymm3[7]
; AVX1-NEXT: vmovups %ymm0, 32(%rdi)
; AVX1-NEXT: vmovups %ymm1, (%rdi)
; XOP-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,1],xmm5[0,2]
; XOP-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3
; XOP-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm1[2,3,2,3]
-; XOP-NEXT: vpermilpd {{.*#+}} ymm5 = ymm5[0,0,3,3]
+; XOP-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[0,0,3,3]
; XOP-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2,3],ymm5[4],ymm3[5,6],ymm5[7]
; XOP-NEXT: vunpckhpd {{.*#+}} xmm5 = xmm4[1],xmm2[1]
; XOP-NEXT: vshufps {{.*#+}} xmm5 = xmm2[1,1],xmm5[0,2]
; XOP-NEXT: vinsertf128 $1, %xmm5, %ymm2, %ymm2
; XOP-NEXT: vbroadcastsd (%rcx), %ymm4
; XOP-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm4[2],ymm2[3,4],ymm4[5],ymm2[6,7]
-; XOP-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
+; XOP-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7]
; XOP-NEXT: vmovups %ymm0, 32(%rdi)
; XOP-NEXT: vmovups %ymm2, (%rdi)
; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-NEXT: vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7]
; AVX1-NEXT: vshufps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1]
; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
; AVX1-NEXT: vmovdqa %xmm0, 16(%rsi)
; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; XOP-NEXT: vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7]
; XOP-NEXT: vshufps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7]
-; XOP-NEXT: vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
+; XOP-NEXT: vshufpd {{.*#+}} ymm5 = ymm1[0,0,3,2]
; XOP-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1]
; XOP-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3]
; XOP-NEXT: vmovdqa %xmm0, 16(%rsi)
; CHECK-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,1,1,0]
; CHECK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; CHECK-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; CHECK-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5],ymm0[6,7]
; CHECK-NEXT: retq
%shuffle = shufflevector <8 x i32> %a, <8 x i32> %b, <8 x i32> <i32 0, i32 13, i32 12, i32 13, i32 3, i32 15, i32 1, i32 4>
; AVX512F_32-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero
; AVX512F_32-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; AVX512F_32-NEXT: vsubpd {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
-; AVX512F_32-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512F_32-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512F_32-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX512F_32-NEXT: vmovsd %xmm0, (%esp)
; AVX512F_32-NEXT: fldl (%esp)
; AVX-LABEL: fadd_splat_splat_nonzero_v2f64:
; AVX: # %bb.0:
; AVX-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; AVX-NEXT: retq
%splatx = shufflevector <2 x double> %vx, <2 x double> undef, <2 x i32> <i32 1, i32 1>
%splaty = shufflevector <2 x double> %vy, <2 x double> undef, <2 x i32> <i32 1, i32 1>
; AVX-LABEL: fadd_splat_splat_mismatch_v2f64:
; AVX: # %bb.0:
; AVX-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,1]
; AVX-NEXT: vaddpd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%splatx = shufflevector <2 x double> %vx, <2 x double> undef, <2 x i32> <i32 0, i32 0>
define <2 x double> @foo2(<2 x double> %v, ptr%p) nounwind {
; AVX2-LABEL: foo2:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; AVX2-NEXT: vmovapd %xmm0, (%rdi)
; AVX2-NEXT: retq
%res = shufflevector <2 x double> %v, <2 x double> undef, <2 x i32> <i32 1, i32 1>
; X86-AVX-LABEL: PR26515:
; X86-AVX: # %bb.0:
; X86-AVX-NEXT: pushl %eax
-; X86-AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X86-AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X86-AVX-NEXT: vaddss %xmm0, %xmm1, %xmm0
; X86-AVX-NEXT: vmovss %xmm0, (%esp)
; X86-AVX-NEXT: flds (%esp)
;
; X64-AVX-LABEL: PR26515:
; X64-AVX: # %bb.0:
-; X64-AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; X64-AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; X64-AVX-NEXT: vaddss %xmm0, %xmm1, %xmm0
; X64-AVX-NEXT: retq
%2 = shufflevector <4 x float> %0, <4 x float> poison, <4 x i32> <i32 2, i32 undef, i32 undef, i32 undef>
; X86-AVX1-LABEL: test_mm_storeh_sd:
; X86-AVX1: # %bb.0:
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X86-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X86-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X86-AVX1-NEXT: vmovsd %xmm0, (%eax) # encoding: [0xc5,0xfb,0x11,0x00]
; X86-AVX1-NEXT: retl # encoding: [0xc3]
; X86-AVX512-LABEL: test_mm_storeh_sd:
; X86-AVX512: # %bb.0:
; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X86-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X86-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X86-AVX512-NEXT: vmovsd %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x11,0x00]
; X86-AVX512-NEXT: retl # encoding: [0xc3]
;
; X64-AVX1-LABEL: test_mm_storeh_sd:
; X64-AVX1: # %bb.0:
-; X64-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X64-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X64-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X64-AVX1-NEXT: vmovsd %xmm0, (%rdi) # encoding: [0xc5,0xfb,0x11,0x07]
; X64-AVX1-NEXT: retq # encoding: [0xc3]
;
; X64-AVX512-LABEL: test_mm_storeh_sd:
; X64-AVX512: # %bb.0:
-; X64-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X64-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X64-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X64-AVX512-NEXT: vmovsd %xmm0, (%rdi) # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x11,0x07]
; X64-AVX512-NEXT: retq # encoding: [0xc3]
;
; X32-AVX1-LABEL: test_mm_storeh_sd:
; X32-AVX1: # %bb.0:
-; X32-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X32-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X32-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X32-AVX1-NEXT: vmovsd %xmm0, (%edi) # encoding: [0x67,0xc5,0xfb,0x11,0x07]
; X32-AVX1-NEXT: retq # encoding: [0xc3]
;
; X32-AVX512-LABEL: test_mm_storeh_sd:
; X32-AVX512: # %bb.0:
-; X32-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X32-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X32-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X32-AVX512-NEXT: vmovsd %xmm0, (%edi) # EVEX TO VEX Compression encoding: [0x67,0xc5,0xfb,0x11,0x07]
; X32-AVX512-NEXT: retq # encoding: [0xc3]
; X86-AVX1-LABEL: test_mm_storer_pd:
; X86-AVX1: # %bb.0:
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X86-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X86-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X86-AVX1-NEXT: vmovapd %xmm0, (%eax) # encoding: [0xc5,0xf9,0x29,0x00]
; X86-AVX1-NEXT: retl # encoding: [0xc3]
; X86-AVX512-LABEL: test_mm_storer_pd:
; X86-AVX512: # %bb.0:
; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
-; X86-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X86-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X86-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X86-AVX512-NEXT: vmovapd %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x29,0x00]
; X86-AVX512-NEXT: retl # encoding: [0xc3]
;
; X64-AVX1-LABEL: test_mm_storer_pd:
; X64-AVX1: # %bb.0:
-; X64-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X64-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X64-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X64-AVX1-NEXT: vmovapd %xmm0, (%rdi) # encoding: [0xc5,0xf9,0x29,0x07]
; X64-AVX1-NEXT: retq # encoding: [0xc3]
;
; X64-AVX512-LABEL: test_mm_storer_pd:
; X64-AVX512: # %bb.0:
-; X64-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X64-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X64-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X64-AVX512-NEXT: vmovapd %xmm0, (%rdi) # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x29,0x07]
; X64-AVX512-NEXT: retq # encoding: [0xc3]
;
; X32-AVX1-LABEL: test_mm_storer_pd:
; X32-AVX1: # %bb.0:
-; X32-AVX1-NEXT: vpermilpd $1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X32-AVX1-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X32-AVX1-NEXT: # xmm0 = xmm0[1,0]
; X32-AVX1-NEXT: vmovapd %xmm0, (%edi) # encoding: [0x67,0xc5,0xf9,0x29,0x07]
; X32-AVX1-NEXT: retq # encoding: [0xc3]
;
; X32-AVX512-LABEL: test_mm_storer_pd:
; X32-AVX512: # %bb.0:
-; X32-AVX512-NEXT: vpermilpd $1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x05,0xc0,0x01]
+; X32-AVX512-NEXT: vshufpd $1, %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc6,0xc0,0x01]
; X32-AVX512-NEXT: # xmm0 = xmm0[1,0]
; X32-AVX512-NEXT: vmovapd %xmm0, (%edi) # EVEX TO VEX Compression encoding: [0x67,0xc5,0xf9,0x29,0x07]
; X32-AVX512-NEXT: retq # encoding: [0xc3]
;
; AVX1-LABEL: test11:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-NEXT: vsubss %xmm1, %xmm0, %xmm0
; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
; AVX1-NEXT: retq
;
; AVX512-LABEL: test11:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vsubss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vbroadcastss %xmm0, %xmm0
; AVX512-NEXT: retq
; AVX-LABEL: test14:
; AVX: # %bb.0:
; AVX-NEXT: vsubss %xmm1, %xmm0, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vsubss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm2[0,1],xmm0[0],xmm2[3]
; AVX-NEXT: retq
; AVX: # %bb.0:
; AVX-NEXT: vmovss {{.*#+}} xmm2 = mem[0],zero,zero,zero
; AVX-NEXT: vsubss %xmm2, %xmm0, %xmm3
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; AVX-NEXT: vsubss %xmm5, %xmm4, %xmm4
; AVX-NEXT: vmovshdup {{.*#+}} xmm5 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm5, %xmm2
define <8 x double> @transform_VPERMILPSZrr(<8 x double> %a) nounwind {
; CHECK-LABEL: transform_VPERMILPSZrr:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,2,3,4,5,7,6]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,1,2,3,4,5,7,6]
; CHECK-NEXT: retq
%shufp = shufflevector <8 x double> %a, <8 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 7, i32 6>
ret <8 x double> %shufp
define <4 x double> @transform_VPERMILPSYrr(<4 x double> %a) nounwind {
; CHECK-LABEL: transform_VPERMILPSYrr:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
; CHECK-NEXT: retq
%shufp = shufflevector <4 x double> %a, <4 x double> poison, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
ret <4 x double> %shufp
define <2 x double> @transform_VPERMILPSrr(<2 x double> %a) nounwind {
; CHECK-LABEL: transform_VPERMILPSrr:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; CHECK-NEXT: retq
%shufp = shufflevector <2 x double> %a, <2 x double> poison, <2 x i32> <i32 1, i32 0>
ret <2 x double> %shufp
; CHECK-LABEL: transform_VPERMILPSZrrkz:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,3,5,4,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm0 {%k1} {z} = zmm0[0,1,2,3,5,4,6,7]
; CHECK-NEXT: retq
%mask = bitcast i8 %mask_int to <8 x i1>
%shufp = shufflevector <8 x double> %a, <8 x double> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
; CHECK-LABEL: transform_VPERMILPSYrrkz:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 {%k1} {z} = ymm0[0,1,3,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 {%k1} {z} = ymm0[0,1,3,2]
; CHECK-NEXT: retq
%mask = bitcast i4 %mask_int to <4 x i1>
%shufp = shufflevector <4 x double> %a, <4 x double> poison, <4 x i32> <i32 0, i32 1, i32 3, i32 2>
; CHECK-LABEL: transform_VPERMILPSrrkz:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 {%k1} {z} = xmm0[1,0]
; CHECK-NEXT: retq
%mask = bitcast i2 %mask_int to <2 x i1>
%shufp = shufflevector <2 x double> %a, <2 x double> poison, <2 x i32> <i32 1, i32 0>
; CHECK-LABEL: transform_VPERMILPSZrrk:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,3,2,4,5,6,7]
+; CHECK-NEXT: vshufpd {{.*#+}} zmm1 {%k1} = zmm0[0,1,3,2,4,5,6,7]
; CHECK-NEXT: vmovapd %zmm1, %zmm0
; CHECK-NEXT: retq
%mask = bitcast i8 %mask_int to <8 x i1>
; CHECK-LABEL: transform_VPERMILPSYrrk:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm1 {%k1} = ymm0[1,0,3,2]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm1 {%k1} = ymm0[1,0,3,2]
; CHECK-NEXT: vmovapd %ymm1, %ymm0
; CHECK-NEXT: retq
%mask = bitcast i4 %mask_int to <4 x i1>
; CHECK-LABEL: transform_VPERMILPSrrk:
; CHECK: # %bb.0:
; CHECK-NEXT: kmovd %edi, %k1
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm1 {%k1} = xmm0[1,0]
; CHECK-NEXT: vmovapd %xmm1, %xmm0
; CHECK-NEXT: retq
%mask = bitcast i2 %mask_int to <2 x i1>
define <4 x double> @transform_VPERMILPDYrr(<4 x double> %a) nounwind {
; CHECK-LABEL: transform_VPERMILPDYrr:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
+; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
; CHECK-NEXT: retq
%shufp = shufflevector <4 x double> %a, <4 x double> poison, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
ret <4 x double> %shufp
define <2 x double> @transform_VPERMILPDrr(<2 x double> %a) nounwind {
; CHECK-LABEL: transform_VPERMILPDrr:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; CHECK-NEXT: retq
%shufp = shufflevector <2 x double> %a, <2 x double> poison, <2 x i32> <i32 1, i32 0>
ret <2 x double> %shufp
; AVX-64: # %bb.0:
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm1
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm0
; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512F-64: # %bb.0:
; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm0
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512VL-64: # %bb.0:
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX-32-NEXT: .cfi_def_cfa_register %ebp
; AVX-32-NEXT: andl $-8, %esp
; AVX-32-NEXT: subl $16, %esp
-; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX-32-NEXT: vcomisd %xmm1, %xmm2
; AVX-32-NEXT: vmovapd %xmm1, %xmm3
; AVX-64-NEXT: shlq $63, %rcx
; AVX-64-NEXT: xorq %rax, %rcx
; AVX-64-NEXT: vmovq %rcx, %xmm3
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcomisd %xmm1, %xmm0
; AVX-64-NEXT: jb .LBB1_4
; AVX-64-NEXT: # %bb.3:
; AVX512F-32-NEXT: .cfi_def_cfa_register %ebp
; AVX512F-32-NEXT: andl $-8, %esp
; AVX512F-32-NEXT: subl $16, %esp
-; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512F-32-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512F-32-NEXT: vmovsd {{.*#+}} xmm2 = mem[0],zero
; AVX512F-32-NEXT: xorl %eax, %eax
; AVX512F-32-NEXT: vcomisd %xmm2, %xmm1
; AVX512F-64: # %bb.0:
; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm0
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512VL-32-NEXT: .cfi_def_cfa_register %ebp
; AVX512VL-32-NEXT: andl $-8, %esp
; AVX512VL-32-NEXT: subl $16, %esp
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512VL-32-NEXT: vmovsd {{.*#+}} xmm2 = mem[0],zero
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomisd %xmm2, %xmm1
; AVX512VL-64: # %bb.0:
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
;
; AVX-64-LABEL: strict_vector_fptoui_v2f64_to_v2i32:
; AVX-64: # %bb.0:
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX-64-NEXT: vcvttsd2si %xmm0, %rcx
; AVX-64-NEXT: vmovd %ecx, %xmm0
; AVX-64: # %bb.0:
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm1
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm0
; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX-32-NEXT: .cfi_def_cfa_register %ebp
; AVX-32-NEXT: andl $-8, %esp
; AVX-32-NEXT: subl $16, %esp
-; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX-32-NEXT: vcomisd %xmm1, %xmm2
; AVX-32-NEXT: vmovapd %xmm1, %xmm3
; AVX-64-NEXT: shlq $63, %rcx
; AVX-64-NEXT: xorq %rax, %rcx
; AVX-64-NEXT: vmovq %rcx, %xmm3
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcomisd %xmm1, %xmm0
; AVX-64-NEXT: jb .LBB19_4
; AVX-64-NEXT: # %bb.3:
; AVX-64-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX-64-NEXT: vmovq %rax, %xmm2
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX-64-NEXT: vmovq %rax, %xmm1
; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm2
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX-64-NEXT: vmovq %rax, %xmm0
; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512F-64-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512F-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512F-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm0
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX-32-NEXT: .cfi_def_cfa_register %ebp
; AVX-32-NEXT: andl $-8, %esp
; AVX-32-NEXT: subl $32, %esp
-; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX-32-NEXT: vcomisd %xmm1, %xmm2
; AVX-32-NEXT: vmovapd %xmm1, %xmm3
; AVX-32-NEXT: shll $31, %eax
; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %eax
; AVX-32-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX-32-NEXT: vcomisd %xmm1, %xmm3
; AVX-32-NEXT: vmovapd %xmm1, %xmm4
; AVX-32-NEXT: jae .LBB1_4
; AVX-64-NEXT: movzbl %al, %eax
; AVX-64-NEXT: shlq $63, %rax
; AVX-64-NEXT: xorq %rcx, %rax
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX-64-NEXT: vcomisd %xmm1, %xmm4
; AVX-64-NEXT: vxorpd %xmm5, %xmm5, %xmm5
; AVX-64-NEXT: jb .LBB1_4
; AVX-64-NEXT: shlq $63, %rcx
; AVX-64-NEXT: xorq %rax, %rcx
; AVX-64-NEXT: vmovq %rcx, %xmm4
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-64-NEXT: vcomisd %xmm1, %xmm0
; AVX-64-NEXT: jb .LBB1_8
; AVX-64-NEXT: # %bb.7:
; AVX512F-32-NEXT: subl $40, %esp
; AVX512F-32-NEXT: .cfi_offset %ebx, -12
; AVX512F-32-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512F-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512F-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX512F-32-NEXT: xorl %eax, %eax
; AVX512F-32-NEXT: vcomisd %xmm1, %xmm3
; AVX512F-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512F-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; AVX512F-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp)
-; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-32-NEXT: xorl %ecx, %ecx
; AVX512F-32-NEXT: vcomisd %xmm1, %xmm2
; AVX512F-32-NEXT: setae %cl
; AVX512F-64-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512F-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512F-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm0
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512VL-32-NEXT: subl $40, %esp
; AVX512VL-32-NEXT: .cfi_offset %ebx, -12
; AVX512VL-32-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512VL-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm3
; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp)
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-32-NEXT: xorl %ecx, %ecx
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm2
; AVX512VL-32-NEXT: setae %cl
; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX-64-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX-64-NEXT: vcvttss2si %xmm1, %rax
; AVX-64-NEXT: vmovq %rax, %xmm1
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-64-NEXT: vcvttss2si %xmm2, %rax
; AVX-64-NEXT: vmovq %rax, %xmm2
; AVX-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-64-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512F-64-NEXT: vcvttss2si %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttss2si %xmm2, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2si %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttss2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX-32-NEXT: movzbl %cl, %ecx
; AVX-32-NEXT: shll $31, %ecx
; AVX-32-NEXT: xorl {{[0-9]+}}(%esp), %ecx
-; AVX-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-32-NEXT: vcomiss %xmm1, %xmm2
; AVX-32-NEXT: vmovaps %xmm1, %xmm3
; AVX-32-NEXT: jae .LBB3_6
; AVX-64-NEXT: movzbl %al, %eax
; AVX-64-NEXT: shlq $63, %rax
; AVX-64-NEXT: xorq %rcx, %rax
-; AVX-64-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-64-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-64-NEXT: vcomiss %xmm1, %xmm4
; AVX-64-NEXT: vxorps %xmm5, %xmm5, %xmm5
; AVX-64-NEXT: jb .LBB3_4
; AVX512F-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512F-32-NEXT: vsubss %xmm3, %xmm2, %xmm2
; AVX512F-32-NEXT: vmovss %xmm2, (%esp)
-; AVX512F-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-32-NEXT: xorl %edx, %edx
; AVX512F-32-NEXT: vcomiss %xmm1, %xmm2
; AVX512F-32-NEXT: setae %dl
; AVX512F-64-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512F-64-NEXT: vcvttss2usi %xmm1, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm1
-; AVX512F-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-64-NEXT: vcvttss2usi %xmm2, %rax
; AVX512F-64-NEXT: vmovq %rax, %xmm2
; AVX512F-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512VL-32-NEXT: vsubss %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovss %xmm2, (%esp)
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-32-NEXT: xorl %edx, %edx
; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm2
; AVX512VL-32-NEXT: setae %dl
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2usi %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttss2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vextractf32x4 $3, %zmm0, %xmm1
; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512VL-64-NEXT: vcvttsd2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512VL-64-NEXT: vcvttsd2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm3[0],xmm0[0]
; AVX512VL-32-NEXT: .cfi_offset %edi, -16
; AVX512VL-32-NEXT: .cfi_offset %ebx, -12
; AVX512VL-32-NEXT: vextractf32x4 $3, %zmm0, %xmm2
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512VL-32-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm3
; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp)
; AVX512VL-32-NEXT: vextractf32x4 $2, %zmm0, %xmm2
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm3
; AVX512VL-32-NEXT: setae %al
; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp)
; AVX512VL-32-NEXT: vextractf128 $1, %ymm0, %xmm2
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm3
; AVX512VL-32-NEXT: setae %al
; AVX512VL-32-NEXT: vmovsd %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512VL-32-NEXT: vsubsd %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovsd %xmm2, {{[0-9]+}}(%esp)
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-32-NEXT: xorl %ebx, %ebx
; AVX512VL-32-NEXT: vcomisd %xmm1, %xmm2
; AVX512VL-32-NEXT: setae %bl
; AVX512VL-64-NEXT: vextractf32x4 $3, %zmm0, %xmm1
; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm1
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-64-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512VL-64-NEXT: vcvttsd2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512VL-64-NEXT: vcvttsd2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm0
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm3[0],xmm0[0]
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttss2si %xmm3, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2si %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttss2si %xmm3, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm4 {%k1} {z}
; AVX512VL-32-NEXT: vsubss %xmm4, %xmm3, %xmm3
; AVX512VL-32-NEXT: vmovss %xmm3, (%esp)
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512VL-32-NEXT: xorl %eax, %eax
; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm3
; AVX512VL-32-NEXT: setae %al
; AVX512VL-32-NEXT: vmovss %xmm1, %xmm1, %xmm3 {%k1} {z}
; AVX512VL-32-NEXT: vsubss %xmm3, %xmm2, %xmm2
; AVX512VL-32-NEXT: vmovss %xmm2, {{[0-9]+}}(%esp)
-; AVX512VL-32-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-32-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-32-NEXT: xorl %ecx, %ecx
; AVX512VL-32-NEXT: vcomiss %xmm1, %xmm2
; AVX512VL-32-NEXT: setae %cl
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512VL-64-NEXT: vcvttss2usi %xmm3, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; AVX512VL-64-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512VL-64-NEXT: vcvttss2usi %xmm2, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm2
-; AVX512VL-64-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX512VL-64-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX512VL-64-NEXT: vcvttss2usi %xmm3, %rax
; AVX512VL-64-NEXT: vmovq %rax, %xmm3
; AVX512VL-64-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
; VEX: # %bb.0:
; VEX-NEXT: vcvttsd2si %xmm0, %rax
; VEX-NEXT: vmovq %rax, %xmm1
-; VEX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; VEX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; VEX-NEXT: vcvttsd2si %xmm0, %rax
; VEX-NEXT: vmovq %rax, %xmm0
; VEX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512F: # %bb.0:
; AVX512F-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm0
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm0
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-NEXT: vcvttsd2si %xmm1, %rax
; AVX1-NEXT: vmovq %rax, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-NEXT: vcvttsd2si %xmm1, %rax
; AVX1-NEXT: vmovq %rax, %xmm1
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-NEXT: vcvttsd2si %xmm0, %rax
; AVX1-NEXT: vmovq %rax, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vcvttsd2si %xmm0, %rax
; AVX1-NEXT: vmovq %rax, %xmm0
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vcvttsd2si %xmm1, %rax
; AVX2-NEXT: vmovq %rax, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-NEXT: vcvttsd2si %xmm1, %rax
; AVX2-NEXT: vmovq %rax, %xmm1
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-NEXT: vcvttsd2si %xmm0, %rax
; AVX2-NEXT: vmovq %rax, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vcvttsd2si %xmm0, %rax
; AVX2-NEXT: vmovq %rax, %xmm0
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512F-NEXT: vcvttsd2si %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512F-NEXT: vcvttsd2si %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-NEXT: vcvttsd2si %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm0
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-NEXT: vcvttsd2si %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-NEXT: vcvttsd2si %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm0
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; VEX-NEXT: andq %rax, %rdx
; VEX-NEXT: orq %rcx, %rdx
; VEX-NEXT: vmovq %rdx, %xmm2
-; VEX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; VEX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; VEX-NEXT: vsubsd %xmm1, %xmm0, %xmm1
; VEX-NEXT: vcvttsd2si %xmm1, %rax
; VEX-NEXT: vcvttsd2si %xmm0, %rcx
; AVX512F: # %bb.0:
; AVX512F-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm0
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm0
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX1-NEXT: andq %rax, %rdx
; AVX1-NEXT: orq %rcx, %rdx
; AVX1-NEXT: vmovq %rdx, %xmm3
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX1-NEXT: vsubsd %xmm1, %xmm2, %xmm4
; AVX1-NEXT: vcvttsd2si %xmm4, %rax
; AVX1-NEXT: vcvttsd2si %xmm2, %rcx
; AVX1-NEXT: andq %rax, %rdx
; AVX1-NEXT: orq %rcx, %rdx
; AVX1-NEXT: vmovq %rdx, %xmm3
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vsubsd %xmm1, %xmm0, %xmm1
; AVX1-NEXT: vcvttsd2si %xmm1, %rax
; AVX1-NEXT: vcvttsd2si %xmm0, %rcx
; AVX2-NEXT: andq %rax, %rdx
; AVX2-NEXT: orq %rcx, %rdx
; AVX2-NEXT: vmovq %rdx, %xmm3
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX2-NEXT: vsubsd %xmm1, %xmm2, %xmm4
; AVX2-NEXT: vcvttsd2si %xmm4, %rax
; AVX2-NEXT: vcvttsd2si %xmm2, %rcx
; AVX2-NEXT: andq %rax, %rdx
; AVX2-NEXT: orq %rcx, %rdx
; AVX2-NEXT: vmovq %rdx, %xmm3
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vsubsd %xmm1, %xmm0, %xmm1
; AVX2-NEXT: vcvttsd2si %xmm1, %rax
; AVX2-NEXT: vcvttsd2si %xmm0, %rcx
; AVX512F-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512F-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512F-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512F-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512F-NEXT: vmovq %rax, %xmm0
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512VL-NEXT: vcvttsd2usi %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512VL-NEXT: vcvttsd2usi %xmm0, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm0
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm2[0],xmm0[0]
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX1-NEXT: vcvttss2si %xmm1, %rax
; AVX1-NEXT: vmovq %rax, %xmm1
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-NEXT: vcvttss2si %xmm2, %rax
; AVX1-NEXT: vmovq %rax, %xmm2
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX2-NEXT: vcvttss2si %xmm1, %rax
; AVX2-NEXT: vmovq %rax, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vcvttss2si %xmm2, %rax
; AVX2-NEXT: vmovq %rax, %xmm2
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512F-NEXT: vcvttss2si %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-NEXT: vcvttss2si %xmm2, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512VL-NEXT: vcvttss2si %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-NEXT: vcvttss2si %xmm2, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX1-NEXT: vcvttss2si %xmm1, %rax
; AVX1-NEXT: vmovq %rax, %xmm1
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-NEXT: vcvttss2si %xmm2, %rax
; AVX1-NEXT: vmovq %rax, %xmm2
; AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX2-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX2-NEXT: vcvttss2si %xmm1, %rax
; AVX2-NEXT: vmovq %rax, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vcvttss2si %xmm2, %rax
; AVX2-NEXT: vmovq %rax, %xmm2
; AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512F-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512F-NEXT: vcvttss2si %xmm1, %rax
; AVX512F-NEXT: vcvttss2si %xmm0, %rcx
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512F-NEXT: vcvttss2si %xmm1, %rdx
; AVX512F-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512F-NEXT: vcvttss2si %xmm0, %rsi
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vcvttss2si %xmm1, %rax
; AVX512VL-NEXT: vcvttss2si %xmm0, %rcx
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512VL-NEXT: vcvttss2si %xmm1, %rdx
; AVX512VL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512VL-NEXT: vcvttss2si %xmm0, %rsi
; AVX1-NEXT: andq %rax, %rdx
; AVX1-NEXT: orq %rcx, %rdx
; AVX1-NEXT: vmovq %rdx, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-NEXT: vsubss %xmm1, %xmm3, %xmm4
; AVX1-NEXT: vcvttss2si %xmm4, %rax
; AVX1-NEXT: vcvttss2si %xmm3, %rcx
; AVX2-NEXT: andq %rax, %rdx
; AVX2-NEXT: orq %rcx, %rdx
; AVX2-NEXT: vmovq %rdx, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX2-NEXT: vsubss %xmm1, %xmm3, %xmm4
; AVX2-NEXT: vcvttss2si %xmm4, %rax
; AVX2-NEXT: vcvttss2si %xmm3, %rcx
; AVX512F-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512F-NEXT: vcvttss2usi %xmm1, %rax
; AVX512F-NEXT: vmovq %rax, %xmm1
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512F-NEXT: vcvttss2usi %xmm2, %rax
; AVX512F-NEXT: vmovq %rax, %xmm2
; AVX512F-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
; AVX512VL-NEXT: vcvttss2usi %xmm1, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512VL-NEXT: vcvttss2usi %xmm2, %rax
; AVX512VL-NEXT: vmovq %rax, %xmm2
; AVX512VL-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
; AVX1-NEXT: andq %rax, %rdx
; AVX1-NEXT: orq %rcx, %rdx
; AVX1-NEXT: vmovq %rdx, %xmm2
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-NEXT: vsubss %xmm1, %xmm3, %xmm4
; AVX1-NEXT: vcvttss2si %xmm4, %rax
; AVX1-NEXT: vcvttss2si %xmm3, %rcx
; AVX2-NEXT: andq %rax, %rdx
; AVX2-NEXT: orq %rcx, %rdx
; AVX2-NEXT: vmovq %rdx, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX2-NEXT: vsubss %xmm1, %xmm3, %xmm4
; AVX2-NEXT: vcvttss2si %xmm4, %rax
; AVX2-NEXT: vcvttss2si %xmm3, %rcx
; AVX512F-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512F-NEXT: vcvttss2usi %xmm1, %rax
; AVX512F-NEXT: vcvttss2usi %xmm0, %rcx
-; AVX512F-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512F-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512F-NEXT: vcvttss2usi %xmm1, %rdx
; AVX512F-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512F-NEXT: vcvttss2usi %xmm0, %rsi
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vcvttss2usi %xmm1, %rax
; AVX512VL-NEXT: vcvttss2usi %xmm0, %rcx
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512VL-NEXT: vcvttss2usi %xmm1, %rdx
; AVX512VL-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512VL-NEXT: vcvttss2usi %xmm0, %rsi
; AVX: # %bb.0:
; AVX-NEXT: subq $40, %rsp
; AVX-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: callq __truncdfhf2@PLT
; AVX-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
; AVX-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; AVX1: # %bb.0:
; AVX1-NEXT: subq $88, %rsp
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2: # %bb.0:
; AVX2-NEXT: subq $88, %rsp
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1: # %bb.0:
; AVX1-NEXT: subq $88, %rsp
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2: # %bb.0:
; AVX2-NEXT: subq $88, %rsp
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1: # %bb.0:
; AVX1-NEXT: subq $88, %rsp
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2: # %bb.0:
; AVX2-NEXT: subq $88, %rsp
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: subq $32, %rsp
; AVX-NEXT: movq %rdi, %rbx
; AVX-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: callq __truncdfhf2@PLT
; AVX-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX-NEXT: vmovdqa (%rsp), %xmm0 # 16-byte Reload
; AVX1-NEXT: subq $80, %rsp
; AVX1-NEXT: movq %rdi, %rbx
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: subq $80, %rsp
; AVX2-NEXT: movq %rdi, %rbx
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: subq $80, %rsp
; AVX1-NEXT: movq %rdi, %rbx
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: subq $80, %rsp
; AVX2-NEXT: movq %rdi, %rbx
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: subq $80, %rsp
; AVX1-NEXT: movq %rdi, %rbx
; AVX1-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-NEXT: vmovups (%rsp), %ymm0 # 32-byte Reload
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: callq __truncdfhf2@PLT
; AVX1-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: subq $80, %rsp
; AVX2-NEXT: movq %rdi, %rbx
; AVX2-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX2-NEXT: vmovupd (%rsp), %ymm0 # 32-byte Reload
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vmovapd %xmm0, (%rsp) # 16-byte Spill
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: callq __truncdfhf2@PLT
; AVX2-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; AVX1-LABEL: splat2_i64:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 $51, (%rdi), %ymm0, %ymm0 # ymm0 = mem[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
; AVX1-NEXT: vbroadcastf128 (%rdi), %ymm1 # ymm1 = mem[0,1,0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-NEXT: vmovupd %ymm0, 32(%rsi)
; AVX1-NEXT: vmovupd %ymm1, (%rsi)
; AVX1-NEXT: vzeroupper
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm1[0,1,2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm7, %xmm7
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm7 = xmm7[0,1],mem[2],xmm7[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm7 = xmm7[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm7 = xmm7[1,0]
; AVX1-ONLY-NEXT: vinsertps {{.*#+}} xmm7 = xmm7[0,1,2],xmm6[1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[3,0],mem[1,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm4[0,1,2,3],ymm3[4,5],ymm4[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm11, %xmm11
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm11 = xmm11[0,1],mem[2],xmm11[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm11 = xmm11[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm11 = xmm11[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm11[0,1,2],ymm8[3,4,5,6,7]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm12 = ymm0[1,0],ymm11[0,0],ymm0[5,4],ymm11[4,4]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1,2,3],ymm13[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm1
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3,0,1]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm12[0,1,2,3],ymm7[4,5],ymm12[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm13 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm2 = ymm12[0,1,2,3],mem[4,5],ymm12[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm5
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm2 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm6
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm4 = ymm15[0,1,2,3],mem[4,5],ymm15[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm4 = xmm4[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm9
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm9[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm4 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm4, %xmm4
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm4 = xmm4[0,1],mem[2],xmm4[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm4 = xmm4[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm4 = xmm4[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm4[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm4
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm10 = ymm4[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm3 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm3, %xmm3
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm3 = xmm3[0,1],mem[2],xmm3[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 288(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 608(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 928(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1248(%rdi), %ymm1
; AVX1-ONLY-NEXT: vmovups %ymm1, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm1 = mem[0,1,2,3],ymm1[4,5],mem[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 1088(%rdi), %ymm2
; AVX1-ONLY-NEXT: vmovups %ymm2, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm1 = xmm1[0,1],mem[2],xmm1[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm0[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 128(%rdi), %ymm3
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm9 = ymm3[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm2 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm2 = xmm2[0,1],mem[2],xmm2[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm1[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 448(%rdi), %ymm6
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm6[2,3,0,1]
; AVX1-ONLY-NEXT: # ymm5 = ymm0[0,1,2,3],mem[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vextractf128 $1, %ymm5, %xmm5
; AVX1-ONLY-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],mem[2],xmm5[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} xmm5 = xmm5[1,0]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} xmm5 = xmm5[1,0]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm5[0,1,2],ymm2[3,4,5,6,7]
; AVX1-ONLY-NEXT: vmovaps 768(%rdi), %ymm8
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,0,1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,1],xmm4[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1],ymm2[2,3],ymm3[4],ymm2[5,6],ymm3[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1],ymm3[2],ymm4[3,4],ymm3[5],ymm4[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm3[1,2],ymm0[3],ymm3[4,5],ymm0[6],ymm3[7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 64(%rcx)
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[1,1],xmm6[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm9, %ymm5, %ymm5
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5,6],ymm6[7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm6 = xmm8[1],xmm4[1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm6 = xmm4[1,1],xmm6[0,2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm3 = xmm3[1,1],xmm7[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm1[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1],ymm3[2,3],ymm6[4],ymm3[5,6],ymm6[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4],ymm6[5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm6[1,2],ymm1[3],ymm6[4,5],ymm1[6],ymm6[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm6 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm7[0,1],ymm6[2],ymm7[3,4],ymm6[5],ymm7[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm6[1,2],ymm0[3],ymm6[4,5],ymm0[6],ymm6[7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 128(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm1, 32(%rcx)
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm5 = xmm5[1,1],xmm6[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm12, %ymm5, %ymm5
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm3[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm5[0],ymm6[1],ymm5[2,3],ymm6[4],ymm5[5,6],ymm6[7]
; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm6
; AVX1-ONLY-NEXT: vmovaps 64(%rdi), %xmm12
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm9 = xmm9[1,1],xmm12[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm9, %ymm9
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm12 = ymm12[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm12[1],ymm9[2,3],ymm12[4],ymm9[5,6],ymm12[7]
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm12 = xmm11[1],xmm10[1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm12 = xmm10[1,1],xmm12[0,2]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm11 = xmm11[1,1],xmm12[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm11, %ymm11
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm12 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm12 = ymm12[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm11[0],ymm12[1],ymm11[2,3],ymm12[4],ymm11[5,6],ymm12[7]
; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm12
; AVX1-ONLY-NEXT: vmovaps 96(%rdi), %xmm13
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm7 = xmm7[1,1],xmm8[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm13, %ymm7, %ymm7
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm8 = ymm8[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm7[0],ymm8[1],ymm7[2,3],ymm8[4],ymm7[5,6],ymm8[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm8[1,2],ymm4[3],ymm8[4,5],ymm4[6],ymm8[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm8[1,2],ymm3[3],ymm8[4,5],ymm3[6],ymm8[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm8[1,2],ymm2[3],ymm8[4,5],ymm2[6],ymm8[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm8 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm13[0,1],ymm8[2],ymm13[3,4],ymm8[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm8[1,2],ymm0[3],ymm8[4,5],ymm0[6],ymm8[7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 320(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 128(%rcx)
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,1],xmm4[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vunpckhpd {{.*#+}} xmm0 = xmm6[1],xmm2[1]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm3[1,1],xmm0[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 64(%rsi), %xmm0
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 96(%rsi), %xmm0
; AVX1-ONLY-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[0,2]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 128(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovapd 128(%rdx), %ymm6
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovaps 160(%rsi), %xmm0
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovapd 160(%rdx), %ymm4
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm15 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7]
; AVX1-ONLY-NEXT: vmovaps 192(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 192(%rdi), %xmm1
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm3, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vmovapd 192(%rdx), %ymm2
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm2[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm0[0],ymm3[1],ymm0[2,3],ymm3[4],ymm0[5,6],ymm3[7]
; AVX1-ONLY-NEXT: vmovaps 224(%rsi), %xmm0
; AVX1-ONLY-NEXT: vmovaps 224(%rdi), %xmm3
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1
; AVX1-ONLY-NEXT: vmovapd 224(%rdx), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm0[2,3,2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm1[0],ymm3[1],ymm1[2,3],ymm3[4],ymm1[5,6],ymm3[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2],ymm13[3,4],ymm1[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm8 = ymm8[0],ymm1[1,2],ymm8[3],ymm1[4,5],ymm8[6],ymm1[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2],ymm13[3,4],ymm1[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm9 = ymm9[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm9 = ymm9[0],ymm1[1,2],ymm9[3],ymm1[4,5],ymm9[6],ymm1[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm1 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm13[0,1],ymm1[2],ymm13[3,4],ymm1[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm11 = ymm11[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm11[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm11[0],ymm1[1,2],ymm11[3],ymm1[4,5],ymm11[6],ymm1[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm11 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm13[0,1],ymm11[2],ymm13[3,4],ymm11[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm12 = ymm12[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0],ymm11[1,2],ymm12[3],ymm11[4,5],ymm12[6],ymm11[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm12[1,2],ymm6[3],ymm12[4,5],ymm6[6],ymm12[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm12[1,2],ymm4[3],ymm12[4,5],ymm4[6],ymm12[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm12[1,2],ymm2[3],ymm12[4,5],ymm2[6],ymm12[7]
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm12 = mem[0,0,3,3,4,4,7,7]
; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm12 = ymm13[0,1],ymm12[2],ymm13[3,4],ymm12[5],ymm13[6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm12[1,2],ymm0[3],ymm12[4,5],ymm0[6],ymm12[7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 704(%rcx)
; AVX1-ONLY-NEXT: vmovaps %ymm2, 608(%rcx)
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1,3,2,4,5,7,6]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm5[0,1,2],ymm2[3],ymm5[4,5],ymm2[6],ymm5[7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,2,2,3,7,6,6,7]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm4[1,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm4[1,0,3,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm3[0],ymm0[1],ymm3[2,3],ymm0[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3],ymm0[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovaps %ymm0, 32(%r8)
; AVX1-ONLY-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[u,u,1,0,u,u,u,6]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm1
; AVX1-ONLY-NEXT: vunpcklps {{.*#+}} ymm7 = ymm4[0],ymm1[0],ymm4[1],ymm1[1],ymm4[4],ymm1[4],ymm4[5],ymm1[5]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm7[0,0,3,3]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[0,0,3,3]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1],ymm5[2,3],ymm7[4,5,6],ymm5[7]
; AVX1-ONLY-NEXT: vbroadcastf128 {{.*#+}} ymm7 = mem[0,1,0,1]
; AVX1-ONLY-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm0
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm3[0,1,2,3],ymm1[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm4[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2],ymm0[3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm3[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2,3],ymm2[4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm9[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm6[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm6[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4,5,6],ymm3[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0],ymm2[1,2,3],ymm1[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1,2],ymm12[3],ymm1[4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm2 = ymm15[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2,3],ymm5[4],ymm2[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm14[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm9 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm13[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm9[0,1],ymm3[2],ymm9[3,4,5,6],ymm3[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: # ymm1 = ymm1[0,1,2,3],mem[4],ymm1[5,6,7]
; AVX1-ONLY-NEXT: vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm0 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm13[0,1],ymm0[2],ymm13[3,4,5,6],ymm0[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0],ymm0[1,2,3],ymm1[4,5],ymm0[6,7]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm6 = ymm6[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2,3],ymm7[4],ymm6[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm9[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm9 = ymm11[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm11[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm7 = ymm9[0,1],ymm7[2],ymm9[3,4,5,6],ymm7[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0],ymm7[1,2,3],ymm6[4,5],ymm7[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm6[0,1,2],ymm8[3],ymm6[4,5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm7 = ymm15[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm7[0,1,2,3],ymm2[4],ymm7[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm14[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm14[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm7[0,1],ymm3[2],ymm7[3,4,5,6],ymm3[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0],ymm3[1,2,3],ymm2[4,5],ymm3[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1,2],ymm4[3],ymm2[4,5,6,7]
; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm10 = ymm15[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm15[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm10[0,1],ymm6[2],ymm10[3,4,5,6],ymm6[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1,2,3],ymm3[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm3 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm3 = ymm3[0,1,2,3],mem[4],ymm3[5,6,7]
; AVX1-ONLY-NEXT: vpermilps $52, {{[-0-9]+}}(%r{{[sb]}}p), %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm6 = mem[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm15 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm13[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm6 = ymm15[0,1],ymm6[2],ymm15[3,4,5,6],ymm6[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0],ymm6[1,2,3],ymm3[4,5],ymm6[6,7]
; AVX1-ONLY-NEXT: vblendps $8, {{[-0-9]+}}(%r{{[sb]}}p), %ymm3, %ymm6 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm0 = ymm0[3,0,2,3,7,4,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4],ymm0[5,6,7]
; AVX1-ONLY-NEXT: vshufps {{.*#+}} ymm1 = ymm4[0,1,3,0,4,5,7,4]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm5[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm5[1,0,2,2]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm1 = ymm4[0,1],ymm1[2],ymm4[3,4,5,6],ymm1[7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4,5],ymm1[6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2],ymm2[3],ymm0[4,5,6,7]
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0,1],ymm2[2,3],ymm3[4,5],ymm2[6,7]
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm3
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],ymm1[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm0[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm4[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2],ymm3[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd %ymm3, 64(%rcx)
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1],ymm5[2,3],ymm6[4,5],ymm5[6,7]
; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm3[2,3],ymm7[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm7[0],ymm6[1],ymm7[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm7
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm8 = ymm0[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm0[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm8 = ymm2[2,3],ymm8[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm8[0],ymm7[1],ymm8[2],ymm7[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovapd %ymm7, 64(%rcx)
; AVX1-ONLY-NEXT: vblendps {{.*#+}} ymm11 = ymm12[0,1],ymm11[2,3],ymm12[4,5],ymm11[6,7]
; AVX1-ONLY-NEXT: vmovapd 80(%rdi), %xmm12
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm12[0,1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = ymm6[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm6[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm13 = ymm8[2,3],ymm13[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm12 = ymm13[0],ymm12[1],ymm13[2],ymm12[3]
; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm13
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],ymm4[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm14 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm14 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm14 = ymm4[2,3],ymm14[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2],ymm13[3]
; AVX1-ONLY-NEXT: vmovapd 112(%rdi), %xmm14
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm3[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm15 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm3[2,3],ymm15[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm15[0],ymm14[1],ymm15[2],ymm14[3]
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0,1],ymm9[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm0 = ymm7[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm9[2,3],ymm0[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0],ymm15[1],ymm0[2],ymm15[3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0],ymm9[1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm6[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm6[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2],ymm6[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0],ymm8[1],ymm6[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0,1],mem[2],ymm2[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm2[0],ymm4[1],ymm2[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0,1],mem[2],ymm1[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm3[1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovapd %ymm0, 64(%rcx)
; AVX1-ONLY-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 16(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm3[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm13[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovupd %ymm4, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm10
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm10[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm10[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 112(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm9
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm9[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm9[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 144(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm8
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm6 = ymm8[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm6 = ymm8[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm2[2,3],ymm6[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm6[0],ymm0[1],ymm6[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 176(%rdi), %xmm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm7
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm11 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm7[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm1[2,3],ymm11[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm11[0],ymm6[1],ymm11[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 208(%rdi), %xmm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 192(%rsi), %ymm5
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm15 = ymm5[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm5[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm0[2,3],ymm15[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm15[0],ymm6[1],ymm15[2],ymm6[3]
; AVX1-ONLY-NEXT: vmovapd 224(%rdx), %ymm15
; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm14
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],ymm15[2,3]
; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm3
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm11 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm11 = ymm3[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm11 = ymm15[2,3],ymm11[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm11 = ymm11[0],ymm14[1],ymm11[2],ymm14[3]
; AVX1-ONLY-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm14 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm14 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm14 = ymm14[0,1],mem[2],ymm14[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm14[0],ymm13[1],ymm14[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2],ymm10[3]
; AVX1-ONLY-NEXT: vblendpd $2, {{[-0-9]+}}(%r{{[sb]}}p), %ymm10, %ymm10 # 32-byte Folded Reload
; AVX1-ONLY-NEXT: # ymm10 = ymm10[0],mem[1],ymm10[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm9 = ymm9[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm9 = ymm9[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm9[0,1],mem[2],ymm9[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm9[0],ymm4[1],ymm9[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm8 = ymm8[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm8 = ymm8[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm8 = ymm8[0,1],mem[2],ymm8[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm8[0],ymm2[1],ymm8[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm5 = ymm5[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm5 = ymm5[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2],ymm5[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm5[0],ymm0[1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0,1],mem[2],ymm3[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm15[1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovapd %ymm11, 736(%rcx)
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd (%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 32(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 64(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 96(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 128(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm2[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm14[2,3]
; AVX1-ONLY-NEXT: vmovapd 160(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm14[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm12[2,3]
; AVX1-ONLY-NEXT: vmovapd 192(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm12[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm11[2,3]
; AVX1-ONLY-NEXT: vmovapd 224(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm11[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm9[2,3]
; AVX1-ONLY-NEXT: vmovapd 256(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm9[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm8[2,3]
; AVX1-ONLY-NEXT: vmovapd 288(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm8[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm6[2,3]
; AVX1-ONLY-NEXT: vmovapd 320(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm6[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
; AVX1-ONLY-NEXT: vmovapd 352(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm5[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm4[2,3]
; AVX1-ONLY-NEXT: vmovapd 384(%rsi), %ymm1
; AVX1-ONLY-NEXT: vmovupd %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm4[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 432(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovapd 416(%rsi), %ymm13
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm13[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm13[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm3[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm1[0],ymm0[1],ymm1[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 448(%rsi), %ymm10
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm10[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm10[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm7 = ymm2[2,3],ymm7[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0],ymm0[1],ymm7[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm7
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm7[0,1],ymm1[2,3]
; AVX1-ONLY-NEXT: vmovapd 480(%rsi), %ymm7
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm15 = ymm7[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm7[0,0,3,2]
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, (%rsp) # 32-byte Spill
; AVX1-ONLY-NEXT: # ymm15 = mem[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2],ymm15[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm15[0],ymm4[1],ymm15[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm13 = ymm13[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm13 = ymm13[0,1],mem[2],ymm13[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm13[0],ymm3[1],ymm13[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm10 = ymm10[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm10 = ymm10[0,1],mem[2],ymm10[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm10[0],ymm2[1],ymm10[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm7 = ymm7[1,0,2,2]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm7 = ymm7[0,1],mem[2],ymm7[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm7[0],ymm1[1],ymm7[2,3]
; AVX1-ONLY-NEXT: vmovups (%rsp), %ymm7 # 32-byte Reload
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm0[0,1],ymm9[2,3]
; AVX1-ONLY-NEXT: vmovapd 48(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm15 = ymm15[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm15 = ymm15[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 32(%rax), %ymm1
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm15 = ymm1[2,3],ymm15[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm15[0],ymm0[1],ymm15[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0,1],ymm3[2,3]
; AVX1-ONLY-NEXT: vmovapd 112(%rdi), %xmm6
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm6 = ymm6[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm4 = ymm4[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm4 = ymm4[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 96(%rax), %ymm8
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm8[2,3],ymm4[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm4[0],ymm6[1],ymm4[2],ymm6[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm4 = ymm15[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 240(%rdi), %xmm15
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm15 = ymm15[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 224(%rax), %ymm0
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm0[2,3],ymm1[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm1 = ymm1[0],ymm15[1],ymm1[2],ymm15[3]
; AVX1-ONLY-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
; AVX1-ONLY-NEXT: vmovapd 432(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 416(%rax), %ymm1
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3],ymm2[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm9 = ymm3[0,1],ymm0[2,3]
; AVX1-ONLY-NEXT: vmovapd 464(%rdi), %xmm0
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm5[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm2 = ymm2[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 448(%rax), %ymm4
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm4[2,3],ymm2[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2],ymm0[3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm2 = ymm5[0,1],ymm2[2,3]
; AVX1-ONLY-NEXT: vmovapd 496(%rdi), %xmm5
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm5 = ymm5[0,1],mem[2,3]
-; AVX1-ONLY-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,0,3,2]
+; AVX1-ONLY-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,0,3,2]
; AVX1-ONLY-NEXT: vmovapd 480(%rax), %ymm11
; AVX1-ONLY-NEXT: vperm2f128 {{.*#+}} ymm3 = ymm11[2,3],ymm3[2,3]
; AVX1-ONLY-NEXT: vblendpd {{.*#+}} ymm3 = ymm3[0],ymm5[1],ymm3[2],ymm5[3]
; AVX1-NEXT: vmulpd %xmm0, %xmm0, %xmm0
; AVX1-NEXT: vmulpd %xmm2, %xmm2, %xmm1
; AVX1-NEXT: vaddpd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-NEXT: retq
;
; AVX2-LABEL: fmul_v2f64:
; AVX2-NEXT: vmulpd %xmm0, %xmm0, %xmm0
; AVX2-NEXT: vmulpd %xmm2, %xmm2, %xmm1
; AVX2-NEXT: vaddpd %xmm0, %xmm1, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: retq
;
; AVX512-LABEL: fmul_v2f64:
; AVX512-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0]
; AVX512-NEXT: vmulpd %xmm0, %xmm0, %xmm0
; AVX512-NEXT: vfmadd231pd {{.*#+}} xmm0 = (xmm2 * xmm2) + xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: retq
%s = shufflevector <2 x double> %x, <2 x double> %y, <4 x i32> <i32 2, i32 0, i32 1, i32 3>
%bo = fmul fast <4 x double> %s, %s
;
; AVX1-SLOW-LABEL: test_v4f32:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
;
; AVX2-LABEL: test_v4f32:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
;
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX2-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddps %ymm2, %ymm1, %ymm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vaddps %ymm2, %ymm1, %ymm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-FAST-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vhaddps %xmm1, %xmm1, %xmm1
; AVX1-FAST-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vaddps %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX2-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddps %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddps %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
;
; AVX1-SLOW-LABEL: test_v4f32_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX2-LABEL: test_v4f32_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vzeroupper
; AVX2-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX1-SLOW-LABEL: test_v4f32_undef:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX2-LABEL: test_v4f32_undef:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm0
; AVX1-FAST-NEXT: vzeroupper
; AVX2-NEXT: vaddps %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
;
; AVX1-SLOW-LABEL: test_v2f64:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: retq
;
; AVX2-LABEL: test_v2f64:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX2-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX1-SLOW-NEXT: vaddpd %ymm2, %ymm1, %ymm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX2-NEXT: vaddpd %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX2-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX512-NEXT: vaddpd %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX1-SLOW-NEXT: vaddpd %ymm2, %ymm1, %ymm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX1-SLOW-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX2-NEXT: vaddpd %ymm2, %ymm1, %ymm1
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX2-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX512-NEXT: vaddpd %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
;
; AVX1-SLOW-LABEL: test_v2f64_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: retq
;
;
; AVX2-LABEL: test_v2f64_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call fast double @llvm.vector.reduce.fadd.f64.v2f64(double 0.0, <2 x double> %a0)
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX1-SLOW-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX1-SLOW-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v2f64_undef:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: retq
;
;
; AVX2-LABEL: test_v2f64_undef:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call fast double @llvm.vector.reduce.fadd.f64.v2f64(double 0.0, <2 x double> %a0)
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2: # %bb.0:
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX1-SLOW-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX1-SLOW-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX1-SLOW-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX2-NEXT: vaddpd %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX2-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
; AVX512-NEXT: vaddpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vaddpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-FAST-LABEL: test_v4f32_zero:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm1
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-FAST-LABEL: test_v8f32_zero:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm1
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX1-FAST-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX1-SLOW: # %bb.0:
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm2
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm3, %xmm2, %xmm2
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX1-SLOW-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-FAST-LABEL: test_v16f32_zero:
; AVX1-FAST: # %bb.0:
; AVX1-FAST-NEXT: vhaddps %xmm0, %xmm0, %xmm2
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm2, %xmm2
; AVX1-FAST-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX1-FAST-NEXT: vaddss %xmm3, %xmm2, %xmm2
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX1-FAST-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-FAST-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX1-FAST-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX1-FAST-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2: # %bb.0:
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX2-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vaddss %xmm0, %xmm2, %xmm2
; AVX2-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX2-NEXT: vaddss %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX2-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX2-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX2-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vaddss %xmm0, %xmm2, %xmm2
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vaddss %xmm3, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vaddss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vaddss %xmm0, %xmm2, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vaddss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vaddss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vaddss %xmm0, %xmm1, %xmm0
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fadd.f64.v2f64(double %a0, <2 x double> %a1)
; AVX-LABEL: test_v4f64:
; AVX: # %bb.0:
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v4f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-LABEL: test_v8f64:
; AVX: # %bb.0:
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm3, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v8f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-LABEL: test_v16f64:
; AVX: # %bb.0:
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm5, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm4, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm4[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm4[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm4, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v16f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm2, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm2, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v2f64_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: retq
;
;
; AVX2-LABEL: test_v2f64_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v2f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fadd.f64.v2f64(double -0.0, <2 x double> %a0)
;
; AVX1-SLOW-LABEL: test_v4f64_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm1
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX1-FAST-NEXT: vhaddpd %xmm0, %xmm0, %xmm1
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v4f64_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm1
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v4f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v8f64_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm0, %xmm2
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX1-FAST-NEXT: vhaddpd %xmm0, %xmm0, %xmm2
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v8f64_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm0, %xmm2
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v8f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX1-SLOW-LABEL: test_v16f64_zero:
; AVX1-SLOW: # %bb.0:
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm4, %xmm0, %xmm4
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm4, %xmm4
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm0, %xmm4, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm4, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-SLOW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-SLOW-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-SLOW-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-SLOW-NEXT: vzeroupper
; AVX1-SLOW-NEXT: retq
; AVX1-FAST-NEXT: vhaddpd %xmm0, %xmm0, %xmm4
; AVX1-FAST-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm4, %xmm4
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm0, %xmm4, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm4, %xmm0, %xmm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX1-FAST-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX1-FAST-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX1-FAST-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX1-FAST-NEXT: vzeroupper
; AVX1-FAST-NEXT: retq
;
; AVX2-LABEL: test_v16f64_zero:
; AVX2: # %bb.0:
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm4, %xmm0, %xmm4
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX2-NEXT: vaddsd %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX2-NEXT: vaddsd %xmm0, %xmm4, %xmm0
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm4, %xmm0, %xmm0
; AVX2-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX2-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX2-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512-LABEL: test_v16f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm2
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v2f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fadd.f64.v2f64(double undef, <2 x double> %a0)
;
; AVX-LABEL: test_v4f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v8f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v16f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm4
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vaddsd %xmm0, %xmm4, %xmm4
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vaddsd %xmm0, %xmm4, %xmm0
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm4, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vaddsd %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v16f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vaddsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm3
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vaddsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vaddsd %xmm0, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vaddsd %xmm0, %xmm2, %xmm0
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vaddsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vaddsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmaxps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call fast double @llvm.vector.reduce.fmin.v2f64(<2 x double> %a0)
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vminpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vminpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmaxps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmaxps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm1, %xmm0, %xmm0
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call nnan double @llvm.vector.reduce.fmax.v2f64(<2 x double> %a0)
;
; AVX-LABEL: test_v3f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmaxsd %xmm0, %xmm1, %xmm0
;
; AVX512-LABEL: test_v3f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX512-NEXT: vmaxsd %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmaxpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmaxpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmaxpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvps %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmaxss %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm0, %xmm1, %xmm0
; AVX-NEXT: retq
; AVX512-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovss %xmm1, %xmm2, %xmm2 {%k1}
; AVX512-NEXT: vcmpunordss %xmm2, %xmm2, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmaxss %xmm2, %xmm1, %xmm0
; AVX512-NEXT: vmovss %xmm1, %xmm0, %xmm0 {%k1}
; AVX512-NEXT: retq
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm0, %xmm3, %xmm4
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm0
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX512-NEXT: vmaxss %xmm0, %xmm3, %xmm4
; AVX512-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX-NEXT: vmaxss %xmm0, %xmm7, %xmm8
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm0
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512BW-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX512BW-NEXT: vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
; AVX512BW-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX512BW-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX512BW-NEXT: vmaxss %xmm0, %xmm7, %xmm8
; AVX512BW-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vmaxss %xmm0, %xmm7, %xmm8
; AVX512VL-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvps %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmaxss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
; AVX-NEXT: vmaxss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmaxss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
; AVX512BW-NEXT: vmaxss %xmm0, %xmm2, %xmm3
; AVX512BW-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vmaxss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vmaxss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vmaxss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vmaxss %xmm3, %xmm1, %xmm3
; AVX512BW-NEXT: vmovss %xmm1, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vmaxss %xmm3, %xmm1, %xmm3
; AVX512BW-NEXT: vmovss %xmm1, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm2, %xmm2, %k1
; AVX512BW-NEXT: vmaxss %xmm2, %xmm1, %xmm2
; AVX512BW-NEXT: vmovss %xmm1, %xmm2, %xmm2 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm2, %xmm2, %k1
; AVX512BW-NEXT: vmaxss %xmm2, %xmm1, %xmm2
; AVX512BW-NEXT: vmovss %xmm1, %xmm2, %xmm2 {%k1}
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf32x4 $3, %zmm0, %xmm3
; AVX512VL-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
; AVX512VL-NEXT: vextractf32x4 $2, %zmm0, %xmm6
; AVX512VL-NEXT: vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm7 = xmm6[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm7 = xmm6[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm8 = xmm6[1,1,3,3]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm9
; AVX512VL-NEXT: vshufps {{.*#+}} xmm10 = xmm9[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm11 = xmm9[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm12 = xmm9[1,1,3,3]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm13 = xmm0[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm14 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm14 = xmm0[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm15 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vmaxss %xmm0, %xmm15, %xmm16
; AVX512VL-NEXT: vcmpunordss %xmm0, %xmm0, %k1
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm0
; AVX-NEXT: vblendvpd %xmm0, %xmm1, %xmm2, %xmm0
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm0, %xmm2, %xmm1
; AVX512-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovsd %xmm2, %xmm1, %xmm1 {%k1}
; AVX-LABEL: test_v4f64:
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm0, %xmm3, %xmm4
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm0
; AVX-NEXT: vblendvpd %xmm0, %xmm3, %xmm4, %xmm0
; AVX512-LABEL: test_v4f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm0, %xmm3, %xmm4
; AVX512-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovsd %xmm3, %xmm4, %xmm4 {%k1}
; AVX-NEXT: vmaxpd %ymm0, %ymm1, %ymm2
; AVX-NEXT: vcmpunordpd %ymm0, %ymm0, %ymm0
; AVX-NEXT: vblendvpd %ymm0, %ymm1, %ymm2, %ymm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvpd %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm1
; AVX-NEXT: vcmpunordsd %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX512BW-LABEL: test_v8f64:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vextractf32x4 $3, %zmm0, %xmm2
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512BW-NEXT: vextractf32x4 $2, %zmm0, %xmm3
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX512BW-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm6 = xmm5[1,0]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm7 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm6 = xmm5[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm7 = xmm0[1,0]
; AVX512BW-NEXT: vmaxsd %xmm0, %xmm7, %xmm8
; AVX512BW-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512BW-NEXT: vmovsd %xmm7, %xmm8, %xmm8 {%k1}
; AVX512VL-LABEL: test_v8f64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf32x4 $3, %zmm0, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512VL-NEXT: vextractf32x4 $2, %zmm0, %xmm3
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm6 = xmm5[1,0]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm7 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm6 = xmm5[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm7 = xmm0[1,0]
; AVX512VL-NEXT: vmaxsd %xmm0, %xmm7, %xmm8
; AVX512VL-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512VL-NEXT: vmovsd %xmm7, %xmm8, %xmm8 {%k1}
; AVX-NEXT: vmaxpd %ymm0, %ymm1, %ymm2
; AVX-NEXT: vcmpunordpd %ymm0, %ymm0, %ymm0
; AVX-NEXT: vblendvpd %ymm0, %ymm1, %ymm2, %ymm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvpd %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm1
; AVX-NEXT: vcmpunordsd %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX512-NEXT: vmaxpd %zmm0, %zmm1, %zmm2
; AVX512-NEXT: vcmpunordpd %zmm0, %zmm0, %k1
; AVX512-NEXT: vmovapd %zmm1, %zmm2 {%k1}
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm2[1,0]
; AVX512-NEXT: vmaxsd %xmm2, %xmm0, %xmm1
; AVX512-NEXT: vcmpunordsd %xmm2, %xmm2, %k1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
; AVX512-NEXT: vmaxsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmaxsd %xmm1, %xmm2, %xmm0
; AVX512-NEXT: vmovsd %xmm2, %xmm0, %xmm0 {%k1}
; AVX512-NEXT: vzeroupper
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm1, %xmm0, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vminss %xmm0, %xmm1, %xmm0
; AVX-NEXT: retq
;
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vminss %xmm0, %xmm1, %xmm0
; AVX512-NEXT: retq
%1 = call nnan float @llvm.vector.reduce.fmin.v3f32(<3 x float> %a0)
;
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vminps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vminps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm1, %xmm0, %xmm0
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call nnan double @llvm.vector.reduce.fmin.v2f64(<2 x double> %a0)
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vminpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vminpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vminpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vminpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vminpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm0, %xmm3, %xmm4
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm0
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm0[3,3,3,3]
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX512-NEXT: vminss %xmm0, %xmm3, %xmm4
; AVX512-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX-NEXT: vminss %xmm0, %xmm7, %xmm8
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm0
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512BW-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX512BW-NEXT: vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
; AVX512BW-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX512BW-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX512BW-NEXT: vminss %xmm0, %xmm7, %xmm8
; AVX512BW-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512VL-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm4 = xmm1[1,1,3,3]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm5 = xmm0[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm6 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm6 = xmm0[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm7 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vminss %xmm0, %xmm7, %xmm8
; AVX512VL-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX-NEXT: vcmpunordss %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvps %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vminss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
; AVX-NEXT: vminss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vminss %xmm1, %xmm3, %xmm1
; AVX-NEXT: vblendvps %xmm2, %xmm3, %xmm1, %xmm1
; AVX-NEXT: vcmpunordss %xmm1, %xmm1, %xmm2
; AVX512BW-NEXT: vminss %xmm0, %xmm2, %xmm3
; AVX512BW-NEXT: vcmpunordss %xmm0, %xmm0, %k1
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vminss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vminss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vminss %xmm3, %xmm2, %xmm3
; AVX512BW-NEXT: vmovss %xmm2, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vminss %xmm3, %xmm1, %xmm3
; AVX512BW-NEXT: vmovss %xmm1, %xmm3, %xmm3 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm3, %xmm3, %k1
; AVX512BW-NEXT: vminss %xmm3, %xmm1, %xmm3
; AVX512BW-NEXT: vmovss %xmm1, %xmm3, %xmm3 {%k1}
; AVX512BW-NEXT: vcmpunordss %xmm2, %xmm2, %k1
; AVX512BW-NEXT: vminss %xmm2, %xmm1, %xmm2
; AVX512BW-NEXT: vmovss %xmm1, %xmm2, %xmm2 {%k1}
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512BW-NEXT: vcmpunordss %xmm2, %xmm2, %k1
; AVX512BW-NEXT: vminss %xmm2, %xmm1, %xmm2
; AVX512BW-NEXT: vmovss %xmm1, %xmm2, %xmm2 {%k1}
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf32x4 $3, %zmm0, %xmm3
; AVX512VL-NEXT: vshufps {{.*#+}} xmm1 = xmm3[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm3[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm3[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm4 = xmm3[1,1,3,3]
; AVX512VL-NEXT: vextractf32x4 $2, %zmm0, %xmm6
; AVX512VL-NEXT: vshufps {{.*#+}} xmm5 = xmm6[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm7 = xmm6[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm7 = xmm6[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm8 = xmm6[1,1,3,3]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm9
; AVX512VL-NEXT: vshufps {{.*#+}} xmm10 = xmm9[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm11 = xmm9[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm11 = xmm9[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm12 = xmm9[1,1,3,3]
; AVX512VL-NEXT: vshufps {{.*#+}} xmm13 = xmm0[3,3,3,3]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm14 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm14 = xmm0[1,0]
; AVX512VL-NEXT: vmovshdup {{.*#+}} xmm15 = xmm0[1,1,3,3]
; AVX512VL-NEXT: vminss %xmm0, %xmm15, %xmm16
; AVX512VL-NEXT: vcmpunordss %xmm0, %xmm0, %k1
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm0
; AVX-NEXT: vblendvpd %xmm0, %xmm1, %xmm2, %xmm0
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm0, %xmm2, %xmm1
; AVX512-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovsd %xmm2, %xmm1, %xmm1 {%k1}
;
; AVX-LABEL: test_v3f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvpd %xmm3, %xmm1, %xmm2, %xmm1
;
; AVX512-LABEL: test_v3f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm0, %xmm1, %xmm2
; AVX512-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovsd %xmm1, %xmm2, %xmm2 {%k1}
; AVX-LABEL: test_v4f64:
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm0, %xmm3, %xmm4
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm0
; AVX-NEXT: vblendvpd %xmm0, %xmm3, %xmm4, %xmm0
; AVX512-LABEL: test_v4f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm0, %xmm3, %xmm4
; AVX512-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512-NEXT: vmovsd %xmm3, %xmm4, %xmm4 {%k1}
; AVX-NEXT: vminpd %ymm0, %ymm1, %ymm2
; AVX-NEXT: vcmpunordpd %ymm0, %ymm0, %ymm0
; AVX-NEXT: vblendvpd %ymm0, %ymm1, %ymm2, %ymm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvpd %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm1
; AVX-NEXT: vcmpunordsd %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX512BW-LABEL: test_v8f64:
; AVX512BW: # %bb.0:
; AVX512BW-NEXT: vextractf32x4 $3, %zmm0, %xmm2
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512BW-NEXT: vextractf32x4 $2, %zmm0, %xmm3
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX512BW-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm6 = xmm5[1,0]
-; AVX512BW-NEXT: vpermilpd {{.*#+}} xmm7 = xmm0[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm6 = xmm5[1,0]
+; AVX512BW-NEXT: vshufpd {{.*#+}} xmm7 = xmm0[1,0]
; AVX512BW-NEXT: vminsd %xmm0, %xmm7, %xmm8
; AVX512BW-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512BW-NEXT: vmovsd %xmm7, %xmm8, %xmm8 {%k1}
; AVX512VL-LABEL: test_v8f64:
; AVX512VL: # %bb.0:
; AVX512VL-NEXT: vextractf32x4 $3, %zmm0, %xmm1
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512VL-NEXT: vextractf32x4 $2, %zmm0, %xmm3
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm4 = xmm3[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm4 = xmm3[1,0]
; AVX512VL-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm6 = xmm5[1,0]
-; AVX512VL-NEXT: vpermilpd {{.*#+}} xmm7 = xmm0[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm6 = xmm5[1,0]
+; AVX512VL-NEXT: vshufpd {{.*#+}} xmm7 = xmm0[1,0]
; AVX512VL-NEXT: vminsd %xmm0, %xmm7, %xmm8
; AVX512VL-NEXT: vcmpunordsd %xmm0, %xmm0, %k1
; AVX512VL-NEXT: vmovsd %xmm7, %xmm8, %xmm8 {%k1}
; AVX-NEXT: vminpd %ymm0, %ymm1, %ymm2
; AVX-NEXT: vcmpunordpd %ymm0, %ymm0, %ymm0
; AVX-NEXT: vblendvpd %ymm0, %ymm1, %ymm2, %ymm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm0, %xmm1, %xmm2
; AVX-NEXT: vcmpunordsd %xmm0, %xmm0, %xmm3
; AVX-NEXT: vblendvpd %xmm3, %xmm1, %xmm2, %xmm1
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm1
; AVX-NEXT: vcmpunordsd %xmm1, %xmm1, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vblendvpd %xmm2, %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX512-NEXT: vminpd %zmm0, %zmm1, %zmm2
; AVX512-NEXT: vcmpunordpd %zmm0, %zmm0, %k1
; AVX512-NEXT: vmovapd %zmm1, %zmm2 {%k1}
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm2[1,0]
; AVX512-NEXT: vminsd %xmm2, %xmm0, %xmm1
; AVX512-NEXT: vcmpunordsd %xmm2, %xmm2, %k1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
; AVX512-NEXT: vminsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vmovsd %xmm0, %xmm1, %xmm1 {%k1}
; AVX512-NEXT: vcmpunordsd %xmm1, %xmm1, %k1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vminsd %xmm1, %xmm2, %xmm0
; AVX512-NEXT: vmovsd %xmm2, %xmm0, %xmm0 {%k1}
; AVX512-NEXT: vzeroupper
;
; AVX-LABEL: test_v4f32:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
;
; AVX512-LABEL: test_v4f32:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT: vmulps %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulps %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulps %ymm2, %ymm1, %ymm1
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT: vmulps %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulps %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulps %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulps %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
;
; AVX-LABEL: test_v4f32_zero:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
;
; AVX-LABEL: test_v4f32_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
;
; AVX512-LABEL: test_v4f32_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulps %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulps %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulps %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
;
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX-NEXT: vmulpd %ymm2, %ymm1, %ymm1
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX512-NEXT: vmulpd %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX-NEXT: vmulpd %ymm2, %ymm1, %ymm1
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX512-NEXT: vmulpd %zmm2, %zmm1, %zmm1
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulpd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
;
; AVX-LABEL: test_v2f64_zero:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64_zero:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call fast double @llvm.vector.reduce.fmul.f64.v2f64(double 1.0, <2 x double> %a0)
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmulpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmulpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v2f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call fast double @llvm.vector.reduce.fmul.f64.v2f64(double 1.0, <2 x double> %a0)
; AVX: # %bb.0:
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512: # %bb.0:
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmulpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmulpd %ymm1, %ymm0, %ymm0
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-NEXT: vmulpd %zmm1, %zmm0, %zmm0
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm1
; AVX512-NEXT: vmulpd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm2[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm2[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vmulss %xmm0, %xmm2, %xmm2
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm1, %xmm0, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX: # %bb.0:
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm3 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vmulss %xmm0, %xmm2, %xmm2
; AVX-NEXT: vmovshdup {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm0[1,0]
; AVX-NEXT: vmulss %xmm3, %xmm2, %xmm2
; AVX-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX-NEXT: vmulss %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm2 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmovshdup {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulss %xmm2, %xmm0, %xmm0
; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm1[3,3,3,3]
; AVX-NEXT: vmulss %xmm1, %xmm0, %xmm0
; AVX512: # %bb.0:
; AVX512-NEXT: vmovshdup {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm3 = xmm2[1,1,3,3]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm2[1,0]
; AVX512-NEXT: vmulss %xmm3, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,3,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm1
; AVX512-NEXT: vmovshdup {{.*#+}} xmm2 = xmm0[1,1,3,3]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulss %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; AVX512-NEXT: vmulss %xmm0, %xmm1, %xmm0
; AVX-LABEL: test_v2f64:
; AVX: # %bb.0:
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fmul.f64.v2f64(double %a0, <2 x double> %a1)
; AVX-LABEL: test_v4f64:
; AVX: # %bb.0:
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v4f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-LABEL: test_v8f64:
; AVX: # %bb.0:
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm3, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v8f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
; AVX-LABEL: test_v16f64:
; AVX: # %bb.0:
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm5, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm4, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm4[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm4[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm4, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
; AVX512-LABEL: test_v16f64:
; AVX512: # %bb.0:
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm2, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm2, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v2f64_one:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64_one:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fmul.f64.v2f64(double 1.0, <2 x double> %a0)
;
; AVX-LABEL: test_v4f64_one:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm1
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4f64_one:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v8f64_one:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm2
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8f64_one:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v16f64_one:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm4, %xmm0, %xmm4
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm4, %xmm4
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm4, %xmm0
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm4, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v16f64_one:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm2
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm2, %xmm0
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v2f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v2f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; AVX512-NEXT: retq
%1 = call double @llvm.vector.reduce.fmul.f64.v2f64(double undef, <2 x double> %a0)
;
; AVX-LABEL: test_v4f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v4f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v8f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm2, %xmm2
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm2, %xmm0
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v8f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm0[1,0]
; AVX512-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm1, %xmm1
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm1
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: test_v16f64_undef:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm0[1,0]
; AVX-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm4
; AVX-NEXT: vextractf128 $1, %ymm0, %xmm0
; AVX-NEXT: vmulsd %xmm0, %xmm4, %xmm4
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: vmulsd %xmm0, %xmm4, %xmm0
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm4 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm4 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm4, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm1, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm2[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm2[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm2, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vmulsd %xmm3, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm3[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm3[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vextractf128 $1, %ymm3, %xmm1
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX-NEXT: vzeroupper
; AVX-NEXT: retq
;
; AVX512-LABEL: test_v16f64_undef:
; AVX512: # %bb.0:
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm0[1,0]
; AVX512-NEXT: vmulsd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm2, %xmm2
; AVX512-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $2, %zmm0, %xmm3
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm3 = xmm3[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm3 = xmm3[1,0]
; AVX512-NEXT: vmulsd %xmm3, %xmm2, %xmm2
; AVX512-NEXT: vextractf32x4 $3, %zmm0, %xmm0
; AVX512-NEXT: vmulsd %xmm0, %xmm2, %xmm2
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX512-NEXT: vmulsd %xmm0, %xmm2, %xmm0
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf128 $1, %ymm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $2, %zmm1, %xmm2
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm2 = xmm2[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm2 = xmm2[1,0]
; AVX512-NEXT: vmulsd %xmm2, %xmm0, %xmm0
; AVX512-NEXT: vextractf32x4 $3, %zmm1, %xmm1
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
-; AVX512-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0]
+; AVX512-NEXT: vshufpd {{.*#+}} xmm1 = xmm1[1,0]
; AVX512-NEXT: vmulsd %xmm1, %xmm0, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
;
; AVX-LABEL: shuffle_v2f64_10:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,0]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 1, i32 0>
;
; AVX-LABEL: shuffle_v2f64_11:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 1, i32 1>
ret <2 x double> %shuffle
;
; AVX-LABEL: shuffle_v2f64_32:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm1[1,0]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 3, i32 2>
;
; AVX-LABEL: shuffle_v2f64_33:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm1[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm1[1,1]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 3, i32 3>
ret <2 x double> %shuffle
;
; AVX-LABEL: shuffle_v2f64_3u:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm1[1,0]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm1[1,0]
; AVX-NEXT: retq
%shuffle = shufflevector <2 x double> %a, <2 x double> %b, <2 x i32> <i32 3, i32 undef>
ret <2 x double> %shuffle
; AVX1-LABEL: shuffle_v4f64_1000:
; AVX1: # %bb.0:
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4f64_1000:
; AVX1-LABEL: shuffle_v4f64_2233:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4f64_2233:
; AVX1-LABEL: shuffle_v4f64_3210:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4f64_3210:
define <4 x double> @shuffle_v4f64_0023(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_0023:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,2,3]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 0, i32 0, i32 2, i32 3>
define <4 x double> @shuffle_v4f64_1032(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_1032:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 3, i32 2>
ret <4 x double> %shuffle
define <4 x double> @shuffle_v4f64_1133(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_1133:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 1, i32 3, i32 3>
ret <4 x double> %shuffle
define <4 x double> @shuffle_v4f64_1023(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_1023:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,3]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 2, i32 3>
ret <4 x double> %shuffle
define <4 x double> @shuffle_v4f64_1022(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_1022:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 2, i32 2>
ret <4 x double> %shuffle
; AVX1OR2-LABEL: shuffle_v4f64_1054:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1OR2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1OR2-NEXT: retq
;
; AVX512VL-SLOW-LABEL: shuffle_v4f64_1054:
; AVX512VL-SLOW: # %bb.0:
; AVX512VL-SLOW-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-SLOW-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-SLOW-NEXT: retq
;
; AVX512VL-FAST-ALL-LABEL: shuffle_v4f64_1054:
; AVX512VL-FAST-PERLANE-LABEL: shuffle_v4f64_1054:
; AVX512VL-FAST-PERLANE: # %bb.0:
; AVX512VL-FAST-PERLANE-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX512VL-FAST-PERLANE-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-FAST-PERLANE-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-FAST-PERLANE-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 0, i32 5, i32 4>
ret <4 x double> %shuffle
; AVX1OR2-LABEL: shuffle_v4f64_3254:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1OR2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1OR2-NEXT: retq
;
; AVX512VL-SLOW-LABEL: shuffle_v4f64_3254:
; AVX512VL-SLOW: # %bb.0:
; AVX512VL-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-SLOW-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-SLOW-NEXT: retq
;
; AVX512VL-FAST-ALL-LABEL: shuffle_v4f64_3254:
; AVX512VL-FAST-PERLANE-LABEL: shuffle_v4f64_3254:
; AVX512VL-FAST-PERLANE: # %bb.0:
; AVX512VL-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX512VL-FAST-PERLANE-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-FAST-PERLANE-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-FAST-PERLANE-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 5, i32 4>
ret <4 x double> %shuffle
; AVX1OR2-LABEL: shuffle_v4f64_3276:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1OR2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1OR2-NEXT: retq
;
; AVX512VL-SLOW-LABEL: shuffle_v4f64_3276:
; AVX512VL-SLOW: # %bb.0:
; AVX512VL-SLOW-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-SLOW-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-SLOW-NEXT: retq
;
; AVX512VL-FAST-ALL-LABEL: shuffle_v4f64_3276:
; AVX512VL-FAST-PERLANE-LABEL: shuffle_v4f64_3276:
; AVX512VL-FAST-PERLANE: # %bb.0:
; AVX512VL-FAST-PERLANE-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX512VL-FAST-PERLANE-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-FAST-PERLANE-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-FAST-PERLANE-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 3, i32 2, i32 7, i32 6>
ret <4 x double> %shuffle
; AVX1OR2-LABEL: shuffle_v4f64_1076:
; AVX1OR2: # %bb.0:
; AVX1OR2-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1OR2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1OR2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1OR2-NEXT: retq
;
; AVX512VL-SLOW-LABEL: shuffle_v4f64_1076:
; AVX512VL-SLOW: # %bb.0:
; AVX512VL-SLOW-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX512VL-SLOW-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX512VL-SLOW-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX512VL-SLOW-NEXT: retq
;
; AVX512VL-FAST-LABEL: shuffle_v4f64_1076:
define <4 x double> @shuffle_v4f64_11uu(<4 x double> %a, <4 x double> %b) {
; ALL-LABEL: shuffle_v4f64_11uu:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; ALL-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; ALL-NEXT: retq
%shuffle = shufflevector <4 x double> %a, <4 x double> %b, <4 x i32> <i32 1, i32 1, i32 undef, i32 undef>
ret <4 x double> %shuffle
; AVX1-LABEL: shuffle_v4f64_3333:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4f64_3333:
; ALL: # %bb.0:
; ALL-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
; ALL-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; ALL-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; ALL-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; ALL-NEXT: retq
%1 = shufflevector <2 x double> %a, <2 x double> undef, <2 x i32> <i32 1, i32 0>
%2 = shufflevector <2 x double> %b, <2 x double> undef, <2 x i32> <i32 1, i32 0>
; AVX1-LABEL: shuffle_v4i64_1000:
; AVX1: # %bb.0:
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,2,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_1000:
; AVX1-LABEL: shuffle_v4i64_3210:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_3210:
; AVX1-LABEL: shuffle_v4i64_1054:
; AVX1: # %bb.0:
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_1054:
; AVX1-LABEL: shuffle_v4i64_3254:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_3254:
; AVX1-LABEL: shuffle_v4i64_3276:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3],ymm1[2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_3276:
; AVX1-LABEL: shuffle_v4i64_1076:
; AVX1: # %bb.0:
; AVX1-NEXT: vblendpd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_1076:
; AVX1-LABEL: shuffle_v4i64_3333:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,3,3]
; AVX1-NEXT: retq
;
; AVX2-LABEL: shuffle_v4i64_3333:
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm1[2,3,0,1]
; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm2[0,0],ymm1[2,2],ymm2[4,4],ymm1[6,6]
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3,4],ymm0[5],ymm1[6],ymm0[7]
; AVX1-NEXT: retq
;
; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[3,1,1,0]
; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; AVX1-NEXT: vperm2f128 {{.*#+}} ymm1 = ymm1[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,2]
; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0],ymm1[1,2,3],ymm0[4],ymm1[5],ymm0[6,7]
; AVX1-NEXT: retq
;
define <16 x float> @shuffle_v16f32_00_01_00_01_06_07_06_07_08_09_10_11_12_13_12_13(<16 x float> %a, <16 x float> %b) {
; ALL-LABEL: shuffle_v16f32_00_01_00_01_06_07_06_07_08_09_10_11_12_13_12_13:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,0,3,3,4,5,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,0,3,3,4,5,6,6]
; ALL-NEXT: retq
%shuffle = shufflevector <16 x float> %a, <16 x float> %b, <16 x i32> <i32 0, i32 1, i32 0, i32 1, i32 6, i32 7, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 12, i32 13>
ret <16 x float> %shuffle
define <8 x double> @shuffle_v8f64_00234467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_00234467:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,0,2,3,4,4,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,0,2,3,4,4,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 0, i32 2, i32 3, i32 4, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_10325476(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10325476:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,0,3,2,5,4,7,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,0,3,2,5,4,7,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_11335577(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_11335577:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,1,3,3,5,5,7,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,1,3,3,5,5,7,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 5, i32 5, i32 7, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_10235467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10235467:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,0,2,3,5,4,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,0,2,3,5,4,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_10225466(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10225466:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,0,2,2,5,4,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,0,2,2,5,4,6,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 2, i32 2, i32 5, i32 4, i32 6, i32 6>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_10324567(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_10324567:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,0,3,2,4,5,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,0,3,2,4,5,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_11334567(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_11334567:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,1,3,3,4,5,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,1,3,3,4,5,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 4, i32 5, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_01235467(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_01235467:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,2,3,5,4,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,1,2,3,5,4,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_01235466(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_01235466:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,1,2,3,5,4,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,1,2,3,5,4,6,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 5, i32 4, i32 6, i32 6>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_103245uu(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_103245uu:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,0,3,2,4,5,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,0,3,2,4,5,6,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 0, i32 3, i32 2, i32 4, i32 5, i32 undef, i32 undef>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_1133uu67(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_1133uu67:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[1,1,3,3,4,4,6,7]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[1,1,3,3,4,4,6,7]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 1, i32 1, i32 3, i32 3, i32 undef, i32 undef, i32 6, i32 7>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_0uu354uu(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_0uu354uu:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,0,2,3,5,4,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,0,2,3,5,4,6,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 0, i32 undef, i32 undef, i32 3, i32 5, i32 4, i32 undef, i32 undef>
ret <8 x double> %shuffle
define <8 x double> @shuffle_v8f64_uuu3uu66(<8 x double> %a, <8 x double> %b) {
; ALL-LABEL: shuffle_v8f64_uuu3uu66:
; ALL: # %bb.0:
-; ALL-NEXT: vpermilpd {{.*#+}} zmm0 = zmm0[0,0,2,3,4,4,6,6]
+; ALL-NEXT: vshufpd {{.*#+}} zmm0 = zmm0[0,0,2,3,4,4,6,6]
; ALL-NEXT: ret{{[l|q]}}
%shuffle = shufflevector <8 x double> %a, <8 x double> %b, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 3, i32 undef, i32 undef, i32 6, i32 6>
ret <8 x double> %shuffle
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %ecx
; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; X86-AVX1-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,1,2,2]
+; X86-AVX1-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,1,2,2]
; X86-AVX1-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
; X86-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm5
; X86-AVX1-NEXT: vshufpd {{.*#+}} ymm4 = ymm5[1],ymm4[0],ymm5[2],ymm4[3]
; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %edx
; X86-AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
; X86-AVX2-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
-; X86-AVX2-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; X86-AVX2-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; X86-AVX2-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm2[0,1]
; X86-AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; X86-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
; X64-AVX1-LABEL: PR48908:
; X64-AVX1: # %bb.0:
; X64-AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
-; X64-AVX1-NEXT: vpermilpd {{.*#+}} ymm3 = ymm3[0,1,2,2]
+; X64-AVX1-NEXT: vshufpd {{.*#+}} ymm3 = ymm3[0,1,2,2]
; X64-AVX1-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
; X64-AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm5
; X64-AVX1-NEXT: vshufpd {{.*#+}} ymm4 = ymm5[1],ymm4[0],ymm5[2],ymm4[3]
; X64-AVX2: # %bb.0:
; X64-AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm3
; X64-AVX2-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm1[2,3],ymm2[0,1]
-; X64-AVX2-NEXT: vpermilpd {{.*#+}} xmm5 = xmm1[1,0]
+; X64-AVX2-NEXT: vshufpd {{.*#+}} xmm5 = xmm1[1,0]
; X64-AVX2-NEXT: vperm2f128 {{.*#+}} ymm6 = ymm0[0,1],ymm2[0,1]
; X64-AVX2-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,2,2,1]
; X64-AVX2-NEXT: vblendpd {{.*#+}} ymm3 = ymm6[0],ymm3[1],ymm6[2],ymm3[3]
; AVX1: # %bb.0:
; AVX1-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
; AVX1-NEXT: ret{{[l|q]}}
;
; AVX2-LABEL: concat_self_v4i64:
; X86-NEXT: vunpcklpd {{.*#+}} xmm2 = xmm2[0],mem[0]
; X86-NEXT: vinsertf128 $1, {{\.?LCPI[0-9]+_[0-9]+}}, %ymm2, %ymm2
; X86-NEXT: vpermil2pd $0, %ymm2, %ymm1, %ymm0, %ymm0
-; X86-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
+; X86-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
; X86-NEXT: retl
;
; X64-AVX-LABEL: demandedelts_vpermil2pd256_as_shufpd:
; X64-AVX2-LABEL: demandedelts_vpermil2pd256_as_shufpd:
; X64-AVX2: # %bb.0:
; X64-AVX2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
-; X64-AVX2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
+; X64-AVX2-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
; X64-AVX2-NEXT: retq
%res0 = insertelement <4 x i64> <i64 0, i64 4, i64 2, i64 7>, i64 %a2, i32 0
%res1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x i64> %res0, i8 0)
;
; AVX-LABEL: combine_undef_input_test9:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; AVX-NEXT: retq
%1 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 5, i32 5>
%2 = shufflevector <4 x float> %1, <4 x float> %a, <4 x i32> <i32 6, i32 7, i32 0, i32 1>
;
; AVX-LABEL: combine_undef_input_test19:
; AVX: # %bb.0:
-; AVX-NEXT: vpermilpd {{.*#+}} xmm0 = xmm0[1,1]
+; AVX-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1,1]
; AVX-NEXT: retq
%1 = shufflevector <4 x float> %a, <4 x float> undef, <4 x i32> <i32 2, i32 3, i32 5, i32 5>
%2 = shufflevector <4 x float> %a, <4 x float> %1, <4 x i32> <i32 2, i32 3, i32 4, i32 5>
; AVX1-LABEL: splat2_v4f64_load_store:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 $51, (%rdi), %ymm0, %ymm0 # ymm0 = mem[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
; AVX1-NEXT: vbroadcastf128 (%rdi), %ymm1 # ymm1 = mem[0,1,0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-NEXT: vmovupd %ymm0, 32(%rsi)
; AVX1-NEXT: vmovupd %ymm1, (%rsi)
; AVX1-NEXT: vzeroupper
; AVX1-LABEL: splat2_v4i64_load_store:
; AVX1: # %bb.0:
; AVX1-NEXT: vperm2f128 $51, (%rdi), %ymm0, %ymm0 # ymm0 = mem[2,3,2,3]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0,0,3,3]
; AVX1-NEXT: vbroadcastf128 (%rdi), %ymm1 # ymm1 = mem[0,1,0,1]
-; AVX1-NEXT: vpermilpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
+; AVX1-NEXT: vshufpd {{.*#+}} ymm1 = ymm1[0,0,3,3]
; AVX1-NEXT: vmovupd %ymm0, 32(%rsi)
; AVX1-NEXT: vmovupd %ymm1, (%rsi)
; AVX1-NEXT: vzeroupper