1433 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefixes=SSE,SSE23; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3 | FileCheck %s --check-prefixes=SSE,SSSE34; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefixes=SSE,SSE415; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=AVX6; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX7 8;9; Unary shuffle indices from registers10;11 12define <2 x double> @var_shuffle_v2f64_v2f64_xx_i64(<2 x double> %x, i64 %i0, i64 %i1) nounwind {13; SSE-LABEL: var_shuffle_v2f64_v2f64_xx_i64:14; SSE: # %bb.0:15; SSE-NEXT: andl $1, %esi16; SSE-NEXT: andl $1, %edi17; SSE-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)18; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero19; SSE-NEXT: movhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]20; SSE-NEXT: retq21;22; AVX-LABEL: var_shuffle_v2f64_v2f64_xx_i64:23; AVX: # %bb.0:24; AVX-NEXT: andl $1, %esi25; AVX-NEXT: andl $1, %edi26; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)27; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero28; AVX-NEXT: vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]29; AVX-NEXT: retq30 %x0 = extractelement <2 x double> %x, i64 %i031 %x1 = extractelement <2 x double> %x, i64 %i132 %r0 = insertelement <2 x double> undef, double %x0, i32 033 %r1 = insertelement <2 x double> %r0, double %x1, i32 134 ret <2 x double> %r135}36 37define <2 x i64> @var_shuffle_v2i64_v2i64_xx_i64(<2 x i64> %x, i32 %i0, i32 %i1) nounwind {38; SSE-LABEL: var_shuffle_v2i64_v2i64_xx_i64:39; SSE: # %bb.0:40; SSE-NEXT: # kill: def $esi killed $esi def $rsi41; SSE-NEXT: # kill: def $edi killed $edi def $rdi42; SSE-NEXT: andl $1, %edi43; SSE-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)44; SSE-NEXT: andl $1, %esi45; SSE-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero46; SSE-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero47; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]48; SSE-NEXT: retq49;50; AVX-LABEL: var_shuffle_v2i64_v2i64_xx_i64:51; AVX: # %bb.0:52; AVX-NEXT: # kill: def $esi killed $esi def $rsi53; AVX-NEXT: # kill: def $edi killed $edi def $rdi54; AVX-NEXT: andl $1, %edi55; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)56; AVX-NEXT: andl $1, %esi57; AVX-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero58; AVX-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero59; AVX-NEXT: vmovlhps {{.*#+}} xmm0 = xmm1[0],xmm0[0]60; AVX-NEXT: retq61 %x0 = extractelement <2 x i64> %x, i32 %i062 %x1 = extractelement <2 x i64> %x, i32 %i163 %r0 = insertelement <2 x i64> undef, i64 %x0, i32 064 %r1 = insertelement <2 x i64> %r0, i64 %x1, i32 165 ret <2 x i64> %r166}67 68define <4 x float> @var_shuffle_v4f32_v4f32_xxxx_i32(<4 x float> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3) nounwind {69; SSE2-LABEL: var_shuffle_v4f32_v4f32_xxxx_i32:70; SSE2: # %bb.0:71; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx72; SSE2-NEXT: # kill: def $edx killed $edx def $rdx73; SSE2-NEXT: # kill: def $esi killed $esi def $rsi74; SSE2-NEXT: # kill: def $edi killed $edi def $rdi75; SSE2-NEXT: andl $3, %edi76; SSE2-NEXT: andl $3, %esi77; SSE2-NEXT: andl $3, %edx78; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)79; SSE2-NEXT: andl $3, %ecx80; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero81; SSE2-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero82; SSE2-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]83; SSE2-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero84; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero85; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]86; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]87; SSE2-NEXT: retq88;89; SSSE3-LABEL: var_shuffle_v4f32_v4f32_xxxx_i32:90; SSSE3: # %bb.0:91; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx92; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx93; SSSE3-NEXT: # kill: def $esi killed $esi def $rsi94; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi95; SSSE3-NEXT: andl $3, %edi96; SSSE3-NEXT: andl $3, %esi97; SSSE3-NEXT: andl $3, %edx98; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)99; SSSE3-NEXT: andl $3, %ecx100; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero101; SSSE3-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero102; SSSE3-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]103; SSSE3-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero104; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero105; SSSE3-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]106; SSSE3-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]107; SSSE3-NEXT: retq108;109; SSE41-LABEL: var_shuffle_v4f32_v4f32_xxxx_i32:110; SSE41: # %bb.0:111; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx112; SSE41-NEXT: # kill: def $edx killed $edx def $rdx113; SSE41-NEXT: # kill: def $esi killed $esi def $rsi114; SSE41-NEXT: # kill: def $edi killed $edi def $rdi115; SSE41-NEXT: andl $3, %edi116; SSE41-NEXT: andl $3, %esi117; SSE41-NEXT: andl $3, %edx118; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)119; SSE41-NEXT: andl $3, %ecx120; SSE41-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero121; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]122; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]123; SSE41-NEXT: insertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]124; SSE41-NEXT: retq125;126; AVX-LABEL: var_shuffle_v4f32_v4f32_xxxx_i32:127; AVX: # %bb.0:128; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx129; AVX-NEXT: # kill: def $edx killed $edx def $rdx130; AVX-NEXT: # kill: def $esi killed $esi def $rsi131; AVX-NEXT: # kill: def $edi killed $edi def $rdi132; AVX-NEXT: andl $3, %edi133; AVX-NEXT: andl $3, %esi134; AVX-NEXT: andl $3, %edx135; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)136; AVX-NEXT: andl $3, %ecx137; AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero138; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[2,3]139; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1],mem[0],xmm0[3]140; AVX-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,1,2],mem[0]141; AVX-NEXT: retq142 %x0 = extractelement <4 x float> %x, i32 %i0143 %x1 = extractelement <4 x float> %x, i32 %i1144 %x2 = extractelement <4 x float> %x, i32 %i2145 %x3 = extractelement <4 x float> %x, i32 %i3146 %r0 = insertelement <4 x float> undef, float %x0, i32 0147 %r1 = insertelement <4 x float> %r0, float %x1, i32 1148 %r2 = insertelement <4 x float> %r1, float %x2, i32 2149 %r3 = insertelement <4 x float> %r2, float %x3, i32 3150 ret <4 x float> %r3151}152 153define <4 x i32> @var_shuffle_v4i32_v4i32_xxxx_i32(<4 x i32> %x, i32 %i0, i32 %i1, i32 %i2, i32 %i3) nounwind {154; SSE2-LABEL: var_shuffle_v4i32_v4i32_xxxx_i32:155; SSE2: # %bb.0:156; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx157; SSE2-NEXT: # kill: def $edx killed $edx def $rdx158; SSE2-NEXT: # kill: def $esi killed $esi def $rsi159; SSE2-NEXT: # kill: def $edi killed $edi def $rdi160; SSE2-NEXT: andl $3, %edi161; SSE2-NEXT: andl $3, %esi162; SSE2-NEXT: andl $3, %edx163; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)164; SSE2-NEXT: andl $3, %ecx165; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero166; SSE2-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero167; SSE2-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]168; SSE2-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero169; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero170; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]171; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]172; SSE2-NEXT: retq173;174; SSSE3-LABEL: var_shuffle_v4i32_v4i32_xxxx_i32:175; SSSE3: # %bb.0:176; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx177; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx178; SSSE3-NEXT: # kill: def $esi killed $esi def $rsi179; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi180; SSSE3-NEXT: andl $3, %edi181; SSSE3-NEXT: andl $3, %esi182; SSSE3-NEXT: andl $3, %edx183; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)184; SSSE3-NEXT: andl $3, %ecx185; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero186; SSSE3-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero187; SSSE3-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]188; SSSE3-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero189; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero190; SSSE3-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]191; SSSE3-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]192; SSSE3-NEXT: retq193;194; SSE41-LABEL: var_shuffle_v4i32_v4i32_xxxx_i32:195; SSE41: # %bb.0:196; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx197; SSE41-NEXT: # kill: def $edx killed $edx def $rdx198; SSE41-NEXT: # kill: def $esi killed $esi def $rsi199; SSE41-NEXT: # kill: def $edi killed $edi def $rdi200; SSE41-NEXT: andl $3, %edi201; SSE41-NEXT: andl $3, %esi202; SSE41-NEXT: andl $3, %edx203; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)204; SSE41-NEXT: andl $3, %ecx205; SSE41-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero206; SSE41-NEXT: pinsrd $1, -24(%rsp,%rsi,4), %xmm0207; SSE41-NEXT: pinsrd $2, -24(%rsp,%rdx,4), %xmm0208; SSE41-NEXT: pinsrd $3, -24(%rsp,%rcx,4), %xmm0209; SSE41-NEXT: retq210;211; AVX-LABEL: var_shuffle_v4i32_v4i32_xxxx_i32:212; AVX: # %bb.0:213; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx214; AVX-NEXT: # kill: def $edx killed $edx def $rdx215; AVX-NEXT: # kill: def $esi killed $esi def $rsi216; AVX-NEXT: # kill: def $edi killed $edi def $rdi217; AVX-NEXT: andl $3, %edi218; AVX-NEXT: andl $3, %esi219; AVX-NEXT: andl $3, %edx220; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)221; AVX-NEXT: andl $3, %ecx222; AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero223; AVX-NEXT: vpinsrd $1, -24(%rsp,%rsi,4), %xmm0, %xmm0224; AVX-NEXT: vpinsrd $2, -24(%rsp,%rdx,4), %xmm0, %xmm0225; AVX-NEXT: vpinsrd $3, -24(%rsp,%rcx,4), %xmm0, %xmm0226; AVX-NEXT: retq227 %x0 = extractelement <4 x i32> %x, i32 %i0228 %x1 = extractelement <4 x i32> %x, i32 %i1229 %x2 = extractelement <4 x i32> %x, i32 %i2230 %x3 = extractelement <4 x i32> %x, i32 %i3231 %r0 = insertelement <4 x i32> undef, i32 %x0, i32 0232 %r1 = insertelement <4 x i32> %r0, i32 %x1, i32 1233 %r2 = insertelement <4 x i32> %r1, i32 %x2, i32 2234 %r3 = insertelement <4 x i32> %r2, i32 %x3, i32 3235 ret <4 x i32> %r3236}237 238define <8 x i16> @var_shuffle_v8i16_v8i16_xxxxxxxx_i16(<8 x i16> %x, i16 %i0, i16 %i1, i16 %i2, i16 %i3, i16 %i4, i16 %i5, i16 %i6, i16 %i7) nounwind {239; SSE2-LABEL: var_shuffle_v8i16_v8i16_xxxxxxxx_i16:240; SSE2: # %bb.0:241; SSE2-NEXT: # kill: def $r9d killed $r9d def $r9242; SSE2-NEXT: # kill: def $r8d killed $r8d def $r8243; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx244; SSE2-NEXT: # kill: def $edx killed $edx def $rdx245; SSE2-NEXT: # kill: def $esi killed $esi def $rsi246; SSE2-NEXT: # kill: def $edi killed $edi def $rdi247; SSE2-NEXT: movzwl {{[0-9]+}}(%rsp), %eax248; SSE2-NEXT: andl $7, %eax249; SSE2-NEXT: movzwl {{[0-9]+}}(%rsp), %r10d250; SSE2-NEXT: andl $7, %r10d251; SSE2-NEXT: andl $7, %edi252; SSE2-NEXT: andl $7, %esi253; SSE2-NEXT: andl $7, %edx254; SSE2-NEXT: andl $7, %ecx255; SSE2-NEXT: andl $7, %r8d256; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)257; SSE2-NEXT: andl $7, %r9d258; SSE2-NEXT: movzwl -24(%rsp,%r10,2), %r10d259; SSE2-NEXT: movd %r10d, %xmm0260; SSE2-NEXT: movzwl -24(%rsp,%rax,2), %eax261; SSE2-NEXT: movd %eax, %xmm1262; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]263; SSE2-NEXT: movzwl -24(%rsp,%r9,2), %eax264; SSE2-NEXT: movd %eax, %xmm0265; SSE2-NEXT: movzwl -24(%rsp,%r8,2), %eax266; SSE2-NEXT: movd %eax, %xmm2267; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]268; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]269; SSE2-NEXT: movzwl -24(%rsp,%rcx,2), %eax270; SSE2-NEXT: movd %eax, %xmm0271; SSE2-NEXT: movzwl -24(%rsp,%rdx,2), %eax272; SSE2-NEXT: movd %eax, %xmm1273; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]274; SSE2-NEXT: movzwl -24(%rsp,%rsi,2), %eax275; SSE2-NEXT: movd %eax, %xmm3276; SSE2-NEXT: movzwl -24(%rsp,%rdi,2), %eax277; SSE2-NEXT: movd %eax, %xmm0278; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]279; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]280; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]281; SSE2-NEXT: retq282;283; SSSE3-LABEL: var_shuffle_v8i16_v8i16_xxxxxxxx_i16:284; SSSE3: # %bb.0:285; SSSE3-NEXT: # kill: def $r9d killed $r9d def $r9286; SSSE3-NEXT: # kill: def $r8d killed $r8d def $r8287; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx288; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx289; SSSE3-NEXT: # kill: def $esi killed $esi def $rsi290; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi291; SSSE3-NEXT: movzwl {{[0-9]+}}(%rsp), %eax292; SSSE3-NEXT: andl $7, %eax293; SSSE3-NEXT: movzwl {{[0-9]+}}(%rsp), %r10d294; SSSE3-NEXT: andl $7, %r10d295; SSSE3-NEXT: andl $7, %edi296; SSSE3-NEXT: andl $7, %esi297; SSSE3-NEXT: andl $7, %edx298; SSSE3-NEXT: andl $7, %ecx299; SSSE3-NEXT: andl $7, %r8d300; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)301; SSSE3-NEXT: andl $7, %r9d302; SSSE3-NEXT: movzwl -24(%rsp,%r10,2), %r10d303; SSSE3-NEXT: movd %r10d, %xmm0304; SSSE3-NEXT: movzwl -24(%rsp,%rax,2), %eax305; SSSE3-NEXT: movd %eax, %xmm1306; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]307; SSSE3-NEXT: movzwl -24(%rsp,%r9,2), %eax308; SSSE3-NEXT: movd %eax, %xmm0309; SSSE3-NEXT: movzwl -24(%rsp,%r8,2), %eax310; SSSE3-NEXT: movd %eax, %xmm2311; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]312; SSSE3-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]313; SSSE3-NEXT: movzwl -24(%rsp,%rcx,2), %eax314; SSSE3-NEXT: movd %eax, %xmm0315; SSSE3-NEXT: movzwl -24(%rsp,%rdx,2), %eax316; SSSE3-NEXT: movd %eax, %xmm1317; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]318; SSSE3-NEXT: movzwl -24(%rsp,%rsi,2), %eax319; SSSE3-NEXT: movd %eax, %xmm3320; SSSE3-NEXT: movzwl -24(%rsp,%rdi,2), %eax321; SSSE3-NEXT: movd %eax, %xmm0322; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm3[0],xmm0[1],xmm3[1],xmm0[2],xmm3[2],xmm0[3],xmm3[3]323; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]324; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]325; SSSE3-NEXT: retq326;327; SSE41-LABEL: var_shuffle_v8i16_v8i16_xxxxxxxx_i16:328; SSE41: # %bb.0:329; SSE41-NEXT: # kill: def $r9d killed $r9d def $r9330; SSE41-NEXT: # kill: def $r8d killed $r8d def $r8331; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx332; SSE41-NEXT: # kill: def $edx killed $edx def $rdx333; SSE41-NEXT: # kill: def $esi killed $esi def $rsi334; SSE41-NEXT: # kill: def $edi killed $edi def $rdi335; SSE41-NEXT: movzwl {{[0-9]+}}(%rsp), %eax336; SSE41-NEXT: andl $7, %eax337; SSE41-NEXT: movzwl {{[0-9]+}}(%rsp), %r10d338; SSE41-NEXT: andl $7, %r10d339; SSE41-NEXT: andl $7, %edi340; SSE41-NEXT: andl $7, %esi341; SSE41-NEXT: andl $7, %edx342; SSE41-NEXT: andl $7, %ecx343; SSE41-NEXT: andl $7, %r8d344; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)345; SSE41-NEXT: andl $7, %r9d346; SSE41-NEXT: movzwl -24(%rsp,%rdi,2), %edi347; SSE41-NEXT: movd %edi, %xmm0348; SSE41-NEXT: pinsrw $1, -24(%rsp,%rsi,2), %xmm0349; SSE41-NEXT: pinsrw $2, -24(%rsp,%rdx,2), %xmm0350; SSE41-NEXT: pinsrw $3, -24(%rsp,%rcx,2), %xmm0351; SSE41-NEXT: pinsrw $4, -24(%rsp,%r8,2), %xmm0352; SSE41-NEXT: pinsrw $5, -24(%rsp,%r9,2), %xmm0353; SSE41-NEXT: pinsrw $6, -24(%rsp,%r10,2), %xmm0354; SSE41-NEXT: pinsrw $7, -24(%rsp,%rax,2), %xmm0355; SSE41-NEXT: retq356;357; AVX-LABEL: var_shuffle_v8i16_v8i16_xxxxxxxx_i16:358; AVX: # %bb.0:359; AVX-NEXT: # kill: def $r9d killed $r9d def $r9360; AVX-NEXT: # kill: def $r8d killed $r8d def $r8361; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx362; AVX-NEXT: # kill: def $edx killed $edx def $rdx363; AVX-NEXT: # kill: def $esi killed $esi def $rsi364; AVX-NEXT: # kill: def $edi killed $edi def $rdi365; AVX-NEXT: movzwl {{[0-9]+}}(%rsp), %eax366; AVX-NEXT: andl $7, %eax367; AVX-NEXT: movzwl {{[0-9]+}}(%rsp), %r10d368; AVX-NEXT: andl $7, %r10d369; AVX-NEXT: andl $7, %edi370; AVX-NEXT: andl $7, %esi371; AVX-NEXT: andl $7, %edx372; AVX-NEXT: andl $7, %ecx373; AVX-NEXT: andl $7, %r8d374; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)375; AVX-NEXT: andl $7, %r9d376; AVX-NEXT: movzwl -24(%rsp,%rdi,2), %edi377; AVX-NEXT: vmovd %edi, %xmm0378; AVX-NEXT: vpinsrw $1, -24(%rsp,%rsi,2), %xmm0, %xmm0379; AVX-NEXT: vpinsrw $2, -24(%rsp,%rdx,2), %xmm0, %xmm0380; AVX-NEXT: vpinsrw $3, -24(%rsp,%rcx,2), %xmm0, %xmm0381; AVX-NEXT: vpinsrw $4, -24(%rsp,%r8,2), %xmm0, %xmm0382; AVX-NEXT: vpinsrw $5, -24(%rsp,%r9,2), %xmm0, %xmm0383; AVX-NEXT: vpinsrw $6, -24(%rsp,%r10,2), %xmm0, %xmm0384; AVX-NEXT: vpinsrw $7, -24(%rsp,%rax,2), %xmm0, %xmm0385; AVX-NEXT: retq386 %x0 = extractelement <8 x i16> %x, i16 %i0387 %x1 = extractelement <8 x i16> %x, i16 %i1388 %x2 = extractelement <8 x i16> %x, i16 %i2389 %x3 = extractelement <8 x i16> %x, i16 %i3390 %x4 = extractelement <8 x i16> %x, i16 %i4391 %x5 = extractelement <8 x i16> %x, i16 %i5392 %x6 = extractelement <8 x i16> %x, i16 %i6393 %x7 = extractelement <8 x i16> %x, i16 %i7394 %r0 = insertelement <8 x i16> undef, i16 %x0, i32 0395 %r1 = insertelement <8 x i16> %r0, i16 %x1, i32 1396 %r2 = insertelement <8 x i16> %r1, i16 %x2, i32 2397 %r3 = insertelement <8 x i16> %r2, i16 %x3, i32 3398 %r4 = insertelement <8 x i16> %r3, i16 %x4, i32 4399 %r5 = insertelement <8 x i16> %r4, i16 %x5, i32 5400 %r6 = insertelement <8 x i16> %r5, i16 %x6, i32 6401 %r7 = insertelement <8 x i16> %r6, i16 %x7, i32 7402 ret <8 x i16> %r7403}404 405define <16 x i8> @var_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8(<16 x i8> %x, i8 %i0, i8 %i1, i8 %i2, i8 %i3, i8 %i4, i8 %i5, i8 %i6, i8 %i7, i8 %i8, i8 %i9, i8 %i10, i8 %i11, i8 %i12, i8 %i13, i8 %i14, i8 %i15) nounwind {406; SSE2-LABEL: var_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:407; SSE2: # %bb.0:408; SSE2-NEXT: # kill: def $r9d killed $r9d def $r9409; SSE2-NEXT: # kill: def $r8d killed $r8d def $r8410; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx411; SSE2-NEXT: # kill: def $edx killed $edx def $rdx412; SSE2-NEXT: # kill: def $esi killed $esi def $rsi413; SSE2-NEXT: # kill: def $edi killed $edi def $rdi414; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax415; SSE2-NEXT: andl $15, %eax416; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)417; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax418; SSE2-NEXT: movd %eax, %xmm1419; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax420; SSE2-NEXT: andl $15, %eax421; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax422; SSE2-NEXT: movd %eax, %xmm2423; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax424; SSE2-NEXT: andl $15, %eax425; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax426; SSE2-NEXT: movd %eax, %xmm4427; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax428; SSE2-NEXT: andl $15, %eax429; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax430; SSE2-NEXT: movd %eax, %xmm3431; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax432; SSE2-NEXT: andl $15, %eax433; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax434; SSE2-NEXT: movd %eax, %xmm6435; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax436; SSE2-NEXT: andl $15, %eax437; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax438; SSE2-NEXT: movd %eax, %xmm7439; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax440; SSE2-NEXT: andl $15, %eax441; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax442; SSE2-NEXT: movd %eax, %xmm8443; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax444; SSE2-NEXT: andl $15, %eax445; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax446; SSE2-NEXT: movd %eax, %xmm5447; SSE2-NEXT: andl $15, %ecx448; SSE2-NEXT: movzbl -24(%rsp,%rcx), %eax449; SSE2-NEXT: movd %eax, %xmm9450; SSE2-NEXT: andl $15, %edx451; SSE2-NEXT: movzbl -24(%rsp,%rdx), %eax452; SSE2-NEXT: movd %eax, %xmm10453; SSE2-NEXT: andl $15, %esi454; SSE2-NEXT: movzbl -24(%rsp,%rsi), %eax455; SSE2-NEXT: movd %eax, %xmm11456; SSE2-NEXT: andl $15, %edi457; SSE2-NEXT: movzbl -24(%rsp,%rdi), %eax458; SSE2-NEXT: movd %eax, %xmm0459; SSE2-NEXT: andl $15, %r9d460; SSE2-NEXT: movzbl -24(%rsp,%r9), %eax461; SSE2-NEXT: movd %eax, %xmm12462; SSE2-NEXT: andl $15, %r8d463; SSE2-NEXT: movzbl -24(%rsp,%r8), %eax464; SSE2-NEXT: movd %eax, %xmm13465; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax466; SSE2-NEXT: andl $15, %eax467; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax468; SSE2-NEXT: movd %eax, %xmm14469; SSE2-NEXT: movzbl {{[0-9]+}}(%rsp), %eax470; SSE2-NEXT: andl $15, %eax471; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax472; SSE2-NEXT: movd %eax, %xmm15473; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]474; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]475; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]476; SSE2-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]477; SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]478; SSE2-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]479; SSE2-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]480; SSE2-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]481; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3],xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]482; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]483; SSE2-NEXT: punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]484; SSE2-NEXT: punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3],xmm15[4],xmm14[4],xmm15[5],xmm14[5],xmm15[6],xmm14[6],xmm15[7],xmm14[7]485; SSE2-NEXT: punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]486; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]487; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]488; SSE2-NEXT: retq489;490; SSSE3-LABEL: var_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:491; SSSE3: # %bb.0:492; SSSE3-NEXT: # kill: def $r9d killed $r9d def $r9493; SSSE3-NEXT: # kill: def $r8d killed $r8d def $r8494; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx495; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx496; SSSE3-NEXT: # kill: def $esi killed $esi def $rsi497; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi498; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax499; SSSE3-NEXT: andl $15, %eax500; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)501; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax502; SSSE3-NEXT: movd %eax, %xmm1503; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax504; SSSE3-NEXT: andl $15, %eax505; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax506; SSSE3-NEXT: movd %eax, %xmm2507; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax508; SSSE3-NEXT: andl $15, %eax509; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax510; SSSE3-NEXT: movd %eax, %xmm4511; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax512; SSSE3-NEXT: andl $15, %eax513; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax514; SSSE3-NEXT: movd %eax, %xmm3515; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax516; SSSE3-NEXT: andl $15, %eax517; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax518; SSSE3-NEXT: movd %eax, %xmm6519; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax520; SSSE3-NEXT: andl $15, %eax521; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax522; SSSE3-NEXT: movd %eax, %xmm7523; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax524; SSSE3-NEXT: andl $15, %eax525; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax526; SSSE3-NEXT: movd %eax, %xmm8527; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax528; SSSE3-NEXT: andl $15, %eax529; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax530; SSSE3-NEXT: movd %eax, %xmm5531; SSSE3-NEXT: andl $15, %ecx532; SSSE3-NEXT: movzbl -24(%rsp,%rcx), %eax533; SSSE3-NEXT: movd %eax, %xmm9534; SSSE3-NEXT: andl $15, %edx535; SSSE3-NEXT: movzbl -24(%rsp,%rdx), %eax536; SSSE3-NEXT: movd %eax, %xmm10537; SSSE3-NEXT: andl $15, %esi538; SSSE3-NEXT: movzbl -24(%rsp,%rsi), %eax539; SSSE3-NEXT: movd %eax, %xmm11540; SSSE3-NEXT: andl $15, %edi541; SSSE3-NEXT: movzbl -24(%rsp,%rdi), %eax542; SSSE3-NEXT: movd %eax, %xmm0543; SSSE3-NEXT: andl $15, %r9d544; SSSE3-NEXT: movzbl -24(%rsp,%r9), %eax545; SSSE3-NEXT: movd %eax, %xmm12546; SSSE3-NEXT: andl $15, %r8d547; SSSE3-NEXT: movzbl -24(%rsp,%r8), %eax548; SSSE3-NEXT: movd %eax, %xmm13549; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax550; SSSE3-NEXT: andl $15, %eax551; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax552; SSSE3-NEXT: movd %eax, %xmm14553; SSSE3-NEXT: movzbl {{[0-9]+}}(%rsp), %eax554; SSSE3-NEXT: andl $15, %eax555; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax556; SSSE3-NEXT: movd %eax, %xmm15557; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]558; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]559; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]560; SSSE3-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]561; SSSE3-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]562; SSSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]563; SSSE3-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]564; SSSE3-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]565; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1],xmm0[2],xmm11[2],xmm0[3],xmm11[3],xmm0[4],xmm11[4],xmm0[5],xmm11[5],xmm0[6],xmm11[6],xmm0[7],xmm11[7]566; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm10[0],xmm0[1],xmm10[1],xmm0[2],xmm10[2],xmm0[3],xmm10[3]567; SSSE3-NEXT: punpcklbw {{.*#+}} xmm13 = xmm13[0],xmm12[0],xmm13[1],xmm12[1],xmm13[2],xmm12[2],xmm13[3],xmm12[3],xmm13[4],xmm12[4],xmm13[5],xmm12[5],xmm13[6],xmm12[6],xmm13[7],xmm12[7]568; SSSE3-NEXT: punpcklbw {{.*#+}} xmm15 = xmm15[0],xmm14[0],xmm15[1],xmm14[1],xmm15[2],xmm14[2],xmm15[3],xmm14[3],xmm15[4],xmm14[4],xmm15[5],xmm14[5],xmm15[6],xmm14[6],xmm15[7],xmm14[7]569; SSSE3-NEXT: punpcklwd {{.*#+}} xmm13 = xmm13[0],xmm15[0],xmm13[1],xmm15[1],xmm13[2],xmm15[2],xmm13[3],xmm15[3]570; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm13[0],xmm0[1],xmm13[1]571; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]572; SSSE3-NEXT: retq573;574; SSE41-LABEL: var_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:575; SSE41: # %bb.0:576; SSE41-NEXT: # kill: def $r9d killed $r9d def $r9577; SSE41-NEXT: # kill: def $r8d killed $r8d def $r8578; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx579; SSE41-NEXT: # kill: def $edx killed $edx def $rdx580; SSE41-NEXT: # kill: def $esi killed $esi def $rsi581; SSE41-NEXT: # kill: def $edi killed $edi def $rdi582; SSE41-NEXT: andl $15, %edi583; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)584; SSE41-NEXT: movzbl -24(%rsp,%rdi), %eax585; SSE41-NEXT: movd %eax, %xmm0586; SSE41-NEXT: andl $15, %esi587; SSE41-NEXT: pinsrb $1, -24(%rsp,%rsi), %xmm0588; SSE41-NEXT: andl $15, %edx589; SSE41-NEXT: pinsrb $2, -24(%rsp,%rdx), %xmm0590; SSE41-NEXT: andl $15, %ecx591; SSE41-NEXT: pinsrb $3, -24(%rsp,%rcx), %xmm0592; SSE41-NEXT: andl $15, %r8d593; SSE41-NEXT: pinsrb $4, -24(%rsp,%r8), %xmm0594; SSE41-NEXT: andl $15, %r9d595; SSE41-NEXT: pinsrb $5, -24(%rsp,%r9), %xmm0596; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax597; SSE41-NEXT: andl $15, %eax598; SSE41-NEXT: pinsrb $6, -24(%rsp,%rax), %xmm0599; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax600; SSE41-NEXT: andl $15, %eax601; SSE41-NEXT: pinsrb $7, -24(%rsp,%rax), %xmm0602; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax603; SSE41-NEXT: andl $15, %eax604; SSE41-NEXT: pinsrb $8, -24(%rsp,%rax), %xmm0605; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax606; SSE41-NEXT: andl $15, %eax607; SSE41-NEXT: pinsrb $9, -24(%rsp,%rax), %xmm0608; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax609; SSE41-NEXT: andl $15, %eax610; SSE41-NEXT: pinsrb $10, -24(%rsp,%rax), %xmm0611; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax612; SSE41-NEXT: andl $15, %eax613; SSE41-NEXT: pinsrb $11, -24(%rsp,%rax), %xmm0614; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax615; SSE41-NEXT: andl $15, %eax616; SSE41-NEXT: pinsrb $12, -24(%rsp,%rax), %xmm0617; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax618; SSE41-NEXT: andl $15, %eax619; SSE41-NEXT: pinsrb $13, -24(%rsp,%rax), %xmm0620; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax621; SSE41-NEXT: andl $15, %eax622; SSE41-NEXT: pinsrb $14, -24(%rsp,%rax), %xmm0623; SSE41-NEXT: movzbl {{[0-9]+}}(%rsp), %eax624; SSE41-NEXT: andl $15, %eax625; SSE41-NEXT: pinsrb $15, -24(%rsp,%rax), %xmm0626; SSE41-NEXT: retq627;628; AVX-LABEL: var_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:629; AVX: # %bb.0:630; AVX-NEXT: # kill: def $r9d killed $r9d def $r9631; AVX-NEXT: # kill: def $r8d killed $r8d def $r8632; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx633; AVX-NEXT: # kill: def $edx killed $edx def $rdx634; AVX-NEXT: # kill: def $esi killed $esi def $rsi635; AVX-NEXT: # kill: def $edi killed $edi def $rdi636; AVX-NEXT: andl $15, %edi637; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)638; AVX-NEXT: movzbl -24(%rsp,%rdi), %eax639; AVX-NEXT: vmovd %eax, %xmm0640; AVX-NEXT: andl $15, %esi641; AVX-NEXT: vpinsrb $1, -24(%rsp,%rsi), %xmm0, %xmm0642; AVX-NEXT: andl $15, %edx643; AVX-NEXT: vpinsrb $2, -24(%rsp,%rdx), %xmm0, %xmm0644; AVX-NEXT: andl $15, %ecx645; AVX-NEXT: vpinsrb $3, -24(%rsp,%rcx), %xmm0, %xmm0646; AVX-NEXT: andl $15, %r8d647; AVX-NEXT: vpinsrb $4, -24(%rsp,%r8), %xmm0, %xmm0648; AVX-NEXT: andl $15, %r9d649; AVX-NEXT: vpinsrb $5, -24(%rsp,%r9), %xmm0, %xmm0650; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax651; AVX-NEXT: andl $15, %eax652; AVX-NEXT: vpinsrb $6, -24(%rsp,%rax), %xmm0, %xmm0653; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax654; AVX-NEXT: andl $15, %eax655; AVX-NEXT: vpinsrb $7, -24(%rsp,%rax), %xmm0, %xmm0656; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax657; AVX-NEXT: andl $15, %eax658; AVX-NEXT: vpinsrb $8, -24(%rsp,%rax), %xmm0, %xmm0659; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax660; AVX-NEXT: andl $15, %eax661; AVX-NEXT: vpinsrb $9, -24(%rsp,%rax), %xmm0, %xmm0662; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax663; AVX-NEXT: andl $15, %eax664; AVX-NEXT: vpinsrb $10, -24(%rsp,%rax), %xmm0, %xmm0665; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax666; AVX-NEXT: andl $15, %eax667; AVX-NEXT: vpinsrb $11, -24(%rsp,%rax), %xmm0, %xmm0668; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax669; AVX-NEXT: andl $15, %eax670; AVX-NEXT: vpinsrb $12, -24(%rsp,%rax), %xmm0, %xmm0671; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax672; AVX-NEXT: andl $15, %eax673; AVX-NEXT: vpinsrb $13, -24(%rsp,%rax), %xmm0, %xmm0674; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax675; AVX-NEXT: andl $15, %eax676; AVX-NEXT: vpinsrb $14, -24(%rsp,%rax), %xmm0, %xmm0677; AVX-NEXT: movzbl {{[0-9]+}}(%rsp), %eax678; AVX-NEXT: andl $15, %eax679; AVX-NEXT: vpinsrb $15, -24(%rsp,%rax), %xmm0, %xmm0680; AVX-NEXT: retq681 %x0 = extractelement <16 x i8> %x, i8 %i0682 %x1 = extractelement <16 x i8> %x, i8 %i1683 %x2 = extractelement <16 x i8> %x, i8 %i2684 %x3 = extractelement <16 x i8> %x, i8 %i3685 %x4 = extractelement <16 x i8> %x, i8 %i4686 %x5 = extractelement <16 x i8> %x, i8 %i5687 %x6 = extractelement <16 x i8> %x, i8 %i6688 %x7 = extractelement <16 x i8> %x, i8 %i7689 %x8 = extractelement <16 x i8> %x, i8 %i8690 %x9 = extractelement <16 x i8> %x, i8 %i9691 %x10 = extractelement <16 x i8> %x, i8 %i10692 %x11 = extractelement <16 x i8> %x, i8 %i11693 %x12 = extractelement <16 x i8> %x, i8 %i12694 %x13 = extractelement <16 x i8> %x, i8 %i13695 %x14 = extractelement <16 x i8> %x, i8 %i14696 %x15 = extractelement <16 x i8> %x, i8 %i15697 %r0 = insertelement <16 x i8> undef, i8 %x0 , i32 0698 %r1 = insertelement <16 x i8> %r0 , i8 %x1 , i32 1699 %r2 = insertelement <16 x i8> %r1 , i8 %x2 , i32 2700 %r3 = insertelement <16 x i8> %r2 , i8 %x3 , i32 3701 %r4 = insertelement <16 x i8> %r3 , i8 %x4 , i32 4702 %r5 = insertelement <16 x i8> %r4 , i8 %x5 , i32 5703 %r6 = insertelement <16 x i8> %r5 , i8 %x6 , i32 6704 %r7 = insertelement <16 x i8> %r6 , i8 %x7 , i32 7705 %r8 = insertelement <16 x i8> %r7 , i8 %x8 , i32 8706 %r9 = insertelement <16 x i8> %r8 , i8 %x9 , i32 9707 %r10 = insertelement <16 x i8> %r9 , i8 %x10, i32 10708 %r11 = insertelement <16 x i8> %r10, i8 %x11, i32 11709 %r12 = insertelement <16 x i8> %r11, i8 %x12, i32 12710 %r13 = insertelement <16 x i8> %r12, i8 %x13, i32 13711 %r14 = insertelement <16 x i8> %r13, i8 %x14, i32 14712 %r15 = insertelement <16 x i8> %r14, i8 %x15, i32 15713 ret <16 x i8> %r15714}715 716;717; Unary shuffle indices from memory718;719 720define <4 x i32> @mem_shuffle_v4i32_v4i32_xxxx_i32(<4 x i32> %x, ptr %i) nounwind {721; SSE2-LABEL: mem_shuffle_v4i32_v4i32_xxxx_i32:722; SSE2: # %bb.0:723; SSE2-NEXT: movl (%rdi), %eax724; SSE2-NEXT: movl 4(%rdi), %ecx725; SSE2-NEXT: andl $3, %eax726; SSE2-NEXT: andl $3, %ecx727; SSE2-NEXT: movl 8(%rdi), %edx728; SSE2-NEXT: andl $3, %edx729; SSE2-NEXT: movl 12(%rdi), %esi730; SSE2-NEXT: andl $3, %esi731; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)732; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero733; SSE2-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero734; SSE2-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]735; SSE2-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero736; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero737; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]738; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]739; SSE2-NEXT: retq740;741; SSSE3-LABEL: mem_shuffle_v4i32_v4i32_xxxx_i32:742; SSSE3: # %bb.0:743; SSSE3-NEXT: movl (%rdi), %eax744; SSSE3-NEXT: movl 4(%rdi), %ecx745; SSSE3-NEXT: andl $3, %eax746; SSSE3-NEXT: andl $3, %ecx747; SSSE3-NEXT: movl 8(%rdi), %edx748; SSSE3-NEXT: andl $3, %edx749; SSSE3-NEXT: movl 12(%rdi), %esi750; SSSE3-NEXT: andl $3, %esi751; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)752; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero753; SSSE3-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero754; SSSE3-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]755; SSSE3-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero756; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero757; SSSE3-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]758; SSSE3-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]759; SSSE3-NEXT: retq760;761; SSE41-LABEL: mem_shuffle_v4i32_v4i32_xxxx_i32:762; SSE41: # %bb.0:763; SSE41-NEXT: movl (%rdi), %eax764; SSE41-NEXT: movl 4(%rdi), %ecx765; SSE41-NEXT: andl $3, %eax766; SSE41-NEXT: andl $3, %ecx767; SSE41-NEXT: movl 8(%rdi), %edx768; SSE41-NEXT: andl $3, %edx769; SSE41-NEXT: movl 12(%rdi), %esi770; SSE41-NEXT: andl $3, %esi771; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)772; SSE41-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero773; SSE41-NEXT: pinsrd $1, -24(%rsp,%rcx,4), %xmm0774; SSE41-NEXT: pinsrd $2, -24(%rsp,%rdx,4), %xmm0775; SSE41-NEXT: pinsrd $3, -24(%rsp,%rsi,4), %xmm0776; SSE41-NEXT: retq777;778; AVX-LABEL: mem_shuffle_v4i32_v4i32_xxxx_i32:779; AVX: # %bb.0:780; AVX-NEXT: movl (%rdi), %eax781; AVX-NEXT: movl 4(%rdi), %ecx782; AVX-NEXT: andl $3, %eax783; AVX-NEXT: andl $3, %ecx784; AVX-NEXT: movl 8(%rdi), %edx785; AVX-NEXT: andl $3, %edx786; AVX-NEXT: movl 12(%rdi), %esi787; AVX-NEXT: andl $3, %esi788; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)789; AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero790; AVX-NEXT: vpinsrd $1, -24(%rsp,%rcx,4), %xmm0, %xmm0791; AVX-NEXT: vpinsrd $2, -24(%rsp,%rdx,4), %xmm0, %xmm0792; AVX-NEXT: vpinsrd $3, -24(%rsp,%rsi,4), %xmm0, %xmm0793; AVX-NEXT: retq794 %p1 = getelementptr inbounds i32, ptr %i, i64 1795 %p2 = getelementptr inbounds i32, ptr %i, i64 2796 %p3 = getelementptr inbounds i32, ptr %i, i64 3797 %i0 = load i32, ptr %i, align 4798 %i1 = load i32, ptr %p1, align 4799 %i2 = load i32, ptr %p2, align 4800 %i3 = load i32, ptr %p3, align 4801 %x0 = extractelement <4 x i32> %x, i32 %i0802 %x1 = extractelement <4 x i32> %x, i32 %i1803 %x2 = extractelement <4 x i32> %x, i32 %i2804 %x3 = extractelement <4 x i32> %x, i32 %i3805 %r0 = insertelement <4 x i32> undef, i32 %x0, i32 0806 %r1 = insertelement <4 x i32> %r0, i32 %x1, i32 1807 %r2 = insertelement <4 x i32> %r1, i32 %x2, i32 2808 %r3 = insertelement <4 x i32> %r2, i32 %x3, i32 3809 ret <4 x i32> %r3810}811 812define <16 x i8> @mem_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8(<16 x i8> %x, ptr %i) nounwind {813; SSE2-LABEL: mem_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:814; SSE2: # %bb.0:815; SSE2-NEXT: pushq %rbp816; SSE2-NEXT: pushq %r15817; SSE2-NEXT: pushq %r14818; SSE2-NEXT: pushq %r13819; SSE2-NEXT: pushq %r12820; SSE2-NEXT: pushq %rbx821; SSE2-NEXT: movzbl (%rdi), %eax822; SSE2-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill823; SSE2-NEXT: movzbl 1(%rdi), %ecx824; SSE2-NEXT: movzbl 2(%rdi), %edx825; SSE2-NEXT: movzbl 3(%rdi), %esi826; SSE2-NEXT: movzbl 4(%rdi), %r8d827; SSE2-NEXT: movzbl 5(%rdi), %r9d828; SSE2-NEXT: movzbl 6(%rdi), %r10d829; SSE2-NEXT: movzbl 7(%rdi), %r11d830; SSE2-NEXT: movzbl 8(%rdi), %ebx831; SSE2-NEXT: movzbl 9(%rdi), %r14d832; SSE2-NEXT: movzbl 10(%rdi), %r15d833; SSE2-NEXT: movzbl 11(%rdi), %r12d834; SSE2-NEXT: movzbl 12(%rdi), %r13d835; SSE2-NEXT: movzbl 13(%rdi), %ebp836; SSE2-NEXT: movzbl 14(%rdi), %eax837; SSE2-NEXT: movzbl 15(%rdi), %edi838; SSE2-NEXT: andl $15, %edi839; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)840; SSE2-NEXT: movzbl -24(%rsp,%rdi), %edi841; SSE2-NEXT: movd %edi, %xmm1842; SSE2-NEXT: andl $15, %eax843; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax844; SSE2-NEXT: movd %eax, %xmm2845; SSE2-NEXT: andl $15, %ebp846; SSE2-NEXT: movzbl -24(%rsp,%rbp), %eax847; SSE2-NEXT: movd %eax, %xmm4848; SSE2-NEXT: andl $15, %r13d849; SSE2-NEXT: movzbl -24(%rsp,%r13), %eax850; SSE2-NEXT: movd %eax, %xmm3851; SSE2-NEXT: andl $15, %r12d852; SSE2-NEXT: movzbl -24(%rsp,%r12), %eax853; SSE2-NEXT: movd %eax, %xmm6854; SSE2-NEXT: andl $15, %r15d855; SSE2-NEXT: movzbl -24(%rsp,%r15), %eax856; SSE2-NEXT: movd %eax, %xmm7857; SSE2-NEXT: andl $15, %r14d858; SSE2-NEXT: movzbl -24(%rsp,%r14), %eax859; SSE2-NEXT: movd %eax, %xmm8860; SSE2-NEXT: andl $15, %ebx861; SSE2-NEXT: movzbl -24(%rsp,%rbx), %eax862; SSE2-NEXT: movd %eax, %xmm5863; SSE2-NEXT: andl $15, %r11d864; SSE2-NEXT: movzbl -24(%rsp,%r11), %eax865; SSE2-NEXT: movd %eax, %xmm9866; SSE2-NEXT: andl $15, %r10d867; SSE2-NEXT: movzbl -24(%rsp,%r10), %eax868; SSE2-NEXT: movd %eax, %xmm10869; SSE2-NEXT: andl $15, %r9d870; SSE2-NEXT: movzbl -24(%rsp,%r9), %eax871; SSE2-NEXT: movd %eax, %xmm12872; SSE2-NEXT: andl $15, %r8d873; SSE2-NEXT: movzbl -24(%rsp,%r8), %eax874; SSE2-NEXT: movd %eax, %xmm11875; SSE2-NEXT: andl $15, %esi876; SSE2-NEXT: movzbl -24(%rsp,%rsi), %eax877; SSE2-NEXT: movd %eax, %xmm13878; SSE2-NEXT: andl $15, %edx879; SSE2-NEXT: movzbl -24(%rsp,%rdx), %eax880; SSE2-NEXT: movd %eax, %xmm14881; SSE2-NEXT: andl $15, %ecx882; SSE2-NEXT: movzbl -24(%rsp,%rcx), %eax883; SSE2-NEXT: movd %eax, %xmm15884; SSE2-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload885; SSE2-NEXT: andl $15, %eax886; SSE2-NEXT: movzbl -24(%rsp,%rax), %eax887; SSE2-NEXT: movd %eax, %xmm0888; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]889; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]890; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]891; SSE2-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]892; SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]893; SSE2-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]894; SSE2-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]895; SSE2-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]896; SSE2-NEXT: punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]897; SSE2-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]898; SSE2-NEXT: punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3],xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]899; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3],xmm0[4],xmm15[4],xmm0[5],xmm15[5],xmm0[6],xmm15[6],xmm0[7],xmm15[7]900; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]901; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]902; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]903; SSE2-NEXT: popq %rbx904; SSE2-NEXT: popq %r12905; SSE2-NEXT: popq %r13906; SSE2-NEXT: popq %r14907; SSE2-NEXT: popq %r15908; SSE2-NEXT: popq %rbp909; SSE2-NEXT: retq910;911; SSSE3-LABEL: mem_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:912; SSSE3: # %bb.0:913; SSSE3-NEXT: pushq %rbp914; SSSE3-NEXT: pushq %r15915; SSSE3-NEXT: pushq %r14916; SSSE3-NEXT: pushq %r13917; SSSE3-NEXT: pushq %r12918; SSSE3-NEXT: pushq %rbx919; SSSE3-NEXT: movzbl (%rdi), %eax920; SSSE3-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill921; SSSE3-NEXT: movzbl 1(%rdi), %ecx922; SSSE3-NEXT: movzbl 2(%rdi), %edx923; SSSE3-NEXT: movzbl 3(%rdi), %esi924; SSSE3-NEXT: movzbl 4(%rdi), %r8d925; SSSE3-NEXT: movzbl 5(%rdi), %r9d926; SSSE3-NEXT: movzbl 6(%rdi), %r10d927; SSSE3-NEXT: movzbl 7(%rdi), %r11d928; SSSE3-NEXT: movzbl 8(%rdi), %ebx929; SSSE3-NEXT: movzbl 9(%rdi), %r14d930; SSSE3-NEXT: movzbl 10(%rdi), %r15d931; SSSE3-NEXT: movzbl 11(%rdi), %r12d932; SSSE3-NEXT: movzbl 12(%rdi), %r13d933; SSSE3-NEXT: movzbl 13(%rdi), %ebp934; SSSE3-NEXT: movzbl 14(%rdi), %eax935; SSSE3-NEXT: movzbl 15(%rdi), %edi936; SSSE3-NEXT: andl $15, %edi937; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)938; SSSE3-NEXT: movzbl -24(%rsp,%rdi), %edi939; SSSE3-NEXT: movd %edi, %xmm1940; SSSE3-NEXT: andl $15, %eax941; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax942; SSSE3-NEXT: movd %eax, %xmm2943; SSSE3-NEXT: andl $15, %ebp944; SSSE3-NEXT: movzbl -24(%rsp,%rbp), %eax945; SSSE3-NEXT: movd %eax, %xmm4946; SSSE3-NEXT: andl $15, %r13d947; SSSE3-NEXT: movzbl -24(%rsp,%r13), %eax948; SSSE3-NEXT: movd %eax, %xmm3949; SSSE3-NEXT: andl $15, %r12d950; SSSE3-NEXT: movzbl -24(%rsp,%r12), %eax951; SSSE3-NEXT: movd %eax, %xmm6952; SSSE3-NEXT: andl $15, %r15d953; SSSE3-NEXT: movzbl -24(%rsp,%r15), %eax954; SSSE3-NEXT: movd %eax, %xmm7955; SSSE3-NEXT: andl $15, %r14d956; SSSE3-NEXT: movzbl -24(%rsp,%r14), %eax957; SSSE3-NEXT: movd %eax, %xmm8958; SSSE3-NEXT: andl $15, %ebx959; SSSE3-NEXT: movzbl -24(%rsp,%rbx), %eax960; SSSE3-NEXT: movd %eax, %xmm5961; SSSE3-NEXT: andl $15, %r11d962; SSSE3-NEXT: movzbl -24(%rsp,%r11), %eax963; SSSE3-NEXT: movd %eax, %xmm9964; SSSE3-NEXT: andl $15, %r10d965; SSSE3-NEXT: movzbl -24(%rsp,%r10), %eax966; SSSE3-NEXT: movd %eax, %xmm10967; SSSE3-NEXT: andl $15, %r9d968; SSSE3-NEXT: movzbl -24(%rsp,%r9), %eax969; SSSE3-NEXT: movd %eax, %xmm12970; SSSE3-NEXT: andl $15, %r8d971; SSSE3-NEXT: movzbl -24(%rsp,%r8), %eax972; SSSE3-NEXT: movd %eax, %xmm11973; SSSE3-NEXT: andl $15, %esi974; SSSE3-NEXT: movzbl -24(%rsp,%rsi), %eax975; SSSE3-NEXT: movd %eax, %xmm13976; SSSE3-NEXT: andl $15, %edx977; SSSE3-NEXT: movzbl -24(%rsp,%rdx), %eax978; SSSE3-NEXT: movd %eax, %xmm14979; SSSE3-NEXT: andl $15, %ecx980; SSSE3-NEXT: movzbl -24(%rsp,%rcx), %eax981; SSSE3-NEXT: movd %eax, %xmm15982; SSSE3-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload983; SSSE3-NEXT: andl $15, %eax984; SSSE3-NEXT: movzbl -24(%rsp,%rax), %eax985; SSSE3-NEXT: movd %eax, %xmm0986; SSSE3-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]987; SSSE3-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3],xmm3[4],xmm4[4],xmm3[5],xmm4[5],xmm3[6],xmm4[6],xmm3[7],xmm4[7]988; SSSE3-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3]989; SSSE3-NEXT: punpcklbw {{.*#+}} xmm7 = xmm7[0],xmm6[0],xmm7[1],xmm6[1],xmm7[2],xmm6[2],xmm7[3],xmm6[3],xmm7[4],xmm6[4],xmm7[5],xmm6[5],xmm7[6],xmm6[6],xmm7[7],xmm6[7]990; SSSE3-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm8[0],xmm5[1],xmm8[1],xmm5[2],xmm8[2],xmm5[3],xmm8[3],xmm5[4],xmm8[4],xmm5[5],xmm8[5],xmm5[6],xmm8[6],xmm5[7],xmm8[7]991; SSSE3-NEXT: punpcklwd {{.*#+}} xmm5 = xmm5[0],xmm7[0],xmm5[1],xmm7[1],xmm5[2],xmm7[2],xmm5[3],xmm7[3]992; SSSE3-NEXT: punpckldq {{.*#+}} xmm5 = xmm5[0],xmm3[0],xmm5[1],xmm3[1]993; SSSE3-NEXT: punpcklbw {{.*#+}} xmm10 = xmm10[0],xmm9[0],xmm10[1],xmm9[1],xmm10[2],xmm9[2],xmm10[3],xmm9[3],xmm10[4],xmm9[4],xmm10[5],xmm9[5],xmm10[6],xmm9[6],xmm10[7],xmm9[7]994; SSSE3-NEXT: punpcklbw {{.*#+}} xmm11 = xmm11[0],xmm12[0],xmm11[1],xmm12[1],xmm11[2],xmm12[2],xmm11[3],xmm12[3],xmm11[4],xmm12[4],xmm11[5],xmm12[5],xmm11[6],xmm12[6],xmm11[7],xmm12[7]995; SSSE3-NEXT: punpcklwd {{.*#+}} xmm11 = xmm11[0],xmm10[0],xmm11[1],xmm10[1],xmm11[2],xmm10[2],xmm11[3],xmm10[3]996; SSSE3-NEXT: punpcklbw {{.*#+}} xmm14 = xmm14[0],xmm13[0],xmm14[1],xmm13[1],xmm14[2],xmm13[2],xmm14[3],xmm13[3],xmm14[4],xmm13[4],xmm14[5],xmm13[5],xmm14[6],xmm13[6],xmm14[7],xmm13[7]997; SSSE3-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm15[0],xmm0[1],xmm15[1],xmm0[2],xmm15[2],xmm0[3],xmm15[3],xmm0[4],xmm15[4],xmm0[5],xmm15[5],xmm0[6],xmm15[6],xmm0[7],xmm15[7]998; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm14[0],xmm0[1],xmm14[1],xmm0[2],xmm14[2],xmm0[3],xmm14[3]999; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm11[0],xmm0[1],xmm11[1]1000; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm5[0]1001; SSSE3-NEXT: popq %rbx1002; SSSE3-NEXT: popq %r121003; SSSE3-NEXT: popq %r131004; SSSE3-NEXT: popq %r141005; SSSE3-NEXT: popq %r151006; SSSE3-NEXT: popq %rbp1007; SSSE3-NEXT: retq1008;1009; SSE41-LABEL: mem_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:1010; SSE41: # %bb.0:1011; SSE41-NEXT: pushq %rbp1012; SSE41-NEXT: pushq %r151013; SSE41-NEXT: pushq %r141014; SSE41-NEXT: pushq %r131015; SSE41-NEXT: pushq %r121016; SSE41-NEXT: pushq %rbx1017; SSE41-NEXT: movzbl (%rdi), %ecx1018; SSE41-NEXT: andl $15, %ecx1019; SSE41-NEXT: movzbl 1(%rdi), %eax1020; SSE41-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill1021; SSE41-NEXT: movzbl 2(%rdi), %edx1022; SSE41-NEXT: movzbl 3(%rdi), %esi1023; SSE41-NEXT: movzbl 4(%rdi), %r8d1024; SSE41-NEXT: movzbl 5(%rdi), %r9d1025; SSE41-NEXT: movzbl 6(%rdi), %r10d1026; SSE41-NEXT: movzbl 7(%rdi), %r11d1027; SSE41-NEXT: movzbl 8(%rdi), %ebx1028; SSE41-NEXT: movzbl 9(%rdi), %r14d1029; SSE41-NEXT: movzbl 10(%rdi), %r15d1030; SSE41-NEXT: movzbl 11(%rdi), %r12d1031; SSE41-NEXT: movzbl 12(%rdi), %r13d1032; SSE41-NEXT: movzbl 13(%rdi), %ebp1033; SSE41-NEXT: movzbl 14(%rdi), %eax1034; SSE41-NEXT: movzbl 15(%rdi), %edi1035; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1036; SSE41-NEXT: movzbl -24(%rsp,%rcx), %ecx1037; SSE41-NEXT: movd %ecx, %xmm01038; SSE41-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload1039; SSE41-NEXT: andl $15, %ecx1040; SSE41-NEXT: pinsrb $1, -24(%rsp,%rcx), %xmm01041; SSE41-NEXT: andl $15, %edx1042; SSE41-NEXT: pinsrb $2, -24(%rsp,%rdx), %xmm01043; SSE41-NEXT: andl $15, %esi1044; SSE41-NEXT: pinsrb $3, -24(%rsp,%rsi), %xmm01045; SSE41-NEXT: andl $15, %r8d1046; SSE41-NEXT: pinsrb $4, -24(%rsp,%r8), %xmm01047; SSE41-NEXT: andl $15, %r9d1048; SSE41-NEXT: pinsrb $5, -24(%rsp,%r9), %xmm01049; SSE41-NEXT: andl $15, %r10d1050; SSE41-NEXT: pinsrb $6, -24(%rsp,%r10), %xmm01051; SSE41-NEXT: andl $15, %r11d1052; SSE41-NEXT: pinsrb $7, -24(%rsp,%r11), %xmm01053; SSE41-NEXT: andl $15, %ebx1054; SSE41-NEXT: pinsrb $8, -24(%rsp,%rbx), %xmm01055; SSE41-NEXT: andl $15, %r14d1056; SSE41-NEXT: pinsrb $9, -24(%rsp,%r14), %xmm01057; SSE41-NEXT: andl $15, %r15d1058; SSE41-NEXT: pinsrb $10, -24(%rsp,%r15), %xmm01059; SSE41-NEXT: andl $15, %r12d1060; SSE41-NEXT: pinsrb $11, -24(%rsp,%r12), %xmm01061; SSE41-NEXT: andl $15, %r13d1062; SSE41-NEXT: pinsrb $12, -24(%rsp,%r13), %xmm01063; SSE41-NEXT: andl $15, %ebp1064; SSE41-NEXT: pinsrb $13, -24(%rsp,%rbp), %xmm01065; SSE41-NEXT: andl $15, %eax1066; SSE41-NEXT: pinsrb $14, -24(%rsp,%rax), %xmm01067; SSE41-NEXT: andl $15, %edi1068; SSE41-NEXT: pinsrb $15, -24(%rsp,%rdi), %xmm01069; SSE41-NEXT: popq %rbx1070; SSE41-NEXT: popq %r121071; SSE41-NEXT: popq %r131072; SSE41-NEXT: popq %r141073; SSE41-NEXT: popq %r151074; SSE41-NEXT: popq %rbp1075; SSE41-NEXT: retq1076;1077; AVX-LABEL: mem_shuffle_v16i8_v16i8_xxxxxxxxxxxxxxxx_i8:1078; AVX: # %bb.0:1079; AVX-NEXT: pushq %rbp1080; AVX-NEXT: pushq %r151081; AVX-NEXT: pushq %r141082; AVX-NEXT: pushq %r131083; AVX-NEXT: pushq %r121084; AVX-NEXT: pushq %rbx1085; AVX-NEXT: movzbl (%rdi), %ecx1086; AVX-NEXT: andl $15, %ecx1087; AVX-NEXT: movzbl 1(%rdi), %eax1088; AVX-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill1089; AVX-NEXT: movzbl 2(%rdi), %edx1090; AVX-NEXT: movzbl 3(%rdi), %esi1091; AVX-NEXT: movzbl 4(%rdi), %r8d1092; AVX-NEXT: movzbl 5(%rdi), %r9d1093; AVX-NEXT: movzbl 6(%rdi), %r10d1094; AVX-NEXT: movzbl 7(%rdi), %r11d1095; AVX-NEXT: movzbl 8(%rdi), %ebx1096; AVX-NEXT: movzbl 9(%rdi), %r14d1097; AVX-NEXT: movzbl 10(%rdi), %r15d1098; AVX-NEXT: movzbl 11(%rdi), %r12d1099; AVX-NEXT: movzbl 12(%rdi), %r13d1100; AVX-NEXT: movzbl 13(%rdi), %ebp1101; AVX-NEXT: movzbl 14(%rdi), %eax1102; AVX-NEXT: movzbl 15(%rdi), %edi1103; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)1104; AVX-NEXT: movzbl -24(%rsp,%rcx), %ecx1105; AVX-NEXT: vmovd %ecx, %xmm01106; AVX-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rcx # 8-byte Reload1107; AVX-NEXT: andl $15, %ecx1108; AVX-NEXT: vpinsrb $1, -24(%rsp,%rcx), %xmm0, %xmm01109; AVX-NEXT: andl $15, %edx1110; AVX-NEXT: vpinsrb $2, -24(%rsp,%rdx), %xmm0, %xmm01111; AVX-NEXT: andl $15, %esi1112; AVX-NEXT: vpinsrb $3, -24(%rsp,%rsi), %xmm0, %xmm01113; AVX-NEXT: andl $15, %r8d1114; AVX-NEXT: vpinsrb $4, -24(%rsp,%r8), %xmm0, %xmm01115; AVX-NEXT: andl $15, %r9d1116; AVX-NEXT: vpinsrb $5, -24(%rsp,%r9), %xmm0, %xmm01117; AVX-NEXT: andl $15, %r10d1118; AVX-NEXT: vpinsrb $6, -24(%rsp,%r10), %xmm0, %xmm01119; AVX-NEXT: andl $15, %r11d1120; AVX-NEXT: vpinsrb $7, -24(%rsp,%r11), %xmm0, %xmm01121; AVX-NEXT: andl $15, %ebx1122; AVX-NEXT: vpinsrb $8, -24(%rsp,%rbx), %xmm0, %xmm01123; AVX-NEXT: andl $15, %r14d1124; AVX-NEXT: vpinsrb $9, -24(%rsp,%r14), %xmm0, %xmm01125; AVX-NEXT: andl $15, %r15d1126; AVX-NEXT: vpinsrb $10, -24(%rsp,%r15), %xmm0, %xmm01127; AVX-NEXT: andl $15, %r12d1128; AVX-NEXT: vpinsrb $11, -24(%rsp,%r12), %xmm0, %xmm01129; AVX-NEXT: andl $15, %r13d1130; AVX-NEXT: vpinsrb $12, -24(%rsp,%r13), %xmm0, %xmm01131; AVX-NEXT: andl $15, %ebp1132; AVX-NEXT: vpinsrb $13, -24(%rsp,%rbp), %xmm0, %xmm01133; AVX-NEXT: andl $15, %eax1134; AVX-NEXT: vpinsrb $14, -24(%rsp,%rax), %xmm0, %xmm01135; AVX-NEXT: andl $15, %edi1136; AVX-NEXT: vpinsrb $15, -24(%rsp,%rdi), %xmm0, %xmm01137; AVX-NEXT: popq %rbx1138; AVX-NEXT: popq %r121139; AVX-NEXT: popq %r131140; AVX-NEXT: popq %r141141; AVX-NEXT: popq %r151142; AVX-NEXT: popq %rbp1143; AVX-NEXT: retq1144 %p1 = getelementptr inbounds i8, ptr %i, i64 11145 %p2 = getelementptr inbounds i8, ptr %i, i64 21146 %p3 = getelementptr inbounds i8, ptr %i, i64 31147 %p4 = getelementptr inbounds i8, ptr %i, i64 41148 %p5 = getelementptr inbounds i8, ptr %i, i64 51149 %p6 = getelementptr inbounds i8, ptr %i, i64 61150 %p7 = getelementptr inbounds i8, ptr %i, i64 71151 %p8 = getelementptr inbounds i8, ptr %i, i64 81152 %p9 = getelementptr inbounds i8, ptr %i, i64 91153 %p10 = getelementptr inbounds i8, ptr %i, i64 101154 %p11 = getelementptr inbounds i8, ptr %i, i64 111155 %p12 = getelementptr inbounds i8, ptr %i, i64 121156 %p13 = getelementptr inbounds i8, ptr %i, i64 131157 %p14 = getelementptr inbounds i8, ptr %i, i64 141158 %p15 = getelementptr inbounds i8, ptr %i, i64 151159 %i0 = load i8, ptr %i , align 41160 %i1 = load i8, ptr %p1 , align 41161 %i2 = load i8, ptr %p2 , align 41162 %i3 = load i8, ptr %p3 , align 41163 %i4 = load i8, ptr %p4 , align 41164 %i5 = load i8, ptr %p5 , align 41165 %i6 = load i8, ptr %p6 , align 41166 %i7 = load i8, ptr %p7 , align 41167 %i8 = load i8, ptr %p8 , align 41168 %i9 = load i8, ptr %p9 , align 41169 %i10 = load i8, ptr %p10, align 41170 %i11 = load i8, ptr %p11, align 41171 %i12 = load i8, ptr %p12, align 41172 %i13 = load i8, ptr %p13, align 41173 %i14 = load i8, ptr %p14, align 41174 %i15 = load i8, ptr %p15, align 41175 %x0 = extractelement <16 x i8> %x, i8 %i01176 %x1 = extractelement <16 x i8> %x, i8 %i11177 %x2 = extractelement <16 x i8> %x, i8 %i21178 %x3 = extractelement <16 x i8> %x, i8 %i31179 %x4 = extractelement <16 x i8> %x, i8 %i41180 %x5 = extractelement <16 x i8> %x, i8 %i51181 %x6 = extractelement <16 x i8> %x, i8 %i61182 %x7 = extractelement <16 x i8> %x, i8 %i71183 %x8 = extractelement <16 x i8> %x, i8 %i81184 %x9 = extractelement <16 x i8> %x, i8 %i91185 %x10 = extractelement <16 x i8> %x, i8 %i101186 %x11 = extractelement <16 x i8> %x, i8 %i111187 %x12 = extractelement <16 x i8> %x, i8 %i121188 %x13 = extractelement <16 x i8> %x, i8 %i131189 %x14 = extractelement <16 x i8> %x, i8 %i141190 %x15 = extractelement <16 x i8> %x, i8 %i151191 %r0 = insertelement <16 x i8> undef, i8 %x0 , i32 01192 %r1 = insertelement <16 x i8> %r0 , i8 %x1 , i32 11193 %r2 = insertelement <16 x i8> %r1 , i8 %x2 , i32 21194 %r3 = insertelement <16 x i8> %r2 , i8 %x3 , i32 31195 %r4 = insertelement <16 x i8> %r3 , i8 %x4 , i32 41196 %r5 = insertelement <16 x i8> %r4 , i8 %x5 , i32 51197 %r6 = insertelement <16 x i8> %r5 , i8 %x6 , i32 61198 %r7 = insertelement <16 x i8> %r6 , i8 %x7 , i32 71199 %r8 = insertelement <16 x i8> %r7 , i8 %x8 , i32 81200 %r9 = insertelement <16 x i8> %r8 , i8 %x9 , i32 91201 %r10 = insertelement <16 x i8> %r9 , i8 %x10, i32 101202 %r11 = insertelement <16 x i8> %r10, i8 %x11, i32 111203 %r12 = insertelement <16 x i8> %r11, i8 %x12, i32 121204 %r13 = insertelement <16 x i8> %r12, i8 %x13, i32 131205 %r14 = insertelement <16 x i8> %r13, i8 %x14, i32 141206 %r15 = insertelement <16 x i8> %r14, i8 %x15, i32 151207 ret <16 x i8> %r151208}1209 1210;1211; Binary shuffle indices from registers1212;1213 1214define <4 x float> @var_shuffle_v4f32_v4f32_x0yx_i32(<4 x float> %x, <4 x float> %y, i32 %i0, i32 %i1, i32 %i2, i32 %i3) nounwind {1215; SSE2-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:1216; SSE2: # %bb.0:1217; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx1218; SSE2-NEXT: # kill: def $edx killed $edx def $rdx1219; SSE2-NEXT: # kill: def $edi killed $edi def $rdi1220; SSE2-NEXT: andl $3, %edi1221; SSE2-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1222; SSE2-NEXT: andl $3, %edx1223; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1224; SSE2-NEXT: andl $3, %ecx1225; SSE2-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero1226; SSE2-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero1227; SSE2-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero1228; SSE2-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]1229; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]1230; SSE2-NEXT: retq1231;1232; SSSE3-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:1233; SSSE3: # %bb.0:1234; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx1235; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx1236; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi1237; SSSE3-NEXT: andl $3, %edi1238; SSSE3-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1239; SSSE3-NEXT: andl $3, %edx1240; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1241; SSSE3-NEXT: andl $3, %ecx1242; SSSE3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero1243; SSSE3-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero1244; SSSE3-NEXT: movss {{.*#+}} xmm2 = mem[0],zero,zero,zero1245; SSSE3-NEXT: unpcklps {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]1246; SSSE3-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]1247; SSSE3-NEXT: retq1248;1249; SSE41-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:1250; SSE41: # %bb.0:1251; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx1252; SSE41-NEXT: # kill: def $edx killed $edx def $rdx1253; SSE41-NEXT: # kill: def $edi killed $edi def $rdi1254; SSE41-NEXT: andl $3, %edi1255; SSE41-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1256; SSE41-NEXT: andl $3, %edx1257; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1258; SSE41-NEXT: andl $3, %ecx1259; SSE41-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero1260; SSE41-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero1261; SSE41-NEXT: insertps {{.*#+}} xmm1 = xmm1[0],mem[0],zero,zero1262; SSE41-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]1263; SSE41-NEXT: retq1264;1265; AVX-LABEL: var_shuffle_v4f32_v4f32_x0yx_i32:1266; AVX: # %bb.0:1267; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx1268; AVX-NEXT: # kill: def $edx killed $edx def $rdx1269; AVX-NEXT: # kill: def $edi killed $edi def $rdi1270; AVX-NEXT: andl $3, %edi1271; AVX-NEXT: vmovaps %xmm1, -{{[0-9]+}}(%rsp)1272; AVX-NEXT: andl $3, %edx1273; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)1274; AVX-NEXT: andl $3, %ecx1275; AVX-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero1276; AVX-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero1277; AVX-NEXT: vinsertps {{.*#+}} xmm1 = xmm1[0],mem[0],zero,zero1278; AVX-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]1279; AVX-NEXT: retq1280 %x0 = extractelement <4 x float> %x, i32 %i01281 %x1 = extractelement <4 x float> %x, i32 %i11282 %y2 = extractelement <4 x float> %y, i32 %i21283 %x3 = extractelement <4 x float> %x, i32 %i31284 %r0 = insertelement <4 x float> undef, float %x0, i32 01285 %r1 = insertelement <4 x float> %r0, float 0.0, i32 11286 %r2 = insertelement <4 x float> %r1, float %y2, i32 21287 %r3 = insertelement <4 x float> %r2, float %x3, i32 31288 ret <4 x float> %r31289}1290 1291define <8 x i16> @var_shuffle_v8i16_v8i16_xyxyxy00_i16(<8 x i16> %x, <8 x i16> %y, i16 %i0, i16 %i1, i16 %i2, i16 %i3, i16 %i4, i16 %i5, i16 %i6, i16 %i7) nounwind {1292; SSE2-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:1293; SSE2: # %bb.0:1294; SSE2-NEXT: # kill: def $r9d killed $r9d def $r91295; SSE2-NEXT: # kill: def $r8d killed $r8d def $r81296; SSE2-NEXT: # kill: def $ecx killed $ecx def $rcx1297; SSE2-NEXT: # kill: def $edx killed $edx def $rdx1298; SSE2-NEXT: # kill: def $esi killed $esi def $rsi1299; SSE2-NEXT: # kill: def $edi killed $edi def $rdi1300; SSE2-NEXT: andl $7, %edi1301; SSE2-NEXT: andl $7, %esi1302; SSE2-NEXT: andl $7, %edx1303; SSE2-NEXT: andl $7, %ecx1304; SSE2-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1305; SSE2-NEXT: andl $7, %r8d1306; SSE2-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1307; SSE2-NEXT: andl $7, %r9d1308; SSE2-NEXT: movzwl -24(%rsp,%rcx,2), %eax1309; SSE2-NEXT: movd %eax, %xmm01310; SSE2-NEXT: movzwl -40(%rsp,%rdx,2), %eax1311; SSE2-NEXT: movd %eax, %xmm11312; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]1313; SSE2-NEXT: movzwl -24(%rsp,%rsi,2), %eax1314; SSE2-NEXT: movd %eax, %xmm21315; SSE2-NEXT: movzwl -40(%rsp,%rdi,2), %eax1316; SSE2-NEXT: movd %eax, %xmm01317; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]1318; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]1319; SSE2-NEXT: movzwl -24(%rsp,%r9,2), %eax1320; SSE2-NEXT: movd %eax, %xmm11321; SSE2-NEXT: movzwl -40(%rsp,%r8,2), %eax1322; SSE2-NEXT: movd %eax, %xmm21323; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]1324; SSE2-NEXT: pxor %xmm1, %xmm11325; SSE2-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]1326; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]1327; SSE2-NEXT: retq1328;1329; SSSE3-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:1330; SSSE3: # %bb.0:1331; SSSE3-NEXT: # kill: def $r9d killed $r9d def $r91332; SSSE3-NEXT: # kill: def $r8d killed $r8d def $r81333; SSSE3-NEXT: # kill: def $ecx killed $ecx def $rcx1334; SSSE3-NEXT: # kill: def $edx killed $edx def $rdx1335; SSSE3-NEXT: # kill: def $esi killed $esi def $rsi1336; SSSE3-NEXT: # kill: def $edi killed $edi def $rdi1337; SSSE3-NEXT: andl $7, %edi1338; SSSE3-NEXT: andl $7, %esi1339; SSSE3-NEXT: andl $7, %edx1340; SSSE3-NEXT: andl $7, %ecx1341; SSSE3-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1342; SSSE3-NEXT: andl $7, %r8d1343; SSSE3-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1344; SSSE3-NEXT: andl $7, %r9d1345; SSSE3-NEXT: movzwl -24(%rsp,%rcx,2), %eax1346; SSSE3-NEXT: movd %eax, %xmm01347; SSSE3-NEXT: movzwl -40(%rsp,%rdx,2), %eax1348; SSSE3-NEXT: movd %eax, %xmm11349; SSSE3-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]1350; SSSE3-NEXT: movzwl -24(%rsp,%rsi,2), %eax1351; SSSE3-NEXT: movd %eax, %xmm21352; SSSE3-NEXT: movzwl -40(%rsp,%rdi,2), %eax1353; SSSE3-NEXT: movd %eax, %xmm01354; SSSE3-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]1355; SSSE3-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]1356; SSSE3-NEXT: movzwl -24(%rsp,%r9,2), %eax1357; SSSE3-NEXT: movd %eax, %xmm11358; SSSE3-NEXT: movzwl -40(%rsp,%r8,2), %eax1359; SSSE3-NEXT: movd %eax, %xmm21360; SSSE3-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]1361; SSSE3-NEXT: pxor %xmm1, %xmm11362; SSSE3-NEXT: movss {{.*#+}} xmm1 = xmm2[0],xmm1[1,2,3]1363; SSSE3-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]1364; SSSE3-NEXT: retq1365;1366; SSE41-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:1367; SSE41: # %bb.0:1368; SSE41-NEXT: # kill: def $r9d killed $r9d def $r91369; SSE41-NEXT: # kill: def $r8d killed $r8d def $r81370; SSE41-NEXT: # kill: def $ecx killed $ecx def $rcx1371; SSE41-NEXT: # kill: def $edx killed $edx def $rdx1372; SSE41-NEXT: # kill: def $esi killed $esi def $rsi1373; SSE41-NEXT: # kill: def $edi killed $edi def $rdi1374; SSE41-NEXT: andl $7, %edi1375; SSE41-NEXT: movaps %xmm0, -{{[0-9]+}}(%rsp)1376; SSE41-NEXT: movzwl -40(%rsp,%rdi,2), %eax1377; SSE41-NEXT: andl $7, %esi1378; SSE41-NEXT: andl $7, %edx1379; SSE41-NEXT: andl $7, %ecx1380; SSE41-NEXT: andl $7, %r8d1381; SSE41-NEXT: movaps %xmm1, -{{[0-9]+}}(%rsp)1382; SSE41-NEXT: andl $7, %r9d1383; SSE41-NEXT: movd %eax, %xmm01384; SSE41-NEXT: pinsrw $1, -24(%rsp,%rsi,2), %xmm01385; SSE41-NEXT: pinsrw $2, -40(%rsp,%rdx,2), %xmm01386; SSE41-NEXT: pinsrw $3, -24(%rsp,%rcx,2), %xmm01387; SSE41-NEXT: pinsrw $4, -40(%rsp,%r8,2), %xmm01388; SSE41-NEXT: pinsrw $5, -24(%rsp,%r9,2), %xmm01389; SSE41-NEXT: retq1390;1391; AVX-LABEL: var_shuffle_v8i16_v8i16_xyxyxy00_i16:1392; AVX: # %bb.0:1393; AVX-NEXT: # kill: def $r9d killed $r9d def $r91394; AVX-NEXT: # kill: def $r8d killed $r8d def $r81395; AVX-NEXT: # kill: def $ecx killed $ecx def $rcx1396; AVX-NEXT: # kill: def $edx killed $edx def $rdx1397; AVX-NEXT: # kill: def $esi killed $esi def $rsi1398; AVX-NEXT: # kill: def $edi killed $edi def $rdi1399; AVX-NEXT: andl $7, %edi1400; AVX-NEXT: vmovaps %xmm0, -{{[0-9]+}}(%rsp)1401; AVX-NEXT: movzwl -40(%rsp,%rdi,2), %eax1402; AVX-NEXT: andl $7, %esi1403; AVX-NEXT: andl $7, %edx1404; AVX-NEXT: andl $7, %ecx1405; AVX-NEXT: andl $7, %r8d1406; AVX-NEXT: vmovaps %xmm1, -{{[0-9]+}}(%rsp)1407; AVX-NEXT: andl $7, %r9d1408; AVX-NEXT: vmovd %eax, %xmm01409; AVX-NEXT: vpinsrw $1, -24(%rsp,%rsi,2), %xmm0, %xmm01410; AVX-NEXT: vpinsrw $2, -40(%rsp,%rdx,2), %xmm0, %xmm01411; AVX-NEXT: vpinsrw $3, -24(%rsp,%rcx,2), %xmm0, %xmm01412; AVX-NEXT: vpinsrw $4, -40(%rsp,%r8,2), %xmm0, %xmm01413; AVX-NEXT: vpinsrw $5, -24(%rsp,%r9,2), %xmm0, %xmm01414; AVX-NEXT: retq1415 %x0 = extractelement <8 x i16> %x, i16 %i01416 %y1 = extractelement <8 x i16> %y, i16 %i11417 %x2 = extractelement <8 x i16> %x, i16 %i21418 %y3 = extractelement <8 x i16> %y, i16 %i31419 %x4 = extractelement <8 x i16> %x, i16 %i41420 %y5 = extractelement <8 x i16> %y, i16 %i51421 %x6 = extractelement <8 x i16> %x, i16 %i61422 %x7 = extractelement <8 x i16> %x, i16 %i71423 %r0 = insertelement <8 x i16> undef, i16 %x0, i32 01424 %r1 = insertelement <8 x i16> %r0, i16 %y1, i32 11425 %r2 = insertelement <8 x i16> %r1, i16 %x2, i32 21426 %r3 = insertelement <8 x i16> %r2, i16 %y3, i32 31427 %r4 = insertelement <8 x i16> %r3, i16 %x4, i32 41428 %r5 = insertelement <8 x i16> %r4, i16 %y5, i32 51429 %r6 = insertelement <8 x i16> %r5, i16 0, i32 61430 %r7 = insertelement <8 x i16> %r6, i16 0, i32 71431 ret <8 x i16> %r71432}1433