1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+sse2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE2 3; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+sse4.2 | FileCheck %s --check-prefix=SSE --check-prefix=SSE42 4; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1 5; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2 6 7define void @v3i64(<2 x i64> %a, <2 x i64> %b, <3 x i64>* %p) nounwind { 8; SSE2-LABEL: v3i64: 9; SSE2: # BB#0: 10; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,0,1] 11; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 12; SSE2-NEXT: movq %xmm2, 16(%rdi) 13; SSE2-NEXT: movdqa %xmm0, (%rdi) 14; SSE2-NEXT: retq 15; 16; SSE42-LABEL: v3i64: 17; SSE42: # BB#0: 18; SSE42-NEXT: pextrq $1, %xmm0, 16(%rdi) 19; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 20; SSE42-NEXT: movdqa %xmm0, (%rdi) 21; SSE42-NEXT: retq 22; 23; AVX1-LABEL: v3i64: 24; AVX1: # BB#0: 25; AVX1-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],xmm1[0] 26; AVX1-NEXT: vpextrq $1, %xmm0, 16(%rdi) 27; AVX1-NEXT: vmovapd %xmm1, (%rdi) 28; AVX1-NEXT: retq 29; 30; AVX2-LABEL: v3i64: 31; AVX2: # BB#0: 32; AVX2-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<def> 33; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 34; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,1,3] 35; AVX2-NEXT: vpextrq $1, %xmm0, 16(%rdi) 36; AVX2-NEXT: vmovdqa %xmm1, (%rdi) 37; AVX2-NEXT: vzeroupper 38; AVX2-NEXT: retq 39 %r = shufflevector <2 x i64> %a, <2 x i64> %b, <3 x i32> <i32 0, i32 2, i32 1> 40 store <3 x i64> %r, <3 x i64>* %p 41 ret void 42} 43define void @v3f64(<2 x double> %a, <2 x double> %b, <3 x double>* %p) nounwind { 44; SSE-LABEL: v3f64: 45; SSE: # BB#0: 46; SSE-NEXT: movhpd %xmm0, 16(%rdi) 47; SSE-NEXT: unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0] 48; SSE-NEXT: movapd %xmm0, (%rdi) 49; SSE-NEXT: retq 50; 51; AVX1-LABEL: v3f64: 52; AVX1: # BB#0: 53; AVX1-NEXT: vunpcklpd {{.*#+}} xmm1 = xmm0[0],xmm1[0] 54; AVX1-NEXT: vmovhpd %xmm0, 16(%rdi) 55; AVX1-NEXT: vmovapd %xmm1, (%rdi) 56; AVX1-NEXT: retq 57; 58; AVX2-LABEL: v3f64: 59; AVX2: # BB#0: 60; AVX2-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<def> 61; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 62; AVX2-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,2,1,3] 63; AVX2-NEXT: vmovhpd %xmm0, 16(%rdi) 64; AVX2-NEXT: vmovapd %xmm1, (%rdi) 65; AVX2-NEXT: vzeroupper 66; AVX2-NEXT: retq 67 %r = shufflevector <2 x double> %a, <2 x double> %b, <3 x i32> <i32 0, i32 2, i32 1> 68 store <3 x double> %r, <3 x double>* %p 69 ret void 70} 71 72define void @v3i32(<2 x i32> %a, <2 x i32> %b, <3 x i32>* %p) nounwind { 73; SSE2-LABEL: v3i32: 74; SSE2: # BB#0: 75; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[0,2,2,3] 76; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1] 77; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] 78; SSE2-NEXT: movd %xmm0, 8(%rdi) 79; SSE2-NEXT: movq %xmm2, (%rdi) 80; SSE2-NEXT: retq 81; 82; SSE42-LABEL: v3i32: 83; SSE42: # BB#0: 84; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] 85; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7] 86; SSE42-NEXT: pextrd $2, %xmm0, 8(%rdi) 87; SSE42-NEXT: movq %xmm1, (%rdi) 88; SSE42-NEXT: retq 89; 90; AVX1-LABEL: v3i32: 91; AVX1: # BB#0: 92; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,1,1] 93; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5,6,7] 94; AVX1-NEXT: vpextrd $2, %xmm0, 8(%rdi) 95; AVX1-NEXT: vmovq %xmm1, (%rdi) 96; AVX1-NEXT: retq 97; 98; AVX2-LABEL: v3i32: 99; AVX2: # BB#0: 100; AVX2-NEXT: vpbroadcastd %xmm1, %xmm1 101; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2,3] 102; AVX2-NEXT: vpextrd $2, %xmm0, 8(%rdi) 103; AVX2-NEXT: vmovq %xmm1, (%rdi) 104; AVX2-NEXT: retq 105 %r = shufflevector <2 x i32> %a, <2 x i32> %b, <3 x i32> <i32 0, i32 2, i32 1> 106 store <3 x i32> %r, <3 x i32>* %p 107 ret void 108} 109 110define void @v5i16(<4 x i16> %a, <4 x i16> %b, <5 x i16>* %p) nounwind { 111; SSE2-LABEL: v5i16: 112; SSE2: # BB#0: 113; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3] 114; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7] 115; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7] 116; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3] 117; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] 118; SSE2-NEXT: pextrw $6, %xmm0, %eax 119; SSE2-NEXT: movw %ax, 8(%rdi) 120; SSE2-NEXT: movq %xmm2, (%rdi) 121; SSE2-NEXT: retq 122; 123; SSE42-LABEL: v5i16: 124; SSE42: # BB#0: 125; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3] 126; SSE42-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7] 127; SSE42-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7] 128; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,3] 129; SSE42-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] 130; SSE42-NEXT: pextrw $6, %xmm0, 8(%rdi) 131; SSE42-NEXT: movq %xmm2, (%rdi) 132; SSE42-NEXT: retq 133; 134; AVX-LABEL: v5i16: 135; AVX: # BB#0: 136; AVX-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[2,1,2,3] 137; AVX-NEXT: vpshuflw {{.*#+}} xmm1 = xmm1[2,0,2,3,4,5,6,7] 138; AVX-NEXT: vpshuflw {{.*#+}} xmm2 = xmm0[0,2,2,3,4,5,6,7] 139; AVX-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[0,3,2,3] 140; AVX-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] 141; AVX-NEXT: vpextrw $6, %xmm0, 8(%rdi) 142; AVX-NEXT: vmovq %xmm1, (%rdi) 143; AVX-NEXT: retq 144 %r = shufflevector <4 x i16> %a, <4 x i16> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 145 store <5 x i16> %r, <5 x i16>* %p 146 ret void 147} 148 149define void @v5i32(<4 x i32> %a, <4 x i32> %b, <5 x i32>* %p) nounwind { 150; SSE2-LABEL: v5i32: 151; SSE2: # BB#0: 152; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,2,2,3] 153; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,1,2,3] 154; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 155; SSE2-NEXT: movd %xmm2, 16(%rdi) 156; SSE2-NEXT: movdqa %xmm0, (%rdi) 157; SSE2-NEXT: retq 158; 159; SSE42-LABEL: v5i32: 160; SSE42: # BB#0: 161; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2] 162; SSE42-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero 163; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7] 164; SSE42-NEXT: pextrd $3, %xmm0, 16(%rdi) 165; SSE42-NEXT: movdqa %xmm2, (%rdi) 166; SSE42-NEXT: retq 167; 168; AVX1-LABEL: v5i32: 169; AVX1: # BB#0: 170; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[0,1],xmm1[1,2] 171; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,3] 172; AVX1-NEXT: vpextrd $3, %xmm0, 16(%rdi) 173; AVX1-NEXT: vmovaps %xmm1, (%rdi) 174; AVX1-NEXT: retq 175; 176; AVX2-LABEL: v5i32: 177; AVX2: # BB#0: 178; AVX2-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<def> 179; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 180; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,5,1,6,3,u,u,u> 181; AVX2-NEXT: vpermd %ymm1, %ymm2, %ymm1 182; AVX2-NEXT: vpextrd $3, %xmm0, 16(%rdi) 183; AVX2-NEXT: vmovdqa %xmm1, (%rdi) 184; AVX2-NEXT: vzeroupper 185; AVX2-NEXT: retq 186 %r = shufflevector <4 x i32> %a, <4 x i32> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 187 store <5 x i32> %r, <5 x i32>* %p 188 ret void 189} 190 191define void @v5f32(<4 x float> %a, <4 x float> %b, <5 x float>* %p) nounwind { 192; SSE2-LABEL: v5f32: 193; SSE2: # BB#0: 194; SSE2-NEXT: movaps %xmm0, %xmm2 195; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[1,2] 196; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2,1,3] 197; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,1,2,3] 198; SSE2-NEXT: movss %xmm0, 16(%rdi) 199; SSE2-NEXT: movaps %xmm2, (%rdi) 200; SSE2-NEXT: retq 201; 202; SSE42-LABEL: v5f32: 203; SSE42: # BB#0: 204; SSE42-NEXT: extractps $3, %xmm0, 16(%rdi) 205; SSE42-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,2] 206; SSE42-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2,1,3] 207; SSE42-NEXT: movaps %xmm0, (%rdi) 208; SSE42-NEXT: retq 209; 210; AVX1-LABEL: v5f32: 211; AVX1: # BB#0: 212; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm0[0,1],xmm1[1,2] 213; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,3] 214; AVX1-NEXT: vextractps $3, %xmm0, 16(%rdi) 215; AVX1-NEXT: vmovaps %xmm1, (%rdi) 216; AVX1-NEXT: retq 217; 218; AVX2-LABEL: v5f32: 219; AVX2: # BB#0: 220; AVX2-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<def> 221; AVX2-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 222; AVX2-NEXT: vmovaps {{.*#+}} ymm2 = <0,5,1,6,3,u,u,u> 223; AVX2-NEXT: vpermps %ymm1, %ymm2, %ymm1 224; AVX2-NEXT: vextractps $3, %xmm0, 16(%rdi) 225; AVX2-NEXT: vmovaps %xmm1, (%rdi) 226; AVX2-NEXT: vzeroupper 227; AVX2-NEXT: retq 228 %r = shufflevector <4 x float> %a, <4 x float> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 229 store <5 x float> %r, <5 x float>* %p 230 ret void 231} 232 233define void @v7i8(<4 x i8> %a, <4 x i8> %b, <7 x i8>* %p) nounwind { 234; SSE2-LABEL: v7i8: 235; SSE2: # BB#0: 236; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 237; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [0,65535,0,65535,0,65535,65535,65535] 238; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,0,3] 239; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,0,4,5,6,7] 240; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,4,7] 241; SSE2-NEXT: pand %xmm2, %xmm1 242; SSE2-NEXT: pandn %xmm0, %xmm2 243; SSE2-NEXT: por %xmm1, %xmm2 244; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [255,255,255,255,255,255,255,255] 245; SSE2-NEXT: pand %xmm2, %xmm0 246; SSE2-NEXT: packuswb %xmm0, %xmm0 247; SSE2-NEXT: movdqa %xmm2, -{{[0-9]+}}(%rsp) 248; SSE2-NEXT: movb -{{[0-9]+}}(%rsp), %al 249; SSE2-NEXT: movb %al, 6(%rdi) 250; SSE2-NEXT: movd %xmm0, (%rdi) 251; SSE2-NEXT: pextrw $2, %xmm0, %eax 252; SSE2-NEXT: movw %ax, 4(%rdi) 253; SSE2-NEXT: retq 254; 255; SSE42-LABEL: v7i8: 256; SSE42: # BB#0: 257; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 258; SSE42-NEXT: pextrb $0, %xmm1, 6(%rdi) 259; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[8,9,8,9,4,5,8,9,0,1,12,13,0,1,14,15] 260; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7] 261; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] 262; SSE42-NEXT: pextrw $2, %xmm1, 4(%rdi) 263; SSE42-NEXT: movd %xmm1, (%rdi) 264; SSE42-NEXT: retq 265; 266; AVX-LABEL: v7i8: 267; AVX: # BB#0: 268; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 269; AVX-NEXT: vpshufb {{.*#+}} xmm2 = xmm1[8,9,8,9,4,5,8,9,0,1,12,13,0,1,14,15] 270; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5,6,7] 271; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u] 272; AVX-NEXT: vpextrb $0, %xmm1, 6(%rdi) 273; AVX-NEXT: vpextrw $2, %xmm0, 4(%rdi) 274; AVX-NEXT: vmovd %xmm0, (%rdi) 275; AVX-NEXT: retq 276 %r = shufflevector <4 x i8> %a, <4 x i8> %b, <7 x i32> <i32 0, i32 6, i32 3, i32 6, i32 1, i32 7, i32 4> 277 store <7 x i8> %r, <7 x i8>* %p 278 ret void 279} 280 281define void @v7i16(<4 x i16> %a, <4 x i16> %b, <7 x i16>* %p) nounwind { 282; SSE2-LABEL: v7i16: 283; SSE2: # BB#0: 284; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 285; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [0,65535,0,65535,0,65535,65535,65535] 286; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[2,1,0,3] 287; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,0,2,0,4,5,6,7] 288; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,4,7] 289; SSE2-NEXT: pand %xmm2, %xmm3 290; SSE2-NEXT: pandn %xmm0, %xmm2 291; SSE2-NEXT: por %xmm3, %xmm2 292; SSE2-NEXT: movd %xmm1, %eax 293; SSE2-NEXT: movw %ax, 12(%rdi) 294; SSE2-NEXT: movq %xmm2, (%rdi) 295; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1] 296; SSE2-NEXT: movd %xmm0, 8(%rdi) 297; SSE2-NEXT: retq 298; 299; SSE42-LABEL: v7i16: 300; SSE42: # BB#0: 301; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 302; SSE42-NEXT: pextrw $0, %xmm1, 12(%rdi) 303; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[8,9,8,9,4,5,8,9,0,1,12,13,0,1,14,15] 304; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3],xmm0[4],xmm1[5,6,7] 305; SSE42-NEXT: pextrd $2, %xmm1, 8(%rdi) 306; SSE42-NEXT: movq %xmm1, (%rdi) 307; SSE42-NEXT: retq 308; 309; AVX-LABEL: v7i16: 310; AVX: # BB#0: 311; AVX-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,3,1,3] 312; AVX-NEXT: vpshufb {{.*#+}} xmm2 = xmm1[8,9,8,9,4,5,8,9,0,1,12,13,0,1,14,15] 313; AVX-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2],xmm2[3],xmm0[4],xmm2[5,6,7] 314; AVX-NEXT: vpextrw $0, %xmm1, 12(%rdi) 315; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 316; AVX-NEXT: vmovq %xmm0, (%rdi) 317; AVX-NEXT: retq 318 %r = shufflevector <4 x i16> %a, <4 x i16> %b, <7 x i32> <i32 0, i32 6, i32 3, i32 6, i32 1, i32 7, i32 4> 319 store <7 x i16> %r, <7 x i16>* %p 320 ret void 321} 322 323 324define void @v7i32(<4 x i32> %a, <4 x i32> %b, <7 x i32>* %p) nounwind { 325; SSE2-LABEL: v7i32: 326; SSE2: # BB#0: 327; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,1,2,2] 328; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,1,0,3] 329; SSE2-NEXT: punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm2[2],xmm3[3],xmm2[3] 330; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm1[3,0] 331; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,3] 332; SSE2-NEXT: movd %xmm1, 24(%rdi) 333; SSE2-NEXT: movlps %xmm0, 16(%rdi) 334; SSE2-NEXT: movdqa %xmm3, (%rdi) 335; SSE2-NEXT: retq 336; 337; SSE42-LABEL: v7i32: 338; SSE42: # BB#0: 339; SSE42-NEXT: movdqa %xmm1, %xmm2 340; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3],xmm2[4,5,6,7] 341; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5],xmm0[6,7] 342; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,3,2] 343; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,3,0,3] 344; SSE42-NEXT: movd %xmm1, 24(%rdi) 345; SSE42-NEXT: movq %xmm2, 16(%rdi) 346; SSE42-NEXT: movdqa %xmm0, (%rdi) 347; SSE42-NEXT: retq 348; 349; AVX1-LABEL: v7i32: 350; AVX1: # BB#0: 351; AVX1-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm1[2],xmm0[3] 352; AVX1-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,2,3,2] 353; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1],xmm1[2,3] 354; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,3,0,3] 355; AVX1-NEXT: vmovss %xmm1, 24(%rdi) 356; AVX1-NEXT: vmovlps %xmm0, 16(%rdi) 357; AVX1-NEXT: vmovaps %xmm2, (%rdi) 358; AVX1-NEXT: retq 359; 360; AVX2-LABEL: v7i32: 361; AVX2: # BB#0: 362; AVX2-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<def> 363; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 364; AVX2-NEXT: vmovdqa {{.*#+}} ymm2 = <0,6,3,6,1,7,4,u> 365; AVX2-NEXT: vpermd %ymm0, %ymm2, %ymm0 366; AVX2-NEXT: vmovd %xmm1, 24(%rdi) 367; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 368; AVX2-NEXT: vmovq %xmm1, 16(%rdi) 369; AVX2-NEXT: vmovdqa %xmm0, (%rdi) 370; AVX2-NEXT: vzeroupper 371; AVX2-NEXT: retq 372 %r = shufflevector <4 x i32> %a, <4 x i32> %b, <7 x i32> <i32 0, i32 6, i32 3, i32 6, i32 1, i32 7, i32 4> 373 store <7 x i32> %r, <7 x i32>* %p 374 ret void 375} 376 377define void @v12i8(<8 x i8> %a, <8 x i8> %b, <12 x i8>* %p) nounwind { 378; SSE2-LABEL: v12i8: 379; SSE2: # BB#0: 380; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] 381; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,2,3,4,5,6,7] 382; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,6,6,7] 383; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,255] 384; SSE2-NEXT: pand {{.*}}(%rip), %xmm0 385; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 386; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 387; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 388; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1] 389; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,1,4,5,6,7] 390; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,6,4] 391; SSE2-NEXT: packuswb %xmm3, %xmm0 392; SSE2-NEXT: pand %xmm2, %xmm0 393; SSE2-NEXT: pandn %xmm1, %xmm2 394; SSE2-NEXT: por %xmm0, %xmm2 395; SSE2-NEXT: movq %xmm2, (%rdi) 396; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1] 397; SSE2-NEXT: movd %xmm0, 8(%rdi) 398; SSE2-NEXT: retq 399; 400; SSE42-LABEL: v12i8: 401; SSE42: # BB#0: 402; SSE42-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,xmm1[0],zero,zero,xmm1[2],zero,zero,xmm1[4],zero,zero,xmm1[6,u,u,u,u] 403; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8],zero,xmm0[2,10],zero,xmm0[4,12],zero,xmm0[6,14],zero,xmm0[u,u,u,u] 404; SSE42-NEXT: por %xmm1, %xmm0 405; SSE42-NEXT: pextrd $2, %xmm0, 8(%rdi) 406; SSE42-NEXT: movq %xmm0, (%rdi) 407; SSE42-NEXT: retq 408; 409; AVX-LABEL: v12i8: 410; AVX: # BB#0: 411; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,xmm1[0],zero,zero,xmm1[2],zero,zero,xmm1[4],zero,zero,xmm1[6,u,u,u,u] 412; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8],zero,xmm0[2,10],zero,xmm0[4,12],zero,xmm0[6,14],zero,xmm0[u,u,u,u] 413; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0 414; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 415; AVX-NEXT: vmovq %xmm0, (%rdi) 416; AVX-NEXT: retq 417 %r = shufflevector <8 x i8> %a, <8 x i8> %b, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 418 store <12 x i8> %r, <12 x i8>* %p 419 ret void 420} 421 422define void @v12i16(<8 x i16> %a, <8 x i16> %b, <12 x i16>* %p) nounwind { 423; SSE2-LABEL: v12i16: 424; SSE2: # BB#0: 425; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,0,0,3] 426; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535] 427; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,6,5,4,7] 428; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1] 429; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7] 430; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,6,4] 431; SSE2-NEXT: pand %xmm3, %xmm4 432; SSE2-NEXT: pandn %xmm2, %xmm3 433; SSE2-NEXT: por %xmm4, %xmm3 434; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,2,3] 435; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,0,65535,65535,65535,65535,65535] 436; SSE2-NEXT: pand %xmm2, %xmm1 437; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3] 438; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,1,3,4,5,6,7] 439; SSE2-NEXT: pandn %xmm0, %xmm2 440; SSE2-NEXT: por %xmm1, %xmm2 441; SSE2-NEXT: movq %xmm2, 16(%rdi) 442; SSE2-NEXT: movdqa %xmm3, (%rdi) 443; SSE2-NEXT: retq 444; 445; SSE42-LABEL: v12i16: 446; SSE42: # BB#0: 447; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,2,3] 448; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 449; SSE42-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 450; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm2[0],xmm3[1,2],xmm2[3,4,5,6,7] 451; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,3] 452; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,2,3,10,11,10,11,4,5,12,13] 453; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 454; SSE42-NEXT: movdqa %xmm0, (%rdi) 455; SSE42-NEXT: movq %xmm3, 16(%rdi) 456; SSE42-NEXT: retq 457; 458; AVX1-LABEL: v12i16: 459; AVX1: # BB#0: 460; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,2,3] 461; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 462; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 463; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3],xmm3[4,5,6,7] 464; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,3] 465; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,2,3,10,11,10,11,4,5,12,13] 466; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 467; AVX1-NEXT: vmovdqa %xmm0, (%rdi) 468; AVX1-NEXT: vmovq %xmm2, 16(%rdi) 469; AVX1-NEXT: retq 470; 471; AVX2-LABEL: v12i16: 472; AVX2: # BB#0: 473; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,2,3] 474; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 475; AVX2-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 476; AVX2-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3],xmm3[4,5,6,7] 477; AVX2-NEXT: vpbroadcastd %xmm1, %xmm1 478; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,8,9,2,3,10,11,10,11,4,5,12,13] 479; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 480; AVX2-NEXT: vmovdqa %xmm0, (%rdi) 481; AVX2-NEXT: vmovq %xmm2, 16(%rdi) 482; AVX2-NEXT: retq 483 %r = shufflevector <8 x i16> %a, <8 x i16> %b, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 484 store <12 x i16> %r, <12 x i16>* %p 485 ret void 486} 487 488define void @v12i32(<8 x i32> %a, <8 x i32> %b, <12 x i32>* %p) nounwind { 489; SSE2-LABEL: v12i32: 490; SSE2: # BB#0: 491; SSE2-NEXT: movdqa %xmm0, %xmm3 492; SSE2-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],xmm1[0],xmm3[1],xmm1[1] 493; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,2] 494; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[0,1,0,1] 495; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm3[3,0] 496; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[0,2] 497; SSE2-NEXT: movdqa %xmm2, %xmm4 498; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm1[1,0] 499; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[2,0],xmm1[2,2] 500; SSE2-NEXT: punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3] 501; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[2,0],xmm4[3,0] 502; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm0[0,2] 503; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,3,2,2] 504; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3] 505; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm0[3,0] 506; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[0,2] 507; SSE2-NEXT: movaps %xmm0, 32(%rdi) 508; SSE2-NEXT: movaps %xmm4, 16(%rdi) 509; SSE2-NEXT: movaps %xmm3, (%rdi) 510; SSE2-NEXT: retq 511; 512; SSE42-LABEL: v12i32: 513; SSE42: # BB#0: 514; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1] 515; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,1,0,1] 516; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3],xmm4[4,5,6,7] 517; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1] 518; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm4[0,1,2,3],xmm3[4,5],xmm4[6,7] 519; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,1,2,2] 520; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3],xmm4[4,5,6,7] 521; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm0[4,5],xmm4[6,7] 522; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,0,1] 523; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3] 524; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3],xmm2[4,5,6,7] 525; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3] 526; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4,5],xmm2[6,7] 527; SSE42-NEXT: movdqa %xmm0, 32(%rdi) 528; SSE42-NEXT: movdqa %xmm4, 16(%rdi) 529; SSE42-NEXT: movdqa %xmm3, (%rdi) 530; SSE42-NEXT: retq 531; 532; AVX1-LABEL: v12i32: 533; AVX1: # BB#0: 534; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1] 535; AVX1-NEXT: vmovsldup {{.*#+}} ymm2 = ymm2[0,0,2,2,4,4,6,6] 536; AVX1-NEXT: vpermilps {{.*#+}} ymm3 = ymm0[0,u,u,1,5,u,u,6] 537; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4,5],ymm2[6],ymm3[7] 538; AVX1-NEXT: vmovddup {{.*#+}} xmm3 = xmm1[0,0] 539; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm3, %ymm3 540; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 541; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 542; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,3],xmm3[3,3] 543; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,1] 544; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 545; AVX1-NEXT: vmovaps %xmm0, 32(%rdi) 546; AVX1-NEXT: vmovaps %ymm2, (%rdi) 547; AVX1-NEXT: vzeroupper 548; AVX1-NEXT: retq 549; 550; AVX2-LABEL: v12i32: 551; AVX2: # BB#0: 552; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[2,3,2,3] 553; AVX2-NEXT: vpshufd {{.*#+}} ymm3 = ymm0[3,3,2,3,7,7,6,7] 554; AVX2-NEXT: vpermq {{.*#+}} ymm3 = ymm3[0,2,2,3] 555; AVX2-NEXT: vpblendd {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3] 556; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = <0,4,u,1,5,u,2,6> 557; AVX2-NEXT: vpermd %ymm0, %ymm3, %ymm0 558; AVX2-NEXT: vpbroadcastq %xmm1, %ymm1 559; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 560; AVX2-NEXT: vmovdqa %ymm0, (%rdi) 561; AVX2-NEXT: vmovdqa %xmm2, 32(%rdi) 562; AVX2-NEXT: vzeroupper 563; AVX2-NEXT: retq 564 %r = shufflevector <8 x i32> %a, <8 x i32> %b, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 565 store <12 x i32> %r, <12 x i32>* %p 566 ret void 567} 568 569define void @pr29025(<4 x i8> %a, <4 x i8> %b, <4 x i8> %c, <12 x i8> *%p) nounwind { 570; SSE2-LABEL: pr29025: 571; SSE2: # BB#0: 572; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,255,255,255] 573; SSE2-NEXT: pand %xmm3, %xmm1 574; SSE2-NEXT: pand %xmm3, %xmm0 575; SSE2-NEXT: packuswb %xmm1, %xmm0 576; SSE2-NEXT: packuswb %xmm0, %xmm0 577; SSE2-NEXT: pxor %xmm1, %xmm1 578; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] 579; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3] 580; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,1,3,4,5,6,7] 581; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 582; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1] 583; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,1,4,5,6,7] 584; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,6,4] 585; SSE2-NEXT: packuswb %xmm1, %xmm0 586; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,255] 587; SSE2-NEXT: pand %xmm1, %xmm0 588; SSE2-NEXT: pand %xmm3, %xmm2 589; SSE2-NEXT: packuswb %xmm2, %xmm2 590; SSE2-NEXT: packuswb %xmm2, %xmm2 591; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,1,1,4,5,6,7] 592; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,1,1] 593; SSE2-NEXT: pandn %xmm2, %xmm1 594; SSE2-NEXT: por %xmm0, %xmm1 595; SSE2-NEXT: movq %xmm1, (%rdi) 596; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1] 597; SSE2-NEXT: movd %xmm0, 8(%rdi) 598; SSE2-NEXT: retq 599; 600; SSE42-LABEL: pr29025: 601; SSE42: # BB#0: 602; SSE42-NEXT: movdqa {{.*#+}} xmm3 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> 603; SSE42-NEXT: pshufb %xmm3, %xmm1 604; SSE42-NEXT: pshufb %xmm3, %xmm0 605; SSE42-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 606; SSE42-NEXT: pshufb %xmm3, %xmm2 607; SSE42-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7] 608; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 609; SSE42-NEXT: pextrd $2, %xmm0, 8(%rdi) 610; SSE42-NEXT: movq %xmm0, (%rdi) 611; SSE42-NEXT: retq 612; 613; AVX-LABEL: pr29025: 614; AVX: # BB#0: 615; AVX-NEXT: vmovdqa {{.*#+}} xmm3 = <0,4,8,12,u,u,u,u,u,u,u,u,u,u,u,u> 616; AVX-NEXT: vpshufb %xmm3, %xmm1, %xmm1 617; AVX-NEXT: vpshufb %xmm3, %xmm0, %xmm0 618; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 619; AVX-NEXT: vpshufb %xmm3, %xmm2, %xmm1 620; AVX-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7] 621; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 622; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 623; AVX-NEXT: vmovq %xmm0, (%rdi) 624; AVX-NEXT: retq 625 %s1 = shufflevector <4 x i8> %a, <4 x i8> %b, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 626 %s2 = shufflevector <4 x i8> %c, <4 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef> 627 %r = shufflevector <8 x i8> %s1, <8 x i8> %s2, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 628 store <12 x i8> %r, <12 x i8>* %p, align 1 629 ret void 630} 631 632define void @interleave_24i8_out(<24 x i8>* %p, <8 x i8>* %q1, <8 x i8>* %q2, <8 x i8>* %q3) nounwind { 633; SSE2-LABEL: interleave_24i8_out: 634; SSE2: # BB#0: 635; SSE2-NEXT: movdqu (%rdi), %xmm0 636; SSE2-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 637; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,255,0,255,255,0,255,255,255,255,255,255,255,255,255,255] 638; SSE2-NEXT: movdqa %xmm0, %xmm2 639; SSE2-NEXT: pand %xmm3, %xmm2 640; SSE2-NEXT: pandn %xmm1, %xmm3 641; SSE2-NEXT: por %xmm2, %xmm3 642; SSE2-NEXT: pxor %xmm2, %xmm2 643; SSE2-NEXT: movdqa %xmm3, %xmm4 644; SSE2-NEXT: punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15] 645; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0] 646; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7] 647; SSE2-NEXT: pand %xmm5, %xmm3 648; SSE2-NEXT: pandn %xmm4, %xmm5 649; SSE2-NEXT: por %xmm3, %xmm5 650; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,2,1,3] 651; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,6,5] 652; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1] 653; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,2,1,4,5,6,7] 654; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7] 655; SSE2-NEXT: packuswb %xmm0, %xmm3 656; SSE2-NEXT: movq %xmm3, (%rsi) 657; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [0,255,255,0,255,255,0,255,255,255,255,255,255,255,255,255] 658; SSE2-NEXT: movdqa %xmm0, %xmm4 659; SSE2-NEXT: pand %xmm3, %xmm4 660; SSE2-NEXT: pandn %xmm1, %xmm3 661; SSE2-NEXT: por %xmm4, %xmm3 662; SSE2-NEXT: movdqa %xmm3, %xmm4 663; SSE2-NEXT: punpckhbw {{.*#+}} xmm4 = xmm4[8],xmm2[8],xmm4[9],xmm2[9],xmm4[10],xmm2[10],xmm4[11],xmm2[11],xmm4[12],xmm2[12],xmm4[13],xmm2[13],xmm4[14],xmm2[14],xmm4[15],xmm2[15] 664; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535] 665; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7] 666; SSE2-NEXT: pand %xmm5, %xmm3 667; SSE2-NEXT: pandn %xmm4, %xmm5 668; SSE2-NEXT: por %xmm3, %xmm5 669; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm5[2,1,0,3,4,5,6,7] 670; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,5,4,7] 671; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,3,2,1] 672; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,2,3,0,4,5,6,7] 673; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,5,6,7,4] 674; SSE2-NEXT: packuswb %xmm0, %xmm3 675; SSE2-NEXT: movq %xmm3, (%rdx) 676; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [255,0,255,255,0,255,255,0,255,255,255,255,255,255,255,255] 677; SSE2-NEXT: pand %xmm3, %xmm0 678; SSE2-NEXT: pandn %xmm1, %xmm3 679; SSE2-NEXT: por %xmm0, %xmm3 680; SSE2-NEXT: movdqa %xmm3, %xmm0 681; SSE2-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm2[8],xmm0[9],xmm2[9],xmm0[10],xmm2[10],xmm0[11],xmm2[11],xmm0[12],xmm2[12],xmm0[13],xmm2[13],xmm0[14],xmm2[14],xmm0[15],xmm2[15] 682; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [0,65535,65535,0,65535,65535,0,65535] 683; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7] 684; SSE2-NEXT: pand %xmm1, %xmm3 685; SSE2-NEXT: pandn %xmm0, %xmm1 686; SSE2-NEXT: por %xmm3, %xmm1 687; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,0] 688; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5] 689; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0] 690; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7] 691; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 692; SSE2-NEXT: packuswb %xmm0, %xmm0 693; SSE2-NEXT: movq %xmm0, (%rcx) 694; SSE2-NEXT: retq 695; 696; SSE42-LABEL: interleave_24i8_out: 697; SSE42: # BB#0: 698; SSE42-NEXT: movdqu (%rdi), %xmm0 699; SSE42-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 700; SSE42-NEXT: movdqa %xmm1, %xmm2 701; SSE42-NEXT: pshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm2[2,5,u,u,u,u,u,u,u,u] 702; SSE42-NEXT: movdqa %xmm0, %xmm3 703; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[0,3,6,9,12,15],zero,zero,xmm3[u,u,u,u,u,u,u,u] 704; SSE42-NEXT: por %xmm2, %xmm3 705; SSE42-NEXT: movq %xmm3, (%rsi) 706; SSE42-NEXT: movdqa %xmm1, %xmm2 707; SSE42-NEXT: pshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,xmm2[0,3,6,u,u,u,u,u,u,u,u] 708; SSE42-NEXT: movdqa %xmm0, %xmm3 709; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[1,4,7,10,13],zero,zero,zero,xmm3[u,u,u,u,u,u,u,u] 710; SSE42-NEXT: por %xmm2, %xmm3 711; SSE42-NEXT: movq %xmm3, (%rdx) 712; SSE42-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u] 713; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 714; SSE42-NEXT: por %xmm1, %xmm0 715; SSE42-NEXT: movq %xmm0, (%rcx) 716; SSE42-NEXT: retq 717; 718; AVX-LABEL: interleave_24i8_out: 719; AVX: # BB#0: 720; AVX-NEXT: vmovdqu (%rdi), %xmm0 721; AVX-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 722; AVX-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u] 723; AVX-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u] 724; AVX-NEXT: vpor %xmm2, %xmm3, %xmm2 725; AVX-NEXT: vmovq %xmm2, (%rsi) 726; AVX-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u] 727; AVX-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 728; AVX-NEXT: vpor %xmm2, %xmm3, %xmm2 729; AVX-NEXT: vmovq %xmm2, (%rdx) 730; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u] 731; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 732; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0 733; AVX-NEXT: vmovq %xmm0, (%rcx) 734; AVX-NEXT: retq 735 %wide.vec = load <24 x i8>, <24 x i8>* %p, align 4 736 %s1 = shufflevector <24 x i8> %wide.vec, <24 x i8> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 737 %s2 = shufflevector <24 x i8> %wide.vec, <24 x i8> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 738 %s3 = shufflevector <24 x i8> %wide.vec, <24 x i8> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 739 store <8 x i8> %s1, <8 x i8>* %q1, align 4 740 store <8 x i8> %s2, <8 x i8>* %q2, align 4 741 store <8 x i8> %s3, <8 x i8>* %q3, align 4 742 ret void 743} 744 745define void @interleave_24i8_in(<24 x i8>* %p, <8 x i8>* %q1, <8 x i8>* %q2, <8 x i8>* %q3) nounwind { 746; SSE2-LABEL: interleave_24i8_in: 747; SSE2: # BB#0: 748; SSE2-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 749; SSE2-NEXT: movq {{.*#+}} xmm2 = mem[0],zero 750; SSE2-NEXT: movq {{.*#+}} xmm0 = mem[0],zero 751; SSE2-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0] 752; SSE2-NEXT: pxor %xmm2, %xmm2 753; SSE2-NEXT: movdqa %xmm1, %xmm3 754; SSE2-NEXT: punpcklbw {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3],xmm3[4],xmm2[4],xmm3[5],xmm2[5],xmm3[6],xmm2[6],xmm3[7],xmm2[7] 755; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[1,1,2,2] 756; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535] 757; SSE2-NEXT: pand %xmm5, %xmm4 758; SSE2-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm2[8],xmm1[9],xmm2[9],xmm1[10],xmm2[10],xmm1[11],xmm2[11],xmm1[12],xmm2[12],xmm1[13],xmm2[13],xmm1[14],xmm2[14],xmm1[15],xmm2[15] 759; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[0,1,3,3,4,5,6,7] 760; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,4,4,6,7] 761; SSE2-NEXT: pandn %xmm2, %xmm5 762; SSE2-NEXT: por %xmm4, %xmm5 763; SSE2-NEXT: movdqa %xmm3, %xmm2 764; SSE2-NEXT: punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3] 765; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,1,2,1] 766; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,1,2,2,4,5,6,7] 767; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,7,5,4,5] 768; SSE2-NEXT: packuswb %xmm5, %xmm2 769; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255] 770; SSE2-NEXT: pand %xmm4, %xmm2 771; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,0,1] 772; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,3,4,5,6,7] 773; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,6,6] 774; SSE2-NEXT: pandn %xmm5, %xmm4 775; SSE2-NEXT: por %xmm2, %xmm4 776; SSE2-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7] 777; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,1,2,3] 778; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,1,0,4,5,6,7] 779; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7] 780; SSE2-NEXT: packuswb %xmm0, %xmm1 781; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,0,255,255,0,255,255,0,255,255,255,255,255,255,255,255] 782; SSE2-NEXT: pand %xmm2, %xmm1 783; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7] 784; SSE2-NEXT: pandn %xmm0, %xmm2 785; SSE2-NEXT: por %xmm1, %xmm2 786; SSE2-NEXT: movq %xmm2, 16(%rdi) 787; SSE2-NEXT: movdqu %xmm4, (%rdi) 788; SSE2-NEXT: retq 789; 790; SSE42-LABEL: interleave_24i8_in: 791; SSE42: # BB#0: 792; SSE42-NEXT: movq {{.*#+}} xmm0 = mem[0],zero 793; SSE42-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 794; SSE42-NEXT: movq {{.*#+}} xmm2 = mem[0],zero 795; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 796; SSE42-NEXT: movdqa %xmm0, %xmm1 797; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[0,8],zero,xmm1[1,9],zero,xmm1[2,10],zero,xmm1[3,11],zero,xmm1[4,12],zero,xmm1[5] 798; SSE42-NEXT: movdqa %xmm2, %xmm3 799; SSE42-NEXT: pshufb {{.*#+}} xmm3 = zero,zero,xmm3[0],zero,zero,xmm3[1],zero,zero,xmm3[2],zero,zero,xmm3[3],zero,zero,xmm3[4],zero 800; SSE42-NEXT: por %xmm1, %xmm3 801; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[13],zero,xmm0[6,14],zero,xmm0[7,15],zero,xmm0[u,u,u,u,u,u,u,u] 802; SSE42-NEXT: pshufb {{.*#+}} xmm2 = zero,xmm2[5],zero,zero,xmm2[6],zero,zero,xmm2[7,u,u,u,u,u,u,u,u] 803; SSE42-NEXT: por %xmm0, %xmm2 804; SSE42-NEXT: movq %xmm2, 16(%rdi) 805; SSE42-NEXT: movdqu %xmm3, (%rdi) 806; SSE42-NEXT: retq 807; 808; AVX-LABEL: interleave_24i8_in: 809; AVX: # BB#0: 810; AVX-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero 811; AVX-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 812; AVX-NEXT: vmovq {{.*#+}} xmm2 = mem[0],zero 813; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 814; AVX-NEXT: vpshufb {{.*#+}} xmm1 = xmm0[0,8],zero,xmm0[1,9],zero,xmm0[2,10],zero,xmm0[3,11],zero,xmm0[4,12],zero,xmm0[5] 815; AVX-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm2[0],zero,zero,xmm2[1],zero,zero,xmm2[2],zero,zero,xmm2[3],zero,zero,xmm2[4],zero 816; AVX-NEXT: vpor %xmm3, %xmm1, %xmm1 817; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[13],zero,xmm0[6,14],zero,xmm0[7,15],zero,xmm0[u,u,u,u,u,u,u,u] 818; AVX-NEXT: vpshufb {{.*#+}} xmm2 = zero,xmm2[5],zero,zero,xmm2[6],zero,zero,xmm2[7,u,u,u,u,u,u,u,u] 819; AVX-NEXT: vpor %xmm2, %xmm0, %xmm0 820; AVX-NEXT: vmovq %xmm0, 16(%rdi) 821; AVX-NEXT: vmovdqu %xmm1, (%rdi) 822; AVX-NEXT: retq 823 %s1 = load <8 x i8>, <8 x i8>* %q1, align 4 824 %s2 = load <8 x i8>, <8 x i8>* %q2, align 4 825 %s3 = load <8 x i8>, <8 x i8>* %q3, align 4 826 %t1 = shufflevector <8 x i8> %s1, <8 x i8> %s2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 827 %t2 = shufflevector <8 x i8> %s3, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 828 %interleaved = shufflevector <16 x i8> %t1, <16 x i8> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 829 store <24 x i8> %interleaved, <24 x i8>* %p, align 4 830 ret void 831} 832 833 834define void @interleave_24i16_out(<24 x i16>* %p, <8 x i16>* %q1, <8 x i16>* %q2, <8 x i16>* %q3) nounwind { 835; SSE2-LABEL: interleave_24i16_out: 836; SSE2: # BB#0: 837; SSE2-NEXT: movdqu (%rdi), %xmm3 838; SSE2-NEXT: movdqu 16(%rdi), %xmm2 839; SSE2-NEXT: movdqu 32(%rdi), %xmm8 840; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0] 841; SSE2-NEXT: movdqa %xmm3, %xmm4 842; SSE2-NEXT: pand %xmm1, %xmm4 843; SSE2-NEXT: pandn %xmm2, %xmm1 844; SSE2-NEXT: por %xmm4, %xmm1 845; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3] 846; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7] 847; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3] 848; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7] 849; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,4,7,6,7] 850; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm8[0,1,2,1] 851; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5] 852; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[3,0],xmm1[2,0] 853; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm4[2,0] 854; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535] 855; SSE2-NEXT: movdqa %xmm4, %xmm5 856; SSE2-NEXT: pandn %xmm2, %xmm5 857; SSE2-NEXT: movdqa %xmm3, %xmm6 858; SSE2-NEXT: pand %xmm4, %xmm6 859; SSE2-NEXT: por %xmm5, %xmm6 860; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm6[2,1,2,3,4,5,6,7] 861; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7] 862; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3] 863; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7] 864; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,6,7] 865; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0] 866; SSE2-NEXT: pand %xmm6, %xmm5 867; SSE2-NEXT: pshuflw {{.*#+}} xmm7 = xmm8[0,3,2,3,4,5,6,7] 868; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3] 869; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,5,6] 870; SSE2-NEXT: movdqa %xmm6, %xmm0 871; SSE2-NEXT: pandn %xmm7, %xmm0 872; SSE2-NEXT: por %xmm5, %xmm0 873; SSE2-NEXT: pand %xmm4, %xmm2 874; SSE2-NEXT: pandn %xmm3, %xmm4 875; SSE2-NEXT: por %xmm2, %xmm4 876; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0] 877; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7] 878; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3] 879; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7] 880; SSE2-NEXT: pand %xmm6, %xmm2 881; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,7,6,7] 882; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0] 883; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,4,5] 884; SSE2-NEXT: pandn %xmm3, %xmm6 885; SSE2-NEXT: por %xmm2, %xmm6 886; SSE2-NEXT: movups %xmm1, (%rsi) 887; SSE2-NEXT: movdqu %xmm0, (%rdx) 888; SSE2-NEXT: movdqu %xmm6, (%rcx) 889; SSE2-NEXT: retq 890; 891; SSE42-LABEL: interleave_24i16_out: 892; SSE42: # BB#0: 893; SSE42-NEXT: movdqu (%rdi), %xmm0 894; SSE42-NEXT: movdqu 16(%rdi), %xmm1 895; SSE42-NEXT: movdqu 32(%rdi), %xmm2 896; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,2,1] 897; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5] 898; SSE42-NEXT: movdqa %xmm0, %xmm4 899; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3],xmm1[4],xmm4[5,6],xmm1[7] 900; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15] 901; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm3[6,7] 902; SSE42-NEXT: movdqa %xmm2, %xmm3 903; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,4,5,6,7,0,1,0,1,6,7,12,13] 904; SSE42-NEXT: movdqa %xmm0, %xmm5 905; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1],xmm1[2],xmm5[3,4],xmm1[5],xmm5[6,7] 906; SSE42-NEXT: pshufb {{.*#+}} xmm5 = xmm5[2,3,8,9,14,15,4,5,10,11,10,11,8,9,14,15] 907; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm3[5,6,7] 908; SSE42-NEXT: pshufb {{.*#+}} xmm2 = xmm2[0,1,2,3,4,5,6,7,8,9,2,3,8,9,14,15] 909; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 910; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3] 911; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7] 912; SSE42-NEXT: movdqu %xmm4, (%rsi) 913; SSE42-NEXT: movdqu %xmm5, (%rdx) 914; SSE42-NEXT: movdqu %xmm1, (%rcx) 915; SSE42-NEXT: retq 916; 917; AVX1-LABEL: interleave_24i16_out: 918; AVX1: # BB#0: 919; AVX1-NEXT: vmovdqu 32(%rdi), %xmm0 920; AVX1-NEXT: vmovdqu (%rdi), %ymm1 921; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2 922; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm2[1],xmm1[2,3],xmm2[4],xmm1[5,6],xmm2[7] 923; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,12,13,2,3,8,9,14,15,12,13,14,15] 924; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm0[0,1,2,1] 925; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6,5] 926; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5],xmm4[6,7] 927; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7] 928; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[2,3,8,9,14,15,4,5,10,11,10,11,8,9,14,15] 929; AVX1-NEXT: vpshufb {{.*#+}} xmm5 = xmm0[0,1,6,7,4,5,6,7,0,1,0,1,6,7,12,13] 930; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4],xmm5[5,6,7] 931; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2],xmm2[3,4],xmm1[5],xmm2[6,7] 932; AVX1-NEXT: vpshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,14,15,0,1,2,3] 933; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,6,7,8,9,2,3,8,9,14,15] 934; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3,4],xmm0[5,6,7] 935; AVX1-NEXT: vmovdqu %xmm3, (%rsi) 936; AVX1-NEXT: vmovdqu %xmm4, (%rdx) 937; AVX1-NEXT: vmovdqu %xmm0, (%rcx) 938; AVX1-NEXT: vzeroupper 939; AVX1-NEXT: retq 940; 941; AVX2-LABEL: interleave_24i16_out: 942; AVX2: # BB#0: 943; AVX2-NEXT: vmovdqu (%rdi), %ymm0 944; AVX2-NEXT: vmovdqu 32(%rdi), %xmm1 945; AVX2-NEXT: vpblendw {{.*#+}} ymm2 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7,8,9],ymm1[10],ymm0[11,12],ymm1[13],ymm0[14,15] 946; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm3 947; AVX2-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1],xmm2[2,3],xmm3[4],xmm2[5,6],xmm3[7] 948; AVX2-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11] 949; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = ymm1[0],ymm0[1,2],ymm1[3],ymm0[4,5],ymm1[6],ymm0[7],ymm1[8],ymm0[9,10],ymm1[11],ymm0[12,13],ymm1[14],ymm0[15] 950; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm4 951; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7] 952; AVX2-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13] 953; AVX2-NEXT: vpblendw {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7],ymm0[8],ymm1[9],ymm0[10,11],ymm1[12],ymm0[13,14],ymm1[15] 954; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1 955; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7] 956; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15] 957; AVX2-NEXT: vmovdqu %xmm2, (%rsi) 958; AVX2-NEXT: vmovdqu %xmm3, (%rdx) 959; AVX2-NEXT: vmovdqu %xmm0, (%rcx) 960; AVX2-NEXT: vzeroupper 961; AVX2-NEXT: retq 962 %wide.vec = load <24 x i16>, <24 x i16>* %p, align 4 963 %s1 = shufflevector <24 x i16> %wide.vec, <24 x i16> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 964 %s2 = shufflevector <24 x i16> %wide.vec, <24 x i16> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 965 %s3 = shufflevector <24 x i16> %wide.vec, <24 x i16> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 966 store <8 x i16> %s1, <8 x i16>* %q1, align 4 967 store <8 x i16> %s2, <8 x i16>* %q2, align 4 968 store <8 x i16> %s3, <8 x i16>* %q3, align 4 969 ret void 970} 971 972define void @interleave_24i16_in(<24 x i16>* %p, <8 x i16>* %q1, <8 x i16>* %q2, <8 x i16>* %q3) nounwind { 973; SSE2-LABEL: interleave_24i16_in: 974; SSE2: # BB#0: 975; SSE2-NEXT: movdqu (%rsi), %xmm3 976; SSE2-NEXT: movdqu (%rdx), %xmm2 977; SSE2-NEXT: movdqu (%rcx), %xmm1 978; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,0,0,3] 979; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,0,65535,65535] 980; SSE2-NEXT: movdqa %xmm0, %xmm5 981; SSE2-NEXT: pandn %xmm4, %xmm5 982; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[0,3,3,3] 983; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm3[1,1,2,2] 984; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3] 985; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1] 986; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7] 987; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,4,5] 988; SSE2-NEXT: pand %xmm0, %xmm3 989; SSE2-NEXT: por %xmm5, %xmm3 990; SSE2-NEXT: movdqa %xmm0, %xmm5 991; SSE2-NEXT: pandn %xmm4, %xmm5 992; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm2[0,1,3,3,4,5,6,7] 993; SSE2-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] 994; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3] 995; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,3,2,0,4,5,6,7] 996; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,6,7] 997; SSE2-NEXT: pand %xmm0, %xmm2 998; SSE2-NEXT: por %xmm5, %xmm2 999; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0] 1000; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,2,2] 1001; SSE2-NEXT: pand %xmm5, %xmm1 1002; SSE2-NEXT: pandn %xmm6, %xmm5 1003; SSE2-NEXT: por %xmm1, %xmm5 1004; SSE2-NEXT: pand %xmm0, %xmm5 1005; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,4,6,7] 1006; SSE2-NEXT: pandn %xmm1, %xmm0 1007; SSE2-NEXT: por %xmm5, %xmm0 1008; SSE2-NEXT: movdqu %xmm0, 16(%rdi) 1009; SSE2-NEXT: movdqu %xmm2, 32(%rdi) 1010; SSE2-NEXT: movdqu %xmm3, (%rdi) 1011; SSE2-NEXT: retq 1012; 1013; SSE42-LABEL: interleave_24i16_in: 1014; SSE42: # BB#0: 1015; SSE42-NEXT: movdqu (%rsi), %xmm0 1016; SSE42-NEXT: movdqu (%rdx), %xmm1 1017; SSE42-NEXT: movdqu (%rcx), %xmm2 1018; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,2,2] 1019; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,3,3,3] 1020; SSE42-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 1021; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11] 1022; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,0,3] 1023; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm0[0,1],xmm5[2],xmm0[3,4],xmm5[5],xmm0[6,7] 1024; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,2] 1025; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6],xmm3[7] 1026; SSE42-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[0,1,3,3,4,5,6,7] 1027; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,6,7] 1028; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm0[0,1],xmm3[2],xmm0[3,4],xmm3[5],xmm0[6,7] 1029; SSE42-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7] 1030; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[4,5,6,7,4,5,8,9,10,11,10,11,12,13,14,15] 1031; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm1[0,1],xmm4[2],xmm1[3,4],xmm4[5],xmm1[6,7] 1032; SSE42-NEXT: movdqu %xmm4, 32(%rdi) 1033; SSE42-NEXT: movdqu %xmm3, 16(%rdi) 1034; SSE42-NEXT: movdqu %xmm5, (%rdi) 1035; SSE42-NEXT: retq 1036; 1037; AVX1-LABEL: interleave_24i16_in: 1038; AVX1: # BB#0: 1039; AVX1-NEXT: vmovdqu (%rsi), %xmm0 1040; AVX1-NEXT: vmovdqu (%rdx), %xmm1 1041; AVX1-NEXT: vmovdqu (%rcx), %xmm2 1042; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,2,2] 1043; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm1[0,1,3,3,4,5,6,7] 1044; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,6,7] 1045; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7] 1046; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2] 1047; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7] 1048; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 1049; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,4,5,6,7,10,11,8,9,10,11] 1050; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[0,0,0,3] 1051; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2],xmm4[3,4],xmm5[5],xmm4[6,7] 1052; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3 1053; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1054; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15] 1055; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3] 1056; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1057; AVX1-NEXT: vmovdqu %xmm0, 32(%rdi) 1058; AVX1-NEXT: vmovups %ymm3, (%rdi) 1059; AVX1-NEXT: vzeroupper 1060; AVX1-NEXT: retq 1061; 1062; AVX2-LABEL: interleave_24i16_in: 1063; AVX2: # BB#0: 1064; AVX2-NEXT: vmovdqu (%rsi), %xmm0 1065; AVX2-NEXT: vmovdqu (%rdx), %xmm1 1066; AVX2-NEXT: vmovdqu (%rcx), %xmm2 1067; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 1068; AVX2-NEXT: vpshufb {{.*#+}} ymm4 = ymm3[0,1,2,3,6,7,2,3,8,9,8,9,4,5,6,7,16,17,18,19,22,23,18,19,24,25,24,25,20,21,22,23] 1069; AVX2-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm3[2,3,0,1] 1070; AVX2-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[u,u,0,1,u,u,u,u,2,3,u,u,u,u,4,5,u,u,22,23,u,u,u,u,24,25,u,u,u,u,26,27] 1071; AVX2-NEXT: vpblendw {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7],ymm4[8],ymm3[9],ymm4[10,11],ymm3[12],ymm4[13,14],ymm3[15] 1072; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2> 1073; AVX2-NEXT: vpermd %ymm2, %ymm4, %ymm4 1074; AVX2-NEXT: vmovdqa {{.*#+}} ymm5 = [255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255,255,255,0,0,255,255] 1075; AVX2-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 1076; AVX2-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1077; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,10,11,8,9,8,9,14,15,12,13,14,15] 1078; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3] 1079; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1080; AVX2-NEXT: vmovdqu %xmm0, 32(%rdi) 1081; AVX2-NEXT: vmovdqu %ymm3, (%rdi) 1082; AVX2-NEXT: vzeroupper 1083; AVX2-NEXT: retq 1084 %s1 = load <8 x i16>, <8 x i16>* %q1, align 4 1085 %s2 = load <8 x i16>, <8 x i16>* %q2, align 4 1086 %s3 = load <8 x i16>, <8 x i16>* %q3, align 4 1087 %t1 = shufflevector <8 x i16> %s1, <8 x i16> %s2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 1088 %t2 = shufflevector <8 x i16> %s3, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 1089 %interleaved = shufflevector <16 x i16> %t1, <16 x i16> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 1090 store <24 x i16> %interleaved, <24 x i16>* %p, align 4 1091 ret void 1092} 1093 1094define void @interleave_24i32_out(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2, <8 x i32>* %q3) nounwind { 1095; SSE2-LABEL: interleave_24i32_out: 1096; SSE2: # BB#0: 1097; SSE2-NEXT: movups 80(%rdi), %xmm5 1098; SSE2-NEXT: movups 64(%rdi), %xmm8 1099; SSE2-NEXT: movups (%rdi), %xmm0 1100; SSE2-NEXT: movups 16(%rdi), %xmm6 1101; SSE2-NEXT: movups 32(%rdi), %xmm2 1102; SSE2-NEXT: movups 48(%rdi), %xmm1 1103; SSE2-NEXT: movaps %xmm1, %xmm3 1104; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm8[2,3] 1105; SSE2-NEXT: movaps %xmm5, %xmm4 1106; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[1,0],xmm3[2,0] 1107; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm4[2,0] 1108; SSE2-NEXT: movaps %xmm0, %xmm4 1109; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm6[2,3] 1110; SSE2-NEXT: movaps %xmm2, %xmm7 1111; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm4[2,0] 1112; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm7[2,0] 1113; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm0[2,3,0,1] 1114; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm6[0,0] 1115; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm6[3,3] 1116; SSE2-NEXT: pshufd {{.*#+}} xmm9 = xmm2[0,1,0,3] 1117; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm0[2,0] 1118; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,0] 1119; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,3,0,1] 1120; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,0],xmm8[0,0] 1121; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm8[3,3] 1122; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm5[0,1,0,3] 1123; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm1[2,0] 1124; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,0] 1125; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm6[1,1,2,3] 1126; SSE2-NEXT: punpckldq {{.*#+}} xmm10 = xmm10[0],xmm5[0],xmm10[1],xmm5[1] 1127; SSE2-NEXT: movsd {{.*#+}} xmm9 = xmm10[0],xmm9[1] 1128; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm8[1,1,2,3] 1129; SSE2-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1] 1130; SSE2-NEXT: movsd {{.*#+}} xmm7 = xmm2[0],xmm7[1] 1131; SSE2-NEXT: movups %xmm3, 16(%rsi) 1132; SSE2-NEXT: movups %xmm4, (%rsi) 1133; SSE2-NEXT: movups %xmm1, 16(%rdx) 1134; SSE2-NEXT: movups %xmm0, (%rdx) 1135; SSE2-NEXT: movupd %xmm7, 16(%rcx) 1136; SSE2-NEXT: movupd %xmm9, (%rcx) 1137; SSE2-NEXT: retq 1138; 1139; SSE42-LABEL: interleave_24i32_out: 1140; SSE42: # BB#0: 1141; SSE42-NEXT: movdqu 80(%rdi), %xmm9 1142; SSE42-NEXT: movdqu 64(%rdi), %xmm10 1143; SSE42-NEXT: movdqu (%rdi), %xmm4 1144; SSE42-NEXT: movdqu 16(%rdi), %xmm2 1145; SSE42-NEXT: movdqu 32(%rdi), %xmm11 1146; SSE42-NEXT: movdqu 48(%rdi), %xmm5 1147; SSE42-NEXT: pshufd {{.*#+}} xmm8 = xmm11[0,1,0,1] 1148; SSE42-NEXT: movdqa %xmm2, %xmm7 1149; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm4[2,3],xmm7[4,5,6,7] 1150; SSE42-NEXT: pshufd {{.*#+}} xmm6 = xmm4[2,3,0,1] 1151; SSE42-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm2[2,3] 1152; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm8[6,7] 1153; SSE42-NEXT: movdqa %xmm10, %xmm1 1154; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3],xmm1[4,5,6,7] 1155; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,3,0,1] 1156; SSE42-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,3],xmm10[2,3] 1157; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm9[0,1,0,1] 1158; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm5[0,1,2,3,4,5],xmm3[6,7] 1159; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm11[0,1,2,2] 1160; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,0,3,3] 1161; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5],xmm5[6,7] 1162; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,0,3,3] 1163; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm9[0,1,2,2] 1164; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm1[0,1,2,3,4,5],xmm5[6,7] 1165; SSE42-NEXT: pblendw {{.*#+}} xmm6 = xmm6[0,1],xmm2[2,3],xmm6[4,5,6,7] 1166; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm11[0,1,0,3] 1167; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm6[0,1,2,3],xmm1[4,5,6,7] 1168; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm10[2,3],xmm0[4,5,6,7] 1169; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm9[0,1,0,3] 1170; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm0[0,1,2,3],xmm2[4,5,6,7] 1171; SSE42-NEXT: movdqu %xmm3, 16(%rsi) 1172; SSE42-NEXT: movdqu %xmm4, (%rsi) 1173; SSE42-NEXT: movdqu %xmm5, 16(%rdx) 1174; SSE42-NEXT: movdqu %xmm7, (%rdx) 1175; SSE42-NEXT: movdqu %xmm2, 16(%rcx) 1176; SSE42-NEXT: movdqu %xmm1, (%rcx) 1177; SSE42-NEXT: retq 1178; 1179; AVX1-LABEL: interleave_24i32_out: 1180; AVX1: # BB#0: 1181; AVX1-NEXT: vmovups (%rdi), %ymm0 1182; AVX1-NEXT: vmovups 32(%rdi), %ymm1 1183; AVX1-NEXT: vmovups 64(%rdi), %ymm2 1184; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3 1185; AVX1-NEXT: vinsertps {{.*#+}} xmm4 = zero,zero,xmm2[2],xmm3[1] 1186; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm0, %ymm4 1187; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7] 1188; AVX1-NEXT: vextractf128 $1, %ymm5, %xmm6 1189; AVX1-NEXT: vblendps {{.*#+}} xmm5 = xmm5[0,1],xmm6[2],xmm5[3] 1190; AVX1-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[0,3,2,1] 1191; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[0,3,2,3] 1192; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm5, %ymm5 1193; AVX1-NEXT: vblendpd {{.*#+}} ymm4 = ymm5[0,1,2],ymm4[3] 1194; AVX1-NEXT: vblendps {{.*#+}} xmm5 = xmm2[0,1],xmm3[2],xmm2[3] 1195; AVX1-NEXT: vpermilps {{.*#+}} xmm5 = xmm5[0,0,3,2] 1196; AVX1-NEXT: vinsertf128 $1, %xmm5, %ymm0, %ymm5 1197; AVX1-NEXT: vblendps {{.*#+}} ymm6 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1198; AVX1-NEXT: vextractf128 $1, %ymm6, %xmm7 1199; AVX1-NEXT: vblendps {{.*#+}} xmm6 = xmm7[0],xmm6[1,2],xmm7[3] 1200; AVX1-NEXT: vpermilps {{.*#+}} xmm6 = xmm6[1,0,3,2] 1201; AVX1-NEXT: vmovshdup {{.*#+}} xmm7 = xmm7[1,1,3,3] 1202; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6 1203; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm6[0,1,2,3,4],ymm5[5,6,7] 1204; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm2[0,1],xmm3[0,3] 1205; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm2 1206; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7] 1207; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 1208; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3] 1209; AVX1-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[2,1,0,3] 1210; AVX1-NEXT: vpermilpd {{.*#+}} xmm1 = xmm1[1,0] 1211; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 1212; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] 1213; AVX1-NEXT: vmovupd %ymm4, (%rsi) 1214; AVX1-NEXT: vmovups %ymm5, (%rdx) 1215; AVX1-NEXT: vmovups %ymm0, (%rcx) 1216; AVX1-NEXT: vzeroupper 1217; AVX1-NEXT: retq 1218; 1219; AVX2-LABEL: interleave_24i32_out: 1220; AVX2: # BB#0: 1221; AVX2-NEXT: vmovdqu (%rdi), %ymm0 1222; AVX2-NEXT: vmovdqu 32(%rdi), %ymm1 1223; AVX2-NEXT: vmovdqu 64(%rdi), %ymm2 1224; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = <u,u,u,u,u,u,2,5> 1225; AVX2-NEXT: vpermd %ymm2, %ymm3, %ymm3 1226; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7] 1227; AVX2-NEXT: vmovdqa {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u> 1228; AVX2-NEXT: vpermd %ymm4, %ymm5, %ymm4 1229; AVX2-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7] 1230; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = <u,u,u,u,u,0,3,6> 1231; AVX2-NEXT: vpermd %ymm2, %ymm4, %ymm4 1232; AVX2-NEXT: vpblendd {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1233; AVX2-NEXT: vmovdqa {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u> 1234; AVX2-NEXT: vpermd %ymm5, %ymm6, %ymm5 1235; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7] 1236; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7] 1237; AVX2-NEXT: vmovdqa {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u> 1238; AVX2-NEXT: vpermd %ymm0, %ymm1, %ymm0 1239; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7] 1240; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,1,0,3] 1241; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] 1242; AVX2-NEXT: vmovdqu %ymm3, (%rsi) 1243; AVX2-NEXT: vmovdqu %ymm4, (%rdx) 1244; AVX2-NEXT: vmovdqu %ymm0, (%rcx) 1245; AVX2-NEXT: vzeroupper 1246; AVX2-NEXT: retq 1247 %wide.vec = load <24 x i32>, <24 x i32>* %p, align 4 1248 %s1 = shufflevector <24 x i32> %wide.vec, <24 x i32> undef, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 1249 %s2 = shufflevector <24 x i32> %wide.vec, <24 x i32> undef, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 1250 %s3 = shufflevector <24 x i32> %wide.vec, <24 x i32> undef, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 1251 store <8 x i32> %s1, <8 x i32>* %q1, align 4 1252 store <8 x i32> %s2, <8 x i32>* %q2, align 4 1253 store <8 x i32> %s3, <8 x i32>* %q3, align 4 1254 ret void 1255} 1256 1257define void @interleave_24i32_in(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2, <8 x i32>* %q3) nounwind { 1258; SSE2-LABEL: interleave_24i32_in: 1259; SSE2: # BB#0: 1260; SSE2-NEXT: movdqu (%rsi), %xmm5 1261; SSE2-NEXT: movdqu 16(%rsi), %xmm2 1262; SSE2-NEXT: movdqu (%rdx), %xmm6 1263; SSE2-NEXT: movdqu 16(%rdx), %xmm1 1264; SSE2-NEXT: movdqu (%rcx), %xmm7 1265; SSE2-NEXT: movdqu 16(%rcx), %xmm4 1266; SSE2-NEXT: movdqa %xmm5, %xmm0 1267; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm6[0],xmm0[1],xmm6[1] 1268; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,1,2,2] 1269; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm7[0,1,0,1] 1270; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm0[3,0] 1271; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm3[0,2] 1272; SSE2-NEXT: movdqa %xmm7, %xmm3 1273; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm6[1,0] 1274; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm6[2,2] 1275; SSE2-NEXT: punpckhdq {{.*#+}} xmm7 = xmm7[2],xmm5[2],xmm7[3],xmm5[3] 1276; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[3,0] 1277; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm5[0,2] 1278; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm7[0,3,2,2] 1279; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[2,2,3,3] 1280; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,0],xmm5[3,0] 1281; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,1],xmm6[0,2] 1282; SSE2-NEXT: movdqa %xmm2, %xmm6 1283; SSE2-NEXT: punpckldq {{.*#+}} xmm6 = xmm6[0],xmm1[0],xmm6[1],xmm1[1] 1284; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[0,1,2,2] 1285; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm4[0,1,0,1] 1286; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm6[3,0] 1287; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm7[0,2] 1288; SSE2-NEXT: movdqa %xmm4, %xmm7 1289; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[1,0],xmm1[1,0] 1290; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[2,0],xmm1[2,2] 1291; SSE2-NEXT: punpckhdq {{.*#+}} xmm4 = xmm4[2],xmm2[2],xmm4[3],xmm2[3] 1292; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm7[3,0] 1293; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm2[0,2] 1294; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[0,3,2,2] 1295; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3] 1296; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[3,0] 1297; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[0,2] 1298; SSE2-NEXT: movups %xmm2, 80(%rdi) 1299; SSE2-NEXT: movups %xmm7, 64(%rdi) 1300; SSE2-NEXT: movups %xmm6, 48(%rdi) 1301; SSE2-NEXT: movups %xmm5, 32(%rdi) 1302; SSE2-NEXT: movups %xmm3, 16(%rdi) 1303; SSE2-NEXT: movups %xmm0, (%rdi) 1304; SSE2-NEXT: retq 1305; 1306; SSE42-LABEL: interleave_24i32_in: 1307; SSE42: # BB#0: 1308; SSE42-NEXT: movdqu (%rsi), %xmm5 1309; SSE42-NEXT: movdqu 16(%rsi), %xmm2 1310; SSE42-NEXT: movdqu (%rdx), %xmm6 1311; SSE42-NEXT: movdqu 16(%rdx), %xmm1 1312; SSE42-NEXT: movdqu (%rcx), %xmm7 1313; SSE42-NEXT: movdqu 16(%rcx), %xmm4 1314; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm6[0,0,1,1] 1315; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm5[0,1,0,1] 1316; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm0[2,3],xmm3[4,5,6,7] 1317; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm7[0,1,0,1] 1318; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm3[0,1,2,3],xmm0[4,5],xmm3[6,7] 1319; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm6[1,1,2,2] 1320; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm7[2,3],xmm3[4,5,6,7] 1321; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1,2,3],xmm5[4,5],xmm3[6,7] 1322; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm5[2,3,0,1] 1323; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm7[2,3,2,3] 1324; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm5[2,3],xmm7[4,5,6,7] 1325; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm6[2,2,3,3] 1326; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm7[0,1,2,3],xmm5[4,5],xmm7[6,7] 1327; SSE42-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,1,1] 1328; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm2[0,1,0,1] 1329; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm6[2,3],xmm7[4,5,6,7] 1330; SSE42-NEXT: pshufd {{.*#+}} xmm6 = xmm4[0,1,0,1] 1331; SSE42-NEXT: pblendw {{.*#+}} xmm6 = xmm7[0,1,2,3],xmm6[4,5],xmm7[6,7] 1332; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm1[1,1,2,2] 1333; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm4[2,3],xmm7[4,5,6,7] 1334; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1,2,3],xmm2[4,5],xmm7[6,7] 1335; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,0,1] 1336; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm4[2,3,2,3] 1337; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3],xmm4[4,5,6,7] 1338; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,2,3,3] 1339; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm4[0,1,2,3],xmm1[4,5],xmm4[6,7] 1340; SSE42-NEXT: movdqu %xmm1, 80(%rdi) 1341; SSE42-NEXT: movdqu %xmm7, 64(%rdi) 1342; SSE42-NEXT: movdqu %xmm6, 48(%rdi) 1343; SSE42-NEXT: movdqu %xmm5, 32(%rdi) 1344; SSE42-NEXT: movdqu %xmm3, 16(%rdi) 1345; SSE42-NEXT: movdqu %xmm0, (%rdi) 1346; SSE42-NEXT: retq 1347; 1348; AVX1-LABEL: interleave_24i32_in: 1349; AVX1: # BB#0: 1350; AVX1-NEXT: vmovups (%rsi), %ymm0 1351; AVX1-NEXT: vmovups (%rdx), %ymm1 1352; AVX1-NEXT: vmovupd (%rcx), %ymm2 1353; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm0[2,0],xmm1[2,0] 1354; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm1[1,1],xmm3[0,2] 1355; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm1[0,0],xmm0[0,0] 1356; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,0],xmm0[2,1] 1357; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3 1358; AVX1-NEXT: vmovddup {{.*#+}} xmm4 = xmm2[0,0] 1359; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm4, %ymm4 1360; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7] 1361; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm4 1362; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5 1363; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm5[3,0],xmm4[3,0] 1364; AVX1-NEXT: vshufps {{.*#+}} xmm6 = xmm4[2,1],xmm6[0,2] 1365; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm4[1,0],xmm5[1,0] 1366; AVX1-NEXT: vshufps {{.*#+}} xmm4 = xmm4[2,0],xmm5[2,2] 1367; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm4, %ymm4 1368; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = ymm0[1,1,3,3] 1369; AVX1-NEXT: vperm2f128 {{.*#+}} ymm5 = ymm5[2,3,2,3] 1370; AVX1-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7] 1371; AVX1-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,2,2] 1372; AVX1-NEXT: vpermilpd {{.*#+}} ymm2 = ymm2[1,1,2,2] 1373; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7] 1374; AVX1-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7] 1375; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1376; AVX1-NEXT: vmovups %ymm0, 32(%rdi) 1377; AVX1-NEXT: vmovups %ymm4, 64(%rdi) 1378; AVX1-NEXT: vmovups %ymm3, (%rdi) 1379; AVX1-NEXT: vzeroupper 1380; AVX1-NEXT: retq 1381; 1382; AVX2-LABEL: interleave_24i32_in: 1383; AVX2: # BB#0: 1384; AVX2-NEXT: vmovdqu (%rsi), %ymm0 1385; AVX2-NEXT: vmovdqu (%rdx), %ymm1 1386; AVX2-NEXT: vmovdqu (%rcx), %ymm2 1387; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm1[1,0,2,2] 1388; AVX2-NEXT: vperm2i128 {{.*#+}} ymm3 = ymm3[0,1,0,1] 1389; AVX2-NEXT: vpermq {{.*#+}} ymm4 = ymm0[0,0,2,1] 1390; AVX2-NEXT: vpblendd {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7] 1391; AVX2-NEXT: vpbroadcastq %xmm2, %ymm4 1392; AVX2-NEXT: vpblendd {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7] 1393; AVX2-NEXT: vpermq {{.*#+}} ymm4 = ymm2[2,1,3,3] 1394; AVX2-NEXT: vpshufd {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7] 1395; AVX2-NEXT: vpermq {{.*#+}} ymm5 = ymm5[2,2,2,3] 1396; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7] 1397; AVX2-NEXT: vpbroadcastq 24(%rsi), %ymm5 1398; AVX2-NEXT: vpblendd {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7] 1399; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,1,2,2] 1400; AVX2-NEXT: vpermq {{.*#+}} ymm2 = ymm2[1,1,2,2] 1401; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7] 1402; AVX2-NEXT: vpshufd {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7] 1403; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1404; AVX2-NEXT: vmovdqu %ymm0, 32(%rdi) 1405; AVX2-NEXT: vmovdqu %ymm4, 64(%rdi) 1406; AVX2-NEXT: vmovdqu %ymm3, (%rdi) 1407; AVX2-NEXT: vzeroupper 1408; AVX2-NEXT: retq 1409 %s1 = load <8 x i32>, <8 x i32>* %q1, align 4 1410 %s2 = load <8 x i32>, <8 x i32>* %q2, align 4 1411 %s3 = load <8 x i32>, <8 x i32>* %q3, align 4 1412 %t1 = shufflevector <8 x i32> %s1, <8 x i32> %s2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 1413 %t2 = shufflevector <8 x i32> %s3, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 1414 %interleaved = shufflevector <16 x i32> %t1, <16 x i32> %t2, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 1415 store <24 x i32> %interleaved, <24 x i32>* %p, align 4 1416 ret void 1417} 1418 1419define <2 x double> @wrongorder(<4 x double> %A, <8 x double>* %P) #0 { 1420; SSE2-LABEL: wrongorder: 1421; SSE2: # BB#0: 1422; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0,0] 1423; SSE2-NEXT: movaps %xmm0, 48(%rdi) 1424; SSE2-NEXT: movaps %xmm0, 32(%rdi) 1425; SSE2-NEXT: movaps %xmm0, 16(%rdi) 1426; SSE2-NEXT: movaps %xmm0, (%rdi) 1427; SSE2-NEXT: retq 1428; 1429; SSE42-LABEL: wrongorder: 1430; SSE42: # BB#0: 1431; SSE42-NEXT: movddup {{.*#+}} xmm0 = xmm0[0,0] 1432; SSE42-NEXT: movapd %xmm0, 48(%rdi) 1433; SSE42-NEXT: movapd %xmm0, 32(%rdi) 1434; SSE42-NEXT: movapd %xmm0, 16(%rdi) 1435; SSE42-NEXT: movapd %xmm0, (%rdi) 1436; SSE42-NEXT: retq 1437; 1438; AVX1-LABEL: wrongorder: 1439; AVX1: # BB#0: 1440; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] 1441; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1 1442; AVX1-NEXT: vmovaps %ymm1, 32(%rdi) 1443; AVX1-NEXT: vmovaps %ymm1, (%rdi) 1444; AVX1-NEXT: # kill: %XMM0<def> %XMM0<kill> %YMM0<kill> 1445; AVX1-NEXT: vzeroupper 1446; AVX1-NEXT: retq 1447; 1448; AVX2-LABEL: wrongorder: 1449; AVX2: # BB#0: 1450; AVX2-NEXT: vbroadcastsd %xmm0, %ymm1 1451; AVX2-NEXT: vmovapd %ymm1, 32(%rdi) 1452; AVX2-NEXT: vmovapd %ymm1, (%rdi) 1453; AVX2-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] 1454; AVX2-NEXT: vzeroupper 1455; AVX2-NEXT: retq 1456 %shuffle = shufflevector <4 x double> %A, <4 x double> %A, <8 x i32> zeroinitializer 1457 store <8 x double> %shuffle, <8 x double>* %P, align 64 1458 %m2 = load <8 x double>, <8 x double>* %P, align 64 1459 store <8 x double> %m2, <8 x double>* %P, align 64 1460 %m3 = load <8 x double>, <8 x double>* %P, align 64 1461 %m4 = shufflevector <8 x double> %m3, <8 x double> undef, <2 x i32> <i32 2, i32 0> 1462 ret <2 x double> %m4 1463} 1464