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-prefixes=AVX,AVX2,AVX2-SLOW 6; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+avx2,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX2-FAST,AVX2-FAST-ALL 7; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+avx2,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX,AVX2,AVX2-FAST,AVX2-FAST-PERLANE 8; RUN: llc < %s -mtriple=x86_64-pc-linux -mattr=+xop | FileCheck %s --check-prefix=XOP 9 10define void @v3i64(<2 x i64> %a, <2 x i64> %b, <3 x i64>* %p) nounwind { 11; SSE2-LABEL: v3i64: 12; SSE2: # %bb.0: 13; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[2,3,2,3] 14; SSE2-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 15; SSE2-NEXT: movq %xmm2, 16(%rdi) 16; SSE2-NEXT: movdqa %xmm0, (%rdi) 17; SSE2-NEXT: retq 18; 19; SSE42-LABEL: v3i64: 20; SSE42: # %bb.0: 21; SSE42-NEXT: pextrq $1, %xmm0, 16(%rdi) 22; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 23; SSE42-NEXT: movdqa %xmm0, (%rdi) 24; SSE42-NEXT: retq 25; 26; AVX-LABEL: v3i64: 27; AVX: # %bb.0: 28; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm0[0],xmm1[0] 29; AVX-NEXT: vpextrq $1, %xmm0, 16(%rdi) 30; AVX-NEXT: vmovdqa %xmm1, (%rdi) 31; AVX-NEXT: retq 32; 33; XOP-LABEL: v3i64: 34; XOP: # %bb.0: 35; XOP-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm0[0],xmm1[0] 36; XOP-NEXT: vpextrq $1, %xmm0, 16(%rdi) 37; XOP-NEXT: vmovdqa %xmm1, (%rdi) 38; XOP-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: movhps %xmm0, 16(%rdi) 47; SSE-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0] 48; SSE-NEXT: movaps %xmm0, (%rdi) 49; SSE-NEXT: retq 50; 51; AVX-LABEL: v3f64: 52; AVX: # %bb.0: 53; AVX-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0] 54; AVX-NEXT: vmovhps %xmm0, 16(%rdi) 55; AVX-NEXT: vmovaps %xmm1, (%rdi) 56; AVX-NEXT: retq 57; 58; XOP-LABEL: v3f64: 59; XOP: # %bb.0: 60; XOP-NEXT: vmovlhps {{.*#+}} xmm1 = xmm0[0],xmm1[0] 61; XOP-NEXT: vmovhps %xmm0, 16(%rdi) 62; XOP-NEXT: vmovaps %xmm1, (%rdi) 63; XOP-NEXT: retq 64 %r = shufflevector <2 x double> %a, <2 x double> %b, <3 x i32> <i32 0, i32 2, i32 1> 65 store <3 x double> %r, <3 x double>* %p 66 ret void 67} 68 69define void @v3i32(<2 x i32> %a, <2 x i32> %b, <3 x i32>* %p) nounwind { 70; SSE2-LABEL: v3i32: 71; SSE2: # %bb.0: 72; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[1,1,1,1] 73; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 74; SSE2-NEXT: movd %xmm2, 8(%rdi) 75; SSE2-NEXT: movq %xmm0, (%rdi) 76; SSE2-NEXT: retq 77; 78; SSE42-LABEL: v3i32: 79; SSE42: # %bb.0: 80; SSE42-NEXT: extractps $1, %xmm0, 8(%rdi) 81; SSE42-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 82; SSE42-NEXT: movlps %xmm0, (%rdi) 83; SSE42-NEXT: retq 84; 85; AVX-LABEL: v3i32: 86; AVX: # %bb.0: 87; AVX-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 88; AVX-NEXT: vextractps $1, %xmm0, 8(%rdi) 89; AVX-NEXT: vmovlps %xmm1, (%rdi) 90; AVX-NEXT: retq 91; 92; XOP-LABEL: v3i32: 93; XOP: # %bb.0: 94; XOP-NEXT: vunpcklps {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 95; XOP-NEXT: vextractps $1, %xmm0, 8(%rdi) 96; XOP-NEXT: vmovlps %xmm1, (%rdi) 97; XOP-NEXT: retq 98 %r = shufflevector <2 x i32> %a, <2 x i32> %b, <3 x i32> <i32 0, i32 2, i32 1> 99 store <3 x i32> %r, <3 x i32>* %p 100 ret void 101} 102 103define void @v5i16(<4 x i16> %a, <4 x i16> %b, <5 x i16>* %p) nounwind { 104; SSE2-LABEL: v5i16: 105; SSE2: # %bb.0: 106; SSE2-NEXT: psrlq $16, %xmm1 107; SSE2-NEXT: pextrw $3, %xmm0, %eax 108; SSE2-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 109; SSE2-NEXT: movw %ax, 8(%rdi) 110; SSE2-NEXT: movq %xmm0, (%rdi) 111; SSE2-NEXT: retq 112; 113; SSE42-LABEL: v5i16: 114; SSE42: # %bb.0: 115; SSE42-NEXT: psrlq $16, %xmm1 116; SSE42-NEXT: pextrw $3, %xmm0, 8(%rdi) 117; SSE42-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 118; SSE42-NEXT: movq %xmm0, (%rdi) 119; SSE42-NEXT: retq 120; 121; AVX-LABEL: v5i16: 122; AVX: # %bb.0: 123; AVX-NEXT: vpsrlq $16, %xmm1, %xmm1 124; AVX-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 125; AVX-NEXT: vpextrw $3, %xmm0, 8(%rdi) 126; AVX-NEXT: vmovq %xmm1, (%rdi) 127; AVX-NEXT: retq 128; 129; XOP-LABEL: v5i16: 130; XOP: # %bb.0: 131; XOP-NEXT: vpsrlq $16, %xmm1, %xmm1 132; XOP-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 133; XOP-NEXT: vpextrw $3, %xmm0, 8(%rdi) 134; XOP-NEXT: vmovq %xmm1, (%rdi) 135; XOP-NEXT: retq 136 %r = shufflevector <4 x i16> %a, <4 x i16> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 137 store <5 x i16> %r, <5 x i16>* %p 138 ret void 139} 140 141define void @v5i32(<4 x i32> %a, <4 x i32> %b, <5 x i32>* %p) nounwind { 142; SSE2-LABEL: v5i32: 143; SSE2: # %bb.0: 144; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,2,2,3] 145; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3] 146; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 147; SSE2-NEXT: movd %xmm2, 16(%rdi) 148; SSE2-NEXT: movdqa %xmm0, (%rdi) 149; SSE2-NEXT: retq 150; 151; SSE42-LABEL: v5i32: 152; SSE42: # %bb.0: 153; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,1,2,2] 154; SSE42-NEXT: pmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero 155; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7] 156; SSE42-NEXT: pextrd $3, %xmm0, 16(%rdi) 157; SSE42-NEXT: movdqa %xmm2, (%rdi) 158; SSE42-NEXT: retq 159; 160; AVX1-LABEL: v5i32: 161; AVX1: # %bb.0: 162; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,2] 163; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero 164; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7] 165; AVX1-NEXT: vpextrd $3, %xmm0, 16(%rdi) 166; AVX1-NEXT: vmovdqa %xmm1, (%rdi) 167; AVX1-NEXT: retq 168; 169; AVX2-LABEL: v5i32: 170; AVX2: # %bb.0: 171; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,1,2,2] 172; AVX2-NEXT: vpmovzxdq {{.*#+}} xmm2 = xmm0[0],zero,xmm0[1],zero 173; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm2[0],xmm1[1],xmm2[2],xmm1[3] 174; AVX2-NEXT: vpextrd $3, %xmm0, 16(%rdi) 175; AVX2-NEXT: vmovdqa %xmm1, (%rdi) 176; AVX2-NEXT: retq 177; 178; XOP-LABEL: v5i32: 179; XOP: # %bb.0: 180; XOP-NEXT: vpperm {{.*#+}} xmm1 = xmm0[0,1,2,3],xmm1[4,5,6,7],xmm0[4,5,6,7],xmm1[8,9,10,11] 181; XOP-NEXT: vpextrd $3, %xmm0, 16(%rdi) 182; XOP-NEXT: vmovdqa %xmm1, (%rdi) 183; XOP-NEXT: retq 184 %r = shufflevector <4 x i32> %a, <4 x i32> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 185 store <5 x i32> %r, <5 x i32>* %p 186 ret void 187} 188 189define void @v5f32(<4 x float> %a, <4 x float> %b, <5 x float>* %p) nounwind { 190; SSE2-LABEL: v5f32: 191; SSE2: # %bb.0: 192; SSE2-NEXT: movaps %xmm0, %xmm2 193; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm1[1,2] 194; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2,1,3] 195; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3] 196; SSE2-NEXT: movss %xmm0, 16(%rdi) 197; SSE2-NEXT: movaps %xmm2, (%rdi) 198; SSE2-NEXT: retq 199; 200; SSE42-LABEL: v5f32: 201; SSE42: # %bb.0: 202; SSE42-NEXT: extractps $3, %xmm0, 16(%rdi) 203; SSE42-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm1[1,2] 204; SSE42-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2,1,3] 205; SSE42-NEXT: movaps %xmm0, (%rdi) 206; SSE42-NEXT: retq 207; 208; AVX-LABEL: v5f32: 209; AVX: # %bb.0: 210; AVX-NEXT: vshufps {{.*#+}} xmm1 = xmm0[0,1],xmm1[1,2] 211; AVX-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,3] 212; AVX-NEXT: vextractps $3, %xmm0, 16(%rdi) 213; AVX-NEXT: vmovaps %xmm1, (%rdi) 214; AVX-NEXT: retq 215; 216; XOP-LABEL: v5f32: 217; XOP: # %bb.0: 218; XOP-NEXT: vshufps {{.*#+}} xmm1 = xmm0[0,1],xmm1[1,2] 219; XOP-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[0,2,1,3] 220; XOP-NEXT: vextractps $3, %xmm0, 16(%rdi) 221; XOP-NEXT: vmovaps %xmm1, (%rdi) 222; XOP-NEXT: retq 223 %r = shufflevector <4 x float> %a, <4 x float> %b, <5 x i32> <i32 0, i32 5, i32 1, i32 6, i32 3> 224 store <5 x float> %r, <5 x float>* %p 225 ret void 226} 227 228define void @v7i8(<4 x i8> %a, <4 x i8> %b, <7 x i8>* %p) nounwind { 229; SSE2-LABEL: v7i8: 230; SSE2: # %bb.0: 231; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] 232; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,1,3,4,5,6,7] 233; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [0,255,0,255,0,255,255,255,255,255,255,255,255,255,255,255] 234; SSE2-NEXT: movdqa %xmm1, -{{[0-9]+}}(%rsp) 235; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] 236; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,2,3,0,4,5,6,7] 237; SSE2-NEXT: pand %xmm2, %xmm1 238; SSE2-NEXT: pandn %xmm0, %xmm2 239; SSE2-NEXT: por %xmm1, %xmm2 240; SSE2-NEXT: movb -{{[0-9]+}}(%rsp), %al 241; SSE2-NEXT: movb %al, 6(%rdi) 242; SSE2-NEXT: movd %xmm2, (%rdi) 243; SSE2-NEXT: pextrw $2, %xmm2, %eax 244; SSE2-NEXT: movw %ax, 4(%rdi) 245; SSE2-NEXT: retq 246; 247; SSE42-LABEL: v7i8: 248; SSE42: # %bb.0: 249; SSE42-NEXT: pextrb $0, %xmm1, 6(%rdi) 250; SSE42-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 251; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[1,4,7,4,3,6,0,u,u,u,u,u,u,u,u,u] 252; SSE42-NEXT: pextrw $2, %xmm1, 4(%rdi) 253; SSE42-NEXT: movd %xmm1, (%rdi) 254; SSE42-NEXT: retq 255; 256; AVX-LABEL: v7i8: 257; AVX: # %bb.0: 258; AVX-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 259; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[1,4,7,4,3,6,0,u,u,u,u,u,u,u,u,u] 260; AVX-NEXT: vpextrb $0, %xmm1, 6(%rdi) 261; AVX-NEXT: vpextrw $2, %xmm0, 4(%rdi) 262; AVX-NEXT: vmovd %xmm0, (%rdi) 263; AVX-NEXT: retq 264; 265; XOP-LABEL: v7i8: 266; XOP: # %bb.0: 267; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[0],xmm1[2],xmm0[3],xmm1[2],xmm0[1],xmm1[3,0,u,u,u,u,u,u,u,u,u] 268; XOP-NEXT: vpextrb $0, %xmm1, 6(%rdi) 269; XOP-NEXT: vpextrw $2, %xmm0, 4(%rdi) 270; XOP-NEXT: vmovd %xmm0, (%rdi) 271; XOP-NEXT: retq 272 %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> 273 store <7 x i8> %r, <7 x i8>* %p 274 ret void 275} 276 277define void @v7i16(<4 x i16> %a, <4 x i16> %b, <7 x i16>* %p) nounwind { 278; SSE2-LABEL: v7i16: 279; SSE2: # %bb.0: 280; SSE2-NEXT: movd %xmm1, %eax 281; SSE2-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3] 282; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm1[0,1,0,3,4,5,6,7] 283; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,7] 284; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,1,3] 285; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,2,3,2,4,5,6,7] 286; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm0[0,1,2,3,5,6,4,7] 287; SSE2-NEXT: movw %ax, 12(%rdi) 288; SSE2-NEXT: movq %xmm0, (%rdi) 289; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3] 290; SSE2-NEXT: movd %xmm0, 8(%rdi) 291; SSE2-NEXT: retq 292; 293; SSE42-LABEL: v7i16: 294; SSE42: # %bb.0: 295; SSE42-NEXT: pextrw $0, %xmm1, 12(%rdi) 296; SSE42-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3] 297; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[2,3,8,9,14,15,8,9,6,7,12,13,0,1,14,15] 298; SSE42-NEXT: pextrd $2, %xmm1, 8(%rdi) 299; SSE42-NEXT: movq %xmm1, (%rdi) 300; SSE42-NEXT: retq 301; 302; AVX-LABEL: v7i16: 303; AVX: # %bb.0: 304; AVX-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3] 305; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,3,8,9,14,15,8,9,6,7,12,13,0,1,14,15] 306; AVX-NEXT: vpextrw $0, %xmm1, 12(%rdi) 307; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 308; AVX-NEXT: vmovq %xmm0, (%rdi) 309; AVX-NEXT: retq 310; 311; XOP-LABEL: v7i16: 312; XOP: # %bb.0: 313; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[0,1],xmm1[4,5],xmm0[6,7],xmm1[4,5],xmm0[2,3],xmm1[6,7,0,1],xmm0[6,7] 314; XOP-NEXT: vpextrw $0, %xmm1, 12(%rdi) 315; XOP-NEXT: vpextrd $2, %xmm0, 8(%rdi) 316; XOP-NEXT: vmovq %xmm0, (%rdi) 317; XOP-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[2,2,2,2] 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: pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3] 331; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1] 332; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1] 333; SSE2-NEXT: movd %xmm1, 24(%rdi) 334; SSE2-NEXT: movq %xmm0, 16(%rdi) 335; SSE2-NEXT: movdqa %xmm3, (%rdi) 336; SSE2-NEXT: retq 337; 338; SSE42-LABEL: v7i32: 339; SSE42: # %bb.0: 340; SSE42-NEXT: movdqa %xmm0, %xmm2 341; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm1[4,5],xmm2[6,7] 342; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,3,2] 343; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7] 344; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,3,2,3] 345; SSE42-NEXT: movd %xmm1, 24(%rdi) 346; SSE42-NEXT: movq %xmm0, 16(%rdi) 347; SSE42-NEXT: movdqa %xmm2, (%rdi) 348; SSE42-NEXT: retq 349; 350; AVX-LABEL: v7i32: 351; AVX: # %bb.0: 352; AVX-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm1[2],xmm0[3] 353; AVX-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,2,3,2] 354; AVX-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3] 355; AVX-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3] 356; AVX-NEXT: vmovss %xmm1, 24(%rdi) 357; AVX-NEXT: vmovlps %xmm0, 16(%rdi) 358; AVX-NEXT: vmovaps %xmm2, (%rdi) 359; AVX-NEXT: retq 360; 361; XOP-LABEL: v7i32: 362; XOP: # %bb.0: 363; XOP-NEXT: vblendps {{.*#+}} xmm2 = xmm0[0,1],xmm1[2],xmm0[3] 364; XOP-NEXT: vpermilps {{.*#+}} xmm2 = xmm2[0,2,3,2] 365; XOP-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3] 366; XOP-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3] 367; XOP-NEXT: vmovss %xmm1, 24(%rdi) 368; XOP-NEXT: vmovlps %xmm0, 16(%rdi) 369; XOP-NEXT: vmovaps %xmm2, (%rdi) 370; XOP-NEXT: retq 371 %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> 372 store <7 x i32> %r, <7 x i32>* %p 373 ret void 374} 375 376define void @v12i8(<8 x i8> %a, <8 x i8> %b, <12 x i8>* %p) nounwind { 377; SSE2-LABEL: v12i8: 378; SSE2: # %bb.0: 379; SSE2-NEXT: pxor %xmm2, %xmm2 380; SSE2-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] 381; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm0[3,1,2,3] 382; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,3,1,3,4,5,6,7] 383; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 384; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1] 385; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,1,4,5,6,7] 386; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,6,4] 387; SSE2-NEXT: packuswb %xmm2, %xmm0 388; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,255] 389; SSE2-NEXT: pand %xmm2, %xmm0 390; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,0,1,1,4,5,6,7] 391; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,1,3] 392; SSE2-NEXT: pandn %xmm1, %xmm2 393; SSE2-NEXT: por %xmm0, %xmm2 394; SSE2-NEXT: movq %xmm2, (%rdi) 395; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,2,3] 396; SSE2-NEXT: movd %xmm0, 8(%rdi) 397; SSE2-NEXT: retq 398; 399; SSE42-LABEL: v12i8: 400; SSE42: # %bb.0: 401; SSE42-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] 402; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 403; SSE42-NEXT: pextrd $2, %xmm0, 8(%rdi) 404; SSE42-NEXT: movq %xmm0, (%rdi) 405; SSE42-NEXT: retq 406; 407; AVX-LABEL: v12i8: 408; AVX: # %bb.0: 409; 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] 410; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 411; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 412; AVX-NEXT: vmovq %xmm0, (%rdi) 413; AVX-NEXT: retq 414; 415; XOP-LABEL: v12i8: 416; XOP: # %bb.0: 417; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[0,4],xmm1[0],xmm0[1,5],xmm1[1],xmm0[2,6],xmm1[2],xmm0[3,7],xmm1[3],xmm0[u,u,u,u] 418; XOP-NEXT: vpextrd $2, %xmm0, 8(%rdi) 419; XOP-NEXT: vmovq %xmm0, (%rdi) 420; XOP-NEXT: retq 421 %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> 422 store <12 x i8> %r, <12 x i8>* %p 423 ret void 424} 425 426define void @v12i16(<8 x i16> %a, <8 x i16> %b, <12 x i16>* %p) nounwind { 427; SSE2-LABEL: v12i16: 428; SSE2: # %bb.0: 429; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[0,0,0,0] 430; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [65535,65535,0,65535,65535,0,65535,65535] 431; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm0[0,1,2,3,6,5,4,7] 432; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1] 433; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,2,2,1,4,5,6,7] 434; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,6,4] 435; SSE2-NEXT: pand %xmm3, %xmm4 436; SSE2-NEXT: pandn %xmm2, %xmm3 437; SSE2-NEXT: por %xmm4, %xmm3 438; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1] 439; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [0,65535,65535,0,65535,65535,65535,65535] 440; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,3] 441; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,3,1,3,4,5,6,7] 442; SSE2-NEXT: pand %xmm2, %xmm0 443; SSE2-NEXT: pandn %xmm1, %xmm2 444; SSE2-NEXT: por %xmm0, %xmm2 445; SSE2-NEXT: movq %xmm2, 16(%rdi) 446; SSE2-NEXT: movdqa %xmm3, (%rdi) 447; SSE2-NEXT: retq 448; 449; SSE42-LABEL: v12i16: 450; SSE42: # %bb.0: 451; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,1,1] 452; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 453; SSE42-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 454; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm2[0],xmm3[1,2],xmm2[3],xmm3[4,5,6,7] 455; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,0,0,0] 456; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,2,3,10,11,u,u,4,5,12,13] 457; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 458; SSE42-NEXT: movdqa %xmm0, (%rdi) 459; SSE42-NEXT: movq %xmm3, 16(%rdi) 460; SSE42-NEXT: retq 461; 462; AVX1-LABEL: v12i16: 463; AVX1: # %bb.0: 464; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1] 465; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 466; AVX1-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 467; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3],xmm3[4,5,6,7] 468; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[0,0,0,0] 469; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,2,3,10,11,u,u,4,5,12,13] 470; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 471; AVX1-NEXT: vmovdqa %xmm0, (%rdi) 472; AVX1-NEXT: vmovq %xmm2, 16(%rdi) 473; AVX1-NEXT: retq 474; 475; AVX2-SLOW-LABEL: v12i16: 476; AVX2-SLOW: # %bb.0: 477; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,1,1] 478; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[3,1,2,3] 479; AVX2-SLOW-NEXT: vpshuflw {{.*#+}} xmm3 = xmm3[0,3,1,3,4,5,6,7] 480; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0],xmm3[1,2],xmm2[3],xmm3[4,5,6,7] 481; AVX2-SLOW-NEXT: vpbroadcastd %xmm1, %xmm1 482; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,8,9,u,u,2,3,10,11,u,u,4,5,12,13] 483; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 484; AVX2-SLOW-NEXT: vmovdqa %xmm0, (%rdi) 485; AVX2-SLOW-NEXT: vmovq %xmm2, 16(%rdi) 486; AVX2-SLOW-NEXT: retq 487; 488; AVX2-FAST-LABEL: v12i16: 489; AVX2-FAST: # %bb.0: 490; AVX2-FAST-NEXT: vpbroadcastd %xmm1, %xmm2 491; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[0,1,8,9,u,u,2,3,10,11,u,u,4,5,12,13] 492; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2],xmm3[3,4],xmm2[5],xmm3[6,7] 493; AVX2-FAST-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,1,1] 494; AVX2-FAST-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,6,7,14,15,u,u,u,u,u,u,u,u,u,u] 495; AVX2-FAST-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3],xmm0[4,5,6,7] 496; AVX2-FAST-NEXT: vmovq %xmm0, 16(%rdi) 497; AVX2-FAST-NEXT: vmovdqa %xmm2, (%rdi) 498; AVX2-FAST-NEXT: retq 499; 500; XOP-LABEL: v12i16: 501; XOP: # %bb.0: 502; XOP-NEXT: vpperm {{.*#+}} xmm2 = xmm0[0,1,8,9],xmm1[0,1],xmm0[2,3,10,11],xmm1[2,3],xmm0[4,5,12,13] 503; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm1[4,5],xmm0[6,7,14,15],xmm1[6,7],xmm0[u,u,u,u,u,u,u,u] 504; XOP-NEXT: vmovq %xmm0, 16(%rdi) 505; XOP-NEXT: vmovdqa %xmm2, (%rdi) 506; XOP-NEXT: retq 507 %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> 508 store <12 x i16> %r, <12 x i16>* %p 509 ret void 510} 511 512define void @v12i32(<8 x i32> %a, <8 x i32> %b, <12 x i32>* %p) nounwind { 513; SSE2-LABEL: v12i32: 514; SSE2: # %bb.0: 515; SSE2-NEXT: movaps %xmm2, %xmm3 516; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,1],xmm0[1,3] 517; SSE2-NEXT: movaps %xmm0, %xmm4 518; SSE2-NEXT: unpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1] 519; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,1],xmm3[0,2] 520; SSE2-NEXT: movaps %xmm0, %xmm3 521; SSE2-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm1[1] 522; SSE2-NEXT: movaps %xmm2, %xmm5 523; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,0],xmm1[1,0] 524; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,0],xmm3[0,2] 525; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm2[3,3] 526; SSE2-NEXT: unpckhps {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3] 527; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm1[0,2] 528; SSE2-NEXT: movaps %xmm2, 32(%rdi) 529; SSE2-NEXT: movaps %xmm5, 16(%rdi) 530; SSE2-NEXT: movaps %xmm4, (%rdi) 531; SSE2-NEXT: retq 532; 533; SSE42-LABEL: v12i32: 534; SSE42: # %bb.0: 535; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm1[0,0,1,1] 536; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm0[0,1,0,1] 537; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm3[2,3],xmm4[4,5,6,7] 538; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,0,1] 539; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm4[0,1,2,3],xmm3[4,5],xmm4[6,7] 540; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,1,2,2] 541; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm2[2,3],xmm4[4,5,6,7] 542; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm0[4,5],xmm4[6,7] 543; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3] 544; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3] 545; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm0[2,3],xmm2[4,5,6,7] 546; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,2,3,3] 547; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm2[0,1,2,3],xmm0[4,5],xmm2[6,7] 548; SSE42-NEXT: movdqa %xmm0, 32(%rdi) 549; SSE42-NEXT: movdqa %xmm4, 16(%rdi) 550; SSE42-NEXT: movdqa %xmm3, (%rdi) 551; SSE42-NEXT: retq 552; 553; AVX1-LABEL: v12i32: 554; AVX1: # %bb.0: 555; AVX1-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1] 556; AVX1-NEXT: vmovsldup {{.*#+}} ymm2 = ymm2[0,0,2,2,4,4,6,6] 557; AVX1-NEXT: vpermilps {{.*#+}} ymm3 = ymm0[0,u,u,1,5,u,u,6] 558; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3,4,5],ymm2[6],ymm3[7] 559; AVX1-NEXT: vpermilps {{.*#+}} xmm3 = xmm1[0,1,0,1] 560; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm3 561; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 562; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 563; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3] 564; AVX1-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3] 565; AVX1-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 566; AVX1-NEXT: vmovaps %xmm0, 32(%rdi) 567; AVX1-NEXT: vmovaps %ymm2, (%rdi) 568; AVX1-NEXT: vzeroupper 569; AVX1-NEXT: retq 570; 571; AVX2-SLOW-LABEL: v12i32: 572; AVX2-SLOW: # %bb.0: 573; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6> 574; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm2, %ymm2 575; AVX2-SLOW-NEXT: vbroadcastsd %xmm1, %ymm3 576; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 577; AVX2-SLOW-NEXT: vextractf128 $1, %ymm0, %xmm3 578; AVX2-SLOW-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3] 579; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3] 580; AVX2-SLOW-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 581; AVX2-SLOW-NEXT: vmovaps %xmm0, 32(%rdi) 582; AVX2-SLOW-NEXT: vmovaps %ymm2, (%rdi) 583; AVX2-SLOW-NEXT: vzeroupper 584; AVX2-SLOW-NEXT: retq 585; 586; AVX2-FAST-ALL-LABEL: v12i32: 587; AVX2-FAST-ALL: # %bb.0: 588; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6> 589; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm2, %ymm2 590; AVX2-FAST-ALL-NEXT: vbroadcastsd %xmm1, %ymm3 591; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 592; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm3 = <u,3,7,u,u,u,u,u> 593; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm3, %ymm0 594; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3] 595; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 596; AVX2-FAST-ALL-NEXT: vmovaps %xmm0, 32(%rdi) 597; AVX2-FAST-ALL-NEXT: vmovaps %ymm2, (%rdi) 598; AVX2-FAST-ALL-NEXT: vzeroupper 599; AVX2-FAST-ALL-NEXT: retq 600; 601; AVX2-FAST-PERLANE-LABEL: v12i32: 602; AVX2-FAST-PERLANE: # %bb.0: 603; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm2 = <0,4,u,1,5,u,2,6> 604; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm2, %ymm2 605; AVX2-FAST-PERLANE-NEXT: vbroadcastsd %xmm1, %ymm3 606; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 607; AVX2-FAST-PERLANE-NEXT: vextractf128 $1, %ymm0, %xmm3 608; AVX2-FAST-PERLANE-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3] 609; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3] 610; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 611; AVX2-FAST-PERLANE-NEXT: vmovaps %xmm0, 32(%rdi) 612; AVX2-FAST-PERLANE-NEXT: vmovaps %ymm2, (%rdi) 613; AVX2-FAST-PERLANE-NEXT: vzeroupper 614; AVX2-FAST-PERLANE-NEXT: retq 615; 616; XOP-LABEL: v12i32: 617; XOP: # %bb.0: 618; XOP-NEXT: vperm2f128 {{.*#+}} ymm2 = ymm0[2,3,0,1] 619; XOP-NEXT: vpermil2ps {{.*#+}} ymm2 = ymm0[0],ymm2[0],ymm0[u,1,5,u],ymm2[6],ymm0[6] 620; XOP-NEXT: vpermilps {{.*#+}} xmm3 = xmm1[0,1,0,1] 621; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm3, %ymm3 622; XOP-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 623; XOP-NEXT: vextractf128 $1, %ymm0, %xmm3 624; XOP-NEXT: vshufps {{.*#+}} xmm0 = xmm0[3,3],xmm3[3,3] 625; XOP-NEXT: vpermilps {{.*#+}} xmm1 = xmm1[2,3,2,3] 626; XOP-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3] 627; XOP-NEXT: vmovaps %xmm0, 32(%rdi) 628; XOP-NEXT: vmovaps %ymm2, (%rdi) 629; XOP-NEXT: vzeroupper 630; XOP-NEXT: retq 631 %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> 632 store <12 x i32> %r, <12 x i32>* %p 633 ret void 634} 635 636define void @pr29025(<4 x i8> %a, <4 x i8> %b, <4 x i8> %c, <12 x i8> *%p) nounwind { 637; SSE2-LABEL: pr29025: 638; SSE2: # %bb.0: 639; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 640; SSE2-NEXT: pxor %xmm1, %xmm1 641; 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] 642; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[3,1,2,3] 643; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,1,3,4,5,6,7] 644; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 645; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,3,2,1] 646; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,1,4,5,6,7] 647; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,5,5,6,4] 648; SSE2-NEXT: packuswb %xmm1, %xmm0 649; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,255,255] 650; SSE2-NEXT: pand %xmm1, %xmm0 651; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,1,1,4,5,6,7] 652; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,0,1,3] 653; SSE2-NEXT: pandn %xmm2, %xmm1 654; SSE2-NEXT: por %xmm0, %xmm1 655; SSE2-NEXT: movq %xmm1, (%rdi) 656; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,2,3] 657; SSE2-NEXT: movd %xmm0, 8(%rdi) 658; SSE2-NEXT: retq 659; 660; SSE42-LABEL: pr29025: 661; SSE42: # %bb.0: 662; SSE42-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 663; 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] 664; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 665; SSE42-NEXT: pextrd $2, %xmm0, 8(%rdi) 666; SSE42-NEXT: movq %xmm0, (%rdi) 667; SSE42-NEXT: retq 668; 669; AVX-LABEL: pr29025: 670; AVX: # %bb.0: 671; AVX-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 672; AVX-NEXT: vpunpcklbw {{.*#+}} 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] 673; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,8,1,2,10,3,4,12,5,6,14,7,u,u,u,u] 674; AVX-NEXT: vpextrd $2, %xmm0, 8(%rdi) 675; AVX-NEXT: vmovq %xmm0, (%rdi) 676; AVX-NEXT: retq 677; 678; XOP-LABEL: pr29025: 679; XOP: # %bb.0: 680; XOP-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 681; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[0,4],xmm2[0],xmm0[1,5],xmm2[1],xmm0[2,6],xmm2[2],xmm0[3,7],xmm2[3],xmm0[u,u,u,u] 682; XOP-NEXT: vpextrd $2, %xmm0, 8(%rdi) 683; XOP-NEXT: vmovq %xmm0, (%rdi) 684; XOP-NEXT: retq 685 %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> 686 %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> 687 %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> 688 store <12 x i8> %r, <12 x i8>* %p, align 1 689 ret void 690} 691 692define void @interleave_24i8_out(<24 x i8>* %p, <8 x i8>* %q1, <8 x i8>* %q2, <8 x i8>* %q3) nounwind { 693; SSE2-LABEL: interleave_24i8_out: 694; SSE2: # %bb.0: 695; SSE2-NEXT: movdqu (%rdi), %xmm0 696; SSE2-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 697; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,255,0,255,255,0,255,255,255,255,255,255,255,255,255,255] 698; SSE2-NEXT: movdqa %xmm0, %xmm2 699; SSE2-NEXT: pand %xmm4, %xmm2 700; SSE2-NEXT: pandn %xmm1, %xmm4 701; SSE2-NEXT: por %xmm2, %xmm4 702; SSE2-NEXT: pxor %xmm2, %xmm2 703; SSE2-NEXT: punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3],xmm4[4],xmm2[4],xmm4[5],xmm2[5],xmm4[6],xmm2[6],xmm4[7],xmm2[7] 704; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0] 705; SSE2-NEXT: pand %xmm5, %xmm4 706; SSE2-NEXT: movdqa %xmm0, %xmm3 707; SSE2-NEXT: punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm2[8],xmm3[9],xmm2[9],xmm3[10],xmm2[10],xmm3[11],xmm2[11],xmm3[12],xmm2[12],xmm3[13],xmm2[13],xmm3[14],xmm2[14],xmm3[15],xmm2[15] 708; SSE2-NEXT: pandn %xmm3, %xmm5 709; SSE2-NEXT: por %xmm4, %xmm5 710; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm5[0,2,1,3] 711; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,7,6,5] 712; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,3,2,1] 713; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,3,2,1,4,5,6,7] 714; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,6,5,4,7] 715; SSE2-NEXT: packuswb %xmm4, %xmm4 716; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [0,255,255,0,255,255,0,255,255,255,255,255,255,255,255,255] 717; SSE2-NEXT: movdqa %xmm0, %xmm6 718; SSE2-NEXT: pand %xmm5, %xmm6 719; SSE2-NEXT: pandn %xmm1, %xmm5 720; SSE2-NEXT: por %xmm6, %xmm5 721; SSE2-NEXT: punpcklbw {{.*#+}} xmm5 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3],xmm5[4],xmm2[4],xmm5[5],xmm2[5],xmm5[6],xmm2[6],xmm5[7],xmm2[7] 722; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,0,65535,65535,0,65535,65535] 723; SSE2-NEXT: pand %xmm6, %xmm5 724; SSE2-NEXT: pandn %xmm3, %xmm6 725; SSE2-NEXT: por %xmm5, %xmm6 726; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm6[2,1,0,3,4,5,6,7] 727; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,4,7] 728; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,1] 729; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7] 730; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,6,7,4] 731; SSE2-NEXT: packuswb %xmm5, %xmm5 732; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [255,0,255,255,0,255,255,0,255,255,255,255,255,255,255,255] 733; SSE2-NEXT: pand %xmm6, %xmm0 734; SSE2-NEXT: pandn %xmm1, %xmm6 735; SSE2-NEXT: por %xmm0, %xmm6 736; SSE2-NEXT: punpcklbw {{.*#+}} xmm6 = xmm6[0],xmm2[0],xmm6[1],xmm2[1],xmm6[2],xmm2[2],xmm6[3],xmm2[3],xmm6[4],xmm2[4],xmm6[5],xmm2[5],xmm6[6],xmm2[6],xmm6[7],xmm2[7] 737; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [0,65535,65535,0,65535,65535,0,65535] 738; SSE2-NEXT: pand %xmm0, %xmm6 739; SSE2-NEXT: pandn %xmm3, %xmm0 740; SSE2-NEXT: por %xmm6, %xmm0 741; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0] 742; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,4,7,6,5] 743; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[3,1,2,0] 744; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,0,3,4,5,6,7] 745; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,5,4,7] 746; SSE2-NEXT: packuswb %xmm0, %xmm0 747; SSE2-NEXT: movq %xmm4, (%rsi) 748; SSE2-NEXT: movq %xmm5, (%rdx) 749; SSE2-NEXT: movq %xmm0, (%rcx) 750; SSE2-NEXT: retq 751; 752; SSE42-LABEL: interleave_24i8_out: 753; SSE42: # %bb.0: 754; SSE42-NEXT: movdqu (%rdi), %xmm0 755; SSE42-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 756; SSE42-NEXT: movdqa %xmm1, %xmm2 757; SSE42-NEXT: pshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm2[2,5,u,u,u,u,u,u,u,u] 758; SSE42-NEXT: movdqa %xmm0, %xmm3 759; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[0,3,6,9,12,15],zero,zero,xmm3[u,u,u,u,u,u,u,u] 760; SSE42-NEXT: por %xmm2, %xmm3 761; SSE42-NEXT: movdqa %xmm1, %xmm2 762; SSE42-NEXT: pshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,xmm2[0,3,6,u,u,u,u,u,u,u,u] 763; SSE42-NEXT: movdqa %xmm0, %xmm4 764; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[1,4,7,10,13],zero,zero,zero,xmm4[u,u,u,u,u,u,u,u] 765; SSE42-NEXT: por %xmm2, %xmm4 766; SSE42-NEXT: pshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u] 767; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 768; SSE42-NEXT: por %xmm1, %xmm0 769; SSE42-NEXT: movq %xmm3, (%rsi) 770; SSE42-NEXT: movq %xmm4, (%rdx) 771; SSE42-NEXT: movq %xmm0, (%rcx) 772; SSE42-NEXT: retq 773; 774; AVX-LABEL: interleave_24i8_out: 775; AVX: # %bb.0: 776; AVX-NEXT: vmovdqu (%rdi), %xmm0 777; AVX-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 778; AVX-NEXT: vpshufb {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,xmm1[2,5,u,u,u,u,u,u,u,u] 779; AVX-NEXT: vpshufb {{.*#+}} xmm3 = xmm0[0,3,6,9,12,15],zero,zero,xmm0[u,u,u,u,u,u,u,u] 780; AVX-NEXT: vpor %xmm2, %xmm3, %xmm2 781; AVX-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,zero,zero,zero,xmm1[0,3,6,u,u,u,u,u,u,u,u] 782; AVX-NEXT: vpshufb {{.*#+}} xmm4 = xmm0[1,4,7,10,13],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 783; AVX-NEXT: vpor %xmm3, %xmm4, %xmm3 784; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,zero,zero,zero,zero,xmm1[1,4,7,u,u,u,u,u,u,u,u] 785; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[2,5,8,11,14],zero,zero,zero,xmm0[u,u,u,u,u,u,u,u] 786; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0 787; AVX-NEXT: vmovq %xmm2, (%rsi) 788; AVX-NEXT: vmovq %xmm3, (%rdx) 789; AVX-NEXT: vmovq %xmm0, (%rcx) 790; AVX-NEXT: retq 791; 792; XOP-LABEL: interleave_24i8_out: 793; XOP: # %bb.0: 794; XOP-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero 795; XOP-NEXT: vmovdqu (%rdi), %xmm1 796; XOP-NEXT: vpperm {{.*#+}} xmm2 = xmm1[0,3,6,9,12,15],xmm0[2,5],xmm1[u,u,u,u,u,u,u,u] 797; XOP-NEXT: vpperm {{.*#+}} xmm3 = xmm1[1,4,7,10,13],xmm0[0,3,6],xmm1[u,u,u,u,u,u,u,u] 798; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm1[2,5,8,11,14],xmm0[1,4,7],xmm1[u,u,u,u,u,u,u,u] 799; XOP-NEXT: vmovq %xmm2, (%rsi) 800; XOP-NEXT: vmovq %xmm3, (%rdx) 801; XOP-NEXT: vmovq %xmm0, (%rcx) 802; XOP-NEXT: retq 803 %wide.vec = load <24 x i8>, <24 x i8>* %p, align 4 804 %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> 805 %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> 806 %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> 807 store <8 x i8> %s1, <8 x i8>* %q1, align 4 808 store <8 x i8> %s2, <8 x i8>* %q2, align 4 809 store <8 x i8> %s3, <8 x i8>* %q3, align 4 810 ret void 811} 812 813define void @interleave_24i8_in(<24 x i8>* %p, <8 x i8>* %q1, <8 x i8>* %q2, <8 x i8>* %q3) nounwind { 814; SSE2-LABEL: interleave_24i8_in: 815; SSE2: # %bb.0: 816; SSE2-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 817; SSE2-NEXT: movq {{.*#+}} xmm2 = mem[0],zero 818; SSE2-NEXT: movq {{.*#+}} xmm0 = mem[0],zero 819; SSE2-NEXT: pxor %xmm3, %xmm3 820; SSE2-NEXT: punpcklbw {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3],xmm1[4],xmm3[4],xmm1[5],xmm3[5],xmm1[6],xmm3[6],xmm1[7],xmm3[7] 821; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[1,1,2,2] 822; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,0,65535,65535,0,65535,65535] 823; SSE2-NEXT: pand %xmm5, %xmm4 824; SSE2-NEXT: punpcklbw {{.*#+}} xmm2 = xmm2[0],xmm3[0],xmm2[1],xmm3[1],xmm2[2],xmm3[2],xmm2[3],xmm3[3],xmm2[4],xmm3[4],xmm2[5],xmm3[5],xmm2[6],xmm3[6],xmm2[7],xmm3[7] 825; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm2[3,3,3,3,4,5,6,7] 826; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4] 827; SSE2-NEXT: pandn %xmm3, %xmm5 828; SSE2-NEXT: por %xmm4, %xmm5 829; SSE2-NEXT: movdqa %xmm1, %xmm3 830; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3] 831; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1] 832; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7] 833; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,4,5] 834; SSE2-NEXT: packuswb %xmm5, %xmm3 835; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [255,255,0,255,255,0,255,255,0,255,255,0,255,255,0,255] 836; SSE2-NEXT: pand %xmm4, %xmm3 837; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,0,1] 838; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,0,4,5,6,7] 839; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,6,6] 840; SSE2-NEXT: pandn %xmm5, %xmm4 841; SSE2-NEXT: por %xmm3, %xmm4 842; SSE2-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] 843; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,1,3,3] 844; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,1,0,4,5,6,7] 845; SSE2-NEXT: packuswb %xmm1, %xmm1 846; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [255,0,255,255,0,255,255,0,255,255,255,255,255,255,255,255] 847; SSE2-NEXT: pand %xmm2, %xmm1 848; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[2,1,3,3,4,5,6,7] 849; SSE2-NEXT: pandn %xmm0, %xmm2 850; SSE2-NEXT: por %xmm1, %xmm2 851; SSE2-NEXT: movq %xmm2, 16(%rdi) 852; SSE2-NEXT: movdqu %xmm4, (%rdi) 853; SSE2-NEXT: retq 854; 855; SSE42-LABEL: interleave_24i8_in: 856; SSE42: # %bb.0: 857; SSE42-NEXT: movq {{.*#+}} xmm0 = mem[0],zero 858; SSE42-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 859; SSE42-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 860; SSE42-NEXT: movq {{.*#+}} xmm1 = mem[0],zero 861; SSE42-NEXT: movdqa %xmm0, %xmm2 862; SSE42-NEXT: pshufb {{.*#+}} xmm2 = xmm2[0,8],zero,xmm2[1,9],zero,xmm2[2,10],zero,xmm2[3,11],zero,xmm2[4,12],zero,xmm2[5] 863; SSE42-NEXT: movdqa %xmm1, %xmm3 864; 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 865; SSE42-NEXT: por %xmm2, %xmm3 866; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[13],zero,xmm0[6,14],zero,xmm0[7,15],zero,xmm0[u,u,u,u,u,u,u,u] 867; SSE42-NEXT: pshufb {{.*#+}} xmm1 = zero,xmm1[5],zero,zero,xmm1[6],zero,zero,xmm1[7,u,u,u,u,u,u,u,u] 868; SSE42-NEXT: por %xmm0, %xmm1 869; SSE42-NEXT: movq %xmm1, 16(%rdi) 870; SSE42-NEXT: movdqu %xmm3, (%rdi) 871; SSE42-NEXT: retq 872; 873; AVX-LABEL: interleave_24i8_in: 874; AVX: # %bb.0: 875; AVX-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero 876; AVX-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 877; AVX-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 878; AVX-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 879; AVX-NEXT: vpshufb {{.*#+}} xmm2 = xmm0[0,8],zero,xmm0[1,9],zero,xmm0[2,10],zero,xmm0[3,11],zero,xmm0[4,12],zero,xmm0[5] 880; AVX-NEXT: vpshufb {{.*#+}} xmm3 = zero,zero,xmm1[0],zero,zero,xmm1[1],zero,zero,xmm1[2],zero,zero,xmm1[3],zero,zero,xmm1[4],zero 881; AVX-NEXT: vpor %xmm3, %xmm2, %xmm2 882; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[13],zero,xmm0[6,14],zero,xmm0[7,15],zero,xmm0[u,u,u,u,u,u,u,u] 883; AVX-NEXT: vpshufb {{.*#+}} xmm1 = zero,xmm1[5],zero,zero,xmm1[6],zero,zero,xmm1[7,u,u,u,u,u,u,u,u] 884; AVX-NEXT: vpor %xmm1, %xmm0, %xmm0 885; AVX-NEXT: vmovq %xmm0, 16(%rdi) 886; AVX-NEXT: vmovdqu %xmm2, (%rdi) 887; AVX-NEXT: retq 888; 889; XOP-LABEL: interleave_24i8_in: 890; XOP: # %bb.0: 891; XOP-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero 892; XOP-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 893; XOP-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 894; XOP-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 895; XOP-NEXT: vpperm {{.*#+}} xmm2 = xmm0[0,8],xmm1[0],xmm0[1,9],xmm1[1],xmm0[2,10],xmm1[2],xmm0[3,11],xmm1[3],xmm0[4,12],xmm1[4],xmm0[5] 896; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[13],xmm1[5],xmm0[6,14],xmm1[6],xmm0[7,15],xmm1[7],xmm0[u,u,u,u,u,u,u,u] 897; XOP-NEXT: vmovq %xmm0, 16(%rdi) 898; XOP-NEXT: vmovdqu %xmm2, (%rdi) 899; XOP-NEXT: retq 900 %s1 = load <8 x i8>, <8 x i8>* %q1, align 4 901 %s2 = load <8 x i8>, <8 x i8>* %q2, align 4 902 %s3 = load <8 x i8>, <8 x i8>* %q3, align 4 903 %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> 904 %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> 905 %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> 906 store <24 x i8> %interleaved, <24 x i8>* %p, align 4 907 ret void 908} 909 910 911define void @interleave_24i16_out(<24 x i16>* %p, <8 x i16>* %q1, <8 x i16>* %q2, <8 x i16>* %q3) nounwind { 912; SSE2-LABEL: interleave_24i16_out: 913; SSE2: # %bb.0: 914; SSE2-NEXT: movdqu (%rdi), %xmm3 915; SSE2-NEXT: movdqu 16(%rdi), %xmm2 916; SSE2-NEXT: movdqu 32(%rdi), %xmm8 917; SSE2-NEXT: movdqa {{.*#+}} xmm1 = [65535,0,65535,65535,0,65535,65535,0] 918; SSE2-NEXT: movdqa %xmm3, %xmm4 919; SSE2-NEXT: pand %xmm1, %xmm4 920; SSE2-NEXT: pandn %xmm2, %xmm1 921; SSE2-NEXT: por %xmm4, %xmm1 922; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3] 923; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,5,6,7] 924; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,1,3] 925; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,3,2,1,4,5,6,7] 926; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,4,7,6,7] 927; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm8[0,1,2,1] 928; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,6,5] 929; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,0],xmm4[2,0] 930; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,0] 931; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535] 932; SSE2-NEXT: movdqa %xmm4, %xmm5 933; SSE2-NEXT: pandn %xmm2, %xmm5 934; SSE2-NEXT: movdqa %xmm3, %xmm6 935; SSE2-NEXT: pand %xmm4, %xmm6 936; SSE2-NEXT: por %xmm5, %xmm6 937; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm6[2,1,2,3,4,5,6,7] 938; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,4,7] 939; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,3,2,3] 940; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[1,2,3,0,4,5,6,7] 941; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,5,5] 942; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0] 943; SSE2-NEXT: pand %xmm6, %xmm5 944; SSE2-NEXT: pshuflw {{.*#+}} xmm7 = xmm8[0,3,2,3,4,5,6,7] 945; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,0,3] 946; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,4,5,6] 947; SSE2-NEXT: movdqa %xmm6, %xmm0 948; SSE2-NEXT: pandn %xmm7, %xmm0 949; SSE2-NEXT: por %xmm5, %xmm0 950; SSE2-NEXT: pand %xmm4, %xmm2 951; SSE2-NEXT: pandn %xmm3, %xmm4 952; SSE2-NEXT: por %xmm2, %xmm4 953; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[3,1,2,0] 954; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,6,5,6,7] 955; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,1,0,3] 956; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7] 957; SSE2-NEXT: pand %xmm6, %xmm2 958; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm8[0,1,2,3,4,7,6,7] 959; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,0] 960; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,7,4,5] 961; SSE2-NEXT: pandn %xmm3, %xmm6 962; SSE2-NEXT: por %xmm2, %xmm6 963; SSE2-NEXT: movups %xmm1, (%rsi) 964; SSE2-NEXT: movdqu %xmm0, (%rdx) 965; SSE2-NEXT: movdqu %xmm6, (%rcx) 966; SSE2-NEXT: retq 967; 968; SSE42-LABEL: interleave_24i16_out: 969; SSE42: # %bb.0: 970; SSE42-NEXT: movdqu (%rdi), %xmm0 971; SSE42-NEXT: movdqu 16(%rdi), %xmm1 972; SSE42-NEXT: movdqu 32(%rdi), %xmm2 973; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm2[0,1,2,1] 974; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5] 975; SSE42-NEXT: movdqa %xmm0, %xmm4 976; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0],xmm1[1],xmm4[2,3],xmm1[4],xmm4[5,6],xmm1[7] 977; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u] 978; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm3[6,7] 979; SSE42-NEXT: movdqa %xmm2, %xmm3 980; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,u,0,1,6,7,12,13] 981; SSE42-NEXT: movdqa %xmm0, %xmm5 982; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1],xmm1[2],xmm5[3,4],xmm1[5],xmm5[6,7] 983; SSE42-NEXT: pshufb {{.*#+}} xmm5 = xmm5[2,3,8,9,14,15,4,5,10,11,u,u,u,u,u,u] 984; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm3[5,6,7] 985; SSE42-NEXT: pshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,u,u,2,3,8,9,14,15] 986; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 987; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[4,5,10,11,0,1,6,7,12,13,u,u,u,u,u,u] 988; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm2[5,6,7] 989; SSE42-NEXT: movdqu %xmm4, (%rsi) 990; SSE42-NEXT: movdqu %xmm5, (%rdx) 991; SSE42-NEXT: movdqu %xmm1, (%rcx) 992; SSE42-NEXT: retq 993; 994; AVX1-LABEL: interleave_24i16_out: 995; AVX1: # %bb.0: 996; AVX1-NEXT: vmovdqu (%rdi), %xmm0 997; AVX1-NEXT: vmovdqu 16(%rdi), %xmm1 998; AVX1-NEXT: vmovdqu 32(%rdi), %xmm2 999; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm2[0,1,2,1] 1000; AVX1-NEXT: vpshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6,5] 1001; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1002; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[0,1,6,7,12,13,2,3,8,9,14,15,u,u,u,u] 1003; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0,1,2,3,4,5],xmm3[6,7] 1004; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm2[u,u,u,u,u,u,u,u,u,u,0,1,6,7,12,13] 1005; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 1006; AVX1-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[2,3,8,9,14,15,4,5,10,11,u,u,u,u,u,u] 1007; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1,2,3,4],xmm4[5,6,7] 1008; AVX1-NEXT: vpshufb {{.*#+}} xmm2 = xmm2[u,u,u,u,u,u,u,u,u,u,2,3,8,9,14,15] 1009; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 1010; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,u,u,u,u,u,u] 1011; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3,4],xmm2[5,6,7] 1012; AVX1-NEXT: vmovdqu %xmm3, (%rsi) 1013; AVX1-NEXT: vmovdqu %xmm4, (%rdx) 1014; AVX1-NEXT: vmovdqu %xmm0, (%rcx) 1015; AVX1-NEXT: retq 1016; 1017; AVX2-LABEL: interleave_24i16_out: 1018; AVX2: # %bb.0: 1019; AVX2-NEXT: vmovdqu (%rdi), %xmm0 1020; AVX2-NEXT: vmovdqu 16(%rdi), %xmm1 1021; AVX2-NEXT: vmovdqu 32(%rdi), %xmm2 1022; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7] 1023; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0],xmm1[1],xmm3[2,3],xmm1[4],xmm3[5,6],xmm1[7] 1024; AVX2-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[0,1,6,7,12,13,2,3,8,9,14,15,4,5,10,11] 1025; AVX2-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1,2],xmm2[3],xmm0[4,5],xmm2[6],xmm0[7] 1026; AVX2-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm1[2],xmm4[3,4],xmm1[5],xmm4[6,7] 1027; AVX2-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[2,3,8,9,14,15,4,5,10,11,0,1,6,7,12,13] 1028; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5,6],xmm2[7] 1029; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0],xmm0[1,2],xmm1[3],xmm0[4,5],xmm1[6],xmm0[7] 1030; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13,2,3,8,9,14,15] 1031; AVX2-NEXT: vmovdqu %xmm3, (%rsi) 1032; AVX2-NEXT: vmovdqu %xmm4, (%rdx) 1033; AVX2-NEXT: vmovdqu %xmm0, (%rcx) 1034; AVX2-NEXT: retq 1035; 1036; XOP-LABEL: interleave_24i16_out: 1037; XOP: # %bb.0: 1038; XOP-NEXT: vmovdqu (%rdi), %xmm0 1039; XOP-NEXT: vmovdqu 16(%rdi), %xmm1 1040; XOP-NEXT: vmovdqu 32(%rdi), %xmm2 1041; XOP-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1042; XOP-NEXT: vpperm {{.*#+}} xmm3 = xmm3[0,1,6,7,12,13,2,3,8,9,14,15],xmm2[4,5,10,11] 1043; XOP-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 1044; XOP-NEXT: vpperm {{.*#+}} xmm4 = xmm4[2,3,8,9,14,15,4,5,10,11],xmm2[0,1,6,7,12,13] 1045; XOP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 1046; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[4,5,10,11,0,1,6,7,12,13],xmm2[2,3,8,9,14,15] 1047; XOP-NEXT: vmovdqu %xmm3, (%rsi) 1048; XOP-NEXT: vmovdqu %xmm4, (%rdx) 1049; XOP-NEXT: vmovdqu %xmm0, (%rcx) 1050; XOP-NEXT: retq 1051 %wide.vec = load <24 x i16>, <24 x i16>* %p, align 4 1052 %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> 1053 %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> 1054 %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> 1055 store <8 x i16> %s1, <8 x i16>* %q1, align 4 1056 store <8 x i16> %s2, <8 x i16>* %q2, align 4 1057 store <8 x i16> %s3, <8 x i16>* %q3, align 4 1058 ret void 1059} 1060 1061define void @interleave_24i16_out_reverse(<24 x i16>* %p, <8 x i16>* %q1, <8 x i16>* %q2, <8 x i16>* %q3) nounwind { 1062; SSE2-LABEL: interleave_24i16_out_reverse: 1063; SSE2: # %bb.0: 1064; SSE2-NEXT: movdqu (%rdi), %xmm8 1065; SSE2-NEXT: movdqu 16(%rdi), %xmm1 1066; SSE2-NEXT: movdqu 32(%rdi), %xmm3 1067; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [65535,0,65535,65535,0,65535,65535,0] 1068; SSE2-NEXT: movdqa %xmm1, %xmm4 1069; SSE2-NEXT: pand %xmm2, %xmm4 1070; SSE2-NEXT: pandn %xmm3, %xmm2 1071; SSE2-NEXT: por %xmm4, %xmm2 1072; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,2,1,3] 1073; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,1,0,3,4,5,6,7] 1074; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[0,3,2,1] 1075; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm2[0,1,2,3,5,6,6,7] 1076; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm8[0,1,2,1] 1077; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,5,5,6] 1078; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,0],xmm4[2,0] 1079; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[3,0,1,2,4,5,6,7] 1080; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,0] 1081; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [65535,65535,0,65535,65535,0,65535,65535] 1082; SSE2-NEXT: movdqa %xmm4, %xmm5 1083; SSE2-NEXT: pandn %xmm1, %xmm5 1084; SSE2-NEXT: movdqa %xmm3, %xmm6 1085; SSE2-NEXT: pand %xmm4, %xmm6 1086; SSE2-NEXT: por %xmm5, %xmm6 1087; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm6[0,3,2,3,4,5,6,7] 1088; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,6,5,6,7] 1089; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,2,1,1] 1090; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[2,1,0,3,4,5,6,7] 1091; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [65535,65535,65535,65535,65535,0,0,0] 1092; SSE2-NEXT: pand %xmm6, %xmm5 1093; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm8[0,1,2,3,4,7,6,7] 1094; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[0,1,2,0] 1095; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,4,5,4,7] 1096; SSE2-NEXT: movdqa %xmm6, %xmm0 1097; SSE2-NEXT: pandn %xmm7, %xmm0 1098; SSE2-NEXT: por %xmm5, %xmm0 1099; SSE2-NEXT: pand %xmm4, %xmm1 1100; SSE2-NEXT: pandn %xmm3, %xmm4 1101; SSE2-NEXT: por %xmm1, %xmm4 1102; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm4[3,1,2,0] 1103; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[2,1,2,3,4,5,6,7] 1104; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[0,2,2,3] 1105; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[3,0,1,2,4,5,6,7] 1106; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,7,7,7,7] 1107; SSE2-NEXT: pand %xmm6, %xmm1 1108; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm8[0,3,2,3,4,5,6,7] 1109; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,0,3] 1110; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,6,5,4] 1111; SSE2-NEXT: pandn %xmm3, %xmm6 1112; SSE2-NEXT: por %xmm1, %xmm6 1113; SSE2-NEXT: movups %xmm2, (%rsi) 1114; SSE2-NEXT: movdqu %xmm0, (%rdx) 1115; SSE2-NEXT: movdqu %xmm6, (%rcx) 1116; SSE2-NEXT: retq 1117; 1118; SSE42-LABEL: interleave_24i16_out_reverse: 1119; SSE42: # %bb.0: 1120; SSE42-NEXT: movdqu (%rdi), %xmm0 1121; SSE42-NEXT: movdqu 16(%rdi), %xmm1 1122; SSE42-NEXT: movdqu 32(%rdi), %xmm2 1123; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,1,2,1] 1124; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,5,6] 1125; SSE42-NEXT: movdqa %xmm1, %xmm4 1126; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0],xmm2[1],xmm4[2,3],xmm2[4],xmm4[5,6],xmm2[7] 1127; SSE42-NEXT: pshufb {{.*#+}} xmm4 = xmm4[14,15,8,9,2,3,12,13,6,7,0,1,u,u,u,u] 1128; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5],xmm3[6,7] 1129; SSE42-NEXT: movdqa %xmm0, %xmm3 1130; SSE42-NEXT: pshufb {{.*#+}} xmm3 = xmm3[u,u,u,u,u,u,u,u,u,u,14,15,8,9,2,3] 1131; SSE42-NEXT: movdqa %xmm2, %xmm5 1132; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1],xmm1[2],xmm5[3,4],xmm1[5],xmm5[6,7] 1133; SSE42-NEXT: pshufb {{.*#+}} xmm5 = xmm5[12,13,6,7,0,1,10,11,4,5,u,u,u,u,u,u] 1134; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm5[0,1,2,3,4],xmm3[5,6,7] 1135; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,u,u,u,u,u,u,12,13,6,7,0,1] 1136; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2],xmm1[3,4],xmm2[5],xmm1[6,7] 1137; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[10,11,4,5,14,15,8,9,2,3,u,u,u,u,u,u] 1138; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1,2,3,4],xmm0[5,6,7] 1139; SSE42-NEXT: movdqu %xmm4, (%rsi) 1140; SSE42-NEXT: movdqu %xmm5, (%rdx) 1141; SSE42-NEXT: movdqu %xmm1, (%rcx) 1142; SSE42-NEXT: retq 1143; 1144; AVX1-LABEL: interleave_24i16_out_reverse: 1145; AVX1: # %bb.0: 1146; AVX1-NEXT: vmovdqu (%rdi), %xmm0 1147; AVX1-NEXT: vmovdqu 16(%rdi), %xmm1 1148; AVX1-NEXT: vmovdqu 32(%rdi), %xmm2 1149; AVX1-NEXT: vpshufb {{.*#+}} xmm3 = xmm2[14,15,8,9,2,3,u,u,u,u,u,u,u,u,u,u] 1150; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 1151; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[u,u,u,u,u,u,12,13,6,7,0,1,10,11,4,5] 1152; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1,2],xmm4[3,4,5,6,7] 1153; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm2[12,13,6,7,0,1,u,u,u,u,u,u,u,u,u,u] 1154; AVX1-NEXT: vpblendw {{.*#+}} xmm5 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 1155; AVX1-NEXT: vpshufb {{.*#+}} xmm5 = xmm5[u,u,u,u,u,u,10,11,4,5,14,15,8,9,2,3] 1156; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1,2],xmm5[3,4,5,6,7] 1157; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[2,1,2,3] 1158; AVX1-NEXT: vpshuflw {{.*#+}} xmm2 = xmm2[1,2,2,3,4,5,6,7] 1159; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1160; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[u,u,u,u,14,15,8,9,2,3,12,13,6,7,0,1] 1161; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3,4,5,6,7] 1162; AVX1-NEXT: vmovdqu %xmm3, (%rsi) 1163; AVX1-NEXT: vmovdqu %xmm4, (%rdx) 1164; AVX1-NEXT: vmovdqu %xmm0, (%rcx) 1165; AVX1-NEXT: retq 1166; 1167; AVX2-LABEL: interleave_24i16_out_reverse: 1168; AVX2: # %bb.0: 1169; AVX2-NEXT: vmovdqu (%rdi), %xmm0 1170; AVX2-NEXT: vmovdqu 16(%rdi), %xmm1 1171; AVX2-NEXT: vmovdqu 32(%rdi), %xmm2 1172; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm0[0],xmm2[1],xmm0[2,3],xmm2[4],xmm0[5,6],xmm2[7] 1173; AVX2-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0],xmm3[1,2],xmm1[3],xmm3[4,5],xmm1[6],xmm3[7] 1174; AVX2-NEXT: vpshufb {{.*#+}} xmm3 = xmm3[14,15,8,9,2,3,12,13,6,7,0,1,10,11,4,5] 1175; AVX2-NEXT: vpblendw {{.*#+}} xmm4 = xmm2[0],xmm0[1,2],xmm2[3],xmm0[4,5],xmm2[6],xmm0[7] 1176; AVX2-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0,1],xmm1[2],xmm4[3,4],xmm1[5],xmm4[6,7] 1177; AVX2-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[12,13,6,7,0,1,10,11,4,5,14,15,8,9,2,3] 1178; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2],xmm0[3,4],xmm2[5],xmm0[6,7] 1179; AVX2-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1180; AVX2-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[10,11,4,5,14,15,8,9,2,3,12,13,6,7,0,1] 1181; AVX2-NEXT: vmovdqu %xmm3, (%rsi) 1182; AVX2-NEXT: vmovdqu %xmm4, (%rdx) 1183; AVX2-NEXT: vmovdqu %xmm0, (%rcx) 1184; AVX2-NEXT: retq 1185; 1186; XOP-LABEL: interleave_24i16_out_reverse: 1187; XOP: # %bb.0: 1188; XOP-NEXT: vmovdqu (%rdi), %xmm0 1189; XOP-NEXT: vmovdqu 16(%rdi), %xmm1 1190; XOP-NEXT: vmovdqu 32(%rdi), %xmm2 1191; XOP-NEXT: vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm0[2],xmm1[3,4],xmm0[5],xmm1[6,7] 1192; XOP-NEXT: vpperm {{.*#+}} xmm3 = xmm2[14,15,8,9,2,3],xmm3[12,13,6,7,0,1,10,11,4,5] 1193; XOP-NEXT: vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 1194; XOP-NEXT: vpperm {{.*#+}} xmm4 = xmm2[12,13,6,7,0,1],xmm4[10,11,4,5,14,15,8,9,2,3] 1195; XOP-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1196; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm2[10,11,4,5],xmm0[14,15,8,9,2,3,12,13,6,7,0,1] 1197; XOP-NEXT: vmovdqu %xmm3, (%rsi) 1198; XOP-NEXT: vmovdqu %xmm4, (%rdx) 1199; XOP-NEXT: vmovdqu %xmm0, (%rcx) 1200; XOP-NEXT: retq 1201 %wide.vec.reverse = load <24 x i16>, <24 x i16>* %p, align 4 1202 %wide.vec = shufflevector <24 x i16> %wide.vec.reverse, <24 x i16> undef, <24 x i32> <i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> 1203 %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> 1204 %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> 1205 %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> 1206 store <8 x i16> %s1, <8 x i16>* %q1, align 4 1207 store <8 x i16> %s2, <8 x i16>* %q2, align 4 1208 store <8 x i16> %s3, <8 x i16>* %q3, align 4 1209 ret void 1210} 1211 1212define void @interleave_24i16_in(<24 x i16>* %p, <8 x i16>* %q1, <8 x i16>* %q2, <8 x i16>* %q3) nounwind { 1213; SSE2-LABEL: interleave_24i16_in: 1214; SSE2: # %bb.0: 1215; SSE2-NEXT: movdqu (%rsi), %xmm3 1216; SSE2-NEXT: movdqu (%rdx), %xmm2 1217; SSE2-NEXT: movdqu (%rcx), %xmm1 1218; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,0,0,0] 1219; SSE2-NEXT: movdqa {{.*#+}} xmm0 = [65535,65535,0,65535,65535,0,65535,65535] 1220; SSE2-NEXT: movdqa %xmm0, %xmm5 1221; SSE2-NEXT: pandn %xmm4, %xmm5 1222; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm3[3,3,3,3] 1223; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm3[1,1,2,2] 1224; SSE2-NEXT: punpcklwd {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1],xmm3[2],xmm2[2],xmm3[3],xmm2[3] 1225; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[0,1,2,1] 1226; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[0,1,2,2,4,5,6,7] 1227; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,7,5,4,5] 1228; SSE2-NEXT: pand %xmm0, %xmm3 1229; SSE2-NEXT: por %xmm5, %xmm3 1230; SSE2-NEXT: movdqa %xmm0, %xmm5 1231; SSE2-NEXT: pandn %xmm4, %xmm5 1232; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm2[3,3,3,3,4,5,6,7] 1233; SSE2-NEXT: punpckhwd {{.*#+}} xmm2 = xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7] 1234; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm2[2,1,2,3] 1235; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[2,3,2,0,4,5,6,7] 1236; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,6,7] 1237; SSE2-NEXT: pand %xmm0, %xmm2 1238; SSE2-NEXT: por %xmm5, %xmm2 1239; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [65535,0,65535,65535,0,65535,65535,0] 1240; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,2,2] 1241; SSE2-NEXT: pand %xmm5, %xmm1 1242; SSE2-NEXT: pandn %xmm6, %xmm5 1243; SSE2-NEXT: por %xmm1, %xmm5 1244; SSE2-NEXT: pand %xmm0, %xmm5 1245; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm4[0,1,2,3,4,4,4,4] 1246; SSE2-NEXT: pandn %xmm1, %xmm0 1247; SSE2-NEXT: por %xmm5, %xmm0 1248; SSE2-NEXT: movdqu %xmm0, 16(%rdi) 1249; SSE2-NEXT: movdqu %xmm2, 32(%rdi) 1250; SSE2-NEXT: movdqu %xmm3, (%rdi) 1251; SSE2-NEXT: retq 1252; 1253; SSE42-LABEL: interleave_24i16_in: 1254; SSE42: # %bb.0: 1255; SSE42-NEXT: movdqu (%rsi), %xmm0 1256; SSE42-NEXT: movdqu (%rdx), %xmm1 1257; SSE42-NEXT: movdqu (%rcx), %xmm2 1258; SSE42-NEXT: pshufd {{.*#+}} xmm3 = xmm0[1,1,2,2] 1259; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3] 1260; SSE42-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 1261; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,u,u,4,5,6,7,u,u,8,9,10,11] 1262; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm2[0,0,0,0] 1263; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm0[0,1],xmm5[2],xmm0[3,4],xmm5[5],xmm0[6,7] 1264; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,2] 1265; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0],xmm3[1],xmm0[2,3],xmm3[4],xmm0[5,6],xmm3[7] 1266; SSE42-NEXT: pshuflw {{.*#+}} xmm3 = xmm1[3,3,3,3,4,5,6,7] 1267; SSE42-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,4,4,4] 1268; SSE42-NEXT: pblendw {{.*#+}} xmm3 = xmm0[0,1],xmm3[2],xmm0[3,4],xmm3[5],xmm0[6,7] 1269; SSE42-NEXT: punpckhwd {{.*#+}} xmm1 = xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7] 1270; SSE42-NEXT: pshufb {{.*#+}} xmm1 = xmm1[4,5,6,7,u,u,8,9,10,11,u,u,12,13,14,15] 1271; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm1[0,1],xmm4[2],xmm1[3,4],xmm4[5],xmm1[6,7] 1272; SSE42-NEXT: movdqu %xmm4, 32(%rdi) 1273; SSE42-NEXT: movdqu %xmm3, 16(%rdi) 1274; SSE42-NEXT: movdqu %xmm5, (%rdi) 1275; SSE42-NEXT: retq 1276; 1277; AVX1-LABEL: interleave_24i16_in: 1278; AVX1: # %bb.0: 1279; AVX1-NEXT: vmovdqu (%rsi), %xmm0 1280; AVX1-NEXT: vmovdqu (%rdx), %xmm1 1281; AVX1-NEXT: vmovdqu (%rcx), %xmm2 1282; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,2,2] 1283; AVX1-NEXT: vpshuflw {{.*#+}} xmm4 = xmm1[3,3,3,3,4,5,6,7] 1284; AVX1-NEXT: vpshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,4] 1285; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3,4],xmm4[5],xmm3[6,7] 1286; AVX1-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2] 1287; AVX1-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7] 1288; AVX1-NEXT: vpunpckhwd {{.*#+}} xmm4 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1289; AVX1-NEXT: vpshufb {{.*#+}} xmm4 = xmm4[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u] 1290; AVX1-NEXT: vpshufd {{.*#+}} xmm5 = xmm2[2,2,3,3] 1291; AVX1-NEXT: vpblendw {{.*#+}} xmm4 = xmm4[0],xmm5[1],xmm4[2,3],xmm5[4],xmm4[5,6],xmm5[7] 1292; AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 1293; AVX1-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,1,2,3,u,u,4,5,6,7,u,u,8,9,10,11] 1294; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[0,0,0,0] 1295; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2],xmm0[3,4],xmm1[5],xmm0[6,7] 1296; AVX1-NEXT: vmovdqu %xmm0, (%rdi) 1297; AVX1-NEXT: vmovdqu %xmm4, 32(%rdi) 1298; AVX1-NEXT: vmovdqu %xmm3, 16(%rdi) 1299; AVX1-NEXT: retq 1300; 1301; AVX2-SLOW-LABEL: interleave_24i16_in: 1302; AVX2-SLOW: # %bb.0: 1303; AVX2-SLOW-NEXT: vmovdqu (%rsi), %xmm0 1304; AVX2-SLOW-NEXT: vmovdqu (%rdx), %xmm1 1305; AVX2-SLOW-NEXT: vmovdqu (%rcx), %xmm2 1306; AVX2-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 1307; AVX2-SLOW-NEXT: vpshufb {{.*#+}} ymm4 = ymm3[0,1,u,u,6,7,2,3,u,u,8,9,4,5,u,u,16,17,u,u,22,23,18,19,u,u,24,25,20,21,u,u] 1308; AVX2-SLOW-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,3,0,1] 1309; AVX2-SLOW-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] 1310; AVX2-SLOW-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] 1311; AVX2-SLOW-NEXT: vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2> 1312; AVX2-SLOW-NEXT: vpermd %ymm2, %ymm4, %ymm4 1313; AVX2-SLOW-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] 1314; AVX2-SLOW-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 1315; AVX2-SLOW-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1316; AVX2-SLOW-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u] 1317; AVX2-SLOW-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3] 1318; AVX2-SLOW-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1319; AVX2-SLOW-NEXT: vmovdqu %xmm0, 32(%rdi) 1320; AVX2-SLOW-NEXT: vmovdqu %ymm3, (%rdi) 1321; AVX2-SLOW-NEXT: vzeroupper 1322; AVX2-SLOW-NEXT: retq 1323; 1324; AVX2-FAST-ALL-LABEL: interleave_24i16_in: 1325; AVX2-FAST-ALL: # %bb.0: 1326; AVX2-FAST-ALL-NEXT: vmovdqu (%rsi), %xmm0 1327; AVX2-FAST-ALL-NEXT: vmovdqu (%rdx), %xmm1 1328; AVX2-FAST-ALL-NEXT: vmovdqu (%rcx), %xmm2 1329; AVX2-FAST-ALL-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 1330; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2> 1331; AVX2-FAST-ALL-NEXT: vpermd %ymm2, %ymm4, %ymm4 1332; AVX2-FAST-ALL-NEXT: vmovdqa {{.*#+}} ymm5 = [0,4,1,5,1,5,2,6] 1333; AVX2-FAST-ALL-NEXT: vpermd %ymm3, %ymm5, %ymm3 1334; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm3 = ymm3[0,1,4,5,u,u,2,3,6,7,u,u,8,9,12,13,u,u,18,19,22,23,u,u,24,25,28,29,u,u,26,27] 1335; AVX2-FAST-ALL-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] 1336; AVX2-FAST-ALL-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 1337; AVX2-FAST-ALL-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1338; AVX2-FAST-ALL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u] 1339; AVX2-FAST-ALL-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3] 1340; AVX2-FAST-ALL-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1341; AVX2-FAST-ALL-NEXT: vmovdqu %xmm0, 32(%rdi) 1342; AVX2-FAST-ALL-NEXT: vmovdqu %ymm3, (%rdi) 1343; AVX2-FAST-ALL-NEXT: vzeroupper 1344; AVX2-FAST-ALL-NEXT: retq 1345; 1346; AVX2-FAST-PERLANE-LABEL: interleave_24i16_in: 1347; AVX2-FAST-PERLANE: # %bb.0: 1348; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rsi), %xmm0 1349; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rdx), %xmm1 1350; AVX2-FAST-PERLANE-NEXT: vmovdqu (%rcx), %xmm2 1351; AVX2-FAST-PERLANE-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 1352; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} ymm4 = ymm3[0,1,u,u,6,7,2,3,u,u,8,9,4,5,u,u,16,17,u,u,22,23,18,19,u,u,24,25,20,21,u,u] 1353; AVX2-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm3 = ymm3[2,3,0,1] 1354; AVX2-FAST-PERLANE-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] 1355; AVX2-FAST-PERLANE-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] 1356; AVX2-FAST-PERLANE-NEXT: vmovdqa {{.*#+}} ymm4 = <u,0,0,u,1,1,u,2> 1357; AVX2-FAST-PERLANE-NEXT: vpermd %ymm2, %ymm4, %ymm4 1358; AVX2-FAST-PERLANE-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] 1359; AVX2-FAST-PERLANE-NEXT: vpblendvb %ymm5, %ymm3, %ymm4, %ymm3 1360; AVX2-FAST-PERLANE-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1361; AVX2-FAST-PERLANE-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[4,5,u,u,10,11,8,9,u,u,14,15,12,13,u,u] 1362; AVX2-FAST-PERLANE-NEXT: vpshufd {{.*#+}} xmm1 = xmm2[2,2,3,3] 1363; AVX2-FAST-PERLANE-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4],xmm0[5,6],xmm1[7] 1364; AVX2-FAST-PERLANE-NEXT: vmovdqu %xmm0, 32(%rdi) 1365; AVX2-FAST-PERLANE-NEXT: vmovdqu %ymm3, (%rdi) 1366; AVX2-FAST-PERLANE-NEXT: vzeroupper 1367; AVX2-FAST-PERLANE-NEXT: retq 1368; 1369; XOP-LABEL: interleave_24i16_in: 1370; XOP: # %bb.0: 1371; XOP-NEXT: vmovdqu (%rsi), %xmm0 1372; XOP-NEXT: vmovdqu (%rdx), %xmm1 1373; XOP-NEXT: vmovdqu (%rcx), %xmm2 1374; XOP-NEXT: vpperm {{.*#+}} xmm3 = xmm0[u,u,6,7],xmm1[6,7],xmm0[u,u,8,9],xmm1[8,9],xmm0[u,u,10,11] 1375; XOP-NEXT: vpshufd {{.*#+}} xmm4 = xmm2[1,1,2,2] 1376; XOP-NEXT: vpblendw {{.*#+}} xmm3 = xmm4[0],xmm3[1,2],xmm4[3],xmm3[4,5],xmm4[6],xmm3[7] 1377; XOP-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3] 1378; XOP-NEXT: vpperm {{.*#+}} xmm4 = xmm4[0,1,2,3],xmm2[0,1],xmm4[4,5,6,7],xmm2[2,3],xmm4[8,9,10,11] 1379; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3 1380; XOP-NEXT: vpunpckhwd {{.*#+}} xmm0 = xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7] 1381; XOP-NEXT: vpperm {{.*#+}} xmm0 = xmm0[4,5],xmm2[10,11],xmm0[10,11,8,9],xmm2[12,13],xmm0[14,15,12,13],xmm2[14,15] 1382; XOP-NEXT: vmovdqu %xmm0, 32(%rdi) 1383; XOP-NEXT: vmovups %ymm3, (%rdi) 1384; XOP-NEXT: vzeroupper 1385; XOP-NEXT: retq 1386 %s1 = load <8 x i16>, <8 x i16>* %q1, align 4 1387 %s2 = load <8 x i16>, <8 x i16>* %q2, align 4 1388 %s3 = load <8 x i16>, <8 x i16>* %q3, align 4 1389 %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> 1390 %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> 1391 %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> 1392 store <24 x i16> %interleaved, <24 x i16>* %p, align 4 1393 ret void 1394} 1395 1396define void @interleave_24i32_out(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2, <8 x i32>* %q3) nounwind { 1397; SSE2-LABEL: interleave_24i32_out: 1398; SSE2: # %bb.0: 1399; SSE2-NEXT: movdqu 64(%rdi), %xmm9 1400; SSE2-NEXT: movups 80(%rdi), %xmm8 1401; SSE2-NEXT: movdqu (%rdi), %xmm0 1402; SSE2-NEXT: movdqu 16(%rdi), %xmm10 1403; SSE2-NEXT: movups 32(%rdi), %xmm5 1404; SSE2-NEXT: movdqu 48(%rdi), %xmm3 1405; SSE2-NEXT: movaps %xmm5, %xmm6 1406; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm0[2,3,2,3] 1407; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm10[1,1,1,1] 1408; SSE2-NEXT: punpckldq {{.*#+}} xmm7 = xmm7[0],xmm4[0],xmm7[1],xmm4[1] 1409; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm5[0,3] 1410; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,1],xmm10[2,3] 1411; SSE2-NEXT: movdqa %xmm0, %xmm4 1412; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm5[2,0] 1413; SSE2-NEXT: movaps %xmm8, %xmm5 1414; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm3[2,3,2,3] 1415; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm9[1,1,1,1] 1416; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1] 1417; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm8[0,3] 1418; SSE2-NEXT: shufps {{.*#+}} xmm8 = xmm8[1,1],xmm9[2,3] 1419; SSE2-NEXT: movdqa %xmm3, %xmm2 1420; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,3],xmm8[2,0] 1421; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[2,1],xmm9[3,3] 1422; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm9[0,0] 1423; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,2],xmm5[2,0] 1424; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[2,1],xmm10[3,3] 1425; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,0],xmm10[0,0] 1426; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,2],xmm6[2,0] 1427; SSE2-NEXT: movups %xmm2, 16(%rsi) 1428; SSE2-NEXT: movups %xmm4, (%rsi) 1429; SSE2-NEXT: movups %xmm3, 16(%rdx) 1430; SSE2-NEXT: movups %xmm0, (%rdx) 1431; SSE2-NEXT: movups %xmm1, 16(%rcx) 1432; SSE2-NEXT: movups %xmm7, (%rcx) 1433; SSE2-NEXT: retq 1434; 1435; SSE42-LABEL: interleave_24i32_out: 1436; SSE42: # %bb.0: 1437; SSE42-NEXT: movups 80(%rdi), %xmm8 1438; SSE42-NEXT: movdqu 64(%rdi), %xmm9 1439; SSE42-NEXT: movdqu (%rdi), %xmm3 1440; SSE42-NEXT: movdqu 16(%rdi), %xmm2 1441; SSE42-NEXT: movups 32(%rdi), %xmm10 1442; SSE42-NEXT: movdqu 48(%rdi), %xmm5 1443; SSE42-NEXT: movdqa %xmm2, %xmm6 1444; SSE42-NEXT: pblendw {{.*#+}} xmm6 = xmm6[0,1],xmm3[2,3],xmm6[4,5,6,7] 1445; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm3[2,3,2,3] 1446; SSE42-NEXT: shufps {{.*#+}} xmm3 = xmm3[0,3],xmm2[2,3] 1447; SSE42-NEXT: insertps {{.*#+}} xmm3 = xmm3[0,1,2],xmm10[1] 1448; SSE42-NEXT: movdqa %xmm9, %xmm1 1449; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm1[0,1],xmm5[2,3],xmm1[4,5,6,7] 1450; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm5[2,3,2,3] 1451; SSE42-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,3],xmm9[2,3] 1452; SSE42-NEXT: insertps {{.*#+}} xmm5 = xmm5[0,1,2],xmm8[1] 1453; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm10[2,2,2,2] 1454; SSE42-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,0,3,3] 1455; SSE42-NEXT: pblendw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,5],xmm4[6,7] 1456; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,0,3,3] 1457; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm8[2,2,2,2] 1458; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm1[0,1,2,3,4,5],xmm4[6,7] 1459; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm2[2,3],xmm7[4,5,6,7] 1460; SSE42-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm10[0,3] 1461; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm9[2,3],xmm0[4,5,6,7] 1462; SSE42-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm8[0,3] 1463; SSE42-NEXT: movups %xmm5, 16(%rsi) 1464; SSE42-NEXT: movups %xmm3, (%rsi) 1465; SSE42-NEXT: movdqu %xmm4, 16(%rdx) 1466; SSE42-NEXT: movdqu %xmm6, (%rdx) 1467; SSE42-NEXT: movups %xmm0, 16(%rcx) 1468; SSE42-NEXT: movups %xmm7, (%rcx) 1469; SSE42-NEXT: retq 1470; 1471; AVX1-LABEL: interleave_24i32_out: 1472; AVX1: # %bb.0: 1473; AVX1-NEXT: vmovups 64(%rdi), %ymm0 1474; AVX1-NEXT: vmovups 32(%rdi), %ymm1 1475; AVX1-NEXT: vmovups (%rdi), %ymm2 1476; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7] 1477; AVX1-NEXT: vmovups 16(%rdi), %xmm4 1478; AVX1-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,1],ymm1[1,3],ymm4[6,5],ymm1[5,7] 1479; AVX1-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm4[0,2],ymm3[4,7],ymm4[4,6] 1480; AVX1-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1] 1481; AVX1-NEXT: vshufps {{.*#+}} ymm5 = ymm0[1,0],ymm4[2,0],ymm0[5,4],ymm4[6,4] 1482; AVX1-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4] 1483; AVX1-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7] 1484; AVX1-NEXT: vshufps {{.*#+}} ymm5 = ymm0[2,0],ymm4[3,0],ymm0[6,4],ymm4[7,4] 1485; AVX1-NEXT: vshufps {{.*#+}} ymm5 = ymm4[0,0],ymm5[2,0],ymm4[4,4],ymm5[6,4] 1486; AVX1-NEXT: vmovups 16(%rdi), %xmm6 1487; AVX1-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7] 1488; AVX1-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7] 1489; AVX1-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5] 1490; AVX1-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7] 1491; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7] 1492; AVX1-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[2,0],ymm6[5,4],ymm2[6,4] 1493; AVX1-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[0,3],ymm2[6,4],ymm1[4,7] 1494; AVX1-NEXT: vshufps {{.*#+}} ymm0 = ymm4[0,1],ymm0[0,3],ymm4[4,5],ymm0[4,7] 1495; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] 1496; AVX1-NEXT: vmovups %ymm3, (%rsi) 1497; AVX1-NEXT: vmovups %ymm5, (%rdx) 1498; AVX1-NEXT: vmovups %ymm0, (%rcx) 1499; AVX1-NEXT: vzeroupper 1500; AVX1-NEXT: retq 1501; 1502; AVX2-SLOW-LABEL: interleave_24i32_out: 1503; AVX2-SLOW: # %bb.0: 1504; AVX2-SLOW-NEXT: vmovups (%rdi), %ymm0 1505; AVX2-SLOW-NEXT: vmovups 32(%rdi), %ymm1 1506; AVX2-SLOW-NEXT: vmovups 64(%rdi), %ymm2 1507; AVX2-SLOW-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482] 1508; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm3, %ymm3 1509; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7] 1510; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u> 1511; AVX2-SLOW-NEXT: vpermps %ymm4, %ymm5, %ymm4 1512; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7] 1513; AVX2-SLOW-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6] 1514; AVX2-SLOW-NEXT: # ymm4 = mem[0,1,0,1] 1515; AVX2-SLOW-NEXT: vpermps %ymm2, %ymm4, %ymm4 1516; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1517; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u> 1518; AVX2-SLOW-NEXT: vpermps %ymm5, %ymm6, %ymm5 1519; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7] 1520; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7] 1521; AVX2-SLOW-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u> 1522; AVX2-SLOW-NEXT: vpermps %ymm0, %ymm1, %ymm0 1523; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7] 1524; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3] 1525; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] 1526; AVX2-SLOW-NEXT: vmovups %ymm3, (%rsi) 1527; AVX2-SLOW-NEXT: vmovups %ymm4, (%rdx) 1528; AVX2-SLOW-NEXT: vmovups %ymm0, (%rcx) 1529; AVX2-SLOW-NEXT: vzeroupper 1530; AVX2-SLOW-NEXT: retq 1531; 1532; AVX2-FAST-ALL-LABEL: interleave_24i32_out: 1533; AVX2-FAST-ALL: # %bb.0: 1534; AVX2-FAST-ALL-NEXT: vmovups (%rdi), %ymm0 1535; AVX2-FAST-ALL-NEXT: vmovups 32(%rdi), %ymm1 1536; AVX2-FAST-ALL-NEXT: vmovups 64(%rdi), %ymm2 1537; AVX2-FAST-ALL-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482] 1538; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm3, %ymm3 1539; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7] 1540; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u> 1541; AVX2-FAST-ALL-NEXT: vpermps %ymm4, %ymm5, %ymm4 1542; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7] 1543; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6] 1544; AVX2-FAST-ALL-NEXT: # ymm4 = mem[0,1,0,1] 1545; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm4, %ymm4 1546; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1547; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u> 1548; AVX2-FAST-ALL-NEXT: vpermps %ymm5, %ymm6, %ymm5 1549; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7] 1550; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm5 = <u,u,u,u,u,1,4,7> 1551; AVX2-FAST-ALL-NEXT: vpermps %ymm2, %ymm5, %ymm2 1552; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7] 1553; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u> 1554; AVX2-FAST-ALL-NEXT: vpermps %ymm0, %ymm1, %ymm0 1555; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm2[5,6,7] 1556; AVX2-FAST-ALL-NEXT: vmovups %ymm3, (%rsi) 1557; AVX2-FAST-ALL-NEXT: vmovups %ymm4, (%rdx) 1558; AVX2-FAST-ALL-NEXT: vmovups %ymm0, (%rcx) 1559; AVX2-FAST-ALL-NEXT: vzeroupper 1560; AVX2-FAST-ALL-NEXT: retq 1561; 1562; AVX2-FAST-PERLANE-LABEL: interleave_24i32_out: 1563; AVX2-FAST-PERLANE: # %bb.0: 1564; AVX2-FAST-PERLANE-NEXT: vmovups (%rdi), %ymm0 1565; AVX2-FAST-PERLANE-NEXT: vmovups 32(%rdi), %ymm1 1566; AVX2-FAST-PERLANE-NEXT: vmovups 64(%rdi), %ymm2 1567; AVX2-FAST-PERLANE-NEXT: vbroadcastsd {{.*#+}} ymm3 = [21474836482,21474836482,21474836482,21474836482] 1568; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm3, %ymm3 1569; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm0[0],ymm1[1],ymm0[2,3],ymm1[4],ymm0[5,6],ymm1[7] 1570; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm5 = <0,3,6,1,4,7,u,u> 1571; AVX2-FAST-PERLANE-NEXT: vpermps %ymm4, %ymm5, %ymm4 1572; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0,1,2,3,4,5],ymm3[6,7] 1573; AVX2-FAST-PERLANE-NEXT: vbroadcastf128 {{.*#+}} ymm4 = [0,0,3,6,0,0,3,6] 1574; AVX2-FAST-PERLANE-NEXT: # ymm4 = mem[0,1,0,1] 1575; AVX2-FAST-PERLANE-NEXT: vpermps %ymm2, %ymm4, %ymm4 1576; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm5 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1577; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm6 = <1,4,7,2,5,u,u,u> 1578; AVX2-FAST-PERLANE-NEXT: vpermps %ymm5, %ymm6, %ymm5 1579; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0,1,2,3,4],ymm4[5,6,7] 1580; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1],ymm0[2],ymm1[3,4],ymm0[5],ymm1[6,7] 1581; AVX2-FAST-PERLANE-NEXT: vmovaps {{.*#+}} ymm1 = <2,5,0,3,6,u,u,u> 1582; AVX2-FAST-PERLANE-NEXT: vpermps %ymm0, %ymm1, %ymm0 1583; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm2[0,1,0,3,4,5,4,7] 1584; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm1 = ymm1[0,1,0,3] 1585; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1,2,3,4],ymm1[5,6,7] 1586; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rsi) 1587; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, (%rdx) 1588; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, (%rcx) 1589; AVX2-FAST-PERLANE-NEXT: vzeroupper 1590; AVX2-FAST-PERLANE-NEXT: retq 1591; 1592; XOP-LABEL: interleave_24i32_out: 1593; XOP: # %bb.0: 1594; XOP-NEXT: vmovups 64(%rdi), %ymm0 1595; XOP-NEXT: vmovups 32(%rdi), %ymm1 1596; XOP-NEXT: vmovups (%rdi), %ymm2 1597; XOP-NEXT: vblendps {{.*#+}} ymm3 = ymm2[0],ymm1[1],ymm2[2,3],ymm1[4],ymm2[5,6],ymm1[7] 1598; XOP-NEXT: vmovups 16(%rdi), %xmm4 1599; XOP-NEXT: vshufps {{.*#+}} ymm4 = ymm4[2,1],ymm1[1,3],ymm4[6,5],ymm1[5,7] 1600; XOP-NEXT: vshufps {{.*#+}} ymm3 = ymm3[0,3],ymm4[0,2],ymm3[4,7],ymm4[4,6] 1601; XOP-NEXT: vperm2f128 {{.*#+}} ymm4 = ymm0[2,3,0,1] 1602; XOP-NEXT: vshufps {{.*#+}} ymm5 = ymm0[1,0],ymm4[2,0],ymm0[5,4],ymm4[6,4] 1603; XOP-NEXT: vpermilps {{.*#+}} ymm5 = ymm5[0,1,2,0,4,5,6,4] 1604; XOP-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1,2,3,4,5],ymm5[6,7] 1605; XOP-NEXT: vshufps {{.*#+}} ymm5 = ymm0[2,0],ymm4[3,0],ymm0[6,4],ymm4[7,4] 1606; XOP-NEXT: vshufps {{.*#+}} ymm5 = ymm4[0,0],ymm5[2,0],ymm4[4,4],ymm5[6,4] 1607; XOP-NEXT: vmovups 16(%rdi), %xmm6 1608; XOP-NEXT: vblendps {{.*#+}} ymm7 = ymm2[0,1],ymm1[2],ymm2[3,4],ymm1[5],ymm2[6,7] 1609; XOP-NEXT: vshufps {{.*#+}} ymm7 = ymm7[1,2],ymm6[0,3],ymm7[5,6],ymm6[4,7] 1610; XOP-NEXT: vpermilps {{.*#+}} ymm7 = ymm7[0,2,3,1,4,6,7,5] 1611; XOP-NEXT: vblendps {{.*#+}} ymm5 = ymm7[0,1,2,3,4],ymm5[5,6,7] 1612; XOP-NEXT: vblendps {{.*#+}} ymm2 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7] 1613; XOP-NEXT: vshufps {{.*#+}} ymm2 = ymm6[1,0],ymm2[2,0],ymm6[5,4],ymm2[6,4] 1614; XOP-NEXT: vshufps {{.*#+}} ymm1 = ymm2[2,0],ymm1[0,3],ymm2[6,4],ymm1[4,7] 1615; XOP-NEXT: vshufps {{.*#+}} ymm0 = ymm4[0,1],ymm0[0,3],ymm4[4,5],ymm0[4,7] 1616; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3,4],ymm0[5,6,7] 1617; XOP-NEXT: vmovups %ymm3, (%rsi) 1618; XOP-NEXT: vmovups %ymm5, (%rdx) 1619; XOP-NEXT: vmovups %ymm0, (%rcx) 1620; XOP-NEXT: vzeroupper 1621; XOP-NEXT: retq 1622 %wide.vec = load <24 x i32>, <24 x i32>* %p, align 4 1623 %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> 1624 %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> 1625 %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> 1626 store <8 x i32> %s1, <8 x i32>* %q1, align 4 1627 store <8 x i32> %s2, <8 x i32>* %q2, align 4 1628 store <8 x i32> %s3, <8 x i32>* %q3, align 4 1629 ret void 1630} 1631 1632define void @interleave_24i32_in(<24 x i32>* %p, <8 x i32>* %q1, <8 x i32>* %q2, <8 x i32>* %q3) nounwind { 1633; SSE2-LABEL: interleave_24i32_in: 1634; SSE2: # %bb.0: 1635; SSE2-NEXT: movups (%rsi), %xmm1 1636; SSE2-NEXT: movups 16(%rsi), %xmm0 1637; SSE2-NEXT: movups (%rdx), %xmm8 1638; SSE2-NEXT: movups 16(%rdx), %xmm5 1639; SSE2-NEXT: movups (%rcx), %xmm3 1640; SSE2-NEXT: movups 16(%rcx), %xmm6 1641; SSE2-NEXT: movaps %xmm3, %xmm7 1642; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[0,1],xmm1[1,3] 1643; SSE2-NEXT: movaps %xmm1, %xmm9 1644; SSE2-NEXT: unpcklps {{.*#+}} xmm9 = xmm9[0],xmm8[0],xmm9[1],xmm8[1] 1645; SSE2-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,1],xmm7[0,2] 1646; SSE2-NEXT: movaps %xmm5, %xmm7 1647; SSE2-NEXT: shufps {{.*#+}} xmm7 = xmm7[3,3],xmm6[3,3] 1648; SSE2-NEXT: movaps %xmm6, %xmm4 1649; SSE2-NEXT: unpckhps {{.*#+}} xmm4 = xmm4[2],xmm0[2],xmm4[3],xmm0[3] 1650; SSE2-NEXT: shufps {{.*#+}} xmm4 = xmm4[0,3],xmm7[0,2] 1651; SSE2-NEXT: movaps %xmm0, %xmm7 1652; SSE2-NEXT: unpckhpd {{.*#+}} xmm7 = xmm7[1],xmm5[1] 1653; SSE2-NEXT: movaps %xmm6, %xmm2 1654; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[1,0],xmm5[1,0] 1655; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[2,0],xmm7[0,2] 1656; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,1],xmm0[1,3] 1657; SSE2-NEXT: unpcklps {{.*#+}} xmm0 = xmm0[0],xmm5[0],xmm0[1],xmm5[1] 1658; SSE2-NEXT: shufps {{.*#+}} xmm0 = xmm0[0,1],xmm6[0,2] 1659; SSE2-NEXT: movaps %xmm8, %xmm5 1660; SSE2-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,3],xmm3[3,3] 1661; SSE2-NEXT: movaps %xmm3, %xmm6 1662; SSE2-NEXT: unpckhps {{.*#+}} xmm6 = xmm6[2],xmm1[2],xmm6[3],xmm1[3] 1663; SSE2-NEXT: shufps {{.*#+}} xmm6 = xmm6[0,3],xmm5[0,2] 1664; SSE2-NEXT: unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm8[1] 1665; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[1,0],xmm8[1,0] 1666; SSE2-NEXT: shufps {{.*#+}} xmm3 = xmm3[2,0],xmm1[0,2] 1667; SSE2-NEXT: movups %xmm3, 16(%rdi) 1668; SSE2-NEXT: movups %xmm6, 32(%rdi) 1669; SSE2-NEXT: movups %xmm0, 48(%rdi) 1670; SSE2-NEXT: movups %xmm2, 64(%rdi) 1671; SSE2-NEXT: movups %xmm4, 80(%rdi) 1672; SSE2-NEXT: movups %xmm9, (%rdi) 1673; SSE2-NEXT: retq 1674; 1675; SSE42-LABEL: interleave_24i32_in: 1676; SSE42: # %bb.0: 1677; SSE42-NEXT: movdqu (%rsi), %xmm8 1678; SSE42-NEXT: movdqu 16(%rsi), %xmm4 1679; SSE42-NEXT: movdqu (%rdx), %xmm2 1680; SSE42-NEXT: movdqu 16(%rdx), %xmm5 1681; SSE42-NEXT: movdqu (%rcx), %xmm3 1682; SSE42-NEXT: movdqu 16(%rcx), %xmm6 1683; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm2[0,0,1,1] 1684; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm8[0,1,0,1] 1685; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm7[0,1],xmm1[2,3],xmm7[4,5,6,7] 1686; SSE42-NEXT: pshufd {{.*#+}} xmm1 = xmm3[0,1,0,1] 1687; SSE42-NEXT: pblendw {{.*#+}} xmm1 = xmm7[0,1,2,3],xmm1[4,5],xmm7[6,7] 1688; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm4[2,3,2,3] 1689; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm6[2,3,2,3] 1690; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm7[2,3],xmm0[4,5,6,7] 1691; SSE42-NEXT: pshufd {{.*#+}} xmm7 = xmm5[2,2,3,3] 1692; SSE42-NEXT: pblendw {{.*#+}} xmm7 = xmm0[0,1,2,3],xmm7[4,5],xmm0[6,7] 1693; SSE42-NEXT: pshufd {{.*#+}} xmm0 = xmm5[1,1,2,2] 1694; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1],xmm6[2,3],xmm0[4,5,6,7] 1695; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm4[4,5],xmm0[6,7] 1696; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm5[0,0,1,1] 1697; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm4[0,1,0,1] 1698; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm4[0,1],xmm5[2,3],xmm4[4,5,6,7] 1699; SSE42-NEXT: pshufd {{.*#+}} xmm5 = xmm6[0,1,0,1] 1700; SSE42-NEXT: pblendw {{.*#+}} xmm5 = xmm4[0,1,2,3],xmm5[4,5],xmm4[6,7] 1701; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm8[2,3,2,3] 1702; SSE42-NEXT: pshufd {{.*#+}} xmm6 = xmm3[2,3,2,3] 1703; SSE42-NEXT: pblendw {{.*#+}} xmm6 = xmm6[0,1],xmm4[2,3],xmm6[4,5,6,7] 1704; SSE42-NEXT: pshufd {{.*#+}} xmm4 = xmm2[2,2,3,3] 1705; SSE42-NEXT: pblendw {{.*#+}} xmm4 = xmm6[0,1,2,3],xmm4[4,5],xmm6[6,7] 1706; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,2,2] 1707; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm3[2,3],xmm2[4,5,6,7] 1708; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1,2,3],xmm8[4,5],xmm2[6,7] 1709; SSE42-NEXT: movdqu %xmm2, 16(%rdi) 1710; SSE42-NEXT: movdqu %xmm4, 32(%rdi) 1711; SSE42-NEXT: movdqu %xmm5, 48(%rdi) 1712; SSE42-NEXT: movdqu %xmm0, 64(%rdi) 1713; SSE42-NEXT: movdqu %xmm7, 80(%rdi) 1714; SSE42-NEXT: movdqu %xmm1, (%rdi) 1715; SSE42-NEXT: retq 1716; 1717; AVX1-LABEL: interleave_24i32_in: 1718; AVX1: # %bb.0: 1719; AVX1-NEXT: vmovups (%rdx), %xmm0 1720; AVX1-NEXT: vmovups (%rsi), %xmm1 1721; AVX1-NEXT: vunpckhpd {{.*#+}} xmm2 = xmm1[1],xmm0[1] 1722; AVX1-NEXT: vshufps {{.*#+}} xmm2 = xmm0[1,1],xmm2[0,2] 1723; AVX1-NEXT: vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0] 1724; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[2,0],xmm1[2,1] 1725; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0 1726; AVX1-NEXT: vbroadcastsd (%rcx), %ymm1 1727; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1728; AVX1-NEXT: vmovups 16(%rcx), %xmm1 1729; AVX1-NEXT: vmovups 16(%rdx), %xmm2 1730; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm2[3,0],xmm1[3,0] 1731; AVX1-NEXT: vshufps {{.*#+}} xmm3 = xmm1[2,1],xmm3[0,2] 1732; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[1,0],xmm2[1,0] 1733; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,2] 1734; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1 1735; AVX1-NEXT: vbroadcastsd 24(%rsi), %ymm2 1736; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7] 1737; AVX1-NEXT: vpermilpd {{.*#+}} ymm2 = mem[1,0,2,2] 1738; AVX1-NEXT: vpermilpd {{.*#+}} ymm3 = mem[1,1,2,2] 1739; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm3[0],ymm2[1],ymm3[2,3],ymm2[4],ymm3[5,6],ymm2[7] 1740; AVX1-NEXT: vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7] 1741; AVX1-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 1742; AVX1-NEXT: vmovups %ymm2, 32(%rdi) 1743; AVX1-NEXT: vmovups %ymm1, 64(%rdi) 1744; AVX1-NEXT: vmovups %ymm0, (%rdi) 1745; AVX1-NEXT: vzeroupper 1746; AVX1-NEXT: retq 1747; 1748; AVX2-SLOW-LABEL: interleave_24i32_in: 1749; AVX2-SLOW: # %bb.0: 1750; AVX2-SLOW-NEXT: vmovups (%rsi), %ymm0 1751; AVX2-SLOW-NEXT: vmovups (%rdx), %ymm1 1752; AVX2-SLOW-NEXT: vmovups (%rcx), %ymm2 1753; AVX2-SLOW-NEXT: vpermilps {{.*#+}} xmm3 = mem[1,0,2,2] 1754; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,1] 1755; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1] 1756; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7] 1757; AVX2-SLOW-NEXT: vbroadcastsd (%rcx), %ymm4 1758; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7] 1759; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3] 1760; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7] 1761; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3] 1762; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7] 1763; AVX2-SLOW-NEXT: vbroadcastsd 24(%rsi), %ymm5 1764; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7] 1765; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2] 1766; AVX2-SLOW-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2] 1767; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7] 1768; AVX2-SLOW-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7] 1769; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1770; AVX2-SLOW-NEXT: vmovups %ymm0, 32(%rdi) 1771; AVX2-SLOW-NEXT: vmovups %ymm4, 64(%rdi) 1772; AVX2-SLOW-NEXT: vmovups %ymm3, (%rdi) 1773; AVX2-SLOW-NEXT: vzeroupper 1774; AVX2-SLOW-NEXT: retq 1775; 1776; AVX2-FAST-ALL-LABEL: interleave_24i32_in: 1777; AVX2-FAST-ALL: # %bb.0: 1778; AVX2-FAST-ALL-NEXT: vmovups (%rsi), %ymm0 1779; AVX2-FAST-ALL-NEXT: vmovups (%rdx), %ymm1 1780; AVX2-FAST-ALL-NEXT: vmovups (%rcx), %ymm2 1781; AVX2-FAST-ALL-NEXT: vbroadcastf128 {{.*#+}} ymm3 = [1,0,2,2,1,0,2,2] 1782; AVX2-FAST-ALL-NEXT: # ymm3 = mem[0,1,0,1] 1783; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm3, %ymm3 1784; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1] 1785; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7] 1786; AVX2-FAST-ALL-NEXT: vbroadcastsd (%rcx), %ymm4 1787; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7] 1788; AVX2-FAST-ALL-NEXT: vmovaps {{.*#+}} ymm4 = <5,u,u,6,u,u,7,u> 1789; AVX2-FAST-ALL-NEXT: vpermps %ymm1, %ymm4, %ymm4 1790; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm5 = ymm2[2,1,3,3] 1791; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0],ymm5[1],ymm4[2,3],ymm5[4],ymm4[5,6],ymm5[7] 1792; AVX2-FAST-ALL-NEXT: vbroadcastsd 24(%rsi), %ymm5 1793; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7] 1794; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2] 1795; AVX2-FAST-ALL-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2] 1796; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7] 1797; AVX2-FAST-ALL-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7] 1798; AVX2-FAST-ALL-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1799; AVX2-FAST-ALL-NEXT: vmovups %ymm0, 32(%rdi) 1800; AVX2-FAST-ALL-NEXT: vmovups %ymm4, 64(%rdi) 1801; AVX2-FAST-ALL-NEXT: vmovups %ymm3, (%rdi) 1802; AVX2-FAST-ALL-NEXT: vzeroupper 1803; AVX2-FAST-ALL-NEXT: retq 1804; 1805; AVX2-FAST-PERLANE-LABEL: interleave_24i32_in: 1806; AVX2-FAST-PERLANE: # %bb.0: 1807; AVX2-FAST-PERLANE-NEXT: vmovups (%rsi), %ymm0 1808; AVX2-FAST-PERLANE-NEXT: vmovups (%rdx), %ymm1 1809; AVX2-FAST-PERLANE-NEXT: vmovups (%rcx), %ymm2 1810; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} xmm3 = mem[1,0,2,2] 1811; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm3 = ymm3[0,1,0,1] 1812; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm0[0,0,2,1] 1813; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm4[0],ymm3[1],ymm4[2,3],ymm3[4],ymm4[5,6],ymm3[7] 1814; AVX2-FAST-PERLANE-NEXT: vbroadcastsd (%rcx), %ymm4 1815; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm3 = ymm3[0,1],ymm4[2],ymm3[3,4],ymm4[5],ymm3[6,7] 1816; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm4 = ymm2[2,1,3,3] 1817; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm5 = ymm1[1,2,3,3,5,6,7,7] 1818; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm5 = ymm5[2,2,2,3] 1819; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm5[0],ymm4[1],ymm5[2,3],ymm4[4],ymm5[5,6],ymm4[7] 1820; AVX2-FAST-PERLANE-NEXT: vbroadcastsd 24(%rsi), %ymm5 1821; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm4 = ymm4[0,1],ymm5[2],ymm4[3,4],ymm5[5],ymm4[6,7] 1822; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm0 = ymm0[1,1,2,2] 1823; AVX2-FAST-PERLANE-NEXT: vpermpd {{.*#+}} ymm2 = ymm2[1,1,2,2] 1824; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm0[1],ymm2[2,3],ymm0[4],ymm2[5,6],ymm0[7] 1825; AVX2-FAST-PERLANE-NEXT: vpermilps {{.*#+}} ymm1 = ymm1[0,0,3,3,4,4,7,7] 1826; AVX2-FAST-PERLANE-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm1[2],ymm0[3,4],ymm1[5],ymm0[6,7] 1827; AVX2-FAST-PERLANE-NEXT: vmovups %ymm0, 32(%rdi) 1828; AVX2-FAST-PERLANE-NEXT: vmovups %ymm4, 64(%rdi) 1829; AVX2-FAST-PERLANE-NEXT: vmovups %ymm3, (%rdi) 1830; AVX2-FAST-PERLANE-NEXT: vzeroupper 1831; AVX2-FAST-PERLANE-NEXT: retq 1832; 1833; XOP-LABEL: interleave_24i32_in: 1834; XOP: # %bb.0: 1835; XOP-NEXT: vmovups (%rsi), %ymm0 1836; XOP-NEXT: vmovups (%rcx), %ymm1 1837; XOP-NEXT: vpermil2ps {{.*#+}} ymm0 = ymm1[2],ymm0[3],ymm1[u,3],ymm0[4],ymm1[u,4],ymm0[5] 1838; XOP-NEXT: vmovups (%rdx), %xmm1 1839; XOP-NEXT: vmovups (%rsi), %xmm2 1840; XOP-NEXT: vunpckhpd {{.*#+}} xmm3 = xmm2[1],xmm1[1] 1841; XOP-NEXT: vshufps {{.*#+}} xmm3 = xmm1[1,1],xmm3[0,2] 1842; XOP-NEXT: vmovlhps {{.*#+}} xmm1 = xmm1[0],xmm2[0] 1843; XOP-NEXT: vshufps {{.*#+}} xmm1 = xmm1[2,0],xmm2[2,1] 1844; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm1, %ymm1 1845; XOP-NEXT: vbroadcastsd (%rcx), %ymm2 1846; XOP-NEXT: vblendps {{.*#+}} ymm1 = ymm1[0,1],ymm2[2],ymm1[3,4],ymm2[5],ymm1[6,7] 1847; XOP-NEXT: vmovups 16(%rcx), %xmm2 1848; XOP-NEXT: vmovups 16(%rdx), %xmm3 1849; XOP-NEXT: vshufps {{.*#+}} xmm4 = xmm3[3,0],xmm2[3,0] 1850; XOP-NEXT: vshufps {{.*#+}} xmm4 = xmm2[2,1],xmm4[0,2] 1851; XOP-NEXT: vshufps {{.*#+}} xmm2 = xmm2[1,0],xmm3[1,0] 1852; XOP-NEXT: vshufps {{.*#+}} xmm2 = xmm2[2,0],xmm3[2,2] 1853; XOP-NEXT: vinsertf128 $1, %xmm4, %ymm2, %ymm2 1854; XOP-NEXT: vbroadcastsd 24(%rsi), %ymm3 1855; XOP-NEXT: vblendps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2],ymm2[3,4],ymm3[5],ymm2[6,7] 1856; XOP-NEXT: vpermilps {{.*#+}} ymm3 = mem[0,0,3,3,4,4,7,7] 1857; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm0[0,1],ymm3[2],ymm0[3,4],ymm3[5],ymm0[6,7] 1858; XOP-NEXT: vmovups %ymm0, 32(%rdi) 1859; XOP-NEXT: vmovups %ymm2, 64(%rdi) 1860; XOP-NEXT: vmovups %ymm1, (%rdi) 1861; XOP-NEXT: vzeroupper 1862; XOP-NEXT: retq 1863 %s1 = load <8 x i32>, <8 x i32>* %q1, align 4 1864 %s2 = load <8 x i32>, <8 x i32>* %q2, align 4 1865 %s3 = load <8 x i32>, <8 x i32>* %q3, align 4 1866 %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> 1867 %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> 1868 %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> 1869 store <24 x i32> %interleaved, <24 x i32>* %p, align 4 1870 ret void 1871} 1872 1873; Repeat each element x 3 of <16 x i8> a0 + a1 to create a <96 x i8>. 1874define void @splat3_128(<16 x i8> %a0, <16 x i8> %a1, <96 x i8> *%a2) { 1875; SSE2-LABEL: splat3_128: 1876; SSE2: # %bb.0: 1877; SSE2-NEXT: pxor %xmm4, %xmm4 1878; SSE2-NEXT: movdqa %xmm0, %xmm3 1879; 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] 1880; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[2,3,3,3,4,5,6,7] 1881; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,4,4,4,5] 1882; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,1] 1883; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,1,4,5,6,7] 1884; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,6,6] 1885; SSE2-NEXT: packuswb %xmm5, %xmm2 1886; SSE2-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15] 1887; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,0,1] 1888; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,1,4,5,6,7] 1889; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,6,6] 1890; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3] 1891; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,1,2,2,4,5,6,7] 1892; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,7,7,7] 1893; SSE2-NEXT: packuswb %xmm5, %xmm3 1894; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm0[2,3,3,3,4,5,6,7] 1895; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,5] 1896; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3] 1897; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,1,2,2,4,5,6,7] 1898; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,7,7] 1899; SSE2-NEXT: packuswb %xmm0, %xmm5 1900; SSE2-NEXT: movdqa %xmm1, %xmm0 1901; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7] 1902; SSE2-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[2,3,3,3,4,5,6,7] 1903; SSE2-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,5] 1904; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,1,0,1] 1905; SSE2-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,0,0,1,4,5,6,7] 1906; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,5,6,6] 1907; SSE2-NEXT: packuswb %xmm6, %xmm7 1908; SSE2-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15] 1909; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,1,0,1] 1910; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,1,4,5,6,7] 1911; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,6,6] 1912; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3] 1913; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,1,2,2,4,5,6,7] 1914; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,7,7] 1915; SSE2-NEXT: packuswb %xmm4, %xmm0 1916; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm1[2,3,3,3,4,5,6,7] 1917; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,5] 1918; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3] 1919; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,1,2,2,4,5,6,7] 1920; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,7,7,7] 1921; SSE2-NEXT: packuswb %xmm1, %xmm4 1922; SSE2-NEXT: movdqa %xmm4, 80(%rdi) 1923; SSE2-NEXT: movdqa %xmm0, 64(%rdi) 1924; SSE2-NEXT: movdqa %xmm7, 48(%rdi) 1925; SSE2-NEXT: movdqa %xmm5, 32(%rdi) 1926; SSE2-NEXT: movdqa %xmm3, 16(%rdi) 1927; SSE2-NEXT: movdqa %xmm2, (%rdi) 1928; SSE2-NEXT: retq 1929; 1930; SSE42-LABEL: splat3_128: 1931; SSE42: # %bb.0: 1932; SSE42-NEXT: movdqa {{.*#+}} xmm2 = [0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5] 1933; SSE42-NEXT: movdqa %xmm0, %xmm3 1934; SSE42-NEXT: pshufb %xmm2, %xmm3 1935; SSE42-NEXT: movdqa {{.*#+}} xmm4 = [5,5,6,6,6,7,7,7,8,8,8,9,9,9,10,10] 1936; SSE42-NEXT: movdqa %xmm0, %xmm5 1937; SSE42-NEXT: pshufb %xmm4, %xmm5 1938; SSE42-NEXT: movdqa {{.*#+}} xmm6 = [10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15] 1939; SSE42-NEXT: pshufb %xmm6, %xmm0 1940; SSE42-NEXT: movdqa %xmm1, %xmm7 1941; SSE42-NEXT: pshufb %xmm2, %xmm7 1942; SSE42-NEXT: movdqa %xmm1, %xmm2 1943; SSE42-NEXT: pshufb %xmm4, %xmm2 1944; SSE42-NEXT: pshufb %xmm6, %xmm1 1945; SSE42-NEXT: movdqa %xmm1, 80(%rdi) 1946; SSE42-NEXT: movdqa %xmm2, 64(%rdi) 1947; SSE42-NEXT: movdqa %xmm7, 48(%rdi) 1948; SSE42-NEXT: movdqa %xmm0, 32(%rdi) 1949; SSE42-NEXT: movdqa %xmm5, 16(%rdi) 1950; SSE42-NEXT: movdqa %xmm3, (%rdi) 1951; SSE42-NEXT: retq 1952; 1953; AVX1-LABEL: splat3_128: 1954; AVX1: # %bb.0: 1955; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm1[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 1956; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 1957; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm1[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 1958; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 1959; AVX1-NEXT: vpalignr {{.*#+}} xmm6 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 1960; AVX1-NEXT: vpalignr {{.*#+}} xmm7 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 1961; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4] 1962; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4] 1963; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm0[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 1964; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm1[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 1965; AVX1-NEXT: vpalignr {{.*#+}} xmm1 = xmm7[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 1966; AVX1-NEXT: vpalignr {{.*#+}} xmm0 = xmm6[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 1967; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 1968; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 1969; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4] 1970; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4] 1971; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5] 1972; AVX1-NEXT: vpshufb %xmm6, %xmm3, %xmm3 1973; AVX1-NEXT: vpshufb %xmm6, %xmm0, %xmm0 1974; AVX1-NEXT: vpshufb %xmm6, %xmm5, %xmm5 1975; AVX1-NEXT: vpshufb %xmm6, %xmm1, %xmm1 1976; AVX1-NEXT: vpshufb %xmm6, %xmm2, %xmm2 1977; AVX1-NEXT: vpshufb %xmm6, %xmm4, %xmm4 1978; AVX1-NEXT: vmovdqa %xmm4, 80(%rdi) 1979; AVX1-NEXT: vmovdqa %xmm2, 64(%rdi) 1980; AVX1-NEXT: vmovdqa %xmm1, 48(%rdi) 1981; AVX1-NEXT: vmovdqa %xmm5, 32(%rdi) 1982; AVX1-NEXT: vmovdqa %xmm3, 16(%rdi) 1983; AVX1-NEXT: vmovdqa %xmm0, (%rdi) 1984; AVX1-NEXT: retq 1985; 1986; AVX2-LABEL: splat3_128: 1987; AVX2: # %bb.0: 1988; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0 1989; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0 1990; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20,21] 1991; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,26] 1992; AVX2-NEXT: vpalignr {{.*#+}} ymm3 = ymm1[5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4],ymm1[21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20] 1993; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[5,6,7,8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4],ymm2[21,22,23,24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20] 1994; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm0[5,6,7,8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4],ymm0[21,22,23,24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20] 1995; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm3[5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4],ymm3[21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20] 1996; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[5,6,7,8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4],ymm1[21,22,23,24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20] 1997; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm2[5,6,7,8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4],ymm2[21,22,23,24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20] 1998; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 1999; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = [0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5,0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5] 2000; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3 2001; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7] 2002; AVX2-NEXT: vpshufb %ymm4, %ymm0, %ymm0 2003; AVX2-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3] 2004; AVX2-NEXT: vpshufb %ymm4, %ymm1, %ymm1 2005; AVX2-NEXT: vmovdqa %ymm1, 64(%rdi) 2006; AVX2-NEXT: vmovdqa %ymm0, 32(%rdi) 2007; AVX2-NEXT: vmovdqa %ymm3, (%rdi) 2008; AVX2-NEXT: vzeroupper 2009; AVX2-NEXT: retq 2010; 2011; XOP-LABEL: splat3_128: 2012; XOP: # %bb.0: 2013; XOP-NEXT: vpalignr {{.*#+}} xmm2 = xmm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2014; XOP-NEXT: vpalignr {{.*#+}} xmm3 = xmm1[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2015; XOP-NEXT: vpalignr {{.*#+}} xmm4 = xmm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2016; XOP-NEXT: vpalignr {{.*#+}} xmm5 = xmm1[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2017; XOP-NEXT: vpalignr {{.*#+}} xmm6 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 2018; XOP-NEXT: vpalignr {{.*#+}} xmm7 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 2019; XOP-NEXT: vpalignr {{.*#+}} xmm8 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4] 2020; XOP-NEXT: vpalignr {{.*#+}} xmm2 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4] 2021; XOP-NEXT: vpalignr {{.*#+}} xmm5 = xmm1[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 2022; XOP-NEXT: vpalignr {{.*#+}} xmm4 = xmm0[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 2023; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [5,16,11,6,17,12,7,18,13,8,19,14,9,20,15,10] 2024; XOP-NEXT: vpperm %xmm3, %xmm4, %xmm2, %xmm2 2025; XOP-NEXT: vpperm %xmm3, %xmm0, %xmm7, %xmm0 2026; XOP-NEXT: vpperm %xmm3, %xmm7, %xmm4, %xmm4 2027; XOP-NEXT: vpperm %xmm3, %xmm1, %xmm6, %xmm1 2028; XOP-NEXT: vpperm %xmm3, %xmm5, %xmm8, %xmm7 2029; XOP-NEXT: vpperm %xmm3, %xmm6, %xmm5, %xmm3 2030; XOP-NEXT: vmovdqa %xmm3, 80(%rdi) 2031; XOP-NEXT: vmovdqa %xmm7, 64(%rdi) 2032; XOP-NEXT: vmovdqa %xmm1, 48(%rdi) 2033; XOP-NEXT: vmovdqa %xmm4, 32(%rdi) 2034; XOP-NEXT: vmovdqa %xmm2, 16(%rdi) 2035; XOP-NEXT: vmovdqa %xmm0, (%rdi) 2036; XOP-NEXT: retq 2037 %1 = shufflevector <16 x i8> %a0, <16 x i8> %a1, <64 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 2038 %2 = shufflevector <16 x i8> %a0, <16 x i8> %a1, <64 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 2039 %3 = shufflevector <64 x i8> %1, <64 x i8> %2, <96 x i32> <i32 0, i32 32, i32 64, i32 1, i32 33, i32 65, i32 2, i32 34, i32 66, i32 3, i32 35, i32 67, i32 4, i32 36, i32 68, i32 5, i32 37, i32 69, i32 6, i32 38, i32 70, i32 7, i32 39, i32 71, i32 8, i32 40, i32 72, i32 9, i32 41, i32 73, i32 10, i32 42, i32 74, i32 11, i32 43, i32 75, i32 12, i32 44, i32 76, i32 13, i32 45, i32 77, i32 14, i32 46, i32 78, i32 15, i32 47, i32 79, i32 16, i32 48, i32 80, i32 17, i32 49, i32 81, i32 18, i32 50, i32 82, i32 19, i32 51, i32 83, i32 20, i32 52, i32 84, i32 21, i32 53, i32 85, i32 22, i32 54, i32 86, i32 23, i32 55, i32 87, i32 24, i32 56, i32 88, i32 25, i32 57, i32 89, i32 26, i32 58, i32 90, i32 27, i32 59, i32 91, i32 28, i32 60, i32 92, i32 29, i32 61, i32 93, i32 30, i32 62, i32 94, i32 31, i32 63, i32 95> 2040 store <96 x i8> %3, <96 x i8>* %a2 2041 ret void 2042} 2043 2044; Repeat each element x 3 of <32 x i8> a0 to create a <96 x i8>. 2045define void @splat3_256(<32 x i8> %a0, <96 x i8> *%a1) { 2046; SSE2-LABEL: splat3_256: 2047; SSE2: # %bb.0: 2048; SSE2-NEXT: pxor %xmm4, %xmm4 2049; SSE2-NEXT: movdqa %xmm0, %xmm3 2050; 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] 2051; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm3[2,3,3,3,4,5,6,7] 2052; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm2[0,1,2,3,4,4,4,5] 2053; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm3[0,1,0,1] 2054; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm2[0,0,0,1,4,5,6,7] 2055; SSE2-NEXT: pshufhw {{.*#+}} xmm2 = xmm2[0,1,2,3,5,5,6,6] 2056; SSE2-NEXT: packuswb %xmm5, %xmm2 2057; SSE2-NEXT: punpckhbw {{.*#+}} xmm0 = xmm0[8],xmm4[8],xmm0[9],xmm4[9],xmm0[10],xmm4[10],xmm0[11],xmm4[11],xmm0[12],xmm4[12],xmm0[13],xmm4[13],xmm0[14],xmm4[14],xmm0[15],xmm4[15] 2058; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm0[0,1,0,1] 2059; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm5[0,0,0,1,4,5,6,7] 2060; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,5,5,6,6] 2061; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[2,3,2,3] 2062; SSE2-NEXT: pshuflw {{.*#+}} xmm3 = xmm3[1,1,2,2,4,5,6,7] 2063; SSE2-NEXT: pshufhw {{.*#+}} xmm3 = xmm3[0,1,2,3,6,7,7,7] 2064; SSE2-NEXT: packuswb %xmm5, %xmm3 2065; SSE2-NEXT: pshuflw {{.*#+}} xmm5 = xmm0[2,3,3,3,4,5,6,7] 2066; SSE2-NEXT: pshufhw {{.*#+}} xmm5 = xmm5[0,1,2,3,4,4,4,5] 2067; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3] 2068; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,1,2,2,4,5,6,7] 2069; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,7,7] 2070; SSE2-NEXT: packuswb %xmm0, %xmm5 2071; SSE2-NEXT: movdqa %xmm1, %xmm0 2072; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm4[0],xmm0[1],xmm4[1],xmm0[2],xmm4[2],xmm0[3],xmm4[3],xmm0[4],xmm4[4],xmm0[5],xmm4[5],xmm0[6],xmm4[6],xmm0[7],xmm4[7] 2073; SSE2-NEXT: pshuflw {{.*#+}} xmm6 = xmm0[2,3,3,3,4,5,6,7] 2074; SSE2-NEXT: pshufhw {{.*#+}} xmm6 = xmm6[0,1,2,3,4,4,4,5] 2075; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm0[0,1,0,1] 2076; SSE2-NEXT: pshuflw {{.*#+}} xmm7 = xmm7[0,0,0,1,4,5,6,7] 2077; SSE2-NEXT: pshufhw {{.*#+}} xmm7 = xmm7[0,1,2,3,5,5,6,6] 2078; SSE2-NEXT: packuswb %xmm6, %xmm7 2079; SSE2-NEXT: punpckhbw {{.*#+}} xmm1 = xmm1[8],xmm4[8],xmm1[9],xmm4[9],xmm1[10],xmm4[10],xmm1[11],xmm4[11],xmm1[12],xmm4[12],xmm1[13],xmm4[13],xmm1[14],xmm4[14],xmm1[15],xmm4[15] 2080; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm1[0,1,0,1] 2081; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm4[0,0,0,1,4,5,6,7] 2082; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,5,5,6,6] 2083; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[2,3,2,3] 2084; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[1,1,2,2,4,5,6,7] 2085; SSE2-NEXT: pshufhw {{.*#+}} xmm0 = xmm0[0,1,2,3,6,7,7,7] 2086; SSE2-NEXT: packuswb %xmm4, %xmm0 2087; SSE2-NEXT: pshuflw {{.*#+}} xmm4 = xmm1[2,3,3,3,4,5,6,7] 2088; SSE2-NEXT: pshufhw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,4,4,5] 2089; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3] 2090; SSE2-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[1,1,2,2,4,5,6,7] 2091; SSE2-NEXT: pshufhw {{.*#+}} xmm1 = xmm1[0,1,2,3,6,7,7,7] 2092; SSE2-NEXT: packuswb %xmm1, %xmm4 2093; SSE2-NEXT: movdqa %xmm4, 80(%rdi) 2094; SSE2-NEXT: movdqa %xmm0, 64(%rdi) 2095; SSE2-NEXT: movdqa %xmm7, 48(%rdi) 2096; SSE2-NEXT: movdqa %xmm5, 32(%rdi) 2097; SSE2-NEXT: movdqa %xmm3, 16(%rdi) 2098; SSE2-NEXT: movdqa %xmm2, (%rdi) 2099; SSE2-NEXT: retq 2100; 2101; SSE42-LABEL: splat3_256: 2102; SSE42: # %bb.0: 2103; SSE42-NEXT: movdqa {{.*#+}} xmm2 = [0,0,0,1,1,1,2,2,2,3,3,3,4,4,4,5] 2104; SSE42-NEXT: movdqa %xmm0, %xmm3 2105; SSE42-NEXT: pshufb %xmm2, %xmm3 2106; SSE42-NEXT: movdqa {{.*#+}} xmm4 = [5,5,6,6,6,7,7,7,8,8,8,9,9,9,10,10] 2107; SSE42-NEXT: movdqa %xmm0, %xmm5 2108; SSE42-NEXT: pshufb %xmm4, %xmm5 2109; SSE42-NEXT: movdqa {{.*#+}} xmm6 = [10,11,11,11,12,12,12,13,13,13,14,14,14,15,15,15] 2110; SSE42-NEXT: pshufb %xmm6, %xmm0 2111; SSE42-NEXT: movdqa %xmm1, %xmm7 2112; SSE42-NEXT: pshufb %xmm2, %xmm7 2113; SSE42-NEXT: movdqa %xmm1, %xmm2 2114; SSE42-NEXT: pshufb %xmm4, %xmm2 2115; SSE42-NEXT: pshufb %xmm6, %xmm1 2116; SSE42-NEXT: movdqa %xmm1, 80(%rdi) 2117; SSE42-NEXT: movdqa %xmm2, 64(%rdi) 2118; SSE42-NEXT: movdqa %xmm7, 48(%rdi) 2119; SSE42-NEXT: movdqa %xmm0, 32(%rdi) 2120; SSE42-NEXT: movdqa %xmm5, 16(%rdi) 2121; SSE42-NEXT: movdqa %xmm3, (%rdi) 2122; SSE42-NEXT: retq 2123; 2124; AVX1-LABEL: splat3_256: 2125; AVX1: # %bb.0: 2126; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm1 2127; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm1[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2128; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2129; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm1[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2130; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2131; AVX1-NEXT: vpalignr {{.*#+}} xmm6 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 2132; AVX1-NEXT: vpalignr {{.*#+}} xmm7 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 2133; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4] 2134; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4] 2135; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm0[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 2136; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm1[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 2137; AVX1-NEXT: vpalignr {{.*#+}} xmm1 = xmm7[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 2138; AVX1-NEXT: vpalignr {{.*#+}} xmm0 = xmm6[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 2139; AVX1-NEXT: vpalignr {{.*#+}} xmm2 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 2140; AVX1-NEXT: vpalignr {{.*#+}} xmm3 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 2141; AVX1-NEXT: vpalignr {{.*#+}} xmm4 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm7[0,1,2,3,4] 2142; AVX1-NEXT: vpalignr {{.*#+}} xmm5 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm6[0,1,2,3,4] 2143; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5] 2144; AVX1-NEXT: vpshufb %xmm6, %xmm3, %xmm3 2145; AVX1-NEXT: vpshufb %xmm6, %xmm0, %xmm0 2146; AVX1-NEXT: vpshufb %xmm6, %xmm5, %xmm5 2147; AVX1-NEXT: vpshufb %xmm6, %xmm1, %xmm1 2148; AVX1-NEXT: vpshufb %xmm6, %xmm2, %xmm2 2149; AVX1-NEXT: vpshufb %xmm6, %xmm4, %xmm4 2150; AVX1-NEXT: vmovdqa %xmm4, 80(%rdi) 2151; AVX1-NEXT: vmovdqa %xmm2, 64(%rdi) 2152; AVX1-NEXT: vmovdqa %xmm1, 48(%rdi) 2153; AVX1-NEXT: vmovdqa %xmm5, 32(%rdi) 2154; AVX1-NEXT: vmovdqa %xmm3, 16(%rdi) 2155; AVX1-NEXT: vmovdqa %xmm0, (%rdi) 2156; AVX1-NEXT: vzeroupper 2157; AVX1-NEXT: retq 2158; 2159; AVX2-LABEL: splat3_256: 2160; AVX2: # %bb.0: 2161; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5,22,23,24,25,26,27,28,29,30,31,16,17,18,19,20,21] 2162; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10,27,28,29,30,31,16,17,18,19,20,21,22,23,24,25,26] 2163; AVX2-NEXT: vpalignr {{.*#+}} ymm3 = ymm1[5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4],ymm1[21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20] 2164; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm2[5,6,7,8,9,10,11,12,13,14,15],ymm1[0,1,2,3,4],ymm2[21,22,23,24,25,26,27,28,29,30,31],ymm1[16,17,18,19,20] 2165; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm0[5,6,7,8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4],ymm0[21,22,23,24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20] 2166; AVX2-NEXT: vpalignr {{.*#+}} ymm0 = ymm3[5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1,2,3,4],ymm3[21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17,18,19,20] 2167; AVX2-NEXT: vpalignr {{.*#+}} ymm1 = ymm1[5,6,7,8,9,10,11,12,13,14,15],ymm2[0,1,2,3,4],ymm1[21,22,23,24,25,26,27,28,29,30,31],ymm2[16,17,18,19,20] 2168; AVX2-NEXT: vpalignr {{.*#+}} ymm2 = ymm2[5,6,7,8,9,10,11,12,13,14,15],ymm3[0,1,2,3,4],ymm2[21,22,23,24,25,26,27,28,29,30,31],ymm3[16,17,18,19,20] 2169; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm3 2170; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = [0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5,0,11,6,1,12,7,2,13,8,3,14,9,4,15,10,5] 2171; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3 2172; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm2[0,1,2,3],ymm0[4,5,6,7] 2173; AVX2-NEXT: vpshufb %ymm4, %ymm0, %ymm0 2174; AVX2-NEXT: vperm2i128 {{.*#+}} ymm1 = ymm1[2,3],ymm2[2,3] 2175; AVX2-NEXT: vpshufb %ymm4, %ymm1, %ymm1 2176; AVX2-NEXT: vmovdqa %ymm1, 64(%rdi) 2177; AVX2-NEXT: vmovdqa %ymm0, 32(%rdi) 2178; AVX2-NEXT: vmovdqa %ymm3, (%rdi) 2179; AVX2-NEXT: vzeroupper 2180; AVX2-NEXT: retq 2181; 2182; XOP-LABEL: splat3_256: 2183; XOP: # %bb.0: 2184; XOP-NEXT: vpalignr {{.*#+}} xmm1 = xmm0[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2185; XOP-NEXT: vextractf128 $1, %ymm0, %xmm2 2186; XOP-NEXT: vpalignr {{.*#+}} xmm3 = xmm2[6,7,8,9,10,11,12,13,14,15,0,1,2,3,4,5] 2187; XOP-NEXT: vpalignr {{.*#+}} xmm4 = xmm0[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2188; XOP-NEXT: vpalignr {{.*#+}} xmm5 = xmm2[11,12,13,14,15,0,1,2,3,4,5,6,7,8,9,10] 2189; XOP-NEXT: vpalignr {{.*#+}} xmm6 = xmm3[5,6,7,8,9,10,11,12,13,14,15],xmm2[0,1,2,3,4] 2190; XOP-NEXT: vpalignr {{.*#+}} xmm7 = xmm1[5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1,2,3,4] 2191; XOP-NEXT: vpalignr {{.*#+}} xmm8 = xmm5[5,6,7,8,9,10,11,12,13,14,15],xmm3[0,1,2,3,4] 2192; XOP-NEXT: vpalignr {{.*#+}} xmm1 = xmm4[5,6,7,8,9,10,11,12,13,14,15],xmm1[0,1,2,3,4] 2193; XOP-NEXT: vpalignr {{.*#+}} xmm5 = xmm2[5,6,7,8,9,10,11,12,13,14,15],xmm5[0,1,2,3,4] 2194; XOP-NEXT: vpalignr {{.*#+}} xmm4 = xmm0[5,6,7,8,9,10,11,12,13,14,15],xmm4[0,1,2,3,4] 2195; XOP-NEXT: vmovdqa {{.*#+}} xmm3 = [5,16,11,6,17,12,7,18,13,8,19,14,9,20,15,10] 2196; XOP-NEXT: vpperm %xmm3, %xmm4, %xmm1, %xmm1 2197; XOP-NEXT: vpperm %xmm3, %xmm0, %xmm7, %xmm0 2198; XOP-NEXT: vpperm %xmm3, %xmm7, %xmm4, %xmm4 2199; XOP-NEXT: vpperm %xmm3, %xmm2, %xmm6, %xmm2 2200; XOP-NEXT: vpperm %xmm3, %xmm5, %xmm8, %xmm7 2201; XOP-NEXT: vpperm %xmm3, %xmm6, %xmm5, %xmm3 2202; XOP-NEXT: vmovdqa %xmm3, 80(%rdi) 2203; XOP-NEXT: vmovdqa %xmm7, 64(%rdi) 2204; XOP-NEXT: vmovdqa %xmm2, 48(%rdi) 2205; XOP-NEXT: vmovdqa %xmm4, 32(%rdi) 2206; XOP-NEXT: vmovdqa %xmm1, 16(%rdi) 2207; XOP-NEXT: vmovdqa %xmm0, (%rdi) 2208; XOP-NEXT: vzeroupper 2209; XOP-NEXT: retq 2210 %1 = shufflevector <32 x i8> %a0, <32 x i8> undef, <64 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 2211 %2 = shufflevector <32 x i8> %a0, <32 x i8> undef, <64 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 2212 %3 = shufflevector <64 x i8> %1, <64 x i8> %2, <96 x i32> <i32 0, i32 32, i32 64, i32 1, i32 33, i32 65, i32 2, i32 34, i32 66, i32 3, i32 35, i32 67, i32 4, i32 36, i32 68, i32 5, i32 37, i32 69, i32 6, i32 38, i32 70, i32 7, i32 39, i32 71, i32 8, i32 40, i32 72, i32 9, i32 41, i32 73, i32 10, i32 42, i32 74, i32 11, i32 43, i32 75, i32 12, i32 44, i32 76, i32 13, i32 45, i32 77, i32 14, i32 46, i32 78, i32 15, i32 47, i32 79, i32 16, i32 48, i32 80, i32 17, i32 49, i32 81, i32 18, i32 50, i32 82, i32 19, i32 51, i32 83, i32 20, i32 52, i32 84, i32 21, i32 53, i32 85, i32 22, i32 54, i32 86, i32 23, i32 55, i32 87, i32 24, i32 56, i32 88, i32 25, i32 57, i32 89, i32 26, i32 58, i32 90, i32 27, i32 59, i32 91, i32 28, i32 60, i32 92, i32 29, i32 61, i32 93, i32 30, i32 62, i32 94, i32 31, i32 63, i32 95> 2213 store <96 x i8> %3, <96 x i8>* %a1 2214 ret void 2215} 2216 2217; D79987 2218define <16 x i32> @splat_v3i32(<3 x i32>* %ptr) { 2219; SSE2-LABEL: splat_v3i32: 2220; SSE2: # %bb.0: 2221; SSE2-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero 2222; SSE2-NEXT: xorps %xmm1, %xmm1 2223; SSE2-NEXT: movss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3] 2224; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm1[1,1,0,1] 2225; SSE2-NEXT: andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 2226; SSE2-NEXT: xorps %xmm1, %xmm1 2227; SSE2-NEXT: xorps %xmm3, %xmm3 2228; SSE2-NEXT: retq 2229; 2230; SSE42-LABEL: splat_v3i32: 2231; SSE42: # %bb.0: 2232; SSE42-NEXT: movq {{.*#+}} xmm0 = mem[0],zero 2233; SSE42-NEXT: pxor %xmm1, %xmm1 2234; SSE42-NEXT: pxor %xmm2, %xmm2 2235; SSE42-NEXT: pblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7] 2236; SSE42-NEXT: pblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3],xmm1[4,5,6,7] 2237; SSE42-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,0,1] 2238; SSE42-NEXT: pxor %xmm1, %xmm1 2239; SSE42-NEXT: xorps %xmm3, %xmm3 2240; SSE42-NEXT: retq 2241; 2242; AVX1-LABEL: splat_v3i32: 2243; AVX1: # %bb.0: 2244; AVX1-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] 2245; AVX1-NEXT: vxorps %xmm2, %xmm2, %xmm2 2246; AVX1-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3,4,5,6,7] 2247; AVX1-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6,7] 2248; AVX1-NEXT: retq 2249; 2250; AVX2-SLOW-LABEL: splat_v3i32: 2251; AVX2-SLOW: # %bb.0: 2252; AVX2-SLOW-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero 2253; AVX2-SLOW-NEXT: vxorps %xmm2, %xmm2, %xmm2 2254; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3,4,5,6,7] 2255; AVX2-SLOW-NEXT: vbroadcastss %xmm1, %xmm1 2256; AVX2-SLOW-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6,7] 2257; AVX2-SLOW-NEXT: retq 2258; 2259; AVX2-FAST-LABEL: splat_v3i32: 2260; AVX2-FAST: # %bb.0: 2261; AVX2-FAST-NEXT: vmovq {{.*#+}} xmm1 = mem[0],zero 2262; AVX2-FAST-NEXT: vpxor %xmm0, %xmm0, %xmm0 2263; AVX2-FAST-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2,3,4,5,6,7] 2264; AVX2-FAST-NEXT: vpshufb {{.*#+}} ymm1 = zero,zero,zero,zero,zero,zero,zero,zero,ymm1[0,1,2,3],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero 2265; AVX2-FAST-NEXT: retq 2266; 2267; XOP-LABEL: splat_v3i32: 2268; XOP: # %bb.0: 2269; XOP-NEXT: vmovddup {{.*#+}} xmm1 = mem[0,0] 2270; XOP-NEXT: vxorps %xmm2, %xmm2, %xmm2 2271; XOP-NEXT: vblendps {{.*#+}} ymm0 = ymm2[0],ymm1[1],ymm2[2,3,4,5,6,7] 2272; XOP-NEXT: vblendps {{.*#+}} ymm1 = ymm2[0,1],ymm1[2],ymm2[3,4,5,6,7] 2273; XOP-NEXT: retq 2274 %1 = load <3 x i32>, <3 x i32>* %ptr, align 1 2275 %2 = shufflevector <3 x i32> %1, <3 x i32> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef> 2276 %3 = shufflevector <16 x i32> <i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0>, <16 x i32> %2, <16 x i32> <i32 0, i32 17, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 11, i32 12, i32 13, i32 14, i32 15> 2277 ret <16 x i32 > %3 2278} 2279 2280define <2 x double> @wrongorder(<4 x double> %A, <8 x double>* %P) #0 { 2281; SSE2-LABEL: wrongorder: 2282; SSE2: # %bb.0: 2283; SSE2-NEXT: movlhps {{.*#+}} xmm0 = xmm0[0,0] 2284; SSE2-NEXT: movaps %xmm0, 48(%rdi) 2285; SSE2-NEXT: movaps %xmm0, 32(%rdi) 2286; SSE2-NEXT: movaps %xmm0, 16(%rdi) 2287; SSE2-NEXT: movaps %xmm0, (%rdi) 2288; SSE2-NEXT: retq 2289; 2290; SSE42-LABEL: wrongorder: 2291; SSE42: # %bb.0: 2292; SSE42-NEXT: movddup {{.*#+}} xmm0 = xmm0[0,0] 2293; SSE42-NEXT: movapd %xmm0, 48(%rdi) 2294; SSE42-NEXT: movapd %xmm0, 32(%rdi) 2295; SSE42-NEXT: movapd %xmm0, 16(%rdi) 2296; SSE42-NEXT: movapd %xmm0, (%rdi) 2297; SSE42-NEXT: retq 2298; 2299; AVX1-LABEL: wrongorder: 2300; AVX1: # %bb.0: 2301; AVX1-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] 2302; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1 2303; AVX1-NEXT: vmovaps %ymm1, 32(%rdi) 2304; AVX1-NEXT: vmovaps %ymm1, (%rdi) 2305; AVX1-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 2306; AVX1-NEXT: vzeroupper 2307; AVX1-NEXT: retq 2308; 2309; AVX2-LABEL: wrongorder: 2310; AVX2: # %bb.0: 2311; AVX2-NEXT: vbroadcastsd %xmm0, %ymm0 2312; AVX2-NEXT: vmovaps %ymm0, 32(%rdi) 2313; AVX2-NEXT: vmovaps %ymm0, (%rdi) 2314; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 2315; AVX2-NEXT: vzeroupper 2316; AVX2-NEXT: retq 2317; 2318; XOP-LABEL: wrongorder: 2319; XOP: # %bb.0: 2320; XOP-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0] 2321; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm1 2322; XOP-NEXT: vmovaps %ymm1, 32(%rdi) 2323; XOP-NEXT: vmovaps %ymm1, (%rdi) 2324; XOP-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 2325; XOP-NEXT: vzeroupper 2326; XOP-NEXT: retq 2327 %shuffle = shufflevector <4 x double> %A, <4 x double> %A, <8 x i32> zeroinitializer 2328 store <8 x double> %shuffle, <8 x double>* %P, align 64 2329 %m2 = load <8 x double>, <8 x double>* %P, align 64 2330 store <8 x double> %m2, <8 x double>* %P, align 64 2331 %m3 = load <8 x double>, <8 x double>* %P, align 64 2332 %m4 = shufflevector <8 x double> %m3, <8 x double> undef, <2 x i32> <i32 2, i32 0> 2333 ret <2 x double> %m4 2334} 2335 2336define void @PR41097() { 2337; SSE2-LABEL: PR41097: 2338; SSE2: # %bb.0: 2339; SSE2-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero 2340; SSE2-NEXT: punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7] 2341; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,0,2,3,4,5,6,7] 2342; SSE2-NEXT: psrad $24, %xmm0 2343; SSE2-NEXT: pxor %xmm1, %xmm1 2344; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1] 2345; SSE2-NEXT: movdqu %xmm0, (%rax) 2346; SSE2-NEXT: retq 2347; 2348; SSE42-LABEL: PR41097: 2349; SSE42: # %bb.0: 2350; SSE42-NEXT: movd {{.*#+}} xmm0 = mem[0],zero,zero,zero 2351; SSE42-NEXT: pshufb {{.*#+}} xmm0 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u] 2352; SSE42-NEXT: pmovsxbd %xmm0, %xmm0 2353; SSE42-NEXT: pmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero 2354; SSE42-NEXT: movdqu %xmm0, (%rax) 2355; SSE42-NEXT: retq 2356; 2357; AVX-LABEL: PR41097: 2358; AVX: # %bb.0: 2359; AVX-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero 2360; AVX-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u] 2361; AVX-NEXT: vpmovsxbd %xmm0, %xmm0 2362; AVX-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero 2363; AVX-NEXT: vmovdqu %xmm0, (%rax) 2364; AVX-NEXT: retq 2365; 2366; XOP-LABEL: PR41097: 2367; XOP: # %bb.0: 2368; XOP-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero 2369; XOP-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,3,u,u,u,u,u,u,u,u,u,u,u,u,u,u] 2370; XOP-NEXT: vpmovsxbd %xmm0, %xmm0 2371; XOP-NEXT: vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero 2372; XOP-NEXT: vmovdqu %xmm0, (%rax) 2373; XOP-NEXT: retq 2374 %wide.vec = load <6 x i8>, <6 x i8>* undef, align 1 2375 %strided.vec = shufflevector <6 x i8> %wide.vec, <6 x i8> undef, <2 x i32> <i32 0, i32 3> 2376 %tmp = sext <2 x i8> %strided.vec to <2 x i32> 2377 %tmp7 = zext <2 x i32> %tmp to <2 x i64> 2378 store <2 x i64> %tmp7, <2 x i64>* undef, align 8 2379 ret void 2380} 2381 2382define void @D107009(<64 x i32>* %input, <64 x i32>* %output) { 2383; SSE-LABEL: D107009: 2384; SSE: # %bb.0: 2385; SSE-NEXT: movdqa 16(%rdi), %xmm0 2386; SSE-NEXT: movdqa 80(%rdi), %xmm1 2387; SSE-NEXT: movdqa 144(%rdi), %xmm2 2388; SSE-NEXT: movdqa 208(%rdi), %xmm3 2389; SSE-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1] 2390; SSE-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1] 2391; SSE-NEXT: punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0] 2392; SSE-NEXT: punpckldq {{.*#+}} xmm3 = xmm3[0],mem[0],xmm3[1],mem[1] 2393; SSE-NEXT: punpckldq {{.*#+}} xmm2 = xmm2[0],mem[0],xmm2[1],mem[1] 2394; SSE-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0] 2395; SSE-NEXT: psrld $16, %xmm2 2396; SSE-NEXT: psrld $16, %xmm0 2397; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1] 2398; SSE-NEXT: pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3] 2399; SSE-NEXT: pshufd {{.*#+}} xmm4 = xmm0[3,3,3,3] 2400; SSE-NEXT: pshufd {{.*#+}} xmm5 = xmm2[1,1,1,1] 2401; SSE-NEXT: pshufd {{.*#+}} xmm6 = xmm2[2,3,2,3] 2402; SSE-NEXT: pshufd {{.*#+}} xmm7 = xmm2[3,3,3,3] 2403; SSE-NEXT: movdqa %xmm0, 128(%rsi) 2404; SSE-NEXT: movdqa %xmm2, 144(%rsi) 2405; SSE-NEXT: movdqa %xmm0, 16(%rsi) 2406; SSE-NEXT: movdqa %xmm7, 240(%rsi) 2407; SSE-NEXT: movdqa %xmm6, 208(%rsi) 2408; SSE-NEXT: movdqa %xmm5, 176(%rsi) 2409; SSE-NEXT: movdqa %xmm4, 112(%rsi) 2410; SSE-NEXT: movdqa %xmm3, 80(%rsi) 2411; SSE-NEXT: movdqa %xmm1, 48(%rsi) 2412; SSE-NEXT: retq 2413; 2414; AVX1-LABEL: D107009: 2415; AVX1: # %bb.0: 2416; AVX1-NEXT: vmovups 96(%rdi), %ymm0 2417; AVX1-NEXT: vmovups (%rdi), %ymm1 2418; AVX1-NEXT: vmovups 128(%rdi), %ymm2 2419; AVX1-NEXT: vmovups 224(%rdi), %ymm3 2420; AVX1-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[2],mem[2] 2421; AVX1-NEXT: vunpcklps {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[1],mem[1],ymm2[4],mem[4],ymm2[5],mem[5] 2422; AVX1-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,0],ymm2[4,5],ymm3[6,4] 2423; AVX1-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5] 2424; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1 2425; AVX1-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] 2426; AVX1-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4] 2427; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0 2428; AVX1-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7] 2429; AVX1-NEXT: vpsrld $16, %xmm0, %xmm0 2430; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1 2431; AVX1-NEXT: vpsrld $16, %xmm1, %xmm1 2432; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 2433; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,1,3] 2434; AVX1-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,1,3,3] 2435; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2 2436; AVX1-NEXT: vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7] 2437; AVX1-NEXT: vpermilps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7] 2438; AVX1-NEXT: vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2] 2439; AVX1-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1] 2440; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 2441; AVX1-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3] 2442; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7 2443; AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 2444; AVX1-NEXT: vmovups %ymm0, (%rsi) 2445; AVX1-NEXT: vmovups %ymm1, 128(%rsi) 2446; AVX1-NEXT: vmovups %ymm7, 96(%rsi) 2447; AVX1-NEXT: vmovups %ymm6, 32(%rsi) 2448; AVX1-NEXT: vmovupd %ymm5, 192(%rsi) 2449; AVX1-NEXT: vmovups %ymm4, 224(%rsi) 2450; AVX1-NEXT: vmovups %ymm3, 160(%rsi) 2451; AVX1-NEXT: vmovups %ymm2, 64(%rsi) 2452; AVX1-NEXT: vzeroupper 2453; AVX1-NEXT: retq 2454; 2455; AVX2-LABEL: D107009: 2456; AVX2: # %bb.0: 2457; AVX2-NEXT: vmovdqu 64(%rdi), %ymm0 2458; AVX2-NEXT: vmovdqu 128(%rdi), %ymm1 2459; AVX2-NEXT: vmovdqu 192(%rdi), %ymm2 2460; AVX2-NEXT: vpunpckldq {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[1],mem[1],ymm2[4],mem[4],ymm2[5],mem[5] 2461; AVX2-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5] 2462; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm1 = ymm1[0],ymm2[0],ymm1[2],ymm2[2] 2463; AVX2-NEXT: vpbroadcastd 48(%rdi), %xmm2 2464; AVX2-NEXT: vpbroadcastd 16(%rdi), %ymm3 2465; AVX2-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1] 2466; AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5] 2467; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[2,2,2,2] 2468; AVX2-NEXT: vpblendd {{.*#+}} xmm0 = xmm2[0,1],xmm0[2,3] 2469; AVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1,2,3],ymm1[4,5,6,7] 2470; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0 2471; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[3,3,3,3] 2472; AVX2-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm1 2473; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[1,1,1,1] 2474; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm2 2475; AVX2-NEXT: vpmovzxdq {{.*#+}} ymm3 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero 2476; AVX2-NEXT: vpbroadcastd %xmm0, %ymm4 2477; AVX2-NEXT: vpshufd {{.*#+}} ymm5 = ymm0[3,3,3,3,7,7,7,7] 2478; AVX2-NEXT: vpshufd {{.*#+}} ymm6 = ymm0[2,3,2,3,6,7,6,7] 2479; AVX2-NEXT: vpshufd {{.*#+}} ymm7 = ymm0[1,1,1,1,5,5,5,5] 2480; AVX2-NEXT: vmovdqu %ymm0, 128(%rsi) 2481; AVX2-NEXT: vmovdqu %ymm7, 160(%rsi) 2482; AVX2-NEXT: vmovdqu %ymm6, 192(%rsi) 2483; AVX2-NEXT: vmovdqu %ymm5, 224(%rsi) 2484; AVX2-NEXT: vmovdqu %ymm4, (%rsi) 2485; AVX2-NEXT: vmovdqu %ymm3, 64(%rsi) 2486; AVX2-NEXT: vmovdqu %ymm2, 32(%rsi) 2487; AVX2-NEXT: vmovdqu %ymm1, 96(%rsi) 2488; AVX2-NEXT: vzeroupper 2489; AVX2-NEXT: retq 2490; 2491; XOP-LABEL: D107009: 2492; XOP: # %bb.0: 2493; XOP-NEXT: vmovups 96(%rdi), %ymm0 2494; XOP-NEXT: vmovups (%rdi), %ymm1 2495; XOP-NEXT: vmovups 128(%rdi), %ymm2 2496; XOP-NEXT: vmovups 224(%rdi), %ymm3 2497; XOP-NEXT: vunpcklpd {{.*#+}} ymm3 = ymm3[0],mem[0],ymm3[2],mem[2] 2498; XOP-NEXT: vunpcklps {{.*#+}} ymm2 = ymm2[0],mem[0],ymm2[1],mem[1],ymm2[4],mem[4],ymm2[5],mem[5] 2499; XOP-NEXT: vshufps {{.*#+}} ymm2 = ymm2[0,1],ymm3[2,0],ymm2[4,5],ymm3[6,4] 2500; XOP-NEXT: vunpcklps {{.*#+}} ymm1 = ymm1[0],mem[0],ymm1[1],mem[1],ymm1[4],mem[4],ymm1[5],mem[5] 2501; XOP-NEXT: vextractf128 $1, %ymm1, %xmm1 2502; XOP-NEXT: vunpcklpd {{.*#+}} ymm0 = ymm0[0],mem[0],ymm0[2],mem[2] 2503; XOP-NEXT: vpermilps {{.*#+}} ymm0 = ymm0[0,1,2,0,4,5,6,4] 2504; XOP-NEXT: vextractf128 $1, %ymm0, %xmm0 2505; XOP-NEXT: vpblendw {{.*#+}} xmm0 = xmm1[0,1,2,3],xmm0[4,5,6,7] 2506; XOP-NEXT: vpsrld $16, %xmm0, %xmm0 2507; XOP-NEXT: vextractf128 $1, %ymm2, %xmm1 2508; XOP-NEXT: vpsrld $16, %xmm1, %xmm1 2509; XOP-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm1 2510; XOP-NEXT: vpshufd {{.*#+}} xmm2 = xmm0[0,1,1,3] 2511; XOP-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[2,1,3,3] 2512; XOP-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2 2513; XOP-NEXT: vmovshdup {{.*#+}} ymm3 = ymm1[1,1,3,3,5,5,7,7] 2514; XOP-NEXT: vpermilps {{.*#+}} ymm4 = ymm1[3,3,3,3,7,7,7,7] 2515; XOP-NEXT: vpermilpd {{.*#+}} ymm5 = ymm1[0,0,3,2] 2516; XOP-NEXT: vpshufd {{.*#+}} xmm6 = xmm0[1,1,1,1] 2517; XOP-NEXT: vinsertf128 $1, %xmm6, %ymm0, %ymm6 2518; XOP-NEXT: vpshufd {{.*#+}} xmm7 = xmm0[3,3,3,3] 2519; XOP-NEXT: vinsertf128 $1, %xmm7, %ymm0, %ymm7 2520; XOP-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0 2521; XOP-NEXT: vmovups %ymm0, (%rsi) 2522; XOP-NEXT: vmovups %ymm1, 128(%rsi) 2523; XOP-NEXT: vmovups %ymm7, 96(%rsi) 2524; XOP-NEXT: vmovups %ymm6, 32(%rsi) 2525; XOP-NEXT: vmovupd %ymm5, 192(%rsi) 2526; XOP-NEXT: vmovups %ymm4, 224(%rsi) 2527; XOP-NEXT: vmovups %ymm3, 160(%rsi) 2528; XOP-NEXT: vmovups %ymm2, 64(%rsi) 2529; XOP-NEXT: vzeroupper 2530; XOP-NEXT: retq 2531 %i = load <64 x i32>, <64 x i32>* %input, align 16 2532 %i2 = shufflevector <64 x i32> %i, <64 x i32> poison, <8 x i32> <i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60> 2533 %i3 = lshr <8 x i32> %i2, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16> 2534 %i4 = add <8 x i32> zeroinitializer, %i3 2535 %i5 = shufflevector <8 x i32> %i4, <8 x i32> poison, <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> 2536 %i6 = shufflevector <16 x i32> %i5, <16 x i32> poison, <32 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, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 2537 %i7 = shufflevector <32 x i32> poison, <32 x i32> %i6, <64 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56, i32 1, i32 9, i32 17, i32 25, i32 33, i32 41, i32 49, i32 57, i32 2, i32 10, i32 18, i32 26, i32 34, i32 42, i32 50, i32 58, i32 3, i32 11, i32 19, i32 27, i32 35, i32 43, i32 51, i32 59, i32 4, i32 12, i32 20, i32 28, i32 36, i32 44, i32 52, i32 60, i32 5, i32 13, i32 21, i32 29, i32 37, i32 45, i32 53, i32 61, i32 6, i32 14, i32 22, i32 30, i32 38, i32 46, i32 54, i32 62, i32 7, i32 15, i32 23, i32 31, i32 39, i32 47, i32 55, i32 63> 2538 store <64 x i32> %i7, <64 x i32>* %output, align 16 2539 ret void 2540} 2541