445 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VL,AVX512VL-FAST-ALL4; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VL,AVX512VL-FAST-PERLANE5; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BW6; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BW7; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL8; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512BWVL9; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VBMI10; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VBMI11; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi,+avx512vl,+fast-variable-crosslane-shuffle,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VBMI12; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vbmi,+avx512vl,+fast-variable-perlane-shuffle | FileCheck %s --check-prefixes=AVX512,AVX512VBMI13 14; PR3155115; Pairs of shufflevector:trunc functions with functional equivalence.16; Ideally, the shuffles should be lowered to code with the same quality as the truncates.17 18define void @shuffle_v64i8_to_v32i8(ptr %L, ptr %S) nounwind {19; AVX512F-LABEL: shuffle_v64i8_to_v32i8:20; AVX512F: # %bb.0:21; AVX512F-NEXT: vmovdqa (%rdi), %ymm022; AVX512F-NEXT: vmovdqa 32(%rdi), %ymm123; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14]24; AVX512F-NEXT: vpshufb %ymm2, %ymm1, %ymm125; AVX512F-NEXT: vpshufb %ymm2, %ymm0, %ymm026; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]27; AVX512F-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]28; AVX512F-NEXT: vmovdqa %ymm0, (%rsi)29; AVX512F-NEXT: vzeroupper30; AVX512F-NEXT: retq31;32; AVX512VL-FAST-ALL-LABEL: shuffle_v64i8_to_v32i8:33; AVX512VL-FAST-ALL: # %bb.0:34; AVX512VL-FAST-ALL-NEXT: vmovdqa (%rdi), %ymm035; AVX512VL-FAST-ALL-NEXT: vmovdqa 32(%rdi), %ymm136; AVX512VL-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u,16,18,20,22,24,26,28,30]37; AVX512VL-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u,16,18,20,22,24,26,28,30,u,u,u,u,u,u,u,u]38; AVX512VL-FAST-ALL-NEXT: vpmovsxbq {{.*#+}} ymm2 = [0,2,5,7]39; AVX512VL-FAST-ALL-NEXT: vpermi2q %ymm1, %ymm0, %ymm240; AVX512VL-FAST-ALL-NEXT: vmovdqa %ymm2, (%rsi)41; AVX512VL-FAST-ALL-NEXT: vzeroupper42; AVX512VL-FAST-ALL-NEXT: retq43;44; AVX512VL-FAST-PERLANE-LABEL: shuffle_v64i8_to_v32i8:45; AVX512VL-FAST-PERLANE: # %bb.0:46; AVX512VL-FAST-PERLANE-NEXT: vmovdqa (%rdi), %ymm047; AVX512VL-FAST-PERLANE-NEXT: vmovdqa 32(%rdi), %ymm148; AVX512VL-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm2 = [0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14,0,2,4,6,8,10,12,14]49; AVX512VL-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm150; AVX512VL-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm051; AVX512VL-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]52; AVX512VL-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]53; AVX512VL-FAST-PERLANE-NEXT: vmovdqa %ymm0, (%rsi)54; AVX512VL-FAST-PERLANE-NEXT: vzeroupper55; AVX512VL-FAST-PERLANE-NEXT: retq56;57; AVX512BW-LABEL: shuffle_v64i8_to_v32i8:58; AVX512BW: # %bb.0:59; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm060; AVX512BW-NEXT: vpmovwb %zmm0, (%rsi)61; AVX512BW-NEXT: vzeroupper62; AVX512BW-NEXT: retq63;64; AVX512BWVL-LABEL: shuffle_v64i8_to_v32i8:65; AVX512BWVL: # %bb.0:66; AVX512BWVL-NEXT: vmovdqa64 (%rdi), %zmm067; AVX512BWVL-NEXT: vpmovwb %zmm0, (%rsi)68; AVX512BWVL-NEXT: vzeroupper69; AVX512BWVL-NEXT: retq70;71; AVX512VBMI-LABEL: shuffle_v64i8_to_v32i8:72; AVX512VBMI: # %bb.0:73; AVX512VBMI-NEXT: vmovdqa64 (%rdi), %zmm074; AVX512VBMI-NEXT: vpmovwb %zmm0, (%rsi)75; AVX512VBMI-NEXT: vzeroupper76; AVX512VBMI-NEXT: retq77 %vec = load <64 x i8>, ptr %L78 %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 32, i32 34, i32 36, i32 38, i32 40, i32 42, i32 44, i32 46, i32 48, i32 50, i32 52, i32 54, i32 56, i32 58, i32 60, i32 62>79 store <32 x i8> %strided.vec, ptr %S80 ret void81}82 83define void @trunc_v32i16_to_v32i8(ptr %L, ptr %S) nounwind {84; AVX512F-LABEL: trunc_v32i16_to_v32i8:85; AVX512F: # %bb.0:86; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero87; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero88; AVX512F-NEXT: vpmovdb %zmm1, 16(%rsi)89; AVX512F-NEXT: vpmovdb %zmm0, (%rsi)90; AVX512F-NEXT: vzeroupper91; AVX512F-NEXT: retq92;93; AVX512VL-LABEL: trunc_v32i16_to_v32i8:94; AVX512VL: # %bb.0:95; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero96; AVX512VL-NEXT: vpmovzxwd {{.*#+}} zmm1 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero,mem[8],zero,mem[9],zero,mem[10],zero,mem[11],zero,mem[12],zero,mem[13],zero,mem[14],zero,mem[15],zero97; AVX512VL-NEXT: vpmovdb %zmm1, 16(%rsi)98; AVX512VL-NEXT: vpmovdb %zmm0, (%rsi)99; AVX512VL-NEXT: vzeroupper100; AVX512VL-NEXT: retq101;102; AVX512BW-LABEL: trunc_v32i16_to_v32i8:103; AVX512BW: # %bb.0:104; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0105; AVX512BW-NEXT: vpmovwb %zmm0, (%rsi)106; AVX512BW-NEXT: vzeroupper107; AVX512BW-NEXT: retq108;109; AVX512BWVL-LABEL: trunc_v32i16_to_v32i8:110; AVX512BWVL: # %bb.0:111; AVX512BWVL-NEXT: vmovdqa64 (%rdi), %zmm0112; AVX512BWVL-NEXT: vpmovwb %zmm0, (%rsi)113; AVX512BWVL-NEXT: vzeroupper114; AVX512BWVL-NEXT: retq115;116; AVX512VBMI-LABEL: trunc_v32i16_to_v32i8:117; AVX512VBMI: # %bb.0:118; AVX512VBMI-NEXT: vmovdqa64 (%rdi), %zmm0119; AVX512VBMI-NEXT: vpmovwb %zmm0, (%rsi)120; AVX512VBMI-NEXT: vzeroupper121; AVX512VBMI-NEXT: retq122 %vec = load <64 x i8>, ptr %L123 %bc = bitcast <64 x i8> %vec to <32 x i16>124 %strided.vec = trunc <32 x i16> %bc to <32 x i8>125 store <32 x i8> %strided.vec, ptr %S126 ret void127}128 129define void @shuffle_v32i16_to_v16i16(ptr %L, ptr %S) nounwind {130; AVX512-LABEL: shuffle_v32i16_to_v16i16:131; AVX512: # %bb.0:132; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0133; AVX512-NEXT: vpmovdw %zmm0, (%rsi)134; AVX512-NEXT: vzeroupper135; AVX512-NEXT: retq136 %vec = load <32 x i16>, ptr %L137 %strided.vec = shufflevector <32 x i16> %vec, <32 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>138 store <16 x i16> %strided.vec, ptr %S139 ret void140}141 142define void @trunc_v16i32_to_v16i16(ptr %L, ptr %S) nounwind {143; AVX512-LABEL: trunc_v16i32_to_v16i16:144; AVX512: # %bb.0:145; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0146; AVX512-NEXT: vpmovdw %zmm0, (%rsi)147; AVX512-NEXT: vzeroupper148; AVX512-NEXT: retq149 %vec = load <32 x i16>, ptr %L150 %bc = bitcast <32 x i16> %vec to <16 x i32>151 %strided.vec = trunc <16 x i32> %bc to <16 x i16>152 store <16 x i16> %strided.vec, ptr %S153 ret void154}155 156define void @shuffle_v16i32_to_v8i32(ptr %L, ptr %S) nounwind {157; AVX512-LABEL: shuffle_v16i32_to_v8i32:158; AVX512: # %bb.0:159; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0160; AVX512-NEXT: vpmovqd %zmm0, (%rsi)161; AVX512-NEXT: vzeroupper162; AVX512-NEXT: retq163 %vec = load <16 x i32>, ptr %L164 %strided.vec = shufflevector <16 x i32> %vec, <16 x i32> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>165 store <8 x i32> %strided.vec, ptr %S166 ret void167}168 169define void @trunc_v8i64_to_v8i32(ptr %L, ptr %S) nounwind {170; AVX512-LABEL: trunc_v8i64_to_v8i32:171; AVX512: # %bb.0:172; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0173; AVX512-NEXT: vpmovqd %zmm0, (%rsi)174; AVX512-NEXT: vzeroupper175; AVX512-NEXT: retq176 %vec = load <16 x i32>, ptr %L177 %bc = bitcast <16 x i32> %vec to <8 x i64>178 %strided.vec = trunc <8 x i64> %bc to <8 x i32>179 store <8 x i32> %strided.vec, ptr %S180 ret void181}182 183define void @shuffle_v64i8_to_v16i8(ptr %L, ptr %S) nounwind {184; AVX512-LABEL: shuffle_v64i8_to_v16i8:185; AVX512: # %bb.0:186; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0187; AVX512-NEXT: vpmovdb %zmm0, (%rsi)188; AVX512-NEXT: vzeroupper189; AVX512-NEXT: retq190 %vec = load <64 x i8>, ptr %L191 %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <16 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28, i32 32, i32 36, i32 40, i32 44, i32 48, i32 52, i32 56, i32 60>192 store <16 x i8> %strided.vec, ptr %S193 ret void194}195 196define void @trunc_v16i32_to_v16i8(ptr %L, ptr %S) nounwind {197; AVX512-LABEL: trunc_v16i32_to_v16i8:198; AVX512: # %bb.0:199; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0200; AVX512-NEXT: vpmovdb %zmm0, (%rsi)201; AVX512-NEXT: vzeroupper202; AVX512-NEXT: retq203 %vec = load <64 x i8>, ptr %L204 %bc = bitcast <64 x i8> %vec to <16 x i32>205 %strided.vec = trunc <16 x i32> %bc to <16 x i8>206 store <16 x i8> %strided.vec, ptr %S207 ret void208}209 210define void @shuffle_v32i16_to_v8i16(ptr %L, ptr %S) nounwind {211; AVX512-LABEL: shuffle_v32i16_to_v8i16:212; AVX512: # %bb.0:213; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0214; AVX512-NEXT: vpmovqw %zmm0, (%rsi)215; AVX512-NEXT: vzeroupper216; AVX512-NEXT: retq217 %vec = load <32 x i16>, ptr %L218 %strided.vec = shufflevector <32 x i16> %vec, <32 x i16> undef, <8 x i32> <i32 0, i32 4, i32 8, i32 12, i32 16, i32 20, i32 24, i32 28>219 store <8 x i16> %strided.vec, ptr %S220 ret void221}222 223define void @trunc_v8i64_to_v8i16(ptr %L, ptr %S) nounwind {224; AVX512-LABEL: trunc_v8i64_to_v8i16:225; AVX512: # %bb.0:226; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0227; AVX512-NEXT: vpmovqw %zmm0, (%rsi)228; AVX512-NEXT: vzeroupper229; AVX512-NEXT: retq230 %vec = load <32 x i16>, ptr %L231 %bc = bitcast <32 x i16> %vec to <8 x i64>232 %strided.vec = trunc <8 x i64> %bc to <8 x i16>233 store <8 x i16> %strided.vec, ptr %S234 ret void235}236 237define void @shuffle_v64i8_to_v8i8(ptr %L, ptr %S) nounwind {238; AVX512-LABEL: shuffle_v64i8_to_v8i8:239; AVX512: # %bb.0:240; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0241; AVX512-NEXT: vpmovqb %zmm0, (%rsi)242; AVX512-NEXT: vzeroupper243; AVX512-NEXT: retq244 %vec = load <64 x i8>, ptr %L245 %strided.vec = shufflevector <64 x i8> %vec, <64 x i8> undef, <8 x i32> <i32 0, i32 8, i32 16, i32 24, i32 32, i32 40, i32 48, i32 56>246 store <8 x i8> %strided.vec, ptr %S247 ret void248}249 250define void @trunc_v8i64_to_v8i8(ptr %L, ptr %S) nounwind {251; AVX512-LABEL: trunc_v8i64_to_v8i8:252; AVX512: # %bb.0:253; AVX512-NEXT: vmovdqa64 (%rdi), %zmm0254; AVX512-NEXT: vpmovqb %zmm0, (%rsi)255; AVX512-NEXT: vzeroupper256; AVX512-NEXT: retq257 %vec = load <64 x i8>, ptr %L258 %bc = bitcast <64 x i8> %vec to <8 x i64>259 %strided.vec = trunc <8 x i64> %bc to <8 x i8>260 store <8 x i8> %strided.vec, ptr %S261 ret void262}263 264define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61(<64 x i8> %x) {265; AVX512-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_61:266; AVX512: # %bb.0:267; AVX512-NEXT: vpsrld $8, %zmm0, %zmm0268; AVX512-NEXT: vpmovdb %zmm0, %xmm0269; AVX512-NEXT: vzeroupper270; AVX512-NEXT: retq271 %res = shufflevector <64 x i8> %x, <64 x i8> %x, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 61>272 ret <16 x i8> %res273}274 275define <16 x i8> @trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62(<64 x i8> %x) {276; AVX512F-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:277; AVX512F: # %bb.0:278; AVX512F-NEXT: vpmovsxbd {{.*#+}} xmm1 = [0,4,16,21]279; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm2280; AVX512F-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,30,u,u,u,u,u,u,u,u]281; AVX512F-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,29,u,u,u,u,u,u,u,u,u,u,u,u]282; AVX512F-NEXT: vpermt2d %zmm2, %zmm1, %zmm0283; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0284; AVX512F-NEXT: vzeroupper285; AVX512F-NEXT: retq286;287; AVX512VL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:288; AVX512VL: # %bb.0:289; AVX512VL-NEXT: vpmovsxbd {{.*#+}} xmm1 = [0,4,8,13]290; AVX512VL-NEXT: vextracti64x4 $1, %zmm0, %ymm2291; AVX512VL-NEXT: vpshufb {{.*#+}} ymm2 = ymm2[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,30,u,u,u,u,u,u,u,u]292; AVX512VL-NEXT: vpshufb {{.*#+}} ymm0 = ymm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,29,u,u,u,u,u,u,u,u,u,u,u,u]293; AVX512VL-NEXT: vpermt2d %ymm2, %ymm1, %ymm0294; AVX512VL-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0295; AVX512VL-NEXT: vzeroupper296; AVX512VL-NEXT: retq297;298; AVX512BW-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:299; AVX512BW: # %bb.0:300; AVX512BW-NEXT: vpmovsxbd {{.*#+}} xmm1 = [0,4,8,13]301; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,29,u,u,u,u,u,u,u,u,u,u,u,u,33,37,41,45,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,49,53,57,62,u,u,u,u,u,u,u,u]302; AVX512BW-NEXT: vpermd %zmm0, %zmm1, %zmm0303; AVX512BW-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0304; AVX512BW-NEXT: vzeroupper305; AVX512BW-NEXT: retq306;307; AVX512BWVL-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:308; AVX512BWVL: # %bb.0:309; AVX512BWVL-NEXT: vpmovsxbd {{.*#+}} xmm1 = [8,12,0,5]310; AVX512BWVL-NEXT: vextracti64x4 $1, %zmm0, %ymm2311; AVX512BWVL-NEXT: vinserti64x4 $1, %ymm0, %zmm2, %zmm0312; AVX512BWVL-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[1,5,9,13,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,u,17,21,25,30,u,u,u,u,u,u,u,u,33,37,41,45,u,u,u,u,u,u,u,u,u,u,u,u,49,53,57,61,u,u,u,u,u,u,u,u,u,u,u,u]313; AVX512BWVL-NEXT: vpermd %zmm0, %zmm1, %zmm0314; AVX512BWVL-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0315; AVX512BWVL-NEXT: vzeroupper316; AVX512BWVL-NEXT: retq317;318; AVX512VBMI-LABEL: trunc_shuffle_v64i8_01_05_09_13_17_21_25_29_33_37_41_45_49_53_57_62:319; AVX512VBMI: # %bb.0:320; AVX512VBMI-NEXT: vmovdqa {{.*#+}} xmm1 = [1,5,9,13,17,21,25,29,33,37,41,45,49,53,57,62]321; AVX512VBMI-NEXT: vpermb %zmm0, %zmm1, %zmm0322; AVX512VBMI-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0323; AVX512VBMI-NEXT: vzeroupper324; AVX512VBMI-NEXT: retq325 %res = shufflevector <64 x i8> %x, <64 x i8> %x, <16 x i32> <i32 1, i32 5, i32 9, i32 13, i32 17, i32 21, i32 25, i32 29, i32 33, i32 37, i32 41, i32 45, i32 49, i32 53, i32 57, i32 62>326 ret <16 x i8> %res327}328 329; PR111611330define <32 x i8> @trunc_shuffle_v32i16_v32i8_ofs1(<32 x i16> %a0) {331; AVX512F-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:332; AVX512F: # %bb.0:333; AVX512F-NEXT: vextracti64x4 $1, %zmm0, %ymm1334; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15]335; AVX512F-NEXT: vpshufb %ymm2, %ymm1, %ymm1336; AVX512F-NEXT: vpshufb %ymm2, %ymm0, %ymm0337; AVX512F-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]338; AVX512F-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]339; AVX512F-NEXT: retq340;341; AVX512VL-FAST-ALL-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:342; AVX512VL-FAST-ALL: # %bb.0:343; AVX512VL-FAST-ALL-NEXT: vextracti64x4 $1, %zmm0, %ymm1344; AVX512VL-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm1 = ymm1[u,u,u,u,u,u,u,u,1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u,17,19,21,23,25,27,29,31]345; AVX512VL-FAST-ALL-NEXT: vpshufb {{.*#+}} ymm2 = ymm0[1,3,5,7,9,11,13,15,u,u,u,u,u,u,u,u,17,19,21,23,25,27,29,31,u,u,u,u,u,u,u,u]346; AVX512VL-FAST-ALL-NEXT: vpmovsxbq {{.*#+}} ymm0 = [0,2,5,7]347; AVX512VL-FAST-ALL-NEXT: vpermi2q %ymm1, %ymm2, %ymm0348; AVX512VL-FAST-ALL-NEXT: retq349;350; AVX512VL-FAST-PERLANE-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:351; AVX512VL-FAST-PERLANE: # %bb.0:352; AVX512VL-FAST-PERLANE-NEXT: vextracti64x4 $1, %zmm0, %ymm1353; AVX512VL-FAST-PERLANE-NEXT: vpbroadcastq {{.*#+}} ymm2 = [1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15,1,3,5,7,9,11,13,15]354; AVX512VL-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm1, %ymm1355; AVX512VL-FAST-PERLANE-NEXT: vpshufb %ymm2, %ymm0, %ymm0356; AVX512VL-FAST-PERLANE-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0,1],ymm1[2,3],ymm0[4,5],ymm1[6,7]357; AVX512VL-FAST-PERLANE-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]358; AVX512VL-FAST-PERLANE-NEXT: retq359;360; AVX512BW-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:361; AVX512BW: # %bb.0:362; AVX512BW-NEXT: vpsrlw $8, %zmm0, %zmm0363; AVX512BW-NEXT: vpmovwb %zmm0, %ymm0364; AVX512BW-NEXT: retq365;366; AVX512BWVL-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:367; AVX512BWVL: # %bb.0:368; AVX512BWVL-NEXT: vpsrlw $8, %zmm0, %zmm0369; AVX512BWVL-NEXT: vpmovwb %zmm0, %ymm0370; AVX512BWVL-NEXT: retq371;372; AVX512VBMI-LABEL: trunc_shuffle_v32i16_v32i8_ofs1:373; AVX512VBMI: # %bb.0:374; AVX512VBMI-NEXT: vpsrlw $8, %zmm0, %zmm0375; AVX512VBMI-NEXT: vpmovwb %zmm0, %ymm0376; AVX512VBMI-NEXT: retq377 %bc = bitcast <32 x i16> %a0 to <64 x i8>378 %res = shufflevector <64 x i8> %bc, <64 x i8> poison, <32 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31, i32 33, i32 35, i32 37, i32 39, i32 41, i32 43, i32 45, i32 47, i32 49, i32 51, i32 53, i32 55, i32 57, i32 59, i32 61, i32 63>379 ret <32 x i8> %res380}381 382define <4 x double> @PR34175(ptr %p) {383; AVX512F-LABEL: PR34175:384; AVX512F: # %bb.0:385; AVX512F-NEXT: vmovdqu (%rdi), %xmm0386; AVX512F-NEXT: vmovdqu 32(%rdi), %xmm1387; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]388; AVX512F-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]389; AVX512F-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]390; AVX512F-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero391; AVX512F-NEXT: vcvtdq2pd %xmm0, %ymm0392; AVX512F-NEXT: retq393;394; AVX512VL-LABEL: PR34175:395; AVX512VL: # %bb.0:396; AVX512VL-NEXT: vmovdqu (%rdi), %xmm0397; AVX512VL-NEXT: vmovdqu 32(%rdi), %xmm1398; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]399; AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]400; AVX512VL-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]401; AVX512VL-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero402; AVX512VL-NEXT: vcvtdq2pd %xmm0, %ymm0403; AVX512VL-NEXT: retq404;405; AVX512BW-LABEL: PR34175:406; AVX512BW: # %bb.0:407; AVX512BW-NEXT: vmovq {{.*#+}} xmm0 = [0,8,16,24,0,0,0,0]408; AVX512BW-NEXT: vpermw (%rdi), %zmm0, %zmm0409; AVX512BW-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero410; AVX512BW-NEXT: vcvtdq2pd %xmm0, %ymm0411; AVX512BW-NEXT: retq412;413; AVX512BWVL-LABEL: PR34175:414; AVX512BWVL: # %bb.0:415; AVX512BWVL-NEXT: vmovq {{.*#+}} xmm0 = [0,8,16,24,0,0,0,0]416; AVX512BWVL-NEXT: vpermw (%rdi), %zmm0, %zmm0417; AVX512BWVL-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero418; AVX512BWVL-NEXT: vcvtdq2pd %xmm0, %ymm0419; AVX512BWVL-NEXT: retq420;421; AVX512VBMI-LABEL: PR34175:422; AVX512VBMI: # %bb.0:423; AVX512VBMI-NEXT: vmovq {{.*#+}} xmm0 = [0,8,16,24,0,0,0,0]424; AVX512VBMI-NEXT: vpermw (%rdi), %zmm0, %zmm0425; AVX512VBMI-NEXT: vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero426; AVX512VBMI-NEXT: vcvtdq2pd %xmm0, %ymm0427; AVX512VBMI-NEXT: retq428 %v = load <32 x i16>, ptr %p, align 2429 %shuf = shufflevector <32 x i16> %v, <32 x i16> undef, <4 x i32> <i32 0, i32 8, i32 16, i32 24>430 %tofp = uitofp <4 x i16> %shuf to <4 x double>431 ret <4 x double> %tofp432}433 434define <16 x i8> @trunc_v8i64_to_v8i8_return_v16i8(<8 x i64> %vec) nounwind {435; AVX512-LABEL: trunc_v8i64_to_v8i8_return_v16i8:436; AVX512: # %bb.0:437; AVX512-NEXT: vpmovqb %zmm0, %xmm0438; AVX512-NEXT: vzeroupper439; AVX512-NEXT: retq440 %truncated = trunc <8 x i64> %vec to <8 x i8>441 %result = shufflevector <8 x i8> %truncated, <8 x i8> zeroinitializer, <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>442 ret <16 x i8> %result443}444 445