73 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc < %s -mtriple=x86_64-unknown-linux-gn -mattr=+avx512f | FileCheck %s --check-prefix=ALL --check-prefix=KNL3; RUN: llc < %s -mtriple=x86_64-unknown-linux-gn -mattr=+avx512f,+avx512vl,+avx512dq,+avx512bw | FileCheck %s --check-prefix=ALL --check-prefix=SKX4 5 6define void @any_extend_load_v8i64(ptr %ptr) {7; ALL-LABEL: any_extend_load_v8i64:8; ALL: # %bb.0:9; ALL-NEXT: vpmovzxbq {{.*#+}} zmm0 = mem[0],zero,zero,zero,zero,zero,zero,zero,mem[1],zero,zero,zero,zero,zero,zero,zero,mem[2],zero,zero,zero,zero,zero,zero,zero,mem[3],zero,zero,zero,zero,zero,zero,zero,mem[4],zero,zero,zero,zero,zero,zero,zero,mem[5],zero,zero,zero,zero,zero,zero,zero,mem[6],zero,zero,zero,zero,zero,zero,zero,mem[7],zero,zero,zero,zero,zero,zero,zero10; ALL-NEXT: vpaddq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm011; ALL-NEXT: vpmovqb %zmm0, (%rdi)12; ALL-NEXT: vzeroupper13; ALL-NEXT: retq14 %wide.load = load <8 x i8>, ptr %ptr, align 115 %1 = zext <8 x i8> %wide.load to <8 x i64>16 %2 = add nuw nsw <8 x i64> %1, <i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4>17 %3 = xor <8 x i64> %2, zeroinitializer18 %4 = trunc <8 x i64> %3 to <8 x i8>19 store <8 x i8> %4, ptr %ptr, align 120 ret void21}22 23define void @any_extend_load_v8i32(ptr %ptr) {24; KNL-LABEL: any_extend_load_v8i32:25; KNL: # %bb.0:26; KNL-NEXT: vpmovzxbd {{.*#+}} ymm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero27; KNL-NEXT: vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]28; KNL-NEXT: vpaddd %ymm1, %ymm0, %ymm029; KNL-NEXT: vpmovdb %zmm0, %xmm030; KNL-NEXT: vmovq %xmm0, (%rdi)31; KNL-NEXT: vzeroupper32; KNL-NEXT: retq33;34; SKX-LABEL: any_extend_load_v8i32:35; SKX: # %bb.0:36; SKX-NEXT: vpmovzxbd {{.*#+}} ymm0 = mem[0],zero,zero,zero,mem[1],zero,zero,zero,mem[2],zero,zero,zero,mem[3],zero,zero,zero,mem[4],zero,zero,zero,mem[5],zero,zero,zero,mem[6],zero,zero,zero,mem[7],zero,zero,zero37; SKX-NEXT: vpaddd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm038; SKX-NEXT: vpmovdb %ymm0, (%rdi)39; SKX-NEXT: vzeroupper40; SKX-NEXT: retq41 %wide.load = load <8 x i8>, ptr %ptr, align 142 %1 = zext <8 x i8> %wide.load to <8 x i32>43 %2 = add nuw nsw <8 x i32> %1, <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>44 %3 = xor <8 x i32> %2, zeroinitializer45 %4 = trunc <8 x i32> %3 to <8 x i8>46 store <8 x i8> %4, ptr %ptr, align 147 ret void48}49 50define void @any_extend_load_v8i16(ptr %ptr) {51; KNL-LABEL: any_extend_load_v8i16:52; KNL: # %bb.0:53; KNL-NEXT: vpmovzxbw {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero54; KNL-NEXT: vpaddw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm055; KNL-NEXT: vpshufb {{.*#+}} xmm0 = xmm0[0,2,4,6,8,10,12,14,u,u,u,u,u,u,u,u]56; KNL-NEXT: vmovq %xmm0, (%rdi)57; KNL-NEXT: retq58;59; SKX-LABEL: any_extend_load_v8i16:60; SKX: # %bb.0:61; SKX-NEXT: vpmovzxbw {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero62; SKX-NEXT: vpaddw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm063; SKX-NEXT: vpmovwb %xmm0, (%rdi)64; SKX-NEXT: retq65 %wide.load = load <8 x i8>, ptr %ptr, align 166 %1 = zext <8 x i8> %wide.load to <8 x i16>67 %2 = add nuw nsw <8 x i16> %1, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>68 %3 = xor <8 x i16> %2, zeroinitializer69 %4 = trunc <8 x i16> %3 to <8 x i8>70 store <8 x i8> %4, ptr %ptr, align 171 ret void72}73