1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s --check-prefix=ALL --check-prefix=KNL 3; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx | FileCheck %s --check-prefix=ALL --check-prefix=SKX 4 5 6define <16 x i32> @vpandd(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp { 7; ALL-LABEL: vpandd: 8; ALL: ## BB#0: ## %entry 9; ALL-NEXT: vpaddd {{.*}}(%rip){1to16}, %zmm0, %zmm0 10; ALL-NEXT: vpandd %zmm1, %zmm0, %zmm0 11; ALL-NEXT: retq 12entry: 13 ; Force the execution domain with an add. 14 %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, 15 i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 16 %x = and <16 x i32> %a2, %b 17 ret <16 x i32> %x 18} 19 20define <16 x i32> @vpord(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp { 21; ALL-LABEL: vpord: 22; ALL: ## BB#0: ## %entry 23; ALL-NEXT: vpaddd {{.*}}(%rip){1to16}, %zmm0, %zmm0 24; ALL-NEXT: vpord %zmm1, %zmm0, %zmm0 25; ALL-NEXT: retq 26entry: 27 ; Force the execution domain with an add. 28 %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, 29 i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 30 %x = or <16 x i32> %a2, %b 31 ret <16 x i32> %x 32} 33 34define <16 x i32> @vpxord(<16 x i32> %a, <16 x i32> %b) nounwind uwtable readnone ssp { 35; ALL-LABEL: vpxord: 36; ALL: ## BB#0: ## %entry 37; ALL-NEXT: vpaddd {{.*}}(%rip){1to16}, %zmm0, %zmm0 38; ALL-NEXT: vpxord %zmm1, %zmm0, %zmm0 39; ALL-NEXT: retq 40entry: 41 ; Force the execution domain with an add. 42 %a2 = add <16 x i32> %a, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, 43 i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> 44 %x = xor <16 x i32> %a2, %b 45 ret <16 x i32> %x 46} 47 48define <8 x i64> @vpandq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp { 49; ALL-LABEL: vpandq: 50; ALL: ## BB#0: ## %entry 51; ALL-NEXT: vpaddq {{.*}}(%rip){1to8}, %zmm0, %zmm0 52; ALL-NEXT: vpandq %zmm1, %zmm0, %zmm0 53; ALL-NEXT: retq 54entry: 55 ; Force the execution domain with an add. 56 %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> 57 %x = and <8 x i64> %a2, %b 58 ret <8 x i64> %x 59} 60 61define <8 x i64> @vporq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp { 62; ALL-LABEL: vporq: 63; ALL: ## BB#0: ## %entry 64; ALL-NEXT: vpaddq {{.*}}(%rip){1to8}, %zmm0, %zmm0 65; ALL-NEXT: vporq %zmm1, %zmm0, %zmm0 66; ALL-NEXT: retq 67entry: 68 ; Force the execution domain with an add. 69 %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> 70 %x = or <8 x i64> %a2, %b 71 ret <8 x i64> %x 72} 73 74define <8 x i64> @vpxorq(<8 x i64> %a, <8 x i64> %b) nounwind uwtable readnone ssp { 75; ALL-LABEL: vpxorq: 76; ALL: ## BB#0: ## %entry 77; ALL-NEXT: vpaddq {{.*}}(%rip){1to8}, %zmm0, %zmm0 78; ALL-NEXT: vpxorq %zmm1, %zmm0, %zmm0 79; ALL-NEXT: retq 80entry: 81 ; Force the execution domain with an add. 82 %a2 = add <8 x i64> %a, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> 83 %x = xor <8 x i64> %a2, %b 84 ret <8 x i64> %x 85} 86 87 88define <8 x i64> @orq_broadcast(<8 x i64> %a) nounwind { 89; ALL-LABEL: orq_broadcast: 90; ALL: ## BB#0: 91; ALL-NEXT: vporq {{.*}}(%rip){1to8}, %zmm0, %zmm0 92; ALL-NEXT: retq 93 %b = or <8 x i64> %a, <i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2> 94 ret <8 x i64> %b 95} 96 97define <16 x i32> @andd512fold(<16 x i32> %y, <16 x i32>* %x) { 98; ALL-LABEL: andd512fold: 99; ALL: ## BB#0: ## %entry 100; ALL-NEXT: vpandd (%rdi), %zmm0, %zmm0 101; ALL-NEXT: retq 102entry: 103 %a = load <16 x i32>, <16 x i32>* %x, align 4 104 %b = and <16 x i32> %y, %a 105 ret <16 x i32> %b 106} 107 108define <8 x i64> @andqbrst(<8 x i64> %p1, i64* %ap) { 109; ALL-LABEL: andqbrst: 110; ALL: ## BB#0: ## %entry 111; ALL-NEXT: vpandq (%rdi){1to8}, %zmm0, %zmm0 112; ALL-NEXT: retq 113entry: 114 %a = load i64, i64* %ap, align 8 115 %b = insertelement <8 x i64> undef, i64 %a, i32 0 116 %c = shufflevector <8 x i64> %b, <8 x i64> undef, <8 x i32> zeroinitializer 117 %d = and <8 x i64> %p1, %c 118 ret <8 x i64>%d 119} 120 121define <64 x i8> @and_v64i8(<64 x i8> %a, <64 x i8> %b) { 122; KNL-LABEL: and_v64i8: 123; KNL: ## BB#0: 124; KNL-NEXT: vandps %ymm2, %ymm0, %ymm0 125; KNL-NEXT: vandps %ymm3, %ymm1, %ymm1 126; KNL-NEXT: retq 127; 128; SKX-LABEL: and_v64i8: 129; SKX: ## BB#0: 130; SKX-NEXT: vpandq %zmm1, %zmm0, %zmm0 131; SKX-NEXT: retq 132 %res = and <64 x i8> %a, %b 133 ret <64 x i8> %res 134} 135 136define <64 x i8> @or_v64i8(<64 x i8> %a, <64 x i8> %b) { 137; KNL-LABEL: or_v64i8: 138; KNL: ## BB#0: 139; KNL-NEXT: vorps %ymm2, %ymm0, %ymm0 140; KNL-NEXT: vorps %ymm3, %ymm1, %ymm1 141; KNL-NEXT: retq 142; 143; SKX-LABEL: or_v64i8: 144; SKX: ## BB#0: 145; SKX-NEXT: vporq %zmm1, %zmm0, %zmm0 146; SKX-NEXT: retq 147 %res = or <64 x i8> %a, %b 148 ret <64 x i8> %res 149} 150 151define <64 x i8> @xor_v64i8(<64 x i8> %a, <64 x i8> %b) { 152; KNL-LABEL: xor_v64i8: 153; KNL: ## BB#0: 154; KNL-NEXT: vxorps %ymm2, %ymm0, %ymm0 155; KNL-NEXT: vxorps %ymm3, %ymm1, %ymm1 156; KNL-NEXT: retq 157; 158; SKX-LABEL: xor_v64i8: 159; SKX: ## BB#0: 160; SKX-NEXT: vpxorq %zmm1, %zmm0, %zmm0 161; SKX-NEXT: retq 162 %res = xor <64 x i8> %a, %b 163 ret <64 x i8> %res 164} 165 166define <32 x i16> @and_v32i16(<32 x i16> %a, <32 x i16> %b) { 167; KNL-LABEL: and_v32i16: 168; KNL: ## BB#0: 169; KNL-NEXT: vandps %ymm2, %ymm0, %ymm0 170; KNL-NEXT: vandps %ymm3, %ymm1, %ymm1 171; KNL-NEXT: retq 172; 173; SKX-LABEL: and_v32i16: 174; SKX: ## BB#0: 175; SKX-NEXT: vpandq %zmm1, %zmm0, %zmm0 176; SKX-NEXT: retq 177 %res = and <32 x i16> %a, %b 178 ret <32 x i16> %res 179} 180 181define <32 x i16> @or_v32i16(<32 x i16> %a, <32 x i16> %b) { 182; KNL-LABEL: or_v32i16: 183; KNL: ## BB#0: 184; KNL-NEXT: vorps %ymm2, %ymm0, %ymm0 185; KNL-NEXT: vorps %ymm3, %ymm1, %ymm1 186; KNL-NEXT: retq 187; 188; SKX-LABEL: or_v32i16: 189; SKX: ## BB#0: 190; SKX-NEXT: vporq %zmm1, %zmm0, %zmm0 191; SKX-NEXT: retq 192 %res = or <32 x i16> %a, %b 193 ret <32 x i16> %res 194} 195 196define <32 x i16> @xor_v32i16(<32 x i16> %a, <32 x i16> %b) { 197; KNL-LABEL: xor_v32i16: 198; KNL: ## BB#0: 199; KNL-NEXT: vxorps %ymm2, %ymm0, %ymm0 200; KNL-NEXT: vxorps %ymm3, %ymm1, %ymm1 201; KNL-NEXT: retq 202; 203; SKX-LABEL: xor_v32i16: 204; SKX: ## BB#0: 205; SKX-NEXT: vpxorq %zmm1, %zmm0, %zmm0 206; SKX-NEXT: retq 207 %res = xor <32 x i16> %a, %b 208 ret <32 x i16> %res 209} 210