1; RUN: llc < %s -march=arm64 -aarch64-neon-syntax=apple | FileCheck %s 2 3define <8 x i8> @shadd8b(<8 x i8>* %A, <8 x i8>* %B) nounwind { 4;CHECK-LABEL: shadd8b: 5;CHECK: shadd.8b 6 %tmp1 = load <8 x i8>, <8 x i8>* %A 7 %tmp2 = load <8 x i8>, <8 x i8>* %B 8 %tmp3 = call <8 x i8> @llvm.aarch64.neon.shadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 9 ret <8 x i8> %tmp3 10} 11 12define <16 x i8> @shadd16b(<16 x i8>* %A, <16 x i8>* %B) nounwind { 13;CHECK-LABEL: shadd16b: 14;CHECK: shadd.16b 15 %tmp1 = load <16 x i8>, <16 x i8>* %A 16 %tmp2 = load <16 x i8>, <16 x i8>* %B 17 %tmp3 = call <16 x i8> @llvm.aarch64.neon.shadd.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 18 ret <16 x i8> %tmp3 19} 20 21define <4 x i16> @shadd4h(<4 x i16>* %A, <4 x i16>* %B) nounwind { 22;CHECK-LABEL: shadd4h: 23;CHECK: shadd.4h 24 %tmp1 = load <4 x i16>, <4 x i16>* %A 25 %tmp2 = load <4 x i16>, <4 x i16>* %B 26 %tmp3 = call <4 x i16> @llvm.aarch64.neon.shadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 27 ret <4 x i16> %tmp3 28} 29 30define <8 x i16> @shadd8h(<8 x i16>* %A, <8 x i16>* %B) nounwind { 31;CHECK-LABEL: shadd8h: 32;CHECK: shadd.8h 33 %tmp1 = load <8 x i16>, <8 x i16>* %A 34 %tmp2 = load <8 x i16>, <8 x i16>* %B 35 %tmp3 = call <8 x i16> @llvm.aarch64.neon.shadd.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 36 ret <8 x i16> %tmp3 37} 38 39define <2 x i32> @shadd2s(<2 x i32>* %A, <2 x i32>* %B) nounwind { 40;CHECK-LABEL: shadd2s: 41;CHECK: shadd.2s 42 %tmp1 = load <2 x i32>, <2 x i32>* %A 43 %tmp2 = load <2 x i32>, <2 x i32>* %B 44 %tmp3 = call <2 x i32> @llvm.aarch64.neon.shadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 45 ret <2 x i32> %tmp3 46} 47 48define <4 x i32> @shadd4s(<4 x i32>* %A, <4 x i32>* %B) nounwind { 49;CHECK-LABEL: shadd4s: 50;CHECK: shadd.4s 51 %tmp1 = load <4 x i32>, <4 x i32>* %A 52 %tmp2 = load <4 x i32>, <4 x i32>* %B 53 %tmp3 = call <4 x i32> @llvm.aarch64.neon.shadd.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 54 ret <4 x i32> %tmp3 55} 56 57define <8 x i8> @uhadd8b(<8 x i8>* %A, <8 x i8>* %B) nounwind { 58;CHECK-LABEL: uhadd8b: 59;CHECK: uhadd.8b 60 %tmp1 = load <8 x i8>, <8 x i8>* %A 61 %tmp2 = load <8 x i8>, <8 x i8>* %B 62 %tmp3 = call <8 x i8> @llvm.aarch64.neon.uhadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 63 ret <8 x i8> %tmp3 64} 65 66define <16 x i8> @uhadd16b(<16 x i8>* %A, <16 x i8>* %B) nounwind { 67;CHECK-LABEL: uhadd16b: 68;CHECK: uhadd.16b 69 %tmp1 = load <16 x i8>, <16 x i8>* %A 70 %tmp2 = load <16 x i8>, <16 x i8>* %B 71 %tmp3 = call <16 x i8> @llvm.aarch64.neon.uhadd.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 72 ret <16 x i8> %tmp3 73} 74 75define <4 x i16> @uhadd4h(<4 x i16>* %A, <4 x i16>* %B) nounwind { 76;CHECK-LABEL: uhadd4h: 77;CHECK: uhadd.4h 78 %tmp1 = load <4 x i16>, <4 x i16>* %A 79 %tmp2 = load <4 x i16>, <4 x i16>* %B 80 %tmp3 = call <4 x i16> @llvm.aarch64.neon.uhadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 81 ret <4 x i16> %tmp3 82} 83 84define <8 x i16> @uhadd8h(<8 x i16>* %A, <8 x i16>* %B) nounwind { 85;CHECK-LABEL: uhadd8h: 86;CHECK: uhadd.8h 87 %tmp1 = load <8 x i16>, <8 x i16>* %A 88 %tmp2 = load <8 x i16>, <8 x i16>* %B 89 %tmp3 = call <8 x i16> @llvm.aarch64.neon.uhadd.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 90 ret <8 x i16> %tmp3 91} 92 93define <2 x i32> @uhadd2s(<2 x i32>* %A, <2 x i32>* %B) nounwind { 94;CHECK-LABEL: uhadd2s: 95;CHECK: uhadd.2s 96 %tmp1 = load <2 x i32>, <2 x i32>* %A 97 %tmp2 = load <2 x i32>, <2 x i32>* %B 98 %tmp3 = call <2 x i32> @llvm.aarch64.neon.uhadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 99 ret <2 x i32> %tmp3 100} 101 102define <4 x i32> @uhadd4s(<4 x i32>* %A, <4 x i32>* %B) nounwind { 103;CHECK-LABEL: uhadd4s: 104;CHECK: uhadd.4s 105 %tmp1 = load <4 x i32>, <4 x i32>* %A 106 %tmp2 = load <4 x i32>, <4 x i32>* %B 107 %tmp3 = call <4 x i32> @llvm.aarch64.neon.uhadd.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 108 ret <4 x i32> %tmp3 109} 110 111declare <8 x i8> @llvm.aarch64.neon.shadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 112declare <4 x i16> @llvm.aarch64.neon.shadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 113declare <2 x i32> @llvm.aarch64.neon.shadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 114 115declare <8 x i8> @llvm.aarch64.neon.uhadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 116declare <4 x i16> @llvm.aarch64.neon.uhadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 117declare <2 x i32> @llvm.aarch64.neon.uhadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 118 119declare <16 x i8> @llvm.aarch64.neon.shadd.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 120declare <8 x i16> @llvm.aarch64.neon.shadd.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 121declare <4 x i32> @llvm.aarch64.neon.shadd.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 122 123declare <16 x i8> @llvm.aarch64.neon.uhadd.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 124declare <8 x i16> @llvm.aarch64.neon.uhadd.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 125declare <4 x i32> @llvm.aarch64.neon.uhadd.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 126 127define <8 x i8> @srhadd8b(<8 x i8>* %A, <8 x i8>* %B) nounwind { 128;CHECK-LABEL: srhadd8b: 129;CHECK: srhadd.8b 130 %tmp1 = load <8 x i8>, <8 x i8>* %A 131 %tmp2 = load <8 x i8>, <8 x i8>* %B 132 %tmp3 = call <8 x i8> @llvm.aarch64.neon.srhadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 133 ret <8 x i8> %tmp3 134} 135 136define <16 x i8> @srhadd16b(<16 x i8>* %A, <16 x i8>* %B) nounwind { 137;CHECK-LABEL: srhadd16b: 138;CHECK: srhadd.16b 139 %tmp1 = load <16 x i8>, <16 x i8>* %A 140 %tmp2 = load <16 x i8>, <16 x i8>* %B 141 %tmp3 = call <16 x i8> @llvm.aarch64.neon.srhadd.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 142 ret <16 x i8> %tmp3 143} 144 145define <4 x i16> @srhadd4h(<4 x i16>* %A, <4 x i16>* %B) nounwind { 146;CHECK-LABEL: srhadd4h: 147;CHECK: srhadd.4h 148 %tmp1 = load <4 x i16>, <4 x i16>* %A 149 %tmp2 = load <4 x i16>, <4 x i16>* %B 150 %tmp3 = call <4 x i16> @llvm.aarch64.neon.srhadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 151 ret <4 x i16> %tmp3 152} 153 154define <8 x i16> @srhadd8h(<8 x i16>* %A, <8 x i16>* %B) nounwind { 155;CHECK-LABEL: srhadd8h: 156;CHECK: srhadd.8h 157 %tmp1 = load <8 x i16>, <8 x i16>* %A 158 %tmp2 = load <8 x i16>, <8 x i16>* %B 159 %tmp3 = call <8 x i16> @llvm.aarch64.neon.srhadd.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 160 ret <8 x i16> %tmp3 161} 162 163define <2 x i32> @srhadd2s(<2 x i32>* %A, <2 x i32>* %B) nounwind { 164;CHECK-LABEL: srhadd2s: 165;CHECK: srhadd.2s 166 %tmp1 = load <2 x i32>, <2 x i32>* %A 167 %tmp2 = load <2 x i32>, <2 x i32>* %B 168 %tmp3 = call <2 x i32> @llvm.aarch64.neon.srhadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 169 ret <2 x i32> %tmp3 170} 171 172define <4 x i32> @srhadd4s(<4 x i32>* %A, <4 x i32>* %B) nounwind { 173;CHECK-LABEL: srhadd4s: 174;CHECK: srhadd.4s 175 %tmp1 = load <4 x i32>, <4 x i32>* %A 176 %tmp2 = load <4 x i32>, <4 x i32>* %B 177 %tmp3 = call <4 x i32> @llvm.aarch64.neon.srhadd.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 178 ret <4 x i32> %tmp3 179} 180 181define <8 x i8> @urhadd8b(<8 x i8>* %A, <8 x i8>* %B) nounwind { 182;CHECK-LABEL: urhadd8b: 183;CHECK: urhadd.8b 184 %tmp1 = load <8 x i8>, <8 x i8>* %A 185 %tmp2 = load <8 x i8>, <8 x i8>* %B 186 %tmp3 = call <8 x i8> @llvm.aarch64.neon.urhadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 187 ret <8 x i8> %tmp3 188} 189 190define <16 x i8> @urhadd16b(<16 x i8>* %A, <16 x i8>* %B) nounwind { 191;CHECK-LABEL: urhadd16b: 192;CHECK: urhadd.16b 193 %tmp1 = load <16 x i8>, <16 x i8>* %A 194 %tmp2 = load <16 x i8>, <16 x i8>* %B 195 %tmp3 = call <16 x i8> @llvm.aarch64.neon.urhadd.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 196 ret <16 x i8> %tmp3 197} 198 199define <4 x i16> @urhadd4h(<4 x i16>* %A, <4 x i16>* %B) nounwind { 200;CHECK-LABEL: urhadd4h: 201;CHECK: urhadd.4h 202 %tmp1 = load <4 x i16>, <4 x i16>* %A 203 %tmp2 = load <4 x i16>, <4 x i16>* %B 204 %tmp3 = call <4 x i16> @llvm.aarch64.neon.urhadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 205 ret <4 x i16> %tmp3 206} 207 208define <8 x i16> @urhadd8h(<8 x i16>* %A, <8 x i16>* %B) nounwind { 209;CHECK-LABEL: urhadd8h: 210;CHECK: urhadd.8h 211 %tmp1 = load <8 x i16>, <8 x i16>* %A 212 %tmp2 = load <8 x i16>, <8 x i16>* %B 213 %tmp3 = call <8 x i16> @llvm.aarch64.neon.urhadd.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 214 ret <8 x i16> %tmp3 215} 216 217define <2 x i32> @urhadd2s(<2 x i32>* %A, <2 x i32>* %B) nounwind { 218;CHECK-LABEL: urhadd2s: 219;CHECK: urhadd.2s 220 %tmp1 = load <2 x i32>, <2 x i32>* %A 221 %tmp2 = load <2 x i32>, <2 x i32>* %B 222 %tmp3 = call <2 x i32> @llvm.aarch64.neon.urhadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 223 ret <2 x i32> %tmp3 224} 225 226define <4 x i32> @urhadd4s(<4 x i32>* %A, <4 x i32>* %B) nounwind { 227;CHECK-LABEL: urhadd4s: 228;CHECK: urhadd.4s 229 %tmp1 = load <4 x i32>, <4 x i32>* %A 230 %tmp2 = load <4 x i32>, <4 x i32>* %B 231 %tmp3 = call <4 x i32> @llvm.aarch64.neon.urhadd.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 232 ret <4 x i32> %tmp3 233} 234 235declare <8 x i8> @llvm.aarch64.neon.srhadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 236declare <4 x i16> @llvm.aarch64.neon.srhadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 237declare <2 x i32> @llvm.aarch64.neon.srhadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 238 239declare <8 x i8> @llvm.aarch64.neon.urhadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 240declare <4 x i16> @llvm.aarch64.neon.urhadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 241declare <2 x i32> @llvm.aarch64.neon.urhadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 242 243declare <16 x i8> @llvm.aarch64.neon.srhadd.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 244declare <8 x i16> @llvm.aarch64.neon.srhadd.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 245declare <4 x i32> @llvm.aarch64.neon.srhadd.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 246 247declare <16 x i8> @llvm.aarch64.neon.urhadd.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 248declare <8 x i16> @llvm.aarch64.neon.urhadd.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 249declare <4 x i32> @llvm.aarch64.neon.urhadd.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 250