1; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s 2 3define <8 x i8> @vmins8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 4;CHECK: vmins8: 5;CHECK: vmin.s8 6 %tmp1 = load <8 x i8>* %A 7 %tmp2 = load <8 x i8>* %B 8 %tmp3 = call <8 x i8> @llvm.arm.neon.vmins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 9 ret <8 x i8> %tmp3 10} 11 12define <4 x i16> @vmins16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 13;CHECK: vmins16: 14;CHECK: vmin.s16 15 %tmp1 = load <4 x i16>* %A 16 %tmp2 = load <4 x i16>* %B 17 %tmp3 = call <4 x i16> @llvm.arm.neon.vmins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 18 ret <4 x i16> %tmp3 19} 20 21define <2 x i32> @vmins32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 22;CHECK: vmins32: 23;CHECK: vmin.s32 24 %tmp1 = load <2 x i32>* %A 25 %tmp2 = load <2 x i32>* %B 26 %tmp3 = call <2 x i32> @llvm.arm.neon.vmins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 27 ret <2 x i32> %tmp3 28} 29 30define <8 x i8> @vminu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 31;CHECK: vminu8: 32;CHECK: vmin.u8 33 %tmp1 = load <8 x i8>* %A 34 %tmp2 = load <8 x i8>* %B 35 %tmp3 = call <8 x i8> @llvm.arm.neon.vminu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 36 ret <8 x i8> %tmp3 37} 38 39define <4 x i16> @vminu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 40;CHECK: vminu16: 41;CHECK: vmin.u16 42 %tmp1 = load <4 x i16>* %A 43 %tmp2 = load <4 x i16>* %B 44 %tmp3 = call <4 x i16> @llvm.arm.neon.vminu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 45 ret <4 x i16> %tmp3 46} 47 48define <2 x i32> @vminu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 49;CHECK: vminu32: 50;CHECK: vmin.u32 51 %tmp1 = load <2 x i32>* %A 52 %tmp2 = load <2 x i32>* %B 53 %tmp3 = call <2 x i32> @llvm.arm.neon.vminu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 54 ret <2 x i32> %tmp3 55} 56 57define <2 x float> @vminf32(<2 x float>* %A, <2 x float>* %B) nounwind { 58;CHECK: vminf32: 59;CHECK: vmin.f32 60 %tmp1 = load <2 x float>* %A 61 %tmp2 = load <2 x float>* %B 62 %tmp3 = call <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float> %tmp1, <2 x float> %tmp2) 63 ret <2 x float> %tmp3 64} 65 66define <16 x i8> @vminQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 67;CHECK: vminQs8: 68;CHECK: vmin.s8 69 %tmp1 = load <16 x i8>* %A 70 %tmp2 = load <16 x i8>* %B 71 %tmp3 = call <16 x i8> @llvm.arm.neon.vmins.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 72 ret <16 x i8> %tmp3 73} 74 75define <8 x i16> @vminQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 76;CHECK: vminQs16: 77;CHECK: vmin.s16 78 %tmp1 = load <8 x i16>* %A 79 %tmp2 = load <8 x i16>* %B 80 %tmp3 = call <8 x i16> @llvm.arm.neon.vmins.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 81 ret <8 x i16> %tmp3 82} 83 84define <4 x i32> @vminQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 85;CHECK: vminQs32: 86;CHECK: vmin.s32 87 %tmp1 = load <4 x i32>* %A 88 %tmp2 = load <4 x i32>* %B 89 %tmp3 = call <4 x i32> @llvm.arm.neon.vmins.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 90 ret <4 x i32> %tmp3 91} 92 93define <16 x i8> @vminQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 94;CHECK: vminQu8: 95;CHECK: vmin.u8 96 %tmp1 = load <16 x i8>* %A 97 %tmp2 = load <16 x i8>* %B 98 %tmp3 = call <16 x i8> @llvm.arm.neon.vminu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 99 ret <16 x i8> %tmp3 100} 101 102define <8 x i16> @vminQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 103;CHECK: vminQu16: 104;CHECK: vmin.u16 105 %tmp1 = load <8 x i16>* %A 106 %tmp2 = load <8 x i16>* %B 107 %tmp3 = call <8 x i16> @llvm.arm.neon.vminu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 108 ret <8 x i16> %tmp3 109} 110 111define <4 x i32> @vminQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 112;CHECK: vminQu32: 113;CHECK: vmin.u32 114 %tmp1 = load <4 x i32>* %A 115 %tmp2 = load <4 x i32>* %B 116 %tmp3 = call <4 x i32> @llvm.arm.neon.vminu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 117 ret <4 x i32> %tmp3 118} 119 120define <4 x float> @vminQf32(<4 x float>* %A, <4 x float>* %B) nounwind { 121;CHECK: vminQf32: 122;CHECK: vmin.f32 123 %tmp1 = load <4 x float>* %A 124 %tmp2 = load <4 x float>* %B 125 %tmp3 = call <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float> %tmp1, <4 x float> %tmp2) 126 ret <4 x float> %tmp3 127} 128 129declare <8 x i8> @llvm.arm.neon.vmins.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 130declare <4 x i16> @llvm.arm.neon.vmins.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 131declare <2 x i32> @llvm.arm.neon.vmins.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 132 133declare <8 x i8> @llvm.arm.neon.vminu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 134declare <4 x i16> @llvm.arm.neon.vminu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 135declare <2 x i32> @llvm.arm.neon.vminu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 136 137declare <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float>, <2 x float>) nounwind readnone 138 139declare <16 x i8> @llvm.arm.neon.vmins.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 140declare <8 x i16> @llvm.arm.neon.vmins.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 141declare <4 x i32> @llvm.arm.neon.vmins.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 142 143declare <16 x i8> @llvm.arm.neon.vminu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 144declare <8 x i16> @llvm.arm.neon.vminu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 145declare <4 x i32> @llvm.arm.neon.vminu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 146 147declare <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float>, <4 x float>) nounwind readnone 148 149define <8 x i8> @vmaxs8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 150;CHECK: vmaxs8: 151;CHECK: vmax.s8 152 %tmp1 = load <8 x i8>* %A 153 %tmp2 = load <8 x i8>* %B 154 %tmp3 = call <8 x i8> @llvm.arm.neon.vmaxs.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 155 ret <8 x i8> %tmp3 156} 157 158define <4 x i16> @vmaxs16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 159;CHECK: vmaxs16: 160;CHECK: vmax.s16 161 %tmp1 = load <4 x i16>* %A 162 %tmp2 = load <4 x i16>* %B 163 %tmp3 = call <4 x i16> @llvm.arm.neon.vmaxs.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 164 ret <4 x i16> %tmp3 165} 166 167define <2 x i32> @vmaxs32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 168;CHECK: vmaxs32: 169;CHECK: vmax.s32 170 %tmp1 = load <2 x i32>* %A 171 %tmp2 = load <2 x i32>* %B 172 %tmp3 = call <2 x i32> @llvm.arm.neon.vmaxs.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 173 ret <2 x i32> %tmp3 174} 175 176define <8 x i8> @vmaxu8(<8 x i8>* %A, <8 x i8>* %B) nounwind { 177;CHECK: vmaxu8: 178;CHECK: vmax.u8 179 %tmp1 = load <8 x i8>* %A 180 %tmp2 = load <8 x i8>* %B 181 %tmp3 = call <8 x i8> @llvm.arm.neon.vmaxu.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2) 182 ret <8 x i8> %tmp3 183} 184 185define <4 x i16> @vmaxu16(<4 x i16>* %A, <4 x i16>* %B) nounwind { 186;CHECK: vmaxu16: 187;CHECK: vmax.u16 188 %tmp1 = load <4 x i16>* %A 189 %tmp2 = load <4 x i16>* %B 190 %tmp3 = call <4 x i16> @llvm.arm.neon.vmaxu.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2) 191 ret <4 x i16> %tmp3 192} 193 194define <2 x i32> @vmaxu32(<2 x i32>* %A, <2 x i32>* %B) nounwind { 195;CHECK: vmaxu32: 196;CHECK: vmax.u32 197 %tmp1 = load <2 x i32>* %A 198 %tmp2 = load <2 x i32>* %B 199 %tmp3 = call <2 x i32> @llvm.arm.neon.vmaxu.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2) 200 ret <2 x i32> %tmp3 201} 202 203define <2 x float> @vmaxf32(<2 x float>* %A, <2 x float>* %B) nounwind { 204;CHECK: vmaxf32: 205;CHECK: vmax.f32 206 %tmp1 = load <2 x float>* %A 207 %tmp2 = load <2 x float>* %B 208 %tmp3 = call <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float> %tmp1, <2 x float> %tmp2) 209 ret <2 x float> %tmp3 210} 211 212define <16 x i8> @vmaxQs8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 213;CHECK: vmaxQs8: 214;CHECK: vmax.s8 215 %tmp1 = load <16 x i8>* %A 216 %tmp2 = load <16 x i8>* %B 217 %tmp3 = call <16 x i8> @llvm.arm.neon.vmaxs.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 218 ret <16 x i8> %tmp3 219} 220 221define <8 x i16> @vmaxQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 222;CHECK: vmaxQs16: 223;CHECK: vmax.s16 224 %tmp1 = load <8 x i16>* %A 225 %tmp2 = load <8 x i16>* %B 226 %tmp3 = call <8 x i16> @llvm.arm.neon.vmaxs.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 227 ret <8 x i16> %tmp3 228} 229 230define <4 x i32> @vmaxQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 231;CHECK: vmaxQs32: 232;CHECK: vmax.s32 233 %tmp1 = load <4 x i32>* %A 234 %tmp2 = load <4 x i32>* %B 235 %tmp3 = call <4 x i32> @llvm.arm.neon.vmaxs.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 236 ret <4 x i32> %tmp3 237} 238 239define <16 x i8> @vmaxQu8(<16 x i8>* %A, <16 x i8>* %B) nounwind { 240;CHECK: vmaxQu8: 241;CHECK: vmax.u8 242 %tmp1 = load <16 x i8>* %A 243 %tmp2 = load <16 x i8>* %B 244 %tmp3 = call <16 x i8> @llvm.arm.neon.vmaxu.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2) 245 ret <16 x i8> %tmp3 246} 247 248define <8 x i16> @vmaxQu16(<8 x i16>* %A, <8 x i16>* %B) nounwind { 249;CHECK: vmaxQu16: 250;CHECK: vmax.u16 251 %tmp1 = load <8 x i16>* %A 252 %tmp2 = load <8 x i16>* %B 253 %tmp3 = call <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2) 254 ret <8 x i16> %tmp3 255} 256 257define <4 x i32> @vmaxQu32(<4 x i32>* %A, <4 x i32>* %B) nounwind { 258;CHECK: vmaxQu32: 259;CHECK: vmax.u32 260 %tmp1 = load <4 x i32>* %A 261 %tmp2 = load <4 x i32>* %B 262 %tmp3 = call <4 x i32> @llvm.arm.neon.vmaxu.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2) 263 ret <4 x i32> %tmp3 264} 265 266define <4 x float> @vmaxQf32(<4 x float>* %A, <4 x float>* %B) nounwind { 267;CHECK: vmaxQf32: 268;CHECK: vmax.f32 269 %tmp1 = load <4 x float>* %A 270 %tmp2 = load <4 x float>* %B 271 %tmp3 = call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> %tmp1, <4 x float> %tmp2) 272 ret <4 x float> %tmp3 273} 274 275declare <8 x i8> @llvm.arm.neon.vmaxs.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 276declare <4 x i16> @llvm.arm.neon.vmaxs.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 277declare <2 x i32> @llvm.arm.neon.vmaxs.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 278 279declare <8 x i8> @llvm.arm.neon.vmaxu.v8i8(<8 x i8>, <8 x i8>) nounwind readnone 280declare <4 x i16> @llvm.arm.neon.vmaxu.v4i16(<4 x i16>, <4 x i16>) nounwind readnone 281declare <2 x i32> @llvm.arm.neon.vmaxu.v2i32(<2 x i32>, <2 x i32>) nounwind readnone 282 283declare <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float>, <2 x float>) nounwind readnone 284 285declare <16 x i8> @llvm.arm.neon.vmaxs.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 286declare <8 x i16> @llvm.arm.neon.vmaxs.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 287declare <4 x i32> @llvm.arm.neon.vmaxs.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 288 289declare <16 x i8> @llvm.arm.neon.vmaxu.v16i8(<16 x i8>, <16 x i8>) nounwind readnone 290declare <8 x i16> @llvm.arm.neon.vmaxu.v8i16(<8 x i16>, <8 x i16>) nounwind readnone 291declare <4 x i32> @llvm.arm.neon.vmaxu.v4i32(<4 x i32>, <4 x i32>) nounwind readnone 292 293declare <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float>, <4 x float>) nounwind readnone 294