1 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -load-store-vectorizer -S -o - %s | FileCheck %s 2 ; Copy of test/CodeGen/AMDGPU/merge-stores.ll with some additions 3 4 target datalayout = "e-p:32:32-p1:64:64-p2:64:64-p3:32:32-p4:64:64-p5:32:32-p24:64:64-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64" 5 6 ; TODO: Vector element tests 7 ; TODO: Non-zero base offset for load and store combinations 8 ; TODO: Same base addrspacecasted 9 10 11 ; CHECK-LABEL: @merge_global_store_2_constants_i8( 12 ; CHECK: store <2 x i8> <i8 -56, i8 123>, <2 x i8> addrspace(1)* %{{[0-9]+}}, align 2 13 define void @merge_global_store_2_constants_i8(i8 addrspace(1)* %out) #0 { 14 %out.gep.1 = getelementptr i8, i8 addrspace(1)* %out, i32 1 15 16 store i8 123, i8 addrspace(1)* %out.gep.1 17 store i8 456, i8 addrspace(1)* %out, align 2 18 ret void 19 } 20 21 ; CHECK-LABEL: @merge_global_store_2_constants_i8_natural_align 22 ; CHECK: store <2 x i8> 23 define void @merge_global_store_2_constants_i8_natural_align(i8 addrspace(1)* %out) #0 { 24 %out.gep.1 = getelementptr i8, i8 addrspace(1)* %out, i32 1 25 26 store i8 123, i8 addrspace(1)* %out.gep.1 27 store i8 456, i8 addrspace(1)* %out 28 ret void 29 } 30 31 ; CHECK-LABEL: @merge_global_store_2_constants_i16 32 ; CHECK: store <2 x i16> <i16 456, i16 123>, <2 x i16> addrspace(1)* %{{[0-9]+}}, align 4 33 define void @merge_global_store_2_constants_i16(i16 addrspace(1)* %out) #0 { 34 %out.gep.1 = getelementptr i16, i16 addrspace(1)* %out, i32 1 35 36 store i16 123, i16 addrspace(1)* %out.gep.1 37 store i16 456, i16 addrspace(1)* %out, align 4 38 ret void 39 } 40 41 ; CHECK-LABEL: @merge_global_store_2_constants_0_i16 42 ; CHECK: store <2 x i16> zeroinitializer, <2 x i16> addrspace(1)* %{{[0-9]+}}, align 4 43 define void @merge_global_store_2_constants_0_i16(i16 addrspace(1)* %out) #0 { 44 %out.gep.1 = getelementptr i16, i16 addrspace(1)* %out, i32 1 45 46 store i16 0, i16 addrspace(1)* %out.gep.1 47 store i16 0, i16 addrspace(1)* %out, align 4 48 ret void 49 } 50 51 ; CHECK-LABEL: @merge_global_store_2_constants_i16_natural_align 52 ; CHECK: store <2 x i16> 53 define void @merge_global_store_2_constants_i16_natural_align(i16 addrspace(1)* %out) #0 { 54 %out.gep.1 = getelementptr i16, i16 addrspace(1)* %out, i32 1 55 56 store i16 123, i16 addrspace(1)* %out.gep.1 57 store i16 456, i16 addrspace(1)* %out 58 ret void 59 } 60 61 ; CHECK-LABEL: @merge_global_store_2_constants_half_natural_align 62 ; CHECK: store <2 x half> 63 define void @merge_global_store_2_constants_half_natural_align(half addrspace(1)* %out) #0 { 64 %out.gep.1 = getelementptr half, half addrspace(1)* %out, i32 1 65 66 store half 2.0, half addrspace(1)* %out.gep.1 67 store half 1.0, half addrspace(1)* %out 68 ret void 69 } 70 71 ; CHECK-LABEL: @merge_global_store_2_constants_i32 72 ; CHECK: store <2 x i32> <i32 456, i32 123>, <2 x i32> addrspace(1)* %{{[0-9]+}}, align 4 73 define void @merge_global_store_2_constants_i32(i32 addrspace(1)* %out) #0 { 74 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 75 76 store i32 123, i32 addrspace(1)* %out.gep.1 77 store i32 456, i32 addrspace(1)* %out 78 ret void 79 } 80 81 ; CHECK-LABEL: @merge_global_store_2_constants_i32_f32 82 ; CHECK: store <2 x i32> <i32 456, i32 1065353216>, <2 x i32> addrspace(1)* %{{[0-9]+}}, align 4 83 define void @merge_global_store_2_constants_i32_f32(i32 addrspace(1)* %out) #0 { 84 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 85 %out.gep.1.bc = bitcast i32 addrspace(1)* %out.gep.1 to float addrspace(1)* 86 store float 1.0, float addrspace(1)* %out.gep.1.bc 87 store i32 456, i32 addrspace(1)* %out 88 ret void 89 } 90 91 ; CHECK-LABEL: @merge_global_store_2_constants_f32_i32 92 ; CHECK store <2 x float> <float 4.000000e+00, float 0x370EC00000000000>, <2 x float> addrspace(1)* %{{[0-9]+$}} 93 define void @merge_global_store_2_constants_f32_i32(float addrspace(1)* %out) #0 { 94 %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1 95 %out.gep.1.bc = bitcast float addrspace(1)* %out.gep.1 to i32 addrspace(1)* 96 store i32 123, i32 addrspace(1)* %out.gep.1.bc 97 store float 4.0, float addrspace(1)* %out 98 ret void 99 } 100 101 ; CHECK-LABEL: @merge_global_store_4_constants_i32 102 ; CHECK: store <4 x i32> <i32 1234, i32 123, i32 456, i32 333>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 103 define void @merge_global_store_4_constants_i32(i32 addrspace(1)* %out) #0 { 104 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 105 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 106 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 3 107 108 store i32 123, i32 addrspace(1)* %out.gep.1 109 store i32 456, i32 addrspace(1)* %out.gep.2 110 store i32 333, i32 addrspace(1)* %out.gep.3 111 store i32 1234, i32 addrspace(1)* %out 112 ret void 113 } 114 115 ; CHECK-LABEL: @merge_global_store_4_constants_f32_order 116 ; CHECK: store <4 x float> <float 8.000000e+00, float 1.000000e+00, float 2.000000e+00, float 4.000000e+00>, <4 x float> addrspace(1)* %{{[0-9]+}} 117 define void @merge_global_store_4_constants_f32_order(float addrspace(1)* %out) #0 { 118 %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1 119 %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2 120 %out.gep.3 = getelementptr float, float addrspace(1)* %out, i32 3 121 122 store float 8.0, float addrspace(1)* %out 123 store float 1.0, float addrspace(1)* %out.gep.1 124 store float 2.0, float addrspace(1)* %out.gep.2 125 store float 4.0, float addrspace(1)* %out.gep.3 126 ret void 127 } 128 129 ; First store is out of order. 130 ; CHECK-LABEL: @merge_global_store_4_constants_f32 131 ; CHECK: store <4 x float> <float 8.000000e+00, float 1.000000e+00, float 2.000000e+00, float 4.000000e+00>, <4 x float> addrspace(1)* %{{[0-9]+}}, align 4 132 define void @merge_global_store_4_constants_f32(float addrspace(1)* %out) #0 { 133 %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1 134 %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2 135 %out.gep.3 = getelementptr float, float addrspace(1)* %out, i32 3 136 137 store float 1.0, float addrspace(1)* %out.gep.1 138 store float 2.0, float addrspace(1)* %out.gep.2 139 store float 4.0, float addrspace(1)* %out.gep.3 140 store float 8.0, float addrspace(1)* %out 141 ret void 142 } 143 144 ; CHECK-LABEL: @merge_global_store_4_constants_mixed_i32_f32 145 ; CHECK: store <4 x i32> <i32 1090519040, i32 11, i32 1073741824, i32 17>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 146 define void @merge_global_store_4_constants_mixed_i32_f32(float addrspace(1)* %out) #0 { 147 %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1 148 %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2 149 %out.gep.3 = getelementptr float, float addrspace(1)* %out, i32 3 150 151 %out.gep.1.bc = bitcast float addrspace(1)* %out.gep.1 to i32 addrspace(1)* 152 %out.gep.3.bc = bitcast float addrspace(1)* %out.gep.3 to i32 addrspace(1)* 153 154 store i32 11, i32 addrspace(1)* %out.gep.1.bc 155 store float 2.0, float addrspace(1)* %out.gep.2 156 store i32 17, i32 addrspace(1)* %out.gep.3.bc 157 store float 8.0, float addrspace(1)* %out 158 ret void 159 } 160 161 ; CHECK-LABEL: @merge_global_store_3_constants_i32 162 ; CHECK: store <3 x i32> <i32 1234, i32 123, i32 456>, <3 x i32> addrspace(1)* %{{[0-9]+}}, align 4 163 define void @merge_global_store_3_constants_i32(i32 addrspace(1)* %out) #0 { 164 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 165 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 166 167 store i32 123, i32 addrspace(1)* %out.gep.1 168 store i32 456, i32 addrspace(1)* %out.gep.2 169 store i32 1234, i32 addrspace(1)* %out 170 ret void 171 } 172 173 ; CHECK-LABEL: @merge_global_store_2_constants_i64 174 ; CHECK: store <2 x i64> <i64 456, i64 123>, <2 x i64> addrspace(1)* %{{[0-9]+}}, align 8 175 define void @merge_global_store_2_constants_i64(i64 addrspace(1)* %out) #0 { 176 %out.gep.1 = getelementptr i64, i64 addrspace(1)* %out, i64 1 177 178 store i64 123, i64 addrspace(1)* %out.gep.1 179 store i64 456, i64 addrspace(1)* %out 180 ret void 181 } 182 183 ; CHECK-LABEL: @merge_global_store_4_constants_i64 184 ; CHECK: store <2 x i64> <i64 456, i64 333>, <2 x i64> addrspace(1)* %{{[0-9]+}}, align 8 185 ; CHECK: store <2 x i64> <i64 1234, i64 123>, <2 x i64> addrspace(1)* %{{[0-9]+}}, align 8 186 define void @merge_global_store_4_constants_i64(i64 addrspace(1)* %out) #0 { 187 %out.gep.1 = getelementptr i64, i64 addrspace(1)* %out, i64 1 188 %out.gep.2 = getelementptr i64, i64 addrspace(1)* %out, i64 2 189 %out.gep.3 = getelementptr i64, i64 addrspace(1)* %out, i64 3 190 191 store i64 123, i64 addrspace(1)* %out.gep.1 192 store i64 456, i64 addrspace(1)* %out.gep.2 193 store i64 333, i64 addrspace(1)* %out.gep.3 194 store i64 1234, i64 addrspace(1)* %out 195 ret void 196 } 197 198 ; CHECK-LABEL: @merge_global_store_2_adjacent_loads_i32 199 ; CHECK: [[LOAD:%[0-9]+]] = load <2 x i32> 200 ; CHECK: [[ELT0:%[0-9]+]] = extractelement <2 x i32> [[LOAD]], i32 0 201 ; CHECK: [[ELT1:%[0-9]+]] = extractelement <2 x i32> [[LOAD]], i32 1 202 ; CHECK: [[INSERT0:%[0-9]+]] = insertelement <2 x i32> undef, i32 [[ELT0]], i32 0 203 ; CHECK: [[INSERT1:%[0-9]+]] = insertelement <2 x i32> [[INSERT0]], i32 [[ELT1]], i32 1 204 ; CHECK: store <2 x i32> [[INSERT1]] 205 define void @merge_global_store_2_adjacent_loads_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 206 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 207 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 208 209 %lo = load i32, i32 addrspace(1)* %in 210 %hi = load i32, i32 addrspace(1)* %in.gep.1 211 212 store i32 %lo, i32 addrspace(1)* %out 213 store i32 %hi, i32 addrspace(1)* %out.gep.1 214 ret void 215 } 216 217 ; CHECK-LABEL: @merge_global_store_2_adjacent_loads_i32_nonzero_base 218 ; CHECK: extractelement 219 ; CHECK: extractelement 220 ; CHECK: insertelement 221 ; CHECK: insertelement 222 ; CHECK: store <2 x i32> 223 define void @merge_global_store_2_adjacent_loads_i32_nonzero_base(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 224 %in.gep.0 = getelementptr i32, i32 addrspace(1)* %in, i32 2 225 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 3 226 227 %out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i32 2 228 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 3 229 %lo = load i32, i32 addrspace(1)* %in.gep.0 230 %hi = load i32, i32 addrspace(1)* %in.gep.1 231 232 store i32 %lo, i32 addrspace(1)* %out.gep.0 233 store i32 %hi, i32 addrspace(1)* %out.gep.1 234 ret void 235 } 236 237 ; CHECK-LABEL: @merge_global_store_2_adjacent_loads_shuffle_i32 238 ; CHECK: [[LOAD:%[0-9]+]] = load <2 x i32> 239 ; CHECK: [[ELT0:%[0-9]+]] = extractelement <2 x i32> [[LOAD]], i32 0 240 ; CHECK: [[ELT1:%[0-9]+]] = extractelement <2 x i32> [[LOAD]], i32 1 241 ; CHECK: [[INSERT0:%[0-9]+]] = insertelement <2 x i32> undef, i32 [[ELT1]], i32 0 242 ; CHECK: [[INSERT1:%[0-9]+]] = insertelement <2 x i32> [[INSERT0]], i32 [[ELT0]], i32 1 243 ; CHECK: store <2 x i32> [[INSERT1]] 244 define void @merge_global_store_2_adjacent_loads_shuffle_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 245 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 246 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 247 248 %lo = load i32, i32 addrspace(1)* %in 249 %hi = load i32, i32 addrspace(1)* %in.gep.1 250 251 store i32 %hi, i32 addrspace(1)* %out 252 store i32 %lo, i32 addrspace(1)* %out.gep.1 253 ret void 254 } 255 256 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_i32 257 ; CHECK: load <4 x i32> 258 ; CHECK: store <4 x i32> 259 define void @merge_global_store_4_adjacent_loads_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 260 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 261 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 262 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 3 263 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 264 %in.gep.2 = getelementptr i32, i32 addrspace(1)* %in, i32 2 265 %in.gep.3 = getelementptr i32, i32 addrspace(1)* %in, i32 3 266 267 %x = load i32, i32 addrspace(1)* %in 268 %y = load i32, i32 addrspace(1)* %in.gep.1 269 %z = load i32, i32 addrspace(1)* %in.gep.2 270 %w = load i32, i32 addrspace(1)* %in.gep.3 271 272 store i32 %x, i32 addrspace(1)* %out 273 store i32 %y, i32 addrspace(1)* %out.gep.1 274 store i32 %z, i32 addrspace(1)* %out.gep.2 275 store i32 %w, i32 addrspace(1)* %out.gep.3 276 ret void 277 } 278 279 ; CHECK-LABEL: @merge_global_store_3_adjacent_loads_i32 280 ; CHECK: load <3 x i32> 281 ; CHECK: store <3 x i32> 282 define void @merge_global_store_3_adjacent_loads_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 283 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 284 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 285 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 286 %in.gep.2 = getelementptr i32, i32 addrspace(1)* %in, i32 2 287 288 %x = load i32, i32 addrspace(1)* %in 289 %y = load i32, i32 addrspace(1)* %in.gep.1 290 %z = load i32, i32 addrspace(1)* %in.gep.2 291 292 store i32 %x, i32 addrspace(1)* %out 293 store i32 %y, i32 addrspace(1)* %out.gep.1 294 store i32 %z, i32 addrspace(1)* %out.gep.2 295 ret void 296 } 297 298 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_f32 299 ; CHECK: load <4 x float> 300 ; CHECK: store <4 x float> 301 define void @merge_global_store_4_adjacent_loads_f32(float addrspace(1)* %out, float addrspace(1)* %in) #0 { 302 %out.gep.1 = getelementptr float, float addrspace(1)* %out, i32 1 303 %out.gep.2 = getelementptr float, float addrspace(1)* %out, i32 2 304 %out.gep.3 = getelementptr float, float addrspace(1)* %out, i32 3 305 %in.gep.1 = getelementptr float, float addrspace(1)* %in, i32 1 306 %in.gep.2 = getelementptr float, float addrspace(1)* %in, i32 2 307 %in.gep.3 = getelementptr float, float addrspace(1)* %in, i32 3 308 309 %x = load float, float addrspace(1)* %in 310 %y = load float, float addrspace(1)* %in.gep.1 311 %z = load float, float addrspace(1)* %in.gep.2 312 %w = load float, float addrspace(1)* %in.gep.3 313 314 store float %x, float addrspace(1)* %out 315 store float %y, float addrspace(1)* %out.gep.1 316 store float %z, float addrspace(1)* %out.gep.2 317 store float %w, float addrspace(1)* %out.gep.3 318 ret void 319 } 320 321 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_i32_nonzero_base 322 ; CHECK: load <4 x i32> 323 ; CHECK: store <4 x i32> 324 define void @merge_global_store_4_adjacent_loads_i32_nonzero_base(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 325 %in.gep.0 = getelementptr i32, i32 addrspace(1)* %in, i32 11 326 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 12 327 %in.gep.2 = getelementptr i32, i32 addrspace(1)* %in, i32 13 328 %in.gep.3 = getelementptr i32, i32 addrspace(1)* %in, i32 14 329 %out.gep.0 = getelementptr i32, i32 addrspace(1)* %out, i32 7 330 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 8 331 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 9 332 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 10 333 334 %x = load i32, i32 addrspace(1)* %in.gep.0 335 %y = load i32, i32 addrspace(1)* %in.gep.1 336 %z = load i32, i32 addrspace(1)* %in.gep.2 337 %w = load i32, i32 addrspace(1)* %in.gep.3 338 339 store i32 %x, i32 addrspace(1)* %out.gep.0 340 store i32 %y, i32 addrspace(1)* %out.gep.1 341 store i32 %z, i32 addrspace(1)* %out.gep.2 342 store i32 %w, i32 addrspace(1)* %out.gep.3 343 ret void 344 } 345 346 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_inverse_i32 347 ; CHECK: load <4 x i32> 348 ; CHECK: store <4 x i32> 349 define void @merge_global_store_4_adjacent_loads_inverse_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 350 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 351 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 352 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 3 353 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 354 %in.gep.2 = getelementptr i32, i32 addrspace(1)* %in, i32 2 355 %in.gep.3 = getelementptr i32, i32 addrspace(1)* %in, i32 3 356 357 %x = load i32, i32 addrspace(1)* %in 358 %y = load i32, i32 addrspace(1)* %in.gep.1 359 %z = load i32, i32 addrspace(1)* %in.gep.2 360 %w = load i32, i32 addrspace(1)* %in.gep.3 361 362 ; Make sure the barrier doesn't stop this 363 tail call void @llvm.amdgcn.s.barrier() #1 364 365 store i32 %w, i32 addrspace(1)* %out.gep.3 366 store i32 %z, i32 addrspace(1)* %out.gep.2 367 store i32 %y, i32 addrspace(1)* %out.gep.1 368 store i32 %x, i32 addrspace(1)* %out 369 370 ret void 371 } 372 373 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_shuffle_i32 374 ; CHECK: load <4 x i32> 375 ; CHECK: store <4 x i32> 376 define void @merge_global_store_4_adjacent_loads_shuffle_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) #0 { 377 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 378 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 379 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 3 380 %in.gep.1 = getelementptr i32, i32 addrspace(1)* %in, i32 1 381 %in.gep.2 = getelementptr i32, i32 addrspace(1)* %in, i32 2 382 %in.gep.3 = getelementptr i32, i32 addrspace(1)* %in, i32 3 383 384 %x = load i32, i32 addrspace(1)* %in 385 %y = load i32, i32 addrspace(1)* %in.gep.1 386 %z = load i32, i32 addrspace(1)* %in.gep.2 387 %w = load i32, i32 addrspace(1)* %in.gep.3 388 389 ; Make sure the barrier doesn't stop this 390 tail call void @llvm.amdgcn.s.barrier() #1 391 392 store i32 %w, i32 addrspace(1)* %out 393 store i32 %z, i32 addrspace(1)* %out.gep.1 394 store i32 %y, i32 addrspace(1)* %out.gep.2 395 store i32 %x, i32 addrspace(1)* %out.gep.3 396 397 ret void 398 } 399 400 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_i8 401 ; CHECK: load <4 x i8> 402 ; CHECK: extractelement <4 x i8> 403 ; CHECK: extractelement <4 x i8> 404 ; CHECK: extractelement <4 x i8> 405 ; CHECK: extractelement <4 x i8> 406 ; CHECK: insertelement <4 x i8> 407 ; CHECK: insertelement <4 x i8> 408 ; CHECK: insertelement <4 x i8> 409 ; CHECK: insertelement <4 x i8> 410 ; CHECK: store <4 x i8> 411 define void @merge_global_store_4_adjacent_loads_i8(i8 addrspace(1)* %out, i8 addrspace(1)* %in) #0 { 412 %out.gep.1 = getelementptr i8, i8 addrspace(1)* %out, i8 1 413 %out.gep.2 = getelementptr i8, i8 addrspace(1)* %out, i8 2 414 %out.gep.3 = getelementptr i8, i8 addrspace(1)* %out, i8 3 415 %in.gep.1 = getelementptr i8, i8 addrspace(1)* %in, i8 1 416 %in.gep.2 = getelementptr i8, i8 addrspace(1)* %in, i8 2 417 %in.gep.3 = getelementptr i8, i8 addrspace(1)* %in, i8 3 418 419 %x = load i8, i8 addrspace(1)* %in, align 4 420 %y = load i8, i8 addrspace(1)* %in.gep.1 421 %z = load i8, i8 addrspace(1)* %in.gep.2 422 %w = load i8, i8 addrspace(1)* %in.gep.3 423 424 store i8 %x, i8 addrspace(1)* %out, align 4 425 store i8 %y, i8 addrspace(1)* %out.gep.1 426 store i8 %z, i8 addrspace(1)* %out.gep.2 427 store i8 %w, i8 addrspace(1)* %out.gep.3 428 ret void 429 } 430 431 ; CHECK-LABEL: @merge_global_store_4_adjacent_loads_i8_natural_align 432 ; CHECK: load <4 x i8> 433 ; CHECK: store <4 x i8> 434 define void @merge_global_store_4_adjacent_loads_i8_natural_align(i8 addrspace(1)* %out, i8 addrspace(1)* %in) #0 { 435 %out.gep.1 = getelementptr i8, i8 addrspace(1)* %out, i8 1 436 %out.gep.2 = getelementptr i8, i8 addrspace(1)* %out, i8 2 437 %out.gep.3 = getelementptr i8, i8 addrspace(1)* %out, i8 3 438 %in.gep.1 = getelementptr i8, i8 addrspace(1)* %in, i8 1 439 %in.gep.2 = getelementptr i8, i8 addrspace(1)* %in, i8 2 440 %in.gep.3 = getelementptr i8, i8 addrspace(1)* %in, i8 3 441 442 %x = load i8, i8 addrspace(1)* %in 443 %y = load i8, i8 addrspace(1)* %in.gep.1 444 %z = load i8, i8 addrspace(1)* %in.gep.2 445 %w = load i8, i8 addrspace(1)* %in.gep.3 446 447 store i8 %x, i8 addrspace(1)* %out 448 store i8 %y, i8 addrspace(1)* %out.gep.1 449 store i8 %z, i8 addrspace(1)* %out.gep.2 450 store i8 %w, i8 addrspace(1)* %out.gep.3 451 ret void 452 } 453 454 ; CHECK-LABEL: @merge_global_store_4_vector_elts_loads_v4i32 455 ; CHECK: load <4 x i32> 456 ; CHECK: store <4 x i32> 457 define void @merge_global_store_4_vector_elts_loads_v4i32(i32 addrspace(1)* %out, <4 x i32> addrspace(1)* %in) #0 { 458 %out.gep.1 = getelementptr i32, i32 addrspace(1)* %out, i32 1 459 %out.gep.2 = getelementptr i32, i32 addrspace(1)* %out, i32 2 460 %out.gep.3 = getelementptr i32, i32 addrspace(1)* %out, i32 3 461 %vec = load <4 x i32>, <4 x i32> addrspace(1)* %in 462 463 %x = extractelement <4 x i32> %vec, i32 0 464 %y = extractelement <4 x i32> %vec, i32 1 465 %z = extractelement <4 x i32> %vec, i32 2 466 %w = extractelement <4 x i32> %vec, i32 3 467 468 store i32 %x, i32 addrspace(1)* %out 469 store i32 %y, i32 addrspace(1)* %out.gep.1 470 store i32 %z, i32 addrspace(1)* %out.gep.2 471 store i32 %w, i32 addrspace(1)* %out.gep.3 472 ret void 473 } 474 475 ; CHECK-LABEL: @merge_local_store_2_constants_i8 476 ; CHECK: store <2 x i8> <i8 -56, i8 123>, <2 x i8> addrspace(3)* %{{[0-9]+}}, align 2 477 define void @merge_local_store_2_constants_i8(i8 addrspace(3)* %out) #0 { 478 %out.gep.1 = getelementptr i8, i8 addrspace(3)* %out, i32 1 479 480 store i8 123, i8 addrspace(3)* %out.gep.1 481 store i8 456, i8 addrspace(3)* %out, align 2 482 ret void 483 } 484 485 ; CHECK-LABEL: @merge_local_store_2_constants_i32 486 ; CHECK: store <2 x i32> <i32 456, i32 123>, <2 x i32> addrspace(3)* %{{[0-9]+}}, align 4 487 define void @merge_local_store_2_constants_i32(i32 addrspace(3)* %out) #0 { 488 %out.gep.1 = getelementptr i32, i32 addrspace(3)* %out, i32 1 489 490 store i32 123, i32 addrspace(3)* %out.gep.1 491 store i32 456, i32 addrspace(3)* %out 492 ret void 493 } 494 495 ; CHECK-LABEL: @merge_local_store_2_constants_i32_align_2 496 ; CHECK: store i32 497 ; CHECK: store i32 498 define void @merge_local_store_2_constants_i32_align_2(i32 addrspace(3)* %out) #0 { 499 %out.gep.1 = getelementptr i32, i32 addrspace(3)* %out, i32 1 500 501 store i32 123, i32 addrspace(3)* %out.gep.1, align 2 502 store i32 456, i32 addrspace(3)* %out, align 2 503 ret void 504 } 505 506 ; CHECK-LABEL: @merge_local_store_4_constants_i32 507 ; CHECK: store <2 x i32> <i32 456, i32 333>, <2 x i32> addrspace(3)* 508 ; CHECK: store <2 x i32> <i32 1234, i32 123>, <2 x i32> addrspace(3)* 509 define void @merge_local_store_4_constants_i32(i32 addrspace(3)* %out) #0 { 510 %out.gep.1 = getelementptr i32, i32 addrspace(3)* %out, i32 1 511 %out.gep.2 = getelementptr i32, i32 addrspace(3)* %out, i32 2 512 %out.gep.3 = getelementptr i32, i32 addrspace(3)* %out, i32 3 513 514 store i32 123, i32 addrspace(3)* %out.gep.1 515 store i32 456, i32 addrspace(3)* %out.gep.2 516 store i32 333, i32 addrspace(3)* %out.gep.3 517 store i32 1234, i32 addrspace(3)* %out 518 ret void 519 } 520 521 ; CHECK-LABEL: @merge_global_store_5_constants_i32 522 ; CHECK: store <4 x i32> <i32 9, i32 12, i32 16, i32 -12>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 523 ; CHECK: store i32 524 define void @merge_global_store_5_constants_i32(i32 addrspace(1)* %out) { 525 store i32 9, i32 addrspace(1)* %out, align 4 526 %idx1 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 1 527 store i32 12, i32 addrspace(1)* %idx1, align 4 528 %idx2 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 2 529 store i32 16, i32 addrspace(1)* %idx2, align 4 530 %idx3 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 3 531 store i32 -12, i32 addrspace(1)* %idx3, align 4 532 %idx4 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 4 533 store i32 11, i32 addrspace(1)* %idx4, align 4 534 ret void 535 } 536 537 ; CHECK-LABEL: @merge_global_store_6_constants_i32 538 ; CHECK: store <4 x i32> <i32 13, i32 15, i32 62, i32 63>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 539 ; CHECK: store <2 x i32> <i32 11, i32 123>, <2 x i32> addrspace(1)* %{{[0-9]+}}, align 4 540 define void @merge_global_store_6_constants_i32(i32 addrspace(1)* %out) { 541 store i32 13, i32 addrspace(1)* %out, align 4 542 %idx1 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 1 543 store i32 15, i32 addrspace(1)* %idx1, align 4 544 %idx2 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 2 545 store i32 62, i32 addrspace(1)* %idx2, align 4 546 %idx3 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 3 547 store i32 63, i32 addrspace(1)* %idx3, align 4 548 %idx4 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 4 549 store i32 11, i32 addrspace(1)* %idx4, align 4 550 %idx5 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 5 551 store i32 123, i32 addrspace(1)* %idx5, align 4 552 ret void 553 } 554 555 ; CHECK-LABEL: @merge_global_store_7_constants_i32 556 ; CHECK: store <4 x i32> <i32 34, i32 999, i32 65, i32 33>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 557 ; CHECK: store <3 x i32> <i32 98, i32 91, i32 212>, <3 x i32> addrspace(1)* %{{[0-9]+}}, align 4 558 define void @merge_global_store_7_constants_i32(i32 addrspace(1)* %out) { 559 store i32 34, i32 addrspace(1)* %out, align 4 560 %idx1 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 1 561 store i32 999, i32 addrspace(1)* %idx1, align 4 562 %idx2 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 2 563 store i32 65, i32 addrspace(1)* %idx2, align 4 564 %idx3 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 3 565 store i32 33, i32 addrspace(1)* %idx3, align 4 566 %idx4 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 4 567 store i32 98, i32 addrspace(1)* %idx4, align 4 568 %idx5 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 5 569 store i32 91, i32 addrspace(1)* %idx5, align 4 570 %idx6 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 6 571 store i32 212, i32 addrspace(1)* %idx6, align 4 572 ret void 573 } 574 575 ; CHECK-LABEL: @merge_global_store_8_constants_i32 576 ; CHECK: store <4 x i32> <i32 34, i32 999, i32 65, i32 33>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 577 ; CHECK: store <4 x i32> <i32 98, i32 91, i32 212, i32 999>, <4 x i32> addrspace(1)* %{{[0-9]+}}, align 4 578 define void @merge_global_store_8_constants_i32(i32 addrspace(1)* %out) { 579 store i32 34, i32 addrspace(1)* %out, align 4 580 %idx1 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 1 581 store i32 999, i32 addrspace(1)* %idx1, align 4 582 %idx2 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 2 583 store i32 65, i32 addrspace(1)* %idx2, align 4 584 %idx3 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 3 585 store i32 33, i32 addrspace(1)* %idx3, align 4 586 %idx4 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 4 587 store i32 98, i32 addrspace(1)* %idx4, align 4 588 %idx5 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 5 589 store i32 91, i32 addrspace(1)* %idx5, align 4 590 %idx6 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 6 591 store i32 212, i32 addrspace(1)* %idx6, align 4 592 %idx7 = getelementptr inbounds i32, i32 addrspace(1)* %out, i64 7 593 store i32 999, i32 addrspace(1)* %idx7, align 4 594 ret void 595 } 596 597 ; CHECK-LABEL: @copy_v3i32_align4 598 ; CHECK: %vec = load <3 x i32>, <3 x i32> addrspace(1)* %in, align 4 599 ; CHECK: store <3 x i32> %vec, <3 x i32> addrspace(1)* %out 600 define void @copy_v3i32_align4(<3 x i32> addrspace(1)* noalias %out, <3 x i32> addrspace(1)* noalias %in) #0 { 601 %vec = load <3 x i32>, <3 x i32> addrspace(1)* %in, align 4 602 store <3 x i32> %vec, <3 x i32> addrspace(1)* %out 603 ret void 604 } 605 606 ; CHECK-LABEL: @copy_v3i64_align4 607 ; CHECK: %vec = load <3 x i64>, <3 x i64> addrspace(1)* %in, align 4 608 ; CHECK: store <3 x i64> %vec, <3 x i64> addrspace(1)* %out 609 define void @copy_v3i64_align4(<3 x i64> addrspace(1)* noalias %out, <3 x i64> addrspace(1)* noalias %in) #0 { 610 %vec = load <3 x i64>, <3 x i64> addrspace(1)* %in, align 4 611 store <3 x i64> %vec, <3 x i64> addrspace(1)* %out 612 ret void 613 } 614 615 ; CHECK-LABEL: @copy_v3f32_align4 616 ; CHECK: %vec = load <3 x float>, <3 x float> addrspace(1)* %in, align 4 617 ; CHECK: store <3 x float> 618 define void @copy_v3f32_align4(<3 x float> addrspace(1)* noalias %out, <3 x float> addrspace(1)* noalias %in) #0 { 619 %vec = load <3 x float>, <3 x float> addrspace(1)* %in, align 4 620 %fadd = fadd <3 x float> %vec, <float 1.0, float 2.0, float 4.0> 621 store <3 x float> %fadd, <3 x float> addrspace(1)* %out 622 ret void 623 } 624 625 ; CHECK-LABEL: @copy_v3f64_align4 626 ; CHECK: %vec = load <3 x double>, <3 x double> addrspace(1)* %in, align 4 627 ; CHECK: store <3 x double> %fadd, <3 x double> addrspace(1)* %out 628 define void @copy_v3f64_align4(<3 x double> addrspace(1)* noalias %out, <3 x double> addrspace(1)* noalias %in) #0 { 629 %vec = load <3 x double>, <3 x double> addrspace(1)* %in, align 4 630 %fadd = fadd <3 x double> %vec, <double 1.0, double 2.0, double 4.0> 631 store <3 x double> %fadd, <3 x double> addrspace(1)* %out 632 ret void 633 } 634 635 declare void @llvm.amdgcn.s.barrier() #1 636 637 attributes #0 = { nounwind } 638 attributes #1 = { convergent nounwind } 639
