1; RUN: llc -verify-machineinstrs -march=amdgcn -mattr=+max-private-element-size-16 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s 2; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=+max-private-element-size-16 < %s | FileCheck -check-prefix=GCN -check-prefix=SI %s 3 4; FIXME: Broken on evergreen 5; FIXME: For some reason the 8 and 16 vectors are being stored as 6; individual elements instead of 128-bit stores. 7 8 9; FIXME: Why is the constant moved into the intermediate register and 10; not just directly into the vector component? 11 12; GCN-LABEL: {{^}}insertelement_v4f32_0: 13; GCN: s_load_dwordx4 s{{\[}}[[LOW_REG:[0-9]+]]: 14; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 15; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 16; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 17; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 18; GCN-DAG: v_mov_b32_e32 [[CONSTREG:v[0-9]+]], 0x40a00000 19; GCN-DAG: v_mov_b32_e32 v[[LOW_REG]], [[CONSTREG]] 20; GCN: buffer_store_dwordx4 v{{\[}}[[LOW_REG]]: 21define void @insertelement_v4f32_0(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind { 22 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 0 23 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 24 ret void 25} 26 27; GCN-LABEL: {{^}}insertelement_v4f32_1: 28define void @insertelement_v4f32_1(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind { 29 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 1 30 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 31 ret void 32} 33 34; GCN-LABEL: {{^}}insertelement_v4f32_2: 35define void @insertelement_v4f32_2(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind { 36 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 2 37 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 38 ret void 39} 40 41; GCN-LABEL: {{^}}insertelement_v4f32_3: 42define void @insertelement_v4f32_3(<4 x float> addrspace(1)* %out, <4 x float> %a) nounwind { 43 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 3 44 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 45 ret void 46} 47 48; GCN-LABEL: {{^}}insertelement_v4i32_0: 49define void @insertelement_v4i32_0(<4 x i32> addrspace(1)* %out, <4 x i32> %a) nounwind { 50 %vecins = insertelement <4 x i32> %a, i32 999, i32 0 51 store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16 52 ret void 53} 54 55; GCN-LABEL: {{^}}insertelement_v3f32_1: 56define void @insertelement_v3f32_1(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind { 57 %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 1 58 store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16 59 ret void 60} 61 62; GCN-LABEL: {{^}}insertelement_v3f32_2: 63define void @insertelement_v3f32_2(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind { 64 %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 2 65 store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16 66 ret void 67} 68 69; GCN-LABEL: {{^}}insertelement_v3f32_3: 70define void @insertelement_v3f32_3(<3 x float> addrspace(1)* %out, <3 x float> %a) nounwind { 71 %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 3 72 store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16 73 ret void 74} 75 76; GCN-LABEL: {{^}}dynamic_insertelement_v2f32: 77; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 78; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 79; GCN: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 80define void @dynamic_insertelement_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, i32 %b) nounwind { 81 %vecins = insertelement <2 x float> %a, float 5.000000e+00, i32 %b 82 store <2 x float> %vecins, <2 x float> addrspace(1)* %out, align 8 83 ret void 84} 85 86; GCN-LABEL: {{^}}dynamic_insertelement_v3f32: 87; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 88; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 89; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 90; GCN-DAG: buffer_store_dword v 91define void @dynamic_insertelement_v3f32(<3 x float> addrspace(1)* %out, <3 x float> %a, i32 %b) nounwind { 92 %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 %b 93 store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16 94 ret void 95} 96 97; GCN-LABEL: {{^}}dynamic_insertelement_v4f32: 98; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 99; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 100; GCN: buffer_store_dwordx4 {{v\[}}[[LOW_RESULT_REG]]: 101define void @dynamic_insertelement_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %a, i32 %b) nounwind { 102 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 %b 103 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 104 ret void 105} 106 107; GCN-LABEL: {{^}}dynamic_insertelement_v8f32: 108; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}} 109; GCN: buffer_store_dwordx4 110; GCN: buffer_store_dwordx4 111define void @dynamic_insertelement_v8f32(<8 x float> addrspace(1)* %out, <8 x float> %a, i32 %b) nounwind { 112 %vecins = insertelement <8 x float> %a, float 5.000000e+00, i32 %b 113 store <8 x float> %vecins, <8 x float> addrspace(1)* %out, align 32 114 ret void 115} 116 117; GCN-LABEL: {{^}}dynamic_insertelement_v16f32: 118; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}} 119; GCN: buffer_store_dwordx4 120; GCN: buffer_store_dwordx4 121; GCN: buffer_store_dwordx4 122; GCN: buffer_store_dwordx4 123define void @dynamic_insertelement_v16f32(<16 x float> addrspace(1)* %out, <16 x float> %a, i32 %b) nounwind { 124 %vecins = insertelement <16 x float> %a, float 5.000000e+00, i32 %b 125 store <16 x float> %vecins, <16 x float> addrspace(1)* %out, align 64 126 ret void 127} 128 129; GCN-LABEL: {{^}}dynamic_insertelement_v2i32: 130; GCN: v_movreld_b32 131; GCN: buffer_store_dwordx2 132define void @dynamic_insertelement_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> %a, i32 %b) nounwind { 133 %vecins = insertelement <2 x i32> %a, i32 5, i32 %b 134 store <2 x i32> %vecins, <2 x i32> addrspace(1)* %out, align 8 135 ret void 136} 137 138; GCN-LABEL: {{^}}dynamic_insertelement_v3i32: 139; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 5 140; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 141; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 142; GCN-DAG: buffer_store_dword v 143define void @dynamic_insertelement_v3i32(<3 x i32> addrspace(1)* %out, <3 x i32> %a, i32 %b) nounwind { 144 %vecins = insertelement <3 x i32> %a, i32 5, i32 %b 145 store <3 x i32> %vecins, <3 x i32> addrspace(1)* %out, align 16 146 ret void 147} 148 149; GCN-LABEL: {{^}}dynamic_insertelement_v4i32: 150; GCN: v_movreld_b32 151; GCN: buffer_store_dwordx4 152define void @dynamic_insertelement_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %a, i32 %b) nounwind { 153 %vecins = insertelement <4 x i32> %a, i32 5, i32 %b 154 store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16 155 ret void 156} 157 158; GCN-LABEL: {{^}}dynamic_insertelement_v8i32: 159; GCN: v_movreld_b32 160; GCN: buffer_store_dwordx4 161; GCN: buffer_store_dwordx4 162define void @dynamic_insertelement_v8i32(<8 x i32> addrspace(1)* %out, <8 x i32> %a, i32 %b) nounwind { 163 %vecins = insertelement <8 x i32> %a, i32 5, i32 %b 164 store <8 x i32> %vecins, <8 x i32> addrspace(1)* %out, align 32 165 ret void 166} 167 168; GCN-LABEL: {{^}}dynamic_insertelement_v16i32: 169; GCN: v_movreld_b32 170; GCN: buffer_store_dwordx4 171; GCN: buffer_store_dwordx4 172; GCN: buffer_store_dwordx4 173; GCN: buffer_store_dwordx4 174define void @dynamic_insertelement_v16i32(<16 x i32> addrspace(1)* %out, <16 x i32> %a, i32 %b) nounwind { 175 %vecins = insertelement <16 x i32> %a, i32 5, i32 %b 176 store <16 x i32> %vecins, <16 x i32> addrspace(1)* %out, align 64 177 ret void 178} 179 180; GCN-LABEL: {{^}}dynamic_insertelement_v2i16: 181define void @dynamic_insertelement_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> %a, i32 %b) nounwind { 182 %vecins = insertelement <2 x i16> %a, i16 5, i32 %b 183 store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out, align 8 184 ret void 185} 186 187; GCN-LABEL: {{^}}dynamic_insertelement_v3i16: 188define void @dynamic_insertelement_v3i16(<3 x i16> addrspace(1)* %out, <3 x i16> %a, i32 %b) nounwind { 189 %vecins = insertelement <3 x i16> %a, i16 5, i32 %b 190 store <3 x i16> %vecins, <3 x i16> addrspace(1)* %out, align 8 191 ret void 192} 193 194; GCN-LABEL: {{^}}dynamic_insertelement_v4i16: 195; GCN: buffer_load_ushort v{{[0-9]+}}, off 196; GCN: buffer_load_ushort v{{[0-9]+}}, off 197; GCN: buffer_load_ushort v{{[0-9]+}}, off 198; GCN: buffer_load_ushort v{{[0-9]+}}, off 199 200; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:6 201; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:4 202; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 203; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 204; GCN: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 205 206; GCN: buffer_load_ushort 207; GCN: buffer_load_ushort 208; GCN: buffer_load_ushort 209; GCN: buffer_load_ushort 210 211; GCN: buffer_store_short v{{[0-9]+}}, off 212; GCN: buffer_store_short v{{[0-9]+}}, off 213; GCN: buffer_store_short v{{[0-9]+}}, off 214; GCN: buffer_store_short v{{[0-9]+}}, off 215define void @dynamic_insertelement_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> %a, i32 %b) nounwind { 216 %vecins = insertelement <4 x i16> %a, i16 5, i32 %b 217 store <4 x i16> %vecins, <4 x i16> addrspace(1)* %out, align 8 218 ret void 219} 220 221; GCN-LABEL: {{^}}dynamic_insertelement_v2i8: 222; GCN: buffer_load_ubyte v{{[0-9]+}}, off 223; GCN: buffer_load_ubyte v{{[0-9]+}}, off 224 225; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 226; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 227 228; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 229 230; GCN: buffer_load_ubyte 231; GCN: buffer_load_ubyte 232 233; GCN: buffer_store_byte v{{[0-9]+}}, off 234; GCN: buffer_store_byte v{{[0-9]+}}, off 235define void @dynamic_insertelement_v2i8(<2 x i8> addrspace(1)* %out, <2 x i8> %a, i32 %b) nounwind { 236 %vecins = insertelement <2 x i8> %a, i8 5, i32 %b 237 store <2 x i8> %vecins, <2 x i8> addrspace(1)* %out, align 8 238 ret void 239} 240 241; GCN-LABEL: {{^}}dynamic_insertelement_v3i8: 242; GCN: buffer_load_ubyte v{{[0-9]+}}, off 243; GCN: buffer_load_ubyte v{{[0-9]+}}, off 244; GCN: buffer_load_ubyte v{{[0-9]+}}, off 245 246; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 247; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 248; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 249 250; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 251 252; GCN: buffer_load_ubyte 253; GCN: buffer_load_ubyte 254; GCN: buffer_load_ubyte 255 256; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off 257; GCN-DAG: buffer_store_short v{{[0-9]+}}, off 258define void @dynamic_insertelement_v3i8(<3 x i8> addrspace(1)* %out, <3 x i8> %a, i32 %b) nounwind { 259 %vecins = insertelement <3 x i8> %a, i8 5, i32 %b 260 store <3 x i8> %vecins, <3 x i8> addrspace(1)* %out, align 4 261 ret void 262} 263 264; GCN-LABEL: {{^}}dynamic_insertelement_v4i8: 265; GCN: buffer_load_ubyte v{{[0-9]+}}, off 266; GCN: buffer_load_ubyte v{{[0-9]+}}, off 267; GCN: buffer_load_ubyte v{{[0-9]+}}, off 268; GCN: buffer_load_ubyte v{{[0-9]+}}, off 269 270; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:3 271; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 272; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 273; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 274 275; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 276 277; GCN: buffer_load_ubyte 278; GCN: buffer_load_ubyte 279; GCN: buffer_load_ubyte 280; GCN: buffer_load_ubyte 281 282; GCN: buffer_store_byte v{{[0-9]+}}, off 283; GCN: buffer_store_byte v{{[0-9]+}}, off 284; GCN: buffer_store_byte v{{[0-9]+}}, off 285; GCN: buffer_store_byte v{{[0-9]+}}, off 286define void @dynamic_insertelement_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> %a, i32 %b) nounwind { 287 %vecins = insertelement <4 x i8> %a, i8 5, i32 %b 288 store <4 x i8> %vecins, <4 x i8> addrspace(1)* %out, align 4 289 ret void 290} 291 292; GCN-LABEL: {{^}}dynamic_insertelement_v8i8: 293define void @dynamic_insertelement_v8i8(<8 x i8> addrspace(1)* %out, <8 x i8> %a, i32 %b) nounwind { 294 %vecins = insertelement <8 x i8> %a, i8 5, i32 %b 295 store <8 x i8> %vecins, <8 x i8> addrspace(1)* %out, align 8 296 ret void 297} 298 299; GCN-LABEL: {{^}}dynamic_insertelement_v16i8: 300define void @dynamic_insertelement_v16i8(<16 x i8> addrspace(1)* %out, <16 x i8> %a, i32 %b) nounwind { 301 %vecins = insertelement <16 x i8> %a, i8 5, i32 %b 302 store <16 x i8> %vecins, <16 x i8> addrspace(1)* %out, align 16 303 ret void 304} 305 306; This test requires handling INSERT_SUBREG in SIFixSGPRCopies. Check that 307; the compiler doesn't crash. 308; GCN-LABEL: {{^}}insert_split_bb: 309define void @insert_split_bb(<2 x i32> addrspace(1)* %out, i32 addrspace(1)* %in, i32 %a, i32 %b) { 310entry: 311 %0 = insertelement <2 x i32> undef, i32 %a, i32 0 312 %1 = icmp eq i32 %a, 0 313 br i1 %1, label %if, label %else 314 315if: 316 %2 = load i32, i32 addrspace(1)* %in 317 %3 = insertelement <2 x i32> %0, i32 %2, i32 1 318 br label %endif 319 320else: 321 %4 = getelementptr i32, i32 addrspace(1)* %in, i32 1 322 %5 = load i32, i32 addrspace(1)* %4 323 %6 = insertelement <2 x i32> %0, i32 %5, i32 1 324 br label %endif 325 326endif: 327 %7 = phi <2 x i32> [%3, %if], [%6, %else] 328 store <2 x i32> %7, <2 x i32> addrspace(1)* %out 329 ret void 330} 331 332; GCN-LABEL: {{^}}dynamic_insertelement_v2f64: 333; GCN: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x11|0x44}}{{$}} 334; GCN-DAG: s_lshl_b32 [[SCALEDIDX:s[0-9]+]], [[IDX]], 1{{$}} 335; GCN-DAG: v_mov_b32_e32 [[ELT0:v[0-9]+]], 0{{$}} 336 337; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 338; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 339; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 340; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 341 342; GCN: s_mov_b32 m0, [[SCALEDIDX]] 343; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT0]] 344 345; Increment to next element. 346; FIXME: Should be able to manipulate m0 directly instead of add and 347; copy. 348 349; GCN: s_or_b32 [[IDX1:s[0-9]+]], [[SCALEDIDX]], 1 350; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0x40200000 351; GCN-DAG: s_mov_b32 m0, [[IDX1]] 352; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]] 353 354; GCN: buffer_store_dwordx4 355; GCN: s_endpgm 356define void @dynamic_insertelement_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %a, i32 %b) nounwind { 357 %vecins = insertelement <2 x double> %a, double 8.0, i32 %b 358 store <2 x double> %vecins, <2 x double> addrspace(1)* %out, align 16 359 ret void 360} 361 362; FIXME: Inline immediate should be folded into v_movreld_b32. 363; GCN-LABEL: {{^}}dynamic_insertelement_v2i64: 364 365; GCN-DAG: v_mov_b32_e32 [[ELT0:v[0-9]+]], 5{{$}} 366; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0{{$}} 367 368; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT0]] 369; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]] 370 371; GCN: buffer_store_dwordx4 372; GCN: s_endpgm 373define void @dynamic_insertelement_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> %a, i32 %b) nounwind { 374 %vecins = insertelement <2 x i64> %a, i64 5, i32 %b 375 store <2 x i64> %vecins, <2 x i64> addrspace(1)* %out, align 8 376 ret void 377} 378 379; GCN-LABEL: {{^}}dynamic_insertelement_v3i64: 380define void @dynamic_insertelement_v3i64(<3 x i64> addrspace(1)* %out, <3 x i64> %a, i32 %b) nounwind { 381 %vecins = insertelement <3 x i64> %a, i64 5, i32 %b 382 store <3 x i64> %vecins, <3 x i64> addrspace(1)* %out, align 32 383 ret void 384} 385 386; FIXME: Should be able to do without stack access. The used stack 387; space is also 2x what should be required. 388 389; GCN-LABEL: {{^}}dynamic_insertelement_v4f64: 390; GCN: SCRATCH_RSRC_DWORD 391 392; Stack store 393 394; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 395; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 396 397; Write element 398; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 399 400; Stack reload 401; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 402; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 403 404; Store result 405; GCN: buffer_store_dwordx4 406; GCN: buffer_store_dwordx4 407; GCN: s_endpgm 408; GCN: ScratchSize: 64 409 410define void @dynamic_insertelement_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %a, i32 %b) nounwind { 411 %vecins = insertelement <4 x double> %a, double 8.0, i32 %b 412 store <4 x double> %vecins, <4 x double> addrspace(1)* %out, align 16 413 ret void 414} 415 416; GCN-LABEL: {{^}}dynamic_insertelement_v8f64: 417; GCN: SCRATCH_RSRC_DWORD 418 419; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 420; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 421; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:32{{$}} 422; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:48{{$}} 423 424; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 425 426; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 427; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 428; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 429; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 430 431; GCN: buffer_store_dwordx4 432; GCN: buffer_store_dwordx4 433; GCN: buffer_store_dwordx4 434; GCN: buffer_store_dwordx4 435; GCN: s_endpgm 436; GCN: ScratchSize: 128 437define void @dynamic_insertelement_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %a, i32 %b) nounwind { 438 %vecins = insertelement <8 x double> %a, double 8.0, i32 %b 439 store <8 x double> %vecins, <8 x double> addrspace(1)* %out, align 16 440 ret void 441} 442