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 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:[0-9]+]], [[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: {{^}}insertelement_to_sgpr: 77; GCN-NOT: v_readfirstlane 78define amdgpu_ps <4 x float> @insertelement_to_sgpr() nounwind { 79 %tmp = load <4 x i32>, <4 x i32> addrspace(2)* undef 80 %tmp1 = insertelement <4 x i32> %tmp, i32 0, i32 0 81 %tmp2 = call <4 x float> @llvm.SI.gather4.lz.v2i32(<2 x i32> undef, <8 x i32> undef, <4 x i32> %tmp1, i32 8, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0) 82 ret <4 x float> %tmp2 83} 84 85; GCN-LABEL: {{^}}dynamic_insertelement_v2f32: 86; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 87; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 88; GCN: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 89define void @dynamic_insertelement_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, i32 %b) nounwind { 90 %vecins = insertelement <2 x float> %a, float 5.000000e+00, i32 %b 91 store <2 x float> %vecins, <2 x float> addrspace(1)* %out, align 8 92 ret void 93} 94 95; GCN-LABEL: {{^}}dynamic_insertelement_v3f32: 96; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 97; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 98; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 99; GCN-DAG: buffer_store_dword v 100define void @dynamic_insertelement_v3f32(<3 x float> addrspace(1)* %out, <3 x float> %a, i32 %b) nounwind { 101 %vecins = insertelement <3 x float> %a, float 5.000000e+00, i32 %b 102 store <3 x float> %vecins, <3 x float> addrspace(1)* %out, align 16 103 ret void 104} 105 106; GCN-LABEL: {{^}}dynamic_insertelement_v4f32: 107; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 0x40a00000 108; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 109; GCN: buffer_store_dwordx4 {{v\[}}[[LOW_RESULT_REG]]: 110define void @dynamic_insertelement_v4f32(<4 x float> addrspace(1)* %out, <4 x float> %a, i32 %b) nounwind { 111 %vecins = insertelement <4 x float> %a, float 5.000000e+00, i32 %b 112 store <4 x float> %vecins, <4 x float> addrspace(1)* %out, align 16 113 ret void 114} 115 116; GCN-LABEL: {{^}}dynamic_insertelement_v8f32: 117; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}} 118; GCN: buffer_store_dwordx4 119; GCN: buffer_store_dwordx4 120define void @dynamic_insertelement_v8f32(<8 x float> addrspace(1)* %out, <8 x float> %a, i32 %b) nounwind { 121 %vecins = insertelement <8 x float> %a, float 5.000000e+00, i32 %b 122 store <8 x float> %vecins, <8 x float> addrspace(1)* %out, align 32 123 ret void 124} 125 126; GCN-LABEL: {{^}}dynamic_insertelement_v16f32: 127; GCN: v_movreld_b32_e32 v{{[0-9]+}}, v{{[0-9]+}} 128; GCN: buffer_store_dwordx4 129; GCN: buffer_store_dwordx4 130; GCN: buffer_store_dwordx4 131; GCN: buffer_store_dwordx4 132define void @dynamic_insertelement_v16f32(<16 x float> addrspace(1)* %out, <16 x float> %a, i32 %b) nounwind { 133 %vecins = insertelement <16 x float> %a, float 5.000000e+00, i32 %b 134 store <16 x float> %vecins, <16 x float> addrspace(1)* %out, align 64 135 ret void 136} 137 138; GCN-LABEL: {{^}}dynamic_insertelement_v2i32: 139; GCN: v_movreld_b32 140; GCN: buffer_store_dwordx2 141define void @dynamic_insertelement_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> %a, i32 %b) nounwind { 142 %vecins = insertelement <2 x i32> %a, i32 5, i32 %b 143 store <2 x i32> %vecins, <2 x i32> addrspace(1)* %out, align 8 144 ret void 145} 146 147; GCN-LABEL: {{^}}dynamic_insertelement_v3i32: 148; GCN: v_mov_b32_e32 [[CONST:v[0-9]+]], 5 149; GCN: v_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], [[CONST]] 150; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 151; GCN-DAG: buffer_store_dword v 152define void @dynamic_insertelement_v3i32(<3 x i32> addrspace(1)* %out, <3 x i32> %a, i32 %b) nounwind { 153 %vecins = insertelement <3 x i32> %a, i32 5, i32 %b 154 store <3 x i32> %vecins, <3 x i32> addrspace(1)* %out, align 16 155 ret void 156} 157 158; GCN-LABEL: {{^}}dynamic_insertelement_v4i32: 159; GCN: v_movreld_b32 160; GCN: buffer_store_dwordx4 161define void @dynamic_insertelement_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %a, i32 %b) nounwind { 162 %vecins = insertelement <4 x i32> %a, i32 5, i32 %b 163 store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16 164 ret void 165} 166 167; GCN-LABEL: {{^}}dynamic_insertelement_v8i32: 168; GCN: v_movreld_b32 169; GCN: buffer_store_dwordx4 170; GCN: buffer_store_dwordx4 171define void @dynamic_insertelement_v8i32(<8 x i32> addrspace(1)* %out, <8 x i32> %a, i32 %b) nounwind { 172 %vecins = insertelement <8 x i32> %a, i32 5, i32 %b 173 store <8 x i32> %vecins, <8 x i32> addrspace(1)* %out, align 32 174 ret void 175} 176 177; GCN-LABEL: {{^}}dynamic_insertelement_v16i32: 178; GCN: v_movreld_b32 179; GCN: buffer_store_dwordx4 180; GCN: buffer_store_dwordx4 181; GCN: buffer_store_dwordx4 182; GCN: buffer_store_dwordx4 183define void @dynamic_insertelement_v16i32(<16 x i32> addrspace(1)* %out, <16 x i32> %a, i32 %b) nounwind { 184 %vecins = insertelement <16 x i32> %a, i32 5, i32 %b 185 store <16 x i32> %vecins, <16 x i32> addrspace(1)* %out, align 64 186 ret void 187} 188 189; GCN-LABEL: {{^}}dynamic_insertelement_v2i16: 190define void @dynamic_insertelement_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> %a, i32 %b) nounwind { 191 %vecins = insertelement <2 x i16> %a, i16 5, i32 %b 192 store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out, align 8 193 ret void 194} 195 196; GCN-LABEL: {{^}}dynamic_insertelement_v3i16: 197define void @dynamic_insertelement_v3i16(<3 x i16> addrspace(1)* %out, <3 x i16> %a, i32 %b) nounwind { 198 %vecins = insertelement <3 x i16> %a, i16 5, i32 %b 199 store <3 x i16> %vecins, <3 x i16> addrspace(1)* %out, align 8 200 ret void 201} 202 203; GCN-LABEL: {{^}}dynamic_insertelement_v4i16: 204; GCN: buffer_load_ushort v{{[0-9]+}}, off 205; GCN: buffer_load_ushort v{{[0-9]+}}, off 206; GCN: buffer_load_ushort v{{[0-9]+}}, off 207; GCN: buffer_load_ushort v{{[0-9]+}}, off 208 209; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:6 210; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:4 211; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 212; GCN-DAG: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 213; GCN: buffer_store_short v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 214 215; GCN: buffer_load_ushort 216; GCN: buffer_load_ushort 217; GCN: buffer_load_ushort 218; GCN: buffer_load_ushort 219 220; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, off 221define void @dynamic_insertelement_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> %a, i32 %b) nounwind { 222 %vecins = insertelement <4 x i16> %a, i16 5, i32 %b 223 store <4 x i16> %vecins, <4 x i16> addrspace(1)* %out, align 8 224 ret void 225} 226 227; GCN-LABEL: {{^}}dynamic_insertelement_v2i8: 228; GCN: buffer_load_ubyte v{{[0-9]+}}, off 229; GCN: buffer_load_ubyte v{{[0-9]+}}, off 230 231; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 232; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 233 234; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 235 236; GCN: buffer_load_ubyte 237; GCN: buffer_load_ubyte 238 239; GCN: buffer_store_short v{{[0-9]+}}, off 240define void @dynamic_insertelement_v2i8(<2 x i8> addrspace(1)* %out, <2 x i8> %a, i32 %b) nounwind { 241 %vecins = insertelement <2 x i8> %a, i8 5, i32 %b 242 store <2 x i8> %vecins, <2 x i8> addrspace(1)* %out, align 8 243 ret void 244} 245 246; GCN-LABEL: {{^}}dynamic_insertelement_v3i8: 247; GCN: buffer_load_ubyte v{{[0-9]+}}, off 248; GCN: buffer_load_ubyte v{{[0-9]+}}, off 249; GCN: buffer_load_ubyte v{{[0-9]+}}, off 250 251; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 252; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 253; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 254 255; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 256 257; GCN: buffer_load_ubyte 258; GCN: buffer_load_ubyte 259; GCN: buffer_load_ubyte 260 261; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off 262; GCN-DAG: buffer_store_short v{{[0-9]+}}, off 263define void @dynamic_insertelement_v3i8(<3 x i8> addrspace(1)* %out, <3 x i8> %a, i32 %b) nounwind { 264 %vecins = insertelement <3 x i8> %a, i8 5, i32 %b 265 store <3 x i8> %vecins, <3 x i8> addrspace(1)* %out, align 4 266 ret void 267} 268 269; GCN-LABEL: {{^}}dynamic_insertelement_v4i8: 270; GCN: buffer_load_ubyte v{{[0-9]+}}, off 271; GCN: buffer_load_ubyte v{{[0-9]+}}, off 272; GCN: buffer_load_ubyte v{{[0-9]+}}, off 273; GCN: buffer_load_ubyte v{{[0-9]+}}, off 274 275; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:3 276; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:2 277; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen offset:1 278; GCN-DAG: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 279 280; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 281 282; GCN: buffer_load_ubyte 283; GCN: buffer_load_ubyte 284; GCN: buffer_load_ubyte 285; GCN: buffer_load_ubyte 286 287; GCN: buffer_store_dword v{{[0-9]+}}, off 288define void @dynamic_insertelement_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> %a, i32 %b) nounwind { 289 %vecins = insertelement <4 x i8> %a, i8 5, i32 %b 290 store <4 x i8> %vecins, <4 x i8> addrspace(1)* %out, align 4 291 ret void 292} 293 294; GCN-LABEL: {{^}}dynamic_insertelement_v8i8: 295define void @dynamic_insertelement_v8i8(<8 x i8> addrspace(1)* %out, <8 x i8> %a, i32 %b) nounwind { 296 %vecins = insertelement <8 x i8> %a, i8 5, i32 %b 297 store <8 x i8> %vecins, <8 x i8> addrspace(1)* %out, align 8 298 ret void 299} 300 301; GCN-LABEL: {{^}}dynamic_insertelement_v16i8: 302define void @dynamic_insertelement_v16i8(<16 x i8> addrspace(1)* %out, <16 x i8> %a, i32 %b) nounwind { 303 %vecins = insertelement <16 x i8> %a, i8 5, i32 %b 304 store <16 x i8> %vecins, <16 x i8> addrspace(1)* %out, align 16 305 ret void 306} 307 308; This test requires handling INSERT_SUBREG in SIFixSGPRCopies. Check that 309; the compiler doesn't crash. 310; GCN-LABEL: {{^}}insert_split_bb: 311define void @insert_split_bb(<2 x i32> addrspace(1)* %out, i32 addrspace(1)* %in, i32 %a, i32 %b) { 312entry: 313 %0 = insertelement <2 x i32> undef, i32 %a, i32 0 314 %1 = icmp eq i32 %a, 0 315 br i1 %1, label %if, label %else 316 317if: 318 %2 = load i32, i32 addrspace(1)* %in 319 %3 = insertelement <2 x i32> %0, i32 %2, i32 1 320 br label %endif 321 322else: 323 %4 = getelementptr i32, i32 addrspace(1)* %in, i32 1 324 %5 = load i32, i32 addrspace(1)* %4 325 %6 = insertelement <2 x i32> %0, i32 %5, i32 1 326 br label %endif 327 328endif: 329 %7 = phi <2 x i32> [%3, %if], [%6, %else] 330 store <2 x i32> %7, <2 x i32> addrspace(1)* %out 331 ret void 332} 333 334; GCN-LABEL: {{^}}dynamic_insertelement_v2f64: 335; GCN: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x11|0x44}}{{$}} 336; GCN-DAG: s_lshl_b32 [[SCALEDIDX:s[0-9]+]], [[IDX]], 1{{$}} 337; GCN-DAG: v_mov_b32_e32 [[ELT0:v[0-9]+]], 0{{$}} 338 339; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 340; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 341; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 342; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 343 344; GCN: s_mov_b32 m0, [[SCALEDIDX]] 345; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT0]] 346 347; Increment to next element. 348; FIXME: Should be able to manipulate m0 directly instead of add and 349; copy. 350 351; FIXME: Should avoid resetting m0 to same value 352; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0x40200000 353; GCN-DAG: s_mov_b32 m0, [[SCALEDIDX]] 354; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]] 355 356; GCN: buffer_store_dwordx4 357; GCN: s_endpgm 358define void @dynamic_insertelement_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %a, i32 %b) nounwind { 359 %vecins = insertelement <2 x double> %a, double 8.0, i32 %b 360 store <2 x double> %vecins, <2 x double> addrspace(1)* %out, align 16 361 ret void 362} 363 364; FIXME: Inline immediate should be folded into v_movreld_b32. 365; GCN-LABEL: {{^}}dynamic_insertelement_v2i64: 366 367; GCN-DAG: v_mov_b32_e32 [[ELT0:v[0-9]+]], 5{{$}} 368; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0{{$}} 369 370; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT0]] 371; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]] 372 373; GCN: buffer_store_dwordx4 374; GCN: s_endpgm 375define void @dynamic_insertelement_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> %a, i32 %b) nounwind { 376 %vecins = insertelement <2 x i64> %a, i64 5, i32 %b 377 store <2 x i64> %vecins, <2 x i64> addrspace(1)* %out, align 8 378 ret void 379} 380 381; GCN-LABEL: {{^}}dynamic_insertelement_v3i64: 382define void @dynamic_insertelement_v3i64(<3 x i64> addrspace(1)* %out, <3 x i64> %a, i32 %b) nounwind { 383 %vecins = insertelement <3 x i64> %a, i64 5, i32 %b 384 store <3 x i64> %vecins, <3 x i64> addrspace(1)* %out, align 32 385 ret void 386} 387 388; FIXME: Should be able to do without stack access. The used stack 389; space is also 2x what should be required. 390 391; GCN-LABEL: {{^}}dynamic_insertelement_v4f64: 392; GCN: SCRATCH_RSRC_DWORD 393 394; Stack store 395 396; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 397; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 398 399; Write element 400; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 401 402; Stack reload 403; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 404; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 405 406; Store result 407; GCN: buffer_store_dwordx4 408; GCN: buffer_store_dwordx4 409; GCN: s_endpgm 410; GCN: ScratchSize: 64 411 412define void @dynamic_insertelement_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %a, i32 %b) nounwind { 413 %vecins = insertelement <4 x double> %a, double 8.0, i32 %b 414 store <4 x double> %vecins, <4 x double> addrspace(1)* %out, align 16 415 ret void 416} 417 418; GCN-LABEL: {{^}}dynamic_insertelement_v8f64: 419; GCN: SCRATCH_RSRC_DWORD 420 421; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 422; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 423; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:32{{$}} 424; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:48{{$}} 425 426; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 427 428; 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; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen offset:16{{$}} 431; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 432 433; GCN: buffer_store_dwordx4 434; GCN: buffer_store_dwordx4 435; GCN: buffer_store_dwordx4 436; GCN: buffer_store_dwordx4 437; GCN: s_endpgm 438; GCN: ScratchSize: 128 439define void @dynamic_insertelement_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %a, i32 %b) nounwind { 440 %vecins = insertelement <8 x double> %a, double 8.0, i32 %b 441 store <8 x double> %vecins, <8 x double> addrspace(1)* %out, align 16 442 ret void 443} 444 445declare <4 x float> @llvm.SI.gather4.lz.v2i32(<2 x i32>, <8 x i32>, <4 x i32>, i32, i32, i32, i32, i32, i32, i32, i32) nounwind readnone 446