1; RUN: llc -verify-machineinstrs -march=amdgcn -mattr=+max-private-element-size-16 < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=GCN-NO-TONGA %s 2; RUN: llc -verify-machineinstrs -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -mattr=+max-private-element-size-16 < %s | FileCheck -check-prefix=GCN -check-prefix=SI -check-prefix=GCN-TONGA %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: s_mov_b32 [[CONSTREG:s[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_movreld_b32_e32 v[[LOW_RESULT_REG:[0-9]+]], 5 149; GCN-DAG: buffer_store_dwordx2 {{v\[}}[[LOW_RESULT_REG]]: 150; GCN-DAG: buffer_store_dword v 151define void @dynamic_insertelement_v3i32(<3 x i32> addrspace(1)* %out, <3 x i32> %a, i32 %b) nounwind { 152 %vecins = insertelement <3 x i32> %a, i32 5, i32 %b 153 store <3 x i32> %vecins, <3 x i32> addrspace(1)* %out, align 16 154 ret void 155} 156 157; GCN-LABEL: {{^}}dynamic_insertelement_v4i32: 158; GCN: s_load_dword [[SVAL:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x12|0x48}} 159; GCN: v_mov_b32_e32 [[VVAL:v[0-9]+]], [[SVAL]] 160; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[VVAL]] 161; GCN: buffer_store_dwordx4 162define void @dynamic_insertelement_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> %a, i32 %b, i32 %val) nounwind { 163 %vecins = insertelement <4 x i32> %a, i32 %val, i32 %b 164 store <4 x i32> %vecins, <4 x i32> addrspace(1)* %out, align 16 165 ret void 166} 167 168; GCN-LABEL: {{^}}dynamic_insertelement_v8i32: 169; GCN: v_movreld_b32 170; GCN: buffer_store_dwordx4 171; GCN: buffer_store_dwordx4 172define void @dynamic_insertelement_v8i32(<8 x i32> addrspace(1)* %out, <8 x i32> %a, i32 %b) nounwind { 173 %vecins = insertelement <8 x i32> %a, i32 5, i32 %b 174 store <8 x i32> %vecins, <8 x i32> addrspace(1)* %out, align 32 175 ret void 176} 177 178; GCN-LABEL: {{^}}dynamic_insertelement_v16i32: 179; GCN: v_movreld_b32 180; GCN: buffer_store_dwordx4 181; GCN: buffer_store_dwordx4 182; GCN: buffer_store_dwordx4 183; GCN: buffer_store_dwordx4 184define void @dynamic_insertelement_v16i32(<16 x i32> addrspace(1)* %out, <16 x i32> %a, i32 %b) nounwind { 185 %vecins = insertelement <16 x i32> %a, i32 5, i32 %b 186 store <16 x i32> %vecins, <16 x i32> addrspace(1)* %out, align 64 187 ret void 188} 189 190; GCN-LABEL: {{^}}dynamic_insertelement_v2i16: 191define void @dynamic_insertelement_v2i16(<2 x i16> addrspace(1)* %out, <2 x i16> %a, i32 %b) nounwind { 192 %vecins = insertelement <2 x i16> %a, i16 5, i32 %b 193 store <2 x i16> %vecins, <2 x i16> addrspace(1)* %out, align 8 194 ret void 195} 196 197; GCN-LABEL: {{^}}dynamic_insertelement_v3i16: 198define void @dynamic_insertelement_v3i16(<3 x i16> addrspace(1)* %out, <3 x i16> %a, i32 %b) nounwind { 199 %vecins = insertelement <3 x i16> %a, i16 5, i32 %b 200 store <3 x i16> %vecins, <3 x i16> addrspace(1)* %out, align 8 201 ret void 202} 203 204; GCN-LABEL: {{^}}dynamic_insertelement_v4i16: 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; GCN: buffer_load_ushort v{{[0-9]+}}, off 209 210; GCN-DAG: v_mov_b32_e32 [[BASE_FI:v[0-9]+]], 8{{$}} 211; GCN-DAG: s_and_b32 [[MASK_IDX:s[0-9]+]], s{{[0-9]+}}, 3{{$}} 212; GCN-DAG: v_or_b32_e32 [[IDX:v[0-9]+]], [[MASK_IDX]], [[BASE_FI]]{{$}} 213 214; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:14 215; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:12 216; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:10 217; GCN-DAG: buffer_store_short v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:8 218; GCN: buffer_store_short v{{[0-9]+}}, [[IDX]], s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 219 220; GCN: s_waitcnt 221 222; GCN: buffer_load_dwordx2 223 224; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, off 225define void @dynamic_insertelement_v4i16(<4 x i16> addrspace(1)* %out, <4 x i16> %a, i32 %b) nounwind { 226 %vecins = insertelement <4 x i16> %a, i16 5, i32 %b 227 store <4 x i16> %vecins, <4 x i16> addrspace(1)* %out, align 8 228 ret void 229} 230 231; GCN-LABEL: {{^}}dynamic_insertelement_v2i8: 232; GCN: buffer_load_ubyte v{{[0-9]+}}, off 233; GCN: buffer_load_ubyte v{{[0-9]+}}, off 234 235; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:5 236; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:4 237 238; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 239 240; GCN-NO-TONGA: buffer_load_ubyte 241; GCN-NO-TONGA: buffer_load_ubyte 242; GCN-TONGA: buffer_load_ushort 243 244; GCN: buffer_store_short v{{[0-9]+}}, off 245define void @dynamic_insertelement_v2i8(<2 x i8> addrspace(1)* %out, <2 x i8> %a, i32 %b) nounwind { 246 %vecins = insertelement <2 x i8> %a, i8 5, i32 %b 247 store <2 x i8> %vecins, <2 x i8> addrspace(1)* %out, align 8 248 ret void 249} 250 251; GCN-LABEL: {{^}}dynamic_insertelement_v3i8: 252; GCN: buffer_load_ubyte v{{[0-9]+}}, off 253; GCN: buffer_load_ubyte v{{[0-9]+}}, off 254; GCN: buffer_load_ubyte v{{[0-9]+}}, off 255 256; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:6 257; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:5 258; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:4 259 260; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 261 262; GCN-NO-TONGA: buffer_load_ubyte 263; GCN-NO-TONGA: buffer_load_ubyte 264; GCN-NO-TONGA: buffer_load_ubyte 265; GCN-TONGA: buffer_load_ushort 266; GCN-TONGA: buffer_load_ubyte 267 268; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off 269; GCN-DAG: buffer_store_short v{{[0-9]+}}, off 270define void @dynamic_insertelement_v3i8(<3 x i8> addrspace(1)* %out, <3 x i8> %a, i32 %b) nounwind { 271 %vecins = insertelement <3 x i8> %a, i8 5, i32 %b 272 store <3 x i8> %vecins, <3 x i8> addrspace(1)* %out, align 4 273 ret void 274} 275 276; GCN-LABEL: {{^}}dynamic_insertelement_v4i8: 277; GCN: buffer_load_ubyte v{{[0-9]+}}, off 278; GCN: buffer_load_ubyte v{{[0-9]+}}, off 279; GCN: buffer_load_ubyte v{{[0-9]+}}, off 280; GCN: buffer_load_ubyte v{{[0-9]+}}, off 281 282; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:7 283; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:6 284; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:5 285; GCN-DAG: buffer_store_byte v{{[0-9]+}}, off, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offset:4 286 287; GCN: buffer_store_byte v{{[0-9]+}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, s{{[0-9]+}} offen{{$}} 288 289; GCN-NO-TONGA: buffer_load_ubyte 290; GCN-NO-TONGA: buffer_load_ubyte 291; GCN-NO-TONGA: buffer_load_ubyte 292; GCN-NO-TONGA: buffer_load_ubyte 293; GCN-TONGA: buffer_load_dword 294 295; GCN: buffer_store_dword v{{[0-9]+}}, off 296define void @dynamic_insertelement_v4i8(<4 x i8> addrspace(1)* %out, <4 x i8> %a, i32 %b) nounwind { 297 %vecins = insertelement <4 x i8> %a, i8 5, i32 %b 298 store <4 x i8> %vecins, <4 x i8> addrspace(1)* %out, align 4 299 ret void 300} 301 302; GCN-LABEL: {{^}}dynamic_insertelement_v8i8: 303define void @dynamic_insertelement_v8i8(<8 x i8> addrspace(1)* %out, <8 x i8> %a, i32 %b) nounwind { 304 %vecins = insertelement <8 x i8> %a, i8 5, i32 %b 305 store <8 x i8> %vecins, <8 x i8> addrspace(1)* %out, align 8 306 ret void 307} 308 309; GCN-LABEL: {{^}}dynamic_insertelement_v16i8: 310define void @dynamic_insertelement_v16i8(<16 x i8> addrspace(1)* %out, <16 x i8> %a, i32 %b) nounwind { 311 %vecins = insertelement <16 x i8> %a, i8 5, i32 %b 312 store <16 x i8> %vecins, <16 x i8> addrspace(1)* %out, align 16 313 ret void 314} 315 316; This test requires handling INSERT_SUBREG in SIFixSGPRCopies. Check that 317; the compiler doesn't crash. 318; GCN-LABEL: {{^}}insert_split_bb: 319define void @insert_split_bb(<2 x i32> addrspace(1)* %out, i32 addrspace(1)* %in, i32 %a, i32 %b) { 320entry: 321 %0 = insertelement <2 x i32> undef, i32 %a, i32 0 322 %1 = icmp eq i32 %a, 0 323 br i1 %1, label %if, label %else 324 325if: 326 %2 = load i32, i32 addrspace(1)* %in 327 %3 = insertelement <2 x i32> %0, i32 %2, i32 1 328 br label %endif 329 330else: 331 %4 = getelementptr i32, i32 addrspace(1)* %in, i32 1 332 %5 = load i32, i32 addrspace(1)* %4 333 %6 = insertelement <2 x i32> %0, i32 %5, i32 1 334 br label %endif 335 336endif: 337 %7 = phi <2 x i32> [%3, %if], [%6, %else] 338 store <2 x i32> %7, <2 x i32> addrspace(1)* %out 339 ret void 340} 341 342; GCN-LABEL: {{^}}dynamic_insertelement_v2f64: 343; GCN-DAG: s_load_dwordx4 s{{\[}}[[A_ELT0:[0-9]+]]:[[A_ELT3:[0-9]+]]{{\]}} 344; GCN-DAG: s_load_dword [[IDX:s[0-9]+]], s{{\[[0-9]+:[0-9]+\]}}, {{0x11|0x44}}{{$}} 345 346; GCN-DAG: s_lshl_b32 [[SCALEDIDX:s[0-9]+]], [[IDX]], 1{{$}} 347 348; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 349; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 350; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 351; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, s{{[0-9]+}} 352; GCN-DAG: v_mov_b32_e32 [[ELT1:v[0-9]+]], 0x40200000 353 354; GCN-DAG: s_mov_b32 m0, [[SCALEDIDX]] 355; GCN: v_movreld_b32_e32 v{{[0-9]+}}, 0 356 357; Increment to next element folded into base register, but FileCheck 358; can't do math expressions 359 360; FIXME: Should be able to manipulate m0 directly instead of s_lshl_b32 + copy to m0 361 362; GCN: v_movreld_b32_e32 v{{[0-9]+}}, [[ELT1]] 363 364; GCN: buffer_store_dwordx4 365; GCN: s_endpgm 366define void @dynamic_insertelement_v2f64(<2 x double> addrspace(1)* %out, <2 x double> %a, i32 %b) nounwind { 367 %vecins = insertelement <2 x double> %a, double 8.0, i32 %b 368 store <2 x double> %vecins, <2 x double> addrspace(1)* %out, align 16 369 ret void 370} 371 372; GCN-LABEL: {{^}}dynamic_insertelement_v2i64: 373 374; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, 5 375; GCN-DAG: v_movreld_b32_e32 v{{[0-9]+}}, 0 376 377; GCN: buffer_store_dwordx4 378; GCN: s_endpgm 379define void @dynamic_insertelement_v2i64(<2 x i64> addrspace(1)* %out, <2 x i64> %a, i32 %b) nounwind { 380 %vecins = insertelement <2 x i64> %a, i64 5, i32 %b 381 store <2 x i64> %vecins, <2 x i64> addrspace(1)* %out, align 8 382 ret void 383} 384 385; GCN-LABEL: {{^}}dynamic_insertelement_v3i64: 386define void @dynamic_insertelement_v3i64(<3 x i64> addrspace(1)* %out, <3 x i64> %a, i32 %b) nounwind { 387 %vecins = insertelement <3 x i64> %a, i64 5, i32 %b 388 store <3 x i64> %vecins, <3 x i64> addrspace(1)* %out, align 32 389 ret void 390} 391 392; FIXME: Should be able to do without stack access. The used stack 393; space is also 2x what should be required. 394 395; GCN-LABEL: {{^}}dynamic_insertelement_v4f64: 396; GCN: SCRATCH_RSRC_DWORD 397 398; Stack store 399 400; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:32{{$}} 401; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:48{{$}} 402 403; Write element 404; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 405 406; Stack reload 407; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:32{{$}} 408; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:48{{$}} 409 410; Store result 411; GCN: buffer_store_dwordx4 412; GCN: buffer_store_dwordx4 413; GCN: s_endpgm 414; GCN: ScratchSize: 64 415 416define void @dynamic_insertelement_v4f64(<4 x double> addrspace(1)* %out, <4 x double> %a, i32 %b) nounwind { 417 %vecins = insertelement <4 x double> %a, double 8.0, i32 %b 418 store <4 x double> %vecins, <4 x double> addrspace(1)* %out, align 16 419 ret void 420} 421 422; GCN-LABEL: {{^}}dynamic_insertelement_v8f64: 423; GCN-DAG: SCRATCH_RSRC_DWORD 424 425; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:64{{$}} 426; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:80{{$}} 427; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:96{{$}} 428; GCN-DAG: buffer_store_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:112{{$}} 429 430; GCN: buffer_store_dwordx2 v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offen{{$}} 431 432; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:64{{$}} 433; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:80{{$}} 434; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:96{{$}} 435; GCN-DAG: buffer_load_dwordx4 v{{\[[0-9]+:[0-9]+\]}}, off, s{{\[[0-9]+:[0-9]+\]}}, {{s[0-9]+}} offset:112{{$}} 436 437; GCN: buffer_store_dwordx4 438; GCN: buffer_store_dwordx4 439; GCN: buffer_store_dwordx4 440; GCN: buffer_store_dwordx4 441; GCN: s_endpgm 442; GCN: ScratchSize: 128 443define void @dynamic_insertelement_v8f64(<8 x double> addrspace(1)* %out, <8 x double> %a, i32 %b) nounwind { 444 %vecins = insertelement <8 x double> %a, double 8.0, i32 %b 445 store <8 x double> %vecins, <8 x double> addrspace(1)* %out, align 16 446 ret void 447} 448 449declare <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 450