1;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck -check-prefix=VERDE %s 2;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s 3 4;CHECK-LABEL: {{^}}buffer_store: 5;CHECK-NOT: s_waitcnt 6;CHECK: buffer_store_dwordx4 v[0:3], off, s[0:3], 0 7;CHECK: buffer_store_dwordx4 v[4:7], off, s[0:3], 0 glc 8;CHECK: buffer_store_dwordx4 v[8:11], off, s[0:3], 0 slc 9define amdgpu_ps void @buffer_store(<4 x i32> inreg, <4 x float>, <4 x float>, <4 x float>) { 10main_body: 11 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 0, i32 0, i1 0, i1 0) 12 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %2, <4 x i32> %0, i32 0, i32 0, i1 1, i1 0) 13 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %3, <4 x i32> %0, i32 0, i32 0, i1 0, i1 1) 14 ret void 15} 16 17;CHECK-LABEL: {{^}}buffer_store_immoffs: 18;CHECK-NOT: s_waitcnt 19;CHECK: buffer_store_dwordx4 v[0:3], off, s[0:3], 0 offset:42 20define amdgpu_ps void @buffer_store_immoffs(<4 x i32> inreg, <4 x float>) { 21main_body: 22 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 0, i32 42, i1 0, i1 0) 23 ret void 24} 25 26;CHECK-LABEL: {{^}}buffer_store_idx: 27;CHECK-NOT: s_waitcnt 28;CHECK: buffer_store_dwordx4 v[0:3], v4, s[0:3], 0 idxen 29define amdgpu_ps void @buffer_store_idx(<4 x i32> inreg, <4 x float>, i32) { 30main_body: 31 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 %2, i32 0, i1 0, i1 0) 32 ret void 33} 34 35;CHECK-LABEL: {{^}}buffer_store_ofs: 36;CHECK-NOT: s_waitcnt 37;CHECK: buffer_store_dwordx4 v[0:3], v4, s[0:3], 0 offen 38define amdgpu_ps void @buffer_store_ofs(<4 x i32> inreg, <4 x float>, i32) { 39main_body: 40 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 0, i32 %2, i1 0, i1 0) 41 ret void 42} 43 44;CHECK-LABEL: {{^}}buffer_store_both: 45;CHECK-NOT: s_waitcnt 46;CHECK: buffer_store_dwordx4 v[0:3], v[4:5], s[0:3], 0 idxen offen 47define amdgpu_ps void @buffer_store_both(<4 x i32> inreg, <4 x float>, i32, i32) { 48main_body: 49 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 %2, i32 %3, i1 0, i1 0) 50 ret void 51} 52 53;CHECK-LABEL: {{^}}buffer_store_both_reversed: 54;CHECK: v_mov_b32_e32 v6, v4 55;CHECK-NOT: s_waitcnt 56;CHECK: buffer_store_dwordx4 v[0:3], v[5:6], s[0:3], 0 idxen offen 57define amdgpu_ps void @buffer_store_both_reversed(<4 x i32> inreg, <4 x float>, i32, i32) { 58main_body: 59 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 %3, i32 %2, i1 0, i1 0) 60 ret void 61} 62 63; Ideally, the register allocator would avoid the wait here 64; 65;CHECK-LABEL: {{^}}buffer_store_wait: 66;CHECK-NOT: s_waitcnt 67;CHECK: buffer_store_dwordx4 v[0:3], v4, s[0:3], 0 idxen 68;VERDE: s_waitcnt expcnt(0) 69;CHECK: buffer_load_dwordx4 v[0:3], v5, s[0:3], 0 idxen 70;CHECK: s_waitcnt vmcnt(0) 71;CHECK: buffer_store_dwordx4 v[0:3], v6, s[0:3], 0 idxen 72define amdgpu_ps void @buffer_store_wait(<4 x i32> inreg, <4 x float>, i32, i32, i32) { 73main_body: 74 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %1, <4 x i32> %0, i32 %2, i32 0, i1 0, i1 0) 75 %data = call <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32> %0, i32 %3, i32 0, i1 0, i1 0) 76 call void @llvm.amdgcn.buffer.store.v4f32(<4 x float> %data, <4 x i32> %0, i32 %4, i32 0, i1 0, i1 0) 77 ret void 78} 79 80;CHECK-LABEL: {{^}}buffer_store_x1: 81;CHECK-NOT: s_waitcnt 82;CHECK: buffer_store_dword v0, v1, s[0:3], 0 idxen 83define amdgpu_ps void @buffer_store_x1(<4 x i32> inreg %rsrc, float %data, i32 %index) { 84main_body: 85 call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0) 86 ret void 87} 88 89;CHECK-LABEL: {{^}}buffer_store_x2: 90;CHECK-NOT: s_waitcnt 91;CHECK: buffer_store_dwordx2 v[0:1], v2, s[0:3], 0 idxen 92define amdgpu_ps void @buffer_store_x2(<4 x i32> inreg %rsrc, <2 x float> %data, i32 %index) #0 { 93main_body: 94 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0) 95 ret void 96} 97 98;CHECK-LABEL: {{^}}buffer_store_x1_offen_merged: 99;CHECK-NOT: s_waitcnt 100;CHECK-DAG: buffer_store_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4 101;CHECK-DAG: buffer_store_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:28 102define amdgpu_ps void @buffer_store_x1_offen_merged(<4 x i32> inreg %rsrc, i32 %a, float %v1, float %v2, float %v3, float %v4, float %v5, float %v6) { 103 %a1 = add i32 %a, 4 104 %a2 = add i32 %a, 8 105 %a3 = add i32 %a, 12 106 %a4 = add i32 %a, 16 107 %a5 = add i32 %a, 28 108 %a6 = add i32 %a, 32 109 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 110 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 111 call void @llvm.amdgcn.buffer.store.f32(float %v3, <4 x i32> %rsrc, i32 0, i32 %a3, i1 0, i1 0) 112 call void @llvm.amdgcn.buffer.store.f32(float %v4, <4 x i32> %rsrc, i32 0, i32 %a4, i1 0, i1 0) 113 call void @llvm.amdgcn.buffer.store.f32(float %v5, <4 x i32> %rsrc, i32 0, i32 %a5, i1 0, i1 0) 114 call void @llvm.amdgcn.buffer.store.f32(float %v6, <4 x i32> %rsrc, i32 0, i32 %a6, i1 0, i1 0) 115 ret void 116} 117 118;CHECK-LABEL: {{^}}buffer_store_x1_offen_merged_glc_slc: 119;CHECK-NOT: s_waitcnt 120;CHECK-DAG: buffer_store_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4{{$}} 121;CHECK-DAG: buffer_store_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:12 glc{{$}} 122;CHECK-DAG: buffer_store_dwordx2 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:28 glc slc{{$}} 123define amdgpu_ps void @buffer_store_x1_offen_merged_glc_slc(<4 x i32> inreg %rsrc, i32 %a, float %v1, float %v2, float %v3, float %v4, float %v5, float %v6) { 124 %a1 = add i32 %a, 4 125 %a2 = add i32 %a, 8 126 %a3 = add i32 %a, 12 127 %a4 = add i32 %a, 16 128 %a5 = add i32 %a, 28 129 %a6 = add i32 %a, 32 130 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 131 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 132 call void @llvm.amdgcn.buffer.store.f32(float %v3, <4 x i32> %rsrc, i32 0, i32 %a3, i1 1, i1 0) 133 call void @llvm.amdgcn.buffer.store.f32(float %v4, <4 x i32> %rsrc, i32 0, i32 %a4, i1 1, i1 0) 134 call void @llvm.amdgcn.buffer.store.f32(float %v5, <4 x i32> %rsrc, i32 0, i32 %a5, i1 1, i1 1) 135 call void @llvm.amdgcn.buffer.store.f32(float %v6, <4 x i32> %rsrc, i32 0, i32 %a6, i1 1, i1 1) 136 ret void 137} 138 139;CHECK-LABEL: {{^}}buffer_store_x2_offen_merged: 140;CHECK-NOT: s_waitcnt 141;CHECK: buffer_store_dwordx4 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4 142define amdgpu_ps void @buffer_store_x2_offen_merged(<4 x i32> inreg %rsrc, i32 %a, <2 x float> %v1, <2 x float> %v2) { 143 %a1 = add i32 %a, 4 144 %a2 = add i32 %a, 12 145 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 146 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 147 ret void 148} 149 150;CHECK-LABEL: {{^}}buffer_store_x3_offen_merged: 151;CHECK-NOT: s_waitcnt 152;CHECK: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:28 153define amdgpu_ps void @buffer_store_x3_offen_merged(<4 x i32> inreg %rsrc, i32 %a, float %v1, float %v2, float %v3) { 154 %a1 = add i32 %a, 28 155 %a2 = add i32 %a, 32 156 %a3 = add i32 %a, 36 157 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 158 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 159 call void @llvm.amdgcn.buffer.store.f32(float %v3, <4 x i32> %rsrc, i32 0, i32 %a3, i1 0, i1 0) 160 ret void 161} 162 163;CHECK-LABEL: {{^}}buffer_store_x3_offen_merged2: 164;CHECK-NOT: s_waitcnt 165;CHECK: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4 166define amdgpu_ps void @buffer_store_x3_offen_merged2(<4 x i32> inreg %rsrc, i32 %a, <2 x float> %v1, float %v2) { 167 %a1 = add i32 %a, 4 168 %a2 = add i32 %a, 12 169 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 170 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 171 ret void 172} 173 174;CHECK-LABEL: {{^}}buffer_store_x3_offen_merged3: 175;CHECK-NOT: s_waitcnt 176;CHECK: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], v0, s[0:3], 0 offen offset:4 177define amdgpu_ps void @buffer_store_x3_offen_merged3(<4 x i32> inreg %rsrc, i32 %a, float %v1, <2 x float> %v2) { 178 %a1 = add i32 %a, 4 179 %a2 = add i32 %a, 8 180 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 %a1, i1 0, i1 0) 181 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v2, <4 x i32> %rsrc, i32 0, i32 %a2, i1 0, i1 0) 182 ret void 183} 184 185;CHECK-LABEL: {{^}}buffer_store_x1_offset_merged: 186;CHECK-NOT: s_waitcnt 187;CHECK-DAG: buffer_store_dwordx4 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4 188;CHECK-DAG: buffer_store_dwordx2 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:28 189define amdgpu_ps void @buffer_store_x1_offset_merged(<4 x i32> inreg %rsrc, float %v1, float %v2, float %v3, float %v4, float %v5, float %v6) { 190 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 4, i1 0, i1 0) 191 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 8, i1 0, i1 0) 192 call void @llvm.amdgcn.buffer.store.f32(float %v3, <4 x i32> %rsrc, i32 0, i32 12, i1 0, i1 0) 193 call void @llvm.amdgcn.buffer.store.f32(float %v4, <4 x i32> %rsrc, i32 0, i32 16, i1 0, i1 0) 194 call void @llvm.amdgcn.buffer.store.f32(float %v5, <4 x i32> %rsrc, i32 0, i32 28, i1 0, i1 0) 195 call void @llvm.amdgcn.buffer.store.f32(float %v6, <4 x i32> %rsrc, i32 0, i32 32, i1 0, i1 0) 196 ret void 197} 198 199;CHECK-LABEL: {{^}}buffer_store_x2_offset_merged: 200;CHECK-NOT: s_waitcnt 201;CHECK: buffer_store_dwordx4 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4 202define amdgpu_ps void @buffer_store_x2_offset_merged(<4 x i32> inreg %rsrc, <2 x float> %v1, <2 x float> %v2) { 203 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v1, <4 x i32> %rsrc, i32 0, i32 4, i1 0, i1 0) 204 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v2, <4 x i32> %rsrc, i32 0, i32 12, i1 0, i1 0) 205 ret void 206} 207 208;CHECK-LABEL: {{^}}buffer_store_x3_offset_merged: 209;CHECK-NOT: s_waitcnt 210;CHECK-DAG: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4 211define amdgpu_ps void @buffer_store_x3_offset_merged(<4 x i32> inreg %rsrc, float %v1, float %v2, float %v3) { 212 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 4, i1 0, i1 0) 213 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 8, i1 0, i1 0) 214 call void @llvm.amdgcn.buffer.store.f32(float %v3, <4 x i32> %rsrc, i32 0, i32 12, i1 0, i1 0) 215 ret void 216} 217 218;CHECK-LABEL: {{^}}buffer_store_x3_offset_merged2: 219;CHECK-NOT: s_waitcnt 220;CHECK-DAG: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:4 221define amdgpu_ps void @buffer_store_x3_offset_merged2(<4 x i32> inreg %rsrc, float %v1, <2 x float> %v2) { 222 call void @llvm.amdgcn.buffer.store.f32(float %v1, <4 x i32> %rsrc, i32 0, i32 4, i1 0, i1 0) 223 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v2, <4 x i32> %rsrc, i32 0, i32 8, i1 0, i1 0) 224 ret void 225} 226 227;CHECK-LABEL: {{^}}buffer_store_x3_offset_merged3: 228;CHECK-NOT: s_waitcnt 229;CHECK-DAG: buffer_store_dwordx3 v[{{[0-9]}}:{{[0-9]}}], off, s[0:3], 0 offset:8 230define amdgpu_ps void @buffer_store_x3_offset_merged3(<4 x i32> inreg %rsrc, <2 x float> %v1, float %v2) { 231 call void @llvm.amdgcn.buffer.store.v2f32(<2 x float> %v1, <4 x i32> %rsrc, i32 0, i32 8, i1 0, i1 0) 232 call void @llvm.amdgcn.buffer.store.f32(float %v2, <4 x i32> %rsrc, i32 0, i32 16, i1 0, i1 0) 233 ret void 234} 235 236;CHECK-LABEL: {{^}}buffer_store_byte: 237;CHECK-NOT: s_waitcnt 238;CHECK-NEXT: %bb. 239;CHECK: buffer_store_byte v{{[0-9]}}, off, s[0:3], 0 offset:8 240define amdgpu_ps void @buffer_store_byte(<4 x i32> inreg %rsrc, float %v1) { 241main_body: 242 %v2 = fptoui float %v1 to i32 243 %v3 = trunc i32 %v2 to i8 244 call void @llvm.amdgcn.buffer.store.i8(i8 %v3, <4 x i32> %rsrc, i32 0, i32 8, i1 0, i1 0) 245 ret void 246} 247 248;CHECK-LABEL: {{^}}buffer_store_short: 249;CHECK-NOT: s_waitcnt 250;CHECK-NEXT: %bb. 251;CHECK: buffer_store_short v{{[0-9]}}, off, s[0:3], 0 offset:16 252define amdgpu_ps void @buffer_store_short(<4 x i32> inreg %rsrc, float %v1) { 253main_body: 254 %v2 = fptoui float %v1 to i32 255 %v3 = trunc i32 %v2 to i16 256 call void @llvm.amdgcn.buffer.store.i16(i16 %v3, <4 x i32> %rsrc, i32 0, i32 16, i1 0, i1 0) 257 ret void 258} 259 260declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1) #0 261declare void @llvm.amdgcn.buffer.store.v2f32(<2 x float>, <4 x i32>, i32, i32, i1, i1) #0 262declare void @llvm.amdgcn.buffer.store.v4f32(<4 x float>, <4 x i32>, i32, i32, i1, i1) #0 263declare void @llvm.amdgcn.buffer.store.i8(i8, <4 x i32>, i32, i32, i1, i1) #0 264declare void @llvm.amdgcn.buffer.store.i16(i16, <4 x i32>, i32, i32, i1, i1) #0 265declare <4 x float> @llvm.amdgcn.buffer.load.v4f32(<4 x i32>, i32, i32, i1, i1) #1 266 267attributes #0 = { nounwind } 268attributes #1 = { nounwind readonly } 269