1; RUN: llc < %s -march=amdgcn -mcpu=tahiti -verify-machineinstrs | FileCheck %s -check-prefixes=GCN,SI,SICI 2; RUN: llc < %s -march=amdgcn -mcpu=hawaii -verify-machineinstrs | FileCheck %s -check-prefixes=GCN,CI,SICI 3; RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s -check-prefixes=GCN,VI 4 5;GCN-LABEL: {{^}}s_buffer_load_imm: 6;GCN-NOT: s_waitcnt; 7;SI: s_buffer_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0x1 8;CI: s_buffer_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0x1 9;VI: s_buffer_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0x4 10define amdgpu_ps void @s_buffer_load_imm(<4 x i32> inreg %desc) { 11main_body: 12 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 4, i32 0) 13 %bitcast = bitcast i32 %load to float 14 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %bitcast, float undef, float undef, float undef, i1 true, i1 true) 15 ret void 16} 17 18;GCN-LABEL: {{^}}s_buffer_load_index: 19;GCN-NOT: s_waitcnt; 20;GCN: s_buffer_load_dword s{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} 21define amdgpu_ps void @s_buffer_load_index(<4 x i32> inreg %desc, i32 inreg %index) { 22main_body: 23 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 %index, i32 0) 24 %bitcast = bitcast i32 %load to float 25 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %bitcast, float undef, float undef, float undef, i1 true, i1 true) 26 ret void 27} 28 29;GCN-LABEL: {{^}}s_buffer_load_index_divergent: 30;GCN-NOT: s_waitcnt; 31;GCN: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 32define amdgpu_ps void @s_buffer_load_index_divergent(<4 x i32> inreg %desc, i32 %index) { 33main_body: 34 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 %index, i32 0) 35 %bitcast = bitcast i32 %load to float 36 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %bitcast, float undef, float undef, float undef, i1 true, i1 true) 37 ret void 38} 39 40;GCN-LABEL: {{^}}s_buffer_loadx2_imm: 41;GCN-NOT: s_waitcnt; 42;SI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x10 43;CI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x10 44;VI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x40 45define amdgpu_ps void @s_buffer_loadx2_imm(<4 x i32> inreg %desc) { 46main_body: 47 %load = call <2 x i32> @llvm.amdgcn.s.buffer.load.v2i32(<4 x i32> %desc, i32 64, i32 0) 48 %bitcast = bitcast <2 x i32> %load to <2 x float> 49 %x = extractelement <2 x float> %bitcast, i32 0 50 %y = extractelement <2 x float> %bitcast, i32 1 51 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float undef, float undef, i1 true, i1 true) 52 ret void 53} 54 55;GCN-LABEL: {{^}}s_buffer_loadx2_index: 56;GCN-NOT: s_waitcnt; 57;GCN: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} 58define amdgpu_ps void @s_buffer_loadx2_index(<4 x i32> inreg %desc, i32 inreg %index) { 59main_body: 60 %load = call <2 x i32> @llvm.amdgcn.s.buffer.load.v2i32(<4 x i32> %desc, i32 %index, i32 0) 61 %bitcast = bitcast <2 x i32> %load to <2 x float> 62 %x = extractelement <2 x float> %bitcast, i32 0 63 %y = extractelement <2 x float> %bitcast, i32 1 64 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float undef, float undef, i1 true, i1 true) 65 ret void 66} 67 68;GCN-LABEL: {{^}}s_buffer_loadx2_index_divergent: 69;GCN-NOT: s_waitcnt; 70;GCN: buffer_load_dwordx2 v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 71define amdgpu_ps void @s_buffer_loadx2_index_divergent(<4 x i32> inreg %desc, i32 %index) { 72main_body: 73 %load = call <2 x i32> @llvm.amdgcn.s.buffer.load.v2i32(<4 x i32> %desc, i32 %index, i32 0) 74 %bitcast = bitcast <2 x i32> %load to <2 x float> 75 %x = extractelement <2 x float> %bitcast, i32 0 76 %y = extractelement <2 x float> %bitcast, i32 1 77 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float undef, float undef, i1 true, i1 true) 78 ret void 79} 80 81;GCN-LABEL: {{^}}s_buffer_loadx3_imm: 82;GCN-NOT: s_waitcnt; 83;SI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x10 84;CI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x10 85;VI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x40 86define amdgpu_ps void @s_buffer_loadx3_imm(<4 x i32> inreg %desc) { 87main_body: 88 %load = call <3 x i32> @llvm.amdgcn.s.buffer.load.v3i32(<4 x i32> %desc, i32 64, i32 0) 89 %bitcast = bitcast <3 x i32> %load to <3 x float> 90 %x = extractelement <3 x float> %bitcast, i32 0 91 %y = extractelement <3 x float> %bitcast, i32 1 92 %z = extractelement <3 x float> %bitcast, i32 2 93 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float undef, i1 true, i1 true) 94 ret void 95} 96 97;GCN-LABEL: {{^}}s_buffer_loadx3_index: 98;GCN-NOT: s_waitcnt; 99;GCN: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} 100define amdgpu_ps void @s_buffer_loadx3_index(<4 x i32> inreg %desc, i32 inreg %index) { 101main_body: 102 %load = call <3 x i32> @llvm.amdgcn.s.buffer.load.v3i32(<4 x i32> %desc, i32 %index, i32 0) 103 %bitcast = bitcast <3 x i32> %load to <3 x float> 104 %x = extractelement <3 x float> %bitcast, i32 0 105 %y = extractelement <3 x float> %bitcast, i32 1 106 %z = extractelement <3 x float> %bitcast, i32 2 107 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float undef, i1 true, i1 true) 108 ret void 109} 110 111;GCN-LABEL: {{^}}s_buffer_loadx3_index_divergent: 112;GCN-NOT: s_waitcnt; 113;SI: buffer_load_dwordx4 v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 114;CI: buffer_load_dwordx3 v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 115;VI: buffer_load_dwordx3 v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 116define amdgpu_ps void @s_buffer_loadx3_index_divergent(<4 x i32> inreg %desc, i32 %index) { 117main_body: 118 %load = call <3 x i32> @llvm.amdgcn.s.buffer.load.v3i32(<4 x i32> %desc, i32 %index, i32 0) 119 %bitcast = bitcast <3 x i32> %load to <3 x float> 120 %x = extractelement <3 x float> %bitcast, i32 0 121 %y = extractelement <3 x float> %bitcast, i32 1 122 %z = extractelement <3 x float> %bitcast, i32 2 123 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float undef, i1 true, i1 true) 124 ret void 125} 126 127;GCN-LABEL: {{^}}s_buffer_loadx4_imm: 128;GCN-NOT: s_waitcnt; 129;SI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x32 130;CI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x32 131;VI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0xc8 132define amdgpu_ps void @s_buffer_loadx4_imm(<4 x i32> inreg %desc) { 133main_body: 134 %load = call <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32> %desc, i32 200, i32 0) 135 %bitcast = bitcast <4 x i32> %load to <4 x float> 136 %x = extractelement <4 x float> %bitcast, i32 0 137 %y = extractelement <4 x float> %bitcast, i32 1 138 %z = extractelement <4 x float> %bitcast, i32 2 139 %w = extractelement <4 x float> %bitcast, i32 3 140 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float %w, i1 true, i1 true) 141 ret void 142} 143 144;GCN-LABEL: {{^}}s_buffer_loadx4_index: 145;GCN-NOT: s_waitcnt; 146;GCN: buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} 147define amdgpu_ps void @s_buffer_loadx4_index(<4 x i32> inreg %desc, i32 inreg %index) { 148main_body: 149 %load = call <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32> %desc, i32 %index, i32 0) 150 %bitcast = bitcast <4 x i32> %load to <4 x float> 151 %x = extractelement <4 x float> %bitcast, i32 0 152 %y = extractelement <4 x float> %bitcast, i32 1 153 %z = extractelement <4 x float> %bitcast, i32 2 154 %w = extractelement <4 x float> %bitcast, i32 3 155 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float %w, i1 true, i1 true) 156 ret void 157} 158 159;GCN-LABEL: {{^}}s_buffer_loadx4_index_divergent: 160;GCN-NOT: s_waitcnt; 161;GCN: buffer_load_dwordx4 v[{{[0-9]+:[0-9]+}}], v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 162define amdgpu_ps void @s_buffer_loadx4_index_divergent(<4 x i32> inreg %desc, i32 %index) { 163main_body: 164 %load = call <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32> %desc, i32 %index, i32 0) 165 %bitcast = bitcast <4 x i32> %load to <4 x float> 166 %x = extractelement <4 x float> %bitcast, i32 0 167 %y = extractelement <4 x float> %bitcast, i32 1 168 %z = extractelement <4 x float> %bitcast, i32 2 169 %w = extractelement <4 x float> %bitcast, i32 3 170 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float %w, i1 true, i1 true) 171 ret void 172} 173 174;GCN-LABEL: {{^}}s_buffer_load_imm_mergex2: 175;GCN-NOT: s_waitcnt; 176;SI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x1 177;CI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x1 178;VI: s_buffer_load_dwordx2 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x4 179define amdgpu_ps void @s_buffer_load_imm_mergex2(<4 x i32> inreg %desc) { 180main_body: 181 %load0 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 4, i32 0) 182 %load1 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 8, i32 0) 183 %x = bitcast i32 %load0 to float 184 %y = bitcast i32 %load1 to float 185 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float undef, float undef, i1 true, i1 true) 186 ret void 187} 188 189;GCN-LABEL: {{^}}s_buffer_load_imm_mergex4: 190;GCN-NOT: s_waitcnt; 191;SI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x2 192;CI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x2 193;VI: s_buffer_load_dwordx4 s[{{[0-9]+:[0-9]+}}], s[{{[0-9]+:[0-9]+}}], 0x8 194define amdgpu_ps void @s_buffer_load_imm_mergex4(<4 x i32> inreg %desc) { 195main_body: 196 %load0 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 8, i32 0) 197 %load1 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 12, i32 0) 198 %load2 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 16, i32 0) 199 %load3 = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 20, i32 0) 200 %x = bitcast i32 %load0 to float 201 %y = bitcast i32 %load1 to float 202 %z = bitcast i32 %load2 to float 203 %w = bitcast i32 %load3 to float 204 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %x, float %y, float %z, float %w, i1 true, i1 true) 205 ret void 206} 207 208;GCN-LABEL: {{^}}s_buffer_load_index_across_bb: 209;GCN-NOT: s_waitcnt; 210;GCN: v_or_b32 211;GCN: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 212define amdgpu_ps void @s_buffer_load_index_across_bb(<4 x i32> inreg %desc, i32 %index) { 213main_body: 214 %tmp = shl i32 %index, 4 215 br label %bb1 216 217bb1: ; preds = %main_body 218 %tmp1 = or i32 %tmp, 8 219 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 %tmp1, i32 0) 220 %bitcast = bitcast i32 %load to float 221 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %bitcast, float undef, float undef, float undef, i1 true, i1 true) 222 ret void 223} 224 225;GCN-LABEL: {{^}}s_buffer_load_index_across_bb_merged: 226;GCN-NOT: s_waitcnt; 227;GCN: v_or_b32 228;GCN: v_or_b32 229;GCN: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 230;GCN: buffer_load_dword v{{[0-9]+}}, v{{[0-9]+}}, s[{{[0-9]+:[0-9]+}}], 0 offen 231define amdgpu_ps void @s_buffer_load_index_across_bb_merged(<4 x i32> inreg %desc, i32 %index) { 232main_body: 233 %tmp = shl i32 %index, 4 234 br label %bb1 235 236bb1: ; preds = %main_body 237 %tmp1 = or i32 %tmp, 8 238 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 %tmp1, i32 0) 239 %tmp2 = or i32 %tmp1, 4 240 %load2 = tail call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 %tmp2, i32 0) 241 %bitcast = bitcast i32 %load to float 242 %bitcast2 = bitcast i32 %load2 to float 243 call void @llvm.amdgcn.exp.f32(i32 0, i32 15, float %bitcast, float %bitcast2, float undef, float undef, i1 true, i1 true) 244 ret void 245} 246 247; GCN-LABEL: {{^}}s_buffer_load_imm_neg1: 248; GCN: s_mov_b32 [[K:s[0-9]+]], -1{{$}} 249; GCN: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 250define amdgpu_ps i32 @s_buffer_load_imm_neg1(<4 x i32> inreg %desc) { 251 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -1, i32 0) 252 ret i32 %load 253} 254 255; GCN-LABEL: {{^}}s_buffer_load_imm_neg4: 256; SI: s_mov_b32 [[K:s[0-9]+]], -4{{$}} 257; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 258 259; CI: s_buffer_load_dword s0, s[0:3], 0x3fffffff{{$}} 260 261; VI: s_mov_b32 [[K:s[0-9]+]], -4{{$}} 262; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 263define amdgpu_ps i32 @s_buffer_load_imm_neg4(<4 x i32> inreg %desc) { 264 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -4, i32 0) 265 ret i32 %load 266} 267 268; GCN-LABEL: {{^}}s_buffer_load_imm_neg8: 269; SI: s_mov_b32 [[K:s[0-9]+]], -8{{$}} 270; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 271 272; CI: s_buffer_load_dword s0, s[0:3], 0x3ffffffe{{$}} 273define amdgpu_ps i32 @s_buffer_load_imm_neg8(<4 x i32> inreg %desc) { 274 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -8, i32 0) 275 ret i32 %load 276} 277 278; GCN-LABEL: {{^}}s_buffer_load_imm_bit31: 279; SI: s_brev_b32 [[K:s[0-9]+]], 1{{$}} 280; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 281 282; CI: s_buffer_load_dword s0, s[0:3], 0x20000000{{$}} 283 284; VI: s_brev_b32 [[K:s[0-9]+]], 1{{$}} 285; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 286define amdgpu_ps i32 @s_buffer_load_imm_bit31(<4 x i32> inreg %desc) { 287 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -2147483648, i32 0) 288 ret i32 %load 289} 290 291; GCN-LABEL: {{^}}s_buffer_load_imm_bit30: 292; SI: s_mov_b32 [[K:s[0-9]+]], 2.0{{$}} 293; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 294 295; CI: s_buffer_load_dword s0, s[0:3], 0x10000000{{$}} 296 297; VI: s_mov_b32 [[K:s[0-9]+]], 2.0{{$}} 298; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 299define amdgpu_ps i32 @s_buffer_load_imm_bit30(<4 x i32> inreg %desc) { 300 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1073741824, i32 0) 301 ret i32 %load 302} 303 304; GCN-LABEL: {{^}}s_buffer_load_imm_bit29: 305; SI: s_brev_b32 [[K:s[0-9]+]], 4{{$}} 306; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 307 308; CI: s_buffer_load_dword s0, s[0:3], 0x8000000{{$}} 309 310; VI: s_brev_b32 [[K:s[0-9]+]], 4{{$}} 311; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 312define amdgpu_ps i32 @s_buffer_load_imm_bit29(<4 x i32> inreg %desc) { 313 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 536870912, i32 0) 314 ret i32 %load 315} 316 317; GCN-LABEL: {{^}}s_buffer_load_imm_bit21: 318; SI: s_mov_b32 [[K:s[0-9]+]], 0x200000{{$}} 319; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 320 321; CI: s_buffer_load_dword s0, s[0:3], 0x80000{{$}} 322 323; VI: s_mov_b32 [[K:s[0-9]+]], 0x200000{{$}} 324; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 325define amdgpu_ps i32 @s_buffer_load_imm_bit21(<4 x i32> inreg %desc) { 326 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 2097152, i32 0) 327 ret i32 %load 328} 329 330; GCN-LABEL: {{^}}s_buffer_load_imm_bit20: 331; SI: s_mov_b32 [[K:s[0-9]+]], 0x100000{{$}} 332; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 333 334; CI: s_buffer_load_dword s0, s[0:3], 0x40000{{$}} 335 336; VI: s_mov_b32 [[K:s[0-9]+]], 0x100000{{$}} 337; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 338define amdgpu_ps i32 @s_buffer_load_imm_bit20(<4 x i32> inreg %desc) { 339 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1048576, i32 0) 340 ret i32 %load 341} 342 343; GCN-LABEL: {{^}}s_buffer_load_imm_neg_bit20: 344; SI: s_mov_b32 [[K:s[0-9]+]], 0xfff00000{{$}} 345; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 346 347; CI: s_buffer_load_dword s0, s[0:3], 0x3ffc0000{{$}} 348 349; VI: s_mov_b32 [[K:s[0-9]+]], 0xfff00000{{$}} 350; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 351define amdgpu_ps i32 @s_buffer_load_imm_neg_bit20(<4 x i32> inreg %desc) { 352 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -1048576, i32 0) 353 ret i32 %load 354} 355 356; GCN-LABEL: {{^}}s_buffer_load_imm_bit19: 357; SI: s_mov_b32 [[K:s[0-9]+]], 0x80000{{$}} 358; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 359 360; CI: s_buffer_load_dword s0, s[0:3], 0x20000{{$}} 361 362; VI: s_buffer_load_dword s0, s[0:3], 0x80000{{$}} 363define amdgpu_ps i32 @s_buffer_load_imm_bit19(<4 x i32> inreg %desc) { 364 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 524288, i32 0) 365 ret i32 %load 366} 367 368; GCN-LABEL: {{^}}s_buffer_load_imm_neg_bit19: 369; SI: s_mov_b32 [[K:s[0-9]+]], 0xfff80000{{$}} 370; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 371 372; CI: s_buffer_load_dword s0, s[0:3], 0x3ffe0000{{$}} 373 374; VI: s_mov_b32 [[K:s[0-9]+]], 0xfff80000{{$}} 375; VI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 376define amdgpu_ps i32 @s_buffer_load_imm_neg_bit19(<4 x i32> inreg %desc) { 377 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 -524288, i32 0) 378 ret i32 %load 379} 380 381; GCN-LABEL: {{^}}s_buffer_load_imm_255: 382; SICI: s_movk_i32 [[K:s[0-9]+]], 0xff{{$}} 383; SICI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 384 385; VI: s_buffer_load_dword s0, s[0:3], 0xff{{$}} 386define amdgpu_ps i32 @s_buffer_load_imm_255(<4 x i32> inreg %desc) { 387 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 255, i32 0) 388 ret i32 %load 389} 390 391; GCN-LABEL: {{^}}s_buffer_load_imm_256: 392; SICI: s_buffer_load_dword s0, s[0:3], 0x40{{$}} 393; VI: s_buffer_load_dword s0, s[0:3], 0x100{{$}} 394define amdgpu_ps i32 @s_buffer_load_imm_256(<4 x i32> inreg %desc) { 395 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 256, i32 0) 396 ret i32 %load 397} 398 399; GCN-LABEL: {{^}}s_buffer_load_imm_1016: 400; SICI: s_buffer_load_dword s0, s[0:3], 0xfe{{$}} 401; VI: s_buffer_load_dword s0, s[0:3], 0x3f8{{$}} 402define amdgpu_ps i32 @s_buffer_load_imm_1016(<4 x i32> inreg %desc) { 403 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1016, i32 0) 404 ret i32 %load 405} 406 407; GCN-LABEL: {{^}}s_buffer_load_imm_1020: 408; SICI: s_buffer_load_dword s0, s[0:3], 0xff{{$}} 409; VI: s_buffer_load_dword s0, s[0:3], 0x3fc{{$}} 410define amdgpu_ps i32 @s_buffer_load_imm_1020(<4 x i32> inreg %desc) { 411 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1020, i32 0) 412 ret i32 %load 413} 414 415; GCN-LABEL: {{^}}s_buffer_load_imm_1021: 416; SICI: s_movk_i32 [[K:s[0-9]+]], 0x3fd{{$}} 417; SICI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 418define amdgpu_ps i32 @s_buffer_load_imm_1021(<4 x i32> inreg %desc) { 419 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1021, i32 0) 420 ret i32 %load 421} 422 423; GCN-LABEL: {{^}}s_buffer_load_imm_1024: 424; SI: s_movk_i32 [[K:s[0-9]+]], 0x400{{$}} 425; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 426 427; CI: s_buffer_load_dword s0, s[0:3], 0x100{{$}} 428 429; VI: s_buffer_load_dword s0, s[0:3], 0x400{{$}} 430define amdgpu_ps i32 @s_buffer_load_imm_1024(<4 x i32> inreg %desc) { 431 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1024, i32 0) 432 ret i32 %load 433} 434 435; GCN-LABEL: {{^}}s_buffer_load_imm_1025: 436; SICI: s_movk_i32 [[K:s[0-9]+]], 0x401{{$}} 437; SICI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 438 439; VI: s_buffer_load_dword s0, s[0:3], 0x401{{$}} 440define amdgpu_ps i32 @s_buffer_load_imm_1025(<4 x i32> inreg %desc) { 441 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1025, i32 0) 442 ret i32 %load 443} 444 445; GCN-LABEL: {{^}}s_buffer_load_imm_1028: 446; SI: s_movk_i32 [[K:s[0-9]+]], 0x400{{$}} 447; SI: s_buffer_load_dword s0, s[0:3], [[K]]{{$}} 448 449; CI: s_buffer_load_dword s0, s[0:3], 0x100{{$}} 450; VI: s_buffer_load_dword s0, s[0:3], 0x400{{$}} 451define amdgpu_ps i32 @s_buffer_load_imm_1028(<4 x i32> inreg %desc) { 452 %load = call i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32> %desc, i32 1024, i32 0) 453 ret i32 %load 454} 455 456declare void @llvm.amdgcn.exp.f32(i32, i32, float, float, float, float, i1, i1) 457declare i32 @llvm.amdgcn.s.buffer.load.i32(<4 x i32>, i32, i32) 458declare <2 x i32> @llvm.amdgcn.s.buffer.load.v2i32(<4 x i32>, i32, i32) 459declare <3 x i32> @llvm.amdgcn.s.buffer.load.v3i32(<4 x i32>, i32, i32) 460declare <4 x i32> @llvm.amdgcn.s.buffer.load.v4i32(<4 x i32>, i32, i32) 461