1 ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SI,SIVI,PRT %s
2 ; RUN: llc -march=amdgcn -mcpu=fiji -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,SIVI,PRT %s
3 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,PRT %s
4 ; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-enable-prt-strict-null -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,NOPRT %s
6 ; GCN-LABEL: {{^}}load_1d:
7 ; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
8 define amdgpu_ps <4 x float> @load_1d(<8 x i32> inreg %rsrc, i32 %s) {
10 %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
14 ; GCN-LABEL: {{^}}load_1d_tfe:
15 ; PRT: v_mov_b32_e32 v0, 0
16 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
17 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
18 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
19 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
20 ; NOPRT: v_mov_b32_e32 v4, 0
21 ; NOPRT-NOT: v_mov_b32_e32 v0
22 ; NOPRT-NOT: v_mov_b32_e32 v1
23 ; NOPRT-NOT: v_mov_b32_e32 v2
24 ; NOPRT-NOT: v_mov_b32_e32 v3
25 ; GCN: image_load v[0:7], v{{[0-9]+}}, s[0:7] dmask:0xf unorm tfe{{$}}
26 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
27 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
28 define amdgpu_ps <4 x float> @load_1d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
30 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
31 %v.vec = extractvalue {<4 x float>, i32} %v, 0
32 %v.err = extractvalue {<4 x float>, i32} %v, 1
33 store i32 %v.err, i32 addrspace(1)* %out, align 4
34 ret <4 x float> %v.vec
37 ; GCN-LABEL: {{^}}load_1d_lwe:
38 ; PRT: v_mov_b32_e32 v0, 0
39 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
40 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
41 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
42 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
43 ; NOPRT: v_mov_b32_e32 v4, 0
44 ; NOPRT-NOT: v_mov_b32_e32 v0
45 ; NOPRT-NOT: v_mov_b32_e32 v1
46 ; NOPRT-NOT: v_mov_b32_e32 v2
47 ; NOPRT-NOT: v_mov_b32_e32 v3
48 ; GCN: image_load v[0:7], v{{[0-9]+}}, s[0:7] dmask:0xf unorm lwe{{$}}
49 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
50 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
51 define amdgpu_ps <4 x float> @load_1d_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
53 %v = call {<4 x float>, i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 2, i32 0)
54 %v.vec = extractvalue {<4 x float>, i32} %v, 0
55 %v.err = extractvalue {<4 x float>, i32} %v, 1
56 store i32 %v.err, i32 addrspace(1)* %out, align 4
57 ret <4 x float> %v.vec
60 ; GCN-LABEL: {{^}}load_2d:
61 ; GCN: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
62 define amdgpu_ps <4 x float> @load_2d(<8 x i32> inreg %rsrc, i32 %s, i32 %t) {
64 %v = call <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
68 ; GCN-LABEL: {{^}}load_2d_tfe:
69 ; PRT: v_mov_b32_e32 v0, 0
70 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
71 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
72 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
73 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
74 ; NOPRT: v_mov_b32_e32 v4, 0
75 ; NOPRT-NOT: v_mov_b32_e32 v0
76 ; NOPRT-NOT: v_mov_b32_e32 v1
77 ; NOPRT-NOT: v_mov_b32_e32 v2
78 ; NOPRT-NOT: v_mov_b32_e32 v3
79 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
80 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
81 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
82 define amdgpu_ps <4 x float> @load_2d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t) {
84 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 1, i32 0)
85 %v.vec = extractvalue {<4 x float>, i32} %v, 0
86 %v.err = extractvalue {<4 x float>, i32} %v, 1
87 store i32 %v.err, i32 addrspace(1)* %out, align 4
88 ret <4 x float> %v.vec
91 ; GCN-LABEL: {{^}}load_3d:
92 ; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
93 define amdgpu_ps <4 x float> @load_3d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %r) {
95 %v = call <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
99 ; GCN-LABEL: {{^}}load_3d_tfe_lwe:
100 ; PRT: v_mov_b32_e32 v0, 0
101 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
102 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
103 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
104 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
105 ; NOPRT: v_mov_b32_e32 v4, 0
106 ; NOPRT-NOT: v_mov_b32_e32 v0
107 ; NOPRT-NOT: v_mov_b32_e32 v1
108 ; NOPRT-NOT: v_mov_b32_e32 v2
109 ; NOPRT-NOT: v_mov_b32_e32 v3
110 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
111 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
112 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
113 define amdgpu_ps <4 x float> @load_3d_tfe_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %r) {
115 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 3, i32 0)
116 %v.vec = extractvalue {<4 x float>, i32} %v, 0
117 %v.err = extractvalue {<4 x float>, i32} %v, 1
118 store i32 %v.err, i32 addrspace(1)* %out, align 4
119 ret <4 x float> %v.vec
122 ; GCN-LABEL: {{^}}load_cube:
123 ; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
124 define amdgpu_ps <4 x float> @load_cube(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice) {
126 %v = call <4 x float> @llvm.amdgcn.image.load.cube.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
130 ; GCN-LABEL: {{^}}load_cube_lwe:
131 ; PRT: v_mov_b32_e32 v0, 0
132 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
133 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
134 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
135 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
136 ; NOPRT: v_mov_b32_e32 v4, 0
137 ; NOPRT-NOT: v_mov_b32_e32 v0
138 ; NOPRT-NOT: v_mov_b32_e32 v1
139 ; NOPRT-NOT: v_mov_b32_e32 v2
140 ; NOPRT-NOT: v_mov_b32_e32 v3
141 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
142 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
143 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
144 define amdgpu_ps <4 x float> @load_cube_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice) {
146 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
147 %v.vec = extractvalue {<4 x float>, i32} %v, 0
148 %v.err = extractvalue {<4 x float>, i32} %v, 1
149 store i32 %v.err, i32 addrspace(1)* %out, align 4
150 ret <4 x float> %v.vec
153 ; GCN-LABEL: {{^}}load_1darray:
154 ; GCN: image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm da{{$}}
155 define amdgpu_ps <4 x float> @load_1darray(<8 x i32> inreg %rsrc, i32 %s, i32 %slice) {
157 %v = call <4 x float> @llvm.amdgcn.image.load.1darray.v4f32.i32(i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
161 ; GCN-LABEL: {{^}}load_1darray_tfe:
162 ; PRT: v_mov_b32_e32 v0, 0
163 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
164 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
165 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
166 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
167 ; NOPRT: v_mov_b32_e32 v4, 0
168 ; NOPRT-NOT: v_mov_b32_e32 v0
169 ; NOPRT-NOT: v_mov_b32_e32 v1
170 ; NOPRT-NOT: v_mov_b32_e32 v2
171 ; NOPRT-NOT: v_mov_b32_e32 v3
172 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
173 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
174 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
175 define amdgpu_ps <4 x float> @load_1darray_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %slice) {
177 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1darray.v4f32i32.i32(i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 1, i32 0)
178 %v.vec = extractvalue {<4 x float>, i32} %v, 0
179 %v.err = extractvalue {<4 x float>, i32} %v, 1
180 store i32 %v.err, i32 addrspace(1)* %out, align 4
181 ret <4 x float> %v.vec
184 ; GCN-LABEL: {{^}}load_2darray:
185 ; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
186 define amdgpu_ps <4 x float> @load_2darray(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice) {
188 %v = call <4 x float> @llvm.amdgcn.image.load.2darray.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
192 ; GCN-LABEL: {{^}}load_2darray_lwe:
193 ; PRT: v_mov_b32_e32 v0, 0
194 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
195 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
196 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
197 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
198 ; NOPRT: v_mov_b32_e32 v4, 0
199 ; NOPRT-NOT: v_mov_b32_e32 v0
200 ; NOPRT-NOT: v_mov_b32_e32 v1
201 ; NOPRT-NOT: v_mov_b32_e32 v2
202 ; NOPRT-NOT: v_mov_b32_e32 v3
203 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe da{{$}}
204 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
205 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
206 define amdgpu_ps <4 x float> @load_2darray_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice) {
208 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
209 %v.vec = extractvalue {<4 x float>, i32} %v, 0
210 %v.err = extractvalue {<4 x float>, i32} %v, 1
211 store i32 %v.err, i32 addrspace(1)* %out, align 4
212 ret <4 x float> %v.vec
215 ; GCN-LABEL: {{^}}load_2dmsaa:
216 ; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
217 define amdgpu_ps <4 x float> @load_2dmsaa(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %fragid) {
219 %v = call <4 x float> @llvm.amdgcn.image.load.2dmsaa.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
223 ; GCN-LABEL: {{^}}load_2dmsaa_both:
224 ; PRT: v_mov_b32_e32 v0, 0
225 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
226 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
227 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
228 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
229 ; NOPRT: v_mov_b32_e32 v4, 0
230 ; NOPRT-NOT: v_mov_b32_e32 v0
231 ; NOPRT-NOT: v_mov_b32_e32 v1
232 ; NOPRT-NOT: v_mov_b32_e32 v2
233 ; NOPRT-NOT: v_mov_b32_e32 v3
234 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe lwe{{$}}
235 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
236 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
237 define amdgpu_ps <4 x float> @load_2dmsaa_both(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %fragid) {
239 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
240 %v.vec = extractvalue {<4 x float>, i32} %v, 0
241 %v.err = extractvalue {<4 x float>, i32} %v, 1
242 store i32 %v.err, i32 addrspace(1)* %out, align 4
243 ret <4 x float> %v.vec
246 ; GCN-LABEL: {{^}}load_2darraymsaa:
247 ; GCN: image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
248 define amdgpu_ps <4 x float> @load_2darraymsaa(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
250 %v = call <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
254 ; GCN-LABEL: {{^}}load_2darraymsaa_tfe:
255 ; PRT: v_mov_b32_e32 v0, 0
256 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
257 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
258 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
259 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
260 ; NOPRT: v_mov_b32_e32 v4, 0
261 ; NOPRT-NOT: v_mov_b32_e32 v0
262 ; NOPRT-NOT: v_mov_b32_e32 v1
263 ; NOPRT-NOT: v_mov_b32_e32 v2
264 ; NOPRT-NOT: v_mov_b32_e32 v3
265 ; GCN: image_load v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe da{{$}}
266 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
267 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
268 define amdgpu_ps <4 x float> @load_2darraymsaa_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
270 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darraymsaa.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
271 %v.vec = extractvalue {<4 x float>, i32} %v, 0
272 %v.err = extractvalue {<4 x float>, i32} %v, 1
273 store i32 %v.err, i32 addrspace(1)* %out, align 4
274 ret <4 x float> %v.vec
277 ; GCN-LABEL: {{^}}load_mip_1d:
278 ; GCN: image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf unorm{{$}}
279 define amdgpu_ps <4 x float> @load_mip_1d(<8 x i32> inreg %rsrc, i32 %s, i32 %mip) {
281 %v = call <4 x float> @llvm.amdgcn.image.load.mip.1d.v4f32.i32(i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
285 ; GCN-LABEL: {{^}}load_mip_1d_lwe:
286 ; PRT: v_mov_b32_e32 v0, 0
287 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
288 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
289 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
290 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
291 ; NOPRT: v_mov_b32_e32 v4, 0
292 ; NOPRT-NOT: v_mov_b32_e32 v0
293 ; NOPRT-NOT: v_mov_b32_e32 v1
294 ; NOPRT-NOT: v_mov_b32_e32 v2
295 ; NOPRT-NOT: v_mov_b32_e32 v3
296 ; GCN: image_load_mip v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm lwe{{$}}
297 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
298 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
299 define amdgpu_ps <4 x float> @load_mip_1d_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %mip) {
301 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.1d.v4f32i32.i32(i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 2, i32 0)
302 %v.vec = extractvalue {<4 x float>, i32} %v, 0
303 %v.err = extractvalue {<4 x float>, i32} %v, 1
304 store i32 %v.err, i32 addrspace(1)* %out, align 4
305 ret <4 x float> %v.vec
308 ; GCN-LABEL: {{^}}load_mip_2d:
309 ; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
310 define amdgpu_ps <4 x float> @load_mip_2d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
312 %v = call <4 x float> @llvm.amdgcn.image.load.mip.2d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
316 ; GCN-LABEL: {{^}}load_mip_2d_tfe:
317 ; PRT: v_mov_b32_e32 v0, 0
318 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
319 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
320 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
321 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
322 ; NOPRT: v_mov_b32_e32 v4, 0
323 ; NOPRT-NOT: v_mov_b32_e32 v0
324 ; NOPRT-NOT: v_mov_b32_e32 v1
325 ; NOPRT-NOT: v_mov_b32_e32 v2
326 ; NOPRT-NOT: v_mov_b32_e32 v3
327 ; GCN: image_load_mip v[0:7], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0xf unorm tfe{{$}}
328 ; SIVI: buffer_store_dword v4, off, s[8:11], 0
329 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v4
330 define amdgpu_ps <4 x float> @load_mip_2d_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %mip) {
332 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
333 %v.vec = extractvalue {<4 x float>, i32} %v, 0
334 %v.err = extractvalue {<4 x float>, i32} %v, 1
335 store i32 %v.err, i32 addrspace(1)* %out, align 4
336 ret <4 x float> %v.vec
339 ; Make sure that error flag is returned even with dmask 0
340 ; GCN-LABEL: {{^}}load_1d_V2_tfe_dmask0:
341 ; GCN: v_mov_b32_e32 v1, 0
342 ; PRT-DAG: v_mov_b32_e32 v2, v1
343 ; PRT: image_load v[1:2], v0, s[0:7] dmask:0x1 unorm tfe{{$}}
344 ; NOPRT-NOT: v_mov_b32_e32 v1
345 ; NOPRT: image_load v[0:1], v0, s[0:7] dmask:0x1 unorm tfe{{$}}
346 define amdgpu_ps float @load_1d_V2_tfe_dmask0(<8 x i32> inreg %rsrc, i32 %s) {
348 %v = call {<2 x float>,i32} @llvm.amdgcn.image.load.1d.v2f32i32.i32(i32 0, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
349 %v.err = extractvalue {<2 x float>, i32} %v, 1
350 %vv = bitcast i32 %v.err to float
354 ; GCN-LABEL: {{^}}load_1d_V1_tfe_dmask0:
355 ; GCN: v_mov_b32_e32 v1, 0
356 ; PRT-DAG: v_mov_b32_e32 v2, v1
357 ; PRT: image_load v[1:2], v0, s[0:7] dmask:0x1 unorm tfe{{$}}
358 ; NOPRT-NOT: v_mov_b32_e32 v1
359 ; NOPRT: image_load v[0:1], v0, s[0:7] dmask:0x1 unorm tfe{{$}}
360 define amdgpu_ps float @load_1d_V1_tfe_dmask0(<8 x i32> inreg %rsrc, i32 %s) {
362 %v = call {float,i32} @llvm.amdgcn.image.load.1d.f32i32.i32(i32 0, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
363 %v.err = extractvalue {float, i32} %v, 1
364 %vv = bitcast i32 %v.err to float
368 ; GCN-LABEL: {{^}}load_mip_2d_tfe_dmask0:
369 ; GCN: v_mov_b32_e32 v3, 0
370 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v3
371 ; PRT: image_load_mip v[3:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
372 ; NOPRT-NOT: v_mov_b32_e32 v2
373 ; NOPRT: image_load_mip v[2:3], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
374 define amdgpu_ps float @load_mip_2d_tfe_dmask0(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
376 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 0, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
377 %v.err = extractvalue {<4 x float>, i32} %v, 1
378 %vv = bitcast i32 %v.err to float
382 ; Do not make dmask 0 even if no result (other than tfe) is used.
383 ; GCN-LABEL: {{^}}load_mip_2d_tfe_nouse:
384 ; GCN: v_mov_b32_e32 v3, 0
385 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v3
386 ; PRT: image_load_mip v[3:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
387 ; NOPRT-NOT: v_mov_b32_e32 v2
388 ; NOPRT: image_load_mip v[2:3], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
389 define amdgpu_ps float @load_mip_2d_tfe_nouse(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
391 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
392 %v.err = extractvalue {<4 x float>, i32} %v, 1
393 %vv = bitcast i32 %v.err to float
397 ; GCN-LABEL: {{^}}load_mip_2d_tfe_nouse_V2:
398 ; GCN: v_mov_b32_e32 v3, 0
399 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v3
400 ; PRT: image_load_mip v[3:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
401 ; NOPRT-NOT: v_mov_b32_e32 v2
402 ; NOPRT: image_load_mip v[2:3], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x1 unorm tfe{{$}}
403 define amdgpu_ps float @load_mip_2d_tfe_nouse_V2(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
405 %v = call {<2 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v2f32i32.i32(i32 6, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
406 %v.err = extractvalue {<2 x float>, i32} %v, 1
407 %vv = bitcast i32 %v.err to float
411 ; GCN-LABEL: {{^}}load_mip_2d_tfe_nouse_V1:
412 ; GCN: v_mov_b32_e32 v3, 0
413 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v3
414 ; PRT: image_load_mip v[3:4], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x2 unorm tfe{{$}}
415 ; NOPRT-NOT: v_mov_b32_e32 v2
416 ; NOPRT: image_load_mip v[2:3], v[{{[0-9]+:[0-9]+}}], s[0:7] dmask:0x2 unorm tfe{{$}}
417 define amdgpu_ps float @load_mip_2d_tfe_nouse_V1(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %mip) {
419 %v = call {float, i32} @llvm.amdgcn.image.load.mip.2d.f32i32.i32(i32 2, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 1, i32 0)
420 %v.err = extractvalue {float, i32} %v, 1
421 %vv = bitcast i32 %v.err to float
425 ; Check for dmask being materially smaller than return type
426 ; GCN-LABEL: {{^}}load_1d_tfe_V4_dmask3:
427 ; PRT: v_mov_b32_e32 v0, 0
428 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
429 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
430 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
431 ; NOPRT: v_mov_b32_e32 v3, 0
432 ; NOPRT-NOT: v_mov_b32_e32 v0
433 ; NOPRT-NOT: v_mov_b32_e32 v1
434 ; NOPRT-NOT: v_mov_b32_e32 v2
435 ; GCN: image_load v[0:3], v{{[0-9]+}}, s[0:7] dmask:0x7 unorm tfe{{$}}
436 ; SIVI: buffer_store_dword v3, off, s[8:11], 0
437 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v3
438 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask3(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
440 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 7, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
441 %v.vec = extractvalue {<4 x float>, i32} %v, 0
442 %v.err = extractvalue {<4 x float>, i32} %v, 1
443 store i32 %v.err, i32 addrspace(1)* %out, align 4
444 ret <4 x float> %v.vec
447 ; GCN-LABEL: {{^}}load_1d_tfe_V4_dmask2:
448 ; PRT: v_mov_b32_e32 v0, 0
449 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
450 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
451 ; NOPRT: v_mov_b32_e32 v2, 0
452 ; NOPRT-NOT: v_mov_b32_e32 v0
453 ; NOPRT-NOT: v_mov_b32_e32 v1
454 ; GCN: image_load v[0:3], v{{[0-9]+}}, s[0:7] dmask:0x6 unorm tfe{{$}}
455 ; SIVI: buffer_store_dword v2, off, s[8:11], 0
456 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v2
457 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask2(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
459 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 6, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
460 %v.vec = extractvalue {<4 x float>, i32} %v, 0
461 %v.err = extractvalue {<4 x float>, i32} %v, 1
462 store i32 %v.err, i32 addrspace(1)* %out, align 4
463 ret <4 x float> %v.vec
466 ; GCN-LABEL: {{^}}load_1d_tfe_V4_dmask1:
467 ; PRT: v_mov_b32_e32 v0, 0
468 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
469 ; NOPRT: v_mov_b32_e32 v1, 0
470 ; NOPRT-NOT: v_mov_b32_e32 v0
471 ; GCN: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
472 ; SIVI: buffer_store_dword v1, off, s[8:11], 0
473 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v1
474 define amdgpu_ps <4 x float> @load_1d_tfe_V4_dmask1(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
476 %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
477 %v.vec = extractvalue {<4 x float>, i32} %v, 0
478 %v.err = extractvalue {<4 x float>, i32} %v, 1
479 store i32 %v.err, i32 addrspace(1)* %out, align 4
480 ret <4 x float> %v.vec
483 ; GCN-LABEL: {{^}}load_1d_tfe_V2_dmask1:
484 ; PRT: v_mov_b32_e32 v0, 0
485 ; PRT-DAG: v_mov_b32_e32 v{{[0-9]+}}, v0
486 ; NOPRT: v_mov_b32_e32 v1, 0
487 ; NOPRT-NOT: v_mov_b32_e32 v0
488 ; GCN: image_load v[0:1], v{{[0-9]+}}, s[0:7] dmask:0x8 unorm tfe{{$}}
489 ; SIVI: buffer_store_dword v1, off, s[8:11], 0
490 ; GFX900: global_store_dword v[{{[0-9]+:[0-9]+}}], v1
491 define amdgpu_ps <2 x float> @load_1d_tfe_V2_dmask1(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s) {
493 %v = call {<2 x float>,i32} @llvm.amdgcn.image.load.1d.v2f32i32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 1, i32 0)
494 %v.vec = extractvalue {<2 x float>, i32} %v, 0
495 %v.err = extractvalue {<2 x float>, i32} %v, 1
496 store i32 %v.err, i32 addrspace(1)* %out, align 4
497 ret <2 x float> %v.vec
501 ; GCN-LABEL: {{^}}load_mip_3d:
502 ; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm{{$}}
503 define amdgpu_ps <4 x float> @load_mip_3d(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %r, i32 %mip) {
505 %v = call <4 x float> @llvm.amdgcn.image.load.mip.3d.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %r, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
509 ; GCN-LABEL: {{^}}load_mip_cube:
510 ; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
511 define amdgpu_ps <4 x float> @load_mip_cube(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %mip) {
513 %v = call <4 x float> @llvm.amdgcn.image.load.mip.cube.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
517 ; GCN-LABEL: {{^}}load_mip_1darray:
518 ; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
519 define amdgpu_ps <4 x float> @load_mip_1darray(<8 x i32> inreg %rsrc, i32 %s, i32 %slice, i32 %mip) {
521 %v = call <4 x float> @llvm.amdgcn.image.load.mip.1darray.v4f32.i32(i32 15, i32 %s, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
525 ; GCN-LABEL: {{^}}load_mip_2darray:
526 ; GCN: image_load_mip v[0:3], v[0:3], s[0:7] dmask:0xf unorm da{{$}}
527 define amdgpu_ps <4 x float> @load_mip_2darray(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %mip) {
529 %v = call <4 x float> @llvm.amdgcn.image.load.mip.2darray.v4f32.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
533 ; GCN-LABEL: {{^}}store_1d:
534 ; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm{{$}}
535 define amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
537 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
541 ; GCN-LABEL: {{^}}store_2d:
542 ; GCN: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
543 define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t) {
545 call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
549 ; GCN-LABEL: {{^}}store_3d:
550 ; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
551 define amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %r) {
553 call void @llvm.amdgcn.image.store.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
557 ; GCN-LABEL: {{^}}store_cube:
558 ; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
559 define amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice) {
561 call void @llvm.amdgcn.image.store.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
565 ; GCN-LABEL: {{^}}store_1darray:
566 ; GCN: image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm da{{$}}
567 define amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %slice) {
569 call void @llvm.amdgcn.image.store.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
573 ; GCN-LABEL: {{^}}store_2darray:
574 ; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
575 define amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice) {
577 call void @llvm.amdgcn.image.store.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
581 ; GCN-LABEL: {{^}}store_2dmsaa:
582 ; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
583 define amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %fragid) {
585 call void @llvm.amdgcn.image.store.2dmsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
589 ; GCN-LABEL: {{^}}store_2darraymsaa:
590 ; GCN: image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
591 define amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
593 call void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
597 ; GCN-LABEL: {{^}}store_mip_1d:
598 ; GCN: image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf unorm{{$}}
599 define amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %mip) {
601 call void @llvm.amdgcn.image.store.mip.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
605 ; GCN-LABEL: {{^}}store_mip_2d:
606 ; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
607 define amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %mip) {
609 call void @llvm.amdgcn.image.store.mip.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
613 ; GCN-LABEL: {{^}}store_mip_3d:
614 ; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm{{$}}
615 define amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %r, i32 %mip) {
617 call void @llvm.amdgcn.image.store.mip.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
621 ; GCN-LABEL: {{^}}store_mip_cube:
622 ; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
623 define amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %mip) {
625 call void @llvm.amdgcn.image.store.mip.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
629 ; GCN-LABEL: {{^}}store_mip_1darray:
630 ; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
631 define amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %slice, i32 %mip) {
633 call void @llvm.amdgcn.image.store.mip.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
637 ; GCN-LABEL: {{^}}store_mip_2darray:
638 ; GCN: image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da{{$}}
639 define amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s, i32 %t, i32 %slice, i32 %mip) {
641 call void @llvm.amdgcn.image.store.mip.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
645 ; GCN-LABEL: {{^}}getresinfo_1d:
646 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
647 define amdgpu_ps <4 x float> @getresinfo_1d(<8 x i32> inreg %rsrc, i32 %mip) {
649 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
653 ; GCN-LABEL: {{^}}getresinfo_2d:
654 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
655 define amdgpu_ps <4 x float> @getresinfo_2d(<8 x i32> inreg %rsrc, i32 %mip) {
657 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
661 ; GCN-LABEL: {{^}}getresinfo_3d:
662 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
663 define amdgpu_ps <4 x float> @getresinfo_3d(<8 x i32> inreg %rsrc, i32 %mip) {
665 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.3d.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
669 ; GCN-LABEL: {{^}}getresinfo_cube:
670 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
671 define amdgpu_ps <4 x float> @getresinfo_cube(<8 x i32> inreg %rsrc, i32 %mip) {
673 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.cube.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
677 ; GCN-LABEL: {{^}}getresinfo_1darray:
678 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
679 define amdgpu_ps <4 x float> @getresinfo_1darray(<8 x i32> inreg %rsrc, i32 %mip) {
681 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1darray.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
685 ; GCN-LABEL: {{^}}getresinfo_2darray:
686 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
687 define amdgpu_ps <4 x float> @getresinfo_2darray(<8 x i32> inreg %rsrc, i32 %mip) {
689 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darray.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
693 ; GCN-LABEL: {{^}}getresinfo_2dmsaa:
694 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm{{$}}
695 define amdgpu_ps <4 x float> @getresinfo_2dmsaa(<8 x i32> inreg %rsrc, i32 %mip) {
697 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2dmsaa.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
701 ; GCN-LABEL: {{^}}getresinfo_2darraymsaa:
702 ; GCN: image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf unorm da{{$}}
703 define amdgpu_ps <4 x float> @getresinfo_2darraymsaa(<8 x i32> inreg %rsrc, i32 %mip) {
705 %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darraymsaa.v4f32.i32(i32 15, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
709 ; GCN-LABEL: {{^}}load_1d_V1:
710 ; GCN: image_load v0, v0, s[0:7] dmask:0x8 unorm{{$}}
711 define amdgpu_ps float @load_1d_V1(<8 x i32> inreg %rsrc, i32 %s) {
713 %v = call float @llvm.amdgcn.image.load.1d.f32.i32(i32 8, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
717 ; GCN-LABEL: {{^}}load_1d_V2:
718 ; GCN: image_load v[0:1], v0, s[0:7] dmask:0x9 unorm{{$}}
719 define amdgpu_ps <2 x float> @load_1d_V2(<8 x i32> inreg %rsrc, i32 %s) {
721 %v = call <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i32(i32 9, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
725 ; GCN-LABEL: {{^}}store_1d_V1:
726 ; GCN: image_store v0, v1, s[0:7] dmask:0x2 unorm{{$}}
727 define amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, i32 %s) {
729 call void @llvm.amdgcn.image.store.1d.f32.i32(float %vdata, i32 2, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
733 ; GCN-LABEL: {{^}}store_1d_V2:
734 ; GCN: image_store v[0:1], v2, s[0:7] dmask:0xc unorm{{$}}
735 define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, i32 %s) {
737 call void @llvm.amdgcn.image.store.1d.v2f32.i32(<2 x float> %vdata, i32 12, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
741 ; GCN-LABEL: {{^}}load_1d_glc:
742 ; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc{{$}}
743 define amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, i32 %s) {
745 %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 1)
749 ; GCN-LABEL: {{^}}load_1d_slc:
750 ; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm slc{{$}}
751 define amdgpu_ps <4 x float> @load_1d_slc(<8 x i32> inreg %rsrc, i32 %s) {
753 %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
757 ; GCN-LABEL: {{^}}load_1d_glc_slc:
758 ; GCN: image_load v[0:3], v0, s[0:7] dmask:0xf unorm glc slc{{$}}
759 define amdgpu_ps <4 x float> @load_1d_glc_slc(<8 x i32> inreg %rsrc, i32 %s) {
761 %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 3)
765 ; GCN-LABEL: {{^}}store_1d_glc:
766 ; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc{{$}}
767 define amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
769 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 1)
773 ; GCN-LABEL: {{^}}store_1d_slc:
774 ; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm slc{{$}}
775 define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
777 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
781 ; GCN-LABEL: {{^}}store_1d_glc_slc:
782 ; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc slc{{$}}
783 define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %s) {
785 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 3)
789 ; GCN-LABEL: {{^}}getresinfo_dmask0:
791 ; GCN: ; return to shader part epilog
792 define amdgpu_ps <4 x float> @getresinfo_dmask0(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %mip) #0 {
794 %r = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i32(i32 0, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
798 ; Ideally, the register allocator would avoid the wait here
800 ; GCN-LABEL: {{^}}image_store_wait:
801 ; GCN: image_store v[0:3], v4, s[0:7] dmask:0xf unorm
802 ; SI: s_waitcnt expcnt(0)
803 ; GCN: image_load v[0:3], v4, s[8:15] dmask:0xf unorm
804 ; GCN: s_waitcnt vmcnt(0)
805 ; GCN: image_store v[0:3], v4, s[16:23] dmask:0xf unorm
806 define amdgpu_ps void @image_store_wait(<8 x i32> inreg %arg, <8 x i32> inreg %arg1, <8 x i32> inreg %arg2, <4 x float> %arg3, i32 %arg4) #0 {
808 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %arg3, i32 15, i32 %arg4, <8 x i32> %arg, i32 0, i32 0)
809 %data = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %arg4, <8 x i32> %arg1, i32 0, i32 0)
810 call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %data, i32 15, i32 %arg4, <8 x i32> %arg2, i32 0, i32 0)
814 ; SI won't merge ds memory operations, because of the signed offset bug, so
815 ; we only have check lines for VI.
816 ; VI-LABEL: image_load_mmo
817 ; VI: v_mov_b32_e32 [[ZERO:v[0-9]+]], 0
818 ; VI: ds_write2_b32 v{{[0-9]+}}, [[ZERO]], [[ZERO]] offset1:4
819 define amdgpu_ps float @image_load_mmo(<8 x i32> inreg %rsrc, float addrspace(3)* %lds, <2 x i32> %c) #0 {
820 store float 0.000000e+00, float addrspace(3)* %lds
821 %c0 = extractelement <2 x i32> %c, i32 0
822 %c1 = extractelement <2 x i32> %c, i32 1
823 %tex = call float @llvm.amdgcn.image.load.2d.f32.i32(i32 1, i32 %c0, i32 %c1, <8 x i32> %rsrc, i32 0, i32 0)
824 %tmp2 = getelementptr float, float addrspace(3)* %lds, i32 4
825 store float 0.000000e+00, float addrspace(3)* %tmp2
829 declare <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #1
830 declare {float,i32} @llvm.amdgcn.image.load.1d.f32i32.i32(i32, i32, <8 x i32>, i32, i32) #1
831 declare {<2 x float>,i32} @llvm.amdgcn.image.load.1d.v2f32i32.i32(i32, i32, <8 x i32>, i32, i32) #1
832 declare {<4 x float>,i32} @llvm.amdgcn.image.load.1d.v4f32i32.i32(i32, i32, <8 x i32>, i32, i32) #1
833 declare <4 x float> @llvm.amdgcn.image.load.2d.v4f32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
834 declare {<4 x float>,i32} @llvm.amdgcn.image.load.2d.v4f32i32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
835 declare <4 x float> @llvm.amdgcn.image.load.3d.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
836 declare {<4 x float>,i32} @llvm.amdgcn.image.load.3d.v4f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
837 declare <4 x float> @llvm.amdgcn.image.load.cube.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
838 declare {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
839 declare <4 x float> @llvm.amdgcn.image.load.1darray.v4f32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
840 declare {<4 x float>,i32} @llvm.amdgcn.image.load.1darray.v4f32i32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
841 declare <4 x float> @llvm.amdgcn.image.load.2darray.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
842 declare {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
843 declare <4 x float> @llvm.amdgcn.image.load.2dmsaa.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
844 declare {<4 x float>,i32} @llvm.amdgcn.image.load.2dmsaa.v4f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
845 declare <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #1
846 declare {<4 x float>,i32} @llvm.amdgcn.image.load.2darraymsaa.v4f32i32.i32(i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #1
848 declare <4 x float> @llvm.amdgcn.image.load.mip.1d.v4f32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
849 declare <4 x float> @llvm.amdgcn.image.load.mip.2d.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
850 declare {<4 x float>,i32} @llvm.amdgcn.image.load.mip.1d.v4f32i32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
851 declare {<4 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v4f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
852 declare {<2 x float>,i32} @llvm.amdgcn.image.load.mip.2d.v2f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
853 declare {float,i32} @llvm.amdgcn.image.load.mip.2d.f32i32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
854 declare <4 x float> @llvm.amdgcn.image.load.mip.3d.v4f32.i32(i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #1
855 declare <4 x float> @llvm.amdgcn.image.load.mip.cube.v4f32.i32(i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #1
856 declare <4 x float> @llvm.amdgcn.image.load.mip.1darray.v4f32.i32(i32, i32, i32, i32, <8 x i32>, i32, i32) #1
857 declare <4 x float> @llvm.amdgcn.image.load.mip.2darray.v4f32.i32(i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #1
859 declare void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float>, i32, i32, <8 x i32>, i32, i32) #0
860 declare void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float>, i32, i32, i32, <8 x i32>, i32, i32) #0
861 declare void @llvm.amdgcn.image.store.3d.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
862 declare void @llvm.amdgcn.image.store.cube.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
863 declare void @llvm.amdgcn.image.store.1darray.v4f32.i32(<4 x float>, i32, i32, i32, <8 x i32>, i32, i32) #0
864 declare void @llvm.amdgcn.image.store.2darray.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
865 declare void @llvm.amdgcn.image.store.2dmsaa.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
866 declare void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i32(<4 x float>, i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
868 declare void @llvm.amdgcn.image.store.mip.1d.v4f32.i32(<4 x float>, i32, i32, i32, <8 x i32>, i32, i32) #0
869 declare void @llvm.amdgcn.image.store.mip.2d.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
870 declare void @llvm.amdgcn.image.store.mip.3d.v4f32.i32(<4 x float>, i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
871 declare void @llvm.amdgcn.image.store.mip.cube.v4f32.i32(<4 x float>, i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
872 declare void @llvm.amdgcn.image.store.mip.1darray.v4f32.i32(<4 x float>, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
873 declare void @llvm.amdgcn.image.store.mip.2darray.v4f32.i32(<4 x float>, i32, i32, i32, i32, i32, <8 x i32>, i32, i32) #0
875 declare <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
876 declare <4 x float> @llvm.amdgcn.image.getresinfo.2d.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
877 declare <4 x float> @llvm.amdgcn.image.getresinfo.3d.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
878 declare <4 x float> @llvm.amdgcn.image.getresinfo.cube.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
879 declare <4 x float> @llvm.amdgcn.image.getresinfo.1darray.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
880 declare <4 x float> @llvm.amdgcn.image.getresinfo.2darray.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
881 declare <4 x float> @llvm.amdgcn.image.getresinfo.2dmsaa.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
882 declare <4 x float> @llvm.amdgcn.image.getresinfo.2darraymsaa.v4f32.i32(i32, i32, <8 x i32>, i32, i32) #2
884 declare float @llvm.amdgcn.image.load.1d.f32.i32(i32, i32, <8 x i32>, i32, i32) #1
885 declare float @llvm.amdgcn.image.load.2d.f32.i32(i32, i32, i32, <8 x i32>, i32, i32) #1
886 declare <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i32(i32, i32, <8 x i32>, i32, i32) #1
887 declare void @llvm.amdgcn.image.store.1d.f32.i32(float, i32, i32, <8 x i32>, i32, i32) #0
888 declare void @llvm.amdgcn.image.store.1d.v2f32.i32(<2 x float>, i32, i32, <8 x i32>, i32, i32) #0
890 attributes #0 = { nounwind }
891 attributes #1 = { nounwind readonly }
892 attributes #2 = { nounwind readnone }