xref: /llvm-project/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.load.2darraymsaa.ll (revision ba52f06f9d92c7ca04b440f618f8d352ea121fcc) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=GFX6 %s
3; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX10 %s
4; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
5; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1200 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX12 %s
6
7define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
8; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw:
9; GFX6:       ; %bb.0:
10; GFX6-NEXT:    s_mov_b32 s0, s2
11; GFX6-NEXT:    s_mov_b32 s1, s3
12; GFX6-NEXT:    s_mov_b32 s2, s4
13; GFX6-NEXT:    s_mov_b32 s3, s5
14; GFX6-NEXT:    s_mov_b32 s4, s6
15; GFX6-NEXT:    s_mov_b32 s5, s7
16; GFX6-NEXT:    s_mov_b32 s6, s8
17; GFX6-NEXT:    s_mov_b32 s7, s9
18; GFX6-NEXT:    image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da
19; GFX6-NEXT:    s_waitcnt vmcnt(0)
20; GFX6-NEXT:    ; return to shader part epilog
21;
22; GFX10PLUS-LABEL: load_2darraymsaa_v4f32_xyzw:
23; GFX10PLUS:       ; %bb.0:
24; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
25; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
26; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
27; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
28; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
29; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
30; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
31; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
32; GFX10PLUS-NEXT:    image_load v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm
33; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
34; GFX10PLUS-NEXT:    ; return to shader part epilog
35;
36; GFX12-LABEL: load_2darraymsaa_v4f32_xyzw:
37; GFX12:       ; %bb.0:
38; GFX12-NEXT:    s_mov_b32 s0, s2
39; GFX12-NEXT:    s_mov_b32 s1, s3
40; GFX12-NEXT:    s_mov_b32 s2, s4
41; GFX12-NEXT:    s_mov_b32 s3, s5
42; GFX12-NEXT:    s_mov_b32 s4, s6
43; GFX12-NEXT:    s_mov_b32 s5, s7
44; GFX12-NEXT:    s_mov_b32 s6, s8
45; GFX12-NEXT:    s_mov_b32 s7, s9
46; GFX12-NEXT:    image_load v[0:3], [v0, v1, v2, v3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY
47; GFX12-NEXT:    s_wait_loadcnt 0x0
48; GFX12-NEXT:    ; return to shader part epilog
49  %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)
50  ret <4 x float> %v
51}
52
53define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe(<8 x i32> inreg %rsrc, ptr addrspace(1) inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
54; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
55; GFX6:       ; %bb.0:
56; GFX6-NEXT:    v_mov_b32_e32 v5, v0
57; GFX6-NEXT:    v_mov_b32_e32 v0, 0
58; GFX6-NEXT:    s_mov_b32 s0, s2
59; GFX6-NEXT:    s_mov_b32 s1, s3
60; GFX6-NEXT:    s_mov_b32 s2, s4
61; GFX6-NEXT:    s_mov_b32 s3, s5
62; GFX6-NEXT:    s_mov_b32 s4, s6
63; GFX6-NEXT:    s_mov_b32 s5, s7
64; GFX6-NEXT:    s_mov_b32 s6, s8
65; GFX6-NEXT:    s_mov_b32 s7, s9
66; GFX6-NEXT:    v_mov_b32_e32 v6, v1
67; GFX6-NEXT:    v_mov_b32_e32 v7, v2
68; GFX6-NEXT:    v_mov_b32_e32 v8, v3
69; GFX6-NEXT:    v_mov_b32_e32 v1, v0
70; GFX6-NEXT:    v_mov_b32_e32 v2, v0
71; GFX6-NEXT:    v_mov_b32_e32 v3, v0
72; GFX6-NEXT:    v_mov_b32_e32 v4, v0
73; GFX6-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf unorm tfe da
74; GFX6-NEXT:    s_mov_b32 s8, s10
75; GFX6-NEXT:    s_mov_b32 s9, s11
76; GFX6-NEXT:    s_mov_b32 s10, -1
77; GFX6-NEXT:    s_mov_b32 s11, 0xf000
78; GFX6-NEXT:    s_waitcnt vmcnt(0)
79; GFX6-NEXT:    buffer_store_dword v4, off, s[8:11], 0
80; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
81; GFX6-NEXT:    ; return to shader part epilog
82;
83; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
84; GFX10:       ; %bb.0:
85; GFX10-NEXT:    v_mov_b32_e32 v9, 0
86; GFX10-NEXT:    v_mov_b32_e32 v5, v0
87; GFX10-NEXT:    v_mov_b32_e32 v6, v1
88; GFX10-NEXT:    v_mov_b32_e32 v7, v2
89; GFX10-NEXT:    v_mov_b32_e32 v8, v3
90; GFX10-NEXT:    v_mov_b32_e32 v10, v9
91; GFX10-NEXT:    v_mov_b32_e32 v11, v9
92; GFX10-NEXT:    v_mov_b32_e32 v12, v9
93; GFX10-NEXT:    v_mov_b32_e32 v13, v9
94; GFX10-NEXT:    s_mov_b32 s0, s2
95; GFX10-NEXT:    s_mov_b32 s1, s3
96; GFX10-NEXT:    s_mov_b32 s2, s4
97; GFX10-NEXT:    s_mov_b32 s3, s5
98; GFX10-NEXT:    s_mov_b32 s4, s6
99; GFX10-NEXT:    s_mov_b32 s5, s7
100; GFX10-NEXT:    s_mov_b32 s6, s8
101; GFX10-NEXT:    s_mov_b32 s7, s9
102; GFX10-NEXT:    v_mov_b32_e32 v0, v9
103; GFX10-NEXT:    v_mov_b32_e32 v1, v10
104; GFX10-NEXT:    v_mov_b32_e32 v2, v11
105; GFX10-NEXT:    v_mov_b32_e32 v3, v12
106; GFX10-NEXT:    v_mov_b32_e32 v4, v13
107; GFX10-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
108; GFX10-NEXT:    s_waitcnt vmcnt(0)
109; GFX10-NEXT:    global_store_dword v9, v4, s[10:11]
110; GFX10-NEXT:    ; return to shader part epilog
111;
112; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
113; GFX11:       ; %bb.0:
114; GFX11-NEXT:    v_mov_b32_e32 v9, 0
115; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
116; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
117; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
118; GFX11-NEXT:    v_mov_b32_e32 v10, v9
119; GFX11-NEXT:    v_mov_b32_e32 v11, v9
120; GFX11-NEXT:    v_mov_b32_e32 v12, v9
121; GFX11-NEXT:    v_mov_b32_e32 v13, v9
122; GFX11-NEXT:    s_mov_b32 s0, s2
123; GFX11-NEXT:    s_mov_b32 s1, s3
124; GFX11-NEXT:    s_mov_b32 s2, s4
125; GFX11-NEXT:    s_mov_b32 s3, s5
126; GFX11-NEXT:    s_mov_b32 s4, s6
127; GFX11-NEXT:    s_mov_b32 s5, s7
128; GFX11-NEXT:    s_mov_b32 s6, s8
129; GFX11-NEXT:    s_mov_b32 s7, s9
130; GFX11-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
131; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
132; GFX11-NEXT:    v_mov_b32_e32 v4, v13
133; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
134; GFX11-NEXT:    s_waitcnt vmcnt(0)
135; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
136; GFX11-NEXT:    ; return to shader part epilog
137;
138; GFX12-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
139; GFX12:       ; %bb.0:
140; GFX12-NEXT:    v_mov_b32_e32 v9, 0
141; GFX12-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
142; GFX12-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
143; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_3)
144; GFX12-NEXT:    v_dual_mov_b32 v10, v9 :: v_dual_mov_b32 v11, v9
145; GFX12-NEXT:    v_dual_mov_b32 v12, v9 :: v_dual_mov_b32 v13, v9
146; GFX12-NEXT:    s_mov_b32 s0, s2
147; GFX12-NEXT:    s_mov_b32 s1, s3
148; GFX12-NEXT:    s_mov_b32 s2, s4
149; GFX12-NEXT:    s_mov_b32 s3, s5
150; GFX12-NEXT:    s_mov_b32 s4, s6
151; GFX12-NEXT:    s_mov_b32 s5, s7
152; GFX12-NEXT:    s_mov_b32 s6, s8
153; GFX12-NEXT:    s_mov_b32 s7, s9
154; GFX12-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
155; GFX12-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
156; GFX12-NEXT:    v_mov_b32_e32 v4, v13
157; GFX12-NEXT:    image_load v[0:4], [v5, v6, v7, v8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY tfe
158; GFX12-NEXT:    s_wait_loadcnt 0x0
159; GFX12-NEXT:    global_store_b32 v9, v4, s[10:11]
160; GFX12-NEXT:    ; return to shader part epilog
161  %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
162  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
163  %v.err = extractvalue { <4 x float>, i32 } %v, 1
164  store i32 %v.err, ptr addrspace(1) %out, align 4
165  ret <4 x float> %v.vec
166}
167
168define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe_lwe(<8 x i32> inreg %rsrc, ptr addrspace(1) inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
169; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
170; GFX6:       ; %bb.0:
171; GFX6-NEXT:    v_mov_b32_e32 v5, v0
172; GFX6-NEXT:    v_mov_b32_e32 v0, 0
173; GFX6-NEXT:    s_mov_b32 s0, s2
174; GFX6-NEXT:    s_mov_b32 s1, s3
175; GFX6-NEXT:    s_mov_b32 s2, s4
176; GFX6-NEXT:    s_mov_b32 s3, s5
177; GFX6-NEXT:    s_mov_b32 s4, s6
178; GFX6-NEXT:    s_mov_b32 s5, s7
179; GFX6-NEXT:    s_mov_b32 s6, s8
180; GFX6-NEXT:    s_mov_b32 s7, s9
181; GFX6-NEXT:    v_mov_b32_e32 v6, v1
182; GFX6-NEXT:    v_mov_b32_e32 v7, v2
183; GFX6-NEXT:    v_mov_b32_e32 v8, v3
184; GFX6-NEXT:    v_mov_b32_e32 v1, v0
185; GFX6-NEXT:    v_mov_b32_e32 v2, v0
186; GFX6-NEXT:    v_mov_b32_e32 v3, v0
187; GFX6-NEXT:    v_mov_b32_e32 v4, v0
188; GFX6-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf unorm tfe lwe da
189; GFX6-NEXT:    s_mov_b32 s8, s10
190; GFX6-NEXT:    s_mov_b32 s9, s11
191; GFX6-NEXT:    s_mov_b32 s10, -1
192; GFX6-NEXT:    s_mov_b32 s11, 0xf000
193; GFX6-NEXT:    s_waitcnt vmcnt(0)
194; GFX6-NEXT:    buffer_store_dword v4, off, s[8:11], 0
195; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
196; GFX6-NEXT:    ; return to shader part epilog
197;
198; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
199; GFX10:       ; %bb.0:
200; GFX10-NEXT:    v_mov_b32_e32 v9, 0
201; GFX10-NEXT:    v_mov_b32_e32 v5, v0
202; GFX10-NEXT:    v_mov_b32_e32 v6, v1
203; GFX10-NEXT:    v_mov_b32_e32 v7, v2
204; GFX10-NEXT:    v_mov_b32_e32 v8, v3
205; GFX10-NEXT:    v_mov_b32_e32 v10, v9
206; GFX10-NEXT:    v_mov_b32_e32 v11, v9
207; GFX10-NEXT:    v_mov_b32_e32 v12, v9
208; GFX10-NEXT:    v_mov_b32_e32 v13, v9
209; GFX10-NEXT:    s_mov_b32 s0, s2
210; GFX10-NEXT:    s_mov_b32 s1, s3
211; GFX10-NEXT:    s_mov_b32 s2, s4
212; GFX10-NEXT:    s_mov_b32 s3, s5
213; GFX10-NEXT:    s_mov_b32 s4, s6
214; GFX10-NEXT:    s_mov_b32 s5, s7
215; GFX10-NEXT:    s_mov_b32 s6, s8
216; GFX10-NEXT:    s_mov_b32 s7, s9
217; GFX10-NEXT:    v_mov_b32_e32 v0, v9
218; GFX10-NEXT:    v_mov_b32_e32 v1, v10
219; GFX10-NEXT:    v_mov_b32_e32 v2, v11
220; GFX10-NEXT:    v_mov_b32_e32 v3, v12
221; GFX10-NEXT:    v_mov_b32_e32 v4, v13
222; GFX10-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe lwe
223; GFX10-NEXT:    s_waitcnt vmcnt(0)
224; GFX10-NEXT:    global_store_dword v9, v4, s[10:11]
225; GFX10-NEXT:    ; return to shader part epilog
226;
227; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
228; GFX11:       ; %bb.0:
229; GFX11-NEXT:    v_mov_b32_e32 v9, 0
230; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
231; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
232; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
233; GFX11-NEXT:    v_mov_b32_e32 v10, v9
234; GFX11-NEXT:    v_mov_b32_e32 v11, v9
235; GFX11-NEXT:    v_mov_b32_e32 v12, v9
236; GFX11-NEXT:    v_mov_b32_e32 v13, v9
237; GFX11-NEXT:    s_mov_b32 s0, s2
238; GFX11-NEXT:    s_mov_b32 s1, s3
239; GFX11-NEXT:    s_mov_b32 s2, s4
240; GFX11-NEXT:    s_mov_b32 s3, s5
241; GFX11-NEXT:    s_mov_b32 s4, s6
242; GFX11-NEXT:    s_mov_b32 s5, s7
243; GFX11-NEXT:    s_mov_b32 s6, s8
244; GFX11-NEXT:    s_mov_b32 s7, s9
245; GFX11-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
246; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
247; GFX11-NEXT:    v_mov_b32_e32 v4, v13
248; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe lwe
249; GFX11-NEXT:    s_waitcnt vmcnt(0)
250; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
251; GFX11-NEXT:    ; return to shader part epilog
252;
253; GFX12-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
254; GFX12:       ; %bb.0:
255; GFX12-NEXT:    v_mov_b32_e32 v9, 0
256; GFX12-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
257; GFX12-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
258; GFX12-NEXT:    s_delay_alu instid0(VALU_DEP_3)
259; GFX12-NEXT:    v_dual_mov_b32 v10, v9 :: v_dual_mov_b32 v11, v9
260; GFX12-NEXT:    v_dual_mov_b32 v12, v9 :: v_dual_mov_b32 v13, v9
261; GFX12-NEXT:    s_mov_b32 s0, s2
262; GFX12-NEXT:    s_mov_b32 s1, s3
263; GFX12-NEXT:    s_mov_b32 s2, s4
264; GFX12-NEXT:    s_mov_b32 s3, s5
265; GFX12-NEXT:    s_mov_b32 s4, s6
266; GFX12-NEXT:    s_mov_b32 s5, s7
267; GFX12-NEXT:    s_mov_b32 s6, s8
268; GFX12-NEXT:    s_mov_b32 s7, s9
269; GFX12-NEXT:    v_dual_mov_b32 v0, v9 :: v_dual_mov_b32 v1, v10
270; GFX12-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
271; GFX12-NEXT:    v_mov_b32_e32 v4, v13
272; GFX12-NEXT:    image_load v[0:4], [v5, v6, v7, v8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY tfe
273; GFX12-NEXT:    s_wait_loadcnt 0x0
274; GFX12-NEXT:    global_store_b32 v9, v4, s[10:11]
275; GFX12-NEXT:    ; return to shader part epilog
276  %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
277  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
278  %v.err = extractvalue { <4 x float>, i32 } %v, 1
279  store i32 %v.err, ptr addrspace(1) %out, align 4
280  ret <4 x float> %v.vec
281}
282
283declare <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32 immarg, i32, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
284declare { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i32(i32 immarg, i32, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
285
286attributes #0 = { nounwind readonly }
287