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
6define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw(<8 x i32> inreg %rsrc, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
7; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw:
8; GFX6:       ; %bb.0:
9; GFX6-NEXT:    s_mov_b32 s0, s2
10; GFX6-NEXT:    s_mov_b32 s1, s3
11; GFX6-NEXT:    s_mov_b32 s2, s4
12; GFX6-NEXT:    s_mov_b32 s3, s5
13; GFX6-NEXT:    s_mov_b32 s4, s6
14; GFX6-NEXT:    s_mov_b32 s5, s7
15; GFX6-NEXT:    s_mov_b32 s6, s8
16; GFX6-NEXT:    s_mov_b32 s7, s9
17; GFX6-NEXT:    image_load v[0:3], v[0:3], s[0:7] dmask:0xf unorm da
18; GFX6-NEXT:    s_waitcnt vmcnt(0)
19; GFX6-NEXT:    ; return to shader part epilog
20;
21; GFX10PLUS-LABEL: load_2darraymsaa_v4f32_xyzw:
22; GFX10PLUS:       ; %bb.0:
23; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
24; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
25; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
26; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
27; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
28; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
29; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
30; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
31; GFX10PLUS-NEXT:    image_load v[0:3], v[0:3], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm
32; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
33; GFX10PLUS-NEXT:    ; return to shader part epilog
34  %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)
35  ret <4 x float> %v
36}
37
38define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
39; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
40; GFX6:       ; %bb.0:
41; GFX6-NEXT:    v_mov_b32_e32 v5, v0
42; GFX6-NEXT:    v_mov_b32_e32 v0, 0
43; GFX6-NEXT:    s_mov_b32 s0, s2
44; GFX6-NEXT:    s_mov_b32 s1, s3
45; GFX6-NEXT:    s_mov_b32 s2, s4
46; GFX6-NEXT:    s_mov_b32 s3, s5
47; GFX6-NEXT:    s_mov_b32 s4, s6
48; GFX6-NEXT:    s_mov_b32 s5, s7
49; GFX6-NEXT:    s_mov_b32 s6, s8
50; GFX6-NEXT:    s_mov_b32 s7, s9
51; GFX6-NEXT:    v_mov_b32_e32 v6, v1
52; GFX6-NEXT:    v_mov_b32_e32 v7, v2
53; GFX6-NEXT:    v_mov_b32_e32 v8, v3
54; GFX6-NEXT:    v_mov_b32_e32 v1, v0
55; GFX6-NEXT:    v_mov_b32_e32 v2, v0
56; GFX6-NEXT:    v_mov_b32_e32 v3, v0
57; GFX6-NEXT:    v_mov_b32_e32 v4, v0
58; GFX6-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf unorm tfe da
59; GFX6-NEXT:    s_mov_b32 s8, s10
60; GFX6-NEXT:    s_mov_b32 s9, s11
61; GFX6-NEXT:    s_mov_b32 s10, -1
62; GFX6-NEXT:    s_mov_b32 s11, 0xf000
63; GFX6-NEXT:    s_waitcnt vmcnt(0)
64; GFX6-NEXT:    buffer_store_dword v4, off, s[8:11], 0
65; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
66; GFX6-NEXT:    ; return to shader part epilog
67;
68; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
69; GFX10:       ; %bb.0:
70; GFX10-NEXT:    v_mov_b32_e32 v9, 0
71; GFX10-NEXT:    v_mov_b32_e32 v5, v0
72; GFX10-NEXT:    v_mov_b32_e32 v6, v1
73; GFX10-NEXT:    v_mov_b32_e32 v7, v2
74; GFX10-NEXT:    v_mov_b32_e32 v8, v3
75; GFX10-NEXT:    v_mov_b32_e32 v10, v9
76; GFX10-NEXT:    v_mov_b32_e32 v11, v9
77; GFX10-NEXT:    v_mov_b32_e32 v12, v9
78; GFX10-NEXT:    v_mov_b32_e32 v13, v9
79; GFX10-NEXT:    s_mov_b32 s0, s2
80; GFX10-NEXT:    s_mov_b32 s1, s3
81; GFX10-NEXT:    s_mov_b32 s2, s4
82; GFX10-NEXT:    s_mov_b32 s3, s5
83; GFX10-NEXT:    s_mov_b32 s4, s6
84; GFX10-NEXT:    s_mov_b32 s5, s7
85; GFX10-NEXT:    s_mov_b32 s6, s8
86; GFX10-NEXT:    s_mov_b32 s7, s9
87; GFX10-NEXT:    v_mov_b32_e32 v0, v9
88; GFX10-NEXT:    v_mov_b32_e32 v1, v10
89; GFX10-NEXT:    v_mov_b32_e32 v2, v11
90; GFX10-NEXT:    v_mov_b32_e32 v3, v12
91; GFX10-NEXT:    v_mov_b32_e32 v4, v13
92; GFX10-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
93; GFX10-NEXT:    s_waitcnt vmcnt(0)
94; GFX10-NEXT:    global_store_dword v9, v4, s[10:11]
95; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
96; GFX10-NEXT:    ; return to shader part epilog
97;
98; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
99; GFX11:       ; %bb.0:
100; GFX11-NEXT:    v_mov_b32_e32 v9, 0
101; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
102; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
103; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
104; GFX11-NEXT:    v_mov_b32_e32 v11, v9
105; GFX11-NEXT:    v_mov_b32_e32 v13, v9
106; GFX11-NEXT:    v_mov_b32_e32 v12, v9
107; GFX11-NEXT:    v_mov_b32_e32 v10, v9
108; GFX11-NEXT:    s_mov_b32 s0, s2
109; GFX11-NEXT:    s_mov_b32 s1, s3
110; GFX11-NEXT:    s_mov_b32 s2, s4
111; GFX11-NEXT:    s_mov_b32 s3, s5
112; GFX11-NEXT:    s_mov_b32 s4, s6
113; GFX11-NEXT:    s_mov_b32 s5, s7
114; GFX11-NEXT:    s_mov_b32 s6, s8
115; GFX11-NEXT:    s_mov_b32 s7, s9
116; GFX11-NEXT:    v_mov_b32_e32 v0, v9
117; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
118; GFX11-NEXT:    v_dual_mov_b32 v1, v10 :: v_dual_mov_b32 v4, v13
119; GFX11-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm tfe
120; GFX11-NEXT:    s_waitcnt vmcnt(0)
121; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
122; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
123; GFX11-NEXT:    ; return to shader part epilog
124  %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)
125  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
126  %v.err = extractvalue { <4 x float>, i32 } %v, 1
127  store i32 %v.err, i32 addrspace(1)* %out, align 4
128  ret <4 x float> %v.vec
129}
130
131define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i32 %s, i32 %t, i32 %slice, i32 %fragid) {
132; GFX6-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
133; GFX6:       ; %bb.0:
134; GFX6-NEXT:    v_mov_b32_e32 v5, v0
135; GFX6-NEXT:    v_mov_b32_e32 v0, 0
136; GFX6-NEXT:    s_mov_b32 s0, s2
137; GFX6-NEXT:    s_mov_b32 s1, s3
138; GFX6-NEXT:    s_mov_b32 s2, s4
139; GFX6-NEXT:    s_mov_b32 s3, s5
140; GFX6-NEXT:    s_mov_b32 s4, s6
141; GFX6-NEXT:    s_mov_b32 s5, s7
142; GFX6-NEXT:    s_mov_b32 s6, s8
143; GFX6-NEXT:    s_mov_b32 s7, s9
144; GFX6-NEXT:    v_mov_b32_e32 v6, v1
145; GFX6-NEXT:    v_mov_b32_e32 v7, v2
146; GFX6-NEXT:    v_mov_b32_e32 v8, v3
147; GFX6-NEXT:    v_mov_b32_e32 v1, v0
148; GFX6-NEXT:    v_mov_b32_e32 v2, v0
149; GFX6-NEXT:    v_mov_b32_e32 v3, v0
150; GFX6-NEXT:    v_mov_b32_e32 v4, v0
151; GFX6-NEXT:    image_load v[0:4], v[5:8], s[0:7] dmask:0xf unorm tfe lwe da
152; GFX6-NEXT:    s_mov_b32 s8, s10
153; GFX6-NEXT:    s_mov_b32 s9, s11
154; GFX6-NEXT:    s_mov_b32 s10, -1
155; GFX6-NEXT:    s_mov_b32 s11, 0xf000
156; GFX6-NEXT:    s_waitcnt vmcnt(0)
157; GFX6-NEXT:    buffer_store_dword v4, off, s[8:11], 0
158; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0)
159; GFX6-NEXT:    ; return to shader part epilog
160;
161; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
162; GFX10:       ; %bb.0:
163; GFX10-NEXT:    v_mov_b32_e32 v9, 0
164; GFX10-NEXT:    v_mov_b32_e32 v5, v0
165; GFX10-NEXT:    v_mov_b32_e32 v6, v1
166; GFX10-NEXT:    v_mov_b32_e32 v7, v2
167; GFX10-NEXT:    v_mov_b32_e32 v8, v3
168; GFX10-NEXT:    v_mov_b32_e32 v10, v9
169; GFX10-NEXT:    v_mov_b32_e32 v11, v9
170; GFX10-NEXT:    v_mov_b32_e32 v12, v9
171; GFX10-NEXT:    v_mov_b32_e32 v13, v9
172; GFX10-NEXT:    s_mov_b32 s0, s2
173; GFX10-NEXT:    s_mov_b32 s1, s3
174; GFX10-NEXT:    s_mov_b32 s2, s4
175; GFX10-NEXT:    s_mov_b32 s3, s5
176; GFX10-NEXT:    s_mov_b32 s4, s6
177; GFX10-NEXT:    s_mov_b32 s5, s7
178; GFX10-NEXT:    s_mov_b32 s6, s8
179; GFX10-NEXT:    s_mov_b32 s7, s9
180; GFX10-NEXT:    v_mov_b32_e32 v0, v9
181; GFX10-NEXT:    v_mov_b32_e32 v1, v10
182; GFX10-NEXT:    v_mov_b32_e32 v2, v11
183; GFX10-NEXT:    v_mov_b32_e32 v3, v12
184; GFX10-NEXT:    v_mov_b32_e32 v4, v13
185; 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
186; GFX10-NEXT:    s_waitcnt vmcnt(0)
187; GFX10-NEXT:    global_store_dword v9, v4, s[10:11]
188; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
189; GFX10-NEXT:    ; return to shader part epilog
190;
191; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
192; GFX11:       ; %bb.0:
193; GFX11-NEXT:    v_mov_b32_e32 v9, 0
194; GFX11-NEXT:    v_dual_mov_b32 v5, v0 :: v_dual_mov_b32 v6, v1
195; GFX11-NEXT:    v_dual_mov_b32 v7, v2 :: v_dual_mov_b32 v8, v3
196; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
197; GFX11-NEXT:    v_mov_b32_e32 v11, v9
198; GFX11-NEXT:    v_mov_b32_e32 v13, v9
199; GFX11-NEXT:    v_mov_b32_e32 v12, v9
200; GFX11-NEXT:    v_mov_b32_e32 v10, v9
201; GFX11-NEXT:    s_mov_b32 s0, s2
202; GFX11-NEXT:    s_mov_b32 s1, s3
203; GFX11-NEXT:    s_mov_b32 s2, s4
204; GFX11-NEXT:    s_mov_b32 s3, s5
205; GFX11-NEXT:    s_mov_b32 s4, s6
206; GFX11-NEXT:    s_mov_b32 s5, s7
207; GFX11-NEXT:    s_mov_b32 s6, s8
208; GFX11-NEXT:    s_mov_b32 s7, s9
209; GFX11-NEXT:    v_mov_b32_e32 v0, v9
210; GFX11-NEXT:    v_dual_mov_b32 v2, v11 :: v_dual_mov_b32 v3, v12
211; GFX11-NEXT:    v_dual_mov_b32 v1, v10 :: v_dual_mov_b32 v4, v13
212; 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
213; GFX11-NEXT:    s_waitcnt vmcnt(0)
214; GFX11-NEXT:    global_store_b32 v9, v4, s[10:11]
215; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
216; GFX11-NEXT:    ; return to shader part epilog
217  %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)
218  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
219  %v.err = extractvalue { <4 x float>, i32 } %v, 1
220  store i32 %v.err, i32 addrspace(1)* %out, align 4
221  ret <4 x float> %v.vec
222}
223
224declare <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i32(i32 immarg, i32, i32, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
225declare { <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
226
227attributes #0 = { nounwind readonly }
228