1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefix=GFX9 %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 -amdgpu-enable-vopd=0 -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, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
7; GFX9-LABEL: load_2darraymsaa_v4f32_xyzw:
8; GFX9:       ; %bb.0:
9; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
10; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
11; GFX9-NEXT:    v_and_or_b32 v0, v0, v4, v1
12; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 16, v3
13; GFX9-NEXT:    s_mov_b32 s0, s2
14; GFX9-NEXT:    s_mov_b32 s1, s3
15; GFX9-NEXT:    s_mov_b32 s2, s4
16; GFX9-NEXT:    s_mov_b32 s3, s5
17; GFX9-NEXT:    s_mov_b32 s4, s6
18; GFX9-NEXT:    s_mov_b32 s5, s7
19; GFX9-NEXT:    s_mov_b32 s6, s8
20; GFX9-NEXT:    s_mov_b32 s7, s9
21; GFX9-NEXT:    v_and_or_b32 v1, v2, v4, v1
22; GFX9-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf unorm a16 da
23; GFX9-NEXT:    s_waitcnt vmcnt(0)
24; GFX9-NEXT:    ; return to shader part epilog
25;
26; GFX10PLUS-LABEL: load_2darraymsaa_v4f32_xyzw:
27; GFX10PLUS:       ; %bb.0:
28; GFX10PLUS-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
29; GFX10PLUS-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
30; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
31; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
32; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
33; GFX10PLUS-NEXT:    v_and_or_b32 v0, 0xffff, v0, v1
34; GFX10PLUS-NEXT:    v_and_or_b32 v1, 0xffff, v2, v3
35; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
36; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
37; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
38; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
39; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
40; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
41; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
42; GFX10PLUS-NEXT:    ; return to shader part epilog
43  %v = call <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i16(i32 15, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
44  ret <4 x float> %v
45}
46
47define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
48; GFX9-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
49; GFX9:       ; %bb.0:
50; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
51; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
52; GFX9-NEXT:    v_and_or_b32 v10, v0, v4, v1
53; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
54; GFX9-NEXT:    v_mov_b32_e32 v5, 0
55; GFX9-NEXT:    v_and_or_b32 v11, v2, v4, v0
56; GFX9-NEXT:    v_mov_b32_e32 v6, v5
57; GFX9-NEXT:    v_mov_b32_e32 v7, v5
58; GFX9-NEXT:    v_mov_b32_e32 v8, v5
59; GFX9-NEXT:    v_mov_b32_e32 v9, v5
60; GFX9-NEXT:    v_mov_b32_e32 v0, v5
61; GFX9-NEXT:    s_mov_b32 s0, s2
62; GFX9-NEXT:    s_mov_b32 s1, s3
63; GFX9-NEXT:    s_mov_b32 s2, s4
64; GFX9-NEXT:    s_mov_b32 s3, s5
65; GFX9-NEXT:    s_mov_b32 s4, s6
66; GFX9-NEXT:    s_mov_b32 s5, s7
67; GFX9-NEXT:    s_mov_b32 s6, s8
68; GFX9-NEXT:    s_mov_b32 s7, s9
69; GFX9-NEXT:    v_mov_b32_e32 v1, v6
70; GFX9-NEXT:    v_mov_b32_e32 v2, v7
71; GFX9-NEXT:    v_mov_b32_e32 v3, v8
72; GFX9-NEXT:    v_mov_b32_e32 v4, v9
73; GFX9-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf unorm a16 tfe da
74; GFX9-NEXT:    s_waitcnt vmcnt(0)
75; GFX9-NEXT:    global_store_dword v5, v4, s[10:11]
76; GFX9-NEXT:    s_waitcnt vmcnt(0)
77; GFX9-NEXT:    ; return to shader part epilog
78;
79; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
80; GFX10:       ; %bb.0:
81; GFX10-NEXT:    v_mov_b32_e32 v5, 0
82; GFX10-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
83; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
84; GFX10-NEXT:    s_mov_b32 s0, s2
85; GFX10-NEXT:    s_mov_b32 s1, s3
86; GFX10-NEXT:    v_mov_b32_e32 v6, v5
87; GFX10-NEXT:    v_mov_b32_e32 v7, v5
88; GFX10-NEXT:    v_mov_b32_e32 v8, v5
89; GFX10-NEXT:    v_mov_b32_e32 v9, v5
90; GFX10-NEXT:    v_and_or_b32 v10, 0xffff, v0, v1
91; GFX10-NEXT:    v_and_or_b32 v11, 0xffff, v2, v3
92; GFX10-NEXT:    s_mov_b32 s2, s4
93; GFX10-NEXT:    s_mov_b32 s3, s5
94; GFX10-NEXT:    s_mov_b32 s4, s6
95; GFX10-NEXT:    s_mov_b32 s5, s7
96; GFX10-NEXT:    s_mov_b32 s6, s8
97; GFX10-NEXT:    s_mov_b32 s7, s9
98; GFX10-NEXT:    v_mov_b32_e32 v0, v5
99; GFX10-NEXT:    v_mov_b32_e32 v1, v6
100; GFX10-NEXT:    v_mov_b32_e32 v2, v7
101; GFX10-NEXT:    v_mov_b32_e32 v3, v8
102; GFX10-NEXT:    v_mov_b32_e32 v4, v9
103; GFX10-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16 tfe
104; GFX10-NEXT:    s_waitcnt vmcnt(0)
105; GFX10-NEXT:    global_store_dword v5, v4, s[10:11]
106; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
107; GFX10-NEXT:    ; return to shader part epilog
108;
109; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe:
110; GFX11:       ; %bb.0:
111; GFX11-NEXT:    v_mov_b32_e32 v5, 0
112; GFX11-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
113; GFX11-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
114; GFX11-NEXT:    s_mov_b32 s0, s2
115; GFX11-NEXT:    s_mov_b32 s1, s3
116; GFX11-NEXT:    v_mov_b32_e32 v6, v5
117; GFX11-NEXT:    v_mov_b32_e32 v7, v5
118; GFX11-NEXT:    v_mov_b32_e32 v8, v5
119; GFX11-NEXT:    v_mov_b32_e32 v9, v5
120; GFX11-NEXT:    v_and_or_b32 v10, 0xffff, v0, v1
121; GFX11-NEXT:    v_and_or_b32 v11, 0xffff, v2, v3
122; GFX11-NEXT:    s_mov_b32 s2, s4
123; GFX11-NEXT:    s_mov_b32 s3, s5
124; GFX11-NEXT:    s_mov_b32 s4, s6
125; GFX11-NEXT:    s_mov_b32 s5, s7
126; GFX11-NEXT:    s_mov_b32 s6, s8
127; GFX11-NEXT:    s_mov_b32 s7, s9
128; GFX11-NEXT:    v_mov_b32_e32 v0, v5
129; GFX11-NEXT:    v_mov_b32_e32 v1, v6
130; GFX11-NEXT:    v_mov_b32_e32 v2, v7
131; GFX11-NEXT:    v_mov_b32_e32 v3, v8
132; GFX11-NEXT:    v_mov_b32_e32 v4, v9
133; GFX11-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16 tfe
134; GFX11-NEXT:    s_waitcnt vmcnt(0)
135; GFX11-NEXT:    global_store_b32 v5, v4, s[10:11]
136; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
137; GFX11-NEXT:    ; return to shader part epilog
138  %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i16(i32 15, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 1, i32 0)
139  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
140  %v.err = extractvalue { <4 x float>, i32 } %v, 1
141  store i32 %v.err, i32 addrspace(1)* %out, align 4
142  ret <4 x float> %v.vec
143}
144
145define amdgpu_ps <4 x float> @load_2darraymsaa_v4f32_xyzw_tfe_lwe(<8 x i32> inreg %rsrc, i32 addrspace(1)* inreg %out, i16 %s, i16 %t, i16 %slice, i16 %fragid) {
146; GFX9-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
147; GFX9:       ; %bb.0:
148; GFX9-NEXT:    v_mov_b32_e32 v4, 0xffff
149; GFX9-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
150; GFX9-NEXT:    v_and_or_b32 v10, v0, v4, v1
151; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 16, v3
152; GFX9-NEXT:    v_mov_b32_e32 v5, 0
153; GFX9-NEXT:    v_and_or_b32 v11, v2, v4, v0
154; GFX9-NEXT:    v_mov_b32_e32 v6, v5
155; GFX9-NEXT:    v_mov_b32_e32 v7, v5
156; GFX9-NEXT:    v_mov_b32_e32 v8, v5
157; GFX9-NEXT:    v_mov_b32_e32 v9, v5
158; GFX9-NEXT:    v_mov_b32_e32 v0, v5
159; GFX9-NEXT:    s_mov_b32 s0, s2
160; GFX9-NEXT:    s_mov_b32 s1, s3
161; GFX9-NEXT:    s_mov_b32 s2, s4
162; GFX9-NEXT:    s_mov_b32 s3, s5
163; GFX9-NEXT:    s_mov_b32 s4, s6
164; GFX9-NEXT:    s_mov_b32 s5, s7
165; GFX9-NEXT:    s_mov_b32 s6, s8
166; GFX9-NEXT:    s_mov_b32 s7, s9
167; GFX9-NEXT:    v_mov_b32_e32 v1, v6
168; GFX9-NEXT:    v_mov_b32_e32 v2, v7
169; GFX9-NEXT:    v_mov_b32_e32 v3, v8
170; GFX9-NEXT:    v_mov_b32_e32 v4, v9
171; GFX9-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf unorm a16 tfe lwe da
172; GFX9-NEXT:    s_waitcnt vmcnt(0)
173; GFX9-NEXT:    global_store_dword v5, v4, s[10:11]
174; GFX9-NEXT:    s_waitcnt vmcnt(0)
175; GFX9-NEXT:    ; return to shader part epilog
176;
177; GFX10-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
178; GFX10:       ; %bb.0:
179; GFX10-NEXT:    v_mov_b32_e32 v5, 0
180; GFX10-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
181; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
182; GFX10-NEXT:    s_mov_b32 s0, s2
183; GFX10-NEXT:    s_mov_b32 s1, s3
184; GFX10-NEXT:    v_mov_b32_e32 v6, v5
185; GFX10-NEXT:    v_mov_b32_e32 v7, v5
186; GFX10-NEXT:    v_mov_b32_e32 v8, v5
187; GFX10-NEXT:    v_mov_b32_e32 v9, v5
188; GFX10-NEXT:    v_and_or_b32 v10, 0xffff, v0, v1
189; GFX10-NEXT:    v_and_or_b32 v11, 0xffff, v2, v3
190; GFX10-NEXT:    s_mov_b32 s2, s4
191; GFX10-NEXT:    s_mov_b32 s3, s5
192; GFX10-NEXT:    s_mov_b32 s4, s6
193; GFX10-NEXT:    s_mov_b32 s5, s7
194; GFX10-NEXT:    s_mov_b32 s6, s8
195; GFX10-NEXT:    s_mov_b32 s7, s9
196; GFX10-NEXT:    v_mov_b32_e32 v0, v5
197; GFX10-NEXT:    v_mov_b32_e32 v1, v6
198; GFX10-NEXT:    v_mov_b32_e32 v2, v7
199; GFX10-NEXT:    v_mov_b32_e32 v3, v8
200; GFX10-NEXT:    v_mov_b32_e32 v4, v9
201; GFX10-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16 tfe lwe
202; GFX10-NEXT:    s_waitcnt vmcnt(0)
203; GFX10-NEXT:    global_store_dword v5, v4, s[10:11]
204; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
205; GFX10-NEXT:    ; return to shader part epilog
206;
207; GFX11-LABEL: load_2darraymsaa_v4f32_xyzw_tfe_lwe:
208; GFX11:       ; %bb.0:
209; GFX11-NEXT:    v_mov_b32_e32 v5, 0
210; GFX11-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
211; GFX11-NEXT:    v_lshlrev_b32_e32 v3, 16, v3
212; GFX11-NEXT:    s_mov_b32 s0, s2
213; GFX11-NEXT:    s_mov_b32 s1, s3
214; GFX11-NEXT:    v_mov_b32_e32 v6, v5
215; GFX11-NEXT:    v_mov_b32_e32 v7, v5
216; GFX11-NEXT:    v_mov_b32_e32 v8, v5
217; GFX11-NEXT:    v_mov_b32_e32 v9, v5
218; GFX11-NEXT:    v_and_or_b32 v10, 0xffff, v0, v1
219; GFX11-NEXT:    v_and_or_b32 v11, 0xffff, v2, v3
220; GFX11-NEXT:    s_mov_b32 s2, s4
221; GFX11-NEXT:    s_mov_b32 s3, s5
222; GFX11-NEXT:    s_mov_b32 s4, s6
223; GFX11-NEXT:    s_mov_b32 s5, s7
224; GFX11-NEXT:    s_mov_b32 s6, s8
225; GFX11-NEXT:    s_mov_b32 s7, s9
226; GFX11-NEXT:    v_mov_b32_e32 v0, v5
227; GFX11-NEXT:    v_mov_b32_e32 v1, v6
228; GFX11-NEXT:    v_mov_b32_e32 v2, v7
229; GFX11-NEXT:    v_mov_b32_e32 v3, v8
230; GFX11-NEXT:    v_mov_b32_e32 v4, v9
231; GFX11-NEXT:    image_load v[0:4], v[10:11], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16 tfe lwe
232; GFX11-NEXT:    s_waitcnt vmcnt(0)
233; GFX11-NEXT:    global_store_b32 v5, v4, s[10:11]
234; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
235; GFX11-NEXT:    ; return to shader part epilog
236  %v = call { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i16(i32 15, i16 %s, i16 %t, i16 %slice, i16 %fragid, <8 x i32> %rsrc, i32 3, i32 0)
237  %v.vec = extractvalue { <4 x float>, i32 } %v, 0
238  %v.err = extractvalue { <4 x float>, i32 } %v, 1
239  store i32 %v.err, i32 addrspace(1)* %out, align 4
240  ret <4 x float> %v.vec
241}
242
243declare <4 x float> @llvm.amdgcn.image.load.2darraymsaa.v4f32.i16(i32 immarg, i16, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
244declare { <4 x float>, i32 } @llvm.amdgcn.image.load.2darraymsaa.sl_v4f32i32s.i16(i32 immarg, i16, i16, i16, i16, <8 x i32>, i32 immarg, i32 immarg) #0
245
246attributes #0 = { nounwind readonly }
247