1; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -stop-after=legalizer -o - %s | FileCheck -check-prefix=UNPACKED %s
3; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx810 -stop-after=legalizer -o - %s | FileCheck -check-prefix=PACKED %s
4
5define amdgpu_ps void @image_store_f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, half %data) {
6  ; UNPACKED-LABEL: name: image_store_f16
7  ; UNPACKED: bb.1 (%ir-block.0):
8  ; UNPACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2
9  ; UNPACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
10  ; UNPACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
11  ; UNPACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
12  ; UNPACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
13  ; UNPACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
14  ; UNPACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
15  ; UNPACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
16  ; UNPACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
17  ; UNPACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
18  ; UNPACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
19  ; UNPACKED:   [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
20  ; UNPACKED:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
21  ; UNPACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
22  ; UNPACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
23  ; UNPACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[TRUNC]](s16), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 2 into custom "TargetCustom8")
24  ; UNPACKED:   S_ENDPGM 0
25  ; PACKED-LABEL: name: image_store_f16
26  ; PACKED: bb.1 (%ir-block.0):
27  ; PACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2
28  ; PACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
29  ; PACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
30  ; PACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
31  ; PACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
32  ; PACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
33  ; PACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
34  ; PACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
35  ; PACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
36  ; PACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
37  ; PACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
38  ; PACKED:   [[COPY10:%[0-9]+]]:_(s32) = COPY $vgpr2
39  ; PACKED:   [[TRUNC:%[0-9]+]]:_(s16) = G_TRUNC [[COPY10]](s32)
40  ; PACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
41  ; PACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
42  ; PACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[TRUNC]](s16), 1, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 2 into custom "TargetCustom8")
43  ; PACKED:   S_ENDPGM 0
44  call void @llvm.amdgcn.image.store.2d.f16.i32(half %data, i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
45  ret void
46}
47
48define amdgpu_ps void @image_store_v2f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <2 x half> %in) {
49  ; UNPACKED-LABEL: name: image_store_v2f16
50  ; UNPACKED: bb.1 (%ir-block.0):
51  ; UNPACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2
52  ; UNPACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
53  ; UNPACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
54  ; UNPACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
55  ; UNPACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
56  ; UNPACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
57  ; UNPACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
58  ; UNPACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
59  ; UNPACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
60  ; UNPACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
61  ; UNPACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
62  ; UNPACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
63  ; UNPACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
64  ; UNPACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
65  ; UNPACKED:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
66  ; UNPACKED:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
67  ; UNPACKED:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
68  ; UNPACKED:   [[COPY11:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
69  ; UNPACKED:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
70  ; UNPACKED:   [[BUILD_VECTOR2:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY11]](s32), [[COPY12]](s32)
71  ; UNPACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR2]](<2 x s32>), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 4 into custom "TargetCustom8")
72  ; UNPACKED:   S_ENDPGM 0
73  ; PACKED-LABEL: name: image_store_v2f16
74  ; PACKED: bb.1 (%ir-block.0):
75  ; PACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2
76  ; PACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
77  ; PACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
78  ; PACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
79  ; PACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
80  ; PACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
81  ; PACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
82  ; PACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
83  ; PACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
84  ; PACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
85  ; PACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
86  ; PACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
87  ; PACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
88  ; PACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
89  ; PACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[COPY10]](<2 x s16>), 3, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 4 into custom "TargetCustom8")
90  ; PACKED:   S_ENDPGM 0
91  call void @llvm.amdgcn.image.store.2d.v2f16.i32(<2 x half> %in, i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
92  ret void
93}
94
95define amdgpu_ps void @image_store_v3f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <3 x half> %in) {
96  ; UNPACKED-LABEL: name: image_store_v3f16
97  ; UNPACKED: bb.1 (%ir-block.0):
98  ; UNPACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
99  ; UNPACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
100  ; UNPACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
101  ; UNPACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
102  ; UNPACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
103  ; UNPACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
104  ; UNPACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
105  ; UNPACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
106  ; UNPACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
107  ; UNPACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
108  ; UNPACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
109  ; UNPACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
110  ; UNPACKED:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
111  ; UNPACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
112  ; UNPACKED:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
113  ; UNPACKED:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
114  ; UNPACKED:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
115  ; UNPACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
116  ; UNPACKED:   [[DEF1:%[0-9]+]]:_(<4 x s16>) = G_IMPLICIT_DEF
117  ; UNPACKED:   [[INSERT:%[0-9]+]]:_(<4 x s16>) = G_INSERT [[DEF1]], [[UV]](<3 x s16>), 0
118  ; UNPACKED:   [[UV2:%[0-9]+]]:_(<2 x s16>), [[UV3:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[INSERT]](<4 x s16>)
119  ; UNPACKED:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[UV2]](<2 x s16>)
120  ; UNPACKED:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
121  ; UNPACKED:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
122  ; UNPACKED:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[UV3]](<2 x s16>)
123  ; UNPACKED:   [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
124  ; UNPACKED:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
125  ; UNPACKED:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
126  ; UNPACKED:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
127  ; UNPACKED:   [[BUILD_VECTOR2:%[0-9]+]]:_(<3 x s32>) = G_BUILD_VECTOR [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32)
128  ; UNPACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR2]](<3 x s32>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 6 into custom "TargetCustom8", align 8)
129  ; UNPACKED:   S_ENDPGM 0
130  ; PACKED-LABEL: name: image_store_v3f16
131  ; PACKED: bb.1 (%ir-block.0):
132  ; PACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
133  ; PACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
134  ; PACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
135  ; PACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
136  ; PACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
137  ; PACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
138  ; PACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
139  ; PACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
140  ; PACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
141  ; PACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
142  ; PACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
143  ; PACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
144  ; PACKED:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
145  ; PACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
146  ; PACKED:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
147  ; PACKED:   [[CONCAT_VECTORS:%[0-9]+]]:_(<6 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>), [[DEF]](<2 x s16>)
148  ; PACKED:   [[UV:%[0-9]+]]:_(<3 x s16>), [[UV1:%[0-9]+]]:_(<3 x s16>) = G_UNMERGE_VALUES [[CONCAT_VECTORS]](<6 x s16>)
149  ; PACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
150  ; PACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[UV]](<3 x s16>), 7, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 6 into custom "TargetCustom8", align 8)
151  ; PACKED:   S_ENDPGM 0
152  call void @llvm.amdgcn.image.store.2d.v3f16.i32(<3 x half> %in, i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
153  ret void
154}
155
156define amdgpu_ps void @image_store_v4f16(<8 x i32> inreg %rsrc, i32 %s, i32 %t, <4 x half> %in) {
157  ; UNPACKED-LABEL: name: image_store_v4f16
158  ; UNPACKED: bb.1 (%ir-block.0):
159  ; UNPACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
160  ; UNPACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
161  ; UNPACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
162  ; UNPACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
163  ; UNPACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
164  ; UNPACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
165  ; UNPACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
166  ; UNPACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
167  ; UNPACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
168  ; UNPACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
169  ; UNPACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
170  ; UNPACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
171  ; UNPACKED:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
172  ; UNPACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
173  ; UNPACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
174  ; UNPACKED:   [[BITCAST:%[0-9]+]]:_(s32) = G_BITCAST [[COPY10]](<2 x s16>)
175  ; UNPACKED:   [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 16
176  ; UNPACKED:   [[LSHR:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST]], [[C]](s32)
177  ; UNPACKED:   [[BITCAST1:%[0-9]+]]:_(s32) = G_BITCAST [[COPY11]](<2 x s16>)
178  ; UNPACKED:   [[LSHR1:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST1]], [[C]](s32)
179  ; UNPACKED:   [[COPY12:%[0-9]+]]:_(s32) = COPY [[BITCAST]](s32)
180  ; UNPACKED:   [[COPY13:%[0-9]+]]:_(s32) = COPY [[LSHR]](s32)
181  ; UNPACKED:   [[COPY14:%[0-9]+]]:_(s32) = COPY [[BITCAST1]](s32)
182  ; UNPACKED:   [[COPY15:%[0-9]+]]:_(s32) = COPY [[LSHR1]](s32)
183  ; UNPACKED:   [[BUILD_VECTOR2:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[COPY12]](s32), [[COPY13]](s32), [[COPY14]](s32), [[COPY15]](s32)
184  ; UNPACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[BUILD_VECTOR2]](<4 x s32>), 15, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 8 into custom "TargetCustom8")
185  ; UNPACKED:   S_ENDPGM 0
186  ; PACKED-LABEL: name: image_store_v4f16
187  ; PACKED: bb.1 (%ir-block.0):
188  ; PACKED:   liveins: $sgpr2, $sgpr3, $sgpr4, $sgpr5, $sgpr6, $sgpr7, $sgpr8, $sgpr9, $vgpr0, $vgpr1, $vgpr2, $vgpr3
189  ; PACKED:   [[COPY:%[0-9]+]]:_(s32) = COPY $sgpr2
190  ; PACKED:   [[COPY1:%[0-9]+]]:_(s32) = COPY $sgpr3
191  ; PACKED:   [[COPY2:%[0-9]+]]:_(s32) = COPY $sgpr4
192  ; PACKED:   [[COPY3:%[0-9]+]]:_(s32) = COPY $sgpr5
193  ; PACKED:   [[COPY4:%[0-9]+]]:_(s32) = COPY $sgpr6
194  ; PACKED:   [[COPY5:%[0-9]+]]:_(s32) = COPY $sgpr7
195  ; PACKED:   [[COPY6:%[0-9]+]]:_(s32) = COPY $sgpr8
196  ; PACKED:   [[COPY7:%[0-9]+]]:_(s32) = COPY $sgpr9
197  ; PACKED:   [[COPY8:%[0-9]+]]:_(s32) = COPY $vgpr0
198  ; PACKED:   [[COPY9:%[0-9]+]]:_(s32) = COPY $vgpr1
199  ; PACKED:   [[COPY10:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr2
200  ; PACKED:   [[COPY11:%[0-9]+]]:_(<2 x s16>) = COPY $vgpr3
201  ; PACKED:   [[BUILD_VECTOR:%[0-9]+]]:_(<8 x s32>) = G_BUILD_VECTOR [[COPY]](s32), [[COPY1]](s32), [[COPY2]](s32), [[COPY3]](s32), [[COPY4]](s32), [[COPY5]](s32), [[COPY6]](s32), [[COPY7]](s32)
202  ; PACKED:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY10]](<2 x s16>), [[COPY11]](<2 x s16>)
203  ; PACKED:   [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY8]](s32), [[COPY9]](s32)
204  ; PACKED:   G_AMDGPU_INTRIN_IMAGE_STORE intrinsic(@llvm.amdgcn.image.store.2d), [[CONCAT_VECTORS]](<4 x s16>), 15, [[BUILD_VECTOR1]](<2 x s32>), $noreg, [[BUILD_VECTOR]](<8 x s32>), 0, 0, 0 :: (dereferenceable store 8 into custom "TargetCustom8")
205  ; PACKED:   S_ENDPGM 0
206  call void @llvm.amdgcn.image.store.2d.v4f16.i32(<4 x half> %in, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
207  ret void
208}
209
210declare void @llvm.amdgcn.image.store.2d.f16.i32(half, i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
211declare void @llvm.amdgcn.image.store.2d.v2f16.i32(<2 x half>, i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
212declare void @llvm.amdgcn.image.store.2d.v3f16.i32(<3 x half>, i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
213declare void @llvm.amdgcn.image.store.2d.v4f16.i32(<4 x half>, i32 immarg, i32, i32, <8 x i32>, i32 immarg, i32 immarg) #0
214
215attributes #0 = { nounwind writeonly }
216