Page MenuHomePhabricator
Differential D64911 Diff 225188 llvm/test/CodeGen/AMDGPU/merge-image-load.mir

Changeset View

Standalone View

View Options

llvm/test/CodeGen/AMDGPU/merge-image-load.mir

  • This file was added.
1​# RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX9 %s
2​
3​# GFX9-LABEL: name: image_load_merged_v1v3
4​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
5​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
6​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
7​
8​name: image_load_merged_v1v3
9​body: |
10​ bb.0.entry:
11​ %0:sgpr_64 = COPY $sgpr0_sgpr1
12​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
13​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
14​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
15​ %4:vgpr_32 = COPY %2.sub3
16​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
17​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
18​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
19​...
20​---
21​# GFX9-LABEL: name: image_load_merged_v1v3_reversed
22​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
23​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub3
24​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub0_sub1_sub2
25​
26​name: image_load_merged_v1v3_reversed
27​body: |
28​ bb.0.entry:
29​ %0:sgpr_64 = COPY $sgpr0_sgpr1
30​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
31​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
32​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
33​ %4:vgpr_32 = COPY %2.sub3
34​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
35​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
36​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
37​...
38​---
39​
40​# GFX9-LABEL: name: image_load_merged_v2v2
41​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, addrspace 4)
42​# GFX9: %{{[0-9]+}}:vreg_64 = COPY %8.sub0_sub1
43​# GFX9: %{{[0-9]+}}:vreg_64 = COPY killed %8.sub2_sub3
44​
45​name: image_load_merged_v2v2
46​body: |
47​ bb.0.entry:
48​ %0:sgpr_64 = COPY $sgpr0_sgpr1
49​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
50​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
51​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
52​ %4:vgpr_32 = COPY %2.sub3
53​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
54​ %6:vreg_64 = IMAGE_LOAD_V2_V4 %5:vreg_128, %3:sreg_256, 3, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
55​ %7:vreg_64 = IMAGE_LOAD_V2_V4 %5:vreg_128, %3:sreg_256, 12, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
56​...
57​---
58​
59​# GFX9-LABEL: name: image_load_merged_v2v2_reversed
60​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, addrspace 4)
61​# GFX9: %{{[0-9]+}}:vreg_64 = COPY %8.sub2_sub3
62​# GFX9: %{{[0-9]+}}:vreg_64 = COPY killed %8.sub0_sub1
63​
64​name: image_load_merged_v2v2_reversed
65​body: |
66​ bb.0.entry:
67​ %0:sgpr_64 = COPY $sgpr0_sgpr1
68​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
69​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
70​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
71​ %4:vgpr_32 = COPY %2.sub3
72​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
73​ %6:vreg_64 = IMAGE_LOAD_V2_V4 %5:vreg_128, %3:sreg_256, 12, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
74​ %7:vreg_64 = IMAGE_LOAD_V2_V4 %5:vreg_128, %3:sreg_256, 3, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
75​...
76​---
77​
78​# GFX9-LABEL: name: image_load_merged_v3v1
79​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, addrspace 4)
80​# GFX9: %{{[0-9]+}}:vreg_96 = COPY %8.sub0_sub1_sub2
81​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %8.sub3
82​
83​name: image_load_merged_v3v1
84​body: |
85​ bb.0.entry:
86​ %0:sgpr_64 = COPY $sgpr0_sgpr1
87​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
88​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
89​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
90​ %4:vgpr_32 = COPY %2.sub3
91​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
92​ %6:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
93​ %7:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
94​...
95​---
96​
97​# GFX9-LABEL: name: image_load_merged_v3v1_reversed
98​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, addrspace 4)
99​# GFX9: %{{[0-9]+}}:vreg_96 = COPY %8.sub1_sub2_sub3
100​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY killed %8.sub0
101​
102​name: image_load_merged_v3v1_reversed
103​body: |
104​ bb.0.entry:
105​ %0:sgpr_64 = COPY $sgpr0_sgpr1
106​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
107​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
108​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
109​ %4:vgpr_32 = COPY %2.sub3
110​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
111​ %6:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
112​ %7:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
113​...
114​---
115​
116​# GFX9-LABEL: name: image_load_divided_merged
117​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
118​
119​name: image_load_divided_merged
120​body: |
121​ bb.0.entry:
122​ %0:sgpr_64 = COPY $sgpr0_sgpr1
123​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
124​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
125​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
126​ %4:vgpr_32 = COPY %2.sub3
127​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
128​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
129​ %7:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
130​ %8:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
131​ %9:vreg_96 = IMAGE_LOAD_V3_V4 %7:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
132​ %10:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
133​ %11:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
134​...
135​---
136​
137​# GFX9-LABEL: name: image_load_divided_not_merged
138​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
139​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
140​
141​name: image_load_divided_not_merged
142​body: |
143​ bb.0.entry:
144​ %0:sgpr_64 = COPY $sgpr0_sgpr1
145​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
146​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
147​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
148​ %4:vreg_128 = COPY %2
149​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
150​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
151​ IMAGE_STORE_V4_V4 %4:vreg_128, %5:vreg_128, %3:sreg_256, 15, -1, 0, 0, 0, 0, 0, 0, 0, implicit $exec
152​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
153​...
154​---
155​
156​# GFX9-LABEL: name: image_load_dmask_overlapped_not_merged
157​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 4, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
158​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
159​
160​name: image_load_dmask_overlapped_not_merged
161​body: |
162​ bb.0.entry:
163​ %0:sgpr_64 = COPY $sgpr0_sgpr1
164​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
165​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
166​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
167​ %4:vgpr_32 = COPY %2.sub3
168​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
169​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 4, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
170​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
171​...
172​---
173​
174​# GFX9-LABEL: name: image_load_dmask_not_disjoint_not_merged
175​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 4, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
176​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 11, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
177​
178​name: image_load_dmask_not_disjoint_not_merged
179​body: |
180​ bb.0.entry:
181​ %0:sgpr_64 = COPY $sgpr0_sgpr1
182​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
183​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
184​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
185​ %4:vgpr_32 = COPY %2.sub3
186​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
187​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5:vreg_128, %3:sreg_256, 4, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
188​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5:vreg_128, %3:sreg_256, 11, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
189​...
190​---
191​
192​# GFX9-LABEL: name: image_load_not_merged_0
193​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
194​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %6, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
195​
196​name: image_load_not_merged_0
197​body: |
198​ bb.0.entry:
199​ %0:sgpr_64 = COPY $sgpr0_sgpr1
200​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
201​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
202​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
203​ %4:vgpr_32 = COPY %2.sub3
204​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
205​ %6:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
206​ %7:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
207​ %8:vreg_96 = IMAGE_LOAD_V3_V4 %6, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
208​...
209​---
210​
211​# GFX9-LABEL: name: image_load_not_merged_1
212​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %6, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
213​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %6, %4, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
214​
215​name: image_load_not_merged_1
216​body: |
217​ bb.0.entry:
218​ %0:sgpr_64 = COPY $sgpr0_sgpr1
219​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
220​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
221​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
222​ %4:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
223​ %5:vgpr_32 = COPY %2.sub3
224​ %6:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
225​ %7:vgpr_32 = IMAGE_LOAD_V1_V4 %6, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
226​ %8:vreg_96 = IMAGE_LOAD_V3_V4 %6, %4, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
227​...
228​---
229​
230​# GFX9-LABEL: name: image_load_not_merged_10
231​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
232​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
233​
234​name: image_load_not_merged_10
235​body: |
236​ bb.0.entry:
237​ %0:sgpr_64 = COPY $sgpr0_sgpr1
238​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
239​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
240​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
241​ %4:vgpr_32 = COPY %2.sub3
242​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
243​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
244​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
245​...
246​---
247​
248​# GFX9-LABEL: name: image_load_not_merged_3
249​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
250​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 1, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
251​
252​name: image_load_not_merged_3
253​body: |
254​ bb.0.entry:
255​ %0:sgpr_64 = COPY $sgpr0_sgpr1
256​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
257​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
258​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
259​ %4:vgpr_32 = COPY %2.sub3
260​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
261​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
262​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 1, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
263​...
264​---
265​
266​# GFX9-LABEL: name: image_load_not_merged_4
267​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
268​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 1, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
269​
270​name: image_load_not_merged_4
271​body: |
272​ bb.0.entry:
273​ %0:sgpr_64 = COPY $sgpr0_sgpr1
274​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
275​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
276​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
277​ %4:vgpr_32 = COPY %2.sub3
278​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
279​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
280​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 1, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
281​...
282​---
283​
284​# GFX9-LABEL: name: image_load_not_merged_5
285​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
286​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 1, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
287​
288​name: image_load_not_merged_5
289​body: |
290​ bb.0.entry:
291​ %0:sgpr_64 = COPY $sgpr0_sgpr1
292​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
293​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
294​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
295​ %4:vgpr_32 = COPY %2.sub3
296​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
297​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
298​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 1, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
299​...
300​---
301​
302​# GFX9-LABEL: name: image_load_not_merged_6
303​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
304​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 1, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
305​
306​name: image_load_not_merged_6
307​body: |
308​ bb.0.entry:
309​ %0:sgpr_64 = COPY $sgpr0_sgpr1
310​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
311​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
312​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
313​ %4:vgpr_32 = COPY %2.sub3
314​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
315​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
316​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 1, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
317​...
318​---
319​
320​# GFX9-LABEL: name: image_load_not_merged_7
321​# GFX9: %{{[0-9]+}}:vreg_64 = IMAGE_LOAD_V2_V4 %5, %3, 8, 0, 0, 0, 0, 1, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
322​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
323​
324​name: image_load_not_merged_7
325​body: |
326​ bb.0.entry:
327​ %0:sgpr_64 = COPY $sgpr0_sgpr1
328​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
329​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
330​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
331​ %4:vgpr_32 = COPY %2.sub3
332​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
333​ %6:vreg_64 = IMAGE_LOAD_V2_V4 %5, %3, 8, 0, 0, 0, 0, 1, 0, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
334​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
335​...
336​---
337​
338​# GFX9-LABEL: name: image_load_not_merged_8
339​# GFX9: %{{[0-9]+}}:vreg_64 = IMAGE_LOAD_V2_V4 %5, %3, 8, 0, 0, 0, 0, 0, 1, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
340​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
341​
342​name: image_load_not_merged_8
343​body: |
344​ bb.0.entry:
345​ %0:sgpr_64 = COPY $sgpr0_sgpr1
346​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
347​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
348​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
349​ %4:vgpr_32 = COPY %2.sub3
350​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
351​ %6:vreg_64 = IMAGE_LOAD_V2_V4 %5, %3, 8, 0, 0, 0, 0, 0, 1, -1, 0, implicit $exec :: (dereferenceable load 8, align 16, addrspace 4)
352​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
353​...
354​---
355​
356​# GFX9-LABEL: name: image_load_not_merged_9
357​# GFX9: %{{[0-9]+}}:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
358​# GFX9: %{{[0-9]+}}:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
359​
360​name: image_load_not_merged_9
361​body: |
362​ bb.0.entry:
363​ %0:sgpr_64 = COPY $sgpr0_sgpr1
364​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
365​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
366​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
367​ %4:vgpr_32 = COPY %2.sub3
368​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
369​ %6:vgpr_32 = IMAGE_LOAD_V1_V4 %5, %3, 8, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
370​ %7:vreg_96 = IMAGE_LOAD_V3_V4 %5, %3, 7, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
371​...
372​---
373​
374​# GFX9-LABEL: name: image_load_mip_merged_v1v3
375​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
376​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
377​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
378​
379​name: image_load_mip_merged_v1v3
380​body: |
381​ bb.0.entry:
382​ %0:sgpr_64 = COPY $sgpr0_sgpr1
383​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
384​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
385​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
386​ %4:vgpr_32 = COPY %2.sub3
387​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
388​ %6:vgpr_32 = IMAGE_LOAD_MIP_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 4, addrspace 4)
389​ %7:vreg_96 = IMAGE_LOAD_MIP_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, 0, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
390​...
391​---
392​
393​
394​
395​# GFX9-LABEL: name: image_load_mip_pck_merged_v1v3
396​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_PCK_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
397​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
398​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
399​
400​name: image_load_mip_pck_merged_v1v3
401​body: |
402​ bb.0.entry:
403​ %0:sgpr_64 = COPY $sgpr0_sgpr1
404​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
405​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
406​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
407​ %4:vgpr_32 = COPY %2.sub3
408​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
409​ %6:vgpr_32 = IMAGE_LOAD_MIP_PCK_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 4, addrspace 4)
410​ %7:vreg_96 = IMAGE_LOAD_MIP_PCK_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
411​...
412​---
413​
414​
415​
416​# GFX9-LABEL: name: image_load_mip_pck_sgn_merged_v1v3
417​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_MIP_PCK_SGN_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
418​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
419​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
420​
421​name: image_load_mip_pck_sgn_merged_v1v3
422​body: |
423​ bb.0.entry:
424​ %0:sgpr_64 = COPY $sgpr0_sgpr1
425​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
426​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
427​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
428​ %4:vgpr_32 = COPY %2.sub3
429​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
430​ %6:vgpr_32 = IMAGE_LOAD_MIP_PCK_SGN_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 4, addrspace 4)
431​ %7:vreg_96 = IMAGE_LOAD_MIP_PCK_SGN_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
432​...
433​---
434​
435​# GFX9-LABEL: name: image_load_pck_merged_v1v3
436​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_PCK_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
437​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
438​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
439​
440​name: image_load_pck_merged_v1v3
441​body: |
442​ bb.0.entry:
443​ %0:sgpr_64 = COPY $sgpr0_sgpr1
444​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
445​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
446​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
447​ %4:vgpr_32 = COPY %2.sub3
448​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
449​ %6:vgpr_32 = IMAGE_LOAD_PCK_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 4, addrspace 4)
450​ %7:vreg_96 = IMAGE_LOAD_PCK_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
451​...
452​---
453​
454​# GFX9-LABEL: name: image_load_pck_sgn_merged_v1v3
455​# GFX9: %{{[0-9]+}}:vreg_128 = IMAGE_LOAD_PCK_SGN_V4_V4 %5, %3, 15, 0, 0, 0, 0, 0, 0, -1, implicit $exec, implicit $exec :: (dereferenceable load 16, align 4, addrspace 4)
456​# GFX9: %{{[0-9]+}}:vgpr_32 = COPY %8.sub0
457​# GFX9: %{{[0-9]+}}:vreg_96 = COPY killed %8.sub1_sub2_sub3
458​
459​name: image_load_pck_sgn_merged_v1v3
460​body: |
461​ bb.0.entry:
462​ %0:sgpr_64 = COPY $sgpr0_sgpr1
463​ %1:sreg_64_xexec = S_LOAD_DWORDX2_IMM %0, 36, 0, 0
464​ %2:sgpr_128 =COPY $sgpr96_sgpr97_sgpr98_sgpr99
465​ %3:sreg_256 = S_LOAD_DWORDX8_IMM %1, 208, 0, 0
466​ %4:vgpr_32 = COPY %2.sub3
467​ %5:vreg_128 = BUFFER_LOAD_DWORDX4_OFFSET %2:sgpr_128, 0, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable invariant load 16)
468​ %6:vgpr_32 = IMAGE_LOAD_PCK_SGN_V1_V4 %5:vreg_128, %3:sreg_256, 1, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 4, addrspace 4)
469​ %7:vreg_96 = IMAGE_LOAD_PCK_SGN_V3_V4 %5:vreg_128, %3:sreg_256, 14, 0, 0, 0, 0, 0, 0, -1, implicit $exec :: (dereferenceable load 12, align 16, addrspace 4)
470​...
471​---