1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes --check-globals
2 ; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | opt -S -mtriple=amdgcn-unknown-unknown -passes=amdgpu-attributor | FileCheck -check-prefixes=CHECK,V4 %s
3 ; RUN: sed 's/CODE_OBJECT_VERSION/500/g' %s | opt -S -mtriple=amdgcn-unknown-unknown -passes=amdgpu-attributor | FileCheck -check-prefixes=CHECK,V5 %s
4 ; RUN: sed 's/CODE_OBJECT_VERSION/600/g' %s | opt -S -mtriple=amdgcn-unknown-unknown -passes=amdgpu-attributor | FileCheck -check-prefixes=CHECK,V6 %s
6 declare ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr() #0
8 declare i32 @llvm.amdgcn.workgroup.id.x() #0
9 declare i32 @llvm.amdgcn.workgroup.id.y() #0
10 declare i32 @llvm.amdgcn.workgroup.id.z() #0
12 declare i32 @llvm.amdgcn.workitem.id.x() #0
13 declare i32 @llvm.amdgcn.workitem.id.y() #0
14 declare i32 @llvm.amdgcn.workitem.id.z() #0
15 declare i32 @llvm.amdgcn.lds.kernel.id() #0
16 declare i64 @llvm.amdgcn.dispatch.id() #0
19 declare ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #0
20 declare ptr addrspace(4) @llvm.amdgcn.queue.ptr() #0
21 declare ptr addrspace(4) @llvm.amdgcn.kernarg.segment.ptr() #0
23 ; Avoid adding all of these to the output attribute sets
24 define void @use_everything_else() {
25 ; CHECK-LABEL: define {{[^@]+}}@use_everything_else
26 ; CHECK-SAME: () #[[ATTR1:[0-9]+]] {
27 ; CHECK-NEXT: [[VAL0:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
28 ; CHECK-NEXT: [[VAL1:%.*]] = call i32 @llvm.amdgcn.workitem.id.y()
29 ; CHECK-NEXT: [[VAL2:%.*]] = call i32 @llvm.amdgcn.workitem.id.z()
30 ; CHECK-NEXT: [[VAL3:%.*]] = call i32 @llvm.amdgcn.workgroup.id.x()
31 ; CHECK-NEXT: [[VAL4:%.*]] = call i32 @llvm.amdgcn.workgroup.id.y()
32 ; CHECK-NEXT: [[VAL5:%.*]] = call i32 @llvm.amdgcn.workgroup.id.z()
33 ; CHECK-NEXT: store volatile i32 [[VAL0]], ptr addrspace(1) null, align 4
34 ; CHECK-NEXT: store volatile i32 [[VAL1]], ptr addrspace(1) null, align 4
35 ; CHECK-NEXT: store volatile i32 [[VAL2]], ptr addrspace(1) null, align 4
36 ; CHECK-NEXT: store volatile i32 [[VAL3]], ptr addrspace(1) null, align 4
37 ; CHECK-NEXT: store volatile i32 [[VAL4]], ptr addrspace(1) null, align 4
38 ; CHECK-NEXT: store volatile i32 [[VAL5]], ptr addrspace(1) null, align 4
39 ; CHECK-NEXT: [[DISPATCH_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
40 ; CHECK-NEXT: [[QUEUE_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.queue.ptr()
41 ; CHECK-NEXT: [[VAL6:%.*]] = load volatile ptr, ptr addrspace(4) [[DISPATCH_PTR]], align 8
42 ; CHECK-NEXT: [[VAL7:%.*]] = load volatile ptr, ptr addrspace(4) [[QUEUE_PTR]], align 8
43 ; CHECK-NEXT: [[VAL8:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()
44 ; CHECK-NEXT: store volatile i32 [[VAL8]], ptr addrspace(1) null, align 4
45 ; CHECK-NEXT: [[VAL9:%.*]] = call i64 @llvm.amdgcn.dispatch.id()
46 ; CHECK-NEXT: store volatile i64 [[VAL9]], ptr addrspace(1) null, align 8
47 ; CHECK-NEXT: ret void
49 %val0 = call i32 @llvm.amdgcn.workitem.id.x()
50 %val1 = call i32 @llvm.amdgcn.workitem.id.y()
51 %val2 = call i32 @llvm.amdgcn.workitem.id.z()
52 %val3 = call i32 @llvm.amdgcn.workgroup.id.x()
53 %val4 = call i32 @llvm.amdgcn.workgroup.id.y()
54 %val5 = call i32 @llvm.amdgcn.workgroup.id.z()
55 store volatile i32 %val0, ptr addrspace(1) null
56 store volatile i32 %val1, ptr addrspace(1) null
57 store volatile i32 %val2, ptr addrspace(1) null
58 store volatile i32 %val3, ptr addrspace(1) null
59 store volatile i32 %val4, ptr addrspace(1) null
60 store volatile i32 %val5, ptr addrspace(1) null
61 %dispatch.ptr = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
62 %queue.ptr = call ptr addrspace(4) @llvm.amdgcn.queue.ptr()
63 %val6 = load volatile ptr, ptr addrspace(4) %dispatch.ptr
64 %val7 = load volatile ptr, ptr addrspace(4) %queue.ptr
65 %val8 = call i32 @llvm.amdgcn.lds.kernel.id()
66 store volatile i32 %val8, ptr addrspace(1) null
67 %val9 = call i64 @llvm.amdgcn.dispatch.id()
68 store volatile i64 %val9, ptr addrspace(1) null
72 define amdgpu_kernel void @test_default_queue_offset_v4_0(ptr addrspace(1) %kernarg) {
73 ; CHECK-LABEL: define {{[^@]+}}@test_default_queue_offset_v4_0
74 ; CHECK-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR2:[0-9]+]] {
75 ; CHECK-NEXT: call void @use_everything_else()
76 ; CHECK-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
77 ; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 32
78 ; CHECK-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
79 ; CHECK-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
80 ; CHECK-NEXT: ret void
82 call void @use_everything_else()
83 %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
84 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 32
85 %load = load ptr, ptr addrspace(4) %gep
86 store ptr %load, ptr addrspace(1) %kernarg
90 define amdgpu_kernel void @test_default_queue_offset_v5_0(ptr addrspace(1) %kernarg) {
91 ; CHECK-LABEL: define {{[^@]+}}@test_default_queue_offset_v5_0
92 ; CHECK-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR3:[0-9]+]] {
93 ; CHECK-NEXT: call void @use_everything_else()
94 ; CHECK-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
95 ; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 104
96 ; CHECK-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
97 ; CHECK-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
98 ; CHECK-NEXT: ret void
100 call void @use_everything_else()
101 %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
102 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 104
103 %load = load ptr, ptr addrspace(4) %gep
104 store ptr %load, ptr addrspace(1) %kernarg
108 define amdgpu_kernel void @test_completion_action_offset_v4_0(ptr addrspace(1) %kernarg) {
109 ; V4-LABEL: define {{[^@]+}}@test_completion_action_offset_v4_0
110 ; V4-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR4:[0-9]+]] {
111 ; V4-NEXT: call void @use_everything_else()
112 ; V4-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
113 ; V4-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 40
114 ; V4-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
115 ; V4-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
118 ; V5-LABEL: define {{[^@]+}}@test_completion_action_offset_v4_0
119 ; V5-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR2]] {
120 ; V5-NEXT: call void @use_everything_else()
121 ; V5-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
122 ; V5-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 40
123 ; V5-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
124 ; V5-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
127 ; V6-LABEL: define {{[^@]+}}@test_completion_action_offset_v4_0
128 ; V6-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR2]] {
129 ; V6-NEXT: call void @use_everything_else()
130 ; V6-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
131 ; V6-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 40
132 ; V6-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
133 ; V6-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
136 call void @use_everything_else()
137 %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
138 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 40
139 %load = load ptr, ptr addrspace(4) %gep
140 store ptr %load, ptr addrspace(1) %kernarg
144 define amdgpu_kernel void @test_completion_action_offset_v5_0(ptr addrspace(1) %kernarg) {
145 ; V4-LABEL: define {{[^@]+}}@test_completion_action_offset_v5_0
146 ; V4-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR3]] {
147 ; V4-NEXT: call void @use_everything_else()
148 ; V4-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
149 ; V4-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 112
150 ; V4-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
151 ; V4-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
154 ; V5-LABEL: define {{[^@]+}}@test_completion_action_offset_v5_0
155 ; V5-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR4:[0-9]+]] {
156 ; V5-NEXT: call void @use_everything_else()
157 ; V5-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
158 ; V5-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 112
159 ; V5-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
160 ; V5-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
163 ; V6-LABEL: define {{[^@]+}}@test_completion_action_offset_v5_0
164 ; V6-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR4:[0-9]+]] {
165 ; V6-NEXT: call void @use_everything_else()
166 ; V6-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
167 ; V6-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 112
168 ; V6-NEXT: [[LOAD:%.*]] = load ptr, ptr addrspace(4) [[GEP]], align 8
169 ; V6-NEXT: store ptr [[LOAD]], ptr addrspace(1) [[KERNARG]], align 8
172 call void @use_everything_else()
173 %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
174 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 112
175 %load = load ptr, ptr addrspace(4) %gep
176 store ptr %load, ptr addrspace(1) %kernarg
180 define amdgpu_kernel void @test_default_queue_completion_action_offset_v3_0(ptr addrspace(1) %kernarg) {
181 ; V4-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v3_0
182 ; V4-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR5:[0-9]+]] {
183 ; V4-NEXT: call void @use_everything_else()
184 ; V4-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
185 ; V4-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 32
186 ; V4-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
187 ; V4-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
190 ; V5-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v3_0
191 ; V5-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR2]] {
192 ; V5-NEXT: call void @use_everything_else()
193 ; V5-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
194 ; V5-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 32
195 ; V5-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
196 ; V5-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
199 ; V6-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v3_0
200 ; V6-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR2]] {
201 ; V6-NEXT: call void @use_everything_else()
202 ; V6-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
203 ; V6-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 32
204 ; V6-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
205 ; V6-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
208 call void @use_everything_else()
209 %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
210 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 32
211 %load = load <2 x ptr>, ptr addrspace(4) %gep
212 store <2 x ptr> %load, ptr addrspace(1) %kernarg
216 define amdgpu_kernel void @test_default_queue_completion_action_offset_v5_0(ptr addrspace(1) %kernarg) {
217 ; V4-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v5_0
218 ; V4-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR3]] {
219 ; V4-NEXT: call void @use_everything_else()
220 ; V4-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
221 ; V4-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 104
222 ; V4-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
223 ; V4-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
226 ; V5-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v5_0
227 ; V5-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR5:[0-9]+]] {
228 ; V5-NEXT: call void @use_everything_else()
229 ; V5-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
230 ; V5-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 104
231 ; V5-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
232 ; V5-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
235 ; V6-LABEL: define {{[^@]+}}@test_default_queue_completion_action_offset_v5_0
236 ; V6-SAME: (ptr addrspace(1) [[KERNARG:%.*]]) #[[ATTR5:[0-9]+]] {
237 ; V6-NEXT: call void @use_everything_else()
238 ; V6-NEXT: [[IMPLICITARG_PTR:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
239 ; V6-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr addrspace(4) [[IMPLICITARG_PTR]], i64 104
240 ; V6-NEXT: [[LOAD:%.*]] = load <2 x ptr>, ptr addrspace(4) [[GEP]], align 16
241 ; V6-NEXT: store <2 x ptr> [[LOAD]], ptr addrspace(1) [[KERNARG]], align 16
245 call void @use_everything_else()%implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
246 %gep = getelementptr inbounds i8, ptr addrspace(4) %implicitarg.ptr, i64 104
247 %load = load <2 x ptr>, ptr addrspace(4) %gep
248 store <2 x ptr> %load, ptr addrspace(1) %kernarg
253 attributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
255 !llvm.module.flags = !{!0}
256 !0 = !{i32 1, !"amdhsa_code_object_version", i32 CODE_OBJECT_VERSION}
260 ; V4: attributes #[[ATTR0:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
261 ; V4: attributes #[[ATTR1]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
262 ; V4: attributes #[[ATTR2]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
263 ; V4: attributes #[[ATTR3]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
264 ; V4: attributes #[[ATTR4]] = { "amdgpu-no-agpr" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
265 ; V4: attributes #[[ATTR5]] = { "amdgpu-no-agpr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
267 ; V5: attributes #[[ATTR0:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
268 ; V5: attributes #[[ATTR1]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
269 ; V5: attributes #[[ATTR2]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
270 ; V5: attributes #[[ATTR3]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
271 ; V5: attributes #[[ATTR4]] = { "amdgpu-no-agpr" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
272 ; V5: attributes #[[ATTR5]] = { "amdgpu-no-agpr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
274 ; V6: attributes #[[ATTR0:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
275 ; V6: attributes #[[ATTR1]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
276 ; V6: attributes #[[ATTR2]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
277 ; V6: attributes #[[ATTR3]] = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
278 ; V6: attributes #[[ATTR4]] = { "amdgpu-no-agpr" "amdgpu-no-default-queue" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
279 ; V6: attributes #[[ATTR5]] = { "amdgpu-no-agpr" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-multigrid-sync-arg" "uniform-work-group-size"="false" }
281 ; V4: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 400}
283 ; V5: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 500}
285 ; V6: [[META0:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 600}