1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-attributes --check-globals --version 3
; RUN: opt -S -mtriple=amdgcn-- -passes=amdgpu-attributor,amdgpu-lower-module-lds < %s --amdgpu-lower-module-lds-strategy=table | FileCheck -check-prefixes=CHECK,TABLE %s
; FIXME: Work around update_test_checks bug in constant expression handling by manually deleting part of the last global pattern
@function.lds = addrspace(3) global i16 poison
@other.kernel.lds = addrspace(3) global i16 poison
@recursive.kernel.lds = addrspace(3) global i16 poison
;.
; CHECK: @llvm.amdgcn.kernel.k0_f0.lds = internal addrspace(3) global %llvm.amdgcn.kernel.k0_f0.lds.t poison, align 2, !absolute_symbol [[META0:![0-9]+]]
; CHECK: @llvm.amdgcn.kernel.k1_f0.lds = internal addrspace(3) global %llvm.amdgcn.kernel.k1_f0.lds.t poison, align 2, !absolute_symbol [[META0]]
; CHECK: @llvm.amdgcn.kernel.kernel_lds.lds = internal addrspace(3) global %llvm.amdgcn.kernel.kernel_lds.lds.t poison, align 2, !absolute_symbol [[META0]]
; CHECK: @llvm.amdgcn.kernel.kernel_lds_recursion.lds = internal addrspace(3) global %llvm.amdgcn.kernel.kernel_lds_recursion.lds.t poison, align 2, !absolute_symbol [[META0]]
; CHECK: @llvm.amdgcn.lds.offset.table = internal addrspace(4) constant [3 x [2 x i32]]
;.
define internal void @lds_use_through_indirect() {
; CHECK-LABEL: define internal void @lds_use_through_indirect(
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()
; CHECK-NEXT: [[FUNCTION_LDS2:%.*]] = getelementptr inbounds [3 x [2 x i32]], ptr addrspace(4) @llvm.amdgcn.lds.offset.table, i32 0, i32 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[FUNCTION_LDS2]], align 4
; CHECK-NEXT: [[FUNCTION_LDS3:%.*]] = inttoptr i32 [[TMP2]] to ptr addrspace(3)
; CHECK-NEXT: [[LD:%.*]] = load i16, ptr addrspace(3) [[FUNCTION_LDS3]], align 2
; CHECK-NEXT: [[MUL:%.*]] = mul i16 [[LD]], 7
; CHECK-NEXT: [[FUNCTION_LDS:%.*]] = getelementptr inbounds [3 x [2 x i32]], ptr addrspace(4) @llvm.amdgcn.lds.offset.table, i32 0, i32 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(4) [[FUNCTION_LDS]], align 4
; CHECK-NEXT: [[FUNCTION_LDS1:%.*]] = inttoptr i32 [[TMP3]] to ptr addrspace(3)
; CHECK-NEXT: store i16 [[MUL]], ptr addrspace(3) [[FUNCTION_LDS1]], align 2
; CHECK-NEXT: ret void
;
%ld = load i16, ptr addrspace(3) @function.lds
%mul = mul i16 %ld, 7
store i16 %mul, ptr addrspace(3) @function.lds
ret void
}
define internal void @indirectly_called() {
; CHECK-LABEL: define internal void @indirectly_called(
; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: store volatile ptr @indirectly_called, ptr addrspace(1) null, align 8
; CHECK-NEXT: call void @lds_use_through_indirect()
; CHECK-NEXT: ret void
;
store volatile ptr @indirectly_called, ptr addrspace(1) null
call void @lds_use_through_indirect()
ret void
}
define internal void @calls_indirectly_called() {
; CHECK-LABEL: define internal void @calls_indirectly_called(
; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: call void @indirectly_called()
; CHECK-NEXT: ret void
;
call void @indirectly_called()
ret void
}
; TODO: Should still have "amdgpu-no-lds-kernel-id" attached
define internal void @no_lds_global_use_leaf() {
; CHECK-LABEL: define internal void @no_lds_global_use_leaf(
; CHECK-SAME: ) #[[ATTR1:[0-9]+]] {
; CHECK-NEXT: ret void
;
ret void
}
; Should have "amdgpu-no-lds-kernel-id" stripped
define internal void @f0() {
; CHECK-LABEL: define internal void @f0(
; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()
; CHECK-NEXT: [[FUNCTION_LDS2:%.*]] = getelementptr inbounds [3 x [2 x i32]], ptr addrspace(4) @llvm.amdgcn.lds.offset.table, i32 0, i32 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[FUNCTION_LDS2]], align 4
; CHECK-NEXT: [[FUNCTION_LDS3:%.*]] = inttoptr i32 [[TMP2]] to ptr addrspace(3)
; CHECK-NEXT: [[LD:%.*]] = load i16, ptr addrspace(3) [[FUNCTION_LDS3]], align 2
; CHECK-NEXT: [[MUL:%.*]] = mul i16 [[LD]], 4
; CHECK-NEXT: [[FUNCTION_LDS:%.*]] = getelementptr inbounds [3 x [2 x i32]], ptr addrspace(4) @llvm.amdgcn.lds.offset.table, i32 0, i32 [[TMP1]], i32 0
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(4) [[FUNCTION_LDS]], align 4
; CHECK-NEXT: [[FUNCTION_LDS1:%.*]] = inttoptr i32 [[TMP3]] to ptr addrspace(3)
; CHECK-NEXT: store i16 [[MUL]], ptr addrspace(3) [[FUNCTION_LDS1]], align 2
; CHECK-NEXT: call void @no_lds_global_use_leaf()
; CHECK-NEXT: ret void
;
%ld = load i16, ptr addrspace(3) @function.lds
%mul = mul i16 %ld, 4
store i16 %mul, ptr addrspace(3) @function.lds
call void @no_lds_global_use_leaf()
ret void
}
; Should have "amdgpu-no-lds-kernel-id" stripped
define internal void @f0_transitive() {
; CHECK-LABEL: define internal void @f0_transitive(
; CHECK-SAME: ) #[[ATTR0]] {
; CHECK-NEXT: call void @f0()
; CHECK-NEXT: call void @no_lds_global_use_leaf()
; CHECK-NEXT: ret void
;
call void @f0()
call void @no_lds_global_use_leaf()
ret void
}
define amdgpu_kernel void @k0_f0() {
; CHECK-LABEL: define amdgpu_kernel void @k0_f0(
; CHECK-SAME: ) #[[ATTR2:[0-9]+]] !llvm.amdgcn.lds.kernel.id [[META2:![0-9]+]] {
; CHECK-NEXT: call void @llvm.donothing() [ "ExplicitUse"(ptr addrspace(3) @llvm.amdgcn.kernel.k0_f0.lds) ]
; CHECK-NEXT: call void @f0_transitive()
; CHECK-NEXT: ret void
;
call void @f0_transitive()
ret void
}
define amdgpu_kernel void @k1_f0() {
; CHECK-LABEL: define amdgpu_kernel void @k1_f0(
; CHECK-SAME: ) #[[ATTR3:[0-9]+]] !llvm.amdgcn.lds.kernel.id [[META3:![0-9]+]] {
; CHECK-NEXT: call void @llvm.donothing() [ "ExplicitUse"(ptr addrspace(3) @llvm.amdgcn.kernel.k1_f0.lds) ], !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]
; CHECK-NEXT: call void @f0_transitive()
; CHECK-NEXT: [[FPTR:%.*]] = load volatile ptr, ptr addrspace(1) null, align 8
; CHECK-NEXT: call void [[FPTR]]()
; CHECK-NEXT: call void @calls_indirectly_called()
; CHECK-NEXT: ret void
;
call void @f0_transitive()
%fptr = load volatile ptr, ptr addrspace(1) null
call void %fptr()
call void @calls_indirectly_called()
ret void
}
; Should still have "amdgpu-no-lds-kernel-id" attached
define amdgpu_kernel void @kernel_lds() {
; CHECK-LABEL: define amdgpu_kernel void @kernel_lds(
; CHECK-SAME: ) #[[ATTR4:[0-9]+]] {
; CHECK-NEXT: [[LD:%.*]] = load i16, ptr addrspace(3) @llvm.amdgcn.kernel.kernel_lds.lds, align 2
; CHECK-NEXT: [[MUL:%.*]] = mul i16 [[LD]], 42
; CHECK-NEXT: store i16 [[MUL]], ptr addrspace(3) @llvm.amdgcn.kernel.kernel_lds.lds, align 2
; CHECK-NEXT: ret void
;
%ld = load i16, ptr addrspace(3) @other.kernel.lds
%mul = mul i16 %ld, 42
store i16 %mul, ptr addrspace(3) @other.kernel.lds
ret void
}
define internal i16 @mutual_recursion_0(i16 %arg) {
; CHECK-LABEL: define internal i16 @mutual_recursion_0(
; CHECK-SAME: i16 [[ARG:%.*]]) #[[ATTR0]] {
; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.lds.kernel.id()
; CHECK-NEXT: [[RECURSIVE_KERNEL_LDS:%.*]] = getelementptr inbounds [3 x [2 x i32]], ptr addrspace(4) @llvm.amdgcn.lds.offset.table, i32 0, i32 [[TMP1]], i32 1
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[RECURSIVE_KERNEL_LDS]], align 4
; CHECK-NEXT: [[RECURSIVE_KERNEL_LDS1:%.*]] = inttoptr i32 [[TMP2]] to ptr addrspace(3)
; CHECK-NEXT: [[LD:%.*]] = load i16, ptr addrspace(3) [[RECURSIVE_KERNEL_LDS1]], align 2
; CHECK-NEXT: [[MUL:%.*]] = mul i16 [[LD]], 7
; CHECK-NEXT: [[RET:%.*]] = call i16 @mutual_recursion_1(i16 [[LD]])
; CHECK-NEXT: [[ADD:%.*]] = add i16 [[RET]], 1
; CHECK-NEXT: ret i16 [[ADD]]
;
%ld = load i16, ptr addrspace(3) @recursive.kernel.lds
%mul = mul i16 %ld, 7
%ret = call i16 @mutual_recursion_1(i16 %ld)
%add = add i16 %ret, 1
ret i16 %add
}
define internal void @mutual_recursion_1(i16 %arg) {
; CHECK-LABEL: define internal void @mutual_recursion_1(
; CHECK-SAME: i16 [[ARG:%.*]]) #[[ATTR0]] {
; CHECK-NEXT: call void @mutual_recursion_0(i16 [[ARG]])
; CHECK-NEXT: ret void
;
call void @mutual_recursion_0(i16 %arg)
ret void
}
define amdgpu_kernel void @kernel_lds_recursion() {
; CHECK-LABEL: define amdgpu_kernel void @kernel_lds_recursion(
; CHECK-SAME: ) #[[ATTR5:[0-9]+]] !llvm.amdgcn.lds.kernel.id [[META9:![0-9]+]] {
; CHECK-NEXT: call void @llvm.donothing() [ "ExplicitUse"(ptr addrspace(3) @llvm.amdgcn.kernel.kernel_lds_recursion.lds) ], !alias.scope [[META10:![0-9]+]], !noalias [[META13:![0-9]+]]
; CHECK-NEXT: call void @mutual_recursion_0(i16 0)
; CHECK-NEXT: ret void
;
call void @mutual_recursion_0(i16 0)
ret void
}
!llvm.module.flags = !{!1}
!1 = !{i32 1, !"amdhsa_code_object_version", i32 400}
;.
; CHECK: attributes #[[ATTR0]] = { "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR1]] = { "amdgpu-agpr-alloc"="0" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR2]] = { "amdgpu-agpr-alloc"="0" "amdgpu-lds-size"="2" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR3]] = { "amdgpu-lds-size"="4" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR4]] = { "amdgpu-agpr-alloc"="0" "amdgpu-lds-size"="2" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR5]] = { "amdgpu-agpr-alloc"="0" "amdgpu-lds-size"="4" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" "amdgpu-waves-per-eu"="4,10" "uniform-work-group-size"="false" }
; CHECK: attributes #[[ATTR6:[0-9]+]] = { nocallback nofree nosync nounwind willreturn memory(none) }
; CHECK: attributes #[[ATTR7:[0-9]+]] = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
;.
; CHECK: [[META0]] = !{i32 0, i32 1}
; CHECK: [[META1:![0-9]+]] = !{i32 1, !"amdhsa_code_object_version", i32 400}
; CHECK: [[META2]] = !{i32 0}
; CHECK: [[META3]] = !{i32 1}
; CHECK: [[META4]] = !{[[META5:![0-9]+]]}
; CHECK: [[META5]] = distinct !{[[META5]], [[META6:![0-9]+]]}
; CHECK: [[META6]] = distinct !{[[META6]]}
; CHECK: [[META7]] = !{[[META8:![0-9]+]]}
; CHECK: [[META8]] = distinct !{[[META8]], [[META6]]}
; CHECK: [[META9]] = !{i32 2}
; CHECK: [[META10]] = !{[[META11:![0-9]+]]}
; CHECK: [[META11]] = distinct !{[[META11]], [[META12:![0-9]+]]}
; CHECK: [[META12]] = distinct !{[[META12]]}
; CHECK: [[META13]] = !{[[META14:![0-9]+]]}
; CHECK: [[META14]] = distinct !{[[META14]], [[META12]]}
;.
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
; TABLE: {{.*}}
|
