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
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx803 < %s | FileCheck --check-prefix=GCN %s
; RUN: llvm-as -data-layout=A5 < %s | llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx803 | FileCheck --check-prefix=GCN %s
declare ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
declare ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
declare i32 @llvm.amdgcn.workitem.id.x()
declare i32 @llvm.amdgcn.workgroup.id.x()
declare void @llvm.amdgcn.s.barrier()
@test_local.temp = internal addrspace(3) global [1 x i32] poison, align 4
@test_global_local.temp = internal addrspace(3) global [1 x i32] poison, align 4
; GCN-LABEL: {{^}}test_local
; GCN: v_mov_b32_e32 v[[VAL:[0-9]+]], 0x777
; GCN: ds_write_b32 v{{[0-9]+}}, v[[VAL]]
; GCN: s_waitcnt lgkmcnt(0){{$}}
; GCN-NEXT: s_barrier
; GCN: flat_store_dword
define amdgpu_kernel void @test_local(ptr addrspace(1) %arg) {
bb:
%i = alloca ptr addrspace(1), align 4, addrspace(5)
store ptr addrspace(1) %arg, ptr addrspace(5) %i, align 4
%i1 = call i32 @llvm.amdgcn.workitem.id.x()
%i2 = zext i32 %i1 to i64
%i3 = icmp eq i64 %i2, 0
br i1 %i3, label %bb4, label %bb5
bb4: ; preds = %bb
store i32 1911, ptr addrspace(3) @test_local.temp, align 4
br label %bb5
bb5: ; preds = %bb4, %bb
fence syncscope("workgroup") release
call void @llvm.amdgcn.s.barrier()
fence syncscope("workgroup") acquire
%i6 = load i32, ptr addrspace(3) @test_local.temp, align 4
%i7 = load ptr addrspace(1), ptr addrspace(5) %i, align 4
%i8 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i9 = call i32 @llvm.amdgcn.workitem.id.x()
%i10 = call i32 @llvm.amdgcn.workgroup.id.x()
%i11 = getelementptr inbounds i8, ptr addrspace(4) %i8, i64 4
%i13 = load i16, ptr addrspace(4) %i11, align 4
%i14 = zext i16 %i13 to i32
%i15 = mul i32 %i10, %i14
%i16 = add i32 %i15, %i9
%i17 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i18 = zext i32 %i16 to i64
%i20 = load i64, ptr addrspace(4) %i17, align 8
%i21 = add i64 %i20, %i18
%i22 = getelementptr inbounds i32, ptr addrspace(1) %i7, i64 %i21
store i32 %i6, ptr addrspace(1) %i22, align 4
ret void
}
; GCN-LABEL: {{^}}test_global
; GCN: v_add_u32_e32 v{{[0-9]+}}, vcc, 0x888, v{{[0-9]+}}
; GCN: flat_store_dword
; GCN: s_waitcnt vmcnt(0){{$}}
; GCN-NEXT: s_barrier
define amdgpu_kernel void @test_global(ptr addrspace(1) %arg) {
bb:
%i = alloca ptr addrspace(1), align 4, addrspace(5)
%i1 = alloca i32, align 4, addrspace(5)
store ptr addrspace(1) %arg, ptr addrspace(5) %i, align 4
store i32 0, ptr addrspace(5) %i1, align 4
br label %bb2
bb2: ; preds = %bb56, %bb
%i3 = load i32, ptr addrspace(5) %i1, align 4
%i4 = sext i32 %i3 to i64
%i5 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i6 = call i32 @llvm.amdgcn.workitem.id.x()
%i7 = call i32 @llvm.amdgcn.workgroup.id.x()
%i8 = getelementptr inbounds i8, ptr addrspace(4) %i5, i64 4
%i10 = load i16, ptr addrspace(4) %i8, align 4
%i11 = zext i16 %i10 to i32
%i12 = mul i32 %i7, %i11
%i13 = add i32 %i12, %i6
%i14 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i15 = zext i32 %i13 to i64
%i17 = load i64, ptr addrspace(4) %i14, align 8
%i18 = add i64 %i17, %i15
%i19 = icmp ult i64 %i4, %i18
br i1 %i19, label %bb20, label %bb59
bb20: ; preds = %bb2
%i21 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i22 = call i32 @llvm.amdgcn.workitem.id.x()
%i23 = call i32 @llvm.amdgcn.workgroup.id.x()
%i24 = getelementptr inbounds i8, ptr addrspace(4) %i21, i64 4
%i26 = load i16, ptr addrspace(4) %i24, align 4
%i27 = zext i16 %i26 to i32
%i28 = mul i32 %i23, %i27
%i29 = add i32 %i28, %i22
%i30 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i31 = zext i32 %i29 to i64
%i33 = load i64, ptr addrspace(4) %i30, align 8
%i34 = add i64 %i33, %i31
%i35 = add i64 %i34, 2184
%i36 = trunc i64 %i35 to i32
%i37 = load ptr addrspace(1), ptr addrspace(5) %i, align 4
%i38 = load i32, ptr addrspace(5) %i1, align 4
%i39 = sext i32 %i38 to i64
%i40 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i41 = call i32 @llvm.amdgcn.workitem.id.x()
%i42 = call i32 @llvm.amdgcn.workgroup.id.x()
%i43 = getelementptr inbounds i8, ptr addrspace(4) %i40, i64 4
%i45 = load i16, ptr addrspace(4) %i43, align 4
%i46 = zext i16 %i45 to i32
%i47 = mul i32 %i42, %i46
%i48 = add i32 %i47, %i41
%i49 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i50 = zext i32 %i48 to i64
%i52 = load i64, ptr addrspace(4) %i49, align 8
%i53 = add i64 %i52, %i50
%i54 = add i64 %i39, %i53
%i55 = getelementptr inbounds i32, ptr addrspace(1) %i37, i64 %i54
store i32 %i36, ptr addrspace(1) %i55, align 4
fence syncscope("workgroup") release
call void @llvm.amdgcn.s.barrier()
fence syncscope("workgroup") acquire
br label %bb56
bb56: ; preds = %bb20
%i57 = load i32, ptr addrspace(5) %i1, align 4
%i58 = add nsw i32 %i57, 1
store i32 %i58, ptr addrspace(5) %i1, align 4
br label %bb2
bb59: ; preds = %bb2
ret void
}
; GCN-LABEL: {{^}}test_global_local
; GCN: v_mov_b32_e32 v[[VAL:[0-9]+]], 0x999
; GCN: ds_write_b32 v{{[0-9]+}}, v[[VAL]]
; GCN: s_waitcnt vmcnt(0) lgkmcnt(0){{$}}
; GCN-NEXT: s_barrier
; GCN: flat_store_dword
define amdgpu_kernel void @test_global_local(ptr addrspace(1) %arg) {
bb:
%i = alloca ptr addrspace(1), align 4, addrspace(5)
store ptr addrspace(1) %arg, ptr addrspace(5) %i, align 4
%i1 = load ptr addrspace(1), ptr addrspace(5) %i, align 4
%i2 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i3 = call i32 @llvm.amdgcn.workitem.id.x()
%i4 = call i32 @llvm.amdgcn.workgroup.id.x()
%i5 = getelementptr inbounds i8, ptr addrspace(4) %i2, i64 4
%i7 = load i16, ptr addrspace(4) %i5, align 4
%i8 = zext i16 %i7 to i32
%i9 = mul i32 %i4, %i8
%i10 = add i32 %i9, %i3
%i11 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i12 = zext i32 %i10 to i64
%i14 = load i64, ptr addrspace(4) %i11, align 8
%i15 = add i64 %i14, %i12
%i16 = getelementptr inbounds i32, ptr addrspace(1) %i1, i64 %i15
store i32 1, ptr addrspace(1) %i16, align 4
%i17 = call i32 @llvm.amdgcn.workitem.id.x()
%i18 = zext i32 %i17 to i64
%i19 = icmp eq i64 %i18, 0
br i1 %i19, label %bb20, label %bb21
bb20: ; preds = %bb
store i32 2457, ptr addrspace(3) @test_global_local.temp, align 4
br label %bb21
bb21: ; preds = %bb20, %bb
fence syncscope("workgroup") release
call void @llvm.amdgcn.s.barrier()
fence syncscope("workgroup") acquire
%i22 = load i32, ptr addrspace(3) @test_global_local.temp, align 4
%i23 = load ptr addrspace(1), ptr addrspace(5) %i, align 4
%i24 = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
%i25 = call i32 @llvm.amdgcn.workitem.id.x()
%i26 = call i32 @llvm.amdgcn.workgroup.id.x()
%i27 = getelementptr inbounds i8, ptr addrspace(4) %i24, i64 4
%i29 = load i16, ptr addrspace(4) %i27, align 4
%i30 = zext i16 %i29 to i32
%i31 = mul i32 %i26, %i30
%i32 = add i32 %i31, %i25
%i33 = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
%i34 = zext i32 %i32 to i64
%i36 = load i64, ptr addrspace(4) %i33, align 8
%i37 = add i64 %i36, %i34
%i38 = getelementptr inbounds i32, ptr addrspace(1) %i23, i64 %i37
store i32 %i22, ptr addrspace(1) %i38, align 4
ret void
}
|
