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
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
|
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -triple amdgcn-unknown-unknown -target-cpu gfx1200 -emit-llvm -o - %s | FileCheck %s
// REQUIRES: amdgpu-registered-target
typedef unsigned int uint;
// CHECK-LABEL: @test_s_sleep_var(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[D_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[D_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[D_ADDR]] to ptr
// CHECK-NEXT: store i32 [[D:%.*]], ptr [[D_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[D_ADDR_ASCAST]], align 4
// CHECK-NEXT: call void @llvm.amdgcn.s.sleep.var(i32 [[TMP0]])
// CHECK-NEXT: call void @llvm.amdgcn.s.sleep.var(i32 15)
// CHECK-NEXT: ret void
//
void test_s_sleep_var(int d)
{
__builtin_amdgcn_s_sleep_var(d);
__builtin_amdgcn_s_sleep_var(15);
}
// CHECK-LABEL: @test_permlane16_var(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[OUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OUT_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(1) [[OUT:%.*]], ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[C:%.*]], ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.amdgcn.permlane16.var(i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i1 false, i1 false)
// CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP3]], ptr addrspace(1) [[TMP4]], align 4
// CHECK-NEXT: ret void
//
void test_permlane16_var(global uint* out, uint a, uint b, uint c) {
*out = __builtin_amdgcn_permlane16_var(a, b, c, 0, 0);
}
// CHECK-LABEL: @test_permlanex16_var(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[OUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OUT_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(1) [[OUT:%.*]], ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[C:%.*]], ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[C_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @llvm.amdgcn.permlanex16.var(i32 [[TMP0]], i32 [[TMP1]], i32 [[TMP2]], i1 false, i1 false)
// CHECK-NEXT: [[TMP4:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP3]], ptr addrspace(1) [[TMP4]], align 4
// CHECK-NEXT: ret void
//
void test_permlanex16_var(global uint* out, uint a, uint b, uint c) {
*out = __builtin_amdgcn_permlanex16_var(a, b, c, 0, 0);
}
// CHECK-LABEL: @test_s_barrier_signal(
// CHECK-NEXT: entry:
// CHECK-NEXT: call void @llvm.amdgcn.s.barrier.signal(i32 -1)
// CHECK-NEXT: call void @llvm.amdgcn.s.barrier.wait(i16 -1)
// CHECK-NEXT: ret void
//
void test_s_barrier_signal()
{
__builtin_amdgcn_s_barrier_signal(-1);
__builtin_amdgcn_s_barrier_wait(-1);
}
// CHECK-LABEL: @test_s_barrier_signal_var(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[BAR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[BAR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[BAR_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: store ptr [[BAR:%.*]], ptr [[BAR_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[BAR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = addrspacecast ptr [[TMP0]] to ptr addrspace(3)
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: call void @llvm.amdgcn.s.barrier.signal.var(ptr addrspace(3) [[TMP1]], i32 [[TMP2]])
// CHECK-NEXT: ret void
//
void test_s_barrier_signal_var(void *bar, int a)
{
__builtin_amdgcn_s_barrier_signal_var(bar, a);
}
// CHECK-LABEL: @test_s_barrier_signal_isfirst(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[C_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: [[C_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[C_ADDR]] to ptr
// CHECK-NEXT: store ptr [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr [[C:%.*]], ptr [[C_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = call i1 @llvm.amdgcn.s.barrier.signal.isfirst(i32 1)
// CHECK-NEXT: br i1 [[TMP0]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
// CHECK: if.then:
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[B_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr [[TMP1]], ptr [[A_ADDR_ASCAST]], align 8
// CHECK-NEXT: br label [[IF_END:%.*]]
// CHECK: if.else:
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[C_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr [[TMP2]], ptr [[A_ADDR_ASCAST]], align 8
// CHECK-NEXT: br label [[IF_END]]
// CHECK: if.end:
// CHECK-NEXT: call void @llvm.amdgcn.s.barrier.wait(i16 1)
// CHECK-NEXT: ret void
//
void test_s_barrier_signal_isfirst(int* a, int* b, int *c)
{
if(__builtin_amdgcn_s_barrier_signal_isfirst(1))
a = b;
else
a = c;
__builtin_amdgcn_s_barrier_wait(1);
}
// CHECK-LABEL: @test_s_get_barrier_state(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[STATE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.amdgcn.s.get.barrier.state(i32 [[TMP0]])
// CHECK-NEXT: store i32 [[TMP1]], ptr addrspace(5) [[STATE]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(5) [[STATE]], align 4
// CHECK-NEXT: ret i32 [[TMP2]]
//
unsigned test_s_get_barrier_state(int a)
{
unsigned State = __builtin_amdgcn_s_get_barrier_state(a);
return State;
}
// CHECK-LABEL: @test_s_get_named_barrier_state(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RETVAL:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[BAR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[STATE:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[RETVAL_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RETVAL]] to ptr
// CHECK-NEXT: [[BAR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[BAR_ADDR]] to ptr
// CHECK-NEXT: store ptr [[BAR:%.*]], ptr [[BAR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[BAR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = addrspacecast ptr [[TMP0]] to ptr addrspace(3)
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.s.get.named.barrier.state(ptr addrspace(3) [[TMP1]])
// CHECK-NEXT: store i32 [[TMP2]], ptr addrspace(5) [[STATE]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(5) [[STATE]], align 4
// CHECK-NEXT: ret i32 [[TMP3]]
//
unsigned test_s_get_named_barrier_state(void *bar)
{
unsigned State = __builtin_amdgcn_s_get_named_barrier_state(bar);
return State;
}
// CHECK-LABEL: @test_s_ttracedata(
// CHECK-NEXT: entry:
// CHECK-NEXT: call void @llvm.amdgcn.s.ttracedata(i32 1)
// CHECK-NEXT: ret void
//
void test_s_ttracedata()
{
__builtin_amdgcn_s_ttracedata(1);
}
// CHECK-LABEL: @test_s_ttracedata_imm(
// CHECK-NEXT: entry:
// CHECK-NEXT: call void @llvm.amdgcn.s.ttracedata.imm(i16 1)
// CHECK-NEXT: ret void
//
void test_s_ttracedata_imm()
{
__builtin_amdgcn_s_ttracedata_imm(1);
}
// CHECK-LABEL: @test_s_prefetch_data(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[FP_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[GP_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[CP_ADDR:%.*]] = alloca ptr addrspace(4), align 8, addrspace(5)
// CHECK-NEXT: [[LEN_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[FP_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[FP_ADDR]] to ptr
// CHECK-NEXT: [[GP_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[GP_ADDR]] to ptr
// CHECK-NEXT: [[CP_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CP_ADDR]] to ptr
// CHECK-NEXT: [[LEN_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[LEN_ADDR]] to ptr
// CHECK-NEXT: store ptr [[FP:%.*]], ptr [[FP_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr addrspace(1) [[GP:%.*]], ptr [[GP_ADDR_ASCAST]], align 8
// CHECK-NEXT: store ptr addrspace(4) [[CP:%.*]], ptr [[CP_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[LEN:%.*]], ptr [[LEN_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[FP_ADDR_ASCAST]], align 8
// CHECK-NEXT: call void @llvm.amdgcn.s.prefetch.data.p0(ptr [[TMP0]], i32 0)
// CHECK-NEXT: [[TMP1:%.*]] = load ptr addrspace(1), ptr [[GP_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[LEN_ADDR_ASCAST]], align 4
// CHECK-NEXT: call void @llvm.amdgcn.s.prefetch.data.p1(ptr addrspace(1) [[TMP1]], i32 [[TMP2]])
// CHECK-NEXT: [[TMP3:%.*]] = load ptr addrspace(4), ptr [[CP_ADDR_ASCAST]], align 8
// CHECK-NEXT: call void @llvm.amdgcn.s.prefetch.data.p4(ptr addrspace(4) [[TMP3]], i32 31)
// CHECK-NEXT: ret void
//
void test_s_prefetch_data(int *fp, global float *gp, constant char *cp, unsigned int len)
{
__builtin_amdgcn_s_prefetch_data(fp, 0);
__builtin_amdgcn_s_prefetch_data(gp, len);
__builtin_amdgcn_s_prefetch_data(cp, 31);
}
// CHECK-LABEL: @test_s_buffer_prefetch_data(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[RSRC_ADDR:%.*]] = alloca ptr addrspace(8), align 16, addrspace(5)
// CHECK-NEXT: [[LEN_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[RSRC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RSRC_ADDR]] to ptr
// CHECK-NEXT: [[LEN_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[LEN_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(8) [[RSRC:%.*]], ptr [[RSRC_ADDR_ASCAST]], align 16
// CHECK-NEXT: store i32 [[LEN:%.*]], ptr [[LEN_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load ptr addrspace(8), ptr [[RSRC_ADDR_ASCAST]], align 16
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[LEN_ADDR_ASCAST]], align 4
// CHECK-NEXT: call void @llvm.amdgcn.s.buffer.prefetch.data(ptr addrspace(8) [[TMP0]], i32 128, i32 [[TMP1]])
// CHECK-NEXT: [[TMP2:%.*]] = load ptr addrspace(8), ptr [[RSRC_ADDR_ASCAST]], align 16
// CHECK-NEXT: call void @llvm.amdgcn.s.buffer.prefetch.data(ptr addrspace(8) [[TMP2]], i32 0, i32 31)
// CHECK-NEXT: ret void
//
void test_s_buffer_prefetch_data(__amdgpu_buffer_rsrc_t rsrc, unsigned int len)
{
__builtin_amdgcn_s_buffer_prefetch_data(rsrc, 128, len);
__builtin_amdgcn_s_buffer_prefetch_data(rsrc, 0, 31);
}
// CHECK-LABEL: @test_ds_bpermute_fi_b32(
// CHECK-NEXT: entry:
// CHECK-NEXT: [[OUT_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[B_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
// CHECK-NEXT: [[OUT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[OUT_ADDR]] to ptr
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
// CHECK-NEXT: [[B_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[B_ADDR]] to ptr
// CHECK-NEXT: store ptr addrspace(1) [[OUT:%.*]], ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[A:%.*]], ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: store i32 [[B:%.*]], ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[A_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[B_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @llvm.amdgcn.ds.bpermute.fi.b32(i32 [[TMP0]], i32 [[TMP1]])
// CHECK-NEXT: [[TMP3:%.*]] = load ptr addrspace(1), ptr [[OUT_ADDR_ASCAST]], align 8
// CHECK-NEXT: store i32 [[TMP2]], ptr addrspace(1) [[TMP3]], align 4
// CHECK-NEXT: ret void
//
void test_ds_bpermute_fi_b32(global int* out, int a, int b)
{
*out = __builtin_amdgcn_ds_bpermute_fi_b32(a, b);
}
|
