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
|
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
// REQUIRES: amdgpu-registered-target
// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx942 -emit-llvm -fcuda-is-device %s -o - | FileCheck %s
#define __device__ __attribute__((device))
#define __shared__ __attribute__((shared))
#define __constant__ __attribute__((constant))
__constant__ float const_float;
__constant__ double const_double;
__device__ float global_float;
__device__ double global_double;
// CHECK-LABEL: define dso_local void @_Z30test_flat_atomic_fadd_f32_flatPff(
// CHECK-SAME: ptr noundef [[PTR:%.*]], float noundef [[VAL:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[PTR_ADDR]] to ptr
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store ptr [[PTR]], ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: store float [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = atomicrmw fadd ptr [[TMP0]], float [[TMP1]] syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META4:![0-9]+]], !amdgpu.ignore.denormal.mode [[META4]]
// CHECK-NEXT: store float [[TMP2]], ptr [[RESULT_ASCAST]], align 4
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f32_flat(float *ptr, float val) {
float result;
result = __builtin_amdgcn_flat_atomic_fadd_f32(ptr, val);
}
// CHECK-LABEL: define dso_local void @_Z30test_flat_atomic_fadd_f64_flatPdd(
// CHECK-SAME: ptr noundef [[PTR:%.*]], double noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[PTR_ADDR]] to ptr
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store ptr [[PTR]], ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: store double [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = load double, ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = atomicrmw fadd ptr [[TMP0]], double [[TMP1]] syncscope("agent") monotonic, align 8, !amdgpu.no.fine.grained.memory [[META4]]
// CHECK-NEXT: store double [[TMP2]], ptr [[RESULT_ASCAST]], align 8
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f64_flat(double *ptr, double val) {
double result;
result = __builtin_amdgcn_flat_atomic_fadd_f64(ptr, val);
}
// CHECK-LABEL: define dso_local void @_Z32test_flat_atomic_fadd_f32_sharedPff(
// CHECK-SAME: ptr noundef [[PTR:%.*]], float noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[PTR_ADDR]] to ptr
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store ptr [[PTR]], ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: store float [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = load float, ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP2:%.*]] = atomicrmw fadd ptr [[TMP0]], float [[TMP1]] syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META4]], !amdgpu.ignore.denormal.mode [[META4]]
// CHECK-NEXT: store float [[TMP2]], ptr [[RESULT_ASCAST]], align 4
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f32_shared(__shared__ float *ptr, float val) {
float result;
result = __builtin_amdgcn_flat_atomic_fadd_f32(ptr, val);
}
// CHECK-LABEL: define dso_local void @_Z32test_flat_atomic_fadd_f64_sharedPdd(
// CHECK-SAME: ptr noundef [[PTR:%.*]], double noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[PTR_ADDR]] to ptr
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store ptr [[PTR]], ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: store double [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[PTR_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = load double, ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP2:%.*]] = atomicrmw fadd ptr [[TMP0]], double [[TMP1]] syncscope("agent") monotonic, align 8, !amdgpu.no.fine.grained.memory [[META4]]
// CHECK-NEXT: store double [[TMP2]], ptr [[RESULT_ASCAST]], align 8
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f64_shared(__shared__ double *ptr, double val) {
double result;
result = __builtin_amdgcn_flat_atomic_fadd_f64(ptr, val);
}
// CHECK-LABEL: define dso_local void @_Z34test_flat_atomic_fadd_f32_constantf(
// CHECK-SAME: float noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store float [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = atomicrmw fadd ptr addrspacecast (ptr addrspace(4) @const_float to ptr), float [[TMP0]] syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META4]], !amdgpu.ignore.denormal.mode [[META4]]
// CHECK-NEXT: store float [[TMP1]], ptr [[RESULT_ASCAST]], align 4
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f32_constant(float val) {
float result;
result = __builtin_amdgcn_flat_atomic_fadd_f32(&const_float, val);
}
// CHECK-LABEL: define dso_local void @_Z34test_flat_atomic_fadd_f64_constantd(
// CHECK-SAME: double noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store double [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load double, ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = atomicrmw fadd ptr addrspacecast (ptr addrspace(4) @const_double to ptr), double [[TMP0]] syncscope("agent") monotonic, align 8, !amdgpu.no.fine.grained.memory [[META4]]
// CHECK-NEXT: store double [[TMP1]], ptr [[RESULT_ASCAST]], align 8
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f64_constant(double val) {
double result;
result = __builtin_amdgcn_flat_atomic_fadd_f64(&const_double, val);
}
// CHECK-LABEL: define dso_local void @_Z32test_flat_atomic_fadd_f32_globalf(
// CHECK-SAME: float noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca float, align 4, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store float [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[VAL_ADDR_ASCAST]], align 4
// CHECK-NEXT: [[TMP1:%.*]] = atomicrmw fadd ptr addrspacecast (ptr addrspace(1) @global_float to ptr), float [[TMP0]] syncscope("agent") monotonic, align 4, !amdgpu.no.fine.grained.memory [[META4]], !amdgpu.ignore.denormal.mode [[META4]]
// CHECK-NEXT: store float [[TMP1]], ptr [[RESULT_ASCAST]], align 4
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f32_global(float val) {
float result;
result = __builtin_amdgcn_flat_atomic_fadd_f32(&global_float, val);
}
// CHECK-LABEL: define dso_local void @_Z32test_flat_atomic_fadd_f64_globald(
// CHECK-SAME: double noundef [[VAL:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[VAL_ADDR:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[RESULT:%.*]] = alloca double, align 8, addrspace(5)
// CHECK-NEXT: [[VAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VAL_ADDR]] to ptr
// CHECK-NEXT: [[RESULT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT]] to ptr
// CHECK-NEXT: store double [[VAL]], ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP0:%.*]] = load double, ptr [[VAL_ADDR_ASCAST]], align 8
// CHECK-NEXT: [[TMP1:%.*]] = atomicrmw fadd ptr addrspacecast (ptr addrspace(1) @global_double to ptr), double [[TMP0]] syncscope("agent") monotonic, align 8, !amdgpu.no.fine.grained.memory [[META4]]
// CHECK-NEXT: store double [[TMP1]], ptr [[RESULT_ASCAST]], align 8
// CHECK-NEXT: ret void
//
__device__ void test_flat_atomic_fadd_f64_global(double val) {
double result;
result = __builtin_amdgcn_flat_atomic_fadd_f64(&global_double, val);
}
//.
// CHECK: [[META4]] = !{}
//.
|
