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
|
#include "clang/Basic/Cuda.h"
#include "llvm/ADT/Twine.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/VersionTuple.h"
namespace clang {
struct CudaVersionMapEntry {
const char *Name;
CudaVersion Version;
llvm::VersionTuple TVersion;
};
#define CUDA_ENTRY(major, minor) \
{ \
#major "." #minor, CudaVersion::CUDA_##major##minor, \
llvm::VersionTuple(major, minor) \
}
static const CudaVersionMapEntry CudaNameVersionMap[] = {
CUDA_ENTRY(7, 0),
CUDA_ENTRY(7, 5),
CUDA_ENTRY(8, 0),
CUDA_ENTRY(9, 0),
CUDA_ENTRY(9, 1),
CUDA_ENTRY(9, 2),
CUDA_ENTRY(10, 0),
CUDA_ENTRY(10, 1),
CUDA_ENTRY(10, 2),
CUDA_ENTRY(11, 0),
CUDA_ENTRY(11, 1),
CUDA_ENTRY(11, 2),
CUDA_ENTRY(11, 3),
CUDA_ENTRY(11, 4),
CUDA_ENTRY(11, 5),
CUDA_ENTRY(11, 6),
CUDA_ENTRY(11, 7),
CUDA_ENTRY(11, 8),
CUDA_ENTRY(12, 0),
CUDA_ENTRY(12, 1),
CUDA_ENTRY(12, 2),
CUDA_ENTRY(12, 3),
CUDA_ENTRY(12, 4),
CUDA_ENTRY(12, 5),
CUDA_ENTRY(12, 6),
CUDA_ENTRY(12, 8),
CUDA_ENTRY(12, 9),
{"", CudaVersion::NEW, llvm::VersionTuple(std::numeric_limits<int>::max())},
{"unknown", CudaVersion::UNKNOWN, {}} // End of list tombstone.
};
#undef CUDA_ENTRY
const char *CudaVersionToString(CudaVersion V) {
for (auto *I = CudaNameVersionMap; I->Version != CudaVersion::UNKNOWN; ++I)
if (I->Version == V)
return I->Name;
return CudaVersionToString(CudaVersion::UNKNOWN);
}
CudaVersion CudaStringToVersion(const llvm::Twine &S) {
std::string VS = S.str();
for (auto *I = CudaNameVersionMap; I->Version != CudaVersion::UNKNOWN; ++I)
if (I->Name == VS)
return I->Version;
return CudaVersion::UNKNOWN;
}
CudaVersion ToCudaVersion(llvm::VersionTuple Version) {
for (auto *I = CudaNameVersionMap; I->Version != CudaVersion::UNKNOWN; ++I)
if (I->TVersion == Version)
return I->Version;
return CudaVersion::UNKNOWN;
}
CudaVersion MinVersionForOffloadArch(OffloadArch A) {
if (A == OffloadArch::UNKNOWN)
return CudaVersion::UNKNOWN;
// AMD GPUs do not depend on CUDA versions.
if (IsAMDOffloadArch(A))
return CudaVersion::CUDA_70;
switch (A) {
case OffloadArch::SM_20:
case OffloadArch::SM_21:
case OffloadArch::SM_30:
case OffloadArch::SM_32_:
case OffloadArch::SM_35:
case OffloadArch::SM_37:
case OffloadArch::SM_50:
case OffloadArch::SM_52:
case OffloadArch::SM_53:
return CudaVersion::CUDA_70;
case OffloadArch::SM_60:
case OffloadArch::SM_61:
case OffloadArch::SM_62:
return CudaVersion::CUDA_80;
case OffloadArch::SM_70:
return CudaVersion::CUDA_90;
case OffloadArch::SM_72:
return CudaVersion::CUDA_91;
case OffloadArch::SM_75:
return CudaVersion::CUDA_100;
case OffloadArch::SM_80:
return CudaVersion::CUDA_110;
case OffloadArch::SM_86:
return CudaVersion::CUDA_111;
case OffloadArch::SM_87:
return CudaVersion::CUDA_114;
case OffloadArch::SM_89:
case OffloadArch::SM_90:
return CudaVersion::CUDA_118;
case OffloadArch::SM_90a:
return CudaVersion::CUDA_120;
case OffloadArch::SM_100:
case OffloadArch::SM_100a:
case OffloadArch::SM_101:
case OffloadArch::SM_101a:
case OffloadArch::SM_120:
case OffloadArch::SM_120a:
return CudaVersion::CUDA_128;
case OffloadArch::SM_103:
case OffloadArch::SM_103a:
case OffloadArch::SM_121:
case OffloadArch::SM_121a:
return CudaVersion::CUDA_129;
default:
llvm_unreachable("invalid enum");
}
}
CudaVersion MaxVersionForOffloadArch(OffloadArch A) {
// AMD GPUs do not depend on CUDA versions.
if (IsAMDOffloadArch(A))
return CudaVersion::NEW;
switch (A) {
case OffloadArch::UNKNOWN:
return CudaVersion::UNKNOWN;
case OffloadArch::SM_20:
case OffloadArch::SM_21:
return CudaVersion::CUDA_80;
case OffloadArch::SM_30:
case OffloadArch::SM_32_:
return CudaVersion::CUDA_102;
case OffloadArch::SM_35:
case OffloadArch::SM_37:
return CudaVersion::CUDA_118;
default:
return CudaVersion::NEW;
}
}
bool CudaFeatureEnabled(llvm::VersionTuple Version, CudaFeature Feature) {
return CudaFeatureEnabled(ToCudaVersion(Version), Feature);
}
bool CudaFeatureEnabled(CudaVersion Version, CudaFeature Feature) {
switch (Feature) {
case CudaFeature::CUDA_USES_NEW_LAUNCH:
return Version >= CudaVersion::CUDA_92;
case CudaFeature::CUDA_USES_FATBIN_REGISTER_END:
return Version >= CudaVersion::CUDA_101;
}
llvm_unreachable("Unknown CUDA feature.");
}
} // namespace clang
|
