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
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
|
//===----------------------------------------------------------------------===//
//
// Part of libcu++, the C++ Standard Library for your entire system,
// under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include <fstream>
#include <vector>
#include <string>
#include <map>
using namespace std::string_literals;
int main() {
std::map<std::string, std::string> scopes{ {"system", ".sys"},
{"device", ".gpu"},
{"block", ".cta"} };
std::map<std::string, std::string> membar_scopes{ {"system", ".sys"},
{"device", ".gl"},
{"block", ".cta"} };
std::map<std::string, std::string> fence_semantics{ { "sc", ".sc" },
{ "acq_rel", ".acq_rel" } };
bool const ld_as_atom = false;
std::vector<int> ld_sizes{ //8,
//16,
32,
64
};
std::map<std::string, std::string> ld_semantics{ { "relaxed", ".relaxed" },
{ "acquire", ".acquire" },
{ "volatile", ".volatile" } };
std::vector<int> st_sizes{ //8,
//16,
32,
64 };
std::map<std::string, std::string> st_semantics{ { "relaxed", ".relaxed" },
{ "release", ".release" },
{ "volatile", ".volatile" } };
std::vector<int> rmw_sizes{ 32, 64 };
std::map<std::string, std::string> rmw_semantics{ { "relaxed", ".relaxed" },
{ "acquire", ".acquire" },
{ "release", ".release" },
{ "acq_rel", ".acq_rel" },
{ "volatile", "" } };
std::map<std::string, std::string> rmw_operations{ { "exchange", ".exch" },
{ "compare_exchange", ".cas" },
{ "fetch_add", ".add" },
{ "fetch_sub", ".add" },
{ "fetch_max", ".max" },
{ "fetch_min", ".min" },
{ "fetch_and", ".and" },
{ "fetch_or", ".or" },
{ "fetch_xor", ".xor" } };
std::vector<std::string> cv_qualifier{ "volatile "/*, ""*/ };
std::map<int, std::string> registers{ //{ 8, "r" } ,
//{ 16, "h" },
{ 32, "r" },
{ 64, "l" } };
std::ofstream out("__atomic_generated");
out << R"XXX(//===----------------------------------------------------------------------===//
//
// Part of libcu++, the C++ Standard Library for your entire system,
// under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
)XXX" << "\n\n";
auto scopenametag = [&](auto scope) {
return "__thread_scope_" + scope + "_tag";
};
auto fencename = [&](auto sem, auto scope) {
return "__cuda_fence_" + sem + "_" + scope;
};
for(auto& s : scopes) {
out << "static inline __device__ void __cuda_membar_" << s.first << "() { asm volatile(\"membar" << membar_scopes[s.first] << ";\":::\"memory\"); }\n";
for(auto& sem : fence_semantics)
out << "static inline __device__ void " << fencename(sem.first, s.first) << "() { asm volatile(\"fence" << sem.second << s.second << ";\":::\"memory\"); }\n";
out << "static inline __device__ void __atomic_thread_fence_cuda(int __memorder, " << scopenametag(s.first) << ") {\n";
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "(); break;\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE:\n";
out << " case __ATOMIC_ACQ_REL:\n";
out << " case __ATOMIC_RELEASE: " << fencename("acq_rel"s, s.first) << "(); break;\n";
out << " case __ATOMIC_RELAXED: break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST:\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE:\n";
out << " case __ATOMIC_ACQ_REL:\n";
out << " case __ATOMIC_RELEASE: __cuda_membar_" << s.first << "(); break;\n";
out << " case __ATOMIC_RELAXED: break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
out << "}\n";
for(auto& sz : ld_sizes) {
for(auto& sem : ld_semantics) {
out << "template<class _CUDA_A, class _CUDA_B> ";
out << "static inline __device__ void __cuda_load_" << sem.first << "_" << sz << "_" << s.first << "(_CUDA_A __ptr, _CUDA_B& __dst) {";
if(ld_as_atom)
out << "asm volatile(\"atom.add" << (sem.first == "volatile" ? "" : sem.second.c_str()) << s.second << ".u" << sz << " %0, [%1], 0;\" : ";
else
out << "asm volatile(\"ld" << sem.second << (sem.first == "volatile" ? "" : s.second.c_str()) << ".b" << sz << " %0,[%1];\" : ";
out << "\"=" << registers[sz] << "\"(__dst) : \"l\"(__ptr)";
out << " : \"memory\"); }\n";
}
for(auto& cv: cv_qualifier) {
out << "template<class _Type, typename _CUDA_VSTD::enable_if<sizeof(_Type)==" << sz/8 << ", int>::type = 0>\n";
out << "__device__ void __atomic_load_cuda(const " << cv << "_Type *__ptr, _Type *__ret, int __memorder, " << scopenametag(s.first) << ") {\n";
out << " uint" << (registers[sz] == "r" ? 32 : sz) << "_t __tmp = 0;\n";
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_load_acquire_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_load_relaxed_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST: __cuda_membar_" << s.first << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_load_volatile_" << sz << "_" << s.first << "(__ptr, __tmp); __cuda_membar_" << s.first << "(); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_load_volatile_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
out << " memcpy(__ret, &__tmp, " << sz/8 << ");\n";
out << "}\n";
}
}
for(auto& sz : st_sizes) {
for(auto& sem : st_semantics) {
out << "template<class _CUDA_A, class _CUDA_B> ";
out << "static inline __device__ void __cuda_store_" << sem.first << "_" << sz << "_" << s.first << "(_CUDA_A __ptr, _CUDA_B __src) { ";
out << "asm volatile(\"st" << sem.second << (sem.first == "volatile" ? "" : s.second.c_str()) << ".b" << sz << " [%0], %1;\" :: ";
out << "\"l\"(__ptr),\"" << registers[sz] << "\"(__src)";
out << " : \"memory\"); }\n";
}
for(auto& cv: cv_qualifier) {
out << "template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==" << sz/8 << ", int>::type = 0>\n";
out << "__device__ void __atomic_store_cuda(" << cv << "_Type *__ptr, _Type *__val, int __memorder, " << scopenametag(s.first) << ") {\n";
out << " uint" << (registers[sz] == "r" ? 32 : sz) << "_t __tmp = 0;\n";
out << " memcpy(&__tmp, __val, " << sz/8 << ");\n";
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_RELEASE: __cuda_store_release_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "();\n";
out << " case __ATOMIC_RELAXED: __cuda_store_relaxed_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_RELEASE:\n";
out << " case __ATOMIC_SEQ_CST: __cuda_membar_" << s.first << "();\n";
out << " case __ATOMIC_RELAXED: __cuda_store_volatile_" << sz << "_" << s.first << "(__ptr, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
out << "}\n";
}
}
for(auto& sz : rmw_sizes) {
for(auto& rmw: rmw_operations) {
for(auto& sem : rmw_semantics) {
if(rmw.first == "compare_exchange")
out << "template<class _CUDA_A, class _CUDA_B, class _CUDA_C, class _CUDA_D> ";
else
out << "template<class _CUDA_A, class _CUDA_B, class _CUDA_C> ";
out << "static inline __device__ void __cuda_" << rmw.first << "_" << sem.first << "_" << sz << "_" << s.first << "(";
if(rmw.first == "compare_exchange")
out << "_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __cmp, _CUDA_D __op";
else
out << "_CUDA_A __ptr, _CUDA_B& __dst, _CUDA_C __op";
out << ") { ";
if(rmw.first == "fetch_sub")
out << "__op = -__op;" << std::endl;
if(rmw.first == "fetch_add" || rmw.first == "fetch_sub" || rmw.first == "fetch_max" || rmw.first == "fetch_min")
out << "asm volatile(\"atom" << rmw.second << sem.second << s.second << ".u" << sz << " ";
else
out << "asm volatile(\"atom" << rmw.second << sem.second << s.second << ".b" << sz << " ";
if(rmw.first == "compare_exchange")
out << "%0,[%1],%2,%3";
else
out << "%0,[%1],%2";
out << ";\" : ";
if(rmw.first == "compare_exchange")
out << "\"=" << registers[sz] << "\"(__dst) : \"l\"(__ptr),\"" << registers[sz] << "\"(__cmp),\"" << registers[sz] << "\"(__op)";
else
out << "\"=" << registers[sz] << "\"(__dst) : \"l\"(__ptr),\"" << registers[sz] << "\"(__op)";
out << " : \"memory\"); }\n";
}
for(auto& cv: cv_qualifier) {
if(rmw.first == "compare_exchange") {
out << "template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==" << sz/8 << ", int>::type = 0>\n";
out << "__device__ bool __atomic_compare_exchange_cuda(" << cv << "_Type *__ptr, _Type *__expected, const _Type *__desired, bool, int __success_memorder, int __failure_memorder, " << scopenametag(s.first) << ") {\n";
out << " uint" << sz << "_t __tmp = 0, __old = 0, __old_tmp;\n";
out << " memcpy(&__tmp, __desired, " << sz/8 << ");\n";
out << " memcpy(&__old, __expected, " << sz/8 << ");\n";
out << " __old_tmp = __old;\n";
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_compare_exchange_acquire_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " case __ATOMIC_ACQ_REL: __cuda_compare_exchange_acq_rel_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_compare_exchange_release_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_compare_exchange_relaxed_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__stronger_order_cuda(__success_memorder, __failure_memorder)) {\n";
out << " case __ATOMIC_SEQ_CST:\n";
out << " case __ATOMIC_ACQ_REL: __cuda_membar_" << s.first << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_compare_exchange_volatile_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); __cuda_membar_" << s.first << "(); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_membar_" << s.first << "(); __cuda_compare_exchange_volatile_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_compare_exchange_volatile_" << sz << "_" << s.first << "(__ptr, __old, __old_tmp, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
out << " bool const __ret = __old == __old_tmp;\n";
out << " memcpy(__expected, &__old, " << sz/8 << ");\n";
out << " return __ret;\n";
out << "}\n";
}
else {
out << "template<class _Type, typename cuda::std::enable_if<sizeof(_Type)==" << sz/8 << ", int>::type = 0>\n";
if(rmw.first == "exchange") {
out << "__device__ void __atomic_exchange_cuda(" << cv << "_Type *__ptr, _Type *__val, _Type *__ret, int __memorder, " << scopenametag(s.first) << ") {\n";
out << " uint" << sz << "_t __tmp = 0;\n";
out << " memcpy(&__tmp, __val, " << sz/8 << ");\n";
}
else {
out << "__device__ _Type __atomic_" << rmw.first << "_cuda(" << cv << "_Type *__ptr, _Type __val, int __memorder, " << scopenametag(s.first) << ") {\n";
out << " _Type __ret;\n";
out << " uint" << sz << "_t __tmp = 0;\n";
out << " memcpy(&__tmp, &__val, " << sz/8 << ");\n";
}
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_" << rmw.first << "_acquire_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_ACQ_REL: __cuda_" << rmw.first << "_acq_rel_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_" << rmw.first << "_release_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_" << rmw.first << "_relaxed_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST:\n";
out << " case __ATOMIC_ACQ_REL: __cuda_membar_" << s.first << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_" << rmw.first << "_volatile_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); __cuda_membar_" << s.first << "(); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_membar_" << s.first << "(); __cuda_" << rmw.first << "_volatile_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_" << rmw.first << "_volatile_" << sz << "_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
if(rmw.first == "exchange")
out << " memcpy(__ret, &__tmp, " << sz/8 << ");\n";
else {
out << " memcpy(&__ret, &__tmp, " << sz/8 << ");\n";
out << " return __ret;\n";
}
out << "}\n";
}
}
}
}
for(auto& cv: cv_qualifier) {
std::vector<std::string> addsub{ "add", "sub" };
for(auto& op : addsub) {
out << "template<class _Type>\n";
out << "__device__ _Type* __atomic_fetch_" << op << "_cuda(_Type *" << cv << "*__ptr, ptrdiff_t __val, int __memorder, " << scopenametag(s.first) << ") {\n";
out << " _Type* __ret;\n";
out << " uint64_t __tmp = 0;\n";
out << " memcpy(&__tmp, &__val, 8);\n";
if(op == "sub")
out << " __tmp = -__tmp;\n";
out << " __tmp *= sizeof(_Type);\n";
out << " NV_DISPATCH_TARGET(\n";
out << " NV_PROVIDES_SM_70, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST: " << fencename("sc"s, s.first) << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_fetch_add_acquire_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_ACQ_REL: __cuda_fetch_add_acq_rel_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_fetch_add_release_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_fetch_add_relaxed_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " }\n";
out << " ),\n";
out << " NV_IS_DEVICE, (\n";
out << " switch (__memorder) {\n";
out << " case __ATOMIC_SEQ_CST:\n";
out << " case __ATOMIC_ACQ_REL: __cuda_membar_" << s.first << "();\n";
out << " case __ATOMIC_CONSUME:\n";
out << " case __ATOMIC_ACQUIRE: __cuda_fetch_add_volatile_64_" << s.first << "(__ptr, __tmp, __tmp); __cuda_membar_" << s.first << "(); break;\n";
out << " case __ATOMIC_RELEASE: __cuda_membar_" << s.first << "(); __cuda_fetch_add_volatile_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " case __ATOMIC_RELAXED: __cuda_fetch_add_volatile_64_" << s.first << "(__ptr, __tmp, __tmp); break;\n";
out << " default: assert(0);\n";
out << " }\n";
out << " )\n";
out << " )\n";
out << " memcpy(&__ret, &__tmp, 8);\n";
out << " return __ret;\n";
out << "}\n";
}
}
}
return 0;
}
|
