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
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
|
/***************************************************************************
ucl_s_obj_help.h
-------------------
W. Michael Brown
Helper routines for allocating memory for s-objects and performing
host/device updates. (Different routines depending on whether the
same type is used on the host and device).
__________________________________________________________________________
This file is part of the Geryon Unified Coprocessor Library (UCL)
__________________________________________________________________________
begin : Mon May 14 2012
copyright : (C) 2012 by W. Michael Brown
email : brownw@ornl.gov
***************************************************************************/
/* -----------------------------------------------------------------------
This software is distributed under the Simplified BSD License.
----------------------------------------------------------------------- */
template <int st> struct _ucl_s_obj_help;
// Host and device containers are same type
// -- Don't need casting buffers
// -- Can potentially use same memory if shared by accelerator
template <> struct _ucl_s_obj_help<1> {
template <class t1, class t2, class t3>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int cols, UCL_Device &acc,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
if (acc.shared_memory()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 1S\n";
#endif
e1=host.alloc(cols,acc,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(host);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 1NS\n";
#endif
e1=host.alloc(cols,acc,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(cols,acc,kind2);
}
}
template <class t1, class t2, class t3, class mat_type>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int cols, mat_type &cq,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
if (cq.shared_mem_device()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 2S\n";
#endif
e1=host.alloc(cols,cq,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(host);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 2NS\n";
#endif
e1=host.alloc(cols,cq,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(cols,cq,kind2);
}
}
template <class t1, class t2, class t3>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int rows, const int cols, UCL_Device &acc,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
if (acc.shared_memory()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 3S\n";
#endif
e1=host.alloc(rows,cols,acc,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(host);
return UCL_SUCCESS;
} else {
e1=host.alloc(rows,cols,acc,kind1);
if (e1!=UCL_SUCCESS)
return e1;
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 3NS\n";
#endif
return device.alloc(rows,cols,acc,kind2);
}
}
template <class t1, class t2, class t3, class mat_type>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int rows, const int cols, mat_type &cq,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
if (cq.shared_mem_device()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 4S\n";
#endif
e1=host.alloc(rows,cols,cq,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(host);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 4NS\n";
#endif
e1=host.alloc(rows,cols,cq,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(rows,cols,cq,kind2);
}
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, t3 &buffer, const bool async) {
ucl_copy(dst,src,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, t3 &buffer, command_queue &cq) {
ucl_copy(dst,src,cq);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int cols, t3 &buffer,
const bool async) {
ucl_copy(dst,src,cols,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int cols, t3 &buffer,
command_queue &cq) {
ucl_copy(dst,src,cols,cq);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int rows, const int cols,
t3 &buffer, const bool async) {
ucl_copy(dst,src,rows,cols,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int rows, const int cols,
t3 &buffer, command_queue &cq) {
ucl_copy(dst,src,rows,cols,cq);
}
template <class t1, class t2, class t3>
static inline int dev_resize(t1 &device, t2 &host, t3 &buff,const int cols) {
if (device.kind()==UCL_VIEW) {
device.view(host);
return UCL_SUCCESS;
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 5S\n";
#endif
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 5NS\n";
#endif
return device.resize(cols);
}
}
template <class t1, class t2, class t3>
static inline int dev_resize(t1 &device, t2 &host, t3 &buff, const int rows,
const int cols) {
if (device.kind()==UCL_VIEW) {
device.view(host);
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 6S\n";
#endif
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 6NS\n";
#endif
return device.resize(rows,cols);
}
}
};
// Host and device containers are different types
template <int st> struct _ucl_s_obj_help {
template <class t1, class t2, class t3>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int cols, UCL_Device &acc,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
e1=host.alloc(cols,acc,UCL_NOT_PINNED);
if (e1!=UCL_SUCCESS)
return e1;
if (acc.shared_memory()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 7S\n";
#endif
e1=_buffer.alloc(cols,acc,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(_buffer);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 7NS\n";
#endif
e1=_buffer.alloc(cols,acc,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(cols,acc,kind2);
}
}
template <class t1, class t2, class t3, class mat_type>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int cols, mat_type &cq,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
e1=host.alloc(cols,cq,UCL_NOT_PINNED);
if (e1!=UCL_SUCCESS)
return e1;
if (cq.shared_mem_device()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 8S\n";
#endif
e1=_buffer.alloc(cols,cq,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(_buffer);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 8NS\n";
#endif
e1=_buffer.alloc(cols,cq,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(cols,cq,kind2);
}
}
template <class t1, class t2, class t3>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int rows, const int cols, UCL_Device &acc,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
e1=host.alloc(rows,cols,acc,UCL_NOT_PINNED);
if (e1!=UCL_SUCCESS)
return e1;
if (acc.shared_memory()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 9S\n";
#endif
e1=_buffer.alloc(rows,cols,acc,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(_buffer);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 9NS\n";
#endif
e1=_buffer.alloc(rows,cols,acc,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(rows,cols,acc,kind2);
}
}
template <class t1, class t2, class t3, class mat_type>
static inline int alloc(t1 &host, t2 &device, t3 &_buffer,
const int rows, const int cols, mat_type &cq,
const enum UCL_MEMOPT kind1,
const enum UCL_MEMOPT kind2) {
int e1;
e1=host.alloc(rows,cols,cq,UCL_NOT_PINNED);
if (e1!=UCL_SUCCESS)
return e1;
if (cq.shared_mem_device()) {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 10S\n";
#endif
e1=_buffer.alloc(rows,cols,cq,kind1,kind2);
if (e1!=UCL_SUCCESS)
return e1;
device.view(_buffer);
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 10NS\n";
#endif
e1=_buffer.alloc(rows,cols,cq,kind1);
if (e1!=UCL_SUCCESS)
return e1;
return device.alloc(rows,cols,cq,kind2);
}
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, t3 &buffer, const bool async) {
ucl_cast_copy(dst,src,buffer,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, t3 &buffer, command_queue &cq) {
ucl_cast_copy(dst,src,buffer,cq);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int cols, t3 &buffer,
const bool async) {
ucl_cast_copy(dst,src,cols,buffer,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int cols, t3 &buffer,
command_queue &cq) {
ucl_cast_copy(dst,src,cols,buffer,cq);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int rows, const int cols,
t3 &buffer, const bool async) {
ucl_cast_copy(dst,src,rows,cols,buffer,async);
}
template <class t1, class t2, class t3>
static inline void copy(t1 &dst, t2 &src, const int rows, const int cols,
t3 &buffer, command_queue &cq) {
ucl_cast_copy(dst,src,rows,cols,buffer,cq);
}
template <class t1, class t2, class t3>
static inline int dev_resize(t1 &device, t2 &host, t3 &buff,const int cols) {
int err=buff.resize(cols);
if (err!=UCL_SUCCESS)
return err;
if (device.kind()==UCL_VIEW) {
device.view(buff);
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 11S\n";
#endif
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 11NS\n";
#endif
return device.resize(cols);
}
}
template <class t1, class t2, class t3>
static inline int dev_resize(t1 &device, t2 &host, t3 &buff, const int rows,
const int cols) {
int err=buff.resize(rows,cols);
if (err!=UCL_SUCCESS)
return err;
if (device.kind()==UCL_VIEW) {
device.view(buff);
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 12S\n";
#endif
return UCL_SUCCESS;
} else {
#ifdef UCL_DBG_MEM_TRACE
std::cerr << "UCL_ALLOC 12NS\n";
#endif
return device.resize(rows,cols);
}
}
};
|
