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
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2017-2021 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
//===- IBiF_VLoadStore.cl - OpenCL explicit conversion functions -====//
//
// This file defines OpenCL vload/vstore built-ins.
//
//===----------------------------------------------------------------------===//
//*****************************************************************************/
// vload/vstore macros
//*****************************************************************************/
#define VLOAD_MACRO(addressSpace, scalarType, numElements) \
INLINE scalarType##numElements OVERLOADABLE vload##numElements(size_t offset, const addressSpace scalarType *p) \
{ \
const addressSpace scalarType *pOffset = p + offset * numElements; \
scalarType##numElements ret; \
__builtin_IB_memcpy_##addressSpace##_to_private((private uchar*)&ret, (addressSpace uchar*)pOffset, sizeof(scalarType) * numElements, sizeof(scalarType)); \
return ret; \
}
#define VLOAD3_MACRO(addressSpace, scalarType, numElements) \
INLINE scalarType##numElements OVERLOADABLE vload##numElements(size_t offset, const addressSpace scalarType *p) \
{ \
const addressSpace scalarType *pOffset = p + offset * numElements; \
scalarType##numElements ret; \
__builtin_IB_memcpy_##addressSpace##_to_private((private uchar*)&ret, (addressSpace uchar*)pOffset, sizeof(scalarType) * numElements, sizeof(scalarType)); \
return ret; \
}
#define VSTORE_MACRO(addressSpace, scalarType, numElements) \
INLINE void OVERLOADABLE vstore##numElements(scalarType##numElements data, size_t offset, addressSpace scalarType *p) \
{ \
addressSpace scalarType *pOffset = p + offset * numElements; \
scalarType##numElements ret = data; \
__builtin_IB_memcpy_private_to_##addressSpace((addressSpace uchar*)pOffset, (private uchar*)&ret, sizeof(scalarType) * numElements, sizeof(scalarType)); \
}
#define VSTORE3_MACRO(addressSpace, scalarType, numElements) \
INLINE void OVERLOADABLE vstore##numElements(scalarType##numElements data, size_t offset, addressSpace scalarType *p) \
{ \
addressSpace scalarType *pOffset = p + offset * numElements; \
scalarType##numElements ret = data; \
__builtin_IB_memcpy_private_to_##addressSpace((addressSpace uchar*)pOffset, (private uchar*)&ret, sizeof(scalarType) * numElements, sizeof(scalarType)); \
}
#define ELEM_ARG(addressSpace, scalarType) \
VLOAD_MACRO(addressSpace, scalarType, 2) \
VLOAD3_MACRO(addressSpace, scalarType, 3) \
VLOAD_MACRO(addressSpace, scalarType, 4) \
VLOAD_MACRO(addressSpace, scalarType, 8) \
VLOAD_MACRO(addressSpace, scalarType, 16)
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define TYPE_ARG(TYPE) \
ELEM_ARG(global, TYPE) \
ELEM_ARG(constant, TYPE) \
ELEM_ARG(local, TYPE) \
ELEM_ARG(private, TYPE) \
ELEM_ARG(generic, TYPE)
#else
#define TYPE_ARG(TYPE) \
ELEM_ARG(global, TYPE) \
ELEM_ARG(constant, TYPE) \
ELEM_ARG(local, TYPE) \
ELEM_ARG(private, TYPE)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
TYPE_ARG(char)
TYPE_ARG(uchar)
TYPE_ARG(short)
TYPE_ARG(ushort)
TYPE_ARG(int)
TYPE_ARG(uint)
TYPE_ARG(long)
TYPE_ARG(ulong)
#if defined(cl_khr_fp16)
TYPE_ARG(half)
#endif
TYPE_ARG(float)
#if defined(cl_khr_fp64)
TYPE_ARG(double)
#endif
#undef TYPE_ARG
#undef ELEM_ARG
#define ELEM_ARG(addressSpace, scalarType) \
VSTORE_MACRO(addressSpace, scalarType, 2) \
VSTORE3_MACRO(addressSpace, scalarType, 3) \
VSTORE_MACRO(addressSpace, scalarType, 4) \
VSTORE_MACRO(addressSpace, scalarType, 8) \
VSTORE_MACRO(addressSpace, scalarType, 16)
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define TYPE_ARG(TYPE) \
ELEM_ARG(global, TYPE) \
ELEM_ARG(local, TYPE) \
ELEM_ARG(private, TYPE) \
ELEM_ARG(generic, TYPE)
#else
#define TYPE_ARG(TYPE) \
ELEM_ARG(global, TYPE) \
ELEM_ARG(local, TYPE) \
ELEM_ARG(private, TYPE)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
TYPE_ARG(char)
TYPE_ARG(uchar)
TYPE_ARG(short)
TYPE_ARG(ushort)
TYPE_ARG(int)
TYPE_ARG(uint)
TYPE_ARG(long)
TYPE_ARG(ulong)
#if defined(cl_khr_fp16)
TYPE_ARG(half)
#endif
TYPE_ARG(float)
#if defined(cl_khr_fp64)
TYPE_ARG(double)
#endif
#undef TYPE_ARG
#undef ELEM_ARG
//*****************************************************************************/
// vload macros
//*****************************************************************************/
static OVERLOADABLE float __intel_half2float(ushort h)
{
return convert_float(as_half(h));
}
#define VLOAD_SHORT(addressSpace) \
static OVERLOADABLE ushort vload(size_t offset, const addressSpace ushort* p) \
{ \
return *(p + offset); \
}
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
VLOAD_SHORT(__generic)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
VLOAD_SHORT(__global)
VLOAD_SHORT(__local)
VLOAD_SHORT(__constant)
VLOAD_SHORT(__private)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_half2float, float, ushort)
#define __CLFN_DEF_F_VLOAD_SCALAR_HALF(addressSpace) \
INLINE half OVERLOADABLE vload(size_t offset, const addressSpace half* p) { \
return *(p + offset); \
}
#define __CLFN_DEF_F_VLOAD_HALFX(addressSpace, numElements) \
INLINE float##numElements OVERLOADABLE vload_half##numElements(size_t offset, const addressSpace half* p) { \
return __intel_half2float(vload##numElements(offset, (const addressSpace ushort*)p)); \
}
#define __CLFN_DEF_F_VLOADA_HALFX(addressSpace, step, numElements) \
INLINE float##numElements OVERLOADABLE vloada_half##numElements(size_t offset, const addressSpace half* p) { \
const addressSpace ushort##numElements* pHalf = (const addressSpace ushort##numElements*)(p + offset * step); \
return __intel_half2float(*pHalf); \
}
#define __CLFN_DEF_F_VLOAD_HALFX_AS(addressSpace) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, ) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, 2) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, 3) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, 4) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, 8) \
__CLFN_DEF_F_VLOAD_HALFX(addressSpace, 16)
#define __CLFN_DEF_F_VLOADA_HALFX_AS(addressSpace) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 1, ) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 2, 2) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 4, 3) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 4, 4) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 8, 8) \
__CLFN_DEF_F_VLOADA_HALFX(addressSpace, 16, 16)
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define __CLFN_DEF_F_VLOAD_HALF_ALL() \
__CLFN_DEF_F_VLOAD_HALFX_AS(__generic) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__global) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__local) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__constant) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__private)
#else
#define __CLFN_DEF_F_VLOAD_HALF_ALL() \
__CLFN_DEF_F_VLOAD_HALFX_AS(__global) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__local) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__constant) \
__CLFN_DEF_F_VLOAD_HALFX_AS(__private)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define __CLFN_DEF_F_VLOADA_HALF_ALL() \
__CLFN_DEF_F_VLOADA_HALFX_AS(__generic) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__global) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__local) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__constant) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__private)
#else
#define __CLFN_DEF_F_VLOADA_HALF_ALL() \
__CLFN_DEF_F_VLOADA_HALFX_AS(__global) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__local) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__constant) \
__CLFN_DEF_F_VLOADA_HALFX_AS(__private)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
//*****************************************************************************/
// vstore macros (half)
//*****************************************************************************/
// These helper functions are used to macro-ize the rounding mode built-ins.
static OVERLOADABLE half __intel_float2half_rtz(float f)
{
return convert_half_rtz(f);
}
static OVERLOADABLE half __intel_float2half_rte(float f)
{
return convert_half_rte(f);
}
static OVERLOADABLE half __intel_float2half_rtp(float f)
{
return convert_half_rtp(f);
}
static OVERLOADABLE half __intel_float2half_rtn(float f)
{
return convert_half_rtn(f);
}
static OVERLOADABLE half __intel_float2half(float f)
{
return __intel_float2half_rte(f);
}
#if defined(cl_khr_fp64)
static OVERLOADABLE half __intel_double2half_rtz(double a)
{
return convert_half_rtz(a);
}
static OVERLOADABLE half __intel_double2half_rte(double a)
{
return convert_half_rte(a);
}
static OVERLOADABLE half __intel_double2half_rtp(double a)
{
return convert_half_rtp(a);
}
static OVERLOADABLE half __intel_double2half_rtn(double a)
{
return convert_half_rtn(a);
}
static OVERLOADABLE half __intel_double2half(double a)
{
return __intel_double2half_rte(a);
}
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_double2half_rtz, half, double)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_double2half_rte, half, double)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_double2half_rtp, half, double)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_double2half_rtn, half, double)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_double2half, half, double)
#endif
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_float2half_rtz, half, float)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_float2half_rte, half, float)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_float2half_rtp, half, float)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_float2half_rtn, half, float)
GENERATE_VECTOR_FUNCTIONS_1ARG(__intel_float2half, half, float)
#if defined(cl_khr_fp64)
#define __CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, numElements) \
INLINE void OVERLOADABLE vstore_half##numElements##rnd(double##numElements data, \
size_t offset, \
addressSpace half* p) { \
vstore##numElements(__intel_double2half##rnd(data), offset, p); \
}
#define __CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, step, numElements) \
INLINE void OVERLOADABLE vstorea_half##numElements##rnd(double##numElements data, \
size_t offset, \
addressSpace half* p) { \
addressSpace half##numElements *pHalf = (addressSpace half##numElements *)(p + offset * step); \
*pHalf = __intel_double2half##rnd(data); \
}
#endif
#define __CLFN_DEF_VSTORE_SCALAR_HALF(addressSpace) \
INLINE void OVERLOADABLE vstore(half data, \
size_t offset, \
addressSpace half* p) { \
addressSpace half *pHalf = (addressSpace half *)(p + offset); \
*pHalf = data; \
}
#define __CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, numElements) \
INLINE void OVERLOADABLE vstore_half##numElements##rnd(float##numElements data, \
size_t offset, \
addressSpace half* p) { \
vstore##numElements(__intel_float2half##rnd(data), offset, p); \
}
#define __CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, step, numElements) \
INLINE void OVERLOADABLE vstorea_half##numElements##rnd(float##numElements data, \
size_t offset, \
addressSpace half* p) { \
addressSpace half##numElements *pHalf = (addressSpace half##numElements *)(p + offset * step); \
*pHalf = __intel_float2half##rnd(data); \
}
#if defined(cl_khr_fp64)
#define __CLFN_DEF_VSTORE_HALF(addressSpace, rnd) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, ) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 2) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 3) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 4) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 8) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 16) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, ) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, 2) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, 3) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, 4) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, 8) \
__CLFN_DEF_D_VSTORE_HALFX(addressSpace, rnd, 16)
#else
#define __CLFN_DEF_VSTORE_HALF(addressSpace, rnd) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, ) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 2) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 3) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 4) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 8) \
__CLFN_DEF_F_VSTORE_HALFX(addressSpace, rnd, 16)
#endif
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define __CLFN_DEF_VSTORE_HALF_AS(rnd) \
__CLFN_DEF_VSTORE_HALF(generic, rnd) \
__CLFN_DEF_VSTORE_HALF(global, rnd) \
__CLFN_DEF_VSTORE_HALF(local, rnd) \
__CLFN_DEF_VSTORE_HALF(private, rnd)
#else
#define __CLFN_DEF_VSTORE_HALF_AS(rnd) \
__CLFN_DEF_VSTORE_HALF(global, rnd) \
__CLFN_DEF_VSTORE_HALF(local, rnd) \
__CLFN_DEF_VSTORE_HALF(private, rnd)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
#define __CLFN_DEF_VSTORE_HALF_ALL() \
__CLFN_DEF_VSTORE_HALF_AS() \
__CLFN_DEF_VSTORE_HALF_AS(_rte) \
__CLFN_DEF_VSTORE_HALF_AS(_rtz) \
__CLFN_DEF_VSTORE_HALF_AS(_rtp) \
__CLFN_DEF_VSTORE_HALF_AS(_rtn)
#if defined(cl_khr_fp64)
#define __CLFN_DEF_VSTOREA_HALF(addressSpace, rnd) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 1, ) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 2, 2) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 4, 3) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 4, 4) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 8, 8) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 16, 16) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 1, ) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 2, 2) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 4, 3) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 4, 4) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 8, 8) \
__CLFN_DEF_D_VSTOREA_HALFX(addressSpace, rnd, 16, 16)
#else
#define __CLFN_DEF_VSTOREA_HALF(addressSpace, rnd) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 1, ) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 2, 2) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 4, 3) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 4, 4) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 8, 8) \
__CLFN_DEF_F_VSTOREA_HALFX(addressSpace, rnd, 16, 16)
#endif
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
#define __CLFN_DEF_VSTOREA_HALF_AS(rnd) \
__CLFN_DEF_VSTOREA_HALF(generic, rnd) \
__CLFN_DEF_VSTOREA_HALF(global, rnd) \
__CLFN_DEF_VSTOREA_HALF(local, rnd) \
__CLFN_DEF_VSTOREA_HALF(private, rnd)
#else
#define __CLFN_DEF_VSTOREA_HALF_AS(rnd) \
__CLFN_DEF_VSTOREA_HALF(global, rnd) \
__CLFN_DEF_VSTOREA_HALF(local, rnd) \
__CLFN_DEF_VSTOREA_HALF(private, rnd)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
#define __CLFN_DEF_VSTOREA_HALF_ALL() \
__CLFN_DEF_VSTOREA_HALF_AS() \
__CLFN_DEF_VSTOREA_HALF_AS(_rte) \
__CLFN_DEF_VSTOREA_HALF_AS(_rtz) \
__CLFN_DEF_VSTOREA_HALF_AS(_rtp) \
__CLFN_DEF_VSTOREA_HALF_AS(_rtn)
//*****************************************************************************/
// vload/vstore HALF functions
//*****************************************************************************/
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
__CLFN_DEF_F_VLOAD_SCALAR_HALF(__generic)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
__CLFN_DEF_F_VLOAD_SCALAR_HALF(__global)
__CLFN_DEF_F_VLOAD_SCALAR_HALF(__local)
__CLFN_DEF_F_VLOAD_SCALAR_HALF(__constant)
__CLFN_DEF_F_VLOAD_SCALAR_HALF(__private)
#if (__OPENCL_C_VERSION__ >= CL_VERSION_2_0)
__CLFN_DEF_VSTORE_SCALAR_HALF(__generic)
#endif // __OPENCL_C_VERSION__ >= CL_VERSION_2_0
__CLFN_DEF_VSTORE_SCALAR_HALF(__global)
__CLFN_DEF_VSTORE_SCALAR_HALF(__local)
__CLFN_DEF_VSTORE_SCALAR_HALF(__private)
//*****************************************************************************/
// vload/vstore HALF functions
//*****************************************************************************/
__CLFN_DEF_F_VLOAD_HALF_ALL()
__CLFN_DEF_F_VLOADA_HALF_ALL()
__CLFN_DEF_VSTORE_HALF_ALL()
__CLFN_DEF_VSTOREA_HALF_ALL()
|
