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
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
|
/*===-------------- avxneconvertintrin.h - AVXNECONVERT --------------------===
*
* Part of the LLVM Project, 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
*
*===-----------------------------------------------------------------------===
*/
#ifndef __IMMINTRIN_H
#error \
"Never use <avxneconvertintrin.h> directly; include <immintrin.h> instead."
#endif // __IMMINTRIN_H
#ifdef __SSE2__
#ifndef __AVXNECONVERTINTRIN_H
#define __AVXNECONVERTINTRIN_H
/* Define the default attributes for the functions in this file. */
#define __DEFAULT_FN_ATTRS128 \
__attribute__((__always_inline__, __nodebug__, __target__("avxneconvert"), \
__min_vector_width__(128)))
#define __DEFAULT_FN_ATTRS256 \
__attribute__((__always_inline__, __nodebug__, __target__("avxneconvert"), \
__min_vector_width__(256)))
/// Convert scalar BF16 (16-bit) floating-point element
/// stored at memory locations starting at location \a __A to a
/// single-precision (32-bit) floating-point, broadcast it to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_bcstnebf16_ps(const void *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VBCSTNEBF162PS instruction.
///
/// \param __A
/// A pointer to a 16-bit memory location. The address of the memory
/// location does not have to be aligned.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// b := Convert_BF16_To_FP32(MEM[__A+15:__A])
/// FOR j := 0 to 3
/// m := j*32
/// dst[m+31:m] := b
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_bcstnebf16_ps(const void *__A) {
return (__m128)__builtin_ia32_vbcstnebf162ps128((const __bf16 *)__A);
}
/// Convert scalar BF16 (16-bit) floating-point element
/// stored at memory locations starting at location \a __A to a
/// single-precision (32-bit) floating-point, broadcast it to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_bcstnebf16_ps(const void *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VBCSTNEBF162PS instruction.
///
/// \param __A
/// A pointer to a 16-bit memory location. The address of the memory
/// location does not have to be aligned.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// b := Convert_BF16_To_FP32(MEM[__A+15:__A])
/// FOR j := 0 to 7
/// m := j*32
/// dst[m+31:m] := b
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_bcstnebf16_ps(const void *__A) {
return (__m256)__builtin_ia32_vbcstnebf162ps256((const __bf16 *)__A);
}
/// Convert scalar half-precision (16-bit) floating-point element
/// stored at memory locations starting at location \a __A to a
/// single-precision (32-bit) floating-point, broadcast it to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_bcstnesh_ps(const void *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VBCSTNESH2PS instruction.
///
/// \param __A
/// A pointer to a 16-bit memory location. The address of the memory
/// location does not have to be aligned.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// b := Convert_FP16_To_FP32(MEM[__A+15:__A])
/// FOR j := 0 to 3
/// m := j*32
/// dst[m+31:m] := b
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_bcstnesh_ps(const void *__A) {
return (__m128)__builtin_ia32_vbcstnesh2ps128((const _Float16 *)__A);
}
/// Convert scalar half-precision (16-bit) floating-point element
/// stored at memory locations starting at location \a __A to a
/// single-precision (32-bit) floating-point, broadcast it to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_bcstnesh_ps(const void *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VBCSTNESH2PS instruction.
///
/// \param __A
/// A pointer to a 16-bit memory location. The address of the memory
/// location does not have to be aligned.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// b := Convert_FP16_To_FP32(MEM[__A+15:__A])
/// FOR j := 0 to 7
/// m := j*32
/// dst[m+31:m] := b
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_bcstnesh_ps(const void *__A) {
return (__m256)__builtin_ia32_vbcstnesh2ps256((const _Float16 *)__A);
}
/// Convert packed BF16 (16-bit) floating-point even-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_cvtneebf16_ps(const __m128bh *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEEBF162PS instruction.
///
/// \param __A
/// A pointer to a 128-bit memory location containing 8 consecutive
/// BF16 (16-bit) floating-point values.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// FOR j := 0 to 3
/// k := j*2
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_BF16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_cvtneebf16_ps(const __m128bh *__A) {
return (__m128)__builtin_ia32_vcvtneebf162ps128((const __v8bf *)__A);
}
/// Convert packed BF16 (16-bit) floating-point even-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_cvtneebf16_ps(const __m256bh *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEEBF162PS instruction.
///
/// \param __A
/// A pointer to a 256-bit memory location containing 16 consecutive
/// BF16 (16-bit) floating-point values.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// FOR j := 0 to 7
/// k := j*2
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_BF16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_cvtneebf16_ps(const __m256bh *__A) {
return (__m256)__builtin_ia32_vcvtneebf162ps256((const __v16bf *)__A);
}
/// Convert packed half-precision (16-bit) floating-point even-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_cvtneeph_ps(const __m128h *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEEPH2PS instruction.
///
/// \param __A
/// A pointer to a 128-bit memory location containing 8 consecutive
/// half-precision (16-bit) floating-point values.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// FOR j := 0 to 3
/// k := j*2
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_FP16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_cvtneeph_ps(const __m128h *__A) {
return (__m128)__builtin_ia32_vcvtneeph2ps128((const __v8hf *)__A);
}
/// Convert packed half-precision (16-bit) floating-point even-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_cvtneeph_ps(const __m256h *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEEPH2PS instruction.
///
/// \param __A
/// A pointer to a 256-bit memory location containing 16 consecutive
/// half-precision (16-bit) floating-point values.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// FOR j := 0 to 7
/// k := j*2
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_FP16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_cvtneeph_ps(const __m256h *__A) {
return (__m256)__builtin_ia32_vcvtneeph2ps256((const __v16hf *)__A);
}
/// Convert packed BF16 (16-bit) floating-point odd-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_cvtneobf16_ps(const __m128bh *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEOBF162PS instruction.
///
/// \param __A
/// A pointer to a 128-bit memory location containing 8 consecutive
/// BF16 (16-bit) floating-point values.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// FOR j := 0 to 3
/// k := j*2+1
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_BF16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_cvtneobf16_ps(const __m128bh *__A) {
return (__m128)__builtin_ia32_vcvtneobf162ps128((const __v8bf *)__A);
}
/// Convert packed BF16 (16-bit) floating-point odd-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_cvtneobf16_ps(const __m256bh *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEOBF162PS instruction.
///
/// \param __A
/// A pointer to a 256-bit memory location containing 16 consecutive
/// BF16 (16-bit) floating-point values.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// FOR j := 0 to 7
/// k := j*2+1
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_BF16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_cvtneobf16_ps(const __m256bh *__A) {
return (__m256)__builtin_ia32_vcvtneobf162ps256((const __v16bf *)__A);
}
/// Convert packed half-precision (16-bit) floating-point odd-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_cvtneoph_ps(const __m128h *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEOPH2PS instruction.
///
/// \param __A
/// A pointer to a 128-bit memory location containing 8 consecutive
/// half-precision (16-bit) floating-point values.
/// \returns
/// A 128-bit vector of [4 x float].
///
/// \code{.operation}
/// FOR j := 0 to 3
/// k := j*2+1
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_FP16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_cvtneoph_ps(const __m128h *__A) {
return (__m128)__builtin_ia32_vcvtneoph2ps128((const __v8hf *)__A);
}
/// Convert packed half-precision (16-bit) floating-point odd-indexed elements
/// stored at memory locations starting at location \a __A to packed
/// single-precision (32-bit) floating-point elements, and store the results in
/// \a dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_cvtneoph_ps(const __m256h *__A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEOPH2PS instruction.
///
/// \param __A
/// A pointer to a 256-bit memory location containing 16 consecutive
/// half-precision (16-bit) floating-point values.
/// \returns
/// A 256-bit vector of [8 x float].
///
/// \code{.operation}
/// FOR j := 0 to 7
/// k := j*2+1
/// i := k*16
/// m := j*32
/// dst[m+31:m] := Convert_FP16_To_FP32(MEM[__A+i+15:__A+i])
/// ENDFOR
/// dst[MAX:256] := 0
/// \endcode
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_cvtneoph_ps(const __m256h *__A) {
return (__m256)__builtin_ia32_vcvtneoph2ps256((const __v16hf *)__A);
}
/// Convert packed single-precision (32-bit) floating-point elements in \a __A
/// to packed BF16 (16-bit) floating-point elements, and store the results in \a
/// dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm_cvtneps_avx_pbh(__m128 __A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEPS2BF16 instruction.
///
/// \param __A
/// A 128-bit vector of [4 x float].
/// \returns
/// A 128-bit vector of [8 x bfloat].
///
/// \code{.operation}
/// FOR j := 0 to 3
/// dst.word[j] := Convert_FP32_To_BF16(__A.fp32[j])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128bh __DEFAULT_FN_ATTRS128
_mm_cvtneps_avx_pbh(__m128 __A) {
return (__m128bh)__builtin_ia32_vcvtneps2bf16128((__v4sf)__A);
}
/// Convert packed single-precision (32-bit) floating-point elements in \a __A
/// to packed BF16 (16-bit) floating-point elements, and store the results in \a
/// dst.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// _mm256_cvtneps_avx_pbh(__m256 __A);
/// \endcode
///
/// This intrinsic corresponds to the \c VCVTNEPS2BF16 instruction.
///
/// \param __A
/// A 256-bit vector of [8 x float].
/// \returns
/// A 128-bit vector of [8 x bfloat].
///
/// \code{.operation}
/// FOR j := 0 to 7
/// dst.word[j] := Convert_FP32_To_BF16(a.fp32[j])
/// ENDFOR
/// dst[MAX:128] := 0
/// \endcode
static __inline__ __m128bh __DEFAULT_FN_ATTRS256
_mm256_cvtneps_avx_pbh(__m256 __A) {
return (__m128bh)__builtin_ia32_vcvtneps2bf16256((__v8sf)__A);
}
#undef __DEFAULT_FN_ATTRS128
#undef __DEFAULT_FN_ATTRS256
#endif // __AVXNECONVERTINTRIN_H
#endif // __SSE2__
|
