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
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
|
/*
* Copyright 2011-2013 Intel Corporation
* Modifications Copyright 2014, Blender Foundation.
*
* Licensed under the Apache License, Version 2.0(the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#ifndef __UTIL_SIMD_TYPES_H__
#define __UTIL_SIMD_TYPES_H__
#include <limits>
#include <stdint.h>
#include "util/util_defines.h"
/* SSE Intrinsics includes
*
* We assume __KERNEL_SSEX__ flags to have been defined at this point.
*
* MinGW64 has conflicting declarations for these SSE headers in <windows.h>.
* Since we can't avoid including <windows.h>, better only include that */
#if defined(FREE_WINDOWS64)
# include "util/util_windows.h"
#elif defined(_MSC_VER)
# include <intrin.h>
#elif (defined(__x86_64__) || defined(__i386__))
# include <x86intrin.h>
#endif
/* Floating Point Control, for Embree. */
#if defined(__x86_64__) || defined(_M_X64)
# define SIMD_SET_FLUSH_TO_ZERO \
_MM_SET_FLUSH_ZERO_MODE(_MM_FLUSH_ZERO_ON); \
_MM_SET_DENORMALS_ZERO_MODE(_MM_DENORMALS_ZERO_ON);
#else
# define SIMD_SET_FLUSH_TO_ZERO
#endif
CCL_NAMESPACE_BEGIN
/* Data structures used by SSE classes. */
#ifdef __KERNEL_SSE2__
extern const __m128 _mm_lookupmask_ps[16];
static struct TrueTy {
__forceinline operator bool() const
{
return true;
}
} True ccl_maybe_unused;
static struct FalseTy {
__forceinline operator bool() const
{
return false;
}
} False ccl_maybe_unused;
static struct ZeroTy {
__forceinline operator float() const
{
return 0;
}
__forceinline operator int() const
{
return 0;
}
} zero ccl_maybe_unused;
static struct OneTy {
__forceinline operator float() const
{
return 1;
}
__forceinline operator int() const
{
return 1;
}
} one ccl_maybe_unused;
static struct NegInfTy {
__forceinline operator float() const
{
return -std::numeric_limits<float>::infinity();
}
__forceinline operator int() const
{
return std::numeric_limits<int>::min();
}
} neg_inf ccl_maybe_unused;
static struct PosInfTy {
__forceinline operator float() const
{
return std::numeric_limits<float>::infinity();
}
__forceinline operator int() const
{
return std::numeric_limits<int>::max();
}
} inf ccl_maybe_unused, pos_inf ccl_maybe_unused;
static struct StepTy {
} step ccl_maybe_unused;
#endif
/* Intrinsics Functions
*
* For fast bit operations. */
#if defined(__BMI__) && defined(__GNUC__)
# ifndef _tzcnt_u32
# define _tzcnt_u32 __tzcnt_u32
# endif
# ifndef _tzcnt_u64
# define _tzcnt_u64 __tzcnt_u64
# endif
#endif
#if defined(__LZCNT__)
# define _lzcnt_u32 __lzcnt32
# define _lzcnt_u64 __lzcnt64
#endif
#if defined(_WIN32) && !defined(__MINGW32__) && !defined(__clang__)
/* Intrinsic functions on Windows. */
__forceinline uint32_t __bsf(uint32_t v)
{
# if defined(__KERNEL_AVX2__)
return _tzcnt_u32(v);
# else
unsigned long r = 0;
_BitScanForward(&r, v);
return r;
# endif
}
__forceinline uint32_t __bsf(uint32_t v)
{
# if defined(__KERNEL_AVX2__)
return _tzcnt_u32(v);
# else
unsigned long r = 0;
_BitScanForward(&r, v);
return r;
# endif
}
__forceinline uint32_t __bsr(uint32_t v)
{
unsigned long r = 0;
_BitScanReverse(&r, v);
return r;
}
__forceinline uint32_t __btc(uint32_t v, uint32_t i)
{
long r = v;
_bittestandcomplement(&r, i);
return r;
}
__forceinline uint32_t bitscan(uint32_t v)
{
# if defined(__KERNEL_AVX2__)
return _tzcnt_u32(v);
# else
return __bsf(v);
# endif
}
# if defined(__KERNEL_64_BIT__)
__forceinline uint64_t __bsf(uint64_t v)
{
# if defined(__KERNEL_AVX2__)
return _tzcnt_u64(v);
# else
unsigned long r = 0;
_BitScanForward64(&r, v);
return r;
# endif
}
__forceinline uint64_t __bsr(uint64_t v)
{
unsigned long r = 0;
_BitScanReverse64(&r, v);
return r;
}
__forceinline uint64_t __btc(uint64_t v, uint64_t i)
{
uint64_t r = v;
_bittestandcomplement64((__int64 *)&r, i);
return r;
}
__forceinline uint64_t bitscan(uint64_t v)
{
# if defined(__KERNEL_AVX2__)
# if defined(__KERNEL_64_BIT__)
return _tzcnt_u64(v);
# else
return _tzcnt_u32(v);
# endif
# else
return __bsf(v);
# endif
}
# endif /* __KERNEL_64_BIT__ */
#elif (defined(__x86_64__) || defined(__i386__)) && defined(__KERNEL_SSE2__)
/* Instrinsic functions with x86 SSE. */
__forceinline uint32_t __bsf(const uint32_t v)
{
uint32_t r = 0;
asm("bsf %1,%0" : "=r"(r) : "r"(v));
return r;
}
__forceinline uint32_t __bsr(const uint32_t v)
{
uint32_t r = 0;
asm("bsr %1,%0" : "=r"(r) : "r"(v));
return r;
}
__forceinline uint32_t __btc(const uint32_t v, uint32_t i)
{
uint32_t r = 0;
asm("btc %1,%0" : "=r"(r) : "r"(i), "0"(v) : "flags");
return r;
}
# if (defined(__KERNEL_64_BIT__) || defined(__APPLE__)) && \
!(defined(__ILP32__) && defined(__x86_64__))
__forceinline uint64_t __bsf(const uint64_t v)
{
uint64_t r = 0;
asm("bsf %1,%0" : "=r"(r) : "r"(v));
return r;
}
# endif
__forceinline uint64_t __bsr(const uint64_t v)
{
uint64_t r = 0;
asm("bsr %1,%0" : "=r"(r) : "r"(v));
return r;
}
__forceinline uint64_t __btc(const uint64_t v, const uint64_t i)
{
uint64_t r = 0;
asm("btc %1,%0" : "=r"(r) : "r"(i), "0"(v) : "flags");
return r;
}
__forceinline uint32_t bitscan(uint32_t v)
{
# if defined(__KERNEL_AVX2__)
return _tzcnt_u32(v);
# else
return __bsf(v);
# endif
}
# if (defined(__KERNEL_64_BIT__) || defined(__APPLE__)) && \
!(defined(__ILP32__) && defined(__x86_64__))
__forceinline uint64_t bitscan(uint64_t v)
{
# if defined(__KERNEL_AVX2__)
# if defined(__KERNEL_64_BIT__)
return _tzcnt_u64(v);
# else
return _tzcnt_u32(v);
# endif
# else
return __bsf(v);
# endif
}
# endif
#else
/* Intrinsic functions fallback for arbitrary processor. */
__forceinline uint32_t __bsf(const uint32_t x)
{
for (int i = 0; i < 32; i++) {
if (x & (1U << i))
return i;
}
return 32;
}
__forceinline uint32_t __bsr(const uint32_t x)
{
for (int i = 0; i < 32; i++) {
if (x & (1U << (31 - i)))
return (31 - i);
}
return 32;
}
__forceinline uint32_t __btc(const uint32_t x, const uint32_t bit)
{
uint32_t mask = 1U << bit;
return x & (~mask);
}
__forceinline uint32_t __bsf(const uint64_t x)
{
for (int i = 0; i < 64; i++) {
if (x & (1UL << i))
return i;
}
return 64;
}
__forceinline uint32_t __bsr(const uint64_t x)
{
for (int i = 0; i < 64; i++) {
if (x & (1UL << (63 - i)))
return (63 - i);
}
return 64;
}
__forceinline uint64_t __btc(const uint64_t x, const uint32_t bit)
{
uint64_t mask = 1UL << bit;
return x & (~mask);
}
__forceinline uint32_t bitscan(uint32_t value)
{
assert(value != 0);
uint32_t bit = 0;
while ((value & (1 << bit)) == 0) {
++bit;
}
return bit;
}
__forceinline uint64_t bitscan(uint64_t value)
{
assert(value != 0);
uint64_t bit = 0;
while ((value & (1 << bit)) == 0) {
++bit;
}
return bit;
}
#endif /* Intrinsics */
/* SSE compatibility.
*
* Various utilities to smooth over differences between SSE versions and
* implementations. */
#ifdef __KERNEL_SSE2__
/* Test __KERNEL_SSE41__ for MSVC which does not define __SSE4_1__, and test
* __SSE4_1__ to avoid OpenImageIO conflicts with our emulation macros on other
* platforms when compiling code outside the kernel. */
# if !(defined(__KERNEL_SSE41__) || defined(__SSE4_1__) || defined(__SSE4_2__))
/* Emulation of SSE4 functions with SSE2 */
# define _MM_FROUND_TO_NEAREST_INT 0x00
# define _MM_FROUND_TO_NEG_INF 0x01
# define _MM_FROUND_TO_POS_INF 0x02
# define _MM_FROUND_TO_ZERO 0x03
# define _MM_FROUND_CUR_DIRECTION 0x04
# undef _mm_blendv_ps
# define _mm_blendv_ps _mm_blendv_ps_emu
__forceinline __m128 _mm_blendv_ps_emu(__m128 value, __m128 input, __m128 mask)
{
__m128i isignmask = _mm_set1_epi32(0x80000000);
__m128 signmask = _mm_castsi128_ps(isignmask);
__m128i iandsign = _mm_castps_si128(_mm_and_ps(mask, signmask));
__m128i icmpmask = _mm_cmpeq_epi32(iandsign, isignmask);
__m128 cmpmask = _mm_castsi128_ps(icmpmask);
return _mm_or_ps(_mm_and_ps(cmpmask, input), _mm_andnot_ps(cmpmask, value));
}
# undef _mm_blend_ps
# define _mm_blend_ps _mm_blend_ps_emu
__forceinline __m128 _mm_blend_ps_emu(__m128 value, __m128 input, const int mask)
{
assert(mask < 0x10);
return _mm_blendv_ps(value, input, _mm_lookupmask_ps[mask]);
}
# undef _mm_blendv_epi8
# define _mm_blendv_epi8 _mm_blendv_epi8_emu
__forceinline __m128i _mm_blendv_epi8_emu(__m128i value, __m128i input, __m128i mask)
{
return _mm_or_si128(_mm_and_si128(mask, input), _mm_andnot_si128(mask, value));
}
# undef _mm_min_epi32
# define _mm_min_epi32 _mm_min_epi32_emu
__forceinline __m128i _mm_min_epi32_emu(__m128i value, __m128i input)
{
return _mm_blendv_epi8(input, value, _mm_cmplt_epi32(value, input));
}
# undef _mm_max_epi32
# define _mm_max_epi32 _mm_max_epi32_emu
__forceinline __m128i _mm_max_epi32_emu(__m128i value, __m128i input)
{
return _mm_blendv_epi8(value, input, _mm_cmplt_epi32(value, input));
}
# undef _mm_extract_epi32
# define _mm_extract_epi32 _mm_extract_epi32_emu
__forceinline int _mm_extract_epi32_emu(__m128i input, const int index)
{
switch (index) {
case 0:
return _mm_cvtsi128_si32(input);
case 1:
return _mm_cvtsi128_si32(_mm_shuffle_epi32(input, _MM_SHUFFLE(1, 1, 1, 1)));
case 2:
return _mm_cvtsi128_si32(_mm_shuffle_epi32(input, _MM_SHUFFLE(2, 2, 2, 2)));
case 3:
return _mm_cvtsi128_si32(_mm_shuffle_epi32(input, _MM_SHUFFLE(3, 3, 3, 3)));
default:
assert(false);
return 0;
}
}
# undef _mm_insert_epi32
# define _mm_insert_epi32 _mm_insert_epi32_emu
__forceinline __m128i _mm_insert_epi32_emu(__m128i value, int input, const int index)
{
assert(index >= 0 && index < 4);
((int *)&value)[index] = input;
return value;
}
# undef _mm_insert_ps
# define _mm_insert_ps _mm_insert_ps_emu
__forceinline __m128 _mm_insert_ps_emu(__m128 value, __m128 input, const int index)
{
assert(index < 0x100);
((float *)&value)[(index >> 4) & 0x3] = ((float *)&input)[index >> 6];
return _mm_andnot_ps(_mm_lookupmask_ps[index & 0xf], value);
}
# undef _mm_round_ps
# define _mm_round_ps _mm_round_ps_emu
__forceinline __m128 _mm_round_ps_emu(__m128 value, const int flags)
{
switch (flags) {
case _MM_FROUND_TO_NEAREST_INT:
return _mm_cvtepi32_ps(_mm_cvtps_epi32(value));
case _MM_FROUND_TO_NEG_INF:
return _mm_cvtepi32_ps(_mm_cvtps_epi32(_mm_add_ps(value, _mm_set1_ps(-0.5f))));
case _MM_FROUND_TO_POS_INF:
return _mm_cvtepi32_ps(_mm_cvtps_epi32(_mm_add_ps(value, _mm_set1_ps(0.5f))));
case _MM_FROUND_TO_ZERO:
return _mm_cvtepi32_ps(_mm_cvttps_epi32(value));
}
return value;
}
# endif /* !(defined(__KERNEL_SSE41__) || defined(__SSE4_1__) || defined(__SSE4_2__)) */
/* Older GCC versions do not have _mm256_cvtss_f32 yet, so define it ourselves.
* _mm256_castps256_ps128 generates no instructions so this is just as efficient. */
# if defined(__KERNEL_AVX__) || defined(__KERNEL_AVX2__)
# undef _mm256_cvtss_f32
# define _mm256_cvtss_f32(a) (_mm_cvtss_f32(_mm256_castps256_ps128(a)))
# endif
#endif /* __KERNEL_SSE2__ */
/* quiet unused define warnings */
#if defined(__KERNEL_SSE2__) || defined(__KERNEL_SSE3__) || defined(__KERNEL_SSSE3__) || \
defined(__KERNEL_SSE41__) || defined(__KERNEL_AVX__) || defined(__KERNEL_AVX2__)
/* do nothing */
#endif
CCL_NAMESPACE_END
#endif /* __UTIL_SIMD_TYPES_H__ */
|