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
|
/*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "Simd/SimdMemory.h"
#include "Simd/SimdBase.h"
#include "Simd/SimdStore.h"
namespace Simd
{
#ifdef SIMD_SSE41_ENABLE
namespace Sse41
{
void ImageErosion(uint8_t * img, const uint8_t * buff, size_t width, size_t height, SimdImageConnexityType connexityType)
{
const size_t buffWidth = width + 2;
const size_t alignedSize = Simd::AlignLo(width, A);
if (connexityType == SimdImageConnexity4) {
size_t offset[5] = {1, buffWidth, buffWidth + 1, buffWidth + 2, buffWidth * 2 + 1};
for (size_t i = 0; i < height; i++) {
const uint8_t *ptr_buff = buff + i * buffWidth;
uint8_t *ptr_img = img + i * width;
for (size_t j = 0; j < alignedSize; j += A) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[0]));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[1])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[2])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[3])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[4])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + j), m);
}
if (alignedSize != width) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[0]));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[1])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[2])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[3])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[4])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + (width - A)), m);
}
}
} else {
size_t offset[9] = { 0,
1,
2,
buffWidth,
buffWidth + 1,
buffWidth + 2,
buffWidth * 2,
buffWidth * 2 + 1,
buffWidth * 2 + 2 };
for (size_t i = 0; i < height; i++) {
const uint8_t *ptr_buff = buff + i * buffWidth;
uint8_t *ptr_img = img + i * width;
for (size_t j = 0; j < alignedSize; j += A) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[0]));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[1])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[2])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[3])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[4])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[5])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[6])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[7])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[8])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + j), m);
}
if (alignedSize != width) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[0]));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[1])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[2])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[3])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[4])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[5])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[6])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[7])));
m = _mm_min_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[8])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + (width - A)), m);
}
}
}
}
void ImageDilatation(uint8_t * img, const uint8_t * buff, size_t width, size_t height, SimdImageConnexityType connexityType)
{
const size_t buffWidth = width + 2;
const size_t alignedSize = Simd::AlignLo(width, A);
if (connexityType == SimdImageConnexity4) {
size_t offset[5] = {1, buffWidth, buffWidth + 1, buffWidth + 2, buffWidth * 2 + 1};
for (size_t i = 0; i < height; i++) {
const uint8_t *ptr_buff = buff + i * buffWidth;
uint8_t *ptr_img = img + i * width;
for (size_t j = 0; j < alignedSize; j += A) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[0]));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[1])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[2])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[3])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[4])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + j), m);
}
if (alignedSize != width) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[0]));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[1])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[2])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[3])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[4])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + (width - A)), m);
}
}
} else {
size_t offset[9] = { 0,
1,
2,
buffWidth,
buffWidth + 1,
buffWidth + 2,
buffWidth * 2,
buffWidth * 2 + 1,
buffWidth * 2 + 2 };
for (size_t i = 0; i < height; i++) {
const uint8_t *ptr_buff = buff + i * buffWidth;
uint8_t *ptr_img = img + i * width;
for (size_t j = 0; j < alignedSize; j += A) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[0]));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[1])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[2])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[3])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[4])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[5])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[6])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[7])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + j + offset[8])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + j), m);
}
if (alignedSize != width) {
__m128i m = _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[0]));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[1])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[2])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[3])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[4])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[5])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[6])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[7])));
m = _mm_max_epu8(m, _mm_loadu_si128(reinterpret_cast<const __m128i *>(ptr_buff + (width - A) + offset[8])));
_mm_storeu_si128(reinterpret_cast<__m128i *>(ptr_img + (width - A)), m);
}
}
}
}
double SimdVectorSum(const double * vec, size_t size)
{
__m128d v_sum1 = _mm_setzero_pd(), v_sum2 = _mm_setzero_pd();
size_t i = 0;
for (; i <= size - 4; i += 4) {
v_sum1 = _mm_add_pd(_mm_loadu_pd(vec + i), v_sum1);
v_sum2 = _mm_add_pd(_mm_loadu_pd(vec + i + 2), v_sum2);
}
__m128d v_sum = _mm_add_pd(v_sum1, v_sum2);
double res[2];
_mm_storeu_pd(res, v_sum);
double sum = res[0] + res[1];
// tail processing
for (; i < size; i++) {
sum += vec[i];
}
return sum;
}
double SimdVectorSumSquare(const double * vec, size_t size)
{
double sum_square = 0.0;
size_t i = 0;
__m128d v_mul1, v_mul2;
__m128d v_sum = _mm_setzero_pd();
for (; i <= size - 4; i += 4) {
v_mul1 = _mm_mul_pd(_mm_loadu_pd(vec + i), _mm_loadu_pd(vec + i));
v_mul2 = _mm_mul_pd(_mm_loadu_pd(vec + i + 2), _mm_loadu_pd(vec + i + 2));
v_sum = _mm_add_pd(v_mul1, v_sum);
v_sum = _mm_add_pd(v_mul2, v_sum);
}
double res[2];
_mm_storeu_pd(res, v_sum);
sum_square = res[0] + res[1];
for (; i < size; i++) {
sum_square += vec[i] * vec[i];
}
return sum_square;
}
double SimdVectorStdev(const double * vec, size_t size, bool useBesselCorrection)
{
double mean_value = SimdVectorSum(vec, size) / size;
double sum_squared_diff = 0.0;
size_t i = 0;
__m128d v_sub, v_mul, v_sum = _mm_setzero_pd();
__m128d v_mean = _mm_set1_pd(mean_value);
for (; i <= size - 4; i += 4) {
v_sub = _mm_sub_pd(_mm_loadu_pd(vec + i), v_mean);
v_mul = _mm_mul_pd(v_sub, v_sub);
v_sum = _mm_add_pd(v_mul, v_sum);
v_sub = _mm_sub_pd(_mm_loadu_pd(vec + i + 2), v_mean);
v_mul = _mm_mul_pd(v_sub, v_sub);
v_sum = _mm_add_pd(v_mul, v_sum);
}
double res[2];
_mm_storeu_pd(res, v_sum);
sum_squared_diff = res[0] + res[1];
for (; i < size; i++) {
sum_squared_diff += (vec[i] - mean_value) * (vec[i] - mean_value);
}
double divisor = (double)size;
if (useBesselCorrection) {
divisor = divisor - 1;
}
return std::sqrt(sum_squared_diff / divisor);
}
void SimdVectorHadamard(const double * src1, const double * src2, size_t size, double * dst)
{
size_t i = 0;
for (; i <= size - 2; i += 2) {
__m128d vout = _mm_mul_pd(_mm_loadu_pd(src1 + i), _mm_loadu_pd(src2 + i));
_mm_storeu_pd(dst + i, vout);
}
for (; i < size; i++) {
dst[i] = src1[i] * src2[i];
}
}
void SimdMatMulTwist(const double * mat, size_t rows, const double * twist, double * dst)
{
// Transpose twist matrix
double transpose[36];
for (size_t i = 0; i < 6; i++) {
for (size_t j = 0; j < 6; j++) {
transpose[i*6 + j] = twist[j*6 + i];
}
}
for (size_t i = 0; i < rows; i++) {
for (size_t j = 0; j < 6; j++) {
__m128d v_mul = _mm_setzero_pd();
for (size_t k = 0; k < 6; k += 2) {
v_mul = _mm_add_pd(v_mul, _mm_mul_pd(_mm_loadu_pd(&mat[i*6 + k]), _mm_loadu_pd(&transpose[j*6 + k])));
}
double v_tmp[2];
_mm_storeu_pd(v_tmp, v_mul);
dst[i*6 + j] = v_tmp[0] + v_tmp[1];
}
}
}
void SimdNormalizedCorrelation(const double * img1, double mean1, const double * img2, double mean2, size_t size,
double& a2, double& b2, double& ab)
{
const __m128d v_mean_a = _mm_set1_pd(mean1);
const __m128d v_mean_b = _mm_set1_pd(mean2);
__m128d v_ab = _mm_setzero_pd();
__m128d v_a2 = _mm_setzero_pd();
__m128d v_b2 = _mm_setzero_pd();
size_t cpt = 0;
for (; cpt <= size - 2; cpt += 2, img1 += 2, img2 += 2) {
const __m128d v1 = _mm_loadu_pd(img1);
const __m128d v2 = _mm_loadu_pd(img2);
const __m128d norm_a = _mm_sub_pd(v1, v_mean_a);
const __m128d norm_b = _mm_sub_pd(v2, v_mean_b);
v_ab = _mm_add_pd(v_ab, _mm_mul_pd(norm_a, norm_b));
v_a2 = _mm_add_pd(v_a2, _mm_mul_pd(norm_a, norm_a));
v_b2 = _mm_add_pd(v_b2, _mm_mul_pd(norm_b, norm_b));
}
double v_res_ab[2], v_res_a2[2], v_res_b2[2];
_mm_storeu_pd(v_res_ab, v_ab);
_mm_storeu_pd(v_res_a2, v_a2);
_mm_storeu_pd(v_res_b2, v_b2);
ab = v_res_ab[0] + v_res_ab[1];
a2 = v_res_a2[0] + v_res_a2[1];
b2 = v_res_b2[0] + v_res_b2[1];
for (; cpt < size; cpt++) {
ab += (img1[cpt] - mean1) * (img2[cpt] - mean2);
a2 += (img1[cpt] - mean1) * (img1[cpt] - mean1);
b2 += (img2[cpt] - mean2) * (img2[cpt] - mean2);
}
}
void SimdNormalizedCorrelation2(const double * img1_, size_t width1, const double * img2,
size_t width2, size_t height2, size_t i0, size_t j0, double& ab)
{
const double *img1 = img1_;
__m128d v_ab = _mm_setzero_pd();
for (size_t i = 0; i < height2; i++) {
size_t j = 0;
img1 = &img1_[(i0 + i) * width1 + j0];
for (; j <= width2 - 2; j += 2, img1 += 2, img2 += 2) {
const __m128d v1 = _mm_loadu_pd(img1);
const __m128d v2 = _mm_loadu_pd(img2);
v_ab = _mm_add_pd(v_ab, _mm_mul_pd(v1, v2));
}
for (; j < width2; j++) {
ab += img1[(i0 + i)*width1 + j0 + j] * img2[i*width2 + j];
}
}
double v_res_ab[2];
_mm_storeu_pd(v_res_ab, v_ab);
ab += v_res_ab[0] + v_res_ab[1];
}
void SimdRemap(const unsigned char * src, size_t channels, size_t width, size_t height, size_t offset,
const int * mapU, const int * mapV, const float * mapDu, const float * mapDv, unsigned char * dst)
{
for (size_t j = 0; j < width; j++) {
int u_round = mapU[offset + j];
int v_round = mapV[offset + j];
const __m128 vdu = _mm_set1_ps(mapDu[offset + j]);
const __m128 vdv = _mm_set1_ps(mapDv[offset + j]);
if (0 <= u_round && 0 <= v_round && u_round < static_cast<int>(width) - 1
&& v_round < static_cast<int>(height) - 1) {
#define VLERP(va, vb, vt) _mm_add_ps(va, _mm_mul_ps(_mm_sub_ps(vb, va), vt))
// process interpolation
const __m128 vdata1 =
_mm_set_ps(static_cast<float>(src[(v_round*width + u_round)*channels + 3]), static_cast<float>(src[(v_round*width + u_round)*channels + 2]),
static_cast<float>(src[(v_round*width + u_round)*channels + 1]), static_cast<float>(src[(v_round*width + u_round)*channels] + 0));
const __m128 vdata2 =
_mm_set_ps(static_cast<float>(src[(v_round*width + u_round + 1)*channels + 3]), static_cast<float>(src[(v_round*width + u_round + 1)*channels + 2]),
static_cast<float>(src[(v_round*width + u_round + 1)*channels + 1]), static_cast<float>(src[(v_round*width + u_round + 1)*channels + 0]));
const __m128 vdata3 =
_mm_set_ps(static_cast<float>(src[((v_round + 1)*width + u_round)*channels + 3]), static_cast<float>(src[((v_round + 1)*width + u_round)*channels + 2]),
static_cast<float>(src[((v_round + 1)*width + u_round)*channels + 1]), static_cast<float>(src[((v_round + 1)*width + u_round)*channels + 0]));
const __m128 vdata4 = _mm_set_ps(
static_cast<float>(src[((v_round + 1)*width + u_round + 1)*channels + 3]), static_cast<float>(src[((v_round + 1)*width + u_round + 1)*channels + 2]),
static_cast<float>(src[((v_round + 1)*width + u_round + 1)*channels + 1]), static_cast<float>(src[((v_round + 1)*width + u_round + 1)*channels + 0]));
const __m128 vcol0 = VLERP(vdata1, vdata2, vdu);
const __m128 vcol1 = VLERP(vdata3, vdata4, vdu);
const __m128 vvalue = VLERP(vcol0, vcol1, vdv);
#undef VLERP
float values[4];
_mm_storeu_ps(values, vvalue);
dst[(offset + j)*channels + 0] = static_cast<unsigned char>(values[0]);
dst[(offset + j)*channels + 1] = static_cast<unsigned char>(values[1]);
dst[(offset + j)*channels + 2] = static_cast<unsigned char>(values[2]);
dst[(offset + j)*channels + 3] = static_cast<unsigned char>(values[3]);
} else {
for (size_t c = 0; c < channels; c++) {
dst[(offset + j)*channels + c] = 0;
}
}
}
}
void SimdComputeJtR(const double * J, size_t rows, const double * R, double * dst)
{
__m128d v_JTR_0_1 = _mm_setzero_pd();
__m128d v_JTR_2_3 = _mm_setzero_pd();
__m128d v_JTR_4_5 = _mm_setzero_pd();
for (size_t i = 0; i < rows; i++) {
const __m128d v_error = _mm_set1_pd(R[i]);
__m128d v_interaction = _mm_loadu_pd(&J[i*6]);
v_JTR_0_1 = _mm_add_pd(v_JTR_0_1, _mm_mul_pd(v_interaction, v_error));
v_interaction = _mm_loadu_pd(&J[i*6 + 2]);
v_JTR_2_3 = _mm_add_pd(v_JTR_2_3, _mm_mul_pd(v_interaction, v_error));
v_interaction = _mm_loadu_pd(&J[i*6 + 4]);
v_JTR_4_5 = _mm_add_pd(v_JTR_4_5, _mm_mul_pd(v_interaction, v_error));
}
_mm_storeu_pd(dst, v_JTR_0_1);
_mm_storeu_pd(dst + 2, v_JTR_2_3);
_mm_storeu_pd(dst + 4, v_JTR_4_5);
}
void SimdImageDifference(const unsigned char * img1, const unsigned char * img2, size_t size, unsigned char * imgDiff)
{
const __m128i mask1 = _mm_set_epi8(-1, 14, -1, 12, -1, 10, -1, 8, -1, 6, -1, 4, -1, 2, -1, 0);
const __m128i mask2 = _mm_set_epi8(-1, 15, -1, 13, -1, 11, -1, 9, -1, 7, -1, 5, -1, 3, -1, 1);
const __m128i mask_out2 = _mm_set_epi8(14, -1, 12, -1, 10, -1, 8, -1, 6, -1, 4, -1, 2, -1, 0, -1);
size_t i = 0;
for (; i <= size-16; i+= 16) {
const __m128i vdata1 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(img1 + i));
const __m128i vdata2 = _mm_loadu_si128(reinterpret_cast<const __m128i *>(img2 + i));
__m128i vdata1_reorg = _mm_shuffle_epi8(vdata1, mask1);
__m128i vdata2_reorg = _mm_shuffle_epi8(vdata2, mask1);
const __m128i vshift = _mm_set1_epi16(128);
__m128i vdata_diff = _mm_add_epi16(_mm_sub_epi16(vdata1_reorg, vdata2_reorg), vshift);
const __m128i v255 = _mm_set1_epi16(255);
const __m128i vzero = _mm_setzero_si128();
const __m128i vdata_diff_min_max1 = _mm_max_epi16(_mm_min_epi16(vdata_diff, v255), vzero);
vdata1_reorg = _mm_shuffle_epi8(vdata1, mask2);
vdata2_reorg = _mm_shuffle_epi8(vdata2, mask2);
vdata_diff = _mm_add_epi16(_mm_sub_epi16(vdata1_reorg, vdata2_reorg), vshift);
const __m128i vdata_diff_min_max2 = _mm_max_epi16(_mm_min_epi16(vdata_diff, v255), vzero);
_mm_storeu_si128(reinterpret_cast<__m128i *>(imgDiff + i), _mm_or_si128(_mm_shuffle_epi8(vdata_diff_min_max1, mask1),
_mm_shuffle_epi8(vdata_diff_min_max2, mask_out2)));
}
if (i < size) {
Base::SimdImageDifference(img1 + i, img2 + i, size - i, imgDiff + i);
}
}
}
#endif// SIMD_SSE41_ENABLE
}
|
