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
|
/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include <emmintrin.h> // SSE2
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"
#include "vpx_ports/mem.h"
typedef void (*getNxMvar_fn_t) (const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse, int *sum);
unsigned int vpx_get_mb_ss_sse2(const int16_t *src) {
__m128i vsum = _mm_setzero_si128();
int i;
for (i = 0; i < 32; ++i) {
const __m128i v = _mm_loadu_si128((const __m128i *)src);
vsum = _mm_add_epi32(vsum, _mm_madd_epi16(v, v));
src += 8;
}
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 4));
return _mm_cvtsi128_si32(vsum);
}
#define READ64(p, stride, i) \
_mm_unpacklo_epi8(_mm_cvtsi32_si128(*(const uint32_t *)(p + i * stride)), \
_mm_cvtsi32_si128(*(const uint32_t *)(p + (i + 1) * stride)))
static void get4x4var_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse, int *sum) {
const __m128i zero = _mm_setzero_si128();
const __m128i src0 = _mm_unpacklo_epi8(READ64(src, src_stride, 0), zero);
const __m128i src1 = _mm_unpacklo_epi8(READ64(src, src_stride, 2), zero);
const __m128i ref0 = _mm_unpacklo_epi8(READ64(ref, ref_stride, 0), zero);
const __m128i ref1 = _mm_unpacklo_epi8(READ64(ref, ref_stride, 2), zero);
const __m128i diff0 = _mm_sub_epi16(src0, ref0);
const __m128i diff1 = _mm_sub_epi16(src1, ref1);
// sum
__m128i vsum = _mm_add_epi16(diff0, diff1);
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 2));
*sum = (int16_t)_mm_extract_epi16(vsum, 0);
// sse
vsum = _mm_add_epi32(_mm_madd_epi16(diff0, diff0),
_mm_madd_epi16(diff1, diff1));
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi32(vsum, _mm_srli_si128(vsum, 4));
*sse = _mm_cvtsi128_si32(vsum);
}
void vpx_get8x8var_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse, int *sum) {
const __m128i zero = _mm_setzero_si128();
__m128i vsum = _mm_setzero_si128();
__m128i vsse = _mm_setzero_si128();
int i;
for (i = 0; i < 8; i += 2) {
const __m128i src0 = _mm_unpacklo_epi8(_mm_loadl_epi64(
(const __m128i *)(src + i * src_stride)), zero);
const __m128i ref0 = _mm_unpacklo_epi8(_mm_loadl_epi64(
(const __m128i *)(ref + i * ref_stride)), zero);
const __m128i diff0 = _mm_sub_epi16(src0, ref0);
const __m128i src1 = _mm_unpacklo_epi8(_mm_loadl_epi64(
(const __m128i *)(src + (i + 1) * src_stride)), zero);
const __m128i ref1 = _mm_unpacklo_epi8(_mm_loadl_epi64(
(const __m128i *)(ref + (i + 1) * ref_stride)), zero);
const __m128i diff1 = _mm_sub_epi16(src1, ref1);
vsum = _mm_add_epi16(vsum, diff0);
vsum = _mm_add_epi16(vsum, diff1);
vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff0, diff0));
vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff1, diff1));
}
// sum
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 2));
*sum = (int16_t)_mm_extract_epi16(vsum, 0);
// sse
vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 8));
vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 4));
*sse = _mm_cvtsi128_si32(vsse);
}
void vpx_get16x16var_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse, int *sum) {
const __m128i zero = _mm_setzero_si128();
__m128i vsum = _mm_setzero_si128();
__m128i vsse = _mm_setzero_si128();
int i;
for (i = 0; i < 16; ++i) {
const __m128i s = _mm_loadu_si128((const __m128i *)src);
const __m128i r = _mm_loadu_si128((const __m128i *)ref);
const __m128i src0 = _mm_unpacklo_epi8(s, zero);
const __m128i ref0 = _mm_unpacklo_epi8(r, zero);
const __m128i diff0 = _mm_sub_epi16(src0, ref0);
const __m128i src1 = _mm_unpackhi_epi8(s, zero);
const __m128i ref1 = _mm_unpackhi_epi8(r, zero);
const __m128i diff1 = _mm_sub_epi16(src1, ref1);
vsum = _mm_add_epi16(vsum, diff0);
vsum = _mm_add_epi16(vsum, diff1);
vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff0, diff0));
vsse = _mm_add_epi32(vsse, _mm_madd_epi16(diff1, diff1));
src += src_stride;
ref += ref_stride;
}
// sum
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 8));
vsum = _mm_add_epi16(vsum, _mm_srli_si128(vsum, 4));
*sum = (int16_t)_mm_extract_epi16(vsum, 0) +
(int16_t)_mm_extract_epi16(vsum, 1);
// sse
vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 8));
vsse = _mm_add_epi32(vsse, _mm_srli_si128(vsse, 4));
*sse = _mm_cvtsi128_si32(vsse);
}
static void variance_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
int w, int h, unsigned int *sse, int *sum,
getNxMvar_fn_t var_fn, int block_size) {
int i, j;
*sse = 0;
*sum = 0;
for (i = 0; i < h; i += block_size) {
for (j = 0; j < w; j += block_size) {
unsigned int sse0;
int sum0;
var_fn(src + src_stride * i + j, src_stride,
ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
*sse += sse0;
*sum += sum0;
}
}
}
unsigned int vpx_variance4x4_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse) {
int sum;
get4x4var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
return *sse - ((sum * sum) >> 4);
}
unsigned int vpx_variance8x4_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 8, 4,
sse, &sum, get4x4var_sse2, 4);
return *sse - ((sum * sum) >> 5);
}
unsigned int vpx_variance4x8_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 4, 8,
sse, &sum, get4x4var_sse2, 4);
return *sse - ((sum * sum) >> 5);
}
unsigned int vpx_variance8x8_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse) {
int sum;
vpx_get8x8var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
return *sse - ((sum * sum) >> 6);
}
unsigned int vpx_variance16x8_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 16, 8,
sse, &sum, vpx_get8x8var_sse2, 8);
return *sse - ((sum * sum) >> 7);
}
unsigned int vpx_variance8x16_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 8, 16,
sse, &sum, vpx_get8x8var_sse2, 8);
return *sse - ((sum * sum) >> 7);
}
unsigned int vpx_variance16x16_sse2(const unsigned char *src, int src_stride,
const unsigned char *ref, int ref_stride,
unsigned int *sse) {
int sum;
vpx_get16x16var_sse2(src, src_stride, ref, ref_stride, sse, &sum);
return *sse - (((uint32_t)((int64_t)sum * sum)) >> 8);
}
unsigned int vpx_variance32x32_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 32, 32,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 10);
}
unsigned int vpx_variance32x16_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 32, 16,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 9);
}
unsigned int vpx_variance16x32_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 16, 32,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 9);
}
unsigned int vpx_variance64x64_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 64, 64,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 12);
}
unsigned int vpx_variance64x32_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 64, 32,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 11);
}
unsigned int vpx_variance32x64_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
int sum;
variance_sse2(src, src_stride, ref, ref_stride, 32, 64,
sse, &sum, vpx_get16x16var_sse2, 16);
return *sse - (((int64_t)sum * sum) >> 11);
}
unsigned int vpx_mse8x8_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
vpx_variance8x8_sse2(src, src_stride, ref, ref_stride, sse);
return *sse;
}
unsigned int vpx_mse8x16_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
vpx_variance8x16_sse2(src, src_stride, ref, ref_stride, sse);
return *sse;
}
unsigned int vpx_mse16x8_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
vpx_variance16x8_sse2(src, src_stride, ref, ref_stride, sse);
return *sse;
}
unsigned int vpx_mse16x16_sse2(const uint8_t *src, int src_stride,
const uint8_t *ref, int ref_stride,
unsigned int *sse) {
vpx_variance16x16_sse2(src, src_stride, ref, ref_stride, sse);
return *sse;
}
#if CONFIG_USE_X86INC
// The 2 unused parameters are place holders for PIC enabled build.
// These definitions are for functions defined in subpel_variance.asm
#define DECL(w, opt) \
int vpx_sub_pixel_variance##w##xh_##opt(const uint8_t *src, \
ptrdiff_t src_stride, \
int x_offset, int y_offset, \
const uint8_t *dst, \
ptrdiff_t dst_stride, \
int height, unsigned int *sse, \
void *unused0, void *unused)
#define DECLS(opt1, opt2) \
DECL(4, opt1); \
DECL(8, opt1); \
DECL(16, opt1)
DECLS(sse2, sse2);
DECLS(ssse3, ssse3);
#undef DECLS
#undef DECL
#define FN(w, h, wf, wlog2, hlog2, opt, cast_prod, cast) \
unsigned int vpx_sub_pixel_variance##w##x##h##_##opt(const uint8_t *src, \
int src_stride, \
int x_offset, \
int y_offset, \
const uint8_t *dst, \
int dst_stride, \
unsigned int *sse_ptr) { \
unsigned int sse; \
int se = vpx_sub_pixel_variance##wf##xh_##opt(src, src_stride, x_offset, \
y_offset, dst, dst_stride, \
h, &sse, NULL, NULL); \
if (w > wf) { \
unsigned int sse2; \
int se2 = vpx_sub_pixel_variance##wf##xh_##opt(src + 16, src_stride, \
x_offset, y_offset, \
dst + 16, dst_stride, \
h, &sse2, NULL, NULL); \
se += se2; \
sse += sse2; \
if (w > wf * 2) { \
se2 = vpx_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
x_offset, y_offset, \
dst + 32, dst_stride, \
h, &sse2, NULL, NULL); \
se += se2; \
sse += sse2; \
se2 = vpx_sub_pixel_variance##wf##xh_##opt(src + 48, src_stride, \
x_offset, y_offset, \
dst + 48, dst_stride, \
h, &sse2, NULL, NULL); \
se += se2; \
sse += sse2; \
} \
} \
*sse_ptr = sse; \
return sse - (cast_prod (cast se * se) >> (wlog2 + hlog2)); \
}
#define FNS(opt1, opt2) \
FN(64, 64, 16, 6, 6, opt1, (int64_t), (int64_t)); \
FN(64, 32, 16, 6, 5, opt1, (int64_t), (int64_t)); \
FN(32, 64, 16, 5, 6, opt1, (int64_t), (int64_t)); \
FN(32, 32, 16, 5, 5, opt1, (int64_t), (int64_t)); \
FN(32, 16, 16, 5, 4, opt1, (int64_t), (int64_t)); \
FN(16, 32, 16, 4, 5, opt1, (int64_t), (int64_t)); \
FN(16, 16, 16, 4, 4, opt1, (uint32_t), (int64_t)); \
FN(16, 8, 16, 4, 3, opt1, (int32_t), (int32_t)); \
FN(8, 16, 8, 3, 4, opt1, (int32_t), (int32_t)); \
FN(8, 8, 8, 3, 3, opt1, (int32_t), (int32_t)); \
FN(8, 4, 8, 3, 2, opt1, (int32_t), (int32_t)); \
FN(4, 8, 4, 2, 3, opt1, (int32_t), (int32_t)); \
FN(4, 4, 4, 2, 2, opt1, (int32_t), (int32_t))
FNS(sse2, sse2);
FNS(ssse3, ssse3);
#undef FNS
#undef FN
// The 2 unused parameters are place holders for PIC enabled build.
#define DECL(w, opt) \
int vpx_sub_pixel_avg_variance##w##xh_##opt(const uint8_t *src, \
ptrdiff_t src_stride, \
int x_offset, int y_offset, \
const uint8_t *dst, \
ptrdiff_t dst_stride, \
const uint8_t *sec, \
ptrdiff_t sec_stride, \
int height, unsigned int *sse, \
void *unused0, void *unused)
#define DECLS(opt1, opt2) \
DECL(4, opt1); \
DECL(8, opt1); \
DECL(16, opt1)
DECLS(sse2, sse2);
DECLS(ssse3, ssse3);
#undef DECL
#undef DECLS
#define FN(w, h, wf, wlog2, hlog2, opt, cast_prod, cast) \
unsigned int vpx_sub_pixel_avg_variance##w##x##h##_##opt(const uint8_t *src, \
int src_stride, \
int x_offset, \
int y_offset, \
const uint8_t *dst, \
int dst_stride, \
unsigned int *sseptr, \
const uint8_t *sec) { \
unsigned int sse; \
int se = vpx_sub_pixel_avg_variance##wf##xh_##opt(src, src_stride, x_offset, \
y_offset, dst, dst_stride, \
sec, w, h, &sse, NULL, \
NULL); \
if (w > wf) { \
unsigned int sse2; \
int se2 = vpx_sub_pixel_avg_variance##wf##xh_##opt(src + 16, src_stride, \
x_offset, y_offset, \
dst + 16, dst_stride, \
sec + 16, w, h, &sse2, \
NULL, NULL); \
se += se2; \
sse += sse2; \
if (w > wf * 2) { \
se2 = vpx_sub_pixel_avg_variance##wf##xh_##opt(src + 32, src_stride, \
x_offset, y_offset, \
dst + 32, dst_stride, \
sec + 32, w, h, &sse2, \
NULL, NULL); \
se += se2; \
sse += sse2; \
se2 = vpx_sub_pixel_avg_variance##wf##xh_##opt(src + 48, src_stride, \
x_offset, y_offset, \
dst + 48, dst_stride, \
sec + 48, w, h, &sse2, \
NULL, NULL); \
se += se2; \
sse += sse2; \
} \
} \
*sseptr = sse; \
return sse - (cast_prod (cast se * se) >> (wlog2 + hlog2)); \
}
#define FNS(opt1, opt2) \
FN(64, 64, 16, 6, 6, opt1, (int64_t), (int64_t)); \
FN(64, 32, 16, 6, 5, opt1, (int64_t), (int64_t)); \
FN(32, 64, 16, 5, 6, opt1, (int64_t), (int64_t)); \
FN(32, 32, 16, 5, 5, opt1, (int64_t), (int64_t)); \
FN(32, 16, 16, 5, 4, opt1, (int64_t), (int64_t)); \
FN(16, 32, 16, 4, 5, opt1, (int64_t), (int64_t)); \
FN(16, 16, 16, 4, 4, opt1, (uint32_t), (int64_t)); \
FN(16, 8, 16, 4, 3, opt1, (uint32_t), (int32_t)); \
FN(8, 16, 8, 3, 4, opt1, (uint32_t), (int32_t)); \
FN(8, 8, 8, 3, 3, opt1, (uint32_t), (int32_t)); \
FN(8, 4, 8, 3, 2, opt1, (uint32_t), (int32_t)); \
FN(4, 8, 4, 2, 3, opt1, (uint32_t), (int32_t)); \
FN(4, 4, 4, 2, 2, opt1, (uint32_t), (int32_t))
FNS(sse2, sse);
FNS(ssse3, ssse3);
#undef FNS
#undef FN
#endif // CONFIG_USE_X86INC
|
