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
505
506
|
// Copyright (c) 2010 The Chromium Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "yuv_row.h"
#include "mozilla/SSE.h"
#define kCoefficientsRgbU kCoefficientsRgbY + 2048
#define kCoefficientsRgbV kCoefficientsRgbY + 4096
extern "C" {
#if defined(MOZILLA_MAY_SUPPORT_SSE) && defined(_M_IX86)
#if defined(__clang__)
// clang-cl has a bug where it doesn't mangle names in inline asm
// so let's do the mangling in the preprocessor (ugh)
// (but we still need to declare a dummy extern for the parser)
extern void* _kCoefficientsRgbY;
#define kCoefficientsRgbY _kCoefficientsRgbY
#endif
__declspec(naked)
void FastConvertYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
mov esi, [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
jmp convertend
convertloop :
movzx eax, byte ptr [edi]
add edi, 1
movzx ebx, byte ptr [esi]
add esi, 1
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [edx]
paddsw mm0, [kCoefficientsRgbV + 8 * ebx]
movzx ebx, byte ptr [edx + 1]
movq mm1, [kCoefficientsRgbY + 8 * eax]
add edx, 2
movq mm2, [kCoefficientsRgbY + 8 * ebx]
paddsw mm1, mm0
paddsw mm2, mm0
psraw mm1, 6
psraw mm2, 6
packuswb mm1, mm2
movntq [ebp], mm1
add ebp, 8
convertend :
sub ecx, 2
jns convertloop
and ecx, 1 // odd number of pixels?
jz convertdone
movzx eax, byte ptr [edi]
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
movd [ebp], mm1
convertdone :
popad
ret
}
}
__declspec(naked)
void ConvertYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int step) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
mov esi, [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
mov ebx, [esp + 32 + 24] // step
jmp wend
wloop :
movzx eax, byte ptr [edi]
add edi, ebx
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
add esi, ebx
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
add edx, ebx
movq mm1, [kCoefficientsRgbY + 8 * eax]
movzx eax, byte ptr [edx]
add edx, ebx
movq mm2, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
paddsw mm2, mm0
psraw mm1, 6
psraw mm2, 6
packuswb mm1, mm2
movntq [ebp], mm1
add ebp, 8
wend :
sub ecx, 2
jns wloop
and ecx, 1 // odd number of pixels?
jz wdone
movzx eax, byte ptr [edi]
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
movd [ebp], mm1
wdone :
popad
ret
}
}
__declspec(naked)
void RotateConvertYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int ystep,
int uvstep) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
mov esi, [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
jmp wend
wloop :
movzx eax, byte ptr [edi]
mov ebx, [esp + 32 + 28] // uvstep
add edi, ebx
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
add esi, ebx
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
mov ebx, [esp + 32 + 24] // ystep
add edx, ebx
movq mm1, [kCoefficientsRgbY + 8 * eax]
movzx eax, byte ptr [edx]
add edx, ebx
movq mm2, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
paddsw mm2, mm0
psraw mm1, 6
psraw mm2, 6
packuswb mm1, mm2
movntq [ebp], mm1
add ebp, 8
wend :
sub ecx, 2
jns wloop
and ecx, 1 // odd number of pixels?
jz wdone
movzx eax, byte ptr [edi]
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
movd [ebp], mm1
wdone :
popad
ret
}
}
__declspec(naked)
void DoubleYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
mov esi, [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
jmp wend
wloop :
movzx eax, byte ptr [edi]
add edi, 1
movzx ebx, byte ptr [esi]
add esi, 1
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [edx]
paddsw mm0, [kCoefficientsRgbV + 8 * ebx]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
punpckldq mm1, mm1
movntq [ebp], mm1
movzx ebx, byte ptr [edx + 1]
add edx, 2
paddsw mm0, [kCoefficientsRgbY + 8 * ebx]
psraw mm0, 6
packuswb mm0, mm0
punpckldq mm0, mm0
movntq [ebp+8], mm0
add ebp, 16
wend :
sub ecx, 4
jns wloop
add ecx, 4
jz wdone
movzx eax, byte ptr [edi]
movq mm0, [kCoefficientsRgbU + 8 * eax]
movzx eax, byte ptr [esi]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
movzx eax, byte ptr [edx]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
jmp wend1
wloop1 :
movd [ebp], mm1
add ebp, 4
wend1 :
sub ecx, 1
jns wloop1
wdone :
popad
ret
}
}
// This version does general purpose scaling by any amount, up or down.
// The only thing it cannot do is rotation by 90 or 270.
// For performance the chroma is under-sampled, reducing cost of a 3x
// 1080p scale from 8.4 ms to 5.4 ms.
__declspec(naked)
void ScaleYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int source_dx) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
mov esi, [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
xor ebx, ebx // x
jmp scaleend
scaleloop :
mov eax, ebx
sar eax, 17
movzx eax, byte ptr [edi + eax]
movq mm0, [kCoefficientsRgbU + 8 * eax]
mov eax, ebx
sar eax, 17
movzx eax, byte ptr [esi + eax]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
mov eax, ebx
add ebx, [esp + 32 + 24] // x += source_dx
sar eax, 16
movzx eax, byte ptr [edx + eax]
movq mm1, [kCoefficientsRgbY + 8 * eax]
mov eax, ebx
add ebx, [esp + 32 + 24] // x += source_dx
sar eax, 16
movzx eax, byte ptr [edx + eax]
movq mm2, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
paddsw mm2, mm0
psraw mm1, 6
psraw mm2, 6
packuswb mm1, mm2
movntq [ebp], mm1
add ebp, 8
scaleend :
sub ecx, 2
jns scaleloop
and ecx, 1 // odd number of pixels?
jz scaledone
mov eax, ebx
sar eax, 17
movzx eax, byte ptr [edi + eax]
movq mm0, [kCoefficientsRgbU + 8 * eax]
mov eax, ebx
sar eax, 17
movzx eax, byte ptr [esi + eax]
paddsw mm0, [kCoefficientsRgbV + 8 * eax]
mov eax, ebx
sar eax, 16
movzx eax, byte ptr [edx + eax]
movq mm1, [kCoefficientsRgbY + 8 * eax]
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
movd [ebp], mm1
scaledone :
popad
ret
}
}
__declspec(naked)
void LinearScaleYUVToRGB32Row_SSE(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int source_dx) {
__asm {
pushad
mov edx, [esp + 32 + 4] // Y
mov edi, [esp + 32 + 8] // U
// [esp + 32 + 12] // V
mov ebp, [esp + 32 + 16] // rgb
mov ecx, [esp + 32 + 20] // width
imul ecx, [esp + 32 + 24] // source_dx
mov [esp + 32 + 20], ecx // source_width = width * source_dx
mov ecx, [esp + 32 + 24] // source_dx
xor ebx, ebx // x = 0
cmp ecx, 0x20000
jl lscaleend
mov ebx, 0x8000 // x = 0.5 for 1/2 or less
jmp lscaleend
lscaleloop:
mov eax, ebx
sar eax, 0x11
movzx ecx, byte ptr [edi + eax]
movzx esi, byte ptr [edi + eax + 1]
mov eax, ebx
and eax, 0x1fffe
imul esi, eax
xor eax, 0x1fffe
imul ecx, eax
add ecx, esi
shr ecx, 17
movq mm0, [kCoefficientsRgbU + 8 * ecx]
mov esi, [esp + 32 + 12]
mov eax, ebx
sar eax, 0x11
movzx ecx, byte ptr [esi + eax]
movzx esi, byte ptr [esi + eax + 1]
mov eax, ebx
and eax, 0x1fffe
imul esi, eax
xor eax, 0x1fffe
imul ecx, eax
add ecx, esi
shr ecx, 17
paddsw mm0, [kCoefficientsRgbV + 8 * ecx]
mov eax, ebx
sar eax, 0x10
movzx ecx, byte ptr [edx + eax]
movzx esi, byte ptr [1 + edx + eax]
mov eax, ebx
add ebx, [esp + 32 + 24]
and eax, 0xffff
imul esi, eax
xor eax, 0xffff
imul ecx, eax
add ecx, esi
shr ecx, 16
movq mm1, [kCoefficientsRgbY + 8 * ecx]
cmp ebx, [esp + 32 + 20]
jge lscalelastpixel
mov eax, ebx
sar eax, 0x10
movzx ecx, byte ptr [edx + eax]
movzx esi, byte ptr [edx + eax + 1]
mov eax, ebx
add ebx, [esp + 32 + 24]
and eax, 0xffff
imul esi, eax
xor eax, 0xffff
imul ecx, eax
add ecx, esi
shr ecx, 16
movq mm2, [kCoefficientsRgbY + 8 * ecx]
paddsw mm1, mm0
paddsw mm2, mm0
psraw mm1, 0x6
psraw mm2, 0x6
packuswb mm1, mm2
movntq [ebp], mm1
add ebp, 0x8
lscaleend:
cmp ebx, [esp + 32 + 20]
jl lscaleloop
popad
ret
lscalelastpixel:
paddsw mm1, mm0
psraw mm1, 6
packuswb mm1, mm1
movd [ebp], mm1
popad
ret
};
}
#endif // if defined(MOZILLA_MAY_SUPPORT_SSE) && defined(_M_IX86)
void FastConvertYUVToRGB32Row(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width) {
#if defined(MOZILLA_MAY_SUPPORT_SSE) && defined(_M_IX86)
if (mozilla::supports_sse()) {
FastConvertYUVToRGB32Row_SSE(y_buf, u_buf, v_buf, rgb_buf, width);
return;
}
#endif
FastConvertYUVToRGB32Row_C(y_buf, u_buf, v_buf, rgb_buf, width, 1);
}
void ScaleYUVToRGB32Row(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int source_dx) {
#if defined(MOZILLA_MAY_SUPPORT_SSE) && defined(_M_IX86)
if (mozilla::supports_sse()) {
ScaleYUVToRGB32Row_SSE(y_buf, u_buf, v_buf, rgb_buf, width, source_dx);
return;
}
#endif
ScaleYUVToRGB32Row_C(y_buf, u_buf, v_buf, rgb_buf, width, source_dx);
}
void LinearScaleYUVToRGB32Row(const uint8* y_buf,
const uint8* u_buf,
const uint8* v_buf,
uint8* rgb_buf,
int width,
int source_dx) {
#if defined(MOZILLA_MAY_SUPPORT_SSE) && defined(_M_IX86)
if (mozilla::supports_sse()) {
LinearScaleYUVToRGB32Row_SSE(y_buf, u_buf, v_buf, rgb_buf, width,
source_dx);
return;
}
#endif
LinearScaleYUVToRGB32Row_C(y_buf, u_buf, v_buf, rgb_buf, width, source_dx);
}
} // extern "C"
|
