32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
36 |
|
* MinChen <chenm001@163.com> |
37 |
|
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
38 |
|
* Fix a little bug for low_delay flage |
39 |
|
* MinChen <chenm001@163.com> |
40 |
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
41 |
* 22.06.2002 added primative N_VOP support |
* 22.06.2002 added primative N_VOP support |
42 |
* #define BFRAMES_DEC now enables Minchenm's bframe decoder |
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
43 |
* 08.05.2002 add low_delay support for B_VOP decode |
* 08.05.2002 add low_delay support for B_VOP decode |
44 |
* MinChen <chenm001@163.com> |
* MinChen <chenm001@163.com> |
45 |
* 05.05.2002 fix some B-frame decode problem |
* 05.05.2002 fix some B-frame decode problem |
55 |
* 22.12.2001 lock based interpolation |
* 22.12.2001 lock based interpolation |
56 |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
57 |
* |
* |
58 |
* $Id: decoder.c,v 1.26 2002-07-09 01:37:22 chenm001 Exp $ |
* $Id: decoder.c,v 1.37.2.16 2002-12-08 06:43:33 suxen_drol Exp $ |
59 |
* |
* |
60 |
*************************************************************************/ |
*************************************************************************/ |
61 |
|
|
62 |
#include <stdlib.h> |
#include <stdlib.h> |
63 |
#include <string.h> |
#include <string.h> |
64 |
|
|
65 |
|
#ifdef BFRAMES_DEC_DEBUG |
66 |
|
#define BFRAMES_DEC |
67 |
|
#endif |
68 |
|
|
69 |
#include "xvid.h" |
#include "xvid.h" |
70 |
#include "portab.h" |
#include "portab.h" |
71 |
|
|
79 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
80 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
81 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
82 |
|
#include "image/reduced.h" |
83 |
|
|
84 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
85 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
86 |
#include "utils/timer.h" |
#include "utils/timer.h" |
87 |
#include "utils/emms.h" |
#include "utils/emms.h" |
88 |
|
#include "motion/motion.h" |
89 |
|
|
90 |
#include "image/image.h" |
#include "image/image.h" |
91 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
92 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
93 |
|
|
94 |
int |
int |
95 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_resize(DECODER * dec) |
96 |
{ |
{ |
97 |
DECODER *dec; |
/* free existing */ |
98 |
|
|
99 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
100 |
if (dec == NULL) { |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
101 |
return XVID_ERR_MEMORY; |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
102 |
} |
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
103 |
param->handle = dec; |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
104 |
|
|
105 |
dec->width = param->width; |
if (dec->last_mbs) |
106 |
dec->height = param->height; |
xvid_free(dec->last_mbs); |
107 |
|
if (dec->mbs) |
108 |
|
xvid_free(dec->mbs); |
109 |
|
|
110 |
|
/* realloc */ |
111 |
|
|
112 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
113 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
114 |
|
|
115 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
116 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
|
dec->low_delay = 0; |
|
117 |
|
|
118 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
119 |
xvid_free(dec); |
xvid_free(dec); |
125 |
xvid_free(dec); |
xvid_free(dec); |
126 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
127 |
} |
} |
128 |
|
|
129 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
130 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
131 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
142 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
143 |
} |
} |
144 |
|
|
145 |
|
if (image_create(&dec->refh, dec->edged_width, dec->edged_height)) { |
146 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
147 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
148 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
149 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
150 |
|
xvid_free(dec); |
151 |
|
return XVID_ERR_MEMORY; |
152 |
|
} |
153 |
|
|
154 |
dec->mbs = |
dec->mbs = |
155 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
156 |
CACHE_LINE); |
CACHE_LINE); |
159 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
160 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
161 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
162 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
163 |
xvid_free(dec); |
xvid_free(dec); |
164 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
165 |
} |
} |
|
|
|
166 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
167 |
|
|
168 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
176 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
177 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
178 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
179 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
180 |
xvid_free(dec); |
xvid_free(dec); |
181 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
182 |
} |
} |
183 |
|
|
184 |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
185 |
|
|
186 |
|
return XVID_ERR_OK; |
187 |
|
} |
188 |
|
|
189 |
|
|
190 |
|
int |
191 |
|
decoder_create(XVID_DEC_PARAM * param) |
192 |
|
{ |
193 |
|
DECODER *dec; |
194 |
|
|
195 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
196 |
|
if (dec == NULL) { |
197 |
|
return XVID_ERR_MEMORY; |
198 |
|
} |
199 |
|
memset(dec, 0, sizeof(DECODER)); |
200 |
|
|
201 |
|
param->handle = dec; |
202 |
|
|
203 |
|
dec->width = param->width; |
204 |
|
dec->height = param->height; |
205 |
|
|
206 |
|
image_null(&dec->cur); |
207 |
|
image_null(&dec->refn[0]); |
208 |
|
image_null(&dec->refn[1]); |
209 |
|
image_null(&dec->refn[2]); |
210 |
|
image_null(&dec->refh); |
211 |
|
|
212 |
|
dec->mbs = NULL; |
213 |
|
dec->last_mbs = NULL; |
214 |
|
|
215 |
init_timer(); |
init_timer(); |
216 |
|
|
217 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
218 |
// for support B-frame to save reference frame's time |
// for support B-frame to save reference frame's time |
219 |
dec->frames = -1; |
dec->frames = -1; |
220 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
221 |
|
dec->low_delay = 0; |
222 |
|
dec->packed_mode = 0; |
223 |
|
|
224 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
225 |
|
|
226 |
|
if (dec->fixed_dimensions) |
227 |
|
return decoder_resize(dec); |
228 |
|
else |
229 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
230 |
} |
} |
231 |
|
|
238 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
239 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
240 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
241 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
242 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
243 |
xvid_free(dec); |
xvid_free(dec); |
244 |
|
|
267 |
Bitstream * bs, |
Bitstream * bs, |
268 |
const uint32_t quant, |
const uint32_t quant, |
269 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
270 |
const unsigned int bound) |
const unsigned int bound, |
271 |
|
const int reduced_resolution) |
272 |
{ |
{ |
273 |
|
|
274 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
281 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
282 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
283 |
|
|
284 |
|
if (reduced_resolution) { |
285 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
286 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
287 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
288 |
|
}else{ |
289 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
290 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
291 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
292 |
|
} |
293 |
|
|
294 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
295 |
|
|
328 |
start_timer(); |
start_timer(); |
329 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
330 |
{ |
{ |
331 |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
int direction = dec->alternate_vertical_scan ? |
332 |
start_coeff); |
2 : pMB->acpred_directions[i]; |
333 |
|
|
334 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
335 |
} |
} |
336 |
stop_coding_timer(); |
stop_coding_timer(); |
337 |
|
|
350 |
start_timer(); |
start_timer(); |
351 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
352 |
stop_idct_timer(); |
stop_idct_timer(); |
353 |
|
|
354 |
} |
} |
355 |
|
|
356 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
359 |
} |
} |
360 |
|
|
361 |
start_timer(); |
start_timer(); |
362 |
|
|
363 |
|
if (reduced_resolution) |
364 |
|
{ |
365 |
|
next_block*=2; |
366 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
367 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
368 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
369 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
370 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
371 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
372 |
|
}else{ |
373 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
374 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
375 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
376 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
377 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
378 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
379 |
|
} |
380 |
stop_transfer_timer(); |
stop_transfer_timer(); |
381 |
} |
} |
382 |
|
|
386 |
|
|
387 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
388 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
|
static const uint32_t roundtab[16] = |
|
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
389 |
|
|
390 |
// decode an inter macroblock |
// decode an inter macroblock |
391 |
|
|
392 |
|
static void |
393 |
|
rrv_mv_scaleup(VECTOR * mv) |
394 |
|
{ |
395 |
|
if (mv->x > 0) { |
396 |
|
mv->x = 2*mv->x - 1; |
397 |
|
} else if (mv->x < 0) { |
398 |
|
mv->x = 2*mv->x + 1; |
399 |
|
} |
400 |
|
|
401 |
|
if (mv->y > 0) { |
402 |
|
mv->y = 2*mv->y - 1; |
403 |
|
} else if (mv->y < 0) { |
404 |
|
mv->y = 2*mv->y + 1; |
405 |
|
} |
406 |
|
} |
407 |
|
|
408 |
|
|
409 |
|
|
410 |
void |
void |
411 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
412 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
416 |
const uint32_t cbp, |
const uint32_t cbp, |
417 |
Bitstream * bs, |
Bitstream * bs, |
418 |
const uint32_t quant, |
const uint32_t quant, |
419 |
const uint32_t rounding) |
const uint32_t rounding, |
420 |
|
const int reduced_resolution) |
421 |
{ |
{ |
422 |
|
|
423 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
425 |
|
|
426 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
427 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
428 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
429 |
uint32_t i; |
uint32_t i; |
430 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
431 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
432 |
|
|
433 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
434 |
|
VECTOR mv[4]; |
435 |
|
|
436 |
|
for (i = 0; i < 4; i++) |
437 |
|
{ |
438 |
|
mv[i] = pMB->mvs[i]; |
439 |
|
//DPRINTF(DPRINTF_MB, "mv[%i] orig=%i,%i local=%i", i, pMB->mvs[i].x, pMB->mvs[i].y, mv[i].x, mv[i].y); |
440 |
|
} |
441 |
|
|
442 |
|
if (reduced_resolution) { |
443 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
444 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
445 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
446 |
|
DPRINTF(DPRINTF_MB,"[%i,%i] %i,%i %i,%i %i,%i %i,%i", |
447 |
|
x_pos, y_pos, |
448 |
|
mv[0].x, mv[0].y, |
449 |
|
mv[1].x, mv[1].y, |
450 |
|
mv[2].x, mv[2].y, |
451 |
|
mv[3].x, mv[3].y); |
452 |
|
|
453 |
|
rrv_mv_scaleup(&mv[0]); |
454 |
|
rrv_mv_scaleup(&mv[1]); |
455 |
|
rrv_mv_scaleup(&mv[2]); |
456 |
|
rrv_mv_scaleup(&mv[3]); |
457 |
|
|
458 |
|
DPRINTF(DPRINTF_MB," %i,%i %i,%i %i,%i %i,%i", |
459 |
|
mv[0].x, mv[0].y, |
460 |
|
mv[1].x, mv[1].y, |
461 |
|
mv[2].x, mv[2].y, |
462 |
|
mv[3].x, mv[3].y); |
463 |
|
}else{ |
464 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
465 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
466 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
467 |
|
} |
468 |
|
|
469 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
470 |
uv_dx = pMB->mvs[0].x; |
uv_dx = mv[0].x; |
471 |
uv_dy = pMB->mvs[0].y; |
uv_dy = mv[0].y; |
472 |
|
|
473 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) |
474 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
{ |
475 |
} else { |
uv_dx /= 2; |
476 |
int sum; |
uv_dy /= 2; |
477 |
|
} |
478 |
|
|
479 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
480 |
uv_dx = |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
481 |
|
|
482 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
start_timer(); |
483 |
uv_dy = |
if (reduced_resolution) |
484 |
(sum == |
{ |
485 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
486 |
(ABS(sum) / 16) * 2)); |
mv[0].x, mv[0].y, stride, rounding); |
487 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
488 |
|
uv_dx, uv_dy, stride2, rounding); |
489 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
490 |
|
uv_dx, uv_dy, stride2, rounding); |
491 |
|
|
492 |
|
} |
493 |
|
else |
494 |
|
{ |
495 |
|
if(dec->quarterpel) { |
496 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
497 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
498 |
|
mv[0].x, mv[0].y, stride, rounding); |
499 |
|
} |
500 |
|
else { |
501 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
502 |
|
mv[0].x, mv[0].y, stride, rounding); |
503 |
} |
} |
504 |
|
|
505 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
506 |
|
uv_dx, uv_dy, stride2, rounding); |
507 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
508 |
|
uv_dx, uv_dy, stride2, rounding); |
509 |
|
} |
510 |
|
stop_comp_timer(); |
511 |
|
|
512 |
|
} else { /* MODE_INTER4V */ |
513 |
|
int sum; |
514 |
|
|
515 |
|
if(dec->quarterpel) |
516 |
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
517 |
|
else |
518 |
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
519 |
|
|
520 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
521 |
|
|
522 |
|
if(dec->quarterpel) |
523 |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
524 |
|
else |
525 |
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
526 |
|
|
527 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
528 |
|
|
529 |
start_timer(); |
start_timer(); |
530 |
|
if (reduced_resolution) |
531 |
|
{ |
532 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
533 |
|
mv[0].x, mv[0].y, stride, rounding); |
534 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos, |
535 |
|
mv[1].x, mv[1].y, stride, rounding); |
536 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos + 16, |
537 |
|
mv[2].x, mv[2].y, stride, rounding); |
538 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos + 16, |
539 |
|
mv[3].x, mv[3].y, stride, rounding); |
540 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
541 |
|
uv_dx, uv_dy, stride2, rounding); |
542 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
543 |
|
uv_dx, uv_dy, stride2, rounding); |
544 |
|
|
545 |
|
// set_block(pY_Cur, stride, 32, 32, 127); |
546 |
|
} |
547 |
|
else |
548 |
|
{ |
549 |
|
if(dec->quarterpel) { |
550 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
551 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
552 |
|
mv[0].x, mv[0].y, stride, rounding); |
553 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
554 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
555 |
|
mv[1].x, mv[1].y, stride, rounding); |
556 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
557 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
558 |
|
mv[2].x, mv[2].y, stride, rounding); |
559 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
560 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
561 |
|
mv[3].x, mv[3].y, stride, rounding); |
562 |
|
} |
563 |
|
else { |
564 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, 16 * y_pos, |
565 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
566 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
567 |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, |
mv[1].x, mv[1].y, stride, rounding); |
568 |
rounding); |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
569 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos, |
mv[2].x, mv[2].y, stride, rounding); |
570 |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
571 |
rounding); |
mv[3].x, mv[3].y, stride, rounding); |
572 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16 * x_pos + 8, |
} |
573 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
|
|
rounding); |
|
574 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
575 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
576 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
577 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
578 |
|
} |
579 |
stop_comp_timer(); |
stop_comp_timer(); |
580 |
|
} |
581 |
|
|
582 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
583 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
584 |
|
|
585 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
586 |
{ |
{ |
587 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
588 |
|
|
589 |
start_timer(); |
start_timer(); |
590 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
591 |
stop_coding_timer(); |
stop_coding_timer(); |
592 |
|
|
593 |
start_timer(); |
start_timer(); |
610 |
} |
} |
611 |
|
|
612 |
start_timer(); |
start_timer(); |
613 |
|
if (reduced_resolution) |
614 |
|
{ |
615 |
|
if (cbp & 32) |
616 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
617 |
|
if (cbp & 16) |
618 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
619 |
|
if (cbp & 8) |
620 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
621 |
|
if (cbp & 4) |
622 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
623 |
|
if (cbp & 2) |
624 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
625 |
|
if (cbp & 1) |
626 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
627 |
|
} |
628 |
|
else |
629 |
|
{ |
630 |
if (cbp & 32) |
if (cbp & 32) |
631 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
632 |
if (cbp & 16) |
if (cbp & 16) |
639 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
640 |
if (cbp & 1) |
if (cbp & 1) |
641 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
642 |
|
} |
643 |
stop_transfer_timer(); |
stop_transfer_timer(); |
644 |
} |
} |
645 |
|
|
647 |
void |
void |
648 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
649 |
Bitstream * bs, |
Bitstream * bs, |
650 |
|
int reduced_resolution, |
651 |
int quant, |
int quant, |
652 |
int intra_dc_threshold) |
int intra_dc_threshold) |
653 |
{ |
{ |
654 |
uint32_t bound; |
uint32_t bound; |
655 |
uint32_t x, y; |
uint32_t x, y; |
656 |
|
int mb_width = dec->mb_width; |
657 |
|
int mb_height = dec->mb_height; |
658 |
|
|
659 |
|
if (reduced_resolution) |
660 |
|
{ |
661 |
|
mb_width /= 2; |
662 |
|
mb_height /= 2; |
663 |
|
} |
664 |
|
|
665 |
bound = 0; |
bound = 0; |
666 |
|
|
667 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
668 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
669 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
670 |
uint32_t mcbpc; |
uint32_t mcbpc; |
671 |
uint32_t cbpc; |
uint32_t cbpc; |
678 |
|
|
679 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
680 |
{ |
{ |
681 |
bound = read_video_packet_header(bs, 0, &quant); |
bound = read_video_packet_header(bs, dec, 0, |
682 |
x = bound % dec->mb_width; |
&quant, NULL, NULL, &intra_dc_threshold); |
683 |
y = bound / dec->mb_width; |
x = bound % mb_width; |
684 |
|
y = bound / mb_width; |
685 |
} |
} |
686 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
687 |
|
|
705 |
} |
} |
706 |
} |
} |
707 |
mb->quant = quant; |
mb->quant = quant; |
708 |
|
mb->mvs[0].x = mb->mvs[0].y = |
709 |
|
mb->mvs[1].x = mb->mvs[1].y = |
710 |
|
mb->mvs[2].x = mb->mvs[2].y = |
711 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
712 |
|
|
713 |
if (dec->interlacing) { |
if (dec->interlacing) { |
714 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
716 |
} |
} |
717 |
|
|
718 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
719 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
720 |
|
|
721 |
} |
} |
722 |
|
if(dec->out_frm) |
723 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
724 |
} |
} |
725 |
|
|
726 |
} |
} |
732 |
int x, |
int x, |
733 |
int y, |
int y, |
734 |
int k, |
int k, |
735 |
VECTOR * mv, |
VECTOR * ret_mv, |
736 |
int fcode, |
int fcode, |
737 |
const int bound) |
const int bound) |
738 |
{ |
{ |
743 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
744 |
|
|
745 |
VECTOR pmv; |
VECTOR pmv; |
746 |
int mv_x, mv_y; |
VECTOR mv; |
747 |
|
|
748 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
749 |
|
|
750 |
mv_x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
751 |
mv_y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
752 |
|
|
753 |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv.x, mv.y, pmv.x, pmv.y); |
754 |
|
|
755 |
mv_x += pmv.x; |
mv.x += pmv.x; |
756 |
mv_y += pmv.y; |
mv.y += pmv.y; |
757 |
|
|
758 |
if (mv_x < low) { |
if (mv.x < low) { |
759 |
mv_x += range; |
mv.x += range; |
760 |
} else if (mv_x > high) { |
} else if (mv.x > high) { |
761 |
mv_x -= range; |
mv.x -= range; |
762 |
} |
} |
763 |
|
|
764 |
if (mv_y < low) { |
if (mv.y < low) { |
765 |
mv_y += range; |
mv.y += range; |
766 |
} else if (mv_y > high) { |
} else if (mv.y > high) { |
767 |
mv_y -= range; |
mv.y -= range; |
768 |
} |
} |
769 |
|
|
770 |
mv->x = mv_x; |
ret_mv->x = mv.x; |
771 |
mv->y = mv_y; |
ret_mv->y = mv.y; |
772 |
|
} |
773 |
|
|
774 |
|
|
775 |
|
|
776 |
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
777 |
|
{ |
778 |
|
int length = 1 << (fcode+4); |
779 |
|
|
780 |
|
if (quarterpel) value *= 2; |
781 |
|
|
782 |
|
if (value < -length) |
783 |
|
return -length; |
784 |
|
else if (value >= length) |
785 |
|
return length-1; |
786 |
|
else return value; |
787 |
} |
} |
788 |
|
|
789 |
|
|
790 |
|
/* for P_VOP set gmc_mv to NULL */ |
791 |
void |
void |
792 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
793 |
Bitstream * bs, |
Bitstream * bs, |
794 |
int rounding, |
int rounding, |
795 |
|
int reduced_resolution, |
796 |
int quant, |
int quant, |
797 |
int fcode, |
int fcode, |
798 |
int intra_dc_threshold) |
int intra_dc_threshold, |
799 |
|
VECTOR * gmc_mv) |
800 |
{ |
{ |
801 |
|
|
802 |
uint32_t x, y; |
uint32_t x, y; |
803 |
uint32_t bound; |
uint32_t bound; |
804 |
|
int cp_mb, st_mb; |
805 |
|
int mb_width = dec->mb_width; |
806 |
|
int mb_height = dec->mb_height; |
807 |
|
|
808 |
|
if (reduced_resolution) |
809 |
|
{ |
810 |
|
mb_width /= 2; |
811 |
|
mb_height /= 2; |
812 |
|
} |
813 |
|
|
814 |
start_timer(); |
start_timer(); |
815 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
816 |
dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
817 |
stop_edges_timer(); |
stop_edges_timer(); |
818 |
|
|
819 |
bound = 0; |
bound = 0; |
820 |
|
|
821 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
822 |
for (x = 0; x < dec->mb_width; x++) { |
cp_mb = st_mb = 0; |
823 |
|
for (x = 0; x < mb_width; x++) { |
824 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
825 |
|
|
826 |
// skip stuffing |
// skip stuffing |
829 |
|
|
830 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) |
831 |
{ |
{ |
832 |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
bound = read_video_packet_header(bs, dec, fcode - 1, |
833 |
x = bound % dec->mb_width; |
&quant, &fcode, NULL, &intra_dc_threshold); |
834 |
y = bound / dec->mb_width; |
x = bound % mb_width; |
835 |
|
y = bound / mb_width; |
836 |
} |
} |
837 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
838 |
|
|
847 |
uint32_t cbpy; |
uint32_t cbpy; |
848 |
uint32_t cbp; |
uint32_t cbp; |
849 |
uint32_t intra; |
uint32_t intra; |
850 |
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
851 |
|
|
852 |
|
cp_mb++; |
853 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
854 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
855 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
864 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
865 |
} |
} |
866 |
|
|
867 |
|
if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
868 |
|
{ |
869 |
|
mcsel = BitstreamGetBit(bs); |
870 |
|
} |
871 |
|
|
872 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
873 |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
874 |
|
|
888 |
mb->quant = quant; |
mb->quant = quant; |
889 |
|
|
890 |
if (dec->interlacing) { |
if (dec->interlacing) { |
891 |
|
if (cbp || intra) { |
892 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
893 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DEBUG1("decp: field_dct: ", mb->field_dct); |
894 |
|
} |
895 |
|
|
896 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
897 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
907 |
} |
} |
908 |
|
|
909 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
910 |
if (dec->interlacing && mb->field_pred) { |
|
911 |
|
if (mcsel) |
912 |
|
{ |
913 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
914 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
915 |
|
|
916 |
|
} else if (dec->interlacing && mb->field_pred) { |
917 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
918 |
fcode, bound); |
fcode, bound); |
919 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
926 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
927 |
mb->mvs[0].y; |
mb->mvs[0].y; |
928 |
} |
} |
929 |
} else if (mb->mode == |
} else if (mb->mode == MODE_INTER4V ) { |
930 |
MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */ ) { |
|
931 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
932 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
933 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
939 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
940 |
0; |
0; |
941 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
942 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
943 |
continue; |
continue; |
944 |
} |
} |
945 |
|
|
946 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
947 |
rounding); |
rounding, reduced_resolution); |
948 |
} else // not coded |
|
949 |
|
} |
950 |
|
else if (gmc_mv) /* not coded S_VOP macroblock */ |
951 |
{ |
{ |
|
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
|
952 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
953 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
954 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
955 |
|
decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution); |
956 |
|
} |
957 |
|
else /* not coded P_VOP macroblock */ |
958 |
|
{ |
959 |
|
mb->mode = MODE_NOT_CODED; |
960 |
|
|
961 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
962 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
|
|
963 |
// copy macroblock directly from ref to cur |
// copy macroblock directly from ref to cur |
964 |
|
|
965 |
start_timer(); |
start_timer(); |
966 |
|
|
967 |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
if (reduced_resolution) |
968 |
(16 * x), |
{ |
969 |
dec->refn[0].y + (16 * y) * dec->edged_width + |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
970 |
(16 * x), dec->edged_width); |
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
971 |
|
dec->edged_width); |
972 |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
|
973 |
(16 * x + 8), |
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
974 |
dec->refn[0].y + (16 * y) * dec->edged_width + |
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
975 |
(16 * x + 8), dec->edged_width); |
dec->edged_width/2); |
976 |
|
|
977 |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
978 |
(16 * x), |
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
979 |
dec->refn[0].y + (16 * y + |
dec->edged_width/2); |
980 |
8) * dec->edged_width + |
} |
981 |
(16 * x), dec->edged_width); |
else |
982 |
|
{ |
983 |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
984 |
(16 * x + 8), |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
985 |
dec->refn[0].y + (16 * y + |
dec->edged_width); |
986 |
8) * dec->edged_width + |
|
987 |
(16 * x + 8), dec->edged_width); |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
988 |
|
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
|
|
(8 * x), |
|
|
dec->refn[0].u + |
|
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
|
989 |
dec->edged_width / 2); |
dec->edged_width / 2); |
990 |
|
|
991 |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
992 |
(8 * x), |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
|
dec->refn[0].v + |
|
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
|
993 |
dec->edged_width / 2); |
dec->edged_width / 2); |
994 |
|
} |
995 |
|
|
996 |
stop_transfer_timer(); |
stop_transfer_timer(); |
997 |
|
|
998 |
|
if(dec->out_frm && cp_mb > 0) { |
999 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1000 |
|
cp_mb = 0; |
1001 |
|
} |
1002 |
|
st_mb = x+1; |
1003 |
} |
} |
1004 |
} |
} |
1005 |
|
if(dec->out_frm && cp_mb > 0) |
1006 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1007 |
} |
} |
1008 |
} |
} |
1009 |
|
|
1081 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1082 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1083 |
|
|
1084 |
|
|
1085 |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1086 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1087 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1088 |
|
|
1089 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) |
1090 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
{ |
1091 |
|
uv_dx /= 2; |
1092 |
|
uv_dy /= 2; |
1093 |
|
} |
1094 |
|
|
1095 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1096 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1097 |
} else { |
} else { |
1098 |
int sum; |
int sum; |
1099 |
|
|
1100 |
|
if(dec->quarterpel) |
1101 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1102 |
|
else |
1103 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
uv_dx = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1104 |
|
|
1105 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1106 |
|
|
1107 |
|
if(dec->quarterpel) |
1108 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1109 |
|
else |
1110 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1111 |
uv_dy = |
|
1112 |
(sum == |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1113 |
} |
} |
1114 |
|
|
1115 |
start_timer(); |
start_timer(); |
1116 |
|
if(dec->quarterpel) { |
1117 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->refh.y, dec->refh.y + 64, |
1118 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
1119 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1120 |
|
} |
1121 |
|
else { |
1122 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
1123 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1124 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1125 |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1126 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1127 |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1128 |
0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1129 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1130 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
} |
1131 |
0); |
|
1132 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
1133 |
uv_dx, uv_dy, stride2, 0); |
uv_dx, uv_dy, stride2, 0); |
1134 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
1136 |
stop_comp_timer(); |
stop_comp_timer(); |
1137 |
|
|
1138 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1139 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1140 |
|
|
1141 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
1142 |
{ |
{ |
1143 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1144 |
|
|
1145 |
start_timer(); |
start_timer(); |
1146 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
1147 |
stop_coding_timer(); |
stop_coding_timer(); |
1148 |
|
|
1149 |
start_timer(); |
start_timer(); |
1181 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1182 |
} |
} |
1183 |
|
|
|
|
|
1184 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
1185 |
// decode an B-frame direct & inter macroblock |
// decode an B-frame direct & inter macroblock |
1186 |
void |
void |
1190 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1191 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1192 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t cbp, |
|
1193 |
Bitstream * bs) |
Bitstream * bs) |
1194 |
{ |
{ |
1195 |
|
|
1204 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
1205 |
uint32_t i; |
uint32_t i; |
1206 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1207 |
|
const uint32_t cbp = pMB->cbp; |
1208 |
|
|
1209 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1210 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1211 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1212 |
|
|
1213 |
|
|
1214 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1215 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1216 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1217 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
|
|
1218 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1219 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1220 |
|
|
1221 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) |
1222 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
{ |
1223 |
|
uv_dx /= 2; |
1224 |
|
uv_dy /= 2; |
1225 |
|
|
1226 |
|
b_uv_dx /= 2; |
1227 |
|
b_uv_dy /= 2; |
1228 |
|
} |
1229 |
|
|
1230 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1231 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1232 |
|
|
1233 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1234 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1235 |
} else { |
} else { |
1236 |
int sum; |
int sum; |
1237 |
|
|
1238 |
|
if(dec->quarterpel) |
1239 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1240 |
|
else |
1241 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
uv_dx = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1242 |
|
|
1243 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1244 |
|
|
1245 |
|
if(dec->quarterpel) |
1246 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1247 |
|
else |
1248 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1249 |
uv_dy = |
|
1250 |
(sum == |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1251 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
1252 |
(ABS(sum) / 16) * 2)); |
|
1253 |
|
if(dec->quarterpel) |
1254 |
sum = |
sum = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1255 |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
else |
1256 |
pMB->b_mvs[3].x; |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1257 |
b_uv_dx = |
|
1258 |
(sum == |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1259 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
1260 |
(ABS(sum) / 16) * 2)); |
if(dec->quarterpel) |
1261 |
|
sum = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1262 |
sum = |
else |
1263 |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1264 |
pMB->b_mvs[3].y; |
|
1265 |
b_uv_dy = |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1266 |
} |
} |
1267 |
|
|
1268 |
|
|
1269 |
start_timer(); |
start_timer(); |
1270 |
|
if(dec->quarterpel) { |
1271 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1272 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
1273 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
1274 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1275 |
|
else { |
1276 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
1277 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
1278 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1279 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
1280 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
1281 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1282 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
1283 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
1284 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1285 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
1286 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1287 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1288 |
|
} |
1289 |
|
} |
1290 |
|
else { |
1291 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1292 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1293 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1297 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1298 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1299 |
0); |
0); |
1300 |
|
} |
1301 |
|
|
1302 |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1303 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1304 |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1305 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1306 |
|
|
1307 |
|
|
1308 |
|
if(dec->quarterpel) { |
1309 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1310 |
|
interpolate16x16_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
1311 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
1312 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1313 |
|
else { |
1314 |
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
1315 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
1316 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1317 |
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
1318 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
1319 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1320 |
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
1321 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
1322 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1323 |
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
1324 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1325 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1326 |
|
} |
1327 |
|
} |
1328 |
|
else { |
1329 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1330 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1331 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1337 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1338 |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1339 |
stride, 0); |
stride, 0); |
1340 |
|
} |
1341 |
|
|
1342 |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1343 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1344 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1345 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1346 |
|
|
1347 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1348 |
stride); |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1349 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, |
1350 |
stride); |
stride, 1, 8); |
1351 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
|
1352 |
stride); |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1353 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1354 |
16 * y_pos + 8, stride); |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1355 |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
stride, 1, 8); |
1356 |
stride2); |
|
1357 |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1358 |
stride2); |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1359 |
|
dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1360 |
|
stride, 1, 8); |
1361 |
|
|
1362 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1363 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1364 |
|
dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1365 |
|
stride, 1, 8); |
1366 |
|
|
1367 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1368 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1369 |
|
dec->refn[2].u + (8 * y_pos * stride2) + 8 * x_pos, |
1370 |
|
stride2, 1, 8); |
1371 |
|
|
1372 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1373 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1374 |
|
dec->refn[2].v + (8 * y_pos * stride2) + 8 * x_pos, |
1375 |
|
stride2, 1, 8); |
1376 |
|
|
1377 |
stop_comp_timer(); |
stop_comp_timer(); |
1378 |
|
|
1379 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1380 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1381 |
|
|
1382 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
1383 |
{ |
{ |
1384 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1385 |
|
|
1386 |
start_timer(); |
start_timer(); |
1387 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
1388 |
stop_coding_timer(); |
stop_coding_timer(); |
1389 |
|
|
1390 |
start_timer(); |
start_timer(); |
1466 |
int fcode_forward, |
int fcode_forward, |
1467 |
int fcode_backward) |
int fcode_backward) |
1468 |
{ |
{ |
|
|
|
1469 |
uint32_t x, y; |
uint32_t x, y; |
1470 |
VECTOR mv, zeromv; |
VECTOR mv; |
1471 |
|
const VECTOR zeromv = {0,0}; |
1472 |
|
#ifdef BFRAMES_DEC_DEBUG |
1473 |
|
FILE *fp; |
1474 |
|
static char first=0; |
1475 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1476 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1477 |
|
} |
1478 |
|
#endif |
1479 |
|
|
1480 |
start_timer(); |
start_timer(); |
1481 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1482 |
dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
1483 |
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1484 |
|
dec->width, dec->height); |
1485 |
stop_edges_timer(); |
stop_edges_timer(); |
1486 |
|
|
1487 |
|
#ifdef BFRAMES_DEC_DEBUG |
1488 |
|
if (!first){ |
1489 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1490 |
|
} |
1491 |
|
#endif |
1492 |
|
|
1493 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1494 |
// Initialize Pred Motion Vector |
// Initialize Pred Motion Vector |
1495 |
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
dec->p_fmv = dec->p_bmv = zeromv; |
1496 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1497 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1498 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1499 |
|
|
1500 |
mb->mvs[0].x = mb->mvs[0].y = zeromv.x = zeromv.y = mv.x = mv.y = |
mv = |
1501 |
0; |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1502 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1503 |
|
|
1504 |
// the last P_VOP is skip macroblock ? |
// the last P_VOP is skip macroblock ? |
1505 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1506 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
|
mb->mb_type = MODE_FORWARD; |
|
1507 |
mb->cbp = 0; |
mb->cbp = 0; |
1508 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
#ifdef BFRAMES_DEC_DEBUG |
1509 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mb_type = MODE_NOT_CODED; |
1510 |
mb->quant = 8; |
BFRAME_DEBUG |
1511 |
|
#endif |
1512 |
|
mb->mb_type = MODE_FORWARD; |
1513 |
|
mb->quant = last_mb->quant; |
1514 |
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1515 |
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1516 |
|
|
1517 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1518 |
continue; |
continue; |
1519 |
} |
} |
|
//t=BitstreamShowBits(bs,32); |
|
1520 |
|
|
1521 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { // modb=='0' |
1522 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1536 |
} else if (quant < 1) { |
} else if (quant < 1) { |
1537 |
quant = 1; |
quant = 1; |
1538 |
} |
} |
|
} else { |
|
|
quant = 8; |
|
1539 |
} |
} |
|
mb->quant = quant; |
|
1540 |
} else { |
} else { |
1541 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mb_type = MODE_DIRECT_NONE_MV; |
1542 |
mb->cbp = 0; |
mb->cbp = 0; |
1543 |
} |
} |
1544 |
|
|
1545 |
mb->mode = MODE_INTER; |
mb->quant = quant; |
1546 |
|
mb->mode = MODE_INTER4V; |
1547 |
//DEBUG1("Switch bm_type=",mb->mb_type); |
//DEBUG1("Switch bm_type=",mb->mb_type); |
1548 |
|
|
1549 |
|
#ifdef BFRAMES_DEC_DEBUG |
1550 |
|
BFRAME_DEBUG |
1551 |
|
#endif |
1552 |
|
|
1553 |
switch (mb->mb_type) { |
switch (mb->mb_type) { |
1554 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1555 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1556 |
|
|
1557 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1558 |
{ // Because this file is a C file not C++ so I use '{' to define var |
{ |
1559 |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
dec->time_pp; |
|
1560 |
int i; |
int i; |
1561 |
|
|
1562 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1563 |
mb->mvs[i].x = |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1564 |
(int32_t) ((TRB * last_mb->mvs[i].x) / TRD + |
/ TRD + mv.x); |
1565 |
mb->mvs[0].x); |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1566 |
mb->b_mvs[i].x = |
? ((TRB - TRD) * last_mb->mvs[i].x) |
1567 |
(int32_t) ((mb->mvs[0].x == |
/ TRD |
1568 |
0) ? ((TRB - |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1569 |
TRD) * last_mb->mvs[i].x) / |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1570 |
TRD : mb->mvs[i].x - last_mb->mvs[i].x); |
/ TRD + mv.y); |
1571 |
mb->mvs[i].y = |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1572 |
(int32_t) ((TRB * last_mb->mvs[i].y) / TRD + |
? ((TRB - TRD) * last_mb->mvs[i].y) |
1573 |
mb->mvs[0].y); |
/ TRD |
1574 |
mb->b_mvs[i].y = |
: mb->mvs[i].y - last_mb->mvs[i].y); |
|
(int32_t) ((mb->mvs[0].y == |
|
|
0) ? ((TRB - |
|
|
TRD) * last_mb->mvs[i].y) / |
|
|
TRD : mb->mvs[i].y - last_mb->mvs[i].y); |
|
1575 |
} |
} |
1576 |
//DEBUG("B-frame Direct!\n"); |
//DEBUG("B-frame Direct!\n"); |
1577 |
} |
} |
|
mb->mode = MODE_INTER4V; |
|
1578 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1579 |
mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1580 |
break; |
break; |
1581 |
|
|
1582 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1583 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1584 |
dec->p_fmv); |
dec->p_fmv); |
1585 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1586 |
|
|
1587 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1588 |
fcode_backward, dec->p_bmv); |
fcode_backward, dec->p_bmv); |
1589 |
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1590 |
mb->b_mvs[3].x = mb->b_mvs[0].x; |
mb->b_mvs[3] = mb->b_mvs[0]; |
|
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = |
|
|
mb->b_mvs[3].y = mb->b_mvs[0].y; |
|
1591 |
|
|
1592 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1593 |
mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1594 |
//DEBUG("B-frame Bidir!\n"); |
//DEBUG("B-frame Bidir!\n"); |
1595 |
break; |
break; |
1596 |
|
|
1597 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1598 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1599 |
dec->p_bmv); |
dec->p_bmv); |
1600 |
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1601 |
|
|
1602 |
|
mb->mode = MODE_INTER; |
1603 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1604 |
//DEBUG("B-frame Backward!\n"); |
//DEBUG("B-frame Backward!\n"); |
1605 |
break; |
break; |
1607 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1608 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1609 |
dec->p_fmv); |
dec->p_fmv); |
1610 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
1611 |
|
|
1612 |
|
mb->mode = MODE_INTER; |
1613 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1614 |
//DEBUG("B-frame Forward!\n"); |
//DEBUG("B-frame Forward!\n"); |
1615 |
break; |
break; |
1620 |
|
|
1621 |
} // end of FOR |
} // end of FOR |
1622 |
} |
} |
1623 |
|
#ifdef BFRAMES_DEC_DEBUG |
1624 |
|
if (!first){ |
1625 |
|
first=1; |
1626 |
|
if (fp) |
1627 |
|
fclose(fp); |
1628 |
|
} |
1629 |
|
#endif |
1630 |
} |
} |
1631 |
|
|
1632 |
// swap two MACROBLOCK array |
// swap two MACROBLOCK array |
1642 |
|
|
1643 |
int |
int |
1644 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1645 |
XVID_DEC_FRAME * frame) |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
1646 |
{ |
{ |
1647 |
|
|
1648 |
Bitstream bs; |
Bitstream bs; |
1649 |
uint32_t rounding; |
uint32_t rounding; |
1650 |
|
uint32_t reduced_resolution; |
1651 |
uint32_t quant; |
uint32_t quant; |
1652 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1653 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1654 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1655 |
|
VECTOR gmc_mv[5]; |
1656 |
uint32_t vop_type; |
uint32_t vop_type; |
1657 |
|
int success = 0; |
1658 |
|
|
1659 |
start_global_timer(); |
start_global_timer(); |
1660 |
|
|
1661 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1662 |
|
|
1663 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1664 |
|
|
1665 |
|
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
1666 |
|
if(frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1667 |
|
{ |
1668 |
|
if (stats) |
1669 |
|
stats->notify = XVID_DEC_VOP; |
1670 |
|
frame->length = 1; |
1671 |
|
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1672 |
|
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
1673 |
|
emms(); |
1674 |
|
return XVID_ERR_OK; |
1675 |
|
} |
1676 |
|
|
1677 |
|
start: |
1678 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
1679 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
1680 |
dec->frames++; |
dec->frames++; |
1681 |
|
|
1682 |
|
xxx: |
1683 |
vop_type = |
vop_type = |
1684 |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1685 |
&fcode_backward, &intra_dc_threshold); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); |
1686 |
|
|
1687 |
|
//DPRINTF(DPRINTF_HEADER, "vop_type=%i", vop_type); |
1688 |
|
|
1689 |
|
if (vop_type == -1 && success) |
1690 |
|
goto done; |
1691 |
|
|
1692 |
|
if (vop_type == -2 || vop_type == -3) |
1693 |
|
{ |
1694 |
|
if (vop_type == -3) |
1695 |
|
decoder_resize(dec); |
1696 |
|
|
1697 |
|
if (stats) |
1698 |
|
{ |
1699 |
|
stats->notify = XVID_DEC_VOL; |
1700 |
|
stats->data.vol.general = 0; |
1701 |
|
if (dec->interlacing) |
1702 |
|
stats->data.vol.general |= XVID_INTERLACING; |
1703 |
|
stats->data.vol.width = dec->width; |
1704 |
|
stats->data.vol.height = dec->height; |
1705 |
|
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
1706 |
|
stats->data.vol.par_width = dec->par_width; |
1707 |
|
stats->data.vol.par_height = dec->par_height; |
1708 |
|
frame->length = BitstreamPos(&bs) / 8; |
1709 |
|
return XVID_ERR_OK; |
1710 |
|
} |
1711 |
|
goto xxx; |
1712 |
|
} |
1713 |
|
|
1714 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
1715 |
|
|
1716 |
switch (vop_type) { |
switch (vop_type) { |
1717 |
case P_VOP: |
case P_VOP: |
1718 |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1719 |
intra_dc_threshold); |
fcode_forward, intra_dc_threshold, NULL); |
1720 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1721 |
DEBUG1("P_VOP Time=", dec->time); |
DEBUG1("P_VOP Time=", dec->time); |
1722 |
#endif |
#endif |
1723 |
break; |
break; |
1724 |
|
|
1725 |
case I_VOP: |
case I_VOP: |
1726 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1727 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1728 |
DEBUG1("I_VOP Time=", dec->time); |
DEBUG1("I_VOP Time=", dec->time); |
1729 |
#endif |
#endif |
1742 |
#endif |
#endif |
1743 |
break; |
break; |
1744 |
|
|
1745 |
|
case S_VOP : |
1746 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1747 |
|
fcode_forward, intra_dc_threshold, gmc_mv); |
1748 |
|
break; |
1749 |
|
|
1750 |
case N_VOP: // vop not coded |
case N_VOP: // vop not coded |
1751 |
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1752 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1753 |
|
#ifdef BFRAMES_DEC |
1754 |
|
DEBUG1("N_VOP Time=", dec->time); |
1755 |
|
#endif |
1756 |
break; |
break; |
1757 |
|
|
1758 |
default: |
default: |
1759 |
|
if (stats) |
1760 |
|
stats->notify = 0; |
1761 |
|
|
1762 |
|
emms(); |
1763 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1764 |
} |
} |
1765 |
|
|
1766 |
frame->length = BitstreamPos(&bs) / 8; |
|
1767 |
|
/* reduced resolution deblocking filter */ |
1768 |
|
|
1769 |
|
if (reduced_resolution) |
1770 |
|
{ |
1771 |
|
const int rmb_height = dec->mb_height / 2; |
1772 |
|
const int rmb_width = dec->mb_width / 2; |
1773 |
|
const int edged_width2 = dec->edged_width /2; |
1774 |
|
int i,j; |
1775 |
|
|
1776 |
|
/* horizontal deblocking */ |
1777 |
|
|
1778 |
|
for (j = 1; j < rmb_height*2; j++) // luma: j,i in block units |
1779 |
|
for (i = 0; i < rmb_width*2; i++) |
1780 |
|
{ |
1781 |
|
if (dec->mbs[(j-1)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED || |
1782 |
|
dec->mbs[(j+0)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED) |
1783 |
|
{ |
1784 |
|
xvid_HFilter_31_C(dec->cur.y + (j*16 - 1)*dec->edged_width + i*16, |
1785 |
|
dec->cur.y + (j*16 + 0)*dec->edged_width + i*16, 2); |
1786 |
|
} |
1787 |
|
} |
1788 |
|
|
1789 |
|
for (j = 1; j < rmb_height; j++) // chroma |
1790 |
|
for (i = 0; i < rmb_width; i++) |
1791 |
|
{ |
1792 |
|
if (dec->mbs[(j-1)*dec->mb_width + i].mode != MODE_NOT_CODED || |
1793 |
|
dec->mbs[(j+0)*dec->mb_width + i].mode != MODE_NOT_CODED) |
1794 |
|
{ |
1795 |
|
hfilter_31(dec->cur.u + (j*16 - 1)*edged_width2 + i*16, |
1796 |
|
dec->cur.u + (j*16 + 0)*edged_width2 + i*16, 2); |
1797 |
|
hfilter_31(dec->cur.v + (j*16 - 1)*edged_width2 + i*16, |
1798 |
|
dec->cur.v + (j*16 + 0)*edged_width2 + i*16, 2); |
1799 |
|
} |
1800 |
|
} |
1801 |
|
|
1802 |
|
/* vertical deblocking */ |
1803 |
|
|
1804 |
|
for (j = 0; j < rmb_height*2; j++) // luma: i,j in block units |
1805 |
|
for (i = 1; i < rmb_width*2; i++) |
1806 |
|
{ |
1807 |
|
if (dec->mbs[(j/2)*dec->mb_width + (i-1)/2].mode != MODE_NOT_CODED || |
1808 |
|
dec->mbs[(j/2)*dec->mb_width + (i+0)/2].mode != MODE_NOT_CODED) |
1809 |
|
{ |
1810 |
|
vfilter_31(dec->cur.y + (j*16)*dec->edged_width + i*16 - 1, |
1811 |
|
dec->cur.y + (j*16)*dec->edged_width + i*16 + 0, |
1812 |
|
dec->edged_width, 2); |
1813 |
|
} |
1814 |
|
} |
1815 |
|
|
1816 |
|
for (j = 0; j < rmb_height; j++) // chroma |
1817 |
|
for (i = 1; i < rmb_width; i++) |
1818 |
|
{ |
1819 |
|
if (dec->mbs[j*dec->mb_width + i - 1].mode != MODE_NOT_CODED || |
1820 |
|
dec->mbs[j*dec->mb_width + i + 0].mode != MODE_NOT_CODED) |
1821 |
|
{ |
1822 |
|
vfilter_31(dec->cur.u + (j*16)*edged_width2 + i*16 - 1, |
1823 |
|
dec->cur.u + (j*16)*edged_width2 + i*16 + 0, |
1824 |
|
edged_width2, 2); |
1825 |
|
vfilter_31(dec->cur.v + (j*16)*edged_width2 + i*16 - 1, |
1826 |
|
dec->cur.v + (j*16)*edged_width2 + i*16 + 0, |
1827 |
|
edged_width2, 2); |
1828 |
|
} |
1829 |
|
} |
1830 |
|
} |
1831 |
|
|
1832 |
|
BitstreamByteAlign(&bs); |
1833 |
|
|
1834 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1835 |
// test if no B_VOP |
// test if no B_VOP |
1836 |
if (dec->low_delay) { |
if (dec->low_delay || dec->frames == 0 || ((dec->packed_mode) && !(frame->length > BitstreamPos(&bs) / 8))) { |
1837 |
#endif |
#endif |
1838 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1839 |
frame->image, frame->stride, frame->colorspace); |
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
1840 |
|
|
1841 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1842 |
} else { |
} else { |
1843 |
if (dec->frames >= 1) { |
if (dec->frames >= 1 && !(dec->packed_mode)) { |
1844 |
start_timer(); |
start_timer(); |
1845 |
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
if ((vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP)) { |
1846 |
image_output(&dec->refn[0], dec->width, dec->height, |
image_output(&dec->refn[0], dec->width, dec->height, |
1847 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, frame->image, frame->stride, |
1848 |
frame->colorspace); |
frame->colorspace, dec->interlacing); |
1849 |
} else if (vop_type == B_VOP) { |
} else if (vop_type == B_VOP) { |
1850 |
image_output(&dec->cur, dec->width, dec->height, |
image_output(&dec->cur, dec->width, dec->height, |
1851 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, frame->image, frame->stride, |
1852 |
frame->colorspace); |
frame->colorspace, dec->interlacing); |
1853 |
} |
} |
1854 |
stop_conv_timer(); |
stop_conv_timer(); |
1855 |
} |
} |
1856 |
} |
} |
1857 |
#endif |
#endif |
1858 |
|
|
1859 |
if (vop_type == I_VOP || vop_type == P_VOP) { |
if (vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP) { |
1860 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1861 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1862 |
|
|
1863 |
// swap MACROBLOCK |
// swap MACROBLOCK |
1864 |
if (dec->low_delay && vop_type == P_VOP) |
// the Divx will not set the low_delay flage some times |
1865 |
|
// so follow code will wrong to not swap at that time |
1866 |
|
// this will broken bitstream! so I'm change it, |
1867 |
|
// But that is not the best way! can anyone tell me how |
1868 |
|
// to do another way? |
1869 |
|
// 18-07-2002 MinChen<chenm001@163.com> |
1870 |
|
//if (!dec->low_delay && vop_type == P_VOP) |
1871 |
|
if (vop_type == P_VOP) |
1872 |
mb_swap(&dec->mbs, &dec->last_mbs); |
mb_swap(&dec->mbs, &dec->last_mbs); |
1873 |
} |
} |
1874 |
|
|
1875 |
|
|
1876 |
|
if (success == 0 && dec->packed_mode) |
1877 |
|
{ |
1878 |
|
success = 1; |
1879 |
|
// if (frame->length > BitstreamPos(&bs) / 8) // multiple vops packed together |
1880 |
|
goto start; |
1881 |
|
} |
1882 |
|
|
1883 |
|
done : |
1884 |
|
|
1885 |
|
frame->length = BitstreamPos(&bs) / 8; |
1886 |
|
|
1887 |
|
if (stats) |
1888 |
|
{ |
1889 |
|
stats->notify = XVID_DEC_VOP; |
1890 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1891 |
|
stats->data.vop.time_increment = 0; //XXX: todo |
1892 |
|
} |
1893 |
|
|
1894 |
emms(); |
emms(); |
1895 |
|
|
1896 |
stop_global_timer(); |
stop_global_timer(); |