1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2002-2003 Peter Ross <pross@xvid.org> |
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
8 |
* |
* |
9 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
20 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
|
* |
|
|
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
|
|
* MinChen <chenm001@163.com> |
|
|
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
|
|
* Fix a little bug for low_delay flage |
|
|
* MinChen <chenm001@163.com> |
|
|
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
|
|
* 22.06.2002 added primative N_VOP support |
|
|
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
|
|
* 08.05.2002 add low_delay support for B_VOP decode |
|
|
* MinChen <chenm001@163.com> |
|
|
* 05.05.2002 fix some B-frame decode problem |
|
|
* 02.05.2002 add B-frame decode support(have some problem); |
|
|
* MinChen <chenm001@163.com> |
|
|
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
|
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
|
|
* reconstructing blocks, thus artifacts |
|
|
* interlacing speedup - used transfers to re-interlace |
|
|
* interlaced decoding should be as fast as progressive now |
|
|
* 26.03.2002 interlacing support - moved transfers outside decode loop |
|
|
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
|
|
* 22.12.2001 lock based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
23 |
* $Id$ |
* $Id$ |
24 |
* |
* |
25 |
*************************************************************************/ |
****************************************************************************/ |
26 |
|
|
27 |
#include <stdio.h> |
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.h> |
40 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
41 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
42 |
|
|
43 |
#include "quant/quant_h263.h" |
#include "quant/quant.h" |
|
#include "quant/quant_mpeg4.h" |
|
44 |
#include "dct/idct.h" |
#include "dct/idct.h" |
45 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
46 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
53 |
#include "utils/timer.h" |
#include "utils/timer.h" |
54 |
#include "utils/emms.h" |
#include "utils/emms.h" |
55 |
#include "motion/motion.h" |
#include "motion/motion.h" |
56 |
|
#include "motion/gmc.h" |
57 |
|
|
58 |
#include "image/image.h" |
#include "image/image.h" |
59 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
63 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
64 |
{ |
{ |
65 |
/* free existing */ |
/* free existing */ |
|
|
|
66 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
67 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
68 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
77 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
78 |
|
|
79 |
/* realloc */ |
/* realloc */ |
|
|
|
80 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
81 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
82 |
|
|
94 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
95 |
} |
} |
96 |
|
|
97 |
// add by chenm001 <chenm001@163.com> |
/* Support B-frame to reference last 2 frame */ |
|
// for support B-frame to reference last 2 frame |
|
98 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
99 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
100 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
142 |
} |
} |
143 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
144 |
|
|
145 |
// add by chenm001 <chenm001@163.com> |
/* For skip MB flag */ |
|
// for skip MB flag |
|
146 |
dec->last_mbs = |
dec->last_mbs = |
147 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
148 |
CACHE_LINE); |
CACHE_LINE); |
159 |
|
|
160 |
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); |
161 |
|
|
162 |
return XVID_ERR_OK; |
return 0; |
163 |
} |
} |
164 |
|
|
165 |
|
|
166 |
int |
int |
167 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_create(xvid_dec_create_t * create) |
168 |
{ |
{ |
169 |
DECODER *dec; |
DECODER *dec; |
170 |
|
|
171 |
|
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
172 |
|
return XVID_ERR_VERSION; |
173 |
|
|
174 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
175 |
if (dec == NULL) { |
if (dec == NULL) { |
176 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
177 |
} |
} |
178 |
memset(dec, 0, sizeof(DECODER)); |
memset(dec, 0, sizeof(DECODER)); |
179 |
|
|
180 |
param->handle = dec; |
create->handle = dec; |
181 |
|
|
182 |
dec->width = param->width; |
dec->width = create->width; |
183 |
dec->height = param->height; |
dec->height = create->height; |
184 |
|
|
185 |
image_null(&dec->cur); |
image_null(&dec->cur); |
186 |
image_null(&dec->refn[0]); |
image_null(&dec->refn[0]); |
197 |
|
|
198 |
init_timer(); |
init_timer(); |
199 |
|
|
200 |
// add by chenm001 <chenm001@163.com> |
/* For B-frame support (used to save reference frame's time */ |
|
// for support B-frame to save reference frame's time |
|
201 |
dec->frames = 0; |
dec->frames = 0; |
202 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
203 |
dec->low_delay = 0; |
dec->low_delay = 0; |
208 |
if (dec->fixed_dimensions) |
if (dec->fixed_dimensions) |
209 |
return decoder_resize(dec); |
return decoder_resize(dec); |
210 |
else |
else |
211 |
return XVID_ERR_OK; |
return 0; |
212 |
} |
} |
213 |
|
|
214 |
|
|
218 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
219 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
220 |
|
|
221 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); /* image based GMC */ |
/* image based GMC */ |
222 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
223 |
|
|
224 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
225 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
229 |
xvid_free(dec); |
xvid_free(dec); |
230 |
|
|
231 |
write_timer(); |
write_timer(); |
232 |
return XVID_ERR_OK; |
return 0; |
233 |
} |
} |
234 |
|
|
235 |
|
|
241 |
|
|
242 |
|
|
243 |
|
|
244 |
// decode an intra macroblock |
/* decode an intra macroblock */ |
|
|
|
245 |
void |
void |
246 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
247 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
276 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
277 |
} |
} |
278 |
|
|
279 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
280 |
|
|
281 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
282 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
299 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
300 |
|
|
301 |
if (dc_size > 8) { |
if (dc_size > 8) { |
302 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); /* marker */ |
303 |
} |
} |
304 |
|
|
305 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
306 |
start_coeff = 1; |
start_coeff = 1; |
307 |
|
|
308 |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
309 |
} else { |
} else { |
310 |
start_coeff = 0; |
start_coeff = 0; |
311 |
} |
} |
312 |
|
|
313 |
start_timer(); |
start_timer(); |
314 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
315 |
{ |
{ |
316 |
int direction = dec->alternate_vertical_scan ? |
int direction = dec->alternate_vertical_scan ? |
317 |
2 : pMB->acpred_directions[i]; |
2 : pMB->acpred_directions[i]; |
326 |
|
|
327 |
start_timer(); |
start_timer(); |
328 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
329 |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
330 |
} else { |
} else { |
331 |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
332 |
} |
} |
333 |
stop_iquant_timer(); |
stop_iquant_timer(); |
334 |
|
|
368 |
|
|
369 |
|
|
370 |
|
|
371 |
// decode an inter macroblock |
/* decode an inter macroblock */ |
|
|
|
372 |
void |
void |
373 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
374 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
375 |
const uint32_t x_pos, |
const uint32_t x_pos, |
376 |
const uint32_t y_pos, |
const uint32_t y_pos, |
377 |
const uint32_t acpred_flag, |
const uint32_t fcode, |
378 |
const uint32_t cbp, |
const uint32_t cbp, |
379 |
Bitstream * bs, |
Bitstream * bs, |
380 |
const uint32_t quant, |
const uint32_t quant, |
381 |
const uint32_t rounding, |
const uint32_t rounding, |
382 |
const int reduced_resolution, |
const int reduced_resolution) |
|
const int mcsel) |
|
383 |
{ |
{ |
384 |
|
|
385 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
391 |
uint32_t i; |
uint32_t i; |
392 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
393 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
uint8_t *pY_Ref, *pU_Ref, *pV_Ref; /* ref for GMC is _not_ pRef itself */ |
|
394 |
|
|
395 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
396 |
VECTOR mv[4]; /* local copy of mvs */ |
VECTOR mv[4]; /* local copy of mvs */ |
411 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
412 |
} |
} |
413 |
|
|
|
if (mcsel) { |
|
|
mv[0].x = mv[0].y = mv[1].x = mv[1].y = mv[2].x = mv[2].y = mv[3].x = mv[3].y = 0; |
|
|
/* position in ref is same as the block, set vector to (0,0) */ |
|
|
pY_Ref = dec->gmc.y; |
|
|
pU_Ref = dec->gmc.u; |
|
|
pV_Ref = dec->gmc.v; |
|
|
/* but reference itself isn't. It's warped... */ |
|
|
/* Btw., this is too slow! For GMC it should simply be transfer_16to8add() */ |
|
|
} else { |
|
|
pY_Ref = dec->refn[0].y; |
|
|
pU_Ref = dec->refn[0].u; |
|
|
pV_Ref = dec->refn[0].v; |
|
|
} |
|
|
|
|
414 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
415 |
|
|
416 |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
422 |
start_timer(); |
start_timer(); |
423 |
if (reduced_resolution) |
if (reduced_resolution) |
424 |
{ |
{ |
425 |
interpolate32x32_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos, |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
426 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
427 |
interpolate16x16_switch(dec->cur.u, pU_Ref , 16 * x_pos, 16 * y_pos, |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
428 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
429 |
interpolate16x16_switch(dec->cur.v, pV_Ref , 16 * x_pos, 16 * y_pos, |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
430 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
431 |
|
|
432 |
} |
} |
433 |
else |
else |
434 |
{ |
{ |
435 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
436 |
interpolate16x16_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
437 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
438 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
439 |
} |
} |
440 |
else { |
else { |
441 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
442 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
443 |
} |
} |
444 |
|
|
445 |
interpolate8x8_switch(dec->cur.u, pU_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
446 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
447 |
interpolate8x8_switch(dec->cur.v, pV_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
448 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
449 |
} |
} |
450 |
stop_comp_timer(); |
stop_comp_timer(); |
469 |
start_timer(); |
start_timer(); |
470 |
if (reduced_resolution) |
if (reduced_resolution) |
471 |
{ |
{ |
472 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
473 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
474 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos + 16, 32*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
475 |
mv[1].x, mv[1].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
476 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos, 32*y_pos + 16, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
477 |
mv[2].x, mv[2].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
478 |
interpolate16x16_switch(dec->cur.y, pY_Ref , 32*x_pos + 16, 32*y_pos + 16, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
479 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
480 |
interpolate16x16_switch(dec->cur.u, pU_Ref , 16 * x_pos, 16 * y_pos, |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
481 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
482 |
interpolate16x16_switch(dec->cur.v, pV_Ref , 16 * x_pos, 16 * y_pos, |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
483 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
484 |
|
|
485 |
// set_block(pY_Cur, stride, 32, 32, 127); |
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
486 |
} |
} |
487 |
else |
else |
488 |
{ |
{ |
489 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
490 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
491 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
492 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
493 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
494 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
495 |
mv[1].x, mv[1].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
496 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
497 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
498 |
mv[2].x, mv[2].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
499 |
interpolate8x8_quarterpel(dec->cur.y, pY_Ref , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
500 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
501 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
502 |
} |
} |
503 |
else { |
else { |
504 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, |
505 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
506 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
507 |
mv[1].x, mv[1].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
508 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
509 |
mv[2].x, mv[2].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
510 |
interpolate8x8_switch(dec->cur.y, pY_Ref , 16*x_pos + 8, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
511 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
512 |
} |
} |
513 |
|
|
514 |
interpolate8x8_switch(dec->cur.u, pU_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
515 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
516 |
interpolate8x8_switch(dec->cur.v, pV_Ref , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v , 8 * x_pos, 8 * y_pos, |
517 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
518 |
} |
} |
519 |
stop_comp_timer(); |
stop_comp_timer(); |
522 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
523 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
524 |
|
|
525 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
526 |
{ |
{ |
527 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
528 |
|
|
529 |
start_timer(); |
start_timer(); |
530 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64], direction); |
532 |
|
|
533 |
start_timer(); |
start_timer(); |
534 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
535 |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_h263_inter(&data[i * 64], &block[i * 64], iQuant); |
536 |
} else { |
} else { |
537 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_mpeg_inter(&data[i * 64], &block[i * 64], iQuant); |
538 |
} |
} |
539 |
stop_iquant_timer(); |
stop_iquant_timer(); |
540 |
|
|
584 |
} |
} |
585 |
|
|
586 |
|
|
587 |
|
static void |
588 |
|
decoder_mbgmc(DECODER * dec, |
589 |
|
MACROBLOCK * const pMB, |
590 |
|
const uint32_t x_pos, |
591 |
|
const uint32_t y_pos, |
592 |
|
const uint32_t fcode, |
593 |
|
const uint32_t cbp, |
594 |
|
Bitstream * bs, |
595 |
|
const uint32_t quant, |
596 |
|
const uint32_t rounding, |
597 |
|
const int reduced_resolution) /* no reduced res support */ |
598 |
|
{ |
599 |
|
|
600 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
601 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
602 |
|
|
603 |
|
const uint32_t stride = dec->edged_width; |
604 |
|
const uint32_t stride2 = stride / 2; |
605 |
|
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
606 |
|
uint32_t i; |
607 |
|
const uint32_t iQuant = pMB->quant; |
608 |
|
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
609 |
|
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
610 |
|
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
611 |
|
|
612 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
613 |
|
|
614 |
|
start_timer(); |
615 |
|
|
616 |
|
/* this is where the calculations are done */ |
617 |
|
|
618 |
|
{ NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
619 |
|
|
620 |
|
gmc_data->predict_16x16(gmc_data, |
621 |
|
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
622 |
|
stride, stride, x_pos, y_pos, rounding); |
623 |
|
|
624 |
|
gmc_data->predict_8x8(gmc_data, |
625 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
626 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
627 |
|
stride2, stride2, x_pos, y_pos, rounding); |
628 |
|
|
629 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
630 |
|
|
631 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
632 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
633 |
|
} |
634 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
635 |
|
|
636 |
|
/* |
637 |
|
transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
638 |
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
639 |
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
640 |
|
*/ |
641 |
|
|
642 |
|
|
643 |
|
stop_transfer_timer(); |
644 |
|
|
645 |
|
if (!cbp) return; |
646 |
|
|
647 |
|
for (i = 0; i < 6; i++) { |
648 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
649 |
|
|
650 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
651 |
|
{ |
652 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
653 |
|
|
654 |
|
start_timer(); |
655 |
|
get_inter_block(bs, &block[i * 64], direction); |
656 |
|
stop_coding_timer(); |
657 |
|
|
658 |
|
start_timer(); |
659 |
|
if (dec->quant_type == 0) { |
660 |
|
dequant_h263_inter(&data[i * 64], &block[i * 64], iQuant); |
661 |
|
} else { |
662 |
|
dequant_mpeg_inter(&data[i * 64], &block[i * 64], iQuant); |
663 |
|
} |
664 |
|
stop_iquant_timer(); |
665 |
|
|
666 |
|
start_timer(); |
667 |
|
idct(&data[i * 64]); |
668 |
|
stop_idct_timer(); |
669 |
|
} |
670 |
|
} |
671 |
|
|
672 |
|
/* interlace + GMC is this possible ??? */ |
673 |
|
/* |
674 |
|
if (dec->interlacing && pMB->field_dct) { |
675 |
|
next_block = stride; |
676 |
|
stride *= 2; |
677 |
|
} |
678 |
|
*/ |
679 |
|
start_timer(); |
680 |
|
if (cbp & 32) |
681 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
682 |
|
if (cbp & 16) |
683 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
684 |
|
if (cbp & 8) |
685 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
686 |
|
if (cbp & 4) |
687 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
688 |
|
if (cbp & 2) |
689 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
690 |
|
if (cbp & 1) |
691 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
692 |
|
stop_transfer_timer(); |
693 |
|
} |
694 |
|
|
695 |
|
|
696 |
void |
void |
697 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
698 |
Bitstream * bs, |
Bitstream * bs, |
734 |
} |
} |
735 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
736 |
|
|
737 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
738 |
|
|
739 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
740 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
761 |
|
|
762 |
if (dec->interlacing) { |
if (dec->interlacing) { |
763 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
764 |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
765 |
} |
} |
766 |
|
|
767 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
799 |
mv.x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
800 |
mv.y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
801 |
|
|
802 |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
803 |
|
|
804 |
mv.x += pmv.x; |
mv.x += pmv.x; |
805 |
mv.y += pmv.y; |
mv.y += pmv.y; |
822 |
|
|
823 |
|
|
824 |
|
|
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
|
|
{ |
|
|
int length = 1 << (fcode+4); |
|
|
|
|
|
if (quarterpel) value *= 2; |
|
|
|
|
|
if (value < -length) |
|
|
return -length; |
|
|
else if (value >= length) |
|
|
return length-1; |
|
|
else return value; |
|
|
} |
|
825 |
|
|
826 |
|
|
827 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
842 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
843 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
844 |
|
|
|
static int framecount=0; |
|
845 |
if (reduced_resolution) |
if (reduced_resolution) |
846 |
{ |
{ |
847 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
855 |
|
|
856 |
if (gmc_warp) |
if (gmc_warp) |
857 |
{ |
{ |
858 |
char filename[80]; |
|
859 |
sprintf(filename,"dGMC%05d.pgm",framecount); |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
860 |
// accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 |
/* { |
861 |
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
fprintf(stderr,"GMC parameters acc=%d(-> 1/%d), %d pts!!!\n", |
|
{ |
|
|
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
|
862 |
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
863 |
dec->sprite_warping_points); |
dec->sprite_warping_points); |
864 |
} |
}*/ |
865 |
|
|
866 |
generate_GMCparameters( dec->sprite_warping_points, |
generate_GMCparameters( dec->sprite_warping_points, |
867 |
(2 << dec->sprite_warping_accuracy), gmc_warp, |
dec->sprite_warping_accuracy, gmc_warp, |
868 |
dec->width, dec->height, &dec->gmc_data); |
dec->width, dec->height, &dec->new_gmc_data); |
|
|
|
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
|
|
mb_width, mb_height, |
|
|
dec->edged_width, dec->edged_width/2, |
|
|
fcode, 0, 0, |
|
|
rounding, dec->mbs, &dec->gmc); |
|
869 |
|
|
870 |
/* |
/* image warping is done block-based in decoder_mbgmc(), now */ |
|
sprintf(filename,"dGMC%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->gmc, |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
|
|
|
sprintf(filename,"dREF%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->refn[0], |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
sprintf(filename,"dCUR%05d.pgm",framecount); |
|
|
image_dump_yuvpgm(&dec->cur, |
|
|
dec->edged_width, dec->width, dec->height, filename); |
|
|
framecount++; |
|
|
*/ |
|
871 |
} |
} |
872 |
|
|
873 |
bound = 0; |
bound = 0; |
877 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
878 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
879 |
|
|
880 |
// skip stuffing |
/* skip stuffing */ |
881 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
882 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
883 |
|
|
890 |
} |
} |
891 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
892 |
|
|
893 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
894 |
|
|
895 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
896 |
if (!(BitstreamGetBit(bs))) // block _is_ coded |
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
897 |
{ |
{ |
898 |
uint32_t mcbpc; |
uint32_t mcbpc; |
899 |
uint32_t cbpc; |
uint32_t cbpc; |
901 |
uint32_t cbpy; |
uint32_t cbpy; |
902 |
uint32_t cbp; |
uint32_t cbp; |
903 |
uint32_t intra; |
uint32_t intra; |
904 |
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
905 |
|
|
906 |
cp_mb++; |
cp_mb++; |
907 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
908 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
909 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
910 |
|
|
911 |
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
912 |
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
913 |
acpred_flag = 0; |
acpred_flag = 0; |
914 |
|
|
915 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
916 |
|
|
|
if (intra) { |
|
|
acpred_flag = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
917 |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
{ |
|
918 |
mcsel = BitstreamGetBit(bs); |
mcsel = BitstreamGetBit(bs); |
919 |
} |
|
920 |
|
if (intra) |
921 |
|
acpred_flag = BitstreamGetBit(bs); |
922 |
|
|
923 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
924 |
DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
925 |
|
|
926 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
927 |
|
|
928 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
929 |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
930 |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
931 |
quant += dquant; |
quant += dquant; |
932 |
if (quant > 31) { |
if (quant > 31) { |
933 |
quant = 31; |
quant = 31; |
934 |
} else if (quant < 1) { |
} else if (quant < 1) { |
935 |
quant = 1; |
quant = 1; |
936 |
} |
} |
937 |
DPRINTF(DPRINTF_MB, "quant %i", quant); |
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
938 |
} |
} |
939 |
mb->quant = quant; |
mb->quant = quant; |
940 |
|
|
941 |
if (dec->interlacing) { |
if (dec->interlacing) { |
942 |
if (cbp || intra) { |
if ((cbp || intra) && !mcsel) { |
943 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
944 |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
945 |
} |
} |
946 |
|
|
947 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
948 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
949 |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
950 |
|
|
951 |
if (mb->field_pred) { |
if (mb->field_pred) { |
952 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
953 |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
954 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
955 |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
956 |
} |
} |
957 |
} |
} |
958 |
} |
} |
959 |
|
|
960 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mcsel) { |
961 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
962 |
|
rounding, reduced_resolution); |
963 |
|
continue; |
964 |
|
|
965 |
if (mcsel) |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->amv; |
|
|
/* already clipped to fcode */ |
|
966 |
|
|
967 |
} else if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
968 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
969 |
fcode, bound); |
fcode, bound); |
970 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
980 |
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); |
981 |
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); |
982 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
983 |
} else // MODE_INTRA, MODE_INTRA_Q |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
984 |
{ |
{ |
985 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
986 |
0; |
0; |
991 |
continue; |
continue; |
992 |
} |
} |
993 |
|
|
994 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
995 |
rounding, reduced_resolution, mcsel); |
rounding, reduced_resolution); |
996 |
|
|
997 |
} |
} |
998 |
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
999 |
{ |
{ |
1000 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->amv; |
|
1001 |
|
|
1002 |
start_timer(); |
start_timer(); |
1003 |
|
|
1004 |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
1005 |
dec->gmc.y + (16*y)*dec->edged_width + (16*x), |
rounding, reduced_resolution); |
|
dec->edged_width); |
|
|
|
|
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->gmc.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->gmc.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
1006 |
|
|
1007 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1008 |
|
|
1018 |
|
|
1019 |
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; |
1020 |
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; |
1021 |
// copy macroblock directly from ref to cur |
/* copy macroblock directly from ref to cur */ |
1022 |
|
|
1023 |
start_timer(); |
start_timer(); |
1024 |
|
|
1066 |
} |
} |
1067 |
|
|
1068 |
|
|
1069 |
// add by MinChen <chenm001@163.com> |
/* decode B-frame motion vector */ |
|
// decode B-frame motion vector |
|
1070 |
void |
void |
1071 |
get_b_motion_vector(DECODER * dec, |
get_b_motion_vector(DECODER * dec, |
1072 |
Bitstream * bs, |
Bitstream * bs, |
1110 |
} |
} |
1111 |
|
|
1112 |
|
|
1113 |
// add by MinChen <chenm001@163.com> |
/* decode an B-frame forward & backward inter macroblock */ |
|
// decode an B-frame forward & backward inter macroblock |
|
1114 |
void |
void |
1115 |
decoder_bf_mbinter(DECODER * dec, |
decoder_bf_mbinter(DECODER * dec, |
1116 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1138 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1139 |
|
|
1140 |
|
|
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
|
1141 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1142 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1143 |
|
|
1144 |
if (dec->quarterpel) |
if (dec->quarterpel) { |
|
{ |
|
1145 |
uv_dx /= 2; |
uv_dx /= 2; |
1146 |
uv_dy /= 2; |
uv_dy /= 2; |
1147 |
} |
} |
1148 |
|
|
1149 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1150 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
} else { |
|
|
int sum; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
|
|
|
|
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
} |
|
1151 |
|
|
1152 |
start_timer(); |
start_timer(); |
1153 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1154 |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
1155 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1156 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1157 |
} |
} else { |
|
else { |
|
1158 |
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, |
1159 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1160 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1161 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1162 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1163 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1164 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1165 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1166 |
} |
} |
1167 |
|
|
1168 |
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, |
1174 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1175 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1176 |
|
|
1177 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) { /* coded */ |
1178 |
{ |
|
1179 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1180 |
|
|
1181 |
start_timer(); |
start_timer(); |
1182 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64], direction); |
1183 |
stop_coding_timer(); |
stop_coding_timer(); |
1184 |
|
|
1185 |
start_timer(); |
start_timer(); |
1186 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) |
1187 |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_h263_inter(&data[i * 64], &block[i * 64], iQuant); |
1188 |
} else { |
else |
1189 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_mpeg_inter(&data[i * 64], &block[i * 64], iQuant); |
1190 |
} |
|
1191 |
stop_iquant_timer(); |
stop_iquant_timer(); |
1192 |
|
|
1193 |
start_timer(); |
start_timer(); |
1217 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1218 |
} |
} |
1219 |
|
|
1220 |
// add by MinChen <chenm001@163.com> |
/* decode an B-frame direct & interpolate macroblock */ |
|
// decode an B-frame direct & inter macroblock |
|
1221 |
void |
void |
1222 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1223 |
IMAGE forward, |
IMAGE forward, |
1225 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1226 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1227 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1228 |
Bitstream * bs) |
Bitstream * bs, |
1229 |
|
const int direct) |
1230 |
{ |
{ |
1231 |
|
|
1232 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1247 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1248 |
|
|
1249 |
|
|
1250 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if (!direct) { |
1251 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1252 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1253 |
|
|
1304 |
|
|
1305 |
start_timer(); |
start_timer(); |
1306 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1307 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) |
1308 |
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1309 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1310 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1342 |
|
|
1343 |
|
|
1344 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1345 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) |
1346 |
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1347 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1348 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1415 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
1416 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1417 |
|
|
1418 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) { /* coded */ |
1419 |
{ |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
|
1420 |
|
|
1421 |
start_timer(); |
start_timer(); |
1422 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64], direction); |
1424 |
|
|
1425 |
start_timer(); |
start_timer(); |
1426 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
1427 |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_h263_inter(&data[i * 64], &block[i * 64], iQuant); |
1428 |
} else { |
} else { |
1429 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
dequant_mpeg_inter(&data[i * 64], &block[i * 64], iQuant); |
1430 |
} |
} |
1431 |
stop_iquant_timer(); |
stop_iquant_timer(); |
1432 |
|
|
1458 |
} |
} |
1459 |
|
|
1460 |
|
|
1461 |
// add by MinChen <chenm001@163.com> |
/* for decode B-frame dbquant */ |
|
// for decode B-frame dbquant |
|
1462 |
int32_t __inline |
int32_t __inline |
1463 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1464 |
{ |
{ |
1465 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1466 |
return (0); |
return (0); |
1467 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1468 |
return (-2); |
return (-2); |
1469 |
else |
else /* '11' */ |
1470 |
return (2); // '11' |
return (2); |
1471 |
} |
} |
1472 |
|
|
1473 |
// add by MinChen <chenm001@163.com> |
/* |
1474 |
// for decode B-frame mb_type |
* For decode B-frame mb_type |
1475 |
// bit ret_value |
* bit ret_value |
1476 |
// 1 0 |
* 1 0 |
1477 |
// 01 1 |
* 01 1 |
1478 |
// 001 2 |
* 001 2 |
1479 |
// 0001 3 |
* 0001 3 |
1480 |
|
*/ |
1481 |
int32_t __inline |
int32_t __inline |
1482 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1483 |
{ |
{ |
1508 |
FILE *fp; |
FILE *fp; |
1509 |
static char first=0; |
static char first=0; |
1510 |
#define BFRAME_DEBUG if (!first && fp){ \ |
#define BFRAME_DEBUG if (!first && fp){ \ |
1511 |
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mode,mb->cbp); \ |
1512 |
} |
} |
1513 |
#endif |
#endif |
1514 |
|
|
1526 |
#endif |
#endif |
1527 |
|
|
1528 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1529 |
// Initialize Pred Motion Vector |
/* Initialize Pred Motion Vector */ |
1530 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1531 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1532 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1536 |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1537 |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1538 |
|
|
1539 |
// skip if the co-located P_VOP macroblock is not coded |
/* |
1540 |
// note: gmc+not_coded isn't skipped |
* skip if the co-located P_VOP macroblock is not coded |
1541 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1542 |
|
* automatically skipped |
1543 |
|
*/ |
1544 |
|
|
1545 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1546 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
/* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
1547 |
mb->cbp = 0; |
mb->cbp = 0; |
1548 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
|
mb->mb_type = MODE_NOT_CODED; |
|
|
BFRAME_DEBUG |
|
|
#endif |
|
|
mb->mb_type = MODE_FORWARD; |
|
1549 |
mb->quant = last_mb->quant; |
mb->quant = last_mb->quant; |
1550 |
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
/* |
1551 |
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1552 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1553 |
|
*/ |
1554 |
|
|
1555 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1556 |
continue; |
continue; |
1557 |
} |
} |
1558 |
|
|
1559 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1560 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1561 |
|
|
1562 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1563 |
|
|
1564 |
if (!modb2) { // modb=='00' |
if (!modb2) { /* modb=='00' */ |
1565 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1566 |
} else { |
} else { |
1567 |
mb->cbp = 0; |
mb->cbp = 0; |
1568 |
} |
} |
1569 |
if (mb->mb_type && mb->cbp) { |
if (mb->mode && mb->cbp) { |
1570 |
quant += get_dbquant(bs); |
quant += get_dbquant(bs); |
1571 |
|
|
1572 |
if (quant > 31) { |
if (quant > 31) { |
1575 |
quant = 1; |
quant = 1; |
1576 |
} |
} |
1577 |
} |
} |
1578 |
|
|
1579 |
|
if (dec->interlacing) { |
1580 |
|
if (mb->cbp) { |
1581 |
|
mb->field_dct = BitstreamGetBit(bs); |
1582 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1583 |
|
} |
1584 |
|
|
1585 |
|
if (mb->mode) { |
1586 |
|
mb->field_pred = BitstreamGetBit(bs); |
1587 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1588 |
|
|
1589 |
|
if (mb->field_pred) { |
1590 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1591 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1592 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1593 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1594 |
|
} |
1595 |
|
} |
1596 |
|
} |
1597 |
|
|
1598 |
} else { |
} else { |
1599 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1600 |
mb->cbp = 0; |
mb->cbp = 0; |
1601 |
} |
} |
1602 |
|
|
1603 |
mb->quant = quant; |
mb->quant = quant; |
1604 |
mb->mode = MODE_INTER4V; |
/* DEBUG1("Switch bm_type=",mb->mode); */ |
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
|
1605 |
|
|
1606 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1607 |
BFRAME_DEBUG |
BFRAME_DEBUG |
1608 |
#endif |
#endif |
1609 |
|
|
1610 |
switch (mb->mb_type) { |
switch (mb->mode) { |
1611 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1612 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1613 |
|
|
1630 |
/ TRD |
/ TRD |
1631 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
: mb->mvs[i].y - last_mb->mvs[i].y); |
1632 |
} |
} |
1633 |
//DEBUG("B-frame Direct!\n"); |
/* DEBUG("B-frame Direct!\n"); */ |
1634 |
} |
} |
1635 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1636 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1637 |
break; |
break; |
1638 |
|
|
1639 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1647 |
mb->b_mvs[3] = mb->b_mvs[0]; |
mb->b_mvs[3] = mb->b_mvs[0]; |
1648 |
|
|
1649 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1650 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
1651 |
//DEBUG("B-frame Bidir!\n"); |
/* DEBUG("B-frame Bidir!\n"); */ |
1652 |
break; |
break; |
1653 |
|
|
1654 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1656 |
dec->p_bmv); |
dec->p_bmv); |
1657 |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1658 |
|
|
|
mb->mode = MODE_INTER; |
|
1659 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1660 |
//DEBUG("B-frame Backward!\n"); |
/* DEBUG("B-frame Backward!\n"); */ |
1661 |
break; |
break; |
1662 |
|
|
1663 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1665 |
dec->p_fmv); |
dec->p_fmv); |
1666 |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1667 |
|
|
|
mb->mode = MODE_INTER; |
|
1668 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1669 |
//DEBUG("B-frame Forward!\n"); |
/* DEBUG("B-frame Forward!\n"); */ |
1670 |
break; |
break; |
1671 |
|
|
1672 |
default: |
default: |
1673 |
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1674 |
} |
} |
1675 |
|
} /* End of for */ |
|
} // end of FOR |
|
1676 |
} |
} |
1677 |
|
|
1678 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1679 |
if (!first){ |
if (!first){ |
1680 |
first=1; |
first=1; |
1684 |
#endif |
#endif |
1685 |
} |
} |
1686 |
|
|
|
// swap two MACROBLOCK array |
|
|
void |
|
|
mb_swap(MACROBLOCK ** mb1, |
|
|
MACROBLOCK ** mb2) |
|
|
{ |
|
|
MACROBLOCK *temp = *mb1; |
|
|
|
|
|
*mb1 = *mb2; |
|
|
*mb2 = temp; |
|
|
} |
|
1687 |
|
|
1688 |
|
|
1689 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1690 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1691 |
const XVID_DEC_FRAME * frame, int pp_disable) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
1692 |
{ |
{ |
1693 |
|
|
|
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
|
|
{ |
|
|
/* note: image is stored to tmp */ |
|
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
|
|
image_deblock_rrv(&dec->tmp, dec->edged_width, |
|
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
|
|
8, frame->general); |
|
|
img = &dec->tmp; |
|
|
} |
|
1694 |
|
|
1695 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1696 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1697 |
frame->colorspace, dec->interlacing); |
frame->output.csp, dec->interlacing); |
1698 |
|
|
1699 |
|
if (stats) |
1700 |
|
{ |
1701 |
|
stats->type = coding2type(coding_type); |
1702 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1703 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1704 |
|
} |
1705 |
} |
} |
1706 |
|
|
1707 |
|
|
1708 |
int |
int |
1709 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1710 |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1711 |
{ |
{ |
1712 |
|
|
1713 |
Bitstream bs; |
Bitstream bs; |
1718 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1719 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1720 |
WARPPOINTS gmc_warp; |
WARPPOINTS gmc_warp; |
1721 |
uint32_t vop_type; |
int coding_type; |
1722 |
int success = 0; |
int success, output, seen_something; |
|
int output = 0; |
|
|
int seen_something = 0; |
|
1723 |
|
|
1724 |
start_global_timer(); |
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1725 |
|
return XVID_ERR_VERSION; |
1726 |
|
|
1727 |
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
start_global_timer(); |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
|
1728 |
|
|
1729 |
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1730 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1731 |
dec->frames = 0; |
dec->frames = 0; |
1732 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1733 |
|
|
1734 |
if (frame->length < 0) /* decoder flush */ |
if (frame->length < 0) /* decoder flush */ |
1735 |
{ |
{ |
1736 |
|
int ret; |
1737 |
/* if not decoding "low_delay/packed", and this isn't low_delay and |
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1738 |
we have a reference frame, then outout the reference frame */ |
we have a reference frame, then outout the reference frame */ |
1739 |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1740 |
{ |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1741 |
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
dec->frames = 0; |
1742 |
output = 1; |
ret = 0; |
1743 |
} |
}else{ |
1744 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1745 |
frame->length = 0; |
ret = XVID_ERR_END; |
|
if (stats) |
|
|
{ |
|
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
1746 |
} |
} |
1747 |
|
|
1748 |
emms(); |
emms(); |
|
|
|
1749 |
stop_global_timer(); |
stop_global_timer(); |
1750 |
return XVID_ERR_OK; |
return ret; |
1751 |
} |
} |
1752 |
|
|
1753 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1754 |
|
|
1755 |
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1756 |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1757 |
{ |
{ |
|
if (stats) |
|
|
stats->notify = XVID_DEC_VOP; |
|
|
frame->length = 1; |
|
1758 |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1759 |
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1760 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1761 |
emms(); |
emms(); |
1762 |
return XVID_ERR_OK; |
return 1; /* one byte consumed */ |
1763 |
} |
} |
1764 |
|
|
1765 |
|
success = 0; |
1766 |
|
output = 0; |
1767 |
|
seen_something = 0; |
1768 |
|
|
1769 |
repeat: |
repeat: |
1770 |
|
|
1771 |
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1772 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1773 |
|
|
1774 |
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1775 |
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1776 |
|
|
1777 |
if (vop_type == - 1) |
if (coding_type == -1) /* nothing */ |
1778 |
{ |
{ |
1779 |
if (success) goto done; |
if (success) goto done; |
1780 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1781 |
emms(); |
emms(); |
1782 |
return XVID_ERR_FAIL; |
return BitstreamPos(&bs)/8; |
1783 |
} |
} |
1784 |
|
|
1785 |
if (vop_type == -2 || vop_type == -3) |
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
1786 |
{ |
{ |
1787 |
if (vop_type == -3) |
if (coding_type == -3) |
1788 |
decoder_resize(dec); |
decoder_resize(dec); |
1789 |
|
|
1790 |
if (stats) |
if (stats) |
1791 |
{ |
{ |
1792 |
stats->notify = XVID_DEC_VOL; |
stats->type = XVID_TYPE_VOL; |
1793 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1794 |
if (dec->interlacing) |
/*XXX: if (dec->interlacing) |
1795 |
stats->data.vol.general |= XVID_INTERLACING; |
stats->data.vol.general |= ++INTERLACING; */ |
1796 |
stats->data.vol.width = dec->width; |
stats->data.vol.width = dec->width; |
1797 |
stats->data.vol.height = dec->height; |
stats->data.vol.height = dec->height; |
1798 |
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
stats->data.vol.par = dec->aspect_ratio; |
1799 |
stats->data.vol.par_width = dec->par_width; |
stats->data.vol.par_width = dec->par_width; |
1800 |
stats->data.vol.par_height = dec->par_height; |
stats->data.vol.par_height = dec->par_height; |
|
frame->length = BitstreamPos(&bs) / 8; |
|
1801 |
emms(); |
emms(); |
1802 |
return XVID_ERR_OK; |
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1803 |
} |
} |
1804 |
goto repeat; |
goto repeat; |
1805 |
} |
} |
1806 |
|
|
1807 |
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 */ |
|
|
|
1808 |
|
|
1809 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1810 |
if (dec->packed_mode && vop_type == N_VOP) |
if (dec->packed_mode && coding_type == N_VOP) |
1811 |
{ |
{ |
1812 |
if (dec->low_delay_default && dec->frames > 0) |
if (dec->low_delay_default && dec->frames > 0) |
1813 |
{ |
{ |
1814 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1815 |
output = 1; |
output = 1; |
1816 |
} |
} |
1817 |
/* ignore otherwise */ |
/* ignore otherwise */ |
1818 |
} |
} |
1819 |
else if (vop_type != B_VOP) |
else if (coding_type != B_VOP) |
1820 |
{ |
{ |
1821 |
switch(vop_type) |
switch(coding_type) |
1822 |
{ |
{ |
1823 |
case I_VOP : |
case I_VOP : |
1824 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1832 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1833 |
break; |
break; |
1834 |
case N_VOP : |
case N_VOP : |
1835 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1836 |
|
/* we should not swap(last_mbs,mbs) */ |
1837 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1838 |
break; |
break; |
1839 |
} |
} |
1842 |
{ |
{ |
1843 |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1844 |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1845 |
16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV); |
16, 0); |
1846 |
} |
} |
1847 |
|
|
1848 |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1850 |
{ |
{ |
1851 |
if (dec->low_delay) |
if (dec->low_delay) |
1852 |
{ |
{ |
1853 |
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1854 |
output = 1; |
output = 1; |
1855 |
} |
} |
1856 |
else if (dec->frames > 0) /* is the reference frame valid? */ |
else if (dec->frames > 0) /* is the reference frame valid? */ |
1857 |
{ |
{ |
1858 |
/* output the reference frame */ |
/* output the reference frame */ |
1859 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1860 |
output = 1; |
output = 1; |
1861 |
} |
} |
1862 |
} |
} |
1863 |
|
|
1864 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1865 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1866 |
mb_swap(&dec->mbs, &dec->last_mbs); |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1867 |
dec->last_reduced_resolution = reduced_resolution; |
dec->last_reduced_resolution = reduced_resolution; |
1868 |
|
dec->last_coding_type = coding_type; |
1869 |
|
|
1870 |
dec->frames++; |
dec->frames++; |
1871 |
seen_something = 1; |
seen_something = 1; |
1874 |
|
|
1875 |
if (dec->low_delay) |
if (dec->low_delay) |
1876 |
{ |
{ |
1877 |
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1878 |
dec->low_delay = 1; |
dec->low_delay = 1; |
1879 |
} |
} |
1880 |
|
|
1892 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1893 |
} |
} |
1894 |
|
|
1895 |
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1896 |
output = 1; |
output = 1; |
1897 |
dec->frames++; |
dec->frames++; |
1898 |
} |
} |
1915 |
if (dec->packed_mode && seen_something) |
if (dec->packed_mode && seen_something) |
1916 |
{ |
{ |
1917 |
/* output the recently decoded frame */ |
/* output the recently decoded frame */ |
1918 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
|
output = 1; |
|
1919 |
} |
} |
1920 |
else |
else |
1921 |
{ |
{ |
1925 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1926 |
"bframe decoder lag"); |
"bframe decoder lag"); |
1927 |
|
|
1928 |
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1929 |
} |
if (stats) stats->type = XVID_TYPE_NOTHING; |
|
} |
|
1930 |
|
|
1931 |
frame->length = BitstreamPos(&bs) / 8; |
} |
|
|
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
1932 |
} |
} |
1933 |
|
|
1934 |
emms(); |
emms(); |
|
|
|
1935 |
stop_global_timer(); |
stop_global_timer(); |
1936 |
|
|
1937 |
return XVID_ERR_OK; |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
1938 |
} |
} |