3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder Module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
|
* 2002-2003 Peter Ross <pross@xvid.org> |
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 |
* |
* |
23 |
* $Id: decoder.c,v 1.49.2.4 2003-05-17 13:24:10 suxen_drol Exp $ |
* $Id: decoder.c,v 1.49.2.17 2003-10-16 12:16:00 syskin Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
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" |
60 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
61 |
|
|
62 |
int |
static int |
63 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
64 |
{ |
{ |
65 |
/* free existing */ |
/* free existing */ |
168 |
{ |
{ |
169 |
DECODER *dec; |
DECODER *dec; |
170 |
|
|
171 |
if (XVID_MAJOR(create->version) != 1) /* v1.x.x */ |
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
172 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
173 |
|
|
174 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
232 |
return 0; |
return 0; |
233 |
} |
} |
234 |
|
|
|
|
|
|
|
|
235 |
static const int32_t dquant_table[4] = { |
static const int32_t dquant_table[4] = { |
236 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
237 |
}; |
}; |
238 |
|
|
|
|
|
|
|
|
|
|
|
239 |
/* decode an intra macroblock */ |
/* decode an intra macroblock */ |
240 |
void |
static void |
241 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
242 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
243 |
const uint32_t x_pos, |
const uint32_t x_pos, |
300 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
301 |
start_coeff = 1; |
start_coeff = 1; |
302 |
|
|
303 |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i", dc_dif); |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
304 |
} else { |
} else { |
305 |
start_coeff = 0; |
start_coeff = 0; |
306 |
} |
} |
321 |
|
|
322 |
start_timer(); |
start_timer(); |
323 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
324 |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
325 |
} else { |
} else { |
326 |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
327 |
} |
} |
328 |
stop_iquant_timer(); |
stop_iquant_timer(); |
329 |
|
|
360 |
stop_transfer_timer(); |
stop_transfer_timer(); |
361 |
} |
} |
362 |
|
|
363 |
|
static void |
364 |
|
decoder_mb_decode(DECODER * dec, |
365 |
|
const uint32_t cbp, |
366 |
|
Bitstream * bs, |
367 |
|
uint8_t * pY_Cur, |
368 |
|
uint8_t * pU_Cur, |
369 |
|
uint8_t * pV_Cur, |
370 |
|
const int reduced_resolution, |
371 |
|
const MACROBLOCK * pMB) |
372 |
|
{ |
373 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
374 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
375 |
|
|
376 |
|
int stride = dec->edged_width, next_block = stride * (reduced_resolution ? 16 : 8); |
377 |
|
const int stride2 = stride/2; |
378 |
|
int i; |
379 |
|
const uint32_t iQuant = pMB->quant; |
380 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
381 |
|
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
382 |
|
|
383 |
|
for (i = 0; i < 6; i++) { |
384 |
|
|
385 |
|
if (cbp & (1 << (5 - i))) { /* coded */ |
386 |
|
|
387 |
|
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
388 |
|
|
389 |
|
start_timer(); |
390 |
|
get_inter_block(bs, block, direction); |
391 |
|
stop_coding_timer(); |
392 |
|
|
393 |
|
start_timer(); |
394 |
|
dequant(&data[i * 64], block, iQuant); |
395 |
|
stop_iquant_timer(); |
396 |
|
|
397 |
|
start_timer(); |
398 |
|
idct(&data[i * 64]); |
399 |
|
stop_idct_timer(); |
400 |
|
} |
401 |
|
} |
402 |
|
|
403 |
|
if (dec->interlacing && pMB->field_dct) { |
404 |
|
next_block = stride; |
405 |
|
stride *= 2; |
406 |
|
} |
407 |
|
|
408 |
|
start_timer(); |
409 |
|
if (reduced_resolution) { |
410 |
|
if (cbp & 32) |
411 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
412 |
|
if (cbp & 16) |
413 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
414 |
|
if (cbp & 8) |
415 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
416 |
|
if (cbp & 4) |
417 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
418 |
|
if (cbp & 2) |
419 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
420 |
|
if (cbp & 1) |
421 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
422 |
|
} else { |
423 |
|
if (cbp & 32) |
424 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
425 |
|
if (cbp & 16) |
426 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
427 |
|
if (cbp & 8) |
428 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
429 |
|
if (cbp & 4) |
430 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
431 |
|
if (cbp & 2) |
432 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
433 |
|
if (cbp & 1) |
434 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
435 |
|
} |
436 |
|
stop_transfer_timer(); |
437 |
|
} |
438 |
|
|
439 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
440 |
void |
static void |
441 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
442 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
443 |
const uint32_t x_pos, |
const uint32_t x_pos, |
444 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t fcode, |
|
445 |
const uint32_t cbp, |
const uint32_t cbp, |
446 |
Bitstream * bs, |
Bitstream * bs, |
|
const uint32_t quant, |
|
447 |
const uint32_t rounding, |
const uint32_t rounding, |
448 |
const int reduced_resolution) |
const int reduced_resolution, |
449 |
|
const int ref) |
450 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
451 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
452 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
453 |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
454 |
uint32_t i; |
uint32_t i; |
455 |
uint32_t iQuant = pMB->quant; |
|
456 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
457 |
|
|
458 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
474 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
475 |
} |
} |
476 |
|
|
477 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
start_timer(); |
478 |
|
|
479 |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
|
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
480 |
|
|
481 |
|
uv_dx = mv[0].x; |
482 |
|
uv_dy = mv[0].y; |
483 |
|
if (dec->quarterpel) { |
484 |
|
uv_dx /= 2; |
485 |
|
uv_dy /= 2; |
486 |
|
} |
487 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
488 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
489 |
|
|
|
start_timer(); |
|
490 |
if (reduced_resolution) |
if (reduced_resolution) |
|
{ |
|
491 |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
492 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
493 |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
else if (dec->quarterpel) |
494 |
uv_dx, uv_dy, stride2, rounding); |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
|
|
|
} |
|
|
else |
|
|
{ |
|
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
495 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
496 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
497 |
} |
else |
498 |
else { |
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
|
499 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
|
} |
|
|
|
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
} |
|
|
stop_comp_timer(); |
|
500 |
|
|
501 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
|
int sum; |
|
502 |
|
|
503 |
if(dec->quarterpel) |
if(dec->quarterpel) { |
504 |
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
505 |
else |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
506 |
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
} else { |
507 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
508 |
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
509 |
|
} |
|
if(dec->quarterpel) |
|
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
|
|
else |
|
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
|
510 |
|
|
511 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
512 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
513 |
|
|
514 |
start_timer(); |
if (reduced_resolution) { |
|
if (reduced_resolution) |
|
|
{ |
|
515 |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
516 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
517 |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
525 |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
526 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
527 |
|
|
528 |
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
} else if (dec->quarterpel) { |
|
} |
|
|
else |
|
|
{ |
|
|
if(dec->quarterpel) { |
|
529 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
530 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
531 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
538 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
539 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
540 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
541 |
} |
} else { |
|
else { |
|
542 |
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, |
543 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
544 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
548 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 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, |
549 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
550 |
} |
} |
551 |
|
} |
552 |
|
|
553 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
/* chroma */ |
554 |
|
if (reduced_resolution) { |
555 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
556 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
557 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v , 8 * x_pos, 8 * y_pos, |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
558 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
|
} |
|
|
stop_comp_timer(); |
|
|
} |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
|
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
{ |
|
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i * 64], direction); |
|
|
stop_coding_timer(); |
|
|
|
|
|
start_timer(); |
|
|
if (dec->quant_type == 0) { |
|
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
|
559 |
} else { |
} else { |
560 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
561 |
} |
uv_dx, uv_dy, stride2, rounding); |
562 |
stop_iquant_timer(); |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
563 |
|
uv_dx, uv_dy, stride2, rounding); |
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if (reduced_resolution) |
|
|
{ |
|
|
if (cbp & 32) |
|
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
} |
|
|
else |
|
|
{ |
|
|
if (cbp & 32) |
|
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
|
} |
|
|
stop_transfer_timer(); |
|
564 |
} |
} |
565 |
|
|
566 |
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
stop_comp_timer(); |
|
{ |
|
|
int length = 1 << (fcode+4); |
|
|
|
|
|
/* if (quarterpel) value *= 2; */ |
|
567 |
|
|
568 |
if (value < -length) |
if (cbp) |
569 |
return -length; |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
570 |
else if (value >= length) |
reduced_resolution, pMB); |
|
return length-1; |
|
|
else return value; |
|
571 |
} |
} |
572 |
|
|
|
|
|
573 |
static void |
static void |
574 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
575 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
578 |
const uint32_t fcode, |
const uint32_t fcode, |
579 |
const uint32_t cbp, |
const uint32_t cbp, |
580 |
Bitstream * bs, |
Bitstream * bs, |
581 |
const uint32_t quant, |
const uint32_t rounding) |
|
const uint32_t rounding, |
|
|
const int reduced_resolution) /* no reduced res support */ |
|
582 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
583 |
const uint32_t stride = dec->edged_width; |
const uint32_t stride = dec->edged_width; |
584 |
const uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
585 |
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
uint32_t i; |
|
|
const uint32_t iQuant = pMB->quant; |
|
586 |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
587 |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
588 |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
589 |
|
|
590 |
|
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
591 |
|
|
592 |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
593 |
|
|
594 |
start_timer(); |
start_timer(); |
595 |
|
|
596 |
/* this is where the calculations are done */ |
/* this is where the calculations are done */ |
597 |
|
|
598 |
{ |
gmc_data->predict_16x16(gmc_data, |
599 |
pMB->amv = generate_GMCimageMB(&dec->gmc_data, &dec->refn[0], x_pos, y_pos, |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
600 |
stride, stride2, dec->quarterpel, rounding, &dec->cur); |
stride, stride, x_pos, y_pos, rounding); |
601 |
|
|
602 |
|
gmc_data->predict_8x8(gmc_data, |
603 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
604 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
605 |
|
stride2, stride2, x_pos, y_pos, rounding); |
606 |
|
|
607 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
608 |
|
|
609 |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
610 |
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
|
} |
|
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
|
|
|
|
/* |
|
|
transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
|
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
|
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
|
|
*/ |
|
611 |
|
|
612 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
613 |
|
|
614 |
stop_transfer_timer(); |
stop_transfer_timer(); |
615 |
|
|
616 |
if (!cbp) return; |
if (cbp) |
617 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
|
for (i = 0; i < 6; i++) { |
|
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
|
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
{ |
|
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i * 64], direction); |
|
|
stop_coding_timer(); |
|
|
|
|
|
start_timer(); |
|
|
if (dec->quant_type == 0) { |
|
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
618 |
|
|
|
/* interlace + GMC is this possible ??? */ |
|
|
/* |
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
*/ |
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
|
stop_transfer_timer(); |
|
619 |
} |
} |
620 |
|
|
621 |
|
|
622 |
void |
static void |
623 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
624 |
Bitstream * bs, |
Bitstream * bs, |
625 |
int reduced_resolution, |
int reduced_resolution, |
631 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
632 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
633 |
|
|
634 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
635 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
636 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
637 |
} |
} |
659 |
} |
} |
660 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
661 |
|
|
662 |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
663 |
|
|
664 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
665 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
686 |
|
|
687 |
if (dec->interlacing) { |
if (dec->interlacing) { |
688 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
689 |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
690 |
} |
} |
691 |
|
|
692 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
700 |
} |
} |
701 |
|
|
702 |
|
|
703 |
void |
static void |
704 |
get_motion_vector(DECODER * dec, |
get_motion_vector(DECODER * dec, |
705 |
Bitstream * bs, |
Bitstream * bs, |
706 |
int x, |
int x, |
711 |
const int bound) |
const int bound) |
712 |
{ |
{ |
713 |
|
|
714 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
715 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
716 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
717 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
718 |
|
|
719 |
VECTOR pmv; |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
720 |
VECTOR mv; |
VECTOR mv; |
721 |
|
|
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
|
|
|
722 |
mv.x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
723 |
mv.y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
724 |
|
|
725 |
DPRINTF(XVID_DEBUG_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); |
726 |
|
|
727 |
mv.x += pmv.x; |
mv.x += pmv.x; |
728 |
mv.y += pmv.y; |
mv.y += pmv.y; |
743 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
744 |
} |
} |
745 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
746 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
747 |
void |
static void |
748 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
749 |
Bitstream * bs, |
Bitstream * bs, |
750 |
int rounding, |
int rounding, |
754 |
int intra_dc_threshold, |
int intra_dc_threshold, |
755 |
const WARPPOINTS *const gmc_warp) |
const WARPPOINTS *const gmc_warp) |
756 |
{ |
{ |
|
|
|
757 |
uint32_t x, y; |
uint32_t x, y; |
758 |
uint32_t bound; |
uint32_t bound; |
759 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
760 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
761 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
762 |
|
|
763 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
764 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
765 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
766 |
} |
} |
770 |
dec->width, dec->height); |
dec->width, dec->height); |
771 |
stop_edges_timer(); |
stop_edges_timer(); |
772 |
|
|
773 |
if (gmc_warp) |
if (gmc_warp) { |
|
{ |
|
|
|
|
774 |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
|
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
|
|
{ |
|
|
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
|
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
|
|
dec->sprite_warping_points); |
|
|
} |
|
|
|
|
775 |
generate_GMCparameters( dec->sprite_warping_points, |
generate_GMCparameters( dec->sprite_warping_points, |
776 |
(2 << dec->sprite_warping_accuracy), gmc_warp, |
dec->sprite_warping_accuracy, gmc_warp, |
777 |
dec->width, dec->height, &dec->gmc_data); |
dec->width, dec->height, &dec->new_gmc_data); |
778 |
|
|
779 |
/* image warping is done block-based in decoder_mbgmc(), now */ |
/* image warping is done block-based in decoder_mbgmc(), now */ |
|
/* |
|
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
|
|
mb_width, mb_height, |
|
|
dec->edged_width, dec->edged_width/2, |
|
|
fcode, dec->quarterpel, 0, |
|
|
rounding, dec->mbs, &dec->gmc); |
|
|
*/ |
|
780 |
} |
} |
781 |
|
|
782 |
bound = 0; |
bound = 0; |
790 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
791 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
792 |
|
|
793 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) { |
|
{ |
|
794 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
795 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
796 |
x = bound % mb_width; |
x = bound % mb_width; |
798 |
} |
} |
799 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
800 |
|
|
801 |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
802 |
|
|
803 |
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
804 |
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
uint32_t mcbpc, cbpc, cbpy, cbp; |
805 |
{ |
uint32_t intra, acpred_flag = 0; |
|
uint32_t mcbpc; |
|
|
uint32_t cbpc; |
|
|
uint32_t acpred_flag; |
|
|
uint32_t cbpy; |
|
|
uint32_t cbp; |
|
|
uint32_t intra; |
|
806 |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
807 |
|
|
808 |
cp_mb++; |
cp_mb++; |
810 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
811 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
812 |
|
|
813 |
DPRINTF(XVID_DEBUG_MB, "mode %i", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
814 |
DPRINTF(XVID_DEBUG_MB, "cbpc %i", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
|
acpred_flag = 0; |
|
815 |
|
|
816 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
817 |
|
|
|
if (intra) { |
|
|
acpred_flag = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
818 |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
{ |
|
819 |
mcsel = BitstreamGetBit(bs); |
mcsel = BitstreamGetBit(bs); |
820 |
} |
else if (intra) |
821 |
|
acpred_flag = BitstreamGetBit(bs); |
822 |
|
|
823 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
824 |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
825 |
|
|
826 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
827 |
|
|
828 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
829 |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
830 |
DPRINTF(XVID_DEBUG_MB, "dquant %i", dquant); |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
831 |
quant += dquant; |
quant += dquant; |
832 |
if (quant > 31) { |
if (quant > 31) { |
833 |
quant = 31; |
quant = 31; |
834 |
} else if (quant < 1) { |
} else if (quant < 1) { |
835 |
quant = 1; |
quant = 1; |
836 |
} |
} |
837 |
DPRINTF(XVID_DEBUG_MB, "quant %i", quant); |
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
838 |
} |
} |
839 |
mb->quant = quant; |
mb->quant = quant; |
840 |
|
|
841 |
if (dec->interlacing) { |
if (dec->interlacing) { |
842 |
if (cbp || intra) { |
if ((cbp || intra) && !mcsel) { |
843 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
844 |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
845 |
} |
} |
846 |
|
|
847 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
848 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
849 |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
850 |
|
|
851 |
if (mb->field_pred) { |
if (mb->field_pred) { |
852 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
853 |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
854 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
855 |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
856 |
} |
} |
857 |
} |
} |
858 |
} |
} |
859 |
|
|
860 |
if (mcsel) { |
if (mcsel) { |
861 |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
|
rounding, reduced_resolution); |
|
862 |
continue; |
continue; |
863 |
|
|
864 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
865 |
|
|
866 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
867 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
868 |
fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
|
|
fcode, bound); |
|
869 |
} else { |
} else { |
870 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
|
fcode, bound); |
|
871 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
872 |
} |
} |
873 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
|
|
|
874 |
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); |
875 |
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); |
876 |
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); |
877 |
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); |
878 |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
879 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
880 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
0; |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
0; |
|
881 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
882 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound, reduced_resolution); |
883 |
continue; |
continue; |
884 |
} |
} |
885 |
|
|
886 |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
887 |
rounding, reduced_resolution); |
rounding, reduced_resolution, 0); |
888 |
|
|
889 |
} |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
|
|
{ |
|
890 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
891 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
|
start_timer(); |
|
|
|
|
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
|
|
rounding, reduced_resolution); |
|
|
|
|
|
stop_transfer_timer(); |
|
892 |
|
|
893 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
894 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
895 |
cp_mb = 0; |
cp_mb = 0; |
896 |
} |
} |
897 |
st_mb = x+1; |
st_mb = x+1; |
898 |
} |
} else { /* not coded P_VOP macroblock */ |
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
899 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
900 |
|
|
901 |
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; |
902 |
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; |
|
/* copy macroblock directly from ref to cur */ |
|
|
|
|
|
start_timer(); |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
} |
|
|
else |
|
|
{ |
|
|
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
|
|
dec->edged_width); |
|
|
|
|
|
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), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
|
} |
|
903 |
|
|
904 |
stop_transfer_timer(); |
decoder_mbinter(dec, mb, x, y, 0, bs, |
905 |
|
rounding, reduced_resolution, 0); |
906 |
|
|
907 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
908 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
911 |
st_mb = x+1; |
st_mb = x+1; |
912 |
} |
} |
913 |
} |
} |
914 |
|
|
915 |
if(dec->out_frm && cp_mb > 0) |
if(dec->out_frm && cp_mb > 0) |
916 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
917 |
} |
} |
919 |
|
|
920 |
|
|
921 |
/* decode B-frame motion vector */ |
/* decode B-frame motion vector */ |
922 |
void |
static void |
923 |
get_b_motion_vector(DECODER * dec, |
get_b_motion_vector(Bitstream * bs, |
|
Bitstream * bs, |
|
|
int x, |
|
|
int y, |
|
924 |
VECTOR * mv, |
VECTOR * mv, |
925 |
int fcode, |
int fcode, |
926 |
const VECTOR pmv) |
const VECTOR pmv) |
927 |
{ |
{ |
928 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
929 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
930 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
931 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
|
|
|
|
int mv_x, mv_y; |
|
|
int pmv_x, pmv_y; |
|
|
|
|
|
pmv_x = pmv.x; |
|
|
pmv_y = pmv.y; |
|
932 |
|
|
933 |
mv_x = get_mv(bs, fcode); |
int mv_x = get_mv(bs, fcode); |
934 |
mv_y = get_mv(bs, fcode); |
int mv_y = get_mv(bs, fcode); |
935 |
|
|
936 |
mv_x += pmv_x; |
mv_x += pmv.x; |
937 |
mv_y += pmv_y; |
mv_y += pmv.y; |
938 |
|
|
939 |
if (mv_x < low) { |
if (mv_x < low) |
940 |
mv_x += range; |
mv_x += range; |
941 |
} else if (mv_x > high) { |
else if (mv_x > high) |
942 |
mv_x -= range; |
mv_x -= range; |
|
} |
|
943 |
|
|
944 |
if (mv_y < low) { |
if (mv_y < low) |
945 |
mv_y += range; |
mv_y += range; |
946 |
} else if (mv_y > high) { |
else if (mv_y > high) |
947 |
mv_y -= range; |
mv_y -= range; |
|
} |
|
948 |
|
|
949 |
mv->x = mv_x; |
mv->x = mv_x; |
950 |
mv->y = mv_y; |
mv->y = mv_y; |
951 |
} |
} |
952 |
|
|
953 |
|
/* decode an B-frame direct & interpolate macroblock */ |
954 |
/* decode an B-frame forward & backward inter macroblock */ |
static void |
|
void |
|
|
decoder_bf_mbinter(DECODER * dec, |
|
|
const MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
const uint32_t cbp, |
|
|
Bitstream * bs, |
|
|
const uint32_t quant, |
|
|
const uint8_t ref) |
|
|
{ |
|
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
|
uint32_t stride = dec->edged_width; |
|
|
uint32_t stride2 = stride / 2; |
|
|
uint32_t next_block = stride * 8; |
|
|
uint32_t i; |
|
|
uint32_t iQuant = pMB->quant; |
|
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
int uv_dx, uv_dy; |
|
|
|
|
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
|
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
|
|
|
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
|
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
|
|
|
|
if (dec->quarterpel) |
|
|
{ |
|
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
} |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 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]; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
} |
|
|
else { |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
} |
|
|
|
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, 0); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, 0); |
|
|
stop_comp_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) { |
|
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
|
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
{ |
|
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i * 64], direction); |
|
|
stop_coding_timer(); |
|
|
|
|
|
start_timer(); |
|
|
if (dec->quant_type == 0) { |
|
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
|
stop_transfer_timer(); |
|
|
} |
|
|
|
|
|
/* decode an B-frame direct & inter macroblock */ |
|
|
void |
|
955 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
956 |
IMAGE forward, |
IMAGE forward, |
957 |
IMAGE backward, |
IMAGE backward, |
958 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
959 |
const uint32_t x_pos, |
const uint32_t x_pos, |
960 |
const uint32_t y_pos, |
const uint32_t y_pos, |
961 |
Bitstream * bs) |
Bitstream * bs, |
962 |
|
const int direct) |
963 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
964 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
965 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
966 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * 8; |
|
uint32_t iQuant = pMB->quant; |
|
967 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
968 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
969 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
970 |
const uint32_t cbp = pMB->cbp; |
const uint32_t cbp = pMB->cbp; |
971 |
|
|
973 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
974 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
975 |
|
|
976 |
|
if (!direct) { |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
|
977 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
978 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
979 |
|
|
980 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
981 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
982 |
|
|
983 |
if (dec->quarterpel) |
if (dec->quarterpel) { |
|
{ |
|
984 |
uv_dx /= 2; |
uv_dx /= 2; |
985 |
uv_dy /= 2; |
uv_dy /= 2; |
|
|
|
986 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
987 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
988 |
} |
} |
992 |
|
|
993 |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
994 |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_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]; |
|
|
|
|
|
|
|
|
if(dec->quarterpel) |
|
|
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); |
|
|
else |
|
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
|
995 |
|
|
996 |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
} else { |
997 |
|
if(dec->quarterpel) { |
998 |
if(dec->quarterpel) |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
999 |
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); |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1000 |
else |
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1001 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1002 |
|
} else { |
1003 |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1004 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1005 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1006 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1007 |
} |
} |
1008 |
|
|
1009 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1010 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1011 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1012 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1013 |
|
} |
1014 |
|
|
1015 |
start_timer(); |
start_timer(); |
1016 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1017 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1018 |
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, |
1019 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1020 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1021 |
else { |
} else { |
1022 |
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1023 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1024 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1032 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1033 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1034 |
} |
} |
1035 |
} |
} else { |
|
else { |
|
1036 |
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, |
1037 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1038 |
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, |
1040 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1041 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1042 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1043 |
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, 0); |
|
0); |
|
1044 |
} |
} |
1045 |
|
|
1046 |
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, |
1050 |
|
|
1051 |
|
|
1052 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1053 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1054 |
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, |
1055 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1056 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1057 |
else { |
} else { |
1058 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1059 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1060 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1068 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1069 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1070 |
} |
} |
1071 |
} |
} else { |
|
else { |
|
1072 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1073 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1074 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1075 |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
|
0); |
|
1076 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1077 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
stride, 0); |
|
1078 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1079 |
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, stride, 0); |
|
stride, 0); |
|
1080 |
} |
} |
1081 |
|
|
1082 |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1116 |
|
|
1117 |
stop_comp_timer(); |
stop_comp_timer(); |
1118 |
|
|
1119 |
for (i = 0; i < 6; i++) { |
if (cbp) |
1120 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
|
|
|
|
if (cbp & (1 << (5 - i))) /* coded */ |
|
|
{ |
|
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i * 64], direction); |
|
|
stop_coding_timer(); |
|
|
|
|
|
start_timer(); |
|
|
if (dec->quant_type == 0) { |
|
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
1121 |
} |
} |
1122 |
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
|
stop_transfer_timer(); |
|
|
} |
|
|
|
|
|
|
|
1123 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1124 |
int32_t __inline |
static __inline int32_t |
1125 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1126 |
{ |
{ |
1127 |
if (!BitstreamGetBit(bs)) /* '0' */ |
if (!BitstreamGetBit(bs)) /* '0' */ |
1133 |
} |
} |
1134 |
|
|
1135 |
/* |
/* |
1136 |
* For decode B-frame mb_type |
* decode B-frame mb_type |
1137 |
* bit ret_value |
* bit ret_value |
1138 |
* 1 0 |
* 1 0 |
1139 |
* 01 1 |
* 01 1 |
1140 |
* 001 2 |
* 001 2 |
1141 |
* 0001 3 |
* 0001 3 |
1142 |
*/ |
*/ |
1143 |
int32_t __inline |
static int32_t __inline |
1144 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1145 |
{ |
{ |
1146 |
int32_t mb_type; |
int32_t mb_type; |
1147 |
|
|
1148 |
for (mb_type = 0; mb_type <= 3; mb_type++) { |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1149 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
|
break; |
|
|
} |
|
|
|
|
|
if (mb_type <= 3) |
|
1150 |
return (mb_type); |
return (mb_type); |
1151 |
else |
|
1152 |
return (-1); |
return -1; |
1153 |
} |
} |
1154 |
|
|
1155 |
void |
static void |
1156 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1157 |
Bitstream * bs, |
Bitstream * bs, |
1158 |
int quant, |
int quant, |
1162 |
uint32_t x, y; |
uint32_t x, y; |
1163 |
VECTOR mv; |
VECTOR mv; |
1164 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1165 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1166 |
|
int i; |
1167 |
|
|
1168 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1169 |
FILE *fp; |
FILE *fp; |
1170 |
static char first=0; |
static char first=0; |
1171 |
#define BFRAME_DEBUG if (!first && fp){ \ |
#define BFRAME_DEBUG |
1172 |
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
if (!first && fp) { \ |
1173 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mode,mb->cbp); \ |
1174 |
} |
} |
1175 |
#endif |
#endif |
1176 |
|
|
1197 |
mv = |
mv = |
1198 |
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] = |
1199 |
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; |
1200 |
|
mb->quant = quant; |
1201 |
|
|
1202 |
/* |
/* |
1203 |
* skip if the co-located P_VOP macroblock is not coded |
* skip if the co-located P_VOP macroblock is not coded |
1206 |
*/ |
*/ |
1207 |
|
|
1208 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
|
/* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
|
1209 |
mb->cbp = 0; |
mb->cbp = 0; |
1210 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
1211 |
mb->mb_type = MODE_NOT_CODED; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
BFRAME_DEBUG |
|
|
#endif |
|
|
mb->mb_type = MODE_FORWARD; |
|
|
mb->quant = last_mb->quant; |
|
|
/* |
|
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
|
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
|
*/ |
|
|
|
|
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
|
1212 |
continue; |
continue; |
1213 |
} |
} |
1214 |
|
|
1215 |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1216 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1217 |
|
|
1218 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1219 |
|
|
1220 |
if (!modb2) { /* modb=='00' */ |
if (!modb2) /* modb=='00' */ |
1221 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1222 |
} else { |
else |
1223 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp) { |
|
|
quant += get_dbquant(bs); |
|
1224 |
|
|
1225 |
if (quant > 31) { |
if (mb->mode && mb->cbp) { |
1226 |
|
quant += get_dbquant(bs); |
1227 |
|
if (quant > 31) |
1228 |
quant = 31; |
quant = 31; |
1229 |
} else if (quant < 1) { |
else if (quant < 1) |
1230 |
quant = 1; |
quant = 1; |
1231 |
} |
} |
1232 |
|
mb->quant = quant; |
1233 |
|
|
1234 |
|
if (dec->interlacing) { |
1235 |
|
if (mb->cbp) { |
1236 |
|
mb->field_dct = BitstreamGetBit(bs); |
1237 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1238 |
|
} |
1239 |
|
|
1240 |
|
if (mb->mode) { |
1241 |
|
mb->field_pred = BitstreamGetBit(bs); |
1242 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1243 |
|
|
1244 |
|
if (mb->field_pred) { |
1245 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1246 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1247 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1248 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1249 |
|
} |
1250 |
} |
} |
1251 |
|
} |
1252 |
|
|
1253 |
} else { |
} else { |
1254 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1255 |
mb->cbp = 0; |
mb->cbp = 0; |
1256 |
} |
} |
1257 |
|
|
1258 |
mb->quant = quant; |
switch (mb->mode) { |
|
mb->mode = MODE_INTER4V; |
|
|
/* DEBUG1("Switch bm_type=",mb->mb_type); */ |
|
|
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
BFRAME_DEBUG |
|
|
#endif |
|
|
|
|
|
switch (mb->mb_type) { |
|
1259 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1260 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv); |
1261 |
|
|
1262 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
|
{ |
|
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
|
int i; |
|
|
|
|
1263 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1264 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
|
/ TRD + mv.x); |
|
1265 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1266 |
? ((TRB - TRD) * last_mb->mvs[i].x) |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
|
/ TRD |
|
1267 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1268 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
|
/ TRD + mv.y); |
|
1269 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1270 |
? ((TRB - TRD) * last_mb->mvs[i].y) |
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
|
/ TRD |
|
1271 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
: mb->mvs[i].y - last_mb->mvs[i].y); |
1272 |
} |
} |
1273 |
/* DEBUG("B-frame Direct!\n"); */ |
|
|
} |
|
1274 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1275 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1276 |
break; |
break; |
1277 |
|
|
1278 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1279 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv); |
|
dec->p_fmv); |
|
1280 |
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]; |
1281 |
|
|
1282 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv); |
1283 |
fcode_backward, dec->p_bmv); |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0]; |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
|
|
mb->b_mvs[3] = mb->b_mvs[0]; |
|
1284 |
|
|
1285 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1286 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
|
/* DEBUG("B-frame Bidir!\n"); */ |
|
1287 |
break; |
break; |
1288 |
|
|
1289 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1290 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv); |
|
dec->p_bmv); |
|
1291 |
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]; |
1292 |
|
|
1293 |
mb->mode = MODE_INTER; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
|
|
/* DEBUG("B-frame Backward!\n"); */ |
|
1294 |
break; |
break; |
1295 |
|
|
1296 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1297 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv); |
|
dec->p_fmv); |
|
1298 |
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]; |
1299 |
|
|
1300 |
mb->mode = MODE_INTER; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
|
|
/* DEBUG("B-frame Forward!\n"); */ |
|
1301 |
break; |
break; |
1302 |
|
|
1303 |
default: |
default: |
1304 |
DPRINTF(XVID_DEBUG_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); |
1305 |
} |
} |
1306 |
} /* End of for */ |
} /* End of for */ |
1307 |
} |
} |
1315 |
#endif |
#endif |
1316 |
} |
} |
1317 |
|
|
|
|
|
|
|
|
1318 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1319 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1320 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
1321 |
{ |
{ |
|
|
|
|
|
|
1322 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1323 |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1324 |
frame->output.csp, dec->interlacing); |
frame->output.csp, dec->interlacing); |
1325 |
|
|
1326 |
if (stats) |
if (stats) { |
|
{ |
|
1327 |
stats->type = coding2type(coding_type); |
stats->type = coding2type(coding_type); |
1328 |
stats->data.vop.time_base = (int)dec->time_base; |
stats->data.vop.time_base = (int)dec->time_base; |
1329 |
stats->data.vop.time_increment = 0; //XXX: todo |
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1330 |
} |
} |
1331 |
} |
} |
1332 |
|
|
1347 |
int coding_type; |
int coding_type; |
1348 |
int success, output, seen_something; |
int success, output, seen_something; |
1349 |
|
|
1350 |
if (XVID_MAJOR(frame->version) != 1 || (stats && XVID_MAJOR(stats->version) != 1)) /* v1.x.x */ |
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1351 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
1352 |
|
|
1353 |
start_global_timer(); |
start_global_timer(); |
1357 |
dec->frames = 0; |
dec->frames = 0; |
1358 |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1359 |
|
|
1360 |
if (frame->length < 0) /* decoder flush */ |
if (frame->length < 0) { /* decoder flush */ |
|
{ |
|
1361 |
int ret; |
int ret; |
1362 |
/* 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 |
1363 |
we have a reference frame, then outout the reference frame */ |
we have a reference frame, then outout the reference frame */ |
1396 |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1397 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1398 |
|
|
1399 |
DPRINTF(XVID_DEBUG_HEADER, "coding_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", |
1400 |
coding_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); |
1401 |
|
|
1402 |
if (coding_type == -1) /* nothing */ |
if (coding_type == -1) { /* nothing */ |
|
{ |
|
1403 |
if (success) goto done; |
if (success) goto done; |
1404 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1405 |
emms(); |
emms(); |
1406 |
return BitstreamPos(&bs)/8; |
return BitstreamPos(&bs)/8; |
1407 |
} |
} |
1408 |
|
|
1409 |
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1410 |
{ |
|
1411 |
if (coding_type == -3) |
if (coding_type == -3) |
1412 |
decoder_resize(dec); |
decoder_resize(dec); |
1413 |
|
|
1414 |
if (stats) |
if (stats) { |
|
{ |
|
1415 |
stats->type = XVID_TYPE_VOL; |
stats->type = XVID_TYPE_VOL; |
1416 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1417 |
/*XXX: if (dec->interlacing) |
/*XXX: if (dec->interlacing) |
1429 |
|
|
1430 |
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 */ |
1431 |
|
|
|
|
|
1432 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1433 |
if (dec->packed_mode && coding_type == N_VOP) |
if (dec->packed_mode && coding_type == N_VOP) { |
1434 |
{ |
if (dec->low_delay_default && dec->frames > 0) { |
|
if (dec->low_delay_default && dec->frames > 0) |
|
|
{ |
|
1435 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1436 |
output = 1; |
output = 1; |
1437 |
} |
} |
1438 |
/* ignore otherwise */ |
/* ignore otherwise */ |
1439 |
} |
} else if (coding_type != B_VOP) { |
1440 |
else if (coding_type != B_VOP) |
switch(coding_type) { |
|
{ |
|
|
switch(coding_type) |
|
|
{ |
|
1441 |
case I_VOP : |
case I_VOP : |
1442 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1443 |
break; |
break; |
1450 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1451 |
break; |
break; |
1452 |
case N_VOP : |
case N_VOP : |
1453 |
// XXX: not_coded vops are not used for forward prediction |
/* XXX: not_coded vops are not used for forward prediction */ |
1454 |
// we should not swap(last_mbs,mbs) |
/* we should not swap(last_mbs,mbs) */ |
1455 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1456 |
break; |
break; |
1457 |
} |
} |
1458 |
|
|
1459 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
1460 |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1461 |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1462 |
16, 0); |
16, 0); |
1463 |
} |
} |
1464 |
|
|
1465 |
/* 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 */ |
1466 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1467 |
{ |
if (dec->low_delay) { |
|
if (dec->low_delay) |
|
|
{ |
|
1468 |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1469 |
output = 1; |
output = 1; |
1470 |
} |
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
1471 |
/* output the reference frame */ |
/* output the reference frame */ |
1472 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1473 |
output = 1; |
output = 1; |
1485 |
|
|
1486 |
}else{ /* B_VOP */ |
}else{ /* B_VOP */ |
1487 |
|
|
1488 |
if (dec->low_delay) |
if (dec->low_delay) { |
1489 |
{ |
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream"); |
|
1490 |
dec->low_delay = 1; |
dec->low_delay = 1; |
1491 |
} |
} |
1492 |
|
|
1493 |
if (dec->frames < 2) |
if (dec->frames < 2) { |
|
{ |
|
1494 |
/* attemping to decode a bvop without atleast 2 reference frames */ |
/* attemping to decode a bvop without atleast 2 reference frames */ |
1495 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1496 |
"broken b-frame, mising ref frames"); |
"broken b-frame, mising ref frames"); |
1511 |
BitstreamByteAlign(&bs); |
BitstreamByteAlign(&bs); |
1512 |
|
|
1513 |
/* low_delay_default mode: repeat in packed_mode */ |
/* low_delay_default mode: repeat in packed_mode */ |
1514 |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
|
{ |
|
1515 |
success = 1; |
success = 1; |
1516 |
goto repeat; |
goto repeat; |
1517 |
} |
} |
1520 |
|
|
1521 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* low_delay_default mode: if we've gotten here without outputting anything, |
1522 |
then output the recently decoded frame, or print an error message */ |
then output the recently decoded frame, or print an error message */ |
1523 |
if (dec->low_delay_default && output == 0) |
if (dec->low_delay_default && output == 0) { |
1524 |
{ |
if (dec->packed_mode && seen_something) { |
|
if (dec->packed_mode && seen_something) |
|
|
{ |
|
1525 |
/* output the recently decoded frame */ |
/* output the recently decoded frame */ |
1526 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1527 |
} |
} else { |
|
else |
|
|
{ |
|
1528 |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1529 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1530 |
"warning: nothing to output"); |
"warning: nothing to output"); |
1533 |
|
|
1534 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1535 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
|
|
|
1536 |
} |
} |
1537 |
} |
} |
1538 |
|
|