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 1.49.2.9 2003-06-28 15:48:39 chl Exp $ |
* $Id: decoder.c,v 1.49.2.23 2003-12-10 12:09:34 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" |
44 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_matrix.h" |
45 |
#include "dct/idct.h" |
#include "dct/idct.h" |
46 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
47 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
60 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
61 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
62 |
|
|
63 |
int |
static int |
64 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
65 |
{ |
{ |
66 |
/* free existing */ |
/* free existing */ |
169 |
{ |
{ |
170 |
DECODER *dec; |
DECODER *dec; |
171 |
|
|
172 |
if (XVID_MAJOR(create->version) != 1) /* v1.x.x */ |
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
173 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
174 |
|
|
175 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
176 |
if (dec == NULL) { |
if (dec == NULL) { |
177 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
178 |
} |
} |
179 |
|
|
180 |
memset(dec, 0, sizeof(DECODER)); |
memset(dec, 0, sizeof(DECODER)); |
181 |
|
|
182 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
183 |
|
if (dec->mpeg_quant_matrices == NULL) { |
184 |
|
xvid_free(dec); |
185 |
|
return XVID_ERR_MEMORY; |
186 |
|
} |
187 |
|
|
188 |
create->handle = dec; |
create->handle = dec; |
189 |
|
|
190 |
dec->width = create->width; |
dec->width = create->width; |
204 |
dec->last_mbs = NULL; |
dec->last_mbs = NULL; |
205 |
|
|
206 |
init_timer(); |
init_timer(); |
207 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
208 |
|
|
209 |
/* For B-frame support (used to save reference frame's time */ |
/* For B-frame support (used to save reference frame's time */ |
210 |
dec->frames = 0; |
dec->frames = 0; |
235 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
236 |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
237 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
238 |
|
xvid_free(dec->mpeg_quant_matrices); |
239 |
xvid_free(dec); |
xvid_free(dec); |
240 |
|
|
241 |
write_timer(); |
write_timer(); |
242 |
return 0; |
return 0; |
243 |
} |
} |
244 |
|
|
|
|
|
|
|
|
245 |
static const int32_t dquant_table[4] = { |
static const int32_t dquant_table[4] = { |
246 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
247 |
}; |
}; |
248 |
|
|
|
|
|
|
|
|
|
|
|
249 |
/* decode an intra macroblock */ |
/* decode an intra macroblock */ |
250 |
void |
static void |
251 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
252 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
253 |
const uint32_t x_pos, |
const uint32_t x_pos, |
331 |
|
|
332 |
start_timer(); |
start_timer(); |
333 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
334 |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
335 |
} else { |
} else { |
336 |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
337 |
} |
} |
338 |
stop_iquant_timer(); |
stop_iquant_timer(); |
339 |
|
|
370 |
stop_transfer_timer(); |
stop_transfer_timer(); |
371 |
} |
} |
372 |
|
|
373 |
|
static void |
374 |
|
decoder_mb_decode(DECODER * dec, |
375 |
|
const uint32_t cbp, |
376 |
|
Bitstream * bs, |
377 |
|
uint8_t * pY_Cur, |
378 |
|
uint8_t * pU_Cur, |
379 |
|
uint8_t * pV_Cur, |
380 |
|
const int reduced_resolution, |
381 |
|
const MACROBLOCK * pMB) |
382 |
|
{ |
383 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
384 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
385 |
|
|
386 |
|
int stride = dec->edged_width; |
387 |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
388 |
|
const int stride2 = stride/2; |
389 |
|
int i; |
390 |
|
const uint32_t iQuant = pMB->quant; |
391 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
392 |
|
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
393 |
|
|
394 |
|
for (i = 0; i < 6; i++) { |
395 |
|
|
396 |
|
if (cbp & (1 << (5 - i))) { /* coded */ |
397 |
|
|
398 |
|
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
399 |
|
|
400 |
|
start_timer(); |
401 |
|
get_inter_block(bs, block, direction); |
402 |
|
stop_coding_timer(); |
403 |
|
|
404 |
|
start_timer(); |
405 |
|
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
406 |
|
stop_iquant_timer(); |
407 |
|
|
408 |
|
start_timer(); |
409 |
|
idct(&data[i * 64]); |
410 |
|
stop_idct_timer(); |
411 |
|
} |
412 |
|
} |
413 |
|
|
414 |
|
if (dec->interlacing && pMB->field_dct) { |
415 |
|
next_block = stride; |
416 |
|
stride *= 2; |
417 |
|
} |
418 |
|
|
419 |
|
start_timer(); |
420 |
|
if (reduced_resolution) { |
421 |
|
if (cbp & 32) |
422 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
423 |
|
if (cbp & 16) |
424 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
425 |
|
if (cbp & 8) |
426 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
427 |
|
if (cbp & 4) |
428 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
429 |
|
if (cbp & 2) |
430 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
431 |
|
if (cbp & 1) |
432 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
433 |
|
} else { |
434 |
|
if (cbp & 32) |
435 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
436 |
|
if (cbp & 16) |
437 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
438 |
|
if (cbp & 8) |
439 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
440 |
|
if (cbp & 4) |
441 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
442 |
|
if (cbp & 2) |
443 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
444 |
|
if (cbp & 1) |
445 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
446 |
|
} |
447 |
|
stop_transfer_timer(); |
448 |
|
} |
449 |
|
|
450 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
451 |
void |
static void |
452 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
453 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
454 |
const uint32_t x_pos, |
const uint32_t x_pos, |
455 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t fcode, |
|
456 |
const uint32_t cbp, |
const uint32_t cbp, |
457 |
Bitstream * bs, |
Bitstream * bs, |
|
const uint32_t quant, |
|
458 |
const uint32_t rounding, |
const uint32_t rounding, |
459 |
const int reduced_resolution) |
const int reduced_resolution, |
460 |
|
const int ref) |
461 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
462 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
463 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
464 |
uint32_t i; |
uint32_t i; |
465 |
uint32_t iQuant = pMB->quant; |
|
466 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
467 |
|
|
468 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
484 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
485 |
} |
} |
486 |
|
|
487 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
start_timer(); |
488 |
|
|
489 |
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); |
|
490 |
|
|
491 |
|
uv_dx = mv[0].x; |
492 |
|
uv_dy = mv[0].y; |
493 |
|
if (dec->quarterpel) { |
494 |
|
uv_dx /= 2; |
495 |
|
uv_dy /= 2; |
496 |
|
} |
497 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
498 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
499 |
|
|
|
start_timer(); |
|
500 |
if (reduced_resolution) |
if (reduced_resolution) |
|
{ |
|
501 |
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, |
502 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
503 |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
else if (dec->quarterpel) |
504 |
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, |
|
505 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
506 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
507 |
} |
else |
508 |
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, |
|
509 |
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(); |
|
510 |
|
|
511 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
|
int sum; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
|
|
else |
|
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
|
|
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
512 |
|
|
513 |
if(dec->quarterpel) |
if(dec->quarterpel) { |
514 |
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
515 |
else |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
516 |
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
} else { |
517 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
518 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
519 |
|
} |
520 |
|
|
521 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
522 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
523 |
|
|
524 |
start_timer(); |
if (reduced_resolution) { |
|
if (reduced_resolution) |
|
|
{ |
|
525 |
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, |
526 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
527 |
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, |
535 |
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, |
536 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
537 |
|
|
538 |
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
} else if (dec->quarterpel) { |
|
} |
|
|
else |
|
|
{ |
|
|
if(dec->quarterpel) { |
|
539 |
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, |
540 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
541 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
548 |
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, |
549 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
550 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
551 |
} |
} else { |
|
else { |
|
552 |
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, |
553 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
554 |
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, |
558 |
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, |
559 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
560 |
} |
} |
561 |
|
} |
562 |
|
|
563 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
/* chroma */ |
564 |
|
if (reduced_resolution) { |
565 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
566 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
567 |
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, |
568 |
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); |
|
569 |
} else { |
} else { |
570 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
571 |
} |
uv_dx, uv_dy, stride2, rounding); |
572 |
stop_iquant_timer(); |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
573 |
|
uv_dx, uv_dy, stride2, rounding); |
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
574 |
} |
} |
575 |
|
|
576 |
if (dec->interlacing && pMB->field_dct) { |
stop_comp_timer(); |
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
577 |
|
|
578 |
start_timer(); |
if (cbp) |
579 |
if (reduced_resolution) |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
580 |
{ |
reduced_resolution, pMB); |
|
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(); |
|
581 |
} |
} |
582 |
|
|
|
|
|
583 |
static void |
static void |
584 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
585 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
588 |
const uint32_t fcode, |
const uint32_t fcode, |
589 |
const uint32_t cbp, |
const uint32_t cbp, |
590 |
Bitstream * bs, |
Bitstream * bs, |
591 |
const uint32_t quant, |
const uint32_t rounding) |
|
const uint32_t rounding, |
|
|
const int reduced_resolution) /* no reduced res support */ |
|
592 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
593 |
const uint32_t stride = dec->edged_width; |
const uint32_t stride = dec->edged_width; |
594 |
const uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
595 |
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
uint32_t i; |
|
|
const uint32_t iQuant = pMB->quant; |
|
596 |
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); |
597 |
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); |
598 |
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); |
599 |
|
|
600 |
|
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
601 |
|
|
602 |
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; |
603 |
|
|
604 |
start_timer(); |
start_timer(); |
605 |
|
|
606 |
/* this is where the calculations are done */ |
/* this is where the calculations are done */ |
607 |
|
|
|
{ NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
|
|
|
|
608 |
gmc_data->predict_16x16(gmc_data, |
gmc_data->predict_16x16(gmc_data, |
609 |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
610 |
stride, stride, x_pos, y_pos, rounding); |
stride, stride, x_pos, y_pos, rounding); |
618 |
|
|
619 |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
620 |
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); |
|
|
*/ |
|
621 |
|
|
622 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
623 |
|
|
624 |
stop_transfer_timer(); |
stop_transfer_timer(); |
625 |
|
|
626 |
if (!cbp) return; |
if (cbp) |
627 |
|
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(); |
|
|
} |
|
|
} |
|
628 |
|
|
|
/* 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(); |
|
629 |
} |
} |
630 |
|
|
631 |
|
|
632 |
void |
static void |
633 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
634 |
Bitstream * bs, |
Bitstream * bs, |
635 |
int reduced_resolution, |
int reduced_resolution, |
641 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
642 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
643 |
|
|
644 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
645 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
646 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
647 |
} |
} |
710 |
} |
} |
711 |
|
|
712 |
|
|
713 |
void |
static void |
714 |
get_motion_vector(DECODER * dec, |
get_motion_vector(DECODER * dec, |
715 |
Bitstream * bs, |
Bitstream * bs, |
716 |
int x, |
int x, |
721 |
const int bound) |
const int bound) |
722 |
{ |
{ |
723 |
|
|
724 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
725 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
726 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
727 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
728 |
|
|
729 |
VECTOR pmv; |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
730 |
VECTOR mv; |
VECTOR mv; |
731 |
|
|
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
|
|
|
732 |
mv.x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
733 |
mv.y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
734 |
|
|
753 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
754 |
} |
} |
755 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
756 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
757 |
void |
static void |
758 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
759 |
Bitstream * bs, |
Bitstream * bs, |
760 |
int rounding, |
int rounding, |
764 |
int intra_dc_threshold, |
int intra_dc_threshold, |
765 |
const WARPPOINTS *const gmc_warp) |
const WARPPOINTS *const gmc_warp) |
766 |
{ |
{ |
|
|
|
767 |
uint32_t x, y; |
uint32_t x, y; |
768 |
uint32_t bound; |
uint32_t bound; |
769 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
770 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
771 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
772 |
|
|
773 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
774 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
775 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
776 |
} |
} |
780 |
dec->width, dec->height); |
dec->width, dec->height); |
781 |
stop_edges_timer(); |
stop_edges_timer(); |
782 |
|
|
783 |
if (gmc_warp) |
if (gmc_warp) { |
|
{ |
|
|
|
|
784 |
/* 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 */ |
|
/* { |
|
|
fprintf(stderr,"GMC parameters acc=%d(-> 1/%d), %d pts!!!\n", |
|
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
|
|
dec->sprite_warping_points); |
|
|
}*/ |
|
|
|
|
785 |
generate_GMCparameters( dec->sprite_warping_points, |
generate_GMCparameters( dec->sprite_warping_points, |
786 |
dec->sprite_warping_accuracy, gmc_warp, |
dec->sprite_warping_accuracy, gmc_warp, |
787 |
dec->width, dec->height, &dec->new_gmc_data); |
dec->width, dec->height, &dec->new_gmc_data); |
800 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
801 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
802 |
|
|
803 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) { |
|
{ |
|
804 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
805 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
806 |
x = bound % mb_width; |
x = bound % mb_width; |
810 |
|
|
811 |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
812 |
|
|
813 |
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
814 |
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
uint32_t mcbpc, cbpc, cbpy, cbp; |
815 |
{ |
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; |
|
816 |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
817 |
|
|
818 |
cp_mb++; |
cp_mb++; |
822 |
|
|
823 |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
824 |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
|
acpred_flag = 0; |
|
825 |
|
|
826 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
827 |
|
|
|
if (intra) { |
|
|
acpred_flag = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
828 |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
{ |
|
829 |
mcsel = BitstreamGetBit(bs); |
mcsel = BitstreamGetBit(bs); |
830 |
} |
else if (intra) |
831 |
|
acpred_flag = BitstreamGetBit(bs); |
832 |
|
|
833 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
834 |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
849 |
mb->quant = quant; |
mb->quant = quant; |
850 |
|
|
851 |
if (dec->interlacing) { |
if (dec->interlacing) { |
852 |
if (cbp || intra) { |
if ((cbp || intra) && !mcsel) { |
853 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
854 |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
855 |
} |
} |
868 |
} |
} |
869 |
|
|
870 |
if (mcsel) { |
if (mcsel) { |
871 |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
|
rounding, reduced_resolution); |
|
872 |
continue; |
continue; |
873 |
|
|
874 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
875 |
|
|
876 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
877 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
878 |
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); |
|
879 |
} else { |
} else { |
880 |
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); |
|
881 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
882 |
} |
} |
883 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
|
|
|
884 |
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); |
885 |
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); |
886 |
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); |
887 |
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); |
888 |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
889 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
890 |
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; |
|
891 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
892 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound, reduced_resolution); |
893 |
continue; |
continue; |
894 |
} |
} |
895 |
|
|
896 |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
897 |
rounding, reduced_resolution); |
rounding, reduced_resolution, 0); |
898 |
|
|
899 |
} |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
|
|
{ |
|
900 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
901 |
|
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(); |
|
902 |
|
|
903 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
904 |
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); |
905 |
cp_mb = 0; |
cp_mb = 0; |
906 |
} |
} |
907 |
st_mb = x+1; |
st_mb = x+1; |
908 |
} |
} else { /* not coded P_VOP macroblock */ |
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
909 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
910 |
|
|
911 |
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; |
912 |
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); |
|
|
} |
|
913 |
|
|
914 |
stop_transfer_timer(); |
decoder_mbinter(dec, mb, x, y, 0, bs, |
915 |
|
rounding, reduced_resolution, 0); |
916 |
|
|
917 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
918 |
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); |
921 |
st_mb = x+1; |
st_mb = x+1; |
922 |
} |
} |
923 |
} |
} |
924 |
|
|
925 |
if(dec->out_frm && cp_mb > 0) |
if(dec->out_frm && cp_mb > 0) |
926 |
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); |
927 |
} |
} |
929 |
|
|
930 |
|
|
931 |
/* decode B-frame motion vector */ |
/* decode B-frame motion vector */ |
932 |
void |
static void |
933 |
get_b_motion_vector(DECODER * dec, |
get_b_motion_vector(Bitstream * bs, |
|
Bitstream * bs, |
|
|
int x, |
|
|
int y, |
|
934 |
VECTOR * mv, |
VECTOR * mv, |
935 |
int fcode, |
int fcode, |
936 |
const VECTOR pmv) |
const VECTOR pmv) |
937 |
{ |
{ |
938 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
939 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
940 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
941 |
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; |
|
942 |
|
|
943 |
mv_x = get_mv(bs, fcode); |
int mv_x = get_mv(bs, fcode); |
944 |
mv_y = get_mv(bs, fcode); |
int mv_y = get_mv(bs, fcode); |
945 |
|
|
946 |
mv_x += pmv_x; |
mv_x += pmv.x; |
947 |
mv_y += pmv_y; |
mv_y += pmv.y; |
948 |
|
|
949 |
if (mv_x < low) { |
if (mv_x < low) |
950 |
mv_x += range; |
mv_x += range; |
951 |
} else if (mv_x > high) { |
else if (mv_x > high) |
952 |
mv_x -= range; |
mv_x -= range; |
|
} |
|
953 |
|
|
954 |
if (mv_y < low) { |
if (mv_y < low) |
955 |
mv_y += range; |
mv_y += range; |
956 |
} else if (mv_y > high) { |
else if (mv_y > high) |
957 |
mv_y -= range; |
mv_y -= range; |
|
} |
|
958 |
|
|
959 |
mv->x = mv_x; |
mv->x = mv_x; |
960 |
mv->y = mv_y; |
mv->y = mv_y; |
961 |
} |
} |
962 |
|
|
963 |
|
/* decode an B-frame direct & interpolate macroblock */ |
964 |
/* 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 |
|
965 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
966 |
IMAGE forward, |
IMAGE forward, |
967 |
IMAGE backward, |
IMAGE backward, |
968 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
969 |
const uint32_t x_pos, |
const uint32_t x_pos, |
970 |
const uint32_t y_pos, |
const uint32_t y_pos, |
971 |
Bitstream * bs) |
Bitstream * bs, |
972 |
|
const int direct) |
973 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
974 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
975 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
|
uint32_t iQuant = pMB->quant; |
|
976 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
977 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
978 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
979 |
const uint32_t cbp = pMB->cbp; |
const uint32_t cbp = pMB->cbp; |
980 |
|
|
982 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
983 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
984 |
|
|
985 |
|
if (!direct) { |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
|
986 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
987 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
988 |
|
|
989 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
990 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
991 |
|
|
992 |
if (dec->quarterpel) |
if (dec->quarterpel) { |
|
{ |
|
993 |
uv_dx /= 2; |
uv_dx /= 2; |
994 |
uv_dy /= 2; |
uv_dy /= 2; |
|
|
|
995 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
996 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
997 |
} |
} |
1001 |
|
|
1002 |
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]; |
1003 |
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; |
|
1004 |
|
|
1005 |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
} else { |
1006 |
|
if(dec->quarterpel) { |
1007 |
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); |
1008 |
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); |
1009 |
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); |
1010 |
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); |
1011 |
|
} else { |
1012 |
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; |
1013 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1014 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1015 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1016 |
} |
} |
1017 |
|
|
1018 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1019 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1020 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1021 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1022 |
|
} |
1023 |
|
|
1024 |
start_timer(); |
start_timer(); |
1025 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1026 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1027 |
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, |
1028 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1029 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1030 |
else { |
} else { |
1031 |
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, |
1032 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1033 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1041 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1042 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1043 |
} |
} |
1044 |
} |
} else { |
|
else { |
|
1045 |
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, |
1046 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1047 |
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, |
1049 |
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, |
1050 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1051 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1052 |
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); |
|
1053 |
} |
} |
1054 |
|
|
1055 |
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, |
1059 |
|
|
1060 |
|
|
1061 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1062 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1063 |
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, |
1064 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1065 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1066 |
else { |
} else { |
1067 |
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, |
1068 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1069 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1077 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1078 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1079 |
} |
} |
1080 |
} |
} else { |
|
else { |
|
1081 |
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, |
1082 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1083 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1084 |
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); |
|
1085 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1086 |
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); |
|
1087 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1088 |
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); |
|
1089 |
} |
} |
1090 |
|
|
1091 |
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, |
1125 |
|
|
1126 |
stop_comp_timer(); |
stop_comp_timer(); |
1127 |
|
|
1128 |
for (i = 0; i < 6; i++) { |
if (cbp) |
1129 |
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(); |
|
|
} |
|
|
} |
|
|
|
|
|
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(); |
|
1130 |
} |
} |
1131 |
|
|
|
|
|
1132 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1133 |
int32_t __inline |
static __inline int32_t |
1134 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1135 |
{ |
{ |
1136 |
if (!BitstreamGetBit(bs)) /* '0' */ |
if (!BitstreamGetBit(bs)) /* '0' */ |
1142 |
} |
} |
1143 |
|
|
1144 |
/* |
/* |
1145 |
* For decode B-frame mb_type |
* decode B-frame mb_type |
1146 |
* bit ret_value |
* bit ret_value |
1147 |
* 1 0 |
* 1 0 |
1148 |
* 01 1 |
* 01 1 |
1149 |
* 001 2 |
* 001 2 |
1150 |
* 0001 3 |
* 0001 3 |
1151 |
*/ |
*/ |
1152 |
int32_t __inline |
static int32_t __inline |
1153 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1154 |
{ |
{ |
1155 |
int32_t mb_type; |
int32_t mb_type; |
1156 |
|
|
1157 |
for (mb_type = 0; mb_type <= 3; mb_type++) { |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1158 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
|
break; |
|
|
} |
|
|
|
|
|
if (mb_type <= 3) |
|
1159 |
return (mb_type); |
return (mb_type); |
1160 |
else |
|
1161 |
return (-1); |
return -1; |
1162 |
} |
} |
1163 |
|
|
1164 |
void |
static void |
1165 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1166 |
Bitstream * bs, |
Bitstream * bs, |
1167 |
int quant, |
int quant, |
1171 |
uint32_t x, y; |
uint32_t x, y; |
1172 |
VECTOR mv; |
VECTOR mv; |
1173 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1174 |
#ifdef BFRAMES_DEC_DEBUG |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1175 |
FILE *fp; |
int i; |
|
static char first=0; |
|
|
#define BFRAME_DEBUG if (!first && fp){ \ |
|
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
|
|
} |
|
|
#endif |
|
1176 |
|
|
1177 |
start_timer(); |
start_timer(); |
1178 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1181 |
dec->width, dec->height); |
dec->width, dec->height); |
1182 |
stop_edges_timer(); |
stop_edges_timer(); |
1183 |
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
|
|
} |
|
|
#endif |
|
|
|
|
1184 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1185 |
/* Initialize Pred Motion Vector */ |
/* Initialize Pred Motion Vector */ |
1186 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1187 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1188 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1189 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1190 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1191 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1192 |
|
|
1193 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1194 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1195 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1196 |
|
x = bound % dec->mb_width; |
1197 |
|
y = bound / dec->mb_width; |
1198 |
|
/* reset predicted macroblocks */ |
1199 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1200 |
|
} |
1201 |
|
|
1202 |
mv = |
mv = |
1203 |
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] = |
1204 |
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; |
1205 |
|
mb->quant = quant; |
1206 |
|
|
1207 |
/* |
/* |
1208 |
* skip if the co-located P_VOP macroblock is not coded |
* skip if the co-located P_VOP macroblock is not coded |
1211 |
*/ |
*/ |
1212 |
|
|
1213 |
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); */ |
|
1214 |
mb->cbp = 0; |
mb->cbp = 0; |
1215 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
1216 |
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); |
|
1217 |
continue; |
continue; |
1218 |
} |
} |
1219 |
|
|
1220 |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1221 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1222 |
|
|
1223 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1224 |
|
|
1225 |
if (!modb2) { /* modb=='00' */ |
if (!modb2) /* modb=='00' */ |
1226 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1227 |
} else { |
else |
1228 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp) { |
|
|
quant += get_dbquant(bs); |
|
1229 |
|
|
1230 |
if (quant > 31) { |
if (mb->mode && mb->cbp) { |
1231 |
|
quant += get_dbquant(bs); |
1232 |
|
if (quant > 31) |
1233 |
quant = 31; |
quant = 31; |
1234 |
} else if (quant < 1) { |
else if (quant < 1) |
1235 |
quant = 1; |
quant = 1; |
1236 |
} |
} |
1237 |
|
mb->quant = quant; |
1238 |
|
|
1239 |
|
if (dec->interlacing) { |
1240 |
|
if (mb->cbp) { |
1241 |
|
mb->field_dct = BitstreamGetBit(bs); |
1242 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1243 |
|
} |
1244 |
|
|
1245 |
|
if (mb->mode) { |
1246 |
|
mb->field_pred = BitstreamGetBit(bs); |
1247 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1248 |
|
|
1249 |
|
if (mb->field_pred) { |
1250 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1251 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1252 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1253 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1254 |
|
} |
1255 |
} |
} |
1256 |
|
} |
1257 |
|
|
1258 |
} else { |
} else { |
1259 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1260 |
mb->cbp = 0; |
mb->cbp = 0; |
1261 |
} |
} |
1262 |
|
|
1263 |
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) { |
|
1264 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1265 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv); |
1266 |
|
|
1267 |
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; |
|
|
|
|
1268 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1269 |
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); |
|
1270 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1271 |
? ((TRB - TRD) * last_mb->mvs[i].x) |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
|
/ TRD |
|
1272 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1273 |
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); |
|
1274 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1275 |
? ((TRB - TRD) * last_mb->mvs[i].y) |
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
|
/ TRD |
|
1276 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
: mb->mvs[i].y - last_mb->mvs[i].y); |
1277 |
} |
} |
1278 |
/* DEBUG("B-frame Direct!\n"); */ |
|
|
} |
|
1279 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1280 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1281 |
break; |
break; |
1282 |
|
|
1283 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1284 |
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); |
|
1285 |
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]; |
1286 |
|
|
1287 |
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); |
1288 |
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]; |
|
1289 |
|
|
1290 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1291 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
|
/* DEBUG("B-frame Bidir!\n"); */ |
|
1292 |
break; |
break; |
1293 |
|
|
1294 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1295 |
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); |
|
1296 |
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]; |
1297 |
|
|
1298 |
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"); */ |
|
1299 |
break; |
break; |
1300 |
|
|
1301 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1302 |
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); |
|
1303 |
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]; |
1304 |
|
|
1305 |
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"); */ |
|
1306 |
break; |
break; |
1307 |
|
|
1308 |
default: |
default: |
1309 |
DPRINTF(XVID_DEBUG_ERROR,"Not support B-frame mb_type = %i\n", mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1310 |
} |
} |
1311 |
} /* End of for */ |
} /* End of for */ |
1312 |
} |
} |
|
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
first=1; |
|
|
if (fp) |
|
|
fclose(fp); |
|
|
} |
|
|
#endif |
|
1313 |
} |
} |
1314 |
|
|
|
|
|
|
|
|
1315 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1316 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1317 |
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) |
1318 |
{ |
{ |
1319 |
|
if (frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV)) /* post process */ |
1320 |
|
{ |
1321 |
|
/* note: image is stored to tmp */ |
1322 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1323 |
|
image_deblock(&dec->tmp, dec->edged_width, |
1324 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1325 |
|
frame->general); |
1326 |
|
img = &dec->tmp; |
1327 |
|
} |
1328 |
|
|
1329 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1330 |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1331 |
frame->output.csp, dec->interlacing); |
frame->output.csp, dec->interlacing); |
1332 |
|
|
1333 |
if (stats) |
if (stats) { |
|
{ |
|
1334 |
stats->type = coding2type(coding_type); |
stats->type = coding2type(coding_type); |
1335 |
stats->data.vop.time_base = (int)dec->time_base; |
stats->data.vop.time_base = (int)dec->time_base; |
1336 |
stats->data.vop.time_increment = 0; /* XXX: todo */ |
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1353 |
WARPPOINTS gmc_warp; |
WARPPOINTS gmc_warp; |
1354 |
int coding_type; |
int coding_type; |
1355 |
int success, output, seen_something; |
int success, output, seen_something; |
|
idctFuncPtr idct_save; |
|
1356 |
|
|
1357 |
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 */ |
1358 |
return XVID_ERR_VERSION; |
return XVID_ERR_VERSION; |
1359 |
|
|
1360 |
start_global_timer(); |
start_global_timer(); |
1364 |
dec->frames = 0; |
dec->frames = 0; |
1365 |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1366 |
|
|
1367 |
if (frame->length < 0) /* decoder flush */ |
if (frame->length < 0) { /* decoder flush */ |
|
{ |
|
1368 |
int ret; |
int ret; |
1369 |
/* 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 |
1370 |
we have a reference frame, then outout the reference frame */ |
we have a reference frame, then outout the reference frame */ |
1397 |
success = 0; |
success = 0; |
1398 |
output = 0; |
output = 0; |
1399 |
seen_something = 0; |
seen_something = 0; |
|
idct_save = idct; |
|
1400 |
|
|
1401 |
repeat: |
repeat: |
1402 |
|
|
1406 |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1407 |
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); |
1408 |
|
|
1409 |
if (coding_type == -1) /* nothing */ |
if (coding_type == -1) { /* nothing */ |
|
{ |
|
1410 |
if (success) goto done; |
if (success) goto done; |
1411 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1412 |
emms(); |
emms(); |
1413 |
return BitstreamPos(&bs)/8; |
return BitstreamPos(&bs)/8; |
1414 |
} |
} |
1415 |
|
|
1416 |
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1417 |
{ |
|
1418 |
if (coding_type == -3) |
if (coding_type == -3) |
1419 |
decoder_resize(dec); |
decoder_resize(dec); |
1420 |
|
|
1421 |
if (stats) |
if (stats) { |
|
{ |
|
1422 |
stats->type = XVID_TYPE_VOL; |
stats->type = XVID_TYPE_VOL; |
1423 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1424 |
/*XXX: if (dec->interlacing) |
/*XXX: if (dec->interlacing) |
1436 |
|
|
1437 |
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 */ |
1438 |
|
|
|
#if defined(ARCH_IS_IA32) |
|
|
/* |
|
|
* /!\ Ugly hack /!\ |
|
|
* IA32: Prior to xvid bitstream 10, we were using Walten's mmx/xmm idct |
|
|
*/ |
|
|
if((idct == simple_idct_mmx) && (dec->bs_version < 10)) |
|
|
idct = idct_mmx; |
|
|
#endif |
|
|
|
|
1439 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1440 |
if (dec->packed_mode && coding_type == N_VOP) |
if (dec->packed_mode && coding_type == N_VOP) { |
1441 |
{ |
if (dec->low_delay_default && dec->frames > 0) { |
|
if (dec->low_delay_default && dec->frames > 0) |
|
|
{ |
|
1442 |
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); |
1443 |
output = 1; |
output = 1; |
1444 |
} |
} |
1445 |
/* ignore otherwise */ |
/* ignore otherwise */ |
1446 |
} |
} else if (coding_type != B_VOP) { |
1447 |
else if (coding_type != B_VOP) |
switch(coding_type) { |
|
{ |
|
|
switch(coding_type) |
|
|
{ |
|
1448 |
case I_VOP : |
case I_VOP : |
1449 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1450 |
break; |
break; |
1460 |
/* XXX: not_coded vops are not used for forward prediction */ |
/* XXX: not_coded vops are not used for forward prediction */ |
1461 |
/* we should not swap(last_mbs,mbs) */ |
/* we should not swap(last_mbs,mbs) */ |
1462 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1463 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1464 |
break; |
break; |
1465 |
} |
} |
1466 |
|
|
1467 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
1468 |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1469 |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1470 |
16, 0); |
16, 0); |
1471 |
} |
} |
1472 |
|
|
1473 |
/* 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 */ |
1474 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1475 |
{ |
if (dec->low_delay) { |
|
if (dec->low_delay) |
|
|
{ |
|
1476 |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1477 |
output = 1; |
output = 1; |
1478 |
} |
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
1479 |
/* output the reference frame */ |
/* output the reference frame */ |
1480 |
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); |
1481 |
output = 1; |
output = 1; |
1493 |
|
|
1494 |
}else{ /* B_VOP */ |
}else{ /* B_VOP */ |
1495 |
|
|
1496 |
if (dec->low_delay) |
if (dec->low_delay) { |
|
{ |
|
1497 |
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\n"); |
1498 |
dec->low_delay = 1; |
dec->low_delay = 1; |
1499 |
} |
} |
1500 |
|
|
1501 |
if (dec->frames < 2) |
if (dec->frames < 2) { |
|
{ |
|
1502 |
/* attemping to decode a bvop without atleast 2 reference frames */ |
/* attemping to decode a bvop without atleast 2 reference frames */ |
1503 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1504 |
"broken b-frame, mising ref frames"); |
"broken b-frame, mising ref frames"); |
1505 |
|
stats->type = XVID_TYPE_NOTHING; |
1506 |
}else if (dec->time_pp <= dec->time_bp) { |
}else if (dec->time_pp <= dec->time_bp) { |
1507 |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1508 |
decoded in vfw. */ |
decoded in vfw. */ |
1509 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1510 |
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1511 |
|
stats->type = XVID_TYPE_NOTHING; |
1512 |
}else{ |
}else{ |
1513 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1514 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1515 |
} |
} |
1516 |
|
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
|
1517 |
output = 1; |
output = 1; |
1518 |
dec->frames++; |
dec->frames++; |
1519 |
} |
} |
1521 |
BitstreamByteAlign(&bs); |
BitstreamByteAlign(&bs); |
1522 |
|
|
1523 |
/* low_delay_default mode: repeat in packed_mode */ |
/* low_delay_default mode: repeat in packed_mode */ |
1524 |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
|
{ |
|
1525 |
success = 1; |
success = 1; |
1526 |
goto repeat; |
goto repeat; |
1527 |
} |
} |
1530 |
|
|
1531 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* low_delay_default mode: if we've gotten here without outputting anything, |
1532 |
then output the recently decoded frame, or print an error message */ |
then output the recently decoded frame, or print an error message */ |
1533 |
if (dec->low_delay_default && output == 0) |
if (dec->low_delay_default && output == 0) { |
1534 |
{ |
if (dec->packed_mode && seen_something) { |
|
if (dec->packed_mode && seen_something) |
|
|
{ |
|
1535 |
/* output the recently decoded frame */ |
/* output the recently decoded frame */ |
1536 |
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); |
1537 |
} |
} else { |
|
else |
|
|
{ |
|
1538 |
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); |
1539 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1540 |
"warning: nothing to output"); |
"warning: nothing to output"); |
1543 |
|
|
1544 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1545 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
|
|
|
1546 |
} |
} |
1547 |
} |
} |
1548 |
|
|
1549 |
emms(); |
emms(); |
1550 |
stop_global_timer(); |
stop_global_timer(); |
1551 |
|
|
|
idct = idct_save; |
|
|
|
|
1552 |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
1553 |
} |
} |