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" |
58 |
|
|
59 |
#include "image/image.h" |
#include "image/image.h" |
60 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
61 |
|
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
int |
static int |
65 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
66 |
{ |
{ |
67 |
/* free existing */ |
/* free existing */ |
77 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
78 |
if (dec->mbs) |
if (dec->mbs) |
79 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
80 |
|
if (dec->qscale) |
81 |
|
xvid_free(dec->qscale); |
82 |
|
|
83 |
/* realloc */ |
/* realloc */ |
84 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
163 |
|
|
164 |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
165 |
|
|
166 |
|
/* nothing happens if that fails */ |
167 |
|
dec->qscale = |
168 |
|
xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE); |
169 |
|
|
170 |
|
if (dec->qscale) |
171 |
|
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
172 |
|
|
173 |
return 0; |
return 0; |
174 |
} |
} |
175 |
|
|
186 |
if (dec == NULL) { |
if (dec == NULL) { |
187 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
188 |
} |
} |
189 |
|
|
190 |
memset(dec, 0, sizeof(DECODER)); |
memset(dec, 0, sizeof(DECODER)); |
191 |
|
|
192 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
193 |
|
if (dec->mpeg_quant_matrices == NULL) { |
194 |
|
xvid_free(dec); |
195 |
|
return XVID_ERR_MEMORY; |
196 |
|
} |
197 |
|
|
198 |
create->handle = dec; |
create->handle = dec; |
199 |
|
|
200 |
dec->width = create->width; |
dec->width = create->width; |
209 |
/* image based GMC */ |
/* image based GMC */ |
210 |
image_null(&dec->gmc); |
image_null(&dec->gmc); |
211 |
|
|
|
|
|
212 |
dec->mbs = NULL; |
dec->mbs = NULL; |
213 |
dec->last_mbs = NULL; |
dec->last_mbs = NULL; |
214 |
|
dec->qscale = NULL; |
215 |
|
|
216 |
init_timer(); |
init_timer(); |
217 |
|
init_postproc(&dec->postproc); |
218 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
219 |
|
|
220 |
/* For B-frame support (used to save reference frame's time */ |
/* For B-frame support (used to save reference frame's time */ |
221 |
dec->frames = 0; |
dec->frames = 0; |
222 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
223 |
dec->low_delay = 0; |
dec->low_delay = 0; |
224 |
dec->packed_mode = 0; |
dec->packed_mode = 0; |
225 |
|
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
226 |
|
|
227 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
228 |
|
|
238 |
{ |
{ |
239 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
240 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
241 |
|
xvid_free(dec->qscale); |
242 |
|
|
243 |
/* image based GMC */ |
/* image based GMC */ |
244 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
248 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
249 |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
250 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
251 |
|
xvid_free(dec->mpeg_quant_matrices); |
252 |
xvid_free(dec); |
xvid_free(dec); |
253 |
|
|
254 |
write_timer(); |
write_timer(); |
255 |
return 0; |
return 0; |
256 |
} |
} |
257 |
|
|
|
|
|
|
|
|
258 |
static const int32_t dquant_table[4] = { |
static const int32_t dquant_table[4] = { |
259 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
260 |
}; |
}; |
261 |
|
|
|
|
|
|
|
|
|
|
|
262 |
/* decode an intra macroblock */ |
/* decode an intra macroblock */ |
263 |
void |
static void |
264 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
265 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
266 |
const uint32_t x_pos, |
const uint32_t x_pos, |
339 |
stop_coding_timer(); |
stop_coding_timer(); |
340 |
|
|
341 |
start_timer(); |
start_timer(); |
342 |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
343 |
stop_prediction_timer(); |
stop_prediction_timer(); |
344 |
|
|
345 |
start_timer(); |
start_timer(); |
346 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
347 |
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); |
348 |
} else { |
} else { |
349 |
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); |
350 |
} |
} |
351 |
stop_iquant_timer(); |
stop_iquant_timer(); |
352 |
|
|
383 |
stop_transfer_timer(); |
stop_transfer_timer(); |
384 |
} |
} |
385 |
|
|
386 |
|
static void |
387 |
|
decoder_mb_decode(DECODER * dec, |
388 |
|
const uint32_t cbp, |
389 |
|
Bitstream * bs, |
390 |
|
uint8_t * pY_Cur, |
391 |
|
uint8_t * pU_Cur, |
392 |
|
uint8_t * pV_Cur, |
393 |
|
const int reduced_resolution, |
394 |
|
const MACROBLOCK * pMB) |
395 |
|
{ |
396 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
397 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
398 |
|
|
399 |
|
int stride = dec->edged_width; |
400 |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
401 |
|
const int stride2 = stride/2; |
402 |
|
int i; |
403 |
|
const uint32_t iQuant = pMB->quant; |
404 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
405 |
|
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
406 |
|
|
407 |
|
for (i = 0; i < 6; i++) { |
408 |
|
|
409 |
|
if (cbp & (1 << (5 - i))) { /* coded */ |
410 |
|
|
411 |
|
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
412 |
|
|
413 |
|
start_timer(); |
414 |
|
get_inter_block(bs, block, direction); |
415 |
|
stop_coding_timer(); |
416 |
|
|
417 |
|
start_timer(); |
418 |
|
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
419 |
|
stop_iquant_timer(); |
420 |
|
|
421 |
|
start_timer(); |
422 |
|
idct(&data[i * 64]); |
423 |
|
stop_idct_timer(); |
424 |
|
} |
425 |
|
} |
426 |
|
|
427 |
|
if (dec->interlacing && pMB->field_dct) { |
428 |
|
next_block = stride; |
429 |
|
stride *= 2; |
430 |
|
} |
431 |
|
|
432 |
|
start_timer(); |
433 |
|
if (reduced_resolution) { |
434 |
|
if (cbp & 32) |
435 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
436 |
|
if (cbp & 16) |
437 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
438 |
|
if (cbp & 8) |
439 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
440 |
|
if (cbp & 4) |
441 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
442 |
|
if (cbp & 2) |
443 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
444 |
|
if (cbp & 1) |
445 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
446 |
|
} else { |
447 |
|
if (cbp & 32) |
448 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
449 |
|
if (cbp & 16) |
450 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
451 |
|
if (cbp & 8) |
452 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
453 |
|
if (cbp & 4) |
454 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
455 |
|
if (cbp & 2) |
456 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
457 |
|
if (cbp & 1) |
458 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
459 |
|
} |
460 |
|
stop_transfer_timer(); |
461 |
|
} |
462 |
|
|
463 |
|
static void __inline |
464 |
|
validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec) |
465 |
|
{ |
466 |
|
/* clip a vector to valid range |
467 |
|
prevents crashes if bitstream is broken |
468 |
|
*/ |
469 |
|
int shift = 5 + dec->quarterpel; |
470 |
|
int xborder_high = (int)(dec->mb_width - x_pos) << shift; |
471 |
|
int xborder_low = (-(int)x_pos-1) << shift; |
472 |
|
int yborder_high = (int)(dec->mb_height - y_pos) << shift; |
473 |
|
int yborder_low = (-(int)y_pos-1) << shift; |
474 |
|
|
475 |
|
#define CHECK_MV(mv) \ |
476 |
|
do { \ |
477 |
|
if ((mv).x > xborder_high) { \ |
478 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \ |
479 |
|
(mv).x = xborder_high; \ |
480 |
|
} else if ((mv).x < xborder_low) { \ |
481 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \ |
482 |
|
(mv).x = xborder_low; \ |
483 |
|
} \ |
484 |
|
if ((mv).y > yborder_high) { \ |
485 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \ |
486 |
|
(mv).y = yborder_high; \ |
487 |
|
} else if ((mv).y < yborder_low) { \ |
488 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \ |
489 |
|
(mv).y = yborder_low; \ |
490 |
|
} \ |
491 |
|
} while (0) |
492 |
|
|
493 |
|
CHECK_MV(mv[0]); |
494 |
|
CHECK_MV(mv[1]); |
495 |
|
CHECK_MV(mv[2]); |
496 |
|
CHECK_MV(mv[3]); |
497 |
|
} |
498 |
|
|
499 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
500 |
void |
static void |
501 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
502 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
503 |
const uint32_t x_pos, |
const uint32_t x_pos, |
504 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t fcode, |
|
505 |
const uint32_t cbp, |
const uint32_t cbp, |
506 |
Bitstream * bs, |
Bitstream * bs, |
|
const uint32_t quant, |
|
507 |
const uint32_t rounding, |
const uint32_t rounding, |
508 |
const int reduced_resolution) |
const int reduced_resolution, |
509 |
|
const int ref) |
510 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
511 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
512 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
513 |
uint32_t i; |
uint32_t i; |
514 |
uint32_t iQuant = pMB->quant; |
|
515 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
516 |
|
|
517 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
533 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
534 |
} |
} |
535 |
|
|
536 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
validate_vector(mv, x_pos, y_pos, dec); |
537 |
|
|
538 |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
start_timer(); |
539 |
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
540 |
|
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
541 |
|
|
542 |
|
uv_dx = mv[0].x; |
543 |
|
uv_dy = mv[0].y; |
544 |
|
if (dec->quarterpel) { |
545 |
|
uv_dx /= 2; |
546 |
|
uv_dy /= 2; |
547 |
|
} |
548 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
549 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
550 |
|
|
|
start_timer(); |
|
551 |
if (reduced_resolution) |
if (reduced_resolution) |
|
{ |
|
552 |
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, |
553 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
554 |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
else if (dec->quarterpel) |
555 |
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, |
|
556 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
557 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
558 |
} |
else |
559 |
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, |
|
560 |
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(); |
|
561 |
|
|
562 |
} 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; |
|
563 |
|
|
564 |
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
if(dec->quarterpel) { |
565 |
|
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
566 |
if(dec->quarterpel) |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
567 |
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
} else { |
568 |
else |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
569 |
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
570 |
|
} |
571 |
|
|
572 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
573 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
574 |
|
|
575 |
start_timer(); |
if (reduced_resolution) { |
|
if (reduced_resolution) |
|
|
{ |
|
576 |
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, |
577 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
578 |
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, |
586 |
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, |
587 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
588 |
|
|
589 |
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
} else if (dec->quarterpel) { |
|
} |
|
|
else |
|
|
{ |
|
|
if(dec->quarterpel) { |
|
590 |
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, |
591 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
592 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
599 |
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, |
600 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
601 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
602 |
} |
} else { |
|
else { |
|
603 |
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, |
604 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
605 |
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, |
609 |
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, |
610 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
611 |
} |
} |
612 |
|
} |
613 |
|
|
614 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
/* chroma */ |
615 |
|
if (reduced_resolution) { |
616 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
617 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
618 |
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, |
619 |
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); |
|
620 |
} else { |
} else { |
621 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
622 |
} |
uv_dx, uv_dy, stride2, rounding); |
623 |
stop_iquant_timer(); |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
624 |
|
uv_dx, uv_dy, stride2, rounding); |
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
625 |
} |
} |
626 |
|
|
627 |
if (dec->interlacing && pMB->field_dct) { |
stop_comp_timer(); |
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
628 |
|
|
629 |
start_timer(); |
if (cbp) |
630 |
if (reduced_resolution) |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
631 |
{ |
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(); |
|
632 |
} |
} |
633 |
|
|
|
|
|
634 |
static void |
static void |
635 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
636 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
639 |
const uint32_t fcode, |
const uint32_t fcode, |
640 |
const uint32_t cbp, |
const uint32_t cbp, |
641 |
Bitstream * bs, |
Bitstream * bs, |
642 |
const uint32_t quant, |
const uint32_t rounding) |
|
const uint32_t rounding, |
|
|
const int reduced_resolution) /* no reduced res support */ |
|
643 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
644 |
const uint32_t stride = dec->edged_width; |
const uint32_t stride = dec->edged_width; |
645 |
const uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
646 |
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
uint32_t i; |
|
|
const uint32_t iQuant = pMB->quant; |
|
647 |
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); |
648 |
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); |
649 |
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); |
650 |
|
|
651 |
|
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
652 |
|
|
653 |
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; |
654 |
|
|
655 |
start_timer(); |
start_timer(); |
656 |
|
|
657 |
/* this is where the calculations are done */ |
/* this is where the calculations are done */ |
658 |
|
|
|
{ NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
|
|
|
|
659 |
gmc_data->predict_16x16(gmc_data, |
gmc_data->predict_16x16(gmc_data, |
660 |
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, |
661 |
stride, stride, x_pos, y_pos, rounding); |
stride, stride, x_pos, y_pos, rounding); |
669 |
|
|
670 |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
671 |
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); |
|
|
*/ |
|
672 |
|
|
673 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
674 |
|
|
675 |
stop_transfer_timer(); |
stop_transfer_timer(); |
676 |
|
|
677 |
if (!cbp) return; |
if (cbp) |
678 |
|
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(); |
|
679 |
|
|
|
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); |
|
680 |
} |
} |
|
stop_iquant_timer(); |
|
681 |
|
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
682 |
|
|
683 |
/* interlace + GMC is this possible ??? */ |
static void |
|
/* |
|
|
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(); |
|
|
} |
|
|
|
|
|
|
|
|
void |
|
684 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
685 |
Bitstream * bs, |
Bitstream * bs, |
686 |
int reduced_resolution, |
int reduced_resolution, |
692 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
693 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
694 |
|
|
695 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
696 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
697 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
698 |
} |
} |
761 |
} |
} |
762 |
|
|
763 |
|
|
764 |
void |
static void |
765 |
get_motion_vector(DECODER * dec, |
get_motion_vector(DECODER * dec, |
766 |
Bitstream * bs, |
Bitstream * bs, |
767 |
int x, |
int x, |
772 |
const int bound) |
const int bound) |
773 |
{ |
{ |
774 |
|
|
775 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
776 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
777 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
778 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
779 |
|
|
780 |
VECTOR pmv; |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
781 |
VECTOR mv; |
VECTOR mv; |
782 |
|
|
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
|
|
|
783 |
mv.x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
784 |
mv.y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
785 |
|
|
804 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
805 |
} |
} |
806 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
807 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
808 |
void |
static void |
809 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
810 |
Bitstream * bs, |
Bitstream * bs, |
811 |
int rounding, |
int rounding, |
815 |
int intra_dc_threshold, |
int intra_dc_threshold, |
816 |
const WARPPOINTS *const gmc_warp) |
const WARPPOINTS *const gmc_warp) |
817 |
{ |
{ |
|
|
|
818 |
uint32_t x, y; |
uint32_t x, y; |
819 |
uint32_t bound; |
uint32_t bound; |
820 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
821 |
uint32_t mb_width = dec->mb_width; |
uint32_t mb_width = dec->mb_width; |
822 |
uint32_t mb_height = dec->mb_height; |
uint32_t mb_height = dec->mb_height; |
823 |
|
|
824 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
825 |
mb_width = (dec->width + 31) / 32; |
mb_width = (dec->width + 31) / 32; |
826 |
mb_height = (dec->height + 31) / 32; |
mb_height = (dec->height + 31) / 32; |
827 |
} |
} |
828 |
|
|
829 |
start_timer(); |
start_timer(); |
830 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
831 |
dec->width, dec->height); |
dec->width, dec->height, dec->bs_version); |
832 |
stop_edges_timer(); |
stop_edges_timer(); |
833 |
|
|
834 |
if (gmc_warp) |
if (gmc_warp) { |
|
{ |
|
|
|
|
835 |
/* 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); |
|
|
}*/ |
|
|
|
|
836 |
generate_GMCparameters( dec->sprite_warping_points, |
generate_GMCparameters( dec->sprite_warping_points, |
837 |
dec->sprite_warping_accuracy, gmc_warp, |
dec->sprite_warping_accuracy, gmc_warp, |
838 |
dec->width, dec->height, &dec->new_gmc_data); |
dec->width, dec->height, &dec->new_gmc_data); |
851 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
852 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
853 |
|
|
854 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) { |
|
{ |
|
855 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
856 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
857 |
x = bound % mb_width; |
x = bound % mb_width; |
861 |
|
|
862 |
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)); |
863 |
|
|
864 |
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
865 |
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
uint32_t mcbpc, cbpc, cbpy, cbp; |
866 |
{ |
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; |
|
867 |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
868 |
|
|
869 |
cp_mb++; |
cp_mb++; |
873 |
|
|
874 |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
875 |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
|
acpred_flag = 0; |
|
876 |
|
|
877 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
878 |
|
|
879 |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
880 |
mcsel = BitstreamGetBit(bs); |
mcsel = BitstreamGetBit(bs); |
881 |
|
else if (intra) |
|
if (intra) |
|
882 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
883 |
|
|
884 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
905 |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
906 |
} |
} |
907 |
|
|
908 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) { |
909 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
910 |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
911 |
|
|
919 |
} |
} |
920 |
|
|
921 |
if (mcsel) { |
if (mcsel) { |
922 |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
|
rounding, reduced_resolution); |
|
923 |
continue; |
continue; |
924 |
|
|
925 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
926 |
|
|
927 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
928 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
929 |
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); |
|
930 |
} else { |
} else { |
931 |
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); |
|
932 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
933 |
} |
} |
934 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
|
|
|
935 |
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); |
936 |
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); |
937 |
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); |
938 |
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); |
939 |
} else /* MODE_INTRA, MODE_INTRA_Q */ |
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
940 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
941 |
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; |
|
942 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
943 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound, reduced_resolution); |
944 |
continue; |
continue; |
945 |
} |
} |
|
|
|
|
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
|
|
rounding, reduced_resolution); |
|
|
|
|
|
} |
|
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
|
|
{ |
|
|
mb->mode = MODE_NOT_CODED_GMC; |
|
|
|
|
|
start_timer(); |
|
|
|
|
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
|
|
rounding, reduced_resolution); |
|
|
|
|
|
stop_transfer_timer(); |
|
|
|
|
|
if(dec->out_frm && cp_mb > 0) { |
|
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
|
cp_mb = 0; |
|
|
} |
|
|
st_mb = x+1; |
|
|
} |
|
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
|
mb->mode = MODE_NOT_CODED; |
|
|
|
|
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 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); |
|
|
} |
|
|
|
|
|
stop_transfer_timer(); |
|
|
|
|
|
if(dec->out_frm && cp_mb > 0) { |
|
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
|
cp_mb = 0; |
|
|
} |
|
|
st_mb = x+1; |
|
|
} |
|
|
} |
|
|
if(dec->out_frm && cp_mb > 0) |
|
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
|
} |
|
|
} |
|
|
|
|
|
|
|
|
/* decode B-frame motion vector */ |
|
|
void |
|
|
get_b_motion_vector(DECODER * dec, |
|
|
Bitstream * bs, |
|
|
int x, |
|
|
int y, |
|
|
VECTOR * mv, |
|
|
int fcode, |
|
|
const VECTOR pmv) |
|
|
{ |
|
|
int scale_fac = 1 << (fcode - 1); |
|
|
int high = (32 * scale_fac) - 1; |
|
|
int low = ((-32) * scale_fac); |
|
|
int range = (64 * scale_fac); |
|
|
|
|
|
int mv_x, mv_y; |
|
|
int pmv_x, pmv_y; |
|
|
|
|
|
pmv_x = pmv.x; |
|
|
pmv_y = pmv.y; |
|
|
|
|
|
mv_x = get_mv(bs, fcode); |
|
|
mv_y = get_mv(bs, fcode); |
|
|
|
|
|
mv_x += pmv_x; |
|
|
mv_y += pmv_y; |
|
|
|
|
|
if (mv_x < low) { |
|
|
mv_x += range; |
|
|
} else if (mv_x > high) { |
|
|
mv_x -= range; |
|
|
} |
|
|
|
|
|
if (mv_y < low) { |
|
|
mv_y += range; |
|
|
} else if (mv_y > high) { |
|
|
mv_y -= range; |
|
|
} |
|
|
|
|
|
mv->x = mv_x; |
|
|
mv->y = mv_y; |
|
|
} |
|
|
|
|
|
|
|
|
/* decode an B-frame forward & backward inter macroblock */ |
|
|
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; |
|
946 |
|
|
947 |
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
948 |
|
rounding, reduced_resolution, 0); |
949 |
|
|
950 |
if(dec->quarterpel) |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
951 |
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
mb->mode = MODE_NOT_CODED_GMC; |
952 |
else |
mb->quant = quant; |
953 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
954 |
|
|
955 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
if(dec->out_frm && cp_mb > 0) { |
956 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
957 |
|
cp_mb = 0; |
958 |
} |
} |
959 |
|
st_mb = x+1; |
960 |
|
} else { /* not coded P_VOP macroblock */ |
961 |
|
mb->mode = MODE_NOT_CODED; |
962 |
|
mb->quant = quant; |
963 |
|
|
964 |
start_timer(); |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
965 |
if(dec->quarterpel) { |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
966 |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
967 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
968 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
rounding, reduced_resolution, 0); |
969 |
|
|
970 |
|
if(dec->out_frm && cp_mb > 0) { |
971 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
972 |
|
cp_mb = 0; |
973 |
|
} |
974 |
|
st_mb = x+1; |
975 |
} |
} |
|
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); |
|
976 |
} |
} |
977 |
|
|
978 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
if(dec->out_frm && cp_mb > 0) |
979 |
uv_dx, uv_dy, stride2, 0); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
980 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
} |
981 |
uv_dx, uv_dy, stride2, 0); |
} |
|
stop_comp_timer(); |
|
982 |
|
|
|
for (i = 0; i < 6; i++) { |
|
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
983 |
|
|
984 |
if (cbp & (1 << (5 - i))) /* coded */ |
/* decode B-frame motion vector */ |
985 |
{ |
static void |
986 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
get_b_motion_vector(Bitstream * bs, |
987 |
|
VECTOR * mv, |
988 |
|
int fcode, |
989 |
|
const VECTOR pmv, |
990 |
|
const DECODER * const dec, |
991 |
|
const int x, const int y) |
992 |
|
{ |
993 |
|
const int scale_fac = 1 << (fcode - 1); |
994 |
|
const int high = (32 * scale_fac) - 1; |
995 |
|
const int low = ((-32) * scale_fac); |
996 |
|
const int range = (64 * scale_fac); |
997 |
|
|
998 |
start_timer(); |
int mv_x = get_mv(bs, fcode); |
999 |
get_inter_block(bs, &block[i * 64], direction); |
int mv_y = get_mv(bs, fcode); |
|
stop_coding_timer(); |
|
1000 |
|
|
1001 |
start_timer(); |
mv_x += pmv.x; |
1002 |
if (dec->quant_type == 0) { |
mv_y += pmv.y; |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
1003 |
|
|
1004 |
start_timer(); |
if (mv_x < low) |
1005 |
idct(&data[i * 64]); |
mv_x += range; |
1006 |
stop_idct_timer(); |
else if (mv_x > high) |
1007 |
} |
mv_x -= range; |
|
} |
|
1008 |
|
|
1009 |
if (dec->interlacing && pMB->field_dct) { |
if (mv_y < low) |
1010 |
next_block = stride; |
mv_y += range; |
1011 |
stride *= 2; |
else if (mv_y > high) |
1012 |
} |
mv_y -= range; |
1013 |
|
|
1014 |
start_timer(); |
mv->x = mv_x; |
1015 |
if (cbp & 32) |
mv->y = mv_y; |
|
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(); |
|
1016 |
} |
} |
1017 |
|
|
1018 |
/* decode an B-frame direct & inter macroblock */ |
/* decode an B-frame direct & interpolate macroblock */ |
1019 |
void |
static void |
1020 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1021 |
IMAGE forward, |
IMAGE forward, |
1022 |
IMAGE backward, |
IMAGE backward, |
1023 |
const MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
1024 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1025 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1026 |
Bitstream * bs) |
Bitstream * bs, |
1027 |
|
const int direct) |
1028 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
1029 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
1030 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
|
uint32_t iQuant = pMB->quant; |
|
1031 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
1032 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
1033 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1034 |
const uint32_t cbp = pMB->cbp; |
const uint32_t cbp = pMB->cbp; |
1035 |
|
|
1037 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1038 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1039 |
|
|
1040 |
|
validate_vector(pMB->mvs, x_pos, y_pos, dec); |
1041 |
|
validate_vector(pMB->b_mvs, x_pos, y_pos, dec); |
1042 |
|
|
1043 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if (!direct) { |
1044 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1045 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1046 |
|
|
1047 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1048 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1049 |
|
|
1050 |
if (dec->quarterpel) |
if (dec->quarterpel) { |
|
{ |
|
1051 |
uv_dx /= 2; |
uv_dx /= 2; |
1052 |
uv_dy /= 2; |
uv_dy /= 2; |
|
|
|
1053 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1054 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1055 |
} |
} |
1059 |
|
|
1060 |
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]; |
1061 |
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]; |
|
1062 |
|
|
1063 |
if(dec->quarterpel) |
} else { |
1064 |
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
if(dec->quarterpel) { |
1065 |
else |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1066 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1067 |
|
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); |
1068 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
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); |
1069 |
|
} else { |
1070 |
|
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1071 |
if(dec->quarterpel) |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1072 |
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); |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1073 |
else |
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
|
|
|
|
|
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
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); |
|
|
else |
|
|
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
|
|
|
|
|
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
1074 |
} |
} |
1075 |
|
|
1076 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1077 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1078 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1079 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1080 |
|
} |
1081 |
|
|
1082 |
start_timer(); |
start_timer(); |
1083 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1084 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1085 |
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, |
1086 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1087 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1088 |
else { |
} else { |
1089 |
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, |
1090 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1091 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1099 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1100 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1101 |
} |
} |
1102 |
} |
} else { |
|
else { |
|
1103 |
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, |
1104 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1105 |
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, |
1106 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1107 |
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, |
1108 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1109 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, |
1110 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
0); |
|
1111 |
} |
} |
1112 |
|
|
1113 |
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, |
1117 |
|
|
1118 |
|
|
1119 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1120 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1121 |
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, |
1122 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1123 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1124 |
else { |
} else { |
1125 |
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, |
1126 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1127 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1135 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1136 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1137 |
} |
} |
1138 |
} |
} else { |
|
else { |
|
1139 |
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, |
1140 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1141 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1142 |
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); |
|
1143 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1144 |
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); |
|
1145 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1146 |
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); |
|
1147 |
} |
} |
1148 |
|
|
1149 |
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, |
1154 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1155 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1156 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1157 |
stride, 1, 8); |
stride, 0, 8); |
1158 |
|
|
1159 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1160 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1161 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1162 |
stride, 1, 8); |
stride, 0, 8); |
1163 |
|
|
1164 |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1165 |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1166 |
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1167 |
stride, 1, 8); |
stride, 0, 8); |
1168 |
|
|
1169 |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1170 |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1171 |
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1172 |
stride, 1, 8); |
stride, 0, 8); |
1173 |
|
|
1174 |
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1175 |
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1176 |
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1177 |
stride2, 1, 8); |
stride2, 0, 8); |
1178 |
|
|
1179 |
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1180 |
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1181 |
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1182 |
stride2, 1, 8); |
stride2, 0, 8); |
1183 |
|
|
1184 |
stop_comp_timer(); |
stop_comp_timer(); |
1185 |
|
|
1186 |
for (i = 0; i < 6; i++) { |
if (cbp) |
1187 |
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(); |
|
1188 |
} |
} |
1189 |
|
|
|
|
|
1190 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1191 |
int32_t __inline |
static __inline int32_t |
1192 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1193 |
{ |
{ |
1194 |
if (!BitstreamGetBit(bs)) /* '0' */ |
if (!BitstreamGetBit(bs)) /* '0' */ |
1200 |
} |
} |
1201 |
|
|
1202 |
/* |
/* |
1203 |
* For decode B-frame mb_type |
* decode B-frame mb_type |
1204 |
* bit ret_value |
* bit ret_value |
1205 |
* 1 0 |
* 1 0 |
1206 |
* 01 1 |
* 01 1 |
1207 |
* 001 2 |
* 001 2 |
1208 |
* 0001 3 |
* 0001 3 |
1209 |
*/ |
*/ |
1210 |
int32_t __inline |
static int32_t __inline |
1211 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1212 |
{ |
{ |
1213 |
int32_t mb_type; |
int32_t mb_type; |
1214 |
|
|
1215 |
for (mb_type = 0; mb_type <= 3; mb_type++) { |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1216 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
|
break; |
|
|
} |
|
|
|
|
|
if (mb_type <= 3) |
|
1217 |
return (mb_type); |
return (mb_type); |
1218 |
else |
|
1219 |
return (-1); |
return -1; |
1220 |
} |
} |
1221 |
|
|
1222 |
void |
static void |
1223 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1224 |
Bitstream * bs, |
Bitstream * bs, |
1225 |
int quant, |
int quant, |
1229 |
uint32_t x, y; |
uint32_t x, y; |
1230 |
VECTOR mv; |
VECTOR mv; |
1231 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1232 |
#ifdef BFRAMES_DEC_DEBUG |
int i; |
|
FILE *fp; |
|
|
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 |
|
1233 |
|
|
1234 |
start_timer(); |
start_timer(); |
1235 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1236 |
dec->width, dec->height); |
dec->width, dec->height, dec->bs_version); |
1237 |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1238 |
dec->width, dec->height); |
dec->width, dec->height, dec->bs_version); |
1239 |
stop_edges_timer(); |
stop_edges_timer(); |
1240 |
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
|
|
} |
|
|
#endif |
|
|
|
|
1241 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1242 |
/* Initialize Pred Motion Vector */ |
/* Initialize Pred Motion Vector */ |
1243 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1244 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1245 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1246 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1247 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1248 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1249 |
|
|
1250 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1251 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1252 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1253 |
|
x = bound % dec->mb_width; |
1254 |
|
y = bound / dec->mb_width; |
1255 |
|
/* reset predicted macroblocks */ |
1256 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1257 |
|
} |
1258 |
|
|
1259 |
mv = |
mv = |
1260 |
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] = |
1261 |
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; |
1262 |
|
mb->quant = quant; |
1263 |
|
|
1264 |
/* |
/* |
1265 |
* skip if the co-located P_VOP macroblock is not coded |
* skip if the co-located P_VOP macroblock is not coded |
1268 |
*/ |
*/ |
1269 |
|
|
1270 |
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); */ |
|
1271 |
mb->cbp = 0; |
mb->cbp = 0; |
1272 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
1273 |
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); |
|
1274 |
continue; |
continue; |
1275 |
} |
} |
1276 |
|
|
1277 |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1278 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1279 |
|
|
1280 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1281 |
|
|
1282 |
if (!modb2) { /* modb=='00' */ |
if (!modb2) /* modb=='00' */ |
1283 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1284 |
} else { |
else |
1285 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp) { |
|
|
quant += get_dbquant(bs); |
|
1286 |
|
|
1287 |
if (quant > 31) { |
if (mb->mode && mb->cbp) { |
1288 |
|
quant += get_dbquant(bs); |
1289 |
|
if (quant > 31) |
1290 |
quant = 31; |
quant = 31; |
1291 |
} else if (quant < 1) { |
else if (quant < 1) |
1292 |
quant = 1; |
quant = 1; |
1293 |
} |
} |
1294 |
|
mb->quant = quant; |
1295 |
|
|
1296 |
|
if (dec->interlacing) { |
1297 |
|
if (mb->cbp) { |
1298 |
|
mb->field_dct = BitstreamGetBit(bs); |
1299 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1300 |
|
} |
1301 |
|
|
1302 |
|
if (mb->mode) { |
1303 |
|
mb->field_pred = BitstreamGetBit(bs); |
1304 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1305 |
|
|
1306 |
|
if (mb->field_pred) { |
1307 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1308 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1309 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1310 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1311 |
|
} |
1312 |
|
} |
1313 |
} |
} |
1314 |
|
|
1315 |
} else { |
} else { |
1316 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1317 |
mb->cbp = 0; |
mb->cbp = 0; |
1318 |
} |
} |
1319 |
|
|
1320 |
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) { |
|
1321 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1322 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y); |
1323 |
|
|
1324 |
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; |
|
|
|
|
1325 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1326 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; |
1327 |
/ TRD + mv.x); |
mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; |
1328 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
|
1329 |
? ((TRB - TRD) * last_mb->mvs[i].x) |
mb->b_mvs[i].x = (mv.x) |
1330 |
/ TRD |
? mb->mvs[i].x - last_mb->mvs[i].x |
1331 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1332 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
mb->b_mvs[i].y = (mv.y) |
1333 |
/ TRD + mv.y); |
? mb->mvs[i].y - last_mb->mvs[i].y |
1334 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
|
|
/ TRD |
|
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
|
|
} |
|
|
/* DEBUG("B-frame Direct!\n"); */ |
|
1335 |
} |
} |
1336 |
|
|
1337 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1338 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1339 |
break; |
break; |
1340 |
|
|
1341 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1342 |
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, x, y); |
|
dec->p_fmv); |
|
1343 |
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]; |
1344 |
|
|
1345 |
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, dec, x, y); |
1346 |
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]; |
|
1347 |
|
|
1348 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1349 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
|
/* DEBUG("B-frame Bidir!\n"); */ |
|
1350 |
break; |
break; |
1351 |
|
|
1352 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1353 |
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, x, y); |
|
dec->p_bmv); |
|
1354 |
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]; |
1355 |
|
|
1356 |
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"); */ |
|
1357 |
break; |
break; |
1358 |
|
|
1359 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1360 |
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, x, y); |
|
dec->p_fmv); |
|
1361 |
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]; |
1362 |
|
|
1363 |
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"); */ |
|
1364 |
break; |
break; |
1365 |
|
|
1366 |
default: |
default: |
1367 |
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); |
1368 |
} |
} |
1369 |
} /* End of for */ |
} /* End of for */ |
1370 |
} |
} |
|
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
first=1; |
|
|
if (fp) |
|
|
fclose(fp); |
|
|
} |
|
|
#endif |
|
1371 |
} |
} |
1372 |
|
|
|
|
|
|
|
|
1373 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1374 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1375 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1376 |
|
int coding_type, int quant) |
1377 |
{ |
{ |
1378 |
|
if (dec->cartoon_mode) |
1379 |
|
frame->general &= ~XVID_FILMEFFECT; |
1380 |
|
|
1381 |
|
if (frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) && mbs != NULL) /* post process */ |
1382 |
|
{ |
1383 |
|
/* note: image is stored to tmp */ |
1384 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1385 |
|
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1386 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1387 |
|
frame->general, dec->frames, (coding_type == B_VOP)); |
1388 |
|
img = &dec->tmp; |
1389 |
|
} |
1390 |
|
|
1391 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1392 |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1393 |
frame->output.csp, dec->interlacing); |
frame->output.csp, dec->interlacing); |
1394 |
|
|
1395 |
if (stats) |
if (stats) { |
|
{ |
|
1396 |
stats->type = coding2type(coding_type); |
stats->type = coding2type(coding_type); |
1397 |
stats->data.vop.time_base = (int)dec->time_base; |
stats->data.vop.time_base = (int)dec->time_base; |
1398 |
stats->data.vop.time_increment = 0; /* XXX: todo */ |
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1399 |
|
stats->data.vop.qscale_stride = dec->mb_width; |
1400 |
|
stats->data.vop.qscale = dec->qscale; |
1401 |
|
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1402 |
|
int i; |
1403 |
|
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1404 |
|
stats->data.vop.qscale[i] = mbs[i].quant; |
1405 |
|
} else |
1406 |
|
stats->data.vop.qscale = NULL; |
1407 |
} |
} |
1408 |
} |
} |
1409 |
|
|
|
|
|
1410 |
int |
int |
1411 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1412 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1415 |
Bitstream bs; |
Bitstream bs; |
1416 |
uint32_t rounding; |
uint32_t rounding; |
1417 |
uint32_t reduced_resolution; |
uint32_t reduced_resolution; |
1418 |
uint32_t quant; |
uint32_t quant = 2; |
1419 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1420 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1421 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1433 |
dec->frames = 0; |
dec->frames = 0; |
1434 |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1435 |
|
|
1436 |
if (frame->length < 0) /* decoder flush */ |
if (frame->length < 0) { /* decoder flush */ |
|
{ |
|
1437 |
int ret; |
int ret; |
1438 |
/* 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 |
1439 |
we have a reference frame, then outout the reference frame */ |
we have a reference frame, then outout the reference frame */ |
1440 |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1441 |
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, quant); |
1442 |
dec->frames = 0; |
dec->frames = 0; |
1443 |
ret = 0; |
ret = 0; |
1444 |
}else{ |
}else{ |
1475 |
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", |
1476 |
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); |
1477 |
|
|
1478 |
if (coding_type == -1) /* nothing */ |
if (coding_type == -1) { /* nothing */ |
|
{ |
|
1479 |
if (success) goto done; |
if (success) goto done; |
1480 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1481 |
emms(); |
emms(); |
1482 |
return BitstreamPos(&bs)/8; |
return BitstreamPos(&bs)/8; |
1483 |
} |
} |
1484 |
|
|
1485 |
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1486 |
{ |
|
1487 |
if (coding_type == -3) |
if (coding_type == -3) |
1488 |
decoder_resize(dec); |
decoder_resize(dec); |
1489 |
|
|
1490 |
if (stats) |
if (stats) { |
|
{ |
|
1491 |
stats->type = XVID_TYPE_VOL; |
stats->type = XVID_TYPE_VOL; |
1492 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1493 |
/*XXX: if (dec->interlacing) |
/*XXX: if (dec->interlacing) |
1503 |
goto repeat; |
goto repeat; |
1504 |
} |
} |
1505 |
|
|
1506 |
|
if(dec->frames == 0 && coding_type != I_VOP) { |
1507 |
|
/* 1st frame is not an i-vop */ |
1508 |
|
goto repeat; |
1509 |
|
} |
1510 |
|
|
1511 |
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 */ |
1512 |
|
|
1513 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1514 |
if (dec->packed_mode && coding_type == N_VOP) |
if (dec->packed_mode && coding_type == N_VOP) { |
1515 |
{ |
if (dec->low_delay_default && dec->frames > 0) { |
1516 |
if (dec->low_delay_default && dec->frames > 0) |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
|
1517 |
output = 1; |
output = 1; |
1518 |
} |
} |
1519 |
/* ignore otherwise */ |
/* ignore otherwise */ |
1520 |
} |
} else if (coding_type != B_VOP) { |
1521 |
else if (coding_type != B_VOP) |
switch(coding_type) { |
|
{ |
|
|
switch(coding_type) |
|
|
{ |
|
1522 |
case I_VOP : |
case I_VOP : |
1523 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1524 |
break; |
break; |
1534 |
/* XXX: not_coded vops are not used for forward prediction */ |
/* XXX: not_coded vops are not used for forward prediction */ |
1535 |
/* we should not swap(last_mbs,mbs) */ |
/* we should not swap(last_mbs,mbs) */ |
1536 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1537 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1538 |
break; |
break; |
1539 |
} |
} |
1540 |
|
|
1541 |
if (reduced_resolution) |
if (reduced_resolution) { |
|
{ |
|
1542 |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1543 |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1544 |
16, 0); |
16, 0); |
1545 |
} |
} |
1546 |
|
|
1547 |
/* 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 */ |
1548 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1549 |
{ |
if (dec->low_delay) { |
1550 |
if (dec->low_delay) |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
|
{ |
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
|
1551 |
output = 1; |
output = 1; |
1552 |
} |
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
1553 |
/* output the reference frame */ |
/* output the reference frame */ |
1554 |
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, quant); |
1555 |
output = 1; |
output = 1; |
1556 |
} |
} |
1557 |
} |
} |
1567 |
|
|
1568 |
}else{ /* B_VOP */ |
}else{ /* B_VOP */ |
1569 |
|
|
1570 |
if (dec->low_delay) |
if (dec->low_delay) { |
|
{ |
|
1571 |
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"); |
1572 |
dec->low_delay = 1; |
dec->low_delay = 0; |
1573 |
} |
} |
1574 |
|
|
1575 |
if (dec->frames < 2) |
if (dec->frames < 2) { |
|
{ |
|
1576 |
/* attemping to decode a bvop without atleast 2 reference frames */ |
/* attemping to decode a bvop without atleast 2 reference frames */ |
1577 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1578 |
"broken b-frame, mising ref frames"); |
"broken b-frame, mising ref frames"); |
1579 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1580 |
}else if (dec->time_pp <= dec->time_bp) { |
}else if (dec->time_pp <= dec->time_bp) { |
1581 |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1582 |
decoded in vfw. */ |
decoded in vfw. */ |
1583 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1584 |
"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); |
1585 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1586 |
}else{ |
}else{ |
1587 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1588 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1589 |
} |
} |
1590 |
|
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
|
1591 |
output = 1; |
output = 1; |
1592 |
dec->frames++; |
dec->frames++; |
1593 |
} |
} |
1594 |
|
|
1595 |
|
#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */ |
1596 |
BitstreamByteAlign(&bs); |
BitstreamByteAlign(&bs); |
1597 |
|
#endif |
1598 |
|
|
1599 |
/* low_delay_default mode: repeat in packed_mode */ |
/* low_delay_default mode: repeat in packed_mode */ |
1600 |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
|
{ |
|
1601 |
success = 1; |
success = 1; |
1602 |
goto repeat; |
goto repeat; |
1603 |
} |
} |
1606 |
|
|
1607 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* low_delay_default mode: if we've gotten here without outputting anything, |
1608 |
then output the recently decoded frame, or print an error message */ |
then output the recently decoded frame, or print an error message */ |
1609 |
if (dec->low_delay_default && output == 0) |
if (dec->low_delay_default && output == 0) { |
1610 |
{ |
if (dec->packed_mode && seen_something) { |
|
if (dec->packed_mode && seen_something) |
|
|
{ |
|
1611 |
/* output the recently decoded frame */ |
/* output the recently decoded frame */ |
1612 |
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, quant); |
1613 |
} |
} else { |
|
else |
|
|
{ |
|
1614 |
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); |
1615 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1616 |
"warning: nothing to output"); |
"warning: nothing to output"); |
1617 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1618 |
"bframe decoder lag"); |
"bframe decoder lag"); |
1619 |
|
|
1620 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1621 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
|
|
|
1622 |
} |
} |
1623 |
} |
} |
1624 |
|
|
1625 |
emms(); |
emms(); |
1626 |
stop_global_timer(); |
stop_global_timer(); |
1627 |
|
|
1628 |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */ |
1629 |
} |
} |