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 1.68 2004-12-05 13:56:13 syskin Exp $ |
* $Id: decoder.c,v 1.78 2006-03-27 11:21:48 Skal Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
61 |
#include "image/postprocessing.h" |
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
|
#define DIV2ROUND(n) (((n)>>1)|((n)&1)) |
65 |
|
#define DIV2(n) ((n)>>1) |
66 |
|
#define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) // |
67 |
|
|
68 |
static int |
static int |
69 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
70 |
{ |
{ |
77 |
|
|
78 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
79 |
|
|
80 |
if (dec->last_mbs) |
image_null(&dec->cur); |
81 |
|
image_null(&dec->refn[0]); |
82 |
|
image_null(&dec->refn[1]); |
83 |
|
image_null(&dec->tmp); |
84 |
|
image_null(&dec->qtmp); |
85 |
|
image_null(&dec->gmc); |
86 |
|
|
87 |
|
|
88 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
|
if (dec->mbs) |
|
89 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
|
if (dec->qscale) |
|
90 |
xvid_free(dec->qscale); |
xvid_free(dec->qscale); |
91 |
|
dec->last_mbs = NULL; |
92 |
|
dec->mbs = NULL; |
93 |
|
dec->qscale = NULL; |
94 |
|
|
95 |
/* realloc */ |
/* realloc */ |
96 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
99 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
100 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
101 |
|
|
102 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if ( image_create(&dec->cur, dec->edged_width, dec->edged_height) |
103 |
xvid_free(dec); |
|| image_create(&dec->refn[0], dec->edged_width, dec->edged_height) |
104 |
return XVID_ERR_MEMORY; |
|| image_create(&dec->refn[1], dec->edged_width, dec->edged_height) /* Support B-frame to reference last 2 frame */ |
105 |
} |
|| image_create(&dec->tmp, dec->edged_width, dec->edged_height) |
106 |
|
|| image_create(&dec->qtmp, dec->edged_width, dec->edged_height) |
107 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|| image_create(&dec->gmc, dec->edged_width, dec->edged_height) ) |
108 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
/* Support B-frame to reference last 2 frame */ |
|
|
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
109 |
|
|
110 |
dec->mbs = |
dec->mbs = |
111 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
112 |
CACHE_LINE); |
CACHE_LINE); |
113 |
if (dec->mbs == NULL) { |
if (dec->mbs == NULL) |
114 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
115 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
116 |
|
|
117 |
/* For skip MB flag */ |
/* For skip MB flag */ |
118 |
dec->last_mbs = |
dec->last_mbs = |
119 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
120 |
CACHE_LINE); |
CACHE_LINE); |
121 |
if (dec->last_mbs == NULL) { |
if (dec->last_mbs == NULL) |
122 |
xvid_free(dec->mbs); |
goto memory_error; |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
123 |
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); |
124 |
|
|
125 |
/* nothing happens if that fails */ |
/* nothing happens if that fails */ |
130 |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
131 |
|
|
132 |
return 0; |
return 0; |
133 |
|
|
134 |
|
memory_error: |
135 |
|
/* Most structures were deallocated / nullifieded, so it should be safe */ |
136 |
|
/* decoder_destroy(dec) minus the write_timer */ |
137 |
|
xvid_free(dec->mbs); |
138 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
139 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
141 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
142 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
143 |
|
|
144 |
|
xvid_free(dec); |
145 |
|
return XVID_ERR_MEMORY; |
146 |
} |
} |
147 |
|
|
148 |
|
|
195 |
dec->low_delay = 0; |
dec->low_delay = 0; |
196 |
dec->packed_mode = 0; |
dec->packed_mode = 0; |
197 |
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
198 |
|
dec->ver_id = 1; |
199 |
|
|
200 |
|
dec->bs_version = 0xffff; /* Initialize to very high value -> assume bugfree stream */ |
201 |
|
|
202 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
203 |
|
|
319 |
stop_iquant_timer(); |
stop_iquant_timer(); |
320 |
|
|
321 |
start_timer(); |
start_timer(); |
322 |
idct(&data[i * 64]); |
idct((short * const)&data[i * 64]); |
323 |
stop_idct_timer(); |
stop_idct_timer(); |
324 |
|
|
325 |
} |
} |
351 |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
352 |
|
|
353 |
int stride = dec->edged_width; |
int stride = dec->edged_width; |
|
int next_block = stride * 8; |
|
354 |
int i; |
int i; |
355 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
356 |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
374 |
|
|
375 |
|
|
376 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
377 |
dst[0] = pY_Cur; |
dst[0] = pY_Cur; |
378 |
dst[2] = pY_Cur + next_block; |
dst[1] = pY_Cur + 8; |
379 |
dst[1] = dst[0] + 8; |
dst[2] = pY_Cur + stride; |
380 |
|
dst[3] = dst[2] + 8; |
381 |
|
dst[4] = pU_Cur; |
382 |
|
dst[5] = pV_Cur; |
383 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride*2; |
384 |
|
strides[4] = stride/2; |
385 |
|
strides[5] = stride/2; |
386 |
|
} else { |
387 |
|
dst[0] = pY_Cur; |
388 |
|
dst[1] = pY_Cur + 8; |
389 |
|
dst[2] = pY_Cur + 8*stride; |
390 |
dst[3] = dst[2] + 8; |
dst[3] = dst[2] + 8; |
391 |
dst[4] = pU_Cur; |
dst[4] = pU_Cur; |
392 |
dst[5] = pV_Cur; |
dst[5] = pV_Cur; |
393 |
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
394 |
strides[4] = stride/2; |
strides[4] = stride/2; |
395 |
strides[5] = stride/2; |
strides[5] = stride/2; |
396 |
|
} |
397 |
|
|
398 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
399 |
/* Process only coded blocks */ |
/* Process only coded blocks */ |
409 |
|
|
410 |
/* iDCT */ |
/* iDCT */ |
411 |
start_timer(); |
start_timer(); |
412 |
idct(&data[0]); |
idct((short * const)&data[0]); |
413 |
stop_idct_timer(); |
stop_idct_timer(); |
414 |
|
|
415 |
/* Add this residual to the predicted block */ |
/* Add this residual to the predicted block */ |
456 |
CHECK_MV(mv[3]); |
CHECK_MV(mv[3]); |
457 |
} |
} |
458 |
|
|
459 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
460 |
|
* So we try to be backward compatible to avoid artifacts */ |
461 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
462 |
|
|
463 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
464 |
static void |
static void |
465 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
469 |
const uint32_t cbp, |
const uint32_t cbp, |
470 |
Bitstream * bs, |
Bitstream * bs, |
471 |
const uint32_t rounding, |
const uint32_t rounding, |
472 |
const int ref) |
const int ref, |
473 |
|
const int bvop) |
474 |
{ |
{ |
475 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
476 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
491 |
|
|
492 |
start_timer(); |
start_timer(); |
493 |
|
|
494 |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
495 |
|
|
496 |
uv_dx = mv[0].x; |
uv_dx = mv[0].x; |
497 |
uv_dy = mv[0].y; |
uv_dy = mv[0].y; |
498 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
499 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
500 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
501 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
502 |
|
} |
503 |
|
else { |
504 |
uv_dx /= 2; |
uv_dx /= 2; |
505 |
uv_dy /= 2; |
uv_dy /= 2; |
506 |
} |
} |
507 |
|
} |
508 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
509 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
510 |
|
|
519 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
520 |
|
|
521 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
522 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
523 |
|
int z; |
524 |
|
uv_dx = 0; uv_dy = 0; |
525 |
|
for (z = 0; z < 4; z++) { |
526 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
527 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
528 |
|
} |
529 |
|
} |
530 |
|
else { |
531 |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
532 |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
533 |
|
} |
534 |
} else { |
} else { |
535 |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
536 |
uv_dy = 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; |
576 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
577 |
} |
} |
578 |
|
|
579 |
|
/* decode an inter macroblock in field mode */ |
580 |
|
static void |
581 |
|
decoder_mbinter_field(DECODER * dec, |
582 |
|
const MACROBLOCK * pMB, |
583 |
|
const uint32_t x_pos, |
584 |
|
const uint32_t y_pos, |
585 |
|
const uint32_t cbp, |
586 |
|
Bitstream * bs, |
587 |
|
const uint32_t rounding, |
588 |
|
const int ref, |
589 |
|
const int bvop) |
590 |
|
{ |
591 |
|
uint32_t stride = dec->edged_width; |
592 |
|
uint32_t stride2 = stride / 2; |
593 |
|
|
594 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
595 |
|
|
596 |
|
int uvtop_dx, uvtop_dy; |
597 |
|
int uvbot_dx, uvbot_dy; |
598 |
|
VECTOR mv[4]; /* local copy of mvs */ |
599 |
|
|
600 |
|
/* Get pointer to memory areas */ |
601 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
602 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
603 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
604 |
|
|
605 |
|
mv[0] = pMB->mvs[0]; |
606 |
|
mv[1] = pMB->mvs[1]; |
607 |
|
memset(&mv[2],0,2*sizeof(VECTOR)); |
608 |
|
|
609 |
|
validate_vector(mv, x_pos, y_pos, dec); |
610 |
|
|
611 |
|
start_timer(); |
612 |
|
|
613 |
|
if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
614 |
|
{ |
615 |
|
/* Prepare top field vector */ |
616 |
|
uvtop_dx = DIV2ROUND(mv[0].x); |
617 |
|
uvtop_dy = DIV2ROUND(mv[0].y); |
618 |
|
|
619 |
|
/* Prepare bottom field vector */ |
620 |
|
uvbot_dx = DIV2ROUND(mv[1].x); |
621 |
|
uvbot_dy = DIV2ROUND(mv[1].y); |
622 |
|
|
623 |
|
if(dec->quarterpel) |
624 |
|
{ |
625 |
|
/* NOT supported */ |
626 |
|
} |
627 |
|
else |
628 |
|
{ |
629 |
|
/* Interpolate top field left part(we use double stride for every 2nd line) */ |
630 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
631 |
|
16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
632 |
|
/* top field right part */ |
633 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
634 |
|
16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
635 |
|
|
636 |
|
/* Interpolate bottom field left part(we use double stride for every 2nd line) */ |
637 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
638 |
|
16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
639 |
|
/* Bottom field right part */ |
640 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
641 |
|
16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
642 |
|
|
643 |
|
/* Interpolate field1 U */ |
644 |
|
interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2, |
645 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
646 |
|
|
647 |
|
/* Interpolate field1 V */ |
648 |
|
interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2, |
649 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
650 |
|
|
651 |
|
/* Interpolate field2 U */ |
652 |
|
interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2, |
653 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
654 |
|
|
655 |
|
/* Interpolate field2 V */ |
656 |
|
interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2, |
657 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
658 |
|
} |
659 |
|
} |
660 |
|
else |
661 |
|
{ |
662 |
|
/* We don't expect 4 motion vectors in interlaced mode */ |
663 |
|
} |
664 |
|
|
665 |
|
stop_comp_timer(); |
666 |
|
|
667 |
|
/* Must add error correction? */ |
668 |
|
if(cbp) |
669 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
670 |
|
} |
671 |
|
|
672 |
static void |
static void |
673 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
674 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
836 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
837 |
} |
} |
838 |
|
|
839 |
|
/* We use this when decoder runs interlaced -> different prediction */ |
840 |
|
|
841 |
|
static void get_motion_vector_interlaced(DECODER * dec, |
842 |
|
Bitstream * bs, |
843 |
|
int x, |
844 |
|
int y, |
845 |
|
int k, |
846 |
|
MACROBLOCK *pMB, |
847 |
|
int fcode, |
848 |
|
const int bound) |
849 |
|
{ |
850 |
|
const int scale_fac = 1 << (fcode - 1); |
851 |
|
const int high = (32 * scale_fac) - 1; |
852 |
|
const int low = ((-32) * scale_fac); |
853 |
|
const int range = (64 * scale_fac); |
854 |
|
|
855 |
|
/* Get interlaced prediction */ |
856 |
|
const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k); |
857 |
|
VECTOR mv,mvf1,mvf2; |
858 |
|
|
859 |
|
if(!pMB->field_pred) |
860 |
|
{ |
861 |
|
mv.x = get_mv(bs,fcode); |
862 |
|
mv.y = get_mv(bs,fcode); |
863 |
|
|
864 |
|
mv.x += pmv.x; |
865 |
|
mv.y += pmv.y; |
866 |
|
|
867 |
|
if(mv.x<low) { |
868 |
|
mv.x += range; |
869 |
|
} else if (mv.x>high) { |
870 |
|
mv.x-=range; |
871 |
|
} |
872 |
|
|
873 |
|
if (mv.y < low) { |
874 |
|
mv.y += range; |
875 |
|
} else if (mv.y > high) { |
876 |
|
mv.y -= range; |
877 |
|
} |
878 |
|
|
879 |
|
pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv; |
880 |
|
} |
881 |
|
else |
882 |
|
{ |
883 |
|
mvf1.x = get_mv(bs, fcode); |
884 |
|
mvf1.y = get_mv(bs, fcode); |
885 |
|
|
886 |
|
mvf1.x += pmv.x; |
887 |
|
mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */ |
888 |
|
|
889 |
|
if (mvf1.x < low) { |
890 |
|
mvf1.x += range; |
891 |
|
} else if (mvf1.x > high) { |
892 |
|
mvf1.x -= range; |
893 |
|
} |
894 |
|
|
895 |
|
if (mvf1.y < low) { |
896 |
|
mvf1.y += range; |
897 |
|
} else if (mvf1.y > high) { |
898 |
|
mvf1.y -= range; |
899 |
|
} |
900 |
|
|
901 |
|
mvf2.x = get_mv(bs, fcode); |
902 |
|
mvf2.y = get_mv(bs, fcode); |
903 |
|
|
904 |
|
mvf2.x += pmv.x; |
905 |
|
mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */ |
906 |
|
|
907 |
|
if (mvf2.x < low) { |
908 |
|
mvf2.x += range; |
909 |
|
} else if (mvf2.x > high) { |
910 |
|
mvf2.x -= range; |
911 |
|
} |
912 |
|
|
913 |
|
if (mvf2.y < low) { |
914 |
|
mvf2.y += range; |
915 |
|
} else if (mvf2.y > high) { |
916 |
|
mvf2.y -= range; |
917 |
|
} |
918 |
|
|
919 |
|
pMB->mvs[0]=mvf1; |
920 |
|
pMB->mvs[1]=mvf2; |
921 |
|
pMB->mvs[2].x=pMB->mvs[3].x=0; |
922 |
|
pMB->mvs[2].y=pMB->mvs[3].y=0; |
923 |
|
|
924 |
|
/* Calculate average for as it is field predicted */ |
925 |
|
pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x); |
926 |
|
pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y); |
927 |
|
} |
928 |
|
} |
929 |
|
|
930 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
931 |
static void |
static void |
932 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
1019 |
} |
} |
1020 |
mb->quant = quant; |
mb->quant = quant; |
1021 |
|
|
1022 |
|
mb->field_pred=0; |
1023 |
if (dec->interlacing) { |
if (dec->interlacing) { |
1024 |
if (cbp || intra) { |
if (cbp || intra) { |
1025 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
1045 |
|
|
1046 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
1047 |
|
|
1048 |
if (dec->interlacing && mb->field_pred) { |
if(dec->interlacing) { |
1049 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
/* Get motion vectors interlaced, field_pred is handled there */ |
1050 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound); |
1051 |
} else { |
} else { |
1052 |
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); |
1053 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1054 |
} |
} |
1055 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
1056 |
|
/* interlaced missing here */ |
1057 |
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); |
1058 |
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); |
1059 |
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); |
1066 |
continue; |
continue; |
1067 |
} |
} |
1068 |
|
|
1069 |
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0); |
/* See how to decode */ |
1070 |
|
if(!mb->field_pred) |
1071 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1072 |
|
else |
1073 |
|
decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1074 |
|
|
1075 |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
1076 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
1088 |
|
|
1089 |
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; |
1090 |
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; |
1091 |
|
mb->field_pred=0; /* (!) */ |
1092 |
|
|
1093 |
decoder_mbinter(dec, mb, x, y, 0, bs, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
1094 |
rounding, 0); |
rounding, 0, 0); |
1095 |
|
|
1096 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1097 |
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); |
1173 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1174 |
|
|
1175 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
1176 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1177 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
1178 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1179 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1180 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1181 |
|
} |
1182 |
|
else { |
1183 |
uv_dx /= 2; |
uv_dx /= 2; |
1184 |
uv_dy /= 2; |
uv_dy /= 2; |
1185 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1186 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1187 |
} |
} |
1188 |
|
} |
1189 |
|
|
1190 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1191 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1193 |
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]; |
1194 |
|
|
1195 |
} else { |
} else { |
1196 |
|
if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */ |
1197 |
|
/* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */ |
1198 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1199 |
|
int z; |
1200 |
|
uv_dx = 0; uv_dy = 0; |
1201 |
|
b_uv_dx = 0; b_uv_dy = 0; |
1202 |
|
for (z = 0; z < 4; z++) { |
1203 |
|
uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1)); |
1204 |
|
uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1)); |
1205 |
|
b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1)); |
1206 |
|
b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1)); |
1207 |
|
} |
1208 |
|
} |
1209 |
|
else { |
1210 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1211 |
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1212 |
|
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); |
1213 |
|
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); |
1214 |
|
} |
1215 |
|
} else { |
1216 |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1217 |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1218 |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1219 |
b_uv_dy = 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 + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
|
|
|
|
if (dec->quarterpel) { |
|
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
b_uv_dx /= 2; |
|
|
b_uv_dy /= 2; |
|
1220 |
} |
} |
1221 |
|
|
1222 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1370 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1371 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1372 |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1373 |
uint32_t intra_dc_threshold; /* fake variable */ |
int intra_dc_threshold; /* fake variable */ |
1374 |
|
|
1375 |
if (check_resync_marker(bs, fcode_max - 1)) { |
if (check_resync_marker(bs, fcode_max - 1)) { |
1376 |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1395 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1396 |
mb->cbp = 0; |
mb->cbp = 0; |
1397 |
mb->mode = MODE_FORWARD; |
mb->mode = MODE_FORWARD; |
1398 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1399 |
continue; |
continue; |
1400 |
} |
} |
1401 |
|
|
1478 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1479 |
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]; |
1480 |
|
|
1481 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1482 |
break; |
break; |
1483 |
|
|
1484 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1485 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1486 |
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]; |
1487 |
|
|
1488 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1489 |
break; |
break; |
1490 |
|
|
1491 |
default: |
default: |
1496 |
} |
} |
1497 |
|
|
1498 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1499 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1500 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1501 |
int coding_type, int quant) |
int coding_type, int quant) |
1502 |
{ |
{ |
1527 |
stats->data.vop.qscale_stride = dec->mb_width; |
stats->data.vop.qscale_stride = dec->mb_width; |
1528 |
stats->data.vop.qscale = dec->qscale; |
stats->data.vop.qscale = dec->qscale; |
1529 |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1530 |
int i; |
unsigned int i; |
1531 |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1532 |
stats->data.vop.qscale[i] = mbs[i].quant; |
stats->data.vop.qscale[i] = mbs[i].quant; |
1533 |
} else |
} else |
1599 |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
1600 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1601 |
|
|
1602 |
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=%" |
1603 |
|
#if defined(_MSC_VER) |
1604 |
|
"I64" |
1605 |
|
#else |
1606 |
|
"ll" |
1607 |
|
#endif |
1608 |
|
"i, time_pp=%i, time_bp=%i\n", |
1609 |
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); |
1610 |
|
|
1611 |
if (coding_type == -1) { /* nothing */ |
if (coding_type == -1) { /* nothing */ |
1732 |
|
|
1733 |
done : |
done : |
1734 |
|
|
1735 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* if we reach here without outputing anything _and_ |
1736 |
then output the recently decoded frame, or print an error message */ |
the calling application has specified low_delay_default, |
1737 |
|
we *must* output something. |
1738 |
|
this always occurs on the first call to decode() call |
1739 |
|
when bframes are present in the bitstream. it may also |
1740 |
|
occur if no vops were seen in the bitstream |
1741 |
|
|
1742 |
|
if packed_mode is enabled, then we output the recently |
1743 |
|
decoded frame (the very first ivop). otherwise we have |
1744 |
|
nothing to display, and therefore output a black screen. |
1745 |
|
*/ |
1746 |
if (dec->low_delay_default && output == 0) { |
if (dec->low_delay_default && output == 0) { |
1747 |
if (dec->packed_mode && seen_something) { |
if (dec->packed_mode && seen_something) { |
|
/* output the recently decoded frame */ |
|
1748 |
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, quant); |
1749 |
} else { |
} else { |
1750 |
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); |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"warning: nothing to output"); |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
|
|
"bframe decoder lag"); |
|
|
|
|
1751 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1752 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1753 |
} |
} |