3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder main module - |
5 |
* |
* |
6 |
|
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
|
* 2002 Peter Ross <pross@xvid.org> |
8 |
|
* |
9 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
10 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
11 |
* to use this software module in hardware or software products are |
* to use this software module in hardware or software products are |
36 |
* History: |
* History: |
37 |
* |
* |
38 |
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
39 |
* MinChen <chenm001@163.com> |
|
40 |
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
41 |
* Fix a little bug for low_delay flage |
* Fix a little bug for low_delay flage |
42 |
* MinChen <chenm001@163.com> |
* MinChen <chenm001@163.com> |
56 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
57 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
58 |
* 22.12.2001 lock based interpolation |
* 22.12.2001 lock based interpolation |
59 |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* 01.12.2001 inital version; (c)2001 peter ross <pross@xvid.org> |
60 |
* |
* |
61 |
* $Id$ |
* $Id$ |
62 |
* |
* |
139 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
140 |
} |
} |
141 |
|
|
|
if (image_create(&dec->refh, 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->refn[2], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
142 |
dec->mbs = |
dec->mbs = |
143 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
144 |
CACHE_LINE); |
CACHE_LINE); |
147 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
148 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
149 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
150 |
xvid_free(dec); |
xvid_free(dec); |
151 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
152 |
} |
} |
164 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
165 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
166 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
167 |
xvid_free(dec); |
xvid_free(dec); |
168 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
169 |
} |
} |
189 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
190 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
191 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
192 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
193 |
xvid_free(dec); |
xvid_free(dec); |
194 |
|
|
271 |
start_timer(); |
start_timer(); |
272 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
273 |
{ |
{ |
274 |
int direction = dec->alternate_vertical_scan ? |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
275 |
2 : pMB->acpred_directions[i]; |
start_coeff); |
|
|
|
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
|
276 |
} |
} |
277 |
stop_coding_timer(); |
stop_coding_timer(); |
278 |
|
|
359 |
|
|
360 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
361 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
|
|
start_timer(); |
|
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
|
|
} |
|
|
else { |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->mvs[3].x, pMB->mvs[3].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(); |
|
|
|
|
362 |
} else { |
} else { |
363 |
int sum; |
int sum; |
364 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
378 |
} |
} |
379 |
|
|
380 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
381 |
|
} |
382 |
|
|
383 |
start_timer(); |
start_timer(); |
384 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
385 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n"); |
386 |
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
387 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
388 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
|
389 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
390 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
|
391 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
392 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
393 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
394 |
} |
} |
395 |
else { |
else { |
408 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
409 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
410 |
stop_comp_timer(); |
stop_comp_timer(); |
|
} |
|
411 |
|
|
412 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
413 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
414 |
{ |
{ |
415 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
416 |
|
|
417 |
start_timer(); |
start_timer(); |
418 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
419 |
stop_coding_timer(); |
stop_coding_timer(); |
420 |
|
|
421 |
start_timer(); |
start_timer(); |
512 |
|
|
513 |
if (dec->interlacing) { |
if (dec->interlacing) { |
514 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
515 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
516 |
} |
} |
517 |
|
|
518 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
658 |
if (dec->interlacing) { |
if (dec->interlacing) { |
659 |
if (cbp || intra) { |
if (cbp || intra) { |
660 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
661 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
662 |
} |
} |
663 |
|
|
664 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
665 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
666 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
667 |
|
|
668 |
if (mb->field_pred) { |
if (mb->field_pred) { |
669 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
670 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
671 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
672 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
673 |
} |
} |
674 |
} |
} |
675 |
} |
} |
709 |
rounding); |
rounding); |
710 |
} else // not coded |
} else // not coded |
711 |
{ |
{ |
712 |
DEBUG2("P-frame MB at (X,Y)=",x,y); |
DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y); |
713 |
|
|
714 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
715 |
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; |
716 |
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; |
880 |
stop_comp_timer(); |
stop_comp_timer(); |
881 |
|
|
882 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
883 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
884 |
{ |
{ |
885 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
886 |
|
|
887 |
start_timer(); |
start_timer(); |
888 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
889 |
stop_coding_timer(); |
stop_coding_timer(); |
890 |
|
|
891 |
start_timer(); |
start_timer(); |
1031 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1032 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1033 |
|
|
1034 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
1035 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
stride); |
1036 |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1037 |
stride, 0); |
stride); |
1038 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1039 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
stride); |
1040 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1041 |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, |
16 * y_pos + 8, stride); |
1042 |
stride, 0); |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1043 |
|
stride2); |
1044 |
interpolate8x8_avg2(dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
1045 |
dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
stride2); |
|
dec->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos, |
|
|
stride, 0); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->refn[2].y + (16 * (y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
stride, 0); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, |
|
|
dec->cur.u + (8 * y_pos * stride) + 8 * x_pos, |
|
|
dec->refn[2].u + (8 * y_pos * stride) + 8 * x_pos, |
|
|
stride2, 0); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, |
|
|
dec->cur.v + (8 * y_pos * stride) + 8 * x_pos, |
|
|
dec->refn[2].v + (8 * y_pos * stride) + 8 * x_pos, |
|
|
stride2, 0); |
|
|
|
|
1046 |
stop_comp_timer(); |
stop_comp_timer(); |
1047 |
|
|
1048 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
1049 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
1050 |
{ |
{ |
1051 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1052 |
|
|
1053 |
start_timer(); |
start_timer(); |
1054 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
1055 |
stop_coding_timer(); |
stop_coding_timer(); |
1056 |
|
|
1057 |
start_timer(); |
start_timer(); |
1282 |
break; |
break; |
1283 |
|
|
1284 |
default: |
default: |
1285 |
DEBUG1("Not support B-frame mb_type =", mb->mb_type); |
DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type); |
1286 |
} |
} |
1287 |
|
|
1288 |
} // end of FOR |
} // end of FOR |
1340 |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1341 |
intra_dc_threshold); |
intra_dc_threshold); |
1342 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1343 |
DEBUG1("P_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "P_VOP Time=%d", dec->time); |
1344 |
#endif |
#endif |
1345 |
break; |
break; |
1346 |
|
|
1347 |
case I_VOP: |
case I_VOP: |
1348 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1349 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1350 |
DEBUG1("I_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "I_VOP Time=%d", dec->time); |
1351 |
#endif |
#endif |
1352 |
break; |
break; |
1353 |
|
|
1354 |
case B_VOP: |
case B_VOP: |
1355 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1356 |
if (dec->time_pp > dec->time_bp) { |
if (dec->time_pp > dec->time_bp) { |
1357 |
DEBUG1("B_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "B_VOP Time=%d", dec->time); |
1358 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1359 |
} else { |
} else { |
1360 |
DEBUG("broken B-frame!"); |
DPRINTF(DPRINTF_DEBUG, "Broken B_VOP"); |
1361 |
} |
} |
1362 |
#else |
#else |
1363 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1375 |
|
|
1376 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1377 |
if (frame->length != BitstreamPos(&bs) / 8){ |
if (frame->length != BitstreamPos(&bs) / 8){ |
1378 |
DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); |
DPRINTF(DPRINTF_DEBUG, "InLen: %d / UseLen: %d", frame->length, BitstreamPos(&bs) / 8); |
1379 |
} |
} |
1380 |
#endif |
#endif |
1381 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |