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 1.58 2004-05-31 21:36:23 edgomez Exp $ |
* $Id: decoder.c,v 1.72 2005-08-01 18:37:46 Isibaar Exp $ |
24 |
* |
* |
25 |
****************************************************************************/ |
****************************************************************************/ |
26 |
|
|
46 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
47 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
48 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
49 |
#include "image/font.h" |
#include "image/font.h" |
50 |
|
#include "image/qpel.h" |
51 |
|
|
52 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
53 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
73 |
|
|
74 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
75 |
|
|
76 |
if (dec->last_mbs) |
image_null(&dec->cur); |
77 |
|
image_null(&dec->refn[0]); |
78 |
|
image_null(&dec->refn[1]); |
79 |
|
image_null(&dec->tmp); |
80 |
|
image_null(&dec->qtmp); |
81 |
|
image_null(&dec->gmc); |
82 |
|
|
83 |
|
|
84 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
|
if (dec->mbs) |
|
85 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
|
if (dec->qscale) |
|
86 |
xvid_free(dec->qscale); |
xvid_free(dec->qscale); |
87 |
|
dec->last_mbs = NULL; |
88 |
|
dec->mbs = NULL; |
89 |
|
dec->qscale = NULL; |
90 |
|
|
91 |
/* realloc */ |
/* realloc */ |
92 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
95 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
96 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
97 |
|
|
98 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if ( image_create(&dec->cur, dec->edged_width, dec->edged_height) |
99 |
xvid_free(dec); |
|| image_create(&dec->refn[0], dec->edged_width, dec->edged_height) |
100 |
return XVID_ERR_MEMORY; |
|| image_create(&dec->refn[1], dec->edged_width, dec->edged_height) /* Support B-frame to reference last 2 frame */ |
101 |
} |
|| image_create(&dec->tmp, dec->edged_width, dec->edged_height) |
102 |
|
|| image_create(&dec->qtmp, dec->edged_width, dec->edged_height) |
103 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|| image_create(&dec->gmc, dec->edged_width, dec->edged_height) ) |
104 |
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; |
|
|
} |
|
105 |
|
|
106 |
dec->mbs = |
dec->mbs = |
107 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
108 |
CACHE_LINE); |
CACHE_LINE); |
109 |
if (dec->mbs == NULL) { |
if (dec->mbs == NULL) |
110 |
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; |
|
|
} |
|
111 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
112 |
|
|
113 |
/* For skip MB flag */ |
/* For skip MB flag */ |
114 |
dec->last_mbs = |
dec->last_mbs = |
115 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
116 |
CACHE_LINE); |
CACHE_LINE); |
117 |
if (dec->last_mbs == NULL) { |
if (dec->last_mbs == NULL) |
118 |
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; |
|
|
} |
|
|
|
|
119 |
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); |
120 |
|
|
121 |
/* nothing happens if that fails */ |
/* nothing happens if that fails */ |
126 |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
127 |
|
|
128 |
return 0; |
return 0; |
129 |
|
|
130 |
|
memory_error: |
131 |
|
/* Most structures were deallocated / nullifieded, so it should be safe */ |
132 |
|
/* decoder_destroy(dec) minus the write_timer */ |
133 |
|
xvid_free(dec->mbs); |
134 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
135 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
136 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
137 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
138 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
139 |
|
|
140 |
|
xvid_free(dec); |
141 |
|
return XVID_ERR_MEMORY; |
142 |
} |
} |
143 |
|
|
144 |
|
|
238 |
Bitstream * bs, |
Bitstream * bs, |
239 |
const uint32_t quant, |
const uint32_t quant, |
240 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
241 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
242 |
{ |
{ |
243 |
|
|
244 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
251 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
252 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
253 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
}else{ |
|
254 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
255 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
256 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
257 |
|
|
258 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
259 |
|
|
264 |
|
|
265 |
start_timer(); |
start_timer(); |
266 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
267 |
iQuant, iDcScaler, predictors, bound, dec->bs_version); |
iQuant, iDcScaler, predictors, bound); |
268 |
if (!acpred_flag) { |
if (!acpred_flag) { |
269 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
270 |
} |
} |
300 |
stop_coding_timer(); |
stop_coding_timer(); |
301 |
|
|
302 |
start_timer(); |
start_timer(); |
303 |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
304 |
stop_prediction_timer(); |
stop_prediction_timer(); |
305 |
|
|
306 |
start_timer(); |
start_timer(); |
323 |
} |
} |
324 |
|
|
325 |
start_timer(); |
start_timer(); |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
next_block*=2; |
|
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
}else{ |
|
326 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
327 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
328 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
329 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
330 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
331 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
332 |
stop_transfer_timer(); |
stop_transfer_timer(); |
333 |
} |
} |
334 |
|
|
339 |
uint8_t * pY_Cur, |
uint8_t * pY_Cur, |
340 |
uint8_t * pU_Cur, |
uint8_t * pU_Cur, |
341 |
uint8_t * pV_Cur, |
uint8_t * pV_Cur, |
|
const int reduced_resolution, |
|
342 |
const MACROBLOCK * pMB) |
const MACROBLOCK * pMB) |
343 |
{ |
{ |
344 |
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
345 |
|
|
346 |
int stride = dec->edged_width; |
int stride = dec->edged_width; |
347 |
int next_block = stride * (reduced_resolution ? 16 : 8); |
int next_block = stride * 8; |
|
const int stride2 = stride/2; |
|
348 |
int i; |
int i; |
349 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
350 |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
351 |
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
typedef void (*get_inter_block_function_t)( |
352 |
|
Bitstream * bs, |
353 |
|
int16_t * block, |
354 |
|
int direction, |
355 |
|
const int quant, |
356 |
|
const uint16_t *matrix); |
357 |
|
typedef void (*add_residual_function_t)( |
358 |
|
uint8_t *predicted_block, |
359 |
|
const int16_t *residual, |
360 |
|
int stride); |
361 |
|
|
362 |
|
const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) |
363 |
|
? (get_inter_block_function_t)get_inter_block_h263 |
364 |
|
: (get_inter_block_function_t)get_inter_block_mpeg; |
365 |
|
|
366 |
for (i = 0; i < 6; i++) { |
uint8_t *dst[6]; |
367 |
|
int strides[6]; |
368 |
|
|
|
if (cbp & (1 << (5 - i))) { /* coded */ |
|
369 |
|
|
370 |
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
if (dec->interlacing && pMB->field_dct) { |
371 |
|
next_block = stride; |
372 |
|
stride *= 2; |
373 |
|
} |
374 |
|
|
375 |
|
dst[0] = pY_Cur; |
376 |
|
dst[2] = pY_Cur + next_block; |
377 |
|
dst[1] = dst[0] + 8; |
378 |
|
dst[3] = dst[2] + 8; |
379 |
|
dst[4] = pU_Cur; |
380 |
|
dst[5] = pV_Cur; |
381 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
382 |
|
strides[4] = stride/2; |
383 |
|
strides[5] = stride/2; |
384 |
|
|
385 |
|
for (i = 0; i < 6; i++) { |
386 |
|
/* Process only coded blocks */ |
387 |
|
if (cbp & (1 << (5 - i))) { |
388 |
|
|
389 |
|
/* Clear the block */ |
390 |
|
memset(&data[0], 0, 64*sizeof(int16_t)); |
391 |
|
|
392 |
|
/* Decode coeffs and dequantize on the fly */ |
393 |
start_timer(); |
start_timer(); |
394 |
get_inter_block(bs, block, direction); |
get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices)); |
395 |
stop_coding_timer(); |
stop_coding_timer(); |
396 |
|
|
397 |
|
/* iDCT */ |
398 |
start_timer(); |
start_timer(); |
399 |
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
idct(&data[0]); |
400 |
stop_iquant_timer(); |
stop_idct_timer(); |
401 |
|
|
402 |
|
/* Add this residual to the predicted block */ |
403 |
start_timer(); |
start_timer(); |
404 |
idct(&data[i * 64]); |
transfer_16to8add(dst[i], &data[0], strides[i]); |
405 |
stop_idct_timer(); |
stop_transfer_timer(); |
406 |
} |
} |
407 |
} |
} |
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
408 |
} |
} |
409 |
|
|
410 |
start_timer(); |
static void __inline |
411 |
if (reduced_resolution) { |
validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec) |
412 |
if (cbp & 32) |
{ |
413 |
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
/* clip a vector to valid range |
414 |
if (cbp & 16) |
prevents crashes if bitstream is broken |
415 |
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
*/ |
416 |
if (cbp & 8) |
int shift = 5 + dec->quarterpel; |
417 |
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
int xborder_high = (int)(dec->mb_width - x_pos) << shift; |
418 |
if (cbp & 4) |
int xborder_low = (-(int)x_pos-1) << shift; |
419 |
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
int yborder_high = (int)(dec->mb_height - y_pos) << shift; |
420 |
if (cbp & 2) |
int yborder_low = (-(int)y_pos-1) << shift; |
421 |
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
422 |
if (cbp & 1) |
#define CHECK_MV(mv) \ |
423 |
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
do { \ |
424 |
} else { |
if ((mv).x > xborder_high) { \ |
425 |
if (cbp & 32) |
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \ |
426 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
(mv).x = xborder_high; \ |
427 |
if (cbp & 16) |
} else if ((mv).x < xborder_low) { \ |
428 |
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \ |
429 |
if (cbp & 8) |
(mv).x = xborder_low; \ |
430 |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
} \ |
431 |
if (cbp & 4) |
if ((mv).y > yborder_high) { \ |
432 |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \ |
433 |
if (cbp & 2) |
(mv).y = yborder_high; \ |
434 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
} else if ((mv).y < yborder_low) { \ |
435 |
if (cbp & 1) |
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \ |
436 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
(mv).y = yborder_low; \ |
437 |
} |
} \ |
438 |
stop_transfer_timer(); |
} while (0) |
439 |
} |
|
440 |
|
CHECK_MV(mv[0]); |
441 |
|
CHECK_MV(mv[1]); |
442 |
|
CHECK_MV(mv[2]); |
443 |
|
CHECK_MV(mv[3]); |
444 |
|
} |
445 |
|
|
446 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
447 |
|
* So we try to be backward compatible to avoid artifacts */ |
448 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
449 |
|
|
450 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
451 |
static void |
static void |
456 |
const uint32_t cbp, |
const uint32_t cbp, |
457 |
Bitstream * bs, |
Bitstream * bs, |
458 |
const uint32_t rounding, |
const uint32_t rounding, |
|
const int reduced_resolution, |
|
459 |
const int ref) |
const int ref) |
460 |
{ |
{ |
461 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
467 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
468 |
VECTOR mv[4]; /* local copy of mvs */ |
VECTOR mv[4]; /* local copy of mvs */ |
469 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
for (i = 0; i < 4; i++) { |
|
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
|
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
|
|
} |
|
|
} else { |
|
470 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
471 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
472 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
473 |
for (i = 0; i < 4; i++) |
for (i = 0; i < 4; i++) |
474 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
|
} |
|
475 |
|
|
476 |
for (i = 0; i < 4; i++) { |
validate_vector(mv, x_pos, y_pos, dec); |
|
/* clip to valid range */ |
|
|
int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel); |
|
|
if (mv[i].x > border) { |
|
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
|
|
mv[i].x = border; |
|
|
} else { |
|
|
border = (-(int)x_pos-1) << (5 + dec->quarterpel); |
|
|
if (mv[i].x < border) { |
|
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
|
|
mv[i].x = border; |
|
|
} |
|
|
} |
|
|
|
|
|
border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel); |
|
|
if (mv[i].y > border) { |
|
|
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
|
|
mv[i].y = border; |
|
|
} else { |
|
|
border = (-(int)y_pos-1) << (5 + dec->quarterpel); |
|
|
if (mv[i].y < border) { |
|
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
|
|
mv[i].y = border; |
|
|
} |
|
|
} |
|
|
} |
|
477 |
|
|
478 |
start_timer(); |
start_timer(); |
479 |
|
|
482 |
uv_dx = mv[0].x; |
uv_dx = mv[0].x; |
483 |
uv_dy = mv[0].y; |
uv_dy = mv[0].y; |
484 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
485 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
486 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
487 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
488 |
|
} |
489 |
|
else { |
490 |
uv_dx /= 2; |
uv_dx /= 2; |
491 |
uv_dy /= 2; |
uv_dy /= 2; |
492 |
} |
} |
493 |
|
} |
494 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
495 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
496 |
|
|
497 |
if (reduced_resolution) |
if (dec->quarterpel) |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
else if (dec->quarterpel) |
|
498 |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
499 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
500 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
505 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
506 |
|
|
507 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
508 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
509 |
|
int z; |
510 |
|
uv_dx = 0; uv_dy = 0; |
511 |
|
for (z = 0; z < 4; z++) { |
512 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
513 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
514 |
|
} |
515 |
|
} |
516 |
|
else { |
517 |
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); |
518 |
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); |
519 |
|
} |
520 |
} else { |
} else { |
521 |
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; |
522 |
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; |
525 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
526 |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
527 |
|
|
528 |
if (reduced_resolution) { |
if (dec->quarterpel) { |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
|
|
mv[1].x, mv[1].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
|
|
mv[2].x, mv[2].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
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) { |
|
529 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
530 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
531 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
551 |
} |
} |
552 |
|
|
553 |
/* chroma */ |
/* chroma */ |
|
if (reduced_resolution) { |
|
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
} else { |
|
554 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
555 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
556 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
557 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
|
} |
|
558 |
|
|
559 |
stop_comp_timer(); |
stop_comp_timer(); |
560 |
|
|
561 |
if (cbp) |
if (cbp) |
562 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
|
reduced_resolution, pMB); |
|
563 |
} |
} |
564 |
|
|
565 |
static void |
static void |
606 |
stop_transfer_timer(); |
stop_transfer_timer(); |
607 |
|
|
608 |
if (cbp) |
if (cbp) |
609 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
610 |
|
|
611 |
} |
} |
612 |
|
|
614 |
static void |
static void |
615 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
616 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
617 |
int quant, |
int quant, |
618 |
int intra_dc_threshold) |
int intra_dc_threshold) |
619 |
{ |
{ |
620 |
uint32_t bound; |
uint32_t bound; |
621 |
uint32_t x, y; |
uint32_t x, y; |
622 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
623 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) { |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
624 |
|
|
625 |
bound = 0; |
bound = 0; |
626 |
|
|
676 |
} |
} |
677 |
|
|
678 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
679 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
680 |
|
|
681 |
} |
} |
682 |
if(dec->out_frm) |
if(dec->out_frm) |
734 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
735 |
Bitstream * bs, |
Bitstream * bs, |
736 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
737 |
int quant, |
int quant, |
738 |
int fcode, |
int fcode, |
739 |
int intra_dc_threshold, |
int intra_dc_threshold, |
742 |
uint32_t x, y; |
uint32_t x, y; |
743 |
uint32_t bound; |
uint32_t bound; |
744 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
745 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
746 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) { |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
747 |
|
|
748 |
if (!dec->is_edged[0]) { |
if (!dec->is_edged[0]) { |
749 |
start_timer(); |
start_timer(); |
862 |
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; |
863 |
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; |
864 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
865 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
866 |
continue; |
continue; |
867 |
} |
} |
868 |
|
|
869 |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0); |
|
rounding, reduced_resolution, 0); |
|
870 |
|
|
871 |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
872 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
886 |
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; |
887 |
|
|
888 |
decoder_mbinter(dec, mb, x, y, 0, bs, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
889 |
rounding, reduced_resolution, 0); |
rounding, 0); |
890 |
|
|
891 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
892 |
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); |
941 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
942 |
IMAGE forward, |
IMAGE forward, |
943 |
IMAGE backward, |
IMAGE backward, |
944 |
const MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
945 |
const uint32_t x_pos, |
const uint32_t x_pos, |
946 |
const uint32_t y_pos, |
const uint32_t y_pos, |
947 |
Bitstream * bs, |
Bitstream * bs, |
958 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
959 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
960 |
|
|
961 |
|
validate_vector(pMB->mvs, x_pos, y_pos, dec); |
962 |
|
validate_vector(pMB->b_mvs, x_pos, y_pos, dec); |
963 |
|
|
964 |
if (!direct) { |
if (!direct) { |
965 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
966 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
|
|
|
967 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
968 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
969 |
|
|
970 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
971 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
972 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
973 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
974 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
975 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
976 |
|
} |
977 |
|
else { |
978 |
uv_dx /= 2; |
uv_dx /= 2; |
979 |
uv_dy /= 2; |
uv_dy /= 2; |
980 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
981 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
982 |
} |
} |
983 |
|
} |
984 |
|
|
985 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
986 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
987 |
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]; |
988 |
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]; |
989 |
|
|
990 |
} else { |
} else { |
|
if(dec->quarterpel) { |
|
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
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); |
|
|
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); |
|
|
} else { |
|
991 |
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; |
992 |
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; |
993 |
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; |
994 |
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; |
995 |
|
|
996 |
|
if (dec->quarterpel) { |
997 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
998 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
999 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1000 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1001 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1002 |
|
} |
1003 |
|
else { |
1004 |
|
uv_dx /= 2; |
1005 |
|
uv_dy /= 2; |
1006 |
|
b_uv_dx /= 2; |
1007 |
|
b_uv_dy /= 2; |
1008 |
|
} |
1009 |
} |
} |
1010 |
|
|
1011 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1053 |
|
|
1054 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1055 |
if(!direct) { |
if(!direct) { |
1056 |
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1057 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1058 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1059 |
} else { |
} else { |
1060 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1061 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1062 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1063 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1064 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1065 |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1066 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1067 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1068 |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1069 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1070 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1071 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1072 |
} |
} |
1073 |
} else { |
} else { |
1074 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos, |
1075 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1076 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1077 |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1078 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, |
1079 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1080 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1081 |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1082 |
} |
} |
1083 |
|
|
1084 |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos, |
1085 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1086 |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos, |
1087 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1088 |
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
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->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
1089 |
stop_comp_timer(); |
stop_comp_timer(); |
1090 |
|
|
1091 |
if (cbp) |
if (cbp) |
1092 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
1093 |
} |
} |
1094 |
|
|
1095 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1159 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1160 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1161 |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1162 |
uint32_t intra_dc_threshold; /* fake variable */ |
int32_t intra_dc_threshold; /* fake variable */ |
1163 |
|
|
1164 |
if (check_resync_marker(bs, fcode_max - 1)) { |
if (check_resync_marker(bs, fcode_max - 1)) { |
1165 |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1184 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1185 |
mb->cbp = 0; |
mb->cbp = 0; |
1186 |
mb->mode = MODE_FORWARD; |
mb->mode = MODE_FORWARD; |
1187 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
1188 |
continue; |
continue; |
1189 |
} |
} |
1190 |
|
|
1267 |
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); |
1268 |
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]; |
1269 |
|
|
1270 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0); |
1271 |
break; |
break; |
1272 |
|
|
1273 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1274 |
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); |
1275 |
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]; |
1276 |
|
|
1277 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1); |
1278 |
break; |
break; |
1279 |
|
|
1280 |
default: |
default: |
1324 |
} |
} |
1325 |
} |
} |
1326 |
|
|
|
|
|
1327 |
int |
int |
1328 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1329 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1331 |
|
|
1332 |
Bitstream bs; |
Bitstream bs; |
1333 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1334 |
uint32_t quant = 2; |
uint32_t quant = 2; |
1335 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1336 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1385 |
|
|
1386 |
repeat: |
repeat: |
1387 |
|
|
1388 |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
1389 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1390 |
|
|
1391 |
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=%" |
1392 |
|
#if defined(_MSC_VER) |
1393 |
|
"I64" |
1394 |
|
#else |
1395 |
|
"ll" |
1396 |
|
#endif |
1397 |
|
"i, time_pp=%i, time_bp=%i\n", |
1398 |
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); |
1399 |
|
|
1400 |
if (coding_type == -1) { /* nothing */ |
if (coding_type == -1) { /* nothing */ |
1442 |
} else if (coding_type != B_VOP) { |
} else if (coding_type != B_VOP) { |
1443 |
switch(coding_type) { |
switch(coding_type) { |
1444 |
case I_VOP : |
case I_VOP : |
1445 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1446 |
break; |
break; |
1447 |
case P_VOP : |
case P_VOP : |
1448 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1449 |
fcode_forward, intra_dc_threshold, NULL); |
fcode_forward, intra_dc_threshold, NULL); |
1450 |
break; |
break; |
1451 |
case S_VOP : |
case S_VOP : |
1452 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1453 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1454 |
break; |
break; |
1455 |
case N_VOP : |
case N_VOP : |
1460 |
break; |
break; |
1461 |
} |
} |
1462 |
|
|
|
if (reduced_resolution) { |
|
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
|
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
|
|
16, 0); |
|
|
} |
|
|
|
|
1463 |
/* 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 */ |
1464 |
if (!(dec->low_delay_default && dec->packed_mode)) { |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1465 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1477 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1478 |
dec->is_edged[0] = 0; |
dec->is_edged[0] = 0; |
1479 |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
|
dec->last_reduced_resolution = reduced_resolution; |
|
1480 |
dec->last_coding_type = coding_type; |
dec->last_coding_type = coding_type; |
1481 |
|
|
1482 |
dec->frames++; |
dec->frames++; |
1486 |
|
|
1487 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1488 |
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"); |
1489 |
dec->low_delay = 1; |
dec->low_delay = 0; |
1490 |
} |
} |
1491 |
|
|
1492 |
if (dec->frames < 2) { |
if (dec->frames < 2) { |
1521 |
|
|
1522 |
done : |
done : |
1523 |
|
|
1524 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* if we reach here without outputing anything _and_ |
1525 |
then output the recently decoded frame, or print an error message */ |
the calling application has specified low_delay_default, |
1526 |
|
we *must* output something. |
1527 |
|
this always occurs on the first call to decode() call |
1528 |
|
when bframes are present in the bitstream. it may also |
1529 |
|
occur if no vops were seen in the bitstream |
1530 |
|
|
1531 |
|
if packed_mode is enabled, then we output the recently |
1532 |
|
decoded frame (the very first ivop). otherwise we have |
1533 |
|
nothing to display, and therefore output a black screen. |
1534 |
|
*/ |
1535 |
if (dec->low_delay_default && output == 0) { |
if (dec->low_delay_default && output == 0) { |
1536 |
if (dec->packed_mode && seen_something) { |
if (dec->packed_mode && seen_something) { |
|
/* output the recently decoded frame */ |
|
1537 |
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); |
1538 |
} else { |
} else { |
1539 |
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"); |
|
|
|
|
1540 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1541 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1542 |
} |
} |