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 1.52 2004-04-01 11:11:28 suxen_drol Exp $ |
* $Id: decoder.c,v 1.84 2010-11-12 10:10:40 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" |
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 |
|
|
194 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
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 */ |
198 |
|
dec->ver_id = 1; |
199 |
|
|
200 |
|
if (create->fourcc == ((int)('X')|((int)('V')<<8)| |
201 |
|
((int)('I')<<16)|((int)('D')<<24))) { /* XVID */ |
202 |
|
dec->bs_version = 0; /* Initially assume oldest xvid version */ |
203 |
|
} |
204 |
|
else { |
205 |
|
dec->bs_version = 0xffff; /* Initialize to very high value -> assume bugfree stream */ |
206 |
|
} |
207 |
|
|
208 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
209 |
|
|
210 |
if (dec->fixed_dimensions) |
if (dec->fixed_dimensions) { |
211 |
return decoder_resize(dec); |
int ret = decoder_resize(dec); |
212 |
|
if (ret == XVID_ERR_MEMORY) create->handle = NULL; |
213 |
|
return ret; |
214 |
|
} |
215 |
else |
else |
216 |
return 0; |
return 0; |
217 |
} |
} |
254 |
Bitstream * bs, |
Bitstream * bs, |
255 |
const uint32_t quant, |
const uint32_t quant, |
256 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
257 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
258 |
{ |
{ |
259 |
|
|
260 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
267 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
268 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
269 |
|
|
|
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{ |
|
270 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
271 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
272 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
273 |
|
|
274 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
275 |
|
|
316 |
stop_coding_timer(); |
stop_coding_timer(); |
317 |
|
|
318 |
start_timer(); |
start_timer(); |
319 |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
320 |
stop_prediction_timer(); |
stop_prediction_timer(); |
321 |
|
|
322 |
start_timer(); |
start_timer(); |
328 |
stop_iquant_timer(); |
stop_iquant_timer(); |
329 |
|
|
330 |
start_timer(); |
start_timer(); |
331 |
idct(&data[i * 64]); |
idct((short * const)&data[i * 64]); |
332 |
stop_idct_timer(); |
stop_idct_timer(); |
333 |
|
|
334 |
} |
} |
339 |
} |
} |
340 |
|
|
341 |
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{ |
|
342 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
343 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
344 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
345 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
346 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
347 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
348 |
stop_transfer_timer(); |
stop_transfer_timer(); |
349 |
} |
} |
350 |
|
|
355 |
uint8_t * pY_Cur, |
uint8_t * pY_Cur, |
356 |
uint8_t * pU_Cur, |
uint8_t * pU_Cur, |
357 |
uint8_t * pV_Cur, |
uint8_t * pV_Cur, |
|
const int reduced_resolution, |
|
358 |
const MACROBLOCK * pMB) |
const MACROBLOCK * pMB) |
359 |
{ |
{ |
360 |
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); |
|
361 |
|
|
362 |
int stride = dec->edged_width; |
int stride = dec->edged_width; |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
const int stride2 = stride/2; |
|
363 |
int i; |
int i; |
364 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
365 |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
366 |
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
typedef void (*get_inter_block_function_t)( |
367 |
|
Bitstream * bs, |
368 |
|
int16_t * block, |
369 |
|
int direction, |
370 |
|
const int quant, |
371 |
|
const uint16_t *matrix); |
372 |
|
typedef void (*add_residual_function_t)( |
373 |
|
uint8_t *predicted_block, |
374 |
|
const int16_t *residual, |
375 |
|
int stride); |
376 |
|
|
377 |
|
const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) |
378 |
|
? (get_inter_block_function_t)get_inter_block_h263 |
379 |
|
: (get_inter_block_function_t)get_inter_block_mpeg; |
380 |
|
|
381 |
for (i = 0; i < 6; i++) { |
uint8_t *dst[6]; |
382 |
|
int strides[6]; |
383 |
|
|
|
if (cbp & (1 << (5 - i))) { /* coded */ |
|
384 |
|
|
385 |
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
if (dec->interlacing && pMB->field_dct) { |
386 |
|
dst[0] = pY_Cur; |
387 |
|
dst[1] = pY_Cur + 8; |
388 |
|
dst[2] = pY_Cur + stride; |
389 |
|
dst[3] = dst[2] + 8; |
390 |
|
dst[4] = pU_Cur; |
391 |
|
dst[5] = pV_Cur; |
392 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride*2; |
393 |
|
strides[4] = stride/2; |
394 |
|
strides[5] = stride/2; |
395 |
|
} else { |
396 |
|
dst[0] = pY_Cur; |
397 |
|
dst[1] = pY_Cur + 8; |
398 |
|
dst[2] = pY_Cur + 8*stride; |
399 |
|
dst[3] = dst[2] + 8; |
400 |
|
dst[4] = pU_Cur; |
401 |
|
dst[5] = pV_Cur; |
402 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
403 |
|
strides[4] = stride/2; |
404 |
|
strides[5] = stride/2; |
405 |
|
} |
406 |
|
|
407 |
|
for (i = 0; i < 6; i++) { |
408 |
|
/* Process only coded blocks */ |
409 |
|
if (cbp & (1 << (5 - i))) { |
410 |
|
|
411 |
|
/* Clear the block */ |
412 |
|
memset(&data[0], 0, 64*sizeof(int16_t)); |
413 |
|
|
414 |
|
/* Decode coeffs and dequantize on the fly */ |
415 |
start_timer(); |
start_timer(); |
416 |
get_inter_block(bs, block, direction); |
get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices)); |
417 |
stop_coding_timer(); |
stop_coding_timer(); |
418 |
|
|
419 |
|
/* iDCT */ |
420 |
start_timer(); |
start_timer(); |
421 |
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
idct((short * const)&data[0]); |
422 |
stop_iquant_timer(); |
stop_idct_timer(); |
423 |
|
|
424 |
|
/* Add this residual to the predicted block */ |
425 |
start_timer(); |
start_timer(); |
426 |
idct(&data[i * 64]); |
transfer_16to8add(dst[i], &data[0], strides[i]); |
427 |
stop_idct_timer(); |
stop_transfer_timer(); |
428 |
} |
} |
429 |
} |
} |
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
430 |
} |
} |
431 |
|
|
432 |
start_timer(); |
static void __inline |
433 |
if (reduced_resolution) { |
validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec) |
434 |
if (cbp & 32) |
{ |
435 |
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
/* clip a vector to valid range |
436 |
if (cbp & 16) |
prevents crashes if bitstream is broken |
437 |
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
*/ |
438 |
if (cbp & 8) |
int shift = 5 + dec->quarterpel; |
439 |
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
int xborder_high = (int)(dec->mb_width - x_pos) << shift; |
440 |
if (cbp & 4) |
int xborder_low = (-(int)x_pos-1) << shift; |
441 |
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
int yborder_high = (int)(dec->mb_height - y_pos) << shift; |
442 |
if (cbp & 2) |
int yborder_low = (-(int)y_pos-1) << shift; |
443 |
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
444 |
if (cbp & 1) |
#define CHECK_MV(mv) \ |
445 |
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
do { \ |
446 |
} else { |
if ((mv).x > xborder_high) { \ |
447 |
if (cbp & 32) |
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \ |
448 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
(mv).x = xborder_high; \ |
449 |
if (cbp & 16) |
} else if ((mv).x < xborder_low) { \ |
450 |
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); \ |
451 |
if (cbp & 8) |
(mv).x = xborder_low; \ |
452 |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
} \ |
453 |
if (cbp & 4) |
if ((mv).y > yborder_high) { \ |
454 |
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); \ |
455 |
if (cbp & 2) |
(mv).y = yborder_high; \ |
456 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
} else if ((mv).y < yborder_low) { \ |
457 |
if (cbp & 1) |
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \ |
458 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
(mv).y = yborder_low; \ |
459 |
} |
} \ |
460 |
stop_transfer_timer(); |
} while (0) |
461 |
} |
|
462 |
|
CHECK_MV(mv[0]); |
463 |
|
CHECK_MV(mv[1]); |
464 |
|
CHECK_MV(mv[2]); |
465 |
|
CHECK_MV(mv[3]); |
466 |
|
} |
467 |
|
|
468 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
469 |
|
* So we try to be backward compatible to avoid artifacts */ |
470 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
471 |
|
|
472 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
473 |
static void |
static void |
478 |
const uint32_t cbp, |
const uint32_t cbp, |
479 |
Bitstream * bs, |
Bitstream * bs, |
480 |
const uint32_t rounding, |
const uint32_t rounding, |
481 |
const int reduced_resolution, |
const int ref, |
482 |
const int ref) |
const int bvop) |
483 |
{ |
{ |
484 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
485 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
490 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
491 |
VECTOR mv[4]; /* local copy of mvs */ |
VECTOR mv[4]; /* local copy of mvs */ |
492 |
|
|
|
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 { |
|
493 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
494 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
495 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
496 |
for (i = 0; i < 4; i++) |
for (i = 0; i < 4; i++) |
497 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
498 |
} |
|
499 |
|
validate_vector(mv, x_pos, y_pos, dec); |
500 |
|
|
501 |
start_timer(); |
start_timer(); |
502 |
|
|
503 |
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 */ |
504 |
|
|
505 |
uv_dx = mv[0].x; |
uv_dx = mv[0].x; |
506 |
uv_dy = mv[0].y; |
uv_dy = mv[0].y; |
507 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
508 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
509 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
510 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
511 |
|
} |
512 |
|
else { |
513 |
uv_dx /= 2; |
uv_dx /= 2; |
514 |
uv_dy /= 2; |
uv_dy /= 2; |
515 |
} |
} |
516 |
|
} |
517 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
518 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
519 |
|
|
520 |
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) |
|
521 |
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, |
522 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
523 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
528 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
529 |
|
|
530 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
531 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
532 |
|
int z; |
533 |
|
uv_dx = 0; uv_dy = 0; |
534 |
|
for (z = 0; z < 4; z++) { |
535 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
536 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
537 |
|
} |
538 |
|
} |
539 |
|
else { |
540 |
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); |
541 |
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); |
542 |
|
} |
543 |
} else { |
} else { |
544 |
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; |
545 |
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; |
548 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
549 |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
550 |
|
|
551 |
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) { |
|
552 |
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, |
553 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
554 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
574 |
} |
} |
575 |
|
|
576 |
/* 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 { |
|
577 |
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, |
578 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
579 |
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, |
580 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
581 |
|
|
582 |
|
stop_comp_timer(); |
583 |
|
|
584 |
|
if (cbp) |
585 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
586 |
|
} |
587 |
|
|
588 |
|
/* decode an inter macroblock in field mode */ |
589 |
|
static void |
590 |
|
decoder_mbinter_field(DECODER * dec, |
591 |
|
const MACROBLOCK * pMB, |
592 |
|
const uint32_t x_pos, |
593 |
|
const uint32_t y_pos, |
594 |
|
const uint32_t cbp, |
595 |
|
Bitstream * bs, |
596 |
|
const uint32_t rounding, |
597 |
|
const int ref, |
598 |
|
const int bvop) |
599 |
|
{ |
600 |
|
uint32_t stride = dec->edged_width; |
601 |
|
uint32_t stride2 = stride / 2; |
602 |
|
|
603 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
604 |
|
|
605 |
|
int uvtop_dx, uvtop_dy; |
606 |
|
int uvbot_dx, uvbot_dy; |
607 |
|
VECTOR mv[4]; /* local copy of mvs */ |
608 |
|
|
609 |
|
/* Get pointer to memory areas */ |
610 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
611 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
612 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
613 |
|
|
614 |
|
mv[0] = pMB->mvs[0]; |
615 |
|
mv[1] = pMB->mvs[1]; |
616 |
|
memset(&mv[2],0,2*sizeof(VECTOR)); |
617 |
|
|
618 |
|
validate_vector(mv, x_pos, y_pos, dec); |
619 |
|
|
620 |
|
start_timer(); |
621 |
|
|
622 |
|
if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
623 |
|
{ |
624 |
|
/* Prepare top field vector */ |
625 |
|
uvtop_dx = DIV2ROUND(mv[0].x); |
626 |
|
uvtop_dy = DIV2ROUND(mv[0].y); |
627 |
|
|
628 |
|
/* Prepare bottom field vector */ |
629 |
|
uvbot_dx = DIV2ROUND(mv[1].x); |
630 |
|
uvbot_dy = DIV2ROUND(mv[1].y); |
631 |
|
|
632 |
|
if(dec->quarterpel) |
633 |
|
{ |
634 |
|
/* NOT supported */ |
635 |
|
} |
636 |
|
else |
637 |
|
{ |
638 |
|
/* Interpolate top field left part(we use double stride for every 2nd line) */ |
639 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
640 |
|
16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
641 |
|
/* top field right part */ |
642 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
643 |
|
16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
644 |
|
|
645 |
|
/* Interpolate bottom field left part(we use double stride for every 2nd line) */ |
646 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
647 |
|
16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
648 |
|
/* Bottom field right part */ |
649 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
650 |
|
16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
651 |
|
|
652 |
|
/* Interpolate field1 U */ |
653 |
|
interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2, |
654 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
655 |
|
|
656 |
|
/* Interpolate field1 V */ |
657 |
|
interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2, |
658 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
659 |
|
|
660 |
|
/* Interpolate field2 U */ |
661 |
|
interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2, |
662 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
663 |
|
|
664 |
|
/* Interpolate field2 V */ |
665 |
|
interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2, |
666 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
667 |
|
} |
668 |
|
} |
669 |
|
else |
670 |
|
{ |
671 |
|
/* We don't expect 4 motion vectors in interlaced mode */ |
672 |
} |
} |
673 |
|
|
674 |
stop_comp_timer(); |
stop_comp_timer(); |
675 |
|
|
676 |
|
/* Must add error correction? */ |
677 |
if (cbp) |
if (cbp) |
678 |
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); |
|
679 |
} |
} |
680 |
|
|
681 |
static void |
static void |
722 |
stop_transfer_timer(); |
stop_transfer_timer(); |
723 |
|
|
724 |
if (cbp) |
if (cbp) |
725 |
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); |
726 |
|
|
727 |
} |
} |
728 |
|
|
730 |
static void |
static void |
731 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
732 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
733 |
int quant, |
int quant, |
734 |
int intra_dc_threshold) |
int intra_dc_threshold) |
735 |
{ |
{ |
736 |
uint32_t bound; |
uint32_t bound; |
737 |
uint32_t x, y; |
uint32_t x, y; |
738 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
739 |
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; |
|
|
} |
|
740 |
|
|
741 |
bound = 0; |
bound = 0; |
742 |
|
|
757 |
bound = read_video_packet_header(bs, dec, 0, |
bound = read_video_packet_header(bs, dec, 0, |
758 |
&quant, NULL, NULL, &intra_dc_threshold); |
&quant, NULL, NULL, &intra_dc_threshold); |
759 |
x = bound % mb_width; |
x = bound % mb_width; |
760 |
y = bound / mb_width; |
y = MIN((bound / mb_width), (mb_height-1)); |
761 |
} |
} |
762 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
763 |
|
|
792 |
} |
} |
793 |
|
|
794 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
795 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
796 |
|
|
797 |
} |
} |
798 |
if(dec->out_frm) |
if(dec->out_frm) |
841 |
mv.y -= range; |
mv.y -= range; |
842 |
} |
} |
843 |
|
|
844 |
/* clip to valid range */ |
ret_mv->x = mv.x; |
845 |
|
ret_mv->y = mv.y; |
846 |
|
} |
847 |
|
|
848 |
|
/* We use this when decoder runs interlaced -> different prediction */ |
849 |
|
|
850 |
|
static void get_motion_vector_interlaced(DECODER * dec, |
851 |
|
Bitstream * bs, |
852 |
|
int x, |
853 |
|
int y, |
854 |
|
int k, |
855 |
|
MACROBLOCK *pMB, |
856 |
|
int fcode, |
857 |
|
const int bound) |
858 |
|
{ |
859 |
|
const int scale_fac = 1 << (fcode - 1); |
860 |
|
const int high = (32 * scale_fac) - 1; |
861 |
|
const int low = ((-32) * scale_fac); |
862 |
|
const int range = (64 * scale_fac); |
863 |
|
|
864 |
if (mv.x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) |
/* Get interlaced prediction */ |
865 |
mv.x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); |
const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k); |
866 |
|
VECTOR mv,mvf1,mvf2; |
867 |
|
|
868 |
else if (mv.x < (int)(-x-1) << (5 + dec->quarterpel)) |
if(!pMB->field_pred) |
869 |
mv.x = (int)(-x-1) << (5 + dec->quarterpel); |
{ |
870 |
|
mv.x = get_mv(bs,fcode); |
871 |
|
mv.y = get_mv(bs,fcode); |
872 |
|
|
873 |
if (mv.y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) |
mv.x += pmv.x; |
874 |
mv.y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); |
mv.y += pmv.y; |
875 |
|
|
876 |
else if (mv.y < ((int)(-y-1)) << (5 + dec->quarterpel) ) |
if(mv.x<low) { |
877 |
mv.y = (int)(-y-1) << (5 + dec->quarterpel); |
mv.x += range; |
878 |
|
} else if (mv.x>high) { |
879 |
|
mv.x-=range; |
880 |
|
} |
881 |
|
|
882 |
ret_mv->x = mv.x; |
if (mv.y < low) { |
883 |
ret_mv->y = mv.y; |
mv.y += range; |
884 |
|
} else if (mv.y > high) { |
885 |
|
mv.y -= range; |
886 |
|
} |
887 |
|
|
888 |
|
pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv; |
889 |
|
} |
890 |
|
else |
891 |
|
{ |
892 |
|
mvf1.x = get_mv(bs, fcode); |
893 |
|
mvf1.y = get_mv(bs, fcode); |
894 |
|
|
895 |
|
mvf1.x += pmv.x; |
896 |
|
mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */ |
897 |
|
|
898 |
|
if (mvf1.x < low) { |
899 |
|
mvf1.x += range; |
900 |
|
} else if (mvf1.x > high) { |
901 |
|
mvf1.x -= range; |
902 |
|
} |
903 |
|
|
904 |
|
if (mvf1.y < low) { |
905 |
|
mvf1.y += range; |
906 |
|
} else if (mvf1.y > high) { |
907 |
|
mvf1.y -= range; |
908 |
|
} |
909 |
|
|
910 |
|
mvf2.x = get_mv(bs, fcode); |
911 |
|
mvf2.y = get_mv(bs, fcode); |
912 |
|
|
913 |
|
mvf2.x += pmv.x; |
914 |
|
mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */ |
915 |
|
|
916 |
|
if (mvf2.x < low) { |
917 |
|
mvf2.x += range; |
918 |
|
} else if (mvf2.x > high) { |
919 |
|
mvf2.x -= range; |
920 |
|
} |
921 |
|
|
922 |
|
if (mvf2.y < low) { |
923 |
|
mvf2.y += range; |
924 |
|
} else if (mvf2.y > high) { |
925 |
|
mvf2.y -= range; |
926 |
|
} |
927 |
|
|
928 |
|
pMB->mvs[0]=mvf1; |
929 |
|
pMB->mvs[1]=mvf2; |
930 |
|
pMB->mvs[2].x=pMB->mvs[3].x=0; |
931 |
|
pMB->mvs[2].y=pMB->mvs[3].y=0; |
932 |
|
|
933 |
|
/* Calculate average for as it is field predicted */ |
934 |
|
pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x); |
935 |
|
pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y); |
936 |
|
} |
937 |
} |
} |
938 |
|
|
939 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
941 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
942 |
Bitstream * bs, |
Bitstream * bs, |
943 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
944 |
int quant, |
int quant, |
945 |
int fcode, |
int fcode, |
946 |
int intra_dc_threshold, |
int intra_dc_threshold, |
949 |
uint32_t x, y; |
uint32_t x, y; |
950 |
uint32_t bound; |
uint32_t bound; |
951 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
952 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
953 |
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; |
|
|
} |
|
954 |
|
|
955 |
|
if (!dec->is_edged[0]) { |
956 |
start_timer(); |
start_timer(); |
957 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
958 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
959 |
|
dec->is_edged[0] = 1; |
960 |
stop_edges_timer(); |
stop_edges_timer(); |
961 |
|
} |
962 |
|
|
963 |
if (gmc_warp) { |
if (gmc_warp) { |
964 |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
984 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
985 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
986 |
x = bound % mb_width; |
x = bound % mb_width; |
987 |
y = bound / mb_width; |
y = MIN((bound / mb_width), (mb_height-1)); |
988 |
} |
} |
989 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
990 |
|
|
1028 |
} |
} |
1029 |
mb->quant = quant; |
mb->quant = quant; |
1030 |
|
|
1031 |
|
mb->field_pred=0; |
1032 |
if (dec->interlacing) { |
if (dec->interlacing) { |
1033 |
if (cbp || intra) { |
if (cbp || intra) { |
1034 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
1054 |
|
|
1055 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
1056 |
|
|
1057 |
if (dec->interlacing && mb->field_pred) { |
if(dec->interlacing) { |
1058 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
/* Get motion vectors interlaced, field_pred is handled there */ |
1059 |
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); |
1060 |
} else { |
} else { |
1061 |
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); |
1062 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1063 |
} |
} |
1064 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
1065 |
|
/* interlaced missing here */ |
1066 |
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); |
1067 |
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); |
1068 |
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); |
1071 |
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; |
1072 |
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; |
1073 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
1074 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
1075 |
continue; |
continue; |
1076 |
} |
} |
1077 |
|
|
1078 |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
/* See how to decode */ |
1079 |
rounding, reduced_resolution, 0); |
if(!mb->field_pred) |
1080 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1081 |
|
else |
1082 |
|
decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1083 |
|
|
1084 |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
1085 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
1097 |
|
|
1098 |
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; |
1099 |
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; |
1100 |
|
mb->field_pred=0; /* (!) */ |
1101 |
|
|
1102 |
decoder_mbinter(dec, mb, x, y, 0, bs, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
1103 |
rounding, reduced_resolution, 0); |
rounding, 0, 0); |
1104 |
|
|
1105 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1106 |
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); |
1146 |
else if (mv_y > high) |
else if (mv_y > high) |
1147 |
mv_y -= range; |
mv_y -= range; |
1148 |
|
|
|
|
|
|
/* clip to valid range */ |
|
|
if (mv_x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) |
|
|
mv_x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); |
|
|
|
|
|
else if (mv_x < (int)(-x-1) << (5 + dec->quarterpel)) |
|
|
mv_x = (int)(-x-1) << (5 + dec->quarterpel); |
|
|
|
|
|
if (mv_y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) |
|
|
mv_y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); |
|
|
|
|
|
else if (mv_y < ((int)(-y-1)) << (5 + dec->quarterpel) ) |
|
|
mv_y = (int)(-y-1) << (5 + dec->quarterpel); |
|
|
|
|
1149 |
mv->x = mv_x; |
mv->x = mv_x; |
1150 |
mv->y = mv_y; |
mv->y = mv_y; |
1151 |
} |
} |
1155 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1156 |
IMAGE forward, |
IMAGE forward, |
1157 |
IMAGE backward, |
IMAGE backward, |
1158 |
const MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
1159 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1160 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1161 |
Bitstream * bs, |
Bitstream * bs, |
1172 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1173 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1174 |
|
|
1175 |
|
validate_vector(pMB->mvs, x_pos, y_pos, dec); |
1176 |
|
validate_vector(pMB->b_mvs, x_pos, y_pos, dec); |
1177 |
|
|
1178 |
if (!direct) { |
if (!direct) { |
1179 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1180 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
|
|
|
1181 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1182 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1183 |
|
|
1184 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
1185 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1186 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
1187 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1188 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1189 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1190 |
|
} |
1191 |
|
else { |
1192 |
uv_dx /= 2; |
uv_dx /= 2; |
1193 |
uv_dy /= 2; |
uv_dy /= 2; |
1194 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1195 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1196 |
} |
} |
1197 |
|
} |
1198 |
|
|
1199 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1200 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
1201 |
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]; |
1202 |
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]; |
1203 |
|
|
1204 |
} else { |
} else { |
1205 |
if(dec->quarterpel) { |
if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */ |
1206 |
|
/* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */ |
1207 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1208 |
|
int z; |
1209 |
|
uv_dx = 0; uv_dy = 0; |
1210 |
|
b_uv_dx = 0; b_uv_dy = 0; |
1211 |
|
for (z = 0; z < 4; z++) { |
1212 |
|
uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1)); |
1213 |
|
uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1)); |
1214 |
|
b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1)); |
1215 |
|
b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1)); |
1216 |
|
} |
1217 |
|
} |
1218 |
|
else { |
1219 |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1220 |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1221 |
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_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); |
1222 |
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); |
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); |
1223 |
|
} |
1224 |
} else { |
} else { |
1225 |
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; |
1226 |
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; |
1273 |
|
|
1274 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1275 |
if(!direct) { |
if(!direct) { |
1276 |
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, |
1277 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1278 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1279 |
} else { |
} else { |
1280 |
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, |
1281 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1282 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1283 |
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, |
1284 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1285 |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1286 |
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, |
1287 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1288 |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1289 |
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, |
1290 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1291 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1292 |
} |
} |
1293 |
} else { |
} else { |
1294 |
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, |
1295 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1296 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1297 |
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); |
1298 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, |
1299 |
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); |
1300 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1301 |
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); |
1302 |
} |
} |
1303 |
|
|
1304 |
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, |
1305 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1306 |
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, |
1307 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1308 |
|
|
|
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); |
|
|
|
|
1309 |
stop_comp_timer(); |
stop_comp_timer(); |
1310 |
|
|
1311 |
if (cbp) |
if (cbp) |
1312 |
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); |
1313 |
} |
} |
1314 |
|
|
1315 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1344 |
return -1; |
return -1; |
1345 |
} |
} |
1346 |
|
|
1347 |
|
static int __inline get_resync_len_b(const int fcode_backward, |
1348 |
|
const int fcode_forward) { |
1349 |
|
int resync_len = ((fcode_forward>fcode_backward) ? fcode_forward : fcode_backward) - 1; |
1350 |
|
if (resync_len < 1) resync_len = 1; |
1351 |
|
return resync_len; |
1352 |
|
} |
1353 |
|
|
1354 |
static void |
static void |
1355 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1356 |
Bitstream * bs, |
Bitstream * bs, |
1361 |
uint32_t x, y; |
uint32_t x, y; |
1362 |
VECTOR mv; |
VECTOR mv; |
1363 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
1364 |
int i; |
int i; |
1365 |
|
int resync_len; |
1366 |
|
|
1367 |
|
if (!dec->is_edged[0]) { |
1368 |
start_timer(); |
start_timer(); |
1369 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1370 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
1371 |
|
dec->is_edged[0] = 1; |
1372 |
|
stop_edges_timer(); |
1373 |
|
} |
1374 |
|
|
1375 |
|
if (!dec->is_edged[1]) { |
1376 |
|
start_timer(); |
1377 |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1378 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
1379 |
|
dec->is_edged[1] = 1; |
1380 |
stop_edges_timer(); |
stop_edges_timer(); |
1381 |
|
} |
1382 |
|
|
1383 |
|
resync_len = get_resync_len_b(fcode_backward, fcode_forward); |
1384 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1385 |
/* Initialize Pred Motion Vector */ |
/* Initialize Pred Motion Vector */ |
1386 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1387 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1388 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1389 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1390 |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
int intra_dc_threshold; /* fake variable */ |
|
uint32_t intra_dc_threshold; /* fake variable */ |
|
1391 |
|
|
1392 |
if (check_resync_marker(bs, fcode_max - 1)) { |
if (check_resync_marker(bs, resync_len)) { |
1393 |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
int bound = read_video_packet_header(bs, dec, resync_len, &quant, |
1394 |
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1395 |
x = bound % dec->mb_width; |
x = bound % dec->mb_width; |
1396 |
y = bound / dec->mb_width; |
y = MIN((bound / dec->mb_width), (dec->mb_height-1)); |
1397 |
/* reset predicted macroblocks */ |
/* reset predicted macroblocks */ |
1398 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1399 |
|
/* update resync len with new fcodes */ |
1400 |
|
resync_len = get_resync_len_b(fcode_backward, fcode_forward); |
1401 |
} |
} |
1402 |
|
|
1403 |
mv = |
mv = |
1414 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1415 |
mb->cbp = 0; |
mb->cbp = 0; |
1416 |
mb->mode = MODE_FORWARD; |
mb->mode = MODE_FORWARD; |
1417 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1418 |
continue; |
continue; |
1419 |
} |
} |
1420 |
|
|
1467 |
|
|
1468 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1469 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1470 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; |
1471 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; |
1472 |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
|
1473 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
mb->b_mvs[i].x = (mv.x) |
1474 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
? mb->mvs[i].x - last_mb->mvs[i].x |
1475 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1476 |
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
mb->b_mvs[i].y = (mv.y) |
1477 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
? mb->mvs[i].y - last_mb->mvs[i].y |
1478 |
|
: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1479 |
} |
} |
1480 |
|
|
1481 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1497 |
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); |
1498 |
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]; |
1499 |
|
|
1500 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1501 |
break; |
break; |
1502 |
|
|
1503 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1504 |
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); |
1505 |
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]; |
1506 |
|
|
1507 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1508 |
break; |
break; |
1509 |
|
|
1510 |
default: |
default: |
1515 |
} |
} |
1516 |
|
|
1517 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1518 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1519 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1520 |
int coding_type, int quant) |
int coding_type, int quant) |
1521 |
{ |
{ |
1522 |
|
const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; |
1523 |
|
|
1524 |
if (dec->cartoon_mode) |
if (dec->cartoon_mode) |
1525 |
frame->general &= ~XVID_FILMEFFECT; |
frame->general &= ~XVID_FILMEFFECT; |
1526 |
|
|
1527 |
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || frame->brightness!=0) |
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) |
1528 |
&& mbs != NULL) /* post process */ |
&& mbs != NULL) /* post process */ |
1529 |
{ |
{ |
1530 |
/* note: image is stored to tmp */ |
/* note: image is stored to tmp */ |
1531 |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1532 |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1533 |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1534 |
frame->general, frame->brightness, dec->frames, (coding_type == B_VOP)); |
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
1535 |
img = &dec->tmp; |
img = &dec->tmp; |
1536 |
} |
} |
1537 |
|
|
1546 |
stats->data.vop.qscale_stride = dec->mb_width; |
stats->data.vop.qscale_stride = dec->mb_width; |
1547 |
stats->data.vop.qscale = dec->qscale; |
stats->data.vop.qscale = dec->qscale; |
1548 |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1549 |
int i; |
unsigned int i; |
1550 |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1551 |
stats->data.vop.qscale[i] = mbs[i].quant; |
stats->data.vop.qscale[i] = mbs[i].quant; |
1552 |
} else |
} else |
1554 |
} |
} |
1555 |
} |
} |
1556 |
|
|
|
|
|
1557 |
int |
int |
1558 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1559 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1561 |
|
|
1562 |
Bitstream bs; |
Bitstream bs; |
1563 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1564 |
uint32_t quant = 2; |
uint32_t quant = 2; |
1565 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1566 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1615 |
|
|
1616 |
repeat: |
repeat: |
1617 |
|
|
1618 |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
1619 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1620 |
|
|
1621 |
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=%" |
1622 |
|
#if defined(_MSC_VER) |
1623 |
|
"I64" |
1624 |
|
#else |
1625 |
|
"ll" |
1626 |
|
#endif |
1627 |
|
"i, time_pp=%i, time_bp=%i\n", |
1628 |
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); |
1629 |
|
|
1630 |
if (coding_type == -1) { /* nothing */ |
if (coding_type == -1) { /* nothing */ |
1637 |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1638 |
|
|
1639 |
if (coding_type == -3) |
if (coding_type == -3) |
1640 |
decoder_resize(dec); |
if (decoder_resize(dec)) return XVID_ERR_MEMORY; |
1641 |
|
|
1642 |
if (stats) { |
if (stats) { |
1643 |
stats->type = XVID_TYPE_VOL; |
stats->type = XVID_TYPE_VOL; |
1655 |
goto repeat; |
goto repeat; |
1656 |
} |
} |
1657 |
|
|
1658 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
if(dec->frames == 0 && coding_type != I_VOP) { |
1659 |
|
/* 1st frame is not an i-vop */ |
1660 |
|
goto repeat; |
1661 |
|
} |
1662 |
|
|
1663 |
|
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.x = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1664 |
|
|
1665 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1666 |
if (dec->packed_mode && coding_type == N_VOP) { |
if (dec->packed_mode && coding_type == N_VOP) { |
1672 |
} else if (coding_type != B_VOP) { |
} else if (coding_type != B_VOP) { |
1673 |
switch(coding_type) { |
switch(coding_type) { |
1674 |
case I_VOP : |
case I_VOP : |
1675 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1676 |
break; |
break; |
1677 |
case P_VOP : |
case P_VOP : |
1678 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1679 |
fcode_forward, intra_dc_threshold, NULL); |
fcode_forward, intra_dc_threshold, NULL); |
1680 |
break; |
break; |
1681 |
case S_VOP : |
case S_VOP : |
1682 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1683 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1684 |
break; |
break; |
1685 |
case N_VOP : |
case N_VOP : |
1690 |
break; |
break; |
1691 |
} |
} |
1692 |
|
|
|
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); |
|
|
} |
|
|
|
|
1693 |
/* 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 */ |
1694 |
if (!(dec->low_delay_default && dec->packed_mode)) { |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1695 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1703 |
} |
} |
1704 |
|
|
1705 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1706 |
|
dec->is_edged[1] = dec->is_edged[0]; |
1707 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1708 |
|
dec->is_edged[0] = 0; |
1709 |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
|
dec->last_reduced_resolution = reduced_resolution; |
|
1710 |
dec->last_coding_type = coding_type; |
dec->last_coding_type = coding_type; |
1711 |
|
|
1712 |
dec->frames++; |
dec->frames++; |
1716 |
|
|
1717 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1718 |
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"); |
1719 |
dec->low_delay = 1; |
dec->low_delay = 0; |
1720 |
} |
} |
1721 |
|
|
1722 |
if (dec->frames < 2) { |
if (dec->frames < 2) { |
1751 |
|
|
1752 |
done : |
done : |
1753 |
|
|
1754 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* if we reach here without outputing anything _and_ |
1755 |
then output the recently decoded frame, or print an error message */ |
the calling application has specified low_delay_default, |
1756 |
|
we *must* output something. |
1757 |
|
this always occurs on the first call to decode() call |
1758 |
|
when bframes are present in the bitstream. it may also |
1759 |
|
occur if no vops were seen in the bitstream |
1760 |
|
|
1761 |
|
if packed_mode is enabled, then we output the recently |
1762 |
|
decoded frame (the very first ivop). otherwise we have |
1763 |
|
nothing to display, and therefore output a black screen. |
1764 |
|
*/ |
1765 |
if (dec->low_delay_default && output == 0) { |
if (dec->low_delay_default && output == 0) { |
1766 |
if (dec->packed_mode && seen_something) { |
if (dec->packed_mode && seen_something) { |
|
/* output the recently decoded frame */ |
|
1767 |
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); |
1768 |
} else { |
} else { |
1769 |
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"); |
|
|
|
|
1770 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1771 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1772 |
} |
} |