32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
36 |
|
* MinChen <chenm001@163.com> |
37 |
|
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
38 |
|
* Fix a little bug for low_delay flage |
39 |
|
* MinChen <chenm001@163.com> |
40 |
|
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
41 |
|
* 22.06.2002 added primative N_VOP support |
42 |
|
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
43 |
|
* 08.05.2002 add low_delay support for B_VOP decode |
44 |
|
* MinChen <chenm001@163.com> |
45 |
|
* 05.05.2002 fix some B-frame decode problem |
46 |
* 02.05.2002 add B-frame decode support(have some problem); |
* 02.05.2002 add B-frame decode support(have some problem); |
47 |
* MinChen <chenm001@163.com> |
* MinChen <chenm001@163.com> |
48 |
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
62 |
#include <stdlib.h> |
#include <stdlib.h> |
63 |
#include <string.h> |
#include <string.h> |
64 |
|
|
65 |
|
#ifdef BFRAMES_DEC_DEBUG |
66 |
|
#define BFRAMES_DEC |
67 |
|
#endif |
68 |
|
|
69 |
#include "xvid.h" |
#include "xvid.h" |
70 |
#include "portab.h" |
#include "portab.h" |
71 |
|
|
89 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
90 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
91 |
|
|
92 |
int decoder_create(XVID_DEC_PARAM * param) |
int |
93 |
|
decoder_create(XVID_DEC_PARAM * param) |
94 |
{ |
{ |
95 |
DECODER * dec; |
DECODER * dec; |
96 |
|
|
97 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
98 |
if (dec == NULL) |
if (dec == NULL) { |
|
{ |
|
99 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
100 |
} |
} |
101 |
param->handle = dec; |
param->handle = dec; |
108 |
|
|
109 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
110 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
111 |
|
dec->low_delay = 0; |
112 |
|
|
113 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
114 |
xvid_free(dec); |
xvid_free(dec); |
115 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
116 |
} |
} |
117 |
|
|
118 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
119 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
120 |
xvid_free(dec); |
xvid_free(dec); |
121 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
122 |
} |
} |
123 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
124 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
125 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
{ |
|
126 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
127 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
128 |
xvid_free(dec); |
xvid_free(dec); |
129 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
130 |
} |
} |
131 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
|
{ |
|
132 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
133 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
134 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
136 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
137 |
} |
} |
138 |
|
|
139 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
dec->mbs = |
140 |
if (dec->mbs == NULL) |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
141 |
{ |
CACHE_LINE); |
142 |
|
if (dec->mbs == NULL) { |
143 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
144 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
145 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
147 |
xvid_free(dec); |
xvid_free(dec); |
148 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
149 |
} |
} |
150 |
|
|
151 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
152 |
|
|
153 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
154 |
// for skip MB flag |
// for skip MB flag |
155 |
dec->last_mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
dec->last_mbs = |
156 |
if (dec->last_mbs == NULL) |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
157 |
{ |
CACHE_LINE); |
158 |
|
if (dec->last_mbs == NULL) { |
159 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
160 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
161 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
165 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
166 |
} |
} |
167 |
|
|
168 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
169 |
|
|
170 |
init_timer(); |
init_timer(); |
171 |
|
|
172 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
178 |
} |
} |
179 |
|
|
180 |
|
|
181 |
int decoder_destroy(DECODER * dec) |
int |
182 |
|
decoder_destroy(DECODER * dec) |
183 |
{ |
{ |
184 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
185 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
195 |
|
|
196 |
|
|
197 |
|
|
198 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
199 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
200 |
}; |
}; |
201 |
|
|
204 |
|
|
205 |
// decode an intra macroblock |
// decode an intra macroblock |
206 |
|
|
207 |
void decoder_mbintra(DECODER * dec, |
void |
208 |
|
decoder_mbintra(DECODER * dec, |
209 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
210 |
const uint32_t x_pos, |
const uint32_t x_pos, |
211 |
const uint32_t y_pos, |
const uint32_t y_pos, |
213 |
const uint32_t cbp, |
const uint32_t cbp, |
214 |
Bitstream * bs, |
Bitstream * bs, |
215 |
const uint32_t quant, |
const uint32_t quant, |
216 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
217 |
|
const unsigned int bound) |
218 |
{ |
{ |
219 |
|
|
220 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
233 |
|
|
234 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
235 |
|
|
236 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
237 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
238 |
int16_t predictors[8]; |
int16_t predictors[8]; |
239 |
int start_coeff; |
int start_coeff; |
240 |
|
|
241 |
start_timer(); |
start_timer(); |
242 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
243 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
244 |
{ |
if (!acpred_flag) { |
245 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
246 |
} |
} |
247 |
stop_prediction_timer(); |
stop_prediction_timer(); |
248 |
|
|
249 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
250 |
int dc_size; |
int dc_size; |
251 |
int dc_dif; |
int dc_dif; |
252 |
|
|
253 |
dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
dc_size = i < 4 ? get_dc_size_lum(bs) : get_dc_size_chrom(bs); |
254 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
255 |
|
|
256 |
if (dc_size > 8) |
if (dc_size > 8) { |
|
{ |
|
257 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
258 |
} |
} |
259 |
|
|
260 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
261 |
start_coeff = 1; |
start_coeff = 1; |
262 |
} |
|
263 |
else |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
264 |
{ |
} else { |
265 |
start_coeff = 0; |
start_coeff = 0; |
266 |
} |
} |
267 |
|
|
268 |
start_timer(); |
start_timer(); |
269 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
270 |
{ |
{ |
271 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
272 |
|
2 : pMB->acpred_directions[i]; |
273 |
|
|
274 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
275 |
} |
} |
276 |
stop_coding_timer(); |
stop_coding_timer(); |
277 |
|
|
280 |
stop_prediction_timer(); |
stop_prediction_timer(); |
281 |
|
|
282 |
start_timer(); |
start_timer(); |
283 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
284 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
285 |
} |
} else { |
|
else |
|
|
{ |
|
286 |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
287 |
} |
} |
288 |
stop_iquant_timer(); |
stop_iquant_timer(); |
292 |
stop_idct_timer(); |
stop_idct_timer(); |
293 |
} |
} |
294 |
|
|
295 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
296 |
next_block = stride; |
next_block = stride; |
297 |
stride *= 2; |
stride *= 2; |
298 |
} |
} |
319 |
|
|
320 |
// decode an inter macroblock |
// decode an inter macroblock |
321 |
|
|
322 |
void decoder_mbinter(DECODER * dec, |
void |
323 |
|
decoder_mbinter(DECODER * dec, |
324 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
325 |
const uint32_t x_pos, |
const uint32_t x_pos, |
326 |
const uint32_t y_pos, |
const uint32_t y_pos, |
346 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
347 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
348 |
|
|
349 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
{ |
|
350 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
351 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
352 |
|
|
353 |
|
if (dec->quarterpel) |
354 |
|
{ |
355 |
|
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
356 |
|
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
357 |
|
} |
358 |
|
|
359 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
360 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
361 |
} |
} else { |
|
else |
|
|
{ |
|
362 |
int sum; |
int sum; |
363 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
364 |
|
|
365 |
|
if (dec->quarterpel) |
366 |
|
{ |
367 |
|
sum /= 2; |
368 |
|
} |
369 |
|
|
370 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
371 |
|
|
372 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
373 |
|
|
374 |
|
if (dec->quarterpel) |
375 |
|
{ |
376 |
|
sum /= 2; |
377 |
|
} |
378 |
|
|
379 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
380 |
} |
} |
381 |
|
|
382 |
start_timer(); |
start_timer(); |
383 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
if(dec->quarterpel) { |
384 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
DEBUG("QUARTERPEL"); |
385 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
386 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
387 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
388 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
389 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
390 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
391 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
392 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
393 |
|
} |
394 |
|
else { |
395 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
396 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
397 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
398 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
399 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
400 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
401 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
402 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
403 |
|
} |
404 |
|
|
405 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
406 |
|
uv_dx, uv_dy, stride2, rounding); |
407 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
408 |
|
uv_dx, uv_dy, stride2, rounding); |
409 |
stop_comp_timer(); |
stop_comp_timer(); |
410 |
|
|
411 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
412 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
413 |
|
|
414 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
415 |
{ |
{ |
416 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
417 |
|
|
418 |
start_timer(); |
start_timer(); |
419 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
420 |
stop_coding_timer(); |
stop_coding_timer(); |
421 |
|
|
422 |
start_timer(); |
start_timer(); |
423 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
424 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
425 |
} |
} else { |
|
else |
|
|
{ |
|
426 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
427 |
} |
} |
428 |
stop_iquant_timer(); |
stop_iquant_timer(); |
433 |
} |
} |
434 |
} |
} |
435 |
|
|
436 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
437 |
next_block = stride; |
next_block = stride; |
438 |
stride *= 2; |
stride *= 2; |
439 |
} |
} |
455 |
} |
} |
456 |
|
|
457 |
|
|
458 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
459 |
|
decoder_iframe(DECODER * dec, |
460 |
|
Bitstream * bs, |
461 |
|
int quant, |
462 |
|
int intra_dc_threshold) |
463 |
{ |
{ |
464 |
|
uint32_t bound; |
465 |
uint32_t x, y; |
uint32_t x, y; |
466 |
|
|
467 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
|
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
|
468 |
|
|
469 |
|
for (y = 0; y < dec->mb_height; y++) { |
470 |
|
for (x = 0; x < dec->mb_width; x++) { |
471 |
|
MACROBLOCK *mb; |
472 |
uint32_t mcbpc; |
uint32_t mcbpc; |
473 |
uint32_t cbpc; |
uint32_t cbpc; |
474 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
475 |
uint32_t cbpy; |
uint32_t cbpy; |
476 |
uint32_t cbp; |
uint32_t cbp; |
477 |
|
|
478 |
|
while (BitstreamShowBits(bs, 9) == 1) |
479 |
|
BitstreamSkip(bs, 9); |
480 |
|
|
481 |
|
if (check_resync_marker(bs, 0)) |
482 |
|
{ |
483 |
|
bound = read_video_packet_header(bs, 0, &quant); |
484 |
|
x = bound % dec->mb_width; |
485 |
|
y = bound / dec->mb_width; |
486 |
|
} |
487 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
488 |
|
|
489 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
490 |
|
|
491 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
492 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
493 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
494 |
|
|
495 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
496 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
497 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
498 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
499 |
|
|
500 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
501 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
502 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
503 |
quant = 31; |
quant = 31; |
504 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
505 |
quant = 1; |
quant = 1; |
506 |
} |
} |
507 |
} |
} |
508 |
mb->quant = quant; |
mb->quant = quant; |
509 |
|
mb->mvs[0].x = mb->mvs[0].y = |
510 |
|
mb->mvs[1].x = mb->mvs[1].y = |
511 |
|
mb->mvs[2].x = mb->mvs[2].y = |
512 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
513 |
|
|
514 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
515 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
516 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DEBUG1("deci: field_dct: ", mb->field_dct); |
517 |
} |
} |
518 |
|
|
519 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
520 |
|
intra_dc_threshold, bound); |
521 |
} |
} |
522 |
|
if(dec->out_frm) |
523 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
524 |
|
|
525 |
} |
} |
526 |
|
|
527 |
} |
} |
528 |
|
|
529 |
|
|
530 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
531 |
|
get_motion_vector(DECODER * dec, |
532 |
|
Bitstream * bs, |
533 |
|
int x, |
534 |
|
int y, |
535 |
|
int k, |
536 |
|
VECTOR * mv, |
537 |
|
int fcode, |
538 |
|
const int bound) |
539 |
{ |
{ |
540 |
|
|
541 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
543 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
544 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
545 |
|
|
546 |
VECTOR pmv[4]; |
VECTOR pmv; |
|
int32_t psad[4]; |
|
|
|
|
547 |
int mv_x, mv_y; |
int mv_x, mv_y; |
|
int pmv_x, pmv_y; |
|
548 |
|
|
549 |
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
|
|
|
|
|
pmv_x = pmv[0].x; |
|
|
pmv_y = pmv[0].y; |
|
550 |
|
|
551 |
mv_x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
552 |
mv_y = get_mv(bs, fcode); |
mv_y = get_mv(bs, fcode); |
553 |
|
|
554 |
mv_x += pmv_x; |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
|
mv_y += pmv_y; |
|
555 |
|
|
556 |
if (mv_x < low) |
mv_x += pmv.x; |
557 |
{ |
mv_y += pmv.y; |
558 |
|
|
559 |
|
if (mv_x < low) { |
560 |
mv_x += range; |
mv_x += range; |
561 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
562 |
mv_x -= range; |
mv_x -= range; |
563 |
} |
} |
564 |
|
|
565 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
566 |
mv_y += range; |
mv_y += range; |
567 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
568 |
mv_y -= range; |
mv_y -= range; |
569 |
} |
} |
570 |
|
|
574 |
} |
} |
575 |
|
|
576 |
|
|
577 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
void |
578 |
|
decoder_pframe(DECODER * dec, |
579 |
|
Bitstream * bs, |
580 |
|
int rounding, |
581 |
|
int quant, |
582 |
|
int fcode, |
583 |
|
int intra_dc_threshold) |
584 |
{ |
{ |
585 |
|
|
586 |
uint32_t x, y; |
uint32_t x, y; |
587 |
|
uint32_t bound; |
588 |
|
int cp_mb, st_mb; |
589 |
|
|
590 |
start_timer(); |
start_timer(); |
591 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
592 |
|
dec->width, dec->height); |
593 |
stop_edges_timer(); |
stop_edges_timer(); |
594 |
|
|
595 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
596 |
{ |
|
597 |
for (x = 0; x < dec->mb_width; x++) |
for (y = 0; y < dec->mb_height; y++) { |
598 |
|
cp_mb = st_mb = 0; |
599 |
|
for (x = 0; x < dec->mb_width; x++) { |
600 |
|
MACROBLOCK *mb; |
601 |
|
|
602 |
|
// skip stuffing |
603 |
|
while (BitstreamShowBits(bs, 10) == 1) |
604 |
|
BitstreamSkip(bs, 10); |
605 |
|
|
606 |
|
if (check_resync_marker(bs, fcode - 1)) |
607 |
{ |
{ |
608 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
609 |
|
x = bound % dec->mb_width; |
610 |
|
y = bound / dec->mb_width; |
611 |
|
} |
612 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
613 |
|
|
614 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
615 |
|
|
616 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
617 |
if (!(BitstreamGetBit(bs))) // not_coded |
if (!(BitstreamGetBit(bs))) // not_coded |
623 |
uint32_t cbp; |
uint32_t cbp; |
624 |
uint32_t intra; |
uint32_t intra; |
625 |
|
|
626 |
|
cp_mb++; |
627 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
628 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
629 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
630 |
|
|
631 |
|
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
632 |
|
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
633 |
acpred_flag = 0; |
acpred_flag = 0; |
634 |
|
|
635 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
636 |
|
|
637 |
if (intra) |
if (intra) { |
|
{ |
|
638 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
639 |
} |
} |
640 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
641 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
642 |
|
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
643 |
|
|
644 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
645 |
|
|
646 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
647 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
648 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
649 |
if (quant > 31) |
quant += dquant; |
650 |
{ |
if (quant > 31) { |
651 |
quant = 31; |
quant = 31; |
652 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
653 |
quant = 1; |
quant = 1; |
654 |
} |
} |
655 |
|
DPRINTF(DPRINTF_MB, "quant %i", quant); |
656 |
} |
} |
657 |
mb->quant = quant; |
mb->quant = quant; |
658 |
|
|
659 |
if (dec->interlacing) |
if (dec->interlacing) { |
660 |
{ |
if (cbp || intra) { |
661 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
662 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DEBUG1("decp: field_dct: ", mb->field_dct); |
663 |
|
} |
664 |
|
|
665 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
666 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
667 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DEBUG1("decp: field_pred: ", mb->field_pred); |
668 |
|
|
669 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
670 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
671 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
672 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
675 |
} |
} |
676 |
} |
} |
677 |
|
|
678 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
679 |
{ |
if (dec->interlacing && mb->field_pred) { |
680 |
if (dec->interlacing && mb->field_pred) |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
681 |
{ |
fcode, bound); |
682 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
683 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
fcode, bound); |
684 |
} |
} else { |
685 |
else |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
686 |
{ |
fcode, bound); |
687 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
688 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
mb->mvs[0].x; |
689 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
690 |
} |
mb->mvs[0].y; |
691 |
} |
} |
692 |
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
} else if (mb->mode == MODE_INTER4V ) { |
693 |
{ |
|
694 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
695 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
696 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
697 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
698 |
} |
} else // MODE_INTRA, MODE_INTRA_Q |
699 |
else // MODE_INTRA, MODE_INTRA_Q |
{ |
700 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
701 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
0; |
702 |
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 = |
703 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
0; |
704 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
705 |
|
intra_dc_threshold, bound); |
706 |
continue; |
continue; |
707 |
} |
} |
708 |
|
|
709 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
710 |
} |
rounding); |
711 |
else // not coded |
} else // not coded |
712 |
{ |
{ |
713 |
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
DEBUG2("P-frame MB at (X,Y)=",x,y); |
714 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
715 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
716 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
719 |
|
|
720 |
start_timer(); |
start_timer(); |
721 |
|
|
722 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
723 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
(16 * x), |
724 |
dec->edged_width); |
dec->refn[0].y + (16 * y) * dec->edged_width + |
725 |
|
(16 * x), dec->edged_width); |
726 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
727 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
728 |
dec->edged_width); |
(16 * x + 8), |
729 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
730 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
(16 * x + 8), dec->edged_width); |
731 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), |
|
732 |
dec->edged_width); |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
733 |
|
(16 * x), |
734 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->refn[0].y + (16 * y + |
735 |
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), |
8) * dec->edged_width + |
736 |
dec->edged_width); |
(16 * x), dec->edged_width); |
737 |
|
|
738 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
739 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), |
740 |
|
dec->refn[0].y + (16 * y + |
741 |
|
8) * dec->edged_width + |
742 |
|
(16 * x + 8), dec->edged_width); |
743 |
|
|
744 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
745 |
|
(8 * x), |
746 |
|
dec->refn[0].u + |
747 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
748 |
dec->edged_width/2); |
dec->edged_width/2); |
749 |
|
|
750 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
751 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
752 |
|
dec->refn[0].v + |
753 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
754 |
dec->edged_width/2); |
dec->edged_width/2); |
|
|
|
755 |
stop_transfer_timer(); |
stop_transfer_timer(); |
756 |
|
if(dec->out_frm && cp_mb > 0) { |
757 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
758 |
|
cp_mb = 0; |
759 |
} |
} |
760 |
|
st_mb = x+1; |
761 |
} |
} |
762 |
} |
} |
763 |
|
if(dec->out_frm && cp_mb > 0) |
764 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
765 |
|
} |
766 |
} |
} |
767 |
|
|
768 |
|
|
769 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
770 |
// decode B-frame motion vector |
// decode B-frame motion vector |
771 |
void get_b_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, VECTOR * mv, int fcode, const VECTOR pmv) |
void |
772 |
|
get_b_motion_vector(DECODER * dec, |
773 |
|
Bitstream * bs, |
774 |
|
int x, |
775 |
|
int y, |
776 |
|
VECTOR * mv, |
777 |
|
int fcode, |
778 |
|
const VECTOR pmv) |
779 |
{ |
{ |
780 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
781 |
int high = (32 * scale_fac) - 1; |
int high = (32 * scale_fac) - 1; |
794 |
mv_x += pmv_x; |
mv_x += pmv_x; |
795 |
mv_y += pmv_y; |
mv_y += pmv_y; |
796 |
|
|
797 |
if (mv_x < low) |
if (mv_x < low) { |
|
{ |
|
798 |
mv_x += range; |
mv_x += range; |
799 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
800 |
mv_x -= range; |
mv_x -= range; |
801 |
} |
} |
802 |
|
|
803 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
804 |
mv_y += range; |
mv_y += range; |
805 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
806 |
mv_y -= range; |
mv_y -= range; |
807 |
} |
} |
808 |
|
|
813 |
|
|
814 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
815 |
// decode an B-frame forward & backward inter macroblock |
// decode an B-frame forward & backward inter macroblock |
816 |
void decoder_bf_mbinter(DECODER * dec, |
void |
817 |
|
decoder_bf_mbinter(DECODER * dec, |
818 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
819 |
const uint32_t x_pos, |
const uint32_t x_pos, |
820 |
const uint32_t y_pos, |
const uint32_t y_pos, |
839 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
840 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
841 |
|
|
842 |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
843 |
{ |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
844 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
845 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
846 |
|
|
847 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
848 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
849 |
} |
} else { |
|
else |
|
|
{ |
|
850 |
int sum; |
int sum; |
851 |
|
|
852 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
853 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = |
854 |
|
(sum == |
855 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
856 |
|
(ABS(sum) / 16) * 2)); |
857 |
|
|
858 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
859 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = |
860 |
|
(sum == |
861 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
862 |
|
(ABS(sum) / 16) * 2)); |
863 |
} |
} |
864 |
|
|
865 |
start_timer(); |
start_timer(); |
866 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
867 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
868 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
869 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
870 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
871 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
872 |
|
0); |
873 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
874 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
875 |
|
0); |
876 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
877 |
|
uv_dx, uv_dy, stride2, 0); |
878 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
879 |
|
uv_dx, uv_dy, stride2, 0); |
880 |
stop_comp_timer(); |
stop_comp_timer(); |
881 |
|
|
882 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
883 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
884 |
|
|
885 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
886 |
{ |
{ |
887 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
888 |
|
|
889 |
start_timer(); |
start_timer(); |
890 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
891 |
stop_coding_timer(); |
stop_coding_timer(); |
892 |
|
|
893 |
start_timer(); |
start_timer(); |
894 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
895 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
896 |
} |
} else { |
|
else |
|
|
{ |
|
897 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
898 |
} |
} |
899 |
stop_iquant_timer(); |
stop_iquant_timer(); |
900 |
|
|
|
data[0]=10; |
|
901 |
start_timer(); |
start_timer(); |
902 |
idct(&data[i*64]); |
idct(&data[i*64]); |
903 |
stop_idct_timer(); |
stop_idct_timer(); |
904 |
} |
} |
905 |
} |
} |
906 |
|
|
907 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
908 |
next_block = stride; |
next_block = stride; |
909 |
stride *= 2; |
stride *= 2; |
910 |
} |
} |
928 |
|
|
929 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
930 |
// decode an B-frame direct & inter macroblock |
// decode an B-frame direct & inter macroblock |
931 |
void decoder_bf_interpolate_mbinter(DECODER * dec, |
void |
932 |
IMAGE forward, IMAGE backward, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
933 |
|
IMAGE forward, |
934 |
|
IMAGE backward, |
935 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
936 |
const uint32_t x_pos, |
const uint32_t x_pos, |
937 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t cbp, |
|
938 |
Bitstream * bs ) |
Bitstream * bs ) |
939 |
{ |
{ |
940 |
|
|
949 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
950 |
uint32_t i; |
uint32_t i; |
951 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
952 |
|
const uint32_t cbp = pMB->cbp; |
953 |
|
|
954 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
955 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
956 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
957 |
|
|
958 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
959 |
{ |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
960 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
961 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
962 |
|
|
968 |
|
|
969 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
970 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
971 |
} |
} else { |
|
else |
|
|
{ |
|
972 |
int sum; |
int sum; |
973 |
|
|
974 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
975 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx = |
976 |
|
(sum == |
977 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
978 |
|
(ABS(sum) / 16) * 2)); |
979 |
|
|
980 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
981 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = |
982 |
|
(sum == |
983 |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
984 |
b_uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
(ABS(sum) / 16) * 2)); |
985 |
|
|
986 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
sum = |
987 |
b_uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
988 |
} |
pMB->b_mvs[3].x; |
989 |
|
b_uv_dx = |
990 |
|
(sum == |
991 |
start_timer(); |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
992 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
(ABS(sum) / 16) * 2)); |
993 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
994 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
sum = |
995 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
996 |
interpolate8x8_switch(dec->cur.u, forward.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
pMB->b_mvs[3].y; |
997 |
interpolate8x8_switch(dec->cur.v, forward.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
b_uv_dy = |
998 |
|
(sum == |
999 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1000 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos , pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
(ABS(sum) / 16) * 2)); |
1001 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos , pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
} |
1002 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
1003 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
1004 |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
start_timer(); |
1005 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1006 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1007 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos , stride); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1008 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos , stride); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1009 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos + 8, stride); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1010 |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos + 8, stride); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1011 |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8*x_pos, 8*y_pos, stride2); |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1012 |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8*x_pos, 8*y_pos, stride2); |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1013 |
|
0); |
1014 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1015 |
|
uv_dy, stride2, 0); |
1016 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1017 |
|
uv_dy, stride2, 0); |
1018 |
|
|
1019 |
|
|
1020 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1021 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1022 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1023 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1024 |
|
0); |
1025 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1026 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1027 |
|
stride, 0); |
1028 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1029 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1030 |
|
stride, 0); |
1031 |
|
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1032 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1033 |
|
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1034 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1035 |
|
|
1036 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
1037 |
|
stride); |
1038 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1039 |
|
stride); |
1040 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1041 |
|
stride); |
1042 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1043 |
|
16 * y_pos + 8, stride); |
1044 |
|
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1045 |
|
stride2); |
1046 |
|
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
1047 |
|
stride2); |
1048 |
stop_comp_timer(); |
stop_comp_timer(); |
1049 |
|
|
1050 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
1051 |
{ |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1052 |
|
|
1053 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
1054 |
{ |
{ |
1055 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
1056 |
|
|
1057 |
start_timer(); |
start_timer(); |
1058 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
1059 |
stop_coding_timer(); |
stop_coding_timer(); |
1060 |
|
|
1061 |
start_timer(); |
start_timer(); |
1062 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
1063 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
1064 |
} |
} else { |
|
else |
|
|
{ |
|
1065 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
1066 |
} |
} |
1067 |
stop_iquant_timer(); |
stop_iquant_timer(); |
1072 |
} |
} |
1073 |
} |
} |
1074 |
|
|
1075 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
1076 |
next_block = stride; |
next_block = stride; |
1077 |
stride *= 2; |
stride *= 2; |
1078 |
} |
} |
1096 |
|
|
1097 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
1098 |
// for decode B-frame dbquant |
// for decode B-frame dbquant |
1099 |
int32_t __inline get_dbquant(Bitstream * bs) |
int32_t __inline |
1100 |
|
get_dbquant(Bitstream * bs) |
1101 |
{ |
{ |
1102 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) // '0' |
1103 |
return(0); |
return(0); |
1114 |
// 01 1 |
// 01 1 |
1115 |
// 001 2 |
// 001 2 |
1116 |
// 0001 3 |
// 0001 3 |
1117 |
int32_t __inline get_mbtype(Bitstream * bs) |
int32_t __inline |
1118 |
|
get_mbtype(Bitstream * bs) |
1119 |
{ |
{ |
1120 |
int32_t mb_type; |
int32_t mb_type; |
1121 |
|
|
1122 |
for(mb_type=0;mb_type<=3;mb_type++){ |
for(mb_type=0;mb_type<=3;mb_type++){ |
1123 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
1124 |
break; |
break; |
1130 |
return(-1); |
return(-1); |
1131 |
} |
} |
1132 |
|
|
1133 |
void decoder_bframe(DECODER * dec, Bitstream * bs, int quant, int fcode_forward, int fcode_backward) |
void |
1134 |
|
decoder_bframe(DECODER * dec, |
1135 |
|
Bitstream * bs, |
1136 |
|
int quant, |
1137 |
|
int fcode_forward, |
1138 |
|
int fcode_backward) |
1139 |
{ |
{ |
|
|
|
1140 |
uint32_t x, y; |
uint32_t x, y; |
1141 |
VECTOR mv, zeromv; |
VECTOR mv; |
1142 |
|
const VECTOR zeromv = {0,0}; |
1143 |
|
#ifdef BFRAMES_DEC_DEBUG |
1144 |
|
FILE *fp; |
1145 |
|
static char first=0; |
1146 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1147 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1148 |
|
} |
1149 |
|
#endif |
1150 |
|
|
1151 |
start_timer(); |
start_timer(); |
1152 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1153 |
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
dec->width, dec->height); |
1154 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1155 |
|
dec->width, dec->height); |
1156 |
stop_edges_timer(); |
stop_edges_timer(); |
1157 |
|
|
1158 |
|
#ifdef BFRAMES_DEC_DEBUG |
1159 |
|
if (!first){ |
1160 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1161 |
|
} |
1162 |
|
#endif |
1163 |
|
|
1164 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) { |
|
{ |
|
1165 |
// Initialize Pred Motion Vector |
// Initialize Pred Motion Vector |
1166 |
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
dec->p_fmv = dec->p_bmv = zeromv; |
1167 |
for (x = 0; x < dec->mb_width; x++) |
for (x = 0; x < dec->mb_width; x++) { |
|
{ |
|
1168 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
1169 |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
1170 |
|
|
1171 |
mb->mvs[0].x=mb->mvs[0].y=zeromv.x = zeromv.y = mv.x = mv.y = 0; |
mv = |
1172 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1173 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1174 |
|
|
1175 |
// the last P_VOP is skip macroblock ? |
// the last P_VOP is skip macroblock ? |
1176 |
if (last_mb->mode == MODE_NOT_CODED){ |
if (last_mb->mode == MODE_NOT_CODED){ |
1177 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
|
mb->mb_type = MODE_FORWARD; |
|
1178 |
mb->cbp = 0; |
mb->cbp = 0; |
1179 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
#ifdef BFRAMES_DEC_DEBUG |
1180 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mb->mb_type = MODE_NOT_CODED; |
1181 |
mb->quant = 8; |
BFRAME_DEBUG |
1182 |
|
#endif |
1183 |
|
mb->mb_type = MODE_FORWARD; |
1184 |
|
mb->quant = last_mb->quant; |
1185 |
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1186 |
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1187 |
|
|
1188 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1189 |
continue; |
continue; |
1190 |
} |
} |
1191 |
|
|
|
//t=BitstreamShowBits(bs,32); |
|
|
|
|
1192 |
if (!BitstreamGetBit(bs)){ // modb=='0' |
if (!BitstreamGetBit(bs)){ // modb=='0' |
1193 |
const uint8_t modb2=BitstreamGetBit(bs); |
const uint8_t modb2=BitstreamGetBit(bs); |
1194 |
|
|
1202 |
if (mb->mb_type && mb->cbp){ |
if (mb->mb_type && mb->cbp){ |
1203 |
quant += get_dbquant(bs); |
quant += get_dbquant(bs); |
1204 |
|
|
1205 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
1206 |
quant = 31; |
quant = 31; |
1207 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
1208 |
quant = 1; |
quant = 1; |
1209 |
} |
} |
|
} else { |
|
|
quant = 8; |
|
1210 |
} |
} |
|
mb->quant = quant; |
|
1211 |
} else { |
} else { |
1212 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mb_type = MODE_DIRECT_NONE_MV; |
1213 |
mb->cbp = 0; |
mb->cbp = 0; |
1214 |
} |
} |
1215 |
|
|
1216 |
mb->mode = MODE_INTER; |
mb->quant = quant; |
1217 |
|
mb->mode = MODE_INTER4V; |
1218 |
//DEBUG1("Switch bm_type=",mb->mb_type); |
//DEBUG1("Switch bm_type=",mb->mb_type); |
1219 |
|
|
1220 |
switch(mb->mb_type) |
#ifdef BFRAMES_DEC_DEBUG |
1221 |
{ |
BFRAME_DEBUG |
1222 |
|
#endif |
1223 |
|
|
1224 |
|
switch (mb->mb_type) { |
1225 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1226 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1227 |
|
|
1228 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1229 |
{ // Because this file is a C file not C++ so I use '{' to define var |
{ |
1230 |
const int64_t TRB=dec->time_pp-dec->time_bp, |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
TRD=dec->time_pp; |
|
1231 |
int i; |
int i; |
1232 |
|
|
1233 |
for(i=0;i<4;i++){ |
for(i=0;i<4;i++){ |
1234 |
mb->mvs[i].x = (int32_t)((TRB * last_mb->mvs[i].x)/TRD+mb->mvs[0].x); |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1235 |
mb->b_mvs[i].x = (int32_t)((mb->mvs[0].x==0)?((TRB-TRD)*last_mb->mvs[i].x)/TRD:mb->mvs[i].x-last_mb->mvs[i].x); |
/ TRD + mv.x); |
1236 |
mb->mvs[i].y = (int32_t)((TRB * last_mb->mvs[i].y)/TRD+mb->mvs[0].y); |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1237 |
mb->b_mvs[i].y = (int32_t)((mb->mvs[0].y==0)?((TRB-TRD)*last_mb->mvs[i].y)/TRD:mb->mvs[i].y-last_mb->mvs[i].y); |
? ((TRB - TRD) * last_mb->mvs[i].x) |
1238 |
|
/ TRD |
1239 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1240 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1241 |
|
/ TRD + mv.y); |
1242 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1243 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1244 |
|
/ TRD |
1245 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1246 |
} |
} |
1247 |
//DEBUG("B-frame Direct!\n"); |
//DEBUG("B-frame Direct!\n"); |
1248 |
} |
} |
1249 |
mb->mode = MODE_INTER4V; |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1250 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
mb, x, y, bs); |
1251 |
break; |
break; |
1252 |
|
|
1253 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1254 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1255 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv); |
1256 |
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1257 |
|
|
1258 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], fcode_backward, dec->p_bmv); |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1259 |
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = mb->b_mvs[3].x = mb->b_mvs[0].x; |
fcode_backward, dec->p_bmv); |
1260 |
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = mb->b_mvs[3].y = mb->b_mvs[0].y; |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1261 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1262 |
|
|
1263 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1264 |
|
mb, x, y, bs); |
1265 |
//DEBUG("B-frame Bidir!\n"); |
//DEBUG("B-frame Bidir!\n"); |
1266 |
break; |
break; |
1267 |
|
|
1268 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1269 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, dec->p_bmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1270 |
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_bmv); |
1271 |
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1272 |
|
|
1273 |
|
mb->mode = MODE_INTER; |
1274 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1275 |
//DEBUG("B-frame Backward!\n"); |
//DEBUG("B-frame Backward!\n"); |
1276 |
break; |
break; |
1277 |
|
|
1278 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1279 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1280 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv); |
1281 |
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1282 |
|
|
1283 |
|
mb->mode = MODE_INTER; |
1284 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1285 |
//DEBUG("B-frame Forward!\n"); |
//DEBUG("B-frame Forward!\n"); |
1286 |
break; |
break; |
1291 |
|
|
1292 |
} // end of FOR |
} // end of FOR |
1293 |
} |
} |
1294 |
|
#ifdef BFRAMES_DEC_DEBUG |
1295 |
|
if (!first){ |
1296 |
|
first=1; |
1297 |
|
if (fp) |
1298 |
|
fclose(fp); |
1299 |
|
} |
1300 |
|
#endif |
1301 |
} |
} |
1302 |
|
|
1303 |
// swap two MACROBLOCK array |
// swap two MACROBLOCK array |
1304 |
void mb_swap(MACROBLOCK **mb1, MACROBLOCK **mb2) |
void |
1305 |
|
mb_swap(MACROBLOCK ** mb1, |
1306 |
|
MACROBLOCK ** mb2) |
1307 |
{ |
{ |
1308 |
MACROBLOCK *temp=*mb1; |
MACROBLOCK *temp=*mb1; |
1309 |
|
|
1310 |
*mb1=*mb2; |
*mb1=*mb2; |
1311 |
*mb2=temp; |
*mb2=temp; |
1312 |
} |
} |
1313 |
|
|
1314 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
int |
1315 |
|
decoder_decode(DECODER * dec, |
1316 |
|
XVID_DEC_FRAME * frame) |
1317 |
{ |
{ |
1318 |
|
|
1319 |
Bitstream bs; |
Bitstream bs; |
1326 |
|
|
1327 |
start_global_timer(); |
start_global_timer(); |
1328 |
|
|
1329 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1330 |
|
|
1331 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1332 |
|
|
1333 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
1334 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
1335 |
dec->frames ++; |
dec->frames ++; |
1336 |
vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold); |
vop_type = |
1337 |
|
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1338 |
|
&fcode_backward, &intra_dc_threshold); |
1339 |
|
|
1340 |
dec->p_bmv.x=dec->p_bmv.y=dec->p_fmv.y=dec->p_fmv.y=0; // init pred vector to 0 |
dec->p_bmv.x=dec->p_bmv.y=dec->p_fmv.y=dec->p_fmv.y=0; // init pred vector to 0 |
1341 |
switch (vop_type) |
|
1342 |
{ |
switch (vop_type) { |
1343 |
case P_VOP : |
case P_VOP : |
1344 |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, intra_dc_threshold); |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1345 |
|
intra_dc_threshold); |
1346 |
|
#ifdef BFRAMES_DEC |
1347 |
DEBUG1("P_VOP Time=",dec->time); |
DEBUG1("P_VOP Time=",dec->time); |
1348 |
|
#endif |
1349 |
break; |
break; |
1350 |
|
|
1351 |
case I_VOP : |
case I_VOP : |
1352 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1353 |
|
#ifdef BFRAMES_DEC |
1354 |
DEBUG1("I_VOP Time=",dec->time); |
DEBUG1("I_VOP Time=",dec->time); |
1355 |
|
#endif |
1356 |
break; |
break; |
1357 |
|
|
1358 |
case B_VOP : |
case B_VOP : |
1359 |
|
#ifdef BFRAMES_DEC |
1360 |
if (dec->time_pp > dec->time_bp){ |
if (dec->time_pp > dec->time_bp){ |
1361 |
DEBUG1("B_VOP Time=",dec->time); |
DEBUG1("B_VOP Time=",dec->time); |
1362 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1363 |
} else { |
} else { |
1364 |
DEBUG("broken B-frame!"); |
DEBUG("broken B-frame!"); |
1365 |
} |
} |
1366 |
|
#else |
1367 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1368 |
|
#endif |
1369 |
break; |
break; |
1370 |
|
|
1371 |
case N_VOP : // vop not coded |
case N_VOP : // vop not coded |
1372 |
|
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1373 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1374 |
break; |
break; |
1375 |
|
|
1376 |
default : |
default : |
1377 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1378 |
} |
} |
1379 |
|
|
1380 |
|
#ifdef BFRAMES_DEC_DEBUG |
1381 |
|
if (frame->length != BitstreamPos(&bs) / 8){ |
1382 |
|
DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); |
1383 |
|
} |
1384 |
|
#endif |
1385 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1386 |
|
|
1387 |
|
|
1388 |
|
#ifdef BFRAMES_DEC |
1389 |
|
// test if no B_VOP |
1390 |
|
if (dec->low_delay || dec->frames == 0) { |
1391 |
|
#endif |
1392 |
|
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1393 |
|
frame->image, frame->stride, frame->colorspace); |
1394 |
|
|
1395 |
|
#ifdef BFRAMES_DEC |
1396 |
|
} else { |
1397 |
if (dec->frames >= 1){ |
if (dec->frames >= 1){ |
1398 |
start_timer(); |
start_timer(); |
1399 |
if ((vop_type == I_VOP || vop_type == P_VOP)) |
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1400 |
{ |
image_output(&dec->refn[0], dec->width, dec->height, |
1401 |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
dec->edged_width, frame->image, frame->stride, |
1402 |
frame->image, frame->stride, frame->colorspace); |
frame->colorspace); |
1403 |
} else if (vop_type == B_VOP) { |
} else if (vop_type == B_VOP) { |
1404 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, |
1405 |
frame->image, frame->stride, frame->colorspace); |
dec->edged_width, frame->image, frame->stride, |
1406 |
|
frame->colorspace); |
1407 |
} |
} |
1408 |
stop_conv_timer(); |
stop_conv_timer(); |
1409 |
} |
} |
1410 |
|
} |
1411 |
|
#endif |
1412 |
|
|
1413 |
if (vop_type==I_VOP || vop_type==P_VOP){ |
if (vop_type==I_VOP || vop_type==P_VOP){ |
1414 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1415 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1416 |
|
|
1417 |
// swap MACROBLOCK |
// swap MACROBLOCK |
1418 |
|
// the Divx will not set the low_delay flage some times |
1419 |
|
// so follow code will wrong to not swap at that time |
1420 |
|
// this will broken bitstream! so I'm change it, |
1421 |
|
// But that is not the best way! can anyone tell me how |
1422 |
|
// to do another way? |
1423 |
|
// 18-07-2002 MinChen<chenm001@163.com> |
1424 |
|
//if (!dec->low_delay && vop_type == P_VOP) |
1425 |
if (vop_type==P_VOP) |
if (vop_type==P_VOP) |
1426 |
mb_swap(&dec->mbs, &dec->last_mbs); |
mb_swap(&dec->mbs, &dec->last_mbs); |
1427 |
} |
} |