1 |
/************************************************************************** |
/************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* decoder main |
* - Decoder main module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
12 |
* editors and their companies, will have no liability for use of this |
* editors and their companies, will have no liability for use of this |
13 |
* software or modifications or derivatives thereof. |
* software or modifications or derivatives thereof. |
14 |
* |
* |
15 |
* This program is xvid_free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
16 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
17 |
* the xvid_free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
18 |
* (at your option) any later version. |
* (at your option) any later version. |
19 |
* |
* |
20 |
* This program is distributed in the hope that it will be useful, |
* This program is distributed in the hope that it will be useful, |
23 |
* GNU General Public License for more details. |
* GNU General Public License for more details. |
24 |
* |
* |
25 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
26 |
* along with this program; if not, write to the xvid_free Software |
* along with this program; if not, write to the Free Software |
27 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
28 |
* |
* |
29 |
*************************************************************************/ |
*************************************************************************/ |
30 |
|
|
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); |
47 |
|
* MinChen <chenm001@163.com> |
48 |
|
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
49 |
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
50 |
|
* reconstructing blocks, thus artifacts |
51 |
|
* interlacing speedup - used transfers to re-interlace |
52 |
|
* interlaced decoding should be as fast as progressive now |
53 |
|
* 26.03.2002 interlacing support - moved transfers outside decode loop |
54 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
55 |
* 22.12.2001 block based interpolation |
* 22.12.2001 lock based interpolation |
56 |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
57 |
* |
* |
58 |
|
* $Id$ |
59 |
|
* |
60 |
*************************************************************************/ |
*************************************************************************/ |
61 |
|
|
62 |
#include <stdlib.h> |
#include <stdlib.h> |
63 |
#include <string.h> // memset |
#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" |
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), 16); |
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, 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); |
120 |
|
xvid_free(dec); |
121 |
|
return XVID_ERR_MEMORY; |
122 |
|
} |
123 |
|
// add by chenm001 <chenm001@163.com> |
124 |
|
// for support B-frame to reference last 2 frame |
125 |
|
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
126 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
127 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
128 |
|
xvid_free(dec); |
129 |
|
return XVID_ERR_MEMORY; |
130 |
|
} |
131 |
|
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); |
134 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
135 |
xvid_free(dec); |
xvid_free(dec); |
136 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
137 |
} |
} |
138 |
|
|
139 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, 16); |
if (image_create(&dec->refh, dec->edged_width, dec->edged_height)) { |
|
if (dec->mbs == NULL) |
|
|
{ |
|
140 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
141 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
142 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
143 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
144 |
xvid_free(dec); |
xvid_free(dec); |
145 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
146 |
} |
} |
147 |
|
|
148 |
|
dec->mbs = |
149 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
150 |
|
CACHE_LINE); |
151 |
|
if (dec->mbs == NULL) { |
152 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
153 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
154 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
155 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
156 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
157 |
|
xvid_free(dec); |
158 |
|
return XVID_ERR_MEMORY; |
159 |
|
} |
160 |
|
|
161 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
162 |
|
|
163 |
|
// add by chenm001 <chenm001@163.com> |
164 |
|
// for skip MB flag |
165 |
|
dec->last_mbs = |
166 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
167 |
|
CACHE_LINE); |
168 |
|
if (dec->last_mbs == NULL) { |
169 |
|
xvid_free(dec->mbs); |
170 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
171 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
172 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
173 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
174 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
175 |
|
xvid_free(dec); |
176 |
|
return XVID_ERR_MEMORY; |
177 |
|
} |
178 |
|
|
179 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
180 |
|
|
181 |
init_timer(); |
init_timer(); |
182 |
create_vlc_tables(); |
|
183 |
|
// add by chenm001 <chenm001@163.com> |
184 |
|
// for support B-frame to save reference frame's time |
185 |
|
dec->frames = -1; |
186 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
187 |
|
|
188 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
189 |
} |
} |
190 |
|
|
191 |
|
|
192 |
int decoder_destroy(DECODER * dec) |
int |
193 |
|
decoder_destroy(DECODER * dec) |
194 |
{ |
{ |
195 |
|
xvid_free(dec->last_mbs); |
196 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
197 |
image_destroy(&dec->refn, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
198 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
199 |
|
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
200 |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
201 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
202 |
xvid_free(dec); |
xvid_free(dec); |
203 |
|
|
|
destroy_vlc_tables(); |
|
|
|
|
204 |
write_timer(); |
write_timer(); |
205 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
206 |
} |
} |
207 |
|
|
208 |
|
|
209 |
|
|
210 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
211 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
212 |
}; |
}; |
213 |
|
|
214 |
|
|
215 |
|
|
216 |
|
|
217 |
// decode an intra macroblock |
// decode an intra macroblock |
218 |
|
|
219 |
void decoder_mbintra(DECODER * dec, MACROBLOCK * mb, int x, int y, uint32_t acpred_flag, uint32_t cbp, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
220 |
{ |
decoder_mbintra(DECODER * dec, |
221 |
uint32_t k; |
MACROBLOCK * pMB, |
222 |
|
const uint32_t x_pos, |
223 |
|
const uint32_t y_pos, |
224 |
|
const uint32_t acpred_flag, |
225 |
|
const uint32_t cbp, |
226 |
|
Bitstream * bs, |
227 |
|
const uint32_t quant, |
228 |
|
const uint32_t intra_dc_threshold, |
229 |
|
const unsigned int bound) |
230 |
|
{ |
231 |
|
|
232 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
233 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
234 |
|
|
235 |
|
uint32_t stride = dec->edged_width; |
236 |
|
uint32_t stride2 = stride / 2; |
237 |
|
uint32_t next_block = stride * 8; |
238 |
|
uint32_t i; |
239 |
|
uint32_t iQuant = pMB->quant; |
240 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
241 |
|
|
242 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
243 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
244 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
245 |
|
|
246 |
for (k = 0; k < 6; k++) |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
247 |
{ |
|
248 |
uint32_t dcscalar; |
for (i = 0; i < 6; i++) { |
249 |
int16_t block[64]; |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
int16_t data[64]; |
|
250 |
int16_t predictors[8]; |
int16_t predictors[8]; |
251 |
int start_coeff; |
int start_coeff; |
252 |
|
|
|
dcscalar = get_dc_scaler(mb->quant, k < 4); |
|
|
|
|
253 |
start_timer(); |
start_timer(); |
254 |
predict_acdc(dec->mbs, x, y, dec->mb_width, k, block, mb->quant, dcscalar, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
255 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
256 |
{ |
if (!acpred_flag) { |
257 |
mb->acpred_directions[k] = 0; |
pMB->acpred_directions[i] = 0; |
258 |
} |
} |
259 |
stop_prediction_timer(); |
stop_prediction_timer(); |
260 |
|
|
261 |
memset(block, 0, 64*sizeof(int16_t)); // clear |
if (quant < intra_dc_threshold) { |
|
|
|
|
if (quant < intra_dc_threshold) |
|
|
{ |
|
262 |
int dc_size; |
int dc_size; |
263 |
int dc_dif; |
int dc_dif; |
264 |
|
|
265 |
dc_size = k < 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); |
266 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
267 |
|
|
268 |
if (dc_size > 8) |
if (dc_size > 8) { |
|
{ |
|
269 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
270 |
} |
} |
271 |
|
|
272 |
block[0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
273 |
start_coeff = 1; |
start_coeff = 1; |
274 |
} |
|
275 |
else |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
276 |
{ |
} else { |
277 |
start_coeff = 0; |
start_coeff = 0; |
278 |
} |
} |
279 |
|
|
280 |
start_timer(); |
start_timer(); |
281 |
if (cbp & (1 << (5-k))) // coded |
if (cbp & (1 << (5 - i))) // coded |
282 |
{ |
{ |
283 |
get_intra_block(bs, block, mb->acpred_directions[k], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
284 |
|
2 : pMB->acpred_directions[i]; |
285 |
|
|
286 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
287 |
} |
} |
288 |
stop_coding_timer(); |
stop_coding_timer(); |
289 |
|
|
290 |
start_timer(); |
start_timer(); |
291 |
add_acdc(mb, k, block, dcscalar, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
292 |
stop_prediction_timer(); |
stop_prediction_timer(); |
293 |
|
|
294 |
start_timer(); |
start_timer(); |
295 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
296 |
{ |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
297 |
dequant_intra(data, block, mb->quant, dcscalar); |
} else { |
298 |
} |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
|
else |
|
|
{ |
|
|
dequant4_intra(data, block, mb->quant, dcscalar); |
|
299 |
} |
} |
300 |
stop_iquant_timer(); |
stop_iquant_timer(); |
301 |
|
|
302 |
start_timer(); |
start_timer(); |
303 |
idct(data); |
idct(&data[i * 64]); |
304 |
stop_idct_timer(); |
stop_idct_timer(); |
|
|
|
|
start_timer(); |
|
|
if (k < 4) |
|
|
{ |
|
|
transfer_16to8copy(dec->cur.y + (16*y*dec->edged_width) + 16*x + (4*(k&2)*dec->edged_width) + 8*(k&1), data, dec->edged_width); |
|
|
} |
|
|
else if (k == 4) |
|
|
{ |
|
|
transfer_16to8copy(dec->cur.u+ 8*y*(dec->edged_width/2) + 8*x, data, (dec->edged_width/2)); |
|
305 |
} |
} |
306 |
else // if (k == 5) |
|
307 |
{ |
if (dec->interlacing && pMB->field_dct) { |
308 |
transfer_16to8copy(dec->cur.v + 8*y*(dec->edged_width/2) + 8*x, data, (dec->edged_width/2)); |
next_block = stride; |
309 |
|
stride *= 2; |
310 |
} |
} |
311 |
|
|
312 |
|
start_timer(); |
313 |
|
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
314 |
|
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
315 |
|
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
316 |
|
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
317 |
|
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
318 |
|
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
319 |
stop_transfer_timer(); |
stop_transfer_timer(); |
320 |
} |
} |
|
} |
|
321 |
|
|
322 |
|
|
323 |
|
|
331 |
|
|
332 |
// decode an inter macroblock |
// decode an inter macroblock |
333 |
|
|
334 |
void decoder_mbinter(DECODER * dec, MACROBLOCK * mb, int x, int y, uint32_t acpred_flag, uint32_t cbp, Bitstream * bs, int quant, int rounding) |
void |
335 |
{ |
decoder_mbinter(DECODER * dec, |
336 |
const uint32_t stride = dec->edged_width; |
const MACROBLOCK * pMB, |
337 |
const uint32_t stride2 = dec->edged_width / 2; |
const uint32_t x_pos, |
338 |
|
const uint32_t y_pos, |
339 |
|
const uint32_t acpred_flag, |
340 |
|
const uint32_t cbp, |
341 |
|
Bitstream * bs, |
342 |
|
const uint32_t quant, |
343 |
|
const uint32_t rounding) |
344 |
|
{ |
345 |
|
|
346 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
347 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
348 |
|
|
349 |
|
uint32_t stride = dec->edged_width; |
350 |
|
uint32_t stride2 = stride / 2; |
351 |
|
uint32_t next_block = stride * 8; |
352 |
|
uint32_t i; |
353 |
|
uint32_t iQuant = pMB->quant; |
354 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
355 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
|
uint32_t k; |
|
356 |
|
|
357 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
358 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
359 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
360 |
|
|
361 |
|
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
362 |
|
uv_dx = pMB->mvs[0].x; |
363 |
|
uv_dy = pMB->mvs[0].y; |
364 |
|
|
365 |
|
if (dec->quarterpel) |
366 |
{ |
{ |
367 |
uv_dx = mb->mvs[0].x; |
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
368 |
uv_dy = mb->mvs[0].y; |
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
369 |
|
} |
370 |
|
|
371 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
372 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
373 |
} |
|
374 |
else |
start_timer(); |
375 |
{ |
if(dec->quarterpel) { |
376 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
377 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
378 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
379 |
|
} |
380 |
|
else { |
381 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
382 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
383 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
384 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
385 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
386 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
387 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
388 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
389 |
|
} |
390 |
|
|
391 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
392 |
|
uv_dx, uv_dy, stride2, rounding); |
393 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
394 |
|
uv_dx, uv_dy, stride2, rounding); |
395 |
|
stop_comp_timer(); |
396 |
|
|
397 |
|
} else { |
398 |
int sum; |
int sum; |
399 |
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
400 |
|
|
401 |
|
if (dec->quarterpel) |
402 |
|
{ |
403 |
|
sum /= 2; |
404 |
|
} |
405 |
|
|
406 |
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) ); |
407 |
|
|
408 |
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
409 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
410 |
|
if (dec->quarterpel) |
411 |
|
{ |
412 |
|
sum /= 2; |
413 |
} |
} |
414 |
|
|
415 |
|
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
416 |
|
|
417 |
start_timer(); |
start_timer(); |
418 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x, 16*y , mb->mvs[0].x, mb->mvs[0].y, stride, rounding); |
if(dec->quarterpel) { |
419 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x + 8, 16*y , mb->mvs[1].x, mb->mvs[1].y, stride, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
420 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x, 16*y + 8, mb->mvs[2].x, mb->mvs[2].y, stride, rounding); |
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
421 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x + 8, 16*y + 8, mb->mvs[3].x, mb->mvs[3].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
422 |
interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x, 8*y, uv_dx, uv_dy, stride2, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
423 |
interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x, 8*y, uv_dx, uv_dy, stride2, rounding); |
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
424 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
425 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
426 |
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
427 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
428 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
429 |
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
430 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
431 |
|
} |
432 |
|
else { |
433 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
434 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
435 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
436 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
437 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
438 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
439 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
440 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
441 |
|
} |
442 |
|
|
443 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
444 |
|
uv_dx, uv_dy, stride2, rounding); |
445 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
446 |
|
uv_dx, uv_dy, stride2, rounding); |
447 |
stop_comp_timer(); |
stop_comp_timer(); |
448 |
|
} |
449 |
|
|
450 |
|
for (i = 0; i < 6; i++) { |
451 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
452 |
|
|
453 |
for (k = 0; k < 6; k++) |
if (cbp & (1 << (5 - i))) // coded |
|
{ |
|
|
int16_t block[64]; |
|
|
int16_t data[64]; |
|
|
|
|
|
if (cbp & (1 << (5-k))) // coded |
|
454 |
{ |
{ |
455 |
memset(block, 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
456 |
|
|
457 |
start_timer(); |
start_timer(); |
458 |
get_inter_block(bs, block); |
get_inter_block(bs, &block[i * 64], direction); |
459 |
stop_coding_timer(); |
stop_coding_timer(); |
460 |
|
|
461 |
start_timer(); |
start_timer(); |
462 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
463 |
{ |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
464 |
dequant_inter(data, block, mb->quant); |
} else { |
465 |
} |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
else |
|
|
{ |
|
|
dequant4_inter(data, block, mb->quant); |
|
466 |
} |
} |
467 |
stop_iquant_timer(); |
stop_iquant_timer(); |
468 |
|
|
469 |
start_timer(); |
start_timer(); |
470 |
idct(data); |
idct(&data[i * 64]); |
471 |
stop_idct_timer(); |
stop_idct_timer(); |
|
|
|
|
start_timer(); |
|
|
if (k < 4) |
|
|
{ |
|
|
transfer_16to8add(dec->cur.y + (16*y + 4*(k&2))*stride + 16*x + 8*(k&1), data, stride); |
|
472 |
} |
} |
|
else if (k == 4) |
|
|
{ |
|
|
transfer_16to8add(dec->cur.u + 8*y*stride2 + 8*x, data, stride2); |
|
473 |
} |
} |
474 |
else // k == 5 |
|
475 |
{ |
if (dec->interlacing && pMB->field_dct) { |
476 |
transfer_16to8add(dec->cur.v + 8*y*stride2 + 8*x, data, stride2); |
next_block = stride; |
477 |
|
stride *= 2; |
478 |
} |
} |
479 |
|
|
480 |
|
start_timer(); |
481 |
|
if (cbp & 32) |
482 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
483 |
|
if (cbp & 16) |
484 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
485 |
|
if (cbp & 8) |
486 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
487 |
|
if (cbp & 4) |
488 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
489 |
|
if (cbp & 2) |
490 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
491 |
|
if (cbp & 1) |
492 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
493 |
stop_transfer_timer(); |
stop_transfer_timer(); |
494 |
} |
} |
|
} |
|
|
} |
|
|
|
|
495 |
|
|
496 |
|
|
497 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
498 |
|
decoder_iframe(DECODER * dec, |
499 |
|
Bitstream * bs, |
500 |
|
int quant, |
501 |
|
int intra_dc_threshold) |
502 |
{ |
{ |
503 |
|
uint32_t bound; |
504 |
uint32_t x, y; |
uint32_t x, y; |
505 |
|
|
506 |
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]; |
|
507 |
|
|
508 |
|
for (y = 0; y < dec->mb_height; y++) { |
509 |
|
for (x = 0; x < dec->mb_width; x++) { |
510 |
|
MACROBLOCK *mb; |
511 |
uint32_t mcbpc; |
uint32_t mcbpc; |
512 |
uint32_t cbpc; |
uint32_t cbpc; |
513 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
514 |
uint32_t cbpy; |
uint32_t cbpy; |
515 |
uint32_t cbp; |
uint32_t cbp; |
516 |
|
|
517 |
|
while (BitstreamShowBits(bs, 9) == 1) |
518 |
|
BitstreamSkip(bs, 9); |
519 |
|
|
520 |
|
if (check_resync_marker(bs, 0)) |
521 |
|
{ |
522 |
|
bound = read_video_packet_header(bs, 0, &quant); |
523 |
|
x = bound % dec->mb_width; |
524 |
|
y = bound / dec->mb_width; |
525 |
|
} |
526 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
527 |
|
|
528 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
529 |
|
|
530 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
531 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
532 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
533 |
|
|
534 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
535 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
536 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
537 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
538 |
|
|
539 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
540 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
541 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
542 |
quant = 31; |
quant = 31; |
543 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
544 |
quant = 1; |
quant = 1; |
545 |
} |
} |
546 |
} |
} |
547 |
mb->quant = quant; |
mb->quant = quant; |
548 |
|
mb->mvs[0].x = mb->mvs[0].y = |
549 |
|
mb->mvs[1].x = mb->mvs[1].y = |
550 |
|
mb->mvs[2].x = mb->mvs[2].y = |
551 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
552 |
|
|
553 |
|
if (dec->interlacing) { |
554 |
|
mb->field_dct = BitstreamGetBit(bs); |
555 |
|
DEBUG1("deci: field_dct: ", mb->field_dct); |
556 |
|
} |
557 |
|
|
558 |
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, |
559 |
|
intra_dc_threshold, bound); |
560 |
} |
} |
561 |
|
if(dec->out_frm) |
562 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
563 |
|
|
564 |
} |
} |
565 |
|
|
566 |
} |
} |
567 |
|
|
568 |
|
|
569 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
570 |
|
get_motion_vector(DECODER * dec, |
571 |
|
Bitstream * bs, |
572 |
|
int x, |
573 |
|
int y, |
574 |
|
int k, |
575 |
|
VECTOR * mv, |
576 |
|
int fcode, |
577 |
|
const int bound) |
578 |
{ |
{ |
579 |
|
|
580 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
581 |
int high = (32 * scale_fac) - 1; |
int high = (32 * scale_fac) - 1; |
582 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
583 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
584 |
|
|
585 |
VECTOR pmv[4]; |
VECTOR pmv; |
|
uint32_t psad[4]; |
|
|
|
|
586 |
int mv_x, mv_y; |
int mv_x, mv_y; |
|
int pmv_x, pmv_y; |
|
587 |
|
|
588 |
|
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; |
|
589 |
|
|
590 |
mv_x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
591 |
mv_y = get_mv(bs, fcode); |
mv_y = get_mv(bs, fcode); |
592 |
|
|
593 |
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; |
|
594 |
|
|
595 |
if (mv_x < low) |
mv_x += pmv.x; |
596 |
{ |
mv_y += pmv.y; |
597 |
|
|
598 |
|
if (mv_x < low) { |
599 |
mv_x += range; |
mv_x += range; |
600 |
} |
} else if (mv_x > high) { |
|
else if (mv_x > high) |
|
|
{ |
|
601 |
mv_x -= range; |
mv_x -= range; |
602 |
} |
} |
603 |
|
|
604 |
if (mv_y < low) |
if (mv_y < low) { |
|
{ |
|
605 |
mv_y += range; |
mv_y += range; |
606 |
} |
} else if (mv_y > high) { |
|
else if (mv_y > high) |
|
|
{ |
|
607 |
mv_y -= range; |
mv_y -= range; |
608 |
} |
} |
609 |
|
|
613 |
} |
} |
614 |
|
|
615 |
|
|
616 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
void |
617 |
|
decoder_pframe(DECODER * dec, |
618 |
|
Bitstream * bs, |
619 |
|
int rounding, |
620 |
|
int quant, |
621 |
|
int fcode, |
622 |
|
int intra_dc_threshold) |
623 |
{ |
{ |
|
uint32_t x, y; |
|
624 |
|
|
625 |
image_swap(&dec->cur, &dec->refn); |
uint32_t x, y; |
626 |
|
uint32_t bound; |
627 |
|
int cp_mb, st_mb; |
628 |
|
|
629 |
start_timer(); |
start_timer(); |
630 |
image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
631 |
|
dec->width, dec->height); |
632 |
stop_edges_timer(); |
stop_edges_timer(); |
633 |
|
|
634 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
635 |
{ |
|
636 |
for (x = 0; x < dec->mb_width; x++) |
for (y = 0; y < dec->mb_height; y++) { |
637 |
|
cp_mb = st_mb = 0; |
638 |
|
for (x = 0; x < dec->mb_width; x++) { |
639 |
|
MACROBLOCK *mb; |
640 |
|
|
641 |
|
// skip stuffing |
642 |
|
while (BitstreamShowBits(bs, 10) == 1) |
643 |
|
BitstreamSkip(bs, 10); |
644 |
|
|
645 |
|
if (check_resync_marker(bs, fcode - 1)) |
646 |
{ |
{ |
647 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
648 |
|
x = bound % dec->mb_width; |
649 |
|
y = bound / dec->mb_width; |
650 |
|
} |
651 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
652 |
|
|
653 |
if (!BitstreamGetBit(bs)) // not_coded |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
654 |
|
|
655 |
|
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
656 |
|
if (!(BitstreamGetBit(bs))) // not_coded |
657 |
{ |
{ |
658 |
uint32_t mcbpc; |
uint32_t mcbpc; |
659 |
uint32_t cbpc; |
uint32_t cbpc; |
662 |
uint32_t cbp; |
uint32_t cbp; |
663 |
uint32_t intra; |
uint32_t intra; |
664 |
|
|
665 |
|
cp_mb++; |
666 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
667 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
668 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
669 |
|
|
670 |
|
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
671 |
|
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
672 |
acpred_flag = 0; |
acpred_flag = 0; |
673 |
|
|
674 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
675 |
|
|
676 |
if (intra) |
if (intra) { |
|
{ |
|
677 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
678 |
} |
} |
679 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
680 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
681 |
|
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
682 |
|
|
683 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
684 |
|
|
685 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
686 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
687 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
688 |
if (quant > 31) |
quant += dquant; |
689 |
{ |
if (quant > 31) { |
690 |
quant = 31; |
quant = 31; |
691 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
692 |
quant = 1; |
quant = 1; |
693 |
} |
} |
694 |
|
DPRINTF(DPRINTF_MB, "quant %i", quant); |
695 |
} |
} |
696 |
mb->quant = quant; |
mb->quant = quant; |
697 |
|
|
698 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (dec->interlacing) { |
699 |
{ |
if (cbp || intra) { |
700 |
|
mb->field_dct = BitstreamGetBit(bs); |
701 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
DEBUG1("decp: field_dct: ", mb->field_dct); |
702 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
} |
703 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
704 |
|
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
705 |
|
mb->field_pred = BitstreamGetBit(bs); |
706 |
|
DEBUG1("decp: field_pred: ", mb->field_pred); |
707 |
|
|
708 |
|
if (mb->field_pred) { |
709 |
|
mb->field_for_top = BitstreamGetBit(bs); |
710 |
|
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
711 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
712 |
|
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
713 |
} |
} |
|
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
|
{ |
|
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
|
714 |
} |
} |
|
else // MODE_INTRA, MODE_INTRA_Q |
|
|
{ |
|
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
|
|
continue; |
|
715 |
} |
} |
716 |
|
|
717 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
718 |
|
if (dec->interlacing && mb->field_pred) { |
719 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
720 |
|
fcode, bound); |
721 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
722 |
|
fcode, bound); |
723 |
|
} else { |
724 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
725 |
|
fcode, bound); |
726 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
727 |
|
mb->mvs[0].x; |
728 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
729 |
|
mb->mvs[0].y; |
730 |
|
} |
731 |
|
} else if (mb->mode == MODE_INTER4V ) { |
732 |
|
|
733 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
734 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
735 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
736 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
737 |
|
} else // MODE_INTRA, MODE_INTRA_Q |
738 |
|
{ |
739 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
740 |
|
0; |
741 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
742 |
|
0; |
743 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
744 |
|
intra_dc_threshold, bound); |
745 |
|
continue; |
746 |
} |
} |
|
else // not coded |
|
|
{ |
|
747 |
|
|
748 |
|
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
749 |
|
rounding); |
750 |
|
} else // not coded |
751 |
|
{ |
752 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
753 |
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; |
754 |
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; |
757 |
|
|
758 |
start_timer(); |
start_timer(); |
759 |
|
|
760 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
761 |
dec->refn.y + (16*y)*dec->edged_width + (16*x), |
(16 * x), |
762 |
dec->edged_width); |
dec->refn[0].y + (16 * y) * dec->edged_width + |
763 |
|
(16 * x), dec->edged_width); |
764 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
765 |
dec->refn.y + (16*y)*dec->edged_width + (16*x+8), |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
766 |
dec->edged_width); |
(16 * x + 8), |
767 |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
768 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
(16 * x + 8), dec->edged_width); |
769 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x), |
|
770 |
dec->edged_width); |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
771 |
|
(16 * x), |
772 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->refn[0].y + (16 * y + |
773 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), |
8) * dec->edged_width + |
774 |
dec->edged_width); |
(16 * x), dec->edged_width); |
775 |
|
|
776 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
777 |
dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), |
778 |
|
dec->refn[0].y + (16 * y + |
779 |
|
8) * dec->edged_width + |
780 |
|
(16 * x + 8), dec->edged_width); |
781 |
|
|
782 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
783 |
|
(8 * x), |
784 |
|
dec->refn[0].u + |
785 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
786 |
dec->edged_width/2); |
dec->edged_width/2); |
787 |
|
|
788 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
789 |
dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
790 |
|
dec->refn[0].v + |
791 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
792 |
dec->edged_width/2); |
dec->edged_width/2); |
793 |
|
stop_transfer_timer(); |
794 |
|
if(dec->out_frm && cp_mb > 0) { |
795 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
796 |
|
cp_mb = 0; |
797 |
|
} |
798 |
|
st_mb = x+1; |
799 |
|
} |
800 |
|
} |
801 |
|
if(dec->out_frm && cp_mb > 0) |
802 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
803 |
|
} |
804 |
|
} |
805 |
|
|
806 |
|
|
807 |
|
// add by MinChen <chenm001@163.com> |
808 |
|
// decode B-frame motion vector |
809 |
|
void |
810 |
|
get_b_motion_vector(DECODER * dec, |
811 |
|
Bitstream * bs, |
812 |
|
int x, |
813 |
|
int y, |
814 |
|
VECTOR * mv, |
815 |
|
int fcode, |
816 |
|
const VECTOR pmv) |
817 |
|
{ |
818 |
|
int scale_fac = 1 << (fcode - 1); |
819 |
|
int high = (32 * scale_fac) - 1; |
820 |
|
int low = ((-32) * scale_fac); |
821 |
|
int range = (64 * scale_fac); |
822 |
|
|
823 |
|
int mv_x, mv_y; |
824 |
|
int pmv_x, pmv_y; |
825 |
|
|
826 |
|
pmv_x = pmv.x; |
827 |
|
pmv_y = pmv.y; |
828 |
|
|
829 |
|
mv_x = get_mv(bs, fcode); |
830 |
|
mv_y = get_mv(bs, fcode); |
831 |
|
|
832 |
|
mv_x += pmv_x; |
833 |
|
mv_y += pmv_y; |
834 |
|
|
835 |
|
if (mv_x < low) { |
836 |
|
mv_x += range; |
837 |
|
} else if (mv_x > high) { |
838 |
|
mv_x -= range; |
839 |
|
} |
840 |
|
|
841 |
|
if (mv_y < low) { |
842 |
|
mv_y += range; |
843 |
|
} else if (mv_y > high) { |
844 |
|
mv_y -= range; |
845 |
|
} |
846 |
|
|
847 |
|
mv->x = mv_x; |
848 |
|
mv->y = mv_y; |
849 |
|
} |
850 |
|
|
851 |
|
|
852 |
|
// add by MinChen <chenm001@163.com> |
853 |
|
// decode an B-frame forward & backward inter macroblock |
854 |
|
void |
855 |
|
decoder_bf_mbinter(DECODER * dec, |
856 |
|
const MACROBLOCK * pMB, |
857 |
|
const uint32_t x_pos, |
858 |
|
const uint32_t y_pos, |
859 |
|
const uint32_t cbp, |
860 |
|
Bitstream * bs, |
861 |
|
const uint32_t quant, |
862 |
|
const uint8_t ref) |
863 |
|
{ |
864 |
|
|
865 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
866 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
867 |
|
|
868 |
|
uint32_t stride = dec->edged_width; |
869 |
|
uint32_t stride2 = stride / 2; |
870 |
|
uint32_t next_block = stride * 8; |
871 |
|
uint32_t i; |
872 |
|
uint32_t iQuant = pMB->quant; |
873 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
874 |
|
int uv_dx, uv_dy; |
875 |
|
|
876 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
877 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
878 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
879 |
|
|
880 |
|
|
881 |
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
882 |
|
uv_dx = pMB->mvs[0].x; |
883 |
|
uv_dy = pMB->mvs[0].y; |
884 |
|
|
885 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
886 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
887 |
|
} else { |
888 |
|
int sum; |
889 |
|
|
890 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
891 |
|
uv_dx = |
892 |
|
(sum == |
893 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
894 |
|
(ABS(sum) / 16) * 2)); |
895 |
|
|
896 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
897 |
|
uv_dy = |
898 |
|
(sum == |
899 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
900 |
|
(ABS(sum) / 16) * 2)); |
901 |
|
} |
902 |
|
|
903 |
|
start_timer(); |
904 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, 16 * y_pos, |
905 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
906 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
907 |
|
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
908 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
909 |
|
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
910 |
|
0); |
911 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
912 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
913 |
|
0); |
914 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
915 |
|
uv_dx, uv_dy, stride2, 0); |
916 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
917 |
|
uv_dx, uv_dy, stride2, 0); |
918 |
|
stop_comp_timer(); |
919 |
|
|
920 |
|
for (i = 0; i < 6; i++) { |
921 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
922 |
|
|
923 |
|
if (cbp & (1 << (5 - i))) // coded |
924 |
|
{ |
925 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
926 |
|
|
927 |
|
start_timer(); |
928 |
|
get_inter_block(bs, &block[i * 64], direction); |
929 |
|
stop_coding_timer(); |
930 |
|
|
931 |
|
start_timer(); |
932 |
|
if (dec->quant_type == 0) { |
933 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
934 |
|
} else { |
935 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
936 |
|
} |
937 |
|
stop_iquant_timer(); |
938 |
|
|
939 |
|
start_timer(); |
940 |
|
idct(&data[i * 64]); |
941 |
|
stop_idct_timer(); |
942 |
|
} |
943 |
|
} |
944 |
|
|
945 |
|
if (dec->interlacing && pMB->field_dct) { |
946 |
|
next_block = stride; |
947 |
|
stride *= 2; |
948 |
|
} |
949 |
|
|
950 |
|
start_timer(); |
951 |
|
if (cbp & 32) |
952 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
953 |
|
if (cbp & 16) |
954 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
955 |
|
if (cbp & 8) |
956 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
957 |
|
if (cbp & 4) |
958 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
959 |
|
if (cbp & 2) |
960 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
961 |
|
if (cbp & 1) |
962 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
963 |
|
stop_transfer_timer(); |
964 |
|
} |
965 |
|
|
966 |
|
// add by MinChen <chenm001@163.com> |
967 |
|
// decode an B-frame direct & inter macroblock |
968 |
|
void |
969 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
970 |
|
IMAGE forward, |
971 |
|
IMAGE backward, |
972 |
|
const MACROBLOCK * pMB, |
973 |
|
const uint32_t x_pos, |
974 |
|
const uint32_t y_pos, |
975 |
|
Bitstream * bs) |
976 |
|
{ |
977 |
|
|
978 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
979 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
980 |
|
|
981 |
|
uint32_t stride = dec->edged_width; |
982 |
|
uint32_t stride2 = stride / 2; |
983 |
|
uint32_t next_block = stride * 8; |
984 |
|
uint32_t iQuant = pMB->quant; |
985 |
|
int uv_dx, uv_dy; |
986 |
|
int b_uv_dx, b_uv_dy; |
987 |
|
uint32_t i; |
988 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
989 |
|
const uint32_t cbp = pMB->cbp; |
990 |
|
|
991 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
992 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
993 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
994 |
|
|
995 |
|
|
996 |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
997 |
|
uv_dx = pMB->mvs[0].x; |
998 |
|
uv_dy = pMB->mvs[0].y; |
999 |
|
|
1000 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
1001 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
1002 |
|
|
1003 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1004 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1005 |
|
|
1006 |
|
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
1007 |
|
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
1008 |
|
} else { |
1009 |
|
int sum; |
1010 |
|
|
1011 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1012 |
|
uv_dx = |
1013 |
|
(sum == |
1014 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1015 |
|
(ABS(sum) / 16) * 2)); |
1016 |
|
|
1017 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1018 |
|
uv_dy = |
1019 |
|
(sum == |
1020 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1021 |
|
(ABS(sum) / 16) * 2)); |
1022 |
|
|
1023 |
|
sum = |
1024 |
|
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
1025 |
|
pMB->b_mvs[3].x; |
1026 |
|
b_uv_dx = |
1027 |
|
(sum == |
1028 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1029 |
|
(ABS(sum) / 16) * 2)); |
1030 |
|
|
1031 |
|
sum = |
1032 |
|
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
1033 |
|
pMB->b_mvs[3].y; |
1034 |
|
b_uv_dy = |
1035 |
|
(sum == |
1036 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
1037 |
|
(ABS(sum) / 16) * 2)); |
1038 |
|
} |
1039 |
|
|
1040 |
|
|
1041 |
|
start_timer(); |
1042 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1043 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1044 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1045 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1046 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1047 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1048 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1049 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1050 |
|
0); |
1051 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1052 |
|
uv_dy, stride2, 0); |
1053 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1054 |
|
uv_dy, stride2, 0); |
1055 |
|
|
1056 |
|
|
1057 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1058 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1059 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1060 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1061 |
|
0); |
1062 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1063 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1064 |
|
stride, 0); |
1065 |
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1066 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1067 |
|
stride, 0); |
1068 |
|
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1069 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1070 |
|
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1071 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1072 |
|
|
1073 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1074 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1075 |
|
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, |
1076 |
|
stride, 0); |
1077 |
|
|
1078 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1079 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1080 |
|
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1081 |
|
stride, 0); |
1082 |
|
|
1083 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1084 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1085 |
|
dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1086 |
|
stride, 0); |
1087 |
|
|
1088 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1089 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1090 |
|
dec->refn[2].y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1091 |
|
stride, 0); |
1092 |
|
|
1093 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1094 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1095 |
|
dec->refn[2].u + (8 * y_pos * stride2) + 8 * x_pos, |
1096 |
|
stride2, 0); |
1097 |
|
|
1098 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1099 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1100 |
|
dec->refn[2].v + (8 * y_pos * stride2) + 8 * x_pos, |
1101 |
|
stride2, 0); |
1102 |
|
|
1103 |
|
stop_comp_timer(); |
1104 |
|
|
1105 |
|
for (i = 0; i < 6; i++) { |
1106 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1107 |
|
|
1108 |
|
if (cbp & (1 << (5 - i))) // coded |
1109 |
|
{ |
1110 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1111 |
|
|
1112 |
|
start_timer(); |
1113 |
|
get_inter_block(bs, &block[i * 64], direction); |
1114 |
|
stop_coding_timer(); |
1115 |
|
|
1116 |
|
start_timer(); |
1117 |
|
if (dec->quant_type == 0) { |
1118 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1119 |
|
} else { |
1120 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1121 |
|
} |
1122 |
|
stop_iquant_timer(); |
1123 |
|
|
1124 |
|
start_timer(); |
1125 |
|
idct(&data[i * 64]); |
1126 |
|
stop_idct_timer(); |
1127 |
|
} |
1128 |
|
} |
1129 |
|
|
1130 |
|
if (dec->interlacing && pMB->field_dct) { |
1131 |
|
next_block = stride; |
1132 |
|
stride *= 2; |
1133 |
|
} |
1134 |
|
|
1135 |
|
start_timer(); |
1136 |
|
if (cbp & 32) |
1137 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1138 |
|
if (cbp & 16) |
1139 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1140 |
|
if (cbp & 8) |
1141 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1142 |
|
if (cbp & 4) |
1143 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1144 |
|
if (cbp & 2) |
1145 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1146 |
|
if (cbp & 1) |
1147 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1148 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1149 |
} |
} |
1150 |
|
|
1151 |
|
|
1152 |
|
// add by MinChen <chenm001@163.com> |
1153 |
|
// for decode B-frame dbquant |
1154 |
|
int32_t __inline |
1155 |
|
get_dbquant(Bitstream * bs) |
1156 |
|
{ |
1157 |
|
if (!BitstreamGetBit(bs)) // '0' |
1158 |
|
return (0); |
1159 |
|
else if (!BitstreamGetBit(bs)) // '10' |
1160 |
|
return (-2); |
1161 |
|
else |
1162 |
|
return (2); // '11' |
1163 |
} |
} |
1164 |
|
|
1165 |
|
// add by MinChen <chenm001@163.com> |
1166 |
|
// for decode B-frame mb_type |
1167 |
|
// bit ret_value |
1168 |
|
// 1 0 |
1169 |
|
// 01 1 |
1170 |
|
// 001 2 |
1171 |
|
// 0001 3 |
1172 |
|
int32_t __inline |
1173 |
|
get_mbtype(Bitstream * bs) |
1174 |
|
{ |
1175 |
|
int32_t mb_type; |
1176 |
|
|
1177 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) { |
1178 |
|
if (BitstreamGetBit(bs)) |
1179 |
|
break; |
1180 |
} |
} |
1181 |
|
|
1182 |
|
if (mb_type <= 3) |
1183 |
|
return (mb_type); |
1184 |
|
else |
1185 |
|
return (-1); |
1186 |
} |
} |
1187 |
|
|
1188 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
void |
1189 |
|
decoder_bframe(DECODER * dec, |
1190 |
|
Bitstream * bs, |
1191 |
|
int quant, |
1192 |
|
int fcode_forward, |
1193 |
|
int fcode_backward) |
1194 |
{ |
{ |
1195 |
|
uint32_t x, y; |
1196 |
|
VECTOR mv; |
1197 |
|
const VECTOR zeromv = {0,0}; |
1198 |
|
#ifdef BFRAMES_DEC_DEBUG |
1199 |
|
FILE *fp; |
1200 |
|
static char first=0; |
1201 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1202 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1203 |
|
} |
1204 |
|
#endif |
1205 |
|
|
1206 |
|
start_timer(); |
1207 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1208 |
|
dec->width, dec->height); |
1209 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1210 |
|
dec->width, dec->height); |
1211 |
|
stop_edges_timer(); |
1212 |
|
|
1213 |
|
#ifdef BFRAMES_DEC_DEBUG |
1214 |
|
if (!first){ |
1215 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1216 |
|
} |
1217 |
|
#endif |
1218 |
|
|
1219 |
|
for (y = 0; y < dec->mb_height; y++) { |
1220 |
|
// Initialize Pred Motion Vector |
1221 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1222 |
|
for (x = 0; x < dec->mb_width; x++) { |
1223 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1224 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1225 |
|
|
1226 |
|
mv = |
1227 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1228 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1229 |
|
|
1230 |
|
// the last P_VOP is skip macroblock ? |
1231 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1232 |
|
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
1233 |
|
mb->cbp = 0; |
1234 |
|
#ifdef BFRAMES_DEC_DEBUG |
1235 |
|
mb->mb_type = MODE_NOT_CODED; |
1236 |
|
BFRAME_DEBUG |
1237 |
|
#endif |
1238 |
|
mb->mb_type = MODE_FORWARD; |
1239 |
|
mb->quant = last_mb->quant; |
1240 |
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1241 |
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1242 |
|
|
1243 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1244 |
|
continue; |
1245 |
|
} |
1246 |
|
|
1247 |
|
if (!BitstreamGetBit(bs)) { // modb=='0' |
1248 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1249 |
|
|
1250 |
|
mb->mb_type = get_mbtype(bs); |
1251 |
|
|
1252 |
|
if (!modb2) { // modb=='00' |
1253 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1254 |
|
} else { |
1255 |
|
mb->cbp = 0; |
1256 |
|
} |
1257 |
|
if (mb->mb_type && mb->cbp) { |
1258 |
|
quant += get_dbquant(bs); |
1259 |
|
|
1260 |
|
if (quant > 31) { |
1261 |
|
quant = 31; |
1262 |
|
} else if (quant < 1) { |
1263 |
|
quant = 1; |
1264 |
|
} |
1265 |
|
} |
1266 |
|
} else { |
1267 |
|
mb->mb_type = MODE_DIRECT_NONE_MV; |
1268 |
|
mb->cbp = 0; |
1269 |
|
} |
1270 |
|
|
1271 |
|
mb->quant = quant; |
1272 |
|
mb->mode = MODE_INTER4V; |
1273 |
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
1274 |
|
|
1275 |
|
#ifdef BFRAMES_DEC_DEBUG |
1276 |
|
BFRAME_DEBUG |
1277 |
|
#endif |
1278 |
|
|
1279 |
|
switch (mb->mb_type) { |
1280 |
|
case MODE_DIRECT: |
1281 |
|
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1282 |
|
|
1283 |
|
case MODE_DIRECT_NONE_MV: |
1284 |
|
{ |
1285 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1286 |
|
int i; |
1287 |
|
|
1288 |
|
for (i = 0; i < 4; i++) { |
1289 |
|
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1290 |
|
/ TRD + mv.x); |
1291 |
|
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1292 |
|
? ((TRB - TRD) * last_mb->mvs[i].x) |
1293 |
|
/ TRD |
1294 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1295 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1296 |
|
/ TRD + mv.y); |
1297 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1298 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1299 |
|
/ TRD |
1300 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1301 |
|
} |
1302 |
|
//DEBUG("B-frame Direct!\n"); |
1303 |
|
} |
1304 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1305 |
|
mb, x, y, bs); |
1306 |
|
break; |
1307 |
|
|
1308 |
|
case MODE_INTERPOLATE: |
1309 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1310 |
|
dec->p_fmv); |
1311 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1312 |
|
|
1313 |
|
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1314 |
|
fcode_backward, dec->p_bmv); |
1315 |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1316 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1317 |
|
|
1318 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1319 |
|
mb, x, y, bs); |
1320 |
|
//DEBUG("B-frame Bidir!\n"); |
1321 |
|
break; |
1322 |
|
|
1323 |
|
case MODE_BACKWARD: |
1324 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1325 |
|
dec->p_bmv); |
1326 |
|
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1327 |
|
|
1328 |
|
mb->mode = MODE_INTER; |
1329 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1330 |
|
//DEBUG("B-frame Backward!\n"); |
1331 |
|
break; |
1332 |
|
|
1333 |
|
case MODE_FORWARD: |
1334 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1335 |
|
dec->p_fmv); |
1336 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1337 |
|
|
1338 |
|
mb->mode = MODE_INTER; |
1339 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1340 |
|
//DEBUG("B-frame Forward!\n"); |
1341 |
|
break; |
1342 |
|
|
1343 |
|
default: |
1344 |
|
DEBUG1("Not support B-frame mb_type =", mb->mb_type); |
1345 |
|
} |
1346 |
|
|
1347 |
|
} // end of FOR |
1348 |
|
} |
1349 |
|
#ifdef BFRAMES_DEC_DEBUG |
1350 |
|
if (!first){ |
1351 |
|
first=1; |
1352 |
|
if (fp) |
1353 |
|
fclose(fp); |
1354 |
|
} |
1355 |
|
#endif |
1356 |
|
} |
1357 |
|
|
1358 |
|
// swap two MACROBLOCK array |
1359 |
|
void |
1360 |
|
mb_swap(MACROBLOCK ** mb1, |
1361 |
|
MACROBLOCK ** mb2) |
1362 |
|
{ |
1363 |
|
MACROBLOCK *temp = *mb1; |
1364 |
|
|
1365 |
|
*mb1 = *mb2; |
1366 |
|
*mb2 = temp; |
1367 |
|
} |
1368 |
|
|
1369 |
|
int |
1370 |
|
decoder_decode(DECODER * dec, |
1371 |
|
XVID_DEC_FRAME * frame) |
1372 |
|
{ |
1373 |
|
|
1374 |
Bitstream bs; |
Bitstream bs; |
1375 |
uint32_t rounding; |
uint32_t rounding; |
1376 |
uint32_t quant; |
uint32_t quant; |
1377 |
uint32_t fcode; |
uint32_t fcode_forward; |
1378 |
|
uint32_t fcode_backward; |
1379 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1380 |
|
uint32_t vop_type; |
1381 |
|
|
1382 |
start_global_timer(); |
start_global_timer(); |
1383 |
|
|
1384 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1385 |
|
|
1386 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1387 |
|
|
1388 |
switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) |
if(BitstreamShowBits(&bs, 8) == 0x7f) { |
1389 |
{ |
BitstreamGetBits(&bs, 8); |
1390 |
|
return XVID_ERR_FAIL; |
1391 |
|
} |
1392 |
|
|
1393 |
|
// add by chenm001 <chenm001@163.com> |
1394 |
|
// for support B-frame to reference last 2 frame |
1395 |
|
dec->frames++; |
1396 |
|
vop_type = |
1397 |
|
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1398 |
|
&fcode_backward, &intra_dc_threshold); |
1399 |
|
|
1400 |
|
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
1401 |
|
|
1402 |
|
switch (vop_type) { |
1403 |
case P_VOP : |
case P_VOP : |
1404 |
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1405 |
|
intra_dc_threshold); |
1406 |
|
#ifdef BFRAMES_DEC |
1407 |
|
DEBUG1("P_VOP Time=", dec->time); |
1408 |
|
#endif |
1409 |
break; |
break; |
1410 |
|
|
1411 |
case I_VOP : |
case I_VOP : |
|
//DEBUG1("",intra_dc_threshold); |
|
1412 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1413 |
|
#ifdef BFRAMES_DEC |
1414 |
|
DEBUG1("I_VOP Time=", dec->time); |
1415 |
|
#endif |
1416 |
break; |
break; |
1417 |
|
|
1418 |
case B_VOP : // ignore |
case B_VOP: |
1419 |
|
#ifdef BFRAMES_DEC |
1420 |
|
if (dec->time_pp > dec->time_bp) { |
1421 |
|
DEBUG1("B_VOP Time=", dec->time); |
1422 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1423 |
|
} else { |
1424 |
|
DEBUG("broken B-frame!"); |
1425 |
|
} |
1426 |
|
#else |
1427 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1428 |
|
#endif |
1429 |
break; |
break; |
1430 |
|
|
1431 |
case N_VOP : // vop not coded |
case N_VOP : // vop not coded |
1432 |
|
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1433 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1434 |
|
#ifdef BFRAMES_DEC |
1435 |
|
DEBUG1("N_VOP Time=", dec->time); |
1436 |
|
#endif |
1437 |
break; |
break; |
1438 |
|
|
1439 |
default : |
default : |
1440 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1441 |
} |
} |
1442 |
|
|
1443 |
|
#ifdef BFRAMES_DEC_DEBUG |
1444 |
|
if (frame->length != BitstreamPos(&bs) / 8){ |
1445 |
|
DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); |
1446 |
|
} |
1447 |
|
#endif |
1448 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1449 |
|
|
1450 |
start_timer(); |
|
1451 |
|
#ifdef BFRAMES_DEC |
1452 |
|
// test if no B_VOP |
1453 |
|
if (dec->low_delay || dec->frames == 0) { |
1454 |
|
#endif |
1455 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1456 |
frame->image, frame->stride, frame->colorspace); |
frame->image, frame->stride, frame->colorspace); |
1457 |
|
|
1458 |
|
#ifdef BFRAMES_DEC |
1459 |
|
} else { |
1460 |
|
if (dec->frames >= 1) { |
1461 |
|
start_timer(); |
1462 |
|
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1463 |
|
image_output(&dec->refn[0], dec->width, dec->height, |
1464 |
|
dec->edged_width, frame->image, frame->stride, |
1465 |
|
frame->colorspace); |
1466 |
|
} else if (vop_type == B_VOP) { |
1467 |
|
image_output(&dec->cur, dec->width, dec->height, |
1468 |
|
dec->edged_width, frame->image, frame->stride, |
1469 |
|
frame->colorspace); |
1470 |
|
} |
1471 |
stop_conv_timer(); |
stop_conv_timer(); |
1472 |
|
} |
1473 |
|
} |
1474 |
|
#endif |
1475 |
|
|
1476 |
|
if (vop_type == I_VOP || vop_type == P_VOP) { |
1477 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1478 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1479 |
|
|
1480 |
|
// swap MACROBLOCK |
1481 |
|
// the Divx will not set the low_delay flage some times |
1482 |
|
// so follow code will wrong to not swap at that time |
1483 |
|
// this will broken bitstream! so I'm change it, |
1484 |
|
// But that is not the best way! can anyone tell me how |
1485 |
|
// to do another way? |
1486 |
|
// 18-07-2002 MinChen<chenm001@163.com> |
1487 |
|
//if (!dec->low_delay && vop_type == P_VOP) |
1488 |
|
if (vop_type == P_VOP) |
1489 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1490 |
|
} |
1491 |
|
|
1492 |
emms(); |
emms(); |
1493 |
|
|