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 |
* 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: decoder.c,v 1.37.2.23 2002-12-16 08:54:44 suxen_drol Exp $ |
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" |
79 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
80 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
81 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
82 |
#include "utils/mbfunctions.h" |
#include "image/reduced.h" |
83 |
|
#include "image/font.h" |
84 |
|
|
85 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
86 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
87 |
#include "utils/timer.h" |
#include "utils/timer.h" |
88 |
#include "utils/emms.h" |
#include "utils/emms.h" |
89 |
|
#include "motion/motion.h" |
90 |
|
|
91 |
#include "image/image.h" |
#include "image/image.h" |
92 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
93 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
94 |
|
|
95 |
int decoder_create(XVID_DEC_PARAM * param) |
int |
96 |
|
decoder_resize(DECODER * dec) |
97 |
{ |
{ |
98 |
DECODER * dec; |
/* free existing */ |
99 |
|
|
100 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
101 |
if (dec == NULL) |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
102 |
{ |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
103 |
return XVID_ERR_MEMORY; |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
104 |
} |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
105 |
param->handle = dec; |
|
106 |
|
if (dec->last_mbs) |
107 |
|
xvid_free(dec->last_mbs); |
108 |
|
if (dec->mbs) |
109 |
|
xvid_free(dec->mbs); |
110 |
|
|
111 |
dec->width = param->width; |
/* realloc */ |
|
dec->height = param->height; |
|
112 |
|
|
113 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
114 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
116 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
117 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
118 |
|
|
119 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&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 |
|
|
124 |
if (image_create(&dec->refn, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
125 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
126 |
xvid_free(dec); |
xvid_free(dec); |
127 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
128 |
} |
} |
129 |
|
|
130 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
// add by chenm001 <chenm001@163.com> |
131 |
if (dec->mbs == NULL) |
// for support B-frame to reference last 2 frame |
132 |
{ |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
133 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
134 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
135 |
|
xvid_free(dec); |
136 |
|
return XVID_ERR_MEMORY; |
137 |
|
} |
138 |
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
139 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
141 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
142 |
|
xvid_free(dec); |
143 |
|
return XVID_ERR_MEMORY; |
144 |
|
} |
145 |
|
|
146 |
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
147 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
148 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
149 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
150 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
151 |
|
xvid_free(dec); |
152 |
|
return XVID_ERR_MEMORY; |
153 |
|
} |
154 |
|
|
155 |
|
dec->mbs = |
156 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
157 |
|
CACHE_LINE); |
158 |
|
if (dec->mbs == NULL) { |
159 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
160 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
161 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
162 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
163 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
164 |
|
xvid_free(dec); |
165 |
|
return XVID_ERR_MEMORY; |
166 |
|
} |
167 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
168 |
|
|
169 |
|
// add by chenm001 <chenm001@163.com> |
170 |
|
// for skip MB flag |
171 |
|
dec->last_mbs = |
172 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
173 |
|
CACHE_LINE); |
174 |
|
if (dec->last_mbs == NULL) { |
175 |
|
xvid_free(dec->mbs); |
176 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
177 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
178 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
179 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
180 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
181 |
xvid_free(dec); |
xvid_free(dec); |
182 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
183 |
} |
} |
184 |
|
|
185 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
186 |
|
|
187 |
|
return XVID_ERR_OK; |
188 |
|
} |
189 |
|
|
190 |
|
|
191 |
|
int |
192 |
|
decoder_create(XVID_DEC_PARAM * param) |
193 |
|
{ |
194 |
|
DECODER *dec; |
195 |
|
|
196 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
197 |
|
if (dec == NULL) { |
198 |
|
return XVID_ERR_MEMORY; |
199 |
|
} |
200 |
|
memset(dec, 0, sizeof(DECODER)); |
201 |
|
|
202 |
|
param->handle = dec; |
203 |
|
|
204 |
|
dec->width = param->width; |
205 |
|
dec->height = param->height; |
206 |
|
|
207 |
|
image_null(&dec->cur); |
208 |
|
image_null(&dec->refn[0]); |
209 |
|
image_null(&dec->refn[1]); |
210 |
|
image_null(&dec->tmp); |
211 |
|
image_null(&dec->qtmp); |
212 |
|
|
213 |
|
dec->mbs = NULL; |
214 |
|
dec->last_mbs = NULL; |
215 |
|
|
216 |
init_timer(); |
init_timer(); |
|
create_vlc_tables(); |
|
217 |
|
|
218 |
|
// add by chenm001 <chenm001@163.com> |
219 |
|
// for support B-frame to save reference frame's time |
220 |
|
dec->frames = 0; |
221 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
222 |
|
dec->low_delay = 0; |
223 |
|
dec->packed_mode = 0; |
224 |
|
|
225 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
226 |
|
|
227 |
|
if (dec->fixed_dimensions) |
228 |
|
return decoder_resize(dec); |
229 |
|
else |
230 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
231 |
} |
} |
232 |
|
|
233 |
|
|
234 |
int decoder_destroy(DECODER * dec) |
int |
235 |
|
decoder_destroy(DECODER * dec) |
236 |
{ |
{ |
237 |
|
xvid_free(dec->last_mbs); |
238 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
239 |
image_destroy(&dec->refn, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
240 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
241 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
242 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
243 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
244 |
xvid_free(dec); |
xvid_free(dec); |
245 |
|
|
|
destroy_vlc_tables(); |
|
|
|
|
246 |
write_timer(); |
write_timer(); |
247 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
248 |
} |
} |
249 |
|
|
250 |
|
|
251 |
|
|
252 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
253 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
254 |
}; |
}; |
255 |
|
|
256 |
|
|
257 |
|
|
258 |
|
|
259 |
// decode an intra macroblock |
// decode an intra macroblock |
260 |
|
|
261 |
void decoder_mbintra(DECODER * dec, |
void |
262 |
|
decoder_mbintra(DECODER * dec, |
263 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
264 |
const uint32_t x_pos, |
const uint32_t x_pos, |
265 |
const uint32_t y_pos, |
const uint32_t y_pos, |
267 |
const uint32_t cbp, |
const uint32_t cbp, |
268 |
Bitstream * bs, |
Bitstream * bs, |
269 |
const uint32_t quant, |
const uint32_t quant, |
270 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
271 |
|
const unsigned int bound, |
272 |
|
const int reduced_resolution) |
273 |
{ |
{ |
|
CACHE_ALIGN int16_t block[6][64]; |
|
|
CACHE_ALIGN int16_t data[6][64]; |
|
274 |
|
|
275 |
const uint32_t stride = dec->edged_width; |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
276 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
277 |
|
|
278 |
|
uint32_t stride = dec->edged_width; |
279 |
|
uint32_t stride2 = stride / 2; |
280 |
|
uint32_t next_block = stride * 8; |
281 |
uint32_t i; |
uint32_t i; |
282 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
283 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
284 |
|
|
285 |
|
if (reduced_resolution) { |
286 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
287 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
288 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
289 |
|
}else{ |
290 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
291 |
pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
292 |
pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
293 |
|
} |
294 |
|
|
295 |
memset(block, 0, sizeof(block)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
296 |
|
|
297 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
298 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
299 |
int16_t predictors[8]; |
int16_t predictors[8]; |
300 |
int start_coeff; |
int start_coeff; |
301 |
|
|
302 |
start_timer(); |
start_timer(); |
303 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, block[i], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
304 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
305 |
{ |
if (!acpred_flag) { |
306 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
307 |
} |
} |
308 |
stop_prediction_timer(); |
stop_prediction_timer(); |
309 |
|
|
310 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
311 |
int dc_size; |
int dc_size; |
312 |
int dc_dif; |
int dc_dif; |
313 |
|
|
314 |
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); |
315 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
316 |
|
|
317 |
if (dc_size > 8) |
if (dc_size > 8) { |
|
{ |
|
318 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
319 |
} |
} |
320 |
|
|
321 |
block[i][0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
322 |
start_coeff = 1; |
start_coeff = 1; |
323 |
} |
|
324 |
else |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
325 |
{ |
} else { |
326 |
start_coeff = 0; |
start_coeff = 0; |
327 |
} |
} |
328 |
|
|
329 |
start_timer(); |
start_timer(); |
330 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
331 |
{ |
{ |
332 |
get_intra_block(bs, block[i], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
333 |
|
2 : pMB->acpred_directions[i]; |
334 |
|
|
335 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
336 |
} |
} |
337 |
stop_coding_timer(); |
stop_coding_timer(); |
338 |
|
|
339 |
start_timer(); |
start_timer(); |
340 |
add_acdc(pMB, i, block[i], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
341 |
stop_prediction_timer(); |
stop_prediction_timer(); |
342 |
|
|
343 |
start_timer(); |
start_timer(); |
344 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
345 |
{ |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
346 |
dequant_intra(data[i], block[i], iQuant, iDcScaler); |
} else { |
347 |
} |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
|
else |
|
|
{ |
|
|
dequant4_intra(data[i], block[i], iQuant, iDcScaler); |
|
348 |
} |
} |
349 |
stop_iquant_timer(); |
stop_iquant_timer(); |
350 |
|
|
351 |
start_timer(); |
start_timer(); |
352 |
idct(data[i]); |
idct(&data[i * 64]); |
353 |
stop_idct_timer(); |
stop_idct_timer(); |
354 |
|
|
355 |
} |
} |
356 |
|
|
357 |
start_timer(); |
if (dec->interlacing && pMB->field_dct) { |
358 |
if (dec->interlacing && pMB->field_dct) |
next_block = stride; |
359 |
{ |
stride *= 2; |
|
MBFieldToFrame(data); |
|
360 |
} |
} |
|
stop_interlacing_timer(); |
|
361 |
|
|
362 |
start_timer(); |
start_timer(); |
363 |
transfer_16to8copy(pY_Cur, data[0], stride); |
|
364 |
transfer_16to8copy(pY_Cur + 8, data[1], stride); |
if (reduced_resolution) |
365 |
transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); |
{ |
366 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); |
next_block*=2; |
367 |
transfer_16to8copy(pU_Cur, data[4], stride / 2); |
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
368 |
transfer_16to8copy(pV_Cur, data[5], stride / 2); |
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
369 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
370 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
371 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
372 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
373 |
|
}else{ |
374 |
|
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
375 |
|
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
376 |
|
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
377 |
|
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
378 |
|
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
379 |
|
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
380 |
|
} |
381 |
stop_transfer_timer(); |
stop_transfer_timer(); |
382 |
} |
} |
383 |
|
|
387 |
|
|
388 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
389 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
|
static const uint32_t roundtab[16] = |
|
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
390 |
|
|
391 |
// decode an inter macroblock |
// decode an inter macroblock |
392 |
|
|
393 |
void decoder_mbinter(DECODER * dec, |
void |
394 |
|
decoder_mbinter(DECODER * dec, |
395 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
396 |
const uint32_t x_pos, |
const uint32_t x_pos, |
397 |
const uint32_t y_pos, |
const uint32_t y_pos, |
399 |
const uint32_t cbp, |
const uint32_t cbp, |
400 |
Bitstream * bs, |
Bitstream * bs, |
401 |
const uint32_t quant, |
const uint32_t quant, |
402 |
const uint32_t rounding) |
const uint32_t rounding, |
403 |
|
const int reduced_resolution) |
404 |
{ |
{ |
|
CACHE_ALIGN int16_t block[6][64]; |
|
|
CACHE_ALIGN int16_t data[6][64]; |
|
405 |
|
|
406 |
const uint32_t stride = dec->edged_width; |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
407 |
const uint32_t stride2 = dec->edged_width / 2; |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
408 |
|
|
409 |
|
uint32_t stride = dec->edged_width; |
410 |
|
uint32_t stride2 = stride / 2; |
411 |
|
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
412 |
uint32_t i; |
uint32_t i; |
413 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
414 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
415 |
|
|
416 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
417 |
|
VECTOR mv[4]; /* local copy of mvs */ |
418 |
|
|
419 |
|
if (reduced_resolution) { |
420 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
421 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
422 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
423 |
|
for (i = 0; i < 4; i++) { |
424 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
425 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
426 |
|
} |
427 |
|
}else{ |
428 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
429 |
pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
430 |
pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
431 |
|
for (i = 0; i < 4; i++) |
432 |
|
mv[i] = pMB->mvs[i]; |
433 |
|
} |
434 |
|
|
435 |
|
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
436 |
|
uv_dx = mv[0].x; |
437 |
|
uv_dy = mv[0].y; |
438 |
|
|
439 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
if (dec->quarterpel) |
440 |
{ |
{ |
441 |
uv_dx = pMB->mvs[0].x; |
uv_dx /= 2; |
442 |
uv_dy = pMB->mvs[0].y; |
uv_dy /= 2; |
443 |
|
} |
444 |
|
|
445 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
446 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
447 |
|
|
448 |
|
start_timer(); |
449 |
|
if (reduced_resolution) |
450 |
|
{ |
451 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
452 |
|
mv[0].x, mv[0].y, stride, rounding); |
453 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
454 |
|
uv_dx, uv_dy, stride2, rounding); |
455 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
456 |
|
uv_dx, uv_dy, stride2, rounding); |
457 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
458 |
} |
} |
459 |
else |
else |
460 |
{ |
{ |
461 |
int sum; |
if(dec->quarterpel) { |
462 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
463 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
464 |
|
mv[0].x, mv[0].y, stride, rounding); |
465 |
|
} |
466 |
|
else { |
467 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
468 |
|
mv[0].x, mv[0].y, stride, rounding); |
469 |
|
} |
470 |
|
|
471 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
472 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx, uv_dy, stride2, rounding); |
473 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
474 |
|
uv_dx, uv_dy, stride2, rounding); |
475 |
} |
} |
476 |
|
stop_comp_timer(); |
477 |
|
|
478 |
|
} else { /* MODE_INTER4V */ |
479 |
|
int sum; |
480 |
|
|
481 |
|
if(dec->quarterpel) |
482 |
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
483 |
|
else |
484 |
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
485 |
|
|
486 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
487 |
|
|
488 |
|
if(dec->quarterpel) |
489 |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
490 |
|
else |
491 |
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
492 |
|
|
493 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
494 |
|
|
495 |
start_timer(); |
start_timer(); |
496 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
if (reduced_resolution) |
497 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
{ |
498 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
499 |
interpolate8x8_switch(dec->cur.y, dec->refn.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
500 |
interpolate8x8_switch(dec->cur.u, dec->refn.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos, |
501 |
interpolate8x8_switch(dec->cur.v, dec->refn.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
502 |
stop_comp_timer(); |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos + 16, |
503 |
|
mv[2].x, mv[2].y, stride, rounding); |
504 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos + 16, 32*y_pos + 16, |
505 |
|
mv[3].x, mv[3].y, stride, rounding); |
506 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
507 |
|
uv_dx, uv_dy, stride2, rounding); |
508 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
509 |
|
uv_dx, uv_dy, stride2, rounding); |
510 |
|
|
511 |
for (i = 0; i < 6; i++) |
// set_block(pY_Cur, stride, 32, 32, 127); |
512 |
|
} |
513 |
|
else |
514 |
{ |
{ |
515 |
|
if(dec->quarterpel) { |
516 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
517 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
518 |
|
mv[0].x, mv[0].y, stride, rounding); |
519 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
520 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
521 |
|
mv[1].x, mv[1].y, stride, rounding); |
522 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
523 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
524 |
|
mv[2].x, mv[2].y, stride, rounding); |
525 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
526 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
527 |
|
mv[3].x, mv[3].y, stride, rounding); |
528 |
|
} |
529 |
|
else { |
530 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
531 |
|
mv[0].x, mv[0].y, stride, rounding); |
532 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
533 |
|
mv[1].x, mv[1].y, stride, rounding); |
534 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
535 |
|
mv[2].x, mv[2].y, stride, rounding); |
536 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
537 |
|
mv[3].x, mv[3].y, stride, rounding); |
538 |
|
} |
539 |
|
|
540 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
541 |
|
uv_dx, uv_dy, stride2, rounding); |
542 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
543 |
|
uv_dx, uv_dy, stride2, rounding); |
544 |
|
} |
545 |
|
stop_comp_timer(); |
546 |
|
} |
547 |
|
|
548 |
|
for (i = 0; i < 6; i++) { |
549 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
550 |
|
|
551 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
552 |
{ |
{ |
553 |
memset(block[i], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
554 |
|
|
555 |
start_timer(); |
start_timer(); |
556 |
get_inter_block(bs, block[i]); |
get_inter_block(bs, &block[i * 64], direction); |
557 |
stop_coding_timer(); |
stop_coding_timer(); |
558 |
|
|
559 |
start_timer(); |
start_timer(); |
560 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
561 |
{ |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
562 |
dequant_inter(data[i], block[i], iQuant); |
} else { |
563 |
} |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
else |
|
|
{ |
|
|
dequant4_inter(data[i], block[i], iQuant); |
|
564 |
} |
} |
565 |
stop_iquant_timer(); |
stop_iquant_timer(); |
566 |
|
|
567 |
start_timer(); |
start_timer(); |
568 |
idct(data[i]); |
idct(&data[i * 64]); |
569 |
stop_idct_timer(); |
stop_idct_timer(); |
570 |
} |
} |
571 |
} |
} |
572 |
|
|
573 |
start_timer(); |
if (dec->interlacing && pMB->field_dct) { |
574 |
if (pMB->field_dct) |
next_block = stride; |
575 |
{ |
stride *= 2; |
|
MBFieldToFrame(data); |
|
576 |
} |
} |
|
stop_interlacing_timer(); |
|
577 |
|
|
578 |
start_timer(); |
start_timer(); |
579 |
|
if (reduced_resolution) |
580 |
|
{ |
581 |
|
if (cbp & 32) |
582 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
583 |
|
if (cbp & 16) |
584 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
585 |
|
if (cbp & 8) |
586 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
587 |
|
if (cbp & 4) |
588 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
589 |
|
if (cbp & 2) |
590 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
591 |
|
if (cbp & 1) |
592 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
593 |
|
} |
594 |
|
else |
595 |
|
{ |
596 |
if (cbp & 32) |
if (cbp & 32) |
597 |
transfer_16to8add(pY_Cur, data[0], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
598 |
if (cbp & 16) |
if (cbp & 16) |
599 |
transfer_16to8add(pY_Cur + 8, data[1], stride); |
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
600 |
if (cbp & 8) |
if (cbp & 8) |
601 |
transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
602 |
if (cbp & 4) |
if (cbp & 4) |
603 |
transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
604 |
if (cbp & 2) |
if (cbp & 2) |
605 |
transfer_16to8add(pU_Cur, data[4], stride / 2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
606 |
if (cbp & 1) |
if (cbp & 1) |
607 |
transfer_16to8add(pV_Cur, data[5], stride / 2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
608 |
|
} |
609 |
stop_transfer_timer(); |
stop_transfer_timer(); |
610 |
} |
} |
611 |
|
|
612 |
|
|
613 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
614 |
|
decoder_iframe(DECODER * dec, |
615 |
|
Bitstream * bs, |
616 |
|
int reduced_resolution, |
617 |
|
int quant, |
618 |
|
int intra_dc_threshold) |
619 |
{ |
{ |
620 |
|
uint32_t bound; |
621 |
uint32_t x, y; |
uint32_t x, y; |
622 |
|
uint32_t mb_width = dec->mb_width; |
623 |
|
uint32_t mb_height = dec->mb_height; |
624 |
|
|
625 |
for (y = 0; y < dec->mb_height; y++) |
if (reduced_resolution) |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
626 |
{ |
{ |
627 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
mb_width = (dec->width + 31) / 32; |
628 |
|
mb_height = (dec->height + 31) / 32; |
629 |
|
} |
630 |
|
|
631 |
|
bound = 0; |
632 |
|
|
633 |
|
for (y = 0; y < mb_height; y++) { |
634 |
|
for (x = 0; x < mb_width; x++) { |
635 |
|
MACROBLOCK *mb; |
636 |
uint32_t mcbpc; |
uint32_t mcbpc; |
637 |
uint32_t cbpc; |
uint32_t cbpc; |
638 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
639 |
uint32_t cbpy; |
uint32_t cbpy; |
640 |
uint32_t cbp; |
uint32_t cbp; |
641 |
|
|
642 |
|
while (BitstreamShowBits(bs, 9) == 1) |
643 |
|
BitstreamSkip(bs, 9); |
644 |
|
|
645 |
|
if (check_resync_marker(bs, 0)) |
646 |
|
{ |
647 |
|
bound = read_video_packet_header(bs, dec, 0, |
648 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
649 |
|
x = bound % mb_width; |
650 |
|
y = bound / mb_width; |
651 |
|
} |
652 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
653 |
|
|
654 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
655 |
|
|
656 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
657 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
658 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
659 |
|
|
660 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
661 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
662 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
663 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
664 |
|
|
665 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
666 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
667 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
668 |
quant = 31; |
quant = 31; |
669 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
670 |
quant = 1; |
quant = 1; |
671 |
} |
} |
672 |
} |
} |
673 |
mb->quant = quant; |
mb->quant = quant; |
674 |
|
mb->mvs[0].x = mb->mvs[0].y = |
675 |
|
mb->mvs[1].x = mb->mvs[1].y = |
676 |
|
mb->mvs[2].x = mb->mvs[2].y = |
677 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
678 |
|
|
679 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
680 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
681 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
682 |
} |
} |
683 |
|
|
684 |
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, |
685 |
|
intra_dc_threshold, bound, reduced_resolution); |
686 |
|
|
687 |
} |
} |
688 |
|
if(dec->out_frm) |
689 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
690 |
} |
} |
691 |
|
|
692 |
} |
} |
693 |
|
|
694 |
|
|
695 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
696 |
|
get_motion_vector(DECODER * dec, |
697 |
|
Bitstream * bs, |
698 |
|
int x, |
699 |
|
int y, |
700 |
|
int k, |
701 |
|
VECTOR * ret_mv, |
702 |
|
int fcode, |
703 |
|
const int bound) |
704 |
{ |
{ |
705 |
|
|
706 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
707 |
int high = (32 * scale_fac) - 1; |
int high = (32 * scale_fac) - 1; |
708 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
709 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
710 |
|
|
711 |
VECTOR pmv[4]; |
VECTOR pmv; |
712 |
uint32_t psad[4]; |
VECTOR mv; |
|
|
|
|
int mv_x, mv_y; |
|
|
int pmv_x, pmv_y; |
|
|
|
|
713 |
|
|
714 |
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
715 |
|
|
716 |
pmv_x = pmv[0].x; |
mv.x = get_mv(bs, fcode); |
717 |
pmv_y = pmv[0].y; |
mv.y = get_mv(bs, fcode); |
718 |
|
|
719 |
mv_x = get_mv(bs, fcode); |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv.x, mv.y, pmv.x, pmv.y); |
|
mv_y = get_mv(bs, fcode); |
|
720 |
|
|
721 |
mv_x += pmv_x; |
mv.x += pmv.x; |
722 |
mv_y += pmv_y; |
mv.y += pmv.y; |
723 |
|
|
724 |
if (mv_x < low) |
if (mv.x < low) { |
725 |
{ |
mv.x += range; |
726 |
mv_x += range; |
} else if (mv.x > high) { |
727 |
|
mv.x -= range; |
728 |
} |
} |
729 |
else if (mv_x > high) |
|
730 |
{ |
if (mv.y < low) { |
731 |
mv_x -= range; |
mv.y += range; |
732 |
|
} else if (mv.y > high) { |
733 |
|
mv.y -= range; |
734 |
} |
} |
735 |
|
|
736 |
if (mv_y < low) |
ret_mv->x = mv.x; |
737 |
{ |
ret_mv->y = mv.y; |
|
mv_y += range; |
|
738 |
} |
} |
739 |
else if (mv_y > high) |
|
740 |
|
|
741 |
|
|
742 |
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
743 |
{ |
{ |
744 |
mv_y -= range; |
int length = 1 << (fcode+4); |
|
} |
|
745 |
|
|
746 |
mv->x = mv_x; |
if (quarterpel) value *= 2; |
|
mv->y = mv_y; |
|
747 |
|
|
748 |
|
if (value < -length) |
749 |
|
return -length; |
750 |
|
else if (value >= length) |
751 |
|
return length-1; |
752 |
|
else return value; |
753 |
} |
} |
754 |
|
|
755 |
|
|
756 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
/* for P_VOP set gmc_mv to NULL */ |
757 |
|
void |
758 |
|
decoder_pframe(DECODER * dec, |
759 |
|
Bitstream * bs, |
760 |
|
int rounding, |
761 |
|
int reduced_resolution, |
762 |
|
int quant, |
763 |
|
int fcode, |
764 |
|
int intra_dc_threshold, |
765 |
|
VECTOR * gmc_mv) |
766 |
{ |
{ |
767 |
|
|
768 |
uint32_t x, y; |
uint32_t x, y; |
769 |
|
uint32_t bound; |
770 |
|
int cp_mb, st_mb; |
771 |
|
uint32_t mb_width = dec->mb_width; |
772 |
|
uint32_t mb_height = dec->mb_height; |
773 |
|
|
774 |
image_swap(&dec->cur, &dec->refn); |
if (reduced_resolution) |
775 |
|
{ |
776 |
|
mb_width = (dec->width + 31) / 32; |
777 |
|
mb_height = (dec->height + 31) / 32; |
778 |
|
} |
779 |
|
|
780 |
start_timer(); |
start_timer(); |
781 |
image_setedges(&dec->refn, dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
782 |
|
dec->width, dec->height); |
783 |
stop_edges_timer(); |
stop_edges_timer(); |
784 |
|
|
785 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
786 |
{ |
|
787 |
for (x = 0; x < dec->mb_width; x++) |
for (y = 0; y < mb_height; y++) { |
788 |
|
cp_mb = st_mb = 0; |
789 |
|
for (x = 0; x < mb_width; x++) { |
790 |
|
MACROBLOCK *mb; |
791 |
|
|
792 |
|
// skip stuffing |
793 |
|
while (BitstreamShowBits(bs, 10) == 1) |
794 |
|
BitstreamSkip(bs, 10); |
795 |
|
|
796 |
|
if (check_resync_marker(bs, fcode - 1)) |
797 |
{ |
{ |
798 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
bound = read_video_packet_header(bs, dec, fcode - 1, |
799 |
|
&quant, &fcode, NULL, &intra_dc_threshold); |
800 |
|
x = bound % mb_width; |
801 |
|
y = bound / mb_width; |
802 |
|
} |
803 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
804 |
|
|
805 |
if (!BitstreamGetBit(bs)) // not_coded |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
806 |
|
|
807 |
|
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
808 |
|
if (!(BitstreamGetBit(bs))) // not_coded |
809 |
{ |
{ |
810 |
uint32_t mcbpc; |
uint32_t mcbpc; |
811 |
uint32_t cbpc; |
uint32_t cbpc; |
813 |
uint32_t cbpy; |
uint32_t cbpy; |
814 |
uint32_t cbp; |
uint32_t cbp; |
815 |
uint32_t intra; |
uint32_t intra; |
816 |
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
817 |
|
|
818 |
|
cp_mb++; |
819 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
820 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
821 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
822 |
|
|
823 |
|
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
824 |
|
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
825 |
acpred_flag = 0; |
acpred_flag = 0; |
826 |
|
|
827 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
828 |
|
|
829 |
if (intra) |
if (intra) { |
|
{ |
|
830 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
831 |
} |
} |
832 |
|
|
833 |
if (mb->mode == MODE_STUFFING) |
if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
834 |
{ |
{ |
835 |
DEBUG("-- STUFFING ?"); |
mcsel = BitstreamGetBit(bs); |
|
continue; |
|
836 |
} |
} |
837 |
|
|
838 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
839 |
|
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
840 |
|
|
841 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
842 |
|
|
843 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
844 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
845 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
846 |
if (quant > 31) |
quant += dquant; |
847 |
{ |
if (quant > 31) { |
848 |
quant = 31; |
quant = 31; |
849 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
850 |
quant = 1; |
quant = 1; |
851 |
} |
} |
852 |
|
DPRINTF(DPRINTF_MB, "quant %i", quant); |
853 |
} |
} |
854 |
mb->quant = quant; |
mb->quant = quant; |
855 |
|
|
856 |
if (dec->interlacing) |
if (dec->interlacing) { |
857 |
{ |
if (cbp || intra) { |
858 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
859 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
860 |
|
} |
861 |
|
|
862 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
863 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
864 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
865 |
|
|
866 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
867 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
868 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
869 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
870 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
871 |
} |
} |
872 |
} |
} |
873 |
} |
} |
874 |
|
|
875 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
876 |
{ |
|
877 |
if (dec->interlacing && mb->field_pred) |
if (mcsel) |
|
{ |
|
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
|
|
} |
|
|
else |
|
|
{ |
|
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
|
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
|
} |
|
|
} |
|
|
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
878 |
{ |
{ |
879 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
880 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
881 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
|
882 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
} else if (dec->interlacing && mb->field_pred) { |
883 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
884 |
|
fcode, bound); |
885 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
886 |
|
fcode, bound); |
887 |
|
} else { |
888 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
889 |
|
fcode, bound); |
890 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
891 |
|
mb->mvs[0].x; |
892 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
893 |
|
mb->mvs[0].y; |
894 |
} |
} |
895 |
else // MODE_INTRA, MODE_INTRA_Q |
} else if (mb->mode == MODE_INTER4V ) { |
896 |
|
|
897 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
898 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
899 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
900 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
901 |
|
} else // MODE_INTRA, MODE_INTRA_Q |
902 |
{ |
{ |
903 |
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 = |
904 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
0; |
905 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
906 |
|
0; |
907 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
908 |
|
intra_dc_threshold, bound, reduced_resolution); |
909 |
continue; |
continue; |
910 |
} |
} |
911 |
|
|
912 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
913 |
|
rounding, reduced_resolution); |
914 |
|
|
915 |
} |
} |
916 |
else // not coded |
else if (gmc_mv) /* not coded S_VOP macroblock */ |
917 |
|
{ |
918 |
|
mb->mode = MODE_NOT_CODED; |
919 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
920 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
921 |
|
decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution); |
922 |
|
} |
923 |
|
else /* not coded P_VOP macroblock */ |
924 |
{ |
{ |
|
|
|
925 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
926 |
|
|
927 |
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; |
928 |
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; |
|
|
|
929 |
// copy macroblock directly from ref to cur |
// copy macroblock directly from ref to cur |
930 |
|
|
931 |
start_timer(); |
start_timer(); |
932 |
|
|
933 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
if (reduced_resolution) |
934 |
dec->refn.y + (16*y)*dec->edged_width + (16*x), |
{ |
935 |
dec->edged_width); |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
936 |
|
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
|
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
|
dec->refn.y + (16*y)*dec->edged_width + (16*x+8), |
|
937 |
dec->edged_width); |
dec->edged_width); |
938 |
|
|
939 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
940 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x), |
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
941 |
dec->edged_width); |
dec->edged_width/2); |
942 |
|
|
943 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
944 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), |
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
945 |
|
dec->edged_width/2); |
946 |
|
} |
947 |
|
else |
948 |
|
{ |
949 |
|
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
950 |
|
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
951 |
dec->edged_width); |
dec->edged_width); |
952 |
|
|
953 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
954 |
dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
955 |
dec->edged_width/2); |
dec->edged_width/2); |
956 |
|
|
957 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
958 |
dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
959 |
dec->edged_width/2); |
dec->edged_width/2); |
960 |
|
} |
961 |
|
|
962 |
stop_transfer_timer(); |
stop_transfer_timer(); |
963 |
|
|
964 |
|
if(dec->out_frm && cp_mb > 0) { |
965 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
966 |
|
cp_mb = 0; |
967 |
|
} |
968 |
|
st_mb = x+1; |
969 |
} |
} |
970 |
} |
} |
971 |
|
if(dec->out_frm && cp_mb > 0) |
972 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
973 |
} |
} |
974 |
} |
} |
975 |
|
|
976 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
|
977 |
|
// add by MinChen <chenm001@163.com> |
978 |
|
// decode B-frame motion vector |
979 |
|
void |
980 |
|
get_b_motion_vector(DECODER * dec, |
981 |
|
Bitstream * bs, |
982 |
|
int x, |
983 |
|
int y, |
984 |
|
VECTOR * mv, |
985 |
|
int fcode, |
986 |
|
const VECTOR pmv) |
987 |
{ |
{ |
988 |
Bitstream bs; |
int scale_fac = 1 << (fcode - 1); |
989 |
uint32_t rounding; |
int high = (32 * scale_fac) - 1; |
990 |
uint32_t quant; |
int low = ((-32) * scale_fac); |
991 |
uint32_t fcode; |
int range = (64 * scale_fac); |
|
uint32_t intra_dc_threshold; |
|
992 |
|
|
993 |
start_global_timer(); |
int mv_x, mv_y; |
994 |
|
int pmv_x, pmv_y; |
995 |
|
|
996 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
pmv_x = pmv.x; |
997 |
|
pmv_y = pmv.y; |
998 |
|
|
999 |
switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) |
mv_x = get_mv(bs, fcode); |
1000 |
{ |
mv_y = get_mv(bs, fcode); |
|
case P_VOP : |
|
|
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
|
|
break; |
|
1001 |
|
|
1002 |
case I_VOP : |
mv_x += pmv_x; |
1003 |
//DEBUG1("",intra_dc_threshold); |
mv_y += pmv_y; |
1004 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
|
1005 |
|
if (mv_x < low) { |
1006 |
|
mv_x += range; |
1007 |
|
} else if (mv_x > high) { |
1008 |
|
mv_x -= range; |
1009 |
|
} |
1010 |
|
|
1011 |
|
if (mv_y < low) { |
1012 |
|
mv_y += range; |
1013 |
|
} else if (mv_y > high) { |
1014 |
|
mv_y -= range; |
1015 |
|
} |
1016 |
|
|
1017 |
|
mv->x = mv_x; |
1018 |
|
mv->y = mv_y; |
1019 |
|
} |
1020 |
|
|
1021 |
|
|
1022 |
|
// add by MinChen <chenm001@163.com> |
1023 |
|
// decode an B-frame forward & backward inter macroblock |
1024 |
|
void |
1025 |
|
decoder_bf_mbinter(DECODER * dec, |
1026 |
|
const MACROBLOCK * pMB, |
1027 |
|
const uint32_t x_pos, |
1028 |
|
const uint32_t y_pos, |
1029 |
|
const uint32_t cbp, |
1030 |
|
Bitstream * bs, |
1031 |
|
const uint32_t quant, |
1032 |
|
const uint8_t ref) |
1033 |
|
{ |
1034 |
|
|
1035 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1036 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
1037 |
|
|
1038 |
|
uint32_t stride = dec->edged_width; |
1039 |
|
uint32_t stride2 = stride / 2; |
1040 |
|
uint32_t next_block = stride * 8; |
1041 |
|
uint32_t i; |
1042 |
|
uint32_t iQuant = pMB->quant; |
1043 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1044 |
|
int uv_dx, uv_dy; |
1045 |
|
|
1046 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1047 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1048 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1049 |
|
|
1050 |
|
|
1051 |
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1052 |
|
uv_dx = pMB->mvs[0].x; |
1053 |
|
uv_dy = pMB->mvs[0].y; |
1054 |
|
|
1055 |
|
if (dec->quarterpel) |
1056 |
|
{ |
1057 |
|
uv_dx /= 2; |
1058 |
|
uv_dy /= 2; |
1059 |
|
} |
1060 |
|
|
1061 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1062 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1063 |
|
} else { |
1064 |
|
int sum; |
1065 |
|
|
1066 |
|
if(dec->quarterpel) |
1067 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1068 |
|
else |
1069 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1070 |
|
|
1071 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1072 |
|
|
1073 |
|
if(dec->quarterpel) |
1074 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1075 |
|
else |
1076 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1077 |
|
|
1078 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1079 |
|
} |
1080 |
|
|
1081 |
|
start_timer(); |
1082 |
|
if(dec->quarterpel) { |
1083 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
1084 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1085 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1086 |
|
} |
1087 |
|
else { |
1088 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
1089 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1090 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1091 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1092 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1093 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1094 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1095 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1096 |
|
} |
1097 |
|
|
1098 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
1099 |
|
uv_dx, uv_dy, stride2, 0); |
1100 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
1101 |
|
uv_dx, uv_dy, stride2, 0); |
1102 |
|
stop_comp_timer(); |
1103 |
|
|
1104 |
|
for (i = 0; i < 6; i++) { |
1105 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1106 |
|
|
1107 |
|
if (cbp & (1 << (5 - i))) // coded |
1108 |
|
{ |
1109 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1110 |
|
|
1111 |
|
start_timer(); |
1112 |
|
get_inter_block(bs, &block[i * 64], direction); |
1113 |
|
stop_coding_timer(); |
1114 |
|
|
1115 |
|
start_timer(); |
1116 |
|
if (dec->quant_type == 0) { |
1117 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1118 |
|
} else { |
1119 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1120 |
|
} |
1121 |
|
stop_iquant_timer(); |
1122 |
|
|
1123 |
|
start_timer(); |
1124 |
|
idct(&data[i * 64]); |
1125 |
|
stop_idct_timer(); |
1126 |
|
} |
1127 |
|
} |
1128 |
|
|
1129 |
|
if (dec->interlacing && pMB->field_dct) { |
1130 |
|
next_block = stride; |
1131 |
|
stride *= 2; |
1132 |
|
} |
1133 |
|
|
1134 |
|
start_timer(); |
1135 |
|
if (cbp & 32) |
1136 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1137 |
|
if (cbp & 16) |
1138 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1139 |
|
if (cbp & 8) |
1140 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1141 |
|
if (cbp & 4) |
1142 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1143 |
|
if (cbp & 2) |
1144 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1145 |
|
if (cbp & 1) |
1146 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1147 |
|
stop_transfer_timer(); |
1148 |
|
} |
1149 |
|
|
1150 |
|
// add by MinChen <chenm001@163.com> |
1151 |
|
// decode an B-frame direct & inter macroblock |
1152 |
|
void |
1153 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
1154 |
|
IMAGE forward, |
1155 |
|
IMAGE backward, |
1156 |
|
const MACROBLOCK * pMB, |
1157 |
|
const uint32_t x_pos, |
1158 |
|
const uint32_t y_pos, |
1159 |
|
Bitstream * bs) |
1160 |
|
{ |
1161 |
|
|
1162 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1163 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
1164 |
|
|
1165 |
|
uint32_t stride = dec->edged_width; |
1166 |
|
uint32_t stride2 = stride / 2; |
1167 |
|
uint32_t next_block = stride * 8; |
1168 |
|
uint32_t iQuant = pMB->quant; |
1169 |
|
int uv_dx, uv_dy; |
1170 |
|
int b_uv_dx, b_uv_dy; |
1171 |
|
uint32_t i; |
1172 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1173 |
|
const uint32_t cbp = pMB->cbp; |
1174 |
|
|
1175 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1176 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1177 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1178 |
|
|
1179 |
|
|
1180 |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1181 |
|
uv_dx = pMB->mvs[0].x; |
1182 |
|
uv_dy = pMB->mvs[0].y; |
1183 |
|
|
1184 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1185 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1186 |
|
|
1187 |
|
if (dec->quarterpel) |
1188 |
|
{ |
1189 |
|
uv_dx /= 2; |
1190 |
|
uv_dy /= 2; |
1191 |
|
|
1192 |
|
b_uv_dx /= 2; |
1193 |
|
b_uv_dy /= 2; |
1194 |
|
} |
1195 |
|
|
1196 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1197 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1198 |
|
|
1199 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1200 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1201 |
|
} else { |
1202 |
|
int sum; |
1203 |
|
|
1204 |
|
if(dec->quarterpel) |
1205 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1206 |
|
else |
1207 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1208 |
|
|
1209 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1210 |
|
|
1211 |
|
if(dec->quarterpel) |
1212 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1213 |
|
else |
1214 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1215 |
|
|
1216 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1217 |
|
|
1218 |
|
|
1219 |
|
if(dec->quarterpel) |
1220 |
|
sum = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1221 |
|
else |
1222 |
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1223 |
|
|
1224 |
|
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1225 |
|
|
1226 |
|
if(dec->quarterpel) |
1227 |
|
sum = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1228 |
|
else |
1229 |
|
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1230 |
|
|
1231 |
|
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1232 |
|
} |
1233 |
|
|
1234 |
|
|
1235 |
|
start_timer(); |
1236 |
|
if(dec->quarterpel) { |
1237 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1238 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1239 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1240 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1241 |
|
else { |
1242 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1243 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1244 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1245 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1246 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1247 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1248 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1249 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1250 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1251 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1252 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1253 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1254 |
|
} |
1255 |
|
} |
1256 |
|
else { |
1257 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1258 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1259 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1260 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1261 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1262 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1263 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1264 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1265 |
|
0); |
1266 |
|
} |
1267 |
|
|
1268 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1269 |
|
uv_dy, stride2, 0); |
1270 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1271 |
|
uv_dy, stride2, 0); |
1272 |
|
|
1273 |
|
|
1274 |
|
if(dec->quarterpel) { |
1275 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1276 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1277 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1278 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1279 |
|
else { |
1280 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1281 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1282 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1283 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1284 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1285 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1286 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1287 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1288 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1289 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1290 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1291 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1292 |
|
} |
1293 |
|
} |
1294 |
|
else { |
1295 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1296 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1297 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1298 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1299 |
|
0); |
1300 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1301 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1302 |
|
stride, 0); |
1303 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1304 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1305 |
|
stride, 0); |
1306 |
|
} |
1307 |
|
|
1308 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1309 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1310 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1311 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1312 |
|
|
1313 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1314 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1315 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1316 |
|
stride, 1, 8); |
1317 |
|
|
1318 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1319 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1320 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1321 |
|
stride, 1, 8); |
1322 |
|
|
1323 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1324 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1325 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1326 |
|
stride, 1, 8); |
1327 |
|
|
1328 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1329 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1330 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1331 |
|
stride, 1, 8); |
1332 |
|
|
1333 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1334 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1335 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1336 |
|
stride2, 1, 8); |
1337 |
|
|
1338 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1339 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1340 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1341 |
|
stride2, 1, 8); |
1342 |
|
|
1343 |
|
stop_comp_timer(); |
1344 |
|
|
1345 |
|
for (i = 0; i < 6; i++) { |
1346 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1347 |
|
|
1348 |
|
if (cbp & (1 << (5 - i))) // coded |
1349 |
|
{ |
1350 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1351 |
|
|
1352 |
|
start_timer(); |
1353 |
|
get_inter_block(bs, &block[i * 64], direction); |
1354 |
|
stop_coding_timer(); |
1355 |
|
|
1356 |
|
start_timer(); |
1357 |
|
if (dec->quant_type == 0) { |
1358 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1359 |
|
} else { |
1360 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1361 |
|
} |
1362 |
|
stop_iquant_timer(); |
1363 |
|
|
1364 |
|
start_timer(); |
1365 |
|
idct(&data[i * 64]); |
1366 |
|
stop_idct_timer(); |
1367 |
|
} |
1368 |
|
} |
1369 |
|
|
1370 |
|
if (dec->interlacing && pMB->field_dct) { |
1371 |
|
next_block = stride; |
1372 |
|
stride *= 2; |
1373 |
|
} |
1374 |
|
|
1375 |
|
start_timer(); |
1376 |
|
if (cbp & 32) |
1377 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1378 |
|
if (cbp & 16) |
1379 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1380 |
|
if (cbp & 8) |
1381 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1382 |
|
if (cbp & 4) |
1383 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1384 |
|
if (cbp & 2) |
1385 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1386 |
|
if (cbp & 1) |
1387 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1388 |
|
stop_transfer_timer(); |
1389 |
|
} |
1390 |
|
|
1391 |
|
|
1392 |
|
// add by MinChen <chenm001@163.com> |
1393 |
|
// for decode B-frame dbquant |
1394 |
|
int32_t __inline |
1395 |
|
get_dbquant(Bitstream * bs) |
1396 |
|
{ |
1397 |
|
if (!BitstreamGetBit(bs)) // '0' |
1398 |
|
return (0); |
1399 |
|
else if (!BitstreamGetBit(bs)) // '10' |
1400 |
|
return (-2); |
1401 |
|
else |
1402 |
|
return (2); // '11' |
1403 |
|
} |
1404 |
|
|
1405 |
|
// add by MinChen <chenm001@163.com> |
1406 |
|
// for decode B-frame mb_type |
1407 |
|
// bit ret_value |
1408 |
|
// 1 0 |
1409 |
|
// 01 1 |
1410 |
|
// 001 2 |
1411 |
|
// 0001 3 |
1412 |
|
int32_t __inline |
1413 |
|
get_mbtype(Bitstream * bs) |
1414 |
|
{ |
1415 |
|
int32_t mb_type; |
1416 |
|
|
1417 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) { |
1418 |
|
if (BitstreamGetBit(bs)) |
1419 |
|
break; |
1420 |
|
} |
1421 |
|
|
1422 |
|
if (mb_type <= 3) |
1423 |
|
return (mb_type); |
1424 |
|
else |
1425 |
|
return (-1); |
1426 |
|
} |
1427 |
|
|
1428 |
|
void |
1429 |
|
decoder_bframe(DECODER * dec, |
1430 |
|
Bitstream * bs, |
1431 |
|
int quant, |
1432 |
|
int fcode_forward, |
1433 |
|
int fcode_backward) |
1434 |
|
{ |
1435 |
|
uint32_t x, y; |
1436 |
|
VECTOR mv; |
1437 |
|
const VECTOR zeromv = {0,0}; |
1438 |
|
#ifdef BFRAMES_DEC_DEBUG |
1439 |
|
FILE *fp; |
1440 |
|
static char first=0; |
1441 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1442 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1443 |
|
} |
1444 |
|
#endif |
1445 |
|
|
1446 |
|
start_timer(); |
1447 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1448 |
|
dec->width, dec->height); |
1449 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1450 |
|
dec->width, dec->height); |
1451 |
|
stop_edges_timer(); |
1452 |
|
|
1453 |
|
#ifdef BFRAMES_DEC_DEBUG |
1454 |
|
if (!first){ |
1455 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1456 |
|
} |
1457 |
|
#endif |
1458 |
|
|
1459 |
|
for (y = 0; y < dec->mb_height; y++) { |
1460 |
|
// Initialize Pred Motion Vector |
1461 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1462 |
|
for (x = 0; x < dec->mb_width; x++) { |
1463 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1464 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1465 |
|
|
1466 |
|
mv = |
1467 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1468 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1469 |
|
|
1470 |
|
// the last P_VOP is skip macroblock ? |
1471 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1472 |
|
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
1473 |
|
mb->cbp = 0; |
1474 |
|
#ifdef BFRAMES_DEC_DEBUG |
1475 |
|
mb->mb_type = MODE_NOT_CODED; |
1476 |
|
BFRAME_DEBUG |
1477 |
|
#endif |
1478 |
|
mb->mb_type = MODE_FORWARD; |
1479 |
|
mb->quant = last_mb->quant; |
1480 |
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1481 |
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1482 |
|
|
1483 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1484 |
|
continue; |
1485 |
|
} |
1486 |
|
|
1487 |
|
if (!BitstreamGetBit(bs)) { // modb=='0' |
1488 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1489 |
|
|
1490 |
|
mb->mb_type = get_mbtype(bs); |
1491 |
|
|
1492 |
|
if (!modb2) { // modb=='00' |
1493 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1494 |
|
} else { |
1495 |
|
mb->cbp = 0; |
1496 |
|
} |
1497 |
|
if (mb->mb_type && mb->cbp) { |
1498 |
|
quant += get_dbquant(bs); |
1499 |
|
|
1500 |
|
if (quant > 31) { |
1501 |
|
quant = 31; |
1502 |
|
} else if (quant < 1) { |
1503 |
|
quant = 1; |
1504 |
|
} |
1505 |
|
} |
1506 |
|
} else { |
1507 |
|
mb->mb_type = MODE_DIRECT_NONE_MV; |
1508 |
|
mb->cbp = 0; |
1509 |
|
} |
1510 |
|
|
1511 |
|
mb->quant = quant; |
1512 |
|
mb->mode = MODE_INTER4V; |
1513 |
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
1514 |
|
|
1515 |
|
#ifdef BFRAMES_DEC_DEBUG |
1516 |
|
BFRAME_DEBUG |
1517 |
|
#endif |
1518 |
|
|
1519 |
|
switch (mb->mb_type) { |
1520 |
|
case MODE_DIRECT: |
1521 |
|
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1522 |
|
|
1523 |
|
case MODE_DIRECT_NONE_MV: |
1524 |
|
{ |
1525 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1526 |
|
int i; |
1527 |
|
|
1528 |
|
for (i = 0; i < 4; i++) { |
1529 |
|
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1530 |
|
/ TRD + mv.x); |
1531 |
|
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1532 |
|
? ((TRB - TRD) * last_mb->mvs[i].x) |
1533 |
|
/ TRD |
1534 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1535 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1536 |
|
/ TRD + mv.y); |
1537 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1538 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1539 |
|
/ TRD |
1540 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1541 |
|
} |
1542 |
|
//DEBUG("B-frame Direct!\n"); |
1543 |
|
} |
1544 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1545 |
|
mb, x, y, bs); |
1546 |
|
break; |
1547 |
|
|
1548 |
|
case MODE_INTERPOLATE: |
1549 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1550 |
|
dec->p_fmv); |
1551 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1552 |
|
|
1553 |
|
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1554 |
|
fcode_backward, dec->p_bmv); |
1555 |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1556 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1557 |
|
|
1558 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1559 |
|
mb, x, y, bs); |
1560 |
|
//DEBUG("B-frame Bidir!\n"); |
1561 |
break; |
break; |
1562 |
|
|
1563 |
case B_VOP : // ignore |
case MODE_BACKWARD: |
1564 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1565 |
|
dec->p_bmv); |
1566 |
|
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1567 |
|
|
1568 |
|
mb->mode = MODE_INTER; |
1569 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1570 |
|
//DEBUG("B-frame Backward!\n"); |
1571 |
break; |
break; |
1572 |
|
|
1573 |
case N_VOP : // vop not coded |
case MODE_FORWARD: |
1574 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1575 |
|
dec->p_fmv); |
1576 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1577 |
|
|
1578 |
|
mb->mode = MODE_INTER; |
1579 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1580 |
|
//DEBUG("B-frame Forward!\n"); |
1581 |
break; |
break; |
1582 |
|
|
1583 |
default : |
default : |
1584 |
|
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
1585 |
|
} |
1586 |
|
|
1587 |
|
} // end of FOR |
1588 |
|
} |
1589 |
|
#ifdef BFRAMES_DEC_DEBUG |
1590 |
|
if (!first){ |
1591 |
|
first=1; |
1592 |
|
if (fp) |
1593 |
|
fclose(fp); |
1594 |
|
} |
1595 |
|
#endif |
1596 |
|
} |
1597 |
|
|
1598 |
|
// swap two MACROBLOCK array |
1599 |
|
void |
1600 |
|
mb_swap(MACROBLOCK ** mb1, |
1601 |
|
MACROBLOCK ** mb2) |
1602 |
|
{ |
1603 |
|
MACROBLOCK *temp = *mb1; |
1604 |
|
|
1605 |
|
*mb1 = *mb2; |
1606 |
|
*mb2 = temp; |
1607 |
|
} |
1608 |
|
|
1609 |
|
|
1610 |
|
/* perform post processing if necessary, and output the image */ |
1611 |
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1612 |
|
const XVID_DEC_FRAME * frame, int pp_disable) |
1613 |
|
{ |
1614 |
|
|
1615 |
|
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
1616 |
|
{ |
1617 |
|
/* note: image is stored to tmp */ |
1618 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1619 |
|
image_deblock_rrv(&dec->tmp, dec->edged_width, |
1620 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1621 |
|
8, frame->general); |
1622 |
|
img = &dec->tmp; |
1623 |
|
} |
1624 |
|
|
1625 |
|
image_output(img, dec->width, dec->height, |
1626 |
|
dec->edged_width, frame->image, frame->stride, |
1627 |
|
frame->colorspace, dec->interlacing); |
1628 |
|
} |
1629 |
|
|
1630 |
|
|
1631 |
|
int |
1632 |
|
decoder_decode(DECODER * dec, |
1633 |
|
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
1634 |
|
{ |
1635 |
|
|
1636 |
|
Bitstream bs; |
1637 |
|
uint32_t rounding; |
1638 |
|
uint32_t reduced_resolution; |
1639 |
|
uint32_t quant; |
1640 |
|
uint32_t fcode_forward; |
1641 |
|
uint32_t fcode_backward; |
1642 |
|
uint32_t intra_dc_threshold; |
1643 |
|
VECTOR gmc_mv[5]; |
1644 |
|
uint32_t vop_type; |
1645 |
|
int success = 0; |
1646 |
|
int output = 0; |
1647 |
|
int seen_something = 0; |
1648 |
|
|
1649 |
|
start_global_timer(); |
1650 |
|
|
1651 |
|
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
1652 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1653 |
|
|
1654 |
|
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
1655 |
|
dec->frames = 0; |
1656 |
|
|
1657 |
|
if (frame->length < 0) /* decoder flush */ |
1658 |
|
{ |
1659 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1660 |
|
we have a reference frame, then outout the reference frame */ |
1661 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
1662 |
|
{ |
1663 |
|
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
1664 |
|
output = 1; |
1665 |
|
} |
1666 |
|
|
1667 |
|
frame->length = 0; |
1668 |
|
if (stats) |
1669 |
|
{ |
1670 |
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
1671 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1672 |
|
stats->data.vop.time_increment = 0; //XXX: todo |
1673 |
|
} |
1674 |
|
|
1675 |
|
emms(); |
1676 |
|
|
1677 |
|
stop_global_timer(); |
1678 |
|
return XVID_ERR_OK; |
1679 |
|
} |
1680 |
|
|
1681 |
|
BitstreamInit(&bs, frame->bitstream, frame->length); |
1682 |
|
|
1683 |
|
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
1684 |
|
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1685 |
|
{ |
1686 |
|
if (stats) |
1687 |
|
stats->notify = XVID_DEC_VOP; |
1688 |
|
frame->length = 1; |
1689 |
|
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1690 |
|
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
1691 |
|
emms(); |
1692 |
|
return XVID_ERR_OK; |
1693 |
|
} |
1694 |
|
|
1695 |
|
repeat: |
1696 |
|
|
1697 |
|
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1698 |
|
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); |
1699 |
|
|
1700 |
|
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%i, time_pp=%i, time_bp=%i", |
1701 |
|
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1702 |
|
|
1703 |
|
if (vop_type == - 1) |
1704 |
|
{ |
1705 |
|
if (success) goto done; |
1706 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1707 |
} |
} |
1708 |
|
|
1709 |
|
if (vop_type == -2 || vop_type == -3) |
1710 |
|
{ |
1711 |
|
if (vop_type == -3) |
1712 |
|
decoder_resize(dec); |
1713 |
|
|
1714 |
|
if (stats) |
1715 |
|
{ |
1716 |
|
stats->notify = XVID_DEC_VOL; |
1717 |
|
stats->data.vol.general = 0; |
1718 |
|
if (dec->interlacing) |
1719 |
|
stats->data.vol.general |= XVID_INTERLACING; |
1720 |
|
stats->data.vol.width = dec->width; |
1721 |
|
stats->data.vol.height = dec->height; |
1722 |
|
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
1723 |
|
stats->data.vol.par_width = dec->par_width; |
1724 |
|
stats->data.vol.par_height = dec->par_height; |
1725 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1726 |
|
return XVID_ERR_OK; |
1727 |
|
} |
1728 |
|
goto repeat; |
1729 |
|
} |
1730 |
|
|
1731 |
start_timer(); |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
1732 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
|
1733 |
frame->image, frame->stride, frame->colorspace); |
|
1734 |
stop_conv_timer(); |
/* packed_mode: special-N_VOP treament */ |
1735 |
|
if (dec->packed_mode && vop_type == N_VOP) |
1736 |
|
{ |
1737 |
|
if (dec->low_delay_default && dec->frames > 0) |
1738 |
|
{ |
1739 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1740 |
|
output = 1; |
1741 |
|
} |
1742 |
|
/* ignore otherwise */ |
1743 |
|
} |
1744 |
|
else if (vop_type != B_VOP) |
1745 |
|
{ |
1746 |
|
switch(vop_type) |
1747 |
|
{ |
1748 |
|
case I_VOP : |
1749 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1750 |
|
break; |
1751 |
|
case P_VOP : |
1752 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1753 |
|
fcode_forward, intra_dc_threshold, NULL); |
1754 |
|
break; |
1755 |
|
case S_VOP : |
1756 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1757 |
|
fcode_forward, intra_dc_threshold, gmc_mv); |
1758 |
|
break; |
1759 |
|
case N_VOP : |
1760 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1761 |
|
break; |
1762 |
|
} |
1763 |
|
|
1764 |
|
if (reduced_resolution) |
1765 |
|
{ |
1766 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1767 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1768 |
|
16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV); |
1769 |
|
} |
1770 |
|
|
1771 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1772 |
|
if (!(dec->low_delay_default && dec->packed_mode)) |
1773 |
|
{ |
1774 |
|
if (dec->low_delay) |
1775 |
|
{ |
1776 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
1777 |
|
output = 1; |
1778 |
|
} |
1779 |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
1780 |
|
{ |
1781 |
|
/* output the reference frame */ |
1782 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1783 |
|
output = 1; |
1784 |
|
} |
1785 |
|
} |
1786 |
|
|
1787 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1788 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1789 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1790 |
|
dec->last_reduced_resolution = reduced_resolution; |
1791 |
|
|
1792 |
|
dec->frames++; |
1793 |
|
seen_something = 1; |
1794 |
|
|
1795 |
|
}else{ /* B_VOP */ |
1796 |
|
|
1797 |
|
if (dec->low_delay) |
1798 |
|
{ |
1799 |
|
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
1800 |
|
dec->low_delay = 1; |
1801 |
|
} |
1802 |
|
|
1803 |
|
if (dec->frames < 2) |
1804 |
|
{ |
1805 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1806 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1807 |
|
"broken b-frame, mising ref frames"); |
1808 |
|
}else if (dec->time_pp <= dec->time_bp) { |
1809 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1810 |
|
decoded in vfw. */ |
1811 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1812 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1813 |
|
}else{ |
1814 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1815 |
|
} |
1816 |
|
|
1817 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
1818 |
|
output = 1; |
1819 |
|
dec->frames++; |
1820 |
|
} |
1821 |
|
|
1822 |
|
BitstreamByteAlign(&bs); |
1823 |
|
|
1824 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1825 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
1826 |
|
{ |
1827 |
|
success = 1; |
1828 |
|
goto repeat; |
1829 |
|
} |
1830 |
|
|
1831 |
|
done : |
1832 |
|
|
1833 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1834 |
|
then output the recently decoded frame, or print an error message */ |
1835 |
|
if (dec->low_delay_default && output == 0) |
1836 |
|
{ |
1837 |
|
if (dec->packed_mode && seen_something) |
1838 |
|
{ |
1839 |
|
/* output the recently decoded frame */ |
1840 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1841 |
|
output = 1; |
1842 |
|
} |
1843 |
|
else |
1844 |
|
{ |
1845 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1846 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1847 |
|
"warning: nothing to output"); |
1848 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1849 |
|
"bframe decoder lag"); |
1850 |
|
|
1851 |
|
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
1852 |
|
} |
1853 |
|
} |
1854 |
|
|
1855 |
|
frame->length = BitstreamPos(&bs) / 8; |
1856 |
|
|
1857 |
|
if (stats) |
1858 |
|
{ |
1859 |
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
1860 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1861 |
|
stats->data.vop.time_increment = 0; //XXX: todo |
1862 |
|
} |
1863 |
|
|
1864 |
emms(); |
emms(); |
1865 |
|
|