1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* decoder main |
* - Decoder Module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
|
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
|
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
7 |
* |
* |
8 |
* 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 |
9 |
* 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 |
10 |
* the xvid_free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
11 |
* (at your option) any later version. |
* (at your option) any later version. |
12 |
* |
* |
13 |
* This program is distributed in the hope that it will be useful, |
* This program is distributed in the hope that it will be useful, |
16 |
* GNU General Public License for more details. |
* GNU General Public License for more details. |
17 |
* |
* |
18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
19 |
* along with this program; if not, write to the xvid_free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
*************************************************************************/ |
* $Id$ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
23 |
* |
* |
24 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
****************************************************************************/ |
|
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
|
|
* 22.12.2001 block based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
*************************************************************************/ |
|
25 |
|
|
26 |
|
#include <stdio.h> |
27 |
#include <stdlib.h> |
#include <stdlib.h> |
28 |
#include <string.h> // memset |
#include <string.h> |
29 |
|
|
30 |
|
#ifdef BFRAMES_DEC_DEBUG |
31 |
|
#define BFRAMES_DEC |
32 |
|
#endif |
33 |
|
|
34 |
#include "xvid.h" |
#include "xvid.h" |
35 |
#include "portab.h" |
#include "portab.h" |
36 |
|
#include "global.h" |
37 |
|
|
38 |
#include "decoder.h" |
#include "decoder.h" |
39 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
45 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
46 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
47 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
48 |
#include "utils/mbfunctions.h" |
#include "image/reduced.h" |
49 |
|
#include "image/font.h" |
50 |
|
|
51 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
52 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
53 |
#include "utils/timer.h" |
#include "utils/timer.h" |
54 |
#include "utils/emms.h" |
#include "utils/emms.h" |
55 |
|
#include "motion/motion.h" |
56 |
|
|
57 |
#include "image/image.h" |
#include "image/image.h" |
58 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
59 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
60 |
|
|
61 |
int decoder_create(XVID_DEC_PARAM * param) |
int |
62 |
{ |
decoder_resize(DECODER * dec) |
|
DECODER * dec; |
|
|
|
|
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
|
|
if (dec == NULL) |
|
63 |
{ |
{ |
64 |
return XVID_ERR_MEMORY; |
/* free existing */ |
65 |
} |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
66 |
param->handle = dec; |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
67 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
68 |
dec->width = param->width; |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
69 |
dec->height = param->height; |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
70 |
|
|
71 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
72 |
|
|
73 |
|
if (dec->last_mbs) |
74 |
|
xvid_free(dec->last_mbs); |
75 |
|
if (dec->mbs) |
76 |
|
xvid_free(dec->mbs); |
77 |
|
|
78 |
|
/* realloc */ |
79 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
80 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
81 |
|
|
82 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
83 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
84 |
|
|
85 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
86 |
xvid_free(dec); |
xvid_free(dec); |
87 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
88 |
} |
} |
89 |
|
|
90 |
if (image_create(&dec->refn, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
91 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
92 |
xvid_free(dec); |
xvid_free(dec); |
93 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
94 |
} |
} |
95 |
|
|
96 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
/* Support B-frame to reference last 2 frame */ |
97 |
if (dec->mbs == NULL) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
98 |
{ |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
99 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
100 |
|
xvid_free(dec); |
101 |
|
return XVID_ERR_MEMORY; |
102 |
|
} |
103 |
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
104 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
105 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
106 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
107 |
|
xvid_free(dec); |
108 |
|
return XVID_ERR_MEMORY; |
109 |
|
} |
110 |
|
|
111 |
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
112 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
113 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
114 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
115 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
116 |
|
xvid_free(dec); |
117 |
|
return XVID_ERR_MEMORY; |
118 |
|
} |
119 |
|
|
120 |
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
121 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
122 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
123 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
124 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
125 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
126 |
|
xvid_free(dec); |
127 |
|
return XVID_ERR_MEMORY; |
128 |
|
} |
129 |
|
|
130 |
|
dec->mbs = |
131 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
132 |
|
CACHE_LINE); |
133 |
|
if (dec->mbs == NULL) { |
134 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
135 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
136 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
137 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
138 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
139 |
|
xvid_free(dec); |
140 |
|
return XVID_ERR_MEMORY; |
141 |
|
} |
142 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
143 |
|
|
144 |
|
/* For skip MB flag */ |
145 |
|
dec->last_mbs = |
146 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
147 |
|
CACHE_LINE); |
148 |
|
if (dec->last_mbs == NULL) { |
149 |
|
xvid_free(dec->mbs); |
150 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
151 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
152 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
153 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
154 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
155 |
xvid_free(dec); |
xvid_free(dec); |
156 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
157 |
} |
} |
158 |
|
|
159 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
160 |
|
|
161 |
|
return XVID_ERR_OK; |
162 |
|
} |
163 |
|
|
164 |
|
|
165 |
|
int |
166 |
|
decoder_create(XVID_DEC_PARAM * param) |
167 |
|
{ |
168 |
|
DECODER *dec; |
169 |
|
|
170 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
171 |
|
if (dec == NULL) { |
172 |
|
return XVID_ERR_MEMORY; |
173 |
|
} |
174 |
|
memset(dec, 0, sizeof(DECODER)); |
175 |
|
|
176 |
|
param->handle = dec; |
177 |
|
|
178 |
|
dec->width = param->width; |
179 |
|
dec->height = param->height; |
180 |
|
|
181 |
|
image_null(&dec->cur); |
182 |
|
image_null(&dec->refn[0]); |
183 |
|
image_null(&dec->refn[1]); |
184 |
|
image_null(&dec->tmp); |
185 |
|
image_null(&dec->qtmp); |
186 |
|
|
187 |
|
/* image based GMC */ |
188 |
|
image_null(&dec->gmc); |
189 |
|
|
190 |
|
|
191 |
|
dec->mbs = NULL; |
192 |
|
dec->last_mbs = NULL; |
193 |
|
|
194 |
init_timer(); |
init_timer(); |
|
create_vlc_tables(); |
|
195 |
|
|
196 |
|
/* For B-frame support (used to save reference frame's time */ |
197 |
|
dec->frames = 0; |
198 |
|
dec->time = dec->time_base = dec->last_time_base = 0; |
199 |
|
dec->low_delay = 0; |
200 |
|
dec->packed_mode = 0; |
201 |
|
|
202 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
203 |
|
|
204 |
|
if (dec->fixed_dimensions) |
205 |
|
return decoder_resize(dec); |
206 |
|
else |
207 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
208 |
} |
} |
209 |
|
|
210 |
|
|
211 |
int decoder_destroy(DECODER * dec) |
int |
212 |
|
decoder_destroy(DECODER * dec) |
213 |
{ |
{ |
214 |
|
xvid_free(dec->last_mbs); |
215 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
216 |
image_destroy(&dec->refn, dec->edged_width, dec->edged_height); |
|
217 |
|
/* image based GMC */ |
218 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
219 |
|
|
220 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
221 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
222 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
223 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
224 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
225 |
xvid_free(dec); |
xvid_free(dec); |
226 |
|
|
|
destroy_vlc_tables(); |
|
|
|
|
227 |
write_timer(); |
write_timer(); |
228 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
229 |
} |
} |
230 |
|
|
231 |
|
|
232 |
|
|
233 |
static const int32_t dquant_table[4] = |
static const int32_t dquant_table[4] = { |
|
{ |
|
234 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
235 |
}; |
}; |
236 |
|
|
237 |
|
|
|
// decode an intra macroblock |
|
238 |
|
|
239 |
void decoder_mbintra(DECODER * dec, |
|
240 |
|
/* decode an intra macroblock */ |
241 |
|
void |
242 |
|
decoder_mbintra(DECODER * dec, |
243 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
244 |
const uint32_t x_pos, |
const uint32_t x_pos, |
245 |
const uint32_t y_pos, |
const uint32_t y_pos, |
247 |
const uint32_t cbp, |
const uint32_t cbp, |
248 |
Bitstream * bs, |
Bitstream * bs, |
249 |
const uint32_t quant, |
const uint32_t quant, |
250 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
251 |
|
const unsigned int bound, |
252 |
|
const int reduced_resolution) |
253 |
{ |
{ |
254 |
|
|
255 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
256 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
257 |
|
|
258 |
const uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
259 |
|
uint32_t stride2 = stride / 2; |
260 |
|
uint32_t next_block = stride * 8; |
261 |
uint32_t i; |
uint32_t i; |
262 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
263 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
264 |
|
|
265 |
|
if (reduced_resolution) { |
266 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
267 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
268 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
269 |
|
}else{ |
270 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
271 |
pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
272 |
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); |
273 |
|
} |
274 |
|
|
275 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
276 |
|
|
277 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
278 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
279 |
int16_t predictors[8]; |
int16_t predictors[8]; |
280 |
int start_coeff; |
int start_coeff; |
281 |
|
|
282 |
start_timer(); |
start_timer(); |
283 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i * 64], |
284 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound); |
285 |
{ |
if (!acpred_flag) { |
286 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
287 |
} |
} |
288 |
stop_prediction_timer(); |
stop_prediction_timer(); |
289 |
|
|
290 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
291 |
int dc_size; |
int dc_size; |
292 |
int dc_dif; |
int dc_dif; |
293 |
|
|
294 |
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); |
295 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
296 |
|
|
297 |
if (dc_size > 8) |
if (dc_size > 8) { |
298 |
{ |
BitstreamSkip(bs, 1); /* marker */ |
|
BitstreamSkip(bs, 1); // marker |
|
299 |
} |
} |
300 |
|
|
301 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
302 |
start_coeff = 1; |
start_coeff = 1; |
303 |
} |
|
304 |
else |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
305 |
{ |
} else { |
306 |
start_coeff = 0; |
start_coeff = 0; |
307 |
} |
} |
308 |
|
|
309 |
start_timer(); |
start_timer(); |
310 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
311 |
{ |
{ |
312 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
313 |
|
2 : pMB->acpred_directions[i]; |
314 |
|
|
315 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
316 |
} |
} |
317 |
stop_coding_timer(); |
stop_coding_timer(); |
318 |
|
|
321 |
stop_prediction_timer(); |
stop_prediction_timer(); |
322 |
|
|
323 |
start_timer(); |
start_timer(); |
324 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
325 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
326 |
} |
} else { |
|
else |
|
|
{ |
|
327 |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
328 |
} |
} |
329 |
stop_iquant_timer(); |
stop_iquant_timer(); |
331 |
start_timer(); |
start_timer(); |
332 |
idct(&data[i*64]); |
idct(&data[i*64]); |
333 |
stop_idct_timer(); |
stop_idct_timer(); |
334 |
|
|
335 |
} |
} |
336 |
|
|
337 |
start_timer(); |
if (dec->interlacing && pMB->field_dct) { |
338 |
if (dec->interlacing && pMB->field_dct) |
next_block = stride; |
339 |
{ |
stride *= 2; |
|
MBFieldToFrame(data); |
|
340 |
} |
} |
|
stop_interlacing_timer(); |
|
341 |
|
|
342 |
start_timer(); |
start_timer(); |
343 |
|
|
344 |
|
if (reduced_resolution) |
345 |
|
{ |
346 |
|
next_block*=2; |
347 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
348 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
349 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
350 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
351 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
352 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
353 |
|
}else{ |
354 |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
355 |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
356 |
transfer_16to8copy(pY_Cur + 8 * stride, &data[2*64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
357 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, &data[3*64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
358 |
transfer_16to8copy(pU_Cur, &data[4*64], stride / 2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
359 |
transfer_16to8copy(pV_Cur, &data[5*64], stride / 2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
360 |
|
} |
361 |
stop_transfer_timer(); |
stop_transfer_timer(); |
362 |
} |
} |
363 |
|
|
364 |
|
|
365 |
|
|
366 |
|
|
367 |
|
/* decode an inter macroblock */ |
368 |
#define SIGN(X) (((X)>0)?1:-1) |
void |
369 |
#define ABS(X) (((X)>0)?(X):-(X)) |
decoder_mbinter(DECODER * dec, |
|
static const uint32_t roundtab[16] = |
|
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
|
|
|
|
// decode an inter macroblock |
|
|
|
|
|
void decoder_mbinter(DECODER * dec, |
|
370 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
371 |
const uint32_t x_pos, |
const uint32_t x_pos, |
372 |
const uint32_t y_pos, |
const uint32_t y_pos, |
373 |
const uint32_t acpred_flag, |
const uint32_t fcode, |
374 |
const uint32_t cbp, |
const uint32_t cbp, |
375 |
Bitstream * bs, |
Bitstream * bs, |
376 |
const uint32_t quant, |
const uint32_t quant, |
377 |
const uint32_t rounding) |
const uint32_t rounding, |
378 |
|
const int reduced_resolution) |
379 |
{ |
{ |
380 |
|
|
381 |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
382 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
383 |
|
|
384 |
const uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
385 |
const uint32_t stride2 = dec->edged_width / 2; |
uint32_t stride2 = stride / 2; |
386 |
|
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
387 |
uint32_t i; |
uint32_t i; |
388 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
389 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
390 |
|
|
391 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
392 |
|
VECTOR mv[4]; /* local copy of mvs */ |
393 |
|
|
394 |
|
if (reduced_resolution) { |
395 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
396 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
397 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
398 |
|
for (i = 0; i < 4; i++) { |
399 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
400 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
401 |
|
} |
402 |
|
} else { |
403 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
404 |
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); |
405 |
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); |
406 |
|
for (i = 0; i < 4; i++) |
407 |
|
mv[i] = pMB->mvs[i]; |
408 |
|
} |
409 |
|
|
410 |
|
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
411 |
|
|
412 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
413 |
|
uv_dy = mv[0].y / (1 + dec->quarterpel); |
414 |
|
|
415 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
416 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
417 |
|
|
418 |
|
start_timer(); |
419 |
|
if (reduced_resolution) |
420 |
{ |
{ |
421 |
uv_dx = pMB->mvs[0].x; |
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
422 |
uv_dy = pMB->mvs[0].y; |
mv[0].x, mv[0].y, stride, rounding); |
423 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
424 |
|
uv_dx, uv_dy, stride2, rounding); |
425 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
426 |
|
uv_dx, uv_dy, stride2, rounding); |
427 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
428 |
} |
} |
429 |
else |
else |
430 |
{ |
{ |
431 |
int sum; |
if(dec->quarterpel) { |
432 |
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, |
433 |
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, |
434 |
|
mv[0].x, mv[0].y, stride, rounding); |
435 |
|
} |
436 |
|
else { |
437 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
438 |
|
mv[0].x, mv[0].y, stride, rounding); |
439 |
|
} |
440 |
|
|
441 |
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, |
442 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dx, uv_dy, stride2, rounding); |
443 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
444 |
|
uv_dx, uv_dy, stride2, rounding); |
445 |
} |
} |
446 |
|
stop_comp_timer(); |
447 |
|
|
448 |
|
} else { /* MODE_INTER4V */ |
449 |
|
int sum; |
450 |
|
|
451 |
|
if(dec->quarterpel) |
452 |
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
453 |
|
else |
454 |
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
455 |
|
|
456 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
457 |
|
|
458 |
|
if(dec->quarterpel) |
459 |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
460 |
|
else |
461 |
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
462 |
|
|
463 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
464 |
|
|
465 |
start_timer(); |
start_timer(); |
466 |
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) |
467 |
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); |
{ |
468 |
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, |
469 |
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); |
470 |
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, |
471 |
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); |
472 |
stop_comp_timer(); |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
473 |
|
mv[2].x, mv[2].y, stride, rounding); |
474 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
475 |
|
mv[3].x, mv[3].y, stride, rounding); |
476 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
477 |
|
uv_dx, uv_dy, stride2, rounding); |
478 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
479 |
|
uv_dx, uv_dy, stride2, rounding); |
480 |
|
|
481 |
for (i = 0; i < 6; i++) |
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
482 |
|
} |
483 |
|
else |
484 |
{ |
{ |
485 |
if (cbp & (1 << (5-i))) // coded |
if(dec->quarterpel) { |
486 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
487 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
488 |
|
mv[0].x, mv[0].y, stride, rounding); |
489 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
490 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
491 |
|
mv[1].x, mv[1].y, stride, rounding); |
492 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
493 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
494 |
|
mv[2].x, mv[2].y, stride, rounding); |
495 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
496 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
497 |
|
mv[3].x, mv[3].y, stride, rounding); |
498 |
|
} |
499 |
|
else { |
500 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, |
501 |
|
mv[0].x, mv[0].y, stride, rounding); |
502 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
503 |
|
mv[1].x, mv[1].y, stride, rounding); |
504 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
505 |
|
mv[2].x, mv[2].y, stride, rounding); |
506 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
507 |
|
mv[3].x, mv[3].y, stride, rounding); |
508 |
|
} |
509 |
|
|
510 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
511 |
|
uv_dx, uv_dy, stride2, rounding); |
512 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v , 8 * x_pos, 8 * y_pos, |
513 |
|
uv_dx, uv_dy, stride2, rounding); |
514 |
|
} |
515 |
|
stop_comp_timer(); |
516 |
|
} |
517 |
|
|
518 |
|
for (i = 0; i < 6; i++) { |
519 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
520 |
|
|
521 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
522 |
{ |
{ |
523 |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
524 |
|
|
525 |
start_timer(); |
start_timer(); |
526 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, &block[i * 64], direction); |
527 |
stop_coding_timer(); |
stop_coding_timer(); |
528 |
|
|
529 |
start_timer(); |
start_timer(); |
530 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
|
{ |
|
531 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
532 |
} |
} else { |
|
else |
|
|
{ |
|
533 |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
534 |
} |
} |
535 |
stop_iquant_timer(); |
stop_iquant_timer(); |
540 |
} |
} |
541 |
} |
} |
542 |
|
|
543 |
|
if (dec->interlacing && pMB->field_dct) { |
544 |
|
next_block = stride; |
545 |
|
stride *= 2; |
546 |
|
} |
547 |
|
|
548 |
|
start_timer(); |
549 |
|
if (reduced_resolution) |
550 |
|
{ |
551 |
|
if (cbp & 32) |
552 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
553 |
|
if (cbp & 16) |
554 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
555 |
|
if (cbp & 8) |
556 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
557 |
|
if (cbp & 4) |
558 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
559 |
|
if (cbp & 2) |
560 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
561 |
|
if (cbp & 1) |
562 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
563 |
|
} |
564 |
|
else |
565 |
|
{ |
566 |
|
if (cbp & 32) |
567 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
568 |
|
if (cbp & 16) |
569 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
570 |
|
if (cbp & 8) |
571 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
572 |
|
if (cbp & 4) |
573 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
574 |
|
if (cbp & 2) |
575 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
576 |
|
if (cbp & 1) |
577 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
578 |
|
} |
579 |
|
stop_transfer_timer(); |
580 |
|
} |
581 |
|
|
582 |
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
583 |
|
{ |
584 |
|
int length = 1 << (fcode+4); |
585 |
|
|
586 |
|
/* if (quarterpel) value *= 2; */ |
587 |
|
|
588 |
|
if (value < -length) |
589 |
|
return -length; |
590 |
|
else if (value >= length) |
591 |
|
return length-1; |
592 |
|
else return value; |
593 |
|
} |
594 |
|
|
595 |
|
|
596 |
|
static void |
597 |
|
decoder_mbgmc(DECODER * dec, |
598 |
|
MACROBLOCK * const pMB, |
599 |
|
const uint32_t x_pos, |
600 |
|
const uint32_t y_pos, |
601 |
|
const uint32_t fcode, |
602 |
|
const uint32_t cbp, |
603 |
|
Bitstream * bs, |
604 |
|
const uint32_t quant, |
605 |
|
const uint32_t rounding, |
606 |
|
const int reduced_resolution) /* no reduced res support */ |
607 |
|
{ |
608 |
|
|
609 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
610 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
611 |
|
|
612 |
|
const uint32_t stride = dec->edged_width; |
613 |
|
const uint32_t stride2 = stride / 2; |
614 |
|
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
615 |
|
uint32_t i; |
616 |
|
const uint32_t iQuant = pMB->quant; |
617 |
|
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
618 |
|
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
619 |
|
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
620 |
|
|
621 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
622 |
|
|
623 |
start_timer(); |
start_timer(); |
624 |
if (pMB->field_dct) |
|
625 |
|
/* this is where the calculations are done */ |
626 |
|
|
627 |
|
{ |
628 |
|
pMB->amv = generate_GMCimageMB(&dec->gmc_data, &dec->refn[0], x_pos, y_pos, |
629 |
|
stride, stride2, dec->quarterpel, rounding, &dec->cur); |
630 |
|
|
631 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
632 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
633 |
|
} |
634 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
635 |
|
|
636 |
|
/* |
637 |
|
transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
638 |
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
639 |
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
640 |
|
*/ |
641 |
|
|
642 |
|
|
643 |
|
stop_transfer_timer(); |
644 |
|
|
645 |
|
if (!cbp) return; |
646 |
|
|
647 |
|
for (i = 0; i < 6; i++) { |
648 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
649 |
|
|
650 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
651 |
{ |
{ |
652 |
MBFieldToFrame(data); |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
653 |
|
|
654 |
|
start_timer(); |
655 |
|
get_inter_block(bs, &block[i * 64], direction); |
656 |
|
stop_coding_timer(); |
657 |
|
|
658 |
|
start_timer(); |
659 |
|
if (dec->quant_type == 0) { |
660 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
661 |
|
} else { |
662 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
663 |
|
} |
664 |
|
stop_iquant_timer(); |
665 |
|
|
666 |
|
start_timer(); |
667 |
|
idct(&data[i * 64]); |
668 |
|
stop_idct_timer(); |
669 |
|
} |
670 |
} |
} |
|
stop_interlacing_timer(); |
|
671 |
|
|
672 |
|
/* interlace + GMC is this possible ??? */ |
673 |
|
/* |
674 |
|
if (dec->interlacing && pMB->field_dct) { |
675 |
|
next_block = stride; |
676 |
|
stride *= 2; |
677 |
|
} |
678 |
|
*/ |
679 |
start_timer(); |
start_timer(); |
680 |
if (cbp & 32) |
if (cbp & 32) |
681 |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
682 |
if (cbp & 16) |
if (cbp & 16) |
683 |
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
684 |
if (cbp & 8) |
if (cbp & 8) |
685 |
transfer_16to8add(pY_Cur + 8 * stride, &data[2*64], stride); |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
686 |
if (cbp & 4) |
if (cbp & 4) |
687 |
transfer_16to8add(pY_Cur + 8 + 8 * stride, &data[3*64], stride); |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
688 |
if (cbp & 2) |
if (cbp & 2) |
689 |
transfer_16to8add(pU_Cur, &data[4*64], stride / 2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
690 |
if (cbp & 1) |
if (cbp & 1) |
691 |
transfer_16to8add(pV_Cur, &data[5*64], stride / 2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
692 |
stop_transfer_timer(); |
stop_transfer_timer(); |
693 |
} |
} |
694 |
|
|
695 |
|
|
696 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void |
697 |
|
decoder_iframe(DECODER * dec, |
698 |
|
Bitstream * bs, |
699 |
|
int reduced_resolution, |
700 |
|
int quant, |
701 |
|
int intra_dc_threshold) |
702 |
{ |
{ |
703 |
|
uint32_t bound; |
704 |
uint32_t x, y; |
uint32_t x, y; |
705 |
|
uint32_t mb_width = dec->mb_width; |
706 |
|
uint32_t mb_height = dec->mb_height; |
707 |
|
|
708 |
for (y = 0; y < dec->mb_height; y++) |
if (reduced_resolution) |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
709 |
{ |
{ |
710 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
mb_width = (dec->width + 31) / 32; |
711 |
|
mb_height = (dec->height + 31) / 32; |
712 |
|
} |
713 |
|
|
714 |
|
bound = 0; |
715 |
|
|
716 |
|
for (y = 0; y < mb_height; y++) { |
717 |
|
for (x = 0; x < mb_width; x++) { |
718 |
|
MACROBLOCK *mb; |
719 |
uint32_t mcbpc; |
uint32_t mcbpc; |
720 |
uint32_t cbpc; |
uint32_t cbpc; |
721 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
722 |
uint32_t cbpy; |
uint32_t cbpy; |
723 |
uint32_t cbp; |
uint32_t cbp; |
724 |
|
|
725 |
|
while (BitstreamShowBits(bs, 9) == 1) |
726 |
|
BitstreamSkip(bs, 9); |
727 |
|
|
728 |
|
if (check_resync_marker(bs, 0)) |
729 |
|
{ |
730 |
|
bound = read_video_packet_header(bs, dec, 0, |
731 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
732 |
|
x = bound % mb_width; |
733 |
|
y = bound / mb_width; |
734 |
|
} |
735 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
736 |
|
|
737 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
738 |
|
|
739 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
740 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
741 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
742 |
|
|
743 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
744 |
|
|
|
if (mb->mode == MODE_STUFFING) |
|
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
|
|
|
745 |
cbpy = get_cbpy(bs, 1); |
cbpy = get_cbpy(bs, 1); |
746 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
747 |
|
|
748 |
if (mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
749 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
750 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
751 |
quant = 31; |
quant = 31; |
752 |
} |
} else if (quant < 1) { |
|
else if (quant < 1) |
|
|
{ |
|
753 |
quant = 1; |
quant = 1; |
754 |
} |
} |
755 |
} |
} |
756 |
mb->quant = quant; |
mb->quant = quant; |
757 |
|
mb->mvs[0].x = mb->mvs[0].y = |
758 |
|
mb->mvs[1].x = mb->mvs[1].y = |
759 |
|
mb->mvs[2].x = mb->mvs[2].y = |
760 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
761 |
|
|
762 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
763 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
764 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
765 |
} |
} |
766 |
|
|
767 |
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, |
768 |
|
intra_dc_threshold, bound, reduced_resolution); |
769 |
|
|
770 |
} |
} |
771 |
|
if(dec->out_frm) |
772 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
773 |
} |
} |
774 |
|
|
775 |
} |
} |
776 |
|
|
777 |
|
|
778 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void |
779 |
|
get_motion_vector(DECODER * dec, |
780 |
|
Bitstream * bs, |
781 |
|
int x, |
782 |
|
int y, |
783 |
|
int k, |
784 |
|
VECTOR * ret_mv, |
785 |
|
int fcode, |
786 |
|
const int bound) |
787 |
{ |
{ |
788 |
|
|
789 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
791 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
792 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
793 |
|
|
794 |
VECTOR pmv[4]; |
VECTOR pmv; |
795 |
uint32_t psad[4]; |
VECTOR mv; |
|
|
|
|
int mv_x, mv_y; |
|
|
int pmv_x, pmv_y; |
|
|
|
|
796 |
|
|
797 |
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
798 |
|
|
799 |
pmv_x = pmv[0].x; |
mv.x = get_mv(bs, fcode); |
800 |
pmv_y = pmv[0].y; |
mv.y = get_mv(bs, fcode); |
801 |
|
|
802 |
mv_x = get_mv(bs, fcode); |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
|
mv_y = get_mv(bs, fcode); |
|
803 |
|
|
804 |
mv_x += pmv_x; |
mv.x += pmv.x; |
805 |
mv_y += pmv_y; |
mv.y += pmv.y; |
806 |
|
|
807 |
if (mv_x < low) |
if (mv.x < low) { |
808 |
{ |
mv.x += range; |
809 |
mv_x += range; |
} else if (mv.x > high) { |
810 |
} |
mv.x -= range; |
|
else if (mv_x > high) |
|
|
{ |
|
|
mv_x -= range; |
|
811 |
} |
} |
812 |
|
|
813 |
if (mv_y < low) |
if (mv.y < low) { |
814 |
{ |
mv.y += range; |
815 |
mv_y += range; |
} else if (mv.y > high) { |
816 |
|
mv.y -= range; |
817 |
} |
} |
818 |
else if (mv_y > high) |
|
819 |
{ |
ret_mv->x = mv.x; |
820 |
mv_y -= range; |
ret_mv->y = mv.y; |
821 |
} |
} |
822 |
|
|
|
mv->x = mv_x; |
|
|
mv->y = mv_y; |
|
823 |
|
|
|
} |
|
824 |
|
|
825 |
|
|
826 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
|
827 |
|
/* for P_VOP set gmc_warp to NULL */ |
828 |
|
void |
829 |
|
decoder_pframe(DECODER * dec, |
830 |
|
Bitstream * bs, |
831 |
|
int rounding, |
832 |
|
int reduced_resolution, |
833 |
|
int quant, |
834 |
|
int fcode, |
835 |
|
int intra_dc_threshold, |
836 |
|
const WARPPOINTS *const gmc_warp) |
837 |
{ |
{ |
838 |
|
|
839 |
uint32_t x, y; |
uint32_t x, y; |
840 |
|
uint32_t bound; |
841 |
|
int cp_mb, st_mb; |
842 |
|
uint32_t mb_width = dec->mb_width; |
843 |
|
uint32_t mb_height = dec->mb_height; |
844 |
|
|
845 |
image_swap(&dec->cur, &dec->refn); |
if (reduced_resolution) |
846 |
|
{ |
847 |
|
mb_width = (dec->width + 31) / 32; |
848 |
|
mb_height = (dec->height + 31) / 32; |
849 |
|
} |
850 |
|
|
851 |
start_timer(); |
start_timer(); |
852 |
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, |
853 |
|
dec->width, dec->height); |
854 |
stop_edges_timer(); |
stop_edges_timer(); |
855 |
|
|
856 |
for (y = 0; y < dec->mb_height; y++) |
if (gmc_warp) |
857 |
{ |
{ |
858 |
for (x = 0; x < dec->mb_width; x++) |
|
859 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
860 |
|
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
861 |
{ |
{ |
862 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
863 |
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
864 |
|
dec->sprite_warping_points); |
865 |
|
} |
866 |
|
|
867 |
|
generate_GMCparameters( dec->sprite_warping_points, |
868 |
|
(2 << dec->sprite_warping_accuracy), gmc_warp, |
869 |
|
dec->width, dec->height, &dec->gmc_data); |
870 |
|
|
871 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
872 |
|
/* |
873 |
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
874 |
|
mb_width, mb_height, |
875 |
|
dec->edged_width, dec->edged_width/2, |
876 |
|
fcode, dec->quarterpel, 0, |
877 |
|
rounding, dec->mbs, &dec->gmc); |
878 |
|
*/ |
879 |
|
} |
880 |
|
|
881 |
|
bound = 0; |
882 |
|
|
883 |
|
for (y = 0; y < mb_height; y++) { |
884 |
|
cp_mb = st_mb = 0; |
885 |
|
for (x = 0; x < mb_width; x++) { |
886 |
|
MACROBLOCK *mb; |
887 |
|
|
888 |
if (!BitstreamGetBit(bs)) // not_coded |
/* skip stuffing */ |
889 |
|
while (BitstreamShowBits(bs, 10) == 1) |
890 |
|
BitstreamSkip(bs, 10); |
891 |
|
|
892 |
|
if (check_resync_marker(bs, fcode - 1)) |
893 |
|
{ |
894 |
|
bound = read_video_packet_header(bs, dec, fcode - 1, |
895 |
|
&quant, &fcode, NULL, &intra_dc_threshold); |
896 |
|
x = bound % mb_width; |
897 |
|
y = bound / mb_width; |
898 |
|
} |
899 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
900 |
|
|
901 |
|
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
902 |
|
|
903 |
|
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
904 |
|
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
905 |
{ |
{ |
906 |
uint32_t mcbpc; |
uint32_t mcbpc; |
907 |
uint32_t cbpc; |
uint32_t cbpc; |
909 |
uint32_t cbpy; |
uint32_t cbpy; |
910 |
uint32_t cbp; |
uint32_t cbp; |
911 |
uint32_t intra; |
uint32_t intra; |
912 |
|
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
913 |
|
|
914 |
|
cp_mb++; |
915 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
916 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
917 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
918 |
|
|
919 |
|
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
920 |
|
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
921 |
acpred_flag = 0; |
acpred_flag = 0; |
922 |
|
|
923 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
924 |
|
|
925 |
if (intra) |
if (intra) { |
|
{ |
|
926 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
927 |
} |
} |
928 |
|
|
929 |
if (mb->mode == MODE_STUFFING) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
930 |
{ |
{ |
931 |
DEBUG("-- STUFFING ?"); |
mcsel = BitstreamGetBit(bs); |
|
continue; |
|
932 |
} |
} |
933 |
|
|
934 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
935 |
|
DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
936 |
|
|
937 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
938 |
|
|
939 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
940 |
{ |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
941 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
942 |
if (quant > 31) |
quant += dquant; |
943 |
{ |
if (quant > 31) { |
944 |
quant = 31; |
quant = 31; |
945 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
946 |
quant = 1; |
quant = 1; |
947 |
} |
} |
948 |
|
DPRINTF(DPRINTF_MB, "quant %i", quant); |
949 |
} |
} |
950 |
mb->quant = quant; |
mb->quant = quant; |
951 |
|
|
952 |
if (dec->interlacing) |
if (dec->interlacing) { |
953 |
{ |
if (cbp || intra) { |
954 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
955 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
956 |
|
} |
957 |
|
|
958 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
{ |
|
959 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
960 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
961 |
|
|
962 |
if (mb->field_pred) |
if (mb->field_pred) { |
|
{ |
|
963 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
964 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
965 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
966 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
967 |
} |
} |
968 |
} |
} |
969 |
} |
} |
970 |
|
|
971 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
if (mcsel) { |
972 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
973 |
|
rounding, reduced_resolution); |
974 |
|
continue; |
975 |
|
|
976 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
977 |
|
|
978 |
|
if (dec->interlacing && mb->field_pred) { |
979 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
980 |
|
fcode, bound); |
981 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
982 |
|
fcode, bound); |
983 |
|
} else { |
984 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
985 |
|
fcode, bound); |
986 |
|
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
987 |
|
} |
988 |
|
} else if (mb->mode == MODE_INTER4V ) { |
989 |
|
|
990 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
991 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
992 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
993 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
994 |
|
} else /* MODE_INTRA, MODE_INTRA_Q */ |
995 |
|
{ |
996 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
997 |
|
0; |
998 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
999 |
|
0; |
1000 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
1001 |
|
intra_dc_threshold, bound, reduced_resolution); |
1002 |
|
continue; |
1003 |
|
} |
1004 |
|
|
1005 |
|
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
1006 |
|
rounding, reduced_resolution); |
1007 |
|
|
1008 |
|
} |
1009 |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
1010 |
{ |
{ |
1011 |
if (dec->interlacing && mb->field_pred) |
mb->mode = MODE_NOT_CODED_GMC; |
1012 |
|
|
1013 |
|
start_timer(); |
1014 |
|
|
1015 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
1016 |
|
rounding, reduced_resolution); |
1017 |
|
|
1018 |
|
stop_transfer_timer(); |
1019 |
|
|
1020 |
|
if(dec->out_frm && cp_mb > 0) { |
1021 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1022 |
|
cp_mb = 0; |
1023 |
|
} |
1024 |
|
st_mb = x+1; |
1025 |
|
} |
1026 |
|
else /* not coded P_VOP macroblock */ |
1027 |
|
{ |
1028 |
|
mb->mode = MODE_NOT_CODED; |
1029 |
|
|
1030 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
1031 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
1032 |
|
/* copy macroblock directly from ref to cur */ |
1033 |
|
|
1034 |
|
start_timer(); |
1035 |
|
|
1036 |
|
if (reduced_resolution) |
1037 |
{ |
{ |
1038 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
1039 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
1040 |
|
dec->edged_width); |
1041 |
|
|
1042 |
|
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
1043 |
|
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
1044 |
|
dec->edged_width/2); |
1045 |
|
|
1046 |
|
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
1047 |
|
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
1048 |
|
dec->edged_width/2); |
1049 |
} |
} |
1050 |
else |
else |
1051 |
{ |
{ |
1052 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
1053 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
1054 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
dec->edged_width); |
1055 |
|
|
1056 |
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
1057 |
|
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
1058 |
|
dec->edged_width/2); |
1059 |
|
|
1060 |
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
1061 |
|
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
1062 |
|
dec->edged_width/2); |
1063 |
} |
} |
1064 |
|
|
1065 |
|
stop_transfer_timer(); |
1066 |
|
|
1067 |
|
if(dec->out_frm && cp_mb > 0) { |
1068 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1069 |
|
cp_mb = 0; |
1070 |
|
} |
1071 |
|
st_mb = x+1; |
1072 |
|
} |
1073 |
|
} |
1074 |
|
if(dec->out_frm && cp_mb > 0) |
1075 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1076 |
} |
} |
1077 |
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
} |
1078 |
|
|
1079 |
|
|
1080 |
|
/* decode B-frame motion vector */ |
1081 |
|
void |
1082 |
|
get_b_motion_vector(DECODER * dec, |
1083 |
|
Bitstream * bs, |
1084 |
|
int x, |
1085 |
|
int y, |
1086 |
|
VECTOR * mv, |
1087 |
|
int fcode, |
1088 |
|
const VECTOR pmv) |
1089 |
{ |
{ |
1090 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
int scale_fac = 1 << (fcode - 1); |
1091 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
int high = (32 * scale_fac) - 1; |
1092 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
int low = ((-32) * scale_fac); |
1093 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
int range = (64 * scale_fac); |
1094 |
|
|
1095 |
|
int mv_x, mv_y; |
1096 |
|
int pmv_x, pmv_y; |
1097 |
|
|
1098 |
|
pmv_x = pmv.x; |
1099 |
|
pmv_y = pmv.y; |
1100 |
|
|
1101 |
|
mv_x = get_mv(bs, fcode); |
1102 |
|
mv_y = get_mv(bs, fcode); |
1103 |
|
|
1104 |
|
mv_x += pmv_x; |
1105 |
|
mv_y += pmv_y; |
1106 |
|
|
1107 |
|
if (mv_x < low) { |
1108 |
|
mv_x += range; |
1109 |
|
} else if (mv_x > high) { |
1110 |
|
mv_x -= range; |
1111 |
|
} |
1112 |
|
|
1113 |
|
if (mv_y < low) { |
1114 |
|
mv_y += range; |
1115 |
|
} else if (mv_y > high) { |
1116 |
|
mv_y -= range; |
1117 |
} |
} |
1118 |
else // MODE_INTRA, MODE_INTRA_Q |
|
1119 |
|
mv->x = mv_x; |
1120 |
|
mv->y = mv_y; |
1121 |
|
} |
1122 |
|
|
1123 |
|
|
1124 |
|
/* decode an B-frame forward & backward inter macroblock */ |
1125 |
|
void |
1126 |
|
decoder_bf_mbinter(DECODER * dec, |
1127 |
|
const MACROBLOCK * pMB, |
1128 |
|
const uint32_t x_pos, |
1129 |
|
const uint32_t y_pos, |
1130 |
|
const uint32_t cbp, |
1131 |
|
Bitstream * bs, |
1132 |
|
const uint32_t quant, |
1133 |
|
const uint8_t ref) |
1134 |
{ |
{ |
1135 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
|
1136 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1137 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
1138 |
continue; |
|
1139 |
|
uint32_t stride = dec->edged_width; |
1140 |
|
uint32_t stride2 = stride / 2; |
1141 |
|
uint32_t next_block = stride * 8; |
1142 |
|
uint32_t i; |
1143 |
|
uint32_t iQuant = pMB->quant; |
1144 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1145 |
|
int uv_dx, uv_dy; |
1146 |
|
|
1147 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1148 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1149 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1150 |
|
|
1151 |
|
|
1152 |
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1153 |
|
uv_dx = pMB->mvs[0].x; |
1154 |
|
uv_dy = pMB->mvs[0].y; |
1155 |
|
|
1156 |
|
if (dec->quarterpel) |
1157 |
|
{ |
1158 |
|
uv_dx /= 2; |
1159 |
|
uv_dy /= 2; |
1160 |
} |
} |
1161 |
|
|
1162 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1163 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1164 |
|
} else { |
1165 |
|
int sum; |
1166 |
|
|
1167 |
|
if(dec->quarterpel) |
1168 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1169 |
|
else |
1170 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1171 |
|
|
1172 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1173 |
|
|
1174 |
|
if(dec->quarterpel) |
1175 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1176 |
|
else |
1177 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1178 |
|
|
1179 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1180 |
} |
} |
1181 |
else // not coded |
|
1182 |
|
start_timer(); |
1183 |
|
if(dec->quarterpel) { |
1184 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
1185 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1186 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1187 |
|
} |
1188 |
|
else { |
1189 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
1190 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1191 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1192 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1193 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1194 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1195 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1196 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1197 |
|
} |
1198 |
|
|
1199 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
1200 |
|
uv_dx, uv_dy, stride2, 0); |
1201 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
1202 |
|
uv_dx, uv_dy, stride2, 0); |
1203 |
|
stop_comp_timer(); |
1204 |
|
|
1205 |
|
for (i = 0; i < 6; i++) { |
1206 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1207 |
|
|
1208 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1209 |
{ |
{ |
1210 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1211 |
|
|
1212 |
mb->mode = MODE_NOT_CODED; |
start_timer(); |
1213 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
get_inter_block(bs, &block[i * 64], direction); |
1214 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
stop_coding_timer(); |
1215 |
|
|
1216 |
|
start_timer(); |
1217 |
|
if (dec->quant_type == 0) { |
1218 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1219 |
|
} else { |
1220 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1221 |
|
} |
1222 |
|
stop_iquant_timer(); |
1223 |
|
|
1224 |
// copy macroblock directly from ref to cur |
start_timer(); |
1225 |
|
idct(&data[i * 64]); |
1226 |
|
stop_idct_timer(); |
1227 |
|
} |
1228 |
|
} |
1229 |
|
|
1230 |
|
if (dec->interlacing && pMB->field_dct) { |
1231 |
|
next_block = stride; |
1232 |
|
stride *= 2; |
1233 |
|
} |
1234 |
|
|
1235 |
start_timer(); |
start_timer(); |
1236 |
|
if (cbp & 32) |
1237 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1238 |
|
if (cbp & 16) |
1239 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1240 |
|
if (cbp & 8) |
1241 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1242 |
|
if (cbp & 4) |
1243 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1244 |
|
if (cbp & 2) |
1245 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1246 |
|
if (cbp & 1) |
1247 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1248 |
|
stop_transfer_timer(); |
1249 |
|
} |
1250 |
|
|
1251 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
/* decode an B-frame direct & inter macroblock */ |
1252 |
dec->refn.y + (16*y)*dec->edged_width + (16*x), |
void |
1253 |
dec->edged_width); |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1254 |
|
IMAGE forward, |
1255 |
|
IMAGE backward, |
1256 |
|
const MACROBLOCK * pMB, |
1257 |
|
const uint32_t x_pos, |
1258 |
|
const uint32_t y_pos, |
1259 |
|
Bitstream * bs) |
1260 |
|
{ |
1261 |
|
|
1262 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1263 |
dec->refn.y + (16*y)*dec->edged_width + (16*x+8), |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
dec->edged_width); |
|
1264 |
|
|
1265 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
uint32_t stride = dec->edged_width; |
1266 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x), |
uint32_t stride2 = stride / 2; |
1267 |
dec->edged_width); |
uint32_t next_block = stride * 8; |
1268 |
|
uint32_t iQuant = pMB->quant; |
1269 |
|
int uv_dx, uv_dy; |
1270 |
|
int b_uv_dx, b_uv_dy; |
1271 |
|
uint32_t i; |
1272 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1273 |
|
const uint32_t cbp = pMB->cbp; |
1274 |
|
|
1275 |
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1276 |
dec->refn.y + (16*y+8)*dec->edged_width + (16*x+8), |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1277 |
dec->edged_width); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1278 |
|
|
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->refn.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
1279 |
|
|
1280 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1281 |
dec->refn.v + (8*y)*dec->edged_width/2 + (8*x), |
uv_dx = pMB->mvs[0].x; |
1282 |
dec->edged_width/2); |
uv_dy = pMB->mvs[0].y; |
1283 |
|
|
1284 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1285 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1286 |
|
|
1287 |
|
if (dec->quarterpel) |
1288 |
|
{ |
1289 |
|
uv_dx /= 2; |
1290 |
|
uv_dy /= 2; |
1291 |
|
|
1292 |
|
b_uv_dx /= 2; |
1293 |
|
b_uv_dy /= 2; |
1294 |
|
} |
1295 |
|
|
1296 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1297 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1298 |
|
|
1299 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1300 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1301 |
|
} else { |
1302 |
|
int sum; |
1303 |
|
|
1304 |
|
if(dec->quarterpel) |
1305 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1306 |
|
else |
1307 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1308 |
|
|
1309 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1310 |
|
|
1311 |
|
if(dec->quarterpel) |
1312 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1313 |
|
else |
1314 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1315 |
|
|
1316 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1317 |
|
|
1318 |
|
|
1319 |
|
if(dec->quarterpel) |
1320 |
|
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); |
1321 |
|
else |
1322 |
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1323 |
|
|
1324 |
|
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1325 |
|
|
1326 |
|
if(dec->quarterpel) |
1327 |
|
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); |
1328 |
|
else |
1329 |
|
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1330 |
|
|
1331 |
|
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1332 |
|
} |
1333 |
|
|
1334 |
|
|
1335 |
|
start_timer(); |
1336 |
|
if(dec->quarterpel) { |
1337 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1338 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1339 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1340 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1341 |
|
else { |
1342 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1343 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1344 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1345 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1346 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1347 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1348 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1349 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1350 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1351 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1352 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1353 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1354 |
|
} |
1355 |
|
} |
1356 |
|
else { |
1357 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1358 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1359 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1360 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1361 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1362 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1363 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1364 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1365 |
|
0); |
1366 |
|
} |
1367 |
|
|
1368 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1369 |
|
uv_dy, stride2, 0); |
1370 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1371 |
|
uv_dy, stride2, 0); |
1372 |
|
|
1373 |
|
|
1374 |
|
if(dec->quarterpel) { |
1375 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1376 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1377 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1378 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1379 |
|
else { |
1380 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1381 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1382 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1383 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1384 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1385 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1386 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1387 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1388 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1389 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1390 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1391 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1392 |
|
} |
1393 |
|
} |
1394 |
|
else { |
1395 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1396 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1397 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1398 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1399 |
|
0); |
1400 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1401 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1402 |
|
stride, 0); |
1403 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1404 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1405 |
|
stride, 0); |
1406 |
|
} |
1407 |
|
|
1408 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1409 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1410 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1411 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1412 |
|
|
1413 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1414 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1415 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1416 |
|
stride, 1, 8); |
1417 |
|
|
1418 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1419 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1420 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1421 |
|
stride, 1, 8); |
1422 |
|
|
1423 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1424 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1425 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1426 |
|
stride, 1, 8); |
1427 |
|
|
1428 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1429 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1430 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1431 |
|
stride, 1, 8); |
1432 |
|
|
1433 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1434 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1435 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1436 |
|
stride2, 1, 8); |
1437 |
|
|
1438 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1439 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1440 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1441 |
|
stride2, 1, 8); |
1442 |
|
|
1443 |
|
stop_comp_timer(); |
1444 |
|
|
1445 |
|
for (i = 0; i < 6; i++) { |
1446 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1447 |
|
|
1448 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1449 |
|
{ |
1450 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1451 |
|
|
1452 |
|
start_timer(); |
1453 |
|
get_inter_block(bs, &block[i * 64], direction); |
1454 |
|
stop_coding_timer(); |
1455 |
|
|
1456 |
|
start_timer(); |
1457 |
|
if (dec->quant_type == 0) { |
1458 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1459 |
|
} else { |
1460 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1461 |
|
} |
1462 |
|
stop_iquant_timer(); |
1463 |
|
|
1464 |
|
start_timer(); |
1465 |
|
idct(&data[i * 64]); |
1466 |
|
stop_idct_timer(); |
1467 |
|
} |
1468 |
|
} |
1469 |
|
|
1470 |
|
if (dec->interlacing && pMB->field_dct) { |
1471 |
|
next_block = stride; |
1472 |
|
stride *= 2; |
1473 |
|
} |
1474 |
|
|
1475 |
|
start_timer(); |
1476 |
|
if (cbp & 32) |
1477 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1478 |
|
if (cbp & 16) |
1479 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1480 |
|
if (cbp & 8) |
1481 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1482 |
|
if (cbp & 4) |
1483 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1484 |
|
if (cbp & 2) |
1485 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1486 |
|
if (cbp & 1) |
1487 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1488 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1489 |
} |
} |
1490 |
|
|
1491 |
|
|
1492 |
|
/* for decode B-frame dbquant */ |
1493 |
|
int32_t __inline |
1494 |
|
get_dbquant(Bitstream * bs) |
1495 |
|
{ |
1496 |
|
if (!BitstreamGetBit(bs)) /* '0' */ |
1497 |
|
return (0); |
1498 |
|
else if (!BitstreamGetBit(bs)) /* '10' */ |
1499 |
|
return (-2); |
1500 |
|
else /* '11' */ |
1501 |
|
return (2); |
1502 |
|
} |
1503 |
|
|
1504 |
|
/* |
1505 |
|
* For decode B-frame mb_type |
1506 |
|
* bit ret_value |
1507 |
|
* 1 0 |
1508 |
|
* 01 1 |
1509 |
|
* 001 2 |
1510 |
|
* 0001 3 |
1511 |
|
*/ |
1512 |
|
int32_t __inline |
1513 |
|
get_mbtype(Bitstream * bs) |
1514 |
|
{ |
1515 |
|
int32_t mb_type; |
1516 |
|
|
1517 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) { |
1518 |
|
if (BitstreamGetBit(bs)) |
1519 |
|
break; |
1520 |
|
} |
1521 |
|
|
1522 |
|
if (mb_type <= 3) |
1523 |
|
return (mb_type); |
1524 |
|
else |
1525 |
|
return (-1); |
1526 |
|
} |
1527 |
|
|
1528 |
|
void |
1529 |
|
decoder_bframe(DECODER * dec, |
1530 |
|
Bitstream * bs, |
1531 |
|
int quant, |
1532 |
|
int fcode_forward, |
1533 |
|
int fcode_backward) |
1534 |
|
{ |
1535 |
|
uint32_t x, y; |
1536 |
|
VECTOR mv; |
1537 |
|
const VECTOR zeromv = {0,0}; |
1538 |
|
#ifdef BFRAMES_DEC_DEBUG |
1539 |
|
FILE *fp; |
1540 |
|
static char first=0; |
1541 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1542 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1543 |
|
} |
1544 |
|
#endif |
1545 |
|
|
1546 |
|
start_timer(); |
1547 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1548 |
|
dec->width, dec->height); |
1549 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1550 |
|
dec->width, dec->height); |
1551 |
|
stop_edges_timer(); |
1552 |
|
|
1553 |
|
#ifdef BFRAMES_DEC_DEBUG |
1554 |
|
if (!first){ |
1555 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1556 |
|
} |
1557 |
|
#endif |
1558 |
|
|
1559 |
|
for (y = 0; y < dec->mb_height; y++) { |
1560 |
|
/* Initialize Pred Motion Vector */ |
1561 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1562 |
|
for (x = 0; x < dec->mb_width; x++) { |
1563 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1564 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1565 |
|
|
1566 |
|
mv = |
1567 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1568 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1569 |
|
|
1570 |
|
/* |
1571 |
|
* skip if the co-located P_VOP macroblock is not coded |
1572 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1573 |
|
* automatically skipped |
1574 |
|
*/ |
1575 |
|
|
1576 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1577 |
|
/* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
1578 |
|
mb->cbp = 0; |
1579 |
|
#ifdef BFRAMES_DEC_DEBUG |
1580 |
|
mb->mb_type = MODE_NOT_CODED; |
1581 |
|
BFRAME_DEBUG |
1582 |
|
#endif |
1583 |
|
mb->mb_type = MODE_FORWARD; |
1584 |
|
mb->quant = last_mb->quant; |
1585 |
|
/* |
1586 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1587 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1588 |
|
*/ |
1589 |
|
|
1590 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1591 |
|
continue; |
1592 |
|
} |
1593 |
|
|
1594 |
|
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1595 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1596 |
|
|
1597 |
|
mb->mb_type = get_mbtype(bs); |
1598 |
|
|
1599 |
|
if (!modb2) { /* modb=='00' */ |
1600 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1601 |
|
} else { |
1602 |
|
mb->cbp = 0; |
1603 |
|
} |
1604 |
|
if (mb->mb_type && mb->cbp) { |
1605 |
|
quant += get_dbquant(bs); |
1606 |
|
|
1607 |
|
if (quant > 31) { |
1608 |
|
quant = 31; |
1609 |
|
} else if (quant < 1) { |
1610 |
|
quant = 1; |
1611 |
|
} |
1612 |
|
} |
1613 |
|
} else { |
1614 |
|
mb->mb_type = MODE_DIRECT_NONE_MV; |
1615 |
|
mb->cbp = 0; |
1616 |
|
} |
1617 |
|
|
1618 |
|
mb->quant = quant; |
1619 |
|
mb->mode = MODE_INTER4V; |
1620 |
|
/* DEBUG1("Switch bm_type=",mb->mb_type); */ |
1621 |
|
|
1622 |
|
#ifdef BFRAMES_DEC_DEBUG |
1623 |
|
BFRAME_DEBUG |
1624 |
|
#endif |
1625 |
|
|
1626 |
|
switch (mb->mb_type) { |
1627 |
|
case MODE_DIRECT: |
1628 |
|
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1629 |
|
|
1630 |
|
case MODE_DIRECT_NONE_MV: |
1631 |
|
{ |
1632 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1633 |
|
int i; |
1634 |
|
|
1635 |
|
for (i = 0; i < 4; i++) { |
1636 |
|
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1637 |
|
/ TRD + mv.x); |
1638 |
|
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1639 |
|
? ((TRB - TRD) * last_mb->mvs[i].x) |
1640 |
|
/ TRD |
1641 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1642 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1643 |
|
/ TRD + mv.y); |
1644 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1645 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1646 |
|
/ TRD |
1647 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1648 |
|
} |
1649 |
|
/* DEBUG("B-frame Direct!\n"); */ |
1650 |
|
} |
1651 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1652 |
|
mb, x, y, bs); |
1653 |
|
break; |
1654 |
|
|
1655 |
|
case MODE_INTERPOLATE: |
1656 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1657 |
|
dec->p_fmv); |
1658 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1659 |
|
|
1660 |
|
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1661 |
|
fcode_backward, dec->p_bmv); |
1662 |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1663 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1664 |
|
|
1665 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1666 |
|
mb, x, y, bs); |
1667 |
|
/* DEBUG("B-frame Bidir!\n"); */ |
1668 |
|
break; |
1669 |
|
|
1670 |
|
case MODE_BACKWARD: |
1671 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1672 |
|
dec->p_bmv); |
1673 |
|
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1674 |
|
|
1675 |
|
mb->mode = MODE_INTER; |
1676 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1677 |
|
/* DEBUG("B-frame Backward!\n"); */ |
1678 |
|
break; |
1679 |
|
|
1680 |
|
case MODE_FORWARD: |
1681 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1682 |
|
dec->p_fmv); |
1683 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1684 |
|
|
1685 |
|
mb->mode = MODE_INTER; |
1686 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1687 |
|
/* DEBUG("B-frame Forward!\n"); */ |
1688 |
|
break; |
1689 |
|
|
1690 |
|
default: |
1691 |
|
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
1692 |
|
} |
1693 |
|
} /* End of for */ |
1694 |
|
} |
1695 |
|
|
1696 |
|
#ifdef BFRAMES_DEC_DEBUG |
1697 |
|
if (!first){ |
1698 |
|
first=1; |
1699 |
|
if (fp) |
1700 |
|
fclose(fp); |
1701 |
|
} |
1702 |
|
#endif |
1703 |
|
} |
1704 |
|
|
1705 |
|
/* swap two MACROBLOCK array */ |
1706 |
|
void |
1707 |
|
mb_swap(MACROBLOCK ** mb1, |
1708 |
|
MACROBLOCK ** mb2) |
1709 |
|
{ |
1710 |
|
MACROBLOCK *temp = *mb1; |
1711 |
|
|
1712 |
|
*mb1 = *mb2; |
1713 |
|
*mb2 = temp; |
1714 |
} |
} |
1715 |
|
|
1716 |
|
|
1717 |
|
/* perform post processing if necessary, and output the image */ |
1718 |
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1719 |
|
const XVID_DEC_FRAME * frame, int pp_disable) |
1720 |
|
{ |
1721 |
|
|
1722 |
|
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
1723 |
|
{ |
1724 |
|
/* note: image is stored to tmp */ |
1725 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1726 |
|
image_deblock_rrv(&dec->tmp, dec->edged_width, |
1727 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1728 |
|
8, frame->general); |
1729 |
|
img = &dec->tmp; |
1730 |
} |
} |
1731 |
|
|
1732 |
|
image_output(img, dec->width, dec->height, |
1733 |
|
dec->edged_width, frame->image, frame->stride, |
1734 |
|
frame->colorspace, dec->interlacing); |
1735 |
} |
} |
1736 |
|
|
1737 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
|
1738 |
|
int |
1739 |
|
decoder_decode(DECODER * dec, |
1740 |
|
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
1741 |
{ |
{ |
1742 |
|
|
1743 |
Bitstream bs; |
Bitstream bs; |
1744 |
uint32_t rounding; |
uint32_t rounding; |
1745 |
|
uint32_t reduced_resolution; |
1746 |
uint32_t quant; |
uint32_t quant; |
1747 |
uint32_t fcode; |
uint32_t fcode_forward; |
1748 |
|
uint32_t fcode_backward; |
1749 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1750 |
|
WARPPOINTS gmc_warp; |
1751 |
|
uint32_t vop_type; |
1752 |
|
int success = 0; |
1753 |
|
int output = 0; |
1754 |
|
int seen_something = 0; |
1755 |
|
|
1756 |
start_global_timer(); |
start_global_timer(); |
1757 |
|
|
1758 |
|
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
1759 |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1760 |
|
|
1761 |
|
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
1762 |
|
dec->frames = 0; |
1763 |
|
|
1764 |
|
if (frame->length < 0) /* decoder flush */ |
1765 |
|
{ |
1766 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1767 |
|
we have a reference frame, then outout the reference frame */ |
1768 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
1769 |
|
{ |
1770 |
|
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
1771 |
|
output = 1; |
1772 |
|
} |
1773 |
|
|
1774 |
|
frame->length = 0; |
1775 |
|
if (stats) |
1776 |
|
{ |
1777 |
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
1778 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1779 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1780 |
|
} |
1781 |
|
|
1782 |
|
emms(); |
1783 |
|
|
1784 |
|
stop_global_timer(); |
1785 |
|
return XVID_ERR_OK; |
1786 |
|
} |
1787 |
|
|
1788 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1789 |
|
|
1790 |
switch (BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode, &intra_dc_threshold)) |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1791 |
|
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1792 |
{ |
{ |
1793 |
case P_VOP : |
if (stats) |
1794 |
decoder_pframe(dec, &bs, rounding, quant, fcode, intra_dc_threshold); |
stats->notify = XVID_DEC_VOP; |
1795 |
break; |
frame->length = 1; |
1796 |
|
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1797 |
|
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
1798 |
|
emms(); |
1799 |
|
return XVID_ERR_OK; |
1800 |
|
} |
1801 |
|
|
1802 |
|
repeat: |
1803 |
|
|
1804 |
|
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1805 |
|
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1806 |
|
|
1807 |
|
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", |
1808 |
|
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1809 |
|
|
1810 |
|
if (vop_type == - 1) |
1811 |
|
{ |
1812 |
|
if (success) goto done; |
1813 |
|
emms(); |
1814 |
|
return XVID_ERR_FAIL; |
1815 |
|
} |
1816 |
|
|
1817 |
|
if (vop_type == -2 || vop_type == -3) |
1818 |
|
{ |
1819 |
|
if (vop_type == -3) |
1820 |
|
decoder_resize(dec); |
1821 |
|
|
1822 |
|
if (stats) |
1823 |
|
{ |
1824 |
|
stats->notify = XVID_DEC_VOL; |
1825 |
|
stats->data.vol.general = 0; |
1826 |
|
if (dec->interlacing) |
1827 |
|
stats->data.vol.general |= XVID_INTERLACING; |
1828 |
|
stats->data.vol.width = dec->width; |
1829 |
|
stats->data.vol.height = dec->height; |
1830 |
|
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
1831 |
|
stats->data.vol.par_width = dec->par_width; |
1832 |
|
stats->data.vol.par_height = dec->par_height; |
1833 |
|
frame->length = BitstreamPos(&bs) / 8; |
1834 |
|
emms(); |
1835 |
|
return XVID_ERR_OK; |
1836 |
|
} |
1837 |
|
goto repeat; |
1838 |
|
} |
1839 |
|
|
1840 |
|
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1841 |
|
|
1842 |
|
|
1843 |
|
/* packed_mode: special-N_VOP treament */ |
1844 |
|
if (dec->packed_mode && vop_type == N_VOP) |
1845 |
|
{ |
1846 |
|
if (dec->low_delay_default && dec->frames > 0) |
1847 |
|
{ |
1848 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1849 |
|
output = 1; |
1850 |
|
} |
1851 |
|
/* ignore otherwise */ |
1852 |
|
} |
1853 |
|
else if (vop_type != B_VOP) |
1854 |
|
{ |
1855 |
|
switch(vop_type) |
1856 |
|
{ |
1857 |
case I_VOP : |
case I_VOP : |
1858 |
//DEBUG1("",intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
|
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
|
1859 |
break; |
break; |
1860 |
|
case P_VOP : |
1861 |
case B_VOP : // ignore |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1862 |
|
fcode_forward, intra_dc_threshold, NULL); |
1863 |
break; |
break; |
1864 |
|
case S_VOP : |
1865 |
case N_VOP : // vop not coded |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1866 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1867 |
|
break; |
1868 |
|
case N_VOP : |
1869 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1870 |
break; |
break; |
1871 |
|
} |
1872 |
|
|
1873 |
default : |
if (reduced_resolution) |
1874 |
return XVID_ERR_FAIL; |
{ |
1875 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1876 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1877 |
|
16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV); |
1878 |
|
} |
1879 |
|
|
1880 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1881 |
|
if (!(dec->low_delay_default && dec->packed_mode)) |
1882 |
|
{ |
1883 |
|
if (dec->low_delay) |
1884 |
|
{ |
1885 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
1886 |
|
output = 1; |
1887 |
|
} |
1888 |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
1889 |
|
{ |
1890 |
|
/* output the reference frame */ |
1891 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1892 |
|
output = 1; |
1893 |
|
} |
1894 |
|
} |
1895 |
|
|
1896 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1897 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1898 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1899 |
|
dec->last_reduced_resolution = reduced_resolution; |
1900 |
|
|
1901 |
|
dec->frames++; |
1902 |
|
seen_something = 1; |
1903 |
|
|
1904 |
|
}else{ /* B_VOP */ |
1905 |
|
|
1906 |
|
if (dec->low_delay) |
1907 |
|
{ |
1908 |
|
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
1909 |
|
dec->low_delay = 1; |
1910 |
|
} |
1911 |
|
|
1912 |
|
if (dec->frames < 2) |
1913 |
|
{ |
1914 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1915 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1916 |
|
"broken b-frame, mising ref frames"); |
1917 |
|
}else if (dec->time_pp <= dec->time_bp) { |
1918 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1919 |
|
decoded in vfw. */ |
1920 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1921 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1922 |
|
}else{ |
1923 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1924 |
|
} |
1925 |
|
|
1926 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
1927 |
|
output = 1; |
1928 |
|
dec->frames++; |
1929 |
|
} |
1930 |
|
|
1931 |
|
BitstreamByteAlign(&bs); |
1932 |
|
|
1933 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1934 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
1935 |
|
{ |
1936 |
|
success = 1; |
1937 |
|
goto repeat; |
1938 |
|
} |
1939 |
|
|
1940 |
|
done : |
1941 |
|
|
1942 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1943 |
|
then output the recently decoded frame, or print an error message */ |
1944 |
|
if (dec->low_delay_default && output == 0) |
1945 |
|
{ |
1946 |
|
if (dec->packed_mode && seen_something) |
1947 |
|
{ |
1948 |
|
/* output the recently decoded frame */ |
1949 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
1950 |
|
output = 1; |
1951 |
|
} |
1952 |
|
else |
1953 |
|
{ |
1954 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1955 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1956 |
|
"warning: nothing to output"); |
1957 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1958 |
|
"bframe decoder lag"); |
1959 |
|
|
1960 |
|
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
1961 |
|
} |
1962 |
} |
} |
1963 |
|
|
1964 |
frame->length = BitstreamPos(&bs) / 8; |
frame->length = BitstreamPos(&bs) / 8; |
1965 |
|
|
1966 |
start_timer(); |
if (stats) |
1967 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
{ |
1968 |
frame->image, frame->stride, frame->colorspace); |
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
1969 |
stop_conv_timer(); |
stats->data.vop.time_base = (int)dec->time_base; |
1970 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1971 |
|
} |
1972 |
|
|
1973 |
emms(); |
emms(); |
1974 |
|
|
1975 |
stop_global_timer(); |
stop_global_timer(); |
1976 |
|
|
1977 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
|
|
|
1978 |
} |
} |