1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2002-2004 Peter Ross <pross@xvid.org> |
|
* 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. |
|
8 |
* |
* |
9 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
10 |
* 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 |
20 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
|
* |
|
|
* 02.05.2002 add B-frame decode support(have some problem); |
|
|
* MinChen <chenm001@163.com> |
|
|
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
|
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
|
|
* reconstructing blocks, thus artifacts |
|
|
* interlacing speedup - used transfers to re-interlace |
|
|
* interlaced decoding should be as fast as progressive now |
|
|
* 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 lock based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
23 |
* $Id$ |
* $Id$ |
24 |
* |
* |
25 |
*************************************************************************/ |
****************************************************************************/ |
26 |
|
|
27 |
|
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.h> |
29 |
#include <string.h> |
#include <string.h> |
30 |
|
|
31 |
|
#ifdef BFRAMES_DEC_DEBUG |
32 |
|
#define BFRAMES_DEC |
33 |
|
#endif |
34 |
|
|
35 |
#include "xvid.h" |
#include "xvid.h" |
36 |
#include "portab.h" |
#include "portab.h" |
37 |
|
#include "global.h" |
38 |
|
|
39 |
#include "decoder.h" |
#include "decoder.h" |
40 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
41 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
42 |
|
|
43 |
#include "quant/quant_h263.h" |
#include "quant/quant.h" |
44 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_matrix.h" |
45 |
#include "dct/idct.h" |
#include "dct/idct.h" |
46 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
47 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
48 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
49 |
|
#include "image/reduced.h" |
50 |
|
#include "image/font.h" |
51 |
|
|
52 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
53 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
54 |
#include "utils/timer.h" |
#include "utils/timer.h" |
55 |
#include "utils/emms.h" |
#include "utils/emms.h" |
56 |
|
#include "motion/motion.h" |
57 |
|
#include "motion/gmc.h" |
58 |
|
|
59 |
#include "image/image.h" |
#include "image/image.h" |
60 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
61 |
|
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
int decoder_create(XVID_DEC_PARAM * param) |
static int |
65 |
|
decoder_resize(DECODER * dec) |
66 |
{ |
{ |
67 |
DECODER * dec; |
/* free existing */ |
68 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
69 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
70 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
71 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
72 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
73 |
|
|
74 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
|
if (dec == NULL) |
|
|
{ |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
param->handle = dec; |
|
75 |
|
|
76 |
dec->width = param->width; |
if (dec->last_mbs) |
77 |
dec->height = param->height; |
xvid_free(dec->last_mbs); |
78 |
|
if (dec->mbs) |
79 |
|
xvid_free(dec->mbs); |
80 |
|
if (dec->qscale) |
81 |
|
xvid_free(dec->qscale); |
82 |
|
|
83 |
|
/* realloc */ |
84 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
85 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
86 |
|
|
87 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
88 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
89 |
|
|
90 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
|
{ |
|
91 |
xvid_free(dec); |
xvid_free(dec); |
92 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
93 |
} |
} |
94 |
|
|
95 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|
{ |
|
96 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
97 |
xvid_free(dec); |
xvid_free(dec); |
98 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
99 |
} |
} |
100 |
// add by chenm001 <chenm001@163.com> |
|
101 |
// for support B-frame to reference last 2 frame |
/* Support B-frame to reference last 2 frame */ |
102 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
{ |
|
103 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
104 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
105 |
xvid_free(dec); |
xvid_free(dec); |
106 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
107 |
} |
} |
108 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
|
{ |
|
109 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
110 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
111 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
113 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
114 |
} |
} |
115 |
|
|
116 |
dec->mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
|
if (dec->mbs == NULL) |
|
|
{ |
|
117 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
118 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
119 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
120 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
121 |
xvid_free(dec); |
xvid_free(dec); |
122 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
123 |
} |
} |
124 |
// add by chenm001 <chenm001@163.com> |
|
125 |
// for skip MB flag |
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
126 |
dec->last_mbs = xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, CACHE_LINE); |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
127 |
if (dec->last_mbs == NULL) |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
128 |
{ |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
129 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
130 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
131 |
|
xvid_free(dec); |
132 |
|
return XVID_ERR_MEMORY; |
133 |
|
} |
134 |
|
|
135 |
|
dec->mbs = |
136 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
137 |
|
CACHE_LINE); |
138 |
|
if (dec->mbs == NULL) { |
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 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
143 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
144 |
|
xvid_free(dec); |
145 |
|
return XVID_ERR_MEMORY; |
146 |
|
} |
147 |
|
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
148 |
|
|
149 |
|
/* For skip MB flag */ |
150 |
|
dec->last_mbs = |
151 |
|
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
152 |
|
CACHE_LINE); |
153 |
|
if (dec->last_mbs == NULL) { |
154 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
155 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
156 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
157 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
158 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
159 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
160 |
|
xvid_free(dec); |
161 |
|
return XVID_ERR_MEMORY; |
162 |
|
} |
163 |
|
|
164 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
165 |
|
|
166 |
|
/* nothing happens if that fails */ |
167 |
|
dec->qscale = |
168 |
|
xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE); |
169 |
|
|
170 |
|
if (dec->qscale) |
171 |
|
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
172 |
|
|
173 |
|
return 0; |
174 |
|
} |
175 |
|
|
176 |
|
|
177 |
|
int |
178 |
|
decoder_create(xvid_dec_create_t * create) |
179 |
|
{ |
180 |
|
DECODER *dec; |
181 |
|
|
182 |
|
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
183 |
|
return XVID_ERR_VERSION; |
184 |
|
|
185 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
186 |
|
if (dec == NULL) { |
187 |
|
return XVID_ERR_MEMORY; |
188 |
|
} |
189 |
|
|
190 |
|
memset(dec, 0, sizeof(DECODER)); |
191 |
|
|
192 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
193 |
|
if (dec->mpeg_quant_matrices == NULL) { |
194 |
xvid_free(dec); |
xvid_free(dec); |
195 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
196 |
} |
} |
197 |
|
|
198 |
|
create->handle = dec; |
199 |
|
|
200 |
|
dec->width = create->width; |
201 |
|
dec->height = create->height; |
202 |
|
|
203 |
|
image_null(&dec->cur); |
204 |
|
image_null(&dec->refn[0]); |
205 |
|
image_null(&dec->refn[1]); |
206 |
|
image_null(&dec->tmp); |
207 |
|
image_null(&dec->qtmp); |
208 |
|
|
209 |
|
/* image based GMC */ |
210 |
|
image_null(&dec->gmc); |
211 |
|
|
212 |
|
dec->mbs = NULL; |
213 |
|
dec->last_mbs = NULL; |
214 |
|
dec->qscale = NULL; |
215 |
|
|
216 |
init_timer(); |
init_timer(); |
217 |
|
init_postproc(&dec->postproc); |
218 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
219 |
|
|
220 |
// add by chenm001 <chenm001@163.com> |
/* For B-frame support (used to save reference frame's time */ |
221 |
// for support B-frame to save reference frame's time |
dec->frames = 0; |
|
dec->frames = -1; |
|
222 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
223 |
|
dec->low_delay = 0; |
224 |
|
dec->packed_mode = 0; |
225 |
|
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
226 |
|
|
227 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
228 |
|
|
229 |
return XVID_ERR_OK; |
if (dec->fixed_dimensions) |
230 |
|
return decoder_resize(dec); |
231 |
|
else |
232 |
|
return 0; |
233 |
} |
} |
234 |
|
|
235 |
|
|
236 |
int decoder_destroy(DECODER * dec) |
int |
237 |
|
decoder_destroy(DECODER * dec) |
238 |
{ |
{ |
239 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
240 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
241 |
|
xvid_free(dec->qscale); |
242 |
|
|
243 |
|
/* image based GMC */ |
244 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
245 |
|
|
246 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
247 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
248 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
249 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
250 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
251 |
|
xvid_free(dec->mpeg_quant_matrices); |
252 |
xvid_free(dec); |
xvid_free(dec); |
253 |
|
|
254 |
write_timer(); |
write_timer(); |
255 |
return XVID_ERR_OK; |
return 0; |
256 |
} |
} |
257 |
|
|
258 |
|
static const int32_t dquant_table[4] = { |
|
|
|
|
static const int32_t dquant_table[4] = |
|
|
{ |
|
259 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
260 |
}; |
}; |
261 |
|
|
262 |
|
/* decode an intra macroblock */ |
263 |
|
static void |
264 |
|
decoder_mbintra(DECODER * dec, |
|
// decode an intra macroblock |
|
|
|
|
|
void decoder_mbintra(DECODER * dec, |
|
265 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
266 |
const uint32_t x_pos, |
const uint32_t x_pos, |
267 |
const uint32_t y_pos, |
const uint32_t y_pos, |
269 |
const uint32_t cbp, |
const uint32_t cbp, |
270 |
Bitstream * bs, |
Bitstream * bs, |
271 |
const uint32_t quant, |
const uint32_t quant, |
272 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold, |
273 |
|
const unsigned int bound, |
274 |
|
const int reduced_resolution) |
275 |
{ |
{ |
276 |
|
|
277 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
284 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
285 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
286 |
|
|
287 |
|
if (reduced_resolution) { |
288 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
289 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
290 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
291 |
|
}else{ |
292 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
293 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
294 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
295 |
|
} |
296 |
|
|
297 |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
298 |
|
|
299 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) { |
|
{ |
|
300 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
301 |
int16_t predictors[8]; |
int16_t predictors[8]; |
302 |
int start_coeff; |
int start_coeff; |
303 |
|
|
304 |
start_timer(); |
start_timer(); |
305 |
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], |
306 |
if (!acpred_flag) |
iQuant, iDcScaler, predictors, bound, dec->bs_version); |
307 |
{ |
if (!acpred_flag) { |
308 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
309 |
} |
} |
310 |
stop_prediction_timer(); |
stop_prediction_timer(); |
311 |
|
|
312 |
if (quant < intra_dc_threshold) |
if (quant < intra_dc_threshold) { |
|
{ |
|
313 |
int dc_size; |
int dc_size; |
314 |
int dc_dif; |
int dc_dif; |
315 |
|
|
316 |
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); |
317 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0 ; |
318 |
|
|
319 |
if (dc_size > 8) |
if (dc_size > 8) { |
320 |
{ |
BitstreamSkip(bs, 1); /* marker */ |
|
BitstreamSkip(bs, 1); // marker |
|
321 |
} |
} |
322 |
|
|
323 |
block[i*64 + 0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
324 |
start_coeff = 1; |
start_coeff = 1; |
325 |
} |
|
326 |
else |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
327 |
{ |
} else { |
328 |
start_coeff = 0; |
start_coeff = 0; |
329 |
} |
} |
330 |
|
|
331 |
start_timer(); |
start_timer(); |
332 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
333 |
{ |
{ |
334 |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
int direction = dec->alternate_vertical_scan ? |
335 |
|
2 : pMB->acpred_directions[i]; |
336 |
|
|
337 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
338 |
} |
} |
339 |
stop_coding_timer(); |
stop_coding_timer(); |
340 |
|
|
343 |
stop_prediction_timer(); |
stop_prediction_timer(); |
344 |
|
|
345 |
start_timer(); |
start_timer(); |
346 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) { |
347 |
{ |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
348 |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
} else { |
349 |
} |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
|
else |
|
|
{ |
|
|
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
|
350 |
} |
} |
351 |
stop_iquant_timer(); |
stop_iquant_timer(); |
352 |
|
|
353 |
start_timer(); |
start_timer(); |
354 |
idct(&data[i*64]); |
idct(&data[i*64]); |
355 |
stop_idct_timer(); |
stop_idct_timer(); |
356 |
|
|
357 |
} |
} |
358 |
|
|
359 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
360 |
next_block = stride; |
next_block = stride; |
361 |
stride *= 2; |
stride *= 2; |
362 |
} |
} |
363 |
|
|
364 |
start_timer(); |
start_timer(); |
365 |
|
|
366 |
|
if (reduced_resolution) |
367 |
|
{ |
368 |
|
next_block*=2; |
369 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
370 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
371 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
372 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
373 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
374 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
375 |
|
}else{ |
376 |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
377 |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
378 |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
379 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
380 |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
381 |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
382 |
|
} |
383 |
stop_transfer_timer(); |
stop_transfer_timer(); |
384 |
} |
} |
385 |
|
|
386 |
|
static void |
387 |
|
decoder_mb_decode(DECODER * dec, |
|
|
|
|
|
|
|
#define SIGN(X) (((X)>0)?1:-1) |
|
|
#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 }; |
|
|
|
|
|
|
|
|
// decode an inter macroblock |
|
|
|
|
|
void decoder_mbinter(DECODER * dec, |
|
|
const MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
const uint32_t acpred_flag, |
|
388 |
const uint32_t cbp, |
const uint32_t cbp, |
389 |
Bitstream * bs, |
Bitstream * bs, |
390 |
const uint32_t quant, |
uint8_t * pY_Cur, |
391 |
const uint32_t rounding) |
uint8_t * pU_Cur, |
392 |
|
uint8_t * pV_Cur, |
393 |
|
const int reduced_resolution, |
394 |
|
const MACROBLOCK * pMB) |
395 |
{ |
{ |
396 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
|
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
|
397 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
398 |
|
|
399 |
uint32_t stride = dec->edged_width; |
int stride = dec->edged_width; |
400 |
uint32_t stride2 = stride / 2; |
int next_block = stride * (reduced_resolution ? 16 : 8); |
401 |
uint32_t next_block = stride * 8; |
const int stride2 = stride/2; |
402 |
uint32_t i; |
int i; |
403 |
uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
404 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
405 |
int uv_dx, uv_dy; |
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
|
|
|
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
|
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
|
|
|
|
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
|
|
{ |
|
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
|
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
} |
|
|
else |
|
|
{ |
|
|
int sum; |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
406 |
|
|
407 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
for (i = 0; i < 6; i++) { |
|
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
|
} |
|
408 |
|
|
409 |
start_timer(); |
if (cbp & (1 << (5 - i))) { /* coded */ |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, rounding); |
|
|
stop_comp_timer(); |
|
410 |
|
|
411 |
for (i = 0; i < 6; i++) |
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
|
{ |
|
|
if (cbp & (1 << (5-i))) // coded |
|
|
{ |
|
|
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
|
412 |
|
|
413 |
start_timer(); |
start_timer(); |
414 |
get_inter_block(bs, &block[i*64]); |
get_inter_block(bs, block, direction); |
415 |
stop_coding_timer(); |
stop_coding_timer(); |
416 |
|
|
417 |
start_timer(); |
start_timer(); |
418 |
if (dec->quant_type == 0) |
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
|
{ |
|
|
dequant_inter(&data[i*64], &block[i*64], iQuant); |
|
|
} |
|
|
else |
|
|
{ |
|
|
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
|
|
} |
|
419 |
stop_iquant_timer(); |
stop_iquant_timer(); |
420 |
|
|
421 |
start_timer(); |
start_timer(); |
424 |
} |
} |
425 |
} |
} |
426 |
|
|
427 |
if (dec->interlacing && pMB->field_dct) |
if (dec->interlacing && pMB->field_dct) { |
|
{ |
|
428 |
next_block = stride; |
next_block = stride; |
429 |
stride *= 2; |
stride *= 2; |
430 |
} |
} |
431 |
|
|
432 |
start_timer(); |
start_timer(); |
433 |
|
if (reduced_resolution) { |
434 |
|
if (cbp & 32) |
435 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
436 |
|
if (cbp & 16) |
437 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
438 |
|
if (cbp & 8) |
439 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
440 |
|
if (cbp & 4) |
441 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
442 |
|
if (cbp & 2) |
443 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
444 |
|
if (cbp & 1) |
445 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
446 |
|
} else { |
447 |
if (cbp & 32) |
if (cbp & 32) |
448 |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
449 |
if (cbp & 16) |
if (cbp & 16) |
456 |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
457 |
if (cbp & 1) |
if (cbp & 1) |
458 |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
459 |
|
} |
460 |
stop_transfer_timer(); |
stop_transfer_timer(); |
461 |
} |
} |
462 |
|
|
463 |
|
/* decode an inter macroblock */ |
464 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
static void |
465 |
{ |
decoder_mbinter(DECODER * dec, |
466 |
|
const MACROBLOCK * pMB, |
467 |
uint32_t x, y; |
const uint32_t x_pos, |
468 |
|
const uint32_t y_pos, |
469 |
for (y = 0; y < dec->mb_height; y++) |
const uint32_t cbp, |
470 |
{ |
Bitstream * bs, |
471 |
for (x = 0; x < dec->mb_width; x++) |
const uint32_t rounding, |
472 |
|
const int reduced_resolution, |
473 |
|
const int ref) |
474 |
{ |
{ |
475 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
uint32_t stride = dec->edged_width; |
476 |
|
uint32_t stride2 = stride / 2; |
477 |
uint32_t mcbpc; |
uint32_t i; |
|
uint32_t cbpc; |
|
|
uint32_t acpred_flag; |
|
|
uint32_t cbpy; |
|
|
uint32_t cbp; |
|
478 |
|
|
479 |
mcbpc = get_mcbpc_intra(bs); |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
mb->mode = mcbpc & 7; |
|
|
cbpc = (mcbpc >> 4); |
|
480 |
|
|
481 |
acpred_flag = BitstreamGetBit(bs); |
int uv_dx, uv_dy; |
482 |
|
VECTOR mv[4]; /* local copy of mvs */ |
483 |
|
|
484 |
if (mb->mode == MODE_STUFFING) |
if (reduced_resolution) { |
485 |
{ |
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
486 |
DEBUG("-- STUFFING ?"); |
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
487 |
continue; |
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
488 |
|
for (i = 0; i < 4; i++) { |
489 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
490 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
491 |
|
} |
492 |
|
} else { |
493 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
494 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
495 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
496 |
|
for (i = 0; i < 4; i++) |
497 |
|
mv[i] = pMB->mvs[i]; |
498 |
} |
} |
499 |
|
|
500 |
cbpy = get_cbpy(bs, 1); |
for (i = 0; i < 4; i++) { |
501 |
cbp = (cbpy << 2) | cbpc; |
/* clip to valid range */ |
502 |
|
int border = (int)(dec->mb_width - x_pos) << (5 + dec->quarterpel); |
503 |
|
if (mv[i].x > border) { |
504 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
505 |
|
mv[i].x = border; |
506 |
|
} else { |
507 |
|
border = (-(int)x_pos-1) << (5 + dec->quarterpel); |
508 |
|
if (mv[i].x < border) { |
509 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", mv[i].x, border, x_pos, y_pos); |
510 |
|
mv[i].x = border; |
511 |
|
} |
512 |
|
} |
513 |
|
|
514 |
if (mb->mode == MODE_INTRA_Q) |
border = (int)(dec->mb_height - y_pos) << (5 + dec->quarterpel); |
515 |
{ |
if (mv[i].y > border) { |
516 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
517 |
if (quant > 31) |
mv[i].y = border; |
518 |
{ |
} else { |
519 |
quant = 31; |
border = (-(int)y_pos-1) << (5 + dec->quarterpel); |
520 |
|
if (mv[i].y < border) { |
521 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", mv[i].y, border, x_pos, y_pos); |
522 |
|
mv[i].y = border; |
523 |
} |
} |
|
else if (quant < 1) |
|
|
{ |
|
|
quant = 1; |
|
524 |
} |
} |
525 |
} |
} |
|
mb->quant = quant; |
|
526 |
|
|
527 |
if (dec->interlacing) |
start_timer(); |
528 |
{ |
|
529 |
mb->field_dct = BitstreamGetBit(bs); |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
530 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
|
531 |
|
uv_dx = mv[0].x; |
532 |
|
uv_dy = mv[0].y; |
533 |
|
if (dec->quarterpel) { |
534 |
|
uv_dx /= 2; |
535 |
|
uv_dy /= 2; |
536 |
|
} |
537 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
538 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
539 |
|
|
540 |
|
if (reduced_resolution) |
541 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
542 |
|
mv[0].x, mv[0].y, stride, rounding); |
543 |
|
else if (dec->quarterpel) |
544 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
545 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
546 |
|
mv[0].x, mv[0].y, stride, rounding); |
547 |
|
else |
548 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
549 |
|
mv[0].x, mv[0].y, stride, rounding); |
550 |
|
|
551 |
|
} else { /* MODE_INTER4V */ |
552 |
|
|
553 |
|
if(dec->quarterpel) { |
554 |
|
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
555 |
|
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
556 |
|
} else { |
557 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
558 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
559 |
} |
} |
560 |
|
|
561 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
562 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
563 |
|
|
564 |
|
if (reduced_resolution) { |
565 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
566 |
|
mv[0].x, mv[0].y, stride, rounding); |
567 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
568 |
|
mv[1].x, mv[1].y, stride, rounding); |
569 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
570 |
|
mv[2].x, mv[2].y, stride, rounding); |
571 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
572 |
|
mv[3].x, mv[3].y, stride, rounding); |
573 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
574 |
|
uv_dx, uv_dy, stride2, rounding); |
575 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
576 |
|
uv_dx, uv_dy, stride2, rounding); |
577 |
|
|
578 |
|
} else if (dec->quarterpel) { |
579 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
580 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
581 |
|
mv[0].x, mv[0].y, stride, rounding); |
582 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
583 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
584 |
|
mv[1].x, mv[1].y, stride, rounding); |
585 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
586 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
587 |
|
mv[2].x, mv[2].y, stride, rounding); |
588 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
589 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
590 |
|
mv[3].x, mv[3].y, stride, rounding); |
591 |
|
} else { |
592 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos, |
593 |
|
mv[0].x, mv[0].y, stride, rounding); |
594 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos, |
595 |
|
mv[1].x, mv[1].y, stride, rounding); |
596 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos, 16*y_pos + 8, |
597 |
|
mv[2].x, mv[2].y, stride, rounding); |
598 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[0].y , 16*x_pos + 8, 16*y_pos + 8, |
599 |
|
mv[3].x, mv[3].y, stride, rounding); |
600 |
} |
} |
601 |
} |
} |
602 |
|
|
603 |
|
/* chroma */ |
604 |
|
if (reduced_resolution) { |
605 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
606 |
|
uv_dx, uv_dy, stride2, rounding); |
607 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
608 |
|
uv_dx, uv_dy, stride2, rounding); |
609 |
|
} else { |
610 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
611 |
|
uv_dx, uv_dy, stride2, rounding); |
612 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
613 |
|
uv_dx, uv_dy, stride2, rounding); |
614 |
} |
} |
615 |
|
|
616 |
|
stop_comp_timer(); |
617 |
|
|
618 |
|
if (cbp) |
619 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
620 |
|
reduced_resolution, pMB); |
621 |
|
} |
622 |
|
|
623 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
static void |
624 |
|
decoder_mbgmc(DECODER * dec, |
625 |
|
MACROBLOCK * const pMB, |
626 |
|
const uint32_t x_pos, |
627 |
|
const uint32_t y_pos, |
628 |
|
const uint32_t fcode, |
629 |
|
const uint32_t cbp, |
630 |
|
Bitstream * bs, |
631 |
|
const uint32_t rounding) |
632 |
{ |
{ |
633 |
|
const uint32_t stride = dec->edged_width; |
634 |
|
const uint32_t stride2 = stride / 2; |
635 |
|
|
636 |
int scale_fac = 1 << (fcode - 1); |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
637 |
int high = (32 * scale_fac) - 1; |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
638 |
int low = ((-32) * scale_fac); |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
int range = (64 * scale_fac); |
|
639 |
|
|
640 |
VECTOR pmv[4]; |
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
|
int32_t psad[4]; |
|
641 |
|
|
642 |
int mv_x, mv_y; |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
int pmv_x, pmv_y; |
|
643 |
|
|
644 |
|
start_timer(); |
645 |
|
|
646 |
get_pmvdata(dec->mbs, x, y, dec->mb_width, k, pmv, psad); |
/* this is where the calculations are done */ |
647 |
|
|
648 |
pmv_x = pmv[0].x; |
gmc_data->predict_16x16(gmc_data, |
649 |
pmv_y = pmv[0].y; |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
650 |
|
stride, stride, x_pos, y_pos, rounding); |
651 |
|
|
652 |
mv_x = get_mv(bs, fcode); |
gmc_data->predict_8x8(gmc_data, |
653 |
mv_y = get_mv(bs, fcode); |
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
654 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
655 |
|
stride2, stride2, x_pos, y_pos, rounding); |
656 |
|
|
657 |
mv_x += pmv_x; |
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
|
mv_y += pmv_y; |
|
658 |
|
|
659 |
if (mv_x < low) |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
660 |
{ |
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
|
mv_x += range; |
|
|
} |
|
|
else if (mv_x > high) |
|
|
{ |
|
|
mv_x -= range; |
|
|
} |
|
661 |
|
|
662 |
if (mv_y < low) |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
{ |
|
|
mv_y += range; |
|
|
} |
|
|
else if (mv_y > high) |
|
|
{ |
|
|
mv_y -= range; |
|
|
} |
|
663 |
|
|
664 |
mv->x = mv_x; |
stop_transfer_timer(); |
665 |
mv->y = mv_y; |
|
666 |
|
if (cbp) |
667 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
668 |
|
|
669 |
} |
} |
670 |
|
|
671 |
|
|
672 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
static void |
673 |
|
decoder_iframe(DECODER * dec, |
674 |
|
Bitstream * bs, |
675 |
|
int reduced_resolution, |
676 |
|
int quant, |
677 |
|
int intra_dc_threshold) |
678 |
{ |
{ |
679 |
|
uint32_t bound; |
680 |
uint32_t x, y; |
uint32_t x, y; |
681 |
|
uint32_t mb_width = dec->mb_width; |
682 |
|
uint32_t mb_height = dec->mb_height; |
683 |
|
|
684 |
start_timer(); |
if (reduced_resolution) { |
685 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
mb_width = (dec->width + 31) / 32; |
686 |
stop_edges_timer(); |
mb_height = (dec->height + 31) / 32; |
687 |
|
} |
688 |
|
|
689 |
for (y = 0; y < dec->mb_height; y++) |
bound = 0; |
|
{ |
|
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
|
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
|
690 |
|
|
691 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
for (y = 0; y < mb_height; y++) { |
692 |
if (!(BitstreamGetBit(bs))) // not_coded |
for (x = 0; x < mb_width; x++) { |
693 |
{ |
MACROBLOCK *mb; |
694 |
uint32_t mcbpc; |
uint32_t mcbpc; |
695 |
uint32_t cbpc; |
uint32_t cbpc; |
696 |
uint32_t acpred_flag; |
uint32_t acpred_flag; |
697 |
uint32_t cbpy; |
uint32_t cbpy; |
698 |
uint32_t cbp; |
uint32_t cbp; |
|
uint32_t intra; |
|
|
|
|
|
mcbpc = get_mcbpc_inter(bs); |
|
|
mb->mode = mcbpc & 7; |
|
|
cbpc = (mcbpc >> 4); |
|
|
acpred_flag = 0; |
|
699 |
|
|
700 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
while (BitstreamShowBits(bs, 9) == 1) |
701 |
|
BitstreamSkip(bs, 9); |
702 |
|
|
703 |
if (intra) |
if (check_resync_marker(bs, 0)) |
704 |
{ |
{ |
705 |
acpred_flag = BitstreamGetBit(bs); |
bound = read_video_packet_header(bs, dec, 0, |
706 |
|
&quant, NULL, NULL, &intra_dc_threshold); |
707 |
|
x = bound % mb_width; |
708 |
|
y = bound / mb_width; |
709 |
} |
} |
710 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
711 |
|
|
712 |
if (mb->mode == MODE_STUFFING) |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
|
{ |
|
|
DEBUG("-- STUFFING ?"); |
|
|
continue; |
|
|
} |
|
713 |
|
|
714 |
cbpy = get_cbpy(bs, intra); |
mcbpc = get_mcbpc_intra(bs); |
715 |
|
mb->mode = mcbpc & 7; |
716 |
|
cbpc = (mcbpc >> 4); |
717 |
|
|
718 |
|
acpred_flag = BitstreamGetBit(bs); |
719 |
|
|
720 |
|
cbpy = get_cbpy(bs, 1); |
721 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
722 |
|
|
723 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) |
if (mb->mode == MODE_INTRA_Q) { |
|
{ |
|
724 |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
quant += dquant_table[BitstreamGetBits(bs,2)]; |
725 |
if (quant > 31) |
if (quant > 31) { |
|
{ |
|
726 |
quant = 31; |
quant = 31; |
727 |
} |
} else if (quant < 1) { |
|
else if (mb->quant < 1) |
|
|
{ |
|
728 |
quant = 1; |
quant = 1; |
729 |
} |
} |
730 |
} |
} |
731 |
mb->quant = quant; |
mb->quant = quant; |
732 |
|
mb->mvs[0].x = mb->mvs[0].y = |
733 |
|
mb->mvs[1].x = mb->mvs[1].y = |
734 |
|
mb->mvs[2].x = mb->mvs[2].y = |
735 |
|
mb->mvs[3].x = mb->mvs[3].y =0; |
736 |
|
|
737 |
if (dec->interlacing) |
if (dec->interlacing) { |
|
{ |
|
738 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
739 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
740 |
|
} |
741 |
|
|
742 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
743 |
{ |
intra_dc_threshold, bound, reduced_resolution); |
|
mb->field_pred = BitstreamGetBit(bs); |
|
|
DEBUG1("decp: field_pred: ", mb->field_pred); |
|
744 |
|
|
|
if (mb->field_pred) |
|
|
{ |
|
|
mb->field_for_top = BitstreamGetBit(bs); |
|
|
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
|
|
mb->field_for_bot = BitstreamGetBit(bs); |
|
|
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
|
745 |
} |
} |
746 |
|
if(dec->out_frm) |
747 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
748 |
} |
} |
749 |
|
|
750 |
} |
} |
751 |
|
|
752 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) |
|
753 |
{ |
static void |
754 |
if (dec->interlacing && mb->field_pred) |
get_motion_vector(DECODER * dec, |
755 |
|
Bitstream * bs, |
756 |
|
int x, |
757 |
|
int y, |
758 |
|
int k, |
759 |
|
VECTOR * ret_mv, |
760 |
|
int fcode, |
761 |
|
const int bound) |
762 |
{ |
{ |
763 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
|
764 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode); |
const int scale_fac = 1 << (fcode - 1); |
765 |
|
const int high = (32 * scale_fac) - 1; |
766 |
|
const int low = ((-32) * scale_fac); |
767 |
|
const int range = (64 * scale_fac); |
768 |
|
|
769 |
|
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
770 |
|
VECTOR mv; |
771 |
|
|
772 |
|
mv.x = get_mv(bs, fcode); |
773 |
|
mv.y = get_mv(bs, fcode); |
774 |
|
|
775 |
|
DPRINTF(XVID_DEBUG_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)\n", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
776 |
|
|
777 |
|
mv.x += pmv.x; |
778 |
|
mv.y += pmv.y; |
779 |
|
|
780 |
|
if (mv.x < low) { |
781 |
|
mv.x += range; |
782 |
|
} else if (mv.x > high) { |
783 |
|
mv.x -= range; |
784 |
} |
} |
785 |
else |
|
786 |
{ |
if (mv.y < low) { |
787 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
mv.y += range; |
788 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
} else if (mv.y > high) { |
789 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
mv.y -= range; |
790 |
} |
} |
791 |
|
|
792 |
|
ret_mv->x = mv.x; |
793 |
|
ret_mv->y = mv.y; |
794 |
} |
} |
795 |
else if (mb->mode == MODE_INTER4V /* || mb->mode == MODE_INTER4V_Q */) |
|
796 |
|
/* for P_VOP set gmc_warp to NULL */ |
797 |
|
static void |
798 |
|
decoder_pframe(DECODER * dec, |
799 |
|
Bitstream * bs, |
800 |
|
int rounding, |
801 |
|
int reduced_resolution, |
802 |
|
int quant, |
803 |
|
int fcode, |
804 |
|
int intra_dc_threshold, |
805 |
|
const WARPPOINTS *const gmc_warp) |
806 |
{ |
{ |
807 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode); |
uint32_t x, y; |
808 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode); |
uint32_t bound; |
809 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode); |
int cp_mb, st_mb; |
810 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode); |
uint32_t mb_width = dec->mb_width; |
811 |
|
uint32_t mb_height = dec->mb_height; |
812 |
|
|
813 |
|
if (reduced_resolution) { |
814 |
|
mb_width = (dec->width + 31) / 32; |
815 |
|
mb_height = (dec->height + 31) / 32; |
816 |
} |
} |
817 |
else // MODE_INTRA, MODE_INTRA_Q |
|
818 |
{ |
if (!dec->is_edged[0]) { |
819 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
start_timer(); |
820 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
821 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
dec->width, dec->height, dec->bs_version); |
822 |
continue; |
dec->is_edged[0] = 1; |
823 |
|
stop_edges_timer(); |
824 |
} |
} |
825 |
|
|
826 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, rounding); |
if (gmc_warp) { |
827 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
828 |
|
generate_GMCparameters( dec->sprite_warping_points, |
829 |
|
dec->sprite_warping_accuracy, gmc_warp, |
830 |
|
dec->width, dec->height, &dec->new_gmc_data); |
831 |
|
|
832 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
833 |
} |
} |
|
else // not coded |
|
|
{ |
|
|
//DEBUG2("P-frame MB at (X,Y)=",x,y); |
|
|
mb->mode = MODE_NOT_CODED; |
|
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
834 |
|
|
835 |
// copy macroblock directly from ref to cur |
bound = 0; |
836 |
|
|
837 |
start_timer(); |
for (y = 0; y < mb_height; y++) { |
838 |
|
cp_mb = st_mb = 0; |
839 |
|
for (x = 0; x < mb_width; x++) { |
840 |
|
MACROBLOCK *mb; |
841 |
|
|
842 |
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
/* skip stuffing */ |
843 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
while (BitstreamShowBits(bs, 10) == 1) |
844 |
dec->edged_width); |
BitstreamSkip(bs, 10); |
|
|
|
|
transfer8x8_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x+8), |
|
|
dec->refn[0].y + (16*y)*dec->edged_width + (16*x+8), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x), |
|
|
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer8x8_copy(dec->cur.y + (16*y+8)*dec->edged_width + (16*x+8), |
|
|
dec->refn[0].y + (16*y+8)*dec->edged_width + (16*x+8), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
|
|
dec->edged_width/2); |
|
845 |
|
|
846 |
stop_transfer_timer(); |
if (check_resync_marker(bs, fcode - 1)) { |
847 |
} |
bound = read_video_packet_header(bs, dec, fcode - 1, |
848 |
} |
&quant, &fcode, NULL, &intra_dc_threshold); |
849 |
} |
x = bound % mb_width; |
850 |
|
y = bound / mb_width; |
851 |
} |
} |
852 |
|
mb = &dec->mbs[y * dec->mb_width + x]; |
853 |
|
|
854 |
|
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
855 |
|
|
856 |
// add by MinChen <chenm001@163.com> |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
857 |
// decode B-frame motion vector |
uint32_t mcbpc, cbpc, cbpy, cbp; |
858 |
void get_b_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, VECTOR * mv, int fcode, const VECTOR pmv) |
uint32_t intra, acpred_flag = 0; |
859 |
{ |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
|
int scale_fac = 1 << (fcode - 1); |
|
|
int high = (32 * scale_fac) - 1; |
|
|
int low = ((-32) * scale_fac); |
|
|
int range = (64 * scale_fac); |
|
|
|
|
|
int mv_x, mv_y; |
|
|
int pmv_x, pmv_y; |
|
860 |
|
|
861 |
pmv_x = pmv.x; |
cp_mb++; |
862 |
pmv_y = pmv.y; |
mcbpc = get_mcbpc_inter(bs); |
863 |
|
mb->mode = mcbpc & 7; |
864 |
|
cbpc = (mcbpc >> 4); |
865 |
|
|
866 |
mv_x = get_mv(bs, fcode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
867 |
mv_y = get_mv(bs, fcode); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
868 |
|
|
869 |
mv_x += pmv_x; |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
|
mv_y += pmv_y; |
|
870 |
|
|
871 |
if (mv_x < low) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
872 |
{ |
mcsel = BitstreamGetBit(bs); |
873 |
mv_x += range; |
else if (intra) |
874 |
} |
acpred_flag = BitstreamGetBit(bs); |
|
else if (mv_x > high) |
|
|
{ |
|
|
mv_x -= range; |
|
|
} |
|
875 |
|
|
876 |
if (mv_y < low) |
cbpy = get_cbpy(bs, intra); |
877 |
{ |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
878 |
mv_y += range; |
|
879 |
|
cbp = (cbpy << 2) | cbpc; |
880 |
|
|
881 |
|
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
882 |
|
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
883 |
|
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
884 |
|
quant += dquant; |
885 |
|
if (quant > 31) { |
886 |
|
quant = 31; |
887 |
|
} else if (quant < 1) { |
888 |
|
quant = 1; |
889 |
} |
} |
890 |
else if (mv_y > high) |
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
|
{ |
|
|
mv_y -= range; |
|
891 |
} |
} |
892 |
|
mb->quant = quant; |
893 |
|
|
894 |
mv->x = mv_x; |
if (dec->interlacing) { |
895 |
mv->y = mv_y; |
if (cbp || intra) { |
896 |
|
mb->field_dct = BitstreamGetBit(bs); |
897 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
898 |
} |
} |
899 |
|
|
900 |
|
if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) { |
901 |
|
mb->field_pred = BitstreamGetBit(bs); |
902 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
903 |
|
|
904 |
// add by MinChen <chenm001@163.com> |
if (mb->field_pred) { |
905 |
// decode an B-frame forward & backward inter macroblock |
mb->field_for_top = BitstreamGetBit(bs); |
906 |
void decoder_bf_mbinter(DECODER * dec, |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
907 |
const MACROBLOCK * pMB, |
mb->field_for_bot = BitstreamGetBit(bs); |
908 |
const uint32_t x_pos, |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
909 |
const uint32_t y_pos, |
} |
910 |
const uint32_t cbp, |
} |
911 |
Bitstream * bs, |
} |
|
const uint32_t quant, |
|
|
const uint8_t ref) |
|
|
{ |
|
912 |
|
|
913 |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
if (mcsel) { |
914 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
915 |
|
continue; |
916 |
|
|
917 |
uint32_t stride = dec->edged_width; |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
|
uint32_t stride2 = stride / 2; |
|
|
uint32_t next_block = stride * 8; |
|
|
uint32_t i; |
|
|
uint32_t iQuant = pMB->quant; |
|
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
int uv_dx, uv_dy; |
|
918 |
|
|
919 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
if (dec->interlacing && mb->field_pred) { |
920 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
921 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
922 |
|
} else { |
923 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
924 |
|
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
925 |
|
} |
926 |
|
} else if (mb->mode == MODE_INTER4V ) { |
927 |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
928 |
|
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
929 |
|
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
930 |
|
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
931 |
|
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
932 |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
933 |
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
934 |
|
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
935 |
|
intra_dc_threshold, bound, reduced_resolution); |
936 |
|
continue; |
937 |
|
} |
938 |
|
|
939 |
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
940 |
{ |
rounding, reduced_resolution, 0); |
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
941 |
|
|
942 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
943 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
mb->mode = MODE_NOT_CODED_GMC; |
944 |
} |
mb->quant = quant; |
945 |
else |
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
|
{ |
|
|
int sum; |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
946 |
|
|
947 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
if(dec->out_frm && cp_mb > 0) { |
948 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
949 |
|
cp_mb = 0; |
950 |
} |
} |
951 |
|
st_mb = x+1; |
952 |
|
} else { /* not coded P_VOP macroblock */ |
953 |
|
mb->mode = MODE_NOT_CODED; |
954 |
|
mb->quant = quant; |
955 |
|
|
956 |
start_timer(); |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
957 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
|
|
stop_comp_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) |
|
|
{ |
|
|
if (cbp & (1 << (5-i))) // coded |
|
|
{ |
|
|
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
|
958 |
|
|
959 |
start_timer(); |
decoder_mbinter(dec, mb, x, y, 0, bs, |
960 |
get_inter_block(bs, &block[i*64]); |
rounding, reduced_resolution, 0); |
|
stop_coding_timer(); |
|
961 |
|
|
962 |
start_timer(); |
if(dec->out_frm && cp_mb > 0) { |
963 |
if (dec->quant_type == 0) |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
964 |
{ |
cp_mb = 0; |
965 |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
} |
966 |
|
st_mb = x+1; |
967 |
} |
} |
|
else |
|
|
{ |
|
|
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
|
968 |
} |
} |
|
stop_iquant_timer(); |
|
969 |
|
|
970 |
data[0]=10; |
if(dec->out_frm && cp_mb > 0) |
971 |
start_timer(); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
idct(&data[i*64]); |
|
|
stop_idct_timer(); |
|
972 |
} |
} |
973 |
} |
} |
974 |
|
|
975 |
if (dec->interlacing && pMB->field_dct) |
|
976 |
|
/* decode B-frame motion vector */ |
977 |
|
static void |
978 |
|
get_b_motion_vector(Bitstream * bs, |
979 |
|
VECTOR * mv, |
980 |
|
int fcode, |
981 |
|
const VECTOR pmv, |
982 |
|
const DECODER * const dec, |
983 |
|
const int x, const int y) |
984 |
{ |
{ |
985 |
next_block = stride; |
const int scale_fac = 1 << (fcode - 1); |
986 |
stride *= 2; |
const int high = (32 * scale_fac) - 1; |
987 |
} |
const int low = ((-32) * scale_fac); |
988 |
|
const int range = (64 * scale_fac); |
989 |
|
|
990 |
start_timer(); |
int mv_x = get_mv(bs, fcode); |
991 |
if (cbp & 32) |
int mv_y = get_mv(bs, fcode); |
992 |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
|
993 |
if (cbp & 16) |
mv_x += pmv.x; |
994 |
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
mv_y += pmv.y; |
995 |
if (cbp & 8) |
|
996 |
transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); |
if (mv_x < low) |
997 |
if (cbp & 4) |
mv_x += range; |
998 |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride); |
else if (mv_x > high) |
999 |
if (cbp & 2) |
mv_x -= range; |
|
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
|
|
stop_transfer_timer(); |
|
|
} |
|
1000 |
|
|
1001 |
|
if (mv_y < low) |
1002 |
|
mv_y += range; |
1003 |
|
else if (mv_y > high) |
1004 |
|
mv_y -= range; |
1005 |
|
|
1006 |
// add by MinChen <chenm001@163.com> |
mv->x = mv_x; |
1007 |
// decode an B-frame direct & inter macroblock |
mv->y = mv_y; |
1008 |
void decoder_bf_interpolate_mbinter(DECODER * dec, |
} |
1009 |
IMAGE forward, IMAGE backward, |
|
1010 |
|
/* decode an B-frame direct & interpolate macroblock */ |
1011 |
|
static void |
1012 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
1013 |
|
IMAGE forward, |
1014 |
|
IMAGE backward, |
1015 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
1016 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1017 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1018 |
const uint32_t cbp, |
Bitstream * bs, |
1019 |
Bitstream * bs ) |
const int direct) |
1020 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
1021 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
1022 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
|
uint32_t iQuant = pMB->quant; |
|
1023 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
1024 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
1025 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1026 |
|
const uint32_t cbp = pMB->cbp; |
1027 |
|
|
1028 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1029 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1030 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1031 |
|
|
1032 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if (!direct) { |
|
{ |
|
1033 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1034 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
1035 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
|
|
1036 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1037 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1038 |
|
|
1039 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) { |
1040 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dx /= 2; |
1041 |
|
uv_dy /= 2; |
1042 |
|
b_uv_dx /= 2; |
1043 |
|
b_uv_dy /= 2; |
1044 |
} |
} |
|
else |
|
|
{ |
|
|
int sum; |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
|
1045 |
|
|
1046 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1047 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1048 |
|
|
1049 |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1050 |
b_uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1051 |
|
|
1052 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
} else { |
1053 |
b_uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) ); |
if(dec->quarterpel) { |
1054 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1055 |
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1056 |
|
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1057 |
|
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1058 |
|
} else { |
1059 |
|
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1060 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1061 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1062 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1063 |
} |
} |
1064 |
|
|
1065 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1066 |
start_timer(); |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1067 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos , pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1068 |
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos , pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos, 16*y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.y, forward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
interpolate8x8_switch(dec->cur.u, forward.u, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
|
|
interpolate8x8_switch(dec->cur.v, forward.v, 8*x_pos, 8*y_pos, uv_dx, uv_dy, stride2, 0); |
|
|
|
|
|
|
|
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos , pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos , pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
|
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos, 16*y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
|
interpolate8x8_switch(dec->refn[2].y, backward.y, 16*x_pos + 8, 16*y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
|
interpolate8x8_switch(dec->refn[2].u, backward.u, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
|
|
interpolate8x8_switch(dec->refn[2].v, backward.v, 8*x_pos, 8*y_pos, b_uv_dx, b_uv_dy, stride2, 0); |
|
|
|
|
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos , stride); |
|
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos , stride); |
|
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos, 16*y_pos + 8, stride); |
|
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16*x_pos + 8, 16*y_pos + 8, stride); |
|
|
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8*x_pos, 8*y_pos, stride2); |
|
|
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8*x_pos, 8*y_pos, stride2); |
|
|
|
|
|
stop_comp_timer(); |
|
|
|
|
|
for (i = 0; i < 6; i++) |
|
|
{ |
|
|
if (cbp & (1 << (5-i))) // coded |
|
|
{ |
|
|
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
|
|
|
|
|
start_timer(); |
|
|
get_inter_block(bs, &block[i*64]); |
|
|
stop_coding_timer(); |
|
|
|
|
|
start_timer(); |
|
|
if (dec->quant_type == 0) |
|
|
{ |
|
|
dequant_inter(&data[i*64], &block[i*64], iQuant); |
|
1069 |
} |
} |
|
else |
|
|
{ |
|
|
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
1070 |
|
|
1071 |
start_timer(); |
start_timer(); |
1072 |
idct(&data[i*64]); |
if(dec->quarterpel) { |
1073 |
stop_idct_timer(); |
if(!direct) { |
1074 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1075 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1076 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1077 |
|
} else { |
1078 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1079 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1080 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1081 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1082 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1083 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1084 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1085 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1086 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1087 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1088 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1089 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1090 |
} |
} |
1091 |
|
} else { |
1092 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1093 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1094 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1095 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1096 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1097 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1098 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, |
1099 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1100 |
|
} |
1101 |
|
|
1102 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1103 |
|
uv_dy, stride2, 0); |
1104 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1105 |
|
uv_dy, stride2, 0); |
1106 |
|
|
1107 |
|
|
1108 |
|
if(dec->quarterpel) { |
1109 |
|
if(!direct) { |
1110 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1111 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1112 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1113 |
|
} else { |
1114 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1115 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1116 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1117 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1118 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1119 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1120 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1121 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1122 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1123 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1124 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1125 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1126 |
} |
} |
1127 |
|
} else { |
1128 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1129 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1130 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1131 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1132 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1133 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1134 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1135 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1136 |
|
} |
1137 |
|
|
1138 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1139 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1140 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1141 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1142 |
|
|
1143 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1144 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1145 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1146 |
|
stride, 1, 8); |
1147 |
|
|
1148 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1149 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1150 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1151 |
|
stride, 1, 8); |
1152 |
|
|
1153 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1154 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1155 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1156 |
|
stride, 1, 8); |
1157 |
|
|
1158 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1159 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1160 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1161 |
|
stride, 1, 8); |
1162 |
|
|
1163 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1164 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1165 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1166 |
|
stride2, 1, 8); |
1167 |
|
|
1168 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1169 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1170 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1171 |
|
stride2, 1, 8); |
1172 |
|
|
1173 |
if (dec->interlacing && pMB->field_dct) |
stop_comp_timer(); |
|
{ |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
1174 |
|
|
1175 |
start_timer(); |
if (cbp) |
1176 |
if (cbp & 32) |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
|
transfer_16to8add(pY_Cur, &data[0*64], stride); |
|
|
if (cbp & 16) |
|
|
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
|
|
if (cbp & 8) |
|
|
transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); |
|
|
if (cbp & 4) |
|
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride); |
|
|
if (cbp & 2) |
|
|
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
|
|
if (cbp & 1) |
|
|
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
|
|
stop_transfer_timer(); |
|
1177 |
} |
} |
1178 |
|
|
1179 |
|
/* for decode B-frame dbquant */ |
1180 |
// add by MinChen <chenm001@163.com> |
static __inline int32_t |
1181 |
// for decode B-frame dbquant |
get_dbquant(Bitstream * bs) |
|
int32_t __inline get_dbquant(Bitstream * bs) |
|
1182 |
{ |
{ |
1183 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1184 |
return(0); |
return(0); |
1185 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1186 |
return(-2); |
return(-2); |
1187 |
else |
else /* '11' */ |
1188 |
return(2); // '11' |
return (2); |
1189 |
} |
} |
1190 |
|
|
1191 |
// add by MinChen <chenm001@163.com> |
/* |
1192 |
// for decode B-frame mb_type |
* decode B-frame mb_type |
1193 |
// bit ret_value |
* bit ret_value |
1194 |
// 1 0 |
* 1 0 |
1195 |
// 01 1 |
* 01 1 |
1196 |
// 001 2 |
* 001 2 |
1197 |
// 0001 3 |
* 0001 3 |
1198 |
int32_t __inline get_mbtype(Bitstream * bs) |
*/ |
1199 |
|
static int32_t __inline |
1200 |
|
get_mbtype(Bitstream * bs) |
1201 |
{ |
{ |
1202 |
int32_t mb_type; |
int32_t mb_type; |
|
for(mb_type=0;mb_type<=3;mb_type++){ |
|
|
if (BitstreamGetBit(bs)) |
|
|
break; |
|
|
} |
|
1203 |
|
|
1204 |
if (mb_type<=3) |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1205 |
|
if (BitstreamGetBit(bs)) |
1206 |
return(mb_type); |
return(mb_type); |
1207 |
else |
|
1208 |
return(-1); |
return -1; |
1209 |
} |
} |
1210 |
|
|
1211 |
void decoder_bframe(DECODER * dec, Bitstream * bs, int quant, int fcode_forward, int fcode_backward) |
static void |
1212 |
|
decoder_bframe(DECODER * dec, |
1213 |
|
Bitstream * bs, |
1214 |
|
int quant, |
1215 |
|
int fcode_forward, |
1216 |
|
int fcode_backward) |
1217 |
{ |
{ |
|
|
|
1218 |
uint32_t x, y; |
uint32_t x, y; |
1219 |
VECTOR mv, zeromv; |
VECTOR mv; |
1220 |
|
const VECTOR zeromv = {0,0}; |
1221 |
|
const int32_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1222 |
|
int i; |
1223 |
|
|
1224 |
|
if (!dec->is_edged[0]) { |
1225 |
start_timer(); |
start_timer(); |
1226 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1227 |
//image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, dec->width, dec->height, dec->interlacing); |
dec->width, dec->height, dec->bs_version); |
1228 |
|
dec->is_edged[0] = 1; |
1229 |
stop_edges_timer(); |
stop_edges_timer(); |
1230 |
|
} |
1231 |
|
|
1232 |
|
if (!dec->is_edged[1]) { |
1233 |
|
start_timer(); |
1234 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1235 |
|
dec->width, dec->height, dec->bs_version); |
1236 |
|
dec->is_edged[1] = 1; |
1237 |
|
stop_edges_timer(); |
1238 |
|
} |
1239 |
|
|
1240 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) { |
1241 |
{ |
/* Initialize Pred Motion Vector */ |
1242 |
// Initialize Pred Motion Vector |
dec->p_fmv = dec->p_bmv = zeromv; |
1243 |
dec->p_fmv.x = dec->p_fmv.y = dec->p_bmv.x = dec->p_bmv.y = 0; |
for (x = 0; x < dec->mb_width; x++) { |
|
for (x = 0; x < dec->mb_width; x++) |
|
|
{ |
|
1244 |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
MACROBLOCK * mb = &dec->mbs[y*dec->mb_width + x]; |
1245 |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
MACROBLOCK * last_mb = &dec->last_mbs[y*dec->mb_width + x]; |
1246 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1247 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1248 |
|
|
1249 |
mb->mvs[0].x=mb->mvs[0].y=zeromv.x = zeromv.y = mv.x = mv.y = 0; |
if (check_resync_marker(bs, fcode_max - 1)) { |
1250 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1251 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1252 |
|
x = bound % dec->mb_width; |
1253 |
|
y = bound / dec->mb_width; |
1254 |
|
/* reset predicted macroblocks */ |
1255 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1256 |
|
} |
1257 |
|
|
1258 |
|
mv = |
1259 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1260 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1261 |
|
mb->quant = quant; |
1262 |
|
|
1263 |
|
/* |
1264 |
|
* skip if the co-located P_VOP macroblock is not coded |
1265 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1266 |
|
* automatically skipped |
1267 |
|
*/ |
1268 |
|
|
|
// the last P_VOP is skip macroblock ? |
|
1269 |
if (last_mb->mode == MODE_NOT_CODED){ |
if (last_mb->mode == MODE_NOT_CODED){ |
|
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
|
|
mb->mb_type = MODE_FORWARD; |
|
1270 |
mb->cbp = 0; |
mb->cbp = 0; |
1271 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
mb->mode = MODE_FORWARD; |
1272 |
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
mb->quant = 8; |
|
|
|
|
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
|
1273 |
continue; |
continue; |
1274 |
} |
} |
1275 |
|
|
1276 |
//t=BitstreamShowBits(bs,32); |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
|
|
|
|
if (!BitstreamGetBit(bs)){ // modb=='0' |
|
1277 |
const uint8_t modb2=BitstreamGetBit(bs); |
const uint8_t modb2=BitstreamGetBit(bs); |
1278 |
|
|
1279 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1280 |
|
|
1281 |
if (!modb2){ // modb=='00' |
if (!modb2) /* modb=='00' */ |
1282 |
mb->cbp = BitstreamGetBits(bs,6); |
mb->cbp = BitstreamGetBits(bs,6); |
1283 |
} else { |
else |
1284 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp){ |
|
|
quant += get_dbquant(bs); |
|
1285 |
|
|
1286 |
|
if (mb->mode && mb->cbp) { |
1287 |
|
quant += get_dbquant(bs); |
1288 |
if (quant > 31) |
if (quant > 31) |
|
{ |
|
1289 |
quant = 31; |
quant = 31; |
1290 |
} |
else if (quant < 1) |
|
else if (mb->quant < 1) |
|
|
{ |
|
1291 |
quant = 1; |
quant = 1; |
1292 |
} |
} |
|
} else { |
|
|
quant = 8; |
|
|
} |
|
1293 |
mb->quant = quant; |
mb->quant = quant; |
1294 |
|
|
1295 |
|
if (dec->interlacing) { |
1296 |
|
if (mb->cbp) { |
1297 |
|
mb->field_dct = BitstreamGetBit(bs); |
1298 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1299 |
|
} |
1300 |
|
|
1301 |
|
if (mb->mode) { |
1302 |
|
mb->field_pred = BitstreamGetBit(bs); |
1303 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1304 |
|
|
1305 |
|
if (mb->field_pred) { |
1306 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1307 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1308 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1309 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1310 |
|
} |
1311 |
|
} |
1312 |
|
} |
1313 |
|
|
1314 |
} else { |
} else { |
1315 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1316 |
mb->cbp = 0; |
mb->cbp = 0; |
1317 |
} |
} |
1318 |
|
|
1319 |
mb->mode = MODE_INTER; |
switch (mb->mode) { |
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
|
|
|
|
|
switch(mb->mb_type) |
|
|
{ |
|
1320 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1321 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y); |
1322 |
|
|
1323 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
|
{ // Because this file is a C file not C++ so I use '{' to define var |
|
|
const int64_t TRB=dec->time_pp-dec->time_bp, |
|
|
TRD=dec->time_pp; |
|
|
int i; |
|
1324 |
for(i=0;i<4;i++){ |
for(i=0;i<4;i++){ |
1325 |
mb->mvs[i].x = (int32_t)((TRB * last_mb->mvs[i].x)/TRD+mb->mvs[0].x); |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
1326 |
mb->b_mvs[i].x = (int32_t)((mb->mvs[0].x==0)?((TRB-TRD)*last_mb->mvs[i].x)/TRD:mb->mvs[i].x-last_mb->mvs[i].x); |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1327 |
mb->mvs[i].y = (int32_t)((TRB * last_mb->mvs[i].y)/TRD+mb->mvs[0].y); |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
1328 |
mb->b_mvs[i].y = (int32_t)((mb->mvs[0].y==0)?((TRB-TRD)*last_mb->mvs[i].y)/TRD:mb->mvs[i].y-last_mb->mvs[i].y); |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1329 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
1330 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1331 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
1332 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1333 |
} |
} |
1334 |
//DEBUG("B-frame Direct!\n"); |
|
1335 |
} |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1336 |
mb->mode = MODE_INTER4V; |
mb, x, y, bs, 1); |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
|
1337 |
break; |
break; |
1338 |
|
|
1339 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1340 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1341 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1342 |
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
1343 |
|
get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1344 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], fcode_backward, dec->p_bmv); |
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = mb->b_mvs[0]; |
|
dec->p_bmv.x = mb->b_mvs[1].x = mb->b_mvs[2].x = mb->b_mvs[3].x = mb->b_mvs[0].x; |
|
|
dec->p_bmv.y = mb->b_mvs[1].y = mb->b_mvs[2].y = mb->b_mvs[3].y = mb->b_mvs[0].y; |
|
1345 |
|
|
1346 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], mb, x, y, mb->cbp, bs); |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1347 |
//DEBUG("B-frame Bidir!\n"); |
mb, x, y, bs, 0); |
1348 |
break; |
break; |
1349 |
|
|
1350 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1351 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, dec->p_bmv); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1352 |
dec->p_bmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
dec->p_bmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
1353 |
|
|
1354 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
|
//DEBUG("B-frame Backward!\n"); |
|
1355 |
break; |
break; |
1356 |
|
|
1357 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1358 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, dec->p_fmv); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1359 |
dec->p_fmv.x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
dec->p_fmv.y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
1360 |
|
|
1361 |
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
//DEBUG("B-frame Forward!\n"); |
|
1362 |
break; |
break; |
1363 |
|
|
1364 |
default: |
default: |
1365 |
DEBUG1("Not support B-frame mb_type =",mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1366 |
} |
} |
1367 |
|
} /* End of for */ |
|
} // end of FOR |
|
1368 |
} |
} |
1369 |
} |
} |
1370 |
|
|
1371 |
// swap two MACROBLOCK array |
/* perform post processing if necessary, and output the image */ |
1372 |
void mb_swap(MACROBLOCK **mb1, MACROBLOCK **mb2) |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1373 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1374 |
|
int coding_type, int quant) |
1375 |
{ |
{ |
1376 |
MACROBLOCK *temp=*mb1; |
const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; |
1377 |
*mb1=*mb2; |
|
1378 |
*mb2=temp; |
if (dec->cartoon_mode) |
1379 |
|
frame->general &= ~XVID_FILMEFFECT; |
1380 |
|
|
1381 |
|
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) |
1382 |
|
&& mbs != NULL) /* post process */ |
1383 |
|
{ |
1384 |
|
/* note: image is stored to tmp */ |
1385 |
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1386 |
|
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1387 |
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1388 |
|
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
1389 |
|
img = &dec->tmp; |
1390 |
} |
} |
1391 |
|
|
1392 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
image_output(img, dec->width, dec->height, |
1393 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1394 |
|
frame->output.csp, dec->interlacing); |
1395 |
|
|
1396 |
|
if (stats) { |
1397 |
|
stats->type = coding2type(coding_type); |
1398 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1399 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1400 |
|
stats->data.vop.qscale_stride = dec->mb_width; |
1401 |
|
stats->data.vop.qscale = dec->qscale; |
1402 |
|
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1403 |
|
int i; |
1404 |
|
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1405 |
|
stats->data.vop.qscale[i] = mbs[i].quant; |
1406 |
|
} else |
1407 |
|
stats->data.vop.qscale = NULL; |
1408 |
|
} |
1409 |
|
} |
1410 |
|
|
1411 |
|
|
1412 |
|
int |
1413 |
|
decoder_decode(DECODER * dec, |
1414 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1415 |
{ |
{ |
1416 |
|
|
1417 |
Bitstream bs; |
Bitstream bs; |
1418 |
uint32_t rounding; |
uint32_t rounding; |
1419 |
uint32_t quant; |
uint32_t reduced_resolution; |
1420 |
|
uint32_t quant = 2; |
1421 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1422 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1423 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1424 |
uint32_t vop_type; |
WARPPOINTS gmc_warp; |
1425 |
|
int coding_type; |
1426 |
|
int success, output, seen_something; |
1427 |
|
|
1428 |
|
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1429 |
|
return XVID_ERR_VERSION; |
1430 |
|
|
1431 |
start_global_timer(); |
start_global_timer(); |
1432 |
|
|
1433 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1434 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1435 |
|
dec->frames = 0; |
1436 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1437 |
|
|
1438 |
|
if (frame->length < 0) { /* decoder flush */ |
1439 |
|
int ret; |
1440 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1441 |
|
we have a reference frame, then outout the reference frame */ |
1442 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1443 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1444 |
|
dec->frames = 0; |
1445 |
|
ret = 0; |
1446 |
|
} else { |
1447 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1448 |
|
ret = XVID_ERR_END; |
1449 |
|
} |
1450 |
|
|
1451 |
// add by chenm001 <chenm001@163.com> |
emms(); |
1452 |
// for support B-frame to reference last 2 frame |
stop_global_timer(); |
1453 |
dec->frames ++; |
return ret; |
1454 |
vop_type=BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, &fcode_backward, &intra_dc_threshold); |
} |
1455 |
|
|
1456 |
|
BitstreamInit(&bs, frame->bitstream, frame->length); |
1457 |
|
|
1458 |
dec->p_bmv.x=dec->p_bmv.y=dec->p_fmv.y=dec->p_fmv.y=0; // init pred vector to 0 |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1459 |
switch (vop_type) |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1460 |
{ |
{ |
1461 |
case P_VOP : |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1462 |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, intra_dc_threshold); |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1463 |
DEBUG1("P_VOP Time=",dec->time); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1464 |
break; |
emms(); |
1465 |
|
return 1; /* one byte consumed */ |
1466 |
|
} |
1467 |
|
|
1468 |
case I_VOP : |
success = 0; |
1469 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
output = 0; |
1470 |
DEBUG1("I_VOP Time=",dec->time); |
seen_something = 0; |
|
break; |
|
1471 |
|
|
1472 |
case B_VOP : |
repeat: |
1473 |
if (dec->time_pp > dec->time_bp){ |
|
1474 |
DEBUG1("B_VOP Time=",dec->time); |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1475 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1476 |
} else { |
|
1477 |
DEBUG("broken B-frame!"); |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1478 |
|
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1479 |
|
|
1480 |
|
if (coding_type == -1) { /* nothing */ |
1481 |
|
if (success) goto done; |
1482 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1483 |
|
emms(); |
1484 |
|
return BitstreamPos(&bs)/8; |
1485 |
} |
} |
|
break; |
|
1486 |
|
|
1487 |
case N_VOP : // vop not coded |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
|
break; |
|
1488 |
|
|
1489 |
default : |
if (coding_type == -3) |
1490 |
return XVID_ERR_FAIL; |
decoder_resize(dec); |
1491 |
|
|
1492 |
|
if (stats) { |
1493 |
|
stats->type = XVID_TYPE_VOL; |
1494 |
|
stats->data.vol.general = 0; |
1495 |
|
/*XXX: if (dec->interlacing) |
1496 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1497 |
|
stats->data.vol.width = dec->width; |
1498 |
|
stats->data.vol.height = dec->height; |
1499 |
|
stats->data.vol.par = dec->aspect_ratio; |
1500 |
|
stats->data.vol.par_width = dec->par_width; |
1501 |
|
stats->data.vol.par_height = dec->par_height; |
1502 |
|
emms(); |
1503 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1504 |
|
} |
1505 |
|
goto repeat; |
1506 |
} |
} |
1507 |
|
|
1508 |
frame->length = BitstreamPos(&bs) / 8; |
if(dec->frames == 0 && coding_type != I_VOP) { |
1509 |
|
/* 1st frame is not an i-vop */ |
1510 |
|
goto repeat; |
1511 |
|
} |
1512 |
|
|
1513 |
if (dec->frames >= 1){ |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1514 |
start_timer(); |
|
1515 |
if ((vop_type == I_VOP || vop_type == P_VOP)) |
/* packed_mode: special-N_VOP treament */ |
1516 |
{ |
if (dec->packed_mode && coding_type == N_VOP) { |
1517 |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
if (dec->low_delay_default && dec->frames > 0) { |
1518 |
frame->image, frame->stride, frame->colorspace); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1519 |
} else if (vop_type == B_VOP) { |
output = 1; |
1520 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
} |
1521 |
frame->image, frame->stride, frame->colorspace); |
/* ignore otherwise */ |
1522 |
|
} else if (coding_type != B_VOP) { |
1523 |
|
switch(coding_type) { |
1524 |
|
case I_VOP : |
1525 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1526 |
|
break; |
1527 |
|
case P_VOP : |
1528 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1529 |
|
fcode_forward, intra_dc_threshold, NULL); |
1530 |
|
break; |
1531 |
|
case S_VOP : |
1532 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1533 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1534 |
|
break; |
1535 |
|
case N_VOP : |
1536 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1537 |
|
/* we should not swap(last_mbs,mbs) */ |
1538 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1539 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1540 |
|
break; |
1541 |
|
} |
1542 |
|
|
1543 |
|
if (reduced_resolution) { |
1544 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1545 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1546 |
|
16, 0); |
1547 |
|
} |
1548 |
|
|
1549 |
|
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1550 |
|
if (!(dec->low_delay_default && dec->packed_mode)) { |
1551 |
|
if (dec->low_delay) { |
1552 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1553 |
|
output = 1; |
1554 |
|
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
1555 |
|
/* output the reference frame */ |
1556 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1557 |
|
output = 1; |
1558 |
} |
} |
|
stop_conv_timer(); |
|
1559 |
} |
} |
1560 |
if (vop_type==I_VOP || vop_type==P_VOP){ |
|
1561 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1562 |
|
dec->is_edged[1] = dec->is_edged[0]; |
1563 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1564 |
// swap MACROBLOCK |
dec->is_edged[0] = 0; |
1565 |
if (vop_type==P_VOP) |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1566 |
mb_swap(&dec->mbs, &dec->last_mbs); |
dec->last_reduced_resolution = reduced_resolution; |
1567 |
|
dec->last_coding_type = coding_type; |
1568 |
|
|
1569 |
|
dec->frames++; |
1570 |
|
seen_something = 1; |
1571 |
|
|
1572 |
|
} else { /* B_VOP */ |
1573 |
|
|
1574 |
|
if (dec->low_delay) { |
1575 |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1576 |
|
dec->low_delay = 1; |
1577 |
|
} |
1578 |
|
|
1579 |
|
if (dec->frames < 2) { |
1580 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1581 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1582 |
|
"broken b-frame, mising ref frames"); |
1583 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1584 |
|
} else if (dec->time_pp <= dec->time_bp) { |
1585 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1586 |
|
decoded in vfw. */ |
1587 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1588 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1589 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1590 |
|
} else { |
1591 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1592 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1593 |
} |
} |
1594 |
|
|
1595 |
emms(); |
output = 1; |
1596 |
|
dec->frames++; |
1597 |
|
} |
1598 |
|
|
1599 |
|
#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */ |
1600 |
|
BitstreamByteAlign(&bs); |
1601 |
|
#endif |
1602 |
|
|
1603 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1604 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
1605 |
|
success = 1; |
1606 |
|
goto repeat; |
1607 |
|
} |
1608 |
|
|
1609 |
|
done : |
1610 |
|
|
1611 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1612 |
|
then output the recently decoded frame, or print an error message */ |
1613 |
|
if (dec->low_delay_default && output == 0) { |
1614 |
|
if (dec->packed_mode && seen_something) { |
1615 |
|
/* output the recently decoded frame */ |
1616 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1617 |
|
} else { |
1618 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1619 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1620 |
|
"warning: nothing to output"); |
1621 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1622 |
|
"bframe decoder lag"); |
1623 |
|
|
1624 |
|
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1625 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1626 |
|
} |
1627 |
|
} |
1628 |
|
|
1629 |
|
emms(); |
1630 |
stop_global_timer(); |
stop_global_timer(); |
1631 |
|
|
1632 |
return XVID_ERR_OK; |
return (BitstreamPos(&bs) + 7) / 8; /* number of bytes consumed */ |
1633 |
} |
} |