1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder Module - |
|
* |
|
|
* Copyright(C) 2002 MinChen <chenm001@163.com> |
|
|
* 2002 Peter Ross <pross@xvid.org> |
|
5 |
* |
* |
6 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
7 |
* |
* |
8 |
* XviD is free software; you can redistribute it and/or modify it |
* This program is free software; you can redistribute it and/or modify |
9 |
* under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
10 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
11 |
* (at your option) any later version. |
* (at your option) any later version. |
12 |
* |
* |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
* |
* |
22 |
* Under section 8 of the GNU General Public License, the copyright |
* $Id: decoder.c,v 1.49.2.3 2003-03-26 10:29:51 suxen_drol Exp $ |
|
* holders of XVID explicitly forbid distribution in the following |
|
|
* countries: |
|
|
* |
|
|
* - Japan |
|
|
* - United States of America |
|
|
* |
|
|
* Linking XviD statically or dynamically with other modules is making a |
|
|
* combined work based on XviD. Thus, the terms and conditions of the |
|
|
* GNU General Public License cover the whole combination. |
|
|
* |
|
|
* As a special exception, the copyright holders of XviD give you |
|
|
* permission to link XviD with independent modules that communicate with |
|
|
* XviD solely through the VFW1.1 and DShow interfaces, regardless of the |
|
|
* license terms of these independent modules, and to copy and distribute |
|
|
* the resulting combined work under terms of your choice, provided that |
|
|
* every copy of the combined work is accompanied by a complete copy of |
|
|
* the source code of XviD (the version of XviD used to produce the |
|
|
* combined work), being distributed under the terms of the GNU General |
|
|
* Public License plus this exception. An independent module is a module |
|
|
* which is not derived from or based on XviD. |
|
23 |
* |
* |
24 |
* Note that people who make modified versions of XviD are not obligated |
****************************************************************************/ |
|
* to grant this special exception for their modified versions; it is |
|
|
* their choice whether to do so. The GNU General Public License gives |
|
|
* permission to release a modified version without this exception; this |
|
|
* exception also makes it possible to release a modified version which |
|
|
* carries forward this exception. |
|
|
* |
|
|
* $Id: decoder.c,v 1.46 2003-02-11 21:56:31 edgomez Exp $ |
|
|
* |
|
|
*************************************************************************/ |
|
25 |
|
|
26 |
|
#include <stdio.h> |
27 |
#include <stdlib.h> |
#include <stdlib.h> |
28 |
#include <string.h> |
#include <string.h> |
29 |
|
|
33 |
|
|
34 |
#include "xvid.h" |
#include "xvid.h" |
35 |
#include "portab.h" |
#include "portab.h" |
36 |
|
#include "global.h" |
37 |
|
|
38 |
#include "decoder.h" |
#include "decoder.h" |
39 |
#include "bitstream/bitstream.h" |
#include "bitstream/bitstream.h" |
45 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
46 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
47 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
48 |
|
#include "image/reduced.h" |
49 |
|
#include "image/font.h" |
50 |
|
|
51 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
52 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
53 |
#include "utils/timer.h" |
#include "utils/timer.h" |
54 |
#include "utils/emms.h" |
#include "utils/emms.h" |
55 |
|
#include "motion/motion.h" |
56 |
|
|
57 |
#include "image/image.h" |
#include "image/image.h" |
58 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
59 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
60 |
|
|
61 |
int |
int |
62 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_resize(DECODER * dec) |
63 |
{ |
{ |
64 |
DECODER *dec; |
/* free existing */ |
65 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
66 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
67 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
68 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
69 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
70 |
|
|
71 |
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; |
|
72 |
|
|
73 |
dec->width = param->width; |
if (dec->last_mbs) |
74 |
dec->height = param->height; |
xvid_free(dec->last_mbs); |
75 |
|
if (dec->mbs) |
76 |
|
xvid_free(dec->mbs); |
77 |
|
|
78 |
|
/* realloc */ |
79 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
80 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
81 |
|
|
82 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
83 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
|
dec->low_delay = 0; |
|
84 |
|
|
85 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
86 |
xvid_free(dec); |
xvid_free(dec); |
92 |
xvid_free(dec); |
xvid_free(dec); |
93 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
94 |
} |
} |
95 |
/* add by chenm001 <chenm001@163.com> */ |
|
96 |
/* for support B-frame to reference last 2 frame */ |
/* Support B-frame to reference last 2 frame */ |
97 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
98 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
99 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
100 |
xvid_free(dec); |
xvid_free(dec); |
101 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
102 |
} |
} |
103 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
104 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
105 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
106 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
107 |
|
xvid_free(dec); |
108 |
|
return XVID_ERR_MEMORY; |
109 |
|
} |
110 |
|
|
111 |
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
112 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
113 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
114 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
115 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
116 |
|
xvid_free(dec); |
117 |
|
return XVID_ERR_MEMORY; |
118 |
|
} |
119 |
|
|
120 |
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
121 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
122 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
123 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
124 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
125 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
126 |
xvid_free(dec); |
xvid_free(dec); |
127 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
128 |
} |
} |
134 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
135 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
136 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
137 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
138 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
139 |
xvid_free(dec); |
xvid_free(dec); |
140 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
141 |
} |
} |
|
|
|
142 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
143 |
|
|
144 |
/* add by chenm001 <chenm001@163.com> */ |
/* For skip MB flag */ |
|
/* for skip MB flag */ |
|
145 |
dec->last_mbs = |
dec->last_mbs = |
146 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
147 |
CACHE_LINE); |
CACHE_LINE); |
150 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
151 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
152 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
153 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
154 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
155 |
xvid_free(dec); |
xvid_free(dec); |
156 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
157 |
} |
} |
158 |
|
|
159 |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
160 |
|
|
161 |
|
return 0; |
162 |
|
} |
163 |
|
|
164 |
|
|
165 |
|
int |
166 |
|
decoder_create(xvid_dec_create_t * create) |
167 |
|
{ |
168 |
|
DECODER *dec; |
169 |
|
|
170 |
|
if (XVID_MAJOR(create->version) != 1) /* v1.x.x */ |
171 |
|
return XVID_ERR_VERSION; |
172 |
|
|
173 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
174 |
|
if (dec == NULL) { |
175 |
|
return XVID_ERR_MEMORY; |
176 |
|
} |
177 |
|
memset(dec, 0, sizeof(DECODER)); |
178 |
|
|
179 |
|
create->handle = dec; |
180 |
|
|
181 |
|
dec->width = create->width; |
182 |
|
dec->height = create->height; |
183 |
|
|
184 |
|
image_null(&dec->cur); |
185 |
|
image_null(&dec->refn[0]); |
186 |
|
image_null(&dec->refn[1]); |
187 |
|
image_null(&dec->tmp); |
188 |
|
image_null(&dec->qtmp); |
189 |
|
|
190 |
|
/* image based GMC */ |
191 |
|
image_null(&dec->gmc); |
192 |
|
|
193 |
|
|
194 |
|
dec->mbs = NULL; |
195 |
|
dec->last_mbs = NULL; |
196 |
|
|
197 |
init_timer(); |
init_timer(); |
198 |
|
|
199 |
/* add by chenm001 <chenm001@163.com> */ |
/* For B-frame support (used to save reference frame's time */ |
200 |
/* for support B-frame to save reference frame's time */ |
dec->frames = 0; |
|
dec->frames = -1; |
|
201 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
202 |
|
dec->low_delay = 0; |
203 |
|
dec->packed_mode = 0; |
204 |
|
|
205 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
206 |
|
|
207 |
return XVID_ERR_OK; |
if (dec->fixed_dimensions) |
208 |
|
return decoder_resize(dec); |
209 |
|
else |
210 |
|
return 0; |
211 |
} |
} |
212 |
|
|
213 |
|
|
216 |
{ |
{ |
217 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
218 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
219 |
|
|
220 |
|
/* image based GMC */ |
221 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
222 |
|
|
223 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
224 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
225 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
226 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
227 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
228 |
xvid_free(dec); |
xvid_free(dec); |
229 |
|
|
230 |
write_timer(); |
write_timer(); |
231 |
return XVID_ERR_OK; |
return 0; |
232 |
} |
} |
233 |
|
|
234 |
|
|
241 |
|
|
242 |
|
|
243 |
/* decode an intra macroblock */ |
/* decode an intra macroblock */ |
|
|
|
244 |
void |
void |
245 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
246 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
251 |
Bitstream * bs, |
Bitstream * bs, |
252 |
const uint32_t quant, |
const uint32_t quant, |
253 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
254 |
const unsigned int bound) |
const unsigned int bound, |
255 |
|
const int reduced_resolution) |
256 |
{ |
{ |
257 |
|
|
258 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
265 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
266 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
267 |
|
|
268 |
|
if (reduced_resolution) { |
269 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
270 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
271 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
272 |
|
}else{ |
273 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
274 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
275 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
276 |
|
} |
277 |
|
|
278 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
279 |
|
|
312 |
start_timer(); |
start_timer(); |
313 |
if (cbp & (1 << (5 - i))) /* coded */ |
if (cbp & (1 << (5 - i))) /* coded */ |
314 |
{ |
{ |
315 |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
int direction = dec->alternate_vertical_scan ? |
316 |
start_coeff); |
2 : pMB->acpred_directions[i]; |
317 |
|
|
318 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
319 |
} |
} |
320 |
stop_coding_timer(); |
stop_coding_timer(); |
321 |
|
|
334 |
start_timer(); |
start_timer(); |
335 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
336 |
stop_idct_timer(); |
stop_idct_timer(); |
337 |
|
|
338 |
} |
} |
339 |
|
|
340 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
343 |
} |
} |
344 |
|
|
345 |
start_timer(); |
start_timer(); |
346 |
|
|
347 |
|
if (reduced_resolution) |
348 |
|
{ |
349 |
|
next_block*=2; |
350 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
351 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
352 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
353 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
354 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
355 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
356 |
|
}else{ |
357 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
358 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
359 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
360 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
361 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
362 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
363 |
|
} |
364 |
stop_transfer_timer(); |
stop_transfer_timer(); |
365 |
} |
} |
366 |
|
|
367 |
|
|
368 |
|
|
369 |
|
|
|
|
|
|
#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 }; |
|
|
|
|
|
|
|
370 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
|
|
|
371 |
void |
void |
372 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
373 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
374 |
const uint32_t x_pos, |
const uint32_t x_pos, |
375 |
const uint32_t y_pos, |
const uint32_t y_pos, |
376 |
const uint32_t acpred_flag, |
const uint32_t fcode, |
377 |
const uint32_t cbp, |
const uint32_t cbp, |
378 |
Bitstream * bs, |
Bitstream * bs, |
379 |
const uint32_t quant, |
const uint32_t quant, |
380 |
const uint32_t rounding) |
const uint32_t rounding, |
381 |
|
const int reduced_resolution) |
382 |
{ |
{ |
383 |
|
|
384 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
386 |
|
|
387 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
388 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
389 |
uint32_t next_block = stride * 8; |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
390 |
uint32_t i; |
uint32_t i; |
391 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
392 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
393 |
|
|
394 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
395 |
|
VECTOR mv[4]; /* local copy of mvs */ |
396 |
|
|
397 |
|
if (reduced_resolution) { |
398 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
399 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
400 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
401 |
|
for (i = 0; i < 4; i++) { |
402 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
403 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
404 |
|
} |
405 |
|
} else { |
406 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
407 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
408 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
409 |
|
for (i = 0; i < 4; i++) |
410 |
|
mv[i] = pMB->mvs[i]; |
411 |
|
} |
412 |
|
|
413 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
414 |
|
|
415 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
uv_dx = mv[0].x / (1 + dec->quarterpel); |
416 |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
} else { |
|
|
int sum; |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
417 |
|
|
418 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
419 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
420 |
|
|
421 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
start_timer(); |
422 |
|
if (reduced_resolution) |
423 |
|
{ |
424 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
425 |
|
mv[0].x, mv[0].y, stride, rounding); |
426 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
427 |
|
uv_dx, uv_dy, stride2, rounding); |
428 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
429 |
|
uv_dx, uv_dy, stride2, rounding); |
430 |
|
|
431 |
|
} |
432 |
|
else |
433 |
|
{ |
434 |
|
if(dec->quarterpel) { |
435 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
436 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
437 |
|
mv[0].x, mv[0].y, stride, rounding); |
438 |
|
} |
439 |
|
else { |
440 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
441 |
|
mv[0].x, mv[0].y, stride, rounding); |
442 |
|
} |
443 |
|
|
444 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
445 |
|
uv_dx, uv_dy, stride2, rounding); |
446 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
447 |
|
uv_dx, uv_dy, stride2, rounding); |
448 |
} |
} |
449 |
|
stop_comp_timer(); |
450 |
|
|
451 |
|
} else { /* MODE_INTER4V */ |
452 |
|
int sum; |
453 |
|
|
454 |
|
if(dec->quarterpel) |
455 |
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
456 |
|
else |
457 |
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
458 |
|
|
459 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
460 |
|
|
461 |
|
if(dec->quarterpel) |
462 |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
463 |
|
else |
464 |
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
465 |
|
|
466 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
467 |
|
|
468 |
start_timer(); |
start_timer(); |
469 |
|
if (reduced_resolution) |
470 |
|
{ |
471 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
472 |
|
mv[0].x, mv[0].y, stride, rounding); |
473 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
474 |
|
mv[1].x, mv[1].y, stride, rounding); |
475 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
476 |
|
mv[2].x, mv[2].y, stride, rounding); |
477 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
478 |
|
mv[3].x, mv[3].y, stride, rounding); |
479 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
480 |
|
uv_dx, uv_dy, stride2, rounding); |
481 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
482 |
|
uv_dx, uv_dy, stride2, rounding); |
483 |
|
|
484 |
|
/* set_block(pY_Cur, stride, 32, 32, 127); */ |
485 |
|
} |
486 |
|
else |
487 |
|
{ |
488 |
|
if(dec->quarterpel) { |
489 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
490 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
491 |
|
mv[0].x, mv[0].y, stride, rounding); |
492 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
493 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
494 |
|
mv[1].x, mv[1].y, stride, rounding); |
495 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
496 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
497 |
|
mv[2].x, mv[2].y, stride, rounding); |
498 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
499 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
500 |
|
mv[3].x, mv[3].y, stride, rounding); |
501 |
|
} |
502 |
|
else { |
503 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
504 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
505 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
506 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
507 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
508 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
509 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
510 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
511 |
|
} |
512 |
|
|
513 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
514 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
515 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
516 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
517 |
|
} |
518 |
stop_comp_timer(); |
stop_comp_timer(); |
519 |
|
} |
520 |
|
|
521 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
522 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
523 |
|
|
524 |
if (cbp & (1 << (5 - i))) /* coded */ |
if (cbp & (1 << (5 - i))) /* coded */ |
525 |
{ |
{ |
526 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
527 |
|
|
528 |
start_timer(); |
start_timer(); |
529 |
get_inter_block(bs, &block[i * 64]); |
get_inter_block(bs, &block[i * 64], direction); |
530 |
stop_coding_timer(); |
stop_coding_timer(); |
531 |
|
|
532 |
start_timer(); |
start_timer(); |
549 |
} |
} |
550 |
|
|
551 |
start_timer(); |
start_timer(); |
552 |
|
if (reduced_resolution) |
553 |
|
{ |
554 |
|
if (cbp & 32) |
555 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
556 |
|
if (cbp & 16) |
557 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
558 |
|
if (cbp & 8) |
559 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
560 |
|
if (cbp & 4) |
561 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
562 |
|
if (cbp & 2) |
563 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
564 |
|
if (cbp & 1) |
565 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
566 |
|
} |
567 |
|
else |
568 |
|
{ |
569 |
|
if (cbp & 32) |
570 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
571 |
|
if (cbp & 16) |
572 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
573 |
|
if (cbp & 8) |
574 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
575 |
|
if (cbp & 4) |
576 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
577 |
|
if (cbp & 2) |
578 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
579 |
|
if (cbp & 1) |
580 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
581 |
|
} |
582 |
|
stop_transfer_timer(); |
583 |
|
} |
584 |
|
|
585 |
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
586 |
|
{ |
587 |
|
int length = 1 << (fcode+4); |
588 |
|
|
589 |
|
/* if (quarterpel) value *= 2; */ |
590 |
|
|
591 |
|
if (value < -length) |
592 |
|
return -length; |
593 |
|
else if (value >= length) |
594 |
|
return length-1; |
595 |
|
else return value; |
596 |
|
} |
597 |
|
|
598 |
|
|
599 |
|
static void |
600 |
|
decoder_mbgmc(DECODER * dec, |
601 |
|
MACROBLOCK * const pMB, |
602 |
|
const uint32_t x_pos, |
603 |
|
const uint32_t y_pos, |
604 |
|
const uint32_t fcode, |
605 |
|
const uint32_t cbp, |
606 |
|
Bitstream * bs, |
607 |
|
const uint32_t quant, |
608 |
|
const uint32_t rounding, |
609 |
|
const int reduced_resolution) /* no reduced res support */ |
610 |
|
{ |
611 |
|
|
612 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
613 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
614 |
|
|
615 |
|
const uint32_t stride = dec->edged_width; |
616 |
|
const uint32_t stride2 = stride / 2; |
617 |
|
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
618 |
|
uint32_t i; |
619 |
|
const uint32_t iQuant = pMB->quant; |
620 |
|
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
621 |
|
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
622 |
|
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
623 |
|
|
624 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
625 |
|
|
626 |
|
start_timer(); |
627 |
|
|
628 |
|
/* this is where the calculations are done */ |
629 |
|
|
630 |
|
{ |
631 |
|
pMB->amv = generate_GMCimageMB(&dec->gmc_data, &dec->refn[0], x_pos, y_pos, |
632 |
|
stride, stride2, dec->quarterpel, rounding, &dec->cur); |
633 |
|
|
634 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
635 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
636 |
|
} |
637 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
638 |
|
|
639 |
|
/* |
640 |
|
transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
641 |
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
642 |
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
643 |
|
*/ |
644 |
|
|
645 |
|
|
646 |
|
stop_transfer_timer(); |
647 |
|
|
648 |
|
if (!cbp) return; |
649 |
|
|
650 |
|
for (i = 0; i < 6; i++) { |
651 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
652 |
|
|
653 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
654 |
|
{ |
655 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
656 |
|
|
657 |
|
start_timer(); |
658 |
|
get_inter_block(bs, &block[i * 64], direction); |
659 |
|
stop_coding_timer(); |
660 |
|
|
661 |
|
start_timer(); |
662 |
|
if (dec->quant_type == 0) { |
663 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
664 |
|
} else { |
665 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
666 |
|
} |
667 |
|
stop_iquant_timer(); |
668 |
|
|
669 |
|
start_timer(); |
670 |
|
idct(&data[i * 64]); |
671 |
|
stop_idct_timer(); |
672 |
|
} |
673 |
|
} |
674 |
|
|
675 |
|
/* interlace + GMC is this possible ??? */ |
676 |
|
/* |
677 |
|
if (dec->interlacing && pMB->field_dct) { |
678 |
|
next_block = stride; |
679 |
|
stride *= 2; |
680 |
|
} |
681 |
|
*/ |
682 |
|
start_timer(); |
683 |
if (cbp & 32) |
if (cbp & 32) |
684 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
685 |
if (cbp & 16) |
if (cbp & 16) |
699 |
void |
void |
700 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
701 |
Bitstream * bs, |
Bitstream * bs, |
702 |
|
int reduced_resolution, |
703 |
int quant, |
int quant, |
704 |
int intra_dc_threshold) |
int intra_dc_threshold) |
705 |
{ |
{ |
706 |
uint32_t bound; |
uint32_t bound; |
707 |
uint32_t x, y; |
uint32_t x, y; |
708 |
|
uint32_t mb_width = dec->mb_width; |
709 |
|
uint32_t mb_height = dec->mb_height; |
710 |
|
|
711 |
|
if (reduced_resolution) |
712 |
|
{ |
713 |
|
mb_width = (dec->width + 31) / 32; |
714 |
|
mb_height = (dec->height + 31) / 32; |
715 |
|
} |
716 |
|
|
717 |
bound = 0; |
bound = 0; |
718 |
|
|
719 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
720 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
721 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
722 |
uint32_t mcbpc; |
uint32_t mcbpc; |
723 |
uint32_t cbpc; |
uint32_t cbpc; |
730 |
|
|
731 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
732 |
{ |
{ |
733 |
bound = read_video_packet_header(bs, 0, &quant); |
bound = read_video_packet_header(bs, dec, 0, |
734 |
x = bound % dec->mb_width; |
&quant, NULL, NULL, &intra_dc_threshold); |
735 |
y = bound / dec->mb_width; |
x = bound % mb_width; |
736 |
|
y = bound / mb_width; |
737 |
} |
} |
738 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
739 |
|
|
764 |
|
|
765 |
if (dec->interlacing) { |
if (dec->interlacing) { |
766 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
767 |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
768 |
} |
} |
769 |
|
|
770 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
771 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
772 |
|
|
773 |
} |
} |
774 |
if(dec->out_frm) |
if(dec->out_frm) |
775 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
|
|
|
776 |
} |
} |
777 |
|
|
778 |
} |
} |
784 |
int x, |
int x, |
785 |
int y, |
int y, |
786 |
int k, |
int k, |
787 |
VECTOR * mv, |
VECTOR * ret_mv, |
788 |
int fcode, |
int fcode, |
789 |
const int bound) |
const int bound) |
790 |
{ |
{ |
795 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
796 |
|
|
797 |
VECTOR pmv; |
VECTOR pmv; |
798 |
int mv_x, mv_y; |
VECTOR mv; |
799 |
|
|
800 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
801 |
|
|
802 |
mv_x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
803 |
mv_y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
804 |
|
|
805 |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i) result (%i,%i)", mv.x, mv.y, pmv.x, pmv.y, mv.x+pmv.x, mv.y+pmv.y); |
806 |
|
|
807 |
mv_x += pmv.x; |
mv.x += pmv.x; |
808 |
mv_y += pmv.y; |
mv.y += pmv.y; |
809 |
|
|
810 |
if (mv_x < low) { |
if (mv.x < low) { |
811 |
mv_x += range; |
mv.x += range; |
812 |
} else if (mv_x > high) { |
} else if (mv.x > high) { |
813 |
mv_x -= range; |
mv.x -= range; |
814 |
} |
} |
815 |
|
|
816 |
if (mv_y < low) { |
if (mv.y < low) { |
817 |
mv_y += range; |
mv.y += range; |
818 |
} else if (mv_y > high) { |
} else if (mv.y > high) { |
819 |
mv_y -= range; |
mv.y -= range; |
820 |
} |
} |
821 |
|
|
822 |
mv->x = mv_x; |
ret_mv->x = mv.x; |
823 |
mv->y = mv_y; |
ret_mv->y = mv.y; |
|
|
|
824 |
} |
} |
825 |
|
|
826 |
|
|
827 |
|
|
828 |
|
|
829 |
|
|
830 |
|
/* for P_VOP set gmc_warp to NULL */ |
831 |
void |
void |
832 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
833 |
Bitstream * bs, |
Bitstream * bs, |
834 |
int rounding, |
int rounding, |
835 |
|
int reduced_resolution, |
836 |
int quant, |
int quant, |
837 |
int fcode, |
int fcode, |
838 |
int intra_dc_threshold) |
int intra_dc_threshold, |
839 |
|
const WARPPOINTS *const gmc_warp) |
840 |
{ |
{ |
841 |
|
|
842 |
uint32_t x, y; |
uint32_t x, y; |
843 |
uint32_t bound; |
uint32_t bound; |
844 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
845 |
|
uint32_t mb_width = dec->mb_width; |
846 |
|
uint32_t mb_height = dec->mb_height; |
847 |
|
|
848 |
|
if (reduced_resolution) |
849 |
|
{ |
850 |
|
mb_width = (dec->width + 31) / 32; |
851 |
|
mb_height = (dec->height + 31) / 32; |
852 |
|
} |
853 |
|
|
854 |
start_timer(); |
start_timer(); |
855 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
856 |
dec->width, dec->height); |
dec->width, dec->height); |
857 |
stop_edges_timer(); |
stop_edges_timer(); |
858 |
|
|
859 |
|
if (gmc_warp) |
860 |
|
{ |
861 |
|
|
862 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
863 |
|
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
864 |
|
{ |
865 |
|
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
866 |
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
867 |
|
dec->sprite_warping_points); |
868 |
|
} |
869 |
|
|
870 |
|
generate_GMCparameters( dec->sprite_warping_points, |
871 |
|
(2 << dec->sprite_warping_accuracy), gmc_warp, |
872 |
|
dec->width, dec->height, &dec->gmc_data); |
873 |
|
|
874 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
875 |
|
/* |
876 |
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
877 |
|
mb_width, mb_height, |
878 |
|
dec->edged_width, dec->edged_width/2, |
879 |
|
fcode, dec->quarterpel, 0, |
880 |
|
rounding, dec->mbs, &dec->gmc); |
881 |
|
*/ |
882 |
|
} |
883 |
|
|
884 |
bound = 0; |
bound = 0; |
885 |
|
|
886 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
887 |
cp_mb = st_mb = 0; |
cp_mb = st_mb = 0; |
888 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
889 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
890 |
|
|
891 |
/* skip stuffing */ |
/* skip stuffing */ |
894 |
|
|
895 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) |
896 |
{ |
{ |
897 |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
bound = read_video_packet_header(bs, dec, fcode - 1, |
898 |
x = bound % dec->mb_width; |
&quant, &fcode, NULL, &intra_dc_threshold); |
899 |
y = bound / dec->mb_width; |
x = bound % mb_width; |
900 |
|
y = bound / mb_width; |
901 |
} |
} |
902 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
903 |
|
|
904 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
905 |
|
|
906 |
/*if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) not_coded */ |
/* if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) */ /* not_coded */ |
907 |
if (!(BitstreamGetBit(bs))) /* not_coded */ |
if (!(BitstreamGetBit(bs))) /* block _is_ coded */ |
908 |
{ |
{ |
909 |
uint32_t mcbpc; |
uint32_t mcbpc; |
910 |
uint32_t cbpc; |
uint32_t cbpc; |
912 |
uint32_t cbpy; |
uint32_t cbpy; |
913 |
uint32_t cbp; |
uint32_t cbp; |
914 |
uint32_t intra; |
uint32_t intra; |
915 |
|
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
916 |
|
|
917 |
cp_mb++; |
cp_mb++; |
918 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
929 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
930 |
} |
} |
931 |
|
|
932 |
|
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
933 |
|
{ |
934 |
|
mcsel = BitstreamGetBit(bs); |
935 |
|
} |
936 |
|
|
937 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
938 |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
939 |
|
|
940 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
941 |
|
|
955 |
if (dec->interlacing) { |
if (dec->interlacing) { |
956 |
if (cbp || intra) { |
if (cbp || intra) { |
957 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
958 |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
959 |
} |
} |
960 |
|
|
961 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
962 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
963 |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
964 |
|
|
965 |
if (mb->field_pred) { |
if (mb->field_pred) { |
966 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
967 |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
968 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
969 |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
970 |
} |
} |
971 |
} |
} |
972 |
} |
} |
973 |
|
|
974 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mcsel) { |
975 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
976 |
|
rounding, reduced_resolution); |
977 |
|
continue; |
978 |
|
|
979 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
980 |
|
|
981 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
982 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
983 |
fcode, bound); |
fcode, bound); |
986 |
} else { |
} else { |
987 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
988 |
fcode, bound); |
fcode, bound); |
989 |
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[0].x; |
|
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
990 |
} |
} |
991 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
992 |
|
|
1001 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
1002 |
0; |
0; |
1003 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
1004 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
1005 |
continue; |
continue; |
1006 |
} |
} |
1007 |
|
|
1008 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
1009 |
rounding); |
rounding, reduced_resolution); |
|
} else /* not coded */ |
|
|
{ |
|
|
DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", 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; |
|
1010 |
|
|
1011 |
/* copy macroblock directly from ref to cur */ |
} |
1012 |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
1013 |
|
{ |
1014 |
|
mb->mode = MODE_NOT_CODED_GMC; |
1015 |
|
|
1016 |
start_timer(); |
start_timer(); |
1017 |
|
|
1018 |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
1019 |
(16 * x), |
rounding, reduced_resolution); |
|
dec->refn[0].y + (16 * y) * dec->edged_width + |
|
|
(16 * x), dec->edged_width); |
|
|
|
|
|
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); |
|
1020 |
|
|
|
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); |
|
1021 |
stop_transfer_timer(); |
stop_transfer_timer(); |
1022 |
|
|
1023 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1024 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1025 |
cp_mb = 0; |
cp_mb = 0; |
1026 |
} |
} |
1027 |
st_mb = x+1; |
st_mb = x+1; |
1028 |
} |
} |
1029 |
} |
else /* not coded P_VOP macroblock */ |
|
if(dec->out_frm && cp_mb > 0) |
|
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
|
|
} |
|
|
} |
|
|
|
|
|
/* swap two MACROBLOCK array */ |
|
|
void |
|
|
mb_swap(MACROBLOCK ** mb1, |
|
|
MACROBLOCK ** mb2) |
|
1030 |
{ |
{ |
1031 |
MACROBLOCK *temp = *mb1; |
mb->mode = MODE_NOT_CODED; |
1032 |
|
|
1033 |
*mb1 = *mb2; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
1034 |
*mb2 = temp; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
1035 |
|
/* copy macroblock directly from ref to cur */ |
1036 |
|
|
1037 |
|
start_timer(); |
1038 |
|
|
1039 |
|
if (reduced_resolution) |
1040 |
|
{ |
1041 |
|
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
1042 |
|
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
1043 |
|
dec->edged_width); |
1044 |
|
|
1045 |
|
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
1046 |
|
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
1047 |
|
dec->edged_width/2); |
1048 |
|
|
1049 |
|
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
1050 |
|
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
1051 |
|
dec->edged_width/2); |
1052 |
|
} |
1053 |
|
else |
1054 |
|
{ |
1055 |
|
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
1056 |
|
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
1057 |
|
dec->edged_width); |
1058 |
|
|
1059 |
|
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
1060 |
|
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
1061 |
|
dec->edged_width/2); |
1062 |
|
|
1063 |
|
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
1064 |
|
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
1065 |
|
dec->edged_width/2); |
1066 |
|
} |
1067 |
|
|
1068 |
|
stop_transfer_timer(); |
1069 |
|
|
1070 |
|
if(dec->out_frm && cp_mb > 0) { |
1071 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1072 |
|
cp_mb = 0; |
1073 |
|
} |
1074 |
|
st_mb = x+1; |
1075 |
|
} |
1076 |
|
} |
1077 |
|
if(dec->out_frm && cp_mb > 0) |
1078 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1079 |
|
} |
1080 |
|
} |
1081 |
|
|
1082 |
|
|
1083 |
|
/* decode B-frame motion vector */ |
1084 |
|
void |
1085 |
|
get_b_motion_vector(DECODER * dec, |
1086 |
|
Bitstream * bs, |
1087 |
|
int x, |
1088 |
|
int y, |
1089 |
|
VECTOR * mv, |
1090 |
|
int fcode, |
1091 |
|
const VECTOR pmv) |
1092 |
|
{ |
1093 |
|
int scale_fac = 1 << (fcode - 1); |
1094 |
|
int high = (32 * scale_fac) - 1; |
1095 |
|
int low = ((-32) * scale_fac); |
1096 |
|
int range = (64 * scale_fac); |
1097 |
|
|
1098 |
|
int mv_x, mv_y; |
1099 |
|
int pmv_x, pmv_y; |
1100 |
|
|
1101 |
|
pmv_x = pmv.x; |
1102 |
|
pmv_y = pmv.y; |
1103 |
|
|
1104 |
|
mv_x = get_mv(bs, fcode); |
1105 |
|
mv_y = get_mv(bs, fcode); |
1106 |
|
|
1107 |
|
mv_x += pmv_x; |
1108 |
|
mv_y += pmv_y; |
1109 |
|
|
1110 |
|
if (mv_x < low) { |
1111 |
|
mv_x += range; |
1112 |
|
} else if (mv_x > high) { |
1113 |
|
mv_x -= range; |
1114 |
|
} |
1115 |
|
|
1116 |
|
if (mv_y < low) { |
1117 |
|
mv_y += range; |
1118 |
|
} else if (mv_y > high) { |
1119 |
|
mv_y -= range; |
1120 |
|
} |
1121 |
|
|
1122 |
|
mv->x = mv_x; |
1123 |
|
mv->y = mv_y; |
1124 |
|
} |
1125 |
|
|
1126 |
|
|
1127 |
|
/* decode an B-frame forward & backward inter macroblock */ |
1128 |
|
void |
1129 |
|
decoder_bf_mbinter(DECODER * dec, |
1130 |
|
const MACROBLOCK * pMB, |
1131 |
|
const uint32_t x_pos, |
1132 |
|
const uint32_t y_pos, |
1133 |
|
const uint32_t cbp, |
1134 |
|
Bitstream * bs, |
1135 |
|
const uint32_t quant, |
1136 |
|
const uint8_t ref) |
1137 |
|
{ |
1138 |
|
|
1139 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1140 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
1141 |
|
|
1142 |
|
uint32_t stride = dec->edged_width; |
1143 |
|
uint32_t stride2 = stride / 2; |
1144 |
|
uint32_t next_block = stride * 8; |
1145 |
|
uint32_t i; |
1146 |
|
uint32_t iQuant = pMB->quant; |
1147 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1148 |
|
int uv_dx, uv_dy; |
1149 |
|
|
1150 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1151 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1152 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1153 |
|
|
1154 |
|
|
1155 |
|
if (!(pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1156 |
|
uv_dx = pMB->mvs[0].x; |
1157 |
|
uv_dy = pMB->mvs[0].y; |
1158 |
|
|
1159 |
|
if (dec->quarterpel) |
1160 |
|
{ |
1161 |
|
uv_dx /= 2; |
1162 |
|
uv_dy /= 2; |
1163 |
|
} |
1164 |
|
|
1165 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1166 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1167 |
|
} else { |
1168 |
|
int sum; |
1169 |
|
|
1170 |
|
if(dec->quarterpel) |
1171 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1172 |
|
else |
1173 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1174 |
|
|
1175 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1176 |
|
|
1177 |
|
if(dec->quarterpel) |
1178 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1179 |
|
else |
1180 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1181 |
|
|
1182 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1183 |
|
} |
1184 |
|
|
1185 |
|
start_timer(); |
1186 |
|
if(dec->quarterpel) { |
1187 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
1188 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1189 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1190 |
|
} |
1191 |
|
else { |
1192 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
1193 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1194 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
1195 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1196 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
1197 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1198 |
|
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
1199 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1200 |
|
} |
1201 |
|
|
1202 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
1203 |
|
uv_dx, uv_dy, stride2, 0); |
1204 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
1205 |
|
uv_dx, uv_dy, stride2, 0); |
1206 |
|
stop_comp_timer(); |
1207 |
|
|
1208 |
|
for (i = 0; i < 6; i++) { |
1209 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1210 |
|
|
1211 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1212 |
|
{ |
1213 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1214 |
|
|
1215 |
|
start_timer(); |
1216 |
|
get_inter_block(bs, &block[i * 64], direction); |
1217 |
|
stop_coding_timer(); |
1218 |
|
|
1219 |
|
start_timer(); |
1220 |
|
if (dec->quant_type == 0) { |
1221 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1222 |
|
} else { |
1223 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1224 |
|
} |
1225 |
|
stop_iquant_timer(); |
1226 |
|
|
1227 |
|
start_timer(); |
1228 |
|
idct(&data[i * 64]); |
1229 |
|
stop_idct_timer(); |
1230 |
|
} |
1231 |
|
} |
1232 |
|
|
1233 |
|
if (dec->interlacing && pMB->field_dct) { |
1234 |
|
next_block = stride; |
1235 |
|
stride *= 2; |
1236 |
|
} |
1237 |
|
|
1238 |
|
start_timer(); |
1239 |
|
if (cbp & 32) |
1240 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1241 |
|
if (cbp & 16) |
1242 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1243 |
|
if (cbp & 8) |
1244 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1245 |
|
if (cbp & 4) |
1246 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1247 |
|
if (cbp & 2) |
1248 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1249 |
|
if (cbp & 1) |
1250 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1251 |
|
stop_transfer_timer(); |
1252 |
} |
} |
1253 |
|
|
1254 |
|
/* decode an B-frame direct & inter macroblock */ |
1255 |
|
void |
1256 |
|
decoder_bf_interpolate_mbinter(DECODER * dec, |
1257 |
|
IMAGE forward, |
1258 |
|
IMAGE backward, |
1259 |
|
const MACROBLOCK * pMB, |
1260 |
|
const uint32_t x_pos, |
1261 |
|
const uint32_t y_pos, |
1262 |
|
Bitstream * bs) |
1263 |
|
{ |
1264 |
|
|
1265 |
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
1266 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
1267 |
|
|
1268 |
|
uint32_t stride = dec->edged_width; |
1269 |
|
uint32_t stride2 = stride / 2; |
1270 |
|
uint32_t next_block = stride * 8; |
1271 |
|
uint32_t iQuant = pMB->quant; |
1272 |
|
int uv_dx, uv_dy; |
1273 |
|
int b_uv_dx, b_uv_dy; |
1274 |
|
uint32_t i; |
1275 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1276 |
|
const uint32_t cbp = pMB->cbp; |
1277 |
|
|
1278 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
1279 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1280 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1281 |
|
|
1282 |
|
|
1283 |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
1284 |
|
uv_dx = pMB->mvs[0].x; |
1285 |
|
uv_dy = pMB->mvs[0].y; |
1286 |
|
|
1287 |
|
b_uv_dx = pMB->b_mvs[0].x; |
1288 |
|
b_uv_dy = pMB->b_mvs[0].y; |
1289 |
|
|
1290 |
|
if (dec->quarterpel) |
1291 |
|
{ |
1292 |
|
uv_dx /= 2; |
1293 |
|
uv_dy /= 2; |
1294 |
|
|
1295 |
|
b_uv_dx /= 2; |
1296 |
|
b_uv_dy /= 2; |
1297 |
|
} |
1298 |
|
|
1299 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1300 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1301 |
|
|
1302 |
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1303 |
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1304 |
|
} else { |
1305 |
|
int sum; |
1306 |
|
|
1307 |
|
if(dec->quarterpel) |
1308 |
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1309 |
|
else |
1310 |
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1311 |
|
|
1312 |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1313 |
|
|
1314 |
|
if(dec->quarterpel) |
1315 |
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1316 |
|
else |
1317 |
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1318 |
|
|
1319 |
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1320 |
|
|
1321 |
|
|
1322 |
|
if(dec->quarterpel) |
1323 |
|
sum = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1324 |
|
else |
1325 |
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1326 |
|
|
1327 |
|
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
1328 |
|
|
1329 |
|
if(dec->quarterpel) |
1330 |
|
sum = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1331 |
|
else |
1332 |
|
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1333 |
|
|
1334 |
|
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
1335 |
|
} |
1336 |
|
|
1337 |
|
|
1338 |
|
start_timer(); |
1339 |
|
if(dec->quarterpel) { |
1340 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1341 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1342 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1343 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1344 |
|
else { |
1345 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1346 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1347 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1348 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1349 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1350 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1351 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1352 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1353 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1354 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1355 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1356 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1357 |
|
} |
1358 |
|
} |
1359 |
|
else { |
1360 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1361 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1362 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1363 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1364 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1365 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1366 |
|
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1367 |
|
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1368 |
|
0); |
1369 |
|
} |
1370 |
|
|
1371 |
|
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1372 |
|
uv_dy, stride2, 0); |
1373 |
|
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1374 |
|
uv_dy, stride2, 0); |
1375 |
|
|
1376 |
|
|
1377 |
|
if(dec->quarterpel) { |
1378 |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
1379 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1380 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1381 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1382 |
|
else { |
1383 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1384 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1385 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1386 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1387 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1388 |
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1389 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1390 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1391 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1392 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1393 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1394 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1395 |
|
} |
1396 |
|
} |
1397 |
|
else { |
1398 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1399 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1400 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1401 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
1402 |
|
0); |
1403 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1404 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1405 |
|
stride, 0); |
1406 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1407 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1408 |
|
stride, 0); |
1409 |
|
} |
1410 |
|
|
1411 |
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1412 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1413 |
|
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1414 |
|
b_uv_dx, b_uv_dy, stride2, 0); |
1415 |
|
|
1416 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1417 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1418 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1419 |
|
stride, 1, 8); |
1420 |
|
|
1421 |
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1422 |
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1423 |
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1424 |
|
stride, 1, 8); |
1425 |
|
|
1426 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1427 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1428 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1429 |
|
stride, 1, 8); |
1430 |
|
|
1431 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1432 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1433 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1434 |
|
stride, 1, 8); |
1435 |
|
|
1436 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1437 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1438 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1439 |
|
stride2, 1, 8); |
1440 |
|
|
1441 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1442 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1443 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1444 |
|
stride2, 1, 8); |
1445 |
|
|
1446 |
|
stop_comp_timer(); |
1447 |
|
|
1448 |
|
for (i = 0; i < 6; i++) { |
1449 |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
1450 |
|
|
1451 |
|
if (cbp & (1 << (5 - i))) /* coded */ |
1452 |
|
{ |
1453 |
|
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); /* clear */ |
1454 |
|
|
1455 |
|
start_timer(); |
1456 |
|
get_inter_block(bs, &block[i * 64], direction); |
1457 |
|
stop_coding_timer(); |
1458 |
|
|
1459 |
|
start_timer(); |
1460 |
|
if (dec->quant_type == 0) { |
1461 |
|
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
1462 |
|
} else { |
1463 |
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
1464 |
|
} |
1465 |
|
stop_iquant_timer(); |
1466 |
|
|
1467 |
|
start_timer(); |
1468 |
|
idct(&data[i * 64]); |
1469 |
|
stop_idct_timer(); |
1470 |
|
} |
1471 |
|
} |
1472 |
|
|
1473 |
|
if (dec->interlacing && pMB->field_dct) { |
1474 |
|
next_block = stride; |
1475 |
|
stride *= 2; |
1476 |
|
} |
1477 |
|
|
1478 |
|
start_timer(); |
1479 |
|
if (cbp & 32) |
1480 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
1481 |
|
if (cbp & 16) |
1482 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
1483 |
|
if (cbp & 8) |
1484 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
1485 |
|
if (cbp & 4) |
1486 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
1487 |
|
if (cbp & 2) |
1488 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
1489 |
|
if (cbp & 1) |
1490 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
1491 |
|
stop_transfer_timer(); |
1492 |
|
} |
1493 |
|
|
1494 |
|
|
1495 |
|
/* for decode B-frame dbquant */ |
1496 |
|
int32_t __inline |
1497 |
|
get_dbquant(Bitstream * bs) |
1498 |
|
{ |
1499 |
|
if (!BitstreamGetBit(bs)) /* '0' */ |
1500 |
|
return (0); |
1501 |
|
else if (!BitstreamGetBit(bs)) /* '10' */ |
1502 |
|
return (-2); |
1503 |
|
else /* '11' */ |
1504 |
|
return (2); |
1505 |
|
} |
1506 |
|
|
1507 |
|
/* |
1508 |
|
* For decode B-frame mb_type |
1509 |
|
* bit ret_value |
1510 |
|
* 1 0 |
1511 |
|
* 01 1 |
1512 |
|
* 001 2 |
1513 |
|
* 0001 3 |
1514 |
|
*/ |
1515 |
|
int32_t __inline |
1516 |
|
get_mbtype(Bitstream * bs) |
1517 |
|
{ |
1518 |
|
int32_t mb_type; |
1519 |
|
|
1520 |
|
for (mb_type = 0; mb_type <= 3; mb_type++) { |
1521 |
|
if (BitstreamGetBit(bs)) |
1522 |
|
break; |
1523 |
|
} |
1524 |
|
|
1525 |
|
if (mb_type <= 3) |
1526 |
|
return (mb_type); |
1527 |
|
else |
1528 |
|
return (-1); |
1529 |
|
} |
1530 |
|
|
1531 |
|
void |
1532 |
|
decoder_bframe(DECODER * dec, |
1533 |
|
Bitstream * bs, |
1534 |
|
int quant, |
1535 |
|
int fcode_forward, |
1536 |
|
int fcode_backward) |
1537 |
|
{ |
1538 |
|
uint32_t x, y; |
1539 |
|
VECTOR mv; |
1540 |
|
const VECTOR zeromv = {0,0}; |
1541 |
|
#ifdef BFRAMES_DEC_DEBUG |
1542 |
|
FILE *fp; |
1543 |
|
static char first=0; |
1544 |
|
#define BFRAME_DEBUG if (!first && fp){ \ |
1545 |
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
1546 |
|
} |
1547 |
|
#endif |
1548 |
|
|
1549 |
|
start_timer(); |
1550 |
|
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1551 |
|
dec->width, dec->height); |
1552 |
|
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1553 |
|
dec->width, dec->height); |
1554 |
|
stop_edges_timer(); |
1555 |
|
|
1556 |
|
#ifdef BFRAMES_DEC_DEBUG |
1557 |
|
if (!first){ |
1558 |
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
1559 |
|
} |
1560 |
|
#endif |
1561 |
|
|
1562 |
|
for (y = 0; y < dec->mb_height; y++) { |
1563 |
|
/* Initialize Pred Motion Vector */ |
1564 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1565 |
|
for (x = 0; x < dec->mb_width; x++) { |
1566 |
|
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1567 |
|
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1568 |
|
|
1569 |
|
mv = |
1570 |
|
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1571 |
|
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1572 |
|
|
1573 |
|
/* |
1574 |
|
* skip if the co-located P_VOP macroblock is not coded |
1575 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1576 |
|
* automatically skipped |
1577 |
|
*/ |
1578 |
|
|
1579 |
|
if (last_mb->mode == MODE_NOT_CODED) { |
1580 |
|
/* DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); */ |
1581 |
|
mb->cbp = 0; |
1582 |
|
#ifdef BFRAMES_DEC_DEBUG |
1583 |
|
mb->mb_type = MODE_NOT_CODED; |
1584 |
|
BFRAME_DEBUG |
1585 |
|
#endif |
1586 |
|
mb->mb_type = MODE_FORWARD; |
1587 |
|
mb->quant = last_mb->quant; |
1588 |
|
/* |
1589 |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
1590 |
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
1591 |
|
*/ |
1592 |
|
|
1593 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
1594 |
|
continue; |
1595 |
|
} |
1596 |
|
|
1597 |
|
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1598 |
|
const uint8_t modb2 = BitstreamGetBit(bs); |
1599 |
|
|
1600 |
|
mb->mb_type = get_mbtype(bs); |
1601 |
|
|
1602 |
|
if (!modb2) { /* modb=='00' */ |
1603 |
|
mb->cbp = BitstreamGetBits(bs, 6); |
1604 |
|
} else { |
1605 |
|
mb->cbp = 0; |
1606 |
|
} |
1607 |
|
if (mb->mb_type && mb->cbp) { |
1608 |
|
quant += get_dbquant(bs); |
1609 |
|
|
1610 |
|
if (quant > 31) { |
1611 |
|
quant = 31; |
1612 |
|
} else if (quant < 1) { |
1613 |
|
quant = 1; |
1614 |
|
} |
1615 |
|
} |
1616 |
|
} else { |
1617 |
|
mb->mb_type = MODE_DIRECT_NONE_MV; |
1618 |
|
mb->cbp = 0; |
1619 |
|
} |
1620 |
|
|
1621 |
|
mb->quant = quant; |
1622 |
|
mb->mode = MODE_INTER4V; |
1623 |
|
/* DEBUG1("Switch bm_type=",mb->mb_type); */ |
1624 |
|
|
1625 |
|
#ifdef BFRAMES_DEC_DEBUG |
1626 |
|
BFRAME_DEBUG |
1627 |
|
#endif |
1628 |
|
|
1629 |
|
switch (mb->mb_type) { |
1630 |
|
case MODE_DIRECT: |
1631 |
|
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
1632 |
|
|
1633 |
|
case MODE_DIRECT_NONE_MV: |
1634 |
|
{ |
1635 |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1636 |
|
int i; |
1637 |
|
|
1638 |
|
for (i = 0; i < 4; i++) { |
1639 |
|
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
1640 |
|
/ TRD + mv.x); |
1641 |
|
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1642 |
|
? ((TRB - TRD) * last_mb->mvs[i].x) |
1643 |
|
/ TRD |
1644 |
|
: mb->mvs[i].x - last_mb->mvs[i].x); |
1645 |
|
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
1646 |
|
/ TRD + mv.y); |
1647 |
|
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1648 |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
1649 |
|
/ TRD |
1650 |
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
1651 |
|
} |
1652 |
|
/* DEBUG("B-frame Direct!\n"); */ |
1653 |
|
} |
1654 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1655 |
|
mb, x, y, bs); |
1656 |
|
break; |
1657 |
|
|
1658 |
|
case MODE_INTERPOLATE: |
1659 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1660 |
|
dec->p_fmv); |
1661 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1662 |
|
|
1663 |
|
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
1664 |
|
fcode_backward, dec->p_bmv); |
1665 |
|
dec->p_bmv = mb->b_mvs[1] = mb->b_mvs[2] = |
1666 |
|
mb->b_mvs[3] = mb->b_mvs[0]; |
1667 |
|
|
1668 |
|
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1669 |
|
mb, x, y, bs); |
1670 |
|
/* DEBUG("B-frame Bidir!\n"); */ |
1671 |
|
break; |
1672 |
|
|
1673 |
|
case MODE_BACKWARD: |
1674 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
1675 |
|
dec->p_bmv); |
1676 |
|
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1677 |
|
|
1678 |
|
mb->mode = MODE_INTER; |
1679 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
1680 |
|
/* DEBUG("B-frame Backward!\n"); */ |
1681 |
|
break; |
1682 |
|
|
1683 |
|
case MODE_FORWARD: |
1684 |
|
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
1685 |
|
dec->p_fmv); |
1686 |
|
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1687 |
|
|
1688 |
|
mb->mode = MODE_INTER; |
1689 |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
1690 |
|
/* DEBUG("B-frame Forward!\n"); */ |
1691 |
|
break; |
1692 |
|
|
1693 |
|
default: |
1694 |
|
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
1695 |
|
} |
1696 |
|
} /* End of for */ |
1697 |
|
} |
1698 |
|
|
1699 |
|
#ifdef BFRAMES_DEC_DEBUG |
1700 |
|
if (!first){ |
1701 |
|
first=1; |
1702 |
|
if (fp) |
1703 |
|
fclose(fp); |
1704 |
|
} |
1705 |
|
#endif |
1706 |
|
} |
1707 |
|
|
1708 |
|
|
1709 |
|
|
1710 |
|
/* perform post processing if necessary, and output the image */ |
1711 |
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1712 |
|
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
1713 |
|
{ |
1714 |
|
|
1715 |
|
|
1716 |
|
image_output(img, dec->width, dec->height, |
1717 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1718 |
|
frame->output.csp, dec->interlacing); |
1719 |
|
|
1720 |
|
if (stats) |
1721 |
|
{ |
1722 |
|
stats->type = coding2type(coding_type); |
1723 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1724 |
|
stats->data.vop.time_increment = 0; //XXX: todo |
1725 |
|
} |
1726 |
|
} |
1727 |
|
|
1728 |
|
|
1729 |
int |
int |
1730 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1731 |
XVID_DEC_FRAME * frame) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1732 |
{ |
{ |
1733 |
|
|
1734 |
Bitstream bs; |
Bitstream bs; |
1735 |
uint32_t rounding; |
uint32_t rounding; |
1736 |
|
uint32_t reduced_resolution; |
1737 |
uint32_t quant; |
uint32_t quant; |
1738 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1739 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1740 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1741 |
uint32_t vop_type; |
WARPPOINTS gmc_warp; |
1742 |
|
int coding_type; |
1743 |
|
int success, output, seen_something; |
1744 |
|
|
1745 |
|
if (XVID_MAJOR(frame->version) != 1 || (stats && XVID_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1746 |
|
return XVID_ERR_VERSION; |
1747 |
|
|
1748 |
start_global_timer(); |
start_global_timer(); |
1749 |
|
|
1750 |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1751 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1752 |
|
dec->frames = 0; |
1753 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1754 |
|
|
1755 |
|
if (frame->length < 0) /* decoder flush */ |
1756 |
|
{ |
1757 |
|
int ret; |
1758 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1759 |
|
we have a reference frame, then outout the reference frame */ |
1760 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1761 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1762 |
|
dec->frames = 0; |
1763 |
|
ret = 0; |
1764 |
|
}else{ |
1765 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1766 |
|
ret = XVID_ERR_END; |
1767 |
|
} |
1768 |
|
|
1769 |
|
emms(); |
1770 |
|
stop_global_timer(); |
1771 |
|
return ret; |
1772 |
|
} |
1773 |
|
|
1774 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1775 |
|
|
1776 |
/* add by chenm001 <chenm001@163.com> */ |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1777 |
/* for support B-frame to reference last 2 frame */ |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1778 |
dec->frames++; |
{ |
1779 |
vop_type = |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1780 |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1781 |
&fcode_backward, &intra_dc_threshold); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1782 |
|
emms(); |
1783 |
|
return 1; /* one byte consumed */ |
1784 |
|
} |
1785 |
|
|
1786 |
|
success = 0; |
1787 |
|
output = 0; |
1788 |
|
seen_something = 0; |
1789 |
|
|
1790 |
|
repeat: |
1791 |
|
|
1792 |
|
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1793 |
|
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1794 |
|
|
1795 |
|
DPRINTF(DPRINTF_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", |
1796 |
|
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1797 |
|
|
1798 |
|
if (coding_type == -1) /* nothing */ |
1799 |
|
{ |
1800 |
|
if (success) goto done; |
1801 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1802 |
|
emms(); |
1803 |
|
return BitstreamPos(&bs)/8; |
1804 |
|
} |
1805 |
|
|
1806 |
|
if (coding_type == -2 || coding_type == -3) /* vol and/or resize */ |
1807 |
|
{ |
1808 |
|
if (coding_type == -3) |
1809 |
|
decoder_resize(dec); |
1810 |
|
|
1811 |
|
if (stats) |
1812 |
|
{ |
1813 |
|
stats->type = XVID_TYPE_VOL; |
1814 |
|
stats->data.vol.general = 0; |
1815 |
|
/*XXX: if (dec->interlacing) |
1816 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1817 |
|
stats->data.vol.width = dec->width; |
1818 |
|
stats->data.vol.height = dec->height; |
1819 |
|
stats->data.vol.par = dec->aspect_ratio; |
1820 |
|
stats->data.vol.par_width = dec->par_width; |
1821 |
|
stats->data.vol.par_height = dec->par_height; |
1822 |
|
emms(); |
1823 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1824 |
|
} |
1825 |
|
goto repeat; |
1826 |
|
} |
1827 |
|
|
1828 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1829 |
|
|
|
switch (vop_type) { |
|
|
case P_VOP: |
|
|
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
|
|
intra_dc_threshold); |
|
|
break; |
|
1830 |
|
|
1831 |
|
/* packed_mode: special-N_VOP treament */ |
1832 |
|
if (dec->packed_mode && coding_type == N_VOP) |
1833 |
|
{ |
1834 |
|
if (dec->low_delay_default && dec->frames > 0) |
1835 |
|
{ |
1836 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1837 |
|
output = 1; |
1838 |
|
} |
1839 |
|
/* ignore otherwise */ |
1840 |
|
} |
1841 |
|
else if (coding_type != B_VOP) |
1842 |
|
{ |
1843 |
|
switch(coding_type) |
1844 |
|
{ |
1845 |
case I_VOP: |
case I_VOP: |
1846 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1847 |
break; |
break; |
1848 |
case B_VOP: |
case P_VOP : |
1849 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1850 |
|
fcode_forward, intra_dc_threshold, NULL); |
1851 |
|
break; |
1852 |
|
case S_VOP : |
1853 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1854 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1855 |
break; |
break; |
1856 |
case N_VOP: |
case N_VOP: |
1857 |
/* when low_delay==0, N_VOP's should interpolate between the past and future frames */ |
// XXX: not_coded vops are not used for forward prediction |
1858 |
|
// we should not swap(last_mbs,mbs) |
1859 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1860 |
break; |
break; |
|
|
|
|
default: |
|
|
return XVID_ERR_FAIL; |
|
1861 |
} |
} |
1862 |
|
|
1863 |
frame->length = BitstreamPos(&bs) / 8; |
if (reduced_resolution) |
1864 |
|
{ |
1865 |
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1866 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1867 |
|
16, 0); |
1868 |
|
} |
1869 |
|
|
1870 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1871 |
frame->image, frame->stride, frame->colorspace); |
if (!(dec->low_delay_default && dec->packed_mode)) |
1872 |
|
{ |
1873 |
|
if (dec->low_delay) |
1874 |
|
{ |
1875 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1876 |
|
output = 1; |
1877 |
|
} |
1878 |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
1879 |
|
{ |
1880 |
|
/* output the reference frame */ |
1881 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1882 |
|
output = 1; |
1883 |
|
} |
1884 |
|
} |
1885 |
|
|
|
if (vop_type == I_VOP || vop_type == P_VOP) { |
|
1886 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1887 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1888 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1889 |
|
dec->last_reduced_resolution = reduced_resolution; |
1890 |
|
dec->last_coding_type = coding_type; |
1891 |
|
|
1892 |
/* swap MACROBLOCK */ |
dec->frames++; |
1893 |
/* the Divx will not set the low_delay flage some times */ |
seen_something = 1; |
1894 |
/* so follow code will wrong to not swap at that time */ |
|
1895 |
/* this will broken bitstream! so I'm change it, */ |
}else{ /* B_VOP */ |
1896 |
/* But that is not the best way! can anyone tell me how */ |
|
1897 |
/* to do another way? */ |
if (dec->low_delay) |
1898 |
/* 18-07-2002 MinChen<chenm001@163.com> */ |
{ |
1899 |
/*if (!dec->low_delay && vop_type == P_VOP) */ |
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
1900 |
if (vop_type == P_VOP) |
dec->low_delay = 1; |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
|
1901 |
} |
} |
1902 |
|
|
1903 |
emms(); |
if (dec->frames < 2) |
1904 |
|
{ |
1905 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1906 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1907 |
|
"broken b-frame, mising ref frames"); |
1908 |
|
}else if (dec->time_pp <= dec->time_bp) { |
1909 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1910 |
|
decoded in vfw. */ |
1911 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1912 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1913 |
|
}else{ |
1914 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1915 |
|
} |
1916 |
|
|
1917 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1918 |
|
output = 1; |
1919 |
|
dec->frames++; |
1920 |
|
} |
1921 |
|
|
1922 |
|
BitstreamByteAlign(&bs); |
1923 |
|
|
1924 |
|
/* low_delay_default mode: repeat in packed_mode */ |
1925 |
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
1926 |
|
{ |
1927 |
|
success = 1; |
1928 |
|
goto repeat; |
1929 |
|
} |
1930 |
|
|
1931 |
|
done : |
1932 |
|
|
1933 |
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
1934 |
|
then output the recently decoded frame, or print an error message */ |
1935 |
|
if (dec->low_delay_default && output == 0) |
1936 |
|
{ |
1937 |
|
if (dec->packed_mode && seen_something) |
1938 |
|
{ |
1939 |
|
/* output the recently decoded frame */ |
1940 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1941 |
|
} |
1942 |
|
else |
1943 |
|
{ |
1944 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1945 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1946 |
|
"warning: nothing to output"); |
1947 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1948 |
|
"bframe decoder lag"); |
1949 |
|
|
1950 |
|
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1951 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1952 |
|
|
1953 |
|
} |
1954 |
|
} |
1955 |
|
|
1956 |
|
emms(); |
1957 |
stop_global_timer(); |
stop_global_timer(); |
1958 |
|
|
1959 |
return XVID_ERR_OK; |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
1960 |
} |
} |