1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* 2002 Peter Ross <pross@xvid.org> |
* 2002-2003 Peter Ross <pross@xvid.org> |
8 |
* |
* |
9 |
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
* 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 |
|
* XviD is free software; you can redistribute it and/or modify it |
|
|
* under the terms of the GNU General Public License as published by |
|
11 |
* the Free Software Foundation; either version 2 of the License, or |
* the Free Software Foundation; either version 2 of the License, or |
12 |
* (at your option) any later version. |
* (at your option) any later version. |
13 |
* |
* |
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 |
* |
* |
23 |
* Under section 8 of the GNU General Public License, the copyright |
* $Id: decoder.c,v 1.49.2.20 2003-11-30 16:13:15 edgomez 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. |
|
24 |
* |
* |
25 |
* 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 1.43 2002-11-16 23:38:15 edgomez Exp $ |
|
|
* |
|
|
*************************************************************************/ |
|
26 |
|
|
27 |
|
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.h> |
29 |
#include <string.h> |
#include <string.h> |
30 |
|
|
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 "utils/mem_align.h" |
#include "utils/mem_align.h" |
62 |
|
|
63 |
int |
static int |
64 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_resize(DECODER * dec) |
65 |
{ |
{ |
66 |
DECODER *dec; |
/* free existing */ |
67 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
68 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
69 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
70 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
71 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
72 |
|
|
73 |
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; |
|
74 |
|
|
75 |
dec->width = param->width; |
if (dec->last_mbs) |
76 |
dec->height = param->height; |
xvid_free(dec->last_mbs); |
77 |
|
if (dec->mbs) |
78 |
|
xvid_free(dec->mbs); |
79 |
|
|
80 |
|
/* realloc */ |
81 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
82 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
83 |
|
|
84 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
85 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
|
dec->low_delay = 0; |
|
86 |
|
|
87 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
88 |
xvid_free(dec); |
xvid_free(dec); |
94 |
xvid_free(dec); |
xvid_free(dec); |
95 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
96 |
} |
} |
97 |
// add by chenm001 <chenm001@163.com> |
|
98 |
// for support B-frame to reference last 2 frame |
/* Support B-frame to reference last 2 frame */ |
99 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
100 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
101 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
102 |
xvid_free(dec); |
xvid_free(dec); |
103 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
104 |
} |
} |
105 |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
106 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
107 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
108 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
109 |
|
xvid_free(dec); |
110 |
|
return XVID_ERR_MEMORY; |
111 |
|
} |
112 |
|
|
113 |
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
114 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
115 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
116 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
117 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
118 |
|
xvid_free(dec); |
119 |
|
return XVID_ERR_MEMORY; |
120 |
|
} |
121 |
|
|
122 |
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
123 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
124 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
125 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
126 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
127 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
128 |
xvid_free(dec); |
xvid_free(dec); |
129 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
130 |
} |
} |
136 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
137 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
138 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
139 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
141 |
xvid_free(dec); |
xvid_free(dec); |
142 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
143 |
} |
} |
|
|
|
144 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
145 |
|
|
146 |
// add by chenm001 <chenm001@163.com> |
/* For skip MB flag */ |
|
// for skip MB flag |
|
147 |
dec->last_mbs = |
dec->last_mbs = |
148 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
149 |
CACHE_LINE); |
CACHE_LINE); |
152 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
153 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
154 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
155 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
156 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
157 |
xvid_free(dec); |
xvid_free(dec); |
158 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
159 |
} |
} |
160 |
|
|
161 |
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); |
162 |
|
|
163 |
|
return 0; |
164 |
|
} |
165 |
|
|
166 |
|
|
167 |
|
int |
168 |
|
decoder_create(xvid_dec_create_t * create) |
169 |
|
{ |
170 |
|
DECODER *dec; |
171 |
|
|
172 |
|
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
173 |
|
return XVID_ERR_VERSION; |
174 |
|
|
175 |
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
176 |
|
if (dec == NULL) { |
177 |
|
return XVID_ERR_MEMORY; |
178 |
|
} |
179 |
|
|
180 |
|
memset(dec, 0, sizeof(DECODER)); |
181 |
|
|
182 |
|
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
183 |
|
if (dec->mpeg_quant_matrices == NULL) { |
184 |
|
xvid_free(dec); |
185 |
|
return XVID_ERR_MEMORY; |
186 |
|
} |
187 |
|
|
188 |
|
create->handle = dec; |
189 |
|
|
190 |
|
dec->width = create->width; |
191 |
|
dec->height = create->height; |
192 |
|
|
193 |
|
image_null(&dec->cur); |
194 |
|
image_null(&dec->refn[0]); |
195 |
|
image_null(&dec->refn[1]); |
196 |
|
image_null(&dec->tmp); |
197 |
|
image_null(&dec->qtmp); |
198 |
|
|
199 |
|
/* image based GMC */ |
200 |
|
image_null(&dec->gmc); |
201 |
|
|
202 |
|
|
203 |
|
dec->mbs = NULL; |
204 |
|
dec->last_mbs = NULL; |
205 |
|
|
206 |
init_timer(); |
init_timer(); |
207 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
208 |
|
|
209 |
// add by chenm001 <chenm001@163.com> |
/* For B-frame support (used to save reference frame's time */ |
210 |
// for support B-frame to save reference frame's time |
dec->frames = 0; |
|
dec->frames = -1; |
|
211 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
212 |
|
dec->low_delay = 0; |
213 |
|
dec->packed_mode = 0; |
214 |
|
|
215 |
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
216 |
|
|
217 |
return XVID_ERR_OK; |
if (dec->fixed_dimensions) |
218 |
|
return decoder_resize(dec); |
219 |
|
else |
220 |
|
return 0; |
221 |
} |
} |
222 |
|
|
223 |
|
|
226 |
{ |
{ |
227 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
228 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
229 |
|
|
230 |
|
/* image based GMC */ |
231 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
232 |
|
|
233 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
234 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
235 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
236 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
237 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
238 |
|
xvid_free(dec->mpeg_quant_matrices); |
239 |
xvid_free(dec); |
xvid_free(dec); |
240 |
|
|
241 |
write_timer(); |
write_timer(); |
242 |
return XVID_ERR_OK; |
return 0; |
243 |
} |
} |
244 |
|
|
|
|
|
|
|
|
245 |
static const int32_t dquant_table[4] = { |
static const int32_t dquant_table[4] = { |
246 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
247 |
}; |
}; |
248 |
|
|
249 |
|
/* decode an intra macroblock */ |
250 |
|
static void |
|
|
|
|
// decode an intra macroblock |
|
|
|
|
|
void |
|
251 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
252 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
253 |
const uint32_t x_pos, |
const uint32_t x_pos, |
257 |
Bitstream * bs, |
Bitstream * bs, |
258 |
const uint32_t quant, |
const uint32_t quant, |
259 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
260 |
const unsigned int bound) |
const unsigned int bound, |
261 |
|
const int reduced_resolution) |
262 |
{ |
{ |
263 |
|
|
264 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
271 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
272 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
273 |
|
|
274 |
|
if (reduced_resolution) { |
275 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
276 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
277 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
278 |
|
}else{ |
279 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
280 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
281 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
282 |
|
} |
283 |
|
|
284 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
285 |
|
|
286 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
287 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
304 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
305 |
|
|
306 |
if (dc_size > 8) { |
if (dc_size > 8) { |
307 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); /* marker */ |
308 |
} |
} |
309 |
|
|
310 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
311 |
start_coeff = 1; |
start_coeff = 1; |
312 |
|
|
313 |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
314 |
} else { |
} else { |
315 |
start_coeff = 0; |
start_coeff = 0; |
316 |
} |
} |
317 |
|
|
318 |
start_timer(); |
start_timer(); |
319 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
320 |
{ |
{ |
321 |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
int direction = dec->alternate_vertical_scan ? |
322 |
start_coeff); |
2 : pMB->acpred_directions[i]; |
323 |
|
|
324 |
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
325 |
} |
} |
326 |
stop_coding_timer(); |
stop_coding_timer(); |
327 |
|
|
331 |
|
|
332 |
start_timer(); |
start_timer(); |
333 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
334 |
dequant_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_h263_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
335 |
} else { |
} else { |
336 |
dequant4_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler); |
dequant_mpeg_intra(&data[i * 64], &block[i * 64], iQuant, iDcScaler, dec->mpeg_quant_matrices); |
337 |
} |
} |
338 |
stop_iquant_timer(); |
stop_iquant_timer(); |
339 |
|
|
340 |
start_timer(); |
start_timer(); |
341 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
342 |
stop_idct_timer(); |
stop_idct_timer(); |
343 |
|
|
344 |
} |
} |
345 |
|
|
346 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
349 |
} |
} |
350 |
|
|
351 |
start_timer(); |
start_timer(); |
352 |
|
|
353 |
|
if (reduced_resolution) |
354 |
|
{ |
355 |
|
next_block*=2; |
356 |
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
357 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
358 |
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
359 |
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
360 |
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
361 |
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
362 |
|
}else{ |
363 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
364 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
365 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
366 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
367 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
368 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
369 |
|
} |
370 |
stop_transfer_timer(); |
stop_transfer_timer(); |
371 |
} |
} |
372 |
|
|
373 |
|
static void |
374 |
|
decoder_mb_decode(DECODER * dec, |
375 |
|
const uint32_t cbp, |
376 |
|
Bitstream * bs, |
377 |
|
uint8_t * pY_Cur, |
378 |
|
uint8_t * pU_Cur, |
379 |
|
uint8_t * pV_Cur, |
380 |
|
const int reduced_resolution, |
381 |
|
const MACROBLOCK * pMB) |
382 |
|
{ |
383 |
|
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
384 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
385 |
|
|
386 |
|
int stride = dec->edged_width; |
387 |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
388 |
|
const int stride2 = stride/2; |
389 |
|
int i; |
390 |
|
const uint32_t iQuant = pMB->quant; |
391 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
392 |
|
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
393 |
|
|
394 |
|
for (i = 0; i < 6; i++) { |
395 |
|
|
396 |
|
if (cbp & (1 << (5 - i))) { /* coded */ |
397 |
|
|
398 |
|
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
399 |
|
|
400 |
|
start_timer(); |
401 |
|
get_inter_block(bs, block, direction); |
402 |
|
stop_coding_timer(); |
403 |
|
|
404 |
|
start_timer(); |
405 |
|
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
406 |
|
stop_iquant_timer(); |
407 |
|
|
408 |
#define SIGN(X) (((X)>0)?1:-1) |
start_timer(); |
409 |
#define ABS(X) (((X)>0)?(X):-(X)) |
idct(&data[i * 64]); |
410 |
static const uint32_t roundtab[16] = |
stop_idct_timer(); |
411 |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
} |
412 |
|
} |
413 |
|
|
414 |
|
if (dec->interlacing && pMB->field_dct) { |
415 |
|
next_block = stride; |
416 |
|
stride *= 2; |
417 |
|
} |
418 |
|
|
419 |
// decode an inter macroblock |
start_timer(); |
420 |
|
if (reduced_resolution) { |
421 |
|
if (cbp & 32) |
422 |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
423 |
|
if (cbp & 16) |
424 |
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
425 |
|
if (cbp & 8) |
426 |
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
427 |
|
if (cbp & 4) |
428 |
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
429 |
|
if (cbp & 2) |
430 |
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
431 |
|
if (cbp & 1) |
432 |
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
433 |
|
} else { |
434 |
|
if (cbp & 32) |
435 |
|
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
436 |
|
if (cbp & 16) |
437 |
|
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
438 |
|
if (cbp & 8) |
439 |
|
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
440 |
|
if (cbp & 4) |
441 |
|
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
442 |
|
if (cbp & 2) |
443 |
|
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
444 |
|
if (cbp & 1) |
445 |
|
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
446 |
|
} |
447 |
|
stop_transfer_timer(); |
448 |
|
} |
449 |
|
|
450 |
void |
/* decode an inter macroblock */ |
451 |
|
static void |
452 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
453 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
454 |
const uint32_t x_pos, |
const uint32_t x_pos, |
455 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t acpred_flag, |
|
456 |
const uint32_t cbp, |
const uint32_t cbp, |
457 |
Bitstream * bs, |
Bitstream * bs, |
458 |
const uint32_t quant, |
const uint32_t rounding, |
459 |
const uint32_t rounding) |
const int reduced_resolution, |
460 |
|
const int ref) |
461 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
462 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
463 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
464 |
uint32_t i; |
uint32_t i; |
465 |
uint32_t iQuant = pMB->quant; |
|
466 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
467 |
|
|
468 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
469 |
|
VECTOR mv[4]; /* local copy of mvs */ |
470 |
|
|
471 |
|
if (reduced_resolution) { |
472 |
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
473 |
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
474 |
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
475 |
|
for (i = 0; i < 4; i++) { |
476 |
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
477 |
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
478 |
|
} |
479 |
|
} else { |
480 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
481 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
482 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
483 |
|
for (i = 0; i < 4; i++) |
484 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
mv[i] = pMB->mvs[i]; |
|
uv_dx = pMB->mvs[0].x; |
|
|
uv_dy = pMB->mvs[0].y; |
|
|
|
|
|
if (dec->quarterpel) |
|
|
{ |
|
|
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
|
|
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
|
485 |
} |
} |
486 |
|
|
487 |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
start_timer(); |
|
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; |
|
488 |
|
|
489 |
if (dec->quarterpel) |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
490 |
{ |
|
491 |
sum /= 2; |
uv_dx = mv[0].x; |
492 |
|
uv_dy = mv[0].y; |
493 |
|
if (dec->quarterpel) { |
494 |
|
uv_dx /= 2; |
495 |
|
uv_dy /= 2; |
496 |
} |
} |
497 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
498 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
499 |
|
|
500 |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
if (reduced_resolution) |
501 |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
502 |
|
mv[0].x, mv[0].y, stride, rounding); |
503 |
|
else if (dec->quarterpel) |
504 |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
505 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
506 |
|
mv[0].x, mv[0].y, stride, rounding); |
507 |
|
else |
508 |
|
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
509 |
|
mv[0].x, mv[0].y, stride, rounding); |
510 |
|
|
511 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
} else { /* MODE_INTER4V */ |
512 |
|
|
513 |
if (dec->quarterpel) |
if(dec->quarterpel) { |
514 |
{ |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
515 |
sum /= 2; |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
516 |
|
} else { |
517 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
518 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
519 |
} |
} |
520 |
|
|
521 |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
522 |
} |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
523 |
|
|
524 |
start_timer(); |
if (reduced_resolution) { |
525 |
if(dec->quarterpel) { |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
526 |
DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n"); |
mv[0].x, mv[0].y, stride, rounding); |
527 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
528 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
529 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
530 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
531 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
532 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
533 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
534 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
uv_dx, uv_dy, stride2, rounding); |
535 |
} |
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
536 |
else { |
uv_dx, uv_dy, stride2, rounding); |
537 |
|
|
538 |
|
} else if (dec->quarterpel) { |
539 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
540 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
541 |
|
mv[0].x, mv[0].y, stride, rounding); |
542 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
543 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
544 |
|
mv[1].x, mv[1].y, stride, rounding); |
545 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
546 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
547 |
|
mv[2].x, mv[2].y, stride, rounding); |
548 |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
549 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
550 |
|
mv[3].x, mv[3].y, stride, rounding); |
551 |
|
} else { |
552 |
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, |
553 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
554 |
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, |
555 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
mv[1].x, mv[1].y, stride, rounding); |
556 |
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, |
557 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
mv[2].x, mv[2].y, stride, rounding); |
558 |
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, |
559 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
560 |
|
} |
561 |
} |
} |
562 |
|
|
563 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
/* chroma */ |
564 |
|
if (reduced_resolution) { |
565 |
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
566 |
|
uv_dx, uv_dy, stride2, rounding); |
567 |
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
568 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
569 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
} else { |
570 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
571 |
|
uv_dx, uv_dy, stride2, rounding); |
572 |
|
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
573 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
574 |
|
} |
575 |
|
|
576 |
stop_comp_timer(); |
stop_comp_timer(); |
577 |
|
|
578 |
for (i = 0; i < 6; i++) { |
if (cbp) |
579 |
if (cbp & (1 << (5 - i))) // coded |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
580 |
|
reduced_resolution, pMB); |
581 |
|
} |
582 |
|
|
583 |
|
static void |
584 |
|
decoder_mbgmc(DECODER * dec, |
585 |
|
MACROBLOCK * const pMB, |
586 |
|
const uint32_t x_pos, |
587 |
|
const uint32_t y_pos, |
588 |
|
const uint32_t fcode, |
589 |
|
const uint32_t cbp, |
590 |
|
Bitstream * bs, |
591 |
|
const uint32_t rounding) |
592 |
{ |
{ |
593 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
const uint32_t stride = dec->edged_width; |
594 |
|
const uint32_t stride2 = stride / 2; |
595 |
|
|
596 |
start_timer(); |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
597 |
get_inter_block(bs, &block[i * 64]); |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
598 |
stop_coding_timer(); |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
599 |
|
|
600 |
start_timer(); |
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
601 |
if (dec->quant_type == 0) { |
|
602 |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
603 |
|
|
604 |
start_timer(); |
start_timer(); |
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
605 |
|
|
606 |
if (dec->interlacing && pMB->field_dct) { |
/* this is where the calculations are done */ |
607 |
next_block = stride; |
|
608 |
stride *= 2; |
gmc_data->predict_16x16(gmc_data, |
609 |
} |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
610 |
|
stride, stride, x_pos, y_pos, rounding); |
611 |
|
|
612 |
|
gmc_data->predict_8x8(gmc_data, |
613 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
614 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
615 |
|
stride2, stride2, x_pos, y_pos, rounding); |
616 |
|
|
617 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
618 |
|
|
619 |
|
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
620 |
|
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
621 |
|
|
622 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
623 |
|
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
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); |
|
624 |
stop_transfer_timer(); |
stop_transfer_timer(); |
625 |
|
|
626 |
|
if (cbp) |
627 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
628 |
|
|
629 |
} |
} |
630 |
|
|
631 |
|
|
632 |
void |
static void |
633 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
634 |
Bitstream * bs, |
Bitstream * bs, |
635 |
|
int reduced_resolution, |
636 |
int quant, |
int quant, |
637 |
int intra_dc_threshold) |
int intra_dc_threshold) |
638 |
{ |
{ |
639 |
uint32_t bound; |
uint32_t bound; |
640 |
uint32_t x, y; |
uint32_t x, y; |
641 |
|
uint32_t mb_width = dec->mb_width; |
642 |
|
uint32_t mb_height = dec->mb_height; |
643 |
|
|
644 |
|
if (reduced_resolution) { |
645 |
|
mb_width = (dec->width + 31) / 32; |
646 |
|
mb_height = (dec->height + 31) / 32; |
647 |
|
} |
648 |
|
|
649 |
bound = 0; |
bound = 0; |
650 |
|
|
651 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
652 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
653 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
654 |
uint32_t mcbpc; |
uint32_t mcbpc; |
655 |
uint32_t cbpc; |
uint32_t cbpc; |
662 |
|
|
663 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
664 |
{ |
{ |
665 |
bound = read_video_packet_header(bs, 0, &quant); |
bound = read_video_packet_header(bs, dec, 0, |
666 |
x = bound % dec->mb_width; |
&quant, NULL, NULL, &intra_dc_threshold); |
667 |
y = bound / dec->mb_width; |
x = bound % mb_width; |
668 |
|
y = bound / mb_width; |
669 |
} |
} |
670 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
671 |
|
|
672 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
673 |
|
|
674 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
675 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
696 |
|
|
697 |
if (dec->interlacing) { |
if (dec->interlacing) { |
698 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
699 |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
700 |
} |
} |
701 |
|
|
702 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
703 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
704 |
|
|
705 |
} |
} |
706 |
if(dec->out_frm) |
if(dec->out_frm) |
707 |
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); |
|
|
|
708 |
} |
} |
709 |
|
|
710 |
} |
} |
711 |
|
|
712 |
|
|
713 |
void |
static void |
714 |
get_motion_vector(DECODER * dec, |
get_motion_vector(DECODER * dec, |
715 |
Bitstream * bs, |
Bitstream * bs, |
716 |
int x, |
int x, |
717 |
int y, |
int y, |
718 |
int k, |
int k, |
719 |
VECTOR * mv, |
VECTOR * ret_mv, |
720 |
int fcode, |
int fcode, |
721 |
const int bound) |
const int bound) |
722 |
{ |
{ |
723 |
|
|
724 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
725 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
726 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
727 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
|
|
|
|
VECTOR pmv; |
|
|
int mv_x, mv_y; |
|
728 |
|
|
729 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
730 |
|
VECTOR mv; |
731 |
|
|
732 |
mv_x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
733 |
mv_y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
734 |
|
|
735 |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
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); |
736 |
|
|
737 |
mv_x += pmv.x; |
mv.x += pmv.x; |
738 |
mv_y += pmv.y; |
mv.y += pmv.y; |
739 |
|
|
740 |
if (mv_x < low) { |
if (mv.x < low) { |
741 |
mv_x += range; |
mv.x += range; |
742 |
} else if (mv_x > high) { |
} else if (mv.x > high) { |
743 |
mv_x -= range; |
mv.x -= range; |
744 |
} |
} |
745 |
|
|
746 |
if (mv_y < low) { |
if (mv.y < low) { |
747 |
mv_y += range; |
mv.y += range; |
748 |
} else if (mv_y > high) { |
} else if (mv.y > high) { |
749 |
mv_y -= range; |
mv.y -= range; |
750 |
} |
} |
751 |
|
|
752 |
mv->x = mv_x; |
ret_mv->x = mv.x; |
753 |
mv->y = mv_y; |
ret_mv->y = mv.y; |
|
|
|
754 |
} |
} |
755 |
|
|
756 |
|
/* for P_VOP set gmc_warp to NULL */ |
757 |
void |
static void |
758 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
759 |
Bitstream * bs, |
Bitstream * bs, |
760 |
int rounding, |
int rounding, |
761 |
|
int reduced_resolution, |
762 |
int quant, |
int quant, |
763 |
int fcode, |
int fcode, |
764 |
int intra_dc_threshold) |
int intra_dc_threshold, |
765 |
|
const WARPPOINTS *const gmc_warp) |
766 |
{ |
{ |
|
|
|
767 |
uint32_t x, y; |
uint32_t x, y; |
768 |
uint32_t bound; |
uint32_t bound; |
769 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
770 |
|
uint32_t mb_width = dec->mb_width; |
771 |
|
uint32_t mb_height = dec->mb_height; |
772 |
|
|
773 |
|
if (reduced_resolution) { |
774 |
|
mb_width = (dec->width + 31) / 32; |
775 |
|
mb_height = (dec->height + 31) / 32; |
776 |
|
} |
777 |
|
|
778 |
start_timer(); |
start_timer(); |
779 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
780 |
dec->width, dec->height); |
dec->width, dec->height); |
781 |
stop_edges_timer(); |
stop_edges_timer(); |
782 |
|
|
783 |
|
if (gmc_warp) { |
784 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
785 |
|
generate_GMCparameters( dec->sprite_warping_points, |
786 |
|
dec->sprite_warping_accuracy, gmc_warp, |
787 |
|
dec->width, dec->height, &dec->new_gmc_data); |
788 |
|
|
789 |
|
/* image warping is done block-based in decoder_mbgmc(), now */ |
790 |
|
} |
791 |
|
|
792 |
bound = 0; |
bound = 0; |
793 |
|
|
794 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < mb_height; y++) { |
795 |
cp_mb = st_mb = 0; |
cp_mb = st_mb = 0; |
796 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
797 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
798 |
|
|
799 |
// skip stuffing |
/* skip stuffing */ |
800 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
801 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
802 |
|
|
803 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) { |
804 |
{ |
bound = read_video_packet_header(bs, dec, fcode - 1, |
805 |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
&quant, &fcode, NULL, &intra_dc_threshold); |
806 |
x = bound % dec->mb_width; |
x = bound % mb_width; |
807 |
y = bound / dec->mb_width; |
y = bound / mb_width; |
808 |
} |
} |
809 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
810 |
|
|
811 |
DPRINTF(DPRINTF_MB, "macroblock (%i,%i) %08x", x, y, BitstreamShowBits(bs, 32)); |
DPRINTF(XVID_DEBUG_MB, "macroblock (%i,%i) %08x\n", x, y, BitstreamShowBits(bs, 32)); |
812 |
|
|
813 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
814 |
if (!(BitstreamGetBit(bs))) // not_coded |
uint32_t mcbpc, cbpc, cbpy, cbp; |
815 |
{ |
uint32_t intra, acpred_flag = 0; |
816 |
uint32_t mcbpc; |
int mcsel = 0; /* mcsel: '0'=local motion, '1'=GMC */ |
|
uint32_t cbpc; |
|
|
uint32_t acpred_flag; |
|
|
uint32_t cbpy; |
|
|
uint32_t cbp; |
|
|
uint32_t intra; |
|
817 |
|
|
818 |
cp_mb++; |
cp_mb++; |
819 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
820 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
821 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
822 |
|
|
823 |
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
824 |
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
|
acpred_flag = 0; |
|
825 |
|
|
826 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
827 |
|
|
828 |
if (intra) { |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
829 |
|
mcsel = BitstreamGetBit(bs); |
830 |
|
else if (intra) |
831 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
|
} |
|
832 |
|
|
833 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
834 |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
835 |
|
|
836 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
837 |
|
|
838 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
839 |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
840 |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
841 |
quant += dquant; |
quant += dquant; |
842 |
if (quant > 31) { |
if (quant > 31) { |
843 |
quant = 31; |
quant = 31; |
844 |
} else if (quant < 1) { |
} else if (quant < 1) { |
845 |
quant = 1; |
quant = 1; |
846 |
} |
} |
847 |
DPRINTF(DPRINTF_MB, "quant %i", quant); |
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
848 |
} |
} |
849 |
mb->quant = quant; |
mb->quant = quant; |
850 |
|
|
851 |
if (dec->interlacing) { |
if (dec->interlacing) { |
852 |
if (cbp || intra) { |
if ((cbp || intra) && !mcsel) { |
853 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
854 |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
855 |
} |
} |
856 |
|
|
857 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
858 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
859 |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
860 |
|
|
861 |
if (mb->field_pred) { |
if (mb->field_pred) { |
862 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
863 |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
864 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
865 |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
866 |
} |
} |
867 |
} |
} |
868 |
} |
} |
869 |
|
|
870 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mcsel) { |
871 |
|
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
872 |
|
continue; |
873 |
|
|
874 |
|
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
875 |
|
|
876 |
if (dec->interlacing && mb->field_pred) { |
if (dec->interlacing && mb->field_pred) { |
877 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
878 |
fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
|
|
fcode, bound); |
|
879 |
} else { |
} else { |
880 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
881 |
fcode, bound); |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
|
mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
|
|
mb->mvs[0].x; |
|
|
mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
mb->mvs[0].y; |
|
882 |
} |
} |
883 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
|
|
|
884 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
885 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
886 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
887 |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
get_motion_vector(dec, bs, x, y, 3, &mb->mvs[3], fcode, bound); |
888 |
} else // MODE_INTRA, MODE_INTRA_Q |
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
889 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
890 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
|
0; |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
|
|
0; |
|
891 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
892 |
intra_dc_threshold, bound); |
intra_dc_threshold, bound, reduced_resolution); |
893 |
continue; |
continue; |
894 |
} |
} |
895 |
|
|
896 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
897 |
rounding); |
rounding, reduced_resolution, 0); |
|
} else // not coded |
|
|
{ |
|
|
DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y); |
|
898 |
|
|
899 |
|
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
900 |
|
mb->mode = MODE_NOT_CODED_GMC; |
901 |
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
902 |
|
|
903 |
|
if(dec->out_frm && cp_mb > 0) { |
904 |
|
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
905 |
|
cp_mb = 0; |
906 |
|
} |
907 |
|
st_mb = x+1; |
908 |
|
} else { /* not coded P_VOP macroblock */ |
909 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
910 |
|
|
911 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
912 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
913 |
|
|
914 |
// copy macroblock directly from ref to cur |
decoder_mbinter(dec, mb, x, y, 0, bs, |
915 |
|
rounding, reduced_resolution, 0); |
916 |
|
|
|
start_timer(); |
|
|
|
|
|
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
|
|
(16 * x), |
|
|
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); |
|
|
|
|
|
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); |
|
|
stop_transfer_timer(); |
|
917 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
918 |
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); |
919 |
cp_mb = 0; |
cp_mb = 0; |
921 |
st_mb = x+1; |
st_mb = x+1; |
922 |
} |
} |
923 |
} |
} |
924 |
|
|
925 |
if(dec->out_frm && cp_mb > 0) |
if(dec->out_frm && cp_mb > 0) |
926 |
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); |
927 |
} |
} |
928 |
} |
} |
929 |
|
|
930 |
|
|
931 |
// add by MinChen <chenm001@163.com> |
/* decode B-frame motion vector */ |
932 |
// decode B-frame motion vector |
static void |
933 |
void |
get_b_motion_vector(Bitstream * bs, |
|
get_b_motion_vector(DECODER * dec, |
|
|
Bitstream * bs, |
|
|
int x, |
|
|
int y, |
|
934 |
VECTOR * mv, |
VECTOR * mv, |
935 |
int fcode, |
int fcode, |
936 |
const VECTOR pmv) |
const VECTOR pmv) |
937 |
{ |
{ |
938 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
939 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
940 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
941 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
942 |
|
|
943 |
int mv_x, mv_y; |
int mv_x = get_mv(bs, fcode); |
944 |
int pmv_x, pmv_y; |
int mv_y = get_mv(bs, fcode); |
945 |
|
|
946 |
pmv_x = pmv.x; |
mv_x += pmv.x; |
947 |
pmv_y = pmv.y; |
mv_y += pmv.y; |
|
|
|
|
mv_x = get_mv(bs, fcode); |
|
|
mv_y = get_mv(bs, fcode); |
|
|
|
|
|
mv_x += pmv_x; |
|
|
mv_y += pmv_y; |
|
948 |
|
|
949 |
if (mv_x < low) { |
if (mv_x < low) |
950 |
mv_x += range; |
mv_x += range; |
951 |
} else if (mv_x > high) { |
else if (mv_x > high) |
952 |
mv_x -= range; |
mv_x -= range; |
|
} |
|
953 |
|
|
954 |
if (mv_y < low) { |
if (mv_y < low) |
955 |
mv_y += range; |
mv_y += range; |
956 |
} else if (mv_y > high) { |
else if (mv_y > high) |
957 |
mv_y -= range; |
mv_y -= range; |
|
} |
|
958 |
|
|
959 |
mv->x = mv_x; |
mv->x = mv_x; |
960 |
mv->y = mv_y; |
mv->y = mv_y; |
961 |
} |
} |
962 |
|
|
963 |
|
/* decode an B-frame direct & interpolate macroblock */ |
964 |
// add by MinChen <chenm001@163.com> |
static void |
|
// decode an B-frame forward & backward inter macroblock |
|
|
void |
|
|
decoder_bf_mbinter(DECODER * dec, |
|
|
const MACROBLOCK * pMB, |
|
|
const uint32_t x_pos, |
|
|
const uint32_t y_pos, |
|
|
const uint32_t cbp, |
|
|
Bitstream * bs, |
|
|
const uint32_t quant, |
|
|
const uint8_t ref) |
|
|
{ |
|
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
|
uint32_t stride = dec->edged_width; |
|
|
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; |
|
|
|
|
|
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)); |
|
|
|
|
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
|
|
uv_dy = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
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); |
|
|
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 |
|
|
|
|
|
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); |
|
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
|
|
|
|
start_timer(); |
|
|
idct(&data[i * 64]); |
|
|
stop_idct_timer(); |
|
|
} |
|
|
} |
|
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if (cbp & 32) |
|
|
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(); |
|
|
} |
|
|
|
|
|
|
|
|
// add by MinChen <chenm001@163.com> |
|
|
// decode an B-frame direct & inter macroblock |
|
|
void |
|
965 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
966 |
IMAGE forward, |
IMAGE forward, |
967 |
IMAGE backward, |
IMAGE backward, |
968 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
969 |
const uint32_t x_pos, |
const uint32_t x_pos, |
970 |
const uint32_t y_pos, |
const uint32_t y_pos, |
971 |
Bitstream * bs) |
Bitstream * bs, |
972 |
|
const int direct) |
973 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
974 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
975 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
|
uint32_t iQuant = pMB->quant; |
|
976 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
977 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
978 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
979 |
const uint32_t cbp = pMB->cbp; |
const uint32_t cbp = pMB->cbp; |
980 |
|
|
982 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
983 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
984 |
|
|
985 |
|
if (!direct) { |
|
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
|
986 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
987 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
988 |
|
|
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
|
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
|
|
|
989 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
990 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
991 |
|
|
992 |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
if (dec->quarterpel) { |
993 |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
uv_dx /= 2; |
994 |
} else { |
uv_dy /= 2; |
995 |
int sum; |
b_uv_dx /= 2; |
996 |
|
b_uv_dy /= 2; |
997 |
|
} |
998 |
|
|
999 |
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1000 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
1001 |
|
|
1002 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1003 |
uv_dx = |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1004 |
(sum == |
|
1005 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
} else { |
1006 |
(ABS(sum) / 16) * 2)); |
if(dec->quarterpel) { |
1007 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1008 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1009 |
uv_dy = |
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); |
1010 |
(sum == |
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); |
1011 |
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
} else { |
1012 |
(ABS(sum) / 16) * 2)); |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1013 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1014 |
sum = |
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1015 |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
|
pMB->b_mvs[3].x; |
|
|
b_uv_dx = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
|
|
|
|
sum = |
|
|
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
|
|
pMB->b_mvs[3].y; |
|
|
b_uv_dy = |
|
|
(sum == |
|
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
|
|
(ABS(sum) / 16) * 2)); |
|
1016 |
} |
} |
1017 |
|
|
1018 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1019 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1020 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1021 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1022 |
|
} |
1023 |
|
|
1024 |
start_timer(); |
start_timer(); |
1025 |
|
if(dec->quarterpel) { |
1026 |
|
if(!direct) { |
1027 |
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1028 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1029 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1030 |
|
} else { |
1031 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1032 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1033 |
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1034 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1035 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1036 |
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1037 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1038 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1039 |
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1040 |
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1041 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1042 |
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1043 |
|
} |
1044 |
|
} else { |
1045 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos, |
1046 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1047 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos, |
1049 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos, 16 * y_pos + 8, |
1050 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1051 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1052 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1053 |
0); |
} |
1054 |
|
|
1055 |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.u, forward.u, 8 * x_pos, 8 * y_pos, uv_dx, |
1056 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1057 |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1058 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1059 |
|
|
1060 |
|
|
1061 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
if(dec->quarterpel) { |
1062 |
|
if(!direct) { |
1063 |
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1064 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1065 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
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); |
|
1066 |
} else { |
} else { |
1067 |
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1068 |
} |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1069 |
stop_iquant_timer(); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1070 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1071 |
start_timer(); |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1072 |
idct(&data[i * 64]); |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1073 |
stop_idct_timer(); |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1074 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1075 |
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1076 |
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1077 |
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1078 |
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1079 |
} |
} |
1080 |
|
} else { |
1081 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
1082 |
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1083 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1084 |
|
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1085 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
1086 |
|
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1087 |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
1088 |
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1089 |
} |
} |
1090 |
|
|
1091 |
if (dec->interlacing && pMB->field_dct) { |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
1092 |
next_block = stride; |
b_uv_dx, b_uv_dy, stride2, 0); |
1093 |
stride *= 2; |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
1094 |
} |
b_uv_dx, b_uv_dy, stride2, 0); |
1095 |
|
|
1096 |
start_timer(); |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1097 |
if (cbp & 32) |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
1098 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
1099 |
if (cbp & 16) |
stride, 1, 8); |
1100 |
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
|
1101 |
if (cbp & 8) |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1102 |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1103 |
if (cbp & 4) |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
1104 |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
stride, 1, 8); |
1105 |
if (cbp & 2) |
|
1106 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1107 |
if (cbp & 1) |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1108 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
1109 |
stop_transfer_timer(); |
stride, 1, 8); |
1110 |
} |
|
1111 |
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1112 |
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1113 |
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
1114 |
|
stride, 1, 8); |
1115 |
|
|
1116 |
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1117 |
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
1118 |
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
1119 |
|
stride2, 1, 8); |
1120 |
|
|
1121 |
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1122 |
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
1123 |
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
1124 |
|
stride2, 1, 8); |
1125 |
|
|
1126 |
|
stop_comp_timer(); |
1127 |
|
|
1128 |
// add by MinChen <chenm001@163.com> |
if (cbp) |
1129 |
// for decode B-frame dbquant |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
1130 |
int32_t __inline |
} |
1131 |
|
|
1132 |
|
/* for decode B-frame dbquant */ |
1133 |
|
static __inline int32_t |
1134 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1135 |
{ |
{ |
1136 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1137 |
return (0); |
return (0); |
1138 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1139 |
return (-2); |
return (-2); |
1140 |
else |
else /* '11' */ |
1141 |
return (2); // '11' |
return (2); |
1142 |
} |
} |
1143 |
|
|
1144 |
// add by MinChen <chenm001@163.com> |
/* |
1145 |
// for decode B-frame mb_type |
* decode B-frame mb_type |
1146 |
// bit ret_value |
* bit ret_value |
1147 |
// 1 0 |
* 1 0 |
1148 |
// 01 1 |
* 01 1 |
1149 |
// 001 2 |
* 001 2 |
1150 |
// 0001 3 |
* 0001 3 |
1151 |
int32_t __inline |
*/ |
1152 |
|
static int32_t __inline |
1153 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1154 |
{ |
{ |
1155 |
int32_t mb_type; |
int32_t mb_type; |
1156 |
|
|
1157 |
for (mb_type = 0; mb_type <= 3; mb_type++) { |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1158 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
|
break; |
|
|
} |
|
|
|
|
|
if (mb_type <= 3) |
|
1159 |
return (mb_type); |
return (mb_type); |
1160 |
else |
|
1161 |
return (-1); |
return -1; |
1162 |
} |
} |
1163 |
|
|
1164 |
void |
static void |
1165 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1166 |
Bitstream * bs, |
Bitstream * bs, |
1167 |
int quant, |
int quant, |
1171 |
uint32_t x, y; |
uint32_t x, y; |
1172 |
VECTOR mv; |
VECTOR mv; |
1173 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1174 |
#ifdef BFRAMES_DEC_DEBUG |
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
1175 |
FILE *fp; |
int i; |
|
static char first=0; |
|
|
#define BFRAME_DEBUG if (!first && fp){ \ |
|
|
fprintf(fp,"Y=%3d X=%3d MB=%2d CBP=%02X\n",y,x,mb->mb_type,mb->cbp); \ |
|
|
} |
|
|
#endif |
|
1176 |
|
|
1177 |
start_timer(); |
start_timer(); |
1178 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1181 |
dec->width, dec->height); |
dec->width, dec->height); |
1182 |
stop_edges_timer(); |
stop_edges_timer(); |
1183 |
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
fp=fopen("C:\\XVIDDBG.TXT","w"); |
|
|
} |
|
|
#endif |
|
|
|
|
1184 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1185 |
// Initialize Pred Motion Vector |
/* Initialize Pred Motion Vector */ |
1186 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1187 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1188 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1189 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1190 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1191 |
|
uint32_t intra_dc_threshold; /* fake variable */ |
1192 |
|
|
1193 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1194 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1195 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1196 |
|
x = bound % dec->mb_width; |
1197 |
|
y = bound / dec->mb_width; |
1198 |
|
/* reset predicted macroblocks */ |
1199 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1200 |
|
} |
1201 |
|
|
1202 |
mv = |
mv = |
1203 |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
mb->b_mvs[0] = mb->b_mvs[1] = mb->b_mvs[2] = mb->b_mvs[3] = |
1204 |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
mb->mvs[0] = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = zeromv; |
1205 |
|
mb->quant = quant; |
1206 |
|
|
1207 |
|
/* |
1208 |
|
* skip if the co-located P_VOP macroblock is not coded |
1209 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1210 |
|
* automatically skipped |
1211 |
|
*/ |
1212 |
|
|
|
// the last P_VOP is skip macroblock ? |
|
1213 |
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); |
|
1214 |
mb->cbp = 0; |
mb->cbp = 0; |
1215 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
1216 |
mb->mb_type = MODE_NOT_CODED; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
BFRAME_DEBUG |
|
|
#endif |
|
|
mb->mb_type = MODE_FORWARD; |
|
|
mb->quant = last_mb->quant; |
|
|
//mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = mb->mvs[0].x; |
|
|
//mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = mb->mvs[0].y; |
|
|
|
|
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, mb->quant, 1); |
|
1217 |
continue; |
continue; |
1218 |
} |
} |
1219 |
|
|
1220 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1221 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1222 |
|
|
1223 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1224 |
|
|
1225 |
if (!modb2) { // modb=='00' |
if (!modb2) /* modb=='00' */ |
1226 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1227 |
} else { |
else |
1228 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp) { |
|
|
quant += get_dbquant(bs); |
|
1229 |
|
|
1230 |
if (quant > 31) { |
if (mb->mode && mb->cbp) { |
1231 |
|
quant += get_dbquant(bs); |
1232 |
|
if (quant > 31) |
1233 |
quant = 31; |
quant = 31; |
1234 |
} else if (quant < 1) { |
else if (quant < 1) |
1235 |
quant = 1; |
quant = 1; |
1236 |
} |
} |
1237 |
|
mb->quant = quant; |
1238 |
|
|
1239 |
|
if (dec->interlacing) { |
1240 |
|
if (mb->cbp) { |
1241 |
|
mb->field_dct = BitstreamGetBit(bs); |
1242 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1243 |
|
} |
1244 |
|
|
1245 |
|
if (mb->mode) { |
1246 |
|
mb->field_pred = BitstreamGetBit(bs); |
1247 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1248 |
|
|
1249 |
|
if (mb->field_pred) { |
1250 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1251 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1252 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1253 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1254 |
|
} |
1255 |
|
} |
1256 |
} |
} |
1257 |
|
|
1258 |
} else { |
} else { |
1259 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1260 |
mb->cbp = 0; |
mb->cbp = 0; |
1261 |
} |
} |
1262 |
|
|
1263 |
mb->quant = quant; |
switch (mb->mode) { |
|
mb->mode = MODE_INTER4V; |
|
|
//DEBUG1("Switch bm_type=",mb->mb_type); |
|
|
|
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
BFRAME_DEBUG |
|
|
#endif |
|
|
|
|
|
switch (mb->mb_type) { |
|
1264 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1265 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv); |
1266 |
|
|
1267 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
|
{ |
|
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
|
int i; |
|
|
|
|
1268 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1269 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
|
/ TRD + mv.x); |
|
1270 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
1271 |
? ((TRB - TRD) * last_mb->mvs[i].x) |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
|
/ TRD |
|
1272 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
: mb->mvs[i].x - last_mb->mvs[i].x); |
1273 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
|
/ TRD + mv.y); |
|
1274 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
1275 |
? ((TRB - TRD) * last_mb->mvs[i].y) |
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
|
/ TRD |
|
1276 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
: mb->mvs[i].y - last_mb->mvs[i].y); |
1277 |
} |
} |
1278 |
//DEBUG("B-frame Direct!\n"); |
|
|
} |
|
1279 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1280 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1281 |
break; |
break; |
1282 |
|
|
1283 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1284 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv); |
|
dec->p_fmv); |
|
1285 |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1286 |
|
|
1287 |
get_b_motion_vector(dec, bs, x, y, &mb->b_mvs[0], |
get_b_motion_vector(bs, &mb->b_mvs[0], fcode_backward, dec->p_bmv); |
1288 |
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 = mb->b_mvs[1] = mb->b_mvs[2] = |
|
|
mb->b_mvs[3] = mb->b_mvs[0]; |
|
1289 |
|
|
1290 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1291 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
|
//DEBUG("B-frame Bidir!\n"); |
|
1292 |
break; |
break; |
1293 |
|
|
1294 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1295 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_backward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv); |
|
dec->p_bmv); |
|
1296 |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1297 |
|
|
1298 |
mb->mode = MODE_INTER; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 0); |
|
|
//DEBUG("B-frame Backward!\n"); |
|
1299 |
break; |
break; |
1300 |
|
|
1301 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1302 |
get_b_motion_vector(dec, bs, x, y, &mb->mvs[0], fcode_forward, |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv); |
|
dec->p_fmv); |
|
1303 |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1304 |
|
|
1305 |
mb->mode = MODE_INTER; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
|
|
//DEBUG("B-frame Forward!\n"); |
|
1306 |
break; |
break; |
1307 |
|
|
1308 |
default: |
default: |
1309 |
DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1310 |
} |
} |
1311 |
|
} /* End of for */ |
|
} // end of FOR |
|
1312 |
} |
} |
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
first=1; |
|
|
if (fp) |
|
|
fclose(fp); |
|
|
} |
|
|
#endif |
|
1313 |
} |
} |
1314 |
|
|
1315 |
// swap two MACROBLOCK array |
/* perform post processing if necessary, and output the image */ |
1316 |
void |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1317 |
mb_swap(MACROBLOCK ** mb1, |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, int coding_type) |
|
MACROBLOCK ** mb2) |
|
1318 |
{ |
{ |
1319 |
MACROBLOCK *temp = *mb1; |
image_output(img, dec->width, dec->height, |
1320 |
|
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1321 |
|
frame->output.csp, dec->interlacing); |
1322 |
|
|
1323 |
*mb1 = *mb2; |
if (stats) { |
1324 |
*mb2 = temp; |
stats->type = coding2type(coding_type); |
1325 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1326 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1327 |
|
} |
1328 |
} |
} |
1329 |
|
|
1330 |
|
|
1331 |
int |
int |
1332 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1333 |
XVID_DEC_FRAME * frame) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1334 |
{ |
{ |
1335 |
|
|
1336 |
Bitstream bs; |
Bitstream bs; |
1337 |
uint32_t rounding; |
uint32_t rounding; |
1338 |
|
uint32_t reduced_resolution; |
1339 |
uint32_t quant; |
uint32_t quant; |
1340 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1341 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1342 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1343 |
uint32_t vop_type; |
WARPPOINTS gmc_warp; |
1344 |
|
int coding_type; |
1345 |
|
int success, output, seen_something; |
1346 |
|
|
1347 |
|
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1348 |
|
return XVID_ERR_VERSION; |
1349 |
|
|
1350 |
start_global_timer(); |
start_global_timer(); |
1351 |
|
|
1352 |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1353 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1354 |
|
dec->frames = 0; |
1355 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1356 |
|
|
1357 |
|
if (frame->length < 0) { /* decoder flush */ |
1358 |
|
int ret; |
1359 |
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1360 |
|
we have a reference frame, then outout the reference frame */ |
1361 |
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1362 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1363 |
|
dec->frames = 0; |
1364 |
|
ret = 0; |
1365 |
|
} else { |
1366 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1367 |
|
ret = XVID_ERR_END; |
1368 |
|
} |
1369 |
|
|
1370 |
|
emms(); |
1371 |
|
stop_global_timer(); |
1372 |
|
return ret; |
1373 |
|
} |
1374 |
|
|
1375 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1376 |
|
|
1377 |
// add by chenm001 <chenm001@163.com> |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1378 |
// for support B-frame to reference last 2 frame |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1379 |
dec->frames++; |
{ |
1380 |
vop_type = |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1381 |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1382 |
&fcode_backward, &intra_dc_threshold); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1383 |
|
emms(); |
1384 |
|
return 1; /* one byte consumed */ |
1385 |
|
} |
1386 |
|
|
1387 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
success = 0; |
1388 |
|
output = 0; |
1389 |
|
seen_something = 0; |
1390 |
|
|
1391 |
switch (vop_type) { |
repeat: |
|
case P_VOP: |
|
|
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
|
|
intra_dc_threshold); |
|
|
#ifdef BFRAMES_DEC |
|
|
DPRINTF(DPRINTF_DEBUG, "P_VOP Time=%d", dec->time); |
|
|
#endif |
|
|
break; |
|
1392 |
|
|
1393 |
case I_VOP: |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
1394 |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
|
#ifdef BFRAMES_DEC |
|
|
DPRINTF(DPRINTF_DEBUG, "I_VOP Time=%d", dec->time); |
|
|
#endif |
|
|
break; |
|
1395 |
|
|
1396 |
case B_VOP: |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
1397 |
#ifdef BFRAMES_DEC |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1398 |
if (dec->time_pp > dec->time_bp) { |
|
1399 |
DPRINTF(DPRINTF_DEBUG, "B_VOP Time=%d", dec->time); |
if (coding_type == -1) { /* nothing */ |
1400 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
if (success) goto done; |
1401 |
} else { |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1402 |
DPRINTF(DPRINTF_DEBUG, "Broken B_VOP"); |
emms(); |
1403 |
|
return BitstreamPos(&bs)/8; |
1404 |
|
} |
1405 |
|
|
1406 |
|
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1407 |
|
|
1408 |
|
if (coding_type == -3) |
1409 |
|
decoder_resize(dec); |
1410 |
|
|
1411 |
|
if (stats) { |
1412 |
|
stats->type = XVID_TYPE_VOL; |
1413 |
|
stats->data.vol.general = 0; |
1414 |
|
/*XXX: if (dec->interlacing) |
1415 |
|
stats->data.vol.general |= ++INTERLACING; */ |
1416 |
|
stats->data.vol.width = dec->width; |
1417 |
|
stats->data.vol.height = dec->height; |
1418 |
|
stats->data.vol.par = dec->aspect_ratio; |
1419 |
|
stats->data.vol.par_width = dec->par_width; |
1420 |
|
stats->data.vol.par_height = dec->par_height; |
1421 |
|
emms(); |
1422 |
|
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1423 |
|
} |
1424 |
|
goto repeat; |
1425 |
} |
} |
|
#else |
|
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
|
|
#endif |
|
|
break; |
|
1426 |
|
|
1427 |
case N_VOP: // vop not coded |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1428 |
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
|
1429 |
|
/* packed_mode: special-N_VOP treament */ |
1430 |
|
if (dec->packed_mode && coding_type == N_VOP) { |
1431 |
|
if (dec->low_delay_default && dec->frames > 0) { |
1432 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1433 |
|
output = 1; |
1434 |
|
} |
1435 |
|
/* ignore otherwise */ |
1436 |
|
} else if (coding_type != B_VOP) { |
1437 |
|
switch(coding_type) { |
1438 |
|
case I_VOP : |
1439 |
|
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
1440 |
|
break; |
1441 |
|
case P_VOP : |
1442 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1443 |
|
fcode_forward, intra_dc_threshold, NULL); |
1444 |
|
break; |
1445 |
|
case S_VOP : |
1446 |
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
1447 |
|
fcode_forward, intra_dc_threshold, &gmc_warp); |
1448 |
|
break; |
1449 |
|
case N_VOP : |
1450 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1451 |
|
/* we should not swap(last_mbs,mbs) */ |
1452 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1453 |
break; |
break; |
1454 |
|
} |
1455 |
|
|
1456 |
default: |
if (reduced_resolution) { |
1457 |
return XVID_ERR_FAIL; |
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
1458 |
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
1459 |
|
16, 0); |
1460 |
} |
} |
1461 |
|
|
1462 |
#ifdef BFRAMES_DEC_DEBUG |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1463 |
if (frame->length != BitstreamPos(&bs) / 8){ |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1464 |
DPRINTF(DPRINTF_DEBUG, "InLen: %d / UseLen: %d", frame->length, BitstreamPos(&bs) / 8); |
if (dec->low_delay) { |
1465 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1466 |
|
output = 1; |
1467 |
|
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
1468 |
|
/* output the reference frame */ |
1469 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1470 |
|
output = 1; |
1471 |
|
} |
1472 |
} |
} |
|
#endif |
|
|
frame->length = BitstreamPos(&bs) / 8; |
|
1473 |
|
|
1474 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1475 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1476 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1477 |
|
dec->last_reduced_resolution = reduced_resolution; |
1478 |
|
dec->last_coding_type = coding_type; |
1479 |
|
|
1480 |
#ifdef BFRAMES_DEC |
dec->frames++; |
1481 |
// test if no B_VOP |
seen_something = 1; |
|
if (dec->low_delay || dec->frames == 0) { |
|
|
#endif |
|
|
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
|
|
frame->image, frame->stride, frame->colorspace); |
|
1482 |
|
|
1483 |
#ifdef BFRAMES_DEC |
} else { /* B_VOP */ |
1484 |
|
|
1485 |
|
if (dec->low_delay) { |
1486 |
|
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1487 |
|
dec->low_delay = 1; |
1488 |
|
} |
1489 |
|
|
1490 |
|
if (dec->frames < 2) { |
1491 |
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
1492 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1493 |
|
"broken b-frame, mising ref frames"); |
1494 |
|
} else if (dec->time_pp <= dec->time_bp) { |
1495 |
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1496 |
|
decoded in vfw. */ |
1497 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1498 |
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1499 |
} else { |
} else { |
1500 |
if (dec->frames >= 1) { |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
|
start_timer(); |
|
|
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
|
|
image_output(&dec->refn[0], dec->width, dec->height, |
|
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace); |
|
|
} else if (vop_type == B_VOP) { |
|
|
image_output(&dec->cur, dec->width, dec->height, |
|
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace); |
|
|
} |
|
|
stop_conv_timer(); |
|
1501 |
} |
} |
1502 |
|
|
1503 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type); |
1504 |
|
output = 1; |
1505 |
|
dec->frames++; |
1506 |
} |
} |
|
#endif |
|
1507 |
|
|
1508 |
if (vop_type == I_VOP || vop_type == P_VOP) { |
BitstreamByteAlign(&bs); |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
|
|
image_swap(&dec->cur, &dec->refn[0]); |
|
1509 |
|
|
1510 |
// swap MACROBLOCK |
/* low_delay_default mode: repeat in packed_mode */ |
1511 |
// the Divx will not set the low_delay flage some times |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
1512 |
// so follow code will wrong to not swap at that time |
success = 1; |
1513 |
// this will broken bitstream! so I'm change it, |
goto repeat; |
1514 |
// But that is not the best way! can anyone tell me how |
} |
1515 |
// to do another way? |
|
1516 |
// 18-07-2002 MinChen<chenm001@163.com> |
done : |
1517 |
//if (!dec->low_delay && vop_type == P_VOP) |
|
1518 |
if (vop_type == P_VOP) |
/* low_delay_default mode: if we've gotten here without outputting anything, |
1519 |
mb_swap(&dec->mbs, &dec->last_mbs); |
then output the recently decoded frame, or print an error message */ |
1520 |
|
if (dec->low_delay_default && output == 0) { |
1521 |
|
if (dec->packed_mode && seen_something) { |
1522 |
|
/* output the recently decoded frame */ |
1523 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type); |
1524 |
|
} else { |
1525 |
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1526 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1527 |
|
"warning: nothing to output"); |
1528 |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
1529 |
|
"bframe decoder lag"); |
1530 |
|
|
1531 |
|
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP); |
1532 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1533 |
|
} |
1534 |
} |
} |
1535 |
|
|
1536 |
emms(); |
emms(); |
|
|
|
1537 |
stop_global_timer(); |
stop_global_timer(); |
1538 |
|
|
1539 |
return XVID_ERR_OK; |
return BitstreamPos(&bs) / 8; /* number of bytes consumed */ |
1540 |
} |
} |