1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* This program is an implementation of a part of one or more MPEG-4 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* 2002-2004 Peter Ross <pross@xvid.org> |
|
* to use this software module in hardware or software products are |
|
|
* advised that its use may infringe existing patents or copyrights, and |
|
|
* any such use would be at such party's own risk. The original |
|
|
* developer of this software module and his/her company, and subsequent |
|
|
* editors and their companies, will have no liability for use of this |
|
|
* software or modifications or derivatives thereof. |
|
8 |
* |
* |
9 |
* This program is free software; you can redistribute it and/or modify |
* This program is free software; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
20 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
21 |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
22 |
* |
* |
23 |
*************************************************************************/ |
* $Id: decoder.c,v 1.76 2005-12-24 01:06:20 Isibaar Exp $ |
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
|
* |
|
|
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
|
|
* MinChen <chenm001@163.com> |
|
|
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
|
|
* Fix a little bug for low_delay flage |
|
|
* MinChen <chenm001@163.com> |
|
|
* 28.06.2002 added basic resync support to iframe/pframe_decode() |
|
|
* 22.06.2002 added primative N_VOP support |
|
|
* #define BFRAMES_DEC now enables Minchen's bframe decoder |
|
|
* 08.05.2002 add low_delay support for B_VOP decode |
|
|
* MinChen <chenm001@163.com> |
|
|
* 05.05.2002 fix some B-frame decode problem |
|
|
* 02.05.2002 add B-frame decode support(have some problem); |
|
|
* MinChen <chenm001@163.com> |
|
|
* 22.04.2002 add some B-frame decode support; chenm001 <chenm001@163.com> |
|
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
|
|
* reconstructing blocks, thus artifacts |
|
|
* interlacing speedup - used transfers to re-interlace |
|
|
* interlaced decoding should be as fast as progressive now |
|
|
* 26.03.2002 interlacing support - moved transfers outside decode loop |
|
|
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
|
|
* 22.12.2001 lock based interpolation |
|
|
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
|
* |
|
|
* $Id: decoder.c,v 1.47 2003-02-15 15:22:17 edgomez Exp $ |
|
24 |
* |
* |
25 |
*************************************************************************/ |
****************************************************************************/ |
26 |
|
|
27 |
#include <stdio.h> |
#include <stdio.h> |
28 |
#include <stdlib.h> |
#include <stdlib.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" |
|
#include "image/reduced.h" |
|
49 |
#include "image/font.h" |
#include "image/font.h" |
50 |
|
#include "image/qpel.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" |
#include "motion/motion.h" |
57 |
|
#include "motion/gmc.h" |
58 |
|
|
59 |
#include "image/image.h" |
#include "image/image.h" |
60 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
61 |
|
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
int |
#define DIV2ROUND(n) (((n)>>1)|((n)&1)) |
65 |
|
#define DIV2(n) ((n)>>1) |
66 |
|
#define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) // |
67 |
|
|
68 |
|
static int |
69 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
70 |
{ |
{ |
71 |
/* free existing */ |
/* free existing */ |
|
|
|
72 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
73 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
74 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
77 |
|
|
78 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
79 |
|
|
80 |
if (dec->last_mbs) |
image_null(&dec->cur); |
81 |
|
image_null(&dec->refn[0]); |
82 |
|
image_null(&dec->refn[1]); |
83 |
|
image_null(&dec->tmp); |
84 |
|
image_null(&dec->qtmp); |
85 |
|
image_null(&dec->gmc); |
86 |
|
|
87 |
|
|
88 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
|
if (dec->mbs) |
|
89 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
90 |
|
xvid_free(dec->qscale); |
91 |
|
dec->last_mbs = NULL; |
92 |
|
dec->mbs = NULL; |
93 |
|
dec->qscale = NULL; |
94 |
|
|
95 |
/* realloc */ |
/* realloc */ |
|
|
|
96 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
97 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
98 |
|
|
99 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
100 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
101 |
|
|
102 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if ( image_create(&dec->cur, dec->edged_width, dec->edged_height) |
103 |
xvid_free(dec); |
|| image_create(&dec->refn[0], dec->edged_width, dec->edged_height) |
104 |
return XVID_ERR_MEMORY; |
|| image_create(&dec->refn[1], dec->edged_width, dec->edged_height) /* Support B-frame to reference last 2 frame */ |
105 |
} |
|| image_create(&dec->tmp, dec->edged_width, dec->edged_height) |
106 |
|
|| image_create(&dec->qtmp, dec->edged_width, dec->edged_height) |
107 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|| image_create(&dec->gmc, dec->edged_width, dec->edged_height) ) |
108 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
// add by chenm001 <chenm001@163.com> |
|
|
// for support B-frame to reference last 2 frame |
|
|
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
109 |
|
|
110 |
dec->mbs = |
dec->mbs = |
111 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
112 |
CACHE_LINE); |
CACHE_LINE); |
113 |
if (dec->mbs == NULL) { |
if (dec->mbs == NULL) |
114 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
115 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
116 |
|
|
117 |
// add by chenm001 <chenm001@163.com> |
/* For skip MB flag */ |
|
// for skip MB flag |
|
118 |
dec->last_mbs = |
dec->last_mbs = |
119 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
120 |
CACHE_LINE); |
CACHE_LINE); |
121 |
if (dec->last_mbs == NULL) { |
if (dec->last_mbs == NULL) |
122 |
|
goto memory_error; |
123 |
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
124 |
|
|
125 |
|
/* nothing happens if that fails */ |
126 |
|
dec->qscale = |
127 |
|
xvid_malloc(sizeof(int) * dec->mb_width * dec->mb_height, CACHE_LINE); |
128 |
|
|
129 |
|
if (dec->qscale) |
130 |
|
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
131 |
|
|
132 |
|
return 0; |
133 |
|
|
134 |
|
memory_error: |
135 |
|
/* Most structures were deallocated / nullifieded, so it should be safe */ |
136 |
|
/* decoder_destroy(dec) minus the write_timer */ |
137 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
138 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
139 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
140 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
141 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
142 |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
143 |
|
|
144 |
xvid_free(dec); |
xvid_free(dec); |
145 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
146 |
} |
} |
147 |
|
|
|
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
|
|
|
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
148 |
|
|
149 |
int |
int |
150 |
decoder_create(XVID_DEC_PARAM * param) |
decoder_create(xvid_dec_create_t * create) |
151 |
{ |
{ |
152 |
DECODER *dec; |
DECODER *dec; |
153 |
|
|
154 |
|
if (XVID_VERSION_MAJOR(create->version) != 1) /* v1.x.x */ |
155 |
|
return XVID_ERR_VERSION; |
156 |
|
|
157 |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
158 |
if (dec == NULL) { |
if (dec == NULL) { |
159 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
160 |
} |
} |
161 |
|
|
162 |
memset(dec, 0, sizeof(DECODER)); |
memset(dec, 0, sizeof(DECODER)); |
163 |
|
|
164 |
param->handle = dec; |
dec->mpeg_quant_matrices = xvid_malloc(sizeof(uint16_t) * 64 * 8, CACHE_LINE); |
165 |
|
if (dec->mpeg_quant_matrices == NULL) { |
166 |
|
xvid_free(dec); |
167 |
|
return XVID_ERR_MEMORY; |
168 |
|
} |
169 |
|
|
170 |
|
create->handle = dec; |
171 |
|
|
172 |
dec->width = param->width; |
dec->width = create->width; |
173 |
dec->height = param->height; |
dec->height = create->height; |
174 |
|
|
175 |
image_null(&dec->cur); |
image_null(&dec->cur); |
176 |
image_null(&dec->refn[0]); |
image_null(&dec->refn[0]); |
181 |
/* image based GMC */ |
/* image based GMC */ |
182 |
image_null(&dec->gmc); |
image_null(&dec->gmc); |
183 |
|
|
|
|
|
184 |
dec->mbs = NULL; |
dec->mbs = NULL; |
185 |
dec->last_mbs = NULL; |
dec->last_mbs = NULL; |
186 |
|
dec->qscale = NULL; |
187 |
|
|
188 |
init_timer(); |
init_timer(); |
189 |
|
init_postproc(&dec->postproc); |
190 |
|
init_mpeg_matrix(dec->mpeg_quant_matrices); |
191 |
|
|
192 |
// add by chenm001 <chenm001@163.com> |
/* For B-frame support (used to save reference frame's time */ |
|
// for support B-frame to save reference frame's time |
|
193 |
dec->frames = 0; |
dec->frames = 0; |
194 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
195 |
dec->low_delay = 0; |
dec->low_delay = 0; |
196 |
dec->packed_mode = 0; |
dec->packed_mode = 0; |
197 |
|
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
198 |
|
|
199 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
200 |
|
|
201 |
if (dec->fixed_dimensions) |
if (dec->fixed_dimensions) |
202 |
return decoder_resize(dec); |
return decoder_resize(dec); |
203 |
else |
else |
204 |
return XVID_ERR_OK; |
return 0; |
205 |
} |
} |
206 |
|
|
207 |
|
|
210 |
{ |
{ |
211 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
212 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
213 |
|
xvid_free(dec->qscale); |
214 |
|
|
215 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); /* image based GMC */ |
/* image based GMC */ |
216 |
|
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
217 |
|
|
218 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
219 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
220 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
221 |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
222 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
223 |
|
xvid_free(dec->mpeg_quant_matrices); |
224 |
xvid_free(dec); |
xvid_free(dec); |
225 |
|
|
226 |
write_timer(); |
write_timer(); |
227 |
return XVID_ERR_OK; |
return 0; |
228 |
} |
} |
229 |
|
|
|
|
|
|
|
|
230 |
static const int32_t dquant_table[4] = { |
static const int32_t dquant_table[4] = { |
231 |
-1, -2, 1, 2 |
-1, -2, 1, 2 |
232 |
}; |
}; |
233 |
|
|
234 |
|
/* decode an intra macroblock */ |
235 |
|
static void |
|
|
|
|
// decode an intra macroblock |
|
|
|
|
|
void |
|
236 |
decoder_mbintra(DECODER * dec, |
decoder_mbintra(DECODER * dec, |
237 |
MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
238 |
const uint32_t x_pos, |
const uint32_t x_pos, |
242 |
Bitstream * bs, |
Bitstream * bs, |
243 |
const uint32_t quant, |
const uint32_t quant, |
244 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
245 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
246 |
{ |
{ |
247 |
|
|
248 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
255 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
256 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
257 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
}else{ |
|
258 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
259 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
260 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
261 |
|
|
262 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
263 |
|
|
264 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
265 |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
282 |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
dc_dif = dc_size ? get_dc_dif(bs, dc_size) : 0; |
283 |
|
|
284 |
if (dc_size > 8) { |
if (dc_size > 8) { |
285 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); /* marker */ |
286 |
} |
} |
287 |
|
|
288 |
block[i * 64 + 0] = dc_dif; |
block[i * 64 + 0] = dc_dif; |
289 |
start_coeff = 1; |
start_coeff = 1; |
290 |
|
|
291 |
DPRINTF(DPRINTF_COEFF,"block[0] %i", dc_dif); |
DPRINTF(XVID_DEBUG_COEFF,"block[0] %i\n", dc_dif); |
292 |
} else { |
} else { |
293 |
start_coeff = 0; |
start_coeff = 0; |
294 |
} |
} |
295 |
|
|
296 |
start_timer(); |
start_timer(); |
297 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) /* coded */ |
298 |
{ |
{ |
299 |
int direction = dec->alternate_vertical_scan ? |
int direction = dec->alternate_vertical_scan ? |
300 |
2 : pMB->acpred_directions[i]; |
2 : pMB->acpred_directions[i]; |
304 |
stop_coding_timer(); |
stop_coding_timer(); |
305 |
|
|
306 |
start_timer(); |
start_timer(); |
307 |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
308 |
stop_prediction_timer(); |
stop_prediction_timer(); |
309 |
|
|
310 |
start_timer(); |
start_timer(); |
311 |
if (dec->quant_type == 0) { |
if (dec->quant_type == 0) { |
312 |
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); |
313 |
} else { |
} else { |
314 |
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); |
315 |
} |
} |
316 |
stop_iquant_timer(); |
stop_iquant_timer(); |
317 |
|
|
318 |
start_timer(); |
start_timer(); |
319 |
idct(&data[i * 64]); |
idct((short * const)&data[i * 64]); |
320 |
stop_idct_timer(); |
stop_idct_timer(); |
321 |
|
|
322 |
} |
} |
327 |
} |
} |
328 |
|
|
329 |
start_timer(); |
start_timer(); |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
next_block*=2; |
|
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
}else{ |
|
330 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
331 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
332 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
333 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
334 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
335 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
336 |
stop_transfer_timer(); |
stop_transfer_timer(); |
337 |
} |
} |
338 |
|
|
339 |
|
static void |
340 |
|
decoder_mb_decode(DECODER * dec, |
341 |
|
const uint32_t cbp, |
342 |
|
Bitstream * bs, |
343 |
|
uint8_t * pY_Cur, |
344 |
|
uint8_t * pU_Cur, |
345 |
|
uint8_t * pV_Cur, |
346 |
|
const MACROBLOCK * pMB) |
347 |
|
{ |
348 |
|
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
349 |
|
|
350 |
|
int stride = dec->edged_width; |
351 |
|
int i; |
352 |
|
const uint32_t iQuant = pMB->quant; |
353 |
|
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
354 |
|
typedef void (*get_inter_block_function_t)( |
355 |
|
Bitstream * bs, |
356 |
|
int16_t * block, |
357 |
|
int direction, |
358 |
|
const int quant, |
359 |
|
const uint16_t *matrix); |
360 |
|
typedef void (*add_residual_function_t)( |
361 |
|
uint8_t *predicted_block, |
362 |
|
const int16_t *residual, |
363 |
|
int stride); |
364 |
|
|
365 |
|
const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) |
366 |
|
? (get_inter_block_function_t)get_inter_block_h263 |
367 |
|
: (get_inter_block_function_t)get_inter_block_mpeg; |
368 |
|
|
369 |
|
uint8_t *dst[6]; |
370 |
|
int strides[6]; |
371 |
|
|
372 |
|
|
373 |
|
if (dec->interlacing && pMB->field_dct) { |
374 |
|
dst[0] = pY_Cur; |
375 |
|
dst[1] = pY_Cur + 8; |
376 |
|
dst[2] = pY_Cur + stride; |
377 |
|
dst[3] = dst[2] + 8; |
378 |
|
dst[4] = pU_Cur; |
379 |
|
dst[5] = pV_Cur; |
380 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride*2; |
381 |
|
strides[4] = stride/2; |
382 |
|
strides[5] = stride/2; |
383 |
|
} else { |
384 |
|
dst[0] = pY_Cur; |
385 |
|
dst[1] = pY_Cur + 8; |
386 |
|
dst[2] = pY_Cur + 8*stride; |
387 |
|
dst[3] = dst[2] + 8; |
388 |
|
dst[4] = pU_Cur; |
389 |
|
dst[5] = pV_Cur; |
390 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
391 |
|
strides[4] = stride/2; |
392 |
|
strides[5] = stride/2; |
393 |
|
} |
394 |
|
|
395 |
|
for (i = 0; i < 6; i++) { |
396 |
|
/* Process only coded blocks */ |
397 |
|
if (cbp & (1 << (5 - i))) { |
398 |
|
|
399 |
|
/* Clear the block */ |
400 |
|
memset(&data[0], 0, 64*sizeof(int16_t)); |
401 |
|
|
402 |
|
/* Decode coeffs and dequantize on the fly */ |
403 |
|
start_timer(); |
404 |
|
get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices)); |
405 |
|
stop_coding_timer(); |
406 |
|
|
407 |
|
/* iDCT */ |
408 |
|
start_timer(); |
409 |
|
idct((short * const)&data[0]); |
410 |
|
stop_idct_timer(); |
411 |
|
|
412 |
|
/* Add this residual to the predicted block */ |
413 |
|
start_timer(); |
414 |
|
transfer_16to8add(dst[i], &data[0], strides[i]); |
415 |
|
stop_transfer_timer(); |
416 |
|
} |
417 |
|
} |
418 |
|
} |
419 |
|
|
420 |
// decode an inter macroblock |
static void __inline |
421 |
|
validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec) |
422 |
|
{ |
423 |
|
/* clip a vector to valid range |
424 |
|
prevents crashes if bitstream is broken |
425 |
|
*/ |
426 |
|
int shift = 5 + dec->quarterpel; |
427 |
|
int xborder_high = (int)(dec->mb_width - x_pos) << shift; |
428 |
|
int xborder_low = (-(int)x_pos-1) << shift; |
429 |
|
int yborder_high = (int)(dec->mb_height - y_pos) << shift; |
430 |
|
int yborder_low = (-(int)y_pos-1) << shift; |
431 |
|
|
432 |
|
#define CHECK_MV(mv) \ |
433 |
|
do { \ |
434 |
|
if ((mv).x > xborder_high) { \ |
435 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \ |
436 |
|
(mv).x = xborder_high; \ |
437 |
|
} else if ((mv).x < xborder_low) { \ |
438 |
|
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \ |
439 |
|
(mv).x = xborder_low; \ |
440 |
|
} \ |
441 |
|
if ((mv).y > yborder_high) { \ |
442 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \ |
443 |
|
(mv).y = yborder_high; \ |
444 |
|
} else if ((mv).y < yborder_low) { \ |
445 |
|
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \ |
446 |
|
(mv).y = yborder_low; \ |
447 |
|
} \ |
448 |
|
} while (0) |
449 |
|
|
450 |
|
CHECK_MV(mv[0]); |
451 |
|
CHECK_MV(mv[1]); |
452 |
|
CHECK_MV(mv[2]); |
453 |
|
CHECK_MV(mv[3]); |
454 |
|
} |
455 |
|
|
456 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
457 |
|
* So we try to be backward compatible to avoid artifacts */ |
458 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
459 |
|
|
460 |
void |
/* decode an inter macroblock */ |
461 |
|
static void |
462 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
463 |
const MACROBLOCK * pMB, |
const MACROBLOCK * pMB, |
464 |
const uint32_t x_pos, |
const uint32_t x_pos, |
465 |
const uint32_t y_pos, |
const uint32_t y_pos, |
|
const uint32_t fcode, |
|
466 |
const uint32_t cbp, |
const uint32_t cbp, |
467 |
Bitstream * bs, |
Bitstream * bs, |
|
const uint32_t quant, |
|
468 |
const uint32_t rounding, |
const uint32_t rounding, |
469 |
const int reduced_resolution) |
const int ref, |
470 |
|
const int bvop) |
471 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
472 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
473 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
474 |
uint32_t i; |
uint32_t i; |
475 |
uint32_t iQuant = pMB->quant; |
|
476 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
477 |
|
|
478 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
479 |
VECTOR mv[4]; /* local copy of mvs */ |
VECTOR mv[4]; /* local copy of mvs */ |
480 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
for (i = 0; i < 4; i++) { |
|
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
|
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
|
|
} |
|
|
} else { |
|
481 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
482 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
483 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
484 |
for (i = 0; i < 4; i++) |
for (i = 0; i < 4; i++) |
485 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
|
} |
|
|
|
|
|
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
|
|
|
|
|
uv_dx = mv[0].x / (1 + dec->quarterpel); |
|
|
uv_dy = mv[0].y / (1 + dec->quarterpel); |
|
486 |
|
|
487 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
validate_vector(mv, x_pos, y_pos, dec); |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
488 |
|
|
489 |
start_timer(); |
start_timer(); |
|
if (reduced_resolution) |
|
|
{ |
|
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u, 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v, 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
490 |
|
|
491 |
} |
if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
492 |
else |
|
493 |
{ |
uv_dx = mv[0].x; |
494 |
|
uv_dy = mv[0].y; |
495 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
496 |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
497 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
uv_dx = (uv_dx>>1) | (uv_dx&1); |
498 |
mv[0].x, mv[0].y, stride, rounding); |
uv_dy = (uv_dy>>1) | (uv_dy&1); |
499 |
} |
} |
500 |
else { |
else { |
501 |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
uv_dx /= 2; |
502 |
mv[0].x, mv[0].y, stride, rounding); |
uv_dy /= 2; |
503 |
} |
} |
|
|
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
504 |
} |
} |
505 |
stop_comp_timer(); |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
506 |
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
} else { /* MODE_INTER4V */ |
|
|
int sum; |
|
507 |
|
|
508 |
if(dec->quarterpel) |
if(dec->quarterpel) |
509 |
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
510 |
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
511 |
|
mv[0].x, mv[0].y, stride, rounding); |
512 |
else |
else |
513 |
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
interpolate16x16_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
514 |
|
mv[0].x, mv[0].y, stride, rounding); |
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
515 |
|
|
516 |
if(dec->quarterpel) |
} else { /* MODE_INTER4V */ |
|
sum = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
|
|
else |
|
|
sum = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
|
517 |
|
|
518 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
if(dec->quarterpel) { |
519 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
520 |
|
int z; |
521 |
|
uv_dx = 0; uv_dy = 0; |
522 |
|
for (z = 0; z < 4; z++) { |
523 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
524 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
525 |
|
} |
526 |
|
} |
527 |
|
else { |
528 |
|
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
529 |
|
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
530 |
|
} |
531 |
|
} else { |
532 |
|
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
533 |
|
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
534 |
|
} |
535 |
|
|
536 |
start_timer(); |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
537 |
if (reduced_resolution) |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
|
{ |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
|
|
mv[1].x, mv[1].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
|
|
mv[2].x, mv[2].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
538 |
|
|
|
// set_block(pY_Cur, stride, 32, 32, 127); |
|
|
} |
|
|
else |
|
|
{ |
|
539 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
540 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
541 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
549 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
550 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
551 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
552 |
} |
} else { |
|
else { |
|
553 |
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, |
554 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
555 |
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, |
559 |
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, |
560 |
mv[3].x, mv[3].y, stride, rounding); |
mv[3].x, mv[3].y, stride, rounding); |
561 |
} |
} |
562 |
|
} |
563 |
|
|
564 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u , 8 * x_pos, 8 * y_pos, |
/* chroma */ |
565 |
|
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
566 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
567 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v , 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
568 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
569 |
} |
|
570 |
stop_comp_timer(); |
stop_comp_timer(); |
|
} |
|
571 |
|
|
572 |
for (i = 0; i < 6; i++) { |
if (cbp) |
573 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
574 |
|
} |
575 |
|
|
576 |
if (cbp & (1 << (5 - i))) // coded |
/* decode an inter macroblock in field mode */ |
577 |
|
static void |
578 |
|
decoder_mbinter_field(DECODER * dec, |
579 |
|
const MACROBLOCK * pMB, |
580 |
|
const uint32_t x_pos, |
581 |
|
const uint32_t y_pos, |
582 |
|
const uint32_t cbp, |
583 |
|
Bitstream * bs, |
584 |
|
const uint32_t rounding, |
585 |
|
const int ref, |
586 |
|
const int bvop) |
587 |
{ |
{ |
588 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
uint32_t stride = dec->edged_width; |
589 |
|
uint32_t stride2 = stride / 2; |
590 |
|
|
591 |
start_timer(); |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
get_inter_block(bs, &block[i * 64], direction); |
|
|
stop_coding_timer(); |
|
592 |
|
|
593 |
start_timer(); |
int uvtop_dx, uvtop_dy; |
594 |
if (dec->quant_type == 0) { |
int uvbot_dx, uvbot_dy; |
595 |
dequant_inter(&data[i * 64], &block[i * 64], iQuant); |
VECTOR mv[4]; /* local copy of mvs */ |
|
} else { |
|
|
dequant4_inter(&data[i * 64], &block[i * 64], iQuant); |
|
|
} |
|
|
stop_iquant_timer(); |
|
596 |
|
|
597 |
start_timer(); |
/* Get pointer to memory areas */ |
598 |
idct(&data[i * 64]); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
599 |
stop_idct_timer(); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
600 |
} |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
601 |
|
|
602 |
if (dec->interlacing && pMB->field_dct) { |
mv[0] = pMB->mvs[0]; |
603 |
next_block = stride; |
mv[1] = pMB->mvs[1]; |
604 |
stride *= 2; |
memset(&mv[2],0,2*sizeof(VECTOR)); |
605 |
} |
|
606 |
|
validate_vector(mv, x_pos, y_pos, dec); |
607 |
|
|
608 |
start_timer(); |
start_timer(); |
609 |
if (reduced_resolution) |
|
610 |
|
if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
611 |
|
{ |
612 |
|
/* Prepare top field vector */ |
613 |
|
uvtop_dx = DIV2ROUND(mv[0].x); |
614 |
|
uvtop_dy = DIV2ROUND(mv[0].y); |
615 |
|
|
616 |
|
/* Prepare bottom field vector */ |
617 |
|
uvbot_dx = DIV2ROUND(mv[1].x); |
618 |
|
uvbot_dy = DIV2ROUND(mv[1].y); |
619 |
|
|
620 |
|
if(dec->quarterpel) |
621 |
{ |
{ |
622 |
if (cbp & 32) |
/* NOT supported */ |
|
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
if (cbp & 16) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
if (cbp & 8) |
|
|
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
if (cbp & 4) |
|
|
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
if (cbp & 2) |
|
|
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
if (cbp & 1) |
|
|
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
623 |
} |
} |
624 |
else |
else |
625 |
{ |
{ |
626 |
if (cbp & 32) |
/* Interpolate top field left part(we use double stride for every 2nd line) */ |
627 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
628 |
if (cbp & 16) |
16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
629 |
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
/* top field right part */ |
630 |
if (cbp & 8) |
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
631 |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
632 |
if (cbp & 4) |
|
633 |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
/* Interpolate bottom field left part(we use double stride for every 2nd line) */ |
634 |
if (cbp & 2) |
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
635 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
636 |
if (cbp & 1) |
/* Bottom field right part */ |
637 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
638 |
|
16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
639 |
|
|
640 |
|
/* Interpolate field1 U */ |
641 |
|
interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2, |
642 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
643 |
|
|
644 |
|
/* Interpolate field1 V */ |
645 |
|
interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2, |
646 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
647 |
|
|
648 |
|
/* Interpolate field2 U */ |
649 |
|
interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2, |
650 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
651 |
|
|
652 |
|
/* Interpolate field2 V */ |
653 |
|
interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2, |
654 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
655 |
} |
} |
|
stop_transfer_timer(); |
|
656 |
} |
} |
657 |
|
else |
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
|
658 |
{ |
{ |
659 |
int length = 1 << (fcode+4); |
/* We don't expect 4 motion vectors in interlaced mode */ |
660 |
|
} |
661 |
|
|
662 |
// if (quarterpel) value *= 2; |
stop_comp_timer(); |
663 |
|
|
664 |
if (value < -length) |
/* Must add error correction? */ |
665 |
return -length; |
if(cbp) |
666 |
else if (value >= length) |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
|
return length-1; |
|
|
else return value; |
|
667 |
} |
} |
668 |
|
|
|
|
|
669 |
static void |
static void |
670 |
decoder_mbgmc(DECODER * dec, |
decoder_mbgmc(DECODER * dec, |
671 |
MACROBLOCK * const pMB, |
MACROBLOCK * const pMB, |
674 |
const uint32_t fcode, |
const uint32_t fcode, |
675 |
const uint32_t cbp, |
const uint32_t cbp, |
676 |
Bitstream * bs, |
Bitstream * bs, |
677 |
const uint32_t quant, |
const uint32_t rounding) |
|
const uint32_t rounding, |
|
|
const int reduced_resolution) /* no reduced res support */ |
|
678 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
679 |
const uint32_t stride = dec->edged_width; |
const uint32_t stride = dec->edged_width; |
680 |
const uint32_t stride2 = stride / 2; |
const uint32_t stride2 = stride / 2; |
681 |
const uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
uint32_t i; |
|
|
const uint32_t iQuant = pMB->quant; |
|
682 |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
uint8_t *const pY_Cur=dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
683 |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
uint8_t *const pU_Cur=dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
684 |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
uint8_t *const pV_Cur=dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
685 |
|
|
686 |
|
NEW_GMC_DATA * gmc_data = &dec->new_gmc_data; |
687 |
|
|
688 |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
689 |
|
|
690 |
start_timer(); |
start_timer(); |
691 |
|
|
692 |
/* this is where the calculations are done */ |
/* this is where the calculations are done */ |
693 |
|
|
694 |
{ |
gmc_data->predict_16x16(gmc_data, |
695 |
pMB->amv = generate_GMCimageMB(&dec->gmc_data, &dec->refn[0], x_pos, y_pos, |
dec->cur.y + y_pos*16*stride + x_pos*16, dec->refn[0].y, |
696 |
stride, stride2, dec->quarterpel, rounding, &dec->cur); |
stride, stride, x_pos, y_pos, rounding); |
697 |
|
|
698 |
|
gmc_data->predict_8x8(gmc_data, |
699 |
|
dec->cur.u + y_pos*8*stride2 + x_pos*8, dec->refn[0].u, |
700 |
|
dec->cur.v + y_pos*8*stride2 + x_pos*8, dec->refn[0].v, |
701 |
|
stride2, stride2, x_pos, y_pos, rounding); |
702 |
|
|
703 |
|
gmc_data->get_average_mv(gmc_data, &pMB->amv, x_pos, y_pos, dec->quarterpel); |
704 |
|
|
705 |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
pMB->amv.x = gmc_sanitize(pMB->amv.x, dec->quarterpel, fcode); |
706 |
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
pMB->amv.y = gmc_sanitize(pMB->amv.y, dec->quarterpel, fcode); |
|
} |
|
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
|
|
|
|
|
/* transfer16x16_copy(pY_Cur, dec->gmc.y + (y_pos << 4)*stride + (x_pos << 4), stride); |
|
|
transfer8x8_copy(pU_Cur, dec->gmc.u + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
|
|
transfer8x8_copy(pV_Cur, dec->gmc.v + (y_pos << 3)*stride2 + (x_pos << 3), stride2); |
|
|
*/ |
|
707 |
|
|
708 |
|
pMB->mvs[0] = pMB->mvs[1] = pMB->mvs[2] = pMB->mvs[3] = pMB->amv; |
709 |
|
|
710 |
stop_transfer_timer(); |
stop_transfer_timer(); |
711 |
|
|
712 |
if (!cbp) return; |
if (cbp) |
713 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
714 |
|
|
715 |
for (i = 0; i < 6; i++) { |
} |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
716 |
|
|
|
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], direction); |
|
|
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(); |
|
|
} |
|
|
} |
|
|
|
|
|
/* interlace + GMC is this possible ??? */ |
|
|
/* 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(); |
|
|
} |
|
717 |
|
|
718 |
|
static void |
|
void |
|
719 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
720 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
721 |
int quant, |
int quant, |
722 |
int intra_dc_threshold) |
int intra_dc_threshold) |
723 |
{ |
{ |
724 |
uint32_t bound; |
uint32_t bound; |
725 |
uint32_t x, y; |
uint32_t x, y; |
726 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
727 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
728 |
|
|
729 |
bound = 0; |
bound = 0; |
730 |
|
|
749 |
} |
} |
750 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
751 |
|
|
752 |
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)); |
753 |
|
|
754 |
mcbpc = get_mcbpc_intra(bs); |
mcbpc = get_mcbpc_intra(bs); |
755 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
776 |
|
|
777 |
if (dec->interlacing) { |
if (dec->interlacing) { |
778 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
779 |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"deci: field_dct: %i\n", mb->field_dct); |
780 |
} |
} |
781 |
|
|
782 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
783 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
784 |
|
|
785 |
} |
} |
786 |
if(dec->out_frm) |
if(dec->out_frm) |
790 |
} |
} |
791 |
|
|
792 |
|
|
793 |
void |
static void |
794 |
get_motion_vector(DECODER * dec, |
get_motion_vector(DECODER * dec, |
795 |
Bitstream * bs, |
Bitstream * bs, |
796 |
int x, |
int x, |
801 |
const int bound) |
const int bound) |
802 |
{ |
{ |
803 |
|
|
804 |
int scale_fac = 1 << (fcode - 1); |
const int scale_fac = 1 << (fcode - 1); |
805 |
int high = (32 * scale_fac) - 1; |
const int high = (32 * scale_fac) - 1; |
806 |
int low = ((-32) * scale_fac); |
const int low = ((-32) * scale_fac); |
807 |
int range = (64 * scale_fac); |
const int range = (64 * scale_fac); |
808 |
|
|
809 |
VECTOR pmv; |
const VECTOR pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
810 |
VECTOR mv; |
VECTOR mv; |
811 |
|
|
|
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
|
|
|
|
812 |
mv.x = get_mv(bs, fcode); |
mv.x = get_mv(bs, fcode); |
813 |
mv.y = get_mv(bs, fcode); |
mv.y = get_mv(bs, fcode); |
814 |
|
|
815 |
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); |
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); |
816 |
|
|
817 |
mv.x += pmv.x; |
mv.x += pmv.x; |
818 |
mv.y += pmv.y; |
mv.y += pmv.y; |
833 |
ret_mv->y = mv.y; |
ret_mv->y = mv.y; |
834 |
} |
} |
835 |
|
|
836 |
|
/* We use this when decoder runs interlaced -> different prediction */ |
837 |
|
|
838 |
|
static void get_motion_vector_interlaced(DECODER * dec, |
839 |
|
Bitstream * bs, |
840 |
|
int x, |
841 |
|
int y, |
842 |
|
int k, |
843 |
|
MACROBLOCK *pMB, |
844 |
|
int fcode, |
845 |
|
const int bound) |
846 |
|
{ |
847 |
|
const int scale_fac = 1 << (fcode - 1); |
848 |
|
const int high = (32 * scale_fac) - 1; |
849 |
|
const int low = ((-32) * scale_fac); |
850 |
|
const int range = (64 * scale_fac); |
851 |
|
|
852 |
|
/* Get interlaced prediction */ |
853 |
|
const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k); |
854 |
|
VECTOR mv,mvf1,mvf2; |
855 |
|
|
856 |
|
if(!pMB->field_pred) |
857 |
|
{ |
858 |
|
mv.x = get_mv(bs,fcode); |
859 |
|
mv.y = get_mv(bs,fcode); |
860 |
|
|
861 |
|
mv.x += pmv.x; |
862 |
|
mv.y += pmv.y; |
863 |
|
|
864 |
|
if(mv.x<low) { |
865 |
|
mv.x += range; |
866 |
|
} else if (mv.x>high) { |
867 |
|
mv.x-=range; |
868 |
|
} |
869 |
|
|
870 |
|
if (mv.y < low) { |
871 |
|
mv.y += range; |
872 |
|
} else if (mv.y > high) { |
873 |
|
mv.y -= range; |
874 |
|
} |
875 |
|
|
876 |
|
pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv; |
877 |
|
} |
878 |
|
else |
879 |
|
{ |
880 |
|
mvf1.x = get_mv(bs, fcode); |
881 |
|
mvf1.y = get_mv(bs, fcode); |
882 |
|
|
883 |
|
mvf1.x += pmv.x; |
884 |
|
mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */ |
885 |
|
|
886 |
|
if (mvf1.x < low) { |
887 |
|
mvf1.x += range; |
888 |
|
} else if (mvf1.x > high) { |
889 |
|
mvf1.x -= range; |
890 |
|
} |
891 |
|
|
892 |
|
if (mvf1.y < low) { |
893 |
|
mvf1.y += range; |
894 |
|
} else if (mvf1.y > high) { |
895 |
|
mvf1.y -= range; |
896 |
|
} |
897 |
|
|
898 |
|
mvf2.x = get_mv(bs, fcode); |
899 |
|
mvf2.y = get_mv(bs, fcode); |
900 |
|
|
901 |
|
mvf2.x += pmv.x; |
902 |
|
mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */ |
903 |
|
|
904 |
|
if (mvf2.x < low) { |
905 |
|
mvf2.x += range; |
906 |
|
} else if (mvf2.x > high) { |
907 |
|
mvf2.x -= range; |
908 |
|
} |
909 |
|
|
910 |
|
if (mvf2.y < low) { |
911 |
|
mvf2.y += range; |
912 |
|
} else if (mvf2.y > high) { |
913 |
|
mvf2.y -= range; |
914 |
|
} |
915 |
|
|
916 |
|
pMB->mvs[0]=mvf1; |
917 |
|
pMB->mvs[1]=mvf2; |
918 |
|
pMB->mvs[2].x=pMB->mvs[3].x=0; |
919 |
|
pMB->mvs[2].y=pMB->mvs[3].y=0; |
920 |
|
|
921 |
|
/* Calculate average for as it is field predicted */ |
922 |
|
pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x); |
923 |
|
pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y); |
924 |
|
} |
925 |
|
} |
926 |
|
|
927 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
928 |
void |
static void |
929 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
930 |
Bitstream * bs, |
Bitstream * bs, |
931 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
932 |
int quant, |
int quant, |
933 |
int fcode, |
int fcode, |
934 |
int intra_dc_threshold, |
int intra_dc_threshold, |
935 |
const WARPPOINTS *const gmc_warp) |
const WARPPOINTS *const gmc_warp) |
936 |
{ |
{ |
|
|
|
937 |
uint32_t x, y; |
uint32_t x, y; |
938 |
uint32_t bound; |
uint32_t bound; |
939 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
940 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
941 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
942 |
|
|
943 |
|
if (!dec->is_edged[0]) { |
944 |
start_timer(); |
start_timer(); |
945 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
946 |
dec->width, dec->height); |
dec->width, dec->height, dec->bs_version); |
947 |
|
dec->is_edged[0] = 1; |
948 |
stop_edges_timer(); |
stop_edges_timer(); |
|
|
|
|
if (gmc_warp) |
|
|
{ |
|
|
|
|
|
// accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 |
|
|
if ( (dec->sprite_warping_accuracy != 3) || (dec->sprite_warping_points != 2) ) |
|
|
{ |
|
|
fprintf(stderr,"Wrong GMC parameters acc=%d(-> 1/%d), %d!!!\n", |
|
|
dec->sprite_warping_accuracy,(2<<dec->sprite_warping_accuracy), |
|
|
dec->sprite_warping_points); |
|
949 |
} |
} |
950 |
|
|
951 |
|
if (gmc_warp) { |
952 |
|
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
953 |
generate_GMCparameters( dec->sprite_warping_points, |
generate_GMCparameters( dec->sprite_warping_points, |
954 |
(2 << dec->sprite_warping_accuracy), gmc_warp, |
dec->sprite_warping_accuracy, gmc_warp, |
955 |
dec->width, dec->height, &dec->gmc_data); |
dec->width, dec->height, &dec->new_gmc_data); |
956 |
|
|
957 |
/* image warping is done block-based in decoder_mbgmc(), now */ |
/* image warping is done block-based in decoder_mbgmc(), now */ |
|
/* |
|
|
generate_GMCimage(&dec->gmc_data, &dec->refn[0], |
|
|
mb_width, mb_height, |
|
|
dec->edged_width, dec->edged_width/2, |
|
|
fcode, dec->quarterpel, 0, |
|
|
rounding, dec->mbs, &dec->gmc); |
|
|
*/ |
|
958 |
} |
} |
959 |
|
|
960 |
bound = 0; |
bound = 0; |
964 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < mb_width; x++) { |
965 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
966 |
|
|
967 |
// skip stuffing |
/* skip stuffing */ |
968 |
while (BitstreamShowBits(bs, 10) == 1) |
while (BitstreamShowBits(bs, 10) == 1) |
969 |
BitstreamSkip(bs, 10); |
BitstreamSkip(bs, 10); |
970 |
|
|
971 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) { |
|
{ |
|
972 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, dec, fcode - 1, |
973 |
&quant, &fcode, NULL, &intra_dc_threshold); |
&quant, &fcode, NULL, &intra_dc_threshold); |
974 |
x = bound % mb_width; |
x = bound % mb_width; |
976 |
} |
} |
977 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
978 |
|
|
979 |
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)); |
980 |
|
|
981 |
//if (!(dec->mb_skip[y*dec->mb_width + x]=BitstreamGetBit(bs))) // not_coded |
if (!(BitstreamGetBit(bs))) { /* block _is_ coded */ |
982 |
if (!(BitstreamGetBit(bs))) // block _is_ coded |
uint32_t mcbpc, cbpc, cbpy, cbp; |
983 |
{ |
uint32_t intra, acpred_flag = 0; |
984 |
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; |
|
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
|
985 |
|
|
986 |
cp_mb++; |
cp_mb++; |
987 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
988 |
mb->mode = mcbpc & 7; |
mb->mode = mcbpc & 7; |
989 |
cbpc = (mcbpc >> 4); |
cbpc = (mcbpc >> 4); |
990 |
|
|
991 |
DPRINTF(DPRINTF_MB, "mode %i", mb->mode); |
DPRINTF(XVID_DEBUG_MB, "mode %i\n", mb->mode); |
992 |
DPRINTF(DPRINTF_MB, "cbpc %i", cbpc); |
DPRINTF(XVID_DEBUG_MB, "cbpc %i\n", cbpc); |
|
acpred_flag = 0; |
|
993 |
|
|
994 |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
intra = (mb->mode == MODE_INTRA || mb->mode == MODE_INTRA_Q); |
995 |
|
|
|
if (intra) { |
|
|
acpred_flag = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
996 |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
if (gmc_warp && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
{ |
|
997 |
mcsel = BitstreamGetBit(bs); |
mcsel = BitstreamGetBit(bs); |
998 |
} |
else if (intra) |
999 |
|
acpred_flag = BitstreamGetBit(bs); |
1000 |
|
|
1001 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
1002 |
DPRINTF(DPRINTF_MB, "cbpy %i mcsel %i ", cbpy,mcsel); |
DPRINTF(XVID_DEBUG_MB, "cbpy %i mcsel %i \n", cbpy,mcsel); |
1003 |
|
|
1004 |
cbp = (cbpy << 2) | cbpc; |
cbp = (cbpy << 2) | cbpc; |
1005 |
|
|
1006 |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
if (mb->mode == MODE_INTER_Q || mb->mode == MODE_INTRA_Q) { |
1007 |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
int dquant = dquant_table[BitstreamGetBits(bs, 2)]; |
1008 |
DPRINTF(DPRINTF_MB, "dquant %i", dquant); |
DPRINTF(XVID_DEBUG_MB, "dquant %i\n", dquant); |
1009 |
quant += dquant; |
quant += dquant; |
1010 |
if (quant > 31) { |
if (quant > 31) { |
1011 |
quant = 31; |
quant = 31; |
1012 |
} else if (quant < 1) { |
} else if (quant < 1) { |
1013 |
quant = 1; |
quant = 1; |
1014 |
} |
} |
1015 |
DPRINTF(DPRINTF_MB, "quant %i", quant); |
DPRINTF(XVID_DEBUG_MB, "quant %i\n", quant); |
1016 |
} |
} |
1017 |
mb->quant = quant; |
mb->quant = quant; |
1018 |
|
|
1019 |
|
mb->field_pred=0; |
1020 |
if (dec->interlacing) { |
if (dec->interlacing) { |
1021 |
if (cbp || intra) { |
if (cbp || intra) { |
1022 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
1023 |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1024 |
} |
} |
1025 |
|
|
1026 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if ((mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) && !mcsel) { |
1027 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
1028 |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1029 |
|
|
1030 |
if (mb->field_pred) { |
if (mb->field_pred) { |
1031 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
1032 |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1033 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
1034 |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1035 |
} |
} |
1036 |
} |
} |
1037 |
} |
} |
1038 |
|
|
1039 |
if (mcsel) { |
if (mcsel) { |
1040 |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, quant, |
decoder_mbgmc(dec, mb, x, y, fcode, cbp, bs, rounding); |
|
rounding, reduced_resolution); |
|
1041 |
continue; |
continue; |
1042 |
|
|
1043 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
1044 |
|
|
1045 |
if (dec->interlacing && mb->field_pred) { |
if(dec->interlacing) { |
1046 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
/* Get motion vectors interlaced, field_pred is handled there */ |
1047 |
fcode, bound); |
get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound); |
|
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
|
|
fcode, bound); |
|
1048 |
} else { |
} else { |
1049 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
|
fcode, bound); |
|
1050 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1051 |
} |
} |
1052 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
1053 |
|
/* interlaced missing here */ |
1054 |
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); |
1055 |
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); |
1056 |
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); |
1057 |
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); |
1058 |
} else // MODE_INTRA, MODE_INTRA_Q |
} else { /* MODE_INTRA, MODE_INTRA_Q */ |
1059 |
{ |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
1060 |
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; |
|
1061 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
1062 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
1063 |
continue; |
continue; |
1064 |
} |
} |
1065 |
|
|
1066 |
decoder_mbinter(dec, mb, x, y, fcode, cbp, bs, quant, |
/* See how to decode */ |
1067 |
rounding, reduced_resolution); |
if(!mb->field_pred) |
1068 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1069 |
|
else |
1070 |
|
decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1071 |
|
|
1072 |
} |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
|
else if (gmc_warp) /* a not coded S(GMC)-VOP macroblock */ |
|
|
{ |
|
1073 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
1074 |
|
mb->quant = quant; |
1075 |
start_timer(); |
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, rounding); |
|
|
|
|
decoder_mbgmc(dec, mb, x, y, fcode, 0x00, bs, quant, |
|
|
rounding, reduced_resolution); |
|
|
|
|
|
stop_transfer_timer(); |
|
1076 |
|
|
1077 |
if(dec->out_frm && cp_mb > 0) { |
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); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1079 |
cp_mb = 0; |
cp_mb = 0; |
1080 |
} |
} |
1081 |
st_mb = x+1; |
st_mb = x+1; |
1082 |
} |
} else { /* not coded P_VOP macroblock */ |
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
1083 |
mb->mode = MODE_NOT_CODED; |
mb->mode = MODE_NOT_CODED; |
1084 |
|
mb->quant = quant; |
1085 |
|
|
1086 |
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; |
1087 |
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; |
1088 |
// copy macroblock directly from ref to cur |
mb->field_pred=0; /* (!) */ |
|
|
|
|
start_timer(); |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
|
|
dec->edged_width); |
|
|
|
|
|
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
|
|
|
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
|
|
dec->edged_width/2); |
|
|
} |
|
|
else |
|
|
{ |
|
|
transfer16x16_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.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); |
|
|
} |
|
1089 |
|
|
1090 |
stop_transfer_timer(); |
decoder_mbinter(dec, mb, x, y, 0, bs, |
1091 |
|
rounding, 0, 0); |
1092 |
|
|
1093 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1094 |
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); |
1097 |
st_mb = x+1; |
st_mb = x+1; |
1098 |
} |
} |
1099 |
} |
} |
1100 |
|
|
1101 |
if(dec->out_frm && cp_mb > 0) |
if(dec->out_frm && cp_mb > 0) |
1102 |
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); |
1103 |
} |
} |
1104 |
} |
} |
1105 |
|
|
1106 |
|
|
1107 |
// add by MinChen <chenm001@163.com> |
/* decode B-frame motion vector */ |
1108 |
// decode B-frame motion vector |
static void |
1109 |
void |
get_b_motion_vector(Bitstream * bs, |
|
get_b_motion_vector(DECODER * dec, |
|
|
Bitstream * bs, |
|
|
int x, |
|
|
int y, |
|
1110 |
VECTOR * mv, |
VECTOR * mv, |
1111 |
int fcode, |
int fcode, |
1112 |
const VECTOR pmv) |
const VECTOR pmv, |
1113 |
{ |
const DECODER * const dec, |
1114 |
int scale_fac = 1 << (fcode - 1); |
const int x, const int y) |
1115 |
int high = (32 * scale_fac) - 1; |
{ |
1116 |
int low = ((-32) * scale_fac); |
const int scale_fac = 1 << (fcode - 1); |
1117 |
int range = (64 * scale_fac); |
const int high = (32 * scale_fac) - 1; |
1118 |
|
const int low = ((-32) * scale_fac); |
1119 |
int mv_x, mv_y; |
const int range = (64 * scale_fac); |
|
int pmv_x, pmv_y; |
|
1120 |
|
|
1121 |
pmv_x = pmv.x; |
int mv_x = get_mv(bs, fcode); |
1122 |
pmv_y = pmv.y; |
int mv_y = get_mv(bs, fcode); |
1123 |
|
|
1124 |
mv_x = get_mv(bs, fcode); |
mv_x += pmv.x; |
1125 |
mv_y = get_mv(bs, fcode); |
mv_y += pmv.y; |
1126 |
|
|
1127 |
mv_x += pmv_x; |
if (mv_x < low) |
|
mv_y += pmv_y; |
|
|
|
|
|
if (mv_x < low) { |
|
1128 |
mv_x += range; |
mv_x += range; |
1129 |
} else if (mv_x > high) { |
else if (mv_x > high) |
1130 |
mv_x -= range; |
mv_x -= range; |
|
} |
|
1131 |
|
|
1132 |
if (mv_y < low) { |
if (mv_y < low) |
1133 |
mv_y += range; |
mv_y += range; |
1134 |
} else if (mv_y > high) { |
else if (mv_y > high) |
1135 |
mv_y -= range; |
mv_y -= range; |
|
} |
|
1136 |
|
|
1137 |
mv->x = mv_x; |
mv->x = mv_x; |
1138 |
mv->y = mv_y; |
mv->y = mv_y; |
1139 |
} |
} |
1140 |
|
|
1141 |
|
/* decode an B-frame direct & interpolate macroblock */ |
1142 |
// 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; |
|
|
|
|
|
if (dec->quarterpel) |
|
|
{ |
|
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
} |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
|
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
} else { |
|
|
int sum; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
|
|
|
|
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
} |
|
|
|
|
|
start_timer(); |
|
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
} |
|
|
else { |
|
|
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++) { |
|
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
|
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], direction); |
|
|
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 |
|
1143 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1144 |
IMAGE forward, |
IMAGE forward, |
1145 |
IMAGE backward, |
IMAGE backward, |
1146 |
const MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
1147 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1148 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1149 |
Bitstream * bs) |
Bitstream * bs, |
1150 |
|
const int direct) |
1151 |
{ |
{ |
|
|
|
|
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
|
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
|
|
|
1152 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
1153 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
|
uint32_t next_block = stride * 8; |
|
|
uint32_t iQuant = pMB->quant; |
|
1154 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
1155 |
int b_uv_dx, b_uv_dy; |
int b_uv_dx, b_uv_dy; |
|
uint32_t i; |
|
1156 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
1157 |
const uint32_t cbp = pMB->cbp; |
const uint32_t cbp = pMB->cbp; |
1158 |
|
|
1160 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1161 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1162 |
|
|
1163 |
|
validate_vector(pMB->mvs, x_pos, y_pos, dec); |
1164 |
|
validate_vector(pMB->b_mvs, x_pos, y_pos, dec); |
1165 |
|
|
1166 |
if ((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) { |
if (!direct) { |
1167 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1168 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
|
|
|
1169 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1170 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1171 |
|
|
1172 |
if (dec->quarterpel) |
if (dec->quarterpel) { |
1173 |
{ |
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1174 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
1175 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1176 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1177 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1178 |
|
} |
1179 |
|
else { |
1180 |
uv_dx /= 2; |
uv_dx /= 2; |
1181 |
uv_dy /= 2; |
uv_dy /= 2; |
|
|
|
1182 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1183 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1184 |
} |
} |
1185 |
|
} |
1186 |
|
|
1187 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1188 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
1189 |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1190 |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
|
} else { |
|
|
int sum; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
|
|
|
|
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
|
|
else |
|
|
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
|
|
|
|
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
|
|
|
if(dec->quarterpel) |
|
|
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); |
|
|
else |
|
|
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
|
|
|
|
|
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
1191 |
|
|
1192 |
if(dec->quarterpel) |
} else { |
1193 |
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); |
if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */ |
1194 |
else |
/* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */ |
1195 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1196 |
|
int z; |
1197 |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
uv_dx = 0; uv_dy = 0; |
1198 |
|
b_uv_dx = 0; b_uv_dy = 0; |
1199 |
|
for (z = 0; z < 4; z++) { |
1200 |
|
uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1)); |
1201 |
|
uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1)); |
1202 |
|
b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1)); |
1203 |
|
b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1)); |
1204 |
|
} |
1205 |
|
} |
1206 |
|
else { |
1207 |
|
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1208 |
|
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1209 |
|
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); |
1210 |
|
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); |
1211 |
|
} |
1212 |
|
} else { |
1213 |
|
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1214 |
|
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1215 |
|
b_uv_dx = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
1216 |
|
b_uv_dy = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
1217 |
} |
} |
1218 |
|
|
1219 |
|
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
1220 |
|
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
1221 |
|
b_uv_dx = (b_uv_dx >> 3) + roundtab_76[b_uv_dx & 0xf]; |
1222 |
|
b_uv_dy = (b_uv_dy >> 3) + roundtab_76[b_uv_dy & 0xf]; |
1223 |
|
} |
1224 |
|
|
1225 |
start_timer(); |
start_timer(); |
1226 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1227 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1228 |
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1229 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1230 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1231 |
else { |
} else { |
1232 |
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1233 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1234 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1242 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1243 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
1244 |
} |
} |
1245 |
} |
} else { |
|
else { |
|
1246 |
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, |
1247 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1248 |
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, |
1249 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
1250 |
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, |
1251 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
1252 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, 16 * y_pos + 8, |
1253 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
0); |
|
1254 |
} |
} |
1255 |
|
|
1256 |
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, |
1260 |
|
|
1261 |
|
|
1262 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1263 |
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
if(!direct) { |
1264 |
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1265 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1266 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1267 |
else { |
} else { |
1268 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1269 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1270 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1271 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1272 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1273 |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1274 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1275 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1276 |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1277 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1278 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1279 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1280 |
} |
} |
1281 |
} |
} else { |
1282 |
else { |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos, |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
|
1283 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1284 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1285 |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1286 |
0); |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, |
1287 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1288 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1289 |
stride, 0); |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
|
|
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
|
|
stride, 0); |
|
1290 |
} |
} |
1291 |
|
|
1292 |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos, |
1293 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1294 |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos, |
1295 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1296 |
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
1297 |
stop_comp_timer(); |
stop_comp_timer(); |
1298 |
|
|
1299 |
for (i = 0; i < 6; i++) { |
if (cbp) |
1300 |
int direction = dec->alternate_vertical_scan ? 2 : 0; |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
|
|
|
|
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], direction); |
|
|
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(); |
|
1301 |
} |
} |
1302 |
|
|
1303 |
|
/* for decode B-frame dbquant */ |
1304 |
// add by MinChen <chenm001@163.com> |
static __inline int32_t |
|
// for decode B-frame dbquant |
|
|
int32_t __inline |
|
1305 |
get_dbquant(Bitstream * bs) |
get_dbquant(Bitstream * bs) |
1306 |
{ |
{ |
1307 |
if (!BitstreamGetBit(bs)) // '0' |
if (!BitstreamGetBit(bs)) /* '0' */ |
1308 |
return (0); |
return (0); |
1309 |
else if (!BitstreamGetBit(bs)) // '10' |
else if (!BitstreamGetBit(bs)) /* '10' */ |
1310 |
return (-2); |
return (-2); |
1311 |
else |
else /* '11' */ |
1312 |
return (2); // '11' |
return (2); |
1313 |
} |
} |
1314 |
|
|
1315 |
// add by MinChen <chenm001@163.com> |
/* |
1316 |
// for decode B-frame mb_type |
* decode B-frame mb_type |
1317 |
// bit ret_value |
* bit ret_value |
1318 |
// 1 0 |
* 1 0 |
1319 |
// 01 1 |
* 01 1 |
1320 |
// 001 2 |
* 001 2 |
1321 |
// 0001 3 |
* 0001 3 |
1322 |
int32_t __inline |
*/ |
1323 |
|
static int32_t __inline |
1324 |
get_mbtype(Bitstream * bs) |
get_mbtype(Bitstream * bs) |
1325 |
{ |
{ |
1326 |
int32_t mb_type; |
int32_t mb_type; |
1327 |
|
|
1328 |
for (mb_type = 0; mb_type <= 3; mb_type++) { |
for (mb_type = 0; mb_type <= 3; mb_type++) |
1329 |
if (BitstreamGetBit(bs)) |
if (BitstreamGetBit(bs)) |
|
break; |
|
|
} |
|
|
|
|
|
if (mb_type <= 3) |
|
1330 |
return (mb_type); |
return (mb_type); |
1331 |
else |
|
1332 |
return (-1); |
return -1; |
1333 |
} |
} |
1334 |
|
|
1335 |
void |
static void |
1336 |
decoder_bframe(DECODER * dec, |
decoder_bframe(DECODER * dec, |
1337 |
Bitstream * bs, |
Bitstream * bs, |
1338 |
int quant, |
int quant, |
1342 |
uint32_t x, y; |
uint32_t x, y; |
1343 |
VECTOR mv; |
VECTOR mv; |
1344 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
1345 |
#ifdef BFRAMES_DEC_DEBUG |
int i; |
|
FILE *fp; |
|
|
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 |
|
1346 |
|
|
1347 |
|
if (!dec->is_edged[0]) { |
1348 |
start_timer(); |
start_timer(); |
1349 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1350 |
dec->width, dec->height); |
dec->width, dec->height, dec->bs_version); |
1351 |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
dec->is_edged[0] = 1; |
|
dec->width, dec->height); |
|
1352 |
stop_edges_timer(); |
stop_edges_timer(); |
1353 |
|
} |
1354 |
|
|
1355 |
#ifdef BFRAMES_DEC_DEBUG |
if (!dec->is_edged[1]) { |
1356 |
if (!first){ |
start_timer(); |
1357 |
fp=fopen("C:\\XVIDDBG.TXT","w"); |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1358 |
|
dec->width, dec->height, dec->bs_version); |
1359 |
|
dec->is_edged[1] = 1; |
1360 |
|
stop_edges_timer(); |
1361 |
} |
} |
|
#endif |
|
1362 |
|
|
1363 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1364 |
// Initialize Pred Motion Vector |
/* Initialize Pred Motion Vector */ |
1365 |
dec->p_fmv = dec->p_bmv = zeromv; |
dec->p_fmv = dec->p_bmv = zeromv; |
1366 |
for (x = 0; x < dec->mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
1367 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1368 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1369 |
|
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1370 |
|
int intra_dc_threshold; /* fake variable */ |
1371 |
|
|
1372 |
|
if (check_resync_marker(bs, fcode_max - 1)) { |
1373 |
|
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1374 |
|
&fcode_forward, &fcode_backward, &intra_dc_threshold); |
1375 |
|
x = bound % dec->mb_width; |
1376 |
|
y = bound / dec->mb_width; |
1377 |
|
/* reset predicted macroblocks */ |
1378 |
|
dec->p_fmv = dec->p_bmv = zeromv; |
1379 |
|
} |
1380 |
|
|
1381 |
mv = |
mv = |
1382 |
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] = |
1383 |
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; |
1384 |
|
mb->quant = quant; |
1385 |
|
|
1386 |
// skip if the co-located P_VOP macroblock is not coded |
/* |
1387 |
// if not codec in co-located S_VOP macroblock is _not_ automatically skipped |
* skip if the co-located P_VOP macroblock is not coded |
1388 |
|
* if not codec in co-located S_VOP macroblock is _not_ |
1389 |
|
* automatically skipped |
1390 |
|
*/ |
1391 |
|
|
1392 |
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); |
|
1393 |
mb->cbp = 0; |
mb->cbp = 0; |
1394 |
#ifdef BFRAMES_DEC_DEBUG |
mb->mode = MODE_FORWARD; |
1395 |
mb->mb_type = MODE_NOT_CODED; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 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); |
|
1396 |
continue; |
continue; |
1397 |
} |
} |
1398 |
|
|
1399 |
if (!BitstreamGetBit(bs)) { // modb=='0' |
if (!BitstreamGetBit(bs)) { /* modb=='0' */ |
1400 |
const uint8_t modb2 = BitstreamGetBit(bs); |
const uint8_t modb2 = BitstreamGetBit(bs); |
1401 |
|
|
1402 |
mb->mb_type = get_mbtype(bs); |
mb->mode = get_mbtype(bs); |
1403 |
|
|
1404 |
if (!modb2) { // modb=='00' |
if (!modb2) /* modb=='00' */ |
1405 |
mb->cbp = BitstreamGetBits(bs, 6); |
mb->cbp = BitstreamGetBits(bs, 6); |
1406 |
} else { |
else |
1407 |
mb->cbp = 0; |
mb->cbp = 0; |
|
} |
|
|
if (mb->mb_type && mb->cbp) { |
|
|
quant += get_dbquant(bs); |
|
1408 |
|
|
1409 |
if (quant > 31) { |
if (mb->mode && mb->cbp) { |
1410 |
|
quant += get_dbquant(bs); |
1411 |
|
if (quant > 31) |
1412 |
quant = 31; |
quant = 31; |
1413 |
} else if (quant < 1) { |
else if (quant < 1) |
1414 |
quant = 1; |
quant = 1; |
1415 |
} |
} |
1416 |
|
mb->quant = quant; |
1417 |
|
|
1418 |
|
if (dec->interlacing) { |
1419 |
|
if (mb->cbp) { |
1420 |
|
mb->field_dct = BitstreamGetBit(bs); |
1421 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_dct: %i\n", mb->field_dct); |
1422 |
|
} |
1423 |
|
|
1424 |
|
if (mb->mode) { |
1425 |
|
mb->field_pred = BitstreamGetBit(bs); |
1426 |
|
DPRINTF(XVID_DEBUG_MB, "decp: field_pred: %i\n", mb->field_pred); |
1427 |
|
|
1428 |
|
if (mb->field_pred) { |
1429 |
|
mb->field_for_top = BitstreamGetBit(bs); |
1430 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_top: %i\n", mb->field_for_top); |
1431 |
|
mb->field_for_bot = BitstreamGetBit(bs); |
1432 |
|
DPRINTF(XVID_DEBUG_MB,"decp: field_for_bot: %i\n", mb->field_for_bot); |
1433 |
} |
} |
1434 |
|
} |
1435 |
|
} |
1436 |
|
|
1437 |
} else { |
} else { |
1438 |
mb->mb_type = MODE_DIRECT_NONE_MV; |
mb->mode = MODE_DIRECT_NONE_MV; |
1439 |
mb->cbp = 0; |
mb->cbp = 0; |
1440 |
} |
} |
1441 |
|
|
1442 |
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) { |
|
1443 |
case MODE_DIRECT: |
case MODE_DIRECT: |
1444 |
get_b_motion_vector(dec, bs, x, y, &mv, 1, zeromv); |
get_b_motion_vector(bs, &mv, 1, zeromv, dec, x, y); |
1445 |
|
|
1446 |
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; |
|
|
|
|
1447 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1448 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) |
mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; |
1449 |
/ TRD + mv.x); |
mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; |
1450 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
|
1451 |
? ((TRB - TRD) * last_mb->mvs[i].x) |
mb->b_mvs[i].x = (mv.x) |
1452 |
/ TRD |
? mb->mvs[i].x - last_mb->mvs[i].x |
1453 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1454 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) |
mb->b_mvs[i].y = (mv.y) |
1455 |
/ TRD + mv.y); |
? mb->mvs[i].y - last_mb->mvs[i].y |
1456 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; |
|
? ((TRB - TRD) * last_mb->mvs[i].y) |
|
|
/ TRD |
|
|
: mb->mvs[i].y - last_mb->mvs[i].y); |
|
|
} |
|
|
//DEBUG("B-frame Direct!\n"); |
|
1457 |
} |
} |
1458 |
|
|
1459 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1460 |
mb, x, y, bs); |
mb, x, y, bs, 1); |
1461 |
break; |
break; |
1462 |
|
|
1463 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: |
1464 |
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, x, y); |
|
dec->p_fmv); |
|
1465 |
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]; |
1466 |
|
|
1467 |
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, dec, x, y); |
1468 |
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]; |
|
1469 |
|
|
1470 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1471 |
mb, x, y, bs); |
mb, x, y, bs, 0); |
|
//DEBUG("B-frame Bidir!\n"); |
|
1472 |
break; |
break; |
1473 |
|
|
1474 |
case MODE_BACKWARD: |
case MODE_BACKWARD: |
1475 |
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, x, y); |
|
dec->p_bmv); |
|
1476 |
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]; |
1477 |
|
|
1478 |
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, 0); |
|
|
//DEBUG("B-frame Backward!\n"); |
|
1479 |
break; |
break; |
1480 |
|
|
1481 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1482 |
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, x, y); |
|
dec->p_fmv); |
|
1483 |
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]; |
1484 |
|
|
1485 |
mb->mode = MODE_INTER; |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
|
decoder_bf_mbinter(dec, mb, x, y, mb->cbp, bs, quant, 1); |
|
|
//DEBUG("B-frame Forward!\n"); |
|
1486 |
break; |
break; |
1487 |
|
|
1488 |
default: |
default: |
1489 |
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
DPRINTF(XVID_DEBUG_ERROR,"Not supported B-frame mb_type = %i\n", mb->mode); |
1490 |
} |
} |
1491 |
|
} /* End of for */ |
|
} // end of FOR |
|
|
} |
|
|
#ifdef BFRAMES_DEC_DEBUG |
|
|
if (!first){ |
|
|
first=1; |
|
|
if (fp) |
|
|
fclose(fp); |
|
1492 |
} |
} |
|
#endif |
|
1493 |
} |
} |
1494 |
|
|
|
// swap two MACROBLOCK array |
|
|
void |
|
|
mb_swap(MACROBLOCK ** mb1, |
|
|
MACROBLOCK ** mb2) |
|
|
{ |
|
|
MACROBLOCK *temp = *mb1; |
|
|
|
|
|
*mb1 = *mb2; |
|
|
*mb2 = temp; |
|
|
} |
|
|
|
|
|
|
|
1495 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1496 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1497 |
const XVID_DEC_FRAME * frame, int pp_disable) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1498 |
|
int coding_type, int quant) |
1499 |
{ |
{ |
1500 |
|
const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; |
1501 |
|
|
1502 |
|
if (dec->cartoon_mode) |
1503 |
|
frame->general &= ~XVID_FILMEFFECT; |
1504 |
|
|
1505 |
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) |
1506 |
|
&& mbs != NULL) /* post process */ |
1507 |
{ |
{ |
1508 |
/* note: image is stored to tmp */ |
/* note: image is stored to tmp */ |
1509 |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1510 |
image_deblock_rrv(&dec->tmp, dec->edged_width, |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1511 |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1512 |
8, frame->general); |
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
1513 |
img = &dec->tmp; |
img = &dec->tmp; |
1514 |
} |
} |
1515 |
|
|
1516 |
image_output(img, dec->width, dec->height, |
image_output(img, dec->width, dec->height, |
1517 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, (uint8_t**)frame->output.plane, frame->output.stride, |
1518 |
frame->colorspace, dec->interlacing); |
frame->output.csp, dec->interlacing); |
|
} |
|
1519 |
|
|
1520 |
|
if (stats) { |
1521 |
|
stats->type = coding2type(coding_type); |
1522 |
|
stats->data.vop.time_base = (int)dec->time_base; |
1523 |
|
stats->data.vop.time_increment = 0; /* XXX: todo */ |
1524 |
|
stats->data.vop.qscale_stride = dec->mb_width; |
1525 |
|
stats->data.vop.qscale = dec->qscale; |
1526 |
|
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1527 |
|
unsigned int i; |
1528 |
|
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1529 |
|
stats->data.vop.qscale[i] = mbs[i].quant; |
1530 |
|
} else |
1531 |
|
stats->data.vop.qscale = NULL; |
1532 |
|
} |
1533 |
|
} |
1534 |
|
|
1535 |
int |
int |
1536 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1537 |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1538 |
{ |
{ |
1539 |
|
|
1540 |
Bitstream bs; |
Bitstream bs; |
1541 |
uint32_t rounding; |
uint32_t rounding; |
1542 |
uint32_t reduced_resolution; |
uint32_t quant = 2; |
|
uint32_t quant; |
|
1543 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1544 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1545 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
1546 |
WARPPOINTS gmc_warp; |
WARPPOINTS gmc_warp; |
1547 |
uint32_t vop_type; |
int coding_type; |
1548 |
int success = 0; |
int success, output, seen_something; |
|
int output = 0; |
|
|
int seen_something = 0; |
|
1549 |
|
|
1550 |
start_global_timer(); |
if (XVID_VERSION_MAJOR(frame->version) != 1 || (stats && XVID_VERSION_MAJOR(stats->version) != 1)) /* v1.x.x */ |
1551 |
|
return XVID_ERR_VERSION; |
1552 |
|
|
1553 |
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
start_global_timer(); |
|
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
|
1554 |
|
|
1555 |
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
dec->low_delay_default = (frame->general & XVID_LOWDELAY); |
1556 |
|
if ((frame->general & XVID_DISCONTINUITY)) |
1557 |
dec->frames = 0; |
dec->frames = 0; |
1558 |
|
dec->out_frm = (frame->output.csp == XVID_CSP_SLICE) ? &frame->output : NULL; |
1559 |
|
|
1560 |
if (frame->length < 0) /* decoder flush */ |
if(frame->length<0) { /* decoder flush */ |
1561 |
{ |
int ret; |
1562 |
/* if not decoding "low_delay/packed", and this isn't low_delay and |
/* if not decoding "low_delay/packed", and this isn't low_delay and |
1563 |
we have a reference frame, then outout the reference frame */ |
we have a reference frame, then outout the reference frame */ |
1564 |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) { |
1565 |
{ |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1566 |
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
dec->frames = 0; |
1567 |
output = 1; |
ret = 0; |
1568 |
} |
} else { |
1569 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1570 |
frame->length = 0; |
ret = XVID_ERR_END; |
|
if (stats) |
|
|
{ |
|
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
1571 |
} |
} |
1572 |
|
|
1573 |
emms(); |
emms(); |
|
|
|
1574 |
stop_global_timer(); |
stop_global_timer(); |
1575 |
return XVID_ERR_OK; |
return ret; |
1576 |
} |
} |
1577 |
|
|
1578 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1579 |
|
|
1580 |
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
/* XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's */ |
1581 |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
1582 |
{ |
{ |
|
if (stats) |
|
|
stats->notify = XVID_DEC_VOP; |
|
|
frame->length = 1; |
|
1583 |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
1584 |
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
(uint8_t**)frame->output.plane, frame->output.stride, frame->output.csp, dec->interlacing); |
1585 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1586 |
emms(); |
emms(); |
1587 |
return XVID_ERR_OK; |
return 1; /* one byte consumed */ |
1588 |
} |
} |
1589 |
|
|
1590 |
|
success = 0; |
1591 |
|
output = 0; |
1592 |
|
seen_something = 0; |
1593 |
|
|
1594 |
repeat: |
repeat: |
1595 |
|
|
1596 |
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
1597 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1598 |
|
|
1599 |
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i", |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%" |
1600 |
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
#if defined(_MSC_VER) |
1601 |
|
"I64" |
1602 |
|
#else |
1603 |
|
"ll" |
1604 |
|
#endif |
1605 |
|
"i, time_pp=%i, time_bp=%i\n", |
1606 |
|
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1607 |
|
|
1608 |
if (vop_type == - 1) |
if (coding_type == -1) { /* nothing */ |
|
{ |
|
1609 |
if (success) goto done; |
if (success) goto done; |
1610 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1611 |
emms(); |
emms(); |
1612 |
return XVID_ERR_FAIL; |
return BitstreamPos(&bs)/8; |
1613 |
} |
} |
1614 |
|
|
1615 |
if (vop_type == -2 || vop_type == -3) |
if (coding_type == -2 || coding_type == -3) { /* vol and/or resize */ |
1616 |
{ |
|
1617 |
if (vop_type == -3) |
if (coding_type == -3) |
1618 |
decoder_resize(dec); |
decoder_resize(dec); |
1619 |
|
|
1620 |
if (stats) |
if(stats) { |
1621 |
{ |
stats->type = XVID_TYPE_VOL; |
|
stats->notify = XVID_DEC_VOL; |
|
1622 |
stats->data.vol.general = 0; |
stats->data.vol.general = 0; |
1623 |
if (dec->interlacing) |
/*XXX: if (dec->interlacing) |
1624 |
stats->data.vol.general |= XVID_INTERLACING; |
stats->data.vol.general |= ++INTERLACING; */ |
1625 |
stats->data.vol.width = dec->width; |
stats->data.vol.width = dec->width; |
1626 |
stats->data.vol.height = dec->height; |
stats->data.vol.height = dec->height; |
1627 |
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
stats->data.vol.par = dec->aspect_ratio; |
1628 |
stats->data.vol.par_width = dec->par_width; |
stats->data.vol.par_width = dec->par_width; |
1629 |
stats->data.vol.par_height = dec->par_height; |
stats->data.vol.par_height = dec->par_height; |
|
frame->length = BitstreamPos(&bs) / 8; |
|
1630 |
emms(); |
emms(); |
1631 |
return XVID_ERR_OK; |
return BitstreamPos(&bs)/8; /* number of bytes consumed */ |
1632 |
} |
} |
1633 |
goto repeat; |
goto repeat; |
1634 |
} |
} |
1635 |
|
|
1636 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
if(dec->frames == 0 && coding_type != I_VOP) { |
1637 |
|
/* 1st frame is not an i-vop */ |
1638 |
|
goto repeat; |
1639 |
|
} |
1640 |
|
|
1641 |
|
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1642 |
|
|
1643 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1644 |
if (dec->packed_mode && vop_type == N_VOP) |
if (dec->packed_mode && coding_type == N_VOP) { |
1645 |
{ |
if (dec->low_delay_default && dec->frames > 0) { |
1646 |
if (dec->low_delay_default && dec->frames > 0) |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
1647 |
output = 1; |
output = 1; |
1648 |
} |
} |
1649 |
/* ignore otherwise */ |
/* ignore otherwise */ |
1650 |
} |
} else if (coding_type != B_VOP) { |
1651 |
else if (vop_type != B_VOP) |
switch(coding_type) { |
|
{ |
|
|
switch(vop_type) |
|
|
{ |
|
1652 |
case I_VOP : |
case I_VOP : |
1653 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1654 |
break; |
break; |
1655 |
case P_VOP : |
case P_VOP : |
1656 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1657 |
fcode_forward, intra_dc_threshold, NULL); |
fcode_forward, intra_dc_threshold, NULL); |
1658 |
break; |
break; |
1659 |
case S_VOP : |
case S_VOP : |
1660 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1661 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1662 |
break; |
break; |
1663 |
case N_VOP : |
case N_VOP : |
1664 |
|
/* XXX: not_coded vops are not used for forward prediction */ |
1665 |
|
/* we should not swap(last_mbs,mbs) */ |
1666 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1667 |
|
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); /* it will be swapped back */ |
1668 |
break; |
break; |
1669 |
} |
} |
1670 |
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
|
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
|
|
16, XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV); |
|
|
} |
|
|
|
|
1671 |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1672 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1673 |
{ |
if(dec->low_delay) { |
1674 |
if (dec->low_delay) |
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
|
{ |
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
|
1675 |
output = 1; |
output = 1; |
1676 |
} |
} else if (dec->frames > 0) { /* is the reference frame valid? */ |
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
1677 |
/* output the reference frame */ |
/* output the reference frame */ |
1678 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1679 |
output = 1; |
output = 1; |
1680 |
} |
} |
1681 |
} |
} |
1682 |
|
|
1683 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1684 |
|
dec->is_edged[1] = dec->is_edged[0]; |
1685 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1686 |
mb_swap(&dec->mbs, &dec->last_mbs); |
dec->is_edged[0] = 0; |
1687 |
dec->last_reduced_resolution = reduced_resolution; |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
1688 |
|
dec->last_coding_type = coding_type; |
1689 |
|
|
1690 |
dec->frames++; |
dec->frames++; |
1691 |
seen_something = 1; |
seen_something = 1; |
1692 |
|
|
1693 |
}else{ /* B_VOP */ |
}else{ /* B_VOP */ |
1694 |
|
|
1695 |
if (dec->low_delay) |
if (dec->low_delay) { |
1696 |
{ |
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1697 |
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
dec->low_delay = 0; |
|
dec->low_delay = 1; |
|
1698 |
} |
} |
1699 |
|
|
1700 |
if (dec->frames < 2) |
if (dec->frames < 2) { |
|
{ |
|
1701 |
/* attemping to decode a bvop without atleast 2 reference frames */ |
/* attemping to decode a bvop without atleast 2 reference frames */ |
1702 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1703 |
"broken b-frame, mising ref frames"); |
"broken b-frame, mising ref frames"); |
1704 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1705 |
}else if (dec->time_pp <= dec->time_bp) { |
}else if (dec->time_pp <= dec->time_bp) { |
1706 |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
/* this occurs when dx50_bvop_compatibility==0 sequences are |
1707 |
decoded in vfw. */ |
decoded in vfw. */ |
1708 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
1709 |
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
1710 |
|
if (stats) stats->type = XVID_TYPE_NOTHING; |
1711 |
}else{ |
}else{ |
1712 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1713 |
|
decoder_output(dec, &dec->cur, dec->mbs, frame, stats, coding_type, quant); |
1714 |
} |
} |
1715 |
|
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
|
1716 |
output = 1; |
output = 1; |
1717 |
dec->frames++; |
dec->frames++; |
1718 |
} |
} |
1719 |
|
|
1720 |
|
#if 0 /* Avoids to read to much data because of 32bit reads in our BS functions */ |
1721 |
BitstreamByteAlign(&bs); |
BitstreamByteAlign(&bs); |
1722 |
|
#endif |
1723 |
|
|
1724 |
/* low_delay_default mode: repeat in packed_mode */ |
/* low_delay_default mode: repeat in packed_mode */ |
1725 |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) { |
|
{ |
|
1726 |
success = 1; |
success = 1; |
1727 |
goto repeat; |
goto repeat; |
1728 |
} |
} |
1729 |
|
|
1730 |
done : |
done : |
1731 |
|
|
1732 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* if we reach here without outputing anything _and_ |
1733 |
then output the recently decoded frame, or print an error message */ |
the calling application has specified low_delay_default, |
1734 |
if (dec->low_delay_default && output == 0) |
we *must* output something. |
1735 |
{ |
this always occurs on the first call to decode() call |
1736 |
if (dec->packed_mode && seen_something) |
when bframes are present in the bitstream. it may also |
1737 |
{ |
occur if no vops were seen in the bitstream |
1738 |
/* output the recently decoded frame */ |
|
1739 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
if packed_mode is enabled, then we output the recently |
1740 |
output = 1; |
decoded frame (the very first ivop). otherwise we have |
1741 |
} |
nothing to display, and therefore output a black screen. |
1742 |
else |
*/ |
1743 |
{ |
if (dec->low_delay_default && output == 0) { |
1744 |
|
if (dec->packed_mode && seen_something) { |
1745 |
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1746 |
|
} else { |
1747 |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
1748 |
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1749 |
"warning: nothing to output"); |
if (stats) stats->type = XVID_TYPE_NOTHING; |
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
|
|
"bframe decoder lag"); |
|
|
|
|
|
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
|
1750 |
} |
} |
1751 |
} |
} |
1752 |
|
|
|
frame->length = BitstreamPos(&bs) / 8; |
|
|
|
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
|
} |
|
|
|
|
1753 |
emms(); |
emms(); |
|
|
|
1754 |
stop_global_timer(); |
stop_global_timer(); |
1755 |
|
|
1756 |
return XVID_ERR_OK; |
return (BitstreamPos(&bs)+7)/8; /* number of bytes consumed */ |
1757 |
} |
} |