3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* - Decoder main module - |
* - Decoder main module - |
5 |
* |
* |
6 |
|
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
|
* 2002 Peter Ross <pross@xvid.org> |
8 |
|
* |
9 |
* This program is an implementation of a part of one or more MPEG-4 |
* This program is an implementation of a part of one or more MPEG-4 |
10 |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
* Video tools as specified in ISO/IEC 14496-2 standard. Those intending |
11 |
* to use this software module in hardware or software products are |
* to use this software module in hardware or software products are |
36 |
* History: |
* History: |
37 |
* |
* |
38 |
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
* 15.07.2002 fix a bug in B-frame decode at DIRECT mode |
39 |
* MinChen <chenm001@163.com> |
|
40 |
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
* 10.07.2002 added BFRAMES_DEC_DEBUG support |
41 |
* Fix a little bug for low_delay flage |
* Fix a little bug for low_delay flage |
42 |
* MinChen <chenm001@163.com> |
* MinChen <chenm001@163.com> |
56 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
57 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
58 |
* 22.12.2001 lock based interpolation |
* 22.12.2001 lock based interpolation |
59 |
* 01.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* 01.12.2001 inital version; (c)2001 peter ross <pross@xvid.org> |
60 |
* |
* |
61 |
* $Id$ |
* $Id$ |
62 |
* |
* |
82 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
83 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
84 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
85 |
|
|
86 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
87 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
88 |
#include "utils/timer.h" |
#include "utils/timer.h" |
89 |
#include "utils/emms.h" |
#include "utils/emms.h" |
|
#include "motion/motion.h" |
|
90 |
|
|
91 |
#include "image/image.h" |
#include "image/image.h" |
92 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
93 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
94 |
|
|
95 |
int |
int |
96 |
decoder_resize(DECODER * dec) |
decoder_create(XVID_DEC_PARAM * param) |
97 |
{ |
{ |
98 |
/* free existing */ |
DECODER *dec; |
|
|
|
|
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->refn[2], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
99 |
|
|
100 |
if (dec->last_mbs) |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
101 |
xvid_free(dec->last_mbs); |
if (dec == NULL) { |
102 |
if (dec->mbs) |
return XVID_ERR_MEMORY; |
103 |
xvid_free(dec->mbs); |
} |
104 |
|
param->handle = dec; |
105 |
|
|
106 |
/* realloc */ |
dec->width = param->width; |
107 |
|
dec->height = param->height; |
108 |
|
|
109 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
110 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
111 |
|
|
112 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
113 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
114 |
|
dec->low_delay = 0; |
115 |
|
|
116 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
117 |
xvid_free(dec); |
xvid_free(dec); |
123 |
xvid_free(dec); |
xvid_free(dec); |
124 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
125 |
} |
} |
|
|
|
126 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
127 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
128 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
139 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
140 |
} |
} |
141 |
|
|
|
if (image_create(&dec->refh, 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->refn[2], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
142 |
dec->mbs = |
dec->mbs = |
143 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
144 |
CACHE_LINE); |
CACHE_LINE); |
147 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
148 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
149 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
150 |
xvid_free(dec); |
xvid_free(dec); |
151 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
152 |
} |
} |
153 |
|
|
154 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
155 |
|
|
156 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
164 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
165 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
166 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
167 |
xvid_free(dec); |
xvid_free(dec); |
168 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
169 |
} |
} |
170 |
|
|
171 |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->last_mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
172 |
|
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
|
|
|
int |
|
|
decoder_create(XVID_DEC_PARAM * param) |
|
|
{ |
|
|
DECODER *dec; |
|
|
|
|
|
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
|
|
if (dec == NULL) { |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
memset(dec, 0, sizeof(DECODER)); |
|
|
|
|
|
param->handle = dec; |
|
|
|
|
|
dec->width = param->width; |
|
|
dec->height = param->height; |
|
|
|
|
|
image_null(&dec->cur); |
|
|
image_null(&dec->refn[0]); |
|
|
image_null(&dec->refn[1]); |
|
|
image_null(&dec->refn[2]); |
|
|
image_null(&dec->refh); |
|
|
|
|
|
dec->mbs = NULL; |
|
|
dec->last_mbs = NULL; |
|
|
|
|
173 |
init_timer(); |
init_timer(); |
174 |
|
|
175 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
176 |
// for support B-frame to save reference frame's time |
// for support B-frame to save reference frame's time |
177 |
dec->frames = -1; |
dec->frames = -1; |
178 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
|
dec->low_delay = 0; |
|
|
dec->packed_mode = 0; |
|
|
|
|
|
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
|
179 |
|
|
|
if (dec->fixed_dimensions) |
|
|
return decoder_resize(dec); |
|
|
else |
|
180 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
181 |
} |
} |
182 |
|
|
189 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
190 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
191 |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->refh, dec->edged_width, dec->edged_height); |
|
192 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
193 |
xvid_free(dec); |
xvid_free(dec); |
194 |
|
|
217 |
Bitstream * bs, |
Bitstream * bs, |
218 |
const uint32_t quant, |
const uint32_t quant, |
219 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
220 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
221 |
{ |
{ |
222 |
|
|
223 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
230 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
231 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
232 |
|
|
|
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{ |
|
233 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
234 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
235 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
236 |
|
|
237 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
238 |
|
|
271 |
start_timer(); |
start_timer(); |
272 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
273 |
{ |
{ |
274 |
int direction = dec->alternate_vertical_scan ? |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
275 |
2 : pMB->acpred_directions[i]; |
start_coeff); |
|
|
|
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
|
276 |
} |
} |
277 |
stop_coding_timer(); |
stop_coding_timer(); |
278 |
|
|
291 |
start_timer(); |
start_timer(); |
292 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
293 |
stop_idct_timer(); |
stop_idct_timer(); |
|
|
|
294 |
} |
} |
295 |
|
|
296 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
299 |
} |
} |
300 |
|
|
301 |
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{ |
|
302 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
303 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
304 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
305 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
306 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
307 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
308 |
stop_transfer_timer(); |
stop_transfer_timer(); |
309 |
} |
} |
310 |
|
|
314 |
|
|
315 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
316 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
317 |
|
static const uint32_t roundtab[16] = |
318 |
// decode an inter macroblock |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
|
|
|
|
static void |
|
|
rrv_mv_scaleup(VECTOR * mv) |
|
|
{ |
|
|
if (mv->x > 0) { |
|
|
mv->x = 2*mv->x - 1; |
|
|
} else if (mv->x < 0) { |
|
|
mv->x = 2*mv->x + 1; |
|
|
} |
|
|
|
|
|
if (mv->y > 0) { |
|
|
mv->y = 2*mv->y - 1; |
|
|
} else if (mv->y < 0) { |
|
|
mv->y = 2*mv->y + 1; |
|
|
} |
|
|
} |
|
319 |
|
|
320 |
|
|
321 |
|
// decode an inter macroblock |
322 |
|
|
323 |
void |
void |
324 |
decoder_mbinter(DECODER * dec, |
decoder_mbinter(DECODER * dec, |
329 |
const uint32_t cbp, |
const uint32_t cbp, |
330 |
Bitstream * bs, |
Bitstream * bs, |
331 |
const uint32_t quant, |
const uint32_t quant, |
332 |
const uint32_t rounding, |
const uint32_t rounding) |
|
const int reduced_resolution) |
|
333 |
{ |
{ |
334 |
|
|
335 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
337 |
|
|
338 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
339 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
340 |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
uint32_t next_block = stride * 8; |
341 |
uint32_t i; |
uint32_t i; |
342 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
343 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
|
344 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
|
VECTOR mv[4]; |
|
345 |
|
|
|
for (i = 0; i < 4; i++) |
|
|
{ |
|
|
mv[i] = pMB->mvs[i]; |
|
|
//DPRINTF(DPRINTF_MB, "mv[%i] orig=%i,%i local=%i", i, pMB->mvs[i].x, pMB->mvs[i].y, mv[i].x, mv[i].y); |
|
|
} |
|
|
|
|
|
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); |
|
|
rrv_mv_scaleup(&mv[0]); |
|
|
rrv_mv_scaleup(&mv[1]); |
|
|
rrv_mv_scaleup(&mv[2]); |
|
|
rrv_mv_scaleup(&mv[3]); |
|
|
}else{ |
|
346 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
347 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
348 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
349 |
|
|
350 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
351 |
uv_dx = mv[0].x; |
uv_dx = pMB->mvs[0].x; |
352 |
uv_dy = mv[0].y; |
uv_dy = pMB->mvs[0].y; |
353 |
|
|
354 |
if (dec->quarterpel) |
if (dec->quarterpel) |
355 |
{ |
{ |
356 |
uv_dx /= 2; |
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
357 |
uv_dy /= 2; |
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
358 |
} |
} |
359 |
|
|
360 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
361 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
362 |
|
} else { |
363 |
start_timer(); |
int sum; |
364 |
if (reduced_resolution) |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
|
{ |
|
|
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); |
|
365 |
|
|
366 |
} |
if (dec->quarterpel) |
|
else |
|
367 |
{ |
{ |
368 |
if(dec->quarterpel) { |
sum /= 2; |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
} |
|
|
else { |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
} |
|
|
|
|
|
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
369 |
} |
} |
|
stop_comp_timer(); |
|
370 |
|
|
371 |
} else { /* MODE_INTER4V */ |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
int sum; |
|
|
|
|
|
if(dec->quarterpel) |
|
|
sum = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
|
|
else |
|
|
sum = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
|
372 |
|
|
373 |
uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
374 |
|
|
375 |
if(dec->quarterpel) |
if(dec->quarterpel) |
|
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; |
|
|
|
|
|
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
|
|
|
|
|
start_timer(); |
|
|
if (reduced_resolution) |
|
376 |
{ |
{ |
377 |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
sum /= 2; |
378 |
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); |
|
379 |
|
|
380 |
// set_block(pY_Cur, stride, 32, 32, 127); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
381 |
} |
} |
382 |
else |
|
383 |
{ |
start_timer(); |
384 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
385 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n"); |
386 |
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
387 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
388 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos, |
389 |
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
390 |
mv[1].x, mv[1].y, stride, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
391 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
392 |
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos + 8, 16*y_pos + 8, |
393 |
mv[2].x, mv[2].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
|
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
394 |
} |
} |
395 |
else { |
else { |
396 |
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, |
397 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
398 |
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, |
399 |
mv[1].x, mv[1].y, stride, rounding); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
400 |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
401 |
mv[2].x, mv[2].y, stride, rounding); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
402 |
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, |
403 |
mv[3].x, mv[3].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
404 |
} |
} |
405 |
|
|
406 |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[0].u, 8 * x_pos, 8 * y_pos, |
407 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
408 |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[0].v, 8 * x_pos, 8 * y_pos, |
409 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
|
} |
|
410 |
stop_comp_timer(); |
stop_comp_timer(); |
|
} |
|
411 |
|
|
412 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
413 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
414 |
{ |
{ |
415 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
416 |
|
|
417 |
start_timer(); |
start_timer(); |
418 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
419 |
stop_coding_timer(); |
stop_coding_timer(); |
420 |
|
|
421 |
start_timer(); |
start_timer(); |
438 |
} |
} |
439 |
|
|
440 |
start_timer(); |
start_timer(); |
|
if (reduced_resolution) |
|
|
{ |
|
|
if (cbp & 32) |
|
|
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); |
|
|
} |
|
|
else |
|
|
{ |
|
441 |
if (cbp & 32) |
if (cbp & 32) |
442 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
443 |
if (cbp & 16) |
if (cbp & 16) |
450 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
451 |
if (cbp & 1) |
if (cbp & 1) |
452 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
453 |
stop_transfer_timer(); |
stop_transfer_timer(); |
454 |
} |
} |
455 |
|
|
457 |
void |
void |
458 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
459 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
460 |
int quant, |
int quant, |
461 |
int intra_dc_threshold) |
int intra_dc_threshold) |
462 |
{ |
{ |
463 |
uint32_t bound; |
uint32_t bound; |
464 |
uint32_t x, y; |
uint32_t x, y; |
|
int mb_width = dec->mb_width; |
|
|
int mb_height = dec->mb_height; |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
465 |
|
|
466 |
bound = 0; |
bound = 0; |
467 |
|
|
468 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
469 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
470 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
471 |
uint32_t mcbpc; |
uint32_t mcbpc; |
472 |
uint32_t cbpc; |
uint32_t cbpc; |
479 |
|
|
480 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
481 |
{ |
{ |
482 |
bound = read_video_packet_header(bs, dec, 0, |
bound = read_video_packet_header(bs, 0, &quant); |
483 |
&quant, NULL, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
484 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
485 |
} |
} |
486 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
487 |
|
|
512 |
|
|
513 |
if (dec->interlacing) { |
if (dec->interlacing) { |
514 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
515 |
DEBUG1("deci: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
516 |
} |
} |
517 |
|
|
518 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
519 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
|
|
|
520 |
} |
} |
521 |
if(dec->out_frm) |
if(dec->out_frm) |
522 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,mb_width); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,0,y,dec->mb_width); |
523 |
|
|
524 |
} |
} |
525 |
|
|
526 |
} |
} |
532 |
int x, |
int x, |
533 |
int y, |
int y, |
534 |
int k, |
int k, |
535 |
VECTOR * ret_mv, |
VECTOR * mv, |
536 |
int fcode, |
int fcode, |
537 |
const int bound) |
const int bound) |
538 |
{ |
{ |
543 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
544 |
|
|
545 |
VECTOR pmv; |
VECTOR pmv; |
546 |
VECTOR mv; |
int mv_x, mv_y; |
547 |
|
|
548 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
549 |
|
|
550 |
mv.x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
551 |
mv.y = get_mv(bs, fcode); |
mv_y = get_mv(bs, fcode); |
|
|
|
|
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv.x, mv.y, pmv.x, pmv.y); |
|
552 |
|
|
553 |
mv.x += pmv.x; |
DPRINTF(DPRINTF_MV,"mv_diff (%i,%i) pred (%i,%i)", mv_x, mv_y, pmv.x, pmv.y); |
|
mv.y += pmv.y; |
|
554 |
|
|
555 |
if (mv.x < low) { |
mv_x += pmv.x; |
556 |
mv.x += range; |
mv_y += pmv.y; |
|
} else if (mv.x > high) { |
|
|
mv.x -= range; |
|
|
} |
|
557 |
|
|
558 |
if (mv.y < low) { |
if (mv_x < low) { |
559 |
mv.y += range; |
mv_x += range; |
560 |
} else if (mv.y > high) { |
} else if (mv_x > high) { |
561 |
mv.y -= range; |
mv_x -= range; |
562 |
} |
} |
563 |
|
|
564 |
ret_mv->x = mv.x; |
if (mv_y < low) { |
565 |
ret_mv->y = mv.y; |
mv_y += range; |
566 |
|
} else if (mv_y > high) { |
567 |
|
mv_y -= range; |
568 |
} |
} |
569 |
|
|
570 |
|
mv->x = mv_x; |
571 |
|
mv->y = mv_y; |
572 |
|
|
|
|
|
|
static __inline int gmc_sanitize(int value, int quarterpel, int fcode) |
|
|
{ |
|
|
int length = 1 << (fcode+4); |
|
|
|
|
|
if (quarterpel) value *= 2; |
|
|
|
|
|
if (value < -length) |
|
|
return -length; |
|
|
else if (value >= length) |
|
|
return length-1; |
|
|
else return value; |
|
573 |
} |
} |
574 |
|
|
575 |
|
|
|
/* for P_VOP set gmc_mv to NULL */ |
|
576 |
void |
void |
577 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
578 |
Bitstream * bs, |
Bitstream * bs, |
579 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
580 |
int quant, |
int quant, |
581 |
int fcode, |
int fcode, |
582 |
int intra_dc_threshold, |
int intra_dc_threshold) |
|
VECTOR * gmc_mv) |
|
583 |
{ |
{ |
584 |
|
|
585 |
uint32_t x, y; |
uint32_t x, y; |
586 |
uint32_t bound; |
uint32_t bound; |
587 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
|
int mb_width = dec->mb_width; |
|
|
int mb_height = dec->mb_height; |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
588 |
|
|
589 |
start_timer(); |
start_timer(); |
590 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
593 |
|
|
594 |
bound = 0; |
bound = 0; |
595 |
|
|
596 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
597 |
cp_mb = st_mb = 0; |
cp_mb = st_mb = 0; |
598 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
599 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
600 |
|
|
601 |
// skip stuffing |
// skip stuffing |
604 |
|
|
605 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) |
606 |
{ |
{ |
607 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
608 |
&quant, &fcode, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
609 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
610 |
} |
} |
611 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
612 |
|
|
621 |
uint32_t cbpy; |
uint32_t cbpy; |
622 |
uint32_t cbp; |
uint32_t cbp; |
623 |
uint32_t intra; |
uint32_t intra; |
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
|
624 |
|
|
625 |
cp_mb++; |
cp_mb++; |
626 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
637 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
638 |
} |
} |
639 |
|
|
|
if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
|
{ |
|
|
mcsel = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
640 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
641 |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
642 |
|
|
658 |
if (dec->interlacing) { |
if (dec->interlacing) { |
659 |
if (cbp || intra) { |
if (cbp || intra) { |
660 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
661 |
DEBUG1("decp: field_dct: ", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
662 |
} |
} |
663 |
|
|
664 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
665 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
666 |
DEBUG1("decp: field_pred: ", mb->field_pred); |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
667 |
|
|
668 |
if (mb->field_pred) { |
if (mb->field_pred) { |
669 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
670 |
DEBUG1("decp: field_for_top: ", mb->field_for_top); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
671 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
672 |
DEBUG1("decp: field_for_bot: ", mb->field_for_bot); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
673 |
} |
} |
674 |
} |
} |
675 |
} |
} |
676 |
|
|
677 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
678 |
|
if (dec->interlacing && mb->field_pred) { |
|
if (mcsel) |
|
|
{ |
|
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
|
|
|
|
|
} else if (dec->interlacing && mb->field_pred) { |
|
679 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
680 |
fcode, bound); |
fcode, bound); |
681 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
701 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = |
702 |
0; |
0; |
703 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
704 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
705 |
continue; |
continue; |
706 |
} |
} |
707 |
|
|
708 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
709 |
rounding, reduced_resolution); |
rounding); |
710 |
|
} else // not coded |
|
} |
|
|
else if (gmc_mv) /* not coded S_VOP macroblock */ |
|
711 |
{ |
{ |
712 |
mb->mode = MODE_NOT_CODED; |
DPRINTF(DPRINTF_DEBUG, "P-frame MB at (X,Y)=(%d,%d)", x, y); |
|
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = gmc_sanitize(gmc_mv[0].x, dec->quarterpel, fcode); |
|
|
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = gmc_sanitize(gmc_mv[0].y, dec->quarterpel, fcode); |
|
|
decoder_mbinter(dec, mb, x, y, 0, 0, bs, quant, rounding, reduced_resolution); |
|
|
} |
|
|
else /* not coded P_VOP macroblock */ |
|
|
{ |
|
|
mb->mode = MODE_NOT_CODED; |
|
713 |
|
|
714 |
|
mb->mode = MODE_NOT_CODED; |
715 |
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; |
716 |
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; |
717 |
|
|
718 |
// copy macroblock directly from ref to cur |
// copy macroblock directly from ref to cur |
719 |
|
|
720 |
start_timer(); |
start_timer(); |
721 |
|
|
722 |
if (reduced_resolution) |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
723 |
{ |
(16 * x), |
724 |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
dec->refn[0].y + (16 * y) * dec->edged_width + |
725 |
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
(16 * x), dec->edged_width); |
726 |
dec->edged_width); |
|
727 |
|
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
728 |
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
(16 * x + 8), |
729 |
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
dec->refn[0].y + (16 * y) * dec->edged_width + |
730 |
dec->edged_width/2); |
(16 * x + 8), dec->edged_width); |
731 |
|
|
732 |
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
733 |
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
(16 * x), |
734 |
dec->edged_width/2); |
dec->refn[0].y + (16 * y + |
735 |
} |
8) * dec->edged_width + |
736 |
else |
(16 * x), dec->edged_width); |
737 |
{ |
|
738 |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
739 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
(16 * x + 8), |
740 |
dec->edged_width); |
dec->refn[0].y + (16 * y + |
741 |
|
8) * dec->edged_width + |
742 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), dec->edged_width); |
743 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
|
744 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
745 |
|
(8 * x), |
746 |
|
dec->refn[0].u + |
747 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
748 |
dec->edged_width/2); |
dec->edged_width/2); |
749 |
|
|
750 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
751 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
752 |
|
dec->refn[0].v + |
753 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
754 |
dec->edged_width/2); |
dec->edged_width/2); |
|
} |
|
|
|
|
755 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
|
|
756 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
757 |
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); |
758 |
cp_mb = 0; |
cp_mb = 0; |
844 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
845 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
846 |
|
|
847 |
if (dec->quarterpel) |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
848 |
{ |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
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]; |
|
849 |
} else { |
} else { |
850 |
int sum; |
int sum; |
851 |
|
|
|
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 |
|
852 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
853 |
|
uv_dx = |
854 |
|
(sum == |
855 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
856 |
|
(ABS(sum) / 16) * 2)); |
857 |
|
|
|
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 |
|
858 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
859 |
|
uv_dy = |
860 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
861 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
862 |
|
(ABS(sum) / 16) * 2)); |
863 |
} |
} |
864 |
|
|
865 |
start_timer(); |
start_timer(); |
|
if(dec->quarterpel) { |
|
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
} |
|
|
else { |
|
866 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos, |
867 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
868 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
869 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
870 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos, 16*y_pos + 8, |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos, |
871 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
872 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
0); |
873 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
874 |
} |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
875 |
|
0); |
876 |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.u, dec->refn[ref].u, 8 * x_pos, 8 * y_pos, |
877 |
uv_dx, uv_dy, stride2, 0); |
uv_dx, uv_dy, stride2, 0); |
878 |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->cur.v, dec->refn[ref].v, 8 * x_pos, 8 * y_pos, |
880 |
stop_comp_timer(); |
stop_comp_timer(); |
881 |
|
|
882 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
883 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
884 |
{ |
{ |
885 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
886 |
|
|
887 |
start_timer(); |
start_timer(); |
888 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
889 |
stop_coding_timer(); |
stop_coding_timer(); |
890 |
|
|
891 |
start_timer(); |
start_timer(); |
923 |
stop_transfer_timer(); |
stop_transfer_timer(); |
924 |
} |
} |
925 |
|
|
926 |
|
|
927 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
928 |
// decode an B-frame direct & inter macroblock |
// decode an B-frame direct & inter macroblock |
929 |
void |
void |
958 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
959 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
960 |
|
|
961 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
962 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
963 |
|
|
964 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
965 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
966 |
|
|
967 |
if (dec->quarterpel) |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
968 |
{ |
b_uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
|
uv_dx /= 2; |
|
|
uv_dy /= 2; |
|
|
|
|
|
b_uv_dx /= 2; |
|
|
b_uv_dy /= 2; |
|
|
} |
|
|
|
|
|
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
|
|
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
|
|
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
|
|
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
|
969 |
} else { |
} else { |
970 |
int sum; |
int sum; |
971 |
|
|
|
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 |
|
972 |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
sum = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
973 |
|
uv_dx = |
974 |
|
(sum == |
975 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
976 |
|
(ABS(sum) / 16) * 2)); |
977 |
|
|
|
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 |
|
978 |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
sum = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
979 |
|
uv_dy = |
980 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
981 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
982 |
|
(ABS(sum) / 16) * 2)); |
983 |
if(dec->quarterpel) |
|
984 |
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); |
sum = |
985 |
else |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
986 |
sum = pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + pMB->b_mvs[3].x; |
pMB->b_mvs[3].x; |
987 |
|
b_uv_dx = |
988 |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
989 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
990 |
if(dec->quarterpel) |
(ABS(sum) / 16) * 2)); |
991 |
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); |
|
992 |
else |
sum = |
993 |
sum = pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + pMB->b_mvs[3].y; |
pMB->b_mvs[0].y + pMB->b_mvs[1].y + pMB->b_mvs[2].y + |
994 |
|
pMB->b_mvs[3].y; |
995 |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
b_uv_dy = |
996 |
|
(sum == |
997 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
998 |
|
(ABS(sum) / 16) * 2)); |
999 |
} |
} |
1000 |
|
|
1001 |
|
|
1002 |
start_timer(); |
start_timer(); |
|
if(dec->quarterpel) { |
|
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
|
interpolate16x16_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
else { |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
1003 |
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, |
1004 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1005 |
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, |
1009 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1010 |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
1011 |
0); |
0); |
|
} |
|
|
|
|
1012 |
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, |
1013 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1014 |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
interpolate8x8_switch(dec->cur.v, forward.v, 8 * x_pos, 8 * y_pos, uv_dx, |
1015 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1016 |
|
|
1017 |
|
|
|
if(dec->quarterpel) { |
|
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
|
interpolate16x16_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
else { |
|
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->refn[2].y, backward.y, dec->refh.y, dec->refh.y + 64, |
|
|
dec->refh.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
1018 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
1019 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1020 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1026 |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1027 |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, |
1028 |
stride, 0); |
stride, 0); |
|
} |
|
|
|
|
1029 |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
1030 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1031 |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1032 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1033 |
|
|
1034 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos, |
1035 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
stride); |
1036 |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1037 |
stride, 1, 8); |
stride); |
1038 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1039 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
stride); |
1040 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1041 |
dec->refn[2].y + (16 * y_pos * stride) + 16 * x_pos + 8, |
16 * y_pos + 8, stride); |
1042 |
stride, 1, 8); |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1043 |
|
stride2); |
1044 |
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.v, dec->refn[2].v, 8 * x_pos, 8 * y_pos, |
1045 |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
stride2); |
|
dec->refn[2].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->refn[2].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->refn[2].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->refn[2].v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
1046 |
stop_comp_timer(); |
stop_comp_timer(); |
1047 |
|
|
1048 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
1049 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
1050 |
{ |
{ |
1051 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1052 |
|
|
1053 |
start_timer(); |
start_timer(); |
1054 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
1055 |
stop_coding_timer(); |
stop_coding_timer(); |
1056 |
|
|
1057 |
start_timer(); |
start_timer(); |
1282 |
break; |
break; |
1283 |
|
|
1284 |
default: |
default: |
1285 |
DEBUG1("Not support B-frame mb_type =", mb->mb_type); |
DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type); |
1286 |
} |
} |
1287 |
|
|
1288 |
} // end of FOR |
} // end of FOR |
1309 |
|
|
1310 |
int |
int |
1311 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1312 |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
XVID_DEC_FRAME * frame) |
1313 |
{ |
{ |
1314 |
|
|
1315 |
Bitstream bs; |
Bitstream bs; |
1316 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1317 |
uint32_t quant; |
uint32_t quant; |
1318 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1319 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1320 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
|
VECTOR gmc_mv[5]; |
|
1321 |
uint32_t vop_type; |
uint32_t vop_type; |
|
int success = 0; |
|
1322 |
|
|
1323 |
start_global_timer(); |
start_global_timer(); |
1324 |
|
|
1326 |
|
|
1327 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1328 |
|
|
|
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
|
|
if(frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
|
|
{ |
|
|
if (stats) |
|
|
stats->notify = XVID_DEC_VOP; |
|
|
frame->length = 1; |
|
|
image_output(&dec->refn[0], dec->width, dec->height, dec->edged_width, |
|
|
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
|
|
emms(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
|
start: |
|
1329 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
1330 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
1331 |
dec->frames++; |
dec->frames++; |
|
|
|
|
xxx: |
|
1332 |
vop_type = |
vop_type = |
1333 |
BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1334 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); |
&fcode_backward, &intra_dc_threshold); |
|
|
|
|
//DPRINTF(DPRINTF_HEADER, "vop_type=%i", vop_type); |
|
|
|
|
|
if (vop_type == -1 && success) |
|
|
goto done; |
|
|
|
|
|
if (vop_type == -2 || vop_type == -3) |
|
|
{ |
|
|
if (vop_type == -3) |
|
|
decoder_resize(dec); |
|
|
|
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = XVID_DEC_VOL; |
|
|
stats->data.vol.general = 0; |
|
|
if (dec->interlacing) |
|
|
stats->data.vol.general |= XVID_INTERLACING; |
|
|
stats->data.vol.width = dec->width; |
|
|
stats->data.vol.height = dec->height; |
|
|
stats->data.vol.aspect_ratio = dec->aspect_ratio; |
|
|
stats->data.vol.par_width = dec->par_width; |
|
|
stats->data.vol.par_height = dec->par_height; |
|
|
frame->length = BitstreamPos(&bs) / 8; |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
goto xxx; |
|
|
} |
|
1335 |
|
|
1336 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; // init pred vector to 0 |
1337 |
|
|
1338 |
switch (vop_type) { |
switch (vop_type) { |
1339 |
case P_VOP: |
case P_VOP: |
1340 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1341 |
fcode_forward, intra_dc_threshold, NULL); |
intra_dc_threshold); |
1342 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1343 |
DEBUG1("P_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "P_VOP Time=%d", dec->time); |
1344 |
#endif |
#endif |
1345 |
break; |
break; |
1346 |
|
|
1347 |
case I_VOP: |
case I_VOP: |
1348 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1349 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1350 |
DEBUG1("I_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "I_VOP Time=%d", dec->time); |
1351 |
#endif |
#endif |
1352 |
break; |
break; |
1353 |
|
|
1354 |
case B_VOP: |
case B_VOP: |
1355 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1356 |
if (dec->time_pp > dec->time_bp) { |
if (dec->time_pp > dec->time_bp) { |
1357 |
DEBUG1("B_VOP Time=", dec->time); |
DPRINTF(DPRINTF_DEBUG, "B_VOP Time=%d", dec->time); |
1358 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1359 |
} else { |
} else { |
1360 |
DEBUG("broken B-frame!"); |
DPRINTF(DPRINTF_DEBUG, "Broken B_VOP"); |
1361 |
} |
} |
1362 |
#else |
#else |
1363 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1364 |
#endif |
#endif |
1365 |
break; |
break; |
1366 |
|
|
|
case S_VOP : |
|
|
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
|
|
fcode_forward, intra_dc_threshold, gmc_mv); |
|
|
break; |
|
|
|
|
1367 |
case N_VOP: // vop not coded |
case N_VOP: // vop not coded |
1368 |
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1369 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
|
#ifdef BFRAMES_DEC |
|
|
DEBUG1("N_VOP Time=", dec->time); |
|
|
#endif |
|
1370 |
break; |
break; |
1371 |
|
|
1372 |
default: |
default: |
|
if (stats) |
|
|
stats->notify = 0; |
|
|
|
|
|
emms(); |
|
1373 |
return XVID_ERR_FAIL; |
return XVID_ERR_FAIL; |
1374 |
} |
} |
1375 |
|
|
1376 |
|
#ifdef BFRAMES_DEC_DEBUG |
1377 |
/* reduced resolution deblocking filter */ |
if (frame->length != BitstreamPos(&bs) / 8){ |
1378 |
|
DPRINTF(DPRINTF_DEBUG, "InLen: %d / UseLen: %d", frame->length, BitstreamPos(&bs) / 8); |
|
if (reduced_resolution) |
|
|
{ |
|
|
const int rmb_width = (dec->width + 31) / 32; |
|
|
const int rmb_height = (dec->height + 31) / 32; |
|
|
const int edged_width2 = dec->edged_width /2; |
|
|
int i,j; |
|
|
|
|
|
/* horizontal deblocking */ |
|
|
|
|
|
for (j = 1; j < rmb_height*2; j++) // luma: j,i in block units |
|
|
for (i = 0; i < rmb_width*2; i++) |
|
|
{ |
|
|
if (dec->mbs[(j-1)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED || |
|
|
dec->mbs[(j+0)/2*dec->mb_width + (i/2)].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
xvid_HFilter_31_C(dec->cur.y + (j*16 - 1)*dec->edged_width + i*16, |
|
|
dec->cur.y + (j*16 + 0)*dec->edged_width + i*16, 2); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 1; j < rmb_height; j++) // chroma |
|
|
for (i = 0; i < rmb_width; i++) |
|
|
{ |
|
|
if (dec->mbs[(j-1)*dec->mb_width + i].mode != MODE_NOT_CODED || |
|
|
dec->mbs[(j+0)*dec->mb_width + i].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
hfilter_31(dec->cur.u + (j*16 - 1)*edged_width2 + i*16, |
|
|
dec->cur.u + (j*16 + 0)*edged_width2 + i*16, 2); |
|
|
hfilter_31(dec->cur.v + (j*16 - 1)*edged_width2 + i*16, |
|
|
dec->cur.v + (j*16 + 0)*edged_width2 + i*16, 2); |
|
|
} |
|
|
} |
|
|
|
|
|
/* vertical deblocking */ |
|
|
|
|
|
for (j = 0; j < rmb_height*2; j++) // luma: i,j in block units |
|
|
for (i = 1; i < rmb_width*2; i++) |
|
|
{ |
|
|
if (dec->mbs[(j/2)*dec->mb_width + (i-1)/2].mode != MODE_NOT_CODED || |
|
|
dec->mbs[(j/2)*dec->mb_width + (i+0)/2].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
vfilter_31(dec->cur.y + (j*16)*dec->edged_width + i*16 - 1, |
|
|
dec->cur.y + (j*16)*dec->edged_width + i*16 + 0, |
|
|
dec->edged_width, 2); |
|
|
} |
|
|
} |
|
|
|
|
|
for (j = 0; j < rmb_height; j++) // chroma |
|
|
for (i = 1; i < rmb_width; i++) |
|
|
{ |
|
|
if (dec->mbs[j*dec->mb_width + i - 1].mode != MODE_NOT_CODED || |
|
|
dec->mbs[j*dec->mb_width + i + 0].mode != MODE_NOT_CODED) |
|
|
{ |
|
|
vfilter_31(dec->cur.u + (j*16)*edged_width2 + i*16 - 1, |
|
|
dec->cur.u + (j*16)*edged_width2 + i*16 + 0, |
|
|
edged_width2, 2); |
|
|
vfilter_31(dec->cur.v + (j*16)*edged_width2 + i*16 - 1, |
|
|
dec->cur.v + (j*16)*edged_width2 + i*16 + 0, |
|
|
edged_width2, 2); |
|
|
} |
|
|
} |
|
1379 |
} |
} |
1380 |
|
#endif |
1381 |
|
frame->length = BitstreamPos(&bs) / 8; |
1382 |
|
|
|
BitstreamByteAlign(&bs); |
|
1383 |
|
|
1384 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1385 |
// test if no B_VOP |
// test if no B_VOP |
1386 |
if (dec->low_delay || dec->frames == 0 || ((dec->packed_mode) && !(frame->length > BitstreamPos(&bs) / 8))) { |
if (dec->low_delay || dec->frames == 0) { |
1387 |
#endif |
#endif |
1388 |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1389 |
frame->image, frame->stride, frame->colorspace, dec->interlacing); |
frame->image, frame->stride, frame->colorspace); |
1390 |
|
|
1391 |
#ifdef BFRAMES_DEC |
#ifdef BFRAMES_DEC |
1392 |
} else { |
} else { |
1393 |
if (dec->frames >= 1 && !(dec->packed_mode)) { |
if (dec->frames >= 1) { |
1394 |
start_timer(); |
start_timer(); |
1395 |
if ((vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP)) { |
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1396 |
image_output(&dec->refn[0], dec->width, dec->height, |
image_output(&dec->refn[0], dec->width, dec->height, |
1397 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, frame->image, frame->stride, |
1398 |
frame->colorspace, dec->interlacing); |
frame->colorspace); |
1399 |
} else if (vop_type == B_VOP) { |
} else if (vop_type == B_VOP) { |
1400 |
image_output(&dec->cur, dec->width, dec->height, |
image_output(&dec->cur, dec->width, dec->height, |
1401 |
dec->edged_width, frame->image, frame->stride, |
dec->edged_width, frame->image, frame->stride, |
1402 |
frame->colorspace, dec->interlacing); |
frame->colorspace); |
1403 |
} |
} |
1404 |
stop_conv_timer(); |
stop_conv_timer(); |
1405 |
} |
} |
1406 |
} |
} |
1407 |
#endif |
#endif |
1408 |
|
|
1409 |
if (vop_type == I_VOP || vop_type == P_VOP || vop_type == S_VOP) { |
if (vop_type == I_VOP || vop_type == P_VOP) { |
1410 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1411 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1412 |
|
|
1422 |
mb_swap(&dec->mbs, &dec->last_mbs); |
mb_swap(&dec->mbs, &dec->last_mbs); |
1423 |
} |
} |
1424 |
|
|
|
|
|
|
if (success == 0 && dec->packed_mode) |
|
|
{ |
|
|
success = 1; |
|
|
// if (frame->length > BitstreamPos(&bs) / 8) // multiple vops packed together |
|
|
goto start; |
|
|
} |
|
|
|
|
|
done : |
|
|
|
|
|
frame->length = BitstreamPos(&bs) / 8; |
|
|
|
|
|
if (stats) |
|
|
{ |
|
|
stats->notify = XVID_DEC_VOP; |
|
|
stats->data.vop.time_base = (int)dec->time_base; |
|
|
stats->data.vop.time_increment = 0; //XXX: todo |
|
|
} |
|
|
|
|
1425 |
emms(); |
emms(); |
1426 |
|
|
1427 |
stop_global_timer(); |
stop_global_timer(); |