53 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
54 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
55 |
* 22.12.2001 lock based interpolation |
* 22.12.2001 lock based interpolation |
56 |
* 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> |
57 |
* |
* |
58 |
* $Id$ |
* $Id$ |
59 |
* |
* |
79 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
80 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
81 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
|
#include "image/font.h" |
|
82 |
|
|
83 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
84 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
85 |
#include "utils/timer.h" |
#include "utils/timer.h" |
86 |
#include "utils/emms.h" |
#include "utils/emms.h" |
|
#include "motion/motion.h" |
|
87 |
|
|
88 |
#include "image/image.h" |
#include "image/image.h" |
89 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
90 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
91 |
|
|
92 |
int |
int |
93 |
decoder_resize(DECODER * dec) |
decoder_create(XVID_DEC_PARAM * param) |
94 |
{ |
{ |
95 |
/* free existing */ |
DECODER *dec; |
|
|
|
|
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); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
96 |
|
|
97 |
if (dec->last_mbs) |
dec = xvid_malloc(sizeof(DECODER), CACHE_LINE); |
98 |
xvid_free(dec->last_mbs); |
if (dec == NULL) { |
99 |
if (dec->mbs) |
return XVID_ERR_MEMORY; |
100 |
xvid_free(dec->mbs); |
} |
101 |
|
param->handle = dec; |
102 |
|
|
103 |
/* realloc */ |
dec->width = param->width; |
104 |
|
dec->height = param->height; |
105 |
|
|
106 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
107 |
dec->mb_height = (dec->height + 15) / 16; |
dec->mb_height = (dec->height + 15) / 16; |
108 |
|
|
109 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
110 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
111 |
|
dec->low_delay = 0; |
112 |
|
|
113 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
114 |
xvid_free(dec); |
xvid_free(dec); |
120 |
xvid_free(dec); |
xvid_free(dec); |
121 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
122 |
} |
} |
|
|
|
123 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
124 |
// for support B-frame to reference last 2 frame |
// for support B-frame to reference last 2 frame |
125 |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
128 |
xvid_free(dec); |
xvid_free(dec); |
129 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
130 |
} |
} |
131 |
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
if (image_create(&dec->refn[2], dec->edged_width, dec->edged_height)) { |
132 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
133 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
134 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
136 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
137 |
} |
} |
138 |
|
|
|
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; |
|
|
} |
|
|
|
|
139 |
dec->mbs = |
dec->mbs = |
140 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
141 |
CACHE_LINE); |
CACHE_LINE); |
143 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
144 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
145 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
146 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
147 |
xvid_free(dec); |
xvid_free(dec); |
148 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
149 |
} |
} |
150 |
|
|
151 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
152 |
|
|
153 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
160 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
161 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
162 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
163 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
164 |
xvid_free(dec); |
xvid_free(dec); |
165 |
return XVID_ERR_MEMORY; |
return XVID_ERR_MEMORY; |
166 |
} |
} |
167 |
|
|
168 |
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); |
169 |
|
|
|
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->tmp); |
|
|
image_null(&dec->qtmp); |
|
|
|
|
|
dec->mbs = NULL; |
|
|
dec->last_mbs = NULL; |
|
|
|
|
170 |
init_timer(); |
init_timer(); |
171 |
|
|
172 |
// add by chenm001 <chenm001@163.com> |
// add by chenm001 <chenm001@163.com> |
173 |
// for support B-frame to save reference frame's time |
// for support B-frame to save reference frame's time |
174 |
dec->frames = 0; |
dec->frames = -1; |
175 |
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); |
|
176 |
|
|
|
if (dec->fixed_dimensions) |
|
|
return decoder_resize(dec); |
|
|
else |
|
177 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
178 |
} |
} |
179 |
|
|
185 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
186 |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
187 |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
188 |
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
image_destroy(&dec->refn[2], dec->edged_width, dec->edged_height); |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
189 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
190 |
xvid_free(dec); |
xvid_free(dec); |
191 |
|
|
214 |
Bitstream * bs, |
Bitstream * bs, |
215 |
const uint32_t quant, |
const uint32_t quant, |
216 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
217 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
218 |
{ |
{ |
219 |
|
|
220 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
227 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
228 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
229 |
|
|
|
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{ |
|
230 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
231 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
232 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
233 |
|
|
234 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
memset(block, 0, 6 * 64 * sizeof(int16_t)); // clear |
235 |
|
|
268 |
start_timer(); |
start_timer(); |
269 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
270 |
{ |
{ |
271 |
int direction = dec->alternate_vertical_scan ? |
get_intra_block(bs, &block[i * 64], pMB->acpred_directions[i], |
272 |
2 : pMB->acpred_directions[i]; |
start_coeff); |
|
|
|
|
get_intra_block(bs, &block[i * 64], direction, start_coeff); |
|
273 |
} |
} |
274 |
stop_coding_timer(); |
stop_coding_timer(); |
275 |
|
|
288 |
start_timer(); |
start_timer(); |
289 |
idct(&data[i * 64]); |
idct(&data[i * 64]); |
290 |
stop_idct_timer(); |
stop_idct_timer(); |
|
|
|
291 |
} |
} |
292 |
|
|
293 |
if (dec->interlacing && pMB->field_dct) { |
if (dec->interlacing && pMB->field_dct) { |
296 |
} |
} |
297 |
|
|
298 |
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{ |
|
299 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
300 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
301 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
302 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
303 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
304 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
305 |
stop_transfer_timer(); |
stop_transfer_timer(); |
306 |
} |
} |
307 |
|
|
311 |
|
|
312 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
313 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
314 |
|
static const uint32_t roundtab[16] = |
315 |
|
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
316 |
|
|
317 |
|
|
318 |
// decode an inter macroblock |
// decode an inter macroblock |
319 |
|
|
326 |
const uint32_t cbp, |
const uint32_t cbp, |
327 |
Bitstream * bs, |
Bitstream * bs, |
328 |
const uint32_t quant, |
const uint32_t quant, |
329 |
const uint32_t rounding, |
const uint32_t rounding) |
|
const int reduced_resolution) |
|
330 |
{ |
{ |
331 |
|
|
332 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
334 |
|
|
335 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
336 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
337 |
uint32_t next_block = stride * (reduced_resolution ? 16 : 8); |
uint32_t next_block = stride * 8; |
338 |
uint32_t i; |
uint32_t i; |
339 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
340 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
|
|
|
341 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
|
VECTOR mv[4]; /* local copy of mvs */ |
|
342 |
|
|
|
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{ |
|
343 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
344 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
345 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
for (i = 0; i < 4; i++) |
|
|
mv[i] = pMB->mvs[i]; |
|
|
} |
|
346 |
|
|
347 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) { |
348 |
uv_dx = mv[0].x; |
uv_dx = pMB->mvs[0].x; |
349 |
uv_dy = mv[0].y; |
uv_dy = pMB->mvs[0].y; |
350 |
|
|
351 |
if (dec->quarterpel) |
if (dec->quarterpel) |
352 |
{ |
{ |
353 |
uv_dx /= 2; |
uv_dx = (uv_dx >> 1) | (uv_dx & 1); |
354 |
uv_dy /= 2; |
uv_dy = (uv_dy >> 1) | (uv_dy & 1); |
355 |
} |
} |
356 |
|
|
357 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
358 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
359 |
|
} else { |
360 |
start_timer(); |
int sum; |
361 |
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); |
|
362 |
|
|
363 |
} |
if (dec->quarterpel) |
|
else |
|
364 |
{ |
{ |
365 |
if(dec->quarterpel) { |
sum /= 2; |
|
interpolate16x16_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
366 |
} |
} |
|
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); |
|
|
} |
|
|
stop_comp_timer(); |
|
|
|
|
|
} else { /* MODE_INTER4V */ |
|
|
int sum; |
|
367 |
|
|
368 |
if(dec->quarterpel) |
uv_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
|
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; |
|
369 |
|
|
370 |
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; |
371 |
|
|
372 |
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) |
|
373 |
{ |
{ |
374 |
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
sum /= 2; |
375 |
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); |
|
376 |
|
|
377 |
// set_block(pY_Cur, stride, 32, 32, 127); |
uv_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
378 |
} |
} |
379 |
else |
|
380 |
{ |
start_timer(); |
381 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
382 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
DPRINTF(DPRINTF_DEBUG, "QUARTERPEL\n"); |
383 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos, |
384 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
385 |
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, 16*x_pos + 8, 16*y_pos, |
386 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
387 |
mv[1].x, mv[1].y, stride, rounding); |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, 16*x_pos, 16*y_pos + 8, |
388 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y, dec->qtmp.y, dec->qtmp.y + 64, |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
389 |
dec->qtmp.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, |
390 |
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->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
391 |
} |
} |
392 |
else { |
else { |
393 |
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, |
394 |
mv[0].x, mv[0].y, stride, rounding); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, rounding); |
395 |
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, |
396 |
mv[1].x, mv[1].y, stride, rounding); |
pMB->mvs[1].x, pMB->mvs[1].y, stride, rounding); |
397 |
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, |
398 |
mv[2].x, mv[2].y, stride, rounding); |
pMB->mvs[2].x, pMB->mvs[2].y, stride, rounding); |
399 |
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, |
400 |
mv[3].x, mv[3].y, stride, rounding); |
pMB->mvs[3].x, pMB->mvs[3].y, stride, rounding); |
401 |
} |
} |
402 |
|
|
403 |
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, |
404 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
405 |
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, |
406 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
|
} |
|
407 |
stop_comp_timer(); |
stop_comp_timer(); |
|
} |
|
408 |
|
|
409 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
410 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
411 |
{ |
{ |
412 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
413 |
|
|
414 |
start_timer(); |
start_timer(); |
415 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
416 |
stop_coding_timer(); |
stop_coding_timer(); |
417 |
|
|
418 |
start_timer(); |
start_timer(); |
435 |
} |
} |
436 |
|
|
437 |
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 |
|
|
{ |
|
438 |
if (cbp & 32) |
if (cbp & 32) |
439 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
440 |
if (cbp & 16) |
if (cbp & 16) |
447 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
448 |
if (cbp & 1) |
if (cbp & 1) |
449 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
450 |
stop_transfer_timer(); |
stop_transfer_timer(); |
451 |
} |
} |
452 |
|
|
454 |
void |
void |
455 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
456 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
457 |
int quant, |
int quant, |
458 |
int intra_dc_threshold) |
int intra_dc_threshold) |
459 |
{ |
{ |
460 |
uint32_t bound; |
uint32_t bound; |
461 |
uint32_t x, y; |
uint32_t x, y; |
|
uint32_t mb_width = dec->mb_width; |
|
|
uint32_t mb_height = dec->mb_height; |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
462 |
|
|
463 |
bound = 0; |
bound = 0; |
464 |
|
|
465 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
466 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
467 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
468 |
uint32_t mcbpc; |
uint32_t mcbpc; |
469 |
uint32_t cbpc; |
uint32_t cbpc; |
476 |
|
|
477 |
if (check_resync_marker(bs, 0)) |
if (check_resync_marker(bs, 0)) |
478 |
{ |
{ |
479 |
bound = read_video_packet_header(bs, dec, 0, |
bound = read_video_packet_header(bs, 0, &quant); |
480 |
&quant, NULL, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
481 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
482 |
} |
} |
483 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
484 |
|
|
509 |
|
|
510 |
if (dec->interlacing) { |
if (dec->interlacing) { |
511 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
512 |
DPRINTF(DPRINTF_MB,"deci: field_dct: %i", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "deci: field_dct: %d", mb->field_dct); |
513 |
} |
} |
514 |
|
|
515 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
516 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
|
|
|
517 |
} |
} |
518 |
if(dec->out_frm) |
if(dec->out_frm) |
519 |
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); |
520 |
|
|
521 |
} |
} |
522 |
|
|
523 |
} |
} |
529 |
int x, |
int x, |
530 |
int y, |
int y, |
531 |
int k, |
int k, |
532 |
VECTOR * ret_mv, |
VECTOR * mv, |
533 |
int fcode, |
int fcode, |
534 |
const int bound) |
const int bound) |
535 |
{ |
{ |
540 |
int range = (64 * scale_fac); |
int range = (64 * scale_fac); |
541 |
|
|
542 |
VECTOR pmv; |
VECTOR pmv; |
543 |
VECTOR mv; |
int mv_x, mv_y; |
544 |
|
|
545 |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
pmv = get_pmv2(dec->mbs, dec->mb_width, bound, x, y, k); |
546 |
|
|
547 |
mv.x = get_mv(bs, fcode); |
mv_x = get_mv(bs, fcode); |
548 |
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); |
|
549 |
|
|
550 |
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; |
|
551 |
|
|
552 |
if (mv.x < low) { |
mv_x += pmv.x; |
553 |
mv.x += range; |
mv_y += pmv.y; |
|
} else if (mv.x > high) { |
|
|
mv.x -= range; |
|
|
} |
|
554 |
|
|
555 |
if (mv.y < low) { |
if (mv_x < low) { |
556 |
mv.y += range; |
mv_x += range; |
557 |
} else if (mv.y > high) { |
} else if (mv_x > high) { |
558 |
mv.y -= range; |
mv_x -= range; |
559 |
} |
} |
560 |
|
|
561 |
ret_mv->x = mv.x; |
if (mv_y < low) { |
562 |
ret_mv->y = mv.y; |
mv_y += range; |
563 |
|
} else if (mv_y > high) { |
564 |
|
mv_y -= range; |
565 |
} |
} |
566 |
|
|
567 |
|
mv->x = mv_x; |
568 |
|
mv->y = mv_y; |
569 |
|
|
|
|
|
|
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; |
|
570 |
} |
} |
571 |
|
|
572 |
|
|
|
/* for P_VOP set gmc_mv to NULL */ |
|
573 |
void |
void |
574 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
575 |
Bitstream * bs, |
Bitstream * bs, |
576 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
577 |
int quant, |
int quant, |
578 |
int fcode, |
int fcode, |
579 |
int intra_dc_threshold, |
int intra_dc_threshold) |
|
VECTOR * gmc_mv) |
|
580 |
{ |
{ |
581 |
|
|
582 |
uint32_t x, y; |
uint32_t x, y; |
583 |
uint32_t bound; |
uint32_t bound; |
584 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
|
uint32_t mb_width = dec->mb_width; |
|
|
uint32_t mb_height = dec->mb_height; |
|
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
585 |
|
|
586 |
start_timer(); |
start_timer(); |
587 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
588 |
dec->width, dec->height); |
dec->width, dec->height, dec->interlacing); |
589 |
stop_edges_timer(); |
stop_edges_timer(); |
590 |
|
|
591 |
bound = 0; |
bound = 0; |
592 |
|
|
593 |
for (y = 0; y < mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
594 |
cp_mb = st_mb = 0; |
cp_mb = st_mb = 0; |
595 |
for (x = 0; x < mb_width; x++) { |
for (x = 0; x < dec->mb_width; x++) { |
596 |
MACROBLOCK *mb; |
MACROBLOCK *mb; |
597 |
|
|
598 |
// skip stuffing |
// skip stuffing |
601 |
|
|
602 |
if (check_resync_marker(bs, fcode - 1)) |
if (check_resync_marker(bs, fcode - 1)) |
603 |
{ |
{ |
604 |
bound = read_video_packet_header(bs, dec, fcode - 1, |
bound = read_video_packet_header(bs, fcode - 1, &quant); |
605 |
&quant, &fcode, NULL, &intra_dc_threshold); |
x = bound % dec->mb_width; |
606 |
x = bound % mb_width; |
y = bound / dec->mb_width; |
|
y = bound / mb_width; |
|
607 |
} |
} |
608 |
mb = &dec->mbs[y * dec->mb_width + x]; |
mb = &dec->mbs[y * dec->mb_width + x]; |
609 |
|
|
618 |
uint32_t cbpy; |
uint32_t cbpy; |
619 |
uint32_t cbp; |
uint32_t cbp; |
620 |
uint32_t intra; |
uint32_t intra; |
|
int mcsel = 0; // mcsel: '0'=local motion, '1'=GMC |
|
621 |
|
|
622 |
cp_mb++; |
cp_mb++; |
623 |
mcbpc = get_mcbpc_inter(bs); |
mcbpc = get_mcbpc_inter(bs); |
634 |
acpred_flag = BitstreamGetBit(bs); |
acpred_flag = BitstreamGetBit(bs); |
635 |
} |
} |
636 |
|
|
|
if (gmc_mv && (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q)) |
|
|
{ |
|
|
mcsel = BitstreamGetBit(bs); |
|
|
} |
|
|
|
|
637 |
cbpy = get_cbpy(bs, intra); |
cbpy = get_cbpy(bs, intra); |
638 |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
DPRINTF(DPRINTF_MB, "cbpy %i", cbpy); |
639 |
|
|
655 |
if (dec->interlacing) { |
if (dec->interlacing) { |
656 |
if (cbp || intra) { |
if (cbp || intra) { |
657 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
658 |
DPRINTF(DPRINTF_MB,"decp: field_dct: %i", mb->field_dct); |
DPRINTF(DPRINTF_DEBUG, "decp: field_dct: %d", mb->field_dct); |
659 |
} |
} |
660 |
|
|
661 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
662 |
mb->field_pred = BitstreamGetBit(bs); |
mb->field_pred = BitstreamGetBit(bs); |
663 |
DPRINTF(DPRINTF_MB, "decp: field_pred: %i", mb->field_pred); |
DPRINTF(DPRINTF_DEBUG, "decp: field_pred: %d", mb->field_pred); |
664 |
|
|
665 |
if (mb->field_pred) { |
if (mb->field_pred) { |
666 |
mb->field_for_top = BitstreamGetBit(bs); |
mb->field_for_top = BitstreamGetBit(bs); |
667 |
DPRINTF(DPRINTF_MB,"decp: field_for_top: %i", mb->field_for_top); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_top: %d", mb->field_for_top); |
668 |
mb->field_for_bot = BitstreamGetBit(bs); |
mb->field_for_bot = BitstreamGetBit(bs); |
669 |
DPRINTF(DPRINTF_MB,"decp: field_for_bot: %i", mb->field_for_bot); |
DPRINTF(DPRINTF_DEBUG, "decp: field_for_bot: %d", mb->field_for_bot); |
670 |
} |
} |
671 |
} |
} |
672 |
} |
} |
673 |
|
|
674 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
675 |
|
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) { |
|
676 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], |
677 |
fcode, bound); |
fcode, bound); |
678 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], |
698 |
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 = |
699 |
0; |
0; |
700 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
701 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
702 |
continue; |
continue; |
703 |
} |
} |
704 |
|
|
705 |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbinter(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
706 |
rounding, reduced_resolution); |
rounding); |
707 |
|
} else // not coded |
|
} |
|
|
else if (gmc_mv) /* not coded S_VOP macroblock */ |
|
708 |
{ |
{ |
709 |
mb->mode = MODE_NOT_CODED_GMC; |
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; |
|
710 |
|
|
711 |
|
mb->mode = MODE_NOT_CODED; |
712 |
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; |
713 |
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; |
714 |
|
|
715 |
// copy macroblock directly from ref to cur |
// copy macroblock directly from ref to cur |
716 |
|
|
717 |
start_timer(); |
start_timer(); |
718 |
|
|
719 |
if (reduced_resolution) |
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
720 |
{ |
(16 * x), |
721 |
transfer32x32_copy(dec->cur.y + (32*y)*dec->edged_width + (32*x), |
dec->refn[0].y + (16 * y) * dec->edged_width + |
722 |
dec->refn[0].y + (32*y)*dec->edged_width + (32*x), |
(16 * x), dec->edged_width); |
723 |
dec->edged_width); |
|
724 |
|
transfer8x8_copy(dec->cur.y + (16 * y) * dec->edged_width + |
725 |
transfer16x16_copy(dec->cur.u + (16*y)*dec->edged_width/2 + (16*x), |
(16 * x + 8), |
726 |
dec->refn[0].u + (16*y)*dec->edged_width/2 + (16*x), |
dec->refn[0].y + (16 * y) * dec->edged_width + |
727 |
dec->edged_width/2); |
(16 * x + 8), dec->edged_width); |
728 |
|
|
729 |
transfer16x16_copy(dec->cur.v + (16*y)*dec->edged_width/2 + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
730 |
dec->refn[0].v + (16*y)*dec->edged_width/2 + (16*x), |
(16 * x), |
731 |
dec->edged_width/2); |
dec->refn[0].y + (16 * y + |
732 |
} |
8) * dec->edged_width + |
733 |
else |
(16 * x), dec->edged_width); |
734 |
{ |
|
735 |
transfer16x16_copy(dec->cur.y + (16*y)*dec->edged_width + (16*x), |
transfer8x8_copy(dec->cur.y + (16 * y + 8) * dec->edged_width + |
736 |
dec->refn[0].y + (16*y)*dec->edged_width + (16*x), |
(16 * x + 8), |
737 |
dec->edged_width); |
dec->refn[0].y + (16 * y + |
738 |
|
8) * dec->edged_width + |
739 |
transfer8x8_copy(dec->cur.u + (8*y)*dec->edged_width/2 + (8*x), |
(16 * x + 8), dec->edged_width); |
740 |
dec->refn[0].u + (8*y)*dec->edged_width/2 + (8*x), |
|
741 |
|
transfer8x8_copy(dec->cur.u + (8 * y) * dec->edged_width / 2 + |
742 |
|
(8 * x), |
743 |
|
dec->refn[0].u + |
744 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
745 |
dec->edged_width/2); |
dec->edged_width/2); |
746 |
|
|
747 |
transfer8x8_copy(dec->cur.v + (8*y)*dec->edged_width/2 + (8*x), |
transfer8x8_copy(dec->cur.v + (8 * y) * dec->edged_width / 2 + |
748 |
dec->refn[0].v + (8*y)*dec->edged_width/2 + (8*x), |
(8 * x), |
749 |
|
dec->refn[0].v + |
750 |
|
(8 * y) * dec->edged_width / 2 + (8 * x), |
751 |
dec->edged_width/2); |
dec->edged_width/2); |
|
} |
|
|
|
|
752 |
stop_transfer_timer(); |
stop_transfer_timer(); |
|
|
|
753 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
754 |
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); |
755 |
cp_mb = 0; |
cp_mb = 0; |
841 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
842 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
843 |
|
|
844 |
if (dec->quarterpel) |
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
845 |
{ |
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]; |
|
846 |
} else { |
} else { |
847 |
int sum; |
int sum; |
848 |
|
|
|
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 |
|
849 |
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; |
850 |
|
uv_dx = |
851 |
|
(sum == |
852 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
853 |
|
(ABS(sum) / 16) * 2)); |
854 |
|
|
|
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 |
|
855 |
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; |
856 |
|
uv_dy = |
857 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
858 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
859 |
|
(ABS(sum) / 16) * 2)); |
860 |
} |
} |
861 |
|
|
862 |
start_timer(); |
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 { |
|
863 |
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, |
864 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
865 |
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, |
866 |
pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
16 * y_pos, pMB->mvs[1].x, pMB->mvs[1].y, stride, 0); |
867 |
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, |
868 |
pMB->mvs[2].x, pMB->mvs[2].y, stride, 0); |
16 * y_pos + 8, pMB->mvs[2].x, pMB->mvs[2].y, stride, |
869 |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16*x_pos + 8, 16*y_pos + 8, |
0); |
870 |
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
interpolate8x8_switch(dec->cur.y, dec->refn[ref].y, 16 * x_pos + 8, |
871 |
} |
16 * y_pos + 8, pMB->mvs[3].x, pMB->mvs[3].y, stride, |
872 |
|
0); |
873 |
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, |
874 |
uv_dx, uv_dy, stride2, 0); |
uv_dx, uv_dy, stride2, 0); |
875 |
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, |
877 |
stop_comp_timer(); |
stop_comp_timer(); |
878 |
|
|
879 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
880 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
881 |
{ |
{ |
882 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
883 |
|
|
884 |
start_timer(); |
start_timer(); |
885 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
886 |
stop_coding_timer(); |
stop_coding_timer(); |
887 |
|
|
888 |
start_timer(); |
start_timer(); |
920 |
stop_transfer_timer(); |
stop_transfer_timer(); |
921 |
} |
} |
922 |
|
|
923 |
|
|
924 |
// add by MinChen <chenm001@163.com> |
// add by MinChen <chenm001@163.com> |
925 |
// decode an B-frame direct & inter macroblock |
// decode an B-frame direct & inter macroblock |
926 |
void |
void |
955 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
956 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
957 |
|
|
958 |
|
uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
959 |
|
uv_dy = (uv_dy & 3) ? (uv_dy >> 1) | 1 : uv_dy / 2; |
960 |
|
|
961 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
962 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
963 |
|
|
964 |
if (dec->quarterpel) |
b_uv_dx = (uv_dx & 3) ? (uv_dx >> 1) | 1 : uv_dx / 2; |
965 |
{ |
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]; |
|
966 |
} else { |
} else { |
967 |
int sum; |
int sum; |
968 |
|
|
|
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 |
|
969 |
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; |
970 |
|
uv_dx = |
971 |
|
(sum == |
972 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
973 |
|
(ABS(sum) / 16) * 2)); |
974 |
|
|
|
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 |
|
975 |
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; |
976 |
|
uv_dy = |
977 |
uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
978 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
979 |
|
(ABS(sum) / 16) * 2)); |
980 |
if(dec->quarterpel) |
|
981 |
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 = |
982 |
else |
pMB->b_mvs[0].x + pMB->b_mvs[1].x + pMB->b_mvs[2].x + |
983 |
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; |
984 |
|
b_uv_dx = |
985 |
b_uv_dx = (sum >> 3) + roundtab_76[sum & 0xf]; |
(sum == |
986 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
987 |
if(dec->quarterpel) |
(ABS(sum) / 16) * 2)); |
988 |
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); |
|
989 |
else |
sum = |
990 |
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 + |
991 |
|
pMB->b_mvs[3].y; |
992 |
b_uv_dy = (sum >> 3) + roundtab_76[sum & 0xf]; |
b_uv_dy = |
993 |
|
(sum == |
994 |
|
0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + |
995 |
|
(ABS(sum) / 16) * 2)); |
996 |
} |
} |
997 |
|
|
998 |
|
|
999 |
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->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_quarterpel(dec->cur.y, forward.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); |
|
|
interpolate8x8_quarterpel(dec->cur.y, forward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.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->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.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->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->mvs[3].x, pMB->mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
1000 |
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, |
1001 |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
pMB->mvs[0].x, pMB->mvs[0].y, stride, 0); |
1002 |
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, |
1006 |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->cur.y, forward.y, 16 * x_pos + 8, |
1007 |
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, |
1008 |
0); |
0); |
|
} |
|
|
|
|
1009 |
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, |
1010 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1011 |
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, |
1012 |
uv_dy, stride2, 0); |
uv_dy, stride2, 0); |
1013 |
|
|
1014 |
|
|
1015 |
if(dec->quarterpel) { |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, 16 * y_pos, |
|
if((pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q)) |
|
|
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
else { |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
|
|
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
|
|
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
|
|
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
|
|
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
|
|
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
|
|
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
|
|
} |
|
|
} |
|
|
else { |
|
|
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
|
1016 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1017 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1018 |
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, |
1019 |
0); |
0); |
1020 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos, |
1021 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, |
1022 |
stride, 0); |
stride, 0); |
1023 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_switch(dec->refn[2].y, backward.y, 16 * x_pos + 8, |
1024 |
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, |
1025 |
stride, 0); |
stride, 0); |
1026 |
} |
interpolate8x8_switch(dec->refn[2].u, backward.u, 8 * x_pos, 8 * y_pos, |
|
|
|
|
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
|
1027 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1028 |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_switch(dec->refn[2].v, backward.v, 8 * x_pos, 8 * y_pos, |
1029 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1030 |
|
|
1031 |
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, |
1032 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
stride); |
1033 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, 16 * y_pos, |
1034 |
stride, 1, 8); |
stride); |
1035 |
|
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos, 16 * y_pos + 8, |
1036 |
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
stride); |
1037 |
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
interpolate8x8_c(dec->cur.y, dec->refn[2].y, 16 * x_pos + 8, |
1038 |
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
16 * y_pos + 8, stride); |
1039 |
stride, 1, 8); |
interpolate8x8_c(dec->cur.u, dec->refn[2].u, 8 * x_pos, 8 * y_pos, |
1040 |
|
stride2); |
1041 |
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, |
1042 |
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
stride2); |
|
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); |
|
|
|
|
1043 |
stop_comp_timer(); |
stop_comp_timer(); |
1044 |
|
|
1045 |
for (i = 0; i < 6; i++) { |
for (i = 0; i < 6; i++) { |
|
int direction = dec->alternate_vertical_scan ? 2 : 0; |
|
|
|
|
1046 |
if (cbp & (1 << (5 - i))) // coded |
if (cbp & (1 << (5 - i))) // coded |
1047 |
{ |
{ |
1048 |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i * 64], 0, 64 * sizeof(int16_t)); // clear |
1049 |
|
|
1050 |
start_timer(); |
start_timer(); |
1051 |
get_inter_block(bs, &block[i * 64], direction); |
get_inter_block(bs, &block[i * 64]); |
1052 |
stop_coding_timer(); |
stop_coding_timer(); |
1053 |
|
|
1054 |
start_timer(); |
start_timer(); |
1143 |
|
|
1144 |
start_timer(); |
start_timer(); |
1145 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1146 |
dec->width, dec->height); |
dec->width, dec->height, dec->interlacing); |
1147 |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1148 |
dec->width, dec->height); |
dec->width, dec->height, dec->interlacing); |
1149 |
stop_edges_timer(); |
stop_edges_timer(); |
1150 |
|
|
1151 |
#ifdef BFRAMES_DEC_DEBUG |
#ifdef BFRAMES_DEC_DEBUG |
1165 |
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] = |
1166 |
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; |
1167 |
|
|
1168 |
// skip if the co-located P_VOP macroblock is not coded |
// the last P_VOP is skip macroblock ? |
|
// note: gmc+not_coded isn't skipped |
|
|
|
|
1169 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1170 |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
//DEBUG2("Skip MB in B-frame at (X,Y)=!",x,y); |
1171 |
mb->cbp = 0; |
mb->cbp = 0; |
1279 |
break; |
break; |
1280 |
|
|
1281 |
default: |
default: |
1282 |
DPRINTF(DPRINTF_ERROR,"Not support B-frame mb_type = %i", mb->mb_type); |
DPRINTF(DPRINTF_ERROR, "Not support B-frame mb_type = %d", mb->mb_type); |
1283 |
} |
} |
1284 |
|
|
1285 |
} // end of FOR |
} // end of FOR |
1304 |
*mb2 = temp; |
*mb2 = temp; |
1305 |
} |
} |
1306 |
|
|
|
|
|
|
/* perform post processing if necessary, and output the image */ |
|
|
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
|
|
const XVID_DEC_FRAME * frame, int pp_disable) |
|
|
{ |
|
|
|
|
|
if ((frame->general & (XVID_DEC_DEBLOCKY|XVID_DEC_DEBLOCKUV)) && !pp_disable) /* post process */ |
|
|
{ |
|
|
/* note: image is stored to tmp */ |
|
|
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
|
|
image_deblock_rrv(&dec->tmp, dec->edged_width, |
|
|
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
|
|
8, frame->general); |
|
|
img = &dec->tmp; |
|
|
} |
|
|
|
|
|
image_output(img, dec->width, dec->height, |
|
|
dec->edged_width, frame->image, frame->stride, |
|
|
frame->colorspace, dec->interlacing); |
|
|
} |
|
|
|
|
|
|
|
1307 |
int |
int |
1308 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1309 |
XVID_DEC_FRAME * frame, XVID_DEC_STATS * stats) |
XVID_DEC_FRAME * frame) |
1310 |
{ |
{ |
1311 |
|
|
1312 |
Bitstream bs; |
Bitstream bs; |
1313 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1314 |
uint32_t quant; |
uint32_t quant; |
1315 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1316 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1317 |
uint32_t intra_dc_threshold; |
uint32_t intra_dc_threshold; |
|
VECTOR gmc_mv[5]; |
|
1318 |
uint32_t vop_type; |
uint32_t vop_type; |
|
int success = 0; |
|
|
int output = 0; |
|
|
int seen_something = 0; |
|
1319 |
|
|
1320 |
start_global_timer(); |
start_global_timer(); |
1321 |
|
|
|
dec->low_delay_default = (frame->general & XVID_DEC_LOWDELAY); |
|
1322 |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
dec->out_frm = (frame->colorspace == XVID_CSP_EXTERN) ? frame->image : NULL; |
1323 |
|
|
|
if ((frame->general & XVID_DEC_DISCONTINUITY)) |
|
|
dec->frames = 0; |
|
|
|
|
|
if (frame->length < 0) /* decoder flush */ |
|
|
{ |
|
|
/* if not decoding "low_delay/packed", and this isn't low_delay and |
|
|
we have a reference frame, then outout the reference frame */ |
|
|
if (!(dec->low_delay_default && dec->packed_mode) && !dec->low_delay && dec->frames>0) |
|
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
|
|
|
frame->length = 0; |
|
|
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 |
|
|
} |
|
|
|
|
|
emms(); |
|
|
|
|
|
stop_global_timer(); |
|
|
return XVID_ERR_OK; |
|
|
} |
|
|
|
|
1324 |
BitstreamInit(&bs, frame->bitstream, frame->length); |
BitstreamInit(&bs, frame->bitstream, frame->length); |
1325 |
|
|
1326 |
// XXX: 0x7f is only valid whilst decoding vfw xvid/divx5 avi's |
// add by chenm001 <chenm001@163.com> |
1327 |
if(dec->low_delay_default && frame->length == 1 && BitstreamShowBits(&bs, 8) == 0x7f) |
// for support B-frame to reference last 2 frame |
1328 |
{ |
dec->frames++; |
1329 |
if (stats) |
vop_type = |
1330 |
stats->notify = XVID_DEC_VOP; |
BitstreamReadHeaders(&bs, dec, &rounding, &quant, &fcode_forward, |
1331 |
frame->length = 1; |
&fcode_backward, &intra_dc_threshold); |
|
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; |
|
|
} |
|
|
|
|
|
repeat: |
|
|
|
|
|
vop_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
|
|
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, gmc_mv); |
|
|
|
|
|
DPRINTF(DPRINTF_HEADER, "vop_type=%i, packed=%i, time=%i, time_pp=%i, time_bp=%i", |
|
|
vop_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
|
|
|
|
|
if (vop_type == - 1) |
|
|
{ |
|
|
if (success) goto done; |
|
|
return XVID_ERR_FAIL; |
|
|
} |
|
|
|
|
|
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 repeat; |
|
|
} |
|
1332 |
|
|
1333 |
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 |
1334 |
|
|
1335 |
|
switch (vop_type) { |
1336 |
|
case P_VOP: |
1337 |
|
decoder_pframe(dec, &bs, rounding, quant, fcode_forward, |
1338 |
|
intra_dc_threshold); |
1339 |
|
#ifdef BFRAMES_DEC |
1340 |
|
DEBUG1("P_VOP Time=", dec->time); |
1341 |
|
#endif |
1342 |
|
break; |
1343 |
|
|
|
/* packed_mode: special-N_VOP treament */ |
|
|
if (dec->packed_mode && vop_type == N_VOP) |
|
|
{ |
|
|
if (dec->low_delay_default && dec->frames > 0) |
|
|
{ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
/* ignore otherwise */ |
|
|
} |
|
|
else if (vop_type != B_VOP) |
|
|
{ |
|
|
switch(vop_type) |
|
|
{ |
|
1344 |
case I_VOP : |
case I_VOP : |
1345 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1346 |
break; |
#ifdef BFRAMES_DEC |
1347 |
case P_VOP : |
DEBUG1("I_VOP Time=", dec->time); |
1348 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
#endif |
|
fcode_forward, intra_dc_threshold, NULL); |
|
1349 |
break; |
break; |
1350 |
case S_VOP : |
|
1351 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
case B_VOP: |
1352 |
fcode_forward, intra_dc_threshold, gmc_mv); |
#ifdef BFRAMES_DEC |
1353 |
|
if (dec->time_pp > dec->time_bp) { |
1354 |
|
DEBUG1("B_VOP Time=", dec->time); |
1355 |
|
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
1356 |
|
} else { |
1357 |
|
DEBUG("broken B-frame!"); |
1358 |
|
} |
1359 |
|
#else |
1360 |
|
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1361 |
|
#endif |
1362 |
break; |
break; |
1363 |
case N_VOP : |
|
1364 |
|
case N_VOP: // vop not coded |
1365 |
|
// when low_delay==0, N_VOP's should interpolate between the past and future frames |
1366 |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
image_copy(&dec->cur, &dec->refn[0], dec->edged_width, dec->height); |
1367 |
break; |
break; |
|
} |
|
1368 |
|
|
1369 |
if (reduced_resolution) |
default: |
1370 |
{ |
return XVID_ERR_FAIL; |
|
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); |
|
1371 |
} |
} |
1372 |
|
|
1373 |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
#ifdef BFRAMES_DEC_DEBUG |
1374 |
if (!(dec->low_delay_default && dec->packed_mode)) |
if (frame->length != BitstreamPos(&bs) / 8){ |
1375 |
{ |
DEBUG2("InLen/UseLen",frame->length, BitstreamPos(&bs) / 8); |
|
if (dec->low_delay) |
|
|
{ |
|
|
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
|
else if (dec->frames > 0) /* is the reference frame valid? */ |
|
|
{ |
|
|
/* output the reference frame */ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
|
} |
|
1376 |
} |
} |
1377 |
|
#endif |
1378 |
|
frame->length = BitstreamPos(&bs) / 8; |
1379 |
|
|
|
image_swap(&dec->refn[0], &dec->refn[1]); |
|
|
image_swap(&dec->cur, &dec->refn[0]); |
|
|
mb_swap(&dec->mbs, &dec->last_mbs); |
|
|
dec->last_reduced_resolution = reduced_resolution; |
|
|
|
|
|
dec->frames++; |
|
|
seen_something = 1; |
|
|
|
|
|
}else{ /* B_VOP */ |
|
1380 |
|
|
1381 |
if (dec->low_delay) |
#ifdef BFRAMES_DEC |
1382 |
{ |
// test if no B_VOP |
1383 |
DPRINTF(DPRINTF_ERROR, "warning: bvop found in low_delay==1 stream"); |
if (dec->low_delay || dec->frames == 0) { |
1384 |
dec->low_delay = 1; |
#endif |
1385 |
} |
image_output(&dec->cur, dec->width, dec->height, dec->edged_width, |
1386 |
|
frame->image, frame->stride, frame->colorspace); |
1387 |
|
|
1388 |
if (dec->frames < 2) |
#ifdef BFRAMES_DEC |
|
{ |
|
|
/* attemping to decode a bvop without atleast 2 reference frames */ |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"broken b-frame, mising ref frames"); |
|
|
}else if (dec->time_pp <= dec->time_bp) { |
|
|
/* this occurs when dx50_bvop_compatibility==0 sequences are |
|
|
decoded in vfw. */ |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"broken b-frame, tpp=%i tbp=%i", dec->time_pp, dec->time_bp); |
|
1389 |
}else{ |
}else{ |
1390 |
decoder_bframe(dec, &bs, quant, fcode_forward, fcode_backward); |
if (dec->frames >= 1) { |
1391 |
} |
start_timer(); |
1392 |
|
if ((vop_type == I_VOP || vop_type == P_VOP)) { |
1393 |
decoder_output(dec, &dec->cur, dec->mbs, frame, reduced_resolution); |
image_output(&dec->refn[0], dec->width, dec->height, |
1394 |
output = 1; |
dec->edged_width, frame->image, frame->stride, |
1395 |
dec->frames++; |
frame->colorspace); |
1396 |
} |
} else if (vop_type == B_VOP) { |
1397 |
|
image_output(&dec->cur, dec->width, dec->height, |
1398 |
BitstreamByteAlign(&bs); |
dec->edged_width, frame->image, frame->stride, |
1399 |
|
frame->colorspace); |
|
/* low_delay_default mode: repeat in packed_mode */ |
|
|
if (dec->low_delay_default && dec->packed_mode && output == 0 && success == 0) |
|
|
{ |
|
|
success = 1; |
|
|
goto repeat; |
|
|
} |
|
|
|
|
|
done : |
|
|
|
|
|
/* low_delay_default mode: if we've gotten here without outputting anything, |
|
|
then output the recently decoded frame, or print an error message */ |
|
|
if (dec->low_delay_default && output == 0) |
|
|
{ |
|
|
if (dec->packed_mode && seen_something) |
|
|
{ |
|
|
/* output the recently decoded frame */ |
|
|
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, dec->last_reduced_resolution); |
|
|
output = 1; |
|
1400 |
} |
} |
1401 |
else |
stop_conv_timer(); |
|
{ |
|
|
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 16, |
|
|
"warning: nothing to output"); |
|
|
image_printf(&dec->cur, dec->edged_width, dec->height, 16, 64, |
|
|
"bframe decoder lag"); |
|
|
|
|
|
decoder_output(dec, &dec->cur, NULL, frame, 1 /*disable pp*/); |
|
1402 |
} |
} |
1403 |
} |
} |
1404 |
|
#endif |
1405 |
|
|
1406 |
frame->length = BitstreamPos(&bs) / 8; |
if (vop_type == I_VOP || vop_type == P_VOP) { |
1407 |
|
image_swap(&dec->refn[0], &dec->refn[1]); |
1408 |
|
image_swap(&dec->cur, &dec->refn[0]); |
1409 |
|
|
1410 |
if (stats) |
// swap MACROBLOCK |
1411 |
{ |
// the Divx will not set the low_delay flage some times |
1412 |
stats->notify = output ? XVID_DEC_VOP : XVID_DEC_NOTHING; |
// so follow code will wrong to not swap at that time |
1413 |
stats->data.vop.time_base = (int)dec->time_base; |
// this will broken bitstream! so I'm change it, |
1414 |
stats->data.vop.time_increment = 0; //XXX: todo |
// But that is not the best way! can anyone tell me how |
1415 |
|
// to do another way? |
1416 |
|
// 18-07-2002 MinChen<chenm001@163.com> |
1417 |
|
//if (!dec->low_delay && vop_type == P_VOP) |
1418 |
|
if (vop_type == P_VOP) |
1419 |
|
mb_swap(&dec->mbs, &dec->last_mbs); |
1420 |
} |
} |
1421 |
|
|
1422 |
emms(); |
emms(); |