4 |
* - Decoder Module - |
* - Decoder Module - |
5 |
* |
* |
6 |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
* Copyright(C) 2002 MinChen <chenm001@163.com> |
7 |
* 2002-2003 Peter Ross <pross@xvid.org> |
* 2002-2004 Peter Ross <pross@xvid.org> |
8 |
* |
* |
9 |
* This program is free software ; you can redistribute it and/or modify |
* This program is free software ; you can redistribute it and/or modify |
10 |
* it under the terms of the GNU General Public License as published by |
* it under the terms of the GNU General Public License as published by |
46 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
47 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
48 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "image/reduced.h" |
|
49 |
#include "image/font.h" |
#include "image/font.h" |
50 |
|
#include "image/qpel.h" |
51 |
|
|
52 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
53 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
61 |
#include "image/postprocessing.h" |
#include "image/postprocessing.h" |
62 |
#include "utils/mem_align.h" |
#include "utils/mem_align.h" |
63 |
|
|
64 |
|
#define DIV2ROUND(n) (((n)>>1)|((n)&1)) |
65 |
|
#define DIV2(n) ((n)>>1) |
66 |
|
#define DIVUVMOV(n) (((n) >> 1) + roundtab_79[(n) & 0x3]) // |
67 |
|
|
68 |
static int |
static int |
69 |
decoder_resize(DECODER * dec) |
decoder_resize(DECODER * dec) |
70 |
{ |
{ |
77 |
|
|
78 |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
image_destroy(&dec->gmc, dec->edged_width, dec->edged_height); |
79 |
|
|
80 |
if (dec->last_mbs) |
image_null(&dec->cur); |
81 |
|
image_null(&dec->refn[0]); |
82 |
|
image_null(&dec->refn[1]); |
83 |
|
image_null(&dec->tmp); |
84 |
|
image_null(&dec->qtmp); |
85 |
|
image_null(&dec->gmc); |
86 |
|
|
87 |
|
|
88 |
xvid_free(dec->last_mbs); |
xvid_free(dec->last_mbs); |
|
if (dec->mbs) |
|
89 |
xvid_free(dec->mbs); |
xvid_free(dec->mbs); |
|
if (dec->qscale) |
|
90 |
xvid_free(dec->qscale); |
xvid_free(dec->qscale); |
91 |
|
dec->last_mbs = NULL; |
92 |
|
dec->mbs = NULL; |
93 |
|
dec->qscale = NULL; |
94 |
|
|
95 |
/* realloc */ |
/* realloc */ |
96 |
dec->mb_width = (dec->width + 15) / 16; |
dec->mb_width = (dec->width + 15) / 16; |
99 |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
dec->edged_width = 16 * dec->mb_width + 2 * EDGE_SIZE; |
100 |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
dec->edged_height = 16 * dec->mb_height + 2 * EDGE_SIZE; |
101 |
|
|
102 |
if (image_create(&dec->cur, dec->edged_width, dec->edged_height)) { |
if ( image_create(&dec->cur, dec->edged_width, dec->edged_height) |
103 |
xvid_free(dec); |
|| image_create(&dec->refn[0], dec->edged_width, dec->edged_height) |
104 |
return XVID_ERR_MEMORY; |
|| image_create(&dec->refn[1], dec->edged_width, dec->edged_height) /* Support B-frame to reference last 2 frame */ |
105 |
} |
|| image_create(&dec->tmp, dec->edged_width, dec->edged_height) |
106 |
|
|| image_create(&dec->qtmp, dec->edged_width, dec->edged_height) |
107 |
if (image_create(&dec->refn[0], dec->edged_width, dec->edged_height)) { |
|| image_create(&dec->gmc, dec->edged_width, dec->edged_height) ) |
108 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
/* Support B-frame to reference last 2 frame */ |
|
|
if (image_create(&dec->refn[1], dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
if (image_create(&dec->tmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->qtmp, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
|
if (image_create(&dec->gmc, dec->edged_width, dec->edged_height)) { |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
109 |
|
|
110 |
dec->mbs = |
dec->mbs = |
111 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
112 |
CACHE_LINE); |
CACHE_LINE); |
113 |
if (dec->mbs == NULL) { |
if (dec->mbs == NULL) |
114 |
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
goto memory_error; |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
|
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
115 |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
memset(dec->mbs, 0, sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height); |
116 |
|
|
117 |
/* For skip MB flag */ |
/* For skip MB flag */ |
118 |
dec->last_mbs = |
dec->last_mbs = |
119 |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
xvid_malloc(sizeof(MACROBLOCK) * dec->mb_width * dec->mb_height, |
120 |
CACHE_LINE); |
CACHE_LINE); |
121 |
if (dec->last_mbs == NULL) { |
if (dec->last_mbs == NULL) |
122 |
xvid_free(dec->mbs); |
goto memory_error; |
|
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); |
|
|
xvid_free(dec); |
|
|
return XVID_ERR_MEMORY; |
|
|
} |
|
|
|
|
123 |
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); |
124 |
|
|
125 |
/* nothing happens if that fails */ |
/* nothing happens if that fails */ |
130 |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
memset(dec->qscale, 0, sizeof(int) * dec->mb_width * dec->mb_height); |
131 |
|
|
132 |
return 0; |
return 0; |
133 |
|
|
134 |
|
memory_error: |
135 |
|
/* Most structures were deallocated / nullifieded, so it should be safe */ |
136 |
|
/* decoder_destroy(dec) minus the write_timer */ |
137 |
|
xvid_free(dec->mbs); |
138 |
|
image_destroy(&dec->cur, dec->edged_width, dec->edged_height); |
139 |
|
image_destroy(&dec->refn[0], dec->edged_width, dec->edged_height); |
140 |
|
image_destroy(&dec->refn[1], dec->edged_width, dec->edged_height); |
141 |
|
image_destroy(&dec->tmp, dec->edged_width, dec->edged_height); |
142 |
|
image_destroy(&dec->qtmp, dec->edged_width, dec->edged_height); |
143 |
|
|
144 |
|
xvid_free(dec); |
145 |
|
return XVID_ERR_MEMORY; |
146 |
} |
} |
147 |
|
|
148 |
|
|
194 |
dec->time = dec->time_base = dec->last_time_base = 0; |
dec->time = dec->time_base = dec->last_time_base = 0; |
195 |
dec->low_delay = 0; |
dec->low_delay = 0; |
196 |
dec->packed_mode = 0; |
dec->packed_mode = 0; |
197 |
|
dec->time_inc_resolution = 1; /* until VOL header says otherwise */ |
198 |
|
|
199 |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
dec->fixed_dimensions = (dec->width > 0 && dec->height > 0); |
200 |
|
|
242 |
Bitstream * bs, |
Bitstream * bs, |
243 |
const uint32_t quant, |
const uint32_t quant, |
244 |
const uint32_t intra_dc_threshold, |
const uint32_t intra_dc_threshold, |
245 |
const unsigned int bound, |
const unsigned int bound) |
|
const int reduced_resolution) |
|
246 |
{ |
{ |
247 |
|
|
248 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
255 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
256 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
257 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
}else{ |
|
258 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
259 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
260 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
|
} |
|
261 |
|
|
262 |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
memset(block, 0, 6 * 64 * sizeof(int16_t)); /* clear */ |
263 |
|
|
304 |
stop_coding_timer(); |
stop_coding_timer(); |
305 |
|
|
306 |
start_timer(); |
start_timer(); |
307 |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i * 64], iDcScaler, predictors, dec->bs_version); |
308 |
stop_prediction_timer(); |
stop_prediction_timer(); |
309 |
|
|
310 |
start_timer(); |
start_timer(); |
316 |
stop_iquant_timer(); |
stop_iquant_timer(); |
317 |
|
|
318 |
start_timer(); |
start_timer(); |
319 |
idct(&data[i * 64]); |
idct((short * const)&data[i * 64]); |
320 |
stop_idct_timer(); |
stop_idct_timer(); |
321 |
|
|
322 |
} |
} |
327 |
} |
} |
328 |
|
|
329 |
start_timer(); |
start_timer(); |
|
|
|
|
if (reduced_resolution) |
|
|
{ |
|
|
next_block*=2; |
|
|
copy_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
|
|
copy_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
|
copy_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
|
|
}else{ |
|
330 |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
transfer_16to8copy(pY_Cur, &data[0 * 64], stride); |
331 |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1 * 64], stride); |
332 |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2 * 64], stride); |
333 |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
334 |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
transfer_16to8copy(pU_Cur, &data[4 * 64], stride2); |
335 |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
transfer_16to8copy(pV_Cur, &data[5 * 64], stride2); |
|
} |
|
336 |
stop_transfer_timer(); |
stop_transfer_timer(); |
337 |
} |
} |
338 |
|
|
343 |
uint8_t * pY_Cur, |
uint8_t * pY_Cur, |
344 |
uint8_t * pU_Cur, |
uint8_t * pU_Cur, |
345 |
uint8_t * pV_Cur, |
uint8_t * pV_Cur, |
|
const int reduced_resolution, |
|
346 |
const MACROBLOCK * pMB) |
const MACROBLOCK * pMB) |
347 |
{ |
{ |
348 |
DECLARE_ALIGNED_MATRIX(block, 1, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 1, 64, int16_t, CACHE_LINE); |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
|
349 |
|
|
350 |
int stride = dec->edged_width; |
int stride = dec->edged_width; |
|
int next_block = stride * (reduced_resolution ? 16 : 8); |
|
|
const int stride2 = stride/2; |
|
351 |
int i; |
int i; |
352 |
const uint32_t iQuant = pMB->quant; |
const uint32_t iQuant = pMB->quant; |
353 |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
const int direction = dec->alternate_vertical_scan ? 2 : 0; |
354 |
const quant_interFuncPtr dequant = dec->quant_type == 0 ? dequant_h263_inter : dequant_mpeg_inter; |
typedef void (*get_inter_block_function_t)( |
355 |
|
Bitstream * bs, |
356 |
|
int16_t * block, |
357 |
|
int direction, |
358 |
|
const int quant, |
359 |
|
const uint16_t *matrix); |
360 |
|
typedef void (*add_residual_function_t)( |
361 |
|
uint8_t *predicted_block, |
362 |
|
const int16_t *residual, |
363 |
|
int stride); |
364 |
|
|
365 |
|
const get_inter_block_function_t get_inter_block = (dec->quant_type == 0) |
366 |
|
? (get_inter_block_function_t)get_inter_block_h263 |
367 |
|
: (get_inter_block_function_t)get_inter_block_mpeg; |
368 |
|
|
369 |
for (i = 0; i < 6; i++) { |
uint8_t *dst[6]; |
370 |
|
int strides[6]; |
371 |
|
|
|
if (cbp & (1 << (5 - i))) { /* coded */ |
|
372 |
|
|
373 |
memset(block, 0, 64 * sizeof(int16_t)); /* clear */ |
if (dec->interlacing && pMB->field_dct) { |
374 |
|
dst[0] = pY_Cur; |
375 |
|
dst[1] = pY_Cur + 8; |
376 |
|
dst[2] = pY_Cur + stride; |
377 |
|
dst[3] = dst[2] + 8; |
378 |
|
dst[4] = pU_Cur; |
379 |
|
dst[5] = pV_Cur; |
380 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride*2; |
381 |
|
strides[4] = stride/2; |
382 |
|
strides[5] = stride/2; |
383 |
|
} else { |
384 |
|
dst[0] = pY_Cur; |
385 |
|
dst[1] = pY_Cur + 8; |
386 |
|
dst[2] = pY_Cur + 8*stride; |
387 |
|
dst[3] = dst[2] + 8; |
388 |
|
dst[4] = pU_Cur; |
389 |
|
dst[5] = pV_Cur; |
390 |
|
strides[0] = strides[1] = strides[2] = strides[3] = stride; |
391 |
|
strides[4] = stride/2; |
392 |
|
strides[5] = stride/2; |
393 |
|
} |
394 |
|
|
395 |
|
for (i = 0; i < 6; i++) { |
396 |
|
/* Process only coded blocks */ |
397 |
|
if (cbp & (1 << (5 - i))) { |
398 |
|
|
399 |
|
/* Clear the block */ |
400 |
|
memset(&data[0], 0, 64*sizeof(int16_t)); |
401 |
|
|
402 |
|
/* Decode coeffs and dequantize on the fly */ |
403 |
start_timer(); |
start_timer(); |
404 |
get_inter_block(bs, block, direction); |
get_inter_block(bs, &data[0], direction, iQuant, get_inter_matrix(dec->mpeg_quant_matrices)); |
405 |
stop_coding_timer(); |
stop_coding_timer(); |
406 |
|
|
407 |
|
/* iDCT */ |
408 |
start_timer(); |
start_timer(); |
409 |
dequant(&data[i * 64], block, iQuant, dec->mpeg_quant_matrices); |
idct((short * const)&data[0]); |
410 |
stop_iquant_timer(); |
stop_idct_timer(); |
411 |
|
|
412 |
|
/* Add this residual to the predicted block */ |
413 |
start_timer(); |
start_timer(); |
414 |
idct(&data[i * 64]); |
transfer_16to8add(dst[i], &data[0], strides[i]); |
415 |
stop_idct_timer(); |
stop_transfer_timer(); |
416 |
} |
} |
417 |
} |
} |
|
|
|
|
if (dec->interlacing && pMB->field_dct) { |
|
|
next_block = stride; |
|
|
stride *= 2; |
|
418 |
} |
} |
419 |
|
|
420 |
start_timer(); |
static void __inline |
421 |
if (reduced_resolution) { |
validate_vector(VECTOR * mv, unsigned int x_pos, unsigned int y_pos, const DECODER * dec) |
422 |
if (cbp & 32) |
{ |
423 |
add_upsampled_8x8_16to8(pY_Cur, &data[0 * 64], stride); |
/* clip a vector to valid range |
424 |
if (cbp & 16) |
prevents crashes if bitstream is broken |
425 |
add_upsampled_8x8_16to8(pY_Cur + 16, &data[1 * 64], stride); |
*/ |
426 |
if (cbp & 8) |
int shift = 5 + dec->quarterpel; |
427 |
add_upsampled_8x8_16to8(pY_Cur + next_block, &data[2 * 64], stride); |
int xborder_high = (int)(dec->mb_width - x_pos) << shift; |
428 |
if (cbp & 4) |
int xborder_low = (-(int)x_pos-1) << shift; |
429 |
add_upsampled_8x8_16to8(pY_Cur + 16 + next_block, &data[3 * 64], stride); |
int yborder_high = (int)(dec->mb_height - y_pos) << shift; |
430 |
if (cbp & 2) |
int yborder_low = (-(int)y_pos-1) << shift; |
431 |
add_upsampled_8x8_16to8(pU_Cur, &data[4 * 64], stride2); |
|
432 |
if (cbp & 1) |
#define CHECK_MV(mv) \ |
433 |
add_upsampled_8x8_16to8(pV_Cur, &data[5 * 64], stride2); |
do { \ |
434 |
} else { |
if ((mv).x > xborder_high) { \ |
435 |
if (cbp & 32) |
DPRINTF(XVID_DEBUG_MV, "mv.x > max -- %d > %d, MB %d, %d", (mv).x, xborder_high, x_pos, y_pos); \ |
436 |
transfer_16to8add(pY_Cur, &data[0 * 64], stride); |
(mv).x = xborder_high; \ |
437 |
if (cbp & 16) |
} else if ((mv).x < xborder_low) { \ |
438 |
transfer_16to8add(pY_Cur + 8, &data[1 * 64], stride); |
DPRINTF(XVID_DEBUG_MV, "mv.x < min -- %d < %d, MB %d, %d", (mv).x, xborder_low, x_pos, y_pos); \ |
439 |
if (cbp & 8) |
(mv).x = xborder_low; \ |
440 |
transfer_16to8add(pY_Cur + next_block, &data[2 * 64], stride); |
} \ |
441 |
if (cbp & 4) |
if ((mv).y > yborder_high) { \ |
442 |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3 * 64], stride); |
DPRINTF(XVID_DEBUG_MV, "mv.y > max -- %d > %d, MB %d, %d", (mv).y, yborder_high, x_pos, y_pos); \ |
443 |
if (cbp & 2) |
(mv).y = yborder_high; \ |
444 |
transfer_16to8add(pU_Cur, &data[4 * 64], stride2); |
} else if ((mv).y < yborder_low) { \ |
445 |
if (cbp & 1) |
DPRINTF(XVID_DEBUG_MV, "mv.y < min -- %d < %d, MB %d, %d", (mv).y, yborder_low, x_pos, y_pos); \ |
446 |
transfer_16to8add(pV_Cur, &data[5 * 64], stride2); |
(mv).y = yborder_low; \ |
447 |
} |
} \ |
448 |
stop_transfer_timer(); |
} while (0) |
449 |
} |
|
450 |
|
CHECK_MV(mv[0]); |
451 |
|
CHECK_MV(mv[1]); |
452 |
|
CHECK_MV(mv[2]); |
453 |
|
CHECK_MV(mv[3]); |
454 |
|
} |
455 |
|
|
456 |
|
/* Up to this version, chroma rounding was wrong with qpel. |
457 |
|
* So we try to be backward compatible to avoid artifacts */ |
458 |
|
#define BS_VERSION_BUGGY_CHROMA_ROUNDING 1 |
459 |
|
|
460 |
/* decode an inter macroblock */ |
/* decode an inter macroblock */ |
461 |
static void |
static void |
466 |
const uint32_t cbp, |
const uint32_t cbp, |
467 |
Bitstream * bs, |
Bitstream * bs, |
468 |
const uint32_t rounding, |
const uint32_t rounding, |
469 |
const int reduced_resolution, |
const int ref, |
470 |
const int ref) |
const int bvop) |
471 |
{ |
{ |
472 |
uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
473 |
uint32_t stride2 = stride / 2; |
uint32_t stride2 = stride / 2; |
478 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
479 |
VECTOR mv[4]; /* local copy of mvs */ |
VECTOR mv[4]; /* local copy of mvs */ |
480 |
|
|
|
if (reduced_resolution) { |
|
|
pY_Cur = dec->cur.y + (y_pos << 5) * stride + (x_pos << 5); |
|
|
pU_Cur = dec->cur.u + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
pV_Cur = dec->cur.v + (y_pos << 4) * stride2 + (x_pos << 4); |
|
|
for (i = 0; i < 4; i++) { |
|
|
mv[i].x = RRV_MV_SCALEUP(pMB->mvs[i].x); |
|
|
mv[i].y = RRV_MV_SCALEUP(pMB->mvs[i].y); |
|
|
} |
|
|
} else { |
|
481 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
482 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
483 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
484 |
for (i = 0; i < 4; i++) |
for (i = 0; i < 4; i++) |
485 |
mv[i] = pMB->mvs[i]; |
mv[i] = pMB->mvs[i]; |
486 |
} |
|
487 |
|
validate_vector(mv, x_pos, y_pos, dec); |
488 |
|
|
489 |
start_timer(); |
start_timer(); |
490 |
|
|
491 |
if (pMB->mode != MODE_INTER4V) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
if ((pMB->mode != MODE_INTER4V) || (bvop)) { /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
492 |
|
|
493 |
uv_dx = mv[0].x; |
uv_dx = mv[0].x; |
494 |
uv_dy = mv[0].y; |
uv_dy = mv[0].y; |
495 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
496 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
497 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
498 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
499 |
|
} |
500 |
|
else { |
501 |
uv_dx /= 2; |
uv_dx /= 2; |
502 |
uv_dy /= 2; |
uv_dy /= 2; |
503 |
} |
} |
504 |
|
} |
505 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
506 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
507 |
|
|
508 |
if (reduced_resolution) |
if (dec->quarterpel) |
|
interpolate32x32_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
else if (dec->quarterpel) |
|
509 |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_quarterpel(dec->cur.y, dec->refn[ref].y, dec->qtmp.y, dec->qtmp.y + 64, |
510 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
511 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
516 |
} else { /* MODE_INTER4V */ |
} else { /* MODE_INTER4V */ |
517 |
|
|
518 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
519 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
520 |
|
int z; |
521 |
|
uv_dx = 0; uv_dy = 0; |
522 |
|
for (z = 0; z < 4; z++) { |
523 |
|
uv_dx += ((mv[z].x>>1) | (mv[z].x&1)); |
524 |
|
uv_dy += ((mv[z].y>>1) | (mv[z].y&1)); |
525 |
|
} |
526 |
|
} |
527 |
|
else { |
528 |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
uv_dx = (mv[0].x / 2) + (mv[1].x / 2) + (mv[2].x / 2) + (mv[3].x / 2); |
529 |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
uv_dy = (mv[0].y / 2) + (mv[1].y / 2) + (mv[2].y / 2) + (mv[3].y / 2); |
530 |
|
} |
531 |
} else { |
} else { |
532 |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
uv_dx = mv[0].x + mv[1].x + mv[2].x + mv[3].x; |
533 |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
uv_dy = mv[0].y + mv[1].y + mv[2].y + mv[3].y; |
536 |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
uv_dx = (uv_dx >> 3) + roundtab_76[uv_dx & 0xf]; |
537 |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
uv_dy = (uv_dy >> 3) + roundtab_76[uv_dy & 0xf]; |
538 |
|
|
539 |
if (reduced_resolution) { |
if (dec->quarterpel) { |
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y, 32*x_pos, 32*y_pos, |
|
|
mv[0].x, mv[0].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos, |
|
|
mv[1].x, mv[1].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos, 32*y_pos + 16, |
|
|
mv[2].x, mv[2].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.y, dec->refn[0].y , 32*x_pos + 16, 32*y_pos + 16, |
|
|
mv[3].x, mv[3].y, stride, rounding); |
|
|
interpolate16x16_switch(dec->cur.u, dec->refn[0].u , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
interpolate16x16_switch(dec->cur.v, dec->refn[0].v , 16 * x_pos, 16 * y_pos, |
|
|
uv_dx, uv_dy, stride2, rounding); |
|
|
|
|
|
} else if (dec->quarterpel) { |
|
540 |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_quarterpel(dec->cur.y, dec->refn[0].y , dec->qtmp.y, dec->qtmp.y + 64, |
541 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
542 |
mv[0].x, mv[0].y, stride, rounding); |
mv[0].x, mv[0].y, stride, rounding); |
562 |
} |
} |
563 |
|
|
564 |
/* chroma */ |
/* chroma */ |
|
if (reduced_resolution) { |
|
|
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); |
|
|
} else { |
|
565 |
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, |
566 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
567 |
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, |
568 |
uv_dx, uv_dy, stride2, rounding); |
uv_dx, uv_dy, stride2, rounding); |
569 |
|
|
570 |
|
stop_comp_timer(); |
571 |
|
|
572 |
|
if (cbp) |
573 |
|
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
574 |
|
} |
575 |
|
|
576 |
|
/* decode an inter macroblock in field mode */ |
577 |
|
static void |
578 |
|
decoder_mbinter_field(DECODER * dec, |
579 |
|
const MACROBLOCK * pMB, |
580 |
|
const uint32_t x_pos, |
581 |
|
const uint32_t y_pos, |
582 |
|
const uint32_t cbp, |
583 |
|
Bitstream * bs, |
584 |
|
const uint32_t rounding, |
585 |
|
const int ref, |
586 |
|
const int bvop) |
587 |
|
{ |
588 |
|
uint32_t stride = dec->edged_width; |
589 |
|
uint32_t stride2 = stride / 2; |
590 |
|
|
591 |
|
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
592 |
|
|
593 |
|
int uvtop_dx, uvtop_dy; |
594 |
|
int uvbot_dx, uvbot_dy; |
595 |
|
VECTOR mv[4]; /* local copy of mvs */ |
596 |
|
|
597 |
|
/* Get pointer to memory areas */ |
598 |
|
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
599 |
|
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
600 |
|
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
601 |
|
|
602 |
|
mv[0] = pMB->mvs[0]; |
603 |
|
mv[1] = pMB->mvs[1]; |
604 |
|
memset(&mv[2],0,2*sizeof(VECTOR)); |
605 |
|
|
606 |
|
validate_vector(mv, x_pos, y_pos, dec); |
607 |
|
|
608 |
|
start_timer(); |
609 |
|
|
610 |
|
if((pMB->mode!=MODE_INTER4V) || (bvop)) /* INTER, INTER_Q, NOT_CODED, FORWARD, BACKWARD */ |
611 |
|
{ |
612 |
|
/* Prepare top field vector */ |
613 |
|
uvtop_dx = DIV2ROUND(mv[0].x); |
614 |
|
uvtop_dy = DIV2ROUND(mv[0].y); |
615 |
|
|
616 |
|
/* Prepare bottom field vector */ |
617 |
|
uvbot_dx = DIV2ROUND(mv[1].x); |
618 |
|
uvbot_dy = DIV2ROUND(mv[1].y); |
619 |
|
|
620 |
|
if(dec->quarterpel) |
621 |
|
{ |
622 |
|
/* NOT supported */ |
623 |
|
} |
624 |
|
else |
625 |
|
{ |
626 |
|
/* Interpolate top field left part(we use double stride for every 2nd line) */ |
627 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
628 |
|
16*x_pos,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
629 |
|
/* top field right part */ |
630 |
|
interpolate8x8_switch(dec->cur.y,dec->refn[ref].y+pMB->field_for_top*stride, |
631 |
|
16*x_pos+8,8*y_pos,mv[0].x, mv[0].y>>1,2*stride, rounding); |
632 |
|
|
633 |
|
/* Interpolate bottom field left part(we use double stride for every 2nd line) */ |
634 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
635 |
|
16*x_pos,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
636 |
|
/* Bottom field right part */ |
637 |
|
interpolate8x8_switch(dec->cur.y+stride,dec->refn[ref].y+pMB->field_for_bot*stride, |
638 |
|
16*x_pos+8,8*y_pos,mv[1].x, mv[1].y>>1,2*stride, rounding); |
639 |
|
|
640 |
|
/* Interpolate field1 U */ |
641 |
|
interpolate8x4_switch(dec->cur.u,dec->refn[ref].u+pMB->field_for_top*stride2, |
642 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
643 |
|
|
644 |
|
/* Interpolate field1 V */ |
645 |
|
interpolate8x4_switch(dec->cur.v,dec->refn[ref].v+pMB->field_for_top*stride2, |
646 |
|
8*x_pos,4*y_pos,uvtop_dx,DIV2ROUND(uvtop_dy),stride,rounding); |
647 |
|
|
648 |
|
/* Interpolate field2 U */ |
649 |
|
interpolate8x4_switch(dec->cur.u+stride2,dec->refn[ref].u+pMB->field_for_bot*stride2, |
650 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
651 |
|
|
652 |
|
/* Interpolate field2 V */ |
653 |
|
interpolate8x4_switch(dec->cur.v+stride2,dec->refn[ref].v+pMB->field_for_bot*stride2, |
654 |
|
8*x_pos,4*y_pos,uvbot_dx,DIV2ROUND(uvbot_dy),stride,rounding); |
655 |
|
} |
656 |
|
} |
657 |
|
else |
658 |
|
{ |
659 |
|
/* We don't expect 4 motion vectors in interlaced mode */ |
660 |
} |
} |
661 |
|
|
662 |
stop_comp_timer(); |
stop_comp_timer(); |
663 |
|
|
664 |
|
/* Must add error correction? */ |
665 |
if (cbp) |
if (cbp) |
666 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
|
reduced_resolution, pMB); |
|
667 |
} |
} |
668 |
|
|
669 |
static void |
static void |
710 |
stop_transfer_timer(); |
stop_transfer_timer(); |
711 |
|
|
712 |
if (cbp) |
if (cbp) |
713 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
714 |
|
|
715 |
} |
} |
716 |
|
|
718 |
static void |
static void |
719 |
decoder_iframe(DECODER * dec, |
decoder_iframe(DECODER * dec, |
720 |
Bitstream * bs, |
Bitstream * bs, |
|
int reduced_resolution, |
|
721 |
int quant, |
int quant, |
722 |
int intra_dc_threshold) |
int intra_dc_threshold) |
723 |
{ |
{ |
724 |
uint32_t bound; |
uint32_t bound; |
725 |
uint32_t x, y; |
uint32_t x, y; |
726 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
727 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) { |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
728 |
|
|
729 |
bound = 0; |
bound = 0; |
730 |
|
|
780 |
} |
} |
781 |
|
|
782 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
783 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
784 |
|
|
785 |
} |
} |
786 |
if(dec->out_frm) |
if(dec->out_frm) |
829 |
mv.y -= range; |
mv.y -= range; |
830 |
} |
} |
831 |
|
|
832 |
/* clip to valid range */ |
ret_mv->x = mv.x; |
833 |
|
ret_mv->y = mv.y; |
834 |
|
} |
835 |
|
|
836 |
if (mv.x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) |
/* We use this when decoder runs interlaced -> different prediction */ |
|
mv.x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); |
|
837 |
|
|
838 |
else if (mv.x < (int)(-x-1) << (5 + dec->quarterpel)) |
static void get_motion_vector_interlaced(DECODER * dec, |
839 |
mv.x = (int)(-x-1) << (5 + dec->quarterpel); |
Bitstream * bs, |
840 |
|
int x, |
841 |
|
int y, |
842 |
|
int k, |
843 |
|
MACROBLOCK *pMB, |
844 |
|
int fcode, |
845 |
|
const int bound) |
846 |
|
{ |
847 |
|
const int scale_fac = 1 << (fcode - 1); |
848 |
|
const int high = (32 * scale_fac) - 1; |
849 |
|
const int low = ((-32) * scale_fac); |
850 |
|
const int range = (64 * scale_fac); |
851 |
|
|
852 |
if (mv.y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) |
/* Get interlaced prediction */ |
853 |
mv.y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); |
const VECTOR pmv=get_pmv2_interlaced(dec->mbs,dec->mb_width,bound,x,y,k); |
854 |
|
VECTOR mv,mvf1,mvf2; |
855 |
|
|
856 |
else if (mv.y < ((int)(-y-1)) << (5 + dec->quarterpel) ) |
if(!pMB->field_pred) |
857 |
mv.y = (int)(-y-1) << (5 + dec->quarterpel); |
{ |
858 |
|
mv.x = get_mv(bs,fcode); |
859 |
|
mv.y = get_mv(bs,fcode); |
860 |
|
|
861 |
ret_mv->x = mv.x; |
mv.x += pmv.x; |
862 |
ret_mv->y = mv.y; |
mv.y += pmv.y; |
863 |
|
|
864 |
|
if(mv.x<low) { |
865 |
|
mv.x += range; |
866 |
|
} else if (mv.x>high) { |
867 |
|
mv.x-=range; |
868 |
|
} |
869 |
|
|
870 |
|
if (mv.y < low) { |
871 |
|
mv.y += range; |
872 |
|
} else if (mv.y > high) { |
873 |
|
mv.y -= range; |
874 |
|
} |
875 |
|
|
876 |
|
pMB->mvs[0]=pMB->mvs[1]=pMB->mvs[2]=pMB->mvs[3]=mv; |
877 |
|
} |
878 |
|
else |
879 |
|
{ |
880 |
|
mvf1.x = get_mv(bs, fcode); |
881 |
|
mvf1.y = get_mv(bs, fcode); |
882 |
|
|
883 |
|
mvf1.x += pmv.x; |
884 |
|
mvf1.y = 2*(mvf1.y+pmv.y/2); /* It's multiple of 2 */ |
885 |
|
|
886 |
|
if (mvf1.x < low) { |
887 |
|
mvf1.x += range; |
888 |
|
} else if (mvf1.x > high) { |
889 |
|
mvf1.x -= range; |
890 |
|
} |
891 |
|
|
892 |
|
if (mvf1.y < low) { |
893 |
|
mvf1.y += range; |
894 |
|
} else if (mvf1.y > high) { |
895 |
|
mvf1.y -= range; |
896 |
|
} |
897 |
|
|
898 |
|
mvf2.x = get_mv(bs, fcode); |
899 |
|
mvf2.y = get_mv(bs, fcode); |
900 |
|
|
901 |
|
mvf2.x += pmv.x; |
902 |
|
mvf2.y = 2*(mvf2.y+pmv.y/2); /* It's multiple of 2 */ |
903 |
|
|
904 |
|
if (mvf2.x < low) { |
905 |
|
mvf2.x += range; |
906 |
|
} else if (mvf2.x > high) { |
907 |
|
mvf2.x -= range; |
908 |
|
} |
909 |
|
|
910 |
|
if (mvf2.y < low) { |
911 |
|
mvf2.y += range; |
912 |
|
} else if (mvf2.y > high) { |
913 |
|
mvf2.y -= range; |
914 |
|
} |
915 |
|
|
916 |
|
pMB->mvs[0]=mvf1; |
917 |
|
pMB->mvs[1]=mvf2; |
918 |
|
pMB->mvs[2].x=pMB->mvs[3].x=0; |
919 |
|
pMB->mvs[2].y=pMB->mvs[3].y=0; |
920 |
|
|
921 |
|
/* Calculate average for as it is field predicted */ |
922 |
|
pMB->mvs_avg.x=DIV2ROUND(pMB->mvs[0].x+pMB->mvs[1].x); |
923 |
|
pMB->mvs_avg.y=DIV2ROUND(pMB->mvs[0].y+pMB->mvs[1].y); |
924 |
|
} |
925 |
} |
} |
926 |
|
|
927 |
/* for P_VOP set gmc_warp to NULL */ |
/* for P_VOP set gmc_warp to NULL */ |
929 |
decoder_pframe(DECODER * dec, |
decoder_pframe(DECODER * dec, |
930 |
Bitstream * bs, |
Bitstream * bs, |
931 |
int rounding, |
int rounding, |
|
int reduced_resolution, |
|
932 |
int quant, |
int quant, |
933 |
int fcode, |
int fcode, |
934 |
int intra_dc_threshold, |
int intra_dc_threshold, |
937 |
uint32_t x, y; |
uint32_t x, y; |
938 |
uint32_t bound; |
uint32_t bound; |
939 |
int cp_mb, st_mb; |
int cp_mb, st_mb; |
940 |
uint32_t mb_width = dec->mb_width; |
const uint32_t mb_width = dec->mb_width; |
941 |
uint32_t mb_height = dec->mb_height; |
const uint32_t mb_height = dec->mb_height; |
|
|
|
|
if (reduced_resolution) { |
|
|
mb_width = (dec->width + 31) / 32; |
|
|
mb_height = (dec->height + 31) / 32; |
|
|
} |
|
942 |
|
|
943 |
|
if (!dec->is_edged[0]) { |
944 |
start_timer(); |
start_timer(); |
945 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
946 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
947 |
|
dec->is_edged[0] = 1; |
948 |
stop_edges_timer(); |
stop_edges_timer(); |
949 |
|
} |
950 |
|
|
951 |
if (gmc_warp) { |
if (gmc_warp) { |
952 |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
/* accuracy: 0==1/2, 1=1/4, 2=1/8, 3=1/16 */ |
1016 |
} |
} |
1017 |
mb->quant = quant; |
mb->quant = quant; |
1018 |
|
|
1019 |
|
mb->field_pred=0; |
1020 |
if (dec->interlacing) { |
if (dec->interlacing) { |
1021 |
if (cbp || intra) { |
if (cbp || intra) { |
1022 |
mb->field_dct = BitstreamGetBit(bs); |
mb->field_dct = BitstreamGetBit(bs); |
1042 |
|
|
1043 |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
} else if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
1044 |
|
|
1045 |
if (dec->interlacing && mb->field_pred) { |
if(dec->interlacing) { |
1046 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
/* Get motion vectors interlaced, field_pred is handled there */ |
1047 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[1], fcode, bound); |
get_motion_vector_interlaced(dec, bs, x, y, 0, mb, fcode, bound); |
1048 |
} else { |
} else { |
1049 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
1050 |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1051 |
} |
} |
1052 |
} else if (mb->mode == MODE_INTER4V ) { |
} else if (mb->mode == MODE_INTER4V ) { |
1053 |
|
/* interlaced missing here */ |
1054 |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
get_motion_vector(dec, bs, x, y, 0, &mb->mvs[0], fcode, bound); |
1055 |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
get_motion_vector(dec, bs, x, y, 1, &mb->mvs[1], fcode, bound); |
1056 |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
get_motion_vector(dec, bs, x, y, 2, &mb->mvs[2], fcode, bound); |
1059 |
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; |
1060 |
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; |
1061 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, |
1062 |
intra_dc_threshold, bound, reduced_resolution); |
intra_dc_threshold, bound); |
1063 |
continue; |
continue; |
1064 |
} |
} |
1065 |
|
|
1066 |
decoder_mbinter(dec, mb, x, y, cbp, bs, |
/* See how to decode */ |
1067 |
rounding, reduced_resolution, 0); |
if(!mb->field_pred) |
1068 |
|
decoder_mbinter(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1069 |
|
else |
1070 |
|
decoder_mbinter_field(dec, mb, x, y, cbp, bs, rounding, 0, 0); |
1071 |
|
|
1072 |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
} else if (gmc_warp) { /* a not coded S(GMC)-VOP macroblock */ |
1073 |
mb->mode = MODE_NOT_CODED_GMC; |
mb->mode = MODE_NOT_CODED_GMC; |
1085 |
|
|
1086 |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
mb->mvs[0].x = mb->mvs[1].x = mb->mvs[2].x = mb->mvs[3].x = 0; |
1087 |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
mb->mvs[0].y = mb->mvs[1].y = mb->mvs[2].y = mb->mvs[3].y = 0; |
1088 |
|
mb->field_pred=0; /* (!) */ |
1089 |
|
|
1090 |
decoder_mbinter(dec, mb, x, y, 0, bs, |
decoder_mbinter(dec, mb, x, y, 0, bs, |
1091 |
rounding, reduced_resolution, 0); |
rounding, 0, 0); |
1092 |
|
|
1093 |
if(dec->out_frm && cp_mb > 0) { |
if(dec->out_frm && cp_mb > 0) { |
1094 |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
output_slice(&dec->cur, dec->edged_width,dec->width,dec->out_frm,st_mb,y,cp_mb); |
1134 |
else if (mv_y > high) |
else if (mv_y > high) |
1135 |
mv_y -= range; |
mv_y -= range; |
1136 |
|
|
|
|
|
|
/* clip to valid range */ |
|
|
if (mv_x > ((int)(dec->mb_width - x) << (5 + dec->quarterpel)) ) |
|
|
mv_x = (int)(dec->mb_width - x) << (5 + dec->quarterpel); |
|
|
|
|
|
else if (mv_x < (int)(-x-1) << (5 + dec->quarterpel)) |
|
|
mv_x = (int)(-x-1) << (5 + dec->quarterpel); |
|
|
|
|
|
if (mv_y > ((int)(dec->mb_height - y) << (5 + dec->quarterpel)) ) |
|
|
mv_y = (int)(dec->mb_height - y) << (5 + dec->quarterpel); |
|
|
|
|
|
else if (mv_y < ((int)(-y-1)) << (5 + dec->quarterpel) ) |
|
|
mv_y = (int)(-y-1) << (5 + dec->quarterpel); |
|
|
|
|
1137 |
mv->x = mv_x; |
mv->x = mv_x; |
1138 |
mv->y = mv_y; |
mv->y = mv_y; |
1139 |
} |
} |
1143 |
decoder_bf_interpolate_mbinter(DECODER * dec, |
decoder_bf_interpolate_mbinter(DECODER * dec, |
1144 |
IMAGE forward, |
IMAGE forward, |
1145 |
IMAGE backward, |
IMAGE backward, |
1146 |
const MACROBLOCK * pMB, |
MACROBLOCK * pMB, |
1147 |
const uint32_t x_pos, |
const uint32_t x_pos, |
1148 |
const uint32_t y_pos, |
const uint32_t y_pos, |
1149 |
Bitstream * bs, |
Bitstream * bs, |
1160 |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
1161 |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
1162 |
|
|
1163 |
|
validate_vector(pMB->mvs, x_pos, y_pos, dec); |
1164 |
|
validate_vector(pMB->b_mvs, x_pos, y_pos, dec); |
1165 |
|
|
1166 |
if (!direct) { |
if (!direct) { |
1167 |
uv_dx = pMB->mvs[0].x; |
uv_dx = pMB->mvs[0].x; |
1168 |
uv_dy = pMB->mvs[0].y; |
uv_dy = pMB->mvs[0].y; |
|
|
|
1169 |
b_uv_dx = pMB->b_mvs[0].x; |
b_uv_dx = pMB->b_mvs[0].x; |
1170 |
b_uv_dy = pMB->b_mvs[0].y; |
b_uv_dy = pMB->b_mvs[0].y; |
1171 |
|
|
1172 |
if (dec->quarterpel) { |
if (dec->quarterpel) { |
1173 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1174 |
|
uv_dx = (uv_dx>>1) | (uv_dx&1); |
1175 |
|
uv_dy = (uv_dy>>1) | (uv_dy&1); |
1176 |
|
b_uv_dx = (b_uv_dx>>1) | (b_uv_dx&1); |
1177 |
|
b_uv_dy = (b_uv_dy>>1) | (b_uv_dy&1); |
1178 |
|
} |
1179 |
|
else { |
1180 |
uv_dx /= 2; |
uv_dx /= 2; |
1181 |
uv_dy /= 2; |
uv_dy /= 2; |
1182 |
b_uv_dx /= 2; |
b_uv_dx /= 2; |
1183 |
b_uv_dy /= 2; |
b_uv_dy /= 2; |
1184 |
} |
} |
1185 |
|
} |
1186 |
|
|
1187 |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
uv_dx = (uv_dx >> 1) + roundtab_79[uv_dx & 0x3]; |
1188 |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
uv_dy = (uv_dy >> 1) + roundtab_79[uv_dy & 0x3]; |
|
|
|
1189 |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
b_uv_dx = (b_uv_dx >> 1) + roundtab_79[b_uv_dx & 0x3]; |
1190 |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
b_uv_dy = (b_uv_dy >> 1) + roundtab_79[b_uv_dy & 0x3]; |
1191 |
|
|
1192 |
} else { |
} else { |
1193 |
if(dec->quarterpel) { |
if (dec->quarterpel) { /* for qpel the /2 shall be done before summation. We've done it right in the encoder in the past. */ |
1194 |
|
/* TODO: figure out if we ever did it wrong on the encoder side. If yes, add some workaround */ |
1195 |
|
if (dec->bs_version <= BS_VERSION_BUGGY_CHROMA_ROUNDING) { |
1196 |
|
int z; |
1197 |
|
uv_dx = 0; uv_dy = 0; |
1198 |
|
b_uv_dx = 0; b_uv_dy = 0; |
1199 |
|
for (z = 0; z < 4; z++) { |
1200 |
|
uv_dx += ((pMB->mvs[z].x>>1) | (pMB->mvs[z].x&1)); |
1201 |
|
uv_dy += ((pMB->mvs[z].y>>1) | (pMB->mvs[z].y&1)); |
1202 |
|
b_uv_dx += ((pMB->b_mvs[z].x>>1) | (pMB->b_mvs[z].x&1)); |
1203 |
|
b_uv_dy += ((pMB->b_mvs[z].y>>1) | (pMB->b_mvs[z].y&1)); |
1204 |
|
} |
1205 |
|
} |
1206 |
|
else { |
1207 |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
uv_dx = (pMB->mvs[0].x / 2) + (pMB->mvs[1].x / 2) + (pMB->mvs[2].x / 2) + (pMB->mvs[3].x / 2); |
1208 |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
uv_dy = (pMB->mvs[0].y / 2) + (pMB->mvs[1].y / 2) + (pMB->mvs[2].y / 2) + (pMB->mvs[3].y / 2); |
1209 |
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
b_uv_dx = (pMB->b_mvs[0].x / 2) + (pMB->b_mvs[1].x / 2) + (pMB->b_mvs[2].x / 2) + (pMB->b_mvs[3].x / 2); |
1210 |
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
b_uv_dy = (pMB->b_mvs[0].y / 2) + (pMB->b_mvs[1].y / 2) + (pMB->b_mvs[2].y / 2) + (pMB->b_mvs[3].y / 2); |
1211 |
|
} |
1212 |
} else { |
} else { |
1213 |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
uv_dx = pMB->mvs[0].x + pMB->mvs[1].x + pMB->mvs[2].x + pMB->mvs[3].x; |
1214 |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
uv_dy = pMB->mvs[0].y + pMB->mvs[1].y + pMB->mvs[2].y + pMB->mvs[3].y; |
1261 |
|
|
1262 |
if(dec->quarterpel) { |
if(dec->quarterpel) { |
1263 |
if(!direct) { |
if(!direct) { |
1264 |
interpolate16x16_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate16x16_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1265 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1266 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1267 |
} else { |
} else { |
1268 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1269 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos, |
1270 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1271 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1272 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos, |
1273 |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1274 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1275 |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos, 16*y_pos + 8, |
1276 |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1277 |
interpolate8x8_quarterpel(dec->tmp.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
interpolate8x8_add_quarterpel(dec->cur.y, backward.y, dec->qtmp.y, dec->qtmp.y + 64, |
1278 |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
dec->qtmp.y + 128, 16*x_pos + 8, 16*y_pos + 8, |
1279 |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1280 |
} |
} |
1281 |
} else { |
} else { |
1282 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, 16 * y_pos, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, 16 * y_pos, |
1283 |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
pMB->b_mvs[0].x, pMB->b_mvs[0].y, stride, 0); |
1284 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1285 |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
16 * y_pos, pMB->b_mvs[1].x, pMB->b_mvs[1].y, stride, 0); |
1286 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos, |
1287 |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
16 * y_pos + 8, pMB->b_mvs[2].x, pMB->b_mvs[2].y, stride, 0); |
1288 |
interpolate8x8_switch(dec->tmp.y, backward.y, 16 * x_pos + 8, |
interpolate8x8_add_switch(dec->cur.y, backward.y, 16 * x_pos + 8, |
1289 |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
16 * y_pos + 8, pMB->b_mvs[3].x, pMB->b_mvs[3].y, stride, 0); |
1290 |
} |
} |
1291 |
|
|
1292 |
interpolate8x8_switch(dec->tmp.u, backward.u, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.u, backward.u, 8 * x_pos, 8 * y_pos, |
1293 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1294 |
interpolate8x8_switch(dec->tmp.v, backward.v, 8 * x_pos, 8 * y_pos, |
interpolate8x8_add_switch(dec->cur.v, backward.v, 8 * x_pos, 8 * y_pos, |
1295 |
b_uv_dx, b_uv_dy, stride2, 0); |
b_uv_dx, b_uv_dy, stride2, 0); |
1296 |
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + (16 * y_pos * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->cur.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
dec->tmp.y + ((16 * y_pos + 8) * stride) + 16 * x_pos + 8, |
|
|
stride, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.u + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
|
interpolate8x8_avg2(dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->cur.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
dec->tmp.v + (8 * y_pos * stride2) + 8 * x_pos, |
|
|
stride2, 1, 8); |
|
|
|
|
1297 |
stop_comp_timer(); |
stop_comp_timer(); |
1298 |
|
|
1299 |
if (cbp) |
if (cbp) |
1300 |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, 0, pMB); |
decoder_mb_decode(dec, cbp, bs, pY_Cur, pU_Cur, pV_Cur, pMB); |
1301 |
} |
} |
1302 |
|
|
1303 |
/* for decode B-frame dbquant */ |
/* for decode B-frame dbquant */ |
1342 |
uint32_t x, y; |
uint32_t x, y; |
1343 |
VECTOR mv; |
VECTOR mv; |
1344 |
const VECTOR zeromv = {0,0}; |
const VECTOR zeromv = {0,0}; |
|
const int64_t TRB = dec->time_pp - dec->time_bp, TRD = dec->time_pp; |
|
1345 |
int i; |
int i; |
1346 |
|
|
1347 |
|
if (!dec->is_edged[0]) { |
1348 |
start_timer(); |
start_timer(); |
1349 |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[0], dec->edged_width, dec->edged_height, |
1350 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
1351 |
|
dec->is_edged[0] = 1; |
1352 |
|
stop_edges_timer(); |
1353 |
|
} |
1354 |
|
|
1355 |
|
if (!dec->is_edged[1]) { |
1356 |
|
start_timer(); |
1357 |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
image_setedges(&dec->refn[1], dec->edged_width, dec->edged_height, |
1358 |
dec->width, dec->height, dec->bs_version); |
dec->width, dec->height, dec->bs_version); |
1359 |
|
dec->is_edged[1] = 1; |
1360 |
stop_edges_timer(); |
stop_edges_timer(); |
1361 |
|
} |
1362 |
|
|
1363 |
for (y = 0; y < dec->mb_height; y++) { |
for (y = 0; y < dec->mb_height; y++) { |
1364 |
/* Initialize Pred Motion Vector */ |
/* Initialize Pred Motion Vector */ |
1367 |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
MACROBLOCK *mb = &dec->mbs[y * dec->mb_width + x]; |
1368 |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
MACROBLOCK *last_mb = &dec->last_mbs[y * dec->mb_width + x]; |
1369 |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
const int fcode_max = (fcode_forward>fcode_backward) ? fcode_forward : fcode_backward; |
1370 |
uint32_t intra_dc_threshold; /* fake variable */ |
int intra_dc_threshold; /* fake variable */ |
1371 |
|
|
1372 |
if (check_resync_marker(bs, fcode_max - 1)) { |
if (check_resync_marker(bs, fcode_max - 1)) { |
1373 |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
int bound = read_video_packet_header(bs, dec, fcode_max - 1, &quant, |
1392 |
if (last_mb->mode == MODE_NOT_CODED) { |
if (last_mb->mode == MODE_NOT_CODED) { |
1393 |
mb->cbp = 0; |
mb->cbp = 0; |
1394 |
mb->mode = MODE_FORWARD; |
mb->mode = MODE_FORWARD; |
1395 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1396 |
continue; |
continue; |
1397 |
} |
} |
1398 |
|
|
1445 |
|
|
1446 |
case MODE_DIRECT_NONE_MV: |
case MODE_DIRECT_NONE_MV: |
1447 |
for (i = 0; i < 4; i++) { |
for (i = 0; i < 4; i++) { |
1448 |
mb->mvs[i].x = (int32_t) ((TRB * last_mb->mvs[i].x) / TRD + mv.x); |
mb->mvs[i].x = last_mb->mvs[i].x*dec->time_bp/dec->time_pp + mv.x; |
1449 |
mb->b_mvs[i].x = (int32_t) ((mv.x == 0) |
mb->mvs[i].y = last_mb->mvs[i].y*dec->time_bp/dec->time_pp + mv.y; |
1450 |
? ((TRB - TRD) * last_mb->mvs[i].x) / TRD |
|
1451 |
: mb->mvs[i].x - last_mb->mvs[i].x); |
mb->b_mvs[i].x = (mv.x) |
1452 |
mb->mvs[i].y = (int32_t) ((TRB * last_mb->mvs[i].y) / TRD + mv.y); |
? mb->mvs[i].x - last_mb->mvs[i].x |
1453 |
mb->b_mvs[i].y = (int32_t) ((mv.y == 0) |
: last_mb->mvs[i].x*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1454 |
? ((TRB - TRD) * last_mb->mvs[i].y) / TRD |
mb->b_mvs[i].y = (mv.y) |
1455 |
: mb->mvs[i].y - last_mb->mvs[i].y); |
? mb->mvs[i].y - last_mb->mvs[i].y |
1456 |
|
: last_mb->mvs[i].y*(dec->time_bp - dec->time_pp)/dec->time_pp; |
1457 |
} |
} |
1458 |
|
|
1459 |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
decoder_bf_interpolate_mbinter(dec, dec->refn[1], dec->refn[0], |
1475 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_backward, dec->p_bmv, dec, x, y); |
1476 |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_bmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1477 |
|
|
1478 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 0); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
1479 |
break; |
break; |
1480 |
|
|
1481 |
case MODE_FORWARD: |
case MODE_FORWARD: |
1482 |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
get_b_motion_vector(bs, &mb->mvs[0], fcode_forward, dec->p_fmv, dec, x, y); |
1483 |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
dec->p_fmv = mb->mvs[1] = mb->mvs[2] = mb->mvs[3] = mb->mvs[0]; |
1484 |
|
|
1485 |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 0, 1); |
decoder_mbinter(dec, mb, x, y, mb->cbp, bs, 0, 1, 1); |
1486 |
break; |
break; |
1487 |
|
|
1488 |
default: |
default: |
1493 |
} |
} |
1494 |
|
|
1495 |
/* perform post processing if necessary, and output the image */ |
/* perform post processing if necessary, and output the image */ |
1496 |
void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
static void decoder_output(DECODER * dec, IMAGE * img, MACROBLOCK * mbs, |
1497 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats, |
1498 |
int coding_type, int quant) |
int coding_type, int quant) |
1499 |
{ |
{ |
1500 |
|
const int brightness = XVID_VERSION_MINOR(frame->version) >= 1 ? frame->brightness : 0; |
1501 |
|
|
1502 |
if (dec->cartoon_mode) |
if (dec->cartoon_mode) |
1503 |
frame->general &= ~XVID_FILMEFFECT; |
frame->general &= ~XVID_FILMEFFECT; |
1504 |
|
|
1505 |
if (frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) && mbs != NULL) /* post process */ |
if ((frame->general & (XVID_DEBLOCKY|XVID_DEBLOCKUV|XVID_FILMEFFECT) || brightness!=0) |
1506 |
|
&& mbs != NULL) /* post process */ |
1507 |
{ |
{ |
1508 |
/* note: image is stored to tmp */ |
/* note: image is stored to tmp */ |
1509 |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
image_copy(&dec->tmp, img, dec->edged_width, dec->height); |
1510 |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
image_postproc(&dec->postproc, &dec->tmp, dec->edged_width, |
1511 |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
mbs, dec->mb_width, dec->mb_height, dec->mb_width, |
1512 |
frame->general, dec->frames, (coding_type == B_VOP)); |
frame->general, brightness, dec->frames, (coding_type == B_VOP)); |
1513 |
img = &dec->tmp; |
img = &dec->tmp; |
1514 |
} |
} |
1515 |
|
|
1524 |
stats->data.vop.qscale_stride = dec->mb_width; |
stats->data.vop.qscale_stride = dec->mb_width; |
1525 |
stats->data.vop.qscale = dec->qscale; |
stats->data.vop.qscale = dec->qscale; |
1526 |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
if (stats->data.vop.qscale != NULL && mbs != NULL) { |
1527 |
int i; |
unsigned int i; |
1528 |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
for (i = 0; i < dec->mb_width*dec->mb_height; i++) |
1529 |
stats->data.vop.qscale[i] = mbs[i].quant; |
stats->data.vop.qscale[i] = mbs[i].quant; |
1530 |
} else |
} else |
1532 |
} |
} |
1533 |
} |
} |
1534 |
|
|
|
|
|
1535 |
int |
int |
1536 |
decoder_decode(DECODER * dec, |
decoder_decode(DECODER * dec, |
1537 |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
xvid_dec_frame_t * frame, xvid_dec_stats_t * stats) |
1539 |
|
|
1540 |
Bitstream bs; |
Bitstream bs; |
1541 |
uint32_t rounding; |
uint32_t rounding; |
|
uint32_t reduced_resolution; |
|
1542 |
uint32_t quant = 2; |
uint32_t quant = 2; |
1543 |
uint32_t fcode_forward; |
uint32_t fcode_forward; |
1544 |
uint32_t fcode_backward; |
uint32_t fcode_backward; |
1593 |
|
|
1594 |
repeat: |
repeat: |
1595 |
|
|
1596 |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, &reduced_resolution, |
coding_type = BitstreamReadHeaders(&bs, dec, &rounding, |
1597 |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
&quant, &fcode_forward, &fcode_backward, &intra_dc_threshold, &gmc_warp); |
1598 |
|
|
1599 |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%lli, time_pp=%i, time_bp=%i\n", |
DPRINTF(XVID_DEBUG_HEADER, "coding_type=%i, packed=%i, time=%" |
1600 |
|
#if defined(_MSC_VER) |
1601 |
|
"I64" |
1602 |
|
#else |
1603 |
|
"ll" |
1604 |
|
#endif |
1605 |
|
"i, time_pp=%i, time_bp=%i\n", |
1606 |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
coding_type, dec->packed_mode, dec->time, dec->time_pp, dec->time_bp); |
1607 |
|
|
1608 |
if (coding_type == -1) { /* nothing */ |
if (coding_type == -1) { /* nothing */ |
1633 |
goto repeat; |
goto repeat; |
1634 |
} |
} |
1635 |
|
|
1636 |
|
if(dec->frames == 0 && coding_type != I_VOP) { |
1637 |
|
/* 1st frame is not an i-vop */ |
1638 |
|
goto repeat; |
1639 |
|
} |
1640 |
|
|
1641 |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
dec->p_bmv.x = dec->p_bmv.y = dec->p_fmv.y = dec->p_fmv.y = 0; /* init pred vector to 0 */ |
1642 |
|
|
1643 |
/* packed_mode: special-N_VOP treament */ |
/* packed_mode: special-N_VOP treament */ |
1650 |
} else if (coding_type != B_VOP) { |
} else if (coding_type != B_VOP) { |
1651 |
switch(coding_type) { |
switch(coding_type) { |
1652 |
case I_VOP : |
case I_VOP : |
1653 |
decoder_iframe(dec, &bs, reduced_resolution, quant, intra_dc_threshold); |
decoder_iframe(dec, &bs, quant, intra_dc_threshold); |
1654 |
break; |
break; |
1655 |
case P_VOP : |
case P_VOP : |
1656 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1657 |
fcode_forward, intra_dc_threshold, NULL); |
fcode_forward, intra_dc_threshold, NULL); |
1658 |
break; |
break; |
1659 |
case S_VOP : |
case S_VOP : |
1660 |
decoder_pframe(dec, &bs, rounding, reduced_resolution, quant, |
decoder_pframe(dec, &bs, rounding, quant, |
1661 |
fcode_forward, intra_dc_threshold, &gmc_warp); |
fcode_forward, intra_dc_threshold, &gmc_warp); |
1662 |
break; |
break; |
1663 |
case N_VOP : |
case N_VOP : |
1668 |
break; |
break; |
1669 |
} |
} |
1670 |
|
|
|
if (reduced_resolution) { |
|
|
image_deblock_rrv(&dec->cur, dec->edged_width, dec->mbs, |
|
|
(dec->width + 31) / 32, (dec->height + 31) / 32, dec->mb_width, |
|
|
16, 0); |
|
|
} |
|
|
|
|
1671 |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
/* note: for packed_mode, output is performed when the special-N_VOP is decoded */ |
1672 |
if (!(dec->low_delay_default && dec->packed_mode)) { |
if (!(dec->low_delay_default && dec->packed_mode)) { |
1673 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1681 |
} |
} |
1682 |
|
|
1683 |
image_swap(&dec->refn[0], &dec->refn[1]); |
image_swap(&dec->refn[0], &dec->refn[1]); |
1684 |
|
dec->is_edged[1] = dec->is_edged[0]; |
1685 |
image_swap(&dec->cur, &dec->refn[0]); |
image_swap(&dec->cur, &dec->refn[0]); |
1686 |
|
dec->is_edged[0] = 0; |
1687 |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
SWAP(MACROBLOCK *, dec->mbs, dec->last_mbs); |
|
dec->last_reduced_resolution = reduced_resolution; |
|
1688 |
dec->last_coding_type = coding_type; |
dec->last_coding_type = coding_type; |
1689 |
|
|
1690 |
dec->frames++; |
dec->frames++; |
1694 |
|
|
1695 |
if (dec->low_delay) { |
if (dec->low_delay) { |
1696 |
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
DPRINTF(XVID_DEBUG_ERROR, "warning: bvop found in low_delay==1 stream\n"); |
1697 |
dec->low_delay = 1; |
dec->low_delay = 0; |
1698 |
} |
} |
1699 |
|
|
1700 |
if (dec->frames < 2) { |
if (dec->frames < 2) { |
1729 |
|
|
1730 |
done : |
done : |
1731 |
|
|
1732 |
/* low_delay_default mode: if we've gotten here without outputting anything, |
/* if we reach here without outputing anything _and_ |
1733 |
then output the recently decoded frame, or print an error message */ |
the calling application has specified low_delay_default, |
1734 |
|
we *must* output something. |
1735 |
|
this always occurs on the first call to decode() call |
1736 |
|
when bframes are present in the bitstream. it may also |
1737 |
|
occur if no vops were seen in the bitstream |
1738 |
|
|
1739 |
|
if packed_mode is enabled, then we output the recently |
1740 |
|
decoded frame (the very first ivop). otherwise we have |
1741 |
|
nothing to display, and therefore output a black screen. |
1742 |
|
*/ |
1743 |
if (dec->low_delay_default && output == 0) { |
if (dec->low_delay_default && output == 0) { |
1744 |
if (dec->packed_mode && seen_something) { |
if (dec->packed_mode && seen_something) { |
|
/* output the recently decoded frame */ |
|
1745 |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
decoder_output(dec, &dec->refn[0], dec->last_mbs, frame, stats, dec->last_coding_type, quant); |
1746 |
} else { |
} else { |
1747 |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
image_clear(&dec->cur, dec->width, dec->height, dec->edged_width, 0, 128, 128); |
|
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"); |
|
|
|
|
1748 |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
decoder_output(dec, &dec->cur, NULL, frame, stats, P_VOP, quant); |
1749 |
if (stats) stats->type = XVID_TYPE_NOTHING; |
if (stats) stats->type = XVID_TYPE_NOTHING; |
1750 |
} |
} |