32 |
* |
* |
33 |
* History: |
* History: |
34 |
* |
* |
35 |
|
* 29.03.2002 interlacing fix - compensated block wasn't being used when |
36 |
|
* reconstructing blocks, thus artifacts |
37 |
|
* interlacing speedup - used transfers to re-interlace |
38 |
|
* interlaced decoding should be as fast as progressive now |
39 |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
* 26.03.2002 interlacing support - moved transfers outside decode loop |
40 |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
* 26.12.2001 decoder_mbinter: dequant/idct moved within if(coded) block |
41 |
* 22.12.2001 block based interpolation |
* 22.12.2001 block based interpolation |
59 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
60 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
61 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
|
#include "utils/mbfunctions.h" |
|
62 |
|
|
63 |
#include "bitstream/mbcoding.h" |
#include "bitstream/mbcoding.h" |
64 |
#include "prediction/mbprediction.h" |
#include "prediction/mbprediction.h" |
150 |
const uint32_t quant, |
const uint32_t quant, |
151 |
const uint32_t intra_dc_threshold) |
const uint32_t intra_dc_threshold) |
152 |
{ |
{ |
|
CACHE_ALIGN int16_t block[6][64]; |
|
|
CACHE_ALIGN int16_t data[6][64]; |
|
153 |
|
|
154 |
const uint32_t stride = dec->edged_width; |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
155 |
|
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
156 |
|
|
157 |
|
uint32_t stride = dec->edged_width; |
158 |
|
uint32_t stride2 = stride / 2; |
159 |
|
uint32_t next_block = stride * 8; |
160 |
uint32_t i; |
uint32_t i; |
161 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
162 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
163 |
|
|
164 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
165 |
pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
166 |
pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
167 |
|
|
168 |
memset(block, 0, sizeof(block)); // clear |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
169 |
|
|
170 |
for (i = 0; i < 6; i++) |
for (i = 0; i < 6; i++) |
171 |
{ |
{ |
174 |
int start_coeff; |
int start_coeff; |
175 |
|
|
176 |
start_timer(); |
start_timer(); |
177 |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, block[i], iQuant, iDcScaler, predictors); |
predict_acdc(dec->mbs, x_pos, y_pos, dec->mb_width, i, &block[i*64], iQuant, iDcScaler, predictors); |
178 |
if (!acpred_flag) |
if (!acpred_flag) |
179 |
{ |
{ |
180 |
pMB->acpred_directions[i] = 0; |
pMB->acpred_directions[i] = 0; |
194 |
BitstreamSkip(bs, 1); // marker |
BitstreamSkip(bs, 1); // marker |
195 |
} |
} |
196 |
|
|
197 |
block[i][0] = dc_dif; |
block[i*64 + 0] = dc_dif; |
198 |
start_coeff = 1; |
start_coeff = 1; |
199 |
} |
} |
200 |
else |
else |
205 |
start_timer(); |
start_timer(); |
206 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
207 |
{ |
{ |
208 |
get_intra_block(bs, block[i], pMB->acpred_directions[i], start_coeff); |
get_intra_block(bs, &block[i*64], pMB->acpred_directions[i], start_coeff); |
209 |
} |
} |
210 |
stop_coding_timer(); |
stop_coding_timer(); |
211 |
|
|
212 |
start_timer(); |
start_timer(); |
213 |
add_acdc(pMB, i, block[i], iDcScaler, predictors); |
add_acdc(pMB, i, &block[i*64], iDcScaler, predictors); |
214 |
stop_prediction_timer(); |
stop_prediction_timer(); |
215 |
|
|
216 |
start_timer(); |
start_timer(); |
217 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) |
218 |
{ |
{ |
219 |
dequant_intra(data[i], block[i], iQuant, iDcScaler); |
dequant_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
220 |
} |
} |
221 |
else |
else |
222 |
{ |
{ |
223 |
dequant4_intra(data[i], block[i], iQuant, iDcScaler); |
dequant4_intra(&data[i*64], &block[i*64], iQuant, iDcScaler); |
224 |
} |
} |
225 |
stop_iquant_timer(); |
stop_iquant_timer(); |
226 |
|
|
227 |
start_timer(); |
start_timer(); |
228 |
idct(data[i]); |
idct(&data[i*64]); |
229 |
stop_idct_timer(); |
stop_idct_timer(); |
230 |
} |
} |
231 |
|
|
232 |
start_timer(); |
if (pMB->field_dct) |
|
if (dec->interlacing && pMB->field_dct) |
|
233 |
{ |
{ |
234 |
MBFieldToFrame(data); |
next_block = stride; |
235 |
|
stride *= 2; |
236 |
} |
} |
|
stop_interlacing_timer(); |
|
237 |
|
|
238 |
start_timer(); |
start_timer(); |
239 |
transfer_16to8copy(pY_Cur, data[0], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
240 |
transfer_16to8copy(pY_Cur + 8, data[1], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
241 |
transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
242 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
243 |
transfer_16to8copy(pU_Cur, data[4], stride / 2); |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
244 |
transfer_16to8copy(pV_Cur, data[5], stride / 2); |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
245 |
stop_transfer_timer(); |
stop_transfer_timer(); |
246 |
} |
} |
247 |
|
|
267 |
const uint32_t quant, |
const uint32_t quant, |
268 |
const uint32_t rounding) |
const uint32_t rounding) |
269 |
{ |
{ |
|
CACHE_ALIGN int16_t block[6][64]; |
|
|
CACHE_ALIGN int16_t data[6][64]; |
|
270 |
|
|
271 |
const uint32_t stride = dec->edged_width; |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
272 |
const uint32_t stride2 = dec->edged_width / 2; |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
273 |
|
|
274 |
|
uint32_t stride = dec->edged_width; |
275 |
|
uint32_t stride2 = stride / 2; |
276 |
|
uint32_t next_block = stride * 8; |
277 |
uint32_t i; |
uint32_t i; |
278 |
uint32_t iQuant = pMB->quant; |
uint32_t iQuant = pMB->quant; |
279 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
280 |
int uv_dx, uv_dy; |
int uv_dx, uv_dy; |
281 |
|
|
282 |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
pY_Cur = dec->cur.y + (y_pos << 4) * stride + (x_pos << 4); |
283 |
pU_Cur = dec->cur.u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = dec->cur.u + (y_pos << 3) * stride2 + (x_pos << 3); |
284 |
pV_Cur = dec->cur.v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = dec->cur.v + (y_pos << 3) * stride2 + (x_pos << 3); |
285 |
|
|
286 |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
if (pMB->mode == MODE_INTER || pMB->mode == MODE_INTER_Q) |
287 |
{ |
{ |
314 |
{ |
{ |
315 |
if (cbp & (1 << (5-i))) // coded |
if (cbp & (1 << (5-i))) // coded |
316 |
{ |
{ |
317 |
memset(block[i], 0, 64 * sizeof(int16_t)); // clear |
memset(&block[i*64], 0, 64 * sizeof(int16_t)); // clear |
318 |
|
|
319 |
start_timer(); |
start_timer(); |
320 |
get_inter_block(bs, block[i]); |
get_inter_block(bs, &block[i*64]); |
321 |
stop_coding_timer(); |
stop_coding_timer(); |
322 |
|
|
323 |
start_timer(); |
start_timer(); |
324 |
if (dec->quant_type == 0) |
if (dec->quant_type == 0) |
325 |
{ |
{ |
326 |
dequant_inter(data[i], block[i], iQuant); |
dequant_inter(&data[i*64], &block[i*64], iQuant); |
327 |
} |
} |
328 |
else |
else |
329 |
{ |
{ |
330 |
dequant4_inter(data[i], block[i], iQuant); |
dequant4_inter(&data[i*64], &block[i*64], iQuant); |
331 |
} |
} |
332 |
stop_iquant_timer(); |
stop_iquant_timer(); |
333 |
|
|
334 |
start_timer(); |
start_timer(); |
335 |
idct(data[i]); |
idct(&data[i*64]); |
336 |
stop_idct_timer(); |
stop_idct_timer(); |
337 |
} |
} |
338 |
} |
} |
339 |
|
|
|
start_timer(); |
|
340 |
if (pMB->field_dct) |
if (pMB->field_dct) |
341 |
{ |
{ |
342 |
MBFieldToFrame(data); |
next_block = stride; |
343 |
|
stride *= 2; |
344 |
} |
} |
|
stop_interlacing_timer(); |
|
345 |
|
|
346 |
start_timer(); |
start_timer(); |
347 |
if (cbp & 32) |
if (cbp & 32) |
348 |
transfer_16to8add(pY_Cur, data[0], stride); |
transfer_16to8add(pY_Cur, &data[0*64], stride); |
349 |
if (cbp & 16) |
if (cbp & 16) |
350 |
transfer_16to8add(pY_Cur + 8, data[1], stride); |
transfer_16to8add(pY_Cur + 8, &data[1*64], stride); |
351 |
if (cbp & 8) |
if (cbp & 8) |
352 |
transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8add(pY_Cur + next_block, &data[2*64], stride); |
353 |
if (cbp & 4) |
if (cbp & 4) |
354 |
transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8add(pY_Cur + 8 + next_block, &data[3*64], stride); |
355 |
if (cbp & 2) |
if (cbp & 2) |
356 |
transfer_16to8add(pU_Cur, data[4], stride / 2); |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
357 |
if (cbp & 1) |
if (cbp & 1) |
358 |
transfer_16to8add(pV_Cur, data[5], stride / 2); |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
359 |
stop_transfer_timer(); |
stop_transfer_timer(); |
360 |
} |
} |
361 |
|
|
362 |
|
|
363 |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
void decoder_iframe(DECODER * dec, Bitstream * bs, int quant, int intra_dc_threshold) |
364 |
{ |
{ |
365 |
|
|
366 |
uint32_t x, y; |
uint32_t x, y; |
367 |
|
|
368 |
for (y = 0; y < dec->mb_height; y++) |
for (y = 0; y < dec->mb_height; y++) |
415 |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
decoder_mbintra(dec, mb, x, y, acpred_flag, cbp, bs, quant, intra_dc_threshold); |
416 |
} |
} |
417 |
} |
} |
418 |
|
|
419 |
} |
} |
420 |
|
|
421 |
|
|
422 |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
void get_motion_vector(DECODER *dec, Bitstream *bs, int x, int y, int k, VECTOR * mv, int fcode) |
423 |
{ |
{ |
424 |
|
|
425 |
int scale_fac = 1 << (fcode - 1); |
int scale_fac = 1 << (fcode - 1); |
426 |
int high = (32 * scale_fac) - 1; |
int high = (32 * scale_fac) - 1; |
427 |
int low = ((-32) * scale_fac); |
int low = ((-32) * scale_fac); |
471 |
|
|
472 |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
void decoder_pframe(DECODER * dec, Bitstream * bs, int rounding, int quant, int fcode, int intra_dc_threshold) |
473 |
{ |
{ |
474 |
|
|
475 |
uint32_t x, y; |
uint32_t x, y; |
476 |
|
|
477 |
image_swap(&dec->cur, &dec->refn); |
image_swap(&dec->cur, &dec->refn); |
624 |
|
|
625 |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
int decoder_decode(DECODER * dec, XVID_DEC_FRAME * frame) |
626 |
{ |
{ |
627 |
|
|
628 |
Bitstream bs; |
Bitstream bs; |
629 |
uint32_t rounding; |
uint32_t rounding; |
630 |
uint32_t quant; |
uint32_t quant; |
668 |
stop_global_timer(); |
stop_global_timer(); |
669 |
|
|
670 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
671 |
|
|
672 |
} |
} |