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" |
154 |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block, 6, 64, int16_t, CACHE_LINE); |
155 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
156 |
|
|
157 |
const uint32_t stride = dec->edged_width; |
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, 6*64*sizeof(int16_t)); // clear |
memset(block, 0, 6*64*sizeof(int16_t)); // clear |
169 |
|
|
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*64], stride); |
transfer_16to8copy(pY_Cur, &data[0*64], stride); |
240 |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
transfer_16to8copy(pY_Cur + 8, &data[1*64], stride); |
241 |
transfer_16to8copy(pY_Cur + 8 * stride, &data[2*64], stride); |
transfer_16to8copy(pY_Cur + next_block, &data[2*64], stride); |
242 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, &data[3*64], stride); |
transfer_16to8copy(pY_Cur + 8 + next_block, &data[3*64], stride); |
243 |
transfer_16to8copy(pU_Cur, &data[4*64], stride / 2); |
transfer_16to8copy(pU_Cur, &data[4*64], stride2); |
244 |
transfer_16to8copy(pV_Cur, &data[5*64], stride / 2); |
transfer_16to8copy(pV_Cur, &data[5*64], stride2); |
245 |
stop_transfer_timer(); |
stop_transfer_timer(); |
246 |
} |
} |
247 |
|
|
271 |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(block,6, 64, int16_t, CACHE_LINE); |
272 |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
DECLARE_ALIGNED_MATRIX(data, 6, 64, int16_t, CACHE_LINE); |
273 |
|
|
274 |
const uint32_t stride = dec->edged_width; |
uint32_t stride = dec->edged_width; |
275 |
const uint32_t stride2 = dec->edged_width / 2; |
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 |
{ |
{ |
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) |
349 |
if (cbp & 16) |
if (cbp & 16) |
350 |
transfer_16to8add(pY_Cur + 8, &data[1*64], 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*64], 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*64], 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*64], stride / 2); |
transfer_16to8add(pU_Cur, &data[4*64], stride2); |
357 |
if (cbp & 1) |
if (cbp & 1) |
358 |
transfer_16to8add(pV_Cur, &data[5*64], stride / 2); |
transfer_16to8add(pV_Cur, &data[5*64], stride2); |
359 |
stop_transfer_timer(); |
stop_transfer_timer(); |
360 |
} |
} |
361 |
|
|