42 |
* * |
* * |
43 |
* Revision history: * |
* Revision history: * |
44 |
* * |
* * |
45 |
|
* 26.03.2002 interlacing support - moved transfers outside loops |
46 |
* 22.12.2001 get_dc_scaler() moved to common.h |
* 22.12.2001 get_dc_scaler() moved to common.h |
47 |
* 19.11.2001 introduced coefficient thresholding (Isibaar) * |
* 19.11.2001 introduced coefficient thresholding (Isibaar) * |
48 |
* 17.11.2001 initial version * |
* 17.11.2001 initial version * |
69 |
/* this isnt pretty, but its better than 20 ifdefs */ |
/* this isnt pretty, but its better than 20 ifdefs */ |
70 |
|
|
71 |
void MBTransQuantIntra(const MBParam *pParam, |
void MBTransQuantIntra(const MBParam *pParam, |
72 |
|
MACROBLOCK * pMB, |
73 |
const uint32_t x_pos, |
const uint32_t x_pos, |
74 |
const uint32_t y_pos, |
const uint32_t y_pos, |
75 |
int16_t data[][64], |
int16_t data[][64], |
86 |
pU_Cur = pCurrent->u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
87 |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
88 |
|
|
|
for(i = 0; i < 6; i++) { |
|
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
|
|
|
|
89 |
start_timer(); |
start_timer(); |
|
|
|
|
switch(i) { |
|
|
case 0 : |
|
90 |
transfer_8to16copy(data[0], pY_Cur, stride); |
transfer_8to16copy(data[0], pY_Cur, stride); |
|
break; |
|
|
case 1 : |
|
91 |
transfer_8to16copy(data[1], pY_Cur + 8, stride); |
transfer_8to16copy(data[1], pY_Cur + 8, stride); |
|
break; |
|
|
case 2 : |
|
92 |
transfer_8to16copy(data[2], pY_Cur + 8 * stride, stride); |
transfer_8to16copy(data[2], pY_Cur + 8 * stride, stride); |
|
break; |
|
|
case 3 : |
|
93 |
transfer_8to16copy(data[3], pY_Cur + 8 * stride + 8, stride); |
transfer_8to16copy(data[3], pY_Cur + 8 * stride + 8, stride); |
|
break; |
|
|
case 4 : |
|
94 |
transfer_8to16copy(data[4], pU_Cur, stride / 2); |
transfer_8to16copy(data[4], pU_Cur, stride / 2); |
|
break; |
|
|
case 5 : |
|
95 |
transfer_8to16copy(data[5], pV_Cur, stride / 2); |
transfer_8to16copy(data[5], pV_Cur, stride / 2); |
|
break; |
|
|
} |
|
96 |
stop_transfer_timer(); |
stop_transfer_timer(); |
97 |
|
|
98 |
start_timer(); |
start_timer(); |
99 |
|
pMB->field_dct = 0; |
100 |
|
if (pParam->global_flags & XVID_INTERLACING) |
101 |
|
{ |
102 |
|
pMB->field_dct = MBDecideFieldDCT(data); |
103 |
|
} |
104 |
|
stop_interlacing_timer(); |
105 |
|
|
106 |
|
for(i = 0; i < 6; i++) |
107 |
|
{ |
108 |
|
uint32_t iDcScaler = get_dc_scaler(iQuant, i < 4); |
109 |
|
|
110 |
|
start_timer(); |
111 |
fdct(data[i]); |
fdct(data[i]); |
112 |
stop_dct_timer(); |
stop_dct_timer(); |
113 |
|
|
135 |
start_timer(); |
start_timer(); |
136 |
idct(data[i]); |
idct(data[i]); |
137 |
stop_idct_timer(); |
stop_idct_timer(); |
138 |
|
} |
139 |
|
|
140 |
start_timer(); |
start_timer(); |
141 |
|
if (pMB->field_dct) |
142 |
|
{ |
143 |
|
MBFieldToFrame(data); |
144 |
|
} |
145 |
|
stop_interlacing_timer(); |
146 |
|
|
147 |
switch(i) { |
start_timer(); |
|
case 0: |
|
148 |
transfer_16to8copy(pY_Cur, data[0], stride); |
transfer_16to8copy(pY_Cur, data[0], stride); |
|
break; |
|
|
case 1: |
|
149 |
transfer_16to8copy(pY_Cur + 8, data[1], stride); |
transfer_16to8copy(pY_Cur + 8, data[1], stride); |
|
break; |
|
|
case 2: |
|
150 |
transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8copy(pY_Cur + 8 * stride, data[2], stride); |
|
break; |
|
|
case 3: |
|
151 |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8copy(pY_Cur + 8 + 8 * stride, data[3], stride); |
|
break; |
|
|
case 4: |
|
152 |
transfer_16to8copy(pU_Cur, data[4], stride / 2); |
transfer_16to8copy(pU_Cur, data[4], stride / 2); |
|
break; |
|
|
case 5: |
|
153 |
transfer_16to8copy(pV_Cur, data[5], stride / 2); |
transfer_16to8copy(pV_Cur, data[5], stride / 2); |
|
break; |
|
|
} |
|
154 |
stop_transfer_timer(); |
stop_transfer_timer(); |
155 |
} |
} |
|
} |
|
156 |
|
|
157 |
|
|
158 |
uint8_t MBTransQuantInter(const MBParam *pParam, |
uint8_t MBTransQuantInter(const MBParam *pParam, |
159 |
|
MACROBLOCK * pMB, |
160 |
const uint32_t x_pos, const uint32_t y_pos, |
const uint32_t x_pos, const uint32_t y_pos, |
161 |
int16_t data[][64], |
int16_t data[][64], |
162 |
int16_t qcoeff[][64], |
int16_t qcoeff[][64], |
164 |
|
|
165 |
{ |
{ |
166 |
const uint32_t stride = pParam->edged_width; |
const uint32_t stride = pParam->edged_width; |
167 |
uint8_t i; |
uint32_t i; |
168 |
uint8_t iQuant = pParam->quant; |
uint32_t iQuant = pParam->quant; |
169 |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
uint8_t *pY_Cur, *pU_Cur, *pV_Cur; |
170 |
uint8_t cbp = 0; |
uint8_t cbp = 0; |
171 |
uint32_t sum; |
uint32_t sum; |
174 |
pU_Cur = pCurrent->u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pU_Cur = pCurrent->u + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
175 |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
pV_Cur = pCurrent->v + (y_pos << 3) * (stride >> 1) + (x_pos << 3); |
176 |
|
|
177 |
for(i = 0; i < 6; i++) { |
start_timer(); |
178 |
|
pMB->field_dct = 0; |
179 |
|
if (pParam->global_flags & XVID_INTERLACING) |
180 |
|
{ |
181 |
|
pMB->field_dct = MBDecideFieldDCT(data); |
182 |
|
} |
183 |
|
stop_interlacing_timer(); |
184 |
|
|
185 |
|
for(i = 0; i < 6; i++) |
186 |
|
{ |
187 |
/* |
/* |
188 |
no need to transfer 8->16-bit |
no need to transfer 8->16-bit |
189 |
(this is performed already in motion compensation) |
(this is performed already in motion compensation) |
225 |
start_timer(); |
start_timer(); |
226 |
idct(data[i]); |
idct(data[i]); |
227 |
stop_idct_timer(); |
stop_idct_timer(); |
228 |
|
} |
229 |
|
} |
230 |
|
|
231 |
start_timer(); |
start_timer(); |
232 |
|
if (pMB->field_dct) |
233 |
|
{ |
234 |
|
MBFieldToFrame(data); |
235 |
|
} |
236 |
|
stop_interlacing_timer(); |
237 |
|
|
238 |
switch(i) { |
start_timer(); |
239 |
case 0: |
if (cbp & 32) |
240 |
transfer_16to8add(pY_Cur, data[0], stride); |
transfer_16to8add(pY_Cur, data[0], stride); |
241 |
break; |
if (cbp & 16) |
|
case 1: |
|
242 |
transfer_16to8add(pY_Cur + 8, data[1], stride); |
transfer_16to8add(pY_Cur + 8, data[1], stride); |
243 |
break; |
if (cbp & 8) |
|
case 2: |
|
244 |
transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); |
transfer_16to8add(pY_Cur + 8 * stride, data[2], stride); |
245 |
break; |
if (cbp & 4) |
|
case 3: |
|
246 |
transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); |
transfer_16to8add(pY_Cur + 8 + 8 * stride, data[3], stride); |
247 |
break; |
if (cbp & 2) |
|
case 4: |
|
248 |
transfer_16to8add(pU_Cur, data[4], stride / 2); |
transfer_16to8add(pU_Cur, data[4], stride / 2); |
249 |
break; |
if (cbp & 1) |
|
case 5: |
|
250 |
transfer_16to8add(pV_Cur, data[5], stride / 2); |
transfer_16to8add(pV_Cur, data[5], stride / 2); |
|
break; |
|
|
} |
|
251 |
stop_transfer_timer(); |
stop_transfer_timer(); |
252 |
|
|
253 |
|
return cbp; |
254 |
} |
} |
255 |
|
|
256 |
|
|
257 |
|
/* if sum(diff between field lines) < sum(diff between frame lines), use field dct */ |
258 |
|
|
259 |
|
#define ABS(X) (X)<0 ? -(X) : (X) |
260 |
|
|
261 |
|
uint32_t MBDecideFieldDCT(int16_t data[][64]) |
262 |
|
{ |
263 |
|
const uint8_t blocks[] = {0, 0, 0, 0, 2, 2, 2, 2}; |
264 |
|
const uint8_t lines[] = {0, 16, 32, 48, 0, 16, 32, 48}; |
265 |
|
|
266 |
|
int frame = 0, field = 0; |
267 |
|
int i, j; |
268 |
|
|
269 |
|
for (i=0 ; i<7 ; ++i) |
270 |
|
{ |
271 |
|
for (j=0 ; j<8 ; ++j) |
272 |
|
{ |
273 |
|
frame += ABS(data[0][(i+1)*8 + j] - data[0][i*8 + j]); |
274 |
|
frame += ABS(data[1][(i+1)*8 + j] - data[1][i*8 + j]); |
275 |
|
frame += ABS(data[2][(i+1)*8 + j] - data[2][i*8 + j]); |
276 |
|
frame += ABS(data[3][(i+1)*8 + j] - data[3][i*8 + j]); |
277 |
|
|
278 |
|
field += ABS(data[blocks[i+1]][lines[i+1] + j] - data[blocks[i]][lines[i] + j]); |
279 |
|
field += ABS(data[blocks[i+1]][lines[i+1] + 8 + j] - data[blocks[i]][lines[i] + 8 + j]); |
280 |
|
field += ABS(data[blocks[i+1]+1][lines[i+1] + j] - data[blocks[i]+1][lines[i] + j]); |
281 |
|
field += ABS(data[blocks[i+1]+1][lines[i+1] + 8 + j] - data[blocks[i]+1][lines[i] + 8 + j]); |
282 |
} |
} |
283 |
return cbp; |
} |
284 |
|
|
285 |
|
if (frame > field) |
286 |
|
{ |
287 |
|
MBFrameToField(data); |
288 |
|
} |
289 |
|
|
290 |
|
return (frame > field); |
291 |
|
} |
292 |
|
|
293 |
|
|
294 |
|
/* deinterlace Y blocks vertically */ |
295 |
|
|
296 |
|
#define MOVLINE(X,Y) memcpy(X, Y, sizeof(tmp)) |
297 |
|
#define LINE(X,Y) &data[X][Y*8] |
298 |
|
|
299 |
|
void MBFrameToField(int16_t data[][64]) |
300 |
|
{ |
301 |
|
int16_t tmp[8]; |
302 |
|
|
303 |
|
/* left blocks */ |
304 |
|
|
305 |
|
// 1=2, 2=4, 4=8, 8=1 |
306 |
|
MOVLINE(tmp, LINE(0,1)); |
307 |
|
MOVLINE(LINE(0,1), LINE(0,2)); |
308 |
|
MOVLINE(LINE(0,2), LINE(0,4)); |
309 |
|
MOVLINE(LINE(0,4), LINE(2,0)); |
310 |
|
MOVLINE(LINE(2,0), tmp); |
311 |
|
|
312 |
|
// 3=6, 6=12, 12=9, 9=3 |
313 |
|
MOVLINE(tmp, LINE(0,3)); |
314 |
|
MOVLINE(LINE(0,3), LINE(0,6)); |
315 |
|
MOVLINE(LINE(0,6), LINE(2,4)); |
316 |
|
MOVLINE(LINE(2,4), LINE(2,1)); |
317 |
|
MOVLINE(LINE(2,1), tmp); |
318 |
|
|
319 |
|
// 5=10, 10=5 |
320 |
|
MOVLINE(tmp, LINE(0,5)); |
321 |
|
MOVLINE(LINE(0,5), LINE(2,2)); |
322 |
|
MOVLINE(LINE(2,2), tmp); |
323 |
|
|
324 |
|
// 7=14, 14=13, 13=11, 11=7 |
325 |
|
MOVLINE(tmp, LINE(0,7)); |
326 |
|
MOVLINE(LINE(0,7), LINE(2,6)); |
327 |
|
MOVLINE(LINE(2,6), LINE(2,5)); |
328 |
|
MOVLINE(LINE(2,5), LINE(2,3)); |
329 |
|
MOVLINE(LINE(2,3), tmp); |
330 |
|
|
331 |
|
/* right blocks */ |
332 |
|
|
333 |
|
// 1=2, 2=4, 4=8, 8=1 |
334 |
|
MOVLINE(tmp, LINE(1,1)); |
335 |
|
MOVLINE(LINE(1,1), LINE(1,2)); |
336 |
|
MOVLINE(LINE(1,2), LINE(1,4)); |
337 |
|
MOVLINE(LINE(1,4), LINE(3,0)); |
338 |
|
MOVLINE(LINE(3,0), tmp); |
339 |
|
|
340 |
|
// 3=6, 6=12, 12=9, 9=3 |
341 |
|
MOVLINE(tmp, LINE(1,3)); |
342 |
|
MOVLINE(LINE(1,3), LINE(1,6)); |
343 |
|
MOVLINE(LINE(1,6), LINE(3,4)); |
344 |
|
MOVLINE(LINE(3,4), LINE(3,1)); |
345 |
|
MOVLINE(LINE(3,1), tmp); |
346 |
|
|
347 |
|
// 5=10, 10=5 |
348 |
|
MOVLINE(tmp, LINE(1,5)); |
349 |
|
MOVLINE(LINE(1,5), LINE(3,2)); |
350 |
|
MOVLINE(LINE(3,2), tmp); |
351 |
|
|
352 |
|
// 7=14, 14=13, 13=11, 11=7 |
353 |
|
MOVLINE(tmp, LINE(1,7)); |
354 |
|
MOVLINE(LINE(1,7), LINE(3,6)); |
355 |
|
MOVLINE(LINE(3,6), LINE(3,5)); |
356 |
|
MOVLINE(LINE(3,5), LINE(3,3)); |
357 |
|
MOVLINE(LINE(3,3), tmp); |
358 |
|
} |
359 |
|
|
360 |
|
|
361 |
|
/* interlace Y blocks vertically */ |
362 |
|
|
363 |
|
void MBFieldToFrame(int16_t data[][64]) |
364 |
|
{ |
365 |
|
uint16_t tmp[8]; |
366 |
|
|
367 |
|
/* left blocks */ |
368 |
|
|
369 |
|
// 1=8, 8=4, 4=2, 2=1 |
370 |
|
MOVLINE(tmp, LINE(0,1)); |
371 |
|
MOVLINE(LINE(0,1), LINE(2,0)); |
372 |
|
MOVLINE(LINE(2,0), LINE(0,4)); |
373 |
|
MOVLINE(LINE(0,4), LINE(0,2)); |
374 |
|
MOVLINE(LINE(0,2), tmp); |
375 |
|
|
376 |
|
// 3=9, 9=12, 12=6, 6=3 |
377 |
|
MOVLINE(tmp, LINE(0,3)); |
378 |
|
MOVLINE(LINE(0,3), LINE(2,1)); |
379 |
|
MOVLINE(LINE(2,1), LINE(2,4)); |
380 |
|
MOVLINE(LINE(2,4), LINE(0,6)); |
381 |
|
MOVLINE(LINE(0,6), tmp); |
382 |
|
|
383 |
|
// 5=10, 10=5 |
384 |
|
MOVLINE(tmp, LINE(0,5)); |
385 |
|
MOVLINE(LINE(0,5), LINE(2,2)); |
386 |
|
MOVLINE(LINE(2,2), tmp); |
387 |
|
|
388 |
|
// 7=11, 11=13, 13=14, 14=7 |
389 |
|
MOVLINE(tmp, LINE(0,7)); |
390 |
|
MOVLINE(LINE(0,7), LINE(2,3)); |
391 |
|
MOVLINE(LINE(2,3), LINE(2,5)); |
392 |
|
MOVLINE(LINE(2,5), LINE(2,6)); |
393 |
|
MOVLINE(LINE(2,6), tmp); |
394 |
|
|
395 |
|
/* right blocks */ |
396 |
|
|
397 |
|
// 1=8, 8=4, 4=2, 2=1 |
398 |
|
MOVLINE(tmp, LINE(1,1)); |
399 |
|
MOVLINE(LINE(1,1), LINE(3,0)); |
400 |
|
MOVLINE(LINE(3,0), LINE(1,4)); |
401 |
|
MOVLINE(LINE(1,4), LINE(1,2)); |
402 |
|
MOVLINE(LINE(1,2), tmp); |
403 |
|
|
404 |
|
// 3=9, 9=12, 12=6, 6=3 |
405 |
|
MOVLINE(tmp, LINE(1,3)); |
406 |
|
MOVLINE(LINE(1,3), LINE(3,1)); |
407 |
|
MOVLINE(LINE(3,1), LINE(3,4)); |
408 |
|
MOVLINE(LINE(3,4), LINE(1,6)); |
409 |
|
MOVLINE(LINE(1,6), tmp); |
410 |
|
|
411 |
|
// 5=10, 10=5 |
412 |
|
MOVLINE(tmp, LINE(1,5)); |
413 |
|
MOVLINE(LINE(1,5), LINE(3,2)); |
414 |
|
MOVLINE(LINE(3,2), tmp); |
415 |
|
|
416 |
|
// 7=11, 11=13, 13=14, 14=7 |
417 |
|
MOVLINE(tmp, LINE(1,7)); |
418 |
|
MOVLINE(LINE(1,7), LINE(3,3)); |
419 |
|
MOVLINE(LINE(3,3), LINE(3,5)); |
420 |
|
MOVLINE(LINE(3,5), LINE(3,6)); |
421 |
|
MOVLINE(LINE(3,6), tmp); |
422 |
} |
} |