37 |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
* - 22.12.2001 API change: added xvid_init() - Isibaar |
38 |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
* - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
39 |
* |
* |
40 |
* $Id: xvid.c,v 1.32 2002-07-21 23:34:08 chl Exp $ |
* $Id: xvid.c,v 1.33.2.19 2002-12-30 10:49:17 suxen_drol Exp $ |
41 |
* |
* |
42 |
****************************************************************************/ |
****************************************************************************/ |
43 |
|
|
49 |
#include "dct/fdct.h" |
#include "dct/fdct.h" |
50 |
#include "image/colorspace.h" |
#include "image/colorspace.h" |
51 |
#include "image/interpolate8x8.h" |
#include "image/interpolate8x8.h" |
52 |
|
#include "image/reduced.h" |
53 |
#include "utils/mem_transfer.h" |
#include "utils/mem_transfer.h" |
54 |
|
#include "utils/mbfunctions.h" |
55 |
#include "quant/quant_h263.h" |
#include "quant/quant_h263.h" |
56 |
#include "quant/quant_mpeg4.h" |
#include "quant/quant_mpeg4.h" |
57 |
#include "motion/motion.h" |
#include "motion/motion.h" |
141 |
* |
* |
142 |
****************************************************************************/ |
****************************************************************************/ |
143 |
|
|
144 |
int |
|
145 |
xvid_init(void *handle, |
static |
146 |
int opt, |
int xvid_init_init(XVID_INIT_PARAM * init_param) |
|
void *param1, |
|
|
void *param2) |
|
147 |
{ |
{ |
148 |
int cpu_flags; |
int cpu_flags; |
|
XVID_INIT_PARAM *init_param; |
|
|
|
|
|
init_param = (XVID_INIT_PARAM *) param1; |
|
149 |
|
|
150 |
/* Inform the client the API version */ |
/* Inform the client the API version */ |
151 |
init_param->api_version = API_VERSION; |
init_param->api_version = API_VERSION; |
213 |
transfer_16to8add = transfer_16to8add_c; |
transfer_16to8add = transfer_16to8add_c; |
214 |
transfer8x8_copy = transfer8x8_copy_c; |
transfer8x8_copy = transfer8x8_copy_c; |
215 |
|
|
216 |
|
/* Interlacing functions */ |
217 |
|
MBFieldTest = MBFieldTest_c; |
218 |
|
|
219 |
/* Image interpolation related functions */ |
/* Image interpolation related functions */ |
220 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; |
221 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; |
222 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
223 |
|
|
224 |
|
interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
225 |
|
interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; |
226 |
|
interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; |
227 |
|
|
228 |
|
interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; |
229 |
|
interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; |
230 |
|
interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; |
231 |
|
|
232 |
|
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; |
233 |
|
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; |
234 |
|
|
235 |
|
interpolate8x8_avg2 = interpolate8x8_avg2_c; |
236 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_c; |
237 |
|
|
238 |
|
/* reduced resoltuion */ |
239 |
|
|
240 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
241 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; |
242 |
|
#ifdef ARCH_X86 |
243 |
|
vfilter_31 = xvid_VFilter_31_x86; |
244 |
|
hfilter_31 = xvid_HFilter_31_x86; |
245 |
|
#else |
246 |
|
vfilter_31 = xvid_VFilter_31_C; |
247 |
|
hfilter_31 = xvid_HFilter_31_C; |
248 |
|
#endif |
249 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; |
250 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; |
251 |
|
|
252 |
/* Initialize internal colorspace transformation tables */ |
/* Initialize internal colorspace transformation tables */ |
253 |
colorspace_init(); |
colorspace_init(); |
254 |
|
|
255 |
/* All colorspace transformation functions User Format->YV12 */ |
/* All colorspace transformation functions User Format->YV12 */ |
256 |
|
yv12_to_yv12 = yv12_to_yv12_c; |
257 |
rgb555_to_yv12 = rgb555_to_yv12_c; |
rgb555_to_yv12 = rgb555_to_yv12_c; |
258 |
rgb565_to_yv12 = rgb565_to_yv12_c; |
rgb565_to_yv12 = rgb565_to_yv12_c; |
259 |
rgb24_to_yv12 = rgb24_to_yv12_c; |
bgr_to_yv12 = bgr_to_yv12_c; |
260 |
rgb32_to_yv12 = rgb32_to_yv12_c; |
bgra_to_yv12 = bgra_to_yv12_c; |
261 |
yuv_to_yv12 = yuv_to_yv12_c; |
abgr_to_yv12 = abgr_to_yv12_c; |
262 |
|
rgba_to_yv12 = rgba_to_yv12_c; |
263 |
yuyv_to_yv12 = yuyv_to_yv12_c; |
yuyv_to_yv12 = yuyv_to_yv12_c; |
264 |
uyvy_to_yv12 = uyvy_to_yv12_c; |
uyvy_to_yv12 = uyvy_to_yv12_c; |
265 |
|
|
266 |
|
rgb555i_to_yv12 = rgb555i_to_yv12_c; |
267 |
|
rgb565i_to_yv12 = rgb565i_to_yv12_c; |
268 |
|
bgri_to_yv12 = bgri_to_yv12_c; |
269 |
|
bgrai_to_yv12 = bgrai_to_yv12_c; |
270 |
|
abgri_to_yv12 = abgri_to_yv12_c; |
271 |
|
rgbai_to_yv12 = rgbai_to_yv12_c; |
272 |
|
yuyvi_to_yv12 = yuyvi_to_yv12_c; |
273 |
|
uyvyi_to_yv12 = uyvyi_to_yv12_c; |
274 |
|
|
275 |
|
|
276 |
/* All colorspace transformation functions YV12->User format */ |
/* All colorspace transformation functions YV12->User format */ |
277 |
yv12_to_rgb555 = yv12_to_rgb555_c; |
yv12_to_rgb555 = yv12_to_rgb555_c; |
278 |
yv12_to_rgb565 = yv12_to_rgb565_c; |
yv12_to_rgb565 = yv12_to_rgb565_c; |
279 |
yv12_to_rgb24 = yv12_to_rgb24_c; |
yv12_to_bgr = yv12_to_bgr_c; |
280 |
yv12_to_rgb32 = yv12_to_rgb32_c; |
yv12_to_bgra = yv12_to_bgra_c; |
281 |
yv12_to_yuv = yv12_to_yuv_c; |
yv12_to_abgr = yv12_to_abgr_c; |
282 |
|
yv12_to_rgba = yv12_to_rgba_c; |
283 |
yv12_to_yuyv = yv12_to_yuyv_c; |
yv12_to_yuyv = yv12_to_yuyv_c; |
284 |
yv12_to_uyvy = yv12_to_uyvy_c; |
yv12_to_uyvy = yv12_to_uyvy_c; |
285 |
|
|
286 |
|
yv12_to_rgb555i = yv12_to_rgb555i_c; |
287 |
|
yv12_to_rgb565i = yv12_to_rgb565i_c; |
288 |
|
yv12_to_bgri = yv12_to_bgri_c; |
289 |
|
yv12_to_bgrai = yv12_to_bgrai_c; |
290 |
|
yv12_to_abgri = yv12_to_abgri_c; |
291 |
|
yv12_to_rgbai = yv12_to_rgbai_c; |
292 |
|
yv12_to_yuyvi = yv12_to_yuyvi_c; |
293 |
|
yv12_to_uyvyi = yv12_to_uyvyi_c; |
294 |
|
|
295 |
/* Functions used in motion estimation algorithms */ |
/* Functions used in motion estimation algorithms */ |
296 |
calc_cbp = calc_cbp_c; |
calc_cbp = calc_cbp_c; |
297 |
sad16 = sad16_c; |
sad16 = sad16_c; |
|
sad16bi = sad16bi_c; |
|
298 |
sad8 = sad8_c; |
sad8 = sad8_c; |
299 |
dev16 = dev16_c; |
sad16bi = sad16bi_c; |
300 |
sad8bi = sad8bi_c; |
sad8bi = sad8bi_c; |
301 |
Halfpel8_Refine = Halfpel8_Refine_c; |
dev16 = dev16_c; |
302 |
|
sad16v = sad16v_c; |
303 |
|
|
304 |
|
// Halfpel8_Refine = Halfpel8_Refine_c; |
305 |
|
|
306 |
#ifdef ARCH_X86 |
#ifdef ARCH_X86 |
307 |
|
|
308 |
|
if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || |
309 |
|
(cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || |
310 |
|
(cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) |
311 |
|
{ |
312 |
|
/* Restore FPU context : emms_c is a nop functions */ |
313 |
|
emms = emms_mmx; |
314 |
|
} |
315 |
|
|
316 |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
if ((cpu_flags & XVID_CPU_MMX) > 0) { |
317 |
|
|
318 |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
/* Forward and Inverse Discrete Cosine Transformation functions */ |
319 |
fdct = fdct_mmx; |
fdct = fdct_mmx; |
320 |
idct = idct_mmx; |
idct = idct_mmx; |
321 |
|
|
|
/* To restore FPU context after mmx use */ |
|
|
emms = emms_mmx; |
|
|
|
|
322 |
/* Quantization related functions */ |
/* Quantization related functions */ |
323 |
quant_intra = quant_intra_mmx; |
quant_intra = quant_intra_mmx; |
324 |
dequant_intra = dequant_intra_mmx; |
dequant_intra = dequant_intra_mmx; |
338 |
transfer_16to8add = transfer_16to8add_mmx; |
transfer_16to8add = transfer_16to8add_mmx; |
339 |
transfer8x8_copy = transfer8x8_copy_mmx; |
transfer8x8_copy = transfer8x8_copy_mmx; |
340 |
|
|
341 |
|
/* Interlacing Functions */ |
342 |
|
MBFieldTest = MBFieldTest_mmx; |
343 |
|
|
344 |
/* Image Interpolation related functions */ |
/* Image Interpolation related functions */ |
345 |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; |
346 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; |
347 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
348 |
|
|
349 |
/* Image RGB->YV12 related functions */ |
interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
350 |
rgb24_to_yv12 = rgb24_to_yv12_mmx; |
interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; |
351 |
rgb32_to_yv12 = rgb32_to_yv12_mmx; |
|
352 |
yuv_to_yv12 = yuv_to_yv12_mmx; |
interpolate8x8_avg2 = interpolate8x8_avg2_mmx; |
353 |
|
interpolate8x8_avg4 = interpolate8x8_avg4_mmx; |
354 |
|
|
355 |
|
/* reduced resolution */ |
356 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; |
357 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; |
358 |
|
hfilter_31 = xvid_HFilter_31_mmx; |
359 |
|
filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; |
360 |
|
filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; |
361 |
|
|
362 |
|
/* image input xxx_to_yv12 related functions */ |
363 |
|
yv12_to_yv12 = yv12_to_yv12_mmx; |
364 |
|
bgr_to_yv12 = bgr_to_yv12_mmx; |
365 |
|
bgra_to_yv12 = bgra_to_yv12_mmx; |
366 |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
yuyv_to_yv12 = yuyv_to_yv12_mmx; |
367 |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
uyvy_to_yv12 = uyvy_to_yv12_mmx; |
368 |
|
|
369 |
/* Image YV12->RGB related functions */ |
/* image output yv12_to_xxx related functions */ |
370 |
yv12_to_rgb24 = yv12_to_rgb24_mmx; |
yv12_to_bgr = yv12_to_bgr_mmx; |
371 |
yv12_to_rgb32 = yv12_to_rgb32_mmx; |
yv12_to_bgra = yv12_to_bgra_mmx; |
372 |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
yv12_to_yuyv = yv12_to_yuyv_mmx; |
373 |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
yv12_to_uyvy = yv12_to_uyvy_mmx; |
374 |
|
|
375 |
|
yv12_to_yuyvi = yv12_to_yuyvi_mmx; |
376 |
|
yv12_to_uyvyi = yv12_to_uyvyi_mmx; |
377 |
|
|
378 |
/* Motion estimation related functions */ |
/* Motion estimation related functions */ |
379 |
calc_cbp = calc_cbp_mmx; |
calc_cbp = calc_cbp_mmx; |
380 |
sad16 = sad16_mmx; |
sad16 = sad16_mmx; |
381 |
sad8 = sad8_mmx; |
sad8 = sad8_mmx; |
382 |
|
sad16bi = sad16bi_mmx; |
383 |
|
sad8bi = sad8bi_mmx; |
384 |
dev16 = dev16_mmx; |
dev16 = dev16_mmx; |
385 |
|
sad16v = sad16v_mmx; |
386 |
|
|
387 |
|
} |
388 |
|
|
389 |
|
/* these 3dnow functions are faster than mmx, but slower than xmm. */ |
390 |
|
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
391 |
|
|
392 |
|
/* ME functions */ |
393 |
|
sad16bi = sad16bi_3dn; |
394 |
|
sad8bi = sad8bi_3dn; |
395 |
|
|
396 |
|
yuyv_to_yv12 = yuyv_to_yv12_3dn; |
397 |
|
uyvy_to_yv12 = uyvy_to_yv12_3dn; |
398 |
} |
} |
399 |
|
|
400 |
|
|
401 |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
if ((cpu_flags & XVID_CPU_MMXEXT) > 0) { |
402 |
|
|
403 |
/* Inverse DCT */ |
/* Inverse DCT */ |
408 |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; |
409 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
410 |
|
|
411 |
|
/* reduced resolution */ |
412 |
|
copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; |
413 |
|
add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; |
414 |
|
|
415 |
/* Quantization */ |
/* Quantization */ |
416 |
|
quant4_intra = quant4_intra_xmm; |
417 |
|
quant4_inter = quant4_inter_xmm; |
418 |
|
|
419 |
dequant_intra = dequant_intra_xmm; |
dequant_intra = dequant_intra_xmm; |
420 |
dequant_inter = dequant_inter_xmm; |
dequant_inter = dequant_inter_xmm; |
421 |
|
|
423 |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
transfer_8to16sub2 = transfer_8to16sub2_xmm; |
424 |
|
|
425 |
/* Colorspace transformation */ |
/* Colorspace transformation */ |
426 |
yuv_to_yv12 = yuv_to_yv12_xmm; |
yv12_to_yv12 = yv12_to_yv12_xmm; |
427 |
|
yuyv_to_yv12 = yuyv_to_yv12_xmm; |
428 |
|
uyvy_to_yv12 = uyvy_to_yv12_xmm; |
429 |
|
|
430 |
/* ME functions */ |
/* ME functions */ |
431 |
sad16 = sad16_xmm; |
sad16 = sad16_xmm; |
|
sad16bi = sad16bi_xmm; |
|
432 |
sad8 = sad8_xmm; |
sad8 = sad8_xmm; |
433 |
|
sad16bi = sad16bi_xmm; |
434 |
|
sad8bi = sad8bi_xmm; |
435 |
dev16 = dev16_xmm; |
dev16 = dev16_xmm; |
436 |
|
sad16v = sad16v_xmm; |
437 |
} |
} |
438 |
|
|
439 |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
if ((cpu_flags & XVID_CPU_3DNOW) > 0) { |
444 |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
445 |
} |
} |
446 |
|
|
447 |
|
if ((cpu_flags & XVID_CPU_3DNOWEXT) > 0) { |
448 |
|
|
449 |
|
/* Inverse DCT */ |
450 |
|
idct = idct_3dne; |
451 |
|
|
452 |
|
/* Buffer transfer */ |
453 |
|
transfer_8to16copy = transfer_8to16copy_3dne; |
454 |
|
transfer_16to8copy = transfer_16to8copy_3dne; |
455 |
|
transfer_8to16sub = transfer_8to16sub_3dne; |
456 |
|
transfer_8to16sub2 = transfer_8to16sub2_3dne; |
457 |
|
transfer_16to8add = transfer_16to8add_3dne; |
458 |
|
transfer8x8_copy = transfer8x8_copy_3dne; |
459 |
|
|
460 |
|
/* Quantization */ |
461 |
|
dequant4_intra = dequant4_intra_3dne; |
462 |
|
dequant4_inter = dequant4_inter_3dne; |
463 |
|
quant_intra = quant_intra_3dne; |
464 |
|
quant_inter = quant_inter_3dne; |
465 |
|
dequant_intra = dequant_intra_3dne; |
466 |
|
dequant_inter = dequant_inter_3dne; |
467 |
|
|
468 |
|
/* ME functions */ |
469 |
|
calc_cbp = calc_cbp_3dne; |
470 |
|
sad16 = sad16_3dne; |
471 |
|
sad8 = sad8_3dne; |
472 |
|
sad16bi = sad16bi_3dne; |
473 |
|
sad8bi = sad8bi_3dne; |
474 |
|
dev16 = dev16_3dne; |
475 |
|
|
476 |
|
/* Interpolation */ |
477 |
|
interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; |
478 |
|
interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; |
479 |
|
interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; |
480 |
|
} |
481 |
|
|
482 |
|
|
483 |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
if ((cpu_flags & XVID_CPU_SSE2) > 0) { |
484 |
#ifdef EXPERIMENTAL_SSE2_CODE |
#ifdef EXPERIMENTAL_SSE2_CODE |
485 |
|
|
515 |
sad16bi = sad16bi_ia64; |
sad16bi = sad16bi_ia64; |
516 |
sad8 = sad8_ia64; |
sad8 = sad8_ia64; |
517 |
dev16 = dev16_ia64; |
dev16 = dev16_ia64; |
518 |
Halfpel8_Refine = Halfpel8_Refine_ia64; |
// Halfpel8_Refine = Halfpel8_Refine_ia64; |
519 |
quant_intra = quant_intra_ia64; |
quant_intra = quant_intra_ia64; |
520 |
dequant_intra = dequant_intra_ia64; |
dequant_intra = dequant_intra_ia64; |
521 |
quant_inter = quant_inter_ia64; |
quant_inter = quant_inter_ia64; |
547 |
return XVID_ERR_OK; |
return XVID_ERR_OK; |
548 |
} |
} |
549 |
|
|
550 |
|
|
551 |
|
|
552 |
|
static int |
553 |
|
xvid_init_convert(XVID_INIT_CONVERTINFO* convert) |
554 |
|
{ |
555 |
|
// const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); |
556 |
|
const int width = convert->width; |
557 |
|
const int height = convert->height; |
558 |
|
const int width2 = convert->width/2; |
559 |
|
const int height2 = convert->height/2; |
560 |
|
IMAGE img; |
561 |
|
|
562 |
|
switch (convert->input.colorspace & ~XVID_CSP_VFLIP) |
563 |
|
{ |
564 |
|
case XVID_CSP_YV12 : |
565 |
|
img.y = convert->input.y; |
566 |
|
img.v = (uint8_t*)convert->input.y + width*height; |
567 |
|
img.u = (uint8_t*)convert->input.y + width*height + width2*height2; |
568 |
|
image_output(&img, width, height, width, |
569 |
|
convert->output.y, convert->output.y_stride, |
570 |
|
convert->output.colorspace, convert->interlacing); |
571 |
|
break; |
572 |
|
|
573 |
|
default : |
574 |
|
return XVID_ERR_FORMAT; |
575 |
|
} |
576 |
|
|
577 |
|
|
578 |
|
emms(); |
579 |
|
return XVID_ERR_OK; |
580 |
|
} |
581 |
|
|
582 |
|
|
583 |
|
|
584 |
|
void fill8(uint8_t * block, int size, int value) |
585 |
|
{ |
586 |
|
int i; |
587 |
|
for (i = 0; i < size; i++) |
588 |
|
block[i] = value; |
589 |
|
} |
590 |
|
|
591 |
|
void fill16(int16_t * block, int size, int value) |
592 |
|
{ |
593 |
|
int i; |
594 |
|
for (i = 0; i < size; i++) |
595 |
|
block[i] = value; |
596 |
|
} |
597 |
|
|
598 |
|
#define RANDOM(min,max) min + (rand() % (max-min)) |
599 |
|
|
600 |
|
void random8(uint8_t * block, int size, int min, int max) |
601 |
|
{ |
602 |
|
int i; |
603 |
|
for (i = 0; i < size; i++) |
604 |
|
block[i] = RANDOM(min,max); |
605 |
|
} |
606 |
|
|
607 |
|
void random16(int16_t * block, int size, int min, int max) |
608 |
|
{ |
609 |
|
int i; |
610 |
|
for (i = 0; i < size; i++) |
611 |
|
block[i] = RANDOM(min,max); |
612 |
|
} |
613 |
|
|
614 |
|
int compare16(const int16_t * blockA, const int16_t * blockB, int size) |
615 |
|
{ |
616 |
|
int i; |
617 |
|
for (i = 0; i < size; i++) |
618 |
|
if (blockA[i] != blockB[i]) |
619 |
|
return 1; |
620 |
|
|
621 |
|
return 0; |
622 |
|
} |
623 |
|
|
624 |
|
|
625 |
|
|
626 |
|
int test_h263_intra(quanth263_intraFunc * funcA, quanth263_intraFunc * funcB, |
627 |
|
const char * nameA, const char * nameB, |
628 |
|
int min, int max) |
629 |
|
{ |
630 |
|
int q,i; |
631 |
|
int64_t timeSTART; |
632 |
|
int64_t timeA = 0; |
633 |
|
int64_t timeB = 0; |
634 |
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
635 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
636 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
637 |
|
|
638 |
|
for (q = 1; q <= 31; q++) /* quantizer */ |
639 |
|
{ |
640 |
|
for (i = min; i < max; i++) /* input coeff */ |
641 |
|
{ |
642 |
|
fill16(arrayX, 64, i); |
643 |
|
|
644 |
|
timeSTART = read_counter(); |
645 |
|
funcA(arrayA, arrayX, q, q); |
646 |
|
timeA += read_counter() - timeSTART; |
647 |
|
|
648 |
|
timeSTART = read_counter(); |
649 |
|
funcB(arrayB, arrayX, q, q); |
650 |
|
timeB += read_counter() - timeSTART; |
651 |
|
|
652 |
|
if (compare16(arrayA, arrayB, 64)) |
653 |
|
{ |
654 |
|
printf("%s/%s error: q=%i, i=%i\n", nameA?nameA:"?", nameB?nameB:"?", q, i); |
655 |
|
return 0; |
656 |
|
} |
657 |
|
} |
658 |
|
} |
659 |
|
|
660 |
|
if (nameA) printf("%s:\t%I64i\n", nameA, timeA); |
661 |
|
if (nameB) printf("%s:\t%I64i\n", nameB, timeB); |
662 |
|
|
663 |
|
return 0; |
664 |
|
} |
665 |
|
|
666 |
|
int test_h263_inter(quanth263_interFunc * funcA, quanth263_interFunc * funcB, |
667 |
|
const char * nameA, const char * nameB, |
668 |
|
int min, int max) |
669 |
|
{ |
670 |
|
int q,i; |
671 |
|
int64_t timeSTART; |
672 |
|
int64_t timeA = 0; |
673 |
|
int64_t timeB = 0; |
674 |
|
DECLARE_ALIGNED_MATRIX(arrayX, 1, 64, int16_t, CACHE_LINE); |
675 |
|
DECLARE_ALIGNED_MATRIX(arrayA, 1, 64, int16_t, CACHE_LINE); |
676 |
|
DECLARE_ALIGNED_MATRIX(arrayB, 1, 64, int16_t, CACHE_LINE); |
677 |
|
|
678 |
|
for (q = 1; q <= 31; q++) /* quantizer */ |
679 |
|
{ |
680 |
|
for (i = min; i < max; i++) /* input coeff */ |
681 |
|
{ |
682 |
|
fill16(arrayX, 64, i); |
683 |
|
|
684 |
|
timeSTART = read_counter(); |
685 |
|
funcA(arrayA, arrayX, q); |
686 |
|
timeA += read_counter() - timeSTART; |
687 |
|
|
688 |
|
timeSTART = read_counter(); |
689 |
|
funcB(arrayB, arrayX, q); |
690 |
|
timeB += read_counter() - timeSTART; |
691 |
|
|
692 |
|
if (compare16(arrayA, arrayB, 64)) |
693 |
|
{ |
694 |
|
printf("%s/%s error: q=%i, i=%i\n", nameA?nameA:"?", nameB?nameB:"?", q, i); |
695 |
|
return 0; |
696 |
|
} |
697 |
|
} |
698 |
|
} |
699 |
|
|
700 |
|
if (nameA) printf("%s:\t%I64i\n", nameA, timeA); |
701 |
|
if (nameB) printf("%s:\t%I64i\n", nameB, timeB); |
702 |
|
|
703 |
|
return 0; |
704 |
|
} |
705 |
|
|
706 |
|
|
707 |
|
|
708 |
|
int xvid_init_test() |
709 |
|
{ |
710 |
|
int cpu_flags; |
711 |
|
|
712 |
|
printf("xvid_init_test\n"); |
713 |
|
|
714 |
|
#if defined(ARCH_X86) |
715 |
|
cpu_flags = check_cpu_features(); |
716 |
|
|
717 |
|
emms_mmx(); |
718 |
|
|
719 |
|
printf("--- quant intra ---\n"); |
720 |
|
if (cpu_flags & XVID_CPU_MMX) |
721 |
|
test_h263_intra(quant_intra_c, quant_intra_mmx, "c", "mmx", -2048, 2047); |
722 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
723 |
|
test_h263_intra(quant_intra_c, quant_intra_3dne, NULL, "3dne", -2048, 2047); |
724 |
|
if (cpu_flags & XVID_CPU_SSE2) |
725 |
|
test_h263_intra(quant_intra_c, quant_intra_sse2, NULL, "sse2", -2048, 2047); |
726 |
|
|
727 |
|
printf("\n--- quant inter ---\n"); |
728 |
|
if (cpu_flags & XVID_CPU_MMX) |
729 |
|
test_h263_inter(quant_inter_c, quant_inter_mmx, "c", "mmx", -2048, 2047); |
730 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
731 |
|
test_h263_inter(quant_inter_c, quant_inter_3dne, NULL, "3dne", -2048, 2047); |
732 |
|
if (cpu_flags & XVID_CPU_SSE2) |
733 |
|
test_h263_inter(quant_inter_c, quant_inter_sse2, NULL, "sse2", -2048, 2047); |
734 |
|
|
735 |
|
printf("\n--- dequan intra ---\n"); |
736 |
|
if (cpu_flags & XVID_CPU_MMX) |
737 |
|
test_h263_intra(dequant_intra_c, dequant_intra_mmx, "c", "mmx", -256, 255); |
738 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
739 |
|
test_h263_intra(dequant_intra_c, dequant_intra_xmm, NULL, "xmm", -256, 255); |
740 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
741 |
|
test_h263_intra(dequant_intra_c, dequant_intra_3dne, NULL, "3dne", -256, 255); |
742 |
|
if (cpu_flags & XVID_CPU_SSE2) |
743 |
|
test_h263_intra(dequant_intra_c, dequant_intra_sse2, NULL, "sse2", -256, 255); |
744 |
|
|
745 |
|
printf("\n--- dequant inter ---\n"); |
746 |
|
if (cpu_flags & XVID_CPU_MMX) |
747 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
748 |
|
(quanth263_interFunc*)dequant_inter_mmx, "c", "mmx", -256, 255); |
749 |
|
|
750 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
751 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
752 |
|
(quanth263_interFunc*)dequant_inter_xmm, NULL, "xmm", -256, 255); |
753 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
754 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
755 |
|
(quanth263_interFunc*)dequant_inter_3dne, NULL, "3dne", -256, 255); |
756 |
|
if (cpu_flags & XVID_CPU_SSE2) |
757 |
|
test_h263_inter((quanth263_interFunc*)dequant_inter_c, |
758 |
|
(quanth263_interFunc*)dequant_inter_sse2, NULL, "sse2", -256, 255); |
759 |
|
|
760 |
|
printf("\n--- quant4_intra ---\n"); |
761 |
|
if (cpu_flags & XVID_CPU_MMX) |
762 |
|
test_h263_intra((quanth263_intraFunc*)quant4_intra_c, |
763 |
|
(quanth263_intraFunc*)quant4_intra_mmx, "c", "mmx", -2048, 2047); |
764 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
765 |
|
test_h263_intra((quanth263_intraFunc*)quant4_intra_c, |
766 |
|
(quanth263_intraFunc*)quant4_intra_xmm, NULL, "xmm", -2048, 2047); |
767 |
|
|
768 |
|
printf("\n--- quant4_inter ---\n"); |
769 |
|
if (cpu_flags & XVID_CPU_MMX) |
770 |
|
test_h263_inter((quanth263_interFunc*)quant4_inter_c, |
771 |
|
(quanth263_interFunc*)quant4_inter_mmx, "c", "mmx", -2048, 2047); |
772 |
|
if (cpu_flags & XVID_CPU_MMXEXT) |
773 |
|
test_h263_inter((quanth263_interFunc*)quant4_inter_c, |
774 |
|
(quanth263_interFunc*)quant4_inter_xmm, NULL, "xmm", -2048, 2047); |
775 |
|
|
776 |
|
|
777 |
|
printf("\n--- dequant4_intra ---\n"); |
778 |
|
if (cpu_flags & XVID_CPU_MMX) |
779 |
|
test_h263_intra((quanth263_intraFunc*)dequant4_intra_c, |
780 |
|
(quanth263_intraFunc*)dequant4_intra_mmx, "c", "mmx", -256, 255); |
781 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
782 |
|
test_h263_intra((quanth263_intraFunc*)dequant4_intra_c, |
783 |
|
(quanth263_intraFunc*)dequant4_intra_3dne, NULL, "sse2", -256, 255); |
784 |
|
|
785 |
|
printf("\n--- dequant4_inter ---\n"); |
786 |
|
if (cpu_flags & XVID_CPU_MMX) |
787 |
|
test_h263_inter((quanth263_interFunc*)dequant4_inter_c, |
788 |
|
(quanth263_interFunc*)dequant4_inter_mmx, "c", "mmx", -256, 255); |
789 |
|
if (cpu_flags & XVID_CPU_3DNOWEXT) |
790 |
|
test_h263_inter((quanth263_interFunc*)dequant4_inter_c, |
791 |
|
(quanth263_interFunc*)dequant4_inter_3dne, NULL, "sse2", -256, 255); |
792 |
|
|
793 |
|
emms_mmx(); |
794 |
|
|
795 |
|
#endif |
796 |
|
|
797 |
|
return XVID_ERR_OK; |
798 |
|
} |
799 |
|
|
800 |
|
|
801 |
|
int |
802 |
|
xvid_init(void *handle, |
803 |
|
int opt, |
804 |
|
void *param1, |
805 |
|
void *param2) |
806 |
|
{ |
807 |
|
switch(opt) |
808 |
|
{ |
809 |
|
case XVID_INIT_INIT : |
810 |
|
return xvid_init_init((XVID_INIT_PARAM*)param1); |
811 |
|
|
812 |
|
case XVID_INIT_CONVERT : |
813 |
|
return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1); |
814 |
|
|
815 |
|
case XVID_INIT_TEST : |
816 |
|
return xvid_init_test(); |
817 |
|
|
818 |
|
default : |
819 |
|
return XVID_ERR_FAIL; |
820 |
|
} |
821 |
|
} |
822 |
|
|
823 |
/***************************************************************************** |
/***************************************************************************** |
824 |
* XviD Native decoder entry point |
* XviD Native decoder entry point |
825 |
* |
* |
838 |
{ |
{ |
839 |
switch (opt) { |
switch (opt) { |
840 |
case XVID_DEC_DECODE: |
case XVID_DEC_DECODE: |
841 |
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1); |
return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2); |
842 |
|
|
843 |
case XVID_DEC_CREATE: |
case XVID_DEC_CREATE: |
844 |
return decoder_create((XVID_DEC_PARAM *) param1); |
return decoder_create((XVID_DEC_PARAM *) param1); |
870 |
{ |
{ |
871 |
switch (opt) { |
switch (opt) { |
872 |
case XVID_ENC_ENCODE: |
case XVID_ENC_ENCODE: |
873 |
#ifdef BFRAMES |
|
874 |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
if (((Encoder *) handle)->mbParam.max_bframes >= 0) |
875 |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
876 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
877 |
else |
else |
|
#endif |
|
878 |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1, |
879 |
(XVID_ENC_STATS *) param2); |
(XVID_ENC_STATS *) param2); |
880 |
|
|