[svn] / trunk / xvidcore / src / xvid.c Repository:
ViewVC logotype

Annotation of /trunk/xvidcore/src/xvid.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 329 - (view) (download)

1 : edgomez 200 /*****************************************************************************
2 : edgomez 201 *
3 :     * XVID MPEG-4 VIDEO CODEC
4 :     * - Native API implementation -
5 :     *
6 :     * This program is an implementation of a part of one or more MPEG-4
7 :     * Video tools as specified in ISO/IEC 14496-2 standard. Those intending
8 :     * to use this software module in hardware or software products are
9 :     * advised that its use may infringe existing patents or copyrights, and
10 :     * any such use would be at such party's own risk. The original
11 :     * developer of this software module and his/her company, and subsequent
12 :     * editors and their companies, will have no liability for use of this
13 :     * software or modifications or derivatives thereof.
14 :     *
15 :     * This program is free software ; you can redistribute it and/or modify
16 :     * it under the terms of the GNU General Public License as published by
17 :     * the Free Software Foundation ; either version 2 of the License, or
18 :     * (at your option) any later version.
19 :     *
20 :     * This program is distributed in the hope that it will be useful,
21 :     * but WITHOUT ANY WARRANTY ; without even the implied warranty of
22 :     * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
23 :     * GNU General Public License for more details.
24 :     *
25 :     * You should have received a copy of the GNU General Public License
26 :     * along with this program ; if not, write to the Free Software
27 :     * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
28 :     *
29 :     ****************************************************************************/
30 : chenm001 274
31 : edgomez 200 /*****************************************************************************
32 : edgomez 201 *
33 :     * History
34 :     *
35 : suxen_drol 234 * - 23.06.2002 added XVID_CPU_CHKONLY
36 : edgomez 201 * - 17.03.2002 Added interpolate8x8_halfpel_hv_xmm
37 :     * - 22.12.2001 API change: added xvid_init() - Isibaar
38 :     * - 16.12.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au>
39 :     *
40 : suxen_drol 329 * $Id: xvid.c,v 1.33 2002-07-23 12:59:57 suxen_drol Exp $
41 : edgomez 201 *
42 :     ****************************************************************************/
43 : Isibaar 3
44 :     #include "xvid.h"
45 :     #include "decoder.h"
46 :     #include "encoder.h"
47 :     #include "bitstream/cbp.h"
48 :     #include "dct/idct.h"
49 :     #include "dct/fdct.h"
50 :     #include "image/colorspace.h"
51 :     #include "image/interpolate8x8.h"
52 :     #include "utils/mem_transfer.h"
53 :     #include "quant/quant_h263.h"
54 :     #include "quant/quant_mpeg4.h"
55 : ia64p 299 #include "motion/motion.h"
56 : Isibaar 3 #include "motion/sad.h"
57 :     #include "utils/emms.h"
58 :     #include "utils/timer.h"
59 : Isibaar 100 #include "bitstream/mbcoding.h"
60 : Isibaar 3
61 : suxen_drol 311 #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
62 :    
63 :     #ifdef WIN32
64 :     #include <windows.h>
65 :     #else
66 :     #include <signal.h>
67 :     #include <setjmp.h>
68 :     #endif
69 :    
70 :    
71 :     #ifndef WIN32
72 :    
73 :     static jmp_buf mark;
74 :    
75 :     static void
76 :     sigill_handler(int signal)
77 :     {
78 :     longjmp(mark, 1);
79 :     }
80 :     #endif
81 :    
82 :    
83 :     /*
84 :     calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled
85 :     return values:
86 :     -1 : could not determine
87 :     0 : SIGILL was *not* signalled
88 :     1 : SIGILL was signalled
89 :     */
90 :    
91 :     int
92 :     sigill_check(void (*func)())
93 :     {
94 :     #ifdef WIN32
95 :     _try {
96 :     func();
97 :     }
98 :     _except(EXCEPTION_EXECUTE_HANDLER) {
99 :    
100 :     if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)
101 :     return 1;
102 :     }
103 :     return 0;
104 :     #else
105 :     void * old_handler;
106 :     int jmpret;
107 :    
108 :    
109 :     old_handler = signal(SIGILL, sigill_handler);
110 :     if (old_handler == SIG_ERR)
111 :     {
112 :     return -1;
113 :     }
114 :    
115 :     jmpret = setjmp(mark);
116 :     if (jmpret == 0)
117 :     {
118 :     func();
119 :     }
120 :    
121 :     signal(SIGILL, old_handler);
122 :    
123 :     return jmpret;
124 :     #endif
125 :     }
126 :     #endif
127 :    
128 : edgomez 200 /*****************************************************************************
129 :     * XviD Init Entry point
130 :     *
131 :     * Well this function initialize all internal function pointers according
132 :     * to the CPU features forced by the library client or autodetected (depending
133 :     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
134 :     * image colorspace transformation tables.
135 :     *
136 :     * Returned value : XVID_ERR_OK
137 :     * + API_VERSION in the input XVID_INIT_PARAM structure
138 :     * + core build " " " " "
139 :     *
140 :     ****************************************************************************/
141 :    
142 : edgomez 195 int
143 :     xvid_init(void *handle,
144 :     int opt,
145 :     void *param1,
146 :     void *param2)
147 : Isibaar 3 {
148 :     int cpu_flags;
149 :     XVID_INIT_PARAM *init_param;
150 :    
151 :     init_param = (XVID_INIT_PARAM *) param1;
152 :    
153 : suxen_drol 234 /* Inform the client the API version */
154 :     init_param->api_version = API_VERSION;
155 :    
156 :     /* Inform the client the core build - unused because we're still alpha */
157 :     init_param->core_build = 1000;
158 :    
159 : suxen_drol 311 /* Do we have to force CPU features ? */
160 :     if ((init_param->cpu_flags & XVID_CPU_FORCE)) {
161 : suxen_drol 234
162 : Isibaar 3 cpu_flags = init_param->cpu_flags;
163 : suxen_drol 311
164 : edgomez 200 } else {
165 : Isibaar 3
166 : chenm001 274 cpu_flags = check_cpu_features();
167 : suxen_drol 311
168 :     #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
169 :     if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))
170 :     cpu_flags &= ~XVID_CPU_SSE;
171 :    
172 :     if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger))
173 :     cpu_flags &= ~XVID_CPU_SSE2;
174 :     #endif
175 :     }
176 :    
177 :     if ((init_param->cpu_flags & XVID_CPU_CHKONLY))
178 :     {
179 : Isibaar 3 init_param->cpu_flags = cpu_flags;
180 : suxen_drol 311 return XVID_ERR_OK;
181 : Isibaar 3 }
182 :    
183 : suxen_drol 311 init_param->cpu_flags = cpu_flags;
184 :    
185 :    
186 : edgomez 200 /* Initialize the function pointers */
187 : Isibaar 3 idct_int32_init();
188 : Isibaar 100 init_vlc_tables();
189 :    
190 : edgomez 200 /* Fixed Point Forward/Inverse DCT transformations */
191 : Isibaar 3 fdct = fdct_int32;
192 :     idct = idct_int32;
193 :    
194 : edgomez 200 /* Only needed on PPC Altivec archs */
195 : canard 115 sadInit = 0;
196 : edgomez 195
197 : edgomez 200 /* Restore FPU context : emms_c is a nop functions */
198 : Isibaar 3 emms = emms_c;
199 :    
200 : edgomez 200 /* Quantization functions */
201 :     quant_intra = quant_intra_c;
202 : Isibaar 3 dequant_intra = dequant_intra_c;
203 : edgomez 200 quant_inter = quant_inter_c;
204 : Isibaar 3 dequant_inter = dequant_inter_c;
205 :    
206 : edgomez 200 quant4_intra = quant4_intra_c;
207 : Isibaar 3 dequant4_intra = dequant4_intra_c;
208 : edgomez 200 quant4_inter = quant4_inter_c;
209 : Isibaar 3 dequant4_inter = dequant4_inter_c;
210 :    
211 : edgomez 200 /* Block transfer related functions */
212 : Isibaar 3 transfer_8to16copy = transfer_8to16copy_c;
213 :     transfer_16to8copy = transfer_16to8copy_c;
214 : edgomez 200 transfer_8to16sub = transfer_8to16sub_c;
215 : suxen_drol 118 transfer_8to16sub2 = transfer_8to16sub2_c;
216 : edgomez 200 transfer_16to8add = transfer_16to8add_c;
217 :     transfer8x8_copy = transfer8x8_copy_c;
218 : Isibaar 3
219 : edgomez 200 /* Image interpolation related functions */
220 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c;
221 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c;
222 : Isibaar 3 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
223 :    
224 : edgomez 200 /* Initialize internal colorspace transformation tables */
225 : Isibaar 3 colorspace_init();
226 :    
227 : edgomez 200 /* All colorspace transformation functions User Format->YV12 */
228 : Isibaar 3 rgb555_to_yv12 = rgb555_to_yv12_c;
229 :     rgb565_to_yv12 = rgb565_to_yv12_c;
230 : edgomez 200 rgb24_to_yv12 = rgb24_to_yv12_c;
231 :     rgb32_to_yv12 = rgb32_to_yv12_c;
232 :     yuv_to_yv12 = yuv_to_yv12_c;
233 :     yuyv_to_yv12 = yuyv_to_yv12_c;
234 :     uyvy_to_yv12 = uyvy_to_yv12_c;
235 : Isibaar 3
236 : edgomez 200 /* All colorspace transformation functions YV12->User format */
237 : Isibaar 3 yv12_to_rgb555 = yv12_to_rgb555_c;
238 :     yv12_to_rgb565 = yv12_to_rgb565_c;
239 : edgomez 200 yv12_to_rgb24 = yv12_to_rgb24_c;
240 :     yv12_to_rgb32 = yv12_to_rgb32_c;
241 :     yv12_to_yuv = yv12_to_yuv_c;
242 :     yv12_to_yuyv = yv12_to_yuyv_c;
243 :     yv12_to_uyvy = yv12_to_uyvy_c;
244 : Isibaar 3
245 : edgomez 200 /* Functions used in motion estimation algorithms */
246 : Isibaar 3 calc_cbp = calc_cbp_c;
247 : edgomez 200 sad16 = sad16_c;
248 : suxen_drol 329 sad8 = sad8_c;
249 : edgomez 200 sad16bi = sad16bi_c;
250 : suxen_drol 329 sad8bi = sad8bi_c;
251 : edgomez 200 dev16 = dev16_c;
252 : suxen_drol 329
253 : ia64p 299 Halfpel8_Refine = Halfpel8_Refine_c;
254 : Isibaar 3
255 :     #ifdef ARCH_X86
256 : edgomez 195 if ((cpu_flags & XVID_CPU_MMX) > 0) {
257 : edgomez 200
258 :     /* Forward and Inverse Discrete Cosine Transformation functions */
259 : Isibaar 3 fdct = fdct_mmx;
260 :     idct = idct_mmx;
261 :    
262 : edgomez 200 /* To restore FPU context after mmx use */
263 : Isibaar 3 emms = emms_mmx;
264 :    
265 : edgomez 200 /* Quantization related functions */
266 :     quant_intra = quant_intra_mmx;
267 : Isibaar 3 dequant_intra = dequant_intra_mmx;
268 : edgomez 200 quant_inter = quant_inter_mmx;
269 : Isibaar 3 dequant_inter = dequant_inter_mmx;
270 :    
271 : edgomez 200 quant4_intra = quant4_intra_mmx;
272 : Isibaar 3 dequant4_intra = dequant4_intra_mmx;
273 : edgomez 200 quant4_inter = quant4_inter_mmx;
274 : Isibaar 3 dequant4_inter = dequant4_inter_mmx;
275 :    
276 : edgomez 200 /* Block related functions */
277 : Isibaar 3 transfer_8to16copy = transfer_8to16copy_mmx;
278 :     transfer_16to8copy = transfer_16to8copy_mmx;
279 : edgomez 200 transfer_8to16sub = transfer_8to16sub_mmx;
280 : edgomez 236 transfer_8to16sub2 = transfer_8to16sub2_mmx;
281 : edgomez 200 transfer_16to8add = transfer_16to8add_mmx;
282 :     transfer8x8_copy = transfer8x8_copy_mmx;
283 : Isibaar 3
284 : edgomez 236
285 : edgomez 200 /* Image Interpolation related functions */
286 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx;
287 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx;
288 : Isibaar 3 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
289 :    
290 : edgomez 200 /* Image RGB->YV12 related functions */
291 : Isibaar 3 rgb24_to_yv12 = rgb24_to_yv12_mmx;
292 :     rgb32_to_yv12 = rgb32_to_yv12_mmx;
293 : edgomez 200 yuv_to_yv12 = yuv_to_yv12_mmx;
294 :     yuyv_to_yv12 = yuyv_to_yv12_mmx;
295 :     uyvy_to_yv12 = uyvy_to_yv12_mmx;
296 : Isibaar 3
297 : edgomez 200 /* Image YV12->RGB related functions */
298 : Isibaar 3 yv12_to_rgb24 = yv12_to_rgb24_mmx;
299 :     yv12_to_rgb32 = yv12_to_rgb32_mmx;
300 : edgomez 200 yv12_to_yuyv = yv12_to_yuyv_mmx;
301 :     yv12_to_uyvy = yv12_to_uyvy_mmx;
302 : Isibaar 3
303 : edgomez 200 /* Motion estimation related functions */
304 : Isibaar 3 calc_cbp = calc_cbp_mmx;
305 : edgomez 200 sad16 = sad16_mmx;
306 :     sad8 = sad8_mmx;
307 : suxen_drol 329 sad16bi = sad16bi_mmx;
308 :     sad8bi = sad8bi_mmx;
309 : edgomez 200 dev16 = dev16_mmx;
310 : Isibaar 3
311 :     }
312 :    
313 : suxen_drol 329 /* these 3dnow functions are faster than mmx, but slower than xmm. */
314 :     if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
315 :    
316 :     /* ME functions */
317 :     sad16bi = sad16bi_3dn;
318 :     sad8bi = sad8bi_3dn;
319 :     }
320 :    
321 :    
322 : edgomez 195 if ((cpu_flags & XVID_CPU_MMXEXT) > 0) {
323 : edgomez 200
324 :     /* Inverse DCT */
325 : Isibaar 3 idct = idct_xmm;
326 : edgomez 200
327 :     /* Interpolation */
328 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm;
329 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm;
330 : h 38 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;
331 : edgomez 200
332 : chenm001 274 /* Quantization */
333 :     dequant_intra = dequant_intra_xmm;
334 :     dequant_inter = dequant_inter_xmm;
335 :    
336 : edgomez 218 /* Buffer transfer */
337 :     transfer_8to16sub2 = transfer_8to16sub2_xmm;
338 :    
339 : edgomez 200 /* Colorspace transformation */
340 : Isibaar 3 yuv_to_yv12 = yuv_to_yv12_xmm;
341 :    
342 : edgomez 200 /* ME functions */
343 : Isibaar 3 sad16 = sad16_xmm;
344 : suxen_drol 329 sad8 = sad8_xmm;
345 : chenm001 274 sad16bi = sad16bi_xmm;
346 : suxen_drol 329 sad8bi = sad8bi_xmm;
347 : Isibaar 3 dev16 = dev16_xmm;
348 :    
349 :     }
350 :    
351 : edgomez 195 if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
352 : edgomez 200
353 :     /* Interpolation */
354 : Isibaar 3 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
355 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
356 : h 40 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;
357 : Isibaar 3 }
358 :    
359 : edgomez 195 if ((cpu_flags & XVID_CPU_SSE2) > 0) {
360 : Isibaar 154 #ifdef EXPERIMENTAL_SSE2_CODE
361 : edgomez 200
362 : chenm001 274 calc_cbp = calc_cbp_sse2;
363 :    
364 : edgomez 200 /* Quantization */
365 :     quant_intra = quant_intra_sse2;
366 : Isibaar 154 dequant_intra = dequant_intra_sse2;
367 : edgomez 200 quant_inter = quant_inter_sse2;
368 : Isibaar 154 dequant_inter = dequant_inter_sse2;
369 : h 135
370 : edgomez 200 /* ME */
371 :     sad16 = sad16_sse2;
372 :     dev16 = dev16_sse2;
373 :    
374 :     /* Forward and Inverse DCT */
375 :     idct = idct_sse2;
376 : Isibaar 154 fdct = fdct_sse2;
377 :     #endif
378 : h 126 }
379 : edgomez 200
380 : Isibaar 3 #endif
381 : edgomez 200
382 : Isibaar 209 #ifdef ARCH_IA64
383 :     if ((cpu_flags & XVID_CPU_IA64) > 0) { //use assembler routines?
384 :     idct_ia64_init();
385 :     fdct = fdct_ia64;
386 :     idct = idct_ia64; //not yet working, crashes
387 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
388 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
389 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
390 :     sad16 = sad16_ia64;
391 :     sad16bi = sad16bi_ia64;
392 :     sad8 = sad8_ia64;
393 :     dev16 = dev16_ia64;
394 : ia64p 299 Halfpel8_Refine = Halfpel8_Refine_ia64;
395 : Isibaar 209 quant_intra = quant_intra_ia64;
396 :     dequant_intra = dequant_intra_ia64;
397 :     quant_inter = quant_inter_ia64;
398 :     dequant_inter = dequant_inter_ia64;
399 :     transfer_8to16copy = transfer_8to16copy_ia64;
400 :     transfer_16to8copy = transfer_16to8copy_ia64;
401 :     transfer_8to16sub = transfer_8to16sub_ia64;
402 :     transfer_8to16sub2 = transfer_8to16sub2_ia64;
403 :     transfer_16to8add = transfer_16to8add_ia64;
404 :     transfer8x8_copy = transfer8x8_copy_ia64;
405 :     DEBUG("Using IA-64 assembler routines.\n");
406 :     }
407 :     #endif
408 :    
409 : canard 52 #ifdef ARCH_PPC
410 : canard 71 #ifdef ARCH_PPC_ALTIVEC
411 :     calc_cbp = calc_cbp_altivec;
412 : canard 76 fdct = fdct_altivec;
413 :     idct = idct_altivec;
414 : canard 115 sadInit = sadInit_altivec;
415 : canard 89 sad16 = sad16_altivec;
416 :     sad8 = sad8_altivec;
417 :     dev16 = dev16_altivec;
418 : canard 71 #else
419 : canard 52 calc_cbp = calc_cbp_ppc;
420 :     #endif
421 : canard 71 #endif
422 : edgomez 195
423 : Isibaar 3 return XVID_ERR_OK;
424 :     }
425 :    
426 : edgomez 200 /*****************************************************************************
427 :     * XviD Native decoder entry point
428 :     *
429 :     * This function is just a wrapper to all the option cases.
430 :     *
431 :     * Returned values : XVID_ERR_FAIL when opt is invalid
432 :     * else returns the wrapped function result
433 :     *
434 :     ****************************************************************************/
435 :    
436 : edgomez 195 int
437 :     xvid_decore(void *handle,
438 :     int opt,
439 :     void *param1,
440 :     void *param2)
441 : Isibaar 3 {
442 : edgomez 195 switch (opt) {
443 :     case XVID_DEC_DECODE:
444 :     return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1);
445 : Isibaar 3
446 : edgomez 195 case XVID_DEC_CREATE:
447 : chenm001 274 return decoder_create((XVID_DEC_PARAM *) param1);
448 : Isibaar 3
449 : edgomez 195 case XVID_DEC_DESTROY:
450 :     return decoder_destroy((DECODER *) handle);
451 :    
452 : Isibaar 3 default:
453 : edgomez 195 return XVID_ERR_FAIL;
454 :     }
455 : Isibaar 3 }
456 :    
457 :    
458 : edgomez 200 /*****************************************************************************
459 :     * XviD Native encoder entry point
460 :     *
461 :     * This function is just a wrapper to all the option cases.
462 :     *
463 :     * Returned values : XVID_ERR_FAIL when opt is invalid
464 :     * else returns the wrapped function result
465 :     *
466 :     ****************************************************************************/
467 :    
468 : edgomez 195 int
469 :     xvid_encore(void *handle,
470 :     int opt,
471 :     void *param1,
472 :     void *param2)
473 : Isibaar 3 {
474 : edgomez 195 switch (opt) {
475 :     case XVID_ENC_ENCODE:
476 : suxen_drol 232 #ifdef BFRAMES
477 :     if (((Encoder *) handle)->mbParam.max_bframes >= 0)
478 :     return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1,
479 :     (XVID_ENC_STATS *) param2);
480 :     else
481 :     #endif
482 : edgomez 195 return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1,
483 :     (XVID_ENC_STATS *) param2);
484 : Isibaar 3
485 : edgomez 195 case XVID_ENC_CREATE:
486 :     return encoder_create((XVID_ENC_PARAM *) param1);
487 : Isibaar 3
488 : edgomez 195 case XVID_ENC_DESTROY:
489 :     return encoder_destroy((Encoder *) handle);
490 :    
491 : Isibaar 3 default:
492 : edgomez 195 return XVID_ERR_FAIL;
493 :     }
494 : Isibaar 3 }

No admin address has been configured
ViewVC Help
Powered by ViewVC 1.0.4