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

Annotation of /xvidcore/src/xvid.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.36 - (view) (download)

1 : edgomez 1.16 /*****************************************************************************
2 : edgomez 1.17 *
3 :     * XVID MPEG-4 VIDEO CODEC
4 :     * - Native API implementation -
5 :     *
6 : edgomez 1.34 * Copyright(C) 2001-2002 Peter Ross <pross@cs.rmit.edu.au>
7 :     *
8 : edgomez 1.17 * This program is an implementation of a part of one or more MPEG-4
9 :     * Video tools as specified in ISO/IEC 14496-2 standard. Those intending
10 :     * to use this software module in hardware or software products are
11 :     * advised that its use may infringe existing patents or copyrights, and
12 :     * any such use would be at such party's own risk. The original
13 :     * developer of this software module and his/her company, and subsequent
14 :     * editors and their companies, will have no liability for use of this
15 :     * software or modifications or derivatives thereof.
16 :     *
17 :     * This program is free software ; you can redistribute it and/or modify
18 :     * it under the terms of the GNU General Public License as published by
19 :     * the Free Software Foundation ; either version 2 of the License, or
20 :     * (at your option) any later version.
21 :     *
22 :     * This program is distributed in the hope that it will be useful,
23 :     * but WITHOUT ANY WARRANTY ; without even the implied warranty of
24 :     * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25 :     * GNU General Public License for more details.
26 :     *
27 :     * You should have received a copy of the GNU General Public License
28 :     * along with this program ; if not, write to the Free Software
29 :     * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
30 :     *
31 : edgomez 1.36 * $Id$
32 :     *
33 : edgomez 1.17 ****************************************************************************/
34 : chenm001 1.29
35 : Isibaar 1.1 #include "xvid.h"
36 :     #include "decoder.h"
37 :     #include "encoder.h"
38 :     #include "bitstream/cbp.h"
39 :     #include "dct/idct.h"
40 :     #include "dct/fdct.h"
41 :     #include "image/colorspace.h"
42 :     #include "image/interpolate8x8.h"
43 :     #include "utils/mem_transfer.h"
44 :     #include "quant/quant_h263.h"
45 :     #include "quant/quant_mpeg4.h"
46 : ia64p 1.30 #include "motion/motion.h"
47 : Isibaar 1.1 #include "motion/sad.h"
48 :     #include "utils/emms.h"
49 :     #include "utils/timer.h"
50 : Isibaar 1.9 #include "bitstream/mbcoding.h"
51 : Isibaar 1.1
52 : suxen_drol 1.31 #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
53 :    
54 :     #ifdef WIN32
55 :     #include <windows.h>
56 :     #else
57 :     #include <signal.h>
58 :     #include <setjmp.h>
59 :     #endif
60 :    
61 :    
62 :     #ifndef WIN32
63 :    
64 :     static jmp_buf mark;
65 :    
66 :     static void
67 :     sigill_handler(int signal)
68 :     {
69 :     longjmp(mark, 1);
70 :     }
71 :     #endif
72 :    
73 :    
74 :     /*
75 : edgomez 1.34 * Calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled
76 :     * Return values:
77 :     * -1 : could not determine
78 :     * 0 : SIGILL was *not* signalled
79 :     * 1 : SIGILL was signalled
80 :     */
81 : suxen_drol 1.31
82 :     int
83 :     sigill_check(void (*func)())
84 :     {
85 :     #ifdef WIN32
86 :     _try {
87 :     func();
88 :     }
89 :     _except(EXCEPTION_EXECUTE_HANDLER) {
90 :    
91 :     if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)
92 :     return 1;
93 :     }
94 :     return 0;
95 :     #else
96 :     void * old_handler;
97 :     int jmpret;
98 :    
99 :    
100 :     old_handler = signal(SIGILL, sigill_handler);
101 :     if (old_handler == SIG_ERR)
102 :     {
103 :     return -1;
104 :     }
105 :    
106 :     jmpret = setjmp(mark);
107 :     if (jmpret == 0)
108 :     {
109 :     func();
110 :     }
111 :    
112 :     signal(SIGILL, old_handler);
113 :    
114 :     return jmpret;
115 :     #endif
116 :     }
117 :     #endif
118 :    
119 : edgomez 1.16 /*****************************************************************************
120 :     * XviD Init Entry point
121 :     *
122 :     * Well this function initialize all internal function pointers according
123 :     * to the CPU features forced by the library client or autodetected (depending
124 :     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
125 :     * image colorspace transformation tables.
126 :     *
127 :     * Returned value : XVID_ERR_OK
128 :     * + API_VERSION in the input XVID_INIT_PARAM structure
129 :     * + core build " " " " "
130 :     *
131 :     ****************************************************************************/
132 :    
133 : edgomez 1.15 int
134 :     xvid_init(void *handle,
135 :     int opt,
136 :     void *param1,
137 :     void *param2)
138 : Isibaar 1.1 {
139 :     int cpu_flags;
140 :     XVID_INIT_PARAM *init_param;
141 :    
142 :     init_param = (XVID_INIT_PARAM *) param1;
143 :    
144 : suxen_drol 1.21 /* Inform the client the API version */
145 :     init_param->api_version = API_VERSION;
146 :    
147 :     /* Inform the client the core build - unused because we're still alpha */
148 :     init_param->core_build = 1000;
149 :    
150 : suxen_drol 1.31 /* Do we have to force CPU features ? */
151 :     if ((init_param->cpu_flags & XVID_CPU_FORCE)) {
152 : suxen_drol 1.21
153 : Isibaar 1.1 cpu_flags = init_param->cpu_flags;
154 : suxen_drol 1.31
155 : edgomez 1.16 } else {
156 : Isibaar 1.1
157 : chenm001 1.29 cpu_flags = check_cpu_features();
158 : suxen_drol 1.31
159 :     #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
160 :     if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))
161 :     cpu_flags &= ~XVID_CPU_SSE;
162 :    
163 :     if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger))
164 :     cpu_flags &= ~XVID_CPU_SSE2;
165 :     #endif
166 :     }
167 :    
168 :     if ((init_param->cpu_flags & XVID_CPU_CHKONLY))
169 :     {
170 : Isibaar 1.1 init_param->cpu_flags = cpu_flags;
171 : suxen_drol 1.31 return XVID_ERR_OK;
172 : Isibaar 1.1 }
173 : suxen_drol 1.31
174 :     init_param->cpu_flags = cpu_flags;
175 :    
176 : Isibaar 1.1
177 : edgomez 1.16 /* Initialize the function pointers */
178 : Isibaar 1.1 idct_int32_init();
179 : Isibaar 1.9 init_vlc_tables();
180 :    
181 : edgomez 1.16 /* Fixed Point Forward/Inverse DCT transformations */
182 : Isibaar 1.1 fdct = fdct_int32;
183 :     idct = idct_int32;
184 :    
185 : edgomez 1.16 /* Only needed on PPC Altivec archs */
186 : canard 1.10 sadInit = 0;
187 : edgomez 1.15
188 : edgomez 1.16 /* Restore FPU context : emms_c is a nop functions */
189 : Isibaar 1.1 emms = emms_c;
190 :    
191 : edgomez 1.16 /* Quantization functions */
192 :     quant_intra = quant_intra_c;
193 : Isibaar 1.1 dequant_intra = dequant_intra_c;
194 : edgomez 1.16 quant_inter = quant_inter_c;
195 : Isibaar 1.1 dequant_inter = dequant_inter_c;
196 :    
197 : edgomez 1.16 quant4_intra = quant4_intra_c;
198 : Isibaar 1.1 dequant4_intra = dequant4_intra_c;
199 : edgomez 1.16 quant4_inter = quant4_inter_c;
200 : Isibaar 1.1 dequant4_inter = dequant4_inter_c;
201 :    
202 : edgomez 1.16 /* Block transfer related functions */
203 : Isibaar 1.1 transfer_8to16copy = transfer_8to16copy_c;
204 :     transfer_16to8copy = transfer_16to8copy_c;
205 : edgomez 1.16 transfer_8to16sub = transfer_8to16sub_c;
206 : suxen_drol 1.11 transfer_8to16sub2 = transfer_8to16sub2_c;
207 : edgomez 1.16 transfer_16to8add = transfer_16to8add_c;
208 :     transfer8x8_copy = transfer8x8_copy_c;
209 : Isibaar 1.1
210 : edgomez 1.16 /* Image interpolation related functions */
211 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c;
212 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c;
213 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
214 :    
215 : edgomez 1.16 /* Initialize internal colorspace transformation tables */
216 : Isibaar 1.1 colorspace_init();
217 :    
218 : edgomez 1.16 /* All colorspace transformation functions User Format->YV12 */
219 : Isibaar 1.1 rgb555_to_yv12 = rgb555_to_yv12_c;
220 :     rgb565_to_yv12 = rgb565_to_yv12_c;
221 : edgomez 1.16 rgb24_to_yv12 = rgb24_to_yv12_c;
222 :     rgb32_to_yv12 = rgb32_to_yv12_c;
223 :     yuv_to_yv12 = yuv_to_yv12_c;
224 :     yuyv_to_yv12 = yuyv_to_yv12_c;
225 :     uyvy_to_yv12 = uyvy_to_yv12_c;
226 : Isibaar 1.1
227 : edgomez 1.16 /* All colorspace transformation functions YV12->User format */
228 : Isibaar 1.1 yv12_to_rgb555 = yv12_to_rgb555_c;
229 :     yv12_to_rgb565 = yv12_to_rgb565_c;
230 : edgomez 1.16 yv12_to_rgb24 = yv12_to_rgb24_c;
231 :     yv12_to_rgb32 = yv12_to_rgb32_c;
232 :     yv12_to_yuv = yv12_to_yuv_c;
233 :     yv12_to_yuyv = yv12_to_yuyv_c;
234 :     yv12_to_uyvy = yv12_to_uyvy_c;
235 : Isibaar 1.1
236 : edgomez 1.16 /* Functions used in motion estimation algorithms */
237 : Isibaar 1.1 calc_cbp = calc_cbp_c;
238 : edgomez 1.16 sad16 = sad16_c;
239 : suxen_drol 1.33 sad8 = sad8_c;
240 : edgomez 1.16 sad16bi = sad16bi_c;
241 : suxen_drol 1.33 sad8bi = sad8bi_c;
242 : edgomez 1.16 dev16 = dev16_c;
243 : suxen_drol 1.33
244 : ia64p 1.30 Halfpel8_Refine = Halfpel8_Refine_c;
245 : Isibaar 1.1
246 :     #ifdef ARCH_X86
247 : edgomez 1.15 if ((cpu_flags & XVID_CPU_MMX) > 0) {
248 : edgomez 1.16
249 :     /* Forward and Inverse Discrete Cosine Transformation functions */
250 : Isibaar 1.1 fdct = fdct_mmx;
251 :     idct = idct_mmx;
252 :    
253 : edgomez 1.16 /* To restore FPU context after mmx use */
254 : Isibaar 1.1 emms = emms_mmx;
255 :    
256 : edgomez 1.16 /* Quantization related functions */
257 :     quant_intra = quant_intra_mmx;
258 : Isibaar 1.1 dequant_intra = dequant_intra_mmx;
259 : edgomez 1.16 quant_inter = quant_inter_mmx;
260 : Isibaar 1.1 dequant_inter = dequant_inter_mmx;
261 :    
262 : edgomez 1.16 quant4_intra = quant4_intra_mmx;
263 : Isibaar 1.1 dequant4_intra = dequant4_intra_mmx;
264 : edgomez 1.16 quant4_inter = quant4_inter_mmx;
265 : Isibaar 1.1 dequant4_inter = dequant4_inter_mmx;
266 :    
267 : edgomez 1.16 /* Block related functions */
268 : Isibaar 1.1 transfer_8to16copy = transfer_8to16copy_mmx;
269 :     transfer_16to8copy = transfer_16to8copy_mmx;
270 : edgomez 1.16 transfer_8to16sub = transfer_8to16sub_mmx;
271 : edgomez 1.22 transfer_8to16sub2 = transfer_8to16sub2_mmx;
272 : edgomez 1.16 transfer_16to8add = transfer_16to8add_mmx;
273 :     transfer8x8_copy = transfer8x8_copy_mmx;
274 : edgomez 1.22
275 : edgomez 1.16
276 :     /* Image Interpolation related functions */
277 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx;
278 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx;
279 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
280 :    
281 : edgomez 1.16 /* Image RGB->YV12 related functions */
282 : Isibaar 1.1 rgb24_to_yv12 = rgb24_to_yv12_mmx;
283 :     rgb32_to_yv12 = rgb32_to_yv12_mmx;
284 : edgomez 1.16 yuv_to_yv12 = yuv_to_yv12_mmx;
285 :     yuyv_to_yv12 = yuyv_to_yv12_mmx;
286 :     uyvy_to_yv12 = uyvy_to_yv12_mmx;
287 : Isibaar 1.1
288 : edgomez 1.16 /* Image YV12->RGB related functions */
289 : Isibaar 1.1 yv12_to_rgb24 = yv12_to_rgb24_mmx;
290 :     yv12_to_rgb32 = yv12_to_rgb32_mmx;
291 : edgomez 1.16 yv12_to_yuyv = yv12_to_yuyv_mmx;
292 :     yv12_to_uyvy = yv12_to_uyvy_mmx;
293 : Isibaar 1.1
294 : edgomez 1.16 /* Motion estimation related functions */
295 : Isibaar 1.1 calc_cbp = calc_cbp_mmx;
296 : edgomez 1.16 sad16 = sad16_mmx;
297 :     sad8 = sad8_mmx;
298 : suxen_drol 1.33 sad16bi = sad16bi_mmx;
299 :     sad8bi = sad8bi_mmx;
300 : edgomez 1.16 dev16 = dev16_mmx;
301 : Isibaar 1.1
302 :     }
303 :    
304 : suxen_drol 1.33 /* these 3dnow functions are faster than mmx, but slower than xmm. */
305 :     if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
306 :    
307 :     /* ME functions */
308 :     sad16bi = sad16bi_3dn;
309 :     sad8bi = sad8bi_3dn;
310 :     }
311 :    
312 :    
313 : edgomez 1.15 if ((cpu_flags & XVID_CPU_MMXEXT) > 0) {
314 : edgomez 1.16
315 :     /* Inverse DCT */
316 : Isibaar 1.1 idct = idct_xmm;
317 : edgomez 1.16
318 :     /* Interpolation */
319 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm;
320 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm;
321 : h 1.3 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;
322 : Isibaar 1.25
323 : chenm001 1.29 /* Quantization */
324 :     dequant_intra = dequant_intra_xmm;
325 :     dequant_inter = dequant_inter_xmm;
326 :    
327 : edgomez 1.19 /* Buffer transfer */
328 :     transfer_8to16sub2 = transfer_8to16sub2_xmm;
329 : edgomez 1.16
330 :     /* Colorspace transformation */
331 : Isibaar 1.1 yuv_to_yv12 = yuv_to_yv12_xmm;
332 :    
333 : edgomez 1.16 /* ME functions */
334 : Isibaar 1.1 sad16 = sad16_xmm;
335 : suxen_drol 1.33 sad8 = sad8_xmm;
336 : chenm001 1.29 sad16bi = sad16bi_xmm;
337 : suxen_drol 1.33 sad8bi = sad8bi_xmm;
338 : Isibaar 1.1 dev16 = dev16_xmm;
339 :    
340 :     }
341 :    
342 : edgomez 1.15 if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
343 : edgomez 1.16
344 :     /* Interpolation */
345 : Isibaar 1.1 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
346 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
347 : h 1.4 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;
348 : Isibaar 1.1 }
349 :    
350 : edgomez 1.15 if ((cpu_flags & XVID_CPU_SSE2) > 0) {
351 : Isibaar 1.14 #ifdef EXPERIMENTAL_SSE2_CODE
352 : edgomez 1.16
353 : chenm001 1.29 calc_cbp = calc_cbp_sse2;
354 :    
355 : edgomez 1.16 /* Quantization */
356 :     quant_intra = quant_intra_sse2;
357 : Isibaar 1.14 dequant_intra = dequant_intra_sse2;
358 : edgomez 1.16 quant_inter = quant_inter_sse2;
359 : Isibaar 1.14 dequant_inter = dequant_inter_sse2;
360 : h 1.13
361 : edgomez 1.16 /* ME */
362 :     sad16 = sad16_sse2;
363 :     dev16 = dev16_sse2;
364 :    
365 :     /* Forward and Inverse DCT */
366 :     idct = idct_sse2;
367 : Isibaar 1.14 fdct = fdct_sse2;
368 :     #endif
369 : h 1.12 }
370 : edgomez 1.16
371 : Isibaar 1.1 #endif
372 : Isibaar 1.18
373 :     #ifdef ARCH_IA64
374 :     if ((cpu_flags & XVID_CPU_IA64) > 0) { //use assembler routines?
375 :     idct_ia64_init();
376 :     fdct = fdct_ia64;
377 :     idct = idct_ia64; //not yet working, crashes
378 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
379 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
380 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
381 :     sad16 = sad16_ia64;
382 :     sad16bi = sad16bi_ia64;
383 :     sad8 = sad8_ia64;
384 :     dev16 = dev16_ia64;
385 : ia64p 1.30 Halfpel8_Refine = Halfpel8_Refine_ia64;
386 : Isibaar 1.18 quant_intra = quant_intra_ia64;
387 :     dequant_intra = dequant_intra_ia64;
388 :     quant_inter = quant_inter_ia64;
389 :     dequant_inter = dequant_inter_ia64;
390 :     transfer_8to16copy = transfer_8to16copy_ia64;
391 :     transfer_16to8copy = transfer_16to8copy_ia64;
392 :     transfer_8to16sub = transfer_8to16sub_ia64;
393 :     transfer_8to16sub2 = transfer_8to16sub2_ia64;
394 :     transfer_16to8add = transfer_16to8add_ia64;
395 :     transfer8x8_copy = transfer8x8_copy_ia64;
396 :     DEBUG("Using IA-64 assembler routines.\n");
397 :     }
398 :     #endif
399 : edgomez 1.16
400 : canard 1.5 #ifdef ARCH_PPC
401 : canard 1.6 #ifdef ARCH_PPC_ALTIVEC
402 :     calc_cbp = calc_cbp_altivec;
403 : canard 1.7 fdct = fdct_altivec;
404 :     idct = idct_altivec;
405 : canard 1.10 sadInit = sadInit_altivec;
406 : canard 1.8 sad16 = sad16_altivec;
407 :     sad8 = sad8_altivec;
408 :     dev16 = dev16_altivec;
409 : canard 1.6 #else
410 : canard 1.5 calc_cbp = calc_cbp_ppc;
411 : canard 1.6 #endif
412 : canard 1.5 #endif
413 : Isibaar 1.1
414 :     return XVID_ERR_OK;
415 :     }
416 :    
417 : edgomez 1.16 /*****************************************************************************
418 :     * XviD Native decoder entry point
419 :     *
420 :     * This function is just a wrapper to all the option cases.
421 :     *
422 :     * Returned values : XVID_ERR_FAIL when opt is invalid
423 :     * else returns the wrapped function result
424 :     *
425 :     ****************************************************************************/
426 :    
427 : edgomez 1.15 int
428 :     xvid_decore(void *handle,
429 :     int opt,
430 :     void *param1,
431 :     void *param2)
432 : Isibaar 1.1 {
433 : edgomez 1.15 switch (opt) {
434 :     case XVID_DEC_DECODE:
435 :     return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1);
436 :    
437 :     case XVID_DEC_CREATE:
438 : chenm001 1.29 return decoder_create((XVID_DEC_PARAM *) param1);
439 : edgomez 1.15
440 :     case XVID_DEC_DESTROY:
441 :     return decoder_destroy((DECODER *) handle);
442 : Isibaar 1.1
443 :     default:
444 : edgomez 1.15 return XVID_ERR_FAIL;
445 :     }
446 : Isibaar 1.1 }
447 :    
448 : edgomez 1.16
449 :     /*****************************************************************************
450 :     * XviD Native encoder entry point
451 :     *
452 :     * This function is just a wrapper to all the option cases.
453 :     *
454 :     * Returned values : XVID_ERR_FAIL when opt is invalid
455 :     * else returns the wrapped function result
456 :     *
457 :     ****************************************************************************/
458 : Isibaar 1.1
459 : edgomez 1.15 int
460 :     xvid_encore(void *handle,
461 :     int opt,
462 :     void *param1,
463 :     void *param2)
464 : Isibaar 1.1 {
465 : edgomez 1.15 switch (opt) {
466 :     case XVID_ENC_ENCODE:
467 :     return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1,
468 :     (XVID_ENC_STATS *) param2);
469 :    
470 :     case XVID_ENC_CREATE:
471 :     return encoder_create((XVID_ENC_PARAM *) param1);
472 :    
473 :     case XVID_ENC_DESTROY:
474 :     return encoder_destroy((Encoder *) handle);
475 : Isibaar 1.1
476 :     default:
477 : edgomez 1.15 return XVID_ERR_FAIL;
478 :     }
479 : Isibaar 1.1 }

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