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

Annotation of /xvidcore/src/xvid.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.33.2.18 - (view) (download)

1 : edgomez 1.16 /*****************************************************************************
2 : edgomez 1.17 *
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 1.29
31 : edgomez 1.16 /*****************************************************************************
32 : edgomez 1.17 *
33 :     * History
34 :     *
35 : suxen_drol 1.21 * - 23.06.2002 added XVID_CPU_CHKONLY
36 : edgomez 1.17 * - 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 : Isibaar 1.33.2.18 * $Id: xvid.c,v 1.33.2.17 2002/12/09 10:47:05 suxen_drol Exp $
41 : edgomez 1.17 *
42 :     ****************************************************************************/
43 : Isibaar 1.1
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 : suxen_drol 1.33.2.15 #include "image/reduced.h"
53 : Isibaar 1.1 #include "utils/mem_transfer.h"
54 : h 1.33.2.5 #include "utils/mbfunctions.h"
55 : Isibaar 1.1 #include "quant/quant_h263.h"
56 :     #include "quant/quant_mpeg4.h"
57 : ia64p 1.30 #include "motion/motion.h"
58 : Isibaar 1.1 #include "motion/sad.h"
59 :     #include "utils/emms.h"
60 :     #include "utils/timer.h"
61 : Isibaar 1.9 #include "bitstream/mbcoding.h"
62 : Isibaar 1.1
63 : suxen_drol 1.31 #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
64 :    
65 :     #ifdef WIN32
66 :     #include <windows.h>
67 :     #else
68 :     #include <signal.h>
69 :     #include <setjmp.h>
70 :     #endif
71 :    
72 :    
73 :     #ifndef WIN32
74 :    
75 :     static jmp_buf mark;
76 :    
77 :     static void
78 :     sigill_handler(int signal)
79 :     {
80 :     longjmp(mark, 1);
81 :     }
82 :     #endif
83 :    
84 :    
85 :     /*
86 :     calls the funcptr, and returns whether SIGILL (illegal instruction) was signalled
87 :     return values:
88 :     -1 : could not determine
89 :     0 : SIGILL was *not* signalled
90 :     1 : SIGILL was signalled
91 :     */
92 :    
93 :     int
94 :     sigill_check(void (*func)())
95 :     {
96 :     #ifdef WIN32
97 :     _try {
98 :     func();
99 :     }
100 :     _except(EXCEPTION_EXECUTE_HANDLER) {
101 :    
102 :     if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)
103 :     return 1;
104 :     }
105 :     return 0;
106 :     #else
107 :     void * old_handler;
108 :     int jmpret;
109 :    
110 :    
111 :     old_handler = signal(SIGILL, sigill_handler);
112 :     if (old_handler == SIG_ERR)
113 :     {
114 :     return -1;
115 :     }
116 :    
117 :     jmpret = setjmp(mark);
118 :     if (jmpret == 0)
119 :     {
120 :     func();
121 :     }
122 :    
123 :     signal(SIGILL, old_handler);
124 :    
125 :     return jmpret;
126 :     #endif
127 :     }
128 :     #endif
129 :    
130 : edgomez 1.16 /*****************************************************************************
131 :     * XviD Init Entry point
132 :     *
133 :     * Well this function initialize all internal function pointers according
134 :     * to the CPU features forced by the library client or autodetected (depending
135 :     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
136 :     * image colorspace transformation tables.
137 :     *
138 :     * Returned value : XVID_ERR_OK
139 :     * + API_VERSION in the input XVID_INIT_PARAM structure
140 :     * + core build " " " " "
141 :     *
142 :     ****************************************************************************/
143 :    
144 : suxen_drol 1.33.2.13
145 :     static
146 :     int xvid_init_init(XVID_INIT_PARAM * init_param)
147 : Isibaar 1.1 {
148 :     int cpu_flags;
149 :    
150 : suxen_drol 1.21 /* Inform the client the API version */
151 :     init_param->api_version = API_VERSION;
152 :    
153 :     /* Inform the client the core build - unused because we're still alpha */
154 :     init_param->core_build = 1000;
155 :    
156 : suxen_drol 1.31 /* Do we have to force CPU features ? */
157 :     if ((init_param->cpu_flags & XVID_CPU_FORCE)) {
158 : suxen_drol 1.21
159 : Isibaar 1.1 cpu_flags = init_param->cpu_flags;
160 : suxen_drol 1.31
161 : edgomez 1.16 } else {
162 : Isibaar 1.1
163 : chenm001 1.29 cpu_flags = check_cpu_features();
164 : suxen_drol 1.31
165 :     #if defined(ARCH_X86) && defined(EXPERIMENTAL_SSE2_CODE)
166 :     if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))
167 :     cpu_flags &= ~XVID_CPU_SSE;
168 :    
169 :     if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger))
170 :     cpu_flags &= ~XVID_CPU_SSE2;
171 :     #endif
172 :     }
173 :    
174 :     if ((init_param->cpu_flags & XVID_CPU_CHKONLY))
175 :     {
176 : Isibaar 1.1 init_param->cpu_flags = cpu_flags;
177 : suxen_drol 1.31 return XVID_ERR_OK;
178 : Isibaar 1.1 }
179 : suxen_drol 1.31
180 :     init_param->cpu_flags = cpu_flags;
181 :    
182 : Isibaar 1.1
183 : edgomez 1.16 /* Initialize the function pointers */
184 : Isibaar 1.1 idct_int32_init();
185 : Isibaar 1.9 init_vlc_tables();
186 :    
187 : edgomez 1.16 /* Fixed Point Forward/Inverse DCT transformations */
188 : Isibaar 1.1 fdct = fdct_int32;
189 :     idct = idct_int32;
190 :    
191 : edgomez 1.16 /* Only needed on PPC Altivec archs */
192 : canard 1.10 sadInit = 0;
193 : edgomez 1.15
194 : edgomez 1.16 /* Restore FPU context : emms_c is a nop functions */
195 : Isibaar 1.1 emms = emms_c;
196 :    
197 : edgomez 1.16 /* Quantization functions */
198 :     quant_intra = quant_intra_c;
199 : Isibaar 1.1 dequant_intra = dequant_intra_c;
200 : edgomez 1.16 quant_inter = quant_inter_c;
201 : Isibaar 1.1 dequant_inter = dequant_inter_c;
202 :    
203 : edgomez 1.16 quant4_intra = quant4_intra_c;
204 : Isibaar 1.1 dequant4_intra = dequant4_intra_c;
205 : edgomez 1.16 quant4_inter = quant4_inter_c;
206 : Isibaar 1.1 dequant4_inter = dequant4_inter_c;
207 :    
208 : edgomez 1.16 /* Block transfer related functions */
209 : Isibaar 1.1 transfer_8to16copy = transfer_8to16copy_c;
210 :     transfer_16to8copy = transfer_16to8copy_c;
211 : edgomez 1.16 transfer_8to16sub = transfer_8to16sub_c;
212 : suxen_drol 1.11 transfer_8to16sub2 = transfer_8to16sub2_c;
213 : edgomez 1.16 transfer_16to8add = transfer_16to8add_c;
214 :     transfer8x8_copy = transfer8x8_copy_c;
215 : Isibaar 1.1
216 : h 1.33.2.5 /* Interlacing functions */
217 :     MBFieldTest = MBFieldTest_c;
218 :    
219 : edgomez 1.16 /* Image interpolation related functions */
220 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c;
221 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c;
222 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
223 : Isibaar 1.33.2.8
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 : Isibaar 1.1
228 : Isibaar 1.33.2.6 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 : suxen_drol 1.33.2.15 /* reduced resoltuion */
239 :    
240 : suxen_drol 1.33.2.17 copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C;
241 :     add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C;
242 : suxen_drol 1.33.2.15 #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 : suxen_drol 1.33.2.17 filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C;
250 :     filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C;
251 : suxen_drol 1.33.2.15
252 : edgomez 1.16 /* Initialize internal colorspace transformation tables */
253 : Isibaar 1.1 colorspace_init();
254 :    
255 : edgomez 1.16 /* All colorspace transformation functions User Format->YV12 */
256 : suxen_drol 1.33.2.11 yv12_to_yv12 = yv12_to_yv12_c;
257 :     rgb555_to_yv12 = rgb555_to_yv12_c;
258 :     rgb565_to_yv12 = rgb565_to_yv12_c;
259 :     bgr_to_yv12 = bgr_to_yv12_c;
260 :     bgra_to_yv12 = bgra_to_yv12_c;
261 :     abgr_to_yv12 = abgr_to_yv12_c;
262 :     rgba_to_yv12 = rgba_to_yv12_c;
263 :     yuyv_to_yv12 = yuyv_to_yv12_c;
264 :     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 : Isibaar 1.1
276 : edgomez 1.16 /* All colorspace transformation functions YV12->User format */
277 : suxen_drol 1.33.2.11 yv12_to_rgb555 = yv12_to_rgb555_c;
278 :     yv12_to_rgb565 = yv12_to_rgb565_c;
279 :     yv12_to_bgr = yv12_to_bgr_c;
280 :     yv12_to_bgra = yv12_to_bgra_c;
281 :     yv12_to_abgr = yv12_to_abgr_c;
282 :     yv12_to_rgba = yv12_to_rgba_c;
283 :     yv12_to_yuyv = yv12_to_yuyv_c;
284 :     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 : Isibaar 1.1
295 : edgomez 1.16 /* Functions used in motion estimation algorithms */
296 : Isibaar 1.1 calc_cbp = calc_cbp_c;
297 : edgomez 1.16 sad16 = sad16_c;
298 : suxen_drol 1.33 sad8 = sad8_c;
299 : edgomez 1.16 sad16bi = sad16bi_c;
300 : suxen_drol 1.33 sad8bi = sad8bi_c;
301 : edgomez 1.16 dev16 = dev16_c;
302 : chl 1.33.2.1 sad16v = sad16v_c;
303 : suxen_drol 1.33
304 : chl 1.33.2.1 // Halfpel8_Refine = Halfpel8_Refine_c;
305 : Isibaar 1.1
306 :     #ifdef ARCH_X86
307 : suxen_drol 1.33.2.11
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 : edgomez 1.15 if ((cpu_flags & XVID_CPU_MMX) > 0) {
317 : edgomez 1.16
318 :     /* Forward and Inverse Discrete Cosine Transformation functions */
319 : Isibaar 1.1 fdct = fdct_mmx;
320 :     idct = idct_mmx;
321 :    
322 : edgomez 1.16 /* Quantization related functions */
323 :     quant_intra = quant_intra_mmx;
324 : Isibaar 1.1 dequant_intra = dequant_intra_mmx;
325 : edgomez 1.16 quant_inter = quant_inter_mmx;
326 : Isibaar 1.1 dequant_inter = dequant_inter_mmx;
327 :    
328 : edgomez 1.16 quant4_intra = quant4_intra_mmx;
329 : Isibaar 1.1 dequant4_intra = dequant4_intra_mmx;
330 : edgomez 1.16 quant4_inter = quant4_inter_mmx;
331 : Isibaar 1.1 dequant4_inter = dequant4_inter_mmx;
332 :    
333 : edgomez 1.16 /* Block related functions */
334 : Isibaar 1.1 transfer_8to16copy = transfer_8to16copy_mmx;
335 :     transfer_16to8copy = transfer_16to8copy_mmx;
336 : edgomez 1.16 transfer_8to16sub = transfer_8to16sub_mmx;
337 : edgomez 1.22 transfer_8to16sub2 = transfer_8to16sub2_mmx;
338 : edgomez 1.16 transfer_16to8add = transfer_16to8add_mmx;
339 :     transfer8x8_copy = transfer8x8_copy_mmx;
340 : edgomez 1.22
341 : h 1.33.2.5 /* Interlacing Functions */
342 :     MBFieldTest = MBFieldTest_mmx;
343 : edgomez 1.16
344 :     /* Image Interpolation related functions */
345 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx;
346 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx;
347 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
348 : Isibaar 1.33.2.6
349 :     interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx;
350 :     interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx;
351 :    
352 : Isibaar 1.33.2.14 interpolate8x8_avg2 = interpolate8x8_avg2_mmx;
353 : Isibaar 1.33.2.6 interpolate8x8_avg4 = interpolate8x8_avg4_mmx;
354 : Isibaar 1.1
355 : suxen_drol 1.33.2.15 /* 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 : suxen_drol 1.33.2.17 filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx;
360 :     filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx;
361 : suxen_drol 1.33.2.15
362 : suxen_drol 1.33.2.11 /* 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 : edgomez 1.16 yuyv_to_yv12 = yuyv_to_yv12_mmx;
367 :     uyvy_to_yv12 = uyvy_to_yv12_mmx;
368 : Isibaar 1.1
369 : suxen_drol 1.33.2.11 /* image output yv12_to_xxx related functions */
370 :     yv12_to_bgr = yv12_to_bgr_mmx;
371 :     yv12_to_bgra = yv12_to_bgra_mmx;
372 : edgomez 1.16 yv12_to_yuyv = yv12_to_yuyv_mmx;
373 :     yv12_to_uyvy = yv12_to_uyvy_mmx;
374 : suxen_drol 1.33.2.11
375 :     yv12_to_yuyvi = yv12_to_yuyvi_mmx;
376 :     yv12_to_uyvyi = yv12_to_uyvyi_mmx;
377 : Isibaar 1.1
378 : edgomez 1.16 /* Motion estimation related functions */
379 : Isibaar 1.1 calc_cbp = calc_cbp_mmx;
380 : edgomez 1.16 sad16 = sad16_mmx;
381 :     sad8 = sad8_mmx;
382 : suxen_drol 1.33 sad16bi = sad16bi_mmx;
383 :     sad8bi = sad8bi_mmx;
384 : edgomez 1.16 dev16 = dev16_mmx;
385 : Isibaar 1.33.2.2 sad16v = sad16v_mmx;
386 : Isibaar 1.1
387 :     }
388 :    
389 : suxen_drol 1.33 /* 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 : suxen_drol 1.33.2.11
396 :     yuyv_to_yv12 = yuyv_to_yv12_3dn;
397 :     uyvy_to_yv12 = uyvy_to_yv12_3dn;
398 : suxen_drol 1.33 }
399 :    
400 :    
401 : edgomez 1.15 if ((cpu_flags & XVID_CPU_MMXEXT) > 0) {
402 : edgomez 1.16
403 :     /* Inverse DCT */
404 : Isibaar 1.1 idct = idct_xmm;
405 : edgomez 1.16
406 :     /* Interpolation */
407 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm;
408 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm;
409 : h 1.3 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;
410 : suxen_drol 1.33.2.15
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 : Isibaar 1.25
415 : chenm001 1.29 /* Quantization */
416 : Isibaar 1.33.2.18 quant4_intra = quant4_intra_xmm;
417 :     quant4_inter = quant4_inter_xmm;
418 :    
419 : chenm001 1.29 dequant_intra = dequant_intra_xmm;
420 :     dequant_inter = dequant_inter_xmm;
421 :    
422 : edgomez 1.19 /* Buffer transfer */
423 :     transfer_8to16sub2 = transfer_8to16sub2_xmm;
424 : edgomez 1.16
425 :     /* Colorspace transformation */
426 : suxen_drol 1.33.2.11 yv12_to_yv12 = yv12_to_yv12_xmm;
427 :     yuyv_to_yv12 = yuyv_to_yv12_xmm;
428 : Isibaar 1.33.2.12 uyvy_to_yv12 = uyvy_to_yv12_xmm;
429 : Isibaar 1.1
430 : edgomez 1.16 /* ME functions */
431 : Isibaar 1.1 sad16 = sad16_xmm;
432 : suxen_drol 1.33 sad8 = sad8_xmm;
433 : chenm001 1.29 sad16bi = sad16bi_xmm;
434 : suxen_drol 1.33 sad8bi = sad8bi_xmm;
435 : Isibaar 1.1 dev16 = dev16_xmm;
436 : chl 1.33.2.1 sad16v = sad16v_xmm;
437 : Isibaar 1.1 }
438 :    
439 : edgomez 1.15 if ((cpu_flags & XVID_CPU_3DNOW) > 0) {
440 : edgomez 1.16
441 :     /* Interpolation */
442 : Isibaar 1.1 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
443 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
444 : h 1.4 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;
445 : Isibaar 1.1 }
446 : Isibaar 1.33.2.18
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 : Isibaar 1.1
483 : edgomez 1.15 if ((cpu_flags & XVID_CPU_SSE2) > 0) {
484 : Isibaar 1.14 #ifdef EXPERIMENTAL_SSE2_CODE
485 : edgomez 1.16
486 : chenm001 1.29 calc_cbp = calc_cbp_sse2;
487 :    
488 : edgomez 1.16 /* Quantization */
489 :     quant_intra = quant_intra_sse2;
490 : Isibaar 1.14 dequant_intra = dequant_intra_sse2;
491 : edgomez 1.16 quant_inter = quant_inter_sse2;
492 : Isibaar 1.14 dequant_inter = dequant_inter_sse2;
493 : h 1.13
494 : edgomez 1.16 /* ME */
495 :     sad16 = sad16_sse2;
496 :     dev16 = dev16_sse2;
497 :    
498 :     /* Forward and Inverse DCT */
499 :     idct = idct_sse2;
500 : Isibaar 1.14 fdct = fdct_sse2;
501 :     #endif
502 : h 1.12 }
503 : edgomez 1.16
504 : Isibaar 1.1 #endif
505 : Isibaar 1.18
506 :     #ifdef ARCH_IA64
507 :     if ((cpu_flags & XVID_CPU_IA64) > 0) { //use assembler routines?
508 :     idct_ia64_init();
509 :     fdct = fdct_ia64;
510 :     idct = idct_ia64; //not yet working, crashes
511 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
512 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
513 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
514 :     sad16 = sad16_ia64;
515 :     sad16bi = sad16bi_ia64;
516 :     sad8 = sad8_ia64;
517 :     dev16 = dev16_ia64;
518 : chl 1.33.2.1 // Halfpel8_Refine = Halfpel8_Refine_ia64;
519 : Isibaar 1.18 quant_intra = quant_intra_ia64;
520 :     dequant_intra = dequant_intra_ia64;
521 :     quant_inter = quant_inter_ia64;
522 :     dequant_inter = dequant_inter_ia64;
523 :     transfer_8to16copy = transfer_8to16copy_ia64;
524 :     transfer_16to8copy = transfer_16to8copy_ia64;
525 :     transfer_8to16sub = transfer_8to16sub_ia64;
526 :     transfer_8to16sub2 = transfer_8to16sub2_ia64;
527 :     transfer_16to8add = transfer_16to8add_ia64;
528 :     transfer8x8_copy = transfer8x8_copy_ia64;
529 :     DEBUG("Using IA-64 assembler routines.\n");
530 :     }
531 :     #endif
532 : edgomez 1.16
533 : canard 1.5 #ifdef ARCH_PPC
534 : canard 1.6 #ifdef ARCH_PPC_ALTIVEC
535 :     calc_cbp = calc_cbp_altivec;
536 : canard 1.7 fdct = fdct_altivec;
537 :     idct = idct_altivec;
538 : canard 1.10 sadInit = sadInit_altivec;
539 : canard 1.8 sad16 = sad16_altivec;
540 :     sad8 = sad8_altivec;
541 :     dev16 = dev16_altivec;
542 : canard 1.6 #else
543 : canard 1.5 calc_cbp = calc_cbp_ppc;
544 : canard 1.6 #endif
545 : canard 1.5 #endif
546 : Isibaar 1.1
547 :     return XVID_ERR_OK;
548 : suxen_drol 1.33.2.13 }
549 :    
550 :    
551 :    
552 :     static int
553 :     xvid_init_convert(XVID_INIT_CONVERTINFO* convert)
554 :     {
555 : suxen_drol 1.33.2.16 // const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);
556 : suxen_drol 1.33.2.13 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 :     int
584 :     xvid_init(void *handle,
585 :     int opt,
586 :     void *param1,
587 :     void *param2)
588 :     {
589 :     switch(opt)
590 :     {
591 :     case XVID_INIT_INIT :
592 :     return xvid_init_init((XVID_INIT_PARAM*)param1);
593 :    
594 :     case XVID_INIT_CONVERT :
595 :     return xvid_init_convert((XVID_INIT_CONVERTINFO*)param1);
596 :    
597 :     default :
598 :     return XVID_ERR_FAIL;
599 :     }
600 : Isibaar 1.1 }
601 :    
602 : edgomez 1.16 /*****************************************************************************
603 :     * XviD Native decoder entry point
604 :     *
605 :     * This function is just a wrapper to all the option cases.
606 :     *
607 :     * Returned values : XVID_ERR_FAIL when opt is invalid
608 :     * else returns the wrapped function result
609 :     *
610 :     ****************************************************************************/
611 :    
612 : edgomez 1.15 int
613 :     xvid_decore(void *handle,
614 :     int opt,
615 :     void *param1,
616 :     void *param2)
617 : Isibaar 1.1 {
618 : edgomez 1.15 switch (opt) {
619 :     case XVID_DEC_DECODE:
620 : suxen_drol 1.33.2.11 return decoder_decode((DECODER *) handle, (XVID_DEC_FRAME *) param1, (XVID_DEC_STATS*) param2);
621 : edgomez 1.15
622 :     case XVID_DEC_CREATE:
623 : chenm001 1.29 return decoder_create((XVID_DEC_PARAM *) param1);
624 : edgomez 1.15
625 :     case XVID_DEC_DESTROY:
626 :     return decoder_destroy((DECODER *) handle);
627 : Isibaar 1.1
628 :     default:
629 : edgomez 1.15 return XVID_ERR_FAIL;
630 :     }
631 : Isibaar 1.1 }
632 :    
633 : edgomez 1.16
634 :     /*****************************************************************************
635 :     * XviD Native encoder entry point
636 :     *
637 :     * This function is just a wrapper to all the option cases.
638 :     *
639 :     * Returned values : XVID_ERR_FAIL when opt is invalid
640 :     * else returns the wrapped function result
641 :     *
642 :     ****************************************************************************/
643 : Isibaar 1.1
644 : edgomez 1.15 int
645 :     xvid_encore(void *handle,
646 :     int opt,
647 :     void *param1,
648 :     void *param2)
649 : Isibaar 1.1 {
650 : edgomez 1.15 switch (opt) {
651 :     case XVID_ENC_ENCODE:
652 : chl 1.33.2.10
653 : suxen_drol 1.20 if (((Encoder *) handle)->mbParam.max_bframes >= 0)
654 :     return encoder_encode_bframes((Encoder *) handle, (XVID_ENC_FRAME *) param1,
655 :     (XVID_ENC_STATS *) param2);
656 :     else
657 : edgomez 1.15 return encoder_encode((Encoder *) handle, (XVID_ENC_FRAME *) param1,
658 :     (XVID_ENC_STATS *) param2);
659 :    
660 :     case XVID_ENC_CREATE:
661 :     return encoder_create((XVID_ENC_PARAM *) param1);
662 :    
663 :     case XVID_ENC_DESTROY:
664 :     return encoder_destroy((Encoder *) handle);
665 : Isibaar 1.1
666 :     default:
667 : edgomez 1.15 return XVID_ERR_FAIL;
668 :     }
669 : Isibaar 1.1 }

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