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

Annotation of /xvidcore/src/xvid.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1.51 - (view) (download)

1 : edgomez 1.16 /*****************************************************************************
2 : edgomez 1.17 *
3 :     * XVID MPEG-4 VIDEO CODEC
4 :     * - Native API implementation -
5 :     *
6 : suxen_drol 1.49 * Copyright(C) 2001-2004 Peter Ross <pross@xvid.org>
7 : edgomez 1.48 *
8 : edgomez 1.41 * This program is free software ; you can redistribute it and/or modify
9 :     * it under the terms of the GNU General Public License as published by
10 :     * the Free Software Foundation ; either version 2 of the License, or
11 : edgomez 1.17 * (at your option) any later version.
12 :     *
13 :     * This program is distributed in the hope that it will be useful,
14 : edgomez 1.41 * but WITHOUT ANY WARRANTY ; without even the implied warranty of
15 : edgomez 1.17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 :     * GNU General Public License for more details.
17 :     *
18 :     * You should have received a copy of the GNU General Public License
19 : edgomez 1.41 * along with this program ; if not, write to the Free Software
20 : edgomez 1.17 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21 :     *
22 : edgomez 1.50 * $Id$
23 : edgomez 1.36 *
24 : edgomez 1.17 ****************************************************************************/
25 : chenm001 1.29
26 : edgomez 1.41 #include <stdio.h>
27 :     #include <stdlib.h>
28 :     #include <string.h>
29 :     #include <time.h>
30 :    
31 : Isibaar 1.1 #include "xvid.h"
32 :     #include "decoder.h"
33 :     #include "encoder.h"
34 :     #include "bitstream/cbp.h"
35 :     #include "dct/idct.h"
36 :     #include "dct/fdct.h"
37 :     #include "image/colorspace.h"
38 :     #include "image/interpolate8x8.h"
39 : edgomez 1.41 #include "image/reduced.h"
40 : Isibaar 1.1 #include "utils/mem_transfer.h"
41 : edgomez 1.41 #include "utils/mbfunctions.h"
42 : edgomez 1.48 #include "quant/quant.h"
43 : ia64p 1.30 #include "motion/motion.h"
44 : Isibaar 1.1 #include "motion/sad.h"
45 :     #include "utils/emms.h"
46 :     #include "utils/timer.h"
47 : Isibaar 1.9 #include "bitstream/mbcoding.h"
48 : edgomez 1.48 #include "image/qpel.h"
49 :     #include "image/postprocessing.h"
50 :    
51 :     #if defined(_DEBUG)
52 :     unsigned int xvid_debug = 0; /* xvid debug mask */
53 :     #endif
54 : Isibaar 1.1
55 : edgomez 1.50 #if defined(ARCH_IS_IA32) && defined(_MSC_VER)
56 : edgomez 1.41 # include <windows.h>
57 : edgomez 1.50 #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC)
58 : edgomez 1.41 # include <signal.h>
59 :     # include <setjmp.h>
60 : suxen_drol 1.31
61 : edgomez 1.41 static jmp_buf mark;
62 : suxen_drol 1.31
63 : edgomez 1.41 static void
64 :     sigill_handler(int signal)
65 :     {
66 :     longjmp(mark, 1);
67 :     }
68 : suxen_drol 1.31 #endif
69 :    
70 :    
71 :     /*
72 : edgomez 1.43 * Calls the funcptr, and returns whether SIGILL (illegal instruction) was
73 :     * signalled
74 :     *
75 :     * Return values:
76 :     * -1 : could not determine
77 :     * 0 : SIGILL was *not* signalled
78 :     * 1 : SIGILL was signalled
79 :     */
80 : edgomez 1.50 #if defined(ARCH_IS_IA32) && defined(_MSC_VER)
81 :     static int
82 : suxen_drol 1.31 sigill_check(void (*func)())
83 :     {
84 :     _try {
85 :     func();
86 : edgomez 1.50 } _except(EXCEPTION_EXECUTE_HANDLER) {
87 : suxen_drol 1.31
88 :     if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION)
89 : edgomez 1.50 return(1);
90 : suxen_drol 1.31 }
91 : edgomez 1.50 return(0);
92 :     }
93 :     #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC)
94 :     static int
95 :     sigill_check(void (*func)())
96 :     {
97 :     void *old_handler;
98 : suxen_drol 1.31 int jmpret;
99 :    
100 : edgomez 1.50 /* Set our SIGILL handler */
101 :     old_handler = signal(SIGILL, sigill_handler);
102 : suxen_drol 1.31
103 : edgomez 1.50 /* Check for error */
104 :     if (old_handler == SIG_ERR) {
105 :     return(-1);
106 : suxen_drol 1.31 }
107 :    
108 : edgomez 1.50 /* Save stack context, so if func triggers a SIGILL, we can still roll
109 :     * back to a valid CPU state */
110 :     jmpret = setjmp(mark);
111 :    
112 :     /* If setjmp returned directly, then its returned value is 0, and we still
113 :     * have to test the passed func. Otherwise it means the stack context has
114 :     * been restored by a longjmp() call, which in our case happens only in the
115 :     * signal handler */
116 :     if (jmpret == 0) {
117 : suxen_drol 1.31 func();
118 :     }
119 :    
120 : edgomez 1.50 /* Restore old signal handler */
121 : suxen_drol 1.31 signal(SIGILL, old_handler);
122 :    
123 : edgomez 1.50 return(jmpret);
124 : suxen_drol 1.31 }
125 :     #endif
126 :    
127 : edgomez 1.41
128 :     /* detect cpu flags */
129 :     static unsigned int
130 :     detect_cpu_flags()
131 :     {
132 :     /* enable native assembly optimizations by default */
133 :     unsigned int cpu_flags = XVID_CPU_ASM;
134 :    
135 :     #if defined(ARCH_IS_IA32)
136 :     cpu_flags |= check_cpu_features();
137 :     if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger))
138 :     cpu_flags &= ~XVID_CPU_SSE;
139 :    
140 :     if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger))
141 :     cpu_flags &= ~XVID_CPU_SSE2;
142 :     #endif
143 :    
144 :     #if defined(ARCH_IS_PPC)
145 : edgomez 1.50 if (!sigill_check(altivec_trigger))
146 :     cpu_flags |= XVID_CPU_ALTIVEC;
147 : edgomez 1.41 #endif
148 :    
149 :     return cpu_flags;
150 :     }
151 :    
152 :    
153 : edgomez 1.16 /*****************************************************************************
154 :     * XviD Init Entry point
155 :     *
156 :     * Well this function initialize all internal function pointers according
157 :     * to the CPU features forced by the library client or autodetected (depending
158 :     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
159 :     * image colorspace transformation tables.
160 : edgomez 1.48 *
161 : edgomez 1.16 * Returned value : XVID_ERR_OK
162 :     * + API_VERSION in the input XVID_INIT_PARAM structure
163 :     * + core build " " " " "
164 :     *
165 :     ****************************************************************************/
166 :    
167 : edgomez 1.41
168 : edgomez 1.48 static
169 :     int xvid_gbl_init(xvid_gbl_init_t * init)
170 : Isibaar 1.1 {
171 : edgomez 1.48 unsigned int cpu_flags;
172 : suxen_drol 1.21
173 : edgomez 1.48 if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */
174 :     return XVID_ERR_VERSION;
175 : suxen_drol 1.31
176 : edgomez 1.48 cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags();
177 : Isibaar 1.1
178 : edgomez 1.16 /* Initialize the function pointers */
179 : Isibaar 1.1 idct_int32_init();
180 : Isibaar 1.9 init_vlc_tables();
181 :    
182 : edgomez 1.16 /* Fixed Point Forward/Inverse DCT transformations */
183 : Isibaar 1.1 fdct = fdct_int32;
184 :     idct = idct_int32;
185 :    
186 : edgomez 1.16 /* Only needed on PPC Altivec archs */
187 : canard 1.10 sadInit = 0;
188 : edgomez 1.15
189 : edgomez 1.16 /* Restore FPU context : emms_c is a nop functions */
190 : Isibaar 1.1 emms = emms_c;
191 :    
192 : edgomez 1.48 /* Qpel stuff */
193 :     xvid_QP_Funcs = &xvid_QP_Funcs_C;
194 :     xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C;
195 :     xvid_Init_QP();
196 :    
197 : edgomez 1.16 /* Quantization functions */
198 : edgomez 1.48 quant_h263_intra = quant_h263_intra_c;
199 :     quant_h263_inter = quant_h263_inter_c;
200 :     dequant_h263_intra = dequant_h263_intra_c;
201 :     dequant_h263_inter = dequant_h263_inter_c;
202 :    
203 :     quant_mpeg_intra = quant_mpeg_intra_c;
204 :     quant_mpeg_inter = quant_mpeg_inter_c;
205 :     dequant_mpeg_intra = dequant_mpeg_intra_c;
206 :     dequant_mpeg_inter = dequant_mpeg_inter_c;
207 : Isibaar 1.1
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 : edgomez 1.41 transfer_8to16subro = transfer_8to16subro_c;
213 : suxen_drol 1.11 transfer_8to16sub2 = transfer_8to16sub2_c;
214 : edgomez 1.16 transfer_16to8add = transfer_16to8add_c;
215 :     transfer8x8_copy = transfer8x8_copy_c;
216 : Isibaar 1.1
217 : edgomez 1.41 /* Interlacing functions */
218 :     MBFieldTest = MBFieldTest_c;
219 :    
220 : edgomez 1.16 /* Image interpolation related functions */
221 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c;
222 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c;
223 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c;
224 :    
225 : edgomez 1.41 interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c;
226 :     interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c;
227 :     interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c;
228 :    
229 :     interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c;
230 :     interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c;
231 :     interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c;
232 :    
233 :     interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c;
234 :     interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c;
235 :    
236 :     interpolate8x8_avg2 = interpolate8x8_avg2_c;
237 :     interpolate8x8_avg4 = interpolate8x8_avg4_c;
238 :    
239 : suxen_drol 1.49 /* postprocessing */
240 :     image_brightness = image_brightness_c;
241 :    
242 : edgomez 1.48 /* reduced resolution */
243 : edgomez 1.41 copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C;
244 :     add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C;
245 :     vfilter_31 = xvid_VFilter_31_C;
246 :     hfilter_31 = xvid_HFilter_31_C;
247 :     filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C;
248 :     filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C;
249 :    
250 : edgomez 1.16 /* Initialize internal colorspace transformation tables */
251 : Isibaar 1.1 colorspace_init();
252 :    
253 : edgomez 1.16 /* All colorspace transformation functions User Format->YV12 */
254 : edgomez 1.41 yv12_to_yv12 = yv12_to_yv12_c;
255 :     rgb555_to_yv12 = rgb555_to_yv12_c;
256 :     rgb565_to_yv12 = rgb565_to_yv12_c;
257 :     bgr_to_yv12 = bgr_to_yv12_c;
258 :     bgra_to_yv12 = bgra_to_yv12_c;
259 :     abgr_to_yv12 = abgr_to_yv12_c;
260 :     rgba_to_yv12 = rgba_to_yv12_c;
261 : edgomez 1.48 argb_to_yv12 = argb_to_yv12_c;
262 : edgomez 1.41 yuyv_to_yv12 = yuyv_to_yv12_c;
263 :     uyvy_to_yv12 = uyvy_to_yv12_c;
264 :    
265 :     rgb555i_to_yv12 = rgb555i_to_yv12_c;
266 :     rgb565i_to_yv12 = rgb565i_to_yv12_c;
267 :     bgri_to_yv12 = bgri_to_yv12_c;
268 :     bgrai_to_yv12 = bgrai_to_yv12_c;
269 :     abgri_to_yv12 = abgri_to_yv12_c;
270 :     rgbai_to_yv12 = rgbai_to_yv12_c;
271 : edgomez 1.48 argbi_to_yv12 = argbi_to_yv12_c;
272 : edgomez 1.41 yuyvi_to_yv12 = yuyvi_to_yv12_c;
273 :     uyvyi_to_yv12 = uyvyi_to_yv12_c;
274 :    
275 : edgomez 1.16 /* All colorspace transformation functions YV12->User format */
276 : edgomez 1.41 yv12_to_rgb555 = yv12_to_rgb555_c;
277 :     yv12_to_rgb565 = yv12_to_rgb565_c;
278 :     yv12_to_bgr = yv12_to_bgr_c;
279 :     yv12_to_bgra = yv12_to_bgra_c;
280 :     yv12_to_abgr = yv12_to_abgr_c;
281 :     yv12_to_rgba = yv12_to_rgba_c;
282 : edgomez 1.48 yv12_to_argb = yv12_to_argb_c;
283 : edgomez 1.41 yv12_to_yuyv = yv12_to_yuyv_c;
284 :     yv12_to_uyvy = yv12_to_uyvy_c;
285 : edgomez 1.48
286 : edgomez 1.41 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 : edgomez 1.48 yv12_to_argbi = yv12_to_argbi_c;
293 : edgomez 1.41 yv12_to_yuyvi = yv12_to_yuyvi_c;
294 :     yv12_to_uyvyi = yv12_to_uyvyi_c;
295 : Isibaar 1.1
296 : edgomez 1.16 /* Functions used in motion estimation algorithms */
297 : edgomez 1.48 calc_cbp = calc_cbp_c;
298 :     sad16 = sad16_c;
299 :     sad8 = sad8_c;
300 :     sad16bi = sad16bi_c;
301 :     sad8bi = sad8bi_c;
302 :     dev16 = dev16_c;
303 :     sad16v = sad16v_c;
304 :     sse8_16bit = sse8_16bit_c;
305 : Isibaar 1.1
306 : edgomez 1.41 #if defined(ARCH_IS_IA32)
307 :    
308 : edgomez 1.48 if ((cpu_flags & XVID_CPU_ASM)) {
309 : edgomez 1.41 vfilter_31 = xvid_VFilter_31_x86;
310 :     hfilter_31 = xvid_HFilter_31_x86;
311 :     }
312 :    
313 :     if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) ||
314 :     (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) ||
315 :     (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2))
316 :     {
317 :     /* Restore FPU context : emms_c is a nop functions */
318 :     emms = emms_mmx;
319 :     }
320 :    
321 :     if ((cpu_flags & XVID_CPU_MMX)) {
322 : edgomez 1.16
323 :     /* Forward and Inverse Discrete Cosine Transformation functions */
324 : edgomez 1.48 fdct = fdct_mmx_skal;
325 : Isibaar 1.47 idct = idct_mmx;
326 : Isibaar 1.1
327 : edgomez 1.48 /* Qpel stuff */
328 :     xvid_QP_Funcs = &xvid_QP_Funcs_mmx;
329 :     xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx;
330 :    
331 : edgomez 1.16 /* Quantization related functions */
332 : edgomez 1.48 quant_h263_intra = quant_h263_intra_mmx;
333 :     quant_h263_inter = quant_h263_inter_mmx;
334 :     dequant_h263_intra = dequant_h263_intra_mmx;
335 :     dequant_h263_inter = dequant_h263_inter_mmx;
336 :    
337 :     quant_mpeg_intra = quant_mpeg_intra_mmx;
338 :     quant_mpeg_inter = quant_mpeg_inter_mmx;
339 :     dequant_mpeg_intra = dequant_mpeg_intra_mmx;
340 :     dequant_mpeg_inter = dequant_mpeg_inter_mmx;
341 : Isibaar 1.1
342 : edgomez 1.16 /* Block related functions */
343 : Isibaar 1.1 transfer_8to16copy = transfer_8to16copy_mmx;
344 :     transfer_16to8copy = transfer_16to8copy_mmx;
345 : edgomez 1.16 transfer_8to16sub = transfer_8to16sub_mmx;
346 : edgomez 1.41 transfer_8to16subro = transfer_8to16subro_mmx;
347 : edgomez 1.22 transfer_8to16sub2 = transfer_8to16sub2_mmx;
348 : edgomez 1.16 transfer_16to8add = transfer_16to8add_mmx;
349 :     transfer8x8_copy = transfer8x8_copy_mmx;
350 : edgomez 1.22
351 : edgomez 1.41 /* Interlacing Functions */
352 :     MBFieldTest = MBFieldTest_mmx;
353 : edgomez 1.16
354 :     /* Image Interpolation related functions */
355 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx;
356 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx;
357 : Isibaar 1.1 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx;
358 :    
359 : edgomez 1.41 interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx;
360 :     interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx;
361 :    
362 :     interpolate8x8_avg2 = interpolate8x8_avg2_mmx;
363 :     interpolate8x8_avg4 = interpolate8x8_avg4_mmx;
364 : suxen_drol 1.49
365 :     /* postprocessing */
366 :     image_brightness = image_brightness_mmx;
367 : edgomez 1.41
368 :     /* reduced resolution */
369 :     copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx;
370 :     add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx;
371 :     hfilter_31 = xvid_HFilter_31_mmx;
372 :     filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx;
373 :     filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx;
374 :    
375 :     /* image input xxx_to_yv12 related functions */
376 :     yv12_to_yv12 = yv12_to_yv12_mmx;
377 :     bgr_to_yv12 = bgr_to_yv12_mmx;
378 :     bgra_to_yv12 = bgra_to_yv12_mmx;
379 : edgomez 1.16 yuyv_to_yv12 = yuyv_to_yv12_mmx;
380 :     uyvy_to_yv12 = uyvy_to_yv12_mmx;
381 : Isibaar 1.1
382 : edgomez 1.41 /* image output yv12_to_xxx related functions */
383 :     yv12_to_bgr = yv12_to_bgr_mmx;
384 :     yv12_to_bgra = yv12_to_bgra_mmx;
385 : edgomez 1.16 yv12_to_yuyv = yv12_to_yuyv_mmx;
386 :     yv12_to_uyvy = yv12_to_uyvy_mmx;
387 : edgomez 1.48
388 : edgomez 1.41 yv12_to_yuyvi = yv12_to_yuyvi_mmx;
389 :     yv12_to_uyvyi = yv12_to_uyvyi_mmx;
390 : Isibaar 1.1
391 : edgomez 1.16 /* Motion estimation related functions */
392 : Isibaar 1.1 calc_cbp = calc_cbp_mmx;
393 : edgomez 1.16 sad16 = sad16_mmx;
394 :     sad8 = sad8_mmx;
395 : suxen_drol 1.33 sad16bi = sad16bi_mmx;
396 :     sad8bi = sad8bi_mmx;
397 : edgomez 1.16 dev16 = dev16_mmx;
398 : edgomez 1.41 sad16v = sad16v_mmx;
399 : edgomez 1.48 sse8_16bit = sse8_16bit_mmx;
400 : Isibaar 1.1 }
401 :    
402 : suxen_drol 1.33 /* these 3dnow functions are faster than mmx, but slower than xmm. */
403 : edgomez 1.41 if ((cpu_flags & XVID_CPU_3DNOW)) {
404 :    
405 :     emms = emms_3dn;
406 : suxen_drol 1.33
407 :     /* ME functions */
408 :     sad16bi = sad16bi_3dn;
409 :     sad8bi = sad8bi_3dn;
410 : edgomez 1.41
411 :     yuyv_to_yv12 = yuyv_to_yv12_3dn;
412 :     uyvy_to_yv12 = uyvy_to_yv12_3dn;
413 : suxen_drol 1.33 }
414 :    
415 :    
416 : edgomez 1.41 if ((cpu_flags & XVID_CPU_MMXEXT)) {
417 : edgomez 1.16
418 : edgomez 1.48 /* DCT */
419 :     fdct = fdct_xmm_skal;
420 : Isibaar 1.47 idct = idct_xmm;
421 : edgomez 1.16
422 :     /* Interpolation */
423 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm;
424 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm;
425 : h 1.3 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm;
426 : Isibaar 1.25
427 : edgomez 1.41 /* reduced resolution */
428 :     copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm;
429 :     add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm;
430 :    
431 : chenm001 1.29 /* Quantization */
432 : edgomez 1.48 quant_mpeg_intra = quant_mpeg_intra_xmm;
433 :     quant_mpeg_inter = quant_mpeg_inter_xmm;
434 : edgomez 1.41
435 : edgomez 1.48 dequant_h263_intra = dequant_h263_intra_xmm;
436 :     dequant_h263_inter = dequant_h263_inter_xmm;
437 : chenm001 1.29
438 : edgomez 1.19 /* Buffer transfer */
439 :     transfer_8to16sub2 = transfer_8to16sub2_xmm;
440 : edgomez 1.16
441 :     /* Colorspace transformation */
442 : edgomez 1.41 yv12_to_yv12 = yv12_to_yv12_xmm;
443 :     yuyv_to_yv12 = yuyv_to_yv12_xmm;
444 :     uyvy_to_yv12 = uyvy_to_yv12_xmm;
445 : Isibaar 1.1
446 : edgomez 1.16 /* ME functions */
447 : Isibaar 1.1 sad16 = sad16_xmm;
448 : suxen_drol 1.33 sad8 = sad8_xmm;
449 : chenm001 1.29 sad16bi = sad16bi_xmm;
450 : suxen_drol 1.33 sad8bi = sad8bi_xmm;
451 : Isibaar 1.1 dev16 = dev16_xmm;
452 : edgomez 1.41 sad16v = sad16v_xmm;
453 : Isibaar 1.1 }
454 :    
455 : edgomez 1.41 if ((cpu_flags & XVID_CPU_3DNOW)) {
456 : edgomez 1.16
457 :     /* Interpolation */
458 : Isibaar 1.1 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn;
459 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn;
460 : h 1.4 interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn;
461 : Isibaar 1.1 }
462 :    
463 : edgomez 1.41 if ((cpu_flags & XVID_CPU_3DNOWEXT)) {
464 :    
465 :     /* Buffer transfer */
466 :     transfer_8to16copy = transfer_8to16copy_3dne;
467 :     transfer_16to8copy = transfer_16to8copy_3dne;
468 :     transfer_8to16sub = transfer_8to16sub_3dne;
469 :     transfer_8to16subro = transfer_8to16subro_3dne;
470 :     transfer_16to8add = transfer_16to8add_3dne;
471 :     transfer8x8_copy = transfer8x8_copy_3dne;
472 :    
473 : edgomez 1.51 if ((cpu_flags & XVID_CPU_MMXEXT)) {
474 :     /* Inverse DCT */
475 :     idct = idct_3dne;
476 :    
477 :     /* Buffer transfer */
478 :     transfer_8to16sub2 = transfer_8to16sub2_3dne;
479 :    
480 :     /* Interpolation */
481 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne;
482 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne;
483 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne;
484 :    
485 :     /* Quantization */
486 :     quant_h263_intra = quant_h263_intra_3dne; /* cmov only */
487 :     quant_h263_inter = quant_h263_inter_3dne;
488 :     dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */
489 :     dequant_mpeg_inter = dequant_mpeg_inter_3dne;
490 :     dequant_h263_intra = dequant_h263_intra_3dne;
491 :     dequant_h263_inter = dequant_h263_inter_3dne;
492 :    
493 :     /* ME functions */
494 :     calc_cbp = calc_cbp_3dne;
495 :    
496 :     sad16 = sad16_3dne;
497 :     sad8 = sad8_3dne;
498 :     sad16bi = sad16bi_3dne;
499 :     sad8bi = sad8bi_3dne;
500 :     dev16 = dev16_3dne;
501 :     }
502 : edgomez 1.41 }
503 :    
504 :     if ((cpu_flags & XVID_CPU_SSE2)) {
505 : edgomez 1.16
506 : chenm001 1.29 calc_cbp = calc_cbp_sse2;
507 :    
508 : edgomez 1.16 /* Quantization */
509 : edgomez 1.48 quant_h263_intra = quant_h263_intra_sse2;
510 :     quant_h263_inter = quant_h263_inter_sse2;
511 :     dequant_h263_intra = dequant_h263_intra_sse2;
512 :     dequant_h263_inter = dequant_h263_inter_sse2;
513 : h 1.13
514 : edgomez 1.48 /* SAD operators */
515 : edgomez 1.16 sad16 = sad16_sse2;
516 :     dev16 = dev16_sse2;
517 : edgomez 1.48
518 :     /* DCT operators
519 :     * no iDCT because it's not "Walken matching" */
520 :     fdct = fdct_sse2_skal;
521 : h 1.12 }
522 : edgomez 1.48 #endif /* ARCH_IS_IA32 */
523 : Isibaar 1.18
524 : edgomez 1.41 #if defined(ARCH_IS_IA64)
525 : edgomez 1.43 if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */
526 : Isibaar 1.18 idct_ia64_init();
527 :     fdct = fdct_ia64;
528 : edgomez 1.43 idct = idct_ia64; /*not yet working, crashes */
529 : Isibaar 1.18 interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64;
530 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64;
531 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64;
532 :     sad16 = sad16_ia64;
533 :     sad16bi = sad16bi_ia64;
534 :     sad8 = sad8_ia64;
535 :     dev16 = dev16_ia64;
536 : edgomez 1.43 /* Halfpel8_Refine = Halfpel8_Refine_ia64; */
537 : edgomez 1.48 quant_h263_intra = quant_h263_intra_ia64;
538 :     quant_h263_inter = quant_h263_inter_ia64;
539 :     dequant_h263_intra = dequant_h263_intra_ia64;
540 :     dequant_h263_inter = dequant_h263_inter_ia64;
541 : Isibaar 1.18 transfer_8to16copy = transfer_8to16copy_ia64;
542 :     transfer_16to8copy = transfer_16to8copy_ia64;
543 :     transfer_8to16sub = transfer_8to16sub_ia64;
544 :     transfer_8to16sub2 = transfer_8to16sub2_ia64;
545 :     transfer_16to8add = transfer_16to8add_ia64;
546 :     transfer8x8_copy = transfer8x8_copy_ia64;
547 :     }
548 : edgomez 1.48 #endif
549 : edgomez 1.16
550 : edgomez 1.41 #if defined(ARCH_IS_PPC)
551 : edgomez 1.50 if ((cpu_flags & XVID_CPU_ALTIVEC)) {
552 :     /* sad operators */
553 :     sad16 = sad16_altivec_c;
554 :     sad16bi = sad16bi_altivec_c;
555 :     sad8 = sad8_altivec_c;
556 :     dev16 = dev16_altivec_c;
557 :    
558 :     sse8_16bit = sse8_16bit_altivec_c;
559 :    
560 :     /* mem transfer */
561 :     transfer_8to16copy = transfer_8to16copy_altivec_c;
562 :     transfer_16to8copy = transfer_16to8copy_altivec_c;
563 :     transfer_8to16sub = transfer_8to16sub_altivec_c;
564 :     transfer_8to16subro = transfer_8to16subro_altivec_c;
565 :     transfer_8to16sub2 = transfer_8to16sub2_altivec_c;
566 :     transfer_16to8add = transfer_16to8add_altivec_c;
567 :     transfer8x8_copy = transfer8x8_copy_altivec_c;
568 :    
569 :     /* Inverse DCT */
570 :     idct = idct_altivec_c;
571 :    
572 :     /* Interpolation */
573 :     interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c;
574 :     interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c;
575 :     interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c;
576 :    
577 :     interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c;
578 :     interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c;
579 :    
580 :     interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_altivec_c;
581 :    
582 :     /* Colorspace conversion */
583 :     bgra_to_yv12 = bgra_to_yv12_altivec_c;
584 :     abgr_to_yv12 = abgr_to_yv12_altivec_c;
585 :     rgba_to_yv12 = rgba_to_yv12_altivec_c;
586 :     argb_to_yv12 = argb_to_yv12_altivec_c;
587 :    
588 :     yuyv_to_yv12 = yuyv_to_yv12_altivec_c;
589 :     uyvy_to_yv12 = uyvy_to_yv12_altivec_c;
590 :    
591 :     yv12_to_yuyv = yv12_to_yuyv_altivec_c;
592 :     yv12_to_uyvy = yv12_to_uyvy_altivec_c;
593 :    
594 :     /* Quantization */
595 :     quant_h263_intra = quant_h263_intra_altivec_c;
596 :     quant_h263_inter = quant_h263_inter_altivec_c;
597 :     dequant_h263_intra = dequant_h263_intra_altivec_c;
598 :     dequant_h263_inter = dequant_h263_inter_altivec_c;
599 :     }
600 : canard 1.6 #endif
601 : edgomez 1.41
602 : edgomez 1.48 #if defined(_DEBUG)
603 :     xvid_debug = init->debug;
604 :     #endif
605 :    
606 : edgomez 1.50 return(0);
607 : edgomez 1.41 }
608 :    
609 :    
610 :     static int
611 : edgomez 1.48 xvid_gbl_info(xvid_gbl_info_t * info)
612 : edgomez 1.41 {
613 : edgomez 1.48 if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */
614 :     return XVID_ERR_VERSION;
615 : edgomez 1.41
616 : edgomez 1.48 info->actual_version = XVID_VERSION;
617 :     info->build = "xvid-1.0.0";
618 :     info->cpu_flags = detect_cpu_flags();
619 : edgomez 1.41
620 : edgomez 1.48 #if defined(_SMP) && defined(WIN32)
621 :     info->num_threads = pthread_num_processors_np();;
622 :     #else
623 :     info->num_threads = 0;
624 :     #endif
625 : edgomez 1.41
626 :     return 0;
627 :     }
628 :    
629 :    
630 : edgomez 1.48 static int
631 :     xvid_gbl_convert(xvid_gbl_convert_t* convert)
632 : edgomez 1.41 {
633 : edgomez 1.48 int width;
634 :     int height;
635 :     int width2;
636 :     int height2;
637 :     IMAGE img;
638 : edgomez 1.41
639 : edgomez 1.48 if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */
640 :     return XVID_ERR_VERSION;
641 : edgomez 1.41
642 : edgomez 1.48 #if 0
643 :     const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP);
644 :     #endif
645 :     width = convert->width;
646 :     height = convert->height;
647 :     width2 = convert->width/2;
648 :     height2 = convert->height/2;
649 : edgomez 1.41
650 : edgomez 1.48 switch (convert->input.csp & ~XVID_CSP_VFLIP)
651 : edgomez 1.41 {
652 : edgomez 1.48 case XVID_CSP_YV12 :
653 :     img.y = convert->input.plane[0];
654 :     img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height;
655 :     img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2;
656 :     image_output(&img, width, height, width,
657 :     (uint8_t**)convert->output.plane, convert->output.stride,
658 :     convert->output.csp, convert->interlacing);
659 :     break;
660 : edgomez 1.41
661 : edgomez 1.48 default :
662 :     return XVID_ERR_FORMAT;
663 : edgomez 1.41 }
664 :    
665 :    
666 : edgomez 1.48 emms();
667 : edgomez 1.41 return 0;
668 :     }
669 :    
670 : edgomez 1.48 /*****************************************************************************
671 :     * XviD Global Entry point
672 :     *
673 :     * Well this function initialize all internal function pointers according
674 :     * to the CPU features forced by the library client or autodetected (depending
675 :     * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all
676 :     * image colorspace transformation tables.
677 :     *
678 :     ****************************************************************************/
679 : Isibaar 1.1
680 : edgomez 1.41
681 :     int
682 : edgomez 1.48 xvid_global(void *handle,
683 : edgomez 1.41 int opt,
684 :     void *param1,
685 :     void *param2)
686 :     {
687 :     switch(opt)
688 :     {
689 : edgomez 1.48 case XVID_GBL_INIT :
690 :     return xvid_gbl_init((xvid_gbl_init_t*)param1);
691 : edgomez 1.41
692 : edgomez 1.48 case XVID_GBL_INFO :
693 :     return xvid_gbl_info((xvid_gbl_info_t*)param1);
694 :    
695 :     case XVID_GBL_CONVERT :
696 :     return xvid_gbl_convert((xvid_gbl_convert_t*)param1);
697 : edgomez 1.41
698 :     default :
699 :     return XVID_ERR_FAIL;
700 :     }
701 :     }
702 :    
703 : edgomez 1.16 /*****************************************************************************
704 :     * XviD Native decoder entry point
705 :     *
706 :     * This function is just a wrapper to all the option cases.
707 :     *
708 :     * Returned values : XVID_ERR_FAIL when opt is invalid
709 :     * else returns the wrapped function result
710 :     *
711 :     ****************************************************************************/
712 :    
713 : edgomez 1.15 int
714 :     xvid_decore(void *handle,
715 :     int opt,
716 :     void *param1,
717 :     void *param2)
718 : Isibaar 1.1 {
719 : edgomez 1.15 switch (opt) {
720 :     case XVID_DEC_CREATE:
721 : edgomez 1.48 return decoder_create((xvid_dec_create_t *) param1);
722 : edgomez 1.15
723 :     case XVID_DEC_DESTROY:
724 :     return decoder_destroy((DECODER *) handle);
725 : Isibaar 1.1
726 : edgomez 1.48 case XVID_DEC_DECODE:
727 :     return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2);
728 :    
729 : Isibaar 1.1 default:
730 : edgomez 1.15 return XVID_ERR_FAIL;
731 :     }
732 : Isibaar 1.1 }
733 :    
734 : edgomez 1.16
735 :     /*****************************************************************************
736 :     * XviD Native encoder entry point
737 :     *
738 :     * This function is just a wrapper to all the option cases.
739 :     *
740 :     * Returned values : XVID_ERR_FAIL when opt is invalid
741 :     * else returns the wrapped function result
742 :     *
743 :     ****************************************************************************/
744 : Isibaar 1.1
745 : edgomez 1.15 int
746 :     xvid_encore(void *handle,
747 :     int opt,
748 :     void *param1,
749 :     void *param2)
750 : Isibaar 1.1 {
751 : edgomez 1.15 switch (opt) {
752 :     case XVID_ENC_ENCODE:
753 : edgomez 1.41
754 : edgomez 1.48 return enc_encode((Encoder *) handle,
755 :     (xvid_enc_frame_t *) param1,
756 :     (xvid_enc_stats_t *) param2);
757 : edgomez 1.15
758 :     case XVID_ENC_CREATE:
759 : edgomez 1.48 return enc_create((xvid_enc_create_t *) param1);
760 : edgomez 1.15
761 :     case XVID_ENC_DESTROY:
762 : edgomez 1.48 return enc_destroy((Encoder *) handle);
763 : Isibaar 1.1
764 :     default:
765 : edgomez 1.15 return XVID_ERR_FAIL;
766 :     }
767 : Isibaar 1.1 }

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