Parent Directory
|
Revision Log
Revision 1493 - (view) (download)
1 : | edgomez | 200 | /***************************************************************************** |
2 : | edgomez | 201 | * |
3 : | * XVID MPEG-4 VIDEO CODEC | ||
4 : | * - Native API implementation - | ||
5 : | * | ||
6 : | suxen_drol | 1397 | * Copyright(C) 2001-2004 Peter Ross <pross@xvid.org> |
7 : | edgomez | 1382 | * |
8 : | edgomez | 851 | * 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 | 201 | * (at your option) any later version. |
12 : | * | ||
13 : | * This program is distributed in the hope that it will be useful, | ||
14 : | edgomez | 851 | * but WITHOUT ANY WARRANTY ; without even the implied warranty of |
15 : | edgomez | 201 | * 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 | 851 | * along with this program ; if not, write to the Free Software |
20 : | edgomez | 201 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 : | * | ||
22 : | suxen_drol | 1493 | * $Id: xvid.c,v 1.54 2004-07-15 10:08:21 suxen_drol Exp $ |
23 : | edgomez | 410 | * |
24 : | edgomez | 201 | ****************************************************************************/ |
25 : | chenm001 | 274 | |
26 : | edgomez | 851 | #include <stdio.h> |
27 : | #include <stdlib.h> | ||
28 : | #include <string.h> | ||
29 : | #include <time.h> | ||
30 : | |||
31 : | Isibaar | 3 | #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 | 851 | #include "image/reduced.h" |
40 : | Isibaar | 3 | #include "utils/mem_transfer.h" |
41 : | edgomez | 851 | #include "utils/mbfunctions.h" |
42 : | edgomez | 1382 | #include "quant/quant.h" |
43 : | ia64p | 299 | #include "motion/motion.h" |
44 : | Isibaar | 3 | #include "motion/sad.h" |
45 : | #include "utils/emms.h" | ||
46 : | #include "utils/timer.h" | ||
47 : | Isibaar | 100 | #include "bitstream/mbcoding.h" |
48 : | edgomez | 1382 | #include "image/qpel.h" |
49 : | #include "image/postprocessing.h" | ||
50 : | Isibaar | 3 | |
51 : | edgomez | 1382 | #if defined(_DEBUG) |
52 : | unsigned int xvid_debug = 0; /* xvid debug mask */ | ||
53 : | #endif | ||
54 : | suxen_drol | 311 | |
55 : | edgomez | 1412 | #if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
56 : | edgomez | 851 | # include <windows.h> |
57 : | edgomez | 1412 | #elif defined(ARCH_IS_IA32) || defined(ARCH_IS_PPC) |
58 : | edgomez | 851 | # include <signal.h> |
59 : | # include <setjmp.h> | ||
60 : | suxen_drol | 311 | |
61 : | edgomez | 851 | static jmp_buf mark; |
62 : | suxen_drol | 311 | |
63 : | edgomez | 851 | static void |
64 : | sigill_handler(int signal) | ||
65 : | { | ||
66 : | longjmp(mark, 1); | ||
67 : | } | ||
68 : | suxen_drol | 311 | #endif |
69 : | |||
70 : | |||
71 : | /* | ||
72 : | edgomez | 874 | * 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 | 1412 | #if defined(ARCH_IS_IA32) && defined(_MSC_VER) |
81 : | static int | ||
82 : | suxen_drol | 311 | sigill_check(void (*func)()) |
83 : | { | ||
84 : | _try { | ||
85 : | func(); | ||
86 : | edgomez | 1412 | } _except(EXCEPTION_EXECUTE_HANDLER) { |
87 : | suxen_drol | 311 | |
88 : | if (_exception_code() == STATUS_ILLEGAL_INSTRUCTION) | ||
89 : | edgomez | 1412 | return(1); |
90 : | suxen_drol | 311 | } |
91 : | edgomez | 1412 | 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 | 311 | int jmpret; |
99 : | |||
100 : | edgomez | 1412 | /* Set our SIGILL handler */ |
101 : | old_handler = signal(SIGILL, sigill_handler); | ||
102 : | suxen_drol | 311 | |
103 : | edgomez | 1412 | /* Check for error */ |
104 : | if (old_handler == SIG_ERR) { | ||
105 : | return(-1); | ||
106 : | suxen_drol | 311 | } |
107 : | |||
108 : | edgomez | 1412 | /* 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 | 311 | func(); |
118 : | } | ||
119 : | |||
120 : | edgomez | 1412 | /* Restore old signal handler */ |
121 : | suxen_drol | 311 | signal(SIGILL, old_handler); |
122 : | |||
123 : | edgomez | 1412 | return(jmpret); |
124 : | suxen_drol | 311 | } |
125 : | #endif | ||
126 : | |||
127 : | edgomez | 851 | |
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 | 1412 | if (!sigill_check(altivec_trigger)) |
146 : | cpu_flags |= XVID_CPU_ALTIVEC; | ||
147 : | edgomez | 851 | #endif |
148 : | |||
149 : | return cpu_flags; | ||
150 : | } | ||
151 : | |||
152 : | |||
153 : | edgomez | 200 | /***************************************************************************** |
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 | 1382 | * |
161 : | edgomez | 200 | * Returned value : XVID_ERR_OK |
162 : | * + API_VERSION in the input XVID_INIT_PARAM structure | ||
163 : | * + core build " " " " " | ||
164 : | * | ||
165 : | ****************************************************************************/ | ||
166 : | |||
167 : | edgomez | 851 | |
168 : | edgomez | 1382 | static |
169 : | int xvid_gbl_init(xvid_gbl_init_t * init) | ||
170 : | Isibaar | 3 | { |
171 : | edgomez | 1382 | unsigned int cpu_flags; |
172 : | Isibaar | 3 | |
173 : | edgomez | 1382 | if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */ |
174 : | return XVID_ERR_VERSION; | ||
175 : | suxen_drol | 234 | |
176 : | edgomez | 1382 | cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
177 : | suxen_drol | 234 | |
178 : | edgomez | 200 | /* Initialize the function pointers */ |
179 : | Isibaar | 3 | idct_int32_init(); |
180 : | Isibaar | 100 | init_vlc_tables(); |
181 : | |||
182 : | edgomez | 200 | /* Fixed Point Forward/Inverse DCT transformations */ |
183 : | Isibaar | 3 | fdct = fdct_int32; |
184 : | idct = idct_int32; | ||
185 : | |||
186 : | edgomez | 200 | /* Only needed on PPC Altivec archs */ |
187 : | canard | 115 | sadInit = 0; |
188 : | edgomez | 195 | |
189 : | edgomez | 200 | /* Restore FPU context : emms_c is a nop functions */ |
190 : | Isibaar | 3 | emms = emms_c; |
191 : | |||
192 : | edgomez | 1382 | /* 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 | 200 | /* Quantization functions */ |
198 : | edgomez | 1382 | 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 : | Isibaar | 3 | |
203 : | edgomez | 1382 | 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 | 3 | |
208 : | edgomez | 200 | /* Block transfer related functions */ |
209 : | Isibaar | 3 | transfer_8to16copy = transfer_8to16copy_c; |
210 : | transfer_16to8copy = transfer_16to8copy_c; | ||
211 : | edgomez | 200 | transfer_8to16sub = transfer_8to16sub_c; |
212 : | edgomez | 851 | transfer_8to16subro = transfer_8to16subro_c; |
213 : | suxen_drol | 118 | transfer_8to16sub2 = transfer_8to16sub2_c; |
214 : | edgomez | 200 | transfer_16to8add = transfer_16to8add_c; |
215 : | transfer8x8_copy = transfer8x8_copy_c; | ||
216 : | Isibaar | 3 | |
217 : | edgomez | 851 | /* Interlacing functions */ |
218 : | MBFieldTest = MBFieldTest_c; | ||
219 : | |||
220 : | edgomez | 200 | /* Image interpolation related functions */ |
221 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; | ||
222 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; | ||
223 : | Isibaar | 3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
224 : | |||
225 : | edgomez | 851 | 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 | 1397 | /* postprocessing */ |
240 : | image_brightness = image_brightness_c; | ||
241 : | |||
242 : | edgomez | 1382 | /* reduced resolution */ |
243 : | edgomez | 851 | 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 | 200 | /* Initialize internal colorspace transformation tables */ |
251 : | Isibaar | 3 | colorspace_init(); |
252 : | |||
253 : | edgomez | 200 | /* All colorspace transformation functions User Format->YV12 */ |
254 : | edgomez | 851 | 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 | 1382 | argb_to_yv12 = argb_to_yv12_c; |
262 : | edgomez | 851 | yuyv_to_yv12 = yuyv_to_yv12_c; |
263 : | uyvy_to_yv12 = uyvy_to_yv12_c; | ||
264 : | Isibaar | 3 | |
265 : | edgomez | 851 | 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 | 1382 | argbi_to_yv12 = argbi_to_yv12_c; |
272 : | edgomez | 851 | yuyvi_to_yv12 = yuyvi_to_yv12_c; |
273 : | uyvyi_to_yv12 = uyvyi_to_yv12_c; | ||
274 : | |||
275 : | edgomez | 200 | /* All colorspace transformation functions YV12->User format */ |
276 : | edgomez | 851 | 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 | 1382 | yv12_to_argb = yv12_to_argb_c; |
283 : | edgomez | 851 | yv12_to_yuyv = yv12_to_yuyv_c; |
284 : | yv12_to_uyvy = yv12_to_uyvy_c; | ||
285 : | edgomez | 1382 | |
286 : | edgomez | 851 | 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 | 1382 | yv12_to_argbi = yv12_to_argbi_c; |
293 : | edgomez | 851 | yv12_to_yuyvi = yv12_to_yuyvi_c; |
294 : | yv12_to_uyvyi = yv12_to_uyvyi_c; | ||
295 : | Isibaar | 3 | |
296 : | edgomez | 200 | /* Functions used in motion estimation algorithms */ |
297 : | edgomez | 1382 | 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 : | edgomez | 1424 | sse8_8bit = sse8_8bit_c; |
306 : | Isibaar | 3 | |
307 : | edgomez | 851 | #if defined(ARCH_IS_IA32) |
308 : | edgomez | 200 | |
309 : | edgomez | 1382 | if ((cpu_flags & XVID_CPU_ASM)) { |
310 : | edgomez | 851 | vfilter_31 = xvid_VFilter_31_x86; |
311 : | hfilter_31 = xvid_HFilter_31_x86; | ||
312 : | } | ||
313 : | |||
314 : | if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || | ||
315 : | (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || | ||
316 : | (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) | ||
317 : | { | ||
318 : | /* Restore FPU context : emms_c is a nop functions */ | ||
319 : | emms = emms_mmx; | ||
320 : | } | ||
321 : | |||
322 : | if ((cpu_flags & XVID_CPU_MMX)) { | ||
323 : | |||
324 : | edgomez | 200 | /* Forward and Inverse Discrete Cosine Transformation functions */ |
325 : | edgomez | 1382 | fdct = fdct_mmx_skal; |
326 : | Isibaar | 1066 | idct = idct_mmx; |
327 : | Isibaar | 3 | |
328 : | edgomez | 1382 | /* Qpel stuff */ |
329 : | xvid_QP_Funcs = &xvid_QP_Funcs_mmx; | ||
330 : | xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx; | ||
331 : | |||
332 : | edgomez | 200 | /* Quantization related functions */ |
333 : | edgomez | 1382 | quant_h263_intra = quant_h263_intra_mmx; |
334 : | quant_h263_inter = quant_h263_inter_mmx; | ||
335 : | dequant_h263_intra = dequant_h263_intra_mmx; | ||
336 : | dequant_h263_inter = dequant_h263_inter_mmx; | ||
337 : | Isibaar | 3 | |
338 : | edgomez | 1382 | quant_mpeg_intra = quant_mpeg_intra_mmx; |
339 : | quant_mpeg_inter = quant_mpeg_inter_mmx; | ||
340 : | dequant_mpeg_intra = dequant_mpeg_intra_mmx; | ||
341 : | dequant_mpeg_inter = dequant_mpeg_inter_mmx; | ||
342 : | Isibaar | 3 | |
343 : | edgomez | 200 | /* Block related functions */ |
344 : | Isibaar | 3 | transfer_8to16copy = transfer_8to16copy_mmx; |
345 : | transfer_16to8copy = transfer_16to8copy_mmx; | ||
346 : | edgomez | 200 | transfer_8to16sub = transfer_8to16sub_mmx; |
347 : | edgomez | 851 | transfer_8to16subro = transfer_8to16subro_mmx; |
348 : | edgomez | 236 | transfer_8to16sub2 = transfer_8to16sub2_mmx; |
349 : | edgomez | 200 | transfer_16to8add = transfer_16to8add_mmx; |
350 : | transfer8x8_copy = transfer8x8_copy_mmx; | ||
351 : | Isibaar | 3 | |
352 : | edgomez | 851 | /* Interlacing Functions */ |
353 : | MBFieldTest = MBFieldTest_mmx; | ||
354 : | edgomez | 236 | |
355 : | edgomez | 200 | /* Image Interpolation related functions */ |
356 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; | ||
357 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; | ||
358 : | Isibaar | 3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
359 : | |||
360 : | edgomez | 851 | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
361 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; | ||
362 : | |||
363 : | interpolate8x8_avg2 = interpolate8x8_avg2_mmx; | ||
364 : | interpolate8x8_avg4 = interpolate8x8_avg4_mmx; | ||
365 : | |||
366 : | suxen_drol | 1397 | /* postprocessing */ |
367 : | image_brightness = image_brightness_mmx; | ||
368 : | |||
369 : | edgomez | 851 | /* reduced resolution */ |
370 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; | ||
371 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; | ||
372 : | hfilter_31 = xvid_HFilter_31_mmx; | ||
373 : | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; | ||
374 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; | ||
375 : | |||
376 : | /* image input xxx_to_yv12 related functions */ | ||
377 : | yv12_to_yv12 = yv12_to_yv12_mmx; | ||
378 : | bgr_to_yv12 = bgr_to_yv12_mmx; | ||
379 : | bgra_to_yv12 = bgra_to_yv12_mmx; | ||
380 : | edgomez | 200 | yuyv_to_yv12 = yuyv_to_yv12_mmx; |
381 : | uyvy_to_yv12 = uyvy_to_yv12_mmx; | ||
382 : | Isibaar | 3 | |
383 : | edgomez | 851 | /* image output yv12_to_xxx related functions */ |
384 : | yv12_to_bgr = yv12_to_bgr_mmx; | ||
385 : | yv12_to_bgra = yv12_to_bgra_mmx; | ||
386 : | edgomez | 200 | yv12_to_yuyv = yv12_to_yuyv_mmx; |
387 : | yv12_to_uyvy = yv12_to_uyvy_mmx; | ||
388 : | edgomez | 1382 | |
389 : | edgomez | 851 | yv12_to_yuyvi = yv12_to_yuyvi_mmx; |
390 : | yv12_to_uyvyi = yv12_to_uyvyi_mmx; | ||
391 : | Isibaar | 3 | |
392 : | edgomez | 200 | /* Motion estimation related functions */ |
393 : | edgomez | 1424 | calc_cbp = calc_cbp_mmx; |
394 : | sad16 = sad16_mmx; | ||
395 : | sad8 = sad8_mmx; | ||
396 : | sad16bi = sad16bi_mmx; | ||
397 : | sad8bi = sad8bi_mmx; | ||
398 : | dev16 = dev16_mmx; | ||
399 : | sad16v = sad16v_mmx; | ||
400 : | edgomez | 1382 | sse8_16bit = sse8_16bit_mmx; |
401 : | edgomez | 1424 | sse8_8bit = sse8_8bit_mmx; |
402 : | Isibaar | 3 | } |
403 : | |||
404 : | suxen_drol | 329 | /* these 3dnow functions are faster than mmx, but slower than xmm. */ |
405 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
406 : | suxen_drol | 329 | |
407 : | edgomez | 851 | emms = emms_3dn; |
408 : | |||
409 : | suxen_drol | 329 | /* ME functions */ |
410 : | sad16bi = sad16bi_3dn; | ||
411 : | sad8bi = sad8bi_3dn; | ||
412 : | edgomez | 851 | |
413 : | yuyv_to_yv12 = yuyv_to_yv12_3dn; | ||
414 : | uyvy_to_yv12 = uyvy_to_yv12_3dn; | ||
415 : | suxen_drol | 329 | } |
416 : | |||
417 : | |||
418 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_MMXEXT)) { |
419 : | edgomez | 200 | |
420 : | edgomez | 1382 | /* DCT */ |
421 : | fdct = fdct_xmm_skal; | ||
422 : | Isibaar | 1066 | idct = idct_xmm; |
423 : | edgomez | 200 | |
424 : | /* Interpolation */ | ||
425 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; | ||
426 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; | ||
427 : | h | 38 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
428 : | edgomez | 200 | |
429 : | edgomez | 851 | /* reduced resolution */ |
430 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; | ||
431 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; | ||
432 : | |||
433 : | chenm001 | 274 | /* Quantization */ |
434 : | edgomez | 1382 | quant_mpeg_intra = quant_mpeg_intra_xmm; |
435 : | quant_mpeg_inter = quant_mpeg_inter_xmm; | ||
436 : | edgomez | 851 | |
437 : | edgomez | 1382 | dequant_h263_intra = dequant_h263_intra_xmm; |
438 : | dequant_h263_inter = dequant_h263_inter_xmm; | ||
439 : | chenm001 | 274 | |
440 : | edgomez | 218 | /* Buffer transfer */ |
441 : | transfer_8to16sub2 = transfer_8to16sub2_xmm; | ||
442 : | |||
443 : | edgomez | 200 | /* Colorspace transformation */ |
444 : | edgomez | 851 | yv12_to_yv12 = yv12_to_yv12_xmm; |
445 : | yuyv_to_yv12 = yuyv_to_yv12_xmm; | ||
446 : | uyvy_to_yv12 = uyvy_to_yv12_xmm; | ||
447 : | Isibaar | 3 | |
448 : | edgomez | 200 | /* ME functions */ |
449 : | Isibaar | 3 | sad16 = sad16_xmm; |
450 : | suxen_drol | 329 | sad8 = sad8_xmm; |
451 : | chenm001 | 274 | sad16bi = sad16bi_xmm; |
452 : | suxen_drol | 329 | sad8bi = sad8bi_xmm; |
453 : | Isibaar | 3 | dev16 = dev16_xmm; |
454 : | edgomez | 851 | sad16v = sad16v_xmm; |
455 : | Isibaar | 3 | } |
456 : | |||
457 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
458 : | edgomez | 200 | |
459 : | /* Interpolation */ | ||
460 : | Isibaar | 3 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
461 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; | ||
462 : | h | 40 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
463 : | Isibaar | 3 | } |
464 : | |||
465 : | edgomez | 851 | if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
466 : | edgomez | 200 | |
467 : | edgomez | 851 | /* Buffer transfer */ |
468 : | transfer_8to16copy = transfer_8to16copy_3dne; | ||
469 : | transfer_16to8copy = transfer_16to8copy_3dne; | ||
470 : | transfer_8to16sub = transfer_8to16sub_3dne; | ||
471 : | transfer_8to16subro = transfer_8to16subro_3dne; | ||
472 : | transfer_16to8add = transfer_16to8add_3dne; | ||
473 : | transfer8x8_copy = transfer8x8_copy_3dne; | ||
474 : | |||
475 : | edgomez | 1423 | if ((cpu_flags & XVID_CPU_MMXEXT)) { |
476 : | /* Inverse DCT */ | ||
477 : | idct = idct_3dne; | ||
478 : | edgomez | 851 | |
479 : | edgomez | 1423 | /* Buffer transfer */ |
480 : | transfer_8to16sub2 = transfer_8to16sub2_3dne; | ||
481 : | edgomez | 851 | |
482 : | edgomez | 1423 | /* Interpolation */ |
483 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; | ||
484 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; | ||
485 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; | ||
486 : | |||
487 : | /* Quantization */ | ||
488 : | quant_h263_intra = quant_h263_intra_3dne; /* cmov only */ | ||
489 : | quant_h263_inter = quant_h263_inter_3dne; | ||
490 : | dequant_mpeg_intra = dequant_mpeg_intra_3dne; /* cmov only */ | ||
491 : | dequant_mpeg_inter = dequant_mpeg_inter_3dne; | ||
492 : | dequant_h263_intra = dequant_h263_intra_3dne; | ||
493 : | dequant_h263_inter = dequant_h263_inter_3dne; | ||
494 : | |||
495 : | /* ME functions */ | ||
496 : | calc_cbp = calc_cbp_3dne; | ||
497 : | |||
498 : | sad16 = sad16_3dne; | ||
499 : | sad8 = sad8_3dne; | ||
500 : | sad16bi = sad16bi_3dne; | ||
501 : | sad8bi = sad8bi_3dne; | ||
502 : | dev16 = dev16_3dne; | ||
503 : | } | ||
504 : | edgomez | 851 | } |
505 : | |||
506 : | if ((cpu_flags & XVID_CPU_SSE2)) { | ||
507 : | |||
508 : | chenm001 | 274 | calc_cbp = calc_cbp_sse2; |
509 : | |||
510 : | edgomez | 200 | /* Quantization */ |
511 : | edgomez | 1382 | quant_h263_intra = quant_h263_intra_sse2; |
512 : | quant_h263_inter = quant_h263_inter_sse2; | ||
513 : | dequant_h263_intra = dequant_h263_intra_sse2; | ||
514 : | dequant_h263_inter = dequant_h263_inter_sse2; | ||
515 : | h | 135 | |
516 : | edgomez | 1382 | /* SAD operators */ |
517 : | edgomez | 200 | sad16 = sad16_sse2; |
518 : | dev16 = dev16_sse2; | ||
519 : | edgomez | 1382 | |
520 : | /* DCT operators | ||
521 : | * no iDCT because it's not "Walken matching" */ | ||
522 : | fdct = fdct_sse2_skal; | ||
523 : | suxen_drol | 1493 | |
524 : | /* postprocessing */ | ||
525 : | image_brightness = image_brightness_sse2; | ||
526 : | h | 126 | } |
527 : | edgomez | 1382 | #endif /* ARCH_IS_IA32 */ |
528 : | edgomez | 200 | |
529 : | edgomez | 851 | #if defined(ARCH_IS_IA64) |
530 : | edgomez | 874 | if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
531 : | Isibaar | 209 | idct_ia64_init(); |
532 : | fdct = fdct_ia64; | ||
533 : | edgomez | 874 | idct = idct_ia64; /*not yet working, crashes */ |
534 : | Isibaar | 209 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
535 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; | ||
536 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; | ||
537 : | sad16 = sad16_ia64; | ||
538 : | sad16bi = sad16bi_ia64; | ||
539 : | sad8 = sad8_ia64; | ||
540 : | dev16 = dev16_ia64; | ||
541 : | edgomez | 874 | /* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
542 : | edgomez | 1382 | quant_h263_intra = quant_h263_intra_ia64; |
543 : | quant_h263_inter = quant_h263_inter_ia64; | ||
544 : | dequant_h263_intra = dequant_h263_intra_ia64; | ||
545 : | dequant_h263_inter = dequant_h263_inter_ia64; | ||
546 : | Isibaar | 209 | transfer_8to16copy = transfer_8to16copy_ia64; |
547 : | transfer_16to8copy = transfer_16to8copy_ia64; | ||
548 : | transfer_8to16sub = transfer_8to16sub_ia64; | ||
549 : | transfer_8to16sub2 = transfer_8to16sub2_ia64; | ||
550 : | transfer_16to8add = transfer_16to8add_ia64; | ||
551 : | transfer8x8_copy = transfer8x8_copy_ia64; | ||
552 : | } | ||
553 : | edgomez | 1382 | #endif |
554 : | Isibaar | 209 | |
555 : | edgomez | 851 | #if defined(ARCH_IS_PPC) |
556 : | edgomez | 1412 | if ((cpu_flags & XVID_CPU_ALTIVEC)) { |
557 : | /* sad operators */ | ||
558 : | sad16 = sad16_altivec_c; | ||
559 : | sad16bi = sad16bi_altivec_c; | ||
560 : | sad8 = sad8_altivec_c; | ||
561 : | dev16 = dev16_altivec_c; | ||
562 : | |||
563 : | sse8_16bit = sse8_16bit_altivec_c; | ||
564 : | |||
565 : | /* mem transfer */ | ||
566 : | transfer_8to16copy = transfer_8to16copy_altivec_c; | ||
567 : | transfer_16to8copy = transfer_16to8copy_altivec_c; | ||
568 : | transfer_8to16sub = transfer_8to16sub_altivec_c; | ||
569 : | transfer_8to16subro = transfer_8to16subro_altivec_c; | ||
570 : | transfer_8to16sub2 = transfer_8to16sub2_altivec_c; | ||
571 : | transfer_16to8add = transfer_16to8add_altivec_c; | ||
572 : | transfer8x8_copy = transfer8x8_copy_altivec_c; | ||
573 : | |||
574 : | /* Inverse DCT */ | ||
575 : | idct = idct_altivec_c; | ||
576 : | |||
577 : | /* Interpolation */ | ||
578 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_altivec_c; | ||
579 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_altivec_c; | ||
580 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_altivec_c; | ||
581 : | |||
582 : | interpolate8x8_avg2 = interpolate8x8_avg2_altivec_c; | ||
583 : | interpolate8x8_avg4 = interpolate8x8_avg4_altivec_c; | ||
584 : | |||
585 : | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_altivec_c; | ||
586 : | |||
587 : | /* Colorspace conversion */ | ||
588 : | bgra_to_yv12 = bgra_to_yv12_altivec_c; | ||
589 : | abgr_to_yv12 = abgr_to_yv12_altivec_c; | ||
590 : | rgba_to_yv12 = rgba_to_yv12_altivec_c; | ||
591 : | argb_to_yv12 = argb_to_yv12_altivec_c; | ||
592 : | |||
593 : | yuyv_to_yv12 = yuyv_to_yv12_altivec_c; | ||
594 : | uyvy_to_yv12 = uyvy_to_yv12_altivec_c; | ||
595 : | |||
596 : | yv12_to_yuyv = yv12_to_yuyv_altivec_c; | ||
597 : | yv12_to_uyvy = yv12_to_uyvy_altivec_c; | ||
598 : | |||
599 : | /* Quantization */ | ||
600 : | quant_h263_intra = quant_h263_intra_altivec_c; | ||
601 : | quant_h263_inter = quant_h263_inter_altivec_c; | ||
602 : | dequant_h263_intra = dequant_h263_intra_altivec_c; | ||
603 : | dequant_h263_inter = dequant_h263_inter_altivec_c; | ||
604 : | } | ||
605 : | canard | 52 | #endif |
606 : | edgomez | 851 | |
607 : | edgomez | 1382 | #if defined(_DEBUG) |
608 : | xvid_debug = init->debug; | ||
609 : | #endif | ||
610 : | |||
611 : | edgomez | 1412 | return(0); |
612 : | edgomez | 851 | } |
613 : | |||
614 : | |||
615 : | static int | ||
616 : | edgomez | 1382 | xvid_gbl_info(xvid_gbl_info_t * info) |
617 : | edgomez | 851 | { |
618 : | edgomez | 1382 | if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */ |
619 : | return XVID_ERR_VERSION; | ||
620 : | edgomez | 851 | |
621 : | edgomez | 1382 | info->actual_version = XVID_VERSION; |
622 : | suxen_drol | 1431 | info->build = "xvid-1.1-cvshead"; |
623 : | edgomez | 1382 | info->cpu_flags = detect_cpu_flags(); |
624 : | edgomez | 851 | |
625 : | edgomez | 1382 | #if defined(_SMP) && defined(WIN32) |
626 : | info->num_threads = pthread_num_processors_np();; | ||
627 : | #else | ||
628 : | info->num_threads = 0; | ||
629 : | #endif | ||
630 : | edgomez | 851 | |
631 : | return 0; | ||
632 : | } | ||
633 : | |||
634 : | |||
635 : | edgomez | 1382 | static int |
636 : | xvid_gbl_convert(xvid_gbl_convert_t* convert) | ||
637 : | edgomez | 851 | { |
638 : | edgomez | 1382 | int width; |
639 : | int height; | ||
640 : | int width2; | ||
641 : | int height2; | ||
642 : | IMAGE img; | ||
643 : | edgomez | 851 | |
644 : | edgomez | 1382 | if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */ |
645 : | return XVID_ERR_VERSION; | ||
646 : | edgomez | 851 | |
647 : | edgomez | 1382 | #if 0 |
648 : | const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); | ||
649 : | #endif | ||
650 : | width = convert->width; | ||
651 : | height = convert->height; | ||
652 : | width2 = convert->width/2; | ||
653 : | height2 = convert->height/2; | ||
654 : | edgomez | 851 | |
655 : | edgomez | 1382 | switch (convert->input.csp & ~XVID_CSP_VFLIP) |
656 : | edgomez | 851 | { |
657 : | edgomez | 1382 | case XVID_CSP_YV12 : |
658 : | img.y = convert->input.plane[0]; | ||
659 : | img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; | ||
660 : | img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; | ||
661 : | image_output(&img, width, height, width, | ||
662 : | (uint8_t**)convert->output.plane, convert->output.stride, | ||
663 : | convert->output.csp, convert->interlacing); | ||
664 : | break; | ||
665 : | edgomez | 851 | |
666 : | edgomez | 1382 | default : |
667 : | return XVID_ERR_FORMAT; | ||
668 : | edgomez | 851 | } |
669 : | |||
670 : | |||
671 : | edgomez | 1382 | emms(); |
672 : | edgomez | 851 | return 0; |
673 : | } | ||
674 : | |||
675 : | edgomez | 1382 | /***************************************************************************** |
676 : | * XviD Global Entry point | ||
677 : | * | ||
678 : | * Well this function initialize all internal function pointers according | ||
679 : | * to the CPU features forced by the library client or autodetected (depending | ||
680 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
681 : | * image colorspace transformation tables. | ||
682 : | * | ||
683 : | ****************************************************************************/ | ||
684 : | edgomez | 851 | |
685 : | |||
686 : | int | ||
687 : | edgomez | 1382 | xvid_global(void *handle, |
688 : | edgomez | 851 | int opt, |
689 : | void *param1, | ||
690 : | void *param2) | ||
691 : | { | ||
692 : | switch(opt) | ||
693 : | { | ||
694 : | edgomez | 1382 | case XVID_GBL_INIT : |
695 : | return xvid_gbl_init((xvid_gbl_init_t*)param1); | ||
696 : | edgomez | 851 | |
697 : | edgomez | 1382 | case XVID_GBL_INFO : |
698 : | return xvid_gbl_info((xvid_gbl_info_t*)param1); | ||
699 : | edgomez | 851 | |
700 : | edgomez | 1382 | case XVID_GBL_CONVERT : |
701 : | return xvid_gbl_convert((xvid_gbl_convert_t*)param1); | ||
702 : | |||
703 : | edgomez | 851 | default : |
704 : | return XVID_ERR_FAIL; | ||
705 : | } | ||
706 : | } | ||
707 : | |||
708 : | edgomez | 200 | /***************************************************************************** |
709 : | * XviD Native decoder entry point | ||
710 : | * | ||
711 : | * This function is just a wrapper to all the option cases. | ||
712 : | * | ||
713 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
714 : | * else returns the wrapped function result | ||
715 : | * | ||
716 : | ****************************************************************************/ | ||
717 : | |||
718 : | edgomez | 195 | int |
719 : | xvid_decore(void *handle, | ||
720 : | int opt, | ||
721 : | void *param1, | ||
722 : | void *param2) | ||
723 : | Isibaar | 3 | { |
724 : | edgomez | 195 | switch (opt) { |
725 : | case XVID_DEC_CREATE: | ||
726 : | edgomez | 1382 | return decoder_create((xvid_dec_create_t *) param1); |
727 : | Isibaar | 3 | |
728 : | edgomez | 195 | case XVID_DEC_DESTROY: |
729 : | return decoder_destroy((DECODER *) handle); | ||
730 : | |||
731 : | edgomez | 1382 | case XVID_DEC_DECODE: |
732 : | return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); | ||
733 : | |||
734 : | Isibaar | 3 | default: |
735 : | edgomez | 195 | return XVID_ERR_FAIL; |
736 : | } | ||
737 : | Isibaar | 3 | } |
738 : | |||
739 : | |||
740 : | edgomez | 200 | /***************************************************************************** |
741 : | * XviD Native encoder entry point | ||
742 : | * | ||
743 : | * This function is just a wrapper to all the option cases. | ||
744 : | * | ||
745 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
746 : | * else returns the wrapped function result | ||
747 : | * | ||
748 : | ****************************************************************************/ | ||
749 : | |||
750 : | edgomez | 195 | int |
751 : | xvid_encore(void *handle, | ||
752 : | int opt, | ||
753 : | void *param1, | ||
754 : | void *param2) | ||
755 : | Isibaar | 3 | { |
756 : | edgomez | 195 | switch (opt) { |
757 : | case XVID_ENC_ENCODE: | ||
758 : | edgomez | 851 | |
759 : | edgomez | 1382 | return enc_encode((Encoder *) handle, |
760 : | (xvid_enc_frame_t *) param1, | ||
761 : | (xvid_enc_stats_t *) param2); | ||
762 : | Isibaar | 3 | |
763 : | edgomez | 195 | case XVID_ENC_CREATE: |
764 : | edgomez | 1382 | return enc_create((xvid_enc_create_t *) param1); |
765 : | Isibaar | 3 | |
766 : | edgomez | 195 | case XVID_ENC_DESTROY: |
767 : | edgomez | 1382 | return enc_destroy((Encoder *) handle); |
768 : | edgomez | 195 | |
769 : | Isibaar | 3 | default: |
770 : | edgomez | 195 | return XVID_ERR_FAIL; |
771 : | } | ||
772 : | Isibaar | 3 | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |