Parent Directory
|
Revision Log
Revision 1.49 - (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 : | suxen_drol | 1.49 | * $Id: xvid.c,v 1.48 2004/03/22 22:36:23 edgomez Exp $ |
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 | 1.41 | #if defined(ARCH_IS_IA32) |
56 : | #if defined(_MSC_VER) | ||
57 : | # include <windows.h> | ||
58 : | suxen_drol | 1.31 | #else |
59 : | edgomez | 1.41 | # include <signal.h> |
60 : | # include <setjmp.h> | ||
61 : | suxen_drol | 1.31 | |
62 : | edgomez | 1.41 | static jmp_buf mark; |
63 : | suxen_drol | 1.31 | |
64 : | edgomez | 1.41 | static void |
65 : | sigill_handler(int signal) | ||
66 : | { | ||
67 : | longjmp(mark, 1); | ||
68 : | } | ||
69 : | suxen_drol | 1.31 | #endif |
70 : | |||
71 : | |||
72 : | /* | ||
73 : | edgomez | 1.43 | * Calls the funcptr, and returns whether SIGILL (illegal instruction) was |
74 : | * signalled | ||
75 : | * | ||
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 : | edgomez | 1.41 | #if defined(_MSC_VER) |
86 : | suxen_drol | 1.31 | _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 | 1.41 | |
120 : | /* detect cpu flags */ | ||
121 : | static unsigned int | ||
122 : | detect_cpu_flags() | ||
123 : | { | ||
124 : | /* enable native assembly optimizations by default */ | ||
125 : | unsigned int cpu_flags = XVID_CPU_ASM; | ||
126 : | |||
127 : | #if defined(ARCH_IS_IA32) | ||
128 : | cpu_flags |= check_cpu_features(); | ||
129 : | if ((cpu_flags & XVID_CPU_SSE) && sigill_check(sse_os_trigger)) | ||
130 : | cpu_flags &= ~XVID_CPU_SSE; | ||
131 : | |||
132 : | if ((cpu_flags & XVID_CPU_SSE2) && sigill_check(sse2_os_trigger)) | ||
133 : | cpu_flags &= ~XVID_CPU_SSE2; | ||
134 : | #endif | ||
135 : | |||
136 : | #if defined(ARCH_IS_PPC) | ||
137 : | #if defined(ARCH_IS_PPC_ALTIVEC) | ||
138 : | cpu_flags |= XVID_CPU_ALTIVEC; | ||
139 : | #endif | ||
140 : | #endif | ||
141 : | |||
142 : | return cpu_flags; | ||
143 : | } | ||
144 : | |||
145 : | |||
146 : | edgomez | 1.16 | /***************************************************************************** |
147 : | * XviD Init Entry point | ||
148 : | * | ||
149 : | * Well this function initialize all internal function pointers according | ||
150 : | * to the CPU features forced by the library client or autodetected (depending | ||
151 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
152 : | * image colorspace transformation tables. | ||
153 : | edgomez | 1.48 | * |
154 : | edgomez | 1.16 | * Returned value : XVID_ERR_OK |
155 : | * + API_VERSION in the input XVID_INIT_PARAM structure | ||
156 : | * + core build " " " " " | ||
157 : | * | ||
158 : | ****************************************************************************/ | ||
159 : | |||
160 : | edgomez | 1.41 | |
161 : | edgomez | 1.48 | static |
162 : | int xvid_gbl_init(xvid_gbl_init_t * init) | ||
163 : | Isibaar | 1.1 | { |
164 : | edgomez | 1.48 | unsigned int cpu_flags; |
165 : | suxen_drol | 1.21 | |
166 : | edgomez | 1.48 | if (XVID_VERSION_MAJOR(init->version) != 1) /* v1.x.x */ |
167 : | return XVID_ERR_VERSION; | ||
168 : | suxen_drol | 1.31 | |
169 : | edgomez | 1.48 | cpu_flags = (init->cpu_flags & XVID_CPU_FORCE) ? init->cpu_flags : detect_cpu_flags(); |
170 : | Isibaar | 1.1 | |
171 : | edgomez | 1.16 | /* Initialize the function pointers */ |
172 : | Isibaar | 1.1 | idct_int32_init(); |
173 : | Isibaar | 1.9 | init_vlc_tables(); |
174 : | |||
175 : | edgomez | 1.16 | /* Fixed Point Forward/Inverse DCT transformations */ |
176 : | Isibaar | 1.1 | fdct = fdct_int32; |
177 : | idct = idct_int32; | ||
178 : | |||
179 : | edgomez | 1.16 | /* Only needed on PPC Altivec archs */ |
180 : | canard | 1.10 | sadInit = 0; |
181 : | edgomez | 1.15 | |
182 : | edgomez | 1.16 | /* Restore FPU context : emms_c is a nop functions */ |
183 : | Isibaar | 1.1 | emms = emms_c; |
184 : | |||
185 : | edgomez | 1.48 | /* Qpel stuff */ |
186 : | xvid_QP_Funcs = &xvid_QP_Funcs_C; | ||
187 : | xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_C; | ||
188 : | xvid_Init_QP(); | ||
189 : | |||
190 : | edgomez | 1.16 | /* Quantization functions */ |
191 : | edgomez | 1.48 | quant_h263_intra = quant_h263_intra_c; |
192 : | quant_h263_inter = quant_h263_inter_c; | ||
193 : | dequant_h263_intra = dequant_h263_intra_c; | ||
194 : | dequant_h263_inter = dequant_h263_inter_c; | ||
195 : | |||
196 : | quant_mpeg_intra = quant_mpeg_intra_c; | ||
197 : | quant_mpeg_inter = quant_mpeg_inter_c; | ||
198 : | dequant_mpeg_intra = dequant_mpeg_intra_c; | ||
199 : | dequant_mpeg_inter = dequant_mpeg_inter_c; | ||
200 : | Isibaar | 1.1 | |
201 : | edgomez | 1.16 | /* Block transfer related functions */ |
202 : | Isibaar | 1.1 | transfer_8to16copy = transfer_8to16copy_c; |
203 : | transfer_16to8copy = transfer_16to8copy_c; | ||
204 : | edgomez | 1.16 | transfer_8to16sub = transfer_8to16sub_c; |
205 : | edgomez | 1.41 | transfer_8to16subro = transfer_8to16subro_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 | 1.41 | /* Interlacing functions */ |
211 : | MBFieldTest = MBFieldTest_c; | ||
212 : | |||
213 : | edgomez | 1.16 | /* Image interpolation related functions */ |
214 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_c; | ||
215 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_c; | ||
216 : | Isibaar | 1.1 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_c; |
217 : | |||
218 : | edgomez | 1.41 | interpolate16x16_lowpass_h = interpolate16x16_lowpass_h_c; |
219 : | interpolate16x16_lowpass_v = interpolate16x16_lowpass_v_c; | ||
220 : | interpolate16x16_lowpass_hv = interpolate16x16_lowpass_hv_c; | ||
221 : | |||
222 : | interpolate8x8_lowpass_h = interpolate8x8_lowpass_h_c; | ||
223 : | interpolate8x8_lowpass_v = interpolate8x8_lowpass_v_c; | ||
224 : | interpolate8x8_lowpass_hv = interpolate8x8_lowpass_hv_c; | ||
225 : | |||
226 : | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_c; | ||
227 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_c; | ||
228 : | |||
229 : | interpolate8x8_avg2 = interpolate8x8_avg2_c; | ||
230 : | interpolate8x8_avg4 = interpolate8x8_avg4_c; | ||
231 : | |||
232 : | suxen_drol | 1.49 | /* postprocessing */ |
233 : | image_brightness = image_brightness_c; | ||
234 : | |||
235 : | edgomez | 1.48 | /* reduced resolution */ |
236 : | edgomez | 1.41 | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_C; |
237 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_C; | ||
238 : | vfilter_31 = xvid_VFilter_31_C; | ||
239 : | hfilter_31 = xvid_HFilter_31_C; | ||
240 : | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_C; | ||
241 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_C; | ||
242 : | |||
243 : | edgomez | 1.16 | /* Initialize internal colorspace transformation tables */ |
244 : | Isibaar | 1.1 | colorspace_init(); |
245 : | |||
246 : | edgomez | 1.16 | /* All colorspace transformation functions User Format->YV12 */ |
247 : | edgomez | 1.41 | yv12_to_yv12 = yv12_to_yv12_c; |
248 : | rgb555_to_yv12 = rgb555_to_yv12_c; | ||
249 : | rgb565_to_yv12 = rgb565_to_yv12_c; | ||
250 : | bgr_to_yv12 = bgr_to_yv12_c; | ||
251 : | bgra_to_yv12 = bgra_to_yv12_c; | ||
252 : | abgr_to_yv12 = abgr_to_yv12_c; | ||
253 : | rgba_to_yv12 = rgba_to_yv12_c; | ||
254 : | edgomez | 1.48 | argb_to_yv12 = argb_to_yv12_c; |
255 : | edgomez | 1.41 | yuyv_to_yv12 = yuyv_to_yv12_c; |
256 : | uyvy_to_yv12 = uyvy_to_yv12_c; | ||
257 : | |||
258 : | rgb555i_to_yv12 = rgb555i_to_yv12_c; | ||
259 : | rgb565i_to_yv12 = rgb565i_to_yv12_c; | ||
260 : | bgri_to_yv12 = bgri_to_yv12_c; | ||
261 : | bgrai_to_yv12 = bgrai_to_yv12_c; | ||
262 : | abgri_to_yv12 = abgri_to_yv12_c; | ||
263 : | rgbai_to_yv12 = rgbai_to_yv12_c; | ||
264 : | edgomez | 1.48 | argbi_to_yv12 = argbi_to_yv12_c; |
265 : | edgomez | 1.41 | yuyvi_to_yv12 = yuyvi_to_yv12_c; |
266 : | uyvyi_to_yv12 = uyvyi_to_yv12_c; | ||
267 : | |||
268 : | edgomez | 1.16 | /* All colorspace transformation functions YV12->User format */ |
269 : | edgomez | 1.41 | yv12_to_rgb555 = yv12_to_rgb555_c; |
270 : | yv12_to_rgb565 = yv12_to_rgb565_c; | ||
271 : | yv12_to_bgr = yv12_to_bgr_c; | ||
272 : | yv12_to_bgra = yv12_to_bgra_c; | ||
273 : | yv12_to_abgr = yv12_to_abgr_c; | ||
274 : | yv12_to_rgba = yv12_to_rgba_c; | ||
275 : | edgomez | 1.48 | yv12_to_argb = yv12_to_argb_c; |
276 : | edgomez | 1.41 | yv12_to_yuyv = yv12_to_yuyv_c; |
277 : | yv12_to_uyvy = yv12_to_uyvy_c; | ||
278 : | edgomez | 1.48 | |
279 : | edgomez | 1.41 | yv12_to_rgb555i = yv12_to_rgb555i_c; |
280 : | yv12_to_rgb565i = yv12_to_rgb565i_c; | ||
281 : | yv12_to_bgri = yv12_to_bgri_c; | ||
282 : | yv12_to_bgrai = yv12_to_bgrai_c; | ||
283 : | yv12_to_abgri = yv12_to_abgri_c; | ||
284 : | yv12_to_rgbai = yv12_to_rgbai_c; | ||
285 : | edgomez | 1.48 | yv12_to_argbi = yv12_to_argbi_c; |
286 : | edgomez | 1.41 | yv12_to_yuyvi = yv12_to_yuyvi_c; |
287 : | yv12_to_uyvyi = yv12_to_uyvyi_c; | ||
288 : | Isibaar | 1.1 | |
289 : | edgomez | 1.16 | /* Functions used in motion estimation algorithms */ |
290 : | edgomez | 1.48 | calc_cbp = calc_cbp_c; |
291 : | sad16 = sad16_c; | ||
292 : | sad8 = sad8_c; | ||
293 : | sad16bi = sad16bi_c; | ||
294 : | sad8bi = sad8bi_c; | ||
295 : | dev16 = dev16_c; | ||
296 : | sad16v = sad16v_c; | ||
297 : | sse8_16bit = sse8_16bit_c; | ||
298 : | Isibaar | 1.1 | |
299 : | edgomez | 1.41 | #if defined(ARCH_IS_IA32) |
300 : | |||
301 : | edgomez | 1.48 | if ((cpu_flags & XVID_CPU_ASM)) { |
302 : | edgomez | 1.41 | vfilter_31 = xvid_VFilter_31_x86; |
303 : | hfilter_31 = xvid_HFilter_31_x86; | ||
304 : | } | ||
305 : | |||
306 : | if ((cpu_flags & XVID_CPU_MMX) || (cpu_flags & XVID_CPU_MMXEXT) || | ||
307 : | (cpu_flags & XVID_CPU_3DNOW) || (cpu_flags & XVID_CPU_3DNOWEXT) || | ||
308 : | (cpu_flags & XVID_CPU_SSE) || (cpu_flags & XVID_CPU_SSE2)) | ||
309 : | { | ||
310 : | /* Restore FPU context : emms_c is a nop functions */ | ||
311 : | emms = emms_mmx; | ||
312 : | } | ||
313 : | |||
314 : | if ((cpu_flags & XVID_CPU_MMX)) { | ||
315 : | edgomez | 1.16 | |
316 : | /* Forward and Inverse Discrete Cosine Transformation functions */ | ||
317 : | edgomez | 1.48 | fdct = fdct_mmx_skal; |
318 : | Isibaar | 1.47 | idct = idct_mmx; |
319 : | Isibaar | 1.1 | |
320 : | edgomez | 1.48 | /* Qpel stuff */ |
321 : | xvid_QP_Funcs = &xvid_QP_Funcs_mmx; | ||
322 : | xvid_QP_Add_Funcs = &xvid_QP_Add_Funcs_mmx; | ||
323 : | |||
324 : | edgomez | 1.16 | /* Quantization related functions */ |
325 : | edgomez | 1.48 | quant_h263_intra = quant_h263_intra_mmx; |
326 : | quant_h263_inter = quant_h263_inter_mmx; | ||
327 : | dequant_h263_intra = dequant_h263_intra_mmx; | ||
328 : | dequant_h263_inter = dequant_h263_inter_mmx; | ||
329 : | |||
330 : | quant_mpeg_intra = quant_mpeg_intra_mmx; | ||
331 : | quant_mpeg_inter = quant_mpeg_inter_mmx; | ||
332 : | dequant_mpeg_intra = dequant_mpeg_intra_mmx; | ||
333 : | dequant_mpeg_inter = dequant_mpeg_inter_mmx; | ||
334 : | Isibaar | 1.1 | |
335 : | edgomez | 1.16 | /* Block related functions */ |
336 : | Isibaar | 1.1 | transfer_8to16copy = transfer_8to16copy_mmx; |
337 : | transfer_16to8copy = transfer_16to8copy_mmx; | ||
338 : | edgomez | 1.16 | transfer_8to16sub = transfer_8to16sub_mmx; |
339 : | edgomez | 1.41 | transfer_8to16subro = transfer_8to16subro_mmx; |
340 : | edgomez | 1.22 | transfer_8to16sub2 = transfer_8to16sub2_mmx; |
341 : | edgomez | 1.16 | transfer_16to8add = transfer_16to8add_mmx; |
342 : | transfer8x8_copy = transfer8x8_copy_mmx; | ||
343 : | edgomez | 1.22 | |
344 : | edgomez | 1.41 | /* Interlacing Functions */ |
345 : | MBFieldTest = MBFieldTest_mmx; | ||
346 : | edgomez | 1.16 | |
347 : | /* Image Interpolation related functions */ | ||
348 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_mmx; | ||
349 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_mmx; | ||
350 : | Isibaar | 1.1 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_mmx; |
351 : | |||
352 : | edgomez | 1.41 | interpolate8x8_6tap_lowpass_h = interpolate8x8_6tap_lowpass_h_mmx; |
353 : | interpolate8x8_6tap_lowpass_v = interpolate8x8_6tap_lowpass_v_mmx; | ||
354 : | |||
355 : | interpolate8x8_avg2 = interpolate8x8_avg2_mmx; | ||
356 : | interpolate8x8_avg4 = interpolate8x8_avg4_mmx; | ||
357 : | suxen_drol | 1.49 | |
358 : | /* postprocessing */ | ||
359 : | image_brightness = image_brightness_mmx; | ||
360 : | edgomez | 1.41 | |
361 : | /* reduced resolution */ | ||
362 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_mmx; | ||
363 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_mmx; | ||
364 : | hfilter_31 = xvid_HFilter_31_mmx; | ||
365 : | filter_18x18_to_8x8 = xvid_Filter_18x18_To_8x8_mmx; | ||
366 : | filter_diff_18x18_to_8x8 = xvid_Filter_Diff_18x18_To_8x8_mmx; | ||
367 : | |||
368 : | /* image input xxx_to_yv12 related functions */ | ||
369 : | yv12_to_yv12 = yv12_to_yv12_mmx; | ||
370 : | bgr_to_yv12 = bgr_to_yv12_mmx; | ||
371 : | bgra_to_yv12 = bgra_to_yv12_mmx; | ||
372 : | edgomez | 1.16 | yuyv_to_yv12 = yuyv_to_yv12_mmx; |
373 : | uyvy_to_yv12 = uyvy_to_yv12_mmx; | ||
374 : | Isibaar | 1.1 | |
375 : | edgomez | 1.41 | /* image output yv12_to_xxx related functions */ |
376 : | yv12_to_bgr = yv12_to_bgr_mmx; | ||
377 : | yv12_to_bgra = yv12_to_bgra_mmx; | ||
378 : | edgomez | 1.16 | yv12_to_yuyv = yv12_to_yuyv_mmx; |
379 : | yv12_to_uyvy = yv12_to_uyvy_mmx; | ||
380 : | edgomez | 1.48 | |
381 : | edgomez | 1.41 | yv12_to_yuyvi = yv12_to_yuyvi_mmx; |
382 : | yv12_to_uyvyi = yv12_to_uyvyi_mmx; | ||
383 : | Isibaar | 1.1 | |
384 : | edgomez | 1.16 | /* Motion estimation related functions */ |
385 : | Isibaar | 1.1 | calc_cbp = calc_cbp_mmx; |
386 : | edgomez | 1.16 | sad16 = sad16_mmx; |
387 : | sad8 = sad8_mmx; | ||
388 : | suxen_drol | 1.33 | sad16bi = sad16bi_mmx; |
389 : | sad8bi = sad8bi_mmx; | ||
390 : | edgomez | 1.16 | dev16 = dev16_mmx; |
391 : | edgomez | 1.41 | sad16v = sad16v_mmx; |
392 : | edgomez | 1.48 | sse8_16bit = sse8_16bit_mmx; |
393 : | Isibaar | 1.1 | } |
394 : | |||
395 : | suxen_drol | 1.33 | /* these 3dnow functions are faster than mmx, but slower than xmm. */ |
396 : | edgomez | 1.41 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
397 : | |||
398 : | emms = emms_3dn; | ||
399 : | suxen_drol | 1.33 | |
400 : | /* ME functions */ | ||
401 : | sad16bi = sad16bi_3dn; | ||
402 : | sad8bi = sad8bi_3dn; | ||
403 : | edgomez | 1.41 | |
404 : | yuyv_to_yv12 = yuyv_to_yv12_3dn; | ||
405 : | uyvy_to_yv12 = uyvy_to_yv12_3dn; | ||
406 : | suxen_drol | 1.33 | } |
407 : | |||
408 : | |||
409 : | edgomez | 1.41 | if ((cpu_flags & XVID_CPU_MMXEXT)) { |
410 : | edgomez | 1.16 | |
411 : | edgomez | 1.48 | /* DCT */ |
412 : | fdct = fdct_xmm_skal; | ||
413 : | Isibaar | 1.47 | idct = idct_xmm; |
414 : | edgomez | 1.16 | |
415 : | /* Interpolation */ | ||
416 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_xmm; | ||
417 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_xmm; | ||
418 : | h | 1.3 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_xmm; |
419 : | Isibaar | 1.25 | |
420 : | edgomez | 1.41 | /* reduced resolution */ |
421 : | copy_upsampled_8x8_16to8 = xvid_Copy_Upsampled_8x8_16To8_xmm; | ||
422 : | add_upsampled_8x8_16to8 = xvid_Add_Upsampled_8x8_16To8_xmm; | ||
423 : | |||
424 : | chenm001 | 1.29 | /* Quantization */ |
425 : | edgomez | 1.48 | quant_mpeg_intra = quant_mpeg_intra_xmm; |
426 : | quant_mpeg_inter = quant_mpeg_inter_xmm; | ||
427 : | edgomez | 1.41 | |
428 : | edgomez | 1.48 | dequant_h263_intra = dequant_h263_intra_xmm; |
429 : | dequant_h263_inter = dequant_h263_inter_xmm; | ||
430 : | chenm001 | 1.29 | |
431 : | edgomez | 1.19 | /* Buffer transfer */ |
432 : | transfer_8to16sub2 = transfer_8to16sub2_xmm; | ||
433 : | edgomez | 1.16 | |
434 : | /* Colorspace transformation */ | ||
435 : | edgomez | 1.41 | yv12_to_yv12 = yv12_to_yv12_xmm; |
436 : | yuyv_to_yv12 = yuyv_to_yv12_xmm; | ||
437 : | uyvy_to_yv12 = uyvy_to_yv12_xmm; | ||
438 : | Isibaar | 1.1 | |
439 : | edgomez | 1.16 | /* ME functions */ |
440 : | Isibaar | 1.1 | sad16 = sad16_xmm; |
441 : | suxen_drol | 1.33 | sad8 = sad8_xmm; |
442 : | chenm001 | 1.29 | sad16bi = sad16bi_xmm; |
443 : | suxen_drol | 1.33 | sad8bi = sad8bi_xmm; |
444 : | Isibaar | 1.1 | dev16 = dev16_xmm; |
445 : | edgomez | 1.41 | sad16v = sad16v_xmm; |
446 : | Isibaar | 1.1 | } |
447 : | |||
448 : | edgomez | 1.41 | if ((cpu_flags & XVID_CPU_3DNOW)) { |
449 : | edgomez | 1.16 | |
450 : | /* Interpolation */ | ||
451 : | Isibaar | 1.1 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dn; |
452 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dn; | ||
453 : | h | 1.4 | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dn; |
454 : | Isibaar | 1.1 | } |
455 : | |||
456 : | edgomez | 1.41 | if ((cpu_flags & XVID_CPU_3DNOWEXT)) { |
457 : | |||
458 : | /* Inverse DCT */ | ||
459 : | Isibaar | 1.47 | idct = idct_3dne; |
460 : | edgomez | 1.41 | |
461 : | /* Buffer transfer */ | ||
462 : | transfer_8to16copy = transfer_8to16copy_3dne; | ||
463 : | transfer_16to8copy = transfer_16to8copy_3dne; | ||
464 : | transfer_8to16sub = transfer_8to16sub_3dne; | ||
465 : | transfer_8to16subro = transfer_8to16subro_3dne; | ||
466 : | transfer_8to16sub2 = transfer_8to16sub2_3dne; | ||
467 : | transfer_16to8add = transfer_16to8add_3dne; | ||
468 : | transfer8x8_copy = transfer8x8_copy_3dne; | ||
469 : | |||
470 : | /* Quantization */ | ||
471 : | edgomez | 1.48 | quant_h263_intra = quant_h263_intra_3dne; |
472 : | quant_h263_inter = quant_h263_inter_3dne; | ||
473 : | dequant_mpeg_intra = dequant_mpeg_intra_3dne; | ||
474 : | dequant_mpeg_inter = dequant_mpeg_inter_3dne; | ||
475 : | dequant_h263_intra = dequant_h263_intra_3dne; | ||
476 : | dequant_h263_inter = dequant_h263_inter_3dne; | ||
477 : | edgomez | 1.41 | |
478 : | /* ME functions */ | ||
479 : | calc_cbp = calc_cbp_3dne; | ||
480 : | sad16 = sad16_3dne; | ||
481 : | sad8 = sad8_3dne; | ||
482 : | sad16bi = sad16bi_3dne; | ||
483 : | sad8bi = sad8bi_3dne; | ||
484 : | dev16 = dev16_3dne; | ||
485 : | |||
486 : | /* Interpolation */ | ||
487 : | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_3dne; | ||
488 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_3dne; | ||
489 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_3dne; | ||
490 : | } | ||
491 : | |||
492 : | if ((cpu_flags & XVID_CPU_SSE2)) { | ||
493 : | edgomez | 1.16 | |
494 : | chenm001 | 1.29 | calc_cbp = calc_cbp_sse2; |
495 : | |||
496 : | edgomez | 1.16 | /* Quantization */ |
497 : | edgomez | 1.48 | quant_h263_intra = quant_h263_intra_sse2; |
498 : | quant_h263_inter = quant_h263_inter_sse2; | ||
499 : | dequant_h263_intra = dequant_h263_intra_sse2; | ||
500 : | dequant_h263_inter = dequant_h263_inter_sse2; | ||
501 : | h | 1.13 | |
502 : | edgomez | 1.48 | /* SAD operators */ |
503 : | edgomez | 1.16 | sad16 = sad16_sse2; |
504 : | dev16 = dev16_sse2; | ||
505 : | edgomez | 1.48 | |
506 : | /* DCT operators | ||
507 : | * no iDCT because it's not "Walken matching" */ | ||
508 : | fdct = fdct_sse2_skal; | ||
509 : | h | 1.12 | } |
510 : | edgomez | 1.48 | #endif /* ARCH_IS_IA32 */ |
511 : | Isibaar | 1.18 | |
512 : | edgomez | 1.41 | #if defined(ARCH_IS_IA64) |
513 : | edgomez | 1.43 | if ((cpu_flags & XVID_CPU_ASM)) { /* use assembler routines? */ |
514 : | Isibaar | 1.18 | idct_ia64_init(); |
515 : | fdct = fdct_ia64; | ||
516 : | edgomez | 1.43 | idct = idct_ia64; /*not yet working, crashes */ |
517 : | Isibaar | 1.18 | interpolate8x8_halfpel_h = interpolate8x8_halfpel_h_ia64; |
518 : | interpolate8x8_halfpel_v = interpolate8x8_halfpel_v_ia64; | ||
519 : | interpolate8x8_halfpel_hv = interpolate8x8_halfpel_hv_ia64; | ||
520 : | sad16 = sad16_ia64; | ||
521 : | sad16bi = sad16bi_ia64; | ||
522 : | sad8 = sad8_ia64; | ||
523 : | dev16 = dev16_ia64; | ||
524 : | edgomez | 1.43 | /* Halfpel8_Refine = Halfpel8_Refine_ia64; */ |
525 : | edgomez | 1.48 | quant_h263_intra = quant_h263_intra_ia64; |
526 : | quant_h263_inter = quant_h263_inter_ia64; | ||
527 : | dequant_h263_intra = dequant_h263_intra_ia64; | ||
528 : | dequant_h263_inter = dequant_h263_inter_ia64; | ||
529 : | Isibaar | 1.18 | transfer_8to16copy = transfer_8to16copy_ia64; |
530 : | transfer_16to8copy = transfer_16to8copy_ia64; | ||
531 : | transfer_8to16sub = transfer_8to16sub_ia64; | ||
532 : | transfer_8to16sub2 = transfer_8to16sub2_ia64; | ||
533 : | transfer_16to8add = transfer_16to8add_ia64; | ||
534 : | transfer8x8_copy = transfer8x8_copy_ia64; | ||
535 : | } | ||
536 : | edgomez | 1.48 | #endif |
537 : | edgomez | 1.16 | |
538 : | edgomez | 1.41 | #if defined(ARCH_IS_PPC) |
539 : | if ((cpu_flags & XVID_CPU_ASM)) | ||
540 : | { | ||
541 : | calc_cbp = calc_cbp_ppc; | ||
542 : | } | ||
543 : | |||
544 : | if ((cpu_flags & XVID_CPU_ALTIVEC)) | ||
545 : | { | ||
546 : | calc_cbp = calc_cbp_altivec; | ||
547 : | fdct = fdct_altivec; | ||
548 : | idct = idct_altivec; | ||
549 : | sadInit = sadInit_altivec; | ||
550 : | sad16 = sad16_altivec; | ||
551 : | sad8 = sad8_altivec; | ||
552 : | dev16 = dev16_altivec; | ||
553 : | } | ||
554 : | canard | 1.6 | #endif |
555 : | edgomez | 1.41 | |
556 : | edgomez | 1.48 | #if defined(_DEBUG) |
557 : | xvid_debug = init->debug; | ||
558 : | #endif | ||
559 : | |||
560 : | return 0; | ||
561 : | edgomez | 1.41 | } |
562 : | |||
563 : | |||
564 : | static int | ||
565 : | edgomez | 1.48 | xvid_gbl_info(xvid_gbl_info_t * info) |
566 : | edgomez | 1.41 | { |
567 : | edgomez | 1.48 | if (XVID_VERSION_MAJOR(info->version) != 1) /* v1.x.x */ |
568 : | return XVID_ERR_VERSION; | ||
569 : | edgomez | 1.41 | |
570 : | edgomez | 1.48 | info->actual_version = XVID_VERSION; |
571 : | info->build = "xvid-1.0.0"; | ||
572 : | info->cpu_flags = detect_cpu_flags(); | ||
573 : | edgomez | 1.41 | |
574 : | edgomez | 1.48 | #if defined(_SMP) && defined(WIN32) |
575 : | info->num_threads = pthread_num_processors_np();; | ||
576 : | #else | ||
577 : | info->num_threads = 0; | ||
578 : | #endif | ||
579 : | edgomez | 1.41 | |
580 : | return 0; | ||
581 : | } | ||
582 : | |||
583 : | |||
584 : | edgomez | 1.48 | static int |
585 : | xvid_gbl_convert(xvid_gbl_convert_t* convert) | ||
586 : | edgomez | 1.41 | { |
587 : | edgomez | 1.48 | int width; |
588 : | int height; | ||
589 : | int width2; | ||
590 : | int height2; | ||
591 : | IMAGE img; | ||
592 : | edgomez | 1.41 | |
593 : | edgomez | 1.48 | if (XVID_VERSION_MAJOR(convert->version) != 1) /* v1.x.x */ |
594 : | return XVID_ERR_VERSION; | ||
595 : | edgomez | 1.41 | |
596 : | edgomez | 1.48 | #if 0 |
597 : | const int flip1 = (convert->input.colorspace & XVID_CSP_VFLIP) ^ (convert->output.colorspace & XVID_CSP_VFLIP); | ||
598 : | #endif | ||
599 : | width = convert->width; | ||
600 : | height = convert->height; | ||
601 : | width2 = convert->width/2; | ||
602 : | height2 = convert->height/2; | ||
603 : | edgomez | 1.41 | |
604 : | edgomez | 1.48 | switch (convert->input.csp & ~XVID_CSP_VFLIP) |
605 : | edgomez | 1.41 | { |
606 : | edgomez | 1.48 | case XVID_CSP_YV12 : |
607 : | img.y = convert->input.plane[0]; | ||
608 : | img.v = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height; | ||
609 : | img.u = (uint8_t*)convert->input.plane[0] + convert->input.stride[0]*height + (convert->input.stride[0]/2)*height2; | ||
610 : | image_output(&img, width, height, width, | ||
611 : | (uint8_t**)convert->output.plane, convert->output.stride, | ||
612 : | convert->output.csp, convert->interlacing); | ||
613 : | break; | ||
614 : | edgomez | 1.41 | |
615 : | edgomez | 1.48 | default : |
616 : | return XVID_ERR_FORMAT; | ||
617 : | edgomez | 1.41 | } |
618 : | |||
619 : | |||
620 : | edgomez | 1.48 | emms(); |
621 : | edgomez | 1.41 | return 0; |
622 : | } | ||
623 : | |||
624 : | edgomez | 1.48 | /***************************************************************************** |
625 : | * XviD Global Entry point | ||
626 : | * | ||
627 : | * Well this function initialize all internal function pointers according | ||
628 : | * to the CPU features forced by the library client or autodetected (depending | ||
629 : | * on the XVID_CPU_FORCE flag). It also initializes vlc coding tables and all | ||
630 : | * image colorspace transformation tables. | ||
631 : | * | ||
632 : | ****************************************************************************/ | ||
633 : | Isibaar | 1.1 | |
634 : | edgomez | 1.41 | |
635 : | int | ||
636 : | edgomez | 1.48 | xvid_global(void *handle, |
637 : | edgomez | 1.41 | int opt, |
638 : | void *param1, | ||
639 : | void *param2) | ||
640 : | { | ||
641 : | switch(opt) | ||
642 : | { | ||
643 : | edgomez | 1.48 | case XVID_GBL_INIT : |
644 : | return xvid_gbl_init((xvid_gbl_init_t*)param1); | ||
645 : | edgomez | 1.41 | |
646 : | edgomez | 1.48 | case XVID_GBL_INFO : |
647 : | return xvid_gbl_info((xvid_gbl_info_t*)param1); | ||
648 : | |||
649 : | case XVID_GBL_CONVERT : | ||
650 : | return xvid_gbl_convert((xvid_gbl_convert_t*)param1); | ||
651 : | edgomez | 1.41 | |
652 : | default : | ||
653 : | return XVID_ERR_FAIL; | ||
654 : | } | ||
655 : | } | ||
656 : | |||
657 : | edgomez | 1.16 | /***************************************************************************** |
658 : | * XviD Native decoder entry point | ||
659 : | * | ||
660 : | * This function is just a wrapper to all the option cases. | ||
661 : | * | ||
662 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
663 : | * else returns the wrapped function result | ||
664 : | * | ||
665 : | ****************************************************************************/ | ||
666 : | |||
667 : | edgomez | 1.15 | int |
668 : | xvid_decore(void *handle, | ||
669 : | int opt, | ||
670 : | void *param1, | ||
671 : | void *param2) | ||
672 : | Isibaar | 1.1 | { |
673 : | edgomez | 1.15 | switch (opt) { |
674 : | case XVID_DEC_CREATE: | ||
675 : | edgomez | 1.48 | return decoder_create((xvid_dec_create_t *) param1); |
676 : | edgomez | 1.15 | |
677 : | case XVID_DEC_DESTROY: | ||
678 : | return decoder_destroy((DECODER *) handle); | ||
679 : | Isibaar | 1.1 | |
680 : | edgomez | 1.48 | case XVID_DEC_DECODE: |
681 : | return decoder_decode((DECODER *) handle, (xvid_dec_frame_t *) param1, (xvid_dec_stats_t*) param2); | ||
682 : | |||
683 : | Isibaar | 1.1 | default: |
684 : | edgomez | 1.15 | return XVID_ERR_FAIL; |
685 : | } | ||
686 : | Isibaar | 1.1 | } |
687 : | |||
688 : | edgomez | 1.16 | |
689 : | /***************************************************************************** | ||
690 : | * XviD Native encoder entry point | ||
691 : | * | ||
692 : | * This function is just a wrapper to all the option cases. | ||
693 : | * | ||
694 : | * Returned values : XVID_ERR_FAIL when opt is invalid | ||
695 : | * else returns the wrapped function result | ||
696 : | * | ||
697 : | ****************************************************************************/ | ||
698 : | Isibaar | 1.1 | |
699 : | edgomez | 1.15 | int |
700 : | xvid_encore(void *handle, | ||
701 : | int opt, | ||
702 : | void *param1, | ||
703 : | void *param2) | ||
704 : | Isibaar | 1.1 | { |
705 : | edgomez | 1.15 | switch (opt) { |
706 : | case XVID_ENC_ENCODE: | ||
707 : | edgomez | 1.41 | |
708 : | edgomez | 1.48 | return enc_encode((Encoder *) handle, |
709 : | (xvid_enc_frame_t *) param1, | ||
710 : | (xvid_enc_stats_t *) param2); | ||
711 : | edgomez | 1.15 | |
712 : | case XVID_ENC_CREATE: | ||
713 : | edgomez | 1.48 | return enc_create((xvid_enc_create_t *) param1); |
714 : | edgomez | 1.15 | |
715 : | case XVID_ENC_DESTROY: | ||
716 : | edgomez | 1.48 | return enc_destroy((Encoder *) handle); |
717 : | Isibaar | 1.1 | |
718 : | default: | ||
719 : | edgomez | 1.15 | return XVID_ERR_FAIL; |
720 : | } | ||
721 : | Isibaar | 1.1 | } |
No admin address has been configured | ViewVC Help |
Powered by ViewVC 1.0.4 |